aco: use nir_intrinsic_has_access

[mesa.git] / src / amd / vulkan / radv_rgp.c

diff --git a/src/amd/vulkan/radv_rgp.c b/src/amd/vulkan/radv_rgp.c
index 3daef62ed5f6f8ce1b00a7f631520306df9a0a61..7003cf456b29ec3b1252366141d6868ef8390155 100644 (file)
--- a/src/amd/vulkan/radv_rgp.c
+++ b/src/amd/vulkan/radv_rgp.c
@@ -27,7 +27,7 @@
 
 #include "util/u_process.h"
 
-#include <amdgpu_drm.h>
+#include "drm-uapi/amdgpu_drm.h"
 
 #define SQTT_FILE_MAGIC_NUMBER 0x50303042
 #define SQTT_FILE_VERSION_MAJOR 1
@@ -166,6 +166,8 @@ static_assert(sizeof(struct sqtt_file_chunk_cpu_info) == 112,
 static void
 radv_sqtt_fill_cpu_info(struct sqtt_file_chunk_cpu_info *chunk)
 {
+       uint64_t system_ram_size = 0;
+
        chunk->header.chunk_id.type = SQTT_FILE_CHUNK_TYPE_CPU_INFO;
        chunk->header.chunk_id.index = 0;
        chunk->header.major_version = 0;
@@ -181,7 +183,10 @@ radv_sqtt_fill_cpu_info(struct sqtt_file_chunk_cpu_info *chunk)
        chunk->clock_speed = 0;
        chunk->num_logical_cores = 0;
        chunk->num_physical_cores = 0;
+
        chunk->system_ram_size = 0;
+       if (os_get_total_physical_memory(&system_ram_size))
+               chunk->system_ram_size = system_ram_size / (1024 * 1024);
 }
 
 /**
@@ -345,20 +350,20 @@ radv_fill_sqtt_asic_info(struct radv_device *device,
        if (rad_info->family == CHIP_FIJI || rad_info->chip_class >= GFX9)
                chunk->flags |= SQTT_FILE_CHUNK_ASIC_INFO_FLAG_PS1_EVENT_TOKENS_ENABLED;
 
-       chunk->trace_shader_core_clock = rad_info->max_shader_clock;
-       chunk->trace_memory_clock = rad_info->max_memory_clock;
+       chunk->trace_shader_core_clock = rad_info->max_shader_clock * 1000000;
+       chunk->trace_memory_clock = rad_info->max_memory_clock * 1000000;
 
        chunk->device_id = rad_info->pci_id;
        chunk->device_revision_id = rad_info->pci_rev_id;
        chunk->vgprs_per_simd = rad_info->num_physical_wave64_vgprs_per_simd;
        chunk->sgprs_per_simd = rad_info->num_physical_sgprs_per_simd;
        chunk->shader_engines = rad_info->max_se;
-       chunk->compute_unit_per_shader_engine = rad_info->num_good_cu_per_sh;
+       chunk->compute_unit_per_shader_engine = rad_info->min_good_cu_per_sa;
        chunk->simd_per_compute_unit = rad_info->num_simd_per_compute_unit;
        chunk->wavefronts_per_simd = rad_info->max_wave64_per_simd;
 
-       chunk->minimum_vgpr_alloc = rad_info->min_vgpr_alloc;
-       chunk->vgpr_alloc_granularity = rad_info->vgpr_alloc_granularity;
+       chunk->minimum_vgpr_alloc = rad_info->min_wave64_vgpr_alloc;
+       chunk->vgpr_alloc_granularity = rad_info->wave64_vgpr_alloc_granularity;
        chunk->minimum_sgpr_alloc = rad_info->min_sgpr_alloc;
        chunk->sgpr_alloc_granularity = rad_info->sgpr_alloc_granularity;
 
@@ -376,7 +381,7 @@ radv_fill_sqtt_asic_info(struct radv_device *device,
        chunk->vram_size = rad_info->vram_size;
        chunk->l2_cache_size = rad_info->l2_cache_size;
        chunk->l1_cache_size = rad_info->l1_cache_size;
-       chunk->lds_size = rad_info->lds_size_per_cu;
+       chunk->lds_size = rad_info->lds_size_per_workgroup;
 
        strncpy(chunk->gpu_name, device->physical_device->name, SQTT_GPU_NAME_MAX_SIZE);
 
@@ -392,9 +397,9 @@ radv_fill_sqtt_asic_info(struct radv_device *device,
        chunk->memory_chip_type = radv_vram_type_to_sqtt_memory_type(rad_info->vram_type);
        chunk->lds_granularity = rad_info->lds_granularity;
 
-       for (unsigned se = 0; se < 32; se++) {
+       for (unsigned se = 0; se < 4; se++) {
                for (unsigned sa = 0; sa < 2; sa++) {
-                       chunk->cu_mask[se][sa] = 0; /* TODO */
+                       chunk->cu_mask[se][sa] = rad_info->cu_mask[se][sa];
                }
        }
 }
@@ -541,7 +546,7 @@ radv_sqtt_fill_sqtt_desc(struct radv_device *device,
 
        chunk->sqtt_version = radv_chip_class_to_sqtt_version(device->physical_device->rad_info.chip_class);
        chunk->shader_engine_index = shader_engine_index;
-       chunk->v1.instrumentation_spec_version = 0;
+       chunk->v1.instrumentation_spec_version = 1;
        chunk->v1.instrumentation_api_version = 0;
        chunk->v1.compute_unit_index = compute_unit_index;
 }