vulkan: add INTEL_XE1 arch enum and enable coopmat1 on Intel Xe-LPG Plus (#24404)

* vulkan: add INTEL_PRE_XE2 arch enum and enable coopmat1 on Intel Xe-LPG Plus (1/3, Xe1-ARLH)

Co-authored-by: Xia, Jie <jie.xia@intel.com>
Co-authored-by: Liu, Russell <russell.liu@intel.com>

* Address comments of bf16 and trailing whitespace

* Rename INTEL_PRE_XE2 to INTEL_XE1 and remove driver workaround

* Add Windows driver check

---------

Co-authored-by: Xia, Jie <jie.xia@intel.com>
Co-authored-by: Liu, Russell <russell.liu@intel.com>

ggml/src/ggml-vulkan/ggml-vulkan.cpp

@@ -308,6 +308,7 @@ enum vk_device_architecture {
     AMD_RDNA1,
     AMD_RDNA2,
     AMD_RDNA3,
+    INTEL_XE1,
     INTEL_XE2,
     NVIDIA_PRE_TURING,
     NVIDIA_TURING,
@@ -365,21 +366,26 @@ static vk_device_architecture get_device_architecture(const vk::PhysicalDevice&
         const std::vector<vk::ExtensionProperties> ext_props = device.enumerateDeviceExtensionProperties();
 
         bool subgroup_size_control = false;
+        bool integer_dot_product = false;
 
         for (const auto& properties : ext_props) {
             if (strcmp("VK_EXT_subgroup_size_control", properties.extensionName) == 0) {
                 subgroup_size_control = true;
+            } else if (strcmp("VK_KHR_shader_integer_dot_product", properties.extensionName) == 0) {
+                integer_dot_product = true;
             }
         }
 
-        if (!subgroup_size_control) {
+        if (!subgroup_size_control || !integer_dot_product) {
             return vk_device_architecture::OTHER;
         }
 
         vk::PhysicalDeviceProperties2 props2;
         vk::PhysicalDeviceSubgroupSizeControlPropertiesEXT subgroup_size_control_props;
+        vk::PhysicalDeviceShaderIntegerDotProductPropertiesKHR integer_dot_props;
 
         props2.pNext = &subgroup_size_control_props;
+        subgroup_size_control_props.pNext = &integer_dot_props;
         device.getProperties2(&props2);
 
         if (subgroup_size_control_props.minSubgroupSize == 16) {
@@ -388,6 +394,9 @@ static vk_device_architecture get_device_architecture(const vk::PhysicalDevice&
             // https://www.intel.com/content/www/us/en/content-details/824434/2024-intel-tech-tour-xe2-and-lunar-lake-s-gpu.html
             // https://www.intel.com/content/www/us/en/docs/oneapi/optimization-guide-gpu/2025-0/intel-xe-gpu-architecture.html
             return vk_device_architecture::INTEL_XE2;
+        } else if (subgroup_size_control_props.minSubgroupSize == 8 &&
+                 integer_dot_product && integer_dot_props.integerDotProduct4x8BitPackedSignedAccelerated) {
+            return vk_device_architecture::INTEL_XE1;
         }
     } else if (props.vendorID == VK_VENDOR_ID_NVIDIA) {
         const std::vector<vk::ExtensionProperties> ext_props = device.enumerateDeviceExtensionProperties();
@@ -3837,7 +3846,7 @@ static void ggml_vk_load_shaders(vk_device& device, vk_pipeline requested) {
             l_warptile = { 256, 128, 128, 16, subgroup_size_8, 64, 2, tm_m, tn_m, tk_m, subgroup_size_8 };
             l_warptile_mmq = l_warptile_mmq_int = { 256, 128, 128, 32, subgroup_size_8, 64, 2, tm_m, tn_m, tk_m, subgroup_size_8 };
             l_warptile_mmq_int_k = { 256, 128, 128, 32, subgroup_size_16, 64, 1, 4, 2, 1, subgroup_size_16 };
-        } else if (device->vendor_id == VK_VENDOR_ID_INTEL && device->coopmat_support && device->architecture == INTEL_XE2) {
+        } else if (device->vendor_id == VK_VENDOR_ID_INTEL && device->coopmat_support) {
             // Xe2/Xe3 with coopmat enabled - warptile performance tuning
             l_warptile = { 512, 128, 128, 16, subgroup_size_8, 32, 2, tm_m, tn_m, tk_m, subgroup_size_8 };
             l_warptile_mmq = { 512, 128, 128, 32, subgroup_size_8, 32, 2, tm_m, tn_m, tk_m, subgroup_size_8 };
@@ -6361,9 +6370,8 @@ static vk_device ggml_vk_get_device(size_t idx) {
                 break;
             case VK_VENDOR_ID_INTEL: {
                 // Current Windows driver does not expose BF16 support.
-                // We only want to use l_warptile if coopmat is available and is Xe2+
-                const bool xe2_with_coopmat = device->coopmat_support && device->architecture == INTEL_XE2;
-                const bool use_l_warptile = (i == GGML_TYPE_BF16) ? (device->coopmat_bf16_support && xe2_with_coopmat) : xe2_with_coopmat;
+                // We only want to use l_warptile if coopmat is available
+                const bool use_l_warptile = (i == GGML_TYPE_BF16) ? (device->coopmat_bf16_support && device->coopmat_support) : device->coopmat_support;
                 device->mul_mat_l[i] = use_l_warptile;
                 device->mul_mat_id_l[i] = use_l_warptile;
                 device->mul_mat_m[i] = true;
@@ -17890,9 +17898,9 @@ static bool ggml_vk_device_is_supported(const vk::PhysicalDevice & vkdev) {
 static bool ggml_vk_khr_cooperative_matrix_support(const vk::PhysicalDeviceProperties& props, const vk::PhysicalDeviceDriverProperties& driver_props, vk_device_architecture arch) {
     switch (props.vendorID) {
     case VK_VENDOR_ID_INTEL:
-        // Only allowing Xe2 GPU at the moment since Xe2 GPU can gain significant performance boost,
-        // while some older hardware (ex. Arc A770) has performance regressions
-        return arch == vk_device_architecture::INTEL_XE2;
+        // Only allowing Xe2/Xe3 GPU and integrated Xe GPUs at the moment since older hardware (ex. Arc A770) has performance regressions.
+        return (arch == vk_device_architecture::INTEL_XE2) ||
+            (arch == vk_device_architecture::INTEL_XE1 && props.deviceType == vk::PhysicalDeviceType::eIntegratedGpu && driver_props.driverID == vk::DriverId::eIntelProprietaryWindows);
     case VK_VENDOR_ID_AMD:
         if (driver_props.driverID == vk::DriverId::eAmdProprietary || driver_props.driverID == vk::DriverId::eAmdOpenSource) {
             // Workaround for AMD proprietary driver reporting support on all GPUs
@@ -17940,6 +17948,8 @@ static uint32_t ggml_vk_intel_shader_core_count(const vk::PhysicalDevice& vkdev)
     case 0xE20B:  // B580
     case 0xE211:  // Pro B60
         return 20;
+    case 0xB080:  // PTL Xe3 LPG 2x6 (12 subslices)
+        return 12;
     default:
         return 0;
     }

ggml/src/ggml-vulkan/vulkan-shaders/conv2d_mm.comp

@@ -158,7 +158,7 @@ const uint32_t Csh_stride = BS_NPQ;
 #ifdef COOPMAT
 const uint32_t Csh_len    = BS_K * Csh_stride;
 #else
-const uint32_t Csh_len    = csh_store != 0 ? BS_K * Csh_stride : 1;
+const uint32_t Csh_len    = csh_store != 0 ? BS_K * Csh_stride : 8; // 8 to workaround compiler bug
 #endif
 shared SHMEM_TYPE Csh[Csh_len];  // K x NPQ
 #endif

ggml/src/ggml-vulkan/vulkan-shaders/conv3d_mm.comp

@@ -144,7 +144,7 @@ const uint32_t Csh_stride = BS_NPQ;
 #ifdef COOPMAT
 const uint32_t Csh_len    = BS_K * Csh_stride;
 #else
-const uint32_t Csh_len    = csh_store != 0 ? BS_K * Csh_stride : 1;
+const uint32_t Csh_len    = csh_store != 0 ? BS_K * Csh_stride : 8; // 8 to workaround compiler bug
 #endif
 shared SHMEM_TYPE Csh[Csh_len];  // K x NPQ
 #endif

tests/test-backend-ops.cpp

@@ -7973,6 +7973,9 @@ static std::vector<std::unique_ptr<test_case>> make_test_cases_eval() {
             }
         }
     }
+    for (auto kernel_type : {GGML_TYPE_F32, GGML_TYPE_F16}) {
+        test_cases.emplace_back(new test_conv_2d({ 256, 256, 192, 1 }, { 3, 3, 192, 96 }, kernel_type, 1, 1, 1, 1, 1, 1, false));
+    }
 
     // sycl backend will limit task global_range < MAX_INT
     // test cases for 2D im2col with large input W and H (occurs in stable-diffusion)

tools/ui/src/lib/components/app/chat/ChatAttachments/ChatAttachmentsList/ChatAttachmentsListItem/ChatAttachmentsListItemMcpPrompt.svelte

@@ -33,7 +33,7 @@
 
 	{#if !readonly && onRemove}
 		<div
-			class="absolute top-10 right-2 flex items-center justify-center opacity-0 transition-opacity group-hover:opacity-100"
+			class="absolute top-10 right-2 flex items-center justify-center opacity-0 transition-opacity group-focus-within:opacity-100 group-hover:opacity-100"
 		>
 			<ActionIcon icon={X} tooltip="Remove" stopPropagationOnClick onclick={() => onRemove?.()} />
 		</div>

tools/ui/src/lib/components/app/chat/ChatMessages/ChatMessage/ChatMessageUser/ChatMessageUserPending.svelte

@@ -56,7 +56,7 @@
 			<div class="relative flex h-6 items-center justify-between">
 				<div class="right-0 flex items-center gap-2 opacity-100 transition-opacity">
 					<div
-						class="pointer-events-auto inset-0 flex items-center gap-1 opacity-0 transition-all duration-150 group-hover:opacity-100"
+						class="pointer-events-auto inset-0 flex items-center gap-1 opacity-0 transition-all duration-150 group-focus-within:opacity-100 group-hover:opacity-100"
 					>
 						<ActionIcon icon={Edit} tooltip="Edit" onclick={editCtx.handleEdit} />
 						<ActionIcon icon={Trash2} tooltip="Delete" onclick={onDelete} />

tools/ui/src/lib/components/app/navigation/SidebarNavigation/SidebarNavigationConversationItem.svelte

@@ -39,7 +39,6 @@
 		depth = 0
 	}: Props = $props();
 
-	let renderActionsDropdown = $state(false);
 	let dropdownOpen = $state(false);
 
 	let isLoading = $derived(getAllLoadingChats().includes(conversation.id));
@@ -71,26 +70,10 @@
 		}
 	}
 
-	function handleMouseLeave() {
-		if (!dropdownOpen) {
-			renderActionsDropdown = false;
-		}
-	}
-
-	function handleMouseOver() {
-		renderActionsDropdown = true;
-	}
-
 	function handleSelect() {
 		onSelect?.(conversation.id);
 	}
 
-	$effect(() => {
-		if (!dropdownOpen) {
-			renderActionsDropdown = false;
-		}
-	});
-
 	onMount(() => {
 		document.addEventListener('edit-active-conversation', handleGlobalEditEvent as EventListener);
 
@@ -103,23 +86,19 @@
 	});
 </script>
 
-<!-- svelte-ignore a11y_mouse_events_have_key_events -->
-<button
-	class="group flex min-h-9 w-full cursor-pointer items-center justify-between space-x-3 rounded-lg py-1.5 text-left transition-colors hover:bg-foreground/10 {isActive
+<div
+	class="conversation-item group relative flex min-h-9 w-full items-center justify-between space-x-3 rounded-lg py-1.5 transition-colors hover:bg-foreground/10 {isActive
 		? 'bg-foreground/5 text-accent-foreground'
 		: ''} px-3"
-	onclick={handleSelect}
-	onmouseover={handleMouseOver}
-	onmouseleave={handleMouseLeave}
-	onfocusin={handleMouseOver}
-	onfocusout={(e) => {
-		if (!e.currentTarget.contains(e.relatedTarget as Node | null)) {
-			handleMouseLeave();
-		}
-	}}
 >
+	<button
+		class="absolute inset-0 z-0 cursor-pointer rounded-lg focus:outline-none focus-visible:ring-2 focus-visible:ring-ring"
+		onclick={handleSelect}
+		aria-label={conversation.name}
+	>
+	</button>
 	<div
-		class="flex min-w-0 flex-1 items-center gap-2"
+		class="pointer-events-none relative z-10 flex min-w-0 flex-1 items-center gap-2"
 		style:padding-left="{depth * FORK_TREE_DEPTH_PADDING}px"
 	>
 		{#if depth > 0}
@@ -130,7 +109,7 @@
 						<a
 							{...props}
 							href={RouterService.chat(conversation.forkedFromConversationId)}
-							class="flex shrink-0 items-center text-muted-foreground transition-colors hover:text-foreground"
+							class="pointer-events-auto flex shrink-0 items-center text-muted-foreground transition-colors hover:text-foreground"
 						>
 							<GitBranch class="h-3.5 w-3.5" />
 						</a>
@@ -146,18 +125,15 @@
 		{#if isLoading}
 			<Tooltip.Root>
 				<Tooltip.Trigger>
-					<div
-						class="stop-button flex h-4 w-4 shrink-0 cursor-pointer items-center justify-center rounded text-muted-foreground transition-colors hover:text-foreground"
+					<button
+						class="stop-button pointer-events-auto flex h-4 w-4 shrink-0 cursor-pointer items-center justify-center rounded text-muted-foreground transition-colors hover:text-foreground"
 						onclick={handleStop}
-						onkeydown={(e) => e.key === 'Enter' && handleStop(e)}
-						role="button"
-						tabindex="0"
 						aria-label="Stop generation"
 					>
 						<Loader2 class="loading-icon h-3.5 w-3.5 animate-spin" />
 
 						<Square class="stop-icon hidden h-3 w-3 fill-current text-destructive" />
-					</div>
+					</button>
 				</Tooltip.Trigger>
 
 				<Tooltip.Content>
@@ -169,52 +145,50 @@
 		<TruncatedText text={conversation.name} class="text-sm font-medium" showTooltip={false} />
 	</div>
 
-	{#if renderActionsDropdown}
-		<div class="actions flex items-center">
-			<DropdownMenuActions
-				triggerIcon={MoreHorizontal}
-				triggerTooltip="More actions"
-				bind:open={dropdownOpen}
-				actions={[
-					{
-						icon: conversation.pinned ? PinOff : Pin,
-						label: conversation.pinned ? 'Unpin' : 'Pin',
-						onclick: (e: Event) => {
-							e.stopPropagation();
-							handleTogglePin();
-						}
-					},
-					{
-						icon: Pencil,
-						label: 'Edit',
-						onclick: handleEdit,
-						shortcut: ['shift', 'cmd', 'e']
-					},
-					{
-						icon: Download,
-						label: 'Export',
-						onclick: (e: Event) => {
-							e.stopPropagation();
-							conversationsStore.downloadConversation(conversation.id);
-						},
-						shortcut: ['shift', 'cmd', 's']
-					},
-					{
-						icon: Trash2,
-						label: 'Delete',
-						onclick: handleDelete,
-						variant: 'destructive',
-						shortcut: ['shift', 'cmd', 'd'],
-						separator: true
+	<div class="actions pointer-events-auto relative z-20 flex items-center">
+		<DropdownMenuActions
+			triggerIcon={MoreHorizontal}
+			triggerTooltip="More actions"
+			bind:open={dropdownOpen}
+			actions={[
+				{
+					icon: conversation.pinned ? PinOff : Pin,
+					label: conversation.pinned ? 'Unpin' : 'Pin',
+					onclick: (e: Event) => {
+						e.stopPropagation();
+						handleTogglePin();
 					}
-				]}
-			/>
-		</div>
-	{/if}
-</button>
+				},
+				{
+					icon: Pencil,
+					label: 'Edit',
+					onclick: handleEdit,
+					shortcut: ['shift', 'cmd', 'e']
+				},
+				{
+					icon: Download,
+					label: 'Export',
+					onclick: (e: Event) => {
+						e.stopPropagation();
+						conversationsStore.downloadConversation(conversation.id);
+					},
+					shortcut: ['shift', 'cmd', 's']
+				},
+				{
+					icon: Trash2,
+					label: 'Delete',
+					onclick: handleDelete,
+					variant: 'destructive',
+					shortcut: ['shift', 'cmd', 'd'],
+					separator: true
+				}
+			]}
+		/>
+	</div>
+</div>
 
 <style>
-	button {
+	.conversation-item {
 		:global([data-slot='dropdown-menu-trigger']:not([data-state='open'])) {
 			opacity: 0;
 		}
@@ -239,7 +213,8 @@
 			}
 		}
 
-		&:is(:hover) .stop-button {
+		&:is(:hover) .stop-button,
+		&:focus-within .stop-button {
 			:global(.stop-icon) {
 				display: block;
 			}