kc3-lang/angle/include/platform/mtl_features.json

• Show log

  Commit

• Hash : 9b00af52
  Author :
  Date : 2023-02-01T11:10:32
  

  Metal: Add an in-memory MTLLibrary cache.
  
  Add a small cache for (msl + compile parameters) -> MTLLibrary at the
  egl::Display level. In regular executions of Chrome, the same shaders
  (particularly vertex) are compiled multiple times in different programs.
  
  Tested for a regular Chrome startup + open wikipedia + motionmark 1.2:
  112/282 (40%) cache hits.
  
  Several different caching methods were profiled (LinkProgram perf test)
   - struct key with std::map                             : 303309
   - struct key with std::unordered_map                   : 308090
   - binary blob key with std::map                        : 263595
   - binary blob key with std::unordered_map              : 286051
   - struct key + is_transparent with std::unordered_map  : 304877
   - struct key + is_transparent with absl::flat_hash_map : 335686
  
  Using is_transparent allows us to search the hash map without
  copying the shader source string to construct the key structure.
  
  Bug: chromium:1385510
  Change-Id: Ieec4ba526fe286276a4af7114d89cde32a8f9e1d
  Reviewed-on: https://chromium-review.googlesource.com/c/angle/angle/+/4214012
  Reviewed-by: Shahbaz Youssefi <syoussefi@chromium.org>
  Commit-Queue: Geoff Lang <geofflang@chromium.org>
  Reviewed-by: Kenneth Russell <kbr@chromium.org>
  

Download

• include/platform/mtl_features.json

• {
      "description": [
          "Copyright 2022 The ANGLE Project Authors. All rights reserved.",
          "Use of this source code is governed by a BSD-style license that can be",
          "found in the LICENSE file.",
          "",
          "mtl_features.json: Optional features for the Metal renderer."
      ],
      "features": [
          {
              "name": "has_base_vertex_instanced_draw",
              "category": "Features",
              "description": [
                  "The renderer supports base vertex instanced draw"
              ]
          },
          {
              "name": "has_explicit_mem_barrier",
              "category": "Features",
              "description": [
                  "The renderer supports explicit memory barrier"
              ]
          },
          {
              "name": "has_cheap_render_pass",
              "category": "Features",
              "description": [
                  "The renderer can cheaply break a render pass."
              ]
          },
          {
              "name": "has_non_uniform_dispatch",
              "category": "Features",
              "description": [
                  "The renderer supports non uniform compute shader dispatch's group size"
              ]
          },
          {
              "name": "has_shader_stencil_output",
              "category": "Features",
              "description": [
                  "The renderer supports stencil output from fragment shader"
              ]
          },
          {
              "name": "has_texture_swizzle",
              "category": "Features",
              "description": [
                  "The renderer supports texture swizzle"
              ]
          },
          {
              "name": "has_depth_auto_resolve",
              "category": "Features",
              "description": [
                  "The renderer supports MSAA depth auto resolve at the end of render pass"
              ]
          },
          {
              "name": "has_stencil_auto_resolve",
              "category": "Features",
              "description": [
                  "The renderer supports MSAA stencil auto resolve at the end of render pass"
              ]
          },
          {
              "name": "has_events",
              "category": "Features",
              "description": [
                  "The renderer supports MTL(Shared)Event"
              ]
          },
          {
              "name": "allow_inline_const_vertex_data",
              "category": "Features",
              "description": [
                  "The renderer supports using inline constant data for small client vertex data"
              ]
          },
          {
              "name": "allow_separate_depth_stencil_buffers",
              "category": "Features",
              "description": [
                  "Some Apple platforms such as iOS allows separate depth and stencil buffers, ",
                  "whereas others such as macOS don't"
              ]
          },
          {
              "name": "allow_runtime_sampler_compare_mode",
              "category": "Features",
              "description": [
                  "The renderer supports changing sampler's compare mode outside shaders"
              ]
          },
          {
              "name": "allow_sampler_compare_gradient",
              "category": "Features",
              "description": [
                  "The renderer supports sample_compare with gradients"
              ]
          },
          {
              "name": "allow_sampler_compare_lod",
              "category": "Features",
              "description": [
                  "The renderer supports sample_compare with lod"
              ]
          },
          {
              "name": "allow_buffer_read_write",
              "category": "Features",
              "description": [
                  "The renderer supports buffer read and write in the same shader"
              ]
          },
          {
              "name": "allow_multisample_store_and_resolve",
              "category": "Features",
              "description": [
                  "The renderer supports MSAA store and resolve in the same pass"
              ]
          },
          {
              "name": "allow_gen_multiple_mips_per_pass",
              "category": "Features",
              "description": [
                  "The renderer supports generating multiple mipmaps per pass"
              ]
          },
          {
              "name": "force_D24S8_as_unsupported",
              "category": "Features",
              "description": [
                  "Force Depth24Stencil8 format as unsupported."
              ]
          },
          {
              "name": "force_buffer_GPU_storage",
              "category": "Features",
              "description": [
                  "On systems that support both buffer' memory allocation on GPU and shared memory (such as ",
                      "macOS), force using GPU memory allocation for buffers everytime or not."
              ]
          },
          {
              "name": "force_non_CS_base_mipmap_generation",
              "category": "Features",
              "description": [
                  "Turn this feature on to disallow Compute Shader based mipmap generation. Compute Shader ",
                  "based mipmap generation might cause GPU hang on some older iOS devices."
              ]
          },
          {
              "name": "emulate_transform_feedback",
              "category": "Features",
              "description": [
                  "Turn this on to allow transform feedback in Metal using a 2-pass VS for GLES3."
              ]
          },
          {
              "name": "rewrite_row_major_matrices",
              "category": "Features",
              "description": [
                  "Rewrite row major matrices in shaders as column major."
              ]
          },
          {
              "name": "intel_explicit_bool_cast_workaround",
              "category": "Workarounds",
              "description": [
                  "Insert explicit casts for float/double/unsigned/signed int on macOS 10.15 with Intel ",
                  "driver"
              ]
          },
          {
              "name": "intel_disable_fast_math",
              "category": "Workarounds",
              "description": [
                  "Disable fast math in atan and invariance cases when running below macOS 12.0"
              ]
          },
          {
              "name": "allow_renderpass_without_attachment",
              "category": "Features",
              "description": [
                  "Allow creation of render passes without any attachments"
              ]
          },
  
          {
              "name": "multisample_color_format_shader_read_workaround",
              "category": "Workarounds",
              "description": [
                  "Add shaderRead usage to some multisampled texture formats"
              ],
              "issue": "http://anglebug.com/7049"
          },
          {
              "name": "copy_IOSurface_to_non_IOSurface_for_read_optimization",
              "category": "Workarounds",
              "description": [
                  "some GPUs are faster to read an IOSurface texture by first copying the texture to a ",
                  "non-IOSurface texture"
              ],
              "issue": "http://anglebug.com/7117 http://anglebug.com/7573"
          },
          {
              "name": "copy_texture_to_buffer_for_read_optimization",
              "category": "Workarounds",
              "description": [
                  "some GPUs are faster to read a texture by first copying the texture to a buffer"
              ],
              "issue": "http://anglebug.com/7117"
          },
          {
              "name": "limit_max_draw_buffers_for_testing",
              "category": "Features",
              "description": [
                  "Used to check the backend works when the device's advertized limit is less than the code's limit"
              ],
              "issue": "http://anglebug.com/7280"
          },
          {
              "name": "limit_max_color_target_bits_for_testing",
              "category": "Features",
              "description": [
                  "Metal iOS has a limit on the number of color target bits per pixel."
              ],
              "issue": "http://anglebug.com/7280"
          },
          {
              "name": "preemptively_start_provoking_vertex_command_buffer",
              "category": "Features",
              "description": [
                  "AMD Metal Drivers appear to have a bug this works around"
              ],
              "issue": "http://anglebug.com/7635"
          },
          {
              "name": "upload_data_to_iosurfaces_with_staging_buffers",
              "category": "Workarounds",
              "description": [
                  "When uploading data to IOSurface-backed textures, use a staging buffer."
              ],
              "issue": "http://anglebug.com/7573"
          },
          {
              "name": "always_use_staged_buffer_updates",
              "category": "Features",
              "description": [
                  "Always update buffers by copying the data to a staging buffer and then blitting it to the actual buffer"
              ],
              "issue": "http://anglebug.com/7544"
          },
          {
              "name": "use_shadow_buffers_when_appropriate",
              "category": "Features",
              "description": [
                  "On some architectures using a shadow buffer can be faster for certain size buffers"
              ],
              "issue": "http://anglebug.com/7544"
          },
          {
              "name": "always_use_managed_storage_mode_for_buffers",
              "category": "Features",
              "description": [
                  "Metal buffers can be managed, shared, or private. Sometimes managed is fastest"
              ],
              "issue": "http://anglebug.com/7544"
          },
          {
              "name": "always_use_shared_storage_mode_for_buffers",
              "category": "Features",
              "description": [
                  "Metal buffers can be managed, shared, or private. Sometimes shared is fastest"
              ],
              "issue": "http://anglebug.com/7544"
          },
          {
              "name": "prefer_cpu_for_buffersubdata",
              "category": "Features",
              "description": [
                  "Makes bufferSubData always update via CPU"
              ],
              "issue": "http://anglebug.com/7544"
          },
          {
              "name": "disable_programmable_blending",
              "category": "Features",
              "description": [
                  "Disable programmable blending in order to test read_write pixel local storage textures"
              ],
              "issue": "http://anglebug.com/7279"
          },
          {
              "name": "disable_r_w_texture_tier2_support",
              "category": "Features",
              "description": [
                  "Disable tier2 read_write textures in order to test tier1 support"
              ],
              "issue": "http://anglebug.com/7279"
          },
          {
              "name": "disable_raster_order_groups",
              "category": "Features",
              "description": [
                  "Disable raster order groups in order to test pixel local storage memory barriers"
              ],
              "issue": "http://anglebug.com/7279"
          },
          {
              "name": "enable_in_memory_mtl_library_cache",
              "category": "Features",
              "description": [
                  "Cache MTLLibrary objects in memory."
              ],
              "issue": "http://crbug.com/1385510"
          }
      ]
  }
  

Download