#include "common/gen_l3_config.h"
#include "common/gen_sample_positions.h"
+#include "nir/nir_xfb_info.h"
#include "vk_util.h"
#include "vk_format_info.h"
__builtin_popcount(elements_double) / 2;
const uint32_t total_elems =
- elem_count + needs_svgs_elem + vs_prog_data->uses_drawid;
- if (total_elems == 0)
- return;
+ MAX2(1, elem_count + needs_svgs_elem + vs_prog_data->uses_drawid);
uint32_t *p;
sf.TriangleStripListProvokingVertexSelect = 0;
sf.LineStripListProvokingVertexSelect = 0;
sf.TriangleFanProvokingVertexSelect = 1;
+ sf.VertexSubPixelPrecisionSelect = _8Bit;
const struct brw_vue_prog_data *last_vue_prog_data =
anv_pipeline_get_last_vue_prog_data(pipeline);
#if GEN_GEN >= 9
/* GEN9+ splits ViewportZClipTestEnable into near and far enable bits */
- raster.ViewportZFarClipTestEnable = !pipeline->depth_clamp_enable;
- raster.ViewportZNearClipTestEnable = !pipeline->depth_clamp_enable;
+ raster.ViewportZFarClipTestEnable = pipeline->depth_clip_enable;
+ raster.ViewportZNearClipTestEnable = pipeline->depth_clip_enable;
#elif GEN_GEN >= 8
- raster.ViewportZClipTestEnable = !pipeline->depth_clamp_enable;
+ raster.ViewportZClipTestEnable = pipeline->depth_clip_enable;
#endif
raster.GlobalDepthOffsetEnableSolid = rs_info->depthBiasEnable;
{
*stencilWriteEnable = state->stencilTestEnable;
- /* If the depth test is disabled, we won't be writing anything. */
- if (!state->depthTestEnable)
- state->depthWriteEnable = false;
-
- /* The Vulkan spec requires that if either depth or stencil is not present,
- * the pipeline is to act as if the test silently passes.
+ /* If the depth test is disabled, we won't be writing anything. Make sure we
+ * treat the test as always passing later on as well.
+ *
+ * Also, the Vulkan spec requires that if either depth or stencil is not
+ * present, the pipeline is to act as if the test silently passes. In that
+ * case we won't write either.
*/
- if (!(ds_aspects & VK_IMAGE_ASPECT_DEPTH_BIT)) {
+ if (!state->depthTestEnable || !(ds_aspects & VK_IMAGE_ASPECT_DEPTH_BIT)) {
state->depthWriteEnable = false;
state->depthCompareOp = VK_COMPARE_OP_ALWAYS;
}
#endif
}
+MAYBE_UNUSED static bool
+is_dual_src_blend_factor(VkBlendFactor factor)
+{
+ return factor == VK_BLEND_FACTOR_SRC1_COLOR ||
+ factor == VK_BLEND_FACTOR_ONE_MINUS_SRC1_COLOR ||
+ factor == VK_BLEND_FACTOR_SRC1_ALPHA ||
+ factor == VK_BLEND_FACTOR_ONE_MINUS_SRC1_ALPHA;
+}
+
static void
emit_cb_state(struct anv_pipeline *pipeline,
const VkPipelineColorBlendStateCreateInfo *info,
const VkPipelineMultisampleStateCreateInfo *ms_info)
{
struct anv_device *device = pipeline->device;
-
+ const struct brw_wm_prog_data *wm_prog_data = get_wm_prog_data(pipeline);
struct GENX(BLEND_STATE) blend_state = {
#if GEN_GEN >= 8
#endif
}
+ /* The Dual Source Blending documentation says:
+ *
+ * "If SRC1 is included in a src/dst blend factor and
+ * a DualSource RT Write message is not used, results
+ * are UNDEFINED. (This reflects the same restriction in DX APIs,
+ * where undefined results are produced if “o1” is not written
+ * by a PS – there are no default values defined)."
+ *
+ * There is no way to gracefully fix this undefined situation
+ * so we just disable the blending to prevent possible issues.
+ */
+ if (!wm_prog_data->dual_src_blend &&
+ (is_dual_src_blend_factor(a->srcColorBlendFactor) ||
+ is_dual_src_blend_factor(a->dstColorBlendFactor) ||
+ is_dual_src_blend_factor(a->srcAlphaBlendFactor) ||
+ is_dual_src_blend_factor(a->dstAlphaBlendFactor))) {
+ vk_debug_report(&device->instance->debug_report_callbacks,
+ VK_DEBUG_REPORT_WARNING_BIT_EXT,
+ VK_DEBUG_REPORT_OBJECT_TYPE_DEVICE_EXT,
+ (uint64_t)(uintptr_t)device,
+ 0, 0, "anv",
+ "Enabled dual-src blend factors without writing both targets "
+ "in the shader. Disabling blending to avoid GPU hangs.");
+ entry.ColorBufferBlendEnable = false;
+ }
+
if (a->colorWriteMask != 0)
has_writeable_rt = true;
#endif
GENX(BLEND_STATE_pack)(NULL, pipeline->blend_state.map, &blend_state);
- anv_state_flush(device, pipeline->blend_state);
anv_batch_emit(&pipeline->batch, GENX(3DSTATE_BLEND_STATE_POINTERS), bsp) {
bsp.BlendStatePointer = pipeline->blend_state.offset;
clip.APIMode = APIMODE_D3D,
clip.ViewportXYClipTestEnable = true;
+#if GEN_GEN >= 8
+ clip.VertexSubPixelPrecisionSelect = _8Bit;
+#endif
+
clip.ClipMode = CLIPMODE_NORMAL;
clip.TriangleStripListProvokingVertexSelect = 0;
#if GEN_GEN == 7
clip.FrontWinding = vk_to_gen_front_face[rs_info->frontFace];
clip.CullMode = vk_to_gen_cullmode[rs_info->cullMode];
- clip.ViewportZClipTestEnable = !pipeline->depth_clamp_enable;
- if (last) {
- clip.UserClipDistanceClipTestEnableBitmask = last->clip_distance_mask;
- clip.UserClipDistanceCullTestEnableBitmask = last->cull_distance_mask;
- }
+ clip.ViewportZClipTestEnable = pipeline->depth_clip_enable;
+ clip.UserClipDistanceClipTestEnableBitmask = last->clip_distance_mask;
+ clip.UserClipDistanceCullTestEnableBitmask = last->cull_distance_mask;
#else
clip.NonPerspectiveBarycentricEnable = wm_prog_data ?
(wm_prog_data->barycentric_interp_modes &
emit_3dstate_streamout(struct anv_pipeline *pipeline,
const VkPipelineRasterizationStateCreateInfo *rs_info)
{
+#if GEN_GEN >= 8
+ const struct brw_vue_prog_data *prog_data =
+ anv_pipeline_get_last_vue_prog_data(pipeline);
+ const struct brw_vue_map *vue_map = &prog_data->vue_map;
+#endif
+
+ nir_xfb_info *xfb_info;
+ if (anv_pipeline_has_stage(pipeline, MESA_SHADER_GEOMETRY))
+ xfb_info = pipeline->shaders[MESA_SHADER_GEOMETRY]->xfb_info;
+ else if (anv_pipeline_has_stage(pipeline, MESA_SHADER_TESS_EVAL))
+ xfb_info = pipeline->shaders[MESA_SHADER_TESS_EVAL]->xfb_info;
+ else
+ xfb_info = pipeline->shaders[MESA_SHADER_VERTEX]->xfb_info;
+
+ pipeline->xfb_used = xfb_info ? xfb_info->buffers_written : 0;
+
anv_batch_emit(&pipeline->batch, GENX(3DSTATE_STREAMOUT), so) {
so.RenderingDisable = rs_info->rasterizerDiscardEnable;
+
+#if GEN_GEN >= 8
+ if (xfb_info) {
+ so.SOFunctionEnable = true;
+ so.SOStatisticsEnable = true;
+
+ const VkPipelineRasterizationStateStreamCreateInfoEXT *stream_info =
+ vk_find_struct_const(rs_info, PIPELINE_RASTERIZATION_STATE_STREAM_CREATE_INFO_EXT);
+ so.RenderStreamSelect = stream_info ?
+ stream_info->rasterizationStream : 0;
+
+ so.Buffer0SurfacePitch = xfb_info->buffers[0].stride;
+ so.Buffer1SurfacePitch = xfb_info->buffers[1].stride;
+ so.Buffer2SurfacePitch = xfb_info->buffers[2].stride;
+ so.Buffer3SurfacePitch = xfb_info->buffers[3].stride;
+
+ int urb_entry_read_offset = 0;
+ int urb_entry_read_length =
+ (prog_data->vue_map.num_slots + 1) / 2 - urb_entry_read_offset;
+
+ /* We always read the whole vertex. This could be reduced at some
+ * point by reading less and offsetting the register index in the
+ * SO_DECLs.
+ */
+ so.Stream0VertexReadOffset = urb_entry_read_offset;
+ so.Stream0VertexReadLength = urb_entry_read_length - 1;
+ so.Stream1VertexReadOffset = urb_entry_read_offset;
+ so.Stream1VertexReadLength = urb_entry_read_length - 1;
+ so.Stream2VertexReadOffset = urb_entry_read_offset;
+ so.Stream2VertexReadLength = urb_entry_read_length - 1;
+ so.Stream3VertexReadOffset = urb_entry_read_offset;
+ so.Stream3VertexReadLength = urb_entry_read_length - 1;
+ }
+#endif /* GEN_GEN >= 8 */
+ }
+
+#if GEN_GEN >= 8
+ if (xfb_info) {
+ struct GENX(SO_DECL) so_decl[MAX_XFB_STREAMS][128];
+ int next_offset[MAX_XFB_BUFFERS] = {0, 0, 0, 0};
+ int decls[MAX_XFB_STREAMS] = {0, 0, 0, 0};
+
+ memset(so_decl, 0, sizeof(so_decl));
+
+ for (unsigned i = 0; i < xfb_info->output_count; i++) {
+ const nir_xfb_output_info *output = &xfb_info->outputs[i];
+ unsigned buffer = output->buffer;
+ unsigned stream = xfb_info->buffer_to_stream[buffer];
+
+ /* Our hardware is unusual in that it requires us to program SO_DECLs
+ * for fake "hole" components, rather than simply taking the offset
+ * for each real varying. Each hole can have size 1, 2, 3, or 4; we
+ * program as many size = 4 holes as we can, then a final hole to
+ * accommodate the final 1, 2, or 3 remaining.
+ */
+ int hole_dwords = (output->offset - next_offset[buffer]) / 4;
+ while (hole_dwords > 0) {
+ so_decl[stream][decls[stream]++] = (struct GENX(SO_DECL)) {
+ .HoleFlag = 1,
+ .OutputBufferSlot = buffer,
+ .ComponentMask = (1 << MIN2(hole_dwords, 4)) - 1,
+ };
+ hole_dwords -= 4;
+ }
+
+ int varying = output->location;
+ uint8_t component_mask = output->component_mask;
+ /* VARYING_SLOT_PSIZ contains three scalar fields packed together:
+ * - VARYING_SLOT_LAYER in VARYING_SLOT_PSIZ.y
+ * - VARYING_SLOT_VIEWPORT in VARYING_SLOT_PSIZ.z
+ * - VARYING_SLOT_PSIZ in VARYING_SLOT_PSIZ.w
+ */
+ if (varying == VARYING_SLOT_LAYER) {
+ varying = VARYING_SLOT_PSIZ;
+ component_mask = 1 << 1; // SO_DECL_COMPMASK_Y
+ } else if (varying == VARYING_SLOT_VIEWPORT) {
+ varying = VARYING_SLOT_PSIZ;
+ component_mask = 1 << 2; // SO_DECL_COMPMASK_Z
+ } else if (varying == VARYING_SLOT_PSIZ) {
+ component_mask = 1 << 3; // SO_DECL_COMPMASK_W
+ }
+
+ next_offset[buffer] = output->offset +
+ __builtin_popcount(component_mask) * 4;
+
+ so_decl[stream][decls[stream]++] = (struct GENX(SO_DECL)) {
+ .OutputBufferSlot = buffer,
+ .RegisterIndex = vue_map->varying_to_slot[varying],
+ .ComponentMask = component_mask,
+ };
+ }
+
+ int max_decls = 0;
+ for (unsigned s = 0; s < MAX_XFB_STREAMS; s++)
+ max_decls = MAX2(max_decls, decls[s]);
+
+ uint8_t sbs[MAX_XFB_STREAMS] = { };
+ for (unsigned b = 0; b < MAX_XFB_BUFFERS; b++) {
+ if (xfb_info->buffers_written & (1 << b))
+ sbs[xfb_info->buffer_to_stream[b]] |= 1 << b;
+ }
+
+ uint32_t *dw = anv_batch_emitn(&pipeline->batch, 3 + 2 * max_decls,
+ GENX(3DSTATE_SO_DECL_LIST),
+ .StreamtoBufferSelects0 = sbs[0],
+ .StreamtoBufferSelects1 = sbs[1],
+ .StreamtoBufferSelects2 = sbs[2],
+ .StreamtoBufferSelects3 = sbs[3],
+ .NumEntries0 = decls[0],
+ .NumEntries1 = decls[1],
+ .NumEntries2 = decls[2],
+ .NumEntries3 = decls[3]);
+
+ for (int i = 0; i < max_decls; i++) {
+ GENX(SO_DECL_ENTRY_pack)(NULL, dw + 3 + i * 2,
+ &(struct GENX(SO_DECL_ENTRY)) {
+ .Stream0Decl = so_decl[0][i],
+ .Stream1Decl = so_decl[1][i],
+ .Stream2Decl = so_decl[2][i],
+ .Stream3Decl = so_decl[3][i],
+ });
+ }
}
+#endif /* GEN_GEN >= 8 */
}
static uint32_t
vs.SingleVertexDispatch = false;
#endif
vs.VectorMaskEnable = false;
- vs.SamplerCount = get_sampler_count(vs_bin);
+ /* WA_1606682166:
+ * Incorrect TDL's SSP address shift in SARB for 16:6 & 18:8 modes.
+ * Disable the Sampler state prefetch functionality in the SARB by
+ * programming 0xB000[30] to '1'.
+ */
+ vs.SamplerCount = GEN_GEN == 11 ? 0 : get_sampler_count(vs_bin);
/* Gen 11 workarounds table #2056 WABTPPrefetchDisable suggests to
* disable prefetching of binding tables on A0 and B0 steppings.
* TODO: Revisit this WA on newer steppings.
vs.IllegalOpcodeExceptionEnable = false;
vs.SoftwareExceptionEnable = false;
vs.MaximumNumberofThreads = devinfo->max_vs_threads - 1;
- vs.VertexCacheDisable = false;
+
+ if (GEN_GEN == 9 && devinfo->gt == 4 &&
+ anv_pipeline_has_stage(pipeline, MESA_SHADER_TESS_EVAL)) {
+ /* On Sky Lake GT4, we have experienced some hangs related to the VS
+ * cache and tessellation. It is unknown exactly what is happening
+ * but the Haswell docs for the "VS Reference Count Full Force Miss
+ * Enable" field of the "Thread Mode" register refer to a HSW bug in
+ * which the VUE handle reference count would overflow resulting in
+ * internal reference counting bugs. My (Jason's) best guess is that
+ * this bug cropped back up on SKL GT4 when we suddenly had more
+ * threads in play than any previous gen9 hardware.
+ *
+ * What we do know for sure is that setting this bit when
+ * tessellation shaders are in use fixes a GPU hang in Batman: Arkham
+ * City when playing with DXVK (https://bugs.freedesktop.org/107280).
+ * Disabling the vertex cache with tessellation shaders should only
+ * have a minor performance impact as the tessellation shaders are
+ * likely generating and processing far more geometry than the vertex
+ * stage.
+ */
+ vs.VertexCacheDisable = true;
+ }
vs.VertexURBEntryReadLength = vs_prog_data->base.urb_read_length;
vs.VertexURBEntryReadOffset = 0;
hs.Enable = true;
hs.StatisticsEnable = true;
hs.KernelStartPointer = tcs_bin->kernel.offset;
-
- hs.SamplerCount = get_sampler_count(tcs_bin);
+ /* WA_1606682166 */
+ hs.SamplerCount = GEN_GEN == 11 ? 0 : get_sampler_count(tcs_bin);
/* Gen 11 workarounds table #2056 WABTPPrefetchDisable */
hs.BindingTableEntryCount = GEN_GEN == 11 ? 0 : get_binding_table_entry_count(tcs_bin);
hs.MaximumNumberofThreads = devinfo->max_tcs_threads - 1;
hs.PerThreadScratchSpace = get_scratch_space(tcs_bin);
hs.ScratchSpaceBasePointer =
get_scratch_address(pipeline, MESA_SHADER_TESS_CTRL, tcs_bin);
+
+#if GEN_GEN >= 9
+ hs.DispatchMode = tcs_prog_data->base.dispatch_mode;
+ hs.IncludePrimitiveID = tcs_prog_data->include_primitive_id;
+#endif
}
- const VkPipelineTessellationDomainOriginStateCreateInfoKHR *domain_origin_state =
- tess_info ? vk_find_struct_const(tess_info, PIPELINE_TESSELLATION_DOMAIN_ORIGIN_STATE_CREATE_INFO_KHR) : NULL;
+ const VkPipelineTessellationDomainOriginStateCreateInfo *domain_origin_state =
+ tess_info ? vk_find_struct_const(tess_info, PIPELINE_TESSELLATION_DOMAIN_ORIGIN_STATE_CREATE_INFO) : NULL;
- VkTessellationDomainOriginKHR uv_origin =
+ VkTessellationDomainOrigin uv_origin =
domain_origin_state ? domain_origin_state->domainOrigin :
- VK_TESSELLATION_DOMAIN_ORIGIN_UPPER_LEFT_KHR;
+ VK_TESSELLATION_DOMAIN_ORIGIN_UPPER_LEFT;
anv_batch_emit(&pipeline->batch, GENX(3DSTATE_TE), te) {
te.Partitioning = tes_prog_data->partitioning;
- if (uv_origin == VK_TESSELLATION_DOMAIN_ORIGIN_LOWER_LEFT_KHR) {
+ if (uv_origin == VK_TESSELLATION_DOMAIN_ORIGIN_LOWER_LEFT) {
te.OutputTopology = tes_prog_data->output_topology;
} else {
/* When the origin is upper-left, we have to flip the winding order */
ds.Enable = true;
ds.StatisticsEnable = true;
ds.KernelStartPointer = tes_bin->kernel.offset;
-
- ds.SamplerCount = get_sampler_count(tes_bin);
+ /* WA_1606682166 */
+ ds.SamplerCount = GEN_GEN == 11 ? 0 : get_sampler_count(tes_bin);
/* Gen 11 workarounds table #2056 WABTPPrefetchDisable */
ds.BindingTableEntryCount = GEN_GEN == 11 ? 0 : get_binding_table_entry_count(tes_bin);
ds.MaximumNumberofThreads = devinfo->max_tes_threads - 1;
gs.SingleProgramFlow = false;
gs.VectorMaskEnable = false;
- gs.SamplerCount = get_sampler_count(gs_bin);
+ /* WA_1606682166 */
+ gs.SamplerCount = GEN_GEN == 11 ? 0 : get_sampler_count(gs_bin);
/* Gen 11 workarounds table #2056 WABTPPrefetchDisable */
gs.BindingTableEntryCount = GEN_GEN == 11 ? 0 : get_binding_table_entry_count(gs_bin);
gs.IncludeVertexHandles = gs_prog_data->base.include_vue_handles;
}
}
-UNUSED static bool
-is_dual_src_blend_factor(VkBlendFactor factor)
-{
- return factor == VK_BLEND_FACTOR_SRC1_COLOR ||
- factor == VK_BLEND_FACTOR_ONE_MINUS_SRC1_COLOR ||
- factor == VK_BLEND_FACTOR_SRC1_ALPHA ||
- factor == VK_BLEND_FACTOR_ONE_MINUS_SRC1_ALPHA;
-}
-
static void
emit_3dstate_ps(struct anv_pipeline *pipeline,
const VkPipelineColorBlendStateCreateInfo *blend,
brw_wm_prog_data_prog_offset(wm_prog_data, ps, 2);
ps.SingleProgramFlow = false;
- ps.VectorMaskEnable = true;
- ps.SamplerCount = get_sampler_count(fs_bin);
+ ps.VectorMaskEnable = GEN_GEN >= 8;
+ /* WA_1606682166 */
+ ps.SamplerCount = GEN_GEN == 11 ? 0 : get_sampler_count(fs_bin);
/* Gen 11 workarounds table #2056 WABTPPrefetchDisable */
ps.BindingTableEntryCount = GEN_GEN == 11 ? 0 : get_binding_table_entry_count(fs_bin);
ps.PushConstantEnable = wm_prog_data->base.nr_params > 0 ||
vfe.URBEntryAllocationSize = GEN_GEN <= 7 ? 0 : 2;
vfe.CURBEAllocationSize = vfe_curbe_allocation;
- vfe.PerThreadScratchSpace = get_scratch_space(cs_bin);
- vfe.ScratchSpaceBasePointer =
- get_scratch_address(pipeline, MESA_SHADER_COMPUTE, cs_bin);
+ if (cs_bin->prog_data->total_scratch) {
+ if (GEN_GEN >= 8) {
+ /* Broadwell's Per Thread Scratch Space is in the range [0, 11]
+ * where 0 = 1k, 1 = 2k, 2 = 4k, ..., 11 = 2M.
+ */
+ vfe.PerThreadScratchSpace =
+ ffs(cs_bin->prog_data->total_scratch) - 11;
+ } else if (GEN_IS_HASWELL) {
+ /* Haswell's Per Thread Scratch Space is in the range [0, 10]
+ * where 0 = 2k, 1 = 4k, 2 = 8k, ..., 10 = 2M.
+ */
+ vfe.PerThreadScratchSpace =
+ ffs(cs_bin->prog_data->total_scratch) - 12;
+ } else {
+ /* IVB and BYT use the range [0, 11] to mean [1kB, 12kB]
+ * where 0 = 1kB, 1 = 2kB, 2 = 3kB, ..., 11 = 12kB.
+ */
+ vfe.PerThreadScratchSpace =
+ cs_bin->prog_data->total_scratch / 1024 - 1;
+ }
+ vfe.ScratchSpaceBasePointer =
+ get_scratch_address(pipeline, MESA_SHADER_COMPUTE, cs_bin);
+ }
}
struct GENX(INTERFACE_DESCRIPTOR_DATA) desc = {
.KernelStartPointer = cs_bin->kernel.offset,
-
- .SamplerCount = get_sampler_count(cs_bin),
- /* Gen 11 workarounds table #2056 WABTPPrefetchDisable */
- .BindingTableEntryCount = GEN_GEN == 11 ? 0 : get_binding_table_entry_count(cs_bin),
+ /* WA_1606682166 */
+ .SamplerCount = GEN_GEN == 11 ? 0 : get_sampler_count(cs_bin),
+ /* Gen 11 workarounds table #2056 WABTPPrefetchDisable
+ *
+ * We add 1 because the CS indirect parameters buffer isn't accounted
+ * for in bind_map.surface_count.
+ */
+ .BindingTableEntryCount = GEN_GEN == 11 ? 0 : 1 + MIN2(cs_bin->bind_map.surface_count, 30),
.BarrierEnable = cs_prog_data->uses_barrier,
.SharedLocalMemorySize =
encode_slm_size(GEN_GEN, cs_prog_data->base.total_shared),