anv: Advertise ray-tracing on DG2

Also disable ray-tracing support if with_intel_vk_rt is not set.

Signed-off-by: Jordan Justen <jordan.l.justen@intel.com>
Reviewed-by: Lionel Landwerlin <lionel.g.landwerlin@intel.com>
Acked-by: Caio Oliveira <caio.oliveira@intel.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/16970>

src/intel/vulkan/anv_device.c

@@ -192,6 +192,8 @@ get_device_extensions(const struct anv_physical_device *device,
       .KHR_8bit_storage                      = true,
       .KHR_16bit_storage                     = true,
       .KHR_acceleration_structure            = device->info.has_ray_tracing,
+      .KHR_acceleration_structure            = ANV_SUPPORT_RT &&
+                                               device->info.has_ray_tracing,
       .KHR_bind_memory2                      = true,
       .KHR_buffer_device_address             = true,
       .KHR_copy_commands2                    = true,
@@ -231,7 +233,10 @@ get_device_extensions(const struct anv_physical_device *device,
       .KHR_pipeline_executable_properties    = true,
       .KHR_pipeline_library                  = true,
       .KHR_push_descriptor                   = true,
-      .KHR_ray_query                         = device->info.has_ray_tracing,
+      .KHR_ray_query                         =
+         ANV_SUPPORT_RT && device->info.has_ray_tracing,
+      .KHR_ray_tracing_pipeline              =
+         ANV_SUPPORT_RT && device->info.has_ray_tracing,
       .KHR_relaxed_block_layout              = true,
       .KHR_sampler_mirror_clamp_to_edge      = true,
       .KHR_sampler_ycbcr_conversion          = true,
@@ -1344,12 +1349,13 @@ void anv_GetPhysicalDeviceFeatures2(
 
       case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_ACCELERATION_STRUCTURE_FEATURES_KHR: {
          VkPhysicalDeviceAccelerationStructureFeaturesKHR *features = (void *)ext;
-         features->accelerationStructure = pdevice->info.has_ray_tracing;
+         features->accelerationStructure =
+            ANV_SUPPORT_RT && pdevice->info.has_ray_tracing;
          features->accelerationStructureCaptureReplay = false; /* TODO */
          features->accelerationStructureIndirectBuild = false; /* TODO */
          features->accelerationStructureHostCommands = false;
          features->descriptorBindingAccelerationStructureUpdateAfterBind =
-            pdevice->info.has_ray_tracing;
+            ANV_SUPPORT_RT && pdevice->info.has_ray_tracing;
          break;
       }
 
@@ -1549,7 +1555,17 @@ void anv_GetPhysicalDeviceFeatures2(
 
       case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_RAY_QUERY_FEATURES_KHR: {
          VkPhysicalDeviceRayQueryFeaturesKHR *features = (void *)ext;
-         features->rayQuery = pdevice->info.has_ray_tracing;
+         features->rayQuery = ANV_SUPPORT_RT && pdevice->info.has_ray_tracing;
+         break;
+      }
+
+      case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_RAY_TRACING_PIPELINE_FEATURES_KHR: {
+         VkPhysicalDeviceRayTracingPipelineFeaturesKHR *features = (void *)ext;
+         features->rayTracingPipeline = pdevice->info.has_ray_tracing;
+         features->rayTracingPipelineShaderGroupHandleCaptureReplay = false;
+         features->rayTracingPipelineShaderGroupHandleCaptureReplayMixed = false;
+         features->rayTracingPipelineTraceRaysIndirect = true;
+         features->rayTraversalPrimitiveCulling = false;
          break;
       }
 
@@ -2559,6 +2575,22 @@ void anv_GetPhysicalDeviceProperties2(
          break;
       }
 
+      case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_RAY_TRACING_PIPELINE_PROPERTIES_KHR: {
+         VkPhysicalDeviceRayTracingPipelinePropertiesKHR *props = (void *)ext;
+         /* TODO */
+         props->shaderGroupHandleSize = 32;
+         props->maxRayRecursionDepth = 31;
+         /* MemRay::hitGroupSRStride is 16 bits */
+         props->maxShaderGroupStride = UINT16_MAX;
+         /* MemRay::hitGroupSRBasePtr requires 16B alignment */
+         props->shaderGroupBaseAlignment = 16;
+         props->shaderGroupHandleAlignment = 16;
+         props->shaderGroupHandleCaptureReplaySize = 32;
+         props->maxRayDispatchInvocationCount = 1U << 30; /* required min limit */
+         props->maxRayHitAttributeSize = BRW_RT_SIZEOF_HIT_ATTRIB_DATA;
+         break;
+      }
+
       case VK_STRUCTURE_TYPE_PHYSICAL_DEVICE_ROBUSTNESS_2_PROPERTIES_EXT: {
          VkPhysicalDeviceRobustness2PropertiesEXT *properties = (void *)ext;
          properties->robustStorageBufferAccessSizeAlignment =
@@ -3461,7 +3493,7 @@ VkResult anv_CreateDevice(
    /* TODO(RT): Do we want some sort of data structure for this? */
    memset(device->rt_scratch_bos, 0, sizeof(device->rt_scratch_bos));
 
-   if (device->info->has_ray_tracing) {
+   if (ANV_SUPPORT_RT && device->info->has_ray_tracing) {
       /* The docs say to always allocate 128KB per DSS */
       const uint32_t btd_fifo_bo_size =
          128 * 1024 * intel_device_info_dual_subslice_id_bound(device->info);
@@ -3523,7 +3555,7 @@ VkResult anv_CreateDevice(
  fail_default_pipeline_cache:
    vk_pipeline_cache_destroy(device->default_pipeline_cache, NULL);
  fail_btd_fifo_bo:
-   if (device->info->has_ray_tracing)
+   if (ANV_SUPPORT_RT && device->info->has_ray_tracing)
       anv_device_release_bo(device, device->btd_fifo_bo);
  fail_trivial_batch_bo_and_scratch_pool:
    anv_scratch_pool_finish(device, &device->scratch_pool);
@@ -3595,7 +3627,7 @@ void anv_DestroyDevice(
    vk_pipeline_cache_destroy(device->internal_cache, NULL);
    vk_pipeline_cache_destroy(device->default_pipeline_cache, NULL);
 
-   if (device->info->has_ray_tracing)
+   if (ANV_SUPPORT_RT && device->info->has_ray_tracing)
       anv_device_release_bo(device, device->btd_fifo_bo);
 
 #ifdef HAVE_VALGRIND

src/intel/vulkan/anv_pipeline.c

@@ -227,8 +227,8 @@ anv_shader_stage_to_nir(struct anv_device *device,
          .post_depth_coverage = true,
          .runtime_descriptor_array = true,
          .float_controls = true,
-         .ray_query = pdevice->info.has_ray_tracing,
-         .ray_tracing = pdevice->info.has_ray_tracing,
+         .ray_query = ANV_SUPPORT_RT && pdevice->info.has_ray_tracing,
+         .ray_tracing = ANV_SUPPORT_RT && pdevice->info.has_ray_tracing,
          .shader_clock = true,
          .shader_viewport_index_layer = true,
          .stencil_export = true,

src/intel/vulkan/genX_query.c

@@ -1510,7 +1510,7 @@ void genX(CmdCopyQueryPoolResults)(
    }
 }
 
-#if GFX_VERx10 >= 125
+#if GFX_VERx10 >= 125 && ANV_SUPPORT_RT
 
 #include "grl/include/GRLRTASCommon.h"
 #include "grl/grl_metakernel_postbuild_info.h"

src/intel/vulkan/genX_state.c

@@ -37,7 +37,7 @@
 
 #include "vk_standard_sample_locations.h"
 
-#if GFX_VERx10 >= 125
+#if GFX_VERx10 >= 125 && ANV_SUPPORT_RT
 #include "grl/genX_grl.h"
 #endif
 
@@ -235,7 +235,7 @@ init_common_queue_state(struct anv_queue *queue, struct anv_batch *batch)
 #endif
 
 #if GFX_VERx10 >= 125
-   if (device->info->has_ray_tracing) {
+   if (ANV_SUPPORT_RT && device->info->has_ray_tracing) {
       anv_batch_emit(batch, GENX(3DSTATE_BTD), btd) {
          /* TODO: This is the timeout after which the bucketed thread
           *       dispatcher will kick off a wave of threads. We go with the
@@ -471,7 +471,7 @@ void
 genX(init_physical_device_state)(ASSERTED struct anv_physical_device *pdevice)
 {
    assert(pdevice->info.verx10 == GFX_VERx10);
-#if GFX_VERx10 >= 125
+#if GFX_VERx10 >= 125 && ANV_SUPPORT_RT
    genX(grl_load_rt_uuid)(pdevice->rt_uuid);
 #endif
 }

src/intel/vulkan/meson.build

@@ -27,7 +27,15 @@ anv_flags = [
 
 anv_cpp_flags = []
 
-subdir('grl')
+if with_intel_vk_rt
+  subdir('grl')
+  optional_libgrl = [libgrl]
+  anv_flags += '-DANV_SUPPORT_RT=1'
+else
+  idep_grl = null_dep
+  optional_libgrl = []
+  anv_flags += '-DANV_SUPPORT_RT=0'
+endif
 
 anv_entrypoints = custom_target(
   'anv_entrypoints',
@@ -80,7 +88,6 @@ endif
 
 libanv_per_hw_ver_libs = []
 anv_per_hw_ver_files = files(
-  'genX_acceleration_structure.c',
   'genX_blorp_exec.c',
   'genX_cmd_buffer.c',
   'genX_gpu_memcpy.c',
@@ -88,6 +95,10 @@ anv_per_hw_ver_files = files(
   'genX_query.c',
   'genX_state.c',
 )
+if with_intel_vk_rt
+  anv_per_hw_ver_files += files('genX_acceleration_structure.c',)
+endif
+
 foreach g : [['90', ['gfx8_cmd_buffer.c']],
              ['110', ['gfx8_cmd_buffer.c']],
              ['120', ['gfx8_cmd_buffer.c']],
@@ -193,7 +204,7 @@ libvulkan_intel = shared_library(
   include_directories : [
     inc_include, inc_src, inc_mapi, inc_mesa, inc_gallium, inc_intel, inc_compiler,
   ],
-  link_whole : [libanv_common, libanv_per_hw_ver_libs, libgrl],
+  link_whole : [libanv_common, libanv_per_hw_ver_libs] + optional_libgrl,
   link_with : [
     libintel_compiler, libintel_dev, libisl, libblorp, libintel_perf,
   ],
@@ -231,9 +242,9 @@ if with_tests
     ],
     link_whole : libanv_common,
     link_with : [
-      libanv_per_hw_ver_libs, libgrl, libintel_compiler, libintel_common, libintel_dev,
+      libanv_per_hw_ver_libs, libintel_compiler, libintel_common, libintel_dev,
       libisl, libblorp, libintel_perf,
-    ],
+    ] + optional_libgrl,
     dependencies : [
       dep_thread, dep_dl, dep_m, anv_deps,
       idep_nir, idep_vulkan_util, idep_vulkan_wsi, idep_vulkan_runtime,