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 | 68 |
1 files changed, 21 insertions, 47 deletions
diff --git a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c index a4309698e76c..e13d9d83767b 100644 --- a/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c +++ b/drivers/gpu/drm/amd/amdgpu/gfx_v7_0.c @@ -1946,7 +1946,7 @@ static void gfx_v7_0_gpu_init(struct amdgpu_device *adev) if (i == 0) sh_mem_base = 0; else - sh_mem_base = adev->mc.shared_aperture_start >> 48; + sh_mem_base = adev->gmc.shared_aperture_start >> 48; cik_srbm_select(adev, 0, 0, 0, i); /* CP and shaders */ WREG32(mmSH_MEM_CONFIG, sh_mem_cfg); @@ -2147,26 +2147,6 @@ static void gfx_v7_0_ring_emit_vgt_flush(struct amdgpu_ring *ring) EVENT_INDEX(0)); } - -/** - * gfx_v7_0_ring_emit_hdp_invalidate - emit an hdp invalidate on the cp - * - * @adev: amdgpu_device pointer - * @ridx: amdgpu ring index - * - * Emits an hdp invalidate on the cp. - */ -static void gfx_v7_0_ring_emit_hdp_invalidate(struct amdgpu_ring *ring) -{ - amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); - amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) | - WRITE_DATA_DST_SEL(0) | - WR_CONFIRM)); - amdgpu_ring_write(ring, mmHDP_DEBUG0); - amdgpu_ring_write(ring, 0); - amdgpu_ring_write(ring, 1); -} - /** * gfx_v7_0_ring_emit_fence_gfx - emit a fence on the gfx ring * @@ -3243,26 +3223,7 @@ static void gfx_v7_0_ring_emit_vm_flush(struct amdgpu_ring *ring, { int usepfp = (ring->funcs->type == AMDGPU_RING_TYPE_GFX); - amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); - amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(usepfp) | - WRITE_DATA_DST_SEL(0))); - if (vmid < 8) { - amdgpu_ring_write(ring, - (mmVM_CONTEXT0_PAGE_TABLE_BASE_ADDR + vmid)); - } else { - amdgpu_ring_write(ring, - (mmVM_CONTEXT8_PAGE_TABLE_BASE_ADDR + vmid - 8)); - } - amdgpu_ring_write(ring, 0); - amdgpu_ring_write(ring, pd_addr >> 12); - - /* bits 0-15 are the VM contexts0-15 */ - amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); - amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(0) | - WRITE_DATA_DST_SEL(0))); - amdgpu_ring_write(ring, mmVM_INVALIDATE_REQUEST); - amdgpu_ring_write(ring, 0); - amdgpu_ring_write(ring, 1 << vmid); + amdgpu_gmc_emit_flush_gpu_tlb(ring, vmid, pd_addr); /* wait for the invalidate to complete */ amdgpu_ring_write(ring, PACKET3(PACKET3_WAIT_REG_MEM, 5)); @@ -3289,6 +3250,19 @@ static void gfx_v7_0_ring_emit_vm_flush(struct amdgpu_ring *ring, } } +static void gfx_v7_0_ring_emit_wreg(struct amdgpu_ring *ring, + uint32_t reg, uint32_t val) +{ + int usepfp = (ring->funcs->type == AMDGPU_RING_TYPE_GFX); + + amdgpu_ring_write(ring, PACKET3(PACKET3_WRITE_DATA, 3)); + amdgpu_ring_write(ring, (WRITE_DATA_ENGINE_SEL(usepfp) | + WRITE_DATA_DST_SEL(0))); + amdgpu_ring_write(ring, reg); + amdgpu_ring_write(ring, 0); + amdgpu_ring_write(ring, val); +} + /* * RLC * The RLC is a multi-purpose microengine that handles a @@ -5089,10 +5063,10 @@ static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_gfx = { .emit_frame_size = 20 + /* gfx_v7_0_ring_emit_gds_switch */ 7 + /* gfx_v7_0_ring_emit_hdp_flush */ - 5 + /* gfx_v7_0_ring_emit_hdp_invalidate */ + 5 + /* hdp invalidate */ 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 */ + CIK_FLUSH_GPU_TLB_NUM_WREG * 5 + 7 + 6 + /* gfx_v7_0_ring_emit_vm_flush */ 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, @@ -5101,12 +5075,12 @@ static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_gfx = { .emit_vm_flush = gfx_v7_0_ring_emit_vm_flush, .emit_gds_switch = gfx_v7_0_ring_emit_gds_switch, .emit_hdp_flush = gfx_v7_0_ring_emit_hdp_flush, - .emit_hdp_invalidate = gfx_v7_0_ring_emit_hdp_invalidate, .test_ring = gfx_v7_0_ring_test_ring, .test_ib = gfx_v7_0_ring_test_ib, .insert_nop = amdgpu_ring_insert_nop, .pad_ib = amdgpu_ring_generic_pad_ib, .emit_cntxcntl = gfx_v7_ring_emit_cntxcntl, + .emit_wreg = gfx_v7_0_ring_emit_wreg, }; static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_compute = { @@ -5120,9 +5094,9 @@ static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_compute = { .emit_frame_size = 20 + /* gfx_v7_0_ring_emit_gds_switch */ 7 + /* gfx_v7_0_ring_emit_hdp_flush */ - 5 + /* gfx_v7_0_ring_emit_hdp_invalidate */ + 5 + /* hdp invalidate */ 7 + /* gfx_v7_0_ring_emit_pipeline_sync */ - 17 + /* gfx_v7_0_ring_emit_vm_flush */ + CIK_FLUSH_GPU_TLB_NUM_WREG * 5 + 7 + /* gfx_v7_0_ring_emit_vm_flush */ 7 + 7 + 7, /* gfx_v7_0_ring_emit_fence_compute x3 for user fence, vm fence */ .emit_ib_size = 4, /* gfx_v7_0_ring_emit_ib_compute */ .emit_ib = gfx_v7_0_ring_emit_ib_compute, @@ -5131,11 +5105,11 @@ static const struct amdgpu_ring_funcs gfx_v7_0_ring_funcs_compute = { .emit_vm_flush = gfx_v7_0_ring_emit_vm_flush, .emit_gds_switch = gfx_v7_0_ring_emit_gds_switch, .emit_hdp_flush = gfx_v7_0_ring_emit_hdp_flush, - .emit_hdp_invalidate = gfx_v7_0_ring_emit_hdp_invalidate, .test_ring = gfx_v7_0_ring_test_ring, .test_ib = gfx_v7_0_ring_test_ib, .insert_nop = amdgpu_ring_insert_nop, .pad_ib = amdgpu_ring_generic_pad_ib, + .emit_wreg = gfx_v7_0_ring_emit_wreg, }; static void gfx_v7_0_set_ring_funcs(struct amdgpu_device *adev) |