OpenHarmony/third_party_mesa3d
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

radv/gfx10: enable wave32 for compute based on shader's wavesize

Browse Source 
This will allow to change wavesize on-demand.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Bas Nieuwenhuizen <bas@basnieuwenhuizen.nl>
This commit is contained in:
Samuel Pitoiset
2019-10-31 09:30:47 +01:00
parent c0f76528ae
commit f010b90ac5
3 changed files with 7 additions and 3 deletions
Download Patch File Download Diff File Expand all files Collapse all files

5
src/amd/vulkan/radv_cmd_buffer.c
View File

@@ -4850,6 +4850,11 @@ radv_emit_dispatch_packets(struct radv_cmd_buffer *cmd_buffer,
ASSERTED unsigned cdw_max = radeon_check_space(ws, cs, 25); ASSERTED unsigned cdw_max = radeon_check_space(ws, cs, 25);
if (compute_shader->info.wave_size == 32) {
assert(cmd_buffer->device->physical_device->rad_info.chip_class >= GFX10);
dispatch_initiator |= S_00B800_CS_W32_EN(1);
}
if (info->indirect) { if (info->indirect) {
uint64_t va = radv_buffer_get_va(info->indirect->bo); uint64_t va = radv_buffer_get_va(info->indirect->bo);

3
src/amd/vulkan/radv_device.c
View File

@@ -2486,8 +2486,7 @@ VkResult radv_CreateDevice(
device->scratch_waves = MAX2(32 * physical_device->rad_info.num_good_compute_units, device->scratch_waves = MAX2(32 * physical_device->rad_info.num_good_compute_units,
max_threads_per_block / 64); max_threads_per_block / 64);
device->dispatch_initiator = S_00B800_COMPUTE_SHADER_EN(1) | device->dispatch_initiator = S_00B800_COMPUTE_SHADER_EN(1);
S_00B800_CS_W32_EN(device->physical_device->cs_wave_size == 32);
if (device->physical_device->rad_info.chip_class >= GFX7) { if (device->physical_device->rad_info.chip_class >= GFX7) {
/* If the KMD allows it (there is a KMD hw register for it), /* If the KMD allows it (there is a KMD hw register for it),

2
src/amd/vulkan/radv_pipeline.c
View File

@@ -5042,7 +5042,7 @@ radv_compute_generate_pm4(struct radv_pipeline *pipeline)
compute_shader->info.cs.block_size[1] * compute_shader->info.cs.block_size[1] *
compute_shader->info.cs.block_size[2]; compute_shader->info.cs.block_size[2];
waves_per_threadgroup = DIV_ROUND_UP(threads_per_threadgroup, waves_per_threadgroup = DIV_ROUND_UP(threads_per_threadgroup,
device->physical_device->cs_wave_size); compute_shader->info.wave_size);
if (device->physical_device->rad_info.chip_class >= GFX10 && if (device->physical_device->rad_info.chip_class >= GFX10 &&
waves_per_threadgroup == 1) waves_per_threadgroup == 1)