Mesa (main): radv: Enable NV_mesh_shader with a perftest flag.

[GitLab Mirror]

Module: Mesa
Branch: main
Commit: 6fee84bc2e7aaa250df00f0731a05310750d2a9c
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=6fee84bc2e7aaa250df00f0731a05310750d2a9c

Author: Timur Kristóf <timur.kristof at gmail.com>
Date:   Thu Oct 28 21:25:48 2021 +0200

radv: Enable NV_mesh_shader with a perftest flag.

We don't plan to support NV_mesh_shader officially on RADV,
because it performs poorly on AMD hardware. However, we are
implementing this extension to get some experience with mesh
shader technology.

Users should not rely on this support because we are going
to remove it if/when a potential cross-vendor extension appears.

Signed-off-by: Timur Kristóf <timur.kristof at gmail.com>
Acked-by: Bas Nieuwenhuizen <bas at basnieuwenhuizen.nl>
Reviewed-by: Samuel Pitoiset <samuel.pitoiset at gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/13580>

---

 docs/envvars.rst             |  2 ++
 src/amd/vulkan/radv_debug.h  |  1 +
 src/amd/vulkan/radv_device.c | 41 +++++++++++++++++++++++++++++++++++++++++
 3 files changed, 44 insertions(+)

diff --git a/docs/envvars.rst b/docs/envvars.rst
index a79a785f6cb..268dad206f6 100644
--- a/docs/envvars.rst
+++ b/docs/envvars.rst
@@ -686,6 +686,8 @@ RADV driver environment variables
       enable local BOs
    ``nosam``
       disable optimizations that get enabled when all VRAM is CPU visible.
+   ``nv_ms``
+      enable unofficial experimental support for NV_mesh_shader.
    ``pswave32``
       enable wave32 for pixel shaders (GFX10+)
    ``nggc``
diff --git a/src/amd/vulkan/radv_debug.h b/src/amd/vulkan/radv_debug.h
index 5c0dd14220e..491efc4b87b 100644
--- a/src/amd/vulkan/radv_debug.h
+++ b/src/amd/vulkan/radv_debug.h
@@ -78,6 +78,7 @@ enum {
    RADV_PERFTEST_RT = 1u << 8,
    RADV_PERFTEST_NGGC = 1u << 9,
    RADV_PERFTEST_FORCE_EMULATE_RT = 1u << 10,
+   RADV_PERFTEST_NV_MS = 1u << 11,
 };
 
 bool radv_init_trace(struct radv_device *device);
diff --git a/src/amd/vulkan/radv_device.c b/src/amd/vulkan/radv_device.c
index 1537f110816..0e1a56b7542 100644
--- a/src/amd/vulkan/radv_device.c
+++ b/src/amd/vulkan/radv_device.c
@@ -568,6 +568,8 @@ radv_physical_device_get_supported_extensions(const struct radv_physical_device
       .GOOGLE_hlsl_functionality1 = true,
       .GOOGLE_user_type = true,
       .NV_compute_shader_derivatives = true,
+      .NV_mesh_shader = device->use_ngg && device->rad_info.chip_class >= GFX10_3 &&
+                        device->instance->perftest_flags & RADV_PERFTEST_NV_MS && !device->use_llvm,
       .VALVE_mutable_descriptor_type = true,
    };
 }
@@ -896,6 +898,7 @@ static const struct debug_control radv_perftest_options[] = {{"localbos", RADV_P
                                                              {"rt", RADV_PERFTEST_RT},
                                                              {"nggc", RADV_PERFTEST_NGGC},
                                                              {"force_emulate_rt", RADV_PERFTEST_FORCE_EMULATE_RT},
+                                                             {"nv_ms", RADV_PERFTEST_NV_MS},
                                                              {NULL, 0}};
 
 const char *
@@ -1674,6 +1677,13 @@ radv_GetPhysicalDeviceFeatures2(VkPhysicalDevice physicalDevice,
          features->dynamicRendering = true;
          break;
       }
+      case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MESH_SHADER_FEATURES_NV: {
+         VkPhysicalDeviceMeshShaderFeaturesNV *features =
+            (VkPhysicalDeviceMeshShaderFeaturesNV *)ext;
+         features->meshShader = true;
+         features->taskShader = false; /* TODO */
+         break;
+      }
       default:
          break;
       }
@@ -2323,6 +2333,37 @@ radv_GetPhysicalDeviceProperties2(VkPhysicalDevice physicalDevice,
          properties->maxBufferSize = RADV_MAX_MEMORY_ALLOCATION_SIZE;
          break;
       }
+      case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_MESH_SHADER_PROPERTIES_NV: {
+         VkPhysicalDeviceMeshShaderPropertiesNV *properties =
+            (VkPhysicalDeviceMeshShaderPropertiesNV *)ext;
+
+         /* Task shader limitations:
+          * Same as compute, because TS are compiled to CS.
+          */
+         properties->maxDrawMeshTasksCount = 65535;
+         properties->maxTaskTotalMemorySize = 65536;
+         properties->maxTaskWorkGroupInvocations = 1024;
+         properties->maxTaskWorkGroupSize[0] = 1024;
+         properties->maxTaskWorkGroupSize[1] = 1024;
+         properties->maxTaskWorkGroupSize[2] = 1024;
+         properties->maxTaskOutputCount = 1024;
+
+         /* Mesh shader limitations:
+          * Same as NGG, because MS are compiled to NGG.
+          */
+         properties->maxMeshMultiviewViewCount = MAX_VIEWS;
+         properties->maxMeshOutputPrimitives = 256;
+         properties->maxMeshOutputVertices = 256;
+         properties->maxMeshTotalMemorySize = 31 * 1024; /* Reserve 1K for prim indices, etc. */
+         properties->maxMeshWorkGroupInvocations = 256;
+         properties->maxMeshWorkGroupSize[0] = 256;
+         properties->maxMeshWorkGroupSize[1] = 256;
+         properties->maxMeshWorkGroupSize[2] = 256;
+         properties->meshOutputPerPrimitiveGranularity = 1;
+         properties->meshOutputPerVertexGranularity = 1;
+
+         break;
+      }
       default:
          break;
       }