diff options
| author |   Nirmoy Das <nirmoy.das@amd.com> | 2021-01-27 11:35:54 +0100 |
|---|---|---|
| committer |   Alex Deucher <alexander.deucher@amd.com> | 2021-02-09 15:27:04 -0500 |
| commit | 0a52a6cacca6030374fca7087e8556b382cca0e3 (patch) | |
| tree | 733da3806d50d8266fd1a5d576acb07defbd6dd5 /drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | |
| parent | drm/amdgpu: enable only one high prio compute queue (diff) | |
| download | linux-dev-0a52a6cacca6030374fca7087e8556b382cca0e3.tar.xz linux-dev-0a52a6cacca6030374fca7087e8556b382cca0e3.zip | |
drm/amdgpu: add wave limit functionality for gfx8,9
Wave limiting can be use to load balance high priority
compute jobs along with gfx jobs. When enabled, this will reserve
~75% of waves for compute jobs.
We do not need this from gfx10 onwards because >=gfx10 has
asynchronous compute tunneling to replace wave limit requirement.
Signed-off-by: Nirmoy Das <nirmoy.das@amd.com>
Reviewed-by: Christian König <christian.koenig@amd.com>
Reviewed-by: Alex Deucher <alexander.deucher@amd.com>
Signed-off-by: Alex Deucher <alexander.deucher@amd.com>
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c')
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c | 18 |
1 files changed, 17 insertions, 1 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c index b0284c4659ba..bdfd29a22b3d 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v8_0.c @@ -29,6 +29,7 @@ #include "amdgpu.h" #include "amdgpu_gfx.h" +#include "amdgpu_ring.h" #include "vi.h" #include "vi_structs.h" #include "vid.h" @@ -6845,6 +6846,19 @@ static void gfx_v8_0_emit_mem_sync_compute(struct amdgpu_ring *ring) amdgpu_ring_write(ring, 0x0000000A); /* poll interval */ } +#define mmSPI_WCL_PIPE_PERCENT_GFX_DEFAULT 0x07ffffff +static void gfx_v8_0_emit_wave_limit(struct amdgpu_ring *ring, bool enable) +{ + uint32_t val; + + /* mmSPI_WCL_PIPE_PERCENT_GFX is 7 bit multiplier register to limit + * number of gfx waves. Setting 5 bit will make sure gfx only gets + * around 25% of gpu resources. + */ + val = enable ? 0x1f : mmSPI_WCL_PIPE_PERCENT_GFX_DEFAULT; + amdgpu_ring_emit_wreg(ring, mmSPI_WCL_PIPE_PERCENT_GFX, val); +} + static const struct amd_ip_funcs gfx_v8_0_ip_funcs = { .name = "gfx_v8_0", .early_init = gfx_v8_0_early_init, @@ -6928,7 +6942,8 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_compute = { 7 + /* gfx_v8_0_ring_emit_pipeline_sync */ VI_FLUSH_GPU_TLB_NUM_WREG * 5 + 7 + /* gfx_v8_0_ring_emit_vm_flush */ 7 + 7 + 7 + /* gfx_v8_0_ring_emit_fence_compute x3 for user fence, vm fence */ - 7, /* gfx_v8_0_emit_mem_sync_compute */ + 7 + /* gfx_v8_0_emit_mem_sync_compute */ + 5, /* gfx_v8_0_emit_wave_limit for updating mmSPI_WCL_PIPE_PERCENT_GFX register */ .emit_ib_size = 7, /* gfx_v8_0_ring_emit_ib_compute */ .emit_ib = gfx_v8_0_ring_emit_ib_compute, .emit_fence = gfx_v8_0_ring_emit_fence_compute, @@ -6942,6 +6957,7 @@ static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_compute = { .pad_ib = amdgpu_ring_generic_pad_ib, .emit_wreg = gfx_v8_0_ring_emit_wreg, .emit_mem_sync = gfx_v8_0_emit_mem_sync_compute, + .emit_wave_limit = gfx_v8_0_emit_wave_limit, }; static const struct amdgpu_ring_funcs gfx_v8_0_ring_funcs_kiq = { |