Mesa (main): radv: Add an indirect dispatch struct to the header.

[GitLab Mirror]

Module: Mesa
Branch: main
Commit: c814e42c123b3138a56b12997224a1269ad28e27
URL:    http://cgit.freedesktop.org/mesa/mesa/commit/?id=c814e42c123b3138a56b12997224a1269ad28e27

Author: Bas Nieuwenhuizen <bas at basnieuwenhuizen.nl>
Date:   Tue Aug 31 02:53:17 2021 +0200

radv: Add an indirect dispatch struct to the header.

Reviewed-by: Samuel Pitoiset <samuel.pitoiset at gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/12840>

---

 src/amd/vulkan/radv_acceleration_structure.c | 25 +++++++++++++++++++++++--
 src/amd/vulkan/radv_acceleration_structure.h |  3 +++
 2 files changed, 26 insertions(+), 2 deletions(-)

diff --git a/src/amd/vulkan/radv_acceleration_structure.c b/src/amd/vulkan/radv_acceleration_structure.c
index cb1f7e94c8c..77c8826e007 100644
--- a/src/amd/vulkan/radv_acceleration_structure.c
+++ b/src/amd/vulkan/radv_acceleration_structure.c
@@ -622,7 +622,12 @@ build_bvh(struct radv_device *device, const VkAccelerationStructureBuildGeometry
 
    compute_bounds(base_ptr, header->root_node_offset, &header->aabb[0][0]);
 
-   /* TODO init sizes and figure out what is needed for serialization. */
+   header->compacted_size = (char *)ctx.curr_ptr - base_ptr;
+
+   /* 16 bytes per invocation, 64 invocations per workgroup */
+   header->copy_dispatch_size[0] = DIV_ROUND_UP(header->compacted_size, 16 * 64);
+   header->copy_dispatch_size[1] = 1;
+   header->copy_dispatch_size[2] = 1;
 
 fail:
    device->ws->buffer_unmap(accel->bo);
@@ -1519,11 +1524,27 @@ radv_CmdBuildAccelerationStructuresKHR(
                                cmd_buffer->device->meta_state.accel_struct_build.internal_p_layout,
                                VK_SHADER_STAGE_COMPUTE_BIT, 0, sizeof(consts), &consts);
          radv_unaligned_dispatch(cmd_buffer, dst_node_count, 1, 1);
-         bvh_states[i].node_offset += dst_node_count * 128;
+         if (!final_iter)
+            bvh_states[i].node_offset += dst_node_count * 128;
          bvh_states[i].node_count = dst_node_count;
          bvh_states[i].scratch_offset = dst_scratch_offset;
       }
    }
+   for (uint32_t i = 0; i < infoCount; ++i) {
+      RADV_FROM_HANDLE(radv_acceleration_structure, accel_struct,
+                       pInfos[i].dstAccelerationStructure);
+      const size_t base = offsetof(struct radv_accel_struct_header, compacted_size);
+      struct radv_accel_struct_header header;
+      header.compacted_size = bvh_states[i].node_offset;
+
+      /* 16 bytes per invocation, 64 invocations per workgroup */
+      header.copy_dispatch_size[0] = DIV_ROUND_UP(header.compacted_size, 16 * 64);
+      header.copy_dispatch_size[1] = 1;
+      header.copy_dispatch_size[2] = 1;
+
+      radv_update_buffer(cmd_buffer, accel_struct->bo, accel_struct->mem_offset + base,
+                         (const char *)&header + base, sizeof(header) - base);
+   }
    free(bvh_states);
    radv_meta_restore(&saved_state, cmd_buffer);
 }
diff --git a/src/amd/vulkan/radv_acceleration_structure.h b/src/amd/vulkan/radv_acceleration_structure.h
index 63d6ed29eb8..7278a9ec136 100644
--- a/src/amd/vulkan/radv_acceleration_structure.h
+++ b/src/amd/vulkan/radv_acceleration_structure.h
@@ -30,8 +30,11 @@ struct radv_accel_struct_header {
    uint32_t root_node_offset;
    uint32_t reserved;
    float aabb[2][3];
+
+   /* Everything after this gets updated/copied from the CPU. */
    uint64_t compacted_size;
    uint64_t serialization_size;
+   uint32_t copy_dispatch_size[3];
 };
 
 struct radv_bvh_triangle_node {