diff options
Diffstat (limited to 'drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c')
| -rw-r--r-- | drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c | 46 |
1 files changed, 44 insertions, 2 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c index 5b631fd1a879..c4e14015ec5b 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c @@ -2105,6 +2105,18 @@ static void gfx_v7_0_ring_emit_hdp_flush(struct amdgpu_ring *ring) amdgpu_ring_write(ring, 0x20); /* poll interval */ } +static void gfx_v7_0_ring_emit_vgt_flush(struct amdgpu_ring *ring) +{ + amdgpu_ring_write(ring, PACKET3(PACKET3_EVENT_WRITE, 0)); + amdgpu_ring_write(ring, EVENT_TYPE(VS_PARTIAL_FLUSH) | + EVENT_INDEX(4)); + + amdgpu_ring_write(ring, PACKET3(PACKET3_EVENT_WRITE, 0)); + amdgpu_ring_write(ring, EVENT_TYPE(VGT_FLUSH) | + EVENT_INDEX(0)); +} + + /** * gfx_v7_0_ring_emit_hdp_invalidate - emit an hdp invalidate on the cp * @@ -2260,6 +2272,7 @@ static void gfx_v7_ring_emit_cntxcntl(struct amdgpu_ring *ring, uint32_t flags) dw2 |= 0x80000000; /* set load_enable otherwise this package is just NOPs */ if (flags & AMDGPU_HAVE_CTX_SWITCH) { + gfx_v7_0_ring_emit_vgt_flush(ring); /* set load_global_config & load_global_uconfig */ dw2 |= 0x8001; /* set load_cs_sh_regs */ @@ -4359,10 +4372,29 @@ static void gfx_v7_0_ring_emit_gds_switch(struct amdgpu_ring *ring, static uint32_t wave_read_ind(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t address) { - WREG32(mmSQ_IND_INDEX, (wave & 0xF) | ((simd & 0x3) << 4) | (address << 16) | (1 << 13)); + WREG32(mmSQ_IND_INDEX, + (wave << SQ_IND_INDEX__WAVE_ID__SHIFT) | + (simd << SQ_IND_INDEX__SIMD_ID__SHIFT) | + (address << SQ_IND_INDEX__INDEX__SHIFT) | + (SQ_IND_INDEX__FORCE_READ_MASK)); return RREG32(mmSQ_IND_DATA); } +static void wave_read_regs(struct amdgpu_device *adev, uint32_t simd, + uint32_t wave, uint32_t thread, + uint32_t regno, uint32_t num, uint32_t *out) +{ + WREG32(mmSQ_IND_INDEX, + (wave << SQ_IND_INDEX__WAVE_ID__SHIFT) | + (simd << SQ_IND_INDEX__SIMD_ID__SHIFT) | + (regno << SQ_IND_INDEX__INDEX__SHIFT) | + (thread << SQ_IND_INDEX__THREAD_ID__SHIFT) | + (SQ_IND_INDEX__FORCE_READ_MASK) | + (SQ_IND_INDEX__AUTO_INCR_MASK)); + while (num--) + *(out++) = RREG32(mmSQ_IND_DATA); +} + static void gfx_v7_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, uint32_t wave, uint32_t *dst, int *no_fields) { /* type 0 wave data */ @@ -4387,10 +4419,20 @@ static void gfx_v7_0_read_wave_data(struct amdgpu_device *adev, uint32_t simd, u dst[(*no_fields)++] = wave_read_ind(adev, simd, wave, ixSQ_WAVE_M0); } +static void gfx_v7_0_read_wave_sgprs(struct amdgpu_device *adev, uint32_t simd, + uint32_t wave, uint32_t start, + uint32_t size, uint32_t *dst) +{ + wave_read_regs( + adev, simd, wave, 0, + start + SQIND_WAVE_SGPRS_OFFSET, size, dst); +} + static const struct amdgpu_gfx_funcs gfx_v7_0_gfx_funcs = { .get_gpu_clock_counter = &gfx_v7_0_get_gpu_clock_counter, .select_se_sh = &gfx_v7_0_select_se_sh, .read_wave_data = &gfx_v7_0_read_wave_data, + .read_wave_sgprs = &gfx_v7_0_read_wave_sgprs, }; static const struct amdgpu_rlc_funcs gfx_v7_0_rlc_funcs = { @@ -5149,7 +5191,7 @@ static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_gfx = { 12 + 12 + 12 + /* gfx_v7_0_ring_emit_fence_gfx x3 for user fence, vm fence */ 7 + 4 + /* gfx_v7_0_ring_emit_pipeline_sync */ 17 + 6 + /* gfx_v7_0_ring_emit_vm_flush */ - 3, /* gfx_v7_ring_emit_cntxcntl */ + 3 + 4, /* gfx_v7_ring_emit_cntxcntl including vgt flush*/ .emit_ib_size = 4, /* gfx_v7_0_ring_emit_ib_gfx */ .emit_ib = gfx_v7_0_ring_emit_ib_gfx, .emit_fence = gfx_v7_0_ring_emit_fence_gfx, |