[ET-VK] Bump Vulkan API requirement to 1.1 and enable 16 bit and 8 bit types in buffer storage

Pull Request resolved: #3058

## Context

Enable use of explicit fp16 and int8 types in GPU storage buffers via the following extensions:

* [VK_KHR_16bit_storage](https://registry.khronos.org/vulkan/specs/1.3-extensions/man/html/VK_KHR_16bit_storage.html)
* [VK_KHR_8bit_storage](https://registry.khronos.org/vulkan/specs/1.3-extensions/man/html/VK_KHR_8bit_storage.html)
* [VK_KHR_shader_float16_int8](https://registry.khronos.org/vulkan/specs/1.3-extensions/man/html/VK_KHR_shader_float16_int8.html)

The first two enables usage of 16-bit and 8-bit types in storage buffers, while the last one enables using those types in arithmetic operations.

By enabling these extensions and checking that the device supports the required features, explicit fp16 and int8 types can be used in compute shaders, as demonstrated by the added test.

Vulkan 1.1 is required in order to access `vkGetPhysicalDeviceFeatures2`, which is required to query whether the device support 16bit and 8bit types. This should be a fairly straightforward version bump as Vulkan 1.1 is supported by the vast majority of Android devices.
ghstack-source-id: 222727208
@exported-using-ghexport

Differential Revision: [D56164239](https://our.internmc.facebook.com/intern/diff/D56164239/)

Author: SS-JIA

backends/vulkan/runtime/api/Adapter.cpp

@@ -6,6 +6,8 @@
  * LICENSE file in the root directory of this source tree.
  */
 
+// @lint-ignore-every CLANGTIDY clang-diagnostic-missing-field-initializers
+
 #include <executorch/backends/vulkan/runtime/api/Adapter.h>
 
 #include <bitset>
@@ -21,15 +23,33 @@ PhysicalDevice::PhysicalDevice(VkPhysicalDevice physical_device_handle)
     : handle(physical_device_handle),
       properties{},
       memory_properties{},
+      shader_16bit_storage{
+          VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_16BIT_STORAGE_FEATURES},
+      shader_8bit_storage{
+          VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_8BIT_STORAGE_FEATURES},
+      shader_float16_int8_types{
+          VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_SHADER_FLOAT16_INT8_FEATURES_KHR},
       queue_families{},
       num_compute_queues(0),
       has_unified_memory(false),
       has_timestamps(properties.limits.timestampComputeAndGraphics),
-      timestamp_period(properties.limits.timestampPeriod) {
+      timestamp_period(properties.limits.timestampPeriod),
+      extension_features(&shader_16bit_storage) {
   // Extract physical device properties
   vkGetPhysicalDeviceProperties(handle, &properties);
   vkGetPhysicalDeviceMemoryProperties(handle, &memory_properties);
 
+  VkPhysicalDeviceFeatures2 features2{
+      VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_FEATURES_2};
+
+  // Create linked list to query availability of extensions
+  features2.pNext = &shader_16bit_storage;
+  shader_16bit_storage.pNext = &shader_8bit_storage;
+  shader_8bit_storage.pNext = &shader_float16_int8_types;
+  shader_float16_int8_types.pNext = nullptr;
+
+  vkGetPhysicalDeviceFeatures2(handle, &features2);
+
   // Check if there are any memory types have both the HOST_VISIBLE and the
   // DEVICE_LOCAL property flags
   const VkMemoryPropertyFlags unified_memory_flags =
@@ -140,6 +160,9 @@ VkDevice create_logical_device(
 #ifdef VK_KHR_portability_subset
       VK_KHR_PORTABILITY_SUBSET_EXTENSION_NAME,
 #endif /* VK_KHR_portability_subset */
+      VK_KHR_16BIT_STORAGE_EXTENSION_NAME,
+      VK_KHR_8BIT_STORAGE_EXTENSION_NAME,
+      VK_KHR_SHADER_FLOAT16_INT8_EXTENSION_NAME,
   };
 
   std::vector<const char*> enabled_device_extensions;
@@ -148,7 +171,7 @@ VkDevice create_logical_device(
       enabled_device_extensions,
       requested_device_extensions);
 
-  const VkDeviceCreateInfo device_create_info{
+  VkDeviceCreateInfo device_create_info{
       VK_STRUCTURE_TYPE_DEVICE_CREATE_INFO, // sType
       nullptr, // pNext
       0u, // flags
@@ -162,6 +185,8 @@ VkDevice create_logical_device(
       nullptr, // pEnabledFeatures
   };
 
+  device_create_info.pNext = physical_device.extension_features;
+
   VkDevice handle = nullptr;
   VK_CHECK(vkCreateDevice(
       physical_device.handle, &device_create_info, nullptr, &handle));
@@ -371,33 +396,53 @@ std::string Adapter::stringize() const {
   ss << "    deviceType:    " << device_type << std::endl;
   ss << "    deviceName:    " << properties.deviceName << std::endl;
 
-#define PRINT_LIMIT_PROP(name)                                         \
-  ss << "      " << std::left << std::setw(36) << #name << limits.name \
+#define PRINT_PROP(struct, name)                                       \
+  ss << "      " << std::left << std::setw(36) << #name << struct.name \
      << std::endl;
 
-#define PRINT_LIMIT_PROP_VEC3(name)                                       \
-  ss << "      " << std::left << std::setw(36) << #name << limits.name[0] \
-     << "," << limits.name[1] << "," << limits.name[2] << std::endl;
+#define PRINT_PROP_VEC3(struct, name)                                     \
+  ss << "      " << std::left << std::setw(36) << #name << struct.name[0] \
+     << "," << struct.name[1] << "," << struct.name[2] << std::endl;
 
   ss << "    Physical Device Limits {" << std::endl;
-  PRINT_LIMIT_PROP(maxImageDimension1D);
-  PRINT_LIMIT_PROP(maxImageDimension2D);
-  PRINT_LIMIT_PROP(maxImageDimension3D);
-  PRINT_LIMIT_PROP(maxTexelBufferElements);
-  PRINT_LIMIT_PROP(maxPushConstantsSize);
-  PRINT_LIMIT_PROP(maxMemoryAllocationCount);
-  PRINT_LIMIT_PROP(maxSamplerAllocationCount);
-  PRINT_LIMIT_PROP(maxComputeSharedMemorySize);
-  PRINT_LIMIT_PROP_VEC3(maxComputeWorkGroupCount);
-  PRINT_LIMIT_PROP(maxComputeWorkGroupInvocations);
-  PRINT_LIMIT_PROP_VEC3(maxComputeWorkGroupSize);
+  PRINT_PROP(limits, maxImageDimension1D);
+  PRINT_PROP(limits, maxImageDimension2D);
+  PRINT_PROP(limits, maxImageDimension3D);
+  PRINT_PROP(limits, maxTexelBufferElements);
+  PRINT_PROP(limits, maxPushConstantsSize);
+  PRINT_PROP(limits, maxMemoryAllocationCount);
+  PRINT_PROP(limits, maxSamplerAllocationCount);
+  PRINT_PROP(limits, maxComputeSharedMemorySize);
+  PRINT_PROP_VEC3(limits, maxComputeWorkGroupCount);
+  PRINT_PROP(limits, maxComputeWorkGroupInvocations);
+  PRINT_PROP_VEC3(limits, maxComputeWorkGroupSize);
+  ss << "    }" << std::endl;
+
+  ss << "    16bit Storage Features {" << std::endl;
+  PRINT_PROP(physical_device_.shader_16bit_storage, storageBuffer16BitAccess);
+  PRINT_PROP(
+      physical_device_.shader_16bit_storage,
+      uniformAndStorageBuffer16BitAccess);
+  PRINT_PROP(physical_device_.shader_16bit_storage, storagePushConstant16);
+  PRINT_PROP(physical_device_.shader_16bit_storage, storageInputOutput16);
+  ss << "    }" << std::endl;
+
+  ss << "    8bit Storage Features {" << std::endl;
+  PRINT_PROP(physical_device_.shader_8bit_storage, storageBuffer8BitAccess);
+  PRINT_PROP(
+      physical_device_.shader_8bit_storage, uniformAndStorageBuffer8BitAccess);
+  PRINT_PROP(physical_device_.shader_8bit_storage, storagePushConstant8);
+  ss << "    }" << std::endl;
+
+  ss << "    Shader 16bit and 8bit Features {" << std::endl;
+  PRINT_PROP(physical_device_.shader_float16_int8_types, shaderFloat16);
+  PRINT_PROP(physical_device_.shader_float16_int8_types, shaderInt8);
   ss << "    }" << std::endl;
-  ss << "  }" << std::endl;
-  ;
 
   const VkPhysicalDeviceMemoryProperties& mem_props =
       physical_device_.memory_properties;
 
+  ss << "  }" << std::endl;
   ss << "  Memory Info {" << std::endl;
   ss << "    Memory Types [" << std::endl;
   for (size_t i = 0; i < mem_props.memoryTypeCount; ++i) {
@@ -432,6 +477,9 @@ std::string Adapter::stringize() const {
   ss << "  ]" << std::endl;
   ss << "}";
 
+#undef PRINT_PROP
+#undef PRINT_PROP_VEC3
+
   return ss.str();
 }
 

backends/vulkan/runtime/api/Adapter.h

@@ -30,6 +30,12 @@ struct PhysicalDevice final {
   // Properties obtained from Vulkan
   VkPhysicalDeviceProperties properties;
   VkPhysicalDeviceMemoryProperties memory_properties;
+  // Additional features available from extensions
+  VkPhysicalDevice16BitStorageFeatures shader_16bit_storage;
+  VkPhysicalDevice8BitStorageFeatures shader_8bit_storage;
+  VkPhysicalDeviceShaderFloat16Int8Features shader_float16_int8_types;
+
+  // Available GPU queues
   std::vector<VkQueueFamilyProperties> queue_families;
 
   // Metadata
@@ -38,6 +44,9 @@ struct PhysicalDevice final {
   bool has_timestamps;
   float timestamp_period;
 
+  // Head of the linked list of extensions to be requested
+  void* extension_features{nullptr};
+
   explicit PhysicalDevice(VkPhysicalDevice);
 };
 
@@ -189,6 +198,34 @@ class Adapter final {
     return vma_;
   }
 
+  // Physical Device Features
+
+  inline bool has_16bit_storage() {
+    return physical_device_.shader_16bit_storage.storageBuffer16BitAccess ==
+        VK_TRUE;
+  }
+
+  inline bool has_8bit_storage() {
+    return physical_device_.shader_8bit_storage.storageBuffer8BitAccess ==
+        VK_TRUE;
+  }
+
+  inline bool has_16bit_compute() {
+    return physical_device_.shader_float16_int8_types.shaderFloat16 == VK_TRUE;
+  }
+
+  inline bool has_8bit_compute() {
+    return physical_device_.shader_float16_int8_types.shaderInt8 == VK_TRUE;
+  }
+
+  inline bool has_full_float16_buffers_support() {
+    return has_16bit_storage() && has_16bit_compute();
+  }
+
+  inline bool has_full_int8_buffers_support() {
+    return has_8bit_storage() && has_8bit_compute();
+  }
+
   // Command Buffer Submission
 
   void

backends/vulkan/runtime/api/Runtime.cpp

@@ -85,7 +85,7 @@ VkInstance create_instance(const RuntimeConfiguration& config) {
       0, // applicationVersion
       nullptr, // pEngineName
       0, // engineVersion
-      VK_API_VERSION_1_0, // apiVersion
+      VK_API_VERSION_1_1, // apiVersion
   };
 
   std::vector<const char*> enabled_layers;

backends/vulkan/runtime/api/Tensor.cpp

@@ -228,7 +228,14 @@ vTensor::vTensor(
           memory_layout_,
           gpu_sizes_,
           dtype_,
-          allocate_memory)) {}
+          allocate_memory)) {
+  if (dtype == api::kHalf) {
+    VK_CHECK_COND(
+        api::context()->adapter_ptr()->has_16bit_storage(),
+        "Half dtype is only available if the physical device supports float16 "
+        "storage buffers!");
+  }
+}
 
 vTensor::vTensor(
     api::Context* const context,

backends/vulkan/runtime/api/Types.h

@@ -23,15 +23,15 @@
 #define VK_FORMAT_FLOAT4 VK_FORMAT_R32G32B32A32_SFLOAT
 #endif /* USE_VULKAN_FP16_INFERENCE */
 
-#define VK_FORALL_SCALAR_TYPES(_)               \
-  _(uint8_t, VK_FORMAT_R8G8B8A8_UINT, Byte)     \
-  _(int8_t, VK_FORMAT_R8G8B8A8_SINT, Char)      \
-  _(int32_t, VK_FORMAT_R32G32B32A32_SINT, Int)  \
-  _(bool, VK_FORMAT_R8G8B8A8_SINT, Bool)        \
-  _(float, VK_FORMAT_R16G16B16A16_SFLOAT, Half) \
-  _(float, VK_FORMAT_FLOAT4, Float)             \
-  _(int8_t, VK_FORMAT_R8G8B8A8_SINT, QInt8)     \
-  _(uint8_t, VK_FORMAT_R8G8B8A8_UINT, QUInt8)   \
+#define VK_FORALL_SCALAR_TYPES(_)                  \
+  _(uint8_t, VK_FORMAT_R8G8B8A8_UINT, Byte)        \
+  _(int8_t, VK_FORMAT_R8G8B8A8_SINT, Char)         \
+  _(int32_t, VK_FORMAT_R32G32B32A32_SINT, Int)     \
+  _(bool, VK_FORMAT_R8G8B8A8_SINT, Bool)           \
+  _(uint16_t, VK_FORMAT_R16G16B16A16_SFLOAT, Half) \
+  _(float, VK_FORMAT_FLOAT4, Float)                \
+  _(int8_t, VK_FORMAT_R8G8B8A8_SINT, QInt8)        \
+  _(uint8_t, VK_FORMAT_R8G8B8A8_UINT, QUInt8)      \
   _(int32_t, VK_FORMAT_R32G32B32A32_SINT, QInt32)
 
 namespace vkcompute {

backends/vulkan/runtime/api/gen_vulkan_spv.py

@@ -100,6 +100,22 @@ def get_buffer_scalar_type(dtype: str) -> str:
     return dtype
 
 
+def get_buffer_gvec_type(dtype: str, n: int) -> str:
+    if n == 1:
+        return get_buffer_scalar_type(dtype)
+
+    if dtype == "float":
+        return f"vec{n}"
+    elif dtype == "half":
+        return f"f16vec{n}"
+    elif dtype == "int8":
+        return f"i8vec{n}"
+    elif dtype == "uint8":
+        return f"u8vec{n}"
+
+    raise AssertionError(f"Invalid dtype: {dtype}")
+
+
 def get_texel_type(dtype: str) -> str:
     image_format = TYPE_MAPPINGS["IMAGE_FORMAT"][dtype]
     if image_format[-1] == "f":
@@ -134,6 +150,7 @@ def get_texel_component_type(dtype: str) -> str:
         2: lambda pos: f"{pos}.xy",
     },
     "buffer_scalar_type": get_buffer_scalar_type,
+    "buffer_gvec_type": get_buffer_gvec_type,
     "texel_type": get_texel_type,
     "gvec_type": get_gvec_type,
     "texel_component_type": get_texel_component_type,
@@ -456,7 +473,7 @@ def generateSPV(self, output_dir: str) -> Dict[str, str]:
                     glsl_out_path,
                     "-o",
                     spv_out_path,
-                    "--target-env=vulkan1.0",
+                    "--target-env=vulkan1.1",
                     "-Werror",
                 ] + [
                     arg

backends/vulkan/test/glsl/all_shaders.yaml

@@ -33,6 +33,17 @@ fill_texture__test:
   shader_variants:
     - NAME: fill_texture__test
 
+idx_fill_buffer:
+  parameter_names_with_default_values:
+    DTYPE: float
+  generate_variant_forall:
+    DTYPE:
+      - VALUE: float
+      - VALUE: half
+      - VALUE: int8
+  shader_variants:
+    - NAME: idx_fill_buffer
+
 idx_fill_texture:
   parameter_names_with_default_values:
     DTYPE: float

backends/vulkan/test/glsl/idx_fill_buffer.glsl

@@ -0,0 +1,48 @@
+/*
+ * Copyright (c) Meta Platforms, Inc. and affiliates.
+ * All rights reserved.
+ *
+ * This source code is licensed under the BSD-style license found in the
+ * LICENSE file in the root directory of this source tree.
+ */
+
+#version 450 core
+
+#define PRECISION ${PRECISION}
+
+#define VEC4_T ${buffer_gvec_type(DTYPE, 4)}
+
+#include "indexing_utils.h"
+
+$if DTYPE == "half":
+  #extension GL_EXT_shader_16bit_storage : require
+  #extension GL_EXT_shader_explicit_arithmetic_types_float16 : require
+$elif DTYPE == "int8":
+  #extension GL_EXT_shader_8bit_storage : require
+  #extension GL_EXT_shader_explicit_arithmetic_types_int8 : require
+$elif DTYPE == "uint8":
+  #extension GL_EXT_shader_8bit_storage : require
+  #extension GL_EXT_shader_explicit_arithmetic_types_uint8 : require
+
+layout(std430) buffer;
+
+layout(set = 0, binding = 0) buffer  PRECISION restrict writeonly Buffer {
+  VEC4_T data[];
+}
+buffer_in;
+
+layout(set = 0, binding = 1) uniform PRECISION restrict Params {
+  int len;
+}
+params;
+
+layout(local_size_x_id = 0, local_size_y_id = 1, local_size_z_id = 2) in;
+
+void main() {
+  const int i = ivec3(gl_GlobalInvocationID).x;
+
+  const int base = 4 * i;
+  if (base < params.len) {
+    buffer_in.data[i] = VEC4_T(base, base + 1, base + 2, base + 3);
+  }
+}