server: split HTTP into its own interface (#17216)

* server: split HTTP into its own interface

* move server-http and httplib to its own file

* add the remaining endpoints

* fix exception/error handling

* renaming

* missing header

* fix missing windows header

* fix error responses from http layer

* fix slot save/restore handler

* fix case where only one stream chunk is returned

* add NOMINMAX

* do not call sink.write on empty data

* use safe_json_to_str for SSE

* clean up

* add some comments

* improve usage of next()

* bring back the "server is listening on" message

* more generic handler

* add req.headers

* move the chat template print to init()

* add req.path

* cont : minor

---------

Co-authored-by: Georgi Gerganov <ggerganov@gmail.com>

.devops/cann.Dockerfile

@@ -49,7 +49,7 @@ RUN source /usr/local/Ascend/ascend-toolkit/set_env.sh --force \
 # -- Organize build artifacts for copying in later stages --
 # Create a lib directory to store all .so files
 RUN mkdir -p /app/lib && \
-    find build -name "*.so" -exec cp {} /app/lib \;
+    find build -name "*.so*" -exec cp -P {} /app/lib \;
 
 # Create a full directory to store all executables and Python scripts
 RUN mkdir -p /app/full && \

.devops/cpu.Dockerfile

@@ -20,7 +20,7 @@ RUN if [ "$TARGETARCH" = "amd64" ] || [ "$TARGETARCH" = "arm64" ]; then \
     cmake --build build -j $(nproc)
 
 RUN mkdir -p /app/lib && \
-    find build -name "*.so" -exec cp {} /app/lib \;
+    find build -name "*.so*" -exec cp -P {} /app/lib \;
 
 RUN mkdir -p /app/full \
     && cp build/bin/* /app/full \

.devops/cuda.Dockerfile

@@ -25,7 +25,7 @@ RUN if [ "${CUDA_DOCKER_ARCH}" != "default" ]; then \
     cmake --build build --config Release -j$(nproc)
 
 RUN mkdir -p /app/lib && \
-    find build -name "*.so" -exec cp {} /app/lib \;
+    find build -name "*.so*" -exec cp -P {} /app/lib \;
 
 RUN mkdir -p /app/full \
     && cp build/bin/* /app/full \

.devops/intel.Dockerfile

@@ -21,7 +21,7 @@ RUN if [ "${GGML_SYCL_F16}" = "ON" ]; then \
     cmake --build build --config Release -j$(nproc)
 
 RUN mkdir -p /app/lib && \
-    find build -name "*.so" -exec cp {} /app/lib \;
+    find build -name "*.so*" -exec cp -P {} /app/lib \;
 
 RUN mkdir -p /app/full \
     && cp build/bin/* /app/full \

.devops/musa.Dockerfile

@@ -32,7 +32,7 @@ RUN if [ "${MUSA_DOCKER_ARCH}" != "default" ]; then \
     cmake --build build --config Release -j$(nproc)
 
 RUN mkdir -p /app/lib && \
-    find build -name "*.so" -exec cp {} /app/lib \;
+    find build -name "*.so*" -exec cp -P {} /app/lib \;
 
 RUN mkdir -p /app/full \
     && cp build/bin/* /app/full \

.devops/rocm.Dockerfile

@@ -45,7 +45,7 @@ RUN HIPCXX="$(hipconfig -l)/clang" HIP_PATH="$(hipconfig -R)" \
     && cmake --build build --config Release -j$(nproc)
 
 RUN mkdir -p /app/lib \
-    && find build -name "*.so" -exec cp {} /app/lib \;
+    && find build -name "*.so*" -exec cp -P {} /app/lib \;
 
 RUN mkdir -p /app/full \
     && cp build/bin/* /app/full \

.devops/vulkan.Dockerfile

@@ -20,7 +20,7 @@ RUN cmake -B build -DGGML_NATIVE=OFF -DGGML_VULKAN=ON -DLLAMA_BUILD_TESTS=OFF -D
     cmake --build build --config Release -j$(nproc)
 
 RUN mkdir -p /app/lib && \
-    find build -name "*.so" -exec cp {} /app/lib \;
+    find build -name "*.so*" -exec cp -P {} /app/lib \;
 
 RUN mkdir -p /app/full \
     && cp build/bin/* /app/full \

.github/copilot-instructions.md

@@ -9,7 +9,7 @@ llama.cpp is a large-scale C/C++ project for efficient LLM (Large Language Model
 - **Size**: ~200k+ lines of code across 1000+ files
 - **Architecture**: Modular design with main library (`libllama`) and 40+ executable tools/examples
 - **Core dependency**: ggml tensor library (vendored in `ggml/` directory)
-- **Backends supported**: CPU (AVX/NEON optimized), CUDA, Metal, Vulkan, SYCL, ROCm, MUSA
+- **Backends supported**: CPU (AVX/NEON/RVV optimized), CUDA, Metal, Vulkan, SYCL, ROCm, MUSA
 - **License**: MIT
 
 ## Build Instructions

.github/workflows/build-amd.yml

@@ -1,52 +0,0 @@
-name: CI (AMD)
-
-on:
-  workflow_dispatch: # allows manual triggering
-  push:
-    branches:
-      - master
-    paths: [
-      '.github/workflows/build-amd.yml',
-      '**/CMakeLists.txt',
-      '**/.cmake',
-      '**/*.h',
-      '**/*.hpp',
-      '**/*.c',
-      '**/*.cpp',
-      '**/*.cu',
-      '**/*.cuh',
-      '**/*.comp'
-    ]
-
-concurrency:
-  group: ${{ github.workflow }}-${{ github.head_ref && github.ref || github.run_id }}
-  cancel-in-progress: true
-
-jobs:
-  ggml-ci-x64-amd-vulkan:
-    runs-on: [self-hosted, Linux, X64, AMD]
-
-    steps:
-      - name: Clone
-        id: checkout
-        uses: actions/checkout@v4
-
-      - name: Test
-        id: ggml-ci
-        run: |
-          vulkaninfo --summary
-          GG_BUILD_VULKAN=1 bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp
-
-  ggml-ci-x64-amd-rocm:
-    runs-on: [self-hosted, Linux, X64, AMD]
-
-    steps:
-      - name: Clone
-        id: checkout
-        uses: actions/checkout@v4
-
-      - name: Test
-        id: ggml-ci
-        run: |
-          amd-smi static
-          GG_BUILD_ROCM=1 GG_BUILD_AMDGPU_TARGETS="gfx1101" bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp

.github/workflows/build.yml

@@ -1599,6 +1599,34 @@ jobs:
         run: |
           bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp
 
+  ggml-ci-x64-amd-vulkan:
+    runs-on: [self-hosted, Linux, X64, AMD]
+
+    steps:
+      - name: Clone
+        id: checkout
+        uses: actions/checkout@v4
+
+      - name: Test
+        id: ggml-ci
+        run: |
+          vulkaninfo --summary
+          GG_BUILD_VULKAN=1 bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp
+
+  ggml-ci-x64-amd-rocm:
+    runs-on: [self-hosted, Linux, X64, AMD]
+
+    steps:
+      - name: Clone
+        id: checkout
+        uses: actions/checkout@v4
+
+      - name: Test
+        id: ggml-ci
+        run: |
+          amd-smi static
+          GG_BUILD_ROCM=1 GG_BUILD_AMDGPU_TARGETS="gfx1101" bash ./ci/run.sh ~/results/llama.cpp /mnt/llama.cpp
+
   ggml-ci-mac-metal:
     runs-on: [self-hosted, macOS, ARM64]
 

.github/workflows/check-vendor.yml

@@ -0,0 +1,52 @@
+name: Check vendor
+
+on:
+  workflow_dispatch: # allows manual triggering
+  push:
+    branches:
+      - master
+    paths: [
+      'vendor/**',
+      'scripts/sync_vendor.py'
+    ]
+
+  pull_request:
+    types: [opened, synchronize, reopened]
+    paths: [
+      'vendor/**',
+      'scripts/sync_vendor.py'
+    ]
+
+jobs:
+  check-vendor:
+    runs-on: ubuntu-latest
+
+    steps:
+      - name: Checkout
+        uses: actions/checkout@v4
+        with:
+          fetch-depth: 0
+
+      - name: Setup Python
+        uses: actions/setup-python@v4
+        with:
+          python-version: '3.x'
+
+      - name: Run vendor sync
+        run: |
+          set -euo pipefail
+          python3 scripts/sync_vendor.py
+
+      - name: Check for changes
+        run: |
+          set -euo pipefail
+          # detect modified or untracked files
+          changed=$(git status --porcelain --untracked-files=all || true)
+          if [ -n "$changed" ]; then
+            echo "Vendor sync modified files:"
+            echo "$changed" | awk '{ print $2 }' | sed '/^$/d'
+            echo "Failing because vendor files mismatch. Please update scripts/sync_vendor.py"
+            exit 1
+          else
+            echo "Vendor files are up-to-date."
+          fi

.github/workflows/server.yml

@@ -209,7 +209,7 @@ jobs:
         working-directory: tools/server/webui
 
       - name: Run UI tests
-        run: npm run test:ui
+        run: npm run test:ui -- --testTimeout=60000
         working-directory: tools/server/webui
 
       - name: Run E2E tests

CMakeLists.txt

@@ -92,6 +92,7 @@ option(LLAMA_TOOLS_INSTALL  "llama: install tools"        ${LLAMA_TOOLS_INSTALL_
 
 # 3rd party libs
 option(LLAMA_CURL       "llama: use libcurl to download model from an URL" ON)
+option(LLAMA_HTTPLIB    "llama: if libcurl is disabled, use httplib to download model from an URL" ON)
 option(LLAMA_OPENSSL    "llama: use openssl to support HTTPS" OFF)
 option(LLAMA_LLGUIDANCE "llama-common: include LLGuidance library for structured output in common utils" OFF)
 
@@ -200,7 +201,9 @@ endif()
 
 if (LLAMA_BUILD_COMMON)
     add_subdirectory(common)
-    add_subdirectory(vendor/cpp-httplib)
+    if (LLAMA_HTTPLIB)
+        add_subdirectory(vendor/cpp-httplib)
+    endif()
 endif()
 
 if (LLAMA_BUILD_COMMON AND LLAMA_BUILD_TESTS AND NOT CMAKE_JS_VERSION)

README.md

@@ -61,6 +61,7 @@ range of hardware - locally and in the cloud.
 - Plain C/C++ implementation without any dependencies
 - Apple silicon is a first-class citizen - optimized via ARM NEON, Accelerate and Metal frameworks
 - AVX, AVX2, AVX512 and AMX support for x86 architectures
+- RVV, ZVFH, ZFH and ZICBOP support for RISC-V architectures
 - 1.5-bit, 2-bit, 3-bit, 4-bit, 5-bit, 6-bit, and 8-bit integer quantization for faster inference and reduced memory use
 - Custom CUDA kernels for running LLMs on NVIDIA GPUs (support for AMD GPUs via HIP and Moore Threads GPUs via MUSA)
 - Vulkan and SYCL backend support

build-xcframework.sh

@@ -454,6 +454,8 @@ cmake -B build-visionos -G Xcode \
     -DCMAKE_C_FLAGS="-D_XOPEN_SOURCE=700 ${COMMON_C_FLAGS}" \
     -DCMAKE_CXX_FLAGS="-D_XOPEN_SOURCE=700 ${COMMON_CXX_FLAGS}" \
     -DLLAMA_CURL=OFF \
+    -DLLAMA_HTTPLIB=OFF \
+    -DLLAMA_BUILD_SERVER=OFF \
     -S .
 cmake --build build-visionos --config Release -- -quiet
 
@@ -468,6 +470,8 @@ cmake -B build-visionos-sim -G Xcode \
     -DCMAKE_C_FLAGS="-D_XOPEN_SOURCE=700 ${COMMON_C_FLAGS}" \
     -DCMAKE_CXX_FLAGS="-D_XOPEN_SOURCE=700 ${COMMON_CXX_FLAGS}" \
     -DLLAMA_CURL=OFF \
+    -DLLAMA_HTTPLIB=OFF \
+    -DLLAMA_BUILD_SERVER=OFF \
     -S .
 cmake --build build-visionos-sim --config Release -- -quiet
 

common/CMakeLists.txt

@@ -91,47 +91,12 @@ if (LLAMA_CURL)
     target_compile_definitions(${TARGET} PUBLIC LLAMA_USE_CURL)
     include_directories(${CURL_INCLUDE_DIRS})
     set(LLAMA_COMMON_EXTRA_LIBS ${LLAMA_COMMON_EXTRA_LIBS} ${CURL_LIBRARIES})
-else()
+elseif (LLAMA_HTTPLIB)
     # otherwise, use cpp-httplib
+    target_compile_definitions(${TARGET} PUBLIC LLAMA_USE_HTTPLIB)
     set(LLAMA_COMMON_EXTRA_LIBS ${LLAMA_COMMON_EXTRA_LIBS} cpp-httplib)
 endif()
 
-if (LLAMA_OPENSSL)
-    find_package(OpenSSL)
-    if (OpenSSL_FOUND)
-        include(CheckCSourceCompiles)
-        set(SAVED_CMAKE_REQUIRED_INCLUDES ${CMAKE_REQUIRED_INCLUDES})
-        set(CMAKE_REQUIRED_INCLUDES ${OPENSSL_INCLUDE_DIR})
-        check_c_source_compiles("
-        #include <openssl/opensslv.h>
-        #if defined(OPENSSL_IS_BORINGSSL) || defined(LIBRESSL_VERSION_NUMBER)
-        #    if OPENSSL_VERSION_NUMBER < 0x1010107f
-        #        error bad version
-        #    endif
-        #else
-        #    if OPENSSL_VERSION_NUMBER < 0x30000000L
-        #        error bad version
-        #    endif
-        #endif
-        int main() { return 0; }
-        " OPENSSL_VERSION_SUPPORTED)
-        set(CMAKE_REQUIRED_INCLUDES ${SAVED_CMAKE_REQUIRED_INCLUDES})
-        if (OPENSSL_VERSION_SUPPORTED)
-            message(STATUS "OpenSSL found: ${OPENSSL_VERSION}")
-            target_compile_definitions(${TARGET} PUBLIC CPPHTTPLIB_OPENSSL_SUPPORT)
-            target_link_libraries(${TARGET} PUBLIC OpenSSL::SSL OpenSSL::Crypto)
-            if (APPLE AND CMAKE_SYSTEM_NAME STREQUAL "Darwin")
-                target_compile_definitions(${TARGET} PUBLIC CPPHTTPLIB_USE_CERTS_FROM_MACOSX_KEYCHAIN)
-                find_library(CORE_FOUNDATION_FRAMEWORK CoreFoundation REQUIRED)
-                find_library(SECURITY_FRAMEWORK Security REQUIRED)
-                target_link_libraries(${TARGET} PUBLIC ${CORE_FOUNDATION_FRAMEWORK} ${SECURITY_FRAMEWORK})
-            endif()
-        endif()
-    else()
-        message(STATUS "OpenSSL not found, SSL support disabled")
-    endif()
-endif()
-
 if (LLAMA_LLGUIDANCE)
     include(ExternalProject)
     set(LLGUIDANCE_SRC ${CMAKE_BINARY_DIR}/llguidance/source)

common/common.cpp

@@ -355,11 +355,7 @@ bool parse_cpu_mask(const std::string & mask, bool (&boolmask)[GGML_MAX_N_THREAD
 }
 
 void common_init() {
-    llama_log_set([](ggml_log_level level, const char * text, void * /*user_data*/) {
-        if (LOG_DEFAULT_LLAMA <= common_log_verbosity_thold) {
-            common_log_add(common_log_main(), level, "%s", text);
-        }
-    }, NULL);
+    llama_log_set(common_log_default_callback, NULL);
 
 #ifdef NDEBUG
     const char * build_type = "";

common/download.cpp

@@ -20,7 +20,7 @@
 #if defined(LLAMA_USE_CURL)
 #include <curl/curl.h>
 #include <curl/easy.h>
-#else
+#elif defined(LLAMA_USE_HTTPLIB)
 #include "http.h"
 #endif
 
@@ -467,7 +467,7 @@ std::pair<long, std::vector<char>> common_remote_get_content(const std::string &
     return { res_code, std::move(res_buffer) };
 }
 
-#else
+#elif defined(LLAMA_USE_HTTPLIB)
 
 static bool is_output_a_tty() {
 #if defined(_WIN32)
@@ -713,6 +713,8 @@ std::pair<long, std::vector<char>> common_remote_get_content(const std::string
 
 #endif // LLAMA_USE_CURL
 
+#if defined(LLAMA_USE_CURL) || defined(LLAMA_USE_HTTPLIB)
+
 static bool common_download_file_single(const std::string & url,
                                         const std::string & path,
                                         const std::string & bearer_token,
@@ -907,33 +909,6 @@ common_hf_file_res common_get_hf_file(const std::string & hf_repo_with_tag, cons
     return { hf_repo, ggufFile, mmprojFile };
 }
 
-std::vector<common_cached_model_info> common_list_cached_models() {
-    std::vector<common_cached_model_info> models;
-    const std::string cache_dir = fs_get_cache_directory();
-    const std::vector<common_file_info> files = fs_list_files(cache_dir);
-    for (const auto & file : files) {
-        if (string_starts_with(file.name, "manifest=") && string_ends_with(file.name, ".json")) {
-            common_cached_model_info model_info;
-            model_info.manifest_path = file.path;
-            std::string fname = file.name;
-            string_replace_all(fname, ".json", ""); // remove extension
-            auto parts = string_split<std::string>(fname, '=');
-            if (parts.size() == 4) {
-                // expect format: manifest=<user>=<model>=<tag>=<other>
-                model_info.user  = parts[1];
-                model_info.model = parts[2];
-                model_info.tag   = parts[3];
-            } else {
-                // invalid format
-                continue;
-            }
-            model_info.size = 0; // TODO: get GGUF size, not manifest size
-            models.push_back(model_info);
-        }
-    }
-    return models;
-}
-
 //
 // Docker registry functions
 //
@@ -1052,3 +1027,46 @@ std::string common_docker_resolve_model(const std::string & docker) {
         throw;
     }
 }
+
+#else
+
+common_hf_file_res common_get_hf_file(const std::string &, const std::string &, bool) {
+    throw std::runtime_error("download functionality is not enabled in this build");
+}
+
+bool common_download_model(const common_params_model &, const std::string &, bool) {
+    throw std::runtime_error("download functionality is not enabled in this build");
+}
+
+std::string common_docker_resolve_model(const std::string &) {
+    throw std::runtime_error("download functionality is not enabled in this build");
+}
+
+#endif // LLAMA_USE_CURL || LLAMA_USE_HTTPLIB
+
+std::vector<common_cached_model_info> common_list_cached_models() {
+    std::vector<common_cached_model_info> models;
+    const std::string cache_dir = fs_get_cache_directory();
+    const std::vector<common_file_info> files = fs_list_files(cache_dir);
+    for (const auto & file : files) {
+        if (string_starts_with(file.name, "manifest=") && string_ends_with(file.name, ".json")) {
+            common_cached_model_info model_info;
+            model_info.manifest_path = file.path;
+            std::string fname = file.name;
+            string_replace_all(fname, ".json", ""); // remove extension
+            auto parts = string_split<std::string>(fname, '=');
+            if (parts.size() == 4) {
+                // expect format: manifest=<user>=<model>=<tag>=<other>
+                model_info.user  = parts[1];
+                model_info.model = parts[2];
+                model_info.tag   = parts[3];
+            } else {
+                // invalid format
+                continue;
+            }
+            model_info.size = 0; // TODO: get GGUF size, not manifest size
+            models.push_back(model_info);
+        }
+    }
+    return models;
+}

common/log.cpp

@@ -442,3 +442,9 @@ void common_log_set_prefix(struct common_log * log, bool prefix) {
 void common_log_set_timestamps(struct common_log * log, bool timestamps) {
     log->set_timestamps(timestamps);
 }
+
+void common_log_default_callback(enum ggml_log_level level, const char * text, void * /*user_data*/) {
+    if (LOG_DEFAULT_LLAMA <= common_log_verbosity_thold) {
+        common_log_add(common_log_main(), level, "%s", text);
+    }
+}

common/log.h

@@ -36,6 +36,8 @@ extern int common_log_verbosity_thold;
 
 void common_log_set_verbosity_thold(int verbosity); // not thread-safe
 
+void common_log_default_callback(enum ggml_log_level level, const char * text, void * user_data);
+
 // the common_log uses an internal worker thread to print/write log messages
 // when the worker thread is paused, incoming log messages are discarded
 struct common_log;

convert_hf_to_gguf.py

@@ -189,10 +189,10 @@ class ModelBase:
             return tensors
 
         prefix = "model" if not self.is_mistral_format else "consolidated"
-        part_names: list[str] = ModelBase.get_model_part_names(self.dir_model, prefix, ".safetensors")
+        part_names: set[str] = set(ModelBase.get_model_part_names(self.dir_model, prefix, ".safetensors"))
         is_safetensors: bool = len(part_names) > 0
         if not is_safetensors:
-            part_names = ModelBase.get_model_part_names(self.dir_model, "pytorch_model", ".bin")
+            part_names = set(ModelBase.get_model_part_names(self.dir_model, "pytorch_model", ".bin"))
 
         tensor_names_from_index: set[str] = set()
 
@@ -209,6 +209,7 @@ class ModelBase:
                     if weight_map is None or not isinstance(weight_map, dict):
                         raise ValueError(f"Can't load 'weight_map' from {index_name!r}")
                     tensor_names_from_index.update(weight_map.keys())
+                    part_names |= set(weight_map.values())
             else:
                 weight_map = {}
         else:
@@ -825,6 +826,15 @@ class TextModel(ModelBase):
             self.gguf_writer.add_expert_group_used_count(n_group_used)
             logger.info(f"gguf: expert groups used count = {n_group_used}")
 
+        if (score_func := self.find_hparam(["score_function", "scoring_func", "score_func"], optional=True)) is not None:
+            if score_func == "sigmoid":
+                self.gguf_writer.add_expert_gating_func(gguf.ExpertGatingFuncType.SIGMOID)
+            elif score_func == "softmax":
+                self.gguf_writer.add_expert_gating_func(gguf.ExpertGatingFuncType.SOFTMAX)
+            else:
+                raise ValueError(f"Unsupported expert score gating function value: {score_func}")
+            logger.info(f"gguf: expert score gating function = {score_func}")
+
         if (head_dim := self.hparams.get("head_dim")) is not None:
             self.gguf_writer.add_key_length(head_dim)
             self.gguf_writer.add_value_length(head_dim)
@@ -1124,6 +1134,9 @@ class TextModel(ModelBase):
         if chkhsh == "a1e163ecab2e718a4c829d1148b6e86824ec36163bb71941c3dca9cd5ac25756":
             # ref: https://huggingface.co/JetBrains/Mellum-4b-base
             res = "mellum"
+        if chkhsh == "49fc0303c9e0d2c2c565c510f64b2d9b271276acdcdadff733249eda9f7d59df":
+            # ref: https://huggingface.co/arcee-ai/Trinity-Tokenizer
+            res = "afmoe"
         if chkhsh == "9b1be57e70d20d9501b2b3186e792d81181ae36ada3903c26f9fea418cf87206":
             # ref: https://huggingface.co/inclusionAI/Ling-mini-base-2.0
             res = "bailingmoe2"
@@ -2533,6 +2546,72 @@ class ArceeModel(LlamaModel):
             self.gguf_writer.add_rope_scaling_orig_ctx_len(rope_scaling["original_max_position_embeddings"])
 
 
+@ModelBase.register("AfmoeForCausalLM")
+class AfmoeModel(LlamaModel):
+    model_arch = gguf.MODEL_ARCH.AFMOE
+
+    def set_gguf_parameters(self):
+        super().set_gguf_parameters()
+
+        # MoE parameters
+        if (n_experts := self.hparams.get("num_experts")) is not None:
+            self.gguf_writer.add_expert_count(n_experts)
+        if (n_shared_experts := self.hparams.get("num_shared_experts")) is not None:
+            self.gguf_writer.add_expert_shared_count(n_shared_experts)
+        if (moe_intermediate_size := self.hparams.get("moe_intermediate_size")) is not None:
+            self.gguf_writer.add_expert_feed_forward_length(moe_intermediate_size)
+        if (n_dense_layers := self.hparams.get("num_dense_layers")) is not None:
+            self.gguf_writer.add_leading_dense_block_count(n_dense_layers)
+
+        # Route normalization and scaling
+        if (route_norm := self.hparams.get("route_norm")) is not None:
+            self.gguf_writer.add_expert_weights_norm(route_norm)
+        if (route_scale := self.hparams.get("route_scale")) is not None:
+            self.gguf_writer.add_expert_weights_scale(route_scale)
+
+        # Sliding window attention
+        if (sliding_window := self.hparams.get("sliding_window")) is not None:
+            self.gguf_writer.add_sliding_window(sliding_window)
+
+    def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None) -> Iterable[tuple[str, Tensor]]:
+        # Handle expert weights - they're already merged in the HF format
+        # process the experts separately
+        if name.find("mlp.experts") != -1:
+            n_experts = self.hparams["num_experts"]
+            assert bid is not None
+
+            if self._experts is None:
+                self._experts = [{} for _ in range(self.block_count)]
+
+            self._experts[bid][name] = data_torch
+
+            if len(self._experts[bid]) >= n_experts * 3:
+                tensors: list[tuple[str, Tensor]] = []
+
+                # merge the experts into a single 3d tensor
+                for w_name in ["gate_proj", "up_proj", "down_proj"]:
+                    datas: list[Tensor] = []
+
+                    for xid in range(n_experts):
+                        ename_to_retrieve = f"model.layers.{bid}.mlp.experts.{xid}.{w_name}.weight"
+                        datas.append(self._experts[bid][ename_to_retrieve])
+                        del self._experts[bid][ename_to_retrieve]
+
+                    data_torch = torch.stack(datas, dim=0)
+                    merged_name = f"model.layers.{bid}.mlp.experts.{w_name}.weight"
+                    new_name = self.map_tensor_name(merged_name)
+                    tensors.append((new_name, data_torch))
+
+                return tensors
+            else:
+                return []
+
+        if name.endswith(".expert_bias"):
+            name = name.replace(".expert_bias", ".expert_bias.bias")
+
+        return [(self.map_tensor_name(name), data_torch)]
+
+
 @ModelBase.register(
     "LlavaForConditionalGeneration", # pixtral
     "Mistral3ForConditionalGeneration", # mistral small 3.1
@@ -7104,13 +7183,6 @@ class DeepseekV2Model(TextModel):
         self.gguf_writer.add_expert_weights_scale(hparams["routed_scaling_factor"])
         self.gguf_writer.add_expert_weights_norm(hparams["norm_topk_prob"])
 
-        if hparams["scoring_func"] == "sigmoid":
-            self.gguf_writer.add_expert_gating_func(gguf.ExpertGatingFuncType.SIGMOID)
-        elif hparams["scoring_func"] == "softmax":
-            self.gguf_writer.add_expert_gating_func(gguf.ExpertGatingFuncType.SOFTMAX)
-        else:
-            raise ValueError(f"Unsupported scoring_func value: {hparams['scoring_func']}")
-
         self.gguf_writer.add_rope_dimension_count(hparams["qk_rope_head_dim"])
 
         rope_scaling = self.hparams.get("rope_scaling") or {}
@@ -7216,12 +7288,6 @@ class MiniMaxM2Model(TextModel):
 
     def set_gguf_parameters(self):
         super().set_gguf_parameters()
-        if self.hparams["scoring_func"] == "sigmoid":
-            self.gguf_writer.add_expert_gating_func(gguf.ExpertGatingFuncType.SIGMOID)
-        elif self.hparams["scoring_func"] == "softmax":
-            self.gguf_writer.add_expert_gating_func(gguf.ExpertGatingFuncType.SOFTMAX)
-        else:
-            raise ValueError(f"Unsupported scoring_func value: {self.hparams['scoring_func']}")
 
         self.gguf_writer.add_expert_feed_forward_length(self.find_hparam(["intermediate_size"]))
         self.gguf_writer.add_rope_dimension_count(self.find_hparam(["rotary_dim"]))
@@ -7314,11 +7380,6 @@ class Dots1Model(Qwen2MoeModel):
         self.gguf_writer.add_expert_weights_scale(self.hparams["routed_scaling_factor"])
         self.gguf_writer.add_expert_weights_norm(self.hparams["norm_topk_prob"])
 
-        if self.hparams["scoring_func"] == "noaux_tc":
-            self.gguf_writer.add_expert_gating_func(gguf.ExpertGatingFuncType.SIGMOID)
-        else:
-            raise ValueError(f"Unsupported scoring_func value: {self.hparams['scoring_func']}")
-
     def modify_tensors(self, data_torch: Tensor, name: str, bid: int | None):
         if name.endswith("e_score_correction_bias"):
             name = name.replace("e_score_correction_bias", "e_score_correction.bias")
@@ -7779,12 +7840,6 @@ class Glm4MoeModel(TextModel):
         special_vocab._set_special_token("unk", tokenizer.get_added_vocab()["<|endoftext|>"]) # 151329
         special_vocab._set_special_token("eom", tokenizer.get_added_vocab()["<|observation|>"])  # 151338
 
-        # Patch broken chat template
-        if isinstance(special_vocab.chat_template, str) and "visible_text(m.content).endswith" in special_vocab.chat_template:
-            special_vocab.chat_template = special_vocab.chat_template.replace(
-                """{{ visible_text(m.content) }}\n{{- '/nothink' if (enable_thinking is defined and not enable_thinking and not visible_text(m.content).endswith("/nothink")) else '' -}}""",
-                """{% set content = visible_text(m.content) %}{{ content }}\n{{- '/nothink' if (enable_thinking is defined and not enable_thinking and not content.endswith("/nothink")) else '' -}}""")
-
         special_vocab.add_to_gguf(self.gguf_writer)
 
     def set_gguf_parameters(self):
@@ -8639,13 +8694,6 @@ class BailingMoeV2Model(TextModel):
         self.gguf_writer.add_expert_shared_count(hparams["num_shared_experts"])
         self.gguf_writer.add_expert_weights_norm(hparams["norm_topk_prob"])
 
-        if hparams["score_function"] == "sigmoid":
-            self.gguf_writer.add_expert_gating_func(gguf.ExpertGatingFuncType.SIGMOID)
-        elif hparams["score_function"] == "softmax":
-            self.gguf_writer.add_expert_gating_func(gguf.ExpertGatingFuncType.SOFTMAX)
-        else:
-            raise ValueError(f"Unsupported score_function value: {hparams['score_function']}")
-
         if (nextn_layers := self.hparams.get("num_nextn_predict_layers")) is not None:
             self.gguf_writer.add_nextn_predict_layers(nextn_layers)
 
@@ -9341,16 +9389,6 @@ class HunYuanModel(TextModel):
 class SmolLM3Model(LlamaModel):
     model_arch = gguf.MODEL_ARCH.SMOLLM3
 
-    def set_vocab(self):
-        super().set_vocab()
-        # remove unsupported array slicing in chat template
-        # ref: https://huggingface.co/ggml-org/SmolLM3-3B-GGUF/discussions/1
-        from transformers import AutoTokenizer
-        tokenizer = AutoTokenizer.from_pretrained(self.dir_model)
-        if tokenizer.chat_template is not None:
-            chat_template = tokenizer.chat_template.replace("[:]", "")
-            self.gguf_writer.add_chat_template(chat_template)
-
 
 @ModelBase.register("GptOssForCausalLM")
 class GptOssModel(TextModel):

convert_hf_to_gguf_update.py

@@ -139,6 +139,7 @@ models = [
     {"name": "lfm2",             "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/LiquidAI/LFM2-Tokenizer"},
     {"name": "exaone4",          "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/LGAI-EXAONE/EXAONE-4.0-32B", },
     {"name": "mellum",           "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/JetBrains/Mellum-4b-base", },
+    {"name": "afmoe",            "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/arcee-ai/Trinity-Tokenizer", },
     {"name": "bailingmoe2",      "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/inclusionAI/Ling-mini-base-2.0", },
     {"name": "granite-docling",  "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/ibm-granite/granite-docling-258M", },
     {"name": "minimax-m2",       "tokt": TOKENIZER_TYPE.BPE, "repo": "https://huggingface.co/MiniMaxAI/MiniMax-M2", },

docs/backend/CANN.md

@@ -313,7 +313,12 @@ Converting the matmul weight format from ND to NZ to improve performance. Enable
 
 ### GGML_CANN_ACL_GRAPH
 
-Operators are executed using ACL graph execution, rather than in op-by-op (eager) mode. Enabled by default.
+Operators are executed using ACL graph execution, rather than in op-by-op (eager) mode. Enabled by default. This option is only effective if `USE_ACL_GRAPH` was enabled at compilation time. To enable it, recompile using:
+
+```sh
+cmake -B build -DGGML_CANN=on -DCMAKE_BUILD_TYPE=release -DUSE_ACL_GRAPH=ON
+cmake --build build --config release
+```
 
 ### GGML_CANN_GRAPH_CACHE_CAPACITY
 

docs/ops.md

@@ -14,103 +14,108 @@ Legend:
 
 | Operation | BLAS | CANN | CPU | CUDA | Metal | OpenCL | SYCL | Vulkan | zDNN |
 |-----------|------|------|------|------|------|------|------|------|------|
-|                              ABS | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | 🟡 | ❌ | ❌ |
+|                              ABS | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | ✅ | 🟡 | ❌ |
 |                              ACC | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ |
 |                              ADD | ❌ | ✅ | ✅ | ✅ | 🟡 | 🟡 | ✅ | ✅ | ❌ |
 |                             ADD1 | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ | ✅ | ❌ | ❌ |
-|                           ADD_ID | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
-|                           ARANGE | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ | ❌ | ❌ |
+|                           ADD_ID | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ | ❌ |
+|                           ARANGE | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ❌ | ❌ |
 |                           ARGMAX | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ |
-|                          ARGSORT | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ |
-|                             CEIL | ❌ | ❌ | ✅ | 🟡 | ❌ | ❌ | ✅ | ❌ | ❌ |
-|                            CLAMP | ❌ | ✅ | ✅ | ✅ | 🟡 | 🟡 | ✅ | 🟡 | ❌ |
+|                          ARGSORT | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | 🟡 | ❌ |
+|                             CEIL | ❌ | ❌ | ✅ | 🟡 | ❌ | ❌ | 🟡 | ❌ | ❌ |
+|                            CLAMP | ❌ | ✅ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | 🟡 | ❌ |
 |                           CONCAT | ❌ | ✅ | ✅ | 🟡 | ✅ | 🟡 | ✅ | ✅ | ❌ |
 |                             CONT | ❌ | 🟡 | ✅ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | ❌ |
-|                          CONV_2D | ❌ | ❌ | ✅ | 🟡 | ❌ | ✅ | ❌ | ✅ | ❌ |
+|                          CONV_2D | ❌ | ❌ | ✅ | ✅ | ❌ | ✅ | ❌ | ✅ | ❌ |
 |                       CONV_2D_DW | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ | ❌ |
-|                          CONV_3D | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
+|                          CONV_3D | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
 |                CONV_TRANSPOSE_1D | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ |
-|                CONV_TRANSPOSE_2D | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ |
-|                              COS | ❌ | ✅ | ✅ | ✅ | 🟡 | ❌ | ✅ | 🟡 | ❌ |
+|                CONV_TRANSPOSE_2D | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ | ❌ |
+|                              COS | ❌ | ✅ | ✅ | ✅ | 🟡 | ❌ | 🟡 | 🟡 | ❌ |
 |                      COUNT_EQUAL | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ | ✅ | ✅ | ❌ |
 |                              CPY | ❌ | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | ❌ |
 |               CROSS_ENTROPY_LOSS | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ |
 |          CROSS_ENTROPY_LOSS_BACK | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ |
+|                           CUMSUM | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
 |                    DIAG_MASK_INF | ❌ | ✅ | ✅ | ✅ | 🟡 | 🟡 | ✅ | ✅ | ❌ |
 |                              DIV | ❌ | ✅ | ✅ | ✅ | 🟡 | 🟡 | ✅ | ✅ | ❌ |
 |                              DUP | ❌ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | ✅ | 🟡 | ❌ |
-|                              ELU | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | 🟡 | ❌ | ❌ |
-|                              EXP | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | 🟡 | ❌ | ❌ |
+|                              ELU | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | ✅ | ❌ | ❌ |
+|                              EXP | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | ✅ | 🟡 | ❌ |
+|                            EXPM1 | ❌ | ❌ | ✅ | 🟡 | ❌ | ❌ | ❌ | ❌ | ❌ |
+|                             FILL | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
 |                   FLASH_ATTN_EXT | ❌ | 🟡 | ✅ | 🟡 | 🟡 | ❌ | ❌ | 🟡 | ❌ |
-|                            FLOOR | ❌ | ❌ | ✅ | 🟡 | ❌ | ❌ | ✅ | ❌ | ❌ |
+|                            FLOOR | ❌ | ❌ | ✅ | 🟡 | ❌ | ❌ | 🟡 | ❌ | ❌ |
 |                GATED_LINEAR_ATTN | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ✅ | ❌ | ❌ |
 |                            GEGLU | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | 🟡 | ❌ |
 |                        GEGLU_ERF | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | 🟡 | ❌ |
 |                      GEGLU_QUICK | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | 🟡 | ❌ |
-|                             GELU | ❌ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | ❌ |
-|                         GELU_ERF | ❌ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | ❌ |
-|                       GELU_QUICK | ❌ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | ❌ |
+|                             GELU | ❌ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | ✅ | 🟡 | ❌ |
+|                         GELU_ERF | ❌ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | ✅ | 🟡 | ❌ |
+|                       GELU_QUICK | ❌ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | ✅ | 🟡 | ❌ |
 |                         GET_ROWS | ❌ | 🟡 | ✅ | 🟡 | ✅ | 🟡 | 🟡 | 🟡 | ❌ |
 |                    GET_ROWS_BACK | ❌ | ❌ | 🟡 | 🟡 | ❌ | ❌ | ❌ | ❌ | ❌ |
 |                       GROUP_NORM | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ |
-|               GROUP_NORM_MUL_ADD | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ✅ | ❌ | ❌ |
-|                      HARDSIGMOID | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | 🟡 | ❌ | ❌ |
-|                        HARDSWISH | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | 🟡 | ❌ | ❌ |
+|               GROUP_NORM_MUL_ADD | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
+|                      HARDSIGMOID | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | ✅ | 🟡 | ❌ |
+|                        HARDSWISH | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | ✅ | 🟡 | ❌ |
 |                           IM2COL | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | ✅ | ❌ |
-|                        IM2COL_3D | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
+|                        IM2COL_3D | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ | ❌ |
 |                          L2_NORM | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ |
-|                       LEAKY_RELU | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ |
-|                              LOG | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ | ✅ | ❌ | ❌ |
-|                             MEAN | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ❌ | ❌ | ❌ |
+|                       LEAKY_RELU | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | 🟡 | ❌ |
+|                              LOG | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ | 🟡 | ✅ | ❌ |
+|                             MEAN | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ |
 |                              MUL | ❌ | ✅ | ✅ | ✅ | 🟡 | 🟡 | ✅ | ✅ | ❌ |
 |                          MUL_MAT | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 |
 |                       MUL_MAT_ID | ❌ | 🟡 | ✅ | ✅ | ✅ | 🟡 | 🟡 | ✅ | ❌ |
-|                              NEG | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | 🟡 | ❌ | ❌ |
+|                              NEG | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | ✅ | 🟡 | ❌ |
 |                             NORM | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | 🟡 | ❌ |
-|                     NORM_MUL_ADD | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ✅ | ❌ | ❌ |
+|                     NORM_MUL_ADD | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
 |                   OPT_STEP_ADAMW | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ | ❌ |
-|                     OPT_STEP_SGD | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
+|                     OPT_STEP_SGD | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ | ❌ |
 |                         OUT_PROD | 🟡 | ❌ | 🟡 | 🟡 | ❌ | ❌ | 🟡 | ❌ | ❌ |
-|                              PAD | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | 🟡 | ✅ | ❌ |
-|                   PAD_REFLECT_1D | ❌ | ✅ | ✅ | ❌ | ✅ | ❌ | ✅ | ❌ | ❌ |
+|                              PAD | ❌ | ✅ | ✅ | 🟡 | ✅ | ✅ | 🟡 | ✅ | ❌ |
+|                   PAD_REFLECT_1D | ❌ | ✅ | ✅ | ✅ | ✅ | ❌ | ✅ | ❌ | ❌ |
 |                          POOL_2D | ❌ | 🟡 | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ |
 |                            REGLU | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | 🟡 | ❌ |
-|                             RELU | ❌ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | ❌ |
+|                             RELU | ❌ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | ✅ | 🟡 | ❌ |
 |                           REPEAT | ❌ | ✅ | ✅ | 🟡 | ✅ | 🟡 | ✅ | 🟡 | ❌ |
-|                      REPEAT_BACK | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ | ❌ |
+|                      REPEAT_BACK | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ✅ | ✅ | ❌ |
 |                         RMS_NORM | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | ✅ | ❌ |
 |                    RMS_NORM_BACK | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ✅ | ✅ | ❌ |
-|                 RMS_NORM_MUL_ADD | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ |
-|                             ROLL | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ | ❌ | ✅ | ❌ |
+|                 RMS_NORM_MUL_ADD | ❌ | ✅ | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ |
+|                             ROLL | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ✅ | ✅ | ❌ |
 |                             ROPE | ❌ | 🟡 | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ |
 |                        ROPE_BACK | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ | ❌ |
-|                            ROUND | ❌ | ❌ | ✅ | 🟡 | ❌ | ❌ | ✅ | ❌ | ❌ |
+|                            ROUND | ❌ | ❌ | ✅ | 🟡 | ❌ | ❌ | 🟡 | ❌ | ❌ |
 |                        RWKV_WKV6 | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ |
 |                        RWKV_WKV7 | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ |
 |                            SCALE | ❌ | 🟡 | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ |
-|                              SET | ❌ | ❌ | ✅ | ❌ | ✅ | ❌ | ❌ | ❌ | ❌ |
+|                              SET | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | 🟡 | ❌ | ❌ |
 |                         SET_ROWS | ❌ | ❌ | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | ❌ |
-|                              SGN | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | 🟡 | ❌ | ❌ |
-|                          SIGMOID | ❌ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | ❌ |
-|                             SILU | ❌ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | 🟡 | 🟡 | ❌ |
+|                              SGN | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | ✅ | ❌ | ❌ |
+|                          SIGMOID | ❌ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | ✅ | 🟡 | ❌ |
+|                             SILU | ❌ | ✅ | ✅ | 🟡 | 🟡 | 🟡 | ✅ | 🟡 | ❌ |
 |                        SILU_BACK | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | ✅ | ❌ |
-|                              SIN | ❌ | ✅ | ✅ | ✅ | 🟡 | ❌ | ✅ | 🟡 | ❌ |
-|                          SOFTCAP | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ✅ | ❌ | ❌ |
+|                              SIN | ❌ | ✅ | ✅ | ✅ | 🟡 | ❌ | 🟡 | 🟡 | ❌ |
+|                          SOFTCAP | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
+|                         SOFTPLUS | ❌ | ❌ | ✅ | 🟡 | ❌ | ❌ | ❌ | ❌ | ❌ |
 |                         SOFT_MAX | ❌ | 🟡 | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ |
 |                    SOFT_MAX_BACK | ❌ | ❌ | 🟡 | 🟡 | ❌ | ❌ | 🟡 | ✅ | ❌ |
-|                              SQR | ❌ | ✅ | ✅ | ✅ | 🟡 | ❌ | ✅ | 🟡 | ❌ |
-|                             SQRT | ❌ | ✅ | ✅ | ✅ | 🟡 | ❌ | ✅ | ❌ | ❌ |
-|                         SSM_CONV | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ | ✅ | ❌ |
-|                         SSM_SCAN | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ | ✅ | ❌ |
-|                             STEP | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | 🟡 | ❌ | ❌ |
+|                        SOLVE_TRI | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
+|                              SQR | ❌ | ✅ | ✅ | ✅ | 🟡 | ❌ | 🟡 | 🟡 | ❌ |
+|                             SQRT | ❌ | ✅ | ✅ | ✅ | 🟡 | ❌ | 🟡 | 🟡 | ❌ |
+|                         SSM_CONV | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ✅ | ✅ | ❌ |
+|                         SSM_SCAN | ❌ | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ | 🟡 | ❌ |
+|                             STEP | ❌ | ✅ | ✅ | 🟡 | 🟡 | ❌ | ✅ | ❌ | ❌ |
 |                              SUB | ❌ | ✅ | ✅ | ✅ | 🟡 | 🟡 | ✅ | ✅ | ❌ |
-|                              SUM | ❌ | ✅ | ✅ | ✅ | ❌ | ❌ | ✅ | ✅ | ❌ |
-|                         SUM_ROWS | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | 🟡 | ✅ | ❌ |
+|                              SUM | ❌ | ✅ | ✅ | 🟡 | ❌ | ❌ | 🟡 | 🟡 | ❌ |
+|                         SUM_ROWS | ❌ | ✅ | ✅ | 🟡 | ✅ | ✅ | 🟡 | ✅ | ❌ |
 |                           SWIGLU | ❌ | ✅ | ✅ | ✅ | 🟡 | ✅ | ✅ | 🟡 | ❌ |
-|                       SWIGLU_OAI | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
-|                             TANH | ❌ | ✅ | ✅ | 🟡 | 🟡 | ✅ | 🟡 | 🟡 | ❌ |
+|                       SWIGLU_OAI | ❌ | ❌ | ✅ | ✅ | ❌ | ❌ | ❌ | 🟡 | ❌ |
+|                             TANH | ❌ | ✅ | ✅ | 🟡 | 🟡 | ✅ | ✅ | 🟡 | ❌ |
 |               TIMESTEP_EMBEDDING | ❌ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ✅ | ❌ |
-|                         TOPK_MOE | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ✅ | ❌ | ❌ |
-|                            TRUNC | ❌ | ❌ | ✅ | 🟡 | ❌ | ❌ | ✅ | ❌ | ❌ |
+|                              TRI | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
+|                            TRUNC | ❌ | ❌ | ✅ | 🟡 | ❌ | ❌ | 🟡 | ❌ | ❌ |
 |                          UPSCALE | ❌ | 🟡 | ✅ | ✅ | 🟡 | ✅ | 🟡 | ✅ | ❌ |
-|                            XIELU | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |
+|                            XIELU | ❌ | ❌ | ✅ | ❌ | ❌ | ❌ | ❌ | ❌ | ❌ |

ggml/include/ggml.h

@@ -475,6 +475,7 @@ extern "C" {
         GGML_OP_COS,
         GGML_OP_SUM,
         GGML_OP_SUM_ROWS,
+        GGML_OP_CUMSUM,
         GGML_OP_MEAN,
         GGML_OP_ARGMAX,
         GGML_OP_COUNT_EQUAL,
@@ -530,6 +531,8 @@ extern "C" {
         GGML_OP_TIMESTEP_EMBEDDING,
         GGML_OP_ARGSORT,
         GGML_OP_LEAKY_RELU,
+        GGML_OP_TRI,
+        GGML_OP_FILL,
 
         GGML_OP_FLASH_ATTN_EXT,
         GGML_OP_FLASH_ATTN_BACK,
@@ -542,6 +545,7 @@ extern "C" {
         GGML_OP_RWKV_WKV6,
         GGML_OP_GATED_LINEAR_ATTN,
         GGML_OP_RWKV_WKV7,
+        GGML_OP_SOLVE_TRI,
 
         GGML_OP_UNARY,
 
@@ -576,6 +580,8 @@ extern "C" {
         GGML_UNARY_OP_HARDSWISH,
         GGML_UNARY_OP_HARDSIGMOID,
         GGML_UNARY_OP_EXP,
+        GGML_UNARY_OP_EXPM1,
+        GGML_UNARY_OP_SOFTPLUS,
         GGML_UNARY_OP_GELU_ERF,
         GGML_UNARY_OP_XIELU,
         GGML_UNARY_OP_FLOOR,
@@ -620,6 +626,13 @@ extern "C" {
         GGML_TENSOR_FLAG_LOSS   =  8, // ...defines loss for numerical optimization (multiple loss tensors add up)
     };
 
+    enum ggml_tri_type {
+        GGML_TRI_TYPE_UPPER_DIAG = 0,
+        GGML_TRI_TYPE_UPPER      = 1,
+        GGML_TRI_TYPE_LOWER_DIAG = 2,
+        GGML_TRI_TYPE_LOWER      = 3
+    };
+
     struct ggml_init_params {
         // memory pool
         size_t mem_size;   // bytes
@@ -957,6 +970,22 @@ extern "C" {
             struct ggml_context * ctx,
             struct ggml_tensor  * a);
 
+    GGML_API struct ggml_tensor * ggml_expm1(
+            struct ggml_context * ctx,
+            struct ggml_tensor  * a);
+
+    GGML_API struct ggml_tensor * ggml_expm1_inplace(
+            struct ggml_context * ctx,
+            struct ggml_tensor  * a);
+
+    GGML_API struct ggml_tensor * ggml_softplus(
+            struct ggml_context * ctx,
+            struct ggml_tensor  * a);
+
+    GGML_API struct ggml_tensor * ggml_softplus_inplace(
+            struct ggml_context * ctx,
+            struct ggml_tensor  * a);
+
     GGML_API struct ggml_tensor * ggml_sin(
             struct ggml_context * ctx,
             struct ggml_tensor  * a);
@@ -983,6 +1012,10 @@ extern "C" {
             struct ggml_context * ctx,
             struct ggml_tensor  * a);
 
+    GGML_API struct ggml_tensor * ggml_cumsum(
+        struct ggml_context * ctx,
+        struct ggml_tensor  * a);
+
     // mean along rows
     GGML_API struct ggml_tensor * ggml_mean(
             struct ggml_context * ctx,
@@ -2187,6 +2220,23 @@ extern "C" {
             int                   shift2,
             int                   shift3);
 
+    // Convert matrix into a triangular one (upper, strict upper, lower or strict lower) by writing
+    // zeroes everywhere outside the masked area
+    GGML_API struct ggml_tensor * ggml_tri(
+            struct ggml_context * ctx,
+            struct ggml_tensor  * a,
+            enum ggml_tri_type    type);
+
+    // Fill tensor a with constant c
+    GGML_API struct ggml_tensor * ggml_fill(
+            struct ggml_context * ctx,
+            struct ggml_tensor  * a,
+            float                 c);
+
+    GGML_API struct ggml_tensor * ggml_fill_inplace(
+            struct ggml_context * ctx,
+            struct ggml_tensor  * a,
+            float                 c);
 
     // Ref: https://github.com/CompVis/stable-diffusion/blob/main/ldm/modules/diffusionmodules/util.py#L151
     // timesteps: [N,]
@@ -2356,6 +2406,27 @@ extern "C" {
             struct ggml_tensor  * b,
             struct ggml_tensor  * state);
 
+    /* Solves a specific equation of the form Ax=B, where A is a triangular matrix
+    *  without zeroes on the diagonal (i.e. invertible).
+    *  B can have any number of columns, but must have the same number of rows as A
+    *  If A is [n, n] and B is [n, m], then the result will be [n, m] as well
+    *  Has O(n^3) complexity (unlike most matrix ops out there), so use on cases
+    *  where n > 100 sparingly, pre-chunk if necessary.
+    *
+    *  If left = false, solves xA=B instead
+    *  If lower = false, assumes upper triangular instead
+    *  If uni = true, assumes diagonal of A to be all ones (will override actual values)
+    *
+    *  TODO: currently only lower, right, non-unitriangular variant is implemented
+    */
+    GGML_API struct ggml_tensor * ggml_solve_tri(
+        struct ggml_context * ctx,
+        struct ggml_tensor  * a,
+        struct ggml_tensor  * b,
+        bool                  left,
+        bool                  lower,
+        bool                  uni);
+
     // custom operators
 
     typedef void (*ggml_custom1_op_t)(struct ggml_tensor * dst , const struct ggml_tensor * a, int ith, int nth, void * userdata);

ggml/src/ggml-backend.cpp

@@ -1698,8 +1698,6 @@ bool ggml_backend_sched_reserve(ggml_backend_sched_t sched, struct ggml_cgraph *
     GGML_ASSERT(sched);
     GGML_ASSERT((int)sched->hash_set.size >= measure_graph->n_nodes + measure_graph->n_leafs);
 
-    ggml_backend_sched_reset(sched);
-
     ggml_backend_sched_synchronize(sched);
 
     ggml_backend_sched_split_graph(sched, measure_graph);

ggml/src/ggml-cann/acl_tensor.cpp

@@ -48,15 +48,14 @@ aclDataType ggml_cann_type_mapping(ggml_type type) {
         default:
             return ACL_DT_UNDEFINED;
     }
-    return ACL_DT_UNDEFINED;
 }
 
-aclTensor * ggml_cann_create_tensor(const ggml_tensor * tensor,
-                                    int64_t *           ne,
-                                    size_t *            nb,
-                                    int64_t             dims,
-                                    aclFormat           format,
-                                    size_t              offset) {
+acl_tensor_ptr ggml_cann_create_tensor(const ggml_tensor * tensor,
+                                       int64_t *           ne,
+                                       size_t *            nb,
+                                       int64_t             dims,
+                                       aclFormat           format,
+                                       size_t              offset) {
     // If tensor is bcasted, Up to GGML_MAX_DIMS additional dimensions will be
     // added.
     int64_t acl_ne[GGML_MAX_DIMS * 2], acl_stride[GGML_MAX_DIMS * 2];
@@ -87,10 +86,20 @@ aclTensor * ggml_cann_create_tensor(const ggml_tensor * tensor,
     std::reverse(acl_ne, acl_ne + final_dims);
     std::reverse(acl_stride, acl_stride + final_dims);
 
-    aclTensor * acl_tensor = aclCreateTensor(acl_ne, final_dims, ggml_cann_type_mapping(tensor->type), acl_stride,
-                                             elem_offset, format, &acl_storage_len, 1, tensor->data);
+    aclTensor * raw = aclCreateTensor(acl_ne, final_dims, ggml_cann_type_mapping(tensor->type), acl_stride, elem_offset,
+                                      format, &acl_storage_len, 1, tensor->data);
 
-    return acl_tensor;
+    return acl_tensor_ptr(raw);
+}
+
+acl_int_array_ptr ggml_cann_create_int_array(const int64_t * value, uint64_t size) {
+    aclIntArray * raw = aclCreateIntArray(value, size);
+    return acl_int_array_ptr(raw);
+}
+
+acl_scalar_ptr ggml_cann_create_scalar(void * value, aclDataType dataType) {
+    aclScalar * raw = aclCreateScalar(value, dataType);
+    return acl_scalar_ptr(raw);
 }
 
 bool ggml_cann_need_bcast(const ggml_tensor * t0, const ggml_tensor * t1) {

ggml/src/ggml-cann/acl_tensor.h

@@ -23,11 +23,12 @@
 #ifndef CANN_ACL_TENSOR_H
 #define CANN_ACL_TENSOR_H
 
-#include <algorithm>
-#include <cstring>
+#include "common.h"
 
 #include <aclnn/aclnn_base.h>
-#include "common.h"
+
+#include <algorithm>
+#include <cstring>
 
 /**
  * @brief	Maps a ggml_type to its corresponding aclDataType.
@@ -43,6 +44,20 @@
  */
 aclDataType ggml_cann_type_mapping(ggml_type type);
 
+// Deleter for acl objects.
+template <typename T, aclError (*DestroyFunc)(const T *)> struct acl_deleter {
+    void operator()(T * ptr) const noexcept {
+        if (ptr) {
+            ACL_CHECK(DestroyFunc(ptr));
+        }
+    }
+};
+
+using acl_tensor_ptr      = std::unique_ptr<aclTensor, acl_deleter<aclTensor, aclDestroyTensor>>;
+using acl_int_array_ptr   = std::unique_ptr<aclIntArray, acl_deleter<aclIntArray, aclDestroyIntArray>>;
+using acl_scalar_ptr      = std::unique_ptr<aclScalar, acl_deleter<aclScalar, aclDestroyScalar>>;
+using acl_tensor_list_ptr = std::unique_ptr<aclTensorList, acl_deleter<aclTensorList, aclDestroyTensorList>>;
+
 /**
  * @brief   Creates an ACL tensor from a ggml_tensor with optional shape.
  *
@@ -62,12 +77,12 @@ aclDataType ggml_cann_type_mapping(ggml_type type);
  * @param   offset      Offset in bytes for the ACL tensor data. Defaults to 0.
  * @return  Pointer to the created ACL tensor.
  */
-aclTensor * ggml_cann_create_tensor(const ggml_tensor * tensor,
-                                    int64_t *           ne     = nullptr,
-                                    size_t *            nb     = nullptr,
-                                    int64_t             dims   = 0,
-                                    aclFormat           format = ACL_FORMAT_ND,
-                                    size_t              offset = 0);
+acl_tensor_ptr ggml_cann_create_tensor(const ggml_tensor * tensor,
+                                       int64_t *           ne     = nullptr,
+                                       size_t *            nb     = nullptr,
+                                       int64_t             dims   = 0,
+                                       aclFormat           format = ACL_FORMAT_ND,
+                                       size_t              offset = 0);
 
 /**
  * @brief   Template for creating an ACL tensor from provided parameters. typename TYPE
@@ -90,14 +105,14 @@ aclTensor * ggml_cann_create_tensor(const ggml_tensor * tensor,
  * @return  Pointer to the created ACL tensor.
  */
 template <typename TYPE>
-aclTensor * ggml_cann_create_tensor(void *      data_ptr,
-                                    aclDataType dtype,
-                                    TYPE        type_size,
-                                    int64_t *   ne,
-                                    TYPE *      nb,
-                                    int64_t     dims,
-                                    aclFormat   format = ACL_FORMAT_ND,
-                                    size_t      offset = 0) {
+acl_tensor_ptr ggml_cann_create_tensor(void *      data_ptr,
+                                       aclDataType dtype,
+                                       TYPE        type_size,
+                                       int64_t *   ne,
+                                       TYPE *      nb,
+                                       int64_t     dims,
+                                       aclFormat   format = ACL_FORMAT_ND,
+                                       size_t      offset = 0) {
     int64_t tmp_ne[GGML_MAX_DIMS * 2];
     int64_t tmp_stride[GGML_MAX_DIMS * 2];
 
@@ -114,10 +129,75 @@ aclTensor * ggml_cann_create_tensor(void *      data_ptr,
     std::reverse(tmp_ne, tmp_ne + dims);
     std::reverse(tmp_stride, tmp_stride + dims);
 
-    aclTensor * acl_tensor =
+    aclTensor * raw =
         aclCreateTensor(tmp_ne, dims, dtype, tmp_stride, offset / type_size, format, &acl_storage_len, 1, data_ptr);
 
-    return acl_tensor;
+    return acl_tensor_ptr(raw);
+}
+
+/**
+ * @brief Create an ACL int array resource wrapped in a smart pointer.
+ *
+ * This function constructs an aclIntArray from the provided int64_t values
+ * and returns it as an acl_int_array_ptr (a std::unique_ptr with a custom
+ * deleter). The returned pointer owns the ACL resource and will automatically
+ * destroy it via aclDestroyIntArray().
+ *
+ * @param value  Pointer to the int64_t elements.
+ * @param size   Number of elements in value.
+ *
+ * @return A smart pointer managing the created ACL int array.
+ */
+acl_int_array_ptr ggml_cann_create_int_array(const int64_t * value, uint64_t size);
+
+/**
+ * @brief Create an ACL scalar resource wrapped in a smart pointer.
+ *
+ * This function constructs an aclScalar from the raw value pointer and ACL
+ * data type, then returns it as an acl_scalar_ptr (a std::unique_ptr with
+ * a custom deleter). The returned pointer owns the ACL scalar and will
+ * automatically destroy it via aclDestroyScalar().
+ *
+ * @param value     Pointer to the raw scalar memory.
+ * @param dataType  ACL data type of the scalar.
+ *
+ * @return A smart pointer managing the created ACL scalar.
+ */
+acl_scalar_ptr ggml_cann_create_scalar(void * value, aclDataType dataType);
+
+/**
+ * @brief Create an ACL tensor list from multiple tensor smart pointers.
+ *
+ * This function accepts a variadic list of acl_tensor_ptr (a unique_ptr with
+ * custom deleter) and produces an aclTensorList using aclCreateTensorList().
+ *
+ * The lifecycle management of the tensor objects changes as follows:
+ *  - aclCreateTensorList() takes ownership of the tensors
+ *  - Each input smart pointer releases ownership using release()
+ *  - As a result, the tensors will NOT be destroyed by unique_ptr
+ *  - Instead, they will be destroyed when aclDestroyTensorList() is called
+ *
+ * This ensures correct ownership transfer and prevents double-free situations.
+ *
+ * @param acl_tensor_ptr  Variadic template parameter; each argument must be
+ *                         a unique_ptr-like type supporting get() and release().
+ *
+ * @param tensors  Variadic list of acl_tensor_ptr objects. Ownership of
+ *                         each tensor is transferred away from these smart pointers.
+ *
+ * @return A smart pointer (acl_tensor_list_ptr) owning the created ACL tensor list.
+ *
+ * @note This implementation is C++11 compatible. The ownership-release process is
+ *       executed using a pack expansion inside an initializer list.
+ */
+template <typename... acl_tensor_ptr> acl_tensor_list_ptr ggml_cann_create_tensor_list(acl_tensor_ptr &&... tensors) {
+    aclTensor *     raw_tensors[] = { tensors.get()... };
+    aclTensorList * raw           = aclCreateTensorList(raw_tensors, sizeof...(tensors));
+    // aclTensor will release by aclTensorList, so release ownership without
+    // destroying the tensor
+    int             dummy[]       = { (tensors.release(), 0)... };
+    GGML_UNUSED(dummy);
+    return acl_tensor_list_ptr(raw);
 }
 
 /**

ggml/src/ggml-cann/aclnn_ops.h

@@ -23,31 +23,35 @@
 #ifndef CANN_ACLNN_OPS
 #define CANN_ACLNN_OPS
 
-#include <unordered_set>
-#include <functional>
+#include "acl_tensor.h"
+#include "common.h"
+
 #include <aclnnop/aclnn_abs.h>
-#include <aclnnop/aclnn_neg.h>
-#include <aclnnop/aclnn_exp.h>
 #include <aclnnop/aclnn_arange.h>
 #include <aclnnop/aclnn_argsort.h>
 #include <aclnnop/aclnn_cat.h>
 #include <aclnnop/aclnn_clamp.h>
+#include <aclnnop/aclnn_cos.h>
+#include <aclnnop/aclnn_exp.h>
 #include <aclnnop/aclnn_gelu.h>
 #include <aclnnop/aclnn_gelu_v2.h>
-#include <aclnnop/aclnn_sigmoid.h>
 #include <aclnnop/aclnn_hardsigmoid.h>
 #include <aclnnop/aclnn_hardswish.h>
 #include <aclnnop/aclnn_leaky_relu.h>
-#include <aclnnop/aclnn_relu.h>
-#include <aclnnop/aclnn_silu.h>
-#include <aclnnop/aclnn_tanh.h>
-#include <aclnnop/aclnn_sqrt.h>
-#include <aclnnop/aclnn_sin.h>
-#include <aclnnop/aclnn_cos.h>
 #include <aclnnop/aclnn_log.h>
+#include <aclnnop/aclnn_logsoftmax.h>
+#include <aclnnop/aclnn_neg.h>
+#include <aclnnop/aclnn_norm.h>
+#include <aclnnop/aclnn_relu.h>
+#include <aclnnop/aclnn_sigmoid.h>
 #include <aclnnop/aclnn_sign.h>
-#include "acl_tensor.h"
-#include "common.h"
+#include <aclnnop/aclnn_silu.h>
+#include <aclnnop/aclnn_sin.h>
+#include <aclnnop/aclnn_sqrt.h>
+#include <aclnnop/aclnn_tanh.h>
+
+#include <functional>
+#include <unordered_set>
 
 /**
  * @brief   Repeats a ggml tensor along each dimension to match the dimensions
@@ -187,6 +191,66 @@ void ggml_cann_argsort(ggml_backend_cann_context & ctx, ggml_tensor * dst);
  */
 void ggml_cann_norm(ggml_backend_cann_context & ctx, ggml_tensor * dst);
 
+/**
+ * @brief   Computes the L2 Normalization for a ggml tensor using the CANN
+ *          backend.
+ *
+ * @details This function applies the L2 Normalization operation on the
+ *          input tensor `src` and stores the result in the destination tensor
+ *          `dst`. L2 Normalization scales the input tensor such that the
+ *          L2 norm along the specified dimension equals 1. This operation
+ *          is commonly used in neural networks for feature normalization
+ *          and vector scaling.
+ *          The operation is defined as:
+ *          \f[
+ *              \text{out} = \frac{x}{\sqrt{\sum{x^2}}}
+ *          \f]
+ *          The normalization is performed along the last dimension by default.
+ *
+ * @param ctx The CANN context used for operations.
+ * @param dst The destination tensor where the normalized values will be stored.
+ * @attention The normalization is performed along the last dimension of the
+ *            input tensor by default.
+ */
+void ggml_cann_l2_norm(ggml_backend_cann_context & ctx, ggml_tensor * dst);
+
+/**
+ * @brief   Computes the Cross Entropy Loss for a ggml tensor using the CANN
+ *          backend.
+ *
+ * @details This function computes the cross entropy loss between the predicted
+ *          logits and target probability distributions. The operation follows
+ *          the same computation pattern as the CPU implementation:
+ *          1. Applies log_softmax to the logits along the class dimension
+ *          2. Element-wise multiplication with target distributions
+ *          3. Summation along the class dimension to get per-sample losses
+ *          4. Global summation and scaling by -1/nr to get final loss
+ *
+ *          The computation can be expressed as:
+ *          \f[
+ *              \text{loss} = -\frac{1}{N} \sum_{i=1}^{N} \sum_{j=1}^{C} y_{ij} \cdot \log(\text{softmax}(x_{ij}))
+ *          \f]
+ *          where \f$N\f$ is the total number of samples, \f$C\f$ is the number
+ *          of classes, \f$x\f$ are the logits, and \f$y\f$ are the target
+ *          probability distributions.
+ *
+ * @param ctx The CANN context used for operations.
+ * @param dst The destination tensor where the computed loss will be stored.
+ *            This should be a scalar tensor containing the final loss value.
+ *
+ * @note This implementation computes cross entropy between probability
+ *       distributions, not the typical classification cross entropy that
+ *       expects class indices as targets. Both input tensors (src0 and src1)
+ *       should have the same shape and represent probability distributions
+ *       over the class dimension.
+ * @note The function expects two source tensors:
+ *       - dst->src[0]: Logits tensor (before softmax)
+ *       - dst->src[1]: Target probability distributions tensor
+ * @note The computation is performed using CANN backend operators including
+ *       LogSoftmax, Mul, ReduceSum, and Muls for the final scaling.
+ */
+void ggml_cann_cross_entropy_loss(ggml_backend_cann_context & ctx, ggml_tensor * dst);
+
 /**
  * @brief  Computes the Group Normalization for a ggml tensor using the CANN
  *         backend.
@@ -626,12 +690,12 @@ void aclnn_sin(ggml_backend_cann_context & ctx, aclTensor * acl_src, aclTensor *
  * @param acl_src1 Output pointer to the created ACL tensor corresponding to src1.
  * @param acl_dst  Output pointer to the created ACL tensor corresponding to dst.
  */
-void bcast_shape(ggml_tensor * src0,
-                 ggml_tensor * src1,
-                 ggml_tensor * dst,
-                 aclTensor **  acl_src0,
-                 aclTensor **  acl_src1,
-                 aclTensor **  acl_dst);
+void bcast_shape(ggml_tensor *    src0,
+                 ggml_tensor *    src1,
+                 ggml_tensor *    dst,
+                 acl_tensor_ptr & acl_src0,
+                 acl_tensor_ptr & acl_src1,
+                 acl_tensor_ptr & acl_dst);
 
 /**
  * @brief   Computes the 1D transposed convolution (deconvolution) of a ggml
@@ -811,83 +875,6 @@ template <typename... Args> void register_acl_resources(std::vector<any_acl_reso
     (vec.emplace_back(make_acl_resource(args)), ...);
 }
 
-/**
- * @brief Task class that wraps the execution of an aclnn function call.
- */
-class aclnn_task : public cann_task {
-  public:
-    aclnn_task(aclnn_func_t    aclnn_func,
-               void *          workspace_addr,
-               uint64_t        workspace_size,
-               aclOpExecutor * executor,
-               aclrtStream     stream) :
-        aclnn_func_(aclnn_func),
-        workspace_addr_(workspace_addr),
-        workspace_size_(workspace_size),
-        executor_(executor),
-        stream_(stream) {}
-
-    virtual void run_task() override { ACL_CHECK(aclnn_func_(workspace_addr_, workspace_size_, executor_, stream_)); }
-  private:
-    aclnn_func_t    aclnn_func_;
-    void *          workspace_addr_;
-    uint64_t        workspace_size_;
-    aclOpExecutor * executor_;
-    aclrtStream     stream_;
-};
-
-/**
- * @brief Task class that releases ACL resources after usage.
- */
-class release_resource_task : public cann_task {
-  public:
-    release_resource_task(std::vector<any_acl_resource> && resources) { resource_ = std::move(resources); }
-
-    virtual void run_task() override { resource_.clear(); }
-  private:
-    std::vector<any_acl_resource> resource_;
-};
-
-/**
- * @brief Task class for performing asynchronous memory copy operations.
- */
-class async_memcpy_task : public cann_task {
-  public:
-    async_memcpy_task(void * dst, const void * src, size_t size, aclrtMemcpyKind kind, aclrtStream stream) :
-        dst_(dst),
-        src_(src),
-        size_(size),
-        kind_(kind),
-        stream_(stream) {}
-
-    virtual void run_task() override { ACL_CHECK(aclrtMemcpyAsync(dst_, size_, src_, size_, kind_, stream_)); }
-  private:
-    void *          dst_;
-    const void *    src_;
-    size_t          size_;
-    aclrtMemcpyKind kind_;
-    aclrtStream     stream_;
-};
-
-/**
- * @brief Task class for performing asynchronous memory set operations.
- */
-class async_memset_task : public cann_task {
-  public:
-    async_memset_task(void * buffer, size_t size, int32_t value, aclrtStream stream) :
-        buffer_(buffer),
-        size_(size),
-        value_(value),
-        stream_(stream) {}
-
-    virtual void run_task() override { ACL_CHECK(aclrtMemsetAsync(buffer_, size_, value_, size_, stream_)); }
-  private:
-    void *      buffer_;
-    size_t      size_;
-    int32_t     value_;
-    aclrtStream stream_;
-};
-
 /**
  * @brief Launches an asynchronous task using the memory allocator.
  *
@@ -906,95 +893,20 @@ class async_memset_task : public cann_task {
  * same stream are executed in queue order.
  */
 
-#define GGML_CANN_CALL_ACLNN_OP(CTX, OP_NAME, ...)                                                                  \
-    do {                                                                                                            \
-        uint64_t        workspaceSize = 0;                                                                          \
-        aclOpExecutor * executor;                                                                                   \
-        void *          workspaceAddr = nullptr;                                                                    \
-        ACL_CHECK(aclnn##OP_NAME##GetWorkspaceSize(__VA_ARGS__, &workspaceSize, &executor));                        \
-        /* workspace should alloced in main thread to keep malloc order when using vmm. */                          \
-        if (workspaceSize > 0) {                                                                                    \
-            ggml_cann_pool_alloc workspace_allocator(CTX.pool(), workspaceSize);                                    \
-            workspaceAddr = workspace_allocator.get();                                                              \
-        }                                                                                                           \
-        if (CTX.async_mode) {                                                                                       \
-            auto task =                                                                                             \
-                std::make_unique<aclnn_task>(aclnn##OP_NAME, workspaceAddr, workspaceSize, executor, CTX.stream()); \
-            CTX.task_queue.submit_task(std::move(task));                                                            \
-        } else {                                                                                                    \
-            ACL_CHECK(aclnn##OP_NAME(workspaceAddr, workspaceSize, executor, CTX.stream()));                        \
-        }                                                                                                           \
+#define GGML_CANN_CALL_ACLNN_OP(CTX, OP_NAME, ...)                                           \
+    do {                                                                                     \
+        uint64_t        workspaceSize = 0;                                                   \
+        aclOpExecutor * executor;                                                            \
+        void *          workspaceAddr = nullptr;                                             \
+        ACL_CHECK(aclnn##OP_NAME##GetWorkspaceSize(__VA_ARGS__, &workspaceSize, &executor)); \
+        /* workspace should alloced in main thread to keep malloc order when using vmm. */   \
+        if (workspaceSize > 0) {                                                             \
+            ggml_cann_pool_alloc workspace_allocator(CTX.pool(), workspaceSize);             \
+            workspaceAddr = workspace_allocator.get();                                       \
+        }                                                                                    \
+        ACL_CHECK(aclnn##OP_NAME(workspaceAddr, workspaceSize, executor, CTX.stream()));     \
     } while (0)
 
-/**
- * @brief Registers and releases multiple ACL resources, optionally deferring the release
- *        using a task.
- *
- * @tparam Args Types of the ACL resources.
- * @param ctx Backend context which manages task submission and async mode.
- * @param args Pointers to ACL resources to be released.
- */
-template <typename... Args> void ggml_cann_release_resources(ggml_backend_cann_context & ctx, Args &&... args) {
-    std::vector<any_acl_resource> resources;
-    register_acl_resources(resources, std::forward<Args>(args)...);
-    if (ctx.async_mode) {
-        auto task = std::make_unique<release_resource_task>(std::move(resources));
-        ctx.task_queue.submit_task(std::move(task));
-    }
-}
-
-/**
- * @brief Performs an asynchronous memory copy operation, optionally deferred via task submission.
- *
- * @param ctx Backend context containing stream and async configuration.
- * @param dst Destination memory address.
- * @param src Source memory address.
- * @param len Size of memory to copy (in bytes).
- * @param kind Type of memory copy (host-to-device, device-to-host, etc).
- */
-inline void ggml_cann_async_memcpy(ggml_backend_cann_context & ctx,
-                                   void *                      dst,
-                                   const void *                src,
-                                   size_t                      len,
-                                   aclrtMemcpyKind             kind) {
-    if (ctx.async_mode) {
-        auto task = std::make_unique<async_memcpy_task>(dst, const_cast<void *>(src), len, kind, ctx.stream());
-        ctx.task_queue.submit_task(std::move(task));
-    } else {
-        ACL_CHECK(aclrtMemcpyAsync(dst, len, src, len, kind, ctx.stream()));
-    }
-}
-
-inline void ggml_cann_async_memcpy(ggml_backend_cann_context * ctx,
-                                   void *                      dst,
-                                   const void *                src,
-                                   size_t                      len,
-                                   aclrtMemcpyKind             kind) {
-    if (ctx->async_mode) {
-        auto task = std::make_unique<async_memcpy_task>(dst, const_cast<void *>(src), len, kind, ctx->stream());
-        ctx->task_queue.submit_task(std::move(task));
-    } else {
-        ACL_CHECK(aclrtMemcpyAsync(dst, len, src, len, kind, ctx->stream()));
-    }
-}
-
-/**
- * @brief Performs an asynchronous memory set operation, optionally deferred via task submission.
- *
- * @param ctx Backend context containing stream and async configuration.
- * @param buffer Memory buffer to be set.
- * @param size Size of the memory buffer (in bytes).
- * @param value Value to set in the buffer.
- */
-inline void ggml_cann_async_memset(ggml_backend_cann_context & ctx, void * buffer, size_t size, int value) {
-    if (ctx.async_mode) {
-        auto task = std::make_unique<async_memset_task>(buffer, size, value, ctx.stream());
-        ctx.task_queue.submit_task(std::move(task));
-    } else {
-        ACL_CHECK(aclrtMemsetAsync(buffer, size, value, size, ctx.stream()));
-    }
-}
-
 /**
  * @brief   Performs sparse expert-based matrix multiplication using the CANN backend.
  *
@@ -1067,15 +979,11 @@ template <auto binary_op> void ggml_cann_binary_op(ggml_backend_cann_context & c
     ggml_tensor * src0 = dst->src[0];
     ggml_tensor * src1 = dst->src[1];
 
-    aclTensor * acl_src0;
-    aclTensor * acl_src1;
-    aclTensor * acl_dst;
+    acl_tensor_ptr acl_src0, acl_src1, acl_dst;
 
     // Need bcast
-    bcast_shape(src0, src1, dst, &acl_src0, &acl_src1, &acl_dst);
-    binary_op(ctx, acl_src0, acl_src1, acl_dst);
-
-    ggml_cann_release_resources(ctx, acl_src0, acl_src1, acl_dst);
+    bcast_shape(src0, src1, dst, acl_src0, acl_src1, acl_dst);
+    binary_op(ctx, acl_src0.get(), acl_src1.get(), acl_dst.get());
 }
 
 /**
@@ -1085,7 +993,7 @@ template <auto binary_op> void ggml_cann_binary_op(ggml_backend_cann_context & c
  * and stores the result in the destination tensor.
  *
  * @tparam unary_op A callable with the signature:
- *         void(ggml_backend_cann_context&, aclTensor*, aclTensor*)
+ *         void(ggml_backend_cann_context&, aclTensor *, aclTensor *)
  *         where the first aclTensor is the source and the second is the destination.
  * @param ctx The CANN backend context for managing resources and execution.
  * @param dst The destination tensor. Its src[0] is treated as the input tensor.
@@ -1094,11 +1002,10 @@ template <void unary_op(ggml_backend_cann_context &, aclTensor *, aclTensor *)>
 void ggml_cann_op_unary(ggml_backend_cann_context & ctx, ggml_tensor * dst) {
     ggml_tensor * src = dst->src[0];
 
-    aclTensor * acl_src = ggml_cann_create_tensor(src);
-    aclTensor * acl_dst = ggml_cann_create_tensor(dst);
+    acl_tensor_ptr acl_src = ggml_cann_create_tensor(src);
+    acl_tensor_ptr acl_dst = ggml_cann_create_tensor(dst);
 
-    unary_op(ctx, acl_src, acl_dst);
-    ggml_cann_release_resources(ctx, acl_src, acl_dst);
+    unary_op(ctx, acl_src.get(), acl_dst.get());
 }
 
 /**

ggml/src/ggml-cann/common.h

@@ -23,26 +23,26 @@
 #ifndef CANN_COMMON_H
 #define CANN_COMMON_H
 
-#include <acl/acl.h>
-
-#include <cstdio>
-#include <iostream>
-#include <map>
-#include <memory>
-#include <string>
-#include <vector>
-#include <atomic>
-#include <condition_variable>
-#include <mutex>
-#include <thread>
-#include <unistd.h>
-#include <functional>
-#include <optional>
-#include <list>
-
+#include "../ggml-impl.h"
 #include "../include/ggml-cann.h"
 #include "../include/ggml.h"
-#include "../ggml-impl.h"
+
+#include <acl/acl.h>
+#include <unistd.h>
+
+#include <atomic>
+#include <condition_variable>
+#include <cstdio>
+#include <functional>
+#include <iostream>
+#include <list>
+#include <map>
+#include <memory>
+#include <mutex>
+#include <optional>
+#include <string>
+#include <thread>
+#include <vector>
 
 #define MATRIX_ROW_PADDING    512
 #define GGML_CANN_MAX_STREAMS 8
@@ -214,130 +214,6 @@ struct ggml_cann_pool_alloc {
     ggml_cann_pool_alloc & operator=(ggml_cann_pool_alloc &&) = delete;
 };
 
-/**
- * @brief Function pointer type for ACLNN operator calls.
- */
-using aclnn_func_t = aclnnStatus (*)(void *, uint64_t, aclOpExecutor *, aclrtStream);
-
-/**
- * @brief Base class for all CANN tasks to be submitted to the task queue.
- *
- * Users should override the run_task() method with actual task logic.
- */
-class cann_task {
-  public:
-    virtual void run_task() {}
-};
-
-/**
- * @brief A lock-free ring-buffer based task queue for asynchronously executing cann_task instances.
- */
-class cann_task_queue {
-  public:
-    /**
-     * @brief Constructs a task queue with a fixed power-of-two capacity for a specific device.
-     *
-     * @param capacity Queue capacity. Must be a power of 2.
-     * @param device Target device ID (used for context setting).
-     */
-    explicit cann_task_queue(size_t capacity, int32_t device) :
-        buffer_(capacity),
-        capacity_(capacity),
-        head_(0),
-        tail_(0),
-        running_(false),
-        device_(device) {
-        GGML_ASSERT((capacity & (capacity - 1)) == 0 && "capacity must be power of 2");
-        mask_ = capacity_ - 1;
-    }
-
-    /**
-     * @brief Attempts to enqueue a task into the queue.
-     *
-     * @param item Unique pointer to the task.
-     * @return true if the task was successfully enqueued, false if the queue was full.
-     */
-    bool enqueue(std::unique_ptr<cann_task> && item) {
-        size_t next_tail = (tail_ + 1) & mask_;
-
-        if (next_tail == head_) {
-            return false;
-        }
-
-        buffer_[tail_] = std::move(item);
-        std::atomic_thread_fence(std::memory_order_release);
-        tail_ = next_tail;
-
-        return true;
-    }
-
-    /**
-     * @brief Submits a task to the queue, and starts the worker thread if not already running.
-     *
-     * @param task Task to be submitted.
-     */
-    void submit_task(std::unique_ptr<cann_task> && task) {
-        while (!enqueue(std::move(task))) {
-            std::this_thread::yield();
-            continue;
-        }
-
-        if (!running_) {
-            running_ = true;
-            thread_  = std::thread(&cann_task_queue::execute, this);
-        }
-    }
-
-    /**
-     * @brief Waits until the queue is completely empty and no tasks are being processed.
-     */
-    void wait() {
-        while (running_ && head_ != tail_) {
-            std::this_thread::yield();
-            continue;
-        }
-    }
-
-    /**
-     * @brief Stops the task queue and joins the worker thread.
-     */
-    void stop() {
-        running_ = false;
-        if (thread_.joinable()) {
-            thread_.join();
-        }
-    }
-
-  private:
-    /**
-     * @brief Worker thread function that continuously dequeues and executes tasks.
-     */
-    void execute() {
-        ggml_cann_set_device(device_);
-
-        while (running_) {
-            if (head_ == tail_) {
-                std::this_thread::yield();
-                continue;
-            }
-
-            std::atomic_thread_fence(std::memory_order_acquire);
-            buffer_[head_]->run_task();
-            buffer_[head_].reset();
-            head_ = (head_ + 1) & mask_;
-        }
-    }
-
-    std::vector<std::unique_ptr<cann_task>> buffer_;
-    const size_t                            capacity_;
-    size_t                                  mask_;
-    size_t                                  head_;
-    size_t                                  tail_;
-    bool                                    running_;
-    std::thread                             thread_;
-    int32_t                                 device_;
-};
-
 #ifdef USE_ACL_GRAPH
 struct ggml_graph_node_properties {
     // dst tensor
@@ -474,7 +350,6 @@ struct ggml_backend_cann_context {
     ggml_cann_graph_lru_cache graph_lru_cache;
     bool                      acl_graph_mode = true;
 #endif
-    cann_task_queue        task_queue;
     bool                   async_mode;
     // Rope Cache
     ggml_cann_rope_cache   rope_cache;
@@ -488,15 +363,10 @@ struct ggml_backend_cann_context {
      * @brief Constructor for initializing the context with a given device.
      * @param device Device ID.
      */
-    explicit ggml_backend_cann_context(int device) :
-        device(device),
-        name("CANN" + std::to_string(device)),
-        task_queue(1024, device) {
+    explicit ggml_backend_cann_context(int device) : device(device), name("CANN" + std::to_string(device)) {
         ggml_cann_set_device(device);
         description = aclrtGetSocName();
 
-        async_mode = parse_bool(get_env("GGML_CANN_ASYNC_MODE").value_or(""));
-        GGML_LOG_INFO("%s: device %d async operator submission is %s\n", __func__, device, async_mode ? "ON" : "OFF");
 #ifdef USE_ACL_GRAPH
         acl_graph_mode = parse_bool(get_env("GGML_CANN_ACL_GRAPH").value_or("on"));
         GGML_LOG_INFO("%s: device %d execution mode is %s (%s)\n", __func__, device, acl_graph_mode ? "GRAPH" : "EAGER",
@@ -509,7 +379,6 @@ struct ggml_backend_cann_context {
      */
     ~ggml_backend_cann_context() {
         ggml_cann_set_device(device);
-        task_queue.stop();
         if (copy_event != nullptr) {
             ACL_CHECK(aclrtDestroyEvent(copy_event));
         }

ggml/src/ggml-cann/ggml-cann.cpp

@@ -22,24 +22,24 @@
 
 #include "ggml-cann.h"
 
-#include <acl/acl.h>
-#include <stdarg.h>
-#include <aclnnop/aclnn_trans_matmul_weight.h>
+#include "ggml-backend-impl.h"
+#include "ggml-cann/aclnn_ops.h"
+#include "ggml-cann/common.h"
+#include "ggml-impl.h"
+#include "ggml.h"
 
+#include <acl/acl.h>
+#include <aclnnop/aclnn_trans_matmul_weight.h>
+#include <stdarg.h>
+
+#include <chrono>
 #include <cmath>
 #include <cstdio>
 #include <cstring>
 #include <mutex>
-#include <queue>
-#include <chrono>
-#include <unordered_set>
 #include <optional>
-
-#include "ggml-impl.h"
-#include "ggml-backend-impl.h"
-#include "ggml-cann/aclnn_ops.h"
-#include "ggml-cann/common.h"
-#include "ggml.h"
+#include <queue>
+#include <unordered_set>
 
 #define GGML_COMMON_DECL_C
 
@@ -1177,19 +1177,18 @@ static ggml_cann_nz_workspace g_nz_workspaces[GGML_CANN_MAX_DEVICES];
  *       across calls. This reduces overhead from repeated memory allocation and deallocation.
  */
 static void weight_format_to_nz(ggml_tensor * tensor, size_t offset, int device) {
-    aclTensor * weightTransposed = ggml_cann_create_tensor(tensor, tensor->ne, tensor->nb, 2, ACL_FORMAT_ND, offset);
-    uint64_t    workspaceSize    = 0;
+    acl_tensor_ptr weightTransposed = ggml_cann_create_tensor(tensor, tensor->ne, tensor->nb, 2, ACL_FORMAT_ND, offset);
+    uint64_t       workspaceSize    = 0;
     aclOpExecutor * executor;
 
     // TransMatmulWeight
-    ACL_CHECK(aclnnTransMatmulWeightGetWorkspaceSize(weightTransposed, &workspaceSize, &executor));
+    ACL_CHECK(aclnnTransMatmulWeightGetWorkspaceSize(weightTransposed.get(), &workspaceSize, &executor));
     // Avoid frequent malloc/free of the workspace.
     g_nz_workspaces[device].realloc(workspaceSize);
 
     void * g_nz_workspace = g_nz_workspaces[device].get();
 
     ACL_CHECK(aclnnTransMatmulWeight(g_nz_workspace, workspaceSize, executor, nullptr));
-    ACL_CHECK(aclDestroyTensor(weightTransposed));
 }
 
 // TODO: need handle tensor which has paddings.
@@ -1641,7 +1640,7 @@ ggml_backend_buffer_type_t ggml_backend_cann_host_buffer_type() {
                            /* .is_host          = */ ggml_backend_cpu_buffer_type()->iface.is_host,
                            },
         /* .device   = */
-         ggml_backend_reg_dev_get(ggml_backend_cann_reg(), 0),
+        ggml_backend_reg_dev_get(ggml_backend_cann_reg(), 0),
         /* .context  = */ nullptr,
     };
 
@@ -1777,6 +1776,12 @@ static bool ggml_cann_compute_forward(ggml_backend_cann_context & ctx, struct gg
         case GGML_OP_GROUP_NORM:
             ggml_cann_group_norm(ctx, dst);
             break;
+        case GGML_OP_L2_NORM:
+            ggml_cann_l2_norm(ctx, dst);
+            break;
+        case GGML_OP_CROSS_ENTROPY_LOSS:
+            ggml_cann_cross_entropy_loss(ctx, dst);
+            break;
         case GGML_OP_CONCAT:
             ggml_cann_concat(ctx, dst);
             break;
@@ -1943,7 +1948,8 @@ static void ggml_backend_cann_set_tensor_async(ggml_backend_t backend,
     GGML_ASSERT(buf->buft == ggml_backend_cann_buffer_type(cann_ctx->device) && "unsupported buffer type");
     GGML_ASSERT(!ggml_is_quantized(tensor->type));
 
-    ggml_cann_async_memcpy(cann_ctx, (char *) tensor->data + offset, data, size, ACL_MEMCPY_HOST_TO_DEVICE);
+    ACL_CHECK(aclrtMemcpyAsync((char *) tensor->data + offset, size, data, size, ACL_MEMCPY_HOST_TO_DEVICE,
+                               cann_ctx->stream()));
 }
 
 /**
@@ -1968,7 +1974,8 @@ static void ggml_backend_cann_get_tensor_async(ggml_backend_t      backend,
     GGML_ASSERT(buf->buft == ggml_backend_cann_buffer_type(cann_ctx->device) && "unsupported buffer type");
     GGML_ASSERT(!ggml_is_quantized(tensor->type));
 
-    ggml_cann_async_memcpy(cann_ctx, data, (char *) tensor->data + offset, size, ACL_MEMCPY_DEVICE_TO_HOST);
+    ACL_CHECK(aclrtMemcpyAsync(data, size, (char *) tensor->data + offset, size, ACL_MEMCPY_DEVICE_TO_HOST,
+                               cann_ctx->stream()));
 }
 
 /**
@@ -2029,7 +2036,6 @@ static bool ggml_backend_cann_cpy_tensor_async(ggml_backend_t      backend_src,
         ACL_CHECK(aclrtDeviceEnablePeerAccess(cann_ctx_dst->device, 0));
 
         // wait for task_queue empty to keep task order.
-        cann_ctx_src->task_queue.wait();
         ACL_CHECK(aclrtMemcpyAsync(dst->data, copy_size, src->data, copy_size, ACL_MEMCPY_DEVICE_TO_DEVICE,
                                    cann_ctx_src->stream()));
         // record event on src stream after the copy
@@ -2062,7 +2068,6 @@ static bool ggml_backend_cann_cpy_tensor_async(ggml_backend_t      backend_src,
  */
 static void ggml_backend_cann_synchronize(ggml_backend_t backend) {
     ggml_backend_cann_context * cann_ctx = (ggml_backend_cann_context *) backend->context;
-    cann_ctx->task_queue.wait();
     ggml_cann_set_device(cann_ctx->device);
     ACL_CHECK(aclrtSynchronizeStream(cann_ctx->stream()));
 }
@@ -2479,6 +2484,9 @@ static bool ggml_backend_cann_supports_op(ggml_backend_dev_t dev, const ggml_ten
                 if (mode & GGML_ROPE_TYPE_VISION) {
                     return false;
                 }
+                if (op->src[0]->ne[0] > 896) {
+                    return false;
+                }
 #ifdef ASCEND_310P
                 if (!ggml_is_contiguous(op->src[0])) {
                     return false;
@@ -2515,8 +2523,11 @@ static bool ggml_backend_cann_supports_op(ggml_backend_dev_t dev, const ggml_ten
                 // value of paddingW should be at most half of kernelW
                 return (p0 <= (k0 / 2)) && (p1 <= (k1 / 2));
             }
-        case GGML_OP_DUP:
         case GGML_OP_SUM:
+            return ggml_is_contiguous_rows(op->src[0]);
+        case GGML_OP_L2_NORM:
+        case GGML_OP_CROSS_ENTROPY_LOSS:
+        case GGML_OP_DUP:
         case GGML_OP_IM2COL:
         case GGML_OP_CONCAT:
         case GGML_OP_REPEAT:

ggml/src/ggml-cpu/CMakeLists.txt

@@ -145,26 +145,27 @@ function(ggml_add_cpu_backend_variant_impl tag_name)
 
                 include(CheckCXXSourceRuns)
 
-                function(check_arm_feature tag code)
+                macro(check_arm_feature tag feature code)
                     set(CMAKE_REQUIRED_FLAGS_SAVE ${CMAKE_REQUIRED_FLAGS})
                     set(CMAKE_REQUIRED_FLAGS "${ARM_NATIVE_FLAG}+${tag}")
                     check_cxx_source_runs("${code}" GGML_MACHINE_SUPPORTS_${tag})
                     if (GGML_MACHINE_SUPPORTS_${tag})
-                        set(ARM_NATIVE_FLAG_FIX "${ARM_NATIVE_FLAG_FIX}+${tag}" PARENT_SCOPE)
+                        set(ARM_NATIVE_FLAG_FIX "${ARM_NATIVE_FLAG_FIX}+${tag}")
                     else()
                         set(CMAKE_REQUIRED_FLAGS "${ARM_NATIVE_FLAG}+no${tag}")
                         check_cxx_source_compiles("int main() { return 0; }" GGML_MACHINE_SUPPORTS_no${tag})
                         if (GGML_MACHINE_SUPPORTS_no${tag})
-                            set(ARM_NATIVE_FLAG_FIX "${ARM_NATIVE_FLAG_FIX}+no${tag}" PARENT_SCOPE)
+                            set(ARM_NATIVE_FLAG_FIX "${ARM_NATIVE_FLAG_FIX}+no${tag}")
+                            list(APPEND ARCH_FLAGS -U__ARM_FEATURE_${feature})
                         endif()
                     endif()
                     set(CMAKE_REQUIRED_FLAGS ${CMAKE_REQUIRED_FLAGS_SAVE})
-                endfunction()
+                endmacro()
 
-                check_arm_feature(dotprod "#include <arm_neon.h>\nint main() { int8x16_t _a, _b; volatile int32x4_t _s = vdotq_s32(_s, _a, _b); return 0; }")
-                check_arm_feature(i8mm    "#include <arm_neon.h>\nint main() { int8x16_t _a, _b; volatile int32x4_t _s = vmmlaq_s32(_s, _a, _b); return 0; }")
-                check_arm_feature(sve     "#include <arm_sve.h>\nint main()  { svfloat32_t _a, _b; volatile svfloat32_t _c = svadd_f32_z(svptrue_b8(), _a, _b); return 0; }")
-                check_arm_feature(sme     "#include <arm_sme.h>\n__arm_locally_streaming int main() { __asm__ volatile(\"smstart; smstop;\"); return 0; }")
+                check_arm_feature(dotprod DOTPROD     "#include <arm_neon.h>\nint main() { int8x16_t _a, _b; volatile int32x4_t _s = vdotq_s32(_s, _a, _b); return 0; }")
+                check_arm_feature(i8mm    MATMUL_INT8 "#include <arm_neon.h>\nint main() { int8x16_t _a, _b; volatile int32x4_t _s = vmmlaq_s32(_s, _a, _b); return 0; }")
+                check_arm_feature(sve     SVE         "#include <arm_sve.h>\nint main()  { svfloat32_t _a, _b; volatile svfloat32_t _c = svadd_f32_z(svptrue_b8(), _a, _b); return 0; }")
+                check_arm_feature(sme     SME         "#include <arm_sme.h>\n__arm_locally_streaming int main() { __asm__ volatile(\"smstart; smstop;\"); return 0; }")
 
                 list(APPEND ARCH_FLAGS "${ARM_NATIVE_FLAG}${ARM_NATIVE_FLAG_FIX}")
             else()
@@ -216,35 +217,27 @@ function(ggml_add_cpu_backend_variant_impl tag_name)
                 endif()
             endif()
 
-            # show enabled features
-            if (CMAKE_HOST_SYSTEM_NAME STREQUAL "Windows")
-                set(FEAT_INPUT_FILE "NUL")
-            else()
-                set(FEAT_INPUT_FILE "/dev/null")
-            endif()
+            message(STATUS "Checking for ARM features using flags:")
+            foreach(flag IN LISTS ARCH_FLAGS)
+                message(STATUS "  ${flag}")
+            endforeach()
 
-            execute_process(
-                COMMAND ${CMAKE_C_COMPILER} ${ARCH_FLAGS} -dM -E -
-                INPUT_FILE ${FEAT_INPUT_FILE}
-                OUTPUT_VARIABLE ARM_FEATURE
-                RESULT_VARIABLE ARM_FEATURE_RESULT
-            )
-            if (ARM_FEATURE_RESULT)
-                message(WARNING "Failed to get ARM features")
-            else()
-                foreach(feature DOTPROD SVE MATMUL_INT8 FMA FP16_VECTOR_ARITHMETIC SME)
-                    string(FIND "${ARM_FEATURE}" "__ARM_FEATURE_${feature} 1" feature_pos)
-                    if (NOT ${feature_pos} EQUAL -1)
-                        # Special handling for MATMUL_INT8 when machine doesn't support i8mm
-                        if ("${feature}" STREQUAL "MATMUL_INT8" AND GGML_MACHINE_SUPPORTS_noi8mm)
-                            message(STATUS "ARM feature ${feature} detected but unsetting due to machine not supporting i8mm")
-                            list(APPEND ARCH_FLAGS -U__ARM_FEATURE_MATMUL_INT8)
-                        else()
-                            message(STATUS "ARM feature ${feature} enabled")
-                        endif()
-                    endif()
-                endforeach()
-            endif()
+            include(CheckCXXSourceCompiles)
+            set(CMAKE_REQUIRED_FLAGS_SAVE ${CMAKE_REQUIRED_FLAGS})
+            set(CMAKE_REQUIRED_FLAGS "${ARCH_FLAGS}")
+            foreach(feature DOTPROD SVE MATMUL_INT8 FMA FP16_VECTOR_ARITHMETIC SME)
+                set(ARM_FEATURE "HAVE_${feature}")
+                check_cxx_source_compiles(
+                    "
+                    #if !defined(__ARM_FEATURE_${feature})
+                    #  error \"Feature ${feature} is not defined\"
+                    #endif
+                    int main() { return 0; }
+                    "
+                    ${ARM_FEATURE}
+                )
+            endforeach()
+            set(CMAKE_REQUIRED_FLAGS ${CMAKE_REQUIRED_FLAGS_SAVE})
         endif()
     elseif (GGML_SYSTEM_ARCH STREQUAL "x86")
         message(STATUS "x86 detected")

ggml/src/ggml-cpu/arch/x86/repack.cpp

@@ -646,7 +646,7 @@ static void gemm_q4_b32_8x8_q8_0_lut_avx(int n, float * GGML_RESTRICT s, size_t
     __m256i requiredOrder = _mm256_set_epi32(3, 2, 1, 0, 7, 6, 5, 4);
     int64_t xstart = 0;
     int anr = nr - nr%16; // Used to align nr with boundary of 16
-#ifdef __AVX512F__
+#if defined(__AVX512BW__) && defined(__AVX512DQ__)
     int anc = nc - nc%16; // Used to align nc with boundary of 16
                           // Mask to mask out nibbles from packed bytes expanded to 512 bit length
     const __m512i m4bexpanded = _mm512_set1_epi8(0x0F);
@@ -1041,7 +1041,7 @@ static void gemm_q4_b32_8x8_q8_0_lut_avx(int n, float * GGML_RESTRICT s, size_t
         xstart = anc/8;
         y = 0;
     }
-#endif // __AVX512F__
+#endif // __AVX512BW__ && __AVX512DQ__
 
     // Take group of four block_q8_0x4 structures at each pass of the loop and perform dot product operation
 
@@ -1989,7 +1989,7 @@ void ggml_gemm_q4_K_8x8_q8_K(int n, float * GGML_RESTRICT s, size_t bs, const vo
     __m256i requiredOrder = _mm256_set_epi32(3, 2, 1, 0, 7, 6, 5, 4);
     int64_t xstart = 0;
     int anr = nr - nr % 16;; // Used to align nr with boundary of 16
-#ifdef __AVX512F__
+#if defined(__AVX512BW__) && defined(__AVX512DQ__)
     int anc = nc - nc % 16; // Used to align nc with boundary of 16
     // Mask to mask out nibbles from packed bytes expanded to 512 bit length
     const __m512i m4bexpanded = _mm512_set1_epi8(0x0F);
@@ -2727,7 +2727,7 @@ void ggml_gemm_q4_K_8x8_q8_K(int n, float * GGML_RESTRICT s, size_t bs, const vo
         xstart = anc/8;
         y = 0;
     }
-#endif //AVX512F
+#endif // __AVX512BW__ && __AVX512DQ__
 
     // Take group of four block_q8_Kx4 structures at each pass of the loop and perform dot product operation
     for (; y < anr / 4; y += 4) {
@@ -3467,7 +3467,7 @@ void ggml_gemm_q2_K_8x8_q8_K(int n, float * GGML_RESTRICT s, size_t bs, const vo
     __m256i scalesmask2 = _mm256_castsi128_si256(scalesmask2_sse);
     scalesmask2 = _mm256_permute2f128_si256(scalesmask2, scalesmask2, 0);
 
-#ifdef __AVX512F__
+#if defined(__AVX512BW__) && defined(__AVX512DQ__)
 
     int anc = nc - nc % 16; // Used to align nc with boundary of 16
 
@@ -4947,7 +4947,7 @@ void ggml_gemm_q2_K_8x8_q8_K(int n, float * GGML_RESTRICT s, size_t bs, const vo
         y = 0;
     }
 
-#endif //AVX512F
+#endif // __AVX512BW__ && __AVX512DQ__
 
     // Take group of four block_q8_Kx4 structures at each pass of the loop and perform dot product operation
     for (; y < anr / 4; y += 4) {

ggml/src/ggml-cpu/ggml-cpu.c

@@ -1731,6 +1731,10 @@ static void ggml_compute_forward(struct ggml_compute_params * params, struct ggm
             {
                 ggml_compute_forward_sum_rows(params, tensor);
             } break;
+        case GGML_OP_CUMSUM:
+            {
+                ggml_compute_forward_cumsum(params, tensor);
+            } break;
         case GGML_OP_MEAN:
             {
                 ggml_compute_forward_mean(params, tensor);
@@ -1927,6 +1931,14 @@ static void ggml_compute_forward(struct ggml_compute_params * params, struct ggm
             {
                 ggml_compute_forward_leaky_relu(params, tensor);
             } break;
+        case GGML_OP_TRI:
+            {
+                ggml_compute_forward_tri(params, tensor);
+            } break;
+        case GGML_OP_FILL:
+            {
+                ggml_compute_forward_fill(params, tensor);
+            } break;
         case GGML_OP_FLASH_ATTN_EXT:
             {
                 ggml_compute_forward_flash_attn_ext(params, tensor);
@@ -1982,6 +1994,10 @@ static void ggml_compute_forward(struct ggml_compute_params * params, struct ggm
             {
                 ggml_compute_forward_rwkv_wkv7(params, tensor);
             } break;
+        case GGML_OP_SOLVE_TRI:
+            {
+                ggml_compute_forward_solve_tri(params, tensor);
+            } break;
         case GGML_OP_MAP_CUSTOM1:
             {
                 ggml_compute_forward_map_custom1(params, tensor);
@@ -2140,6 +2156,9 @@ static int ggml_get_n_tasks(struct ggml_tensor * node, int n_threads) {
         case GGML_OP_ADD_ID:
         case GGML_OP_ADD1:
         case GGML_OP_ACC:
+        case GGML_OP_CUMSUM:
+        case GGML_OP_TRI:
+        case GGML_OP_FILL:
             {
                 n_tasks = n_threads;
             } break;
@@ -2157,6 +2176,7 @@ static int ggml_get_n_tasks(struct ggml_tensor * node, int n_threads) {
                 n_tasks = 1;
             } break;
         case GGML_OP_COUNT_EQUAL:
+        case GGML_OP_SOLVE_TRI:
             {
                 n_tasks = n_threads;
             } break;
@@ -2179,6 +2199,8 @@ static int ggml_get_n_tasks(struct ggml_tensor * node, int n_threads) {
                 case GGML_UNARY_OP_HARDSWISH:
                 case GGML_UNARY_OP_HARDSIGMOID:
                 case GGML_UNARY_OP_EXP:
+                case GGML_UNARY_OP_SOFTPLUS:
+                case GGML_UNARY_OP_EXPM1:
                 case GGML_UNARY_OP_FLOOR:
                 case GGML_UNARY_OP_CEIL:
                 case GGML_UNARY_OP_ROUND:

ggml/src/ggml-cpu/ops.cpp

@@ -7,8 +7,10 @@
 #include "unary-ops.h"
 #include "vec.h"
 
-#include <float.h>
+#include <cfloat>
 #include <algorithm>
+#include <cmath>
+#include <functional>
 
 // ggml_compute_forward_dup
 
@@ -1394,6 +1396,56 @@ void ggml_compute_forward_sum(
     }
 }
 
+// ggml_compute_forward_cumsum
+
+static void ggml_compute_forward_cumsum_f32(
+        const ggml_compute_params * params,
+        ggml_tensor * dst) {
+
+    const ggml_tensor * src0 = dst->src[0];
+
+    GGML_ASSERT(src0->nb[0] == sizeof(float));
+    GGML_ASSERT(dst->nb[0] == sizeof(float));
+
+    GGML_TENSOR_UNARY_OP_LOCALS
+
+    GGML_ASSERT(ne0 == ne00);
+    GGML_ASSERT(ne1 == ne01);
+    GGML_ASSERT(ne2 == ne02);
+    GGML_ASSERT(ne3 == ne03);
+
+    const auto [ir0, ir1] = get_thread_range(params, src0);
+
+    for (int64_t ir = ir0; ir < ir1; ++ir) {
+        const int64_t i03 = ir/(ne02*ne01);
+        const int64_t i02 = (ir - i03*ne02*ne01)/ne01;
+        const int64_t i01 = (ir - i03*ne02*ne01 - i02*ne01);
+
+        float * src_row = (float *) ((char *) src0->data + i01*nb01 + i02*nb02 + i03*nb03);
+        float * dst_row = (float *) ((char *) dst->data  + i01*nb1  + i02*nb2  + i03*nb3);
+
+        ggml_vec_cumsum_f32(ne00, dst_row, src_row);
+    }
+}
+
+void ggml_compute_forward_cumsum(
+        const ggml_compute_params * params,
+        ggml_tensor * dst) {
+
+    const ggml_tensor * src0 = dst->src[0];
+
+    switch (src0->type) {
+        case GGML_TYPE_F32:
+            {
+                ggml_compute_forward_cumsum_f32(params, dst);
+            } break;
+        default:
+            {
+                GGML_ABORT("fatal error");
+            }
+    }
+}
+
 // ggml_compute_forward_sum_rows
 
 static void ggml_compute_forward_sum_rows_f32(
@@ -2140,6 +2192,83 @@ static void ggml_compute_forward_gelu(
     }
 }
 
+// ggml_compute_fill
+
+static void ggml_compute_forward_fill_f32(const ggml_compute_params * params, ggml_tensor * dst) {
+    const float c = ggml_get_op_params_f32(dst, 0);
+
+    GGML_TENSOR_LOCALS(int64_t, ne, dst, ne);
+    GGML_TENSOR_LOCALS(size_t,  nb, dst, nb);
+
+    const auto [ir0, ir1] = get_thread_range(params, dst);
+
+    for (int64_t ir = ir0; ir < ir1; ++ir) {
+        const int64_t i03 = ir/(ne2*ne1);
+        const int64_t i02 = (ir - i03*ne2*ne1)/ne1;
+        const int64_t i01 = (ir - i03*ne2*ne1 - i02*ne1);
+
+        float * dst_ptr  = (float *) ((char *) dst->data + i03*nb3 + i02*nb2 + i01*nb1);
+
+        ggml_vec_set_f32(ne0, dst_ptr, c);
+    }
+}
+
+void ggml_compute_forward_fill(const ggml_compute_params * params, ggml_tensor * dst) {
+    ggml_compute_forward_fill_f32(params, dst);
+}
+
+// ggml_compute_tri
+
+static void ggml_compute_forward_tri_f32(const ggml_compute_params * params, ggml_tensor * dst) {
+    const ggml_tensor * src0 = dst->src[0];
+
+    const ggml_tri_type ttype = (ggml_tri_type) ggml_get_op_params_i32(dst, 0);
+
+    GGML_ASSERT(ggml_is_contiguous(src0));
+
+    GGML_TENSOR_UNARY_OP_LOCALS
+
+    const auto [ir0, ir1] = get_thread_range(params, src0);
+
+    bool (*bipred)(int, int);
+
+    switch (ttype) {
+        case GGML_TRI_TYPE_LOWER:      bipred = [](int i, int r) { return i <  r; }; break;
+        case GGML_TRI_TYPE_LOWER_DIAG: bipred = [](int i, int r) { return i <= r; }; break;
+        case GGML_TRI_TYPE_UPPER:      bipred = [](int i, int r) { return i >  r; }; break;
+        case GGML_TRI_TYPE_UPPER_DIAG: bipred = [](int i, int r) { return i >= r; }; break;
+        default: GGML_ABORT("invalid tri type");
+    }
+
+    for (int64_t ir = ir0; ir < ir1; ++ir) {
+        const int64_t i03 = ir/(ne02*ne01);
+        const int64_t i02 = (ir - i03*ne02*ne01)/ne01;
+        const int64_t i01 = (ir - i03*ne02*ne01 - i02*ne01);
+
+        const float * src_ptr = (const float  *) ((const char *) src0->data + i03*nb03 + i02*nb02 + i01*nb01);
+              float * dst_ptr = (      float  *) ((      char *) dst->data  + i03*nb3  + i02*nb2  + i01*nb1);
+
+        for (int i0 = 0; i0 < ne0; ++i0) {
+            dst_ptr[i0] = bipred(i0, i01) ? src_ptr[i0] : 0.0f;
+        }
+    }
+}
+
+void ggml_compute_forward_tri(const ggml_compute_params * params, ggml_tensor * dst) {
+    const ggml_tensor * src0 = dst->src[0];
+
+    switch (src0->type) {
+        case GGML_TYPE_F32:
+            {
+                ggml_compute_forward_tri_f32(params, dst);
+            } break;
+        default:
+            {
+                GGML_ABORT("fatal error");
+            }
+    }
+}
+
 // ggml_compute_forward_gelu_erf
 
 static void ggml_compute_forward_gelu_erf_f32(
@@ -7664,6 +7793,18 @@ void ggml_compute_forward_timestep_embedding(
 
 // ggml_compute_forward_argsort
 
+template<enum ggml_sort_order order>
+struct argsort_cmp {
+    const float * data;
+    bool operator()(int32_t a, int32_t b) const {
+        if constexpr (order == GGML_SORT_ORDER_ASC) {
+            return data[a] < data[b];
+        } else {
+            return data[a] > data[b];
+        }
+    }
+};
+
 static void ggml_compute_forward_argsort_f32(
     const ggml_compute_params * params,
     ggml_tensor * dst) {
@@ -7682,23 +7823,25 @@ static void ggml_compute_forward_argsort_f32(
     ggml_sort_order order = (ggml_sort_order) ggml_get_op_params_i32(dst, 0);
 
     for (int64_t i = ith; i < nr; i += nth) {
-        int32_t * dst_data = (int32_t *)((char *) dst->data + i*nb1);
         const float * src_data = (float *)((char *) src0->data + i*nb01);
 
+        int32_t * dst_data = (int32_t *)((char *) dst->data + i*nb1);
+
         for (int64_t j = 0; j < ne0; j++) {
             dst_data[j] = j;
         }
 
-        // C doesn't have a functional sort, so we do a bubble sort instead
-        for (int64_t j = 0; j < ne0; j++) {
-            for (int64_t k = j + 1; k < ne0; k++) {
-                if ((order == GGML_SORT_ORDER_ASC  && src_data[dst_data[j]] > src_data[dst_data[k]]) ||
-                    (order == GGML_SORT_ORDER_DESC && src_data[dst_data[j]] < src_data[dst_data[k]])) {
-                    int32_t tmp = dst_data[j];
-                    dst_data[j] = dst_data[k];
-                    dst_data[k] = tmp;
-                }
-            }
+        switch (order) {
+            case GGML_SORT_ORDER_ASC:
+                std::sort(dst_data, dst_data + ne0, argsort_cmp<GGML_SORT_ORDER_ASC>{src_data});
+                break;
+
+            case GGML_SORT_ORDER_DESC:
+                std::sort(dst_data, dst_data + ne0, argsort_cmp<GGML_SORT_ORDER_DESC>{src_data});
+                break;
+
+            default:
+                GGML_ABORT("invalid sort order");
         }
     }
 }
@@ -8521,7 +8664,7 @@ static void ggml_compute_forward_ssm_scan_f32(
                 // n_head
                 for (int h = ih0; h < ih1; ++h) {
                     // ref: https://github.com/state-spaces/mamba/blob/62db608da60f6fc790b8ed9f4b3225e95ca15fde/mamba_ssm/ops/triton/softplus.py#L16
-                    const float dt_soft_plus = ggml_softplus(dt[h]);
+                    const float dt_soft_plus = ggml_compute_softplus_f32(dt[h]);
                     const float dA = expf(dt_soft_plus * A[h]);
                     const int g = h / (nh / ng); // repeat_interleave
 
@@ -8618,7 +8761,7 @@ static void ggml_compute_forward_ssm_scan_f32(
                 // n_head
                 for (int h = ih0; h < ih1; ++h) {
                     // ref: https://github.com/state-spaces/mamba/blob/62db608da60f6fc790b8ed9f4b3225e95ca15fde/mamba_ssm/ops/triton/softplus.py#L16
-                    const float dt_soft_plus = ggml_softplus(dt[h]);
+                    const float dt_soft_plus = ggml_compute_softplus_f32(dt[h]);
                     const int g = h / (nh / ng); // repeat_interleave
 
                     // dim
@@ -8901,6 +9044,14 @@ void ggml_compute_forward_unary(
             {
                 ggml_compute_forward_xielu(params, dst);
             } break;
+        case GGML_UNARY_OP_EXPM1:
+            {
+                ggml_compute_forward_expm1(params, dst);
+            } break;
+        case GGML_UNARY_OP_SOFTPLUS:
+            {
+                ggml_compute_forward_softplus(params, dst);
+            } break;
         default:
             {
                 GGML_ABORT("fatal error");
@@ -9497,6 +9648,76 @@ void ggml_compute_forward_gla(
     }
 }
 
+static void ggml_compute_forward_solve_tri_f32(const struct ggml_compute_params * params, struct ggml_tensor * dst) {
+    const struct ggml_tensor * src0 = dst->src[0];  // A (lower triangular)
+    const struct ggml_tensor * src1 = dst->src[1];  // B (RHS)
+
+    GGML_TENSOR_BINARY_OP_LOCALS;
+
+    GGML_ASSERT(src0->type == GGML_TYPE_F32);
+    GGML_ASSERT(src1->type == GGML_TYPE_F32);
+    GGML_ASSERT(dst->type  == GGML_TYPE_F32);
+
+    GGML_ASSERT(ne00 == ne01); // A must be square
+    GGML_ASSERT(ne0  == ne10); // solution cols == B cols
+    GGML_ASSERT(ne1  == ne11); // solution rows == B rows
+
+    GGML_ASSERT(ne02 == ne12 && ne12 == ne2);
+    GGML_ASSERT(ne03 == ne13 && ne13 == ne3);
+
+    const int ith = params->ith;
+    const int nth = params->nth;
+
+    const int64_t k = ne10;   // number of RHS columns
+    const int64_t n = ne11;   // A is n×n
+    const int64_t nr = ne02 * ne03 * k; // we're parallelizing on columns here, so seq x token x column will be the unit
+
+    // chunks per thread
+    const int64_t dr = (nr + nth - 1)/nth;
+
+    // chunk range for this thread
+    const int64_t ir0 = dr*ith;
+    const int64_t ir1 = MIN(ir0 + dr, nr);
+
+    const float * A = (const float *) src0->data;  // [n, n, B1, B2]
+    const float * B = (const float *) src1->data;  // [n, k, B1, B2]
+          float * X = (      float *) dst->data;   // [n, k, B1, B2]
+
+    for (int64_t ir = ir0; ir < ir1; ++ir) {
+        const int64_t i03 = ir/(ne02*k);
+        const int64_t i02 = (ir - i03*ne02*k)/k;
+        const int64_t i01 = (ir - i03*ne02*k - i02*k);
+
+        const float * A_batch = A + i02 * nb02 / sizeof(float) + i03 * nb03 / sizeof(float);
+        const float * B_batch = B + i02 * nb12 / sizeof(float) + i03 * nb13 / sizeof(float);
+
+        float * X_batch = X + i02 * nb2 / sizeof(float) + i03 * nb3 / sizeof(float);
+
+        for (int64_t i00 = 0; i00 < n; ++i00) {
+            float sum = 0.0f;
+            for (int64_t t = 0; t < i00; ++t) {
+                sum += A_batch[i00 * n + t] * X_batch[i01 * n + t];
+            }
+
+            const float diag = A_batch[i00 * n + i00];
+            GGML_ASSERT(diag != 0.0f && "Zero diagonal in triangular matrix");
+
+            X_batch[i01 * n + i00] = (B_batch[i00 * k + i01] - sum) / diag;
+        }
+    }
+}
+
+void ggml_compute_forward_solve_tri(const struct ggml_compute_params * params, struct ggml_tensor * dst) {
+    const ggml_tensor * src0 = dst->src[0];
+    const ggml_tensor * src1 = dst->src[1];
+
+    if (src0->type == GGML_TYPE_F32 && src1->type == GGML_TYPE_F32) {
+        ggml_compute_forward_solve_tri_f32(params, dst);
+    } else {
+        GGML_ABORT("fatal error");
+    }
+}
+
 // ggml_compute_forward_rwkv_wkv7
 
 static void ggml_compute_forward_rwkv_wkv7_f32(

ggml/src/ggml-cpu/ops.h

@@ -34,6 +34,7 @@ void ggml_compute_forward_add1(const struct ggml_compute_params * params, struct
 void ggml_compute_forward_acc(const struct ggml_compute_params * params, struct ggml_tensor * dst);
 void ggml_compute_forward_sum(const struct ggml_compute_params * params, struct ggml_tensor * dst);
 void ggml_compute_forward_sum_rows(const struct ggml_compute_params * params, struct ggml_tensor * dst);
+void ggml_compute_forward_cumsum(const struct ggml_compute_params * params, struct ggml_tensor * dst);
 void ggml_compute_forward_mean(const struct ggml_compute_params * params, struct ggml_tensor * dst);
 void ggml_compute_forward_argmax(const struct ggml_compute_params * params, struct ggml_tensor * dst);
 void ggml_compute_forward_count_equal(const struct ggml_compute_params * params, struct ggml_tensor * dst);
@@ -81,6 +82,8 @@ void ggml_compute_forward_arange(const struct ggml_compute_params * params, stru
 void ggml_compute_forward_timestep_embedding(const struct ggml_compute_params * params, struct ggml_tensor * dst);
 void ggml_compute_forward_argsort(const struct ggml_compute_params * params, struct ggml_tensor * dst);
 void ggml_compute_forward_leaky_relu(const struct ggml_compute_params * params, struct ggml_tensor * dst);
+void ggml_compute_forward_tri(const struct ggml_compute_params * params, struct ggml_tensor * dst);
+void ggml_compute_forward_fill(const struct ggml_compute_params * params, struct ggml_tensor * dst);
 void ggml_compute_forward_flash_attn_ext(const struct ggml_compute_params * params, struct ggml_tensor * dst);
 void ggml_compute_forward_flash_attn_back(
         const struct ggml_compute_params * params,
@@ -96,6 +99,7 @@ void ggml_compute_forward_get_rel_pos(const struct ggml_compute_params * params,
 void ggml_compute_forward_add_rel_pos(const struct ggml_compute_params * params, struct ggml_tensor * dst);
 void ggml_compute_forward_rwkv_wkv6(const struct ggml_compute_params * params, struct ggml_tensor * dst);
 void ggml_compute_forward_rwkv_wkv7(const struct ggml_compute_params * params, struct ggml_tensor * dst);
+void ggml_compute_forward_solve_tri(const struct ggml_compute_params * params, struct ggml_tensor * dst);
 void ggml_compute_forward_gla(const struct ggml_compute_params * params, struct ggml_tensor * dst);
 void ggml_compute_forward_map_custom1(const struct ggml_compute_params * params, struct ggml_tensor * dst);
 void ggml_compute_forward_map_custom2(const struct ggml_compute_params * params, struct ggml_tensor * dst);

ggml/src/ggml-cpu/repack.cpp

@@ -1600,29 +1600,52 @@ template <typename BLOC_TYPE, int64_t INTER_SIZE, int64_t NB_COLS, ggml_type PAR
         return false;
     }
 
-    void forward_mul_mat_one_chunk(ggml_compute_params * params, ggml_tensor * op, int64_t src0_start, int64_t src0_end) {
+    void forward_mul_mat_one_chunk(ggml_compute_params * params,
+                                   ggml_tensor *         op,
+                                   int64_t               src0_start,
+                                   int64_t               src0_end,
+                                   int64_t               src1_start,
+                                   int64_t               src1_end) {
         const ggml_tensor * src0 = op->src[0];
         const ggml_tensor * src1 = op->src[1];
         ggml_tensor *       dst  = op;
 
         GGML_TENSOR_BINARY_OP_LOCALS
 
-        const void * src1_wdata      = params->wdata;
         const size_t src1_col_stride = ggml_row_size(PARAM_TYPE, ne10);
 
+        GGML_ASSERT(ne03 == 1 && ne13 == 1);
+        GGML_ASSERT(ne12 % ne02 == 0);
+        const int64_t r2 = ne12 / ne02;
+
+        const int64_t i12 = src1_start / ne1;
+        const int64_t i11 = src1_start - i12 * ne1;
+
+        // Determine batch index
+        const int64_t i02 = i12 / r2;
+
+        const int64_t i1 = i11;
+        const int64_t i2 = i12;
+
+        const char * src0_ptr = (const char *) src0->data + i02 * nb02;
+        const char * src1_ptr = (const char *) params->wdata + (i11 + i12 * ne11) * src1_col_stride;
+        char *       dst_ptr  = ((char *) dst->data + (i1 * nb1 + i2 * nb2));
+
+        const int64_t nrows = src1_end - src1_start;
+        const int64_t ncols = src0_end - src0_start;
+
+        GGML_ASSERT(src1_ptr + src1_col_stride * nrows <= (const char *) params->wdata + params->wsize);
+
         // If there are more than three rows in src1, use gemm; otherwise, use gemv.
-        if (ne11 > 3) {
-            gemm<BLOC_TYPE, INTER_SIZE, NB_COLS, PARAM_TYPE>(ne00,
-                    (float *) ((char *) dst->data) + src0_start, ne01,
-                    (const char *) src0->data + src0_start * nb01,
-                    (const char *) src1_wdata, ne11 - ne11 % 4, src0_end - src0_start);
+        if (nrows > 3) {
+            gemm<BLOC_TYPE, INTER_SIZE, NB_COLS, PARAM_TYPE>(ne00, (float *) (dst_ptr) + src0_start, nb1 / nb0,
+                                                             src0_ptr + src0_start * nb01, src1_ptr,
+                                                             nrows - (nrows % 4), ncols);
         }
-        for (int iter = ne11 - ne11 % 4; iter < ne11; iter++) {
-            gemv<BLOC_TYPE, INTER_SIZE, NB_COLS, PARAM_TYPE>(ne00,
-                    (float *) ((char *) dst->data + (iter * nb1)) + src0_start, ne01,
-                    (const char *) src0->data + src0_start * nb01,
-                    (const char *) src1_wdata + (src1_col_stride * iter), 1,
-                    src0_end - src0_start);
+        for (int iter = nrows - (nrows % 4); iter < nrows; iter++) {
+            gemv<BLOC_TYPE, INTER_SIZE, NB_COLS, PARAM_TYPE>(ne00, (float *) (dst_ptr + (iter * nb1)) + src0_start,
+                                                             ne01, src0_ptr + src0_start * nb01,
+                                                             src1_ptr + (src1_col_stride * iter), 1 /* nrows */, ncols);
         }
     }
 
@@ -1647,6 +1670,12 @@ template <typename BLOC_TYPE, int64_t INTER_SIZE, int64_t NB_COLS, ggml_type PAR
         GGML_ASSERT(nb1 <= nb2);
         GGML_ASSERT(nb2 <= nb3);
 
+        // TODO: General batched mul mat for 4D tensors
+        // Currently only supports 3D tensors
+        GGML_ASSERT(ne03 == 1);
+        GGML_ASSERT(ne13 == 1);
+        GGML_ASSERT(ne3 == 1);
+
         GGML_ASSERT(src1->type == GGML_TYPE_F32);
 
         GGML_ASSERT(ggml_n_dims(op->src[0]) == 2);
@@ -1654,47 +1683,64 @@ template <typename BLOC_TYPE, int64_t INTER_SIZE, int64_t NB_COLS, ggml_type PAR
 
         char *       wdata = static_cast<char *>(params->wdata);
         const size_t nbw1  = ggml_row_size(PARAM_TYPE, ne10);
+        const size_t nbw2  = nbw1 * ne11;
 
-        assert(params->wsize >= nbw1 * ne11);
+        assert(params->wsize >= nbw2 * ne12);
 
         const ggml_from_float_t from_float = ggml_get_type_traits_cpu(PARAM_TYPE)->from_float;
 
-        int64_t i11_processed = 0;
-        for (int64_t i11 = ith * 4; i11 < ne11 - ne11 % 4; i11 += nth * 4) {
-            ggml_quantize_mat_t<INTER_SIZE, PARAM_TYPE>((float *) ((char *) src1->data + i11 * nb11), (void *) (wdata + i11 * nbw1), 4, ne10);
-        }
+        // INFO: Quantization is done in planes to avoid extra complexity in chunking.
+        // Flattening dimensions not multiple of INTER_SIZE would require extra handling depending on how
+        // the planes are broadcast.
+        for (int64_t i12 = 0; i12 < ne12; i12++) {
+            char * data_ptr  = (char *) src1->data + i12 * nb12;
+            char * wdata_ptr = wdata + i12 * nbw2;
 
-        i11_processed = ne11 - ne11 % 4;
-        for (int64_t i11 = i11_processed + ith; i11 < ne11; i11 += nth) {
-            from_float((float *) ((char *) src1->data + i11 * nb11), (void *) (wdata + i11 * nbw1), ne10);
+            for (int64_t i11 = ith * 4; i11 < ne11 - ne11 % 4; i11 += nth * 4) {
+                ggml_quantize_mat_t<INTER_SIZE, PARAM_TYPE>((float *) (data_ptr + i11 * nb11),
+                                                            (void *) (wdata_ptr + i11 * nbw1), 4, ne10);
+            }
+
+            const int64_t i11_processed = ne11 - ne11 % 4;
+            for (int64_t i11 = i11_processed + ith; i11 < ne11; i11 += nth) {
+                from_float((float *) (data_ptr + i11 * nb11), (void *) (wdata_ptr + i11 * nbw1), ne10);
+            }
         }
 
         // disable for NUMA
         const bool disable_chunking = ggml_is_numa();
 
         // 4x chunks per thread
-        int64_t nr = ggml_nrows(op->src[0]);
-        int nth_scaled = nth * 4;
-        int64_t chunk_size = (nr + nth_scaled - 1) / nth_scaled;
-        int64_t nchunk     = (nr + chunk_size - 1) / chunk_size;
+        const int64_t nr0 = ggml_nrows(op->src[0]);
+
+        int     nth_scaled  = nth * 4;
+        int64_t chunk_size0 = (nr0 + nth_scaled - 1) / nth_scaled;
+        int64_t nchunk0     = (nr0 + chunk_size0 - 1) / chunk_size0;
+
+        // src1 is chunked only by full planes.
+        // When we flatten we need to address dimensions not multiple of the q8 INTER_SIZE
+        // to route them thorugh GEMV.
+        // nchunk1 = ne12 also avoids messing the chunking for models with no 3d tensors
+        // to avoid affecting their performance
+        int64_t nchunk1 = ne12;
 
         // Ensure minimum chunk size to avoid alignment issues with high thread counts
         // Minimum chunk size should be at least NB_COLS to prevent overlapping chunks after alignment
         const int64_t min_chunk_size = NB_COLS;
-        if (nchunk > 0 && (nr / nchunk) < min_chunk_size && nr >= min_chunk_size) {
-            nchunk = (nr + min_chunk_size - 1) / min_chunk_size;
+        if (nchunk0 > 0 && (nr0 / nchunk0) < min_chunk_size && nr0 >= min_chunk_size) {
+            nchunk0 = (nr0 + min_chunk_size - 1) / min_chunk_size;
         }
 
-        if (nth == 1 || nchunk < nth || disable_chunking) {
-            nchunk = nth;
+        if (nth == 1 || nchunk0 < nth || disable_chunking) {
+            nchunk0 = nth;
         }
 
+        const int64_t dr0 = (nr0 + nchunk0 - 1) / nchunk0;
+
         // Ensure nchunk doesn't exceed the number of rows divided by minimum chunk size
         // This prevents creating too many tiny chunks that could overlap after alignment
-        const int64_t max_nchunk = (nr + min_chunk_size - 1) / min_chunk_size;
-        if (nchunk > max_nchunk) {
-            nchunk = max_nchunk;
-        }
+        const int64_t max_nchunk = (nr0 + min_chunk_size - 1) / min_chunk_size;
+        nchunk0                  = MIN(nchunk0, max_nchunk);
 
         if (ith == 0) {
             // Every thread starts at ith, so the first unprocessed chunk is nth.  This save a bit of coordination right at the start.
@@ -1706,23 +1752,30 @@ template <typename BLOC_TYPE, int64_t INTER_SIZE, int64_t NB_COLS, ggml_type PAR
         // The first chunk comes from our thread_id, the rest will get auto-assigned.
         int current_chunk = ith;
 
-        while (current_chunk < nchunk) {
-            int64_t src0_start = (current_chunk * ne01) / nchunk;
-            int64_t src0_end   = ((current_chunk + 1) * ne01) / nchunk;
+        while (current_chunk < nchunk0 * nchunk1) {
+            const int64_t ith0 = current_chunk % nchunk0;
+            const int64_t ith1 = current_chunk / nchunk0;
+
+            int64_t src0_start = dr0 * ith0;
+            int64_t src0_end   = MIN(src0_start + dr0, nr0);
+
+            // full-plane range for src1
+            int64_t src1_start = ith1 * ne11;
+            int64_t src1_end = (ith1 + 1) * ne11;
 
             // Align boundaries to NB_COLS - round up to ensure all data is included
             // The chunk size limiting above ensures chunks are large enough to prevent overlaps
             src0_start = (src0_start % NB_COLS) ? src0_start + NB_COLS - (src0_start % NB_COLS) : src0_start;
-            src0_end   = (src0_end   % NB_COLS) ? src0_end   + NB_COLS - (src0_end   % NB_COLS) : src0_end;
-            if (src0_end > ne01) {
-                src0_end = ne01;
-            }
+            src0_end   = (src0_end % NB_COLS) ? src0_end + NB_COLS - (src0_end % NB_COLS) : src0_end;
+            src0_end   = MIN(src0_end, ne01);
 
+            // Make sure current plane is the last one before exiting
             if (src0_start >= src0_end) {
-                break;
+                current_chunk = ggml_threadpool_chunk_add(params->threadpool, 1);
+                continue;
             }
 
-            forward_mul_mat_one_chunk(params, dst, src0_start, src0_end);
+            forward_mul_mat_one_chunk(params, dst, src0_start, src0_end, src1_start, src1_end);
 
             current_chunk = ggml_threadpool_chunk_add(params->threadpool, 1);
         }

ggml/src/ggml-cpu/unary-ops.cpp

@@ -73,6 +73,14 @@ static inline float op_log(float x) {
     return logf(x);
 }
 
+static inline float op_expm1(float x) {
+    return expf(x) - 1.0f;
+}
+
+static inline float op_softplus(float x) {
+    return (x > 20.0f) ? x : logf(1.0f + expf(x));
+}
+
 static inline float op_floor(float x) {
     return floorf(x);
 }
@@ -290,6 +298,14 @@ void ggml_compute_forward_log(const ggml_compute_params * params, ggml_tensor *
     unary_op<op_log>(params, dst);
 }
 
+void ggml_compute_forward_expm1(const ggml_compute_params * params, ggml_tensor * dst) {
+    unary_op<op_expm1>(params, dst);
+}
+
+void ggml_compute_forward_softplus(const ggml_compute_params * params, ggml_tensor * dst) {
+    unary_op<op_softplus>(params, dst);
+}
+
 void ggml_compute_forward_floor(const ggml_compute_params * params, ggml_tensor * dst) {
     unary_op<op_floor>(params, dst);
 }

ggml/src/ggml-cpu/unary-ops.h

@@ -22,6 +22,8 @@ void ggml_compute_forward_sqrt(const struct ggml_compute_params * params, struct
 void ggml_compute_forward_sin(const struct ggml_compute_params * params, struct ggml_tensor * dst);
 void ggml_compute_forward_cos(const struct ggml_compute_params * params, struct ggml_tensor * dst);
 void ggml_compute_forward_log(const struct ggml_compute_params * params, struct ggml_tensor * dst);
+void ggml_compute_forward_expm1(const struct ggml_compute_params * params, struct ggml_tensor * dst);
+void ggml_compute_forward_softplus(const struct ggml_compute_params * params, struct ggml_tensor * dst);
 void ggml_compute_forward_floor(const struct ggml_compute_params * params, struct ggml_tensor * dst);
 void ggml_compute_forward_ceil(const struct ggml_compute_params * params, struct ggml_tensor * dst);
 void ggml_compute_forward_round(const struct ggml_compute_params * params, struct ggml_tensor * dst);

ggml/src/ggml-cpu/vec.cpp

@@ -360,6 +360,13 @@ void ggml_vec_silu_f32(const int n, float * y, const float * x) {
     for (; i + 3 < n; i += 4) {
         vst1q_f32(y + i, ggml_v_silu(vld1q_f32(x + i)));
     }
+#elif defined(__riscv_v_intrinsic)
+    for (int vl; i < n; i += vl) {
+        vl = __riscv_vsetvl_e32m2(n - i);
+        vfloat32m2_t vx = __riscv_vle32_v_f32m2(&x[i], vl);
+        vfloat32m2_t vy = ggml_v_silu_m2(vx, vl);
+        __riscv_vse32_v_f32m2(&y[i], vy, vl);
+    }
 #endif
     for (; i < n; ++i) {
         y[i] = ggml_silu_f32(x[i]);
@@ -460,6 +467,16 @@ ggml_float ggml_vec_cvar_f32(const int n, float * y, const float * x, const floa
         val = vec_mul(val, val);
         sum += (ggml_float)vec_hsum_f32x4(val);
     }
+#elif defined(__riscv_v_intrinsic)
+    vfloat64m1_t vsum = __riscv_vfmv_v_f_f64m1(0, 1);
+    for (int vl; i < n; i += vl) {
+        vl = __riscv_vsetvl_e32m2(n - i);
+        vfloat32m2_t val = __riscv_vfsub_vf_f32m2(__riscv_vle32_v_f32m2(&x[i], vl), mean, vl);
+        __riscv_vse32_v_f32m2(&y[i], val, vl);
+        val = __riscv_vfmul_vv_f32m2(val, val, vl);
+        vsum = __riscv_vfwredusum_vs_f32m2_f64m1(val, vsum, vl);
+    }
+    sum = (ggml_float)__riscv_vfmv_f_s_f64m1_f64(vsum);
 #endif
     for (; i < n; ++i) {
         float val = x[i] - mean;

ggml/src/ggml-cpu/vec.h

@@ -1416,6 +1416,16 @@ inline static void ggml_vec_sum_f32(const int n, float * s, const float * x) {
 #endif
 }
 
+inline static void ggml_vec_cumsum_f32(const int n, float * y, const float * x) {
+    for (int i = 0; i < n; ++i) {
+        if (i == 0) {
+            y[i] = x[i];
+        } else {
+            y[i] = y[i - 1] + x[i];
+        }
+    }
+}
+
 inline static void ggml_vec_sum_f32_ggf(const int n, ggml_float * s, const float * x) {
     ggml_float sum = 0.0;
     for (int i = 0; i < n; ++i) {

ggml/src/ggml-cuda/common.cuh

@@ -586,6 +586,12 @@ static __device__ __forceinline__ void ggml_cuda_mad(half2 & acc, const half2 v,
 //     If dst and src point at different address spaces then they are guaranteed to not be aliased.
 template <int nbytes, int alignment = 0>
 static __device__ __forceinline__ void ggml_cuda_memcpy_1(void * __restrict__ dst, const void * __restrict__ src) {
+    static_assert(
+        nbytes <= ggml_cuda_get_max_cpy_bytes() || alignment == 0,
+        "You are misusing the alignment parameter for ggml_cuda_memcpy_1. "
+        "The intent is for the parameter is only as a workaround if either one of the pointers is not properly aligned. "
+        "If you use it to do more bytes per copy than ggml_cuda_max_cpy_bytes() the reads and writes may not be coalesced. "
+        "Call ggml_cuda_memcpy_1 in a loop instead.");
     if constexpr (alignment != 0) {
         static_assert(nbytes % alignment == 0, "bad alignment");
     }

ggml/src/ggml-cuda/ggml-cuda.cu

@@ -2527,6 +2527,12 @@ static bool ggml_cuda_compute_forward(ggml_backend_cuda_context & ctx, struct gg
                 case GGML_UNARY_OP_TRUNC:
                     ggml_cuda_op_trunc(ctx, dst);
                     break;
+                case GGML_UNARY_OP_EXPM1:
+                    ggml_cuda_op_expm1(ctx, dst);
+                    break;
+                case GGML_UNARY_OP_SOFTPLUS:
+                    ggml_cuda_op_softplus(ctx, dst);
+                    break;
                 default:
                     return false;
             }
@@ -2992,6 +2998,36 @@ static void update_cuda_graph_executable(ggml_backend_cuda_context * cuda_ctx) {
 }
 #endif
 
+static bool ggml_cuda_should_fuse_rope_set_rows(const ggml_tensor * rope,
+                                                const ggml_tensor * view,
+                                                const ggml_tensor * set_rows) {
+    // ne3 not tested
+    if (rope->src[0]->ne[3] != 1) {
+        return false;
+    }
+
+    if (set_rows->type != GGML_TYPE_F32 && set_rows->type != GGML_TYPE_F16) {
+        return false;
+    }
+
+    if (set_rows->src[1]->type != GGML_TYPE_I64) {
+        return false;
+    }
+
+    // The view should flatten two dims of rope into one dim
+    if (!ggml_is_contiguous(view) || view->ne[0] != rope->ne[0] * rope->ne[1]) {
+        return false;
+    }
+
+    // Only norm/neox shaders have the fusion code
+    const int mode = ((const int32_t *) rope->op_params)[2];
+    if (mode != GGML_ROPE_TYPE_NORMAL && mode != GGML_ROPE_TYPE_NEOX) {
+        return false;
+    }
+
+    return true;
+}
+
 static bool ggml_cuda_can_fuse(const struct ggml_cgraph * cgraph, int node_idx, std::initializer_list<enum ggml_op> ops, std::initializer_list<enum ggml_unary_op> unary_ops) {
 #ifndef NDEBUG
     const size_t num_unary = std::count(ops.begin(), ops.end(), GGML_OP_UNARY);
@@ -3067,6 +3103,16 @@ static bool ggml_cuda_can_fuse(const struct ggml_cgraph * cgraph, int node_idx,
         }
     }
 
+    if (ops.size() == 3 && ggml_can_fuse_subgraph(cgraph, node_idx, ops, { node_idx + 2 })) {
+        const ggml_tensor * rope     = cgraph->nodes[node_idx];
+        const ggml_tensor * view     = cgraph->nodes[node_idx + 1];
+        const ggml_tensor * set_rows = cgraph->nodes[node_idx + 2];
+
+        if (ggml_cuda_should_fuse_rope_set_rows(rope, view, set_rows)) {
+            return true;
+        }
+    }
+
     if (!ggml_can_fuse(cgraph, node_idx, ops)) {
         return false;
     }
@@ -3196,6 +3242,15 @@ static void evaluate_and_capture_cuda_graph(ggml_backend_cuda_context * cuda_ctx
                         continue;
                     }
 
+                    if (ggml_cuda_can_fuse(cgraph, i, { GGML_OP_ROPE, GGML_OP_VIEW, GGML_OP_SET_ROWS }, {})) {
+                        ggml_tensor * rope = cgraph->nodes[i];
+                        ggml_tensor * set_rows = cgraph->nodes[i + 2];
+
+                        ggml_cuda_op_rope_fused(*cuda_ctx, rope, set_rows);
+                        i += 2;
+                        continue;
+                    }
+
                     if (node->op == GGML_OP_ADD) {
                         int n_fuse = 0;
                         ggml_op ops[8];
@@ -3780,6 +3835,8 @@ static bool ggml_backend_cuda_device_supports_op(ggml_backend_dev_t dev, const g
                 case GGML_UNARY_OP_GELU_QUICK:
                 case GGML_UNARY_OP_TANH:
                 case GGML_UNARY_OP_EXP:
+                case GGML_UNARY_OP_EXPM1:
+                case GGML_UNARY_OP_SOFTPLUS:
                 case GGML_UNARY_OP_ELU:
                 case GGML_UNARY_OP_FLOOR:
                 case GGML_UNARY_OP_CEIL:

ggml/src/ggml-cuda/rope.cu

@@ -1,3 +1,6 @@
+#include "convert.cuh"
+#include "ggml-cuda/common.cuh"
+#include "ggml.h"
 #include "rope.cuh"
 
 struct rope_corr_dims {
@@ -37,11 +40,23 @@ static __device__ void rope_yarn(
     }
 }
 
-template<bool forward, bool has_ff, typename T>
-static __global__ void rope_norm(
-        const T * x, T * dst, const int ne0, const int ne1, const int s1, const int s2, const int n_dims,
-        const int32_t * pos, const float freq_scale, const float ext_factor, const float attn_factor,
-        const rope_corr_dims corr_dims, const float theta_scale, const float * freq_factors) {
+template <bool forward, bool has_ff, typename T, typename D>
+static __global__ void rope_norm(const T *            x,
+                                 D *                  dst,
+                                 const int            ne0,
+                                 const int            ne1,
+                                 const int            s1,
+                                 const int            s2,
+                                 const int            n_dims,
+                                 const int32_t *      pos,
+                                 const float          freq_scale,
+                                 const float          ext_factor,
+                                 const float          attn_factor,
+                                 const rope_corr_dims corr_dims,
+                                 const float          theta_scale,
+                                 const float *        freq_factors,
+                                 const int64_t *      row_indices,
+                                 const int            set_rows_stride) {
     const int i0 = 2*(blockDim.y*blockIdx.y + threadIdx.y);
 
     if (i0 >= ne0) {
@@ -53,13 +68,27 @@ static __global__ void rope_norm(
     const int row_x     = row_dst % ne1;
     const int channel_x = row_dst / ne1;
 
-    const int idst = row_dst*ne0 + i0;
+    int       idst = row_dst * ne0 + i0;
     const int ix   = channel_x*s2 + row_x*s1 + i0;
 
-    if (i0 >= n_dims) {
-        dst[idst + 0] = x[ix + 0];
-        dst[idst + 1] = x[ix + 1];
+    // Fusion optimization: ROPE + VIEW + SET_ROWS.
+    // The rope output is viewed as a 1D tensor and offset based on a row index in row_indices.
+    if (set_rows_stride != 0) {
+        idst = row_x * ne0 + i0;
+        idst += row_indices[channel_x] * set_rows_stride;
+    }
 
+    const auto & store_coaelsced = [&](float x0, float x1) {
+        if constexpr (std::is_same_v<float, D>) {
+            float2 v = make_float2(x0, x1);
+            ggml_cuda_memcpy_1<8>(dst + idst, &v);
+        } else if constexpr (std::is_same_v<half, D>) {
+            half2 v = make_half2(x0, x1);
+            ggml_cuda_memcpy_1<4>(dst + idst, &v);
+        }
+    };
+    if (i0 >= n_dims) {
+        store_coaelsced(x[ix + 0], x[ix + 1]);
         return;
     }
 
@@ -75,15 +104,26 @@ static __global__ void rope_norm(
     const float x0 = x[ix + 0];
     const float x1 = x[ix + 1];
 
-    dst[idst + 0] = x0*cos_theta - x1*sin_theta;
-    dst[idst + 1] = x0*sin_theta + x1*cos_theta;
+    store_coaelsced(x0 * cos_theta - x1 * sin_theta, x0 * sin_theta + x1 * cos_theta);
 }
 
-template<bool forward, bool has_ff, typename T>
-static __global__ void rope_neox(
-        const T * x, T * dst, const int ne0, const int ne1, const int s1, const int s2, const int n_dims,
-        const int32_t * pos, const float freq_scale, const float ext_factor, const float attn_factor,
-        const rope_corr_dims corr_dims, const float theta_scale, const float * freq_factors) {
+template <bool forward, bool has_ff, typename T, typename D>
+static __global__ void rope_neox(const T *            x,
+                                 D *                  dst,
+                                 const int            ne0,
+                                 const int            ne1,
+                                 const int            s1,
+                                 const int            s2,
+                                 const int            n_dims,
+                                 const int32_t *      pos,
+                                 const float          freq_scale,
+                                 const float          ext_factor,
+                                 const float          attn_factor,
+                                 const rope_corr_dims corr_dims,
+                                 const float          theta_scale,
+                                 const float *        freq_factors,
+                                 const int64_t *      row_indices,
+                                 const int            set_rows_stride) {
     const int i0 = 2*(blockDim.y*blockIdx.y + threadIdx.y);
 
     if (i0 >= ne0) {
@@ -95,12 +135,19 @@ static __global__ void rope_neox(
     const int row_x     = row_dst % ne1;
     const int channel_x = row_dst / ne1;
 
-    const int idst = row_dst*ne0 + i0/2;
+    int       idst = row_dst * ne0 + i0 / 2;
     const int ix   = channel_x*s2 + row_x*s1 + i0/2;
 
+    // Fusion optimization: ROPE + VIEW + SET_ROWS.
+    // The rope output is viewed as a 1D tensor and offset based on a row index in row_indices.
+    if (set_rows_stride != 0) {
+        idst = row_x * ne0 + i0 / 2;
+        idst += row_indices[channel_x] * set_rows_stride;
+    }
+
     if (i0 >= n_dims) {
-        dst[idst + i0/2 + 0] = x[ix + i0/2 + 0];
-        dst[idst + i0/2 + 1] = x[ix + i0/2 + 1];
+        dst[idst + i0 / 2 + 0] = ggml_cuda_cast<D>(x[ix + i0 / 2 + 0]);
+        dst[idst + i0 / 2 + 1] = ggml_cuda_cast<D>(x[ix + i0 / 2 + 1]);
 
         return;
     }
@@ -117,8 +164,8 @@ static __global__ void rope_neox(
     const float x0 = x[ix + 0];
     const float x1 = x[ix + n_dims/2];
 
-    dst[idst + 0]        = x0*cos_theta - x1*sin_theta;
-    dst[idst + n_dims/2] = x0*sin_theta + x1*cos_theta;
+    dst[idst + 0]          = ggml_cuda_cast<D>(x0 * cos_theta - x1 * sin_theta);
+    dst[idst + n_dims / 2] = ggml_cuda_cast<D>(x0 * sin_theta + x1 * cos_theta);
 }
 
 template<bool forward, bool has_ff, typename T>
@@ -238,11 +285,25 @@ static __global__ void rope_vision(
     dst[idst + n_dims] = x0*sin_theta + x1*cos_theta;
 }
 
-template<bool forward, typename T>
-static void rope_norm_cuda(
-        const T * x, T * dst, const int ne0, const int ne1, const int s1, const int s2, const int n_dims, const int nr,
-        const int32_t * pos, const float freq_scale, const float freq_base, const float ext_factor, const float attn_factor,
-        const rope_corr_dims corr_dims, const float * freq_factors, cudaStream_t stream) {
+template <bool forward, typename T, typename D>
+static void rope_norm_cuda(const T *            x,
+                           D *                  dst,
+                           const int            ne0,
+                           const int            ne1,
+                           const int            s1,
+                           const int            s2,
+                           const int            n_dims,
+                           const int            nr,
+                           const int32_t *      pos,
+                           const float          freq_scale,
+                           const float          freq_base,
+                           const float          ext_factor,
+                           const float          attn_factor,
+                           const rope_corr_dims corr_dims,
+                           const float *        freq_factors,
+                           const int64_t *      row_indices,
+                           const int            set_rows_stride,
+                           cudaStream_t         stream) {
     GGML_ASSERT(ne0 % 2 == 0);
     const dim3 block_dims(1, CUDA_ROPE_BLOCK_SIZE, 1);
     const int n_blocks_x = (ne0 + 2*CUDA_ROPE_BLOCK_SIZE - 1) / (2*CUDA_ROPE_BLOCK_SIZE);
@@ -252,20 +313,34 @@ static void rope_norm_cuda(
 
     if (freq_factors == nullptr) {
         rope_norm<forward, false><<<block_nums, block_dims, 0, stream>>>(
-            x, dst, ne0, ne1, s1, s2, n_dims, pos, freq_scale, ext_factor,
-            attn_factor, corr_dims, theta_scale, freq_factors);
+            x, dst, ne0, ne1, s1, s2, n_dims, pos, freq_scale, ext_factor, attn_factor, corr_dims, theta_scale,
+            freq_factors, row_indices, set_rows_stride);
     } else {
         rope_norm<forward, true><<<block_nums, block_dims, 0, stream>>>(
-            x, dst, ne0, ne1, s1, s2, n_dims, pos, freq_scale, ext_factor,
-            attn_factor, corr_dims, theta_scale, freq_factors);
+            x, dst, ne0, ne1, s1, s2, n_dims, pos, freq_scale, ext_factor, attn_factor, corr_dims, theta_scale,
+            freq_factors, row_indices, set_rows_stride);
     }
 }
 
-template<bool forward, typename T>
-static void rope_neox_cuda(
-        const T * x, T * dst, const int ne0, const int ne1, const int s1, const int s2, const int n_dims, const int nr,
-        const int32_t * pos, const float freq_scale, const float freq_base, const float ext_factor, const float attn_factor,
-        const rope_corr_dims corr_dims, const float * freq_factors, cudaStream_t stream) {
+template <bool forward, typename T, typename D>
+static void rope_neox_cuda(const T *            x,
+                           D *                  dst,
+                           const int            ne0,
+                           const int            ne1,
+                           const int            s1,
+                           const int            s2,
+                           const int            n_dims,
+                           const int            nr,
+                           const int32_t *      pos,
+                           const float          freq_scale,
+                           const float          freq_base,
+                           const float          ext_factor,
+                           const float          attn_factor,
+                           const rope_corr_dims corr_dims,
+                           const float *        freq_factors,
+                           const int64_t *      row_indices,
+                           const int            set_rows_stride,
+                           cudaStream_t         stream) {
     GGML_ASSERT(ne0 % 2 == 0);
     const dim3 block_dims(1, CUDA_ROPE_BLOCK_SIZE, 1);
     const int n_blocks_x = (ne0 + 2*CUDA_ROPE_BLOCK_SIZE - 1) / (2*CUDA_ROPE_BLOCK_SIZE);
@@ -274,13 +349,13 @@ static void rope_neox_cuda(
     const float theta_scale = powf(freq_base, -2.0f/n_dims);
 
     if (freq_factors == nullptr) {
-        rope_neox<forward, false, T><<<block_nums, block_dims, 0, stream>>>(
-            x, dst, ne0, ne1, s1, s2, n_dims, pos, freq_scale, ext_factor,
-            attn_factor, corr_dims, theta_scale, freq_factors);
+        rope_neox<forward, false><<<block_nums, block_dims, 0, stream>>>(
+            x, dst, ne0, ne1, s1, s2, n_dims, pos, freq_scale, ext_factor, attn_factor, corr_dims, theta_scale,
+            freq_factors, row_indices, set_rows_stride);
     } else {
-        rope_neox<forward, true, T><<<block_nums, block_dims, 0, stream>>>(
-            x, dst, ne0, ne1, s1, s2, n_dims, pos, freq_scale, ext_factor,
-            attn_factor, corr_dims, theta_scale, freq_factors);
+        rope_neox<forward, true><<<block_nums, block_dims, 0, stream>>>(
+            x, dst, ne0, ne1, s1, s2, n_dims, pos, freq_scale, ext_factor, attn_factor, corr_dims, theta_scale,
+            freq_factors, row_indices, set_rows_stride);
     }
 }
 
@@ -333,7 +408,9 @@ static void rope_vision_cuda(
 }
 
 template <bool forward>
-void ggml_cuda_op_rope_impl(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
+void ggml_cuda_op_rope_impl(ggml_backend_cuda_context & ctx,
+                            ggml_tensor *               dst,
+                            const ggml_tensor *         set_rows = nullptr) {
     const ggml_tensor * src0 = dst->src[0];
     const ggml_tensor * src1 = dst->src[1];
     const ggml_tensor * src2 = dst->src[2];
@@ -341,12 +418,25 @@ void ggml_cuda_op_rope_impl(ggml_backend_cuda_context & ctx, ggml_tensor * dst)
     const float * src0_d = (const float *)src0->data;
     const float * src1_d = (const float *)src1->data;
 
-    float * dst_d = (float *)dst->data;
+    void *          dst_d           = dst->data;
+    const int64_t * row_indices     = nullptr;
+    ggml_type       dst_type        = dst->type;
+    int             set_rows_stride = 0;
+
+    if (set_rows != nullptr) {
+        GGML_ASSERT(forward);
+        dst_d           = set_rows->data;
+        row_indices     = (const int64_t *) set_rows->src[1]->data;
+        dst_type        = set_rows->type;
+        set_rows_stride = set_rows->nb[1] / ggml_type_size(set_rows->type);
+    }
     cudaStream_t stream = ctx.stream();
 
     GGML_ASSERT(src0->type == GGML_TYPE_F32 || src0->type == GGML_TYPE_F16);
     GGML_ASSERT( dst->type == GGML_TYPE_F32 ||  dst->type == GGML_TYPE_F16);
-    GGML_ASSERT(src0->type == dst->type);
+    // When not fused, src0 and dst types must match
+    // When fused (ROPE+VIEW+SET_ROWS), src0 may be F32 and dst may be F16
+    GGML_ASSERT(src0->type == dst->type || (src0->type == GGML_TYPE_F32 && dst->type == GGML_TYPE_F16));
 
     const int64_t ne00 = src0->ne[0]; // head dims
     const int64_t ne01 = src0->ne[1]; // num heads
@@ -404,14 +494,18 @@ void ggml_cuda_op_rope_impl(ggml_backend_cuda_context & ctx, ggml_tensor * dst)
 
     // compute
     if (is_neox) {
-        if (src0->type == GGML_TYPE_F32) {
-            rope_neox_cuda<forward>(
-                (const float *) src0_d, (float *) dst_d, ne00, ne01, s01, s02, n_dims, nr, pos, freq_scale,
-                freq_base, ext_factor, attn_factor, corr_dims, freq_factors, stream);
-        } else if (src0->type == GGML_TYPE_F16) {
-            rope_neox_cuda<forward>(
-                (const half *) src0_d, (half *) dst_d, ne00, ne01, s01, s02, n_dims, nr, pos, freq_scale,
-                freq_base, ext_factor, attn_factor, corr_dims, freq_factors, stream);
+        if (src0->type == GGML_TYPE_F32 && dst_type == GGML_TYPE_F32) {
+            rope_neox_cuda<forward, float, float>((const float *) src0_d, (float *) dst_d, ne00, ne01, s01, s02, n_dims,
+                                                  nr, pos, freq_scale, freq_base, ext_factor, attn_factor, corr_dims,
+                                                  freq_factors, row_indices, set_rows_stride, stream);
+        } else if (src0->type == GGML_TYPE_F32 && dst_type == GGML_TYPE_F16) {
+            rope_neox_cuda<forward, float, half>((const float *) src0_d, (half *) dst_d, ne00, ne01, s01, s02, n_dims,
+                                                 nr, pos, freq_scale, freq_base, ext_factor, attn_factor, corr_dims,
+                                                 freq_factors, row_indices, set_rows_stride, stream);
+        } else if (src0->type == GGML_TYPE_F16 && dst_type == GGML_TYPE_F16) {
+            rope_neox_cuda<forward, half, half>((const half *) src0_d, (half *) dst_d, ne00, ne01, s01, s02, n_dims, nr,
+                                                pos, freq_scale, freq_base, ext_factor, attn_factor, corr_dims,
+                                                freq_factors, row_indices, set_rows_stride, stream);
         } else {
             GGML_ABORT("fatal error");
         }
@@ -440,14 +534,18 @@ void ggml_cuda_op_rope_impl(ggml_backend_cuda_context & ctx, ggml_tensor * dst)
             GGML_ABORT("fatal error");
         }
     } else {
-        if (src0->type == GGML_TYPE_F32) {
-            rope_norm_cuda<forward>(
-                (const float *) src0_d, (float *) dst_d, ne00, ne01, s01, s02, n_dims, nr, pos, freq_scale,
-                freq_base, ext_factor, attn_factor, corr_dims, freq_factors, stream);
-        } else if (src0->type == GGML_TYPE_F16) {
-            rope_norm_cuda<forward>(
-                (const half *) src0_d, (half *) dst_d, ne00, ne01, s01, s02, n_dims, nr, pos, freq_scale,
-                freq_base, ext_factor, attn_factor, corr_dims, freq_factors, stream);
+        if (src0->type == GGML_TYPE_F32 && dst_type == GGML_TYPE_F32) {
+            rope_norm_cuda<forward, float, float>((const float *) src0_d, (float *) dst_d, ne00, ne01, s01, s02, n_dims,
+                                                  nr, pos, freq_scale, freq_base, ext_factor, attn_factor, corr_dims,
+                                                  freq_factors, row_indices, set_rows_stride, stream);
+        } else if (src0->type == GGML_TYPE_F32 && dst_type == GGML_TYPE_F16) {
+            rope_norm_cuda<forward, float, half>((const float *) src0_d, (half *) dst_d, ne00, ne01, s01, s02, n_dims,
+                                                 nr, pos, freq_scale, freq_base, ext_factor, attn_factor, corr_dims,
+                                                 freq_factors, row_indices, set_rows_stride, stream);
+        } else if (src0->type == GGML_TYPE_F16 && dst_type == GGML_TYPE_F16) {
+            rope_norm_cuda<forward, half, half>((const half *) src0_d, (half *) dst_d, ne00, ne01, s01, s02, n_dims, nr,
+                                                pos, freq_scale, freq_base, ext_factor, attn_factor, corr_dims,
+                                                freq_factors, row_indices, set_rows_stride, stream);
         } else {
             GGML_ABORT("fatal error");
         }
@@ -461,3 +559,7 @@ void ggml_cuda_op_rope(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
 void ggml_cuda_op_rope_back(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
     ggml_cuda_op_rope_impl<false>(ctx, dst);
 }
+
+void ggml_cuda_op_rope_fused(ggml_backend_cuda_context & ctx, ggml_tensor * rope, ggml_tensor * set_rows) {
+    ggml_cuda_op_rope_impl<true>(ctx, rope, set_rows);
+}

ggml/src/ggml-cuda/rope.cuh

@@ -5,3 +5,5 @@
 void ggml_cuda_op_rope(ggml_backend_cuda_context & ctx, ggml_tensor * dst);
 
 void ggml_cuda_op_rope_back(ggml_backend_cuda_context & ctx, ggml_tensor * dst);
+
+void ggml_cuda_op_rope_fused(ggml_backend_cuda_context & ctx, ggml_tensor * dst, ggml_tensor * set_rows);

ggml/src/ggml-cuda/unary.cu

@@ -81,6 +81,14 @@ static __device__ __forceinline__ float op_log(float x) {
     return logf(x);
 }
 
+static __device__ __forceinline__ float op_expm1(float x) {
+    return expm1f(x);
+}
+
+static __device__ __forceinline__ float op_softplus(float x) {
+    return (x > 20.0f) ? x : logf(1.0f + expf(x));
+}
+
 static __device__ __forceinline__ float op_elu(float x) {
     return (x > 0.f) ? x : expm1f(x);
 }
@@ -233,6 +241,14 @@ void ggml_cuda_op_round(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
 void ggml_cuda_op_trunc(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
     ggml_cuda_op_unary<op_trunc>(ctx, dst);
 }
+
+void ggml_cuda_op_expm1(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
+    ggml_cuda_op_unary<op_expm1>(ctx, dst);
+}
+
+void ggml_cuda_op_softplus(ggml_backend_cuda_context & ctx, ggml_tensor * dst) {
+    ggml_cuda_op_unary<op_softplus>(ctx, dst);
+}
 /* gated ops */
 
 template <float (*op)(float), typename T>

ggml/src/ggml-cuda/unary.cuh

@@ -61,6 +61,10 @@ void ggml_cuda_op_cos(ggml_backend_cuda_context & ctx, ggml_tensor * dst);
 
 void ggml_cuda_op_log(ggml_backend_cuda_context & ctx, ggml_tensor * dst);
 
+void ggml_cuda_op_expm1(ggml_backend_cuda_context & ctx, ggml_tensor * dst);
+
+void ggml_cuda_op_softplus(ggml_backend_cuda_context & ctx, ggml_tensor * dst);
+
 void ggml_cuda_op_elu(ggml_backend_cuda_context & ctx, ggml_tensor * dst);
 
 void ggml_cuda_op_floor(ggml_backend_cuda_context & ctx, ggml_tensor * dst);

ggml/src/ggml-hexagon/ggml-hexagon.cpp

@@ -3156,26 +3156,17 @@ static inline bool op_reuse_src1(const ggml_tensor * op1, const ggml_tensor * op
     return (op0 && op0->src[1] == op1->src[1]);
 }
 
+static inline bool is_compute_op(ggml_tensor *node)
+{
+    return !(ggml_op_is_empty(node->op) || ggml_is_empty(node));
+}
+
 // scan the graph and figure out last compute op index
 static inline int last_compute_op(ggml_cgraph * graph) {
-    int last;
+    int last = 0;
     for (int i = 0; i < graph->n_nodes; ++i) {
-        ggml_tensor * node = graph->nodes[i];
-
-        switch (node->op) {
-            case GGML_OP_MUL_MAT:
-            case GGML_OP_MUL_MAT_ID:
-            case GGML_OP_MUL:
-            case GGML_OP_ADD:
-            case GGML_OP_SUB:
-            case GGML_OP_RMS_NORM:
-            case GGML_OP_GLU:
-            case GGML_OP_ADD_ID:
-                last = i;
-                break;
-
-            default:
-                break;
+        if (is_compute_op(graph->nodes[i])) {
+            last = i;
         }
     }
 
@@ -3194,6 +3185,10 @@ static ggml_status ggml_backend_hexagon_graph_compute(ggml_backend_t backend, gg
     for (int i = 0; i < graph->n_nodes; ++i) {
         ggml_tensor * node = graph->nodes[i];
 
+        if (!is_compute_op(node)) {
+            continue;
+        }
+
         uint32_t flags = 0;
 
         // skip quantizer if src1 is reused
@@ -3245,14 +3240,6 @@ static ggml_status ggml_backend_hexagon_graph_compute(ggml_backend_t backend, gg
                 ggml_hexagon_rope(node, flags);
                 break;
 
-            // non-compute ops
-            case GGML_OP_NONE:
-            case GGML_OP_RESHAPE:
-            case GGML_OP_VIEW:
-            case GGML_OP_PERMUTE:
-            case GGML_OP_TRANSPOSE:
-                break;
-
             default:
                 GGML_ABORT("\nggml-hex: graph-compute %s is not supported\n", ggml_op_desc(node));
         }

ggml/src/ggml-hexagon/htp/binary-ops.c

@@ -34,6 +34,11 @@ static hvx_elemwise_f32_func func_table_HVX[]     = { hvx_mul_f32, hvx_add_f32,
 static hvx_elemwise_f32_func func_table_HVX_opt[] = { hvx_mul_f32_opt, hvx_add_f32_opt, hvx_sub_f32_opt };
 
 #define htp_binary_preamble            \
+    const struct htp_tensor * src0 = &octx->src0; \
+    const struct htp_tensor * src1 = &octx->src1; \
+    const struct htp_tensor * src2 = &octx->src2; \
+    struct htp_tensor *       dst  = &octx->dst;  \
+                                       \
     const uint32_t ne00 = src0->ne[0]; \
     const uint32_t ne01 = src0->ne[1]; \
     const uint32_t ne02 = src0->ne[2]; \
@@ -62,16 +67,15 @@ static hvx_elemwise_f32_func func_table_HVX_opt[] = { hvx_mul_f32_opt, hvx_add_f
     const uint32_t nb0 = dst->nb[0];   \
     const uint32_t nb1 = dst->nb[1];   \
     const uint32_t nb2 = dst->nb[2];   \
-    const uint32_t nb3 = dst->nb[3];
+    const uint32_t nb3 = dst->nb[3];   \
+                                       \
+    const uint32_t src0_nrows_per_thread = octx->src0_nrows_per_thread;
 
-static void binary_job_f32_per_thread(const struct htp_tensor * src0,
-                                      const struct htp_tensor * src1,
-                                      struct htp_tensor *       dst,
-                                      uint8_t *                 spad_data,
-                                      uint32_t                  nth,
-                                      uint32_t                  ith,
-                                      uint32_t                  src0_nrows_per_thread,
-                                      enum htp_op               op) {
+static void binary_job_f32_per_thread(struct htp_ops_context * octx,
+                                      uint8_t *                spad_data,
+                                      uint32_t                 nth,
+                                      uint32_t                 ith,
+                                      enum htp_op              op) {
     htp_binary_preamble;
 
     const size_t src0_row_size = nb01;
@@ -107,16 +111,23 @@ static void binary_job_f32_per_thread(const struct htp_tensor * src0,
 
     uint8_t * restrict spad_data_th = spad_data + (ith * src0_row_size);
 
-    const uint32_t nr0 = ne00 / ne10;
-
     const uint8_t * restrict src0_ptr = (const uint8_t *) src0->data + (src0_start_row * src0_row_size);
     uint8_t * restrict dst_ptr        = (uint8_t *) dst->data + (src0_start_row * dst_row_size);
 
     const uint8_t * restrict data_src1 = (const uint8_t *) src1->data;
-    const uint8_t * restrict src1_ptr  = NULL;
+
+    const uint32_t ne02_ne01 = ne02 * ne01;
 
     for (uint32_t ir = src0_start_row; ir < src0_end_row; ir++) {
-        src1_ptr = data_src1 + (ir % src1_nrows) * src1_row_size;
+        const uint32_t i03 = fastdiv(ir, &octx->src0_div21);
+        const uint32_t i02 = fastdiv(ir - i03 * ne02_ne01, &octx->src0_div1);
+        const uint32_t i01 = (ir - i03 * ne02_ne01 - i02 * ne01);
+
+        const uint32_t i13 = fastmodulo(i03, ne13, &octx->src1_div3);
+        const uint32_t i12 = fastmodulo(i02, ne12, &octx->src1_div2);
+        const uint32_t i11 = fastmodulo(i01, ne11, &octx->src1_div1);
+
+        const uint8_t * restrict src1_ptr = data_src1 + i13 * nb13 + i12 * nb12 + i11 * src1_row_size;
 
         if (ir + 1 < src0_end_row) {
             htp_l2fetch(src0_ptr + ne00, 1, src0_row_size, src0_row_size);
@@ -125,6 +136,7 @@ static void binary_job_f32_per_thread(const struct htp_tensor * src0,
             }
         }
 
+        const uint32_t nr0 = ne00 / ne10;
         if (nr0 > 1) {
             if ((1 == is_aligned) && (nr0 == ne00)) {
                 hvx_bcast_fp32_a(spad_data_th, *(float *) src1_ptr, nr0);
@@ -149,22 +161,17 @@ static void binary_job_f32_per_thread(const struct htp_tensor * src0,
          (unsigned) HAP_perf_qtimer_count_to_us(t2 - t1));
 }
 
-static void binary_add_id_job_f32_per_thread(const struct htp_tensor * src0,
-                                             const struct htp_tensor * src1,
-                                             const struct htp_tensor * src2,
-                                             struct htp_tensor *       dst,
-                                             uint8_t *                 spad_data,
-                                             uint32_t                  nth,
-                                             uint32_t                  ith,
-                                             uint32_t                  src0_nrows_per_thread,
-                                             hvx_elemwise_f32_func     func_HVX) {
+static void binary_add_id_job_f32_per_thread(struct htp_ops_context * octx,
+                                             uint8_t *                spad_data,
+                                             uint32_t                 nth,
+                                             uint32_t                 ith,
+                                             hvx_elemwise_f32_func    func_HVX) {
     htp_binary_preamble;
 
     const size_t src0_row_size = nb01;
     const size_t src1_row_size = nb11;
     const size_t dst_row_size  = nb1;
 
-    const uint32_t ne02_ne01  = ne02 * ne01;
     const uint32_t src0_nrows = ne01 * ne02 * ne03;  // src0 rows
 
     const uint32_t src0_start_row = src0_nrows_per_thread * ith;
@@ -187,10 +194,11 @@ static void binary_add_id_job_f32_per_thread(const struct htp_tensor * src0,
     const uint8_t * restrict data_src1 = (const uint8_t *) src1->data;
     uint8_t * restrict data_dst        = (uint8_t *) dst->data;
 
+    const uint32_t ne02_ne01  = ne02 * ne01;
     for (uint32_t ir = src0_start_row; ir < src0_end_row; ir++) {
         // src0 indices
-        const uint32_t i03 = ir / ne02_ne01;
-        const uint32_t i02 = (ir - i03 * ne02_ne01) / ne01;
+        const uint32_t i03 = fastdiv(ir, &octx->src0_div21);
+        const uint32_t i02 = fastdiv(ir - i03 * ne02_ne01, &octx->src0_div1);
         const uint32_t i01 = (ir - i03 * ne02_ne01 - i02 * ne01);
 
         // src1 indices
@@ -234,13 +242,11 @@ static void binary_job_dispatcher_f32(unsigned int n, unsigned int i, void * dat
         case HTP_OP_MUL:
         case HTP_OP_ADD:
         case HTP_OP_SUB:
-            binary_job_f32_per_thread(&octx->src0, &octx->src1, &octx->dst, octx->src1_spad.data, n, i,
-                                      octx->src0_nrows_per_thread, octx->op);
+            binary_job_f32_per_thread(octx, octx->src1_spad.data, n, i, octx->op);
             break;
 
         case HTP_OP_ADD_ID:
-            binary_add_id_job_f32_per_thread(&octx->src0, &octx->src1, &octx->src2, &octx->dst, octx->src0_spad.data, n,
-                                             i, octx->src0_nrows_per_thread, hvx_add_f32);
+            binary_add_id_job_f32_per_thread(octx, octx->src0_spad.data, n, i, hvx_add_f32);
             break;
 
         default:
@@ -321,6 +327,16 @@ static int execute_op_binary_f32(struct htp_ops_context * octx) {
 
         octx->src0_nrows_per_thread = (src0_nrows + n_jobs - 1) / n_jobs;
 
+        octx->src0_div21 = init_fastdiv_values(src0->ne[2] * src0->ne[1]);
+        octx->src0_div3  = init_fastdiv_values(src0->ne[3]);
+        octx->src0_div2  = init_fastdiv_values(src0->ne[2]);
+        octx->src0_div1  = init_fastdiv_values(src0->ne[1]);
+
+        octx->src1_div21 = init_fastdiv_values(src1->ne[2] * src1->ne[1]);
+        octx->src1_div3  = init_fastdiv_values(src1->ne[3]);
+        octx->src1_div2  = init_fastdiv_values(src1->ne[2]);
+        octx->src1_div1  = init_fastdiv_values(src1->ne[1]);
+
         worker_pool_run_func(octx->ctx->worker_pool, binary_op_func, octx, n_jobs);
     }
 

ggml/src/ggml-hexagon/htp/htp-msg.h

@@ -119,10 +119,10 @@ static const char * htp_type_name(uint32_t t) {
 #define HTP_MAX_DIMS 4
 
 struct htp_tensor {
-    uint32_t data;              // Buffer offset in the messages, and data pointer on the NSP
-    uint32_t type;              // Data type
-    uint32_t ne[HTP_MAX_DIMS];  // Number of elements
-    uint32_t nb[HTP_MAX_DIMS];  // Stride in bytes (see ggml.h ggml_tensor)
+    uint32_t data;                // Buffer offset in the messages, and data pointer on the NSP
+    uint32_t type;                // Data type
+    uint32_t ne[HTP_MAX_DIMS];    // Number of elements
+    uint32_t nb[HTP_MAX_DIMS];    // Stride in bytes (see ggml.h ggml_tensor)
 };
 
 #define HTP_MAX_OP_PARAMS 64

ggml/src/ggml-hexagon/htp/htp-ops.h

@@ -4,6 +4,7 @@
 #include "htp-ctx.h"
 #include "htp-msg.h"
 #include "worker-pool.h"
+#include "ops-utils.h"
 
 #include <assert.h>
 #include <stdint.h>
@@ -38,6 +39,16 @@ struct htp_ops_context {
     uint32_t src0_nrows_per_thread;
     uint32_t src1_nrows_per_thread;
 
+    struct fastdiv_values src0_div1;  // fastdiv values for ne1
+    struct fastdiv_values src0_div2;  // fastdiv values for ne2
+    struct fastdiv_values src0_div3;  // fastdiv values for ne3
+    struct fastdiv_values src0_div21; // fastdiv values for ne2 * ne1
+
+    struct fastdiv_values src1_div1;  // fastdiv values for ne1
+    struct fastdiv_values src1_div2;  // fastdiv values for ne2
+    struct fastdiv_values src1_div3;  // fastdiv values for ne3
+    struct fastdiv_values src1_div21; // fastdiv values for ne2 * ne1
+
     uint32_t flags;
 };
 

ggml/src/ggml-hexagon/htp/ops-utils.h

@@ -31,6 +31,39 @@ static inline uint32_t htp_round_up(uint32_t n, uint32_t m) {
     return m * ((n + m - 1) / m);
 }
 
+// See https://gmplib.org/~tege/divcnst-pldi94.pdf figure 4.1.
+// Precompute mp (m' in the paper) and L such that division
+// can be computed using a multiply (high 32b of 64b result)
+// and a shift:
+//
+// n/d = (mulhi(n, mp) + n) >> L;
+struct fastdiv_values {
+    uint32_t mp;
+    uint32_t l;
+};
+
+static inline struct fastdiv_values init_fastdiv_values(uint32_t d) {
+    struct fastdiv_values result = { 0, 0 };
+    // compute L = ceil(log2(d));
+    while (result.l < 32 && ((uint32_t) 1 << result.l) < d) {
+        ++(result.l);
+    }
+
+    result.mp = (uint32_t) (((uint64_t) 1 << 32) * (((uint64_t) 1 << result.l) - d) / d + 1);
+    return result;
+}
+
+static inline uint32_t fastdiv(uint32_t n, const struct fastdiv_values * vals) {
+    // Compute high 32 bits of n * mp
+    const uint32_t hi = (uint32_t) (((uint64_t) n * vals->mp) >> 32);  // mulhi(n, mp)
+    // add n, apply bit shift
+    return (hi + n) >> vals->l;
+}
+
+static inline uint32_t fastmodulo(uint32_t n, uint32_t d, const struct fastdiv_values * vals) {
+    return n - fastdiv(n, vals) * d;
+}
+
 static inline void htp_l2fetch(const void * p, uint32_t height, uint32_t width, uint32_t stride) {
     const uint64_t control = Q6_P_combine_RR(stride, Q6_R_combine_RlRl(width, height));
     asm volatile(" l2fetch(%0,%1) " : : "r"(p), "r"(control));

ggml/src/ggml-impl.h

@@ -102,7 +102,7 @@ static bool ggml_op_is_empty(enum ggml_op op) {
     }
 }
 
-static inline float ggml_softplus(float input) {
+static inline float ggml_compute_softplus_f32(float input) {
     return (input > 20.0f) ? input : logf(1 + expf(input));
 }
 //

ggml/src/ggml-metal/ggml-metal-device.cpp

@@ -318,6 +318,44 @@ ggml_metal_pipeline_t ggml_metal_library_get_pipeline_sum_rows(ggml_metal_librar
     return res;
 }
 
+ggml_metal_pipeline_t ggml_metal_library_get_pipeline_cumsum_blk(ggml_metal_library_t lib, const ggml_tensor * op) {
+    GGML_ASSERT(op->op == GGML_OP_CUMSUM);
+
+    char base[256];
+    char name[256];
+
+    snprintf(base, 256, "kernel_cumsum_blk_%s", ggml_type_name(op->src[0]->type));
+    snprintf(name, 256, "%s", base);
+
+    ggml_metal_pipeline_t res = ggml_metal_library_get_pipeline(lib, name);
+    if (res) {
+        return res;
+    }
+
+    res = ggml_metal_library_compile_pipeline(lib, base, name, nullptr);
+
+    return res;
+}
+
+ggml_metal_pipeline_t ggml_metal_library_get_pipeline_cumsum_add(ggml_metal_library_t lib, const ggml_tensor * op) {
+    GGML_ASSERT(op->op == GGML_OP_CUMSUM);
+
+    char base[256];
+    char name[256];
+
+    snprintf(base, 256, "kernel_cumsum_add_%s", ggml_type_name(op->src[0]->type));
+    snprintf(name, 256, "%s", base);
+
+    ggml_metal_pipeline_t res = ggml_metal_library_get_pipeline(lib, name);
+    if (res) {
+        return res;
+    }
+
+    res = ggml_metal_library_compile_pipeline(lib, base, name, nullptr);
+
+    return res;
+}
+
 ggml_metal_pipeline_t ggml_metal_library_get_pipeline_soft_max(ggml_metal_library_t lib, const ggml_tensor * op) {
     GGML_ASSERT(!op->src[1] || op->src[1]->type == GGML_TYPE_F16 || op->src[1]->type == GGML_TYPE_F32);
 
@@ -943,6 +981,34 @@ ggml_metal_pipeline_t ggml_metal_library_get_pipeline_argsort(ggml_metal_library
     return res;
 }
 
+ggml_metal_pipeline_t ggml_metal_library_get_pipeline_argsort_merge(ggml_metal_library_t lib, const ggml_tensor * op) {
+    assert(op->op == GGML_OP_ARGSORT);
+
+    char base[256];
+    char name[256];
+
+    ggml_sort_order order = (ggml_sort_order) op->op_params[0];
+
+    const char * order_str = "undefined";
+    switch (order) {
+        case GGML_SORT_ORDER_ASC:  order_str = "asc";  break;
+        case GGML_SORT_ORDER_DESC: order_str = "desc"; break;
+        default: GGML_ABORT("fatal error");
+    };
+
+    snprintf(base, 256, "kernel_argsort_merge_%s_%s_%s", ggml_type_name(op->src[0]->type), ggml_type_name(op->type), order_str);
+    snprintf(name, 256, "%s", base);
+
+    ggml_metal_pipeline_t res = ggml_metal_library_get_pipeline(lib, name);
+    if (res) {
+        return res;
+    }
+
+    res = ggml_metal_library_compile_pipeline(lib, base, name, nullptr);
+
+    return res;
+}
+
 ggml_metal_pipeline_t ggml_metal_library_get_pipeline_flash_attn_ext_pad(
         ggml_metal_library_t lib,
         const struct ggml_tensor * op,
@@ -1438,6 +1504,30 @@ ggml_metal_pipeline_t ggml_metal_library_get_pipeline_conv_transpose_2d(ggml_met
     return res;
 }
 
+ggml_metal_pipeline_t ggml_metal_library_get_pipeline_conv_2d(ggml_metal_library_t lib, const ggml_tensor * op) {
+    assert(op->op == GGML_OP_CONV_2D);
+
+    GGML_ASSERT(ggml_is_contiguous(op->src[0]));
+    GGML_ASSERT(op->src[0]->type == GGML_TYPE_F16 || op->src[0]->type == GGML_TYPE_F32);
+    GGML_ASSERT(op->src[1]->type == GGML_TYPE_F32);
+    GGML_ASSERT(op->type         == GGML_TYPE_F32);
+
+    char base[256];
+    char name[256];
+
+    snprintf(base, 256, "kernel_conv_2d_%s_%s", ggml_type_name(op->src[0]->type), ggml_type_name(op->src[1]->type));
+    snprintf(name, 256, "%s", base);
+
+    ggml_metal_pipeline_t res = ggml_metal_library_get_pipeline(lib, name);
+    if (res) {
+        return res;
+    }
+
+    res = ggml_metal_library_compile_pipeline(lib, base, name, nullptr);
+
+    return res;
+}
+
 ggml_metal_pipeline_t ggml_metal_library_get_pipeline_upscale(ggml_metal_library_t lib, const ggml_tensor * op) {
     assert(op->op == GGML_OP_UPSCALE);
 

ggml/src/ggml-metal/ggml-metal-device.h

@@ -113,6 +113,8 @@ ggml_metal_pipeline_t ggml_metal_library_get_pipeline_unary             (ggml_me
 ggml_metal_pipeline_t ggml_metal_library_get_pipeline_glu               (ggml_metal_library_t lib, const struct ggml_tensor * op);
 ggml_metal_pipeline_t ggml_metal_library_get_pipeline_sum               (ggml_metal_library_t lib, const struct ggml_tensor * op);
 ggml_metal_pipeline_t ggml_metal_library_get_pipeline_sum_rows          (ggml_metal_library_t lib, const struct ggml_tensor * op);
+ggml_metal_pipeline_t ggml_metal_library_get_pipeline_cumsum_blk        (ggml_metal_library_t lib, const struct ggml_tensor * op);
+ggml_metal_pipeline_t ggml_metal_library_get_pipeline_cumsum_add        (ggml_metal_library_t lib, const struct ggml_tensor * op);
 ggml_metal_pipeline_t ggml_metal_library_get_pipeline_soft_max          (ggml_metal_library_t lib, const struct ggml_tensor * op);
 ggml_metal_pipeline_t ggml_metal_library_get_pipeline_ssm_conv          (ggml_metal_library_t lib, const struct ggml_tensor * op);
 ggml_metal_pipeline_t ggml_metal_library_get_pipeline_ssm_scan          (ggml_metal_library_t lib, const struct ggml_tensor * op);
@@ -125,6 +127,7 @@ ggml_metal_pipeline_t ggml_metal_library_get_pipeline_mul_mm_id         (ggml_me
 ggml_metal_pipeline_t ggml_metal_library_get_pipeline_mul_mv_id         (ggml_metal_library_t lib, const struct ggml_tensor * op);
 ggml_metal_pipeline_t ggml_metal_library_get_pipeline_argmax            (ggml_metal_library_t lib, const struct ggml_tensor * op);
 ggml_metal_pipeline_t ggml_metal_library_get_pipeline_argsort           (ggml_metal_library_t lib, const struct ggml_tensor * op);
+ggml_metal_pipeline_t ggml_metal_library_get_pipeline_argsort_merge     (ggml_metal_library_t lib, const struct ggml_tensor * op);
 ggml_metal_pipeline_t ggml_metal_library_get_pipeline_bin               (ggml_metal_library_t lib, enum ggml_op op, int32_t n_fuse, bool row);
 ggml_metal_pipeline_t ggml_metal_library_get_pipeline_l2_norm           (ggml_metal_library_t lib, const struct ggml_tensor * op);
 ggml_metal_pipeline_t ggml_metal_library_get_pipeline_group_norm        (ggml_metal_library_t lib, const struct ggml_tensor * op);
@@ -133,6 +136,7 @@ ggml_metal_pipeline_t ggml_metal_library_get_pipeline_rope              (ggml_me
 ggml_metal_pipeline_t ggml_metal_library_get_pipeline_im2col            (ggml_metal_library_t lib, const struct ggml_tensor * op);
 ggml_metal_pipeline_t ggml_metal_library_get_pipeline_conv_transpose_1d (ggml_metal_library_t lib, const struct ggml_tensor * op);
 ggml_metal_pipeline_t ggml_metal_library_get_pipeline_conv_transpose_2d (ggml_metal_library_t lib, const struct ggml_tensor * op);
+ggml_metal_pipeline_t ggml_metal_library_get_pipeline_conv_2d           (ggml_metal_library_t lib, const struct ggml_tensor * op);
 ggml_metal_pipeline_t ggml_metal_library_get_pipeline_upscale           (ggml_metal_library_t lib, const struct ggml_tensor * op);
 ggml_metal_pipeline_t ggml_metal_library_get_pipeline_pad               (ggml_metal_library_t lib, const struct ggml_tensor * op);
 ggml_metal_pipeline_t ggml_metal_library_get_pipeline_pad_reflect_1d    (ggml_metal_library_t lib, const struct ggml_tensor * op);

ggml/src/ggml-metal/ggml-metal-device.m

@@ -870,6 +870,7 @@ bool ggml_metal_device_supports_op(ggml_metal_device_t dev, const struct ggml_te
         case GGML_OP_SUM:
             return has_simdgroup_reduction && ggml_is_contiguous(op->src[0]);
         case GGML_OP_SUM_ROWS:
+        case GGML_OP_CUMSUM:
         case GGML_OP_MEAN:
         case GGML_OP_SOFT_MAX:
         case GGML_OP_GROUP_NORM:
@@ -885,6 +886,11 @@ bool ggml_metal_device_supports_op(ggml_metal_device_t dev, const struct ggml_te
             return true;
         case GGML_OP_IM2COL:
             return ggml_is_contiguous(op->src[1]) && op->src[1]->type == GGML_TYPE_F32 && (op->type == GGML_TYPE_F16 || op->type == GGML_TYPE_F32);
+        case GGML_OP_CONV_2D:
+            return ggml_is_contiguous(op->src[0]) &&
+                   op->src[1]->type == GGML_TYPE_F32 &&
+                   op->type == GGML_TYPE_F32 &&
+                   (op->src[0]->type == GGML_TYPE_F16 || op->src[0]->type == GGML_TYPE_F32);
         case GGML_OP_POOL_1D:
             return false;
         case GGML_OP_UPSCALE:
@@ -899,8 +905,6 @@ bool ggml_metal_device_supports_op(ggml_metal_device_t dev, const struct ggml_te
         case GGML_OP_LEAKY_RELU:
             return op->src[0]->type == GGML_TYPE_F32;
         case GGML_OP_ARGSORT:
-            // TODO: Support arbitrary column width
-            return op->src[0]->ne[0] <= 1024;
         case GGML_OP_ARANGE:
             return true;
         case GGML_OP_FLASH_ATTN_EXT:
@@ -985,7 +989,7 @@ bool ggml_metal_device_supports_op(ggml_metal_device_t dev, const struct ggml_te
                                 return false;
                         }
                     case GGML_TYPE_I32:
-                        return op->type == GGML_TYPE_F32;
+                        return op->type == GGML_TYPE_F32 || op->type == GGML_TYPE_I32;
                     default:
                         return false;
                 };

ggml/src/ggml-metal/ggml-metal-impl.h

@@ -528,6 +528,36 @@ typedef struct {
     uint64_t nb2;
 } ggml_metal_kargs_conv_transpose_2d;
 
+typedef struct {
+    uint64_t nb00;
+    uint64_t nb01;
+    uint64_t nb02;
+    uint64_t nb03;
+    uint64_t nb10;
+    uint64_t nb11;
+    uint64_t nb12;
+    uint64_t nb13;
+    uint64_t nb0;
+    uint64_t nb1;
+    uint64_t nb2;
+    uint64_t nb3;
+    int32_t  IW;
+    int32_t  IH;
+    int32_t  KW;
+    int32_t  KH;
+    int32_t  IC;
+    int32_t  OC;
+    int32_t  OW;
+    int32_t  OH;
+    int32_t  N;
+    int32_t  s0;
+    int32_t  s1;
+    int32_t  p0;
+    int32_t  p1;
+    int32_t  d0;
+    int32_t  d1;
+} ggml_metal_kargs_conv_2d;
+
 typedef struct {
     uint64_t  ofs0;
     uint64_t  ofs1;
@@ -582,6 +612,45 @@ typedef struct {
     uint64_t nb3;
 } ggml_metal_kargs_sum_rows;
 
+typedef struct {
+    int64_t  ne00;
+    int64_t  ne01;
+    int64_t  ne02;
+    int64_t  ne03;
+    uint64_t nb00;
+    uint64_t nb01;
+    uint64_t nb02;
+    uint64_t nb03;
+    int64_t  net0;
+    int64_t  net1;
+    int64_t  net2;
+    int64_t  net3;
+    uint64_t nbt0;
+    uint64_t nbt1;
+    uint64_t nbt2;
+    uint64_t nbt3;
+    bool     outb;
+} ggml_metal_kargs_cumsum_blk;
+
+typedef struct {
+    int64_t  ne00;
+    int64_t  ne01;
+    int64_t  ne02;
+    int64_t  ne03;
+    uint64_t nb00;
+    uint64_t nb01;
+    uint64_t nb02;
+    uint64_t nb03;
+    int64_t  net0;
+    int64_t  net1;
+    int64_t  net2;
+    int64_t  net3;
+    uint64_t nbt0;
+    uint64_t nbt1;
+    uint64_t nbt2;
+    uint64_t nbt3;
+} ggml_metal_kargs_cumsum_add;
+
 typedef struct {
     int32_t  ne00;
     int32_t  ne01;
@@ -763,10 +832,28 @@ typedef struct {
 } ggml_metal_kargs_leaky_relu;
 
 typedef struct {
-    int64_t  ncols;
-    int64_t  ncols_pad;
+    int64_t  ne00;
+    int64_t  ne01;
+    int64_t  ne02;
+    int64_t  ne03;
+    uint64_t nb00;
+    uint64_t nb01;
+    uint64_t nb02;
+    uint64_t nb03;
 } ggml_metal_kargs_argsort;
 
+typedef struct {
+    int64_t  ne00;
+    int64_t  ne01;
+    int64_t  ne02;
+    int64_t  ne03;
+    uint64_t nb00;
+    uint64_t nb01;
+    uint64_t nb02;
+    uint64_t nb03;
+    int32_t  len;
+} ggml_metal_kargs_argsort_merge;
+
 typedef struct {
     int64_t  ne0;
     float    start;

ggml/src/ggml-metal/ggml-metal-ops.cpp

@@ -10,6 +10,7 @@
 
 #include <cassert>
 #include <algorithm>
+#include <limits>
 
 static ggml_metal_buffer_id ggml_metal_get_buffer_id(const ggml_tensor * t) {
     if (!t) {
@@ -310,6 +311,10 @@ static int ggml_metal_op_encode_impl(ggml_metal_op_t ctx, int idx) {
             {
                 n_fuse = ggml_metal_op_sum_rows(ctx, idx);
             } break;
+        case GGML_OP_CUMSUM:
+            {
+                n_fuse = ggml_metal_op_cumsum(ctx, idx);
+            } break;
         case GGML_OP_SOFT_MAX:
             {
                 n_fuse = ggml_metal_op_soft_max(ctx, idx);
@@ -364,6 +369,10 @@ static int ggml_metal_op_encode_impl(ggml_metal_op_t ctx, int idx) {
             {
                 n_fuse = ggml_metal_op_im2col(ctx, idx);
             } break;
+        case GGML_OP_CONV_2D:
+            {
+                n_fuse = ggml_metal_op_conv_2d(ctx, idx);
+            } break;
         case GGML_OP_CONV_TRANSPOSE_1D:
             {
                 n_fuse = ggml_metal_op_conv_transpose_1d(ctx, idx);
@@ -534,7 +543,7 @@ int ggml_metal_op_repeat(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne0, op->src[0], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb0, op->src[0], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     ggml_metal_pipeline_t pipeline = ggml_metal_library_get_pipeline_repeat(lib, op->type);
 
@@ -580,7 +589,7 @@ int ggml_metal_op_acc(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne1, op->src[1], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb1, op->src[1], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     GGML_ASSERT(op->src[0]->type == GGML_TYPE_F32);
     GGML_ASSERT(op->src[1]->type == GGML_TYPE_F32);
@@ -689,7 +698,7 @@ int ggml_metal_op_scale(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne0, op->src[0], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb0, op->src[0], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     float scale;
     float bias;
@@ -728,7 +737,7 @@ int ggml_metal_op_clamp(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne0, op->src[0], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb0, op->src[0], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     float min;
     float max;
@@ -767,7 +776,7 @@ int ggml_metal_op_unary(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne0, op->src[0], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb0, op->src[0], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     int64_t n = ggml_nelements(op);
 
@@ -797,7 +806,7 @@ int ggml_metal_op_glu(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne1, op->src[1], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb1, op->src[1], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     if (op->src[1]) {
         GGML_ASSERT(ggml_are_same_shape(op->src[0], op->src[1]));
@@ -829,18 +838,6 @@ int ggml_metal_op_glu(ggml_metal_op_t ctx, int idx) {
 
     const int32_t nth = std::min(ggml_metal_pipeline_max_theads_per_threadgroup(pipeline), ne00/2);
 
-    //[encoder setComputePipelineState:pipeline];
-    //[encoder setBuffer:id_src0 offset:offs_src0 atIndex:0];
-    //if (src1) {
-    //    [encoder setBuffer:id_src1 offset:offs_src1 atIndex:1];
-    //} else {
-    //    [encoder setBuffer:id_src0 offset:offs_src0 atIndex:1];
-    //}
-    //[encoder setBuffer:id_dst  offset:offs_dst  atIndex:2];
-    //[encoder setBytes:&args length:sizeof(args) atIndex:3];
-
-    //[encoder dispatchThreadgroups:MTLSizeMake(nrows, 1, 1) threadsPerThreadgroup:MTLSizeMake(nth, 1, 1)];
-
     ggml_metal_encoder_set_pipeline(enc, pipeline);
     ggml_metal_encoder_set_bytes   (enc, &args, sizeof(args), 0);
     ggml_metal_encoder_set_buffer  (enc, ggml_metal_get_buffer_id(op->src[0]), 1);
@@ -902,7 +899,7 @@ int ggml_metal_op_sum_rows(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne0, op->src[0], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb0, op->src[0], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     ggml_metal_kargs_sum_rows args = {
         /*.ne00 =*/ ne00,
@@ -936,14 +933,6 @@ int ggml_metal_op_sum_rows(ggml_metal_op_t ctx, int idx) {
 
     const size_t smem = ggml_metal_pipeline_get_smem(pipeline);
 
-    //[encoder setComputePipelineState:pipeline];
-    //[encoder setBytes:&args length:sizeof(args) atIndex:0];
-    //[encoder setBuffer:id_src0 offset:offs_src0 atIndex:1];
-    //[encoder setBuffer:id_dst  offset:offs_dst  atIndex:2];
-    //[encoder setThreadgroupMemoryLength:32*sizeof(float) atIndex:0];
-
-    //[encoder dispatchThreadgroups:MTLSizeMake(ne01, ne02, ne03) threadsPerThreadgroup:MTLSizeMake(nth, 1, 1)];
-
     ggml_metal_encoder_set_pipeline(enc, pipeline);
     ggml_metal_encoder_set_bytes   (enc, &args, sizeof(args), 0);
     ggml_metal_encoder_set_buffer  (enc, ggml_metal_get_buffer_id(op->src[0]), 1);
@@ -956,6 +945,149 @@ int ggml_metal_op_sum_rows(ggml_metal_op_t ctx, int idx) {
     return 1;
 }
 
+int ggml_metal_op_cumsum(ggml_metal_op_t ctx, int idx) {
+    ggml_tensor * op = ctx->node(idx);
+
+    ggml_metal_library_t lib = ctx->lib;
+    ggml_metal_encoder_t enc = ctx->enc;
+
+    GGML_ASSERT(ggml_is_contiguous_rows(op->src[0]));
+
+    GGML_TENSOR_LOCALS( int32_t, ne0, op->src[0], ne);
+    GGML_TENSOR_LOCALS(uint64_t, nb0, op->src[0], nb);
+    GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
+
+    ggml_metal_pipeline_t pipeline_blk = ggml_metal_library_get_pipeline_cumsum_blk(lib, op);
+
+    int nth = 1;
+    while (nth < ne00 && 2*nth <= ggml_metal_pipeline_max_theads_per_threadgroup(pipeline_blk)) {
+        nth *= 2;
+    }
+
+    GGML_ASSERT(ne00 <= nth*nth);
+
+    const int64_t net0 = (ne00 + nth - 1) / nth;
+    const int64_t net1 = ne01;
+    const int64_t net2 = ne02;
+    const int64_t net3 = ne03;
+
+    const uint64_t nbt0 = sizeof(float);
+    const uint64_t nbt1 = net0*nbt0;
+    const uint64_t nbt2 = net1*nbt1;
+    const uint64_t nbt3 = net2*nbt2;
+
+    const size_t smem = GGML_PAD(32*sizeof(float), 16);
+
+    ggml_metal_buffer_id bid_src0 = ggml_metal_get_buffer_id(op->src[0]);
+    ggml_metal_buffer_id bid_dst  = ggml_metal_get_buffer_id(op);
+
+    ggml_metal_buffer_id bid_tmp = bid_dst;
+    bid_tmp.offs += ggml_nbytes(op);
+
+    {
+        ggml_metal_kargs_cumsum_blk args = {
+            /*.ne00 =*/ ne00,
+            /*.ne01 =*/ ne01,
+            /*.ne02 =*/ ne02,
+            /*.ne03 =*/ ne03,
+            /*.nb00 =*/ nb00,
+            /*.nb01 =*/ nb01,
+            /*.nb02 =*/ nb02,
+            /*.nb03 =*/ nb03,
+            /*.net0 =*/ net0,
+            /*.net1 =*/ net1,
+            /*.net2 =*/ net2,
+            /*.net3 =*/ net3,
+            /*.nbt0 =*/ nbt0,
+            /*.nbt1 =*/ nbt1,
+            /*.nbt2 =*/ nbt2,
+            /*.nbt3 =*/ nbt3,
+            /*.outb =*/ ne00 > nth,
+        };
+
+        ggml_metal_encoder_set_pipeline(enc, pipeline_blk);
+        ggml_metal_encoder_set_bytes   (enc, &args, sizeof(args), 0);
+        ggml_metal_encoder_set_buffer  (enc, bid_src0, 1);
+        ggml_metal_encoder_set_buffer  (enc, bid_tmp,  2);
+        ggml_metal_encoder_set_buffer  (enc, bid_dst,  3);
+
+        ggml_metal_encoder_set_threadgroup_memory_size(enc, smem, 0);
+
+        ggml_metal_encoder_dispatch_threadgroups(enc, net0*ne01, ne02, ne03, nth, 1, 1);
+    }
+
+    if (ne00 > nth) {
+        ggml_metal_op_concurrency_reset(ctx);
+
+        {
+            ggml_metal_kargs_cumsum_blk args = {
+                /*.ne00 =*/ net0,
+                /*.ne01 =*/ net1,
+                /*.ne02 =*/ net2,
+                /*.ne03 =*/ net3,
+                /*.nb00 =*/ nbt0,
+                /*.nb01 =*/ nbt1,
+                /*.nb02 =*/ nbt2,
+                /*.nb03 =*/ nbt3,
+                /*.net0 =*/ net0,
+                /*.net1 =*/ net1,
+                /*.net2 =*/ net2,
+                /*.net3 =*/ net3,
+                /*.nbt0 =*/ nbt0,
+                /*.nbt1 =*/ nbt1,
+                /*.nbt2 =*/ nbt2,
+                /*.nbt3 =*/ nbt3,
+                /*.outb =*/ false,
+            };
+
+            ggml_metal_encoder_set_pipeline(enc, pipeline_blk);
+            ggml_metal_encoder_set_bytes   (enc, &args, sizeof(args), 0);
+            ggml_metal_encoder_set_buffer  (enc, bid_tmp, 1);
+            ggml_metal_encoder_set_buffer  (enc, bid_tmp, 2);
+            ggml_metal_encoder_set_buffer  (enc, bid_tmp, 3);
+
+            ggml_metal_encoder_set_threadgroup_memory_size(enc, smem, 0);
+
+            ggml_metal_encoder_dispatch_threadgroups(enc, net1, net2, net3, nth, 1, 1);
+        }
+
+        ggml_metal_op_concurrency_reset(ctx);
+
+        {
+            ggml_metal_pipeline_t pipeline_add = ggml_metal_library_get_pipeline_cumsum_add(lib, op);
+
+            ggml_metal_kargs_cumsum_add args = {
+                /*.ne00 =*/ ne00,
+                /*.ne01 =*/ ne01,
+                /*.ne02 =*/ ne02,
+                /*.ne03 =*/ ne03,
+                /*.nb00 =*/ nb00,
+                /*.nb01 =*/ nb01,
+                /*.nb02 =*/ nb02,
+                /*.nb03 =*/ nb03,
+                /*.net0 =*/ net0,
+                /*.net1 =*/ net1,
+                /*.net2 =*/ net2,
+                /*.net3 =*/ net3,
+                /*.nbt0 =*/ nbt0,
+                /*.nbt1 =*/ nbt1,
+                /*.nbt2 =*/ nbt2,
+                /*.nbt3 =*/ nbt3,
+            };
+
+            ggml_metal_encoder_set_pipeline(enc, pipeline_add);
+            ggml_metal_encoder_set_bytes   (enc, &args, sizeof(args), 0);
+            ggml_metal_encoder_set_buffer  (enc, bid_tmp, 1);
+            ggml_metal_encoder_set_buffer  (enc, bid_dst, 2);
+
+            ggml_metal_encoder_dispatch_threadgroups(enc, net0*ne01, ne02, ne03, nth, 1, 1);
+        }
+    }
+
+    return 1;
+}
+
 int ggml_metal_op_get_rows(ggml_metal_op_t ctx, int idx) {
     ggml_tensor * op = ctx->node(idx);
 
@@ -967,7 +1099,7 @@ int ggml_metal_op_get_rows(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne1, op->src[1], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb1, op->src[1], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     ggml_metal_pipeline_t pipeline = ggml_metal_library_get_pipeline_get_rows(lib, op->src[0]->type);
 
@@ -1012,7 +1144,7 @@ int ggml_metal_op_set_rows(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne1, op->src[1], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb1, op->src[1], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     ggml_metal_pipeline_t pipeline = ggml_metal_library_get_pipeline_set_rows(lib, op->src[1]->type, op->type);
 
@@ -1036,11 +1168,6 @@ int ggml_metal_op_set_rows(ggml_metal_op_t ctx, int idx) {
 
     nth = std::min(nth, nk0);
 
-    if (nth*nrptg > ggml_metal_pipeline_max_theads_per_threadgroup(pipeline)) {
-        nth = ggml_metal_pipeline_max_theads_per_threadgroup(pipeline);
-        nrptg = 1;
-    }
-
     ggml_metal_kargs_set_rows args = {
         /*.nk0  =*/ nk0,
         /*.ne01 =*/ ne01,
@@ -1081,7 +1208,7 @@ int ggml_metal_op_soft_max(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne2, op->src[2], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb2, op->src[2], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     float scale;
     float max_bias;
@@ -1169,7 +1296,7 @@ int ggml_metal_op_ssm_conv(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne1, op->src[1], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb1, op->src[1], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     ggml_metal_kargs_ssm_conv args = {
         /*.ne00 =*/ ne00,
@@ -1224,7 +1351,7 @@ int ggml_metal_op_ssm_scan(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne6, op->src[6], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb6, op->src[6], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     const ggml_tensor * src3 = op->src[3];
     const ggml_tensor * src4 = op->src[4];
@@ -1310,7 +1437,7 @@ int ggml_metal_op_rwkv(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne0, op->src[0], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb0, op->src[0], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     const int64_t B = op->op == GGML_OP_RWKV_WKV6 ? op->src[5]->ne[1] : op->src[6]->ne[1];
     const int64_t T = op->src[0]->ne[2];
@@ -1351,7 +1478,7 @@ int ggml_metal_op_cpy(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne0, op->src[0], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb0, op->src[0], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     ggml_metal_pipeline_t pipeline = ggml_metal_library_get_pipeline_cpy(lib, op->src[0]->type, op->type);
 
@@ -1424,7 +1551,7 @@ int ggml_metal_op_pool_2d(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne0, op->src[0], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb0, op->src[0], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     const int32_t * opts = op->op_params;
     ggml_op_pool op_pool = (ggml_op_pool) opts[0];
@@ -1488,7 +1615,7 @@ int ggml_metal_op_mul_mat(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne1, op->src[1], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb1, op->src[1], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     GGML_ASSERT(ne00 == ne10);
 
@@ -1729,7 +1856,7 @@ int ggml_metal_op_mul_mat_id(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne2, op->src[2], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb2, op->src[2], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     // src2 = ids
     GGML_ASSERT(op->src[2]->type == GGML_TYPE_I32);
@@ -1975,7 +2102,9 @@ size_t ggml_metal_op_flash_attn_ext_extra_pad(const ggml_tensor * op) {
     const bool has_mask = op->src[3] != nullptr;
 
     if (ggml_metal_op_flash_attn_ext_use_vec(op)) {
-        const bool has_kvpad = ne11 % OP_FLASH_ATTN_EXT_VEC_NCPSG != 0;
+        // note: always reserve the padding space to avoid graph reallocations
+        //const bool has_kvpad = ne11 % OP_FLASH_ATTN_EXT_VEC_NCPSG != 0;
+        const bool has_kvpad = true;
 
         if (has_kvpad) {
             res += OP_FLASH_ATTN_EXT_VEC_NCPSG*(
@@ -1984,7 +2113,8 @@ size_t ggml_metal_op_flash_attn_ext_extra_pad(const ggml_tensor * op) {
                 (has_mask ? ggml_type_size(GGML_TYPE_F16)*ne31*ne32*ne33 : 0));
         }
     } else {
-        const bool has_kvpad = ne11 % OP_FLASH_ATTN_EXT_NCPSG != 0;
+        //const bool has_kvpad = ne11 % OP_FLASH_ATTN_EXT_NCPSG != 0;
+        const bool has_kvpad = true;
 
         if (has_kvpad) {
             res += OP_FLASH_ATTN_EXT_NCPSG*(
@@ -2020,9 +2150,10 @@ size_t ggml_metal_op_flash_attn_ext_extra_blk(const ggml_tensor * op) {
     const bool is_vec = ggml_metal_op_flash_attn_ext_use_vec(op);
 
     // this optimization is not useful for the vector kernels
-    if (is_vec) {
-        return res;
-    }
+    // note: always reserve the blk buffer to avoid graph reallocations
+    //if (is_vec) {
+    //    return res;
+    //}
 
     const int nqptg = is_vec ? OP_FLASH_ATTN_EXT_VEC_NQPTG : OP_FLASH_ATTN_EXT_NQPTG;
     const int ncpsg = is_vec ? OP_FLASH_ATTN_EXT_VEC_NCPSG : OP_FLASH_ATTN_EXT_NCPSG;
@@ -2049,13 +2180,16 @@ size_t ggml_metal_op_flash_attn_ext_extra_tmp(const ggml_tensor * op) {
 
     size_t res = 0;
 
-    if (ggml_metal_op_flash_attn_ext_use_vec(op)) {
+    // note: always reserve the temp buffer to avoid graph reallocations
+    //if (ggml_metal_op_flash_attn_ext_use_vec(op)) {
+    if (true) {
         const int64_t nwg = 32;
+        const int64_t ne01_max = std::min(ne01, 32);
 
         // temp buffer for writing the results from each workgroup
         // - ne20: the size of the Value head
         // -  + 2: the S and M values for each intermediate result
-        res += ggml_type_size(GGML_TYPE_F32)*(ne01*ne02*ne03*nwg*(ne20 + 2));
+        res += ggml_type_size(GGML_TYPE_F32)*(ne01_max*ne02*ne03*nwg*(ne20 + 2));
     }
 
     return res;
@@ -2184,8 +2318,6 @@ int ggml_metal_op_flash_attn_ext(ggml_metal_op_t ctx, int idx) {
             ggml_metal_encoder_dispatch_threadgroups(enc, ncpsg, std::max(ne12, ne32), std::max(ne13, ne33), 32, 1, 1);
 
             need_sync = true;
-        } else {
-            assert(ggml_metal_op_flash_attn_ext_extra_pad(op) == 0);
         }
 
         if (has_mask) {
@@ -2215,8 +2347,6 @@ int ggml_metal_op_flash_attn_ext(ggml_metal_op_t ctx, int idx) {
             ggml_metal_encoder_dispatch_threadgroups(enc, nblk0, nblk1, ne32*ne33, 32, 1, 1);
 
             need_sync = true;
-        } else {
-            assert(ggml_metal_op_flash_attn_ext_extra_blk(op) == 0);
         }
 
         if (need_sync) {
@@ -2356,8 +2486,6 @@ int ggml_metal_op_flash_attn_ext(ggml_metal_op_t ctx, int idx) {
             ggml_metal_encoder_dispatch_threadgroups(enc, ncpsg, std::max(ne12, ne32), std::max(ne13, ne33), 32, 1, 1);
 
             need_sync = true;
-        } else {
-            assert(ggml_metal_op_flash_attn_ext_extra_pad(op) == 0);
         }
 
         if (need_sync) {
@@ -2688,7 +2816,7 @@ int ggml_metal_op_l2_norm(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne0, op->src[0], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb0, op->src[0], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     float eps;
     memcpy(&eps, op->op_params, sizeof(float));
@@ -2736,7 +2864,7 @@ int ggml_metal_op_group_norm(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne0, op->src[0], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb0, op->src[0], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     const int32_t ngrp = ((const int32_t *) op->op_params)[0];
 
@@ -2791,7 +2919,7 @@ int ggml_metal_op_norm(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne0, op->src[0], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb0, op->src[0], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     float eps;
     memcpy(&eps, op->op_params, sizeof(float));
@@ -2927,7 +3055,7 @@ int ggml_metal_op_rope(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne1, op->src[1], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb1, op->src[1], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     // make sure we have one or more position id(ne10) per token(ne02)
     GGML_ASSERT(ne10 % ne02 == 0);
@@ -3021,7 +3149,7 @@ int ggml_metal_op_im2col(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne0, op->src[0], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb0, op->src[0], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     const int32_t s0 = ((const int32_t *)(op->op_params))[0];
     const int32_t s1 = ((const int32_t *)(op->op_params))[1];
@@ -3082,6 +3210,84 @@ int ggml_metal_op_im2col(ggml_metal_op_t ctx, int idx) {
     return 1;
 }
 
+int ggml_metal_op_conv_2d(ggml_metal_op_t ctx, int idx) {
+    ggml_tensor * op = ctx->node(idx);
+
+    ggml_metal_library_t lib = ctx->lib;
+    ggml_metal_encoder_t enc = ctx->enc;
+
+    GGML_TENSOR_LOCALS( int32_t, ne0, op->src[0], ne);
+    GGML_TENSOR_LOCALS(uint64_t, nb0, op->src[0], nb);
+    GGML_TENSOR_LOCALS( int32_t, ne1, op->src[1], ne);
+    GGML_TENSOR_LOCALS(uint64_t, nb1, op->src[1], nb);
+    GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
+
+    GGML_ASSERT(ggml_is_contiguous(op->src[0]));
+    GGML_ASSERT(op->src[1]->type == GGML_TYPE_F32);
+    GGML_ASSERT(op->type == GGML_TYPE_F32);
+    GGML_ASSERT(op->src[0]->type == GGML_TYPE_F16 || op->src[0]->type == GGML_TYPE_F32);
+
+    const int32_t s0 = ((const int32_t *) op->op_params)[0];
+    const int32_t s1 = ((const int32_t *) op->op_params)[1];
+    const int32_t p0 = ((const int32_t *) op->op_params)[2];
+    const int32_t p1 = ((const int32_t *) op->op_params)[3];
+    const int32_t d0 = ((const int32_t *) op->op_params)[4];
+    const int32_t d1 = ((const int32_t *) op->op_params)[5];
+
+    ggml_metal_kargs_conv_2d args = {
+        /*.nb00 =*/ nb00,
+        /*.nb01 =*/ nb01,
+        /*.nb02 =*/ nb02,
+        /*.nb03 =*/ nb03,
+        /*.nb10 =*/ nb10,
+        /*.nb11 =*/ nb11,
+        /*.nb12 =*/ nb12,
+        /*.nb13 =*/ nb13,
+        /*.nb0  =*/ nb0,
+        /*.nb1  =*/ nb1,
+        /*.nb2  =*/ nb2,
+        /*.nb3  =*/ nb3,
+        /*.IW   =*/ ne10,
+        /*.IH   =*/ ne11,
+        /*.KW   =*/ ne00,
+        /*.KH   =*/ ne01,
+        /*.IC   =*/ ne02,
+        /*.OC   =*/ ne03,
+        /*.OW   =*/ ne0,
+        /*.OH   =*/ ne1,
+        /*.N    =*/ ne3,
+        /*.s0   =*/ s0,
+        /*.s1   =*/ s1,
+        /*.p0   =*/ p0,
+        /*.p1   =*/ p1,
+        /*.d0   =*/ d0,
+        /*.d1   =*/ d1,
+    };
+
+    ggml_metal_pipeline_t pipeline = ggml_metal_library_get_pipeline_conv_2d(lib, op);
+
+    int nth = ggml_metal_pipeline_max_theads_per_threadgroup(pipeline);
+    nth = std::min(nth, 256);
+    nth = std::max(nth, 1);
+
+    const uint64_t n_out = ggml_nelements(op);
+
+    uint64_t tg = (n_out + nth - 1)/nth;
+    tg = std::max<uint64_t>(tg, 1);
+    tg = std::min<uint64_t>(tg, (uint64_t) std::numeric_limits<int>::max());
+
+    ggml_metal_encoder_set_pipeline(enc, pipeline);
+    ggml_metal_encoder_set_bytes   (enc, &args, sizeof(args), 0);
+    ggml_metal_encoder_set_buffer  (enc, ggml_metal_get_buffer_id(op->src[0]), 1);
+    ggml_metal_encoder_set_buffer  (enc, ggml_metal_get_buffer_id(op->src[1]), 2);
+    ggml_metal_encoder_set_buffer  (enc, ggml_metal_get_buffer_id(op),         3);
+
+    ggml_metal_encoder_dispatch_threadgroups(enc, tg, 1, 1, nth, 1, 1);
+
+    return 1;
+}
+
 int ggml_metal_op_conv_transpose_1d(ggml_metal_op_t ctx, int idx) {
     ggml_tensor * op = ctx->node(idx);
 
@@ -3093,7 +3299,7 @@ int ggml_metal_op_conv_transpose_1d(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne1, op->src[1], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb1, op->src[1], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     const int32_t s0 = ((const int32_t *)(op->op_params))[0];
 
@@ -3138,7 +3344,7 @@ int ggml_metal_op_conv_transpose_2d(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne1, op->src[1], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb1, op->src[1], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     const int32_t s0 = ((const int32_t *)(op->op_params))[0];
 
@@ -3192,7 +3398,7 @@ int ggml_metal_op_upscale(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne0, op->src[0], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb0, op->src[0], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     const float sf0 = (float)ne0/op->src[0]->ne[0];
     const float sf1 = (float)ne1/op->src[0]->ne[1];
@@ -3245,7 +3451,7 @@ int ggml_metal_op_pad(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne0, op->src[0], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb0, op->src[0], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     ggml_metal_kargs_pad args = {
         /*.ne00 =*/ ne00,
@@ -3289,7 +3495,7 @@ int ggml_metal_op_pad_reflect_1d(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne0, op->src[0], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb0, op->src[0], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     ggml_metal_kargs_pad_reflect_1d args = {
         /*.ne00 =*/ ne00,
@@ -3333,7 +3539,7 @@ int ggml_metal_op_arange(ggml_metal_op_t ctx, int idx) {
     ggml_metal_encoder_t enc = ctx->enc;
 
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     float start;
     float step;
@@ -3351,12 +3557,6 @@ int ggml_metal_op_arange(ggml_metal_op_t ctx, int idx) {
 
     ggml_metal_pipeline_t pipeline = ggml_metal_library_get_pipeline_arange(lib, op);
 
-    //[encoder setComputePipelineState:pipeline];
-    //[encoder setBuffer:id_dst  offset:offs_dst  atIndex:0];
-    //[encoder setBytes:&args length:sizeof(args) atIndex:1];
-
-    //[encoder dispatchThreadgroups:MTLSizeMake(1, 1, 1) threadsPerThreadgroup:MTLSizeMake(nth, 1, 1)];
-
     ggml_metal_encoder_set_pipeline(enc, pipeline);
     ggml_metal_encoder_set_bytes   (enc, &args, sizeof(args), 0);
     ggml_metal_encoder_set_buffer  (enc, ggml_metal_get_buffer_id(op), 1);
@@ -3375,7 +3575,7 @@ int ggml_metal_op_timestep_embedding(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne0, op->src[0], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb0, op->src[0], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     const int dim        = op->op_params[0];
     const int max_period = op->op_params[1];
@@ -3409,7 +3609,7 @@ int ggml_metal_op_argmax(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne0, op->src[0], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb0, op->src[0], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     ggml_metal_kargs_argmax args = {
         /*.ne00 = */ ne00,
@@ -3445,38 +3645,93 @@ int ggml_metal_op_argsort(ggml_metal_op_t ctx, int idx) {
     ggml_metal_library_t lib = ctx->lib;
     ggml_metal_encoder_t enc = ctx->enc;
 
+    GGML_ASSERT(ggml_is_contiguous_rows(op->src[0]));
+
     GGML_TENSOR_LOCALS( int32_t, ne0, op->src[0], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb0, op->src[0], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
-
-    // bitonic sort requires the number of elements to be power of 2
-    int64_t ne00_padded = 1;
-    while (ne00_padded < ne00) {
-        ne00_padded *= 2;
-    }
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     ggml_metal_pipeline_t pipeline = ggml_metal_library_get_pipeline_argsort(lib, op);
 
-    const int64_t nrows = ggml_nrows(op->src[0]);
+    // bitonic sort requires the number of elements to be power of 2
+    int nth = 1;
+    while (nth < ne00 && 2*nth <= ggml_metal_pipeline_max_theads_per_threadgroup(pipeline)) {
+        nth *= 2;
+    }
+
+    const int npr = (ne00 + nth - 1)/nth;
 
     // Metal kernels require the buffer size to be multiple of 16 bytes
     // https://developer.apple.com/documentation/metal/mtlcomputecommandencoder/1443142-setthreadgroupmemorylength
-    const size_t smem = GGML_PAD(ne00_padded*sizeof(int32_t), 16);
+    const size_t smem = GGML_PAD(nth*sizeof(int32_t), 16);
+
+    ggml_metal_buffer_id bid_src0 = ggml_metal_get_buffer_id(op->src[0]);
+    ggml_metal_buffer_id bid_dst  = ggml_metal_get_buffer_id(op);
+
+    ggml_metal_buffer_id bid_tmp = bid_dst;
+    bid_tmp.offs += ggml_nbytes(op);
+
+    if ((int) ceil(std::log(npr) / std::log(2)) % 2 == 1) {
+        std::swap(bid_dst, bid_tmp);
+    }
 
     ggml_metal_kargs_argsort args = {
-        /*.ncols =*/ ne00,
-        /*.ncols_pad =*/ ne00_padded
+        /*.ne00 =*/ ne00,
+        /*.ne01 =*/ ne01,
+        /*.ne02 =*/ ne02,
+        /*.ne03 =*/ ne03,
+        /*.nb00 =*/ nb00,
+        /*.nb01 =*/ nb01,
+        /*.nb02 =*/ nb02,
+        /*.nb03 =*/ nb03,
     };
 
     ggml_metal_encoder_set_pipeline(enc, pipeline);
     ggml_metal_encoder_set_bytes   (enc, &args, sizeof(args), 0);
-    ggml_metal_encoder_set_buffer  (enc, ggml_metal_get_buffer_id(op->src[0]), 1);
-    ggml_metal_encoder_set_buffer  (enc, ggml_metal_get_buffer_id(op),         2);
+    ggml_metal_encoder_set_buffer  (enc, bid_src0, 1);
+    ggml_metal_encoder_set_buffer  (enc, bid_dst,  2);
 
     ggml_metal_encoder_set_threadgroup_memory_size(enc, smem, 0);
 
-    ggml_metal_encoder_dispatch_threadgroups(enc, 1, nrows, 1, ne00_padded, 1, 1);
+    ggml_metal_encoder_dispatch_threadgroups(enc, npr*ne01, ne02, ne03, nth, 1, 1);
+
+    ggml_metal_pipeline_t pipeline_merge = ggml_metal_library_get_pipeline_argsort_merge(lib, op);
+
+    int len = nth;
+
+    while (len < ne00) {
+        ggml_metal_op_concurrency_reset(ctx);
+
+        ggml_metal_kargs_argsort_merge args_merge = {
+            .ne00 = ne00,
+            .ne01 = ne01,
+            .ne02 = ne02,
+            .ne03 = ne03,
+            .nb00 = nb00,
+            .nb01 = nb01,
+            .nb02 = nb02,
+            .nb03 = nb03,
+            .len  = len,
+        };
+
+        // merges per row
+        const int nm = (ne00 + 2*len - 1) / (2*len);
+
+        const int nth = std::min(512, ggml_metal_pipeline_max_theads_per_threadgroup(pipeline_merge));
+
+        ggml_metal_encoder_set_pipeline(enc, pipeline_merge);
+        ggml_metal_encoder_set_bytes   (enc, &args_merge, sizeof(args_merge), 0);
+        ggml_metal_encoder_set_buffer  (enc, bid_src0, 1);
+        ggml_metal_encoder_set_buffer  (enc, bid_dst,  2);
+        ggml_metal_encoder_set_buffer  (enc, bid_tmp,  3);
+
+        ggml_metal_encoder_dispatch_threadgroups(enc, nm*ne01, ne02, ne03, nth, 1, 1);
+
+        std::swap(bid_dst, bid_tmp);
+
+        len <<= 1;
+    }
 
     return 1;
 }
@@ -3490,7 +3745,7 @@ int ggml_metal_op_leaky_relu(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne0, op->src[0], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb0, op->src[0], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     float slope;
     memcpy(&slope, op->op_params, sizeof(float));
@@ -3526,7 +3781,7 @@ int ggml_metal_op_opt_step_adamw(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne0, op->src[0], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb0, op->src[0], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     ggml_metal_pipeline_t pipeline = ggml_metal_library_get_pipeline_opt_step_adamw(lib, op);
 
@@ -3562,7 +3817,7 @@ int ggml_metal_op_opt_step_sgd(ggml_metal_op_t ctx, int idx) {
     GGML_TENSOR_LOCALS( int32_t, ne0, op->src[0], ne);
     GGML_TENSOR_LOCALS(uint64_t, nb0, op->src[0], nb);
     GGML_TENSOR_LOCALS( int32_t, ne,  op,         ne);
-    GGML_TENSOR_LOCALS(uint32_t, nb,  op,         nb);
+    GGML_TENSOR_LOCALS(uint64_t, nb,  op,         nb);
 
     ggml_metal_pipeline_t pipeline = ggml_metal_library_get_pipeline_opt_step_sgd(lib, op);
 

ggml/src/ggml-metal/ggml-metal-ops.h

@@ -52,6 +52,7 @@ int ggml_metal_op_unary             (ggml_metal_op_t ctx, int idx);
 int ggml_metal_op_glu               (ggml_metal_op_t ctx, int idx);
 int ggml_metal_op_sum               (ggml_metal_op_t ctx, int idx);
 int ggml_metal_op_sum_rows          (ggml_metal_op_t ctx, int idx);
+int ggml_metal_op_cumsum            (ggml_metal_op_t ctx, int idx);
 int ggml_metal_op_get_rows          (ggml_metal_op_t ctx, int idx);
 int ggml_metal_op_set_rows          (ggml_metal_op_t ctx, int idx);
 int ggml_metal_op_soft_max          (ggml_metal_op_t ctx, int idx);
@@ -70,6 +71,7 @@ int ggml_metal_op_group_norm        (ggml_metal_op_t ctx, int idx);
 int ggml_metal_op_norm              (ggml_metal_op_t ctx, int idx);
 int ggml_metal_op_rope              (ggml_metal_op_t ctx, int idx);
 int ggml_metal_op_im2col            (ggml_metal_op_t ctx, int idx);
+int ggml_metal_op_conv_2d           (ggml_metal_op_t ctx, int idx);
 int ggml_metal_op_conv_transpose_1d (ggml_metal_op_t ctx, int idx);
 int ggml_metal_op_conv_transpose_2d (ggml_metal_op_t ctx, int idx);
 int ggml_metal_op_upscale           (ggml_metal_op_t ctx, int idx);

ggml/src/ggml-metal/ggml-metal.cpp

@@ -197,6 +197,11 @@ static size_t ggml_backend_metal_buffer_type_get_alloc_size(ggml_backend_buffer_
                 res += ggml_metal_op_flash_attn_ext_extra_blk(tensor);
                 res += ggml_metal_op_flash_attn_ext_extra_tmp(tensor);
             } break;
+        case GGML_OP_CUMSUM:
+        case GGML_OP_ARGSORT:
+            {
+                res *= 2;
+            } break;
         default:
             break;
     }

ggml/src/ggml-metal/ggml-metal.metal

@@ -1832,6 +1832,117 @@ typedef decltype(kernel_sum_rows<false>) kernel_sum_rows_t;
 template [[host_name("kernel_sum_rows_f32")]] kernel kernel_sum_rows_t kernel_sum_rows<false>;
 template [[host_name("kernel_mean_f32")]]     kernel kernel_sum_rows_t kernel_sum_rows<true>;
 
+template<typename T>
+kernel void kernel_cumsum_blk(
+        constant ggml_metal_kargs_cumsum_blk & args,
+        device const char * src0,
+        device       char * tmp,
+        device       char * dst,
+        threadgroup  char * shmem [[threadgroup(0)]],
+        uint3   tgpig[[threadgroup_position_in_grid]],
+        ushort3 tpitg[[thread_position_in_threadgroup]],
+        ushort  sgitg[[simdgroup_index_in_threadgroup]],
+        ushort  tiisg[[thread_index_in_simdgroup]],
+        ushort3   ntg[[threads_per_threadgroup]]) {
+    const int ib = tgpig[0]/args.ne01;
+
+    const int i00 = ib*ntg.x;
+    const int i01 = tgpig[0]%args.ne01;
+    const int i02 = tgpig[1];
+    const int i03 = tgpig[2];
+
+    device const float * src0_row = (device const float *) (src0 +
+            args.nb01*i01 +
+            args.nb02*i02 +
+            args.nb03*i03);
+
+    threadgroup float * shmem_f32 = (threadgroup float *) shmem;
+
+    float v = 0.0f;
+
+    if (i00 + tpitg.x < args.ne00) {
+        v = src0_row[i00 + tpitg.x];
+    }
+
+    float s = simd_prefix_inclusive_sum(v);
+
+    if (tiisg == N_SIMDWIDTH - 1) {
+        shmem_f32[sgitg] = s;
+    }
+
+    threadgroup_barrier(mem_flags::mem_threadgroup);
+
+    if (sgitg == 0) {
+        shmem_f32[tiisg] = simd_prefix_exclusive_sum(shmem_f32[tiisg]);
+    }
+
+    threadgroup_barrier(mem_flags::mem_threadgroup);
+
+    s += shmem_f32[sgitg];
+
+    device float * dst_row = (device float *) dst +
+        args.ne00*i01 +
+        args.ne00*args.ne01*i02 +
+        args.ne00*args.ne01*args.ne02*i03;
+
+    if (i00 + tpitg.x < args.ne00) {
+        dst_row[i00 + tpitg.x] = s;
+    }
+
+    if (args.outb && tpitg.x == ntg.x - 1) {
+        device float * tmp_row = (device float *) tmp +
+            args.net0*i01 +
+            args.net0*args.net1*i02 +
+            args.net0*args.net1*args.net2*i03;
+
+        tmp_row[ib] = s;
+    }
+}
+
+typedef decltype(kernel_cumsum_blk<float>) kernel_cumsum_blk_t;
+
+template [[host_name("kernel_cumsum_blk_f32")]] kernel kernel_cumsum_blk_t kernel_cumsum_blk<float>;
+
+template<typename T>
+kernel void kernel_cumsum_add(
+        constant ggml_metal_kargs_cumsum_add & args,
+        device const char * tmp,
+        device       char * dst,
+        uint3   tgpig[[threadgroup_position_in_grid]],
+        ushort3 tpitg[[thread_position_in_threadgroup]],
+        ushort  sgitg[[simdgroup_index_in_threadgroup]],
+        ushort  tiisg[[thread_index_in_simdgroup]],
+        ushort3   ntg[[threads_per_threadgroup]]) {
+    const int ib = tgpig[0]/args.ne01;
+
+    if (ib == 0) {
+        return;
+    }
+
+    const int i00 = ib*ntg.x;
+    const int i01 = tgpig[0]%args.ne01;
+    const int i02 = tgpig[1];
+    const int i03 = tgpig[2];
+
+    device const float * tmp_row = (device const float *) (tmp +
+            args.nbt1*i01 +
+            args.nbt2*i02 +
+            args.nbt3*i03);
+
+    device float * dst_row = (device float *) dst +
+        args.ne00*i01 +
+        args.ne00*args.ne01*i02 +
+        args.ne00*args.ne01*args.ne02*i03;
+
+    if (i00 + tpitg.x < args.ne00) {
+        dst_row[i00 + tpitg.x] += tmp_row[ib - 1];
+    }
+}
+
+typedef decltype(kernel_cumsum_add<float>) kernel_cumsum_add_t;
+
+template [[host_name("kernel_cumsum_add_f32")]] kernel kernel_cumsum_add_t kernel_cumsum_add<float>;
+
 template<typename T>
 kernel void kernel_soft_max(
         constant ggml_metal_kargs_soft_max & args,
@@ -4146,6 +4257,120 @@ template [[host_name("kernel_im2col_f16")]] kernel im2col_t kernel_im2col<half>;
 //template [[host_name("kernel_im2col_ext_f32")]] kernel im2col_ext_t kernel_im2col_ext<float>;
 //template [[host_name("kernel_im2col_ext_f16")]] kernel im2col_ext_t kernel_im2col_ext<half>;
 
+template <typename TK>
+kernel void kernel_conv_2d(
+        constant ggml_metal_kargs_conv_2d & args,
+        device const char * weights,
+        device const char * src,
+        device       char * dst,
+        uint3   tgpig[[threadgroup_position_in_grid]],
+        uint3    tgpg[[threadgroups_per_grid]],
+        uint3   tpitg[[thread_position_in_threadgroup]],
+        uint3     ntg[[threads_per_threadgroup]]) {
+
+    const uint threads_per_tg = ntg.x * ntg.y * ntg.z;
+    const uint tg_index = (tgpig.z * tgpg.y + tgpig.y) * tgpg.x + tgpig.x;
+    const uint local_thread = tpitg.z * (ntg.x * ntg.y) + tpitg.y * ntg.x + tpitg.x;
+    const uint thread_index = tg_index * threads_per_tg + local_thread;
+    const uint64_t total_threads = (uint64_t) threads_per_tg * tgpg.x * tgpg.y * tgpg.z;
+    const uint64_t total_outputs = (uint64_t) args.N * args.OC * args.OH * args.OW;
+
+    for (uint64_t index = thread_index; index < total_outputs; index += total_threads) {
+        uint64_t tmp = index;
+
+        const int32_t ow = tmp % args.OW; tmp /= args.OW;
+        const int32_t oh = tmp % args.OH; tmp /= args.OH;
+        const int32_t oc = tmp % args.OC; tmp /= args.OC;
+        const int32_t  n = tmp;
+
+        float acc = 0.0f;
+
+        const int32_t base_x = ow*args.s0 - args.p0;
+        const int32_t base_y = oh*args.s1 - args.p1;
+
+        int32_t ky_start = 0;
+        if (base_y < 0) {
+            ky_start = (-base_y + args.d1 - 1)/args.d1;
+        }
+        int32_t ky_end = args.KH;
+        const int32_t y_max = args.IH - 1 - base_y;
+        if (y_max < 0) {
+            ky_end = ky_start;
+        } else if (base_y + (args.KH - 1)*args.d1 >= args.IH) {
+            ky_end = min(ky_end, y_max/args.d1 + 1);
+        }
+
+        int32_t kx_start = 0;
+        if (base_x < 0) {
+            kx_start = (-base_x + args.d0 - 1)/args.d0;
+        }
+        int32_t kx_end = args.KW;
+        const int32_t x_max = args.IW - 1 - base_x;
+        if (x_max < 0) {
+            kx_end = kx_start;
+        } else if (base_x + (args.KW - 1)*args.d0 >= args.IW) {
+            kx_end = min(kx_end, x_max/args.d0 + 1);
+        }
+
+        if (ky_start < ky_end && kx_start < kx_end) {
+            const uint64_t src_base_n = (uint64_t) n  * args.nb13;
+            const uint64_t w_base_oc  = (uint64_t) oc * args.nb03;
+
+            for (int32_t ic = 0; ic < args.IC; ++ic) {
+                const uint64_t src_base_nc = src_base_n + (uint64_t) ic * args.nb12;
+                const uint64_t w_base_ocic = w_base_oc  + (uint64_t) ic * args.nb02;
+
+                for (int32_t ky = ky_start; ky < ky_end; ++ky) {
+                    const int32_t iy = base_y + ky*args.d1;
+                    const uint64_t src_base_row = src_base_nc + (uint64_t) iy * args.nb11;
+                    const uint64_t w_base_row   = w_base_ocic + (uint64_t) ky * args.nb01;
+
+                    for (int32_t kx = kx_start; kx < kx_end; ++kx) {
+                        const int32_t ix = base_x + kx*args.d0;
+                        const uint64_t src_offs = src_base_row + (uint64_t) ix * args.nb10;
+                        const uint64_t w_offs   = w_base_row   + (uint64_t) kx * args.nb00;
+
+                        const float x = *(device const float *)(src + src_offs);
+                        const float w = (float) (*(device const TK *)(weights + w_offs));
+
+                        acc += x * w;
+                    }
+                }
+            }
+        }
+
+        const uint64_t dst_offs =
+            (uint64_t) n  * args.nb3 +
+            (uint64_t) oc * args.nb2 +
+            (uint64_t) oh * args.nb1 +
+            (uint64_t) ow * args.nb0;
+
+        *(device float *)(dst + dst_offs) = acc;
+    }
+}
+
+template [[host_name("kernel_conv_2d_f32_f32")]]
+kernel void kernel_conv_2d<float>(
+        constant ggml_metal_kargs_conv_2d & args,
+        device const char * weights,
+        device const char * src,
+        device       char * dst,
+        uint3   tgpig[[threadgroup_position_in_grid]],
+        uint3    tgpg[[threadgroups_per_grid]],
+        uint3   tpitg[[thread_position_in_threadgroup]],
+        uint3     ntg[[threads_per_threadgroup]]);
+
+template [[host_name("kernel_conv_2d_f16_f32")]]
+kernel void kernel_conv_2d<half>(
+        constant ggml_metal_kargs_conv_2d & args,
+        device const char * weights,
+        device const char * src,
+        device       char * dst,
+        uint3   tgpig[[threadgroup_position_in_grid]],
+        uint3    tgpg[[threadgroups_per_grid]],
+        uint3   tpitg[[thread_position_in_threadgroup]],
+        uint3     ntg[[threads_per_threadgroup]]);
+
 typedef void (conv_transpose_1d_t)(
         constant ggml_metal_kargs_conv_transpose_1d & args,
         device const float * src0,
@@ -4427,69 +4652,227 @@ kernel void kernel_timestep_embedding_f32(
 // bitonic sort implementation following the CUDA kernels as reference
 typedef void (argsort_t)(
         constant   ggml_metal_kargs_argsort & args,
-        device  const float * x,
+        device   const char * src0,
         device      int32_t * dst,
-        threadgroup int32_t * shared_values [[threadgroup(0)]],
-        uint3 tgpig[[threadgroup_position_in_grid]],
-        uint3 tpitg[[thread_position_in_threadgroup]]);
+        threadgroup int32_t * shmem_i32 [[threadgroup(0)]],
+        uint3   tgpig[[threadgroup_position_in_grid]],
+        ushort3 tpitg[[thread_position_in_threadgroup]],
+        ushort3   ntg[[threads_per_threadgroup]]);
 
 template<ggml_sort_order order>
 kernel void kernel_argsort_f32_i32(
         constant   ggml_metal_kargs_argsort & args,
-        device const float  * x,
+        device   const char * src0,
         device      int32_t * dst,
-        threadgroup int32_t * shared_values [[threadgroup(0)]],
-        uint3 tgpig[[threadgroup_position_in_grid]],
-        uint3 tpitg[[thread_position_in_threadgroup]]) {
+        threadgroup int32_t * shmem_i32 [[threadgroup(0)]],
+        uint3   tgpig[[threadgroup_position_in_grid]],
+        ushort3 tpitg[[thread_position_in_threadgroup]],
+        ushort3   ntg[[threads_per_threadgroup]]) {
     // bitonic sort
-    int col = tpitg[0];
-    int row = tgpig[1];
+    const int col = tpitg[0];
 
-    if (col >= args.ncols_pad) return;
+    const int i00 = (tgpig[0]/args.ne01)*ntg.x;
+    const int i01 =  tgpig[0]%args.ne01;
+    const int i02 =  tgpig[1];
+    const int i03 =  tgpig[2];
 
-    device const float   * x_row   = x + row * args.ncols;
-    threadgroup int32_t  * dst_row = shared_values;
+    device const float * src0_row = (device const float *) (src0 + args.nb01*i01 + args.nb02*i02 + args.nb03*i03);
 
     // initialize indices
-    dst_row[col] = col;
+    shmem_i32[col] = i00 + col;
 
     threadgroup_barrier(mem_flags::mem_threadgroup);
 
-    for (int k = 2; k <= args.ncols_pad; k *= 2) {
+    for (int k = 2; k <= ntg.x; k *= 2) {
         for (int j = k / 2; j > 0; j /= 2) {
             int ixj = col ^ j;
             if (ixj > col) {
                 if ((col & k) == 0) {
-                    if (dst_row[col] >= args.ncols ||
-                        (dst_row[ixj] < args.ncols && (order == GGML_SORT_ORDER_ASC ?
-                            x_row[dst_row[col]] > x_row[dst_row[ixj]] :
-                            x_row[dst_row[col]] < x_row[dst_row[ixj]]))
+                    if (shmem_i32[col] >= args.ne00 ||
+                       (shmem_i32[ixj] <  args.ne00 && (order == GGML_SORT_ORDER_ASC ?
+                            src0_row[shmem_i32[col]] > src0_row[shmem_i32[ixj]] :
+                            src0_row[shmem_i32[col]] < src0_row[shmem_i32[ixj]]))
                     ) {
-                        SWAP(dst_row[col], dst_row[ixj]);
+                        SWAP(shmem_i32[col], shmem_i32[ixj]);
                     }
                 } else {
-                    if (dst_row[ixj] >= args.ncols ||
-                        (dst_row[col] < args.ncols && (order == GGML_SORT_ORDER_ASC ?
-                            x_row[dst_row[col]] < x_row[dst_row[ixj]] :
-                            x_row[dst_row[col]] > x_row[dst_row[ixj]]))
+                    if (shmem_i32[ixj] >= args.ne00 ||
+                       (shmem_i32[col] <  args.ne00 && (order == GGML_SORT_ORDER_ASC ?
+                            src0_row[shmem_i32[col]] < src0_row[shmem_i32[ixj]] :
+                            src0_row[shmem_i32[col]] > src0_row[shmem_i32[ixj]]))
                     ) {
-                        SWAP(dst_row[col], dst_row[ixj]);
+                        SWAP(shmem_i32[col], shmem_i32[ixj]);
                     }
                 }
             }
+
             threadgroup_barrier(mem_flags::mem_threadgroup);
         }
     }
 
     // copy the result to dst without the padding
-    if (col < args.ncols) {
-        dst[row * args.ncols + col] = dst_row[col];
+    if (i00 + col < args.ne00) {
+        dst += i00 + args.ne00*i01 + args.ne00*args.ne01*i02 + args.ne00*args.ne01*args.ne02*i03;
+
+        dst[col] = shmem_i32[col];
     }
 }
 
 template [[host_name("kernel_argsort_f32_i32_asc")]]  kernel argsort_t kernel_argsort_f32_i32<GGML_SORT_ORDER_ASC>;
 template [[host_name("kernel_argsort_f32_i32_desc")]] kernel argsort_t kernel_argsort_f32_i32<GGML_SORT_ORDER_DESC>;
 
+typedef void (argsort_merge_t)(
+        constant   ggml_metal_kargs_argsort_merge & args,
+        device const char    * src0,
+        device const int32_t * tmp,
+        device       int32_t * dst,
+        uint3   tgpig[[threadgroup_position_in_grid]],
+        ushort3 tpitg[[thread_position_in_threadgroup]],
+        ushort3   ntg[[threads_per_threadgroup]]);
+
+template<ggml_sort_order order>
+kernel void kernel_argsort_merge_f32_i32(
+        constant   ggml_metal_kargs_argsort_merge & args,
+        device const char    * src0,
+        device const int32_t * tmp,
+        device       int32_t * dst,
+        uint3   tgpig[[threadgroup_position_in_grid]],
+        ushort3 tpitg[[thread_position_in_threadgroup]],
+        ushort3   ntg[[threads_per_threadgroup]]) {
+
+    const int im  = tgpig[0] / args.ne01;
+    const int i01 = tgpig[0] % args.ne01;
+    const int i02 = tgpig[1];
+    const int i03 = tgpig[2];
+
+    const int start = im * (2 * args.len);
+
+    const int len0 = MIN(args.len, MAX(0, args.ne00 - (int)(start)));
+    const int len1 = MIN(args.len, MAX(0, args.ne00 - (int)(start + args.len)));
+
+    const int total = len0 + len1;
+
+    device const int32_t * tmp0 = tmp + start
+        + i01*args.ne00
+        + i02*args.ne00*args.ne01
+        + i03*args.ne00*args.ne01*args.ne02;
+
+    device const int32_t * tmp1 = tmp0 + args.len;
+
+    dst += start
+        + i01*args.ne00
+        + i02*args.ne00*args.ne01
+        + i03*args.ne00*args.ne01*args.ne02;
+
+    device const float * src0_row = (device const float *)(src0
+        + args.nb01*i01
+        + args.nb02*i02
+        + args.nb03*i03);
+
+    if (total == 0) {
+        return;
+    }
+
+    const int chunk = (total + ntg.x - 1) / ntg.x;
+
+    const int k0 = tpitg.x * chunk;
+    const int k1 = min(k0 + chunk, total);
+
+    if (k0 >= total) {
+        return;
+    }
+
+    int low  = k0 > len1 ? k0 - len1 : 0;
+    int high = MIN(k0, len0);
+
+    // binary-search partition (i, j) such that i + j = k
+    while (low < high) {
+        const int mid = (low + high) >> 1;
+
+        const int32_t idx0 = tmp0[mid];
+        const int32_t idx1 = tmp1[k0 - mid - 1];
+
+        const float val0 = src0_row[idx0];
+        const float val1 = src0_row[idx1];
+
+        bool take_left;
+        if (order == GGML_SORT_ORDER_ASC) {
+            take_left = (val0 <= val1);
+        } else {
+            take_left = (val0 >= val1);
+        }
+
+        if (take_left) {
+            low = mid + 1;
+        } else {
+            high = mid;
+        }
+    }
+
+    int i = low;
+    int j = k0 - i;
+
+    // keep the merge fronts into registers
+    int32_t idx0 = 0;
+    float   val0 = 0.0f;
+    if (i < len0) {
+        idx0 = tmp0[i];
+        val0 = src0_row[idx0];
+    }
+
+    int32_t idx1 = 0;
+    float   val1 = 0.0f;
+    if (j < len1) {
+        idx1 = tmp1[j];
+        val1 = src0_row[idx1];
+    }
+
+    for (int k = k0; k < k1; ++k) {
+        int32_t out_idx;
+
+        if (i >= len0) {
+            while (k < k1) {
+                dst[k++] = tmp1[j++];
+            }
+            break;
+        } else if (j >= len1) {
+            while (k < k1) {
+                dst[k++] = tmp0[i++];
+            }
+            break;
+        } else {
+            bool take_left;
+
+            if (order == GGML_SORT_ORDER_ASC) {
+                take_left = (val0 <= val1);
+            } else {
+                take_left = (val0 >= val1);
+            }
+
+            if (take_left) {
+                out_idx = idx0;
+                ++i;
+                if (i < len0) {
+                    idx0 = tmp0[i];
+                    val0 = src0_row[idx0];
+                }
+            } else {
+                out_idx = idx1;
+                ++j;
+                if (j < len1) {
+                    idx1 = tmp1[j];
+                    val1 = src0_row[idx1];
+                }
+            }
+        }
+
+        dst[k] = out_idx;
+    }
+}
+
+template [[host_name("kernel_argsort_merge_f32_i32_asc")]]  kernel argsort_merge_t kernel_argsort_merge_f32_i32<GGML_SORT_ORDER_ASC>;
+template [[host_name("kernel_argsort_merge_f32_i32_desc")]] kernel argsort_merge_t kernel_argsort_merge_f32_i32<GGML_SORT_ORDER_DESC>;
+
 kernel void kernel_leaky_relu_f32(
         constant     ggml_metal_kargs_leaky_relu & args,
         device const float * src0,
@@ -6177,6 +6560,7 @@ template [[host_name("kernel_cpy_f32_f32")]]   kernel kernel_cpy_t kernel_cpy_t_
 template [[host_name("kernel_cpy_f32_f16")]]   kernel kernel_cpy_t kernel_cpy_t_t<float,   half>;
 template [[host_name("kernel_cpy_f32_i32")]]   kernel kernel_cpy_t kernel_cpy_t_t<float,   int32_t>;
 template [[host_name("kernel_cpy_i32_f32")]]   kernel kernel_cpy_t kernel_cpy_t_t<int32_t, float>;
+template [[host_name("kernel_cpy_i32_i32")]]   kernel kernel_cpy_t kernel_cpy_t_t<int32_t, int32_t>;
 #if defined(GGML_METAL_HAS_BF16)
 template [[host_name("kernel_cpy_f32_bf16")]]  kernel kernel_cpy_t kernel_cpy_t_t<float,   bfloat>;
 #endif

ggml/src/ggml-opencl/CMakeLists.txt

@@ -119,6 +119,7 @@ set(GGML_OPENCL_KERNELS
     pad
     repeat
     mul_mat_f16_f32
+    mul_mm_f16_f32_kq_kqv
     conv2d
     conv2d_f16_f32
     flash_attn_f32_f16

ggml/src/ggml-opencl/ggml-opencl.cpp

@@ -407,6 +407,8 @@ struct ggml_backend_opencl_context {
     cl_program program_mul_mv_f32_f32;
     cl_program program_mul;
     cl_program program_mul_mat_f16_f32_tiled;
+    cl_program program_mul_mm_f16_f32_kqv;
+    cl_program program_mul_mm_f16_f32_kq;
     cl_program program_div;
     cl_program program_sub;
     cl_program program_norm;
@@ -481,6 +483,8 @@ struct ggml_backend_opencl_context {
     cl_kernel kernel_mul_mat_f16_f32;
     cl_kernel kernel_mul_mat_f16_f32_l4;
     cl_kernel kernel_mul_mat_f16_f32_tiled;
+    cl_kernel kernel_mul_mm_f16_f32_kqv;
+    cl_kernel kernel_mul_mm_f16_f32_kq;
     cl_kernel kernel_mul_mat_q4_0_f32, kernel_mul_mat_q4_0_f32_v;
     cl_kernel kernel_convert_block_q4_0, kernel_restore_block_q4_0;
     cl_kernel kernel_convert_block_mxfp4, kernel_convert_block_mxfp4_trans, kernel_restore_block_mxfp4, kernel_restore_block_mxfp4_trans;
@@ -1235,6 +1239,25 @@ static void load_cl_kernels(ggml_backend_opencl_context *backend_ctx, ggml_cl_ve
         GGML_LOG_CONT(".");
     }
 
+    // mul_mm_f16_f32_kq_kqv
+    {
+#ifdef GGML_OPENCL_EMBED_KERNELS
+        const std::string kernel_src {
+            #include "mul_mm_f16_f32_kq_kqv.cl.h"
+        };
+#else
+        const std::string kernel_src = read_file("mul_mm_f16_f32_kq_kqv.cl");
+#endif
+        backend_ctx->program_mul_mm_f16_f32_kqv =
+            build_program_from_source(backend_ctx->context, backend_ctx->device, kernel_src.c_str(), compile_opts+" -DKQV ");
+        backend_ctx->program_mul_mm_f16_f32_kq =
+            build_program_from_source(backend_ctx->context, backend_ctx->device, kernel_src.c_str(), compile_opts);
+
+        CL_CHECK((backend_ctx->kernel_mul_mm_f16_f32_kqv = clCreateKernel(backend_ctx->program_mul_mm_f16_f32_kqv, "mul_mm_f16_f32_kqv", &err), err));
+        CL_CHECK((backend_ctx->kernel_mul_mm_f16_f32_kq = clCreateKernel(backend_ctx->program_mul_mm_f16_f32_kq, "mul_mm_f16_f32_kq", &err), err));
+        GGML_LOG_CONT(".");
+    }
+
     // mul
     {
 #ifdef GGML_OPENCL_EMBED_KERNELS
@@ -5682,7 +5705,7 @@ static void ggml_opencl_op_rms_norm_fused(ggml_backend_t backend, ggml_tensor *
     CL_CHECK(clSetKernelArg(kernel, 21, sizeof(cl_ulong),      &nb2));
     CL_CHECK(clSetKernelArg(kernel, 22, sizeof(cl_ulong),      &nb3));
     CL_CHECK(clSetKernelArg(kernel, 23, sizeof(float),         &eps));
-    CL_CHECK(clSetKernelArg(kernel, 24, sizeof(float)*nth/sgs, NULL));
+    CL_CHECK(clSetKernelArg(kernel, 24, sizeof(float)*sgs,     NULL));
 
     backend_ctx->enqueue_ndrange_kernel(kernel, 3, global_work_size, local_work_size, dst);
 }
@@ -6665,6 +6688,146 @@ static void ggml_cl_conv_2d(ggml_backend_t backend, const ggml_tensor * src0, co
     backend_ctx->enqueue_ndrange_kernel(kernel, 2, global_work_size, local_work_size, dst);
 }
 
+static void ggml_cl_mul_mat_kq_kqv_adreno(ggml_backend_t backend, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
+    ggml_backend_opencl_context *backend_ctx = (ggml_backend_opencl_context *)backend->context;
+
+    ggml_tensor_extra_cl * extra0 = (ggml_tensor_extra_cl *)src0->extra;
+    ggml_tensor_extra_cl * extra1 = (ggml_tensor_extra_cl *)src1->extra;
+    ggml_tensor_extra_cl * extrad = (ggml_tensor_extra_cl *)dst->extra;
+
+    const int  ne00 = src0->ne[0];
+    const int  ne01 = src0->ne[1];
+    const int  ne02 = src0->ne[2];
+
+    const cl_ulong nb01 = src0->nb[1];
+    const cl_ulong nb02 = src0->nb[2];
+
+    const int  ne10 = src1->ne[0];
+    const int  ne11 = src1->ne[1];
+    const int  ne12 = src1->ne[2];
+
+    const cl_ulong nb10 = src1->nb[0];
+
+    const int  ne0 = dst->ne[0];
+    const int  ne1 = dst->ne[1];
+
+    GGML_ASSERT(ne00 == ne10);
+
+    cl_kernel kernel;
+    cl_context context = backend_ctx->context;
+
+    cl_int              status;
+    cl_image_format     img_fmt_1d;
+    cl_image_desc       img_desc_1d;
+    cl_buffer_region    region;
+    cl_mem              A_image1d;
+    cl_mem              A_sub_buffer;
+    cl_mem              B_sub_buffer;
+    cl_mem              D_image1d;
+    cl_mem              D_sub_buffer;
+
+    int M = ne01;
+    int N = ne1;
+    int K = ne00;
+
+    if (nb01 > nb02) {
+        // KQ
+        kernel = backend_ctx->kernel_mul_mm_f16_f32_kq;
+    } else {
+        // KQV
+        kernel = backend_ctx->kernel_mul_mm_f16_f32_kqv;
+    }
+    // create sub-buffer for A
+    // <--------------------------------------------> //
+    extra0 = src0->view_src ? (ggml_tensor_extra_cl *)src0->view_src->extra : (ggml_tensor_extra_cl *)src0->extra;
+
+    region.origin = (extra0->offset);
+    if (nb01 > nb02) {
+        // KQ
+        region.size = nb01 * ne01;
+    } else {
+        // KQV
+        region.size = nb02 * ne02;
+    }
+
+    A_sub_buffer = clCreateSubBuffer((extra0->data_device), 0, CL_BUFFER_CREATE_TYPE_REGION, &region, &status);
+    CL_CHECK(status);
+
+    // <--------------------------------------------> //
+
+    // create sub-buffer for B
+    // <--------------------------------------------> //
+    region.origin = (extra1->offset);
+    region.size = nb10 * ne10 * ne11 * ne12;
+    B_sub_buffer = clCreateSubBuffer((extra1->data_device), 0, CL_BUFFER_CREATE_TYPE_REGION, &region, &status);
+    CL_CHECK(status);
+    // <--------------------------------------------> //
+
+    img_fmt_1d = {CL_RGBA, CL_FLOAT};
+    memset(&img_desc_1d, 0, sizeof(img_desc_1d));
+    img_desc_1d.image_type = CL_MEM_OBJECT_IMAGE1D_BUFFER;
+    if (nb01 > nb02) {
+        img_desc_1d.image_width = (nb01 * ne01 / 4)/4;
+    }
+    else {
+        img_desc_1d.image_width = (nb02 * ne02 / 4)/4;
+    }
+    img_desc_1d.buffer = A_sub_buffer;
+    A_image1d = clCreateImage(context, CL_MEM_READ_ONLY, &img_fmt_1d, &img_desc_1d, NULL, &status);
+    CL_CHECK(status);
+
+    // create sub-buffer for output C
+    // <--------------------------------------------> //
+    region.origin = (extrad->offset);
+    region.size = ne0 * ne1 * dst->ne[2] * dst->nb[0]; // size of C in bytes
+    D_sub_buffer = clCreateSubBuffer((extrad->data_device), 0, CL_BUFFER_CREATE_TYPE_REGION, &region, &status);
+    CL_CHECK(status);
+    // <--------------------------------------------> //
+
+    // create image for C output
+    // <--------------------------------------------> //
+    img_fmt_1d = {CL_R, CL_FLOAT};
+    memset(&img_desc_1d, 0, sizeof(img_desc_1d));
+    img_desc_1d.image_type = CL_MEM_OBJECT_IMAGE1D_BUFFER;
+    img_desc_1d.image_width = ne0 * ne1 * dst->ne[2] * dst->nb[0] / 4;
+    img_desc_1d.buffer = D_sub_buffer;
+    D_image1d = clCreateImage(context, CL_MEM_WRITE_ONLY, &img_fmt_1d, &img_desc_1d, NULL, &status);
+    CL_CHECK(status);
+    // <--------------------------------------------> //
+
+    int offset_src0 = 0;
+    int offset_src1 = 0;
+
+    // set kernel args
+    // <--------------------------------------------> //
+    cl_uint k_arg = 0;
+    CL_CHECK(clSetKernelArg(kernel,  k_arg++, sizeof(cl_mem), &A_image1d));
+    CL_CHECK(clSetKernelArg(kernel,  k_arg++, sizeof(int),    &offset_src0));
+    CL_CHECK(clSetKernelArg(kernel,  k_arg++, sizeof(cl_mem), &B_sub_buffer));
+    CL_CHECK(clSetKernelArg(kernel,  k_arg++, sizeof(int),    &offset_src1));
+    CL_CHECK(clSetKernelArg(kernel,  k_arg++, sizeof(cl_mem), &D_image1d));
+    CL_CHECK(clSetKernelArg(kernel,  k_arg++, sizeof(int),    &extrad->offset));
+    CL_CHECK(clSetKernelArg(kernel,  k_arg++, sizeof(int),    &M));
+    CL_CHECK(clSetKernelArg(kernel,  k_arg++, sizeof(int),    &K));
+    CL_CHECK(clSetKernelArg(kernel,  k_arg++, sizeof(int),    &N));
+    CL_CHECK(clSetKernelArg(kernel,  k_arg++, sizeof(int),    &ne02));
+    CL_CHECK(clSetKernelArg(kernel,  k_arg++, sizeof(int),    &ne12));
+    CL_CHECK(clSetKernelArg(kernel,  k_arg++, sizeof(int),    &nb01));
+
+    size_t global_work_size[3] = {64, static_cast<size_t>(((M+63)/64)), static_cast<size_t>(((N+31)/32)*ne12)};
+    size_t local_work_size[3] = {64, 1, 2};
+
+    backend_ctx->enqueue_ndrange_kernel(kernel, 3, global_work_size, local_work_size, dst);
+
+    // deallocate sub buffers and images
+    // <--------------------------------------------> //
+    CL_CHECK(clReleaseMemObject(A_image1d));
+    CL_CHECK(clReleaseMemObject(D_image1d));
+    CL_CHECK(clReleaseMemObject(A_sub_buffer));
+    CL_CHECK(clReleaseMemObject(B_sub_buffer));
+    CL_CHECK(clReleaseMemObject(D_sub_buffer));
+}
+
 static void ggml_cl_mul_mat(ggml_backend_t backend, const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst) {
     GGML_ASSERT(src0);
     GGML_ASSERT(src0->extra);
@@ -6731,6 +6894,13 @@ static void ggml_cl_mul_mat(ggml_backend_t backend, const ggml_tensor * src0, co
 #ifdef GGML_OPENCL_USE_ADRENO_KERNELS
     cl_context context = backend_ctx->context;
 
+    if(src0t == GGML_TYPE_F16 && src1t == GGML_TYPE_F32){
+        if (ne01 >= 64 && ne1 >= 32 && ne00 >= 16 && (ne12 % ne02) == 0){
+            ggml_cl_mul_mat_kq_kqv_adreno(backend, src0, src1, dst);
+            return;
+        }
+    }
+
     if (ne01 && ne1 && use_adreno_kernels(backend_ctx, src0)) {
 
     // init CL objects

ggml/src/ggml-opencl/kernels/mul_mm_f16_f32_kq_kqv.cl

@@ -0,0 +1,273 @@
+#pragma OPENCL EXTENSION cl_khr_fp16 : enable
+#pragma OPENCL EXTENSION cl_khr_subgroups : enable
+
+#define LM_FIRST_256B   0
+#define LM_SECOND_256B  64
+#define LM_THIRD_256B   128
+#define LM_FOURTH_256B  192
+
+
+inline float16 mm_load_a(
+    image1d_buffer_t matrix_A,
+    uint subMatrixAStartInElements,
+    int nb01,
+    int line_stride_matrix_A_in_bytes
+) {
+    __private float8 regA;
+    size_t sub_block_id_m = get_local_id(0);
+
+#ifdef KQV
+    uint a_texCoord = subMatrixAStartInElements/2 + (sub_block_id_m * nb01/4);
+#else // KQ
+    uint a_texCoord = subMatrixAStartInElements/2 + (sub_block_id_m * line_stride_matrix_A_in_bytes/4);
+#endif
+
+    regA.s0123  = read_imagef(matrix_A, a_texCoord/4);
+    regA.s4567  = read_imagef(matrix_A, (a_texCoord+4)/4);
+
+    return convert_float16(as_half16(regA));
+}
+
+inline float4 alu_32(
+    float16 regA,
+    __local float4* matrix_B_vec
+) {
+
+    __private float4 rC = 0;
+    int i = get_sub_group_id() * 64;
+
+    rC += regA.s0  * matrix_B_vec[i];
+    rC += regA.s1  * matrix_B_vec[i + 16];
+    rC += regA.s4  * matrix_B_vec[i + 1];
+    rC += regA.s5  * matrix_B_vec[i + 17];
+    rC += regA.s8  * matrix_B_vec[i + 2];
+    rC += regA.s9  * matrix_B_vec[i + 18];
+    rC += regA.sc  * matrix_B_vec[i + 3];
+    rC += regA.sd  * matrix_B_vec[i + 19];
+
+    i += 32;
+
+    rC += regA.s2  * matrix_B_vec[i];
+     rC += regA.s3  * matrix_B_vec[i + 16];
+    rC += regA.s6  * matrix_B_vec[i + 1];
+    rC += regA.s7  * matrix_B_vec[i + 17];
+    rC += regA.sa  * matrix_B_vec[i + 2];
+    rC += regA.sb  * matrix_B_vec[i + 18];
+    rC += regA.se  * matrix_B_vec[i + 3];
+    rC += regA.sf  * matrix_B_vec[i + 19];
+
+    return rC;
+}
+
+inline float16 alu_16(
+    float16 regA,
+    __local float* matrix_B_local
+) {
+    float16 out;
+    __local float4* matrix_B_vec = (__local float4*)matrix_B_local;
+
+    out.s0123 = alu_32(regA, matrix_B_vec);
+    out.s4567 = alu_32(regA, matrix_B_vec + 4);
+    out.s89ab = alu_32(regA, matrix_B_vec + 8);
+    out.scdef = alu_32(regA, matrix_B_vec + 12);
+
+    return out;
+}
+
+inline void mm_mad(
+    __local float* matrix_B_local,
+    float16 regA,
+    float8 regB,
+    uint b_localOffsetInWords,
+    float16* regC0_ptr,
+    float16* regC1_ptr
+) {
+    int offset = b_localOffsetInWords + get_sub_group_id() * 256;
+
+    matrix_B_local[offset + LM_FIRST_256B] = regB.s0;
+    matrix_B_local[offset + LM_SECOND_256B] = regB.s1;
+    matrix_B_local[offset + LM_THIRD_256B] = regB.s2;
+    matrix_B_local[offset + LM_FOURTH_256B] = regB.s3;
+
+    float16 add0 = alu_16(regA, matrix_B_local);
+    *regC0_ptr += add0;
+
+    matrix_B_local[offset + LM_FIRST_256B] = regB.s4;
+    matrix_B_local[offset + LM_SECOND_256B] = regB.s5;
+    matrix_B_local[offset + LM_THIRD_256B] = regB.s6;
+    matrix_B_local[offset + LM_FOURTH_256B] = regB.s7;
+
+    float16 add1 = alu_16(regA, matrix_B_local);
+    *regC1_ptr += add1;
+}
+
+inline void mm_store_c_N(
+    __write_only image1d_buffer_t matrix_C,
+    float16 regC0,
+    float16 regC1,
+    uint subMatrixCStartInElements,
+    int line_stride_matrix_C_in_bytes,
+    int mask
+) {
+    size_t sub_block_id_m = get_local_id(0);
+
+    uint strideInWords     = line_stride_matrix_C_in_bytes/4;
+    uint c_coordInWords_0  = (subMatrixCStartInElements + sub_block_id_m);
+
+    uint c_coordInWords_1  = c_coordInWords_0 + 1  * strideInWords;
+    uint c_coordInWords_2  = c_coordInWords_0 + 2  * strideInWords;
+    uint c_coordInWords_3  = c_coordInWords_0 + 3  * strideInWords;
+    uint c_coordInWords_4  = c_coordInWords_0 + 4  * strideInWords;
+    uint c_coordInWords_5  = c_coordInWords_0 + 5  * strideInWords;
+    uint c_coordInWords_6  = c_coordInWords_0 + 6  * strideInWords;
+    uint c_coordInWords_7  = c_coordInWords_0 + 7  * strideInWords;
+    uint c_coordInWords_8  = c_coordInWords_0 + 8  * strideInWords;
+    uint c_coordInWords_9  = c_coordInWords_0 + 9  * strideInWords;
+    uint c_coordInWords_10 = c_coordInWords_0 + 10 * strideInWords;
+    uint c_coordInWords_11 = c_coordInWords_0 + 11 * strideInWords;
+    uint c_coordInWords_12 = c_coordInWords_0 + 12 * strideInWords;
+    uint c_coordInWords_13 = c_coordInWords_0 + 13 * strideInWords;
+    uint c_coordInWords_14 = c_coordInWords_0 + 14 * strideInWords;
+    uint c_coordInWords_15 = c_coordInWords_0 + 15 * strideInWords;
+    uint c_coordInWords_16 = c_coordInWords_0 + 16 * strideInWords;
+    uint c_coordInWords_17 = c_coordInWords_0 + 17 * strideInWords;
+    uint c_coordInWords_18 = c_coordInWords_0 + 18 * strideInWords;
+    uint c_coordInWords_19 = c_coordInWords_0 + 19 * strideInWords;
+    uint c_coordInWords_20 = c_coordInWords_0 + 20 * strideInWords;
+    uint c_coordInWords_21 = c_coordInWords_0 + 21 * strideInWords;
+    uint c_coordInWords_22 = c_coordInWords_0 + 22 * strideInWords;
+    uint c_coordInWords_23 = c_coordInWords_0 + 23 * strideInWords;
+    uint c_coordInWords_24 = c_coordInWords_0 + 24 * strideInWords;
+    uint c_coordInWords_25 = c_coordInWords_0 + 25 * strideInWords;
+    uint c_coordInWords_26 = c_coordInWords_0 + 26 * strideInWords;
+    uint c_coordInWords_27 = c_coordInWords_0 + 27 * strideInWords;
+    uint c_coordInWords_28 = c_coordInWords_0 + 28 * strideInWords;
+    uint c_coordInWords_29 = c_coordInWords_0 + 29 * strideInWords;
+    uint c_coordInWords_30 = c_coordInWords_0 + 30 * strideInWords;
+    uint c_coordInWords_31 = c_coordInWords_0 + 31 * strideInWords;
+
+    if (mask > 0)  { write_imagef(matrix_C, c_coordInWords_0, regC0.s0);  }
+    if (mask > 1)  { write_imagef(matrix_C, c_coordInWords_1, regC0.s1);  }
+    if (mask > 2)  { write_imagef(matrix_C, c_coordInWords_2, regC0.s2);  }
+    if (mask > 3)  { write_imagef(matrix_C, c_coordInWords_3, regC0.s3);  }
+    if (mask > 4)  { write_imagef(matrix_C, c_coordInWords_4, regC0.s4);  }
+    if (mask > 5)  { write_imagef(matrix_C, c_coordInWords_5, regC0.s5);  }
+    if (mask > 6)  { write_imagef(matrix_C, c_coordInWords_6, regC0.s6);  }
+    if (mask > 7)  { write_imagef(matrix_C, c_coordInWords_7, regC0.s7);  }
+    if (mask > 8)  { write_imagef(matrix_C, c_coordInWords_8, regC0.s8);  }
+    if (mask > 9)  { write_imagef(matrix_C, c_coordInWords_9, regC0.s9);  }
+    if (mask > 10) { write_imagef(matrix_C, c_coordInWords_10, regC0.sa); }
+    if (mask > 11) { write_imagef(matrix_C, c_coordInWords_11, regC0.sb); }
+    if (mask > 12) { write_imagef(matrix_C, c_coordInWords_12, regC0.sc); }
+    if (mask > 13) { write_imagef(matrix_C, c_coordInWords_13, regC0.sd); }
+    if (mask > 14) { write_imagef(matrix_C, c_coordInWords_14, regC0.se); }
+    if (mask > 15) { write_imagef(matrix_C, c_coordInWords_15, regC0.sf); }
+    if (mask > 16) { write_imagef(matrix_C, c_coordInWords_16, regC1.s0); }
+    if (mask > 17) { write_imagef(matrix_C, c_coordInWords_17, regC1.s1); }
+    if (mask > 18) { write_imagef(matrix_C, c_coordInWords_18, regC1.s2); }
+    if (mask > 19) { write_imagef(matrix_C, c_coordInWords_19, regC1.s3); }
+    if (mask > 20) { write_imagef(matrix_C, c_coordInWords_20, regC1.s4); }
+    if (mask > 21) { write_imagef(matrix_C, c_coordInWords_21, regC1.s5); }
+    if (mask > 22) { write_imagef(matrix_C, c_coordInWords_22, regC1.s6); }
+    if (mask > 23) { write_imagef(matrix_C, c_coordInWords_23, regC1.s7); }
+    if (mask > 24) { write_imagef(matrix_C, c_coordInWords_24, regC1.s8); }
+    if (mask > 25) { write_imagef(matrix_C, c_coordInWords_25, regC1.s9); }
+    if (mask > 26) { write_imagef(matrix_C, c_coordInWords_26, regC1.sa); }
+    if (mask > 27) { write_imagef(matrix_C, c_coordInWords_27, regC1.sb); }
+    if (mask > 28) { write_imagef(matrix_C, c_coordInWords_28, regC1.sc); }
+    if (mask > 29) { write_imagef(matrix_C, c_coordInWords_29, regC1.sd); }
+    if (mask > 30) { write_imagef(matrix_C, c_coordInWords_30, regC1.se); }
+    if (mask > 31) { write_imagef(matrix_C, c_coordInWords_31, regC1.sf); }
+}
+
+#define TILESIZE_K 16
+#define TILESIZE_M 64
+#define TILESIZE_N 32
+#ifdef KQV
+__kernel void mul_mm_f16_f32_kqv(
+#else
+__kernel void mul_mm_f16_f32_kq(
+#endif
+        __read_only  image1d_buffer_t matrix_A,
+        int offset0,
+        __global float* matrix_B,
+        int offset1,
+        __write_only image1d_buffer_t matrix_C,
+        int offsetd,
+        int M, int K, int N,
+        int D_A,
+        int D_B,
+        int nb01
+) {
+
+    uint block_id_m = get_global_id(1);
+    uint block_id_n = get_global_id(2) % ((N+TILESIZE_N-1)/TILESIZE_N);
+    uint block_id_d = get_global_id(2) / ((N+TILESIZE_N-1)/TILESIZE_N);
+
+    __private float16  regA;
+    __private float8   regB;
+    __private float16 regC0;
+    __private float16 regC1;
+
+    const uint col   = block_id_m * TILESIZE_M;
+    const uint row   = block_id_n * TILESIZE_N;
+    const uint depth_A = block_id_d / (D_B/D_A);
+    const uint depth_B = block_id_d;
+
+#ifdef KQV
+    int line_stride_matrix_A_in_bytes = nb01 * M;
+    int line_stride_matrix_B_in_bytes = K * N * 4;
+#else
+    int line_stride_matrix_A_in_bytes = K * D_A * 2;
+    int line_stride_matrix_B_in_bytes = K * D_B * 4;
+#endif
+
+    int line_stride_matrix_C_in_bytes = M * 4;
+
+    const uint strideAinElements = line_stride_matrix_A_in_bytes / 2;
+    const uint strideBinElements = line_stride_matrix_B_in_bytes / 4;
+
+    size_t sub_block_id_m = get_local_id(0);
+
+    uint b_localOffsetInWords = (sub_block_id_m/16)*16
+                           + ((((sub_block_id_m)>>0)&1)<<2)
+                           + ((((sub_block_id_m)>>1)&1)<<3)
+                           + ((((sub_block_id_m)>>2)&1)<<0)
+                           + ((((sub_block_id_m)>>3)&1)<<1);
+
+    uint2 b_globalOffsetInWords_xy = {((sub_block_id_m%4)*4), (sub_block_id_m>>2)};
+    uint b_globalOffsetInWords00, b_globalOffsetInWords16;
+#ifdef KQV
+    b_globalOffsetInWords00 = b_globalOffsetInWords_xy.x + b_globalOffsetInWords_xy.y*K;
+    b_globalOffsetInWords16 = b_globalOffsetInWords00 + (16 * K);
+    uint subMatrixAStartInElements = depth_A * strideAinElements + col * nb01 / 2;
+    uint subMatrixBStartInElements = depth_B * strideBinElements + row * K;
+#else
+    b_globalOffsetInWords00 = b_globalOffsetInWords_xy.x + b_globalOffsetInWords_xy.y*line_stride_matrix_B_in_bytes/4;
+    b_globalOffsetInWords16 = b_globalOffsetInWords00 + (16 * line_stride_matrix_B_in_bytes/4);
+    uint subMatrixAStartInElements = col * strideAinElements + depth_A * K;
+    uint subMatrixBStartInElements = row * strideBinElements + depth_B * K;
+#endif
+
+    __local float matrix_B_local[1024];
+
+    for (uint step=0; step < K; step+=TILESIZE_K) {
+        size_t sub_block_id_m = get_local_id(0);
+        regA = mm_load_a(matrix_A, subMatrixAStartInElements, nb01, line_stride_matrix_A_in_bytes);
+
+        uint b_coordInWords00 = subMatrixBStartInElements + b_globalOffsetInWords00;
+        uint b_coordInWords16 = subMatrixBStartInElements + b_globalOffsetInWords16;
+
+        regB.s0123 = vload4(b_coordInWords00/4, matrix_B);
+        regB.s4567 = vload4(b_coordInWords16/4, matrix_B);
+
+        mm_mad(matrix_B_local, regA, regB, b_localOffsetInWords, &regC0, &regC1);
+
+        subMatrixAStartInElements += TILESIZE_K;
+        subMatrixBStartInElements += TILESIZE_K;
+    }
+
+    uint subMatrixCStartInElements = depth_B * N * M + row * M + col;
+    mm_store_c_N(matrix_C, regC0, regC1, subMatrixCStartInElements, line_stride_matrix_C_in_bytes, (N-block_id_n*32));
+}
+

ggml/src/ggml-opencl/kernels/rms_norm.cl

@@ -134,6 +134,15 @@ kernel void kernel_rms_norm_mul(
     src1 = src1 + offset1;
     dst  = dst  + offsetd;
 
+    // The size of sum is sizeof(float)*subgroup_size.
+    // Each subgroup writes its partial sum to this array.
+    // So the number of subgroups per workgroup for this kernel cannot exceed the subgroup size.
+    // This is generally true -
+    // for subgroup size 64, workgroup size should be less than 4096 (the max is usually 1024).
+    if (get_sub_group_id() == 0) {
+        sum[get_sub_group_local_id()] = 0.0f;
+    }
+
     int i03 = get_group_id(2);
     int i02 = get_group_id(1);
     int i01 = get_group_id(0);
@@ -148,24 +157,30 @@ kernel void kernel_rms_norm_mul(
         sumf += dot(x[i00], x[i00]);
     }
     sumf = sub_group_reduce_add(sumf);
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
     if (get_sub_group_local_id() == 0) {
         sum[get_sub_group_id()] = sumf;
     }
 
     barrier(CLK_LOCAL_MEM_FENCE);
 
-    for (uint i = get_local_size(0) / get_max_sub_group_size() / 2; i > 0; i /= 2) {
-       if (get_local_id(0) < i) {
-           sum[get_local_id(0)] += sum[get_local_id(0) + i];
-       }
-    }
-    if (get_local_id(0) == 0) {
-        sum[0] /= ne00;
-    }
+    //for (uint i = get_local_size(0) / get_max_sub_group_size() / 2; i > 0; i /= 2) {
+    //   if (get_local_id(0) < i) {
+    //       sum[get_local_id(0)] += sum[get_local_id(0) + i];
+    //   }
+    //}
+    //if (get_local_id(0) == 0) {
+    //    sum[0] /= ne00;
+    //}
 
-    barrier(CLK_LOCAL_MEM_FENCE);
+    //barrier(CLK_LOCAL_MEM_FENCE);
 
-    float mean  = sum[0];
+    sumf = sum[get_sub_group_local_id()];
+    sumf = sub_group_reduce_add(sumf);
+
+    float mean  = sumf / ne00;
     float scale = 1.0f/sqrt(mean + eps);
 
     global float4 * y = (global float4 *) (dst + i03*nb3 + i02*nb2 + i01*nb1);

ggml/src/ggml-sycl/element_wise.cpp

@@ -170,73 +170,31 @@ static __dpct_inline__ T op_trunc(T x) {
     return sycl::trunc(x);
 }
 
-template<typename T>
-static void unary_op_sgn_kernel(const T * x, T * dst, const int k, const sycl::nd_item<1> &item_ct1) {
-    SYCL_GLOBAL_ID_LOOP(k, item_ct1) {
-        dst[i] = op_sgn(x[i]);
-    }
-}
+template<typename T, typename F>
+static void unary_op_generic_kernel(
+        const T * x,
+        T * dst,
+        const int k,
+        const int64_t ne0, const int64_t ne1, const int64_t ne2, const int64_t ne3,
+        const size_t nb0,  const size_t nb1,  const size_t nb2,  const size_t nb3,
+        const size_t nbd0, const size_t nbd1, const size_t nbd2, const size_t nbd3,
+        const sycl::nd_item<1> & item_ct1,
+        F func) {
 
-template<typename T>
-static void unary_op_abs_kernel(const T * x, T * dst, const int k, const sycl::nd_item<1> &item_ct1) {
+        (void) ne3;
     SYCL_GLOBAL_ID_LOOP(k, item_ct1) {
-        dst[i] = op_abs(x[i]);
-    }
-}
+        const int64_t i0 =  i % ne0;
+        const int64_t i1 = (i / ne0)        % ne1;
+        const int64_t i2 = (i / (ne0*ne1))  % ne2;
+        const int64_t i3 =  i / (ne0*ne1*ne2);
 
-template<typename T>
-static void unary_op_elu_kernel(const T * x, T * dst, const int k, const sycl::nd_item<1> &item_ct1) {
-    SYCL_GLOBAL_ID_LOOP(k, item_ct1) {
-        dst[i] = op_elu(x[i]);
-    }
-}
+        const char * src_base = (const char *) x;
+        char       * dst_base = (char *) dst;
 
-template<typename T>
-static void unary_op_gelu_kernel(const T * x, T * dst, const int k, const sycl::nd_item<1> &item_ct1) {
-    SYCL_GLOBAL_ID_LOOP(k, item_ct1) {
-        dst[i] = op_gelu(x[i]);
-    }
-}
+        const T * srcp = (const T *)(src_base + i0*nb0  + i1*nb1  + i2*nb2  + i3*nb3 );
+        T *       dstp = (T *)(dst_base + i0*nbd0 + i1*nbd1 + i2*nbd2 + i3*nbd3);
 
-template<typename T>
-static void unary_op_silu_kernel(const T * x, T * dst, const int k, const sycl::nd_item<1> &item_ct1) {
-    SYCL_GLOBAL_ID_LOOP(k, item_ct1) {
-        dst[i] = op_silu(x[i]);
-    }
-}
-
-template<typename T>
-static void unary_op_gelu_quick_kernel(const T * x, T * dst, const int k, const sycl::nd_item<1> &item_ct1) {
-    SYCL_GLOBAL_ID_LOOP(k, item_ct1) {
-        dst[i] = op_gelu_quick(x[i]);
-    }
-}
-
-template<typename T>
-static void unary_op_gelu_erf_kernel(const T * x, T * dst, const int k, const sycl::nd_item<1> &item_ct1) {
-    SYCL_GLOBAL_ID_LOOP(k, item_ct1) {
-        dst[i] = op_gelu_erf(x[i]);
-    }
-}
-
-template<typename T>
-static void unary_op_tanh_kernel(const T * x, T * dst, const int k, const sycl::nd_item<1> &item_ct1) {
-    SYCL_GLOBAL_ID_LOOP(k, item_ct1) {
-        dst[i] = op_tanh(x[i]);
-    }
-}
-
-template<typename T>
-static void unary_op_relu_kernel(const T * x, T * dst, const int k, const sycl::nd_item<1> &item_ct1) {
-    SYCL_GLOBAL_ID_LOOP(k, item_ct1) {
-        dst[i] = op_relu(x[i]);
-    }
-}
-
-template<typename T>
-static void unary_op_sigmoid_kernel(const T * x, T * dst, const int k, const sycl::nd_item<1> &item_ct1) {
-    SYCL_GLOBAL_ID_LOOP(k, item_ct1) {
-        dst[i] = op_sigmoid(x[i]);
+        *dstp = func(*srcp);
     }
 }
 
@@ -261,27 +219,6 @@ static void unary_op_cos_kernel(const T * x, T * dst, const int k, const sycl::n
     }
 }
 
-template<typename T>
-static void unary_op_hardsigmoid_kernel(const T * x, T * dst, const int k, const sycl::nd_item<1> &item_ct1) {
-    SYCL_GLOBAL_ID_LOOP(k, item_ct1) {
-        dst[i] = op_hardsigmoid(x[i]);
-    }
-}
-
-template<typename T>
-static void unary_op_hardswish_kernel(const T * x, T * dst, const int k, const sycl::nd_item<1> &item_ct1) {
-    SYCL_GLOBAL_ID_LOOP(k, item_ct1) {
-        dst[i] = op_hardswish(x[i]);
-    }
-}
-
-template<typename T>
-static void unary_op_exp_kernel(const T * x, T * dst, const int k, const sycl::nd_item<1> &item_ct1) {
-    SYCL_GLOBAL_ID_LOOP(k, item_ct1) {
-        dst[i] = op_exp(x[i]);
-    }
-}
-
 template<typename T>
 static void unary_op_log_kernel(const T * x, T * dst, const int k, const sycl::nd_item<1> &item_ct1) {
     SYCL_GLOBAL_ID_LOOP(k, item_ct1) {
@@ -289,19 +226,6 @@ static void unary_op_log_kernel(const T * x, T * dst, const int k, const sycl::n
     }
 }
 
-template<typename T>
-static void unary_op_neg_kernel(const T * x, T * dst, const int k, const sycl::nd_item<1> &item_ct1) {
-    SYCL_GLOBAL_ID_LOOP(k, item_ct1) {
-        dst[i] = op_neg(x[i]);
-    }
-}
-
-template<typename T>
-static void unary_op_step_kernel(const T * x, T * dst, const int k, const sycl::nd_item<1> &item_ct1) {
-    SYCL_GLOBAL_ID_LOOP(k, item_ct1) {
-        dst[i] = op_step(x[i]);
-    }
-}
 
 template<typename T>
 static void unary_op_leaky_relu_kernel(const T * x, T * dst, const int k, float negative_slope, const sycl::nd_item<1> &item_ct1) {
@@ -620,6 +544,48 @@ static inline void dispatch_ggml_sycl_op_upscale(ggml_backend_sycl_context & ctx
     }
 }
 
+template<typename F>
+static inline void ggml_sycl_op_unary(
+        ggml_backend_sycl_context & ctx, ggml_tensor * dst, F func) {
+
+    ggml_tensor * src0 = dst->src[0];
+
+    const int64_t ne0  = dst->ne[0];
+    const int64_t ne1  = dst->ne[1];
+    const int64_t ne2  = dst->ne[2];
+    const int64_t ne3  = dst->ne[3];
+
+    const size_t  nb0  = src0->nb[0];
+    const size_t  nb1  = src0->nb[1];
+    const size_t  nb2  = src0->nb[2];
+    const size_t  nb3  = src0->nb[3];
+
+    const size_t  nbd0 = dst->nb[0];
+    const size_t  nbd1 = dst->nb[1];
+    const size_t  nbd2 = dst->nb[2];
+    const size_t  nbd3 = dst->nb[3];
+
+    ggml_sycl_detail::dispatch_ggml_sycl_op_unary(ctx, dst,
+        [=](const auto* src, auto* dst_ptr, int k_elements, queue_ptr stream) {
+
+            const int num_blocks = ceil_div(k_elements, 256);
+
+            stream->parallel_for(
+                sycl::nd_range<1>(sycl::range<1>(num_blocks) * sycl::range<1>(256),
+                                  sycl::range<1>(256)),
+                [=](sycl::nd_item<1> item_ct1) {
+                    unary_op_generic_kernel(
+                        src, dst_ptr, k_elements,
+                        ne0, ne1, ne2, ne3,
+                        nb0, nb1, nb2, nb3,
+                        nbd0, nbd1, nbd2, nbd3,
+                        item_ct1,
+                        func
+                    );
+                });
+        });
+}
+
 
 static inline void ggml_sycl_op_arange(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
     GGML_ASSERT(dst->type == GGML_TYPE_F32);
@@ -645,159 +611,75 @@ static inline void ggml_sycl_op_arange(ggml_backend_sycl_context & ctx, ggml_ten
 
 
 static inline void ggml_sycl_op_sgn(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
-    ggml_sycl_detail::dispatch_ggml_sycl_op_unary(ctx, dst,
-        [](const auto* src, auto* dst_ptr, int k_elements, queue_ptr stream) {
-            const int num_blocks = ceil_div(k_elements, 256);
-            stream->parallel_for(
-                sycl::nd_range<1>(sycl::range<1>(num_blocks) * sycl::range<1>(256),
-                                  sycl::range<1>(256)),
-                [=](sycl::nd_item<1> item_ct1) {
-                    unary_op_sgn_kernel(src, dst_ptr, k_elements, item_ct1);
-                });
-        });
+    ggml_sycl_detail::ggml_sycl_op_unary(ctx, dst, [](auto x) {
+        return op_sgn(x);
+    });
 }
 
+
 static inline void ggml_sycl_op_abs(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
-    ggml_sycl_detail::dispatch_ggml_sycl_op_unary(ctx, dst,
-        [](const auto* src, auto* dst_ptr, int k_elements, queue_ptr stream) {
-            const int num_blocks = ceil_div(k_elements, 256);
-            stream->parallel_for(
-                sycl::nd_range<1>(sycl::range<1>(num_blocks) * sycl::range<1>(256),
-                                  sycl::range<1>(256)),
-                [=](sycl::nd_item<1> item_ct1) {
-                    unary_op_abs_kernel(src, dst_ptr, k_elements, item_ct1);
-                });
-        });
+    ggml_sycl_detail::ggml_sycl_op_unary(ctx, dst, [](auto x) {
+        return op_abs(x);
+    });
 }
 
 static inline void ggml_sycl_op_elu(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
-    ggml_sycl_detail::dispatch_ggml_sycl_op_unary(ctx, dst,
-        [](const auto* src, auto* dst_ptr, int k_elements, queue_ptr stream) {
-            const int num_blocks = ceil_div(k_elements, 256);
-            stream->parallel_for(
-                sycl::nd_range<1>(sycl::range<1>(num_blocks) * sycl::range<1>(256),
-                                  sycl::range<1>(256)),
-                [=](sycl::nd_item<1> item_ct1) {
-                    unary_op_elu_kernel(src, dst_ptr, k_elements, item_ct1);
-                });
-        });
+    ggml_sycl_detail::ggml_sycl_op_unary(ctx, dst, [](auto x) {
+        return op_elu(x);
+    });
 }
-
 static inline void ggml_sycl_op_silu(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
-    ggml_sycl_detail::dispatch_ggml_sycl_op_unary(ctx, dst,
-        [](const auto* src, auto* dst_ptr, int k_elements, queue_ptr stream) {
-            const int num_blocks = ceil_div(k_elements, SYCL_SILU_BLOCK_SIZE);
-            stream->parallel_for(
-                sycl::nd_range<1>(sycl::range<1>(num_blocks) * sycl::range<1>(SYCL_SILU_BLOCK_SIZE),
-                                  sycl::range<1>(SYCL_SILU_BLOCK_SIZE)),
-                [=](sycl::nd_item<1> item_ct1) {
-                    unary_op_silu_kernel(src, dst_ptr, k_elements, item_ct1);
-                });
-        });
+    ggml_sycl_detail::ggml_sycl_op_unary(ctx, dst, [](auto x) {
+        return op_silu(x);
+    });
 }
 
 static inline void ggml_sycl_op_gelu(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
-    ggml_sycl_detail::dispatch_ggml_sycl_op_unary(ctx, dst,
-        [](const auto* src, auto* dst_ptr, int k_elements, queue_ptr stream) {
-            const int num_blocks = ceil_div(k_elements, SYCL_GELU_BLOCK_SIZE);
-            stream->parallel_for(
-                sycl::nd_range<1>(sycl::range<1>(num_blocks) * sycl::range<1>(SYCL_GELU_BLOCK_SIZE),
-                                  sycl::range<1>(SYCL_GELU_BLOCK_SIZE)),
-                [=](sycl::nd_item<1> item_ct1) {
-                    unary_op_gelu_kernel(src, dst_ptr, k_elements, item_ct1);
-                });
-        });
+    ggml_sycl_detail::ggml_sycl_op_unary(ctx, dst, [](auto x) {
+        return op_gelu(x);
+    });
 }
 
-static inline void ggml_sycl_op_gelu_quick(ggml_backend_sycl_context & ctx, ggml_tensor *dst) {
-    ggml_sycl_detail::dispatch_ggml_sycl_op_unary(ctx, dst,
-        [](const auto* src, auto* dst_ptr, int k_elements, queue_ptr stream) {
-            const int num_blocks = ceil_div(k_elements, SYCL_GELU_BLOCK_SIZE);
-            stream->parallel_for(
-                sycl::nd_range<1>(sycl::range<1>(num_blocks) * sycl::range<1>(SYCL_GELU_BLOCK_SIZE),
-                                  sycl::range<1>(SYCL_GELU_BLOCK_SIZE)),
-                [=](sycl::nd_item<1> item_ct1) {
-                    unary_op_gelu_quick_kernel(src, dst_ptr, k_elements, item_ct1);
-                });
-        });
+static inline void ggml_sycl_op_gelu_quick(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
+    ggml_sycl_detail::ggml_sycl_op_unary(ctx, dst, [](auto x) {
+        return op_gelu_quick(x);
+    });
 }
 
-static inline void ggml_sycl_op_gelu_erf(ggml_backend_sycl_context & ctx, ggml_tensor *dst) {
-    ggml_sycl_detail::dispatch_ggml_sycl_op_unary(ctx, dst,
-        [](const auto* src, auto* dst_ptr, int k_elements, queue_ptr stream) {
-            const int num_blocks = ceil_div(k_elements, SYCL_GELU_BLOCK_SIZE);
-            stream->parallel_for(
-                sycl::nd_range<1>(sycl::range<1>(num_blocks) * sycl::range<1>(SYCL_GELU_BLOCK_SIZE),
-                                  sycl::range<1>(SYCL_GELU_BLOCK_SIZE)),
-                [=](sycl::nd_item<1> item_ct1) {
-                    unary_op_gelu_erf_kernel(src, dst_ptr, k_elements, item_ct1);
-                });
-        });
+static inline void ggml_sycl_op_gelu_erf(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
+    ggml_sycl_detail::ggml_sycl_op_unary(ctx, dst, [](auto x) {
+        return op_gelu_erf(x);
+    });
 }
 
 static inline void ggml_sycl_op_tanh(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
-    ggml_sycl_detail::dispatch_ggml_sycl_op_unary(ctx, dst,
-        [](const auto* src, auto* dst_ptr, int k_elements, queue_ptr stream) {
-            const int num_blocks = ceil_div(k_elements, SYCL_TANH_BLOCK_SIZE);
-            stream->parallel_for(
-                sycl::nd_range<1>(sycl::range<1>(num_blocks) * sycl::range<1>(SYCL_TANH_BLOCK_SIZE),
-                                  sycl::range<1>(SYCL_TANH_BLOCK_SIZE)),
-                [=](sycl::nd_item<1> item_ct1) {
-                    unary_op_tanh_kernel(src, dst_ptr, k_elements, item_ct1);
-                });
-        });
+    ggml_sycl_detail::ggml_sycl_op_unary(ctx, dst, [](auto x) {
+        return op_tanh(x);
+    });
 }
 
 static inline void ggml_sycl_op_relu(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
-    ggml_sycl_detail::dispatch_ggml_sycl_op_unary(ctx, dst,
-        [](const auto* src, auto* dst_ptr, int k_elements, queue_ptr stream) {
-            const int num_blocks = ceil_div(k_elements, SYCL_RELU_BLOCK_SIZE);
-            stream->parallel_for(
-                sycl::nd_range<1>(sycl::range<1>(num_blocks) * sycl::range<1>(SYCL_RELU_BLOCK_SIZE),
-                                  sycl::range<1>(SYCL_RELU_BLOCK_SIZE)),
-                [=](sycl::nd_item<1> item_ct1) {
-                    unary_op_relu_kernel(src, dst_ptr, k_elements, item_ct1);
-                });
-        });
+    ggml_sycl_detail::ggml_sycl_op_unary(ctx, dst, [](auto x) {
+        return op_relu(x);
+    });
 }
 
 static inline void ggml_sycl_op_hardsigmoid(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
-    ggml_sycl_detail::dispatch_ggml_sycl_op_unary(ctx, dst,
-        [](const auto* src, auto* dst_ptr, int k_elements, queue_ptr stream) {
-            const int num_blocks = ceil_div(k_elements, SYCL_HARDSIGMOID_BLOCK_SIZE);
-            stream->parallel_for(
-                sycl::nd_range<1>(sycl::range<1>(num_blocks) * sycl::range<1>(SYCL_HARDSIGMOID_BLOCK_SIZE),
-                                  sycl::range<1>(SYCL_HARDSIGMOID_BLOCK_SIZE)),
-                [=](sycl::nd_item<1> item_ct1) {
-                    unary_op_hardsigmoid_kernel(src, dst_ptr, k_elements, item_ct1);
-                });
-        });
+    ggml_sycl_detail::ggml_sycl_op_unary(ctx, dst, [](auto x) {
+        return op_hardsigmoid(x);
+    });
 }
 
 static inline void ggml_sycl_op_hardswish(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
-    ggml_sycl_detail::dispatch_ggml_sycl_op_unary(ctx, dst,
-        [](const auto* src, auto* dst_ptr, int k_elements, queue_ptr stream) {
-            const int num_blocks = ceil_div(k_elements, SYCL_HARDSWISH_BLOCK_SIZE);
-            stream->parallel_for(
-                sycl::nd_range<1>(sycl::range<1>(num_blocks) * sycl::range<1>(SYCL_HARDSWISH_BLOCK_SIZE),
-                                  sycl::range<1>(SYCL_HARDSWISH_BLOCK_SIZE)),
-                [=](sycl::nd_item<1> item_ct1) {
-                    unary_op_hardswish_kernel(src, dst_ptr, k_elements, item_ct1);
-                });
-        });
+    ggml_sycl_detail::ggml_sycl_op_unary(ctx, dst, [](auto x) {
+        return op_hardswish(x);
+    });
 }
 
 static inline void ggml_sycl_op_exp(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
-    ggml_sycl_detail::dispatch_ggml_sycl_op_unary(ctx, dst,
-        [](const auto* src, auto* dst_ptr, int k_elements, queue_ptr stream) {
-            const int num_blocks = ceil_div(k_elements, SYCL_EXP_BLOCK_SIZE);
-            stream->parallel_for(
-                sycl::nd_range<1>(sycl::range<1>(num_blocks) * sycl::range<1>(SYCL_EXP_BLOCK_SIZE),
-                                  sycl::range<1>(SYCL_EXP_BLOCK_SIZE)),
-                [=](sycl::nd_item<1> item_ct1) {
-                    unary_op_exp_kernel(src, dst_ptr, k_elements, item_ct1);
-                });
-        });
+    ggml_sycl_detail::ggml_sycl_op_unary(ctx, dst, [](auto x) {
+        return op_exp(x);
+    });
 }
 
 static inline void ggml_sycl_op_log(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
@@ -814,42 +696,22 @@ static inline void ggml_sycl_op_log(ggml_backend_sycl_context & ctx, ggml_tensor
 }
 
 static inline void ggml_sycl_op_neg(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
-    ggml_sycl_detail::dispatch_ggml_sycl_op_unary(ctx, dst,
-        [](const auto* src, auto* dst_ptr, int k_elements, queue_ptr stream) {
-            const int num_blocks = ceil_div(k_elements, SYCL_NEG_BLOCK_SIZE);
-            stream->parallel_for(
-                sycl::nd_range<1>(sycl::range<1>(num_blocks) * sycl::range<1>(SYCL_NEG_BLOCK_SIZE),
-                                  sycl::range<1>(SYCL_NEG_BLOCK_SIZE)),
-                [=](sycl::nd_item<1> item_ct1) {
-                    unary_op_neg_kernel(src, dst_ptr, k_elements, item_ct1);
-                });
-        });
+    ggml_sycl_detail::ggml_sycl_op_unary(ctx, dst, [](auto x) {
+        return op_neg(x);
+    });
 }
 
+
 static inline void ggml_sycl_op_step(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
-    ggml_sycl_detail::dispatch_ggml_sycl_op_unary(ctx, dst,
-        [](const auto* src, auto* dst_ptr, int k_elements, queue_ptr stream) {
-            const int num_blocks = ceil_div(k_elements, SYCL_NEG_BLOCK_SIZE); // Using NEG block size
-            stream->parallel_for(
-                sycl::nd_range<1>(sycl::range<1>(num_blocks) * sycl::range<1>(SYCL_NEG_BLOCK_SIZE),
-                                  sycl::range<1>(SYCL_NEG_BLOCK_SIZE)),
-                [=](sycl::nd_item<1> item_ct1) {
-                    unary_op_step_kernel(src, dst_ptr, k_elements, item_ct1);
-                });
-        });
+    ggml_sycl_detail::ggml_sycl_op_unary(ctx, dst, [](auto x) {
+        return op_step(x);
+    });
 }
 
 static inline void ggml_sycl_op_sigmoid(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {
-    ggml_sycl_detail::dispatch_ggml_sycl_op_unary(ctx, dst,
-        [](const auto* src, auto* dst_ptr, int k_elements, queue_ptr stream) {
-            const int num_blocks = ceil_div(k_elements, SYCL_SIGMOID_BLOCK_SIZE);
-            stream->parallel_for(
-                sycl::nd_range<1>(sycl::range<1>(num_blocks) * sycl::range<1>(SYCL_SIGMOID_BLOCK_SIZE),
-                                  sycl::range<1>(SYCL_SIGMOID_BLOCK_SIZE)),
-                [=](sycl::nd_item<1> item_ct1) {
-                    unary_op_sigmoid_kernel(src, dst_ptr, k_elements, item_ct1);
-                });
-        });
+    ggml_sycl_detail::ggml_sycl_op_unary(ctx, dst, [](auto x) {
+        return op_sigmoid(x);
+    });
 }
 
 static inline void ggml_sycl_op_sqrt(ggml_backend_sycl_context & ctx, ggml_tensor * dst) {

ggml/src/ggml-sycl/ggml-sycl.cpp

@@ -3933,6 +3933,7 @@ static bool ggml_sycl_compute_forward(ggml_backend_sycl_context & ctx, struct gg
             break;
         case GGML_OP_SSM_CONV:
             ggml_sycl_ssm_conv(ctx, dst);
+            break;
         case GGML_OP_ROLL:
             ggml_sycl_roll(ctx, dst);
             break;
@@ -4359,21 +4360,22 @@ static bool ggml_backend_sycl_device_supports_op(ggml_backend_dev_t dev, const g
             }
         case GGML_OP_UNARY:
             switch (ggml_get_unary_op(op)) {
+                case GGML_UNARY_OP_SGN:
+                case GGML_UNARY_OP_ABS:
                 case GGML_UNARY_OP_NEG:
                 case GGML_UNARY_OP_STEP:
+                case GGML_UNARY_OP_RELU:
+                case GGML_UNARY_OP_HARDSIGMOID:
+                case GGML_UNARY_OP_TANH:
                 case GGML_UNARY_OP_GELU:
                 case GGML_UNARY_OP_SILU:
-                case GGML_UNARY_OP_RELU:
                 case GGML_UNARY_OP_SIGMOID:
-                case GGML_UNARY_OP_HARDSIGMOID:
                 case GGML_UNARY_OP_HARDSWISH:
                 case GGML_UNARY_OP_GELU_QUICK:
                 case GGML_UNARY_OP_GELU_ERF:
-                case GGML_UNARY_OP_TANH:
                 case GGML_UNARY_OP_EXP:
-                case GGML_UNARY_OP_SGN:
-                case GGML_UNARY_OP_ABS:
                 case GGML_UNARY_OP_ELU:
+                    return true;
                 case GGML_UNARY_OP_FLOOR:
                 case GGML_UNARY_OP_CEIL:
                 case GGML_UNARY_OP_ROUND:

ggml/src/ggml-vulkan/vulkan-shaders/abs.comp

@@ -0,0 +1,21 @@
+#version 450
+
+#include "generic_head.glsl"
+#include "types.glsl"
+
+#extension GL_EXT_control_flow_attributes : enable
+
+layout(local_size_x = 512, local_size_y = 1, local_size_z = 1) in;
+
+layout (binding = 0) readonly buffer X {A_TYPE data_a[];};
+layout (binding = 1) writeonly buffer D {D_TYPE data_d[];};
+
+void main() {
+    const uint i = gl_GlobalInvocationID.z * 262144 + gl_GlobalInvocationID.y * 512 + gl_GlobalInvocationID.x;
+
+    if (i >= p.KX) {
+        return;
+    }
+
+    data_d[i] = D_TYPE(abs(float(data_a[i])));
+}

ggml/src/ggml-vulkan/vulkan-shaders/flash_attn.comp

@@ -7,6 +7,7 @@
 #extension GL_EXT_shader_explicit_arithmetic_types_int32 : require
 
 #extension GL_KHR_shader_subgroup_shuffle : enable
+#extension GL_KHR_shader_subgroup_vote : enable
 
 #include "types.glsl"
 #include "flash_attn_base.glsl"
@@ -108,6 +109,38 @@ void main() {
     [[dont_unroll]]
     for (uint32_t j = start_j; j < end_j; ++j) {
 
+        if ((p.mask_n_head_log2 & MASK_ENABLE_BIT) != 0) {
+            bool nem1_bounds_check = !(p.gqa_ratio > 1) && (p.nem1 % Br) != 0;
+
+            float max_mask = NEG_FLT_MAX_OVER_2;
+            [[unroll]] for (uint32_t idx = 0; idx < Bc * Br; idx += gl_WorkGroupSize.x) {
+                uint32_t c = (idx + tid) % Bc;
+                uint32_t r = (idx + tid) / Bc;
+                if (idx + tid < Bc * Br) {
+                    if ((!KV_bounds_check || j * Bc + c < KV) && (!nem1_bounds_check || i * Br + r < p.nem1)) {
+                        float m = float(data_m[m_offset + (i * Br + r) * m_stride + (j * Bc + c)]);
+                        masksh[c][r] = m;
+                        max_mask = max(max_mask, m);
+                    } else {
+                        masksh[c][r] = float(0);
+                    }
+                }
+            }
+            // skip the block if the mask is entirely -inf
+            bool all_less = subgroupAll(max_mask <= NEG_FLT_MAX_OVER_2);
+            barrier();
+            if (gl_SubgroupInvocationID == 0) {
+                tmpsh[gl_SubgroupID] = all_less ? NEG_FLT_MAX_OVER_2 : 0.0f;
+            }
+            barrier();
+            [[unroll]] for (uint s = 0; s < gl_NumSubgroups; ++s) {
+                max_mask = max(max_mask, tmpsh[s]);
+            }
+            if (max_mask <= NEG_FLT_MAX_OVER_2) {
+                continue;
+            }
+        }
+
         float Sf[Br][cols_per_thread];
         [[unroll]] for (uint32_t r = 0; r < Br; ++r) {
             [[unroll]] for (uint32_t c = 0; c < cols_per_thread; ++c) {
@@ -153,21 +186,6 @@ void main() {
         }
 
         if ((p.mask_n_head_log2 & MASK_ENABLE_BIT) != 0) {
-            bool nem1_bounds_check = !(p.gqa_ratio > 1) && (p.nem1 % Br) != 0;
-
-            [[unroll]] for (uint32_t idx = 0; idx < Bc * Br; idx += gl_WorkGroupSize.x) {
-                uint32_t c = (idx + tid) % Bc;
-                uint32_t r = (idx + tid) / Bc;
-                if (idx + tid < Bc * Br) {
-                    if ((!KV_bounds_check || j * Bc + c < KV) && (!nem1_bounds_check || i * Br + r < p.nem1)) {
-                        masksh[c][r] = float(data_m[m_offset + (i * Br + r) * m_stride + (j * Bc + c)]);
-                    } else {
-                        masksh[c][r] = float(0);
-                    }
-                }
-            }
-            barrier();
-
             [[unroll]] for (uint32_t c = 0; c < cols_per_thread; ++c) {
                 [[unroll]] for (uint32_t r = 0; r < Br; ++r) {
                     float mvf = masksh[c * cols_per_iter + col_tid][r];

ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_cm1.comp

@@ -7,6 +7,7 @@
 #extension GL_EXT_shader_explicit_arithmetic_types_int32 : require
 
 #extension GL_KHR_shader_subgroup_basic : enable
+#extension GL_KHR_shader_subgroup_vote : enable
 #extension GL_KHR_memory_scope_semantics : enable
 #extension GL_KHR_cooperative_matrix : enable
 
@@ -148,6 +149,37 @@ void main() {
     [[dont_unroll]]
     for (uint32_t j = start_j; j < end_j; ++j) {
 
+        float mask_cache[Bc * Br / WorkGroupSize];
+        if ((p.mask_n_head_log2 & MASK_ENABLE_BIT) != 0) {
+            bool nem1_bounds_check = !(p.gqa_ratio > 1) && (p.nem1 % Br) != 0;
+
+            float max_mask = NEG_FLT_MAX_OVER_2;
+            [[unroll]] for (uint32_t idx = 0; idx < Bc * Br; idx += gl_WorkGroupSize.x) {
+                uint32_t c = (idx + tid) % Bc;
+                uint32_t r = (idx + tid) / Bc;
+                if (idx + tid < Bc * Br || idx + gl_WorkGroupSize.x <= Bc * Br) {
+                    if ((!KV_bounds_check || j * Bc + c < KV) && (!nem1_bounds_check || i * Br + r < p.nem1)) {
+                        float m = float(data_m[m_offset + (i * Br + r) * m_stride + (j * Bc + c)]);
+                        mask_cache[idx / WorkGroupSize] = m;
+                        max_mask = max(max_mask, m);
+                    }
+                }
+            }
+            // skip the block if the mask is entirely -inf
+            bool all_less = subgroupAll(max_mask <= NEG_FLT_MAX_OVER_2);
+            barrier();
+            if (gl_SubgroupInvocationID == 0) {
+                tmpsh[gl_SubgroupID] = all_less ? NEG_FLT_MAX_OVER_2 : 0.0f;
+            }
+            barrier();
+            [[unroll]] for (uint s = 0; s < gl_NumSubgroups; ++s) {
+                max_mask = max(max_mask, tmpsh[s]);
+            }
+            if (max_mask <= NEG_FLT_MAX_OVER_2) {
+                continue;
+            }
+        }
+
         [[unroll]] for (uint32_t idx = 0; idx < Bc * HSK / 4; idx += gl_WorkGroupSize.x) {
             uint32_t d = (idx + tid) % (HSK / 4);
             uint32_t c = (idx + tid) / (HSK / 4);
@@ -208,7 +240,8 @@ void main() {
                 uint32_t r = (idx + tid) / Bc;
                 if (idx + tid < Bc * Br || idx + gl_WorkGroupSize.x <= Bc * Br) {
                     if ((!KV_bounds_check || j * Bc + c < KV) && (!nem1_bounds_check || i * Br + r < p.nem1)) {
-                        sfsh[c * sfshstride + r] += ACC_TYPE(slope[r] * float(data_m[m_offset + (i * Br + r) * m_stride + (j * Bc + c)]));
+                        float f = mask_cache[idx / WorkGroupSize];
+                        sfsh[c * sfshstride + r] += ACC_TYPE(slope[r] * f);
                     }
                 }
             }

ggml/src/ggml-vulkan/vulkan-shaders/flash_attn_cm2.comp

@@ -29,6 +29,10 @@ ACC_TYPE maxReduce(const in ACC_TYPE x, const in ACC_TYPE y) {
     return max(x, y);
 }
 
+float16_t maxReduceFp16(const in float16_t x, const in float16_t y) {
+    return max(x, y);
+}
+
 ACC_TYPE smearReduce(const in ACC_TYPE x, const in ACC_TYPE y) {
     return x;
 }
@@ -142,6 +146,44 @@ void main() {
     [[dont_unroll]]
     for (uint32_t j = start_j; j < end_j; ++j) {
 
+        coopmat<float16_t, gl_ScopeWorkgroup, Br, Bc, gl_MatrixUseAccumulator> mv;
+        if ((p.mask_n_head_log2 & MASK_ENABLE_BIT) != 0) {
+            bool nem1_bounds_check = !(p.gqa_ratio > 1) && (p.nem1 % Br) != 0;
+
+            if (nem1_bounds_check) {
+                tensorLayoutNV<2, gl_CooperativeMatrixClampModeConstantNV> tensorLayoutM = createTensorLayoutNV(2, gl_CooperativeMatrixClampModeConstantNV);
+                tensorLayoutM = setTensorLayoutDimensionNV(tensorLayoutM, p.nem1, KV);
+                tensorLayoutM = setTensorLayoutStrideNV(tensorLayoutM, m_stride, 1);
+                tensorLayoutM = setTensorLayoutClampValueNV(tensorLayoutM, 0xfc00); // -inf in float16_t
+
+                coopmat<float16_t, gl_ScopeWorkgroup, Br, Bc, gl_MatrixUseAccumulator> mv, mvmax;
+
+                coopMatLoadTensorNV(mv, data_m, m_offset, sliceTensorLayoutNV(tensorLayoutM, i * Br, Br, j * Bc, Bc));
+
+                // skip the block if the mask is entirely -inf
+                coopMatReduceNV(mvmax, mv, gl_CooperativeMatrixReduceRowAndColumnNV, maxReduceFp16);
+                if (mvmax[0] <= NEG_FLT_MAX_OVER_2) {
+                    continue;
+                }
+            } else {
+                tensorLayoutNV<2, Clamp> tensorLayoutM = createTensorLayoutNV(2, Clamp);
+                // Don't clamp against nem1 when GQA is enabled
+                uint32_t m_height = p.gqa_ratio > 1 ? ~0 : p.nem1;
+                tensorLayoutM = setTensorLayoutDimensionNV(tensorLayoutM, m_height, KV);
+                tensorLayoutM = setTensorLayoutStrideNV(tensorLayoutM, m_stride, 1);
+
+                coopmat<float16_t, gl_ScopeWorkgroup, Br, Bc, gl_MatrixUseAccumulator> mvmax;
+
+                coopMatLoadTensorNV(mv, data_m, m_offset, sliceTensorLayoutNV(tensorLayoutM, i * Br, Br, j * Bc, Bc));
+
+                // skip the block if the mask is entirely -inf
+                coopMatReduceNV(mvmax, mv, gl_CooperativeMatrixReduceRowAndColumnNV, maxReduceFp16);
+                if (mvmax[0] <= NEG_FLT_MAX_OVER_2) {
+                    continue;
+                }
+            }
+        }
+
         coopmat<ACC_TYPE, gl_ScopeWorkgroup, Br, Bc, gl_MatrixUseAccumulator> S = coopmat<ACC_TYPE, gl_ScopeWorkgroup, Br, Bc, gl_MatrixUseAccumulator>(0);
 
         coopmat<float16_t, gl_ScopeWorkgroup, HSK_pad, Bc, gl_MatrixUseB> K_T;
@@ -158,31 +200,7 @@ void main() {
         }
 
         if ((p.mask_n_head_log2 & MASK_ENABLE_BIT) != 0) {
-            bool nem1_bounds_check = !(p.gqa_ratio > 1) && (p.nem1 % Br) != 0;
-
-            if (nem1_bounds_check) {
-                tensorLayoutNV<2, gl_CooperativeMatrixClampModeConstantNV> tensorLayoutM = createTensorLayoutNV(2, gl_CooperativeMatrixClampModeConstantNV);
-                tensorLayoutM = setTensorLayoutDimensionNV(tensorLayoutM, p.nem1, KV);
-                tensorLayoutM = setTensorLayoutStrideNV(tensorLayoutM, m_stride, 1);
-
-                coopmat<float16_t, gl_ScopeWorkgroup, Br, Bc, gl_MatrixUseAccumulator> mv;
-
-                coopMatLoadTensorNV(mv, data_m, m_offset, sliceTensorLayoutNV(tensorLayoutM, i * Br, Br, j * Bc, Bc));
-
-                S += slopeMat*coopmat<ACC_TYPE, gl_ScopeWorkgroup, Br, Bc, gl_MatrixUseAccumulator>(mv);
-            } else {
-                tensorLayoutNV<2, Clamp> tensorLayoutM = createTensorLayoutNV(2, Clamp);
-                // Don't clamp against nem1 when GQA is enabled
-                uint32_t m_height = p.gqa_ratio > 1 ? ~0 : p.nem1;
-                tensorLayoutM = setTensorLayoutDimensionNV(tensorLayoutM, m_height, KV);
-                tensorLayoutM = setTensorLayoutStrideNV(tensorLayoutM, m_stride, 1);
-
-                coopmat<float16_t, gl_ScopeWorkgroup, Br, Bc, gl_MatrixUseAccumulator> mv;
-
-                coopMatLoadTensorNV(mv, data_m, m_offset, sliceTensorLayoutNV(tensorLayoutM, i * Br, Br, j * Bc, Bc));
-
-                S += slopeMat*coopmat<ACC_TYPE, gl_ScopeWorkgroup, Br, Bc, gl_MatrixUseAccumulator>(mv);
-            }
+            S += slopeMat*coopmat<ACC_TYPE, gl_ScopeWorkgroup, Br, Bc, gl_MatrixUseAccumulator>(mv);
         }
 
         // Clear padding elements to -inf, so they don't contribute to rowmax

ggml/src/ggml-vulkan/vulkan-shaders/log.comp

@@ -0,0 +1,18 @@
+#version 450
+
+#include "rte.glsl"
+#include "types.glsl"
+#include "generic_unary_head.glsl"
+
+layout(local_size_x = 512, local_size_y = 1, local_size_z = 1) in;
+
+void main() {
+    const uint idx = get_idx();
+
+    if (idx >= p.ne) {
+        return;
+    }
+
+    const float val = float(data_a[get_aoffset() + src0_idx(idx)]);
+    data_d[get_doffset() + dst_idx(idx)] = D_TYPE(log(val));
+}

ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_base.glsl

@@ -11,29 +11,7 @@
 #define EXPERT_COUNT 8
 #endif
 
-#include "types.glsl"
-
-#ifndef MMQ
-layout (binding = 0) readonly buffer A {A_TYPE data_a[];};
-#else
-layout (binding = 0) readonly buffer A {A_TYPE_PACKED16 data_a[];};
-#endif
-
-layout (binding = 1) readonly buffer B {B_TYPE data_b[];};
-#ifdef B_TYPE_VEC2
-layout (binding = 1) readonly buffer BV2 {B_TYPE_VEC2 data_b_v2[];};
-#endif
-#ifdef B_TYPE_VEC4
-layout (binding = 1) readonly buffer BV4 {B_TYPE_VEC4 data_b_v4[];};
-#endif
-
-layout (binding = 2) writeonly buffer D {D_TYPE data_d[];};
-
-layout (binding = 3) readonly buffer Bias {D_TYPE data_bias[];};
-
-#ifdef MUL_MAT_ID
-layout (binding = 4) readonly buffer IDS {int data_ids[];};
-#endif
+#include "mul_mat_vec_iface.glsl"
 
 #include "dequant_funcs.glsl"
 
@@ -48,8 +26,7 @@ layout (push_constant) uniform parameter
     uint batch_stride_b;
     uint batch_stride_d;
 
-    uint enable_bias;
-    uint enable_scale;
+    uint fusion_flags;
 
 #ifdef MUL_MAT_ID
     uint nei0;
@@ -123,17 +100,24 @@ void reduce_result(inout FLOAT_TYPE temp[NUM_COLS][NUM_ROWS], const in uint32_t
     if (tid == 0) {
         [[unroll]] for (uint j = 0; j < NUM_COLS; ++j) {
             [[unroll]] for (uint n = 0; n < num_rows; ++n) {
-                if (p.enable_bias != 0) {
 #ifdef MUL_MAT_ID
-                    temp[j][n] += FLOAT_TYPE(data_bias[expert_id*p.stride_d + first_row + n]);
-#else
-                    temp[j][n] += FLOAT_TYPE(data_bias[j*p.batch_stride_d + d_offset + first_row + n]);
-#endif
+                if ((p.fusion_flags & MAT_VEC_FUSION_FLAGS_BIAS0) != 0) {
+                    temp[j][n] += FLOAT_TYPE(data_fuse0[expert_id*p.stride_d + first_row + n]);
                 }
-#ifdef MUL_MAT_ID
-                if (p.enable_scale != 0) {
+                if ((p.fusion_flags & MAT_VEC_FUSION_FLAGS_SCALE0) != 0) {
                     const uint expert_idx = gl_GlobalInvocationID.y;
-                    temp[j][n] *= FLOAT_TYPE(data_bias[expert_idx]);
+                    temp[j][n] *= FLOAT_TYPE(data_fuse0[expert_idx]);
+                }
+                if ((p.fusion_flags & MAT_VEC_FUSION_FLAGS_SCALE1) != 0) {
+                    const uint expert_idx = gl_GlobalInvocationID.y;
+                    temp[j][n] *= FLOAT_TYPE(data_fuse1[expert_idx]);
+                }
+#else
+                if ((p.fusion_flags & MAT_VEC_FUSION_FLAGS_BIAS0) != 0) {
+                    temp[j][n] += FLOAT_TYPE(data_fuse0[j*p.batch_stride_d + d_offset + first_row + n]);
+                }
+                if ((p.fusion_flags & MAT_VEC_FUSION_FLAGS_BIAS1) != 0) {
+                    temp[j][n] += FLOAT_TYPE(data_fuse1[j*p.batch_stride_d + d_offset + first_row + n]);
                 }
 #endif
                 data_d[j*p.batch_stride_d + d_offset + first_row + n] = D_TYPE(temp[j][n]);
@@ -171,17 +155,24 @@ void reduce_result(FLOAT_TYPE temp[NUM_COLS][NUM_ROWS], const in uint32_t d_offs
                 [[unroll]] for (uint s = 0; s < gl_NumSubgroups; ++s) {
                     temp[j][n] += tmpsh[j][n][s];
                 }
-                if (p.enable_bias != 0) {
 #ifdef MUL_MAT_ID
-                    temp[j][n] += FLOAT_TYPE(data_bias[expert_id*p.stride_d + first_row + n]);
-#else
-                    temp[j][n] += FLOAT_TYPE(data_bias[j*p.batch_stride_d + d_offset + first_row + n]);
-#endif
+                if ((p.fusion_flags & MAT_VEC_FUSION_FLAGS_BIAS0) != 0) {
+                    temp[j][n] += FLOAT_TYPE(data_fuse0[expert_id*p.stride_d + first_row + n]);
                 }
-#ifdef MUL_MAT_ID
-                if (p.enable_scale != 0) {
+                if ((p.fusion_flags & MAT_VEC_FUSION_FLAGS_SCALE0) != 0) {
                     const uint expert_idx = gl_GlobalInvocationID.y;
-                    temp[j][n] *= FLOAT_TYPE(data_bias[expert_idx]);
+                    temp[j][n] *= FLOAT_TYPE(data_fuse0[expert_idx]);
+                }
+                if ((p.fusion_flags & MAT_VEC_FUSION_FLAGS_SCALE1) != 0) {
+                    const uint expert_idx = gl_GlobalInvocationID.y;
+                    temp[j][n] *= FLOAT_TYPE(data_fuse1[expert_idx]);
+                }
+#else
+                if ((p.fusion_flags & MAT_VEC_FUSION_FLAGS_BIAS0) != 0) {
+                    temp[j][n] += FLOAT_TYPE(data_fuse0[j*p.batch_stride_d + d_offset + first_row + n]);
+                }
+                if ((p.fusion_flags & MAT_VEC_FUSION_FLAGS_BIAS1) != 0) {
+                    temp[j][n] += FLOAT_TYPE(data_fuse1[j*p.batch_stride_d + d_offset + first_row + n]);
                 }
 #endif
                 data_d[j*p.batch_stride_d + d_offset + first_row + n] = D_TYPE(temp[j][n]);
@@ -209,17 +200,24 @@ void reduce_result(FLOAT_TYPE temp[NUM_COLS][NUM_ROWS], const in uint32_t d_offs
     if (tid == 0) {
         [[unroll]] for (uint j = 0; j < NUM_COLS; ++j) {
             [[unroll]] for (uint n = 0; n < num_rows; ++n) {
-                if (p.enable_bias != 0) {
 #ifdef MUL_MAT_ID
-                    tmpsh[j][n][0] += FLOAT_TYPE(data_bias[expert_id*p.stride_d + first_row + n]);
-#else
-                    tmpsh[j][n][0] += FLOAT_TYPE(data_bias[j*p.batch_stride_d + d_offset + first_row + n]);
-#endif
+                if ((p.fusion_flags & MAT_VEC_FUSION_FLAGS_BIAS0) != 0) {
+                    tmpsh[j][n][0] += FLOAT_TYPE(data_fuse0[expert_id*p.stride_d + first_row + n]);
                 }
-#ifdef MUL_MAT_ID
-                if (p.enable_scale != 0) {
+                if ((p.fusion_flags & MAT_VEC_FUSION_FLAGS_SCALE0) != 0) {
                     const uint expert_idx = gl_GlobalInvocationID.y;
-                    tmpsh[j][n][0] *= FLOAT_TYPE(data_bias[expert_idx]);
+                    tmpsh[j][n][0] *= FLOAT_TYPE(data_fuse0[expert_idx]);
+                }
+                if ((p.fusion_flags & MAT_VEC_FUSION_FLAGS_SCALE1) != 0) {
+                    const uint expert_idx = gl_GlobalInvocationID.y;
+                    tmpsh[j][n][0] *= FLOAT_TYPE(data_fuse1[expert_idx]);
+                }
+#else
+                if ((p.fusion_flags & MAT_VEC_FUSION_FLAGS_BIAS0) != 0) {
+                    tmpsh[j][n][0] += FLOAT_TYPE(data_fuse0[j*p.batch_stride_d + d_offset + first_row + n]);
+                }
+                if ((p.fusion_flags & MAT_VEC_FUSION_FLAGS_BIAS1) != 0) {
+                    tmpsh[j][n][0] += FLOAT_TYPE(data_fuse1[j*p.batch_stride_d + d_offset + first_row + n]);
                 }
 #endif
                 data_d[j*p.batch_stride_d + d_offset + first_row + n] = D_TYPE(tmpsh[j][n][0]);

ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_iface.glsl

@@ -0,0 +1,33 @@
+#include "types.glsl"
+
+#define MAT_VEC_FUSION_FLAGS_BIAS0 0x1
+#define MAT_VEC_FUSION_FLAGS_BIAS1 0x2
+#define MAT_VEC_FUSION_FLAGS_SCALE0 0x4
+#define MAT_VEC_FUSION_FLAGS_SCALE1 0x8
+
+#ifndef MMQ
+layout (binding = 0) readonly buffer A {A_TYPE data_a[];};
+#if defined(A_TYPE_VEC4)
+layout (binding = 0) readonly buffer AV4 {A_TYPE_VEC4 data_a_v4[];};
+#endif
+#else
+layout (binding = 0) readonly buffer A {A_TYPE_PACKED16 data_a[];};
+#endif
+
+layout (binding = 1) readonly buffer B {B_TYPE data_b[];};
+#ifdef B_TYPE_VEC2
+layout (binding = 1) readonly buffer BV2 {B_TYPE_VEC2 data_b_v2[];};
+#endif
+#ifdef B_TYPE_VEC4
+layout (binding = 1) readonly buffer BV4 {B_TYPE_VEC4 data_b_v4[];};
+#endif
+
+layout (binding = 2) writeonly buffer D {D_TYPE data_d[];};
+
+layout (binding = 3) readonly buffer Fuse0 {D_TYPE data_fuse0[];};
+layout (binding = 4) readonly buffer Fuse1 {D_TYPE data_fuse1[];};
+
+#ifdef MUL_MAT_ID
+layout (binding = 5) readonly buffer IDS {int data_ids[];};
+#endif
+

ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_nc.comp

@@ -8,14 +8,7 @@
 
 layout(local_size_x = BLOCK_SIZE, local_size_y = 1, local_size_z = 1) in;
 
-layout (binding = 0) readonly buffer A {A_TYPE data_a[];};
-layout (binding = 1) readonly buffer B {B_TYPE data_b[];};
-layout (binding = 2) writeonly buffer D {D_TYPE dst[];};
-
-layout (binding = 0) readonly buffer AV4 {A_TYPE_VEC4 data_a_v4[];};
-layout (binding = 1) readonly buffer BV4 {B_TYPE_VEC4 data_b_v4[];};
-
-layout (binding = 3) readonly buffer Bias {D_TYPE data_bias[];};
+#include "mul_mat_vec_iface.glsl"
 
 layout (push_constant) uniform parameter
 {
@@ -31,7 +24,7 @@ layout (push_constant) uniform parameter
     uint nb03;
     uint nb13;
     uint nb23;
-    uint enable_bias;
+    uint fusion_flags;
 } p;
 
 shared FLOAT_TYPE tmp[BLOCK_SIZE];
@@ -120,9 +113,12 @@ void main() {
     }
 
     if (tid == 0) {
-        if (p.enable_bias != 0) {
-            tmp[0] += FLOAT_TYPE(data_bias[idst]);
+        if ((p.fusion_flags & MAT_VEC_FUSION_FLAGS_BIAS0) != 0) {
+            tmp[0] += FLOAT_TYPE(data_fuse0[idst]);
         }
-        dst[idst] = tmp[0];
+        if ((p.fusion_flags & MAT_VEC_FUSION_FLAGS_BIAS1) != 0) {
+            tmp[0] += FLOAT_TYPE(data_fuse1[idst]);
+        }
+        data_d[idst] = tmp[0];
     }
 }

ggml/src/ggml-vulkan/vulkan-shaders/mul_mat_vec_p021.comp

@@ -10,14 +10,7 @@
 
 layout(local_size_x_id = 0, local_size_y = 1, local_size_z = 1) in;
 
-layout (binding = 0) readonly buffer A {A_TYPE data_a[];};
-layout (binding = 1) readonly buffer B {B_TYPE data_b[];};
-layout (binding = 2) writeonly buffer D {D_TYPE dst[];};
-
-layout (binding = 0) readonly buffer AV4 {A_TYPE_VEC4 data_a_v4[];};
-layout (binding = 1) readonly buffer BV4 {B_TYPE_VEC4 data_b_v4[];};
-
-layout (binding = 3) readonly buffer Bias {D_TYPE data_bias[];};
+#include "mul_mat_vec_iface.glsl"
 
 layout(constant_id = 0) const int BLOCK_SIZE = 32;
 // gqa_ratio is in the range [1,8]
@@ -31,7 +24,7 @@ layout (push_constant) uniform parameter
     uint nchannels_y;
     uint b_offset;
     uint d_offset;
-    uint enable_bias;
+    uint fusion_flags;
 } p;
 
 #if !USE_SUBGROUP_ADD
@@ -151,10 +144,13 @@ void main() {
         [[unroll]] for (uint c = 0; c < gqa_ratio; ++c) {
             // dst is not transposed and not permuted
             const uint idst = (channel + c)*nrows_dst + row_dst;
-            if (p.enable_bias != 0) {
-                temp[c] += FLOAT_TYPE(data_bias[idst]);
+            if ((p.fusion_flags & MAT_VEC_FUSION_FLAGS_BIAS0) != 0) {
+                temp[c] += FLOAT_TYPE(data_fuse0[idst]);
             }
-            dst[idst] = temp[c];
+            if ((p.fusion_flags & MAT_VEC_FUSION_FLAGS_BIAS1) != 0) {
+                temp[c] += FLOAT_TYPE(data_fuse1[idst]);
+            }
+            data_d[idst] = temp[c];
         }
     }
 }

ggml/src/ggml-vulkan/vulkan-shaders/mul_mmq_funcs.glsl

@@ -300,7 +300,7 @@ void block_a_to_shmem(const uint buf_ib, const uint ib, const uint iqs) {
 
     if (iqs == 0) {
         buf_a[buf_ib].dm = FLOAT_TYPE_VEC2(data_a_packed32[ib_k].dm);
-        buf_a[buf_ib].scales = unpack8(data_a_packed16[ib_k].scales[iqs_k / 8]);
+        buf_a[buf_ib].scales = unpack8(uint32_t(data_a_packed16[ib_k].scales[iqs_k / 8])).xy; // vec4 used due to #12147
     }
 }
 
@@ -345,21 +345,22 @@ void block_a_to_shmem(const uint buf_ib, const uint ib, const uint iqs) {
 
     // Repack 2x4 quants into one int
     // Add the 3rd bit instead of subtracting it to allow packing the quants
-    const i8vec2 vals00 = unpack8(int16_t((data_a_packed16[ib_k].qs[qs_idx * 2        ] >> qs_shift) & uint16_t(0x0303))) |
-                          unpack8(int16_t(((data_a_packed16[ib_k].hmask[hm_idx * 2    ] >> hm_shift) & uint16_t(0x0101)) << 2));
-    const i8vec2 vals01 = unpack8(int16_t((data_a_packed16[ib_k].qs[qs_idx * 2 + 1    ] >> qs_shift) & uint16_t(0x0303))) |
-                          unpack8(int16_t(((data_a_packed16[ib_k].hmask[hm_idx * 2 + 1] >> hm_shift) & uint16_t(0x0101)) << 2));
-    const i8vec2 vals10 = unpack8(int16_t((data_a_packed16[ib_k].qs[qs_idx * 2 + 2    ] >> qs_shift) & uint16_t(0x0303))) |
-                          unpack8(int16_t(((data_a_packed16[ib_k].hmask[hm_idx * 2 + 2] >> hm_shift) & uint16_t(0x0101)) << 2));
-    const i8vec2 vals11 = unpack8(int16_t((data_a_packed16[ib_k].qs[qs_idx * 2 + 3    ] >> qs_shift) & uint16_t(0x0303))) |
-                          unpack8(int16_t(((data_a_packed16[ib_k].hmask[hm_idx * 2 + 3] >> hm_shift) & uint16_t(0x0101)) << 2));
+    // vec4 for unpack8 used due to #12147
+    const i8vec2 vals00 = unpack8(int32_t(int16_t((data_a_packed16[ib_k].qs[qs_idx * 2        ] >> qs_shift) & uint16_t(0x0303)))).xy |
+                          unpack8(int32_t(int16_t(((data_a_packed16[ib_k].hmask[hm_idx * 2    ] >> hm_shift) & uint16_t(0x0101))) << 2)).xy;
+    const i8vec2 vals01 = unpack8(int32_t(int16_t((data_a_packed16[ib_k].qs[qs_idx * 2 + 1    ] >> qs_shift) & uint16_t(0x0303)))).xy |
+                          unpack8(int32_t(int16_t(((data_a_packed16[ib_k].hmask[hm_idx * 2 + 1] >> hm_shift) & uint16_t(0x0101))) << 2)).xy;
+    const i8vec2 vals10 = unpack8(int32_t(int16_t((data_a_packed16[ib_k].qs[qs_idx * 2 + 2    ] >> qs_shift) & uint16_t(0x0303)))).xy |
+                          unpack8(int32_t(int16_t(((data_a_packed16[ib_k].hmask[hm_idx * 2 + 2] >> hm_shift) & uint16_t(0x0101))) << 2)).xy;
+    const i8vec2 vals11 = unpack8(int32_t(int16_t((data_a_packed16[ib_k].qs[qs_idx * 2 + 3    ] >> qs_shift) & uint16_t(0x0303)))).xy |
+                          unpack8(int32_t(int16_t(((data_a_packed16[ib_k].hmask[hm_idx * 2 + 3] >> hm_shift) & uint16_t(0x0101))) << 2)).xy;
     buf_a[buf_ib].qs[iqs] = pack32(u8vec4(vals00.x, vals00.y, vals01.x, vals01.y)) |
                            (pack32(u8vec4(vals10.x, vals10.y, vals11.x, vals11.y)) << 4);
 
     if (iqs == 0) {
         const uint is = iqs_k / 4;
-        const i8vec2 scales = i8vec2(unpack8(((data_a_packed16[ib_k].scales[(is % 8      ) / 2] >> (4 * (is / 8))) & 0x0F0F) |
-                                            (((data_a_packed16[ib_k].scales[(8 + (is % 4)) / 2] >> (2 * (is / 4))) & 0x0303) << 4)));
+        const i8vec2 scales = i8vec2(unpack8(uint32_t(((data_a_packed16[ib_k].scales[(is % 8      ) / 2] >> (4 * (is / 8))) & 0x0F0F) |
+                                                     (((data_a_packed16[ib_k].scales[(8 + (is % 4)) / 2] >> (2 * (is / 4))) & 0x0303) << 4))).xy); // vec4 used due to #12147
 
         buf_a[buf_ib].d_scales = FLOAT_TYPE(data_a_packed16[ib_k].d) * FLOAT_TYPE_VEC2(scales - 32);
     }
@@ -516,15 +517,15 @@ void block_a_to_shmem(const uint buf_ib, const uint ib, const uint iqs) {
     const uint qh_idx = (iqs_k / 32) * 8 + iqs;
     const uint qh_shift = ((iqs_k % 32) / 8) * 2;
 
-    const i8vec2 vals00 = (unpack8(int16_t((data_a_packed16[ib_k].ql[ql_idx * 2    ] >> ql_shift) & uint16_t(0x0F0F))) |
-                          unpack8(int16_t(((data_a_packed16[ib_k].qh[qh_idx * 2    ] >> qh_shift) & uint16_t(0x0303)) << 4))) - int8_t(32);
-    const i8vec2 vals01 = (unpack8(int16_t((data_a_packed16[ib_k].ql[ql_idx * 2 + 1] >> ql_shift) & uint16_t(0x0F0F))) |
-                          unpack8(int16_t(((data_a_packed16[ib_k].qh[qh_idx * 2 + 1] >> qh_shift) & uint16_t(0x0303)) << 4))) - int8_t(32);
+    const i8vec2 vals00 = (unpack8(int32_t((data_a_packed16[ib_k].ql[ql_idx * 2    ] >> ql_shift) & uint16_t(0x0F0F))).xy |
+                          unpack8(int32_t(((data_a_packed16[ib_k].qh[qh_idx * 2    ] >> qh_shift) & uint16_t(0x0303)) << 4)).xy) - int8_t(32);
+    const i8vec2 vals01 = (unpack8(int32_t((data_a_packed16[ib_k].ql[ql_idx * 2 + 1] >> ql_shift) & uint16_t(0x0F0F))).xy |
+                          unpack8(int32_t(((data_a_packed16[ib_k].qh[qh_idx * 2 + 1] >> qh_shift) & uint16_t(0x0303)) << 4)).xy) - int8_t(32);
     buf_a[buf_ib].qs[iqs] = pack32(i8vec4(vals00.x, vals00.y, vals01.x, vals01.y));
 
     if (iqs == 0) {
         const uint is = iqs_k / 4;
-        const i8vec2 scales = unpack8(data_a_packed16[ib_k].scales[is / 2]);
+        const i8vec2 scales = unpack8(int32_t(data_a_packed16[ib_k].scales[is / 2])).xy;
 
         buf_a[buf_ib].d_scales = FLOAT_TYPE(data_a_packed16[ib_k].d) * FLOAT_TYPE_VEC2(scales);
     }

ggml/src/ggml-vulkan/vulkan-shaders/neg.comp

@@ -0,0 +1,20 @@
+#version 450
+
+#include "generic_head.glsl"
+#include "types.glsl"
+
+#extension GL_EXT_control_flow_attributes : enable
+
+layout(local_size_x = 512, local_size_y = 1, local_size_z = 1) in;
+
+layout (binding = 0) readonly buffer X {A_TYPE data_a[];};
+layout (binding = 1) writeonly buffer D {D_TYPE data_d[];};
+
+void main() {
+    const uint i = gl_GlobalInvocationID.z * 262144 + gl_GlobalInvocationID.y * 512 + gl_GlobalInvocationID.x;
+
+    if (i >= p.KX) {
+        return;
+    }
+    data_d[i] = D_TYPE(-float(data_a[i]));
+}

ggml/src/ggml-vulkan/vulkan-shaders/vulkan-shaders-gen.cpp

@@ -76,7 +76,7 @@ enum MatMulIdType {
 
 namespace {
 
-void execute_command(const std::string& command, std::string& stdout_str, std::string& stderr_str) {
+void execute_command(std::vector<std::string>& command, std::string& stdout_str, std::string& stderr_str) {
 #ifdef _WIN32
     HANDLE stdout_read, stdout_write;
     HANDLE stderr_read, stderr_write;
@@ -99,8 +99,10 @@ void execute_command(const std::string& command, std::string& stdout_str, std::s
     si.hStdOutput = stdout_write;
     si.hStdError = stderr_write;
 
-    std::vector<char> cmd(command.begin(), command.end());
-    cmd.push_back('\0');
+    std::string cmd;
+    for (const auto& part : command) {
+        cmd += part + " ";
+    }
 
     if (!CreateProcessA(NULL, cmd.data(), NULL, NULL, TRUE, 0, NULL, NULL, &si, &pi)) {
         throw std::runtime_error("Failed to create process");
@@ -138,6 +140,12 @@ void execute_command(const std::string& command, std::string& stdout_str, std::s
         throw std::runtime_error("Failed to fork process");
     }
 
+    std::vector<char*> argv;
+    for (std::string& part : command) {
+        argv.push_back(part.data());
+    }
+    argv.push_back(nullptr);
+
     if (pid == 0) {
         close(stdout_pipe[0]);
         close(stderr_pipe[0]);
@@ -145,7 +153,7 @@ void execute_command(const std::string& command, std::string& stdout_str, std::s
         dup2(stderr_pipe[1], STDERR_FILENO);
         close(stdout_pipe[1]);
         close(stderr_pipe[1]);
-        execl("/bin/sh", "sh", "-c", command.c_str(), (char*) nullptr);
+        execvp(argv[0], argv.data());
         _exit(EXIT_FAILURE);
     } else {
         close(stdout_pipe[1]);
@@ -316,21 +324,27 @@ compile_count_guard acquire_compile_slot() {
 void string_to_spv_func(std::string name, std::string in_path, std::string out_path, std::map<std::string, std::string> defines, bool coopmat, bool dep_file, compile_count_guard slot) {
     std::string target_env = (name.find("_cm2") != std::string::npos) ? "--target-env=vulkan1.3" : "--target-env=vulkan1.2";
 
+    #ifdef _WIN32
+        std::vector<std::string> cmd = {GLSLC, "-fshader-stage=compute", target_env, "\"" + in_path + "\"", "-o", "\"" + out_path + "\""};
+    #else
+        std::vector<std::string> cmd = {GLSLC, "-fshader-stage=compute", target_env, in_path, "-o", out_path};
+    #endif
+
     // disable spirv-opt for coopmat shaders for https://github.com/ggerganov/llama.cpp/issues/10734
     // disable spirv-opt for bf16 shaders for https://github.com/ggml-org/llama.cpp/issues/15344
     // disable spirv-opt for rope shaders for https://github.com/ggml-org/llama.cpp/issues/16860
-    std::string opt_level = (coopmat || name.find("bf16") != std::string::npos || name.find("rope") != std::string::npos) ? "" : "-O";
-
-    #ifdef _WIN32
-        std::vector<std::string> cmd = {GLSLC, "-fshader-stage=compute", target_env, opt_level, "\"" + in_path + "\"", "-o", "\"" + out_path + "\""};
-    #else
-        std::vector<std::string> cmd = {GLSLC, "-fshader-stage=compute", target_env, opt_level, in_path, "-o",  out_path};
-    #endif
+    if (!coopmat && name.find("bf16") == std::string::npos && name.find("rope") == std::string::npos) {
+        cmd.push_back("-O");
+    }
 
     if (dep_file) {
         cmd.push_back("-MD");
         cmd.push_back("-MF");
+#ifdef _WIN32
         cmd.push_back("\"" + target_cpp + ".d\"");
+#else
+        cmd.push_back(target_cpp + ".d");
+#endif
     }
 
     #ifdef GGML_VULKAN_SHADER_DEBUG_INFO
@@ -354,9 +368,13 @@ void string_to_spv_func(std::string name, std::string in_path, std::string out_p
         // }
         // std::cout << std::endl;
 
-        execute_command(command, stdout_str, stderr_str);
+        execute_command(cmd, stdout_str, stderr_str);
         if (!stderr_str.empty()) {
-            std::cerr << "cannot compile " << name << "\n\n" << command << "\n\n" << stderr_str << std::endl;
+            std::cerr << "cannot compile " << name << "\n\n";
+            for (const auto& part : cmd) {
+                std::cerr << part << " ";
+            }
+            std::cerr << "\n\n" << stderr_str << std::endl;
             return;
         }
 
@@ -430,7 +448,7 @@ void matmul_shaders(bool fp16, MatMulIdType matmul_id_type, bool coopmat, bool c
     base_dict["ACC_TYPE"     ] = f16acc ? "float16_t" : "float";
     base_dict["ACC_TYPE_VEC2"] = f16acc ? "f16vec2"   : "vec2";
     if (f16acc) {
-        base_dict["ACC_TYPE_MAX"] = "\"float16_t(65504.0)\"";
+        base_dict["ACC_TYPE_MAX"] = "float16_t(65504.0)";
     }
 
     if (coopmat) {
@@ -610,7 +628,7 @@ void process_shaders() {
         fa_base_dict["ACC_TYPE"] = f16acc ? "float16_t" : "float";
         fa_base_dict["ACC_TYPEV4"] = f16acc ? "f16vec4" : "vec4";
         if (f16acc) {
-            fa_base_dict["ACC_TYPE_MAX"] = "\"float16_t(65504.0)\"";
+            fa_base_dict["ACC_TYPE_MAX"] = "float16_t(65504.0)";
         }
 
         for (const auto& tname : type_names) {
@@ -798,6 +816,9 @@ void process_shaders() {
         std::string suffix = rte ? "_rte" : "";
         string_to_spv("exp_f16" + suffix,        "exp.comp",         {{"A_TYPE", "float16_t"},   {"D_TYPE", "float16_t"},   {"RTE16", rte ? "1" : "0"}});
         string_to_spv("exp_f32" + suffix,        "exp.comp",         {{"A_TYPE", "float"},       {"D_TYPE", "float"}    ,   {"RTE16", rte ? "1" : "0"}});
+
+        string_to_spv("log_f16" + suffix,        "log.comp",         {{"A_TYPE", "float16_t"},   {"D_TYPE", "float16_t"},   {"RTE16", rte ? "1" : "0"}});
+        string_to_spv("log_f32" + suffix,        "log.comp",         {{"A_TYPE", "float"},       {"D_TYPE", "float"},       {"RTE16", rte ? "1" : "0"}});
     }
     string_to_spv("gelu_f16",       "gelu.comp",        {{"A_TYPE", "float16_t"},   {"D_TYPE", "float16_t"}});
     string_to_spv("gelu_f32",       "gelu.comp",        {{"A_TYPE", "float"},       {"D_TYPE", "float"}});
@@ -809,6 +830,8 @@ void process_shaders() {
     string_to_spv("silu_f32",       "silu.comp",        {{"A_TYPE", "float"},       {"D_TYPE", "float"}});
     string_to_spv("relu_f16",       "relu.comp",        {{"A_TYPE", "float16_t"},   {"D_TYPE", "float16_t"}});
     string_to_spv("relu_f32",       "relu.comp",        {{"A_TYPE", "float"},       {"D_TYPE", "float"}});
+    string_to_spv("neg_f16",        "neg.comp",         {{"A_TYPE", "float16_t"},   {"D_TYPE", "float16_t"}});
+    string_to_spv("neg_f32",        "neg.comp",         {{"A_TYPE", "float"},       {"D_TYPE", "float"}});
     string_to_spv("tanh_f16",       "tanh.comp",        {{"A_TYPE", "float16_t"},   {"D_TYPE", "float16_t"}});
     string_to_spv("tanh_f32",       "tanh.comp",        {{"A_TYPE", "float"},       {"D_TYPE", "float"}});
     string_to_spv("sigmoid_f16",    "sigmoid.comp",     {{"A_TYPE", "float16_t"},   {"D_TYPE", "float16_t"}});
@@ -817,6 +840,8 @@ void process_shaders() {
     string_to_spv("hardsigmoid_f32","hardsigmoid.comp", {{"A_TYPE", "float"},       {"D_TYPE", "float"}});
     string_to_spv("hardswish_f16",  "hardswish.comp",   {{"A_TYPE", "float16_t"},   {"D_TYPE", "float16_t"}});
     string_to_spv("hardswish_f32",  "hardswish.comp",   {{"A_TYPE", "float"},       {"D_TYPE", "float"}});
+    string_to_spv("abs_f16",        "abs.comp",         {{"A_TYPE", "float16_t"},   {"D_TYPE", "float16_t"}});
+    string_to_spv("abs_f32",        "abs.comp",         {{"A_TYPE", "float"},       {"D_TYPE", "float"}});
 
     for (auto rte : {false, true}) {
         std::string suffix = rte ? "_rte" : "";
@@ -1081,11 +1106,6 @@ int main(int argc, char** argv) {
 
     if (args.find("--glslc") != args.end()) {
         GLSLC = args["--glslc"]; // Path to glslc
-
-        if (!std::filesystem::exists(GLSLC) || !std::filesystem::is_regular_file(GLSLC)) {
-            std::cerr << "Error: glslc not found at " << GLSLC << std::endl;
-            return EXIT_FAILURE;
-        }
     }
     if (args.find("--source") != args.end()) {
         input_filepath = args["--source"]; // The shader source file to compile

ggml/src/ggml.c

@@ -935,6 +935,7 @@ static const char * GGML_OP_NAME[GGML_OP_COUNT] = {
     "COS",
     "SUM",
     "SUM_ROWS",
+    "CUMSUM",
     "MEAN",
     "ARGMAX",
     "COUNT_EQUAL",
@@ -990,6 +991,8 @@ static const char * GGML_OP_NAME[GGML_OP_COUNT] = {
     "TIMESTEP_EMBEDDING",
     "ARGSORT",
     "LEAKY_RELU",
+    "TRI",
+    "FILL",
 
     "FLASH_ATTN_EXT",
     "FLASH_ATTN_BACK",
@@ -1002,6 +1005,7 @@ static const char * GGML_OP_NAME[GGML_OP_COUNT] = {
     "RWKV_WKV6",
     "GATED_LINEAR_ATTN",
     "RWKV_WKV7",
+    "SOLVE_TRI",
 
     "UNARY",
 
@@ -1019,7 +1023,7 @@ static const char * GGML_OP_NAME[GGML_OP_COUNT] = {
     "GLU",
 };
 
-static_assert(GGML_OP_COUNT == 90, "GGML_OP_COUNT != 90");
+static_assert(GGML_OP_COUNT == 94, "GGML_OP_COUNT != 94");
 
 static const char * GGML_OP_SYMBOL[GGML_OP_COUNT] = {
     "none",
@@ -1039,6 +1043,7 @@ static const char * GGML_OP_SYMBOL[GGML_OP_COUNT] = {
     "cos(x)",
     "Σx",
     "Σx_k",
+    "cumsum(x)",
     "Σx/n",
     "argmax(x)",
     "count_equal(x)",
@@ -1094,6 +1099,8 @@ static const char * GGML_OP_SYMBOL[GGML_OP_COUNT] = {
     "timestep_embedding(timesteps, dim, max_period)",
     "argsort(x)",
     "leaky_relu(x)",
+    "tri(x)",
+    "fill(x, c)",
 
     "flash_attn_ext(x)",
     "flash_attn_back(x)",
@@ -1106,6 +1113,7 @@ static const char * GGML_OP_SYMBOL[GGML_OP_COUNT] = {
     "rwkv_wkv6(k, v, r, tf, td, s)",
     "gated_linear_attn(k, v, q, gate, s)",
     "rwkv_wkv7(r, w, k, v, a, b, s)",
+    "A X = B, A triangular, solve X",
 
     "unary(x)",
 
@@ -1123,7 +1131,7 @@ static const char * GGML_OP_SYMBOL[GGML_OP_COUNT] = {
     "glu(x)",
 };
 
-static_assert(GGML_OP_COUNT == 90, "GGML_OP_COUNT != 90");
+static_assert(GGML_OP_COUNT == 94, "GGML_OP_COUNT != 94");
 
 static_assert(GGML_OP_POOL_COUNT == 2, "GGML_OP_POOL_COUNT != 2");
 
@@ -1142,6 +1150,8 @@ static const char * GGML_UNARY_OP_NAME[GGML_UNARY_OP_COUNT] = {
     "HARDSWISH",
     "HARDSIGMOID",
     "EXP",
+    "EXPM1",
+    "SOFTPLUS",
     "GELU_ERF",
     "XIELU",
     "FLOOR",
@@ -1150,7 +1160,7 @@ static const char * GGML_UNARY_OP_NAME[GGML_UNARY_OP_COUNT] = {
     "TRUNC",
 };
 
-static_assert(GGML_UNARY_OP_COUNT == 20, "GGML_UNARY_OP_COUNT != 20");
+static_assert(GGML_UNARY_OP_COUNT == 22, "GGML_UNARY_OP_COUNT != 22");
 
 static const char * GGML_GLU_OP_NAME[GGML_GLU_OP_COUNT] = {
     "REGLU",
@@ -2258,6 +2268,30 @@ struct ggml_tensor * ggml_log_inplace(
     return ggml_log_impl(ctx, a, true);
 }
 
+struct ggml_tensor * ggml_expm1(
+        struct ggml_context * ctx,
+        struct ggml_tensor  * a) {
+    return ggml_unary(ctx, a, GGML_UNARY_OP_EXPM1);
+}
+
+struct ggml_tensor * ggml_expm1_inplace(
+        struct ggml_context * ctx,
+        struct ggml_tensor  * a) {
+    return ggml_unary_inplace(ctx, a, GGML_UNARY_OP_EXPM1);
+}
+
+struct ggml_tensor * ggml_softplus(
+        struct ggml_context * ctx,
+        struct ggml_tensor  * a) {
+    return ggml_unary(ctx, a, GGML_UNARY_OP_SOFTPLUS);
+}
+
+struct ggml_tensor * ggml_softplus_inplace(
+        struct ggml_context * ctx,
+        struct ggml_tensor  * a) {
+    return ggml_unary_inplace(ctx, a, GGML_UNARY_OP_SOFTPLUS);
+}
+
 // ggml_sin
 
 static struct ggml_tensor * ggml_sin_impl(
@@ -2341,6 +2375,21 @@ struct ggml_tensor * ggml_sum_rows(
     return result;
 }
 
+// ggml_cumsum
+
+struct ggml_tensor * ggml_cumsum(
+        struct ggml_context * ctx,
+        struct ggml_tensor  * a) {
+    GGML_ASSERT(a->type == GGML_TYPE_F32);
+
+    struct ggml_tensor * result = ggml_dup_tensor(ctx, a);
+
+    result->op     = GGML_OP_CUMSUM;
+    result->src[0] = a;
+
+    return result;
+}
+
 // ggml_mean
 
 struct ggml_tensor * ggml_mean(
@@ -2668,8 +2717,8 @@ struct ggml_tensor * ggml_xielu(
     struct ggml_tensor * result = ggml_dup_tensor(ctx, a);
 
     ggml_set_op_params_i32(result, 0, (int32_t) GGML_UNARY_OP_XIELU);
-    ggml_set_op_params_f32(result, 1, beta + ggml_softplus(alpha_n));
-    ggml_set_op_params_f32(result, 2, ggml_softplus(alpha_p));
+    ggml_set_op_params_f32(result, 1, beta + ggml_compute_softplus_f32(alpha_n));
+    ggml_set_op_params_f32(result, 2, ggml_compute_softplus_f32(alpha_p));
     ggml_set_op_params_f32(result, 3, beta);
     ggml_set_op_params_f32(result, 4, eps);
 
@@ -5028,6 +5077,61 @@ struct ggml_tensor * ggml_timestep_embedding(
     return result;
 }
 
+// ggml_tri
+
+struct ggml_tensor * ggml_tri(
+    struct ggml_context * ctx,
+    struct ggml_tensor  * a,
+    enum ggml_tri_type    type) {
+    GGML_ASSERT(a->type == GGML_TYPE_F32);
+
+    GGML_ASSERT(ggml_is_contiguous(a));
+    GGML_ASSERT(a->ne[0] == a->ne[1]);
+
+    struct ggml_tensor * result = ggml_dup_tensor(ctx, a);
+
+    ggml_set_op_params_i32(result, 0, type);
+
+    result->op = GGML_OP_TRI;
+    result->src[0] = a;
+
+    return result;
+}
+
+// ggml_fill
+
+static struct ggml_tensor * ggml_fill_impl(
+    struct ggml_context * ctx,
+    struct ggml_tensor  * a,
+    float                 c,
+    bool                  inplace) {
+    GGML_ASSERT(a->type == GGML_TYPE_F32);
+    GGML_ASSERT(ggml_is_contiguous(a));
+
+    struct ggml_tensor * result = inplace ? ggml_view_tensor(ctx, a) : ggml_dup_tensor(ctx, a);
+
+    ggml_set_op_params_f32(result, 0, c);
+
+    result->op = GGML_OP_FILL;
+    result->src[0] = a;
+
+    return result;
+}
+
+struct ggml_tensor * ggml_fill(
+    struct ggml_context * ctx,
+    struct ggml_tensor  * a,
+    float                 c) {
+    return ggml_fill_impl(ctx, a, c, false);
+}
+
+struct ggml_tensor * ggml_fill_inplace(
+    struct ggml_context * ctx,
+    struct ggml_tensor  * a,
+    float                 c) {
+    return ggml_fill_impl(ctx, a, c, true);
+}
+
 // ggml_argsort
 
 struct ggml_tensor * ggml_argsort(
@@ -5882,6 +5986,41 @@ struct ggml_tensor * ggml_opt_step_sgd(
     return result;
 }
 
+// solve_tri
+
+struct ggml_tensor * ggml_solve_tri(
+        struct ggml_context * ctx,
+        struct ggml_tensor  * a,
+        struct ggml_tensor  * b,
+        bool                  left,
+        bool                  lower,
+        bool                  uni) {
+    GGML_ASSERT(a->type == GGML_TYPE_F32);
+    GGML_ASSERT(b->type == GGML_TYPE_F32);
+
+    // A must be square and lower diagonal
+    GGML_ASSERT(a->ne[0] == a->ne[1]);
+    // B must have same outer dimension as A
+    GGML_ASSERT(a->ne[1] == b->ne[1]);
+
+    // batch dimensions must be equal
+    GGML_ASSERT(a->ne[2] == b->ne[2]);
+    GGML_ASSERT(a->ne[3] == b->ne[3]);
+
+    GGML_ASSERT(ggml_is_contiguous(a));
+    GGML_ASSERT(ggml_is_contiguous(b));
+
+    GGML_ASSERT(lower && left && !uni); // TODO: support other variants
+
+    struct ggml_tensor * result = ggml_new_tensor_4d(ctx, GGML_TYPE_F32, b->ne[0], b->ne[1], b->ne[2], b->ne[3]);
+
+    result->op     = GGML_OP_SOLVE_TRI;
+    result->src[0] = a;
+    result->src[1] = b;
+
+    return result;
+}
+
 ////////////////////////////////////////////////////////////////////////////////
 
 struct ggml_hash_set ggml_hash_set_new(size_t size) {
@@ -6454,6 +6593,16 @@ static void ggml_compute_backward(
                         ggml_add_or_set(ctx, cgraph, isrc0, ggml_mul(ctx, tensor, grad));
                     }
                 } break;
+                case GGML_UNARY_OP_EXPM1: {
+                    if (src0_needs_grads) {
+                        ggml_add_or_set(ctx, cgraph, isrc0, ggml_mul(ctx, grad, ggml_exp(ctx, src0)));
+                    }
+                } break;
+                case GGML_UNARY_OP_SOFTPLUS: {
+                    if (src0_needs_grads) {
+                        ggml_add_or_set(ctx, cgraph, isrc0, ggml_mul(ctx, grad, ggml_sigmoid(ctx, src0)));
+                    }
+                } break;
                 default: {
                     fprintf(stderr, "%s: unsupported unary op for backward pass: %s\n",
                         __func__, ggml_unary_op_name(ggml_get_unary_op(tensor)));

gguf-py/gguf/constants.py

@@ -409,6 +409,7 @@ class MODEL_ARCH(IntEnum):
     BAILINGMOE2      = auto()
     DOTS1            = auto()
     ARCEE            = auto()
+    AFMOE            = auto()
     ERNIE4_5         = auto()
     ERNIE4_5_MOE     = auto()
     HUNYUAN_MOE      = auto()
@@ -464,6 +465,7 @@ class MODEL_TENSOR(IntEnum):
     ATTN_POST_NORM       = auto()
     ATTN_ROT_EMBD        = auto()
     ATTN_SINKS           = auto()
+    ATTN_GATE            = auto()
     FFN_GATE_INP         = auto()
     FFN_GATE_INP_SHEXP   = auto()
     FFN_NORM             = auto()
@@ -776,6 +778,7 @@ MODEL_ARCH_NAMES: dict[MODEL_ARCH, str] = {
     MODEL_ARCH.BAILINGMOE2:      "bailingmoe2",
     MODEL_ARCH.DOTS1:            "dots1",
     MODEL_ARCH.ARCEE:            "arcee",
+    MODEL_ARCH.AFMOE:            "afmoe",
     MODEL_ARCH.ERNIE4_5:         "ernie4_5",
     MODEL_ARCH.ERNIE4_5_MOE:     "ernie4_5-moe",
     MODEL_ARCH.FALCON_H1:        "falcon-h1",
@@ -828,6 +831,7 @@ TENSOR_NAMES: dict[MODEL_TENSOR, str] = {
     MODEL_TENSOR.ATTN_OUT:                  "blk.{bid}.attn_output",
     MODEL_TENSOR.ATTN_ROT_EMBD:             "blk.{bid}.attn_rot_embd",
     MODEL_TENSOR.ATTN_SINKS:                "blk.{bid}.attn_sinks",
+    MODEL_TENSOR.ATTN_GATE:                 "blk.{bid}.attn_gate",
     MODEL_TENSOR.ATTN_Q_NORM:               "blk.{bid}.attn_q_norm",
     MODEL_TENSOR.ATTN_K_NORM:               "blk.{bid}.attn_k_norm",
     MODEL_TENSOR.ATTN_OUT_NORM:             "blk.{bid}.attn_output_norm",
@@ -2693,6 +2697,33 @@ MODEL_TENSORS: dict[MODEL_ARCH, list[MODEL_TENSOR]] = {
         MODEL_TENSOR.FFN_DOWN,
         MODEL_TENSOR.FFN_UP,
     ],
+    MODEL_ARCH.AFMOE: [
+        MODEL_TENSOR.TOKEN_EMBD,
+        MODEL_TENSOR.OUTPUT_NORM,
+        MODEL_TENSOR.OUTPUT,
+        MODEL_TENSOR.ATTN_NORM,
+        MODEL_TENSOR.ATTN_POST_NORM,
+        MODEL_TENSOR.ATTN_Q,
+        MODEL_TENSOR.ATTN_K,
+        MODEL_TENSOR.ATTN_V,
+        MODEL_TENSOR.ATTN_OUT,
+        MODEL_TENSOR.ATTN_Q_NORM,
+        MODEL_TENSOR.ATTN_K_NORM,
+        MODEL_TENSOR.ATTN_GATE,
+        MODEL_TENSOR.FFN_GATE,
+        MODEL_TENSOR.FFN_DOWN,
+        MODEL_TENSOR.FFN_UP,
+        MODEL_TENSOR.FFN_GATE_INP,
+        MODEL_TENSOR.FFN_GATE_EXP,
+        MODEL_TENSOR.FFN_DOWN_EXP,
+        MODEL_TENSOR.FFN_UP_EXP,
+        MODEL_TENSOR.FFN_GATE_SHEXP,
+        MODEL_TENSOR.FFN_UP_SHEXP,
+        MODEL_TENSOR.FFN_DOWN_SHEXP,
+        MODEL_TENSOR.FFN_PRE_NORM,
+        MODEL_TENSOR.FFN_POST_NORM,
+        MODEL_TENSOR.FFN_EXP_PROBS_B,
+    ],
     MODEL_ARCH.ERNIE4_5: [
         MODEL_TENSOR.TOKEN_EMBD,
         MODEL_TENSOR.OUTPUT_NORM,

gguf-py/gguf/tensor_mapping.py

@@ -314,6 +314,10 @@ class TensorNameMap:
             "model.layers.{bid}.self_attn.sinks", # openai-moe
         ),
 
+        MODEL_TENSOR.ATTN_GATE: (
+            "model.layers.{bid}.self_attn.gate_proj", # afmoe
+        ),
+
         # Feed-forward norm
         MODEL_TENSOR.FFN_NORM: (
             "gpt_neox.layers.{bid}.post_attention_layernorm",                # gptneox
@@ -340,11 +344,12 @@ class TensorNameMap:
             "model.layers.{bid}.feedforward_layernorm",                      # apertus
         ),
 
-        # Post feed-forward norm
+        # Pre feed-forward norm
         MODEL_TENSOR.FFN_PRE_NORM: (
             "model.layers.{bid}.pre_feedforward_layernorm", # gemma2
             "layers.{bid}.pre_feedforward_layernorm",       # embeddinggemma
             "model.layers.{bid}.pre_ff_layernorm.weight",
+            "model.layers.{bid}.pre_mlp_layernorm",        # afmoe
         ),
 
         # Post feed-forward norm
@@ -370,6 +375,7 @@ class TensorNameMap:
             "model.layers.{bid}.mlp.gate.wg",                   # hunyuan
             "model.layers.{bid}.block_sparse_moe.primary_router", # smallthinker
             "model.layers.{bid}.feed_forward.gate",               # lfm2moe
+            "model.layers.{bid}.mlp.router.gate",               # afmoe
         ),
 
         MODEL_TENSOR.FFN_GATE_INP_SHEXP: (
@@ -380,6 +386,7 @@ class TensorNameMap:
             "model.layers.{bid}.mlp.gate.e_score_correction",               # deepseek-v3 dots1
             "model.layers.{bid}.mlp.moe_statics.e_score_correction",        # ernie4.5-moe
             "model.layers.{bid}.mlp.gate.expert_bias",                      # bailingmoe2
+            "model.layers.{bid}.mlp.expert_bias",                           # afmoe
             "model.layers.{bid}.feed_forward.expert_bias",                  # lfm2moe
             "model.layers.{bid}.block_sparse_moe.e_score_correction",       # minimax-m2
         ),

scripts/sync_vendor.py

@@ -12,9 +12,11 @@ vendor = {
 
     "https://raw.githubusercontent.com/nothings/stb/refs/heads/master/stb_image.h": "vendor/stb/stb_image.h",
 
-    "https://github.com/mackron/miniaudio/raw/refs/tags/0.11.22/miniaudio.h": "vendor/miniaudio/miniaudio.h",
+    # not using latest tag to avoid this issue: https://github.com/ggml-org/llama.cpp/pull/17179#discussion_r2515877926
+    # "https://github.com/mackron/miniaudio/raw/refs/tags/0.11.23/miniaudio.h": "vendor/miniaudio/miniaudio.h",
+    "https://github.com/mackron/miniaudio/raw/669ed3e844524fcd883231b13095baee9f6de304/miniaudio.h": "vendor/miniaudio/miniaudio.h",
 
-    "https://raw.githubusercontent.com/yhirose/cpp-httplib/refs/tags/v0.20.1/httplib.h": "vendor/cpp-httplib/httplib.h",
+    "https://raw.githubusercontent.com/yhirose/cpp-httplib/refs/tags/v0.27.0/httplib.h": "vendor/cpp-httplib/httplib.h",
 }
 
 for url, filename in vendor.items():

src/CMakeLists.txt

@@ -35,6 +35,7 @@ add_library(llama
             unicode-data.cpp
             unicode.cpp
             unicode.h
+            models/afmoe.cpp
             models/apertus.cpp
             models/arcee.cpp
             models/arctic.cpp

src/llama-arch.cpp

@@ -90,6 +90,7 @@ static const std::map<llm_arch, const char *> LLM_ARCH_NAMES = {
     { LLM_ARCH_BAILINGMOE2,      "bailingmoe2"      },
     { LLM_ARCH_DOTS1,            "dots1"            },
     { LLM_ARCH_ARCEE,            "arcee"            },
+    { LLM_ARCH_AFMOE,            "afmoe"            },
     { LLM_ARCH_ERNIE4_5,         "ernie4_5"         },
     { LLM_ARCH_ERNIE4_5_MOE,     "ernie4_5-moe"     },
     { LLM_ARCH_HUNYUAN_MOE,      "hunyuan-moe"      },
@@ -333,6 +334,36 @@ static const std::map<llm_arch, std::map<llm_tensor, const char *>> LLM_TENSOR_N
             { LLM_TENSOR_FFN_UP,          "blk.%d.ffn_up" },
         },
     },
+    {
+        LLM_ARCH_AFMOE,
+        {
+            { LLM_TENSOR_TOKEN_EMBD,      "token_embd" },
+            { LLM_TENSOR_OUTPUT_NORM,     "output_norm" },
+            { LLM_TENSOR_OUTPUT,          "output" },
+            { LLM_TENSOR_ATTN_NORM,       "blk.%d.attn_norm" },
+            { LLM_TENSOR_ATTN_POST_NORM,  "blk.%d.post_attention_norm" },
+            { LLM_TENSOR_ATTN_Q,          "blk.%d.attn_q" },
+            { LLM_TENSOR_ATTN_K,          "blk.%d.attn_k" },
+            { LLM_TENSOR_ATTN_V,          "blk.%d.attn_v" },
+            { LLM_TENSOR_ATTN_OUT,        "blk.%d.attn_output" },
+            { LLM_TENSOR_ATTN_Q_NORM,     "blk.%d.attn_q_norm" },
+            { LLM_TENSOR_ATTN_K_NORM,     "blk.%d.attn_k_norm" },
+            { LLM_TENSOR_ATTN_GATE,       "blk.%d.attn_gate" },
+            { LLM_TENSOR_FFN_NORM,        "blk.%d.ffn_norm" },
+            { LLM_TENSOR_FFN_POST_NORM,   "blk.%d.post_ffw_norm" },
+            { LLM_TENSOR_FFN_GATE_INP,    "blk.%d.ffn_gate_inp" },
+            { LLM_TENSOR_FFN_GATE,        "blk.%d.ffn_gate" },
+            { LLM_TENSOR_FFN_DOWN,        "blk.%d.ffn_down" },
+            { LLM_TENSOR_FFN_UP,          "blk.%d.ffn_up" },
+            { LLM_TENSOR_FFN_GATE_EXPS,   "blk.%d.ffn_gate_exps" },
+            { LLM_TENSOR_FFN_DOWN_EXPS,   "blk.%d.ffn_down_exps" },
+            { LLM_TENSOR_FFN_UP_EXPS,     "blk.%d.ffn_up_exps" },
+            { LLM_TENSOR_FFN_GATE_SHEXP,  "blk.%d.ffn_gate_shexp" },
+            { LLM_TENSOR_FFN_UP_SHEXP,    "blk.%d.ffn_up_shexp" },
+            { LLM_TENSOR_FFN_DOWN_SHEXP,  "blk.%d.ffn_down_shexp" },
+            { LLM_TENSOR_FFN_EXP_PROBS_B, "blk.%d.exp_probs_b" },
+        },
+    },
     {
         LLM_ARCH_LLAMA4,
         {
@@ -2444,6 +2475,7 @@ static const std::map<llm_tensor, llm_tensor_info> LLM_TENSOR_INFOS = {
     {LLM_TENSOR_ATTN_V,                     {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
     {LLM_TENSOR_ATTN_QKV,                   {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
     {LLM_TENSOR_ATTN_OUT,                   {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
+    {LLM_TENSOR_ATTN_GATE,                  {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
     {LLM_TENSOR_FFN_GATE,                   {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
     {LLM_TENSOR_FFN_DOWN,                   {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},
     {LLM_TENSOR_FFN_UP,                     {LLM_TENSOR_LAYER_REPEATING, GGML_OP_MUL_MAT}},

src/llama-arch.h

@@ -94,6 +94,7 @@ enum llm_arch {
     LLM_ARCH_BAILINGMOE2,
     LLM_ARCH_DOTS1,
     LLM_ARCH_ARCEE,
+    LLM_ARCH_AFMOE,
     LLM_ARCH_ERNIE4_5,
     LLM_ARCH_ERNIE4_5_MOE,
     LLM_ARCH_HUNYUAN_MOE,
@@ -312,6 +313,7 @@ enum llm_tensor {
     LLM_TENSOR_ATTN_POST_NORM,
     LLM_TENSOR_ATTN_ROT_EMBD,
     LLM_TENSOR_ATTN_SINKS,
+    LLM_TENSOR_ATTN_GATE,
     LLM_TENSOR_FFN_GATE_INP,
     LLM_TENSOR_FFN_GATE_INP_SHEXP,
     LLM_TENSOR_FFN_NORM,

src/llama-graph.cpp

@@ -1592,9 +1592,10 @@ ggml_tensor * llm_graph_context::build_attn(
             int       il) const {
     // these nodes are added to the graph together so that they are not reordered
     // by doing so, the number of splits in the graph is reduced
+    // expand k later to enable rope fusion which directly writes into k-v cache
     ggml_build_forward_expand(gf, q_cur);
-    ggml_build_forward_expand(gf, k_cur);
     ggml_build_forward_expand(gf, v_cur);
+    ggml_build_forward_expand(gf, k_cur);
 
     const auto * mctx_cur = inp->mctx;
 

src/llama-model.cpp

@@ -84,6 +84,7 @@ const char * llm_type_name(llm_type type) {
         case LLM_TYPE_15B:           return "15B";
         case LLM_TYPE_16B:           return "16B";
         case LLM_TYPE_20B:           return "20B";
+        case LLM_TYPE_26B:           return "26B";
         case LLM_TYPE_27B:           return "27B";
         case LLM_TYPE_30B:           return "30B";
         case LLM_TYPE_32B:           return "32B";
@@ -695,6 +696,37 @@ void llama_model::load_hparams(llama_model_loader & ml) {
                     default: type = LLM_TYPE_UNKNOWN;
                 }
             } break;
+        case LLM_ARCH_AFMOE:
+            {
+                ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
+                ml.get_key(LLM_KV_LEADING_DENSE_BLOCK_COUNT,   hparams.n_layer_dense_lead);
+                ml.get_key(LLM_KV_EXPERT_FEED_FORWARD_LENGTH,  hparams.n_ff_exp);
+                ml.get_key(LLM_KV_EXPERT_SHARED_COUNT,         hparams.n_expert_shared);
+                ml.get_key(LLM_KV_EXPERT_GATING_FUNC,          hparams.expert_gating_func, false);
+                ml.get_key(LLM_KV_EXPERT_WEIGHTS_SCALE,        hparams.expert_weights_scale, false);
+                ml.get_key(LLM_KV_EXPERT_WEIGHTS_NORM,         hparams.expert_weights_norm, false);
+                ml.get_key(LLM_KV_ATTENTION_SLIDING_WINDOW,    hparams.n_swa, false);
+
+                // Set up interleaved sliding window attention (ISWA)
+                // Pattern: 3 sliding - 1 full (global_attn_every_n_layers = 4)
+                if (hparams.n_swa > 0) {
+                    hparams.swa_type = LLAMA_SWA_TYPE_STANDARD;
+                    hparams.set_swa_pattern(4);
+                } else {
+                    hparams.swa_type = LLAMA_SWA_TYPE_NONE;
+                }
+
+                // Default to sigmoid if not set
+                if (hparams.expert_gating_func == LLAMA_EXPERT_GATING_FUNC_TYPE_NONE) {
+                    hparams.expert_gating_func = LLAMA_EXPERT_GATING_FUNC_TYPE_SIGMOID;
+                }
+
+                switch (hparams.n_layer) {
+                    case 56: type = LLM_TYPE_6B; break;
+                    case 32: type = LLM_TYPE_26B; break;
+                    default: type = LLM_TYPE_UNKNOWN;
+                }
+            } break;
         case LLM_ARCH_DECI:
             {
                 ml.get_key(LLM_KV_ATTENTION_LAYERNORM_RMS_EPS, hparams.f_norm_rms_eps);
@@ -5749,6 +5781,71 @@ bool llama_model::load_tensors(llama_model_loader & ml) {
                         layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd,   n_ff}, 0);
                     }
                 } break;
+            case LLM_ARCH_AFMOE:
+                {
+                    tok_embd = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, 0);
+
+                    // output
+                    output_norm = create_tensor(tn(LLM_TENSOR_OUTPUT_NORM, "weight"), {n_embd}, 0);
+                    output      = create_tensor(tn(LLM_TENSOR_OUTPUT,      "weight"), {n_embd, n_vocab}, TENSOR_NOT_REQUIRED);
+
+                    // if output is NULL, init from the input tok embed
+                    if (output == NULL) {
+                        output = create_tensor(tn(LLM_TENSOR_TOKEN_EMBD, "weight"), {n_embd, n_vocab}, TENSOR_DUPLICATED);
+                    }
+
+                    const int64_t n_ff_exp = hparams.n_ff_exp;
+                    const int64_t n_expert_shared = hparams.n_expert_shared;
+
+                    for (int i = 0; i < n_layer; ++i) {
+                        auto & layer = layers[i];
+
+                        // dual attention normalization
+                        layer.attn_norm      = create_tensor(tn(LLM_TENSOR_ATTN_NORM,      "weight", i), {n_embd}, 0);
+                        layer.attn_post_norm = create_tensor(tn(LLM_TENSOR_ATTN_POST_NORM, "weight", i), {n_embd}, 0);
+
+                        // attention projections
+                        layer.wq = create_tensor(tn(LLM_TENSOR_ATTN_Q,   "weight", i), {n_embd, n_embd_head_k * n_head}, 0);
+                        layer.wk = create_tensor(tn(LLM_TENSOR_ATTN_K,   "weight", i), {n_embd, n_embd_k_gqa}, 0);
+                        layer.wv = create_tensor(tn(LLM_TENSOR_ATTN_V,   "weight", i), {n_embd, n_embd_v_gqa}, 0);
+                        layer.wo = create_tensor(tn(LLM_TENSOR_ATTN_OUT, "weight", i), {n_embd_head_k * n_head, n_embd}, 0);
+
+                        // Q/K normalization
+                        layer.attn_q_norm = create_tensor(tn(LLM_TENSOR_ATTN_Q_NORM, "weight", i), {n_embd_head_k}, 0);
+                        layer.attn_k_norm = create_tensor(tn(LLM_TENSOR_ATTN_K_NORM, "weight", i), {n_embd_head_k}, 0);
+
+                        // attention gating
+                        layer.wqkv_gate = create_tensor(tn(LLM_TENSOR_ATTN_GATE, "weight", i), {n_embd, n_embd_head_k * n_head}, 0);
+
+                        // dual ffn normalization
+                        layer.ffn_norm      = create_tensor(tn(LLM_TENSOR_FFN_NORM,      "weight", i), {n_embd}, 0);
+                        layer.ffn_post_norm = create_tensor(tn(LLM_TENSOR_FFN_POST_NORM, "weight", i), {n_embd}, 0);
+
+                        if (static_cast<uint32_t>(i) >= hparams.n_layer_dense_lead) {
+                            // MoE layers
+                            layer.ffn_gate_inp = create_tensor(tn(LLM_TENSOR_FFN_GATE_INP, "weight", i), {n_embd, n_expert}, 0);
+                            layer.ffn_exp_probs_b = create_tensor(tn(LLM_TENSOR_FFN_EXP_PROBS_B, "bias", i), {n_expert}, 0);
+
+                            // grouped expert weights
+                            layer.ffn_gate_exps = create_tensor(tn(LLM_TENSOR_FFN_GATE_EXPS, "weight", i), {n_embd, n_ff_exp, n_expert}, 0);
+                            layer.ffn_down_exps = create_tensor(tn(LLM_TENSOR_FFN_DOWN_EXPS, "weight", i), {n_ff_exp, n_embd, n_expert}, 0);
+                            layer.ffn_up_exps   = create_tensor(tn(LLM_TENSOR_FFN_UP_EXPS,   "weight", i), {n_embd, n_ff_exp, n_expert}, 0);
+
+                            // shared expert
+                            if (n_expert_shared > 0) {
+                                const int64_t n_ff_shexp = n_ff_exp * n_expert_shared;
+                                layer.ffn_gate_shexp = create_tensor(tn(LLM_TENSOR_FFN_GATE_SHEXP, "weight", i), {n_embd, n_ff_shexp}, 0);
+                                layer.ffn_down_shexp = create_tensor(tn(LLM_TENSOR_FFN_DOWN_SHEXP, "weight", i), {n_ff_shexp, n_embd}, 0);
+                                layer.ffn_up_shexp   = create_tensor(tn(LLM_TENSOR_FFN_UP_SHEXP,   "weight", i), {n_embd, n_ff_shexp}, 0);
+                            }
+                        } else {
+                            // Dense layers
+                            layer.ffn_gate = create_tensor(tn(LLM_TENSOR_FFN_GATE, "weight", i), {n_embd, n_ff}, 0);
+                            layer.ffn_down = create_tensor(tn(LLM_TENSOR_FFN_DOWN, "weight", i), {n_ff, n_embd}, 0);
+                            layer.ffn_up   = create_tensor(tn(LLM_TENSOR_FFN_UP,   "weight", i), {n_embd, n_ff}, 0);
+                        }
+                    }
+                } break;
             case LLM_ARCH_ERNIE4_5:
             case LLM_ARCH_ERNIE4_5_MOE:
                 {
@@ -7243,6 +7340,10 @@ ggml_cgraph * llama_model::build_graph(const llm_graph_params & params) const {
             {
                 llm = std::make_unique<llm_build_arcee>(*this, params);
             } break;
+        case LLM_ARCH_AFMOE:
+            {
+                llm = std::make_unique<llm_build_afmoe>(*this, params);
+            } break;
         case LLM_ARCH_ERNIE4_5:
             {
                 llm = std::make_unique<llm_build_ernie4_5>(*this, params);
@@ -7528,6 +7629,7 @@ llama_rope_type llama_model_rope_type(const llama_model * model) {
         case LLM_ARCH_MINIMAX_M2:
         case LLM_ARCH_COGVLM:
         case LLM_ARCH_PANGU_EMBED:
+        case LLM_ARCH_AFMOE:
             return LLAMA_ROPE_TYPE_NEOX;
 
         case LLM_ARCH_QWEN2VL:

src/llama-model.h

@@ -76,6 +76,7 @@ enum llm_type {
     LLM_TYPE_15B,
     LLM_TYPE_16B,
     LLM_TYPE_20B,
+    LLM_TYPE_26B,
     LLM_TYPE_27B,
     LLM_TYPE_30B,
     LLM_TYPE_32B,
@@ -234,6 +235,7 @@ struct llama_layer {
     struct ggml_tensor * wk_enc    = nullptr;
     struct ggml_tensor * wv_enc    = nullptr;
     struct ggml_tensor * wo_enc    = nullptr;
+    struct ggml_tensor * wqkv_gate = nullptr;
 
     // attention bias
     struct ggml_tensor * bq   = nullptr;

src/llama-sampling.cpp

@@ -4,6 +4,7 @@
 #include "llama-vocab.h"
 #include "llama-grammar.h"
 
+#include <array>
 #include <algorithm>
 #include <cassert>
 #include <cfloat>
@@ -1625,10 +1626,12 @@ static struct llama_sampler * llama_sampler_init_grammar_impl(
     auto * ctx = new llama_sampler_grammar;
 
     if (grammar_str != nullptr && grammar_str[0] != '\0') {
+        std::string trigger_pattern;
+        llama_grammar * grammar = nullptr;
         // TODO: remove trigger_words support.
         if (trigger_words != nullptr && num_trigger_words > 0) {
             GGML_ASSERT(trigger_patterns == nullptr && num_trigger_patterns == 0);
-            std::string trigger_pattern("[\\s\\S]*?(");
+            trigger_pattern = "[\\s\\S]*?(";
             for (size_t i = 0; i < num_trigger_words; ++i) {
                 static const std::regex special_chars("[.^$|()*+?\\[\\]{}\\\\]");
                 if (i > 0) {
@@ -1637,15 +1640,17 @@ static struct llama_sampler * llama_sampler_init_grammar_impl(
                 trigger_pattern += std::regex_replace(trigger_words[i], special_chars, "\\$0");
             }
             trigger_pattern += ")[\\s\\S]*";
-            const auto * trigger_pattern_c = trigger_pattern.c_str();
-            trigger_patterns = &trigger_pattern_c;
-            num_trigger_patterns = 1;
+
+            std::array<const char *, 1> tmp_trigger_patterns = { trigger_pattern.c_str() };
+            grammar = llama_grammar_init_impl(vocab, grammar_str, grammar_root, lazy, tmp_trigger_patterns.data(), tmp_trigger_patterns.size(), trigger_tokens, num_trigger_tokens);
+        } else {
+            grammar = llama_grammar_init_impl(vocab, grammar_str, grammar_root, lazy, trigger_patterns, num_trigger_patterns, trigger_tokens, num_trigger_tokens);
         }
         *ctx = {
             /* .vocab        = */ vocab,
             /* .grammar_str  = */ grammar_str,
             /* .grammar_root = */ grammar_root,
-            /* .grammar      = */ llama_grammar_init_impl(vocab, grammar_str, grammar_root, lazy, trigger_patterns, num_trigger_patterns, trigger_tokens, num_trigger_tokens),
+            /* .grammar      = */ grammar,
         };
         if (!ctx->grammar) {
             delete ctx;

src/llama-vocab.cpp

@@ -443,6 +443,17 @@ struct llm_tokenizer_bpe : llm_tokenizer {
                     "(?:'[sS]|'[tT]|'[rR][eE]|'[vV][eE]|'[mM]|'[lL][lL]|'[dD])|[^\\r\\n\\p{L}\\p{N}]?\\p{L}+|\\p{N}| ?[^\\s\\p{L}\\p{N}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
                 };
                 break;
+            case LLAMA_VOCAB_PRE_TYPE_AFMOE:
+                regex_exprs = {
+                    // Digit handling - uses custom implementation in unicode.cpp
+                    // Groups digits with leading 1-2 based on total length modulo 3
+                    "\\p{AFMoE_digits}",
+                    // CJK and Asian scripts (using direct Unicode literals)
+                    "[一-鿿㐀-䶿豈-﫿぀-ゟ゠-ヿ･-ﾟ⼀-⿟เ-๿຀-໿ក-៿က-႟ꩠ-ꩿꧠ-꧿가-힯ᄀ-ᇿ]+",
+                    // Main BPE pattern
+                    "[!\"#$%&'()*+,\\-./:;<=>?@\\[\\\\\\]^_`{|}~][A-Za-z]+|[^\\r\\n\\p{L}\\p{P}\\p{S}]?[\\p{L}\\p{M}]+| ?[\\p{P}\\p{S}]+[\\r\\n]*|\\s*[\\r\\n]+|\\s+(?!\\S)|\\s+",
+                };
+                break;
             default:
                 // default regex for BPE tokenization pre-processing
                 regex_exprs = {
@@ -1013,7 +1024,7 @@ private:
         }
     private:
         uint32_t get_node(size_t index) {
-            if (index > xcda_array_size) {
+            if (index >= xcda_array_size) {
                 throw std::runtime_error("Index out of array bounds in XCDA array!");
             }
             return xcda_array[index];
@@ -1993,6 +2004,10 @@ void llama_vocab::impl::load(llama_model_loader & ml, const LLM_KV & kv) {
                 tokenizer_pre == "grok-2") {
                 pre_type = LLAMA_VOCAB_PRE_TYPE_GROK_2;
                 clean_spaces = false;
+            } else if (
+                tokenizer_pre == "afmoe") {
+                pre_type = LLAMA_VOCAB_PRE_TYPE_AFMOE;
+                clean_spaces = false;
             } else if (
                 tokenizer_pre == "minimax-m2") {
                 pre_type = LLAMA_VOCAB_PRE_TYPE_MINIMAX_M2;

src/llama-vocab.h

@@ -50,6 +50,7 @@ enum llama_vocab_pre_type {
     LLAMA_VOCAB_PRE_TYPE_GROK_2          = 39,
     LLAMA_VOCAB_PRE_TYPE_GRANITE_DOCLING = 40,
     LLAMA_VOCAB_PRE_TYPE_MINIMAX_M2      = 41,
+    LLAMA_VOCAB_PRE_TYPE_AFMOE           = 42,
 };
 
 struct LLM_KV;

src/models/afmoe.cpp

@@ -0,0 +1,187 @@
+#include "models.h"
+
+llm_build_afmoe::llm_build_afmoe(const llama_model & model, const llm_graph_params & params) : llm_graph_context(params) {
+    const int64_t n_embd_head = hparams.n_embd_head_v;
+    GGML_ASSERT(n_embd_head == hparams.n_embd_head_k);
+
+    ggml_tensor * cur;
+    ggml_tensor * inpL;
+
+    inpL = build_inp_embd(model.tok_embd);
+
+    // MuP scaling: embeddings * sqrt(hidden_size)
+    // mup_enabled = true, hidden_size = 1024, scale = 32.0
+    inpL = ggml_scale(ctx0, inpL, sqrtf(float(n_embd)));
+    cb(inpL, "inp_embd_scaled", -1);
+
+    // inp_pos - contains the positions
+    ggml_tensor * inp_pos = build_inp_pos();
+    auto * inp_attn = build_attn_inp_kv_iswa();
+    ggml_tensor * inp_out_ids = build_inp_out_ids();
+
+    const float kq_scale = 1.0f/sqrtf(float(n_embd_head));
+
+    for (int il = 0; il < n_layer; ++il) {
+        ggml_tensor * inpSA = inpL;
+
+        // dual attention normalization (pre)
+        cur = build_norm(inpL,
+                model.layers[il].attn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_norm", il);
+
+        // self-attention
+        {
+            ggml_tensor * attn_inp = cur;  // save input for gate computation
+
+            ggml_tensor * Qcur = build_lora_mm(model.layers[il].wq, cur);
+            cb(Qcur, "Qcur", il);
+
+            ggml_tensor * Kcur = build_lora_mm(model.layers[il].wk, cur);
+            cb(Kcur, "Kcur", il);
+
+            ggml_tensor * Vcur = build_lora_mm(model.layers[il].wv, cur);
+            cb(Vcur, "Vcur", il);
+
+            // compute gate from input
+            ggml_tensor * gate = build_lora_mm(model.layers[il].wqkv_gate, attn_inp);
+            cb(gate, "attn_gate_proj", il);
+
+            Qcur = ggml_reshape_3d(ctx0, Qcur, n_embd_head, n_head,    n_tokens);
+            Kcur = ggml_reshape_3d(ctx0, Kcur, n_embd_head, n_head_kv, n_tokens);
+
+            // Q/K normalization
+            Qcur = build_norm(Qcur, model.layers[il].attn_q_norm, NULL, LLM_NORM_RMS, il);
+            Kcur = build_norm(Kcur, model.layers[il].attn_k_norm, NULL, LLM_NORM_RMS, il);
+            cb(Qcur, "Qcur_normed", il);
+            cb(Kcur, "Kcur_normed", il);
+
+            // RoPE only for sliding_attention layers
+            const bool use_rope = hparams.n_no_rope_layer_step > 0 &&
+                                ((il + 1) % hparams.n_no_rope_layer_step) != 0;
+            if (use_rope) {
+                Qcur = ggml_rope_ext(
+                        ctx0, Qcur, inp_pos, nullptr,
+                        n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                        ext_factor, attn_factor, beta_fast, beta_slow);
+                cb(Qcur, "Qcur_rope", il);
+
+                Kcur = ggml_rope_ext(
+                        ctx0, Kcur, inp_pos, nullptr,
+                        n_rot, rope_type, n_ctx_orig, freq_base, freq_scale,
+                        ext_factor, attn_factor, beta_fast, beta_slow);
+                cb(Kcur, "Kcur_rope", il);
+            }
+
+            Vcur = ggml_reshape_3d(ctx0, Vcur, n_embd_head, n_head_kv, n_tokens);
+
+            cur = build_attn(inp_attn,
+                    NULL, NULL,  // wo will be applied after gating
+                    Qcur, Kcur, Vcur, nullptr, nullptr, nullptr, kq_scale, il);
+            cb(cur, "attn_out", il);
+
+            // attention gating: attn_out * sigmoid(gate) BEFORE o_proj
+            gate = ggml_sigmoid(ctx0, gate);
+            cb(gate, "attn_gate_sig", il);
+            cur = ggml_mul(ctx0, cur, gate);
+            cb(cur, "attn_gated", il);
+
+            // now apply output projection
+            cur = build_lora_mm(model.layers[il].wo, cur);
+            cb(cur, "attn_o_proj", il);
+        }
+
+        // dual attention normalization (post)
+        cur = build_norm(cur,
+                model.layers[il].attn_post_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "attn_post_norm", il);
+
+        if (il == n_layer - 1 && inp_out_ids) {
+            cur   = ggml_get_rows(ctx0,   cur, inp_out_ids);
+            inpSA = ggml_get_rows(ctx0, inpSA, inp_out_ids);
+        }
+
+        ggml_tensor * ffn_inp = ggml_add(ctx0, cur, inpSA);
+        cb(ffn_inp, "ffn_inp", il);
+
+        // dual ffn normalization (pre)
+        cur = build_norm(ffn_inp,
+                model.layers[il].ffn_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "ffn_norm", il);
+
+        // MoE or dense FFN
+        if ((uint32_t)il >= hparams.n_layer_dense_lead) {
+            // MoE layer with sigmoid routing, normalization, and scaling
+            ggml_tensor * moe_out = build_moe_ffn(cur,
+                    model.layers[il].ffn_gate_inp,
+                    model.layers[il].ffn_up_exps,
+                    model.layers[il].ffn_gate_exps,
+                    model.layers[il].ffn_down_exps,
+                    model.layers[il].ffn_exp_probs_b,
+                    n_expert, n_expert_used,
+                    LLM_FFN_SILU,
+                    hparams.expert_weights_norm,           // norm_w (route_norm=True)
+                    hparams.expert_weights_scale,          // scale_w
+                    hparams.expert_weights_scale,          // w_scale (route_scale=2.826)
+                    (llama_expert_gating_func_type) hparams.expert_gating_func,
+                    il);
+            cb(moe_out, "ffn_moe_out", il);
+
+            // shared expert
+            if (hparams.n_expert_shared > 0) {
+                ggml_tensor * ffn_shexp = build_ffn(cur,
+                        model.layers[il].ffn_up_shexp,   NULL, NULL,
+                        model.layers[il].ffn_gate_shexp, NULL, NULL,
+                        model.layers[il].ffn_down_shexp, NULL, NULL,
+                        NULL,
+                        LLM_FFN_SILU, LLM_FFN_PAR, il);
+                cb(ffn_shexp, "ffn_shexp", il);
+
+                cur = ggml_add(ctx0, moe_out, ffn_shexp);
+                cb(cur, "ffn_out", il);
+            } else {
+                cur = moe_out;
+            }
+        } else {
+            // dense layer
+            cur = build_ffn(cur,
+                    model.layers[il].ffn_up,   NULL, NULL,
+                    model.layers[il].ffn_gate, NULL, NULL,
+                    model.layers[il].ffn_down, NULL, NULL,
+                    NULL,
+                    LLM_FFN_SILU, LLM_FFN_PAR, il);
+            cb(cur, "ffn_out", il);
+        }
+
+        // dual ffn normalization (post)
+        cur = build_norm(cur,
+                model.layers[il].ffn_post_norm, NULL,
+                LLM_NORM_RMS, il);
+        cb(cur, "ffn_post_norm", il);
+
+        cur = ggml_add(ctx0, cur, ffn_inp);
+        cur = build_cvec(cur, il);
+        cb(cur, "l_out", il);
+
+        // input for next layer
+        inpL = cur;
+    }
+
+    cur = inpL;
+
+    cur = build_norm(cur,
+            model.output_norm, NULL,
+            LLM_NORM_RMS, -1);
+    cb(cur, "result_norm", -1);
+
+    res->t_embd = cur;
+
+    // lm_head
+    cur = build_lora_mm(model.output, cur);
+    cb(cur, "result_output", -1);
+    res->t_logits = cur;
+
+    ggml_build_forward_expand(gf, cur);
+}