#include "main/state.h"
#include "brw_context.h"
-#if GEN_GEN == 6
-#include "brw_defines.h"
-#endif
#include "brw_draw.h"
#include "brw_multisample_state.h"
#include "brw_state.h"
struct brw_address {
struct brw_bo *bo;
- uint32_t read_domains;
- uint32_t write_domain;
+ unsigned reloc_flags;
uint32_t offset;
};
-static uint64_t
-emit_reloc(struct brw_context *brw,
- void *location, struct brw_address address, uint32_t delta)
-{
- uint32_t offset = (char *) location - (char *) brw->batch.map;
-
- return brw_emit_reloc(&brw->batch, offset, address.bo,
- address.offset + delta,
- address.read_domains,
- address.write_domain);
-}
-
#define __gen_address_type struct brw_address
#define __gen_user_data struct brw_context
__gen_combine_address(struct brw_context *brw, void *location,
struct brw_address address, uint32_t delta)
{
+ struct intel_batchbuffer *batch = &brw->batch;
+ uint32_t offset;
+
if (address.bo == NULL) {
return address.offset + delta;
} else {
- return emit_reloc(brw, location, address, delta);
- }
-}
+ if (GEN_GEN < 6 && brw_ptr_in_state_buffer(batch, location)) {
+ offset = (char *) location - (char *) brw->batch.state.map;
+ return brw_state_reloc(batch, offset, address.bo,
+ address.offset + delta,
+ address.reloc_flags);
+ }
-static inline struct brw_address
-render_bo(struct brw_bo *bo, uint32_t offset)
-{
- return (struct brw_address) {
- .bo = bo,
- .offset = offset,
- .read_domains = I915_GEM_DOMAIN_RENDER,
- .write_domain = I915_GEM_DOMAIN_RENDER,
- };
-}
+ assert(!brw_ptr_in_state_buffer(batch, location));
-static inline struct brw_address
-render_ro_bo(struct brw_bo *bo, uint32_t offset)
-{
- return (struct brw_address) {
- .bo = bo,
- .offset = offset,
- .read_domains = I915_GEM_DOMAIN_RENDER,
- .write_domain = 0,
- };
+ offset = (char *) location - (char *) brw->batch.batch.map;
+ return brw_batch_reloc(batch, offset, address.bo,
+ address.offset + delta,
+ address.reloc_flags);
+ }
}
-static inline struct brw_address
-instruction_bo(struct brw_bo *bo, uint32_t offset)
+static struct brw_address
+rw_bo(struct brw_bo *bo, uint32_t offset)
{
return (struct brw_address) {
.bo = bo,
.offset = offset,
- .read_domains = I915_GEM_DOMAIN_INSTRUCTION,
- .write_domain = I915_GEM_DOMAIN_INSTRUCTION,
+ .reloc_flags = RELOC_WRITE,
};
}
-static inline struct brw_address
-instruction_ro_bo(struct brw_bo *bo, uint32_t offset)
+static struct brw_address
+ro_bo(struct brw_bo *bo, uint32_t offset)
{
return (struct brw_address) {
.bo = bo,
.offset = offset,
- .read_domains = I915_GEM_DOMAIN_INSTRUCTION,
- .write_domain = 0,
};
}
-static inline struct brw_address
-vertex_bo(struct brw_bo *bo, uint32_t offset)
+UNUSED static struct brw_address
+ggtt_bo(struct brw_bo *bo, uint32_t offset)
{
return (struct brw_address) {
.bo = bo,
.offset = offset,
- .read_domains = I915_GEM_DOMAIN_VERTEX,
- .write_domain = 0,
+ .reloc_flags = RELOC_WRITE | RELOC_NEEDS_GGTT,
};
}
#if GEN_GEN == 4
-static inline struct brw_address
+static struct brw_address
KSP(struct brw_context *brw, uint32_t offset)
{
- return instruction_bo(brw->cache.bo, offset);
-}
-
-static inline struct brw_address
-KSP_ro(struct brw_context *brw, uint32_t offset)
-{
- return instruction_ro_bo(brw->cache.bo, offset);
+ return ro_bo(brw->cache.bo, offset);
}
#else
-static inline uint32_t
+static uint32_t
KSP(struct brw_context *brw, uint32_t offset)
{
return offset;
}
-
-#define KSP_ro KSP
-
#endif
#include "genxml/genX_pack.h"
})
#define brw_state_emit(brw, cmd, align, offset, name) \
- for (struct cmd name = { 0, }, \
+ for (struct cmd name = {}, \
*_dst = brw_state_batch(brw, _brw_cmd_length(cmd) * 4, \
align, offset); \
__builtin_expect(_dst != NULL, 1); \
struct GENX(VERTEX_BUFFER_STATE) buf_state = {
.VertexBufferIndex = buffer_nr,
.BufferPitch = stride,
- .BufferStartingAddress = vertex_bo(bo, start_offset),
+ .BufferStartingAddress = ro_bo(bo, start_offset),
#if GEN_GEN >= 8
.BufferSize = end_offset - start_offset,
#endif
.BufferAccessType = step_rate ? INSTANCEDATA : VERTEXDATA,
.InstanceDataStepRate = step_rate,
#if GEN_GEN >= 5
- .EndAddress = vertex_bo(bo, end_offset - 1),
+ .EndAddress = ro_bo(bo, end_offset - 1),
#endif
#endif
static void
genX(emit_vertices)(struct brw_context *brw)
{
+ const struct gen_device_info *devinfo = &brw->screen->devinfo;
uint32_t *dw;
brw_prepare_vertices(brw);
1 + GENX(VERTEX_ELEMENT_STATE_length));
struct GENX(VERTEX_ELEMENT_STATE) elem = {
.Valid = true,
- .SourceElementFormat = ISL_FORMAT_R32G32B32A32_FLOAT,
+ .SourceElementFormat = (enum GENX(SURFACE_FORMAT)) ISL_FORMAT_R32G32B32A32_FLOAT,
.Component0Control = VFCOMP_STORE_0,
.Component1Control = VFCOMP_STORE_0,
.Component2Control = VFCOMP_STORE_0,
* vertex element may poke over the end of the buffer by 2 bytes.
*/
const unsigned padding =
- (GEN_GEN <= 7 && !brw->is_baytrail && !brw->is_haswell) * 2;
+ (GEN_GEN <= 7 && !GEN_IS_HASWELL && !devinfo->is_baytrail) * 2;
const unsigned end = buffer->offset + buffer->size + padding;
dw = genX(emit_vertex_buffer_state)(brw, dw, i, buffer->bo,
buffer->offset,
if (vs_prog_data->uses_basevertex ||
vs_prog_data->uses_baseinstance) {
elem_state.VertexBufferIndex = brw->vb.nr_buffers;
- elem_state.SourceElementFormat = ISL_FORMAT_R32G32_UINT;
+ elem_state.SourceElementFormat = (enum GENX(SURFACE_FORMAT)) ISL_FORMAT_R32G32_UINT;
elem_state.Component0Control = VFCOMP_STORE_SRC;
elem_state.Component1Control = VFCOMP_STORE_SRC;
}
#else
elem_state.VertexBufferIndex = brw->vb.nr_buffers;
- elem_state.SourceElementFormat = ISL_FORMAT_R32G32_UINT;
+ elem_state.SourceElementFormat = (enum GENX(SURFACE_FORMAT)) ISL_FORMAT_R32G32_UINT;
if (vs_prog_data->uses_basevertex)
elem_state.Component0Control = VFCOMP_STORE_SRC;
struct GENX(VERTEX_ELEMENT_STATE) elem_state = {
.Valid = true,
.VertexBufferIndex = brw->vb.nr_buffers + 1,
- .SourceElementFormat = ISL_FORMAT_R32_UINT,
+ .SourceElementFormat = (enum GENX(SURFACE_FORMAT)) ISL_FORMAT_R32_UINT,
.Component0Control = VFCOMP_STORE_SRC,
.Component1Control = VFCOMP_STORE_0,
.Component2Control = VFCOMP_STORE_0,
ib.CutIndexEnable = brw->prim_restart.enable_cut_index;
#endif
ib.IndexFormat = brw_get_index_type(index_buffer->index_size);
- ib.BufferStartingAddress = vertex_bo(brw->ib.bo, 0);
+ ib.BufferStartingAddress = ro_bo(brw->ib.bo, 0);
#if GEN_GEN >= 8
ib.IndexBufferMOCS = GEN_GEN >= 9 ? SKL_MOCS_WB : BDW_MOCS_WB;
ib.BufferSize = brw->ib.size;
#else
- ib.BufferEndingAddress = vertex_bo(brw->ib.bo, brw->ib.size - 1);
+ ib.BufferEndingAddress = ro_bo(brw->ib.bo, brw->ib.size - 1);
#endif
}
}
/* _NEW_POINT */
const struct gl_point_attrib *point = &ctx->Point;
+ /* BRW_NEW_FRAGMENT_PROGRAM */
+ const struct gl_program *fp = brw->programs[MESA_SHADER_FRAGMENT];
+
/* BRW_NEW_FS_PROG_DATA */
const struct brw_wm_prog_data *wm_prog_data =
brw_wm_prog_data(brw->wm.base.prog_data);
*point_sprite_enables = 0;
- /* BRW_NEW_FRAGMENT_PROGRAM
- *
- * If the fragment shader reads VARYING_SLOT_LAYER, then we need to pass in
- * the full vertex header. Otherwise, we can program the SF to start
- * reading at an offset of 1 (2 varying slots) to skip unnecessary data:
- * - VARYING_SLOT_PSIZ and BRW_VARYING_SLOT_NDC on gen4-5
- * - VARYING_SLOT_{PSIZ,LAYER} and VARYING_SLOT_POS on gen6+
- */
-
- bool fs_needs_vue_header = brw->fragment_program->info.inputs_read &
- (VARYING_BIT_LAYER | VARYING_BIT_VIEWPORT);
+ int first_slot =
+ brw_compute_first_urb_slot_required(fp->info.inputs_read,
+ &brw->vue_map_geom_out);
- *urb_entry_read_offset = fs_needs_vue_header ? 0 : 1;
+ /* Each URB offset packs two varying slots */
+ assert(first_slot % 2 == 0);
+ *urb_entry_read_offset = first_slot / 2;
/* From the Ivybridge PRM, Vol 2 Part 1, 3DSTATE_SBE,
* description of dw10 Point Sprite Texture Coordinate Enable:
/* ---------------------------------------------------------------------- */
-#if GEN_GEN >= 6
+#if GEN_GEN <= 5
+
+static void
+genX(upload_clip_state)(struct brw_context *brw)
+{
+ struct gl_context *ctx = &brw->ctx;
+
+ ctx->NewDriverState |= BRW_NEW_GEN4_UNIT_STATE;
+ brw_state_emit(brw, GENX(CLIP_STATE), 32, &brw->clip.state_offset, clip) {
+ clip.KernelStartPointer = KSP(brw, brw->clip.prog_offset);
+ clip.GRFRegisterCount =
+ DIV_ROUND_UP(brw->clip.prog_data->total_grf, 16) - 1;
+ clip.FloatingPointMode = FLOATING_POINT_MODE_Alternate;
+ clip.SingleProgramFlow = true;
+ clip.VertexURBEntryReadLength = brw->clip.prog_data->urb_read_length;
+ clip.ConstantURBEntryReadLength = brw->clip.prog_data->curb_read_length;
+
+ /* BRW_NEW_PUSH_CONSTANT_ALLOCATION */
+ clip.ConstantURBEntryReadOffset = brw->curbe.clip_start * 2;
+ clip.DispatchGRFStartRegisterForURBData = 1;
+ clip.VertexURBEntryReadOffset = 0;
+
+ /* BRW_NEW_URB_FENCE */
+ clip.NumberofURBEntries = brw->urb.nr_clip_entries;
+ clip.URBEntryAllocationSize = brw->urb.vsize - 1;
+
+ if (brw->urb.nr_clip_entries >= 10) {
+ /* Half of the URB entries go to each thread, and it has to be an
+ * even number.
+ */
+ assert(brw->urb.nr_clip_entries % 2 == 0);
+
+ /* Although up to 16 concurrent Clip threads are allowed on Ironlake,
+ * only 2 threads can output VUEs at a time.
+ */
+ clip.MaximumNumberofThreads = (GEN_GEN == 5 ? 16 : 2) - 1;
+ } else {
+ assert(brw->urb.nr_clip_entries >= 5);
+ clip.MaximumNumberofThreads = 1 - 1;
+ }
+
+ clip.VertexPositionSpace = VPOS_NDCSPACE;
+ clip.UserClipFlagsMustClipEnable = true;
+ clip.GuardbandClipTestEnable = true;
+
+ clip.ClipperViewportStatePointer =
+ ro_bo(brw->batch.state.bo, brw->clip.vp_offset);
+
+ clip.ScreenSpaceViewportXMin = -1;
+ clip.ScreenSpaceViewportXMax = 1;
+ clip.ScreenSpaceViewportYMin = -1;
+ clip.ScreenSpaceViewportYMax = 1;
+
+ clip.ViewportXYClipTestEnable = true;
+ clip.ViewportZClipTestEnable = !ctx->Transform.DepthClamp;
+
+ /* _NEW_TRANSFORM */
+ if (GEN_GEN == 5 || GEN_IS_G4X) {
+ clip.UserClipDistanceClipTestEnableBitmask =
+ ctx->Transform.ClipPlanesEnabled;
+ } else {
+ /* Up to 6 actual clip flags, plus the 7th for the negative RHW
+ * workaround.
+ */
+ clip.UserClipDistanceClipTestEnableBitmask =
+ (ctx->Transform.ClipPlanesEnabled & 0x3f) | 0x40;
+ }
+
+ if (ctx->Transform.ClipDepthMode == GL_ZERO_TO_ONE)
+ clip.APIMode = APIMODE_D3D;
+ else
+ clip.APIMode = APIMODE_OGL;
+
+ clip.GuardbandClipTestEnable = true;
+
+ clip.ClipMode = brw->clip.prog_data->clip_mode;
+
+#if GEN_IS_G4X
+ clip.NegativeWClipTestEnable = true;
+#endif
+ }
+}
+
+const struct brw_tracked_state genX(clip_state) = {
+ .dirty = {
+ .mesa = _NEW_TRANSFORM |
+ _NEW_VIEWPORT,
+ .brw = BRW_NEW_BATCH |
+ BRW_NEW_BLORP |
+ BRW_NEW_CLIP_PROG_DATA |
+ BRW_NEW_PUSH_CONSTANT_ALLOCATION |
+ BRW_NEW_PROGRAM_CACHE |
+ BRW_NEW_URB_FENCE,
+ },
+ .emit = genX(upload_clip_state),
+};
+
+#else
+
static void
genX(upload_clip_state)(struct brw_context *brw)
{
ctx->NewDriverState |= BRW_NEW_GEN4_UNIT_STATE;
brw_state_emit(brw, GENX(SF_STATE), 64, &brw->sf.state_offset, sf) {
- sf.KernelStartPointer = KSP_ro(brw, brw->sf.prog_offset);
+ sf.KernelStartPointer = KSP(brw, brw->sf.prog_offset);
sf.FloatingPointMode = FLOATING_POINT_MODE_Alternate;
sf.GRFRegisterCount = DIV_ROUND_UP(sf_prog_data->total_grf, 16) - 1;
sf.DispatchGRFStartRegisterForURBData = 3;
* domain.
*/
sf.SetupViewportStateOffset =
- instruction_ro_bo(brw->batch.bo, brw->sf.vp_offset);
+ ro_bo(brw->batch.state.bo, brw->sf.vp_offset);
sf.PointRasterizationRule = RASTRULE_UPPER_RIGHT;
/* _NEW_LINE */
#if GEN_GEN == 8
- if (brw->is_cherryview)
+ const struct gen_device_info *devinfo = &brw->screen->devinfo;
+
+ if (devinfo->is_cherryview)
sf.CHVLineWidth = brw_get_line_width(brw);
else
sf.LineWidth = brw_get_line_width(brw);
sf.SmoothPointEnable = true;
#endif
+#if GEN_GEN == 10
+ /* _NEW_BUFFERS
+ * Smooth Point Enable bit MUST not be set when NUM_MULTISAMPLES > 1.
+ */
+ const bool multisampled_fbo =
+ _mesa_geometric_samples(ctx->DrawBuffer) > 1;
+ if (multisampled_fbo)
+ sf.SmoothPointEnable = false;
+#endif
+
#if GEN_IS_G4X || GEN_GEN >= 5
sf.AALineDistanceMode = AALINEDISTANCE_TRUE;
#endif
_NEW_POINT |
_NEW_PROGRAM |
(GEN_GEN >= 6 ? _NEW_MULTISAMPLE : 0) |
- (GEN_GEN <= 7 ? _NEW_BUFFERS | _NEW_POLYGON : 0),
+ (GEN_GEN <= 7 ? _NEW_BUFFERS | _NEW_POLYGON : 0) |
+ (GEN_GEN == 10 ? _NEW_BUFFERS : 0),
.brw = BRW_NEW_BLORP |
BRW_NEW_VUE_MAP_GEOM_OUT |
(GEN_GEN <= 5 ? BRW_NEW_BATCH |
/* ---------------------------------------------------------------------- */
-#if GEN_GEN >= 6
+static bool
+brw_color_buffer_write_enabled(struct brw_context *brw)
+{
+ struct gl_context *ctx = &brw->ctx;
+ /* BRW_NEW_FRAGMENT_PROGRAM */
+ const struct gl_program *fp = brw->programs[MESA_SHADER_FRAGMENT];
+ unsigned i;
+
+ /* _NEW_BUFFERS */
+ for (i = 0; i < ctx->DrawBuffer->_NumColorDrawBuffers; i++) {
+ struct gl_renderbuffer *rb = ctx->DrawBuffer->_ColorDrawBuffers[i];
+ uint64_t outputs_written = fp->info.outputs_written;
+
+ /* _NEW_COLOR */
+ if (rb && (outputs_written & BITFIELD64_BIT(FRAG_RESULT_COLOR) ||
+ outputs_written & BITFIELD64_BIT(FRAG_RESULT_DATA0 + i)) &&
+ (ctx->Color.ColorMask[i][0] ||
+ ctx->Color.ColorMask[i][1] ||
+ ctx->Color.ColorMask[i][2] ||
+ ctx->Color.ColorMask[i][3])) {
+ return true;
+ }
+ }
+
+ return false;
+}
+
static void
genX(upload_wm)(struct brw_context *brw)
{
UNUSED bool writes_depth =
wm_prog_data->computed_depth_mode != BRW_PSCDEPTH_OFF;
+ UNUSED struct brw_stage_state *stage_state = &brw->wm.base;
+ UNUSED const struct gen_device_info *devinfo = &brw->screen->devinfo;
-#if GEN_GEN < 7
- const struct brw_stage_state *stage_state = &brw->wm.base;
- const struct gen_device_info *devinfo = &brw->screen->devinfo;
-
+#if GEN_GEN == 6
/* We can't fold this into gen6_upload_wm_push_constants(), because
* according to the SNB PRM, vol 2 part 1 section 7.2.2
* (3DSTATE_CONSTANT_PS [DevSNB]):
}
#endif
+#if GEN_GEN >= 6
brw_batch_emit(brw, GENX(3DSTATE_WM), wm) {
- wm.StatisticsEnable = true;
wm.LineAntialiasingRegionWidth = _10pixels;
wm.LineEndCapAntialiasingRegionWidth = _05pixels;
+ wm.PointRasterizationRule = RASTRULE_UPPER_RIGHT;
+ wm.BarycentricInterpolationMode = wm_prog_data->barycentric_interp_modes;
+#else
+ ctx->NewDriverState |= BRW_NEW_GEN4_UNIT_STATE;
+ brw_state_emit(brw, GENX(WM_STATE), 64, &stage_state->state_offset, wm) {
+ if (wm_prog_data->dispatch_8 && wm_prog_data->dispatch_16) {
+ /* These two fields should be the same pre-gen6, which is why we
+ * only have one hardware field to program for both dispatch
+ * widths.
+ */
+ assert(wm_prog_data->base.dispatch_grf_start_reg ==
+ wm_prog_data->dispatch_grf_start_reg_2);
+ }
+
+ if (wm_prog_data->dispatch_8 || wm_prog_data->dispatch_16)
+ wm.GRFRegisterCount0 = wm_prog_data->reg_blocks_0;
+
+ if (stage_state->sampler_count)
+ wm.SamplerStatePointer =
+ ro_bo(brw->batch.state.bo, stage_state->sampler_offset);
+#if GEN_GEN == 5
+ if (wm_prog_data->prog_offset_2)
+ wm.GRFRegisterCount2 = wm_prog_data->reg_blocks_2;
+#endif
+
+ wm.SetupURBEntryReadLength = wm_prog_data->num_varying_inputs * 2;
+ wm.ConstantURBEntryReadLength = wm_prog_data->base.curb_read_length;
+ /* BRW_NEW_PUSH_CONSTANT_ALLOCATION */
+ wm.ConstantURBEntryReadOffset = brw->curbe.wm_start * 2;
+ wm.EarlyDepthTestEnable = true;
+ wm.LineAntialiasingRegionWidth = _05pixels;
+ wm.LineEndCapAntialiasingRegionWidth = _10pixels;
+
+ /* _NEW_POLYGON */
+ if (ctx->Polygon.OffsetFill) {
+ wm.GlobalDepthOffsetEnable = true;
+ /* Something weird going on with legacy_global_depth_bias,
+ * offset_constant, scaling and MRD. This value passes glean
+ * but gives some odd results elsewere (eg. the
+ * quad-offset-units test).
+ */
+ wm.GlobalDepthOffsetConstant = ctx->Polygon.OffsetUnits * 2;
+
+ /* This is the only value that passes glean:
+ */
+ wm.GlobalDepthOffsetScale = ctx->Polygon.OffsetFactor;
+ }
+
+ wm.DepthCoefficientURBReadOffset = 1;
+#endif
+
+ /* BRW_NEW_STATS_WM */
+ wm.StatisticsEnable = GEN_GEN >= 6 || brw->stats_wm;
+
#if GEN_GEN < 7
if (wm_prog_data->base.use_alt_mode)
- wm.FloatingPointMode = Alternate;
+ wm.FloatingPointMode = FLOATING_POINT_MODE_Alternate;
- wm.SamplerCount = DIV_ROUND_UP(stage_state->sampler_count, 4);
- wm.BindingTableEntryCount = wm_prog_data->base.binding_table.size_bytes / 4;
+ wm.SamplerCount = GEN_GEN == 5 ?
+ 0 : DIV_ROUND_UP(stage_state->sampler_count, 4);
+
+ wm.BindingTableEntryCount =
+ wm_prog_data->base.binding_table.size_bytes / 4;
wm.MaximumNumberofThreads = devinfo->max_wm_threads - 1;
wm._8PixelDispatchEnable = wm_prog_data->dispatch_8;
wm._16PixelDispatchEnable = wm_prog_data->dispatch_16;
wm.DispatchGRFStartRegisterForConstantSetupData0 =
wm_prog_data->base.dispatch_grf_start_reg;
- wm.DispatchGRFStartRegisterForConstantSetupData2 =
- wm_prog_data->dispatch_grf_start_reg_2;
- wm.KernelStartPointer0 = stage_state->prog_offset;
- wm.KernelStartPointer2 = stage_state->prog_offset +
- wm_prog_data->prog_offset_2;
+ if (GEN_GEN == 6 ||
+ wm_prog_data->dispatch_8 || wm_prog_data->dispatch_16) {
+ wm.KernelStartPointer0 = KSP(brw, stage_state->prog_offset);
+ }
+
+#if GEN_GEN >= 5
+ if (GEN_GEN == 6 || wm_prog_data->prog_offset_2) {
+ wm.KernelStartPointer2 =
+ KSP(brw, stage_state->prog_offset + wm_prog_data->prog_offset_2);
+ }
+#endif
+
+#if GEN_GEN == 6
wm.DualSourceBlendEnable =
wm_prog_data->dual_src_blend && (ctx->Color.BlendEnabled & 1) &&
ctx->Color.Blend[0]._UsesDualSrc;
else
wm.PositionXYOffsetSelect = POSOFFSET_NONE;
+ wm.DispatchGRFStartRegisterForConstantSetupData2 =
+ wm_prog_data->dispatch_grf_start_reg_2;
+#endif
+
if (wm_prog_data->base.total_scratch) {
- wm.ScratchSpaceBasePointer =
- render_bo(stage_state->scratch_bo,
- ffs(stage_state->per_thread_scratch) - 11);
+ wm.ScratchSpaceBasePointer = rw_bo(stage_state->scratch_bo, 0);
+ wm.PerThreadScratchSpace =
+ ffs(stage_state->per_thread_scratch) - 11;
}
wm.PixelShaderComputedDepth = writes_depth;
#endif
- wm.PointRasterizationRule = RASTRULE_UPPER_RIGHT;
-
/* _NEW_LINE */
wm.LineStippleEnable = ctx->Line.StippleFlag;
/* _NEW_POLYGON */
wm.PolygonStippleEnable = ctx->Polygon.StippleFlag;
- wm.BarycentricInterpolationMode = wm_prog_data->barycentric_interp_modes;
#if GEN_GEN < 8
- /* _NEW_BUFFERS */
- const bool multisampled_fbo = _mesa_geometric_samples(ctx->DrawBuffer) > 1;
- wm.PixelShaderUsesSourceDepth = wm_prog_data->uses_src_depth;
+#if GEN_GEN >= 6
wm.PixelShaderUsesSourceW = wm_prog_data->uses_src_w;
- if (wm_prog_data->uses_kill ||
- _mesa_is_alpha_test_enabled(ctx) ||
- _mesa_is_alpha_to_coverage_enabled(ctx) ||
- wm_prog_data->uses_omask) {
- wm.PixelShaderKillsPixel = true;
- }
- /* _NEW_BUFFERS | _NEW_COLOR */
- if (brw_color_buffer_write_enabled(brw) || writes_depth ||
- wm_prog_data->has_side_effects || wm.PixelShaderKillsPixel) {
- wm.ThreadDispatchEnable = true;
- }
+ /* _NEW_BUFFERS */
+ const bool multisampled_fbo = _mesa_geometric_samples(ctx->DrawBuffer) > 1;
+
if (multisampled_fbo) {
/* _NEW_MULTISAMPLE */
if (ctx->Multisample.Enabled)
wm.MultisampleRasterizationMode = MSRASTMODE_OFF_PIXEL;
wm.MultisampleDispatchMode = MSDISPMODE_PERSAMPLE;
}
+#endif
+ wm.PixelShaderUsesSourceDepth = wm_prog_data->uses_src_depth;
+ if (wm_prog_data->uses_kill ||
+ _mesa_is_alpha_test_enabled(ctx) ||
+ _mesa_is_alpha_to_coverage_enabled(ctx) ||
+ (GEN_GEN >= 6 && wm_prog_data->uses_omask)) {
+ wm.PixelShaderKillsPixel = true;
+ }
+
+ /* _NEW_BUFFERS | _NEW_COLOR */
+ if (brw_color_buffer_write_enabled(brw) || writes_depth ||
+ wm.PixelShaderKillsPixel ||
+ (GEN_GEN >= 6 && wm_prog_data->has_side_effects)) {
+ wm.ThreadDispatchEnable = true;
+ }
#if GEN_GEN >= 7
wm.PixelShaderComputedDepthMode = wm_prog_data->computed_depth_mode;
wm.EarlyDepthStencilControl = EDSC_PSEXEC;
#endif
}
+
+#if GEN_GEN <= 5
+ if (brw->wm.offset_clamp != ctx->Polygon.OffsetClamp) {
+ brw_batch_emit(brw, GENX(3DSTATE_GLOBAL_DEPTH_OFFSET_CLAMP), clamp) {
+ clamp.GlobalDepthOffsetClamp = ctx->Polygon.OffsetClamp;
+ }
+
+ brw->wm.offset_clamp = ctx->Polygon.OffsetClamp;
+ }
+#endif
}
static const struct brw_tracked_state genX(wm_state) = {
.mesa = _NEW_LINE |
_NEW_POLYGON |
(GEN_GEN < 8 ? _NEW_BUFFERS |
- _NEW_COLOR |
- _NEW_MULTISAMPLE :
+ _NEW_COLOR :
0) |
- (GEN_GEN < 7 ? _NEW_PROGRAM_CONSTANTS : 0),
+ (GEN_GEN == 6 ? _NEW_PROGRAM_CONSTANTS : 0) |
+ (GEN_GEN < 6 ? _NEW_POLYGONSTIPPLE : 0) |
+ (GEN_GEN < 8 && GEN_GEN >= 6 ? _NEW_MULTISAMPLE : 0),
.brw = BRW_NEW_BLORP |
BRW_NEW_FS_PROG_DATA |
+ (GEN_GEN < 6 ? BRW_NEW_PUSH_CONSTANT_ALLOCATION |
+ BRW_NEW_FRAGMENT_PROGRAM |
+ BRW_NEW_PROGRAM_CACHE |
+ BRW_NEW_SAMPLER_STATE_TABLE |
+ BRW_NEW_STATS_WM
+ : 0) |
(GEN_GEN < 7 ? BRW_NEW_BATCH : BRW_NEW_CONTEXT),
},
.emit = genX(upload_wm),
};
-#endif
/* ---------------------------------------------------------------------- */
pkt.FloatingPointMode = stage_prog_data->use_alt_mode; \
\
if (stage_prog_data->total_scratch) { \
- pkt.ScratchSpaceBasePointer = \
- render_bo(stage_state->scratch_bo, 0); \
+ pkt.ScratchSpaceBasePointer = rw_bo(stage_state->scratch_bo, 0); \
pkt.PerThreadScratchSpace = \
ffs(stage_state->per_thread_scratch) - 11; \
} \
vs.StatisticsEnable = false;
vs.SamplerStatePointer =
- instruction_ro_bo(brw->batch.bo, stage_state->sampler_offset);
+ ro_bo(brw->batch.state.bo, stage_state->sampler_offset);
#endif
#if GEN_GEN == 5
/* ---------------------------------------------------------------------- */
-static inline void
+static void
set_scissor_bits(const struct gl_context *ctx, int i,
bool render_to_fbo, unsigned fb_width, unsigned fb_height,
struct GENX(SCISSOR_RECT) *sc)
/* ---------------------------------------------------------------------- */
-#if GEN_GEN >= 6
static void
genX(upload_gs_state)(struct brw_context *brw)
{
- const struct gen_device_info *devinfo = &brw->screen->devinfo;
+ UNUSED struct gl_context *ctx = &brw->ctx;
+ UNUSED const struct gen_device_info *devinfo = &brw->screen->devinfo;
const struct brw_stage_state *stage_state = &brw->gs.base;
+ const struct gl_program *gs_prog = brw->programs[MESA_SHADER_GEOMETRY];
/* BRW_NEW_GEOMETRY_PROGRAM */
- bool active = brw->geometry_program;
+ bool active = GEN_GEN >= 6 && gs_prog;
/* BRW_NEW_GS_PROG_DATA */
struct brw_stage_prog_data *stage_prog_data = stage_state->prog_data;
- const struct brw_vue_prog_data *vue_prog_data =
+ UNUSED const struct brw_vue_prog_data *vue_prog_data =
brw_vue_prog_data(stage_prog_data);
#if GEN_GEN >= 7
const struct brw_gs_prog_data *gs_prog_data =
* whole fixed function pipeline" means to emit a PIPE_CONTROL with the "CS
* Stall" bit set.
*/
- if (brw->gt == 2 && brw->gs.enabled != active)
+ if (devinfo->gt == 2 && brw->gs.enabled != active)
gen7_emit_cs_stall_flush(brw);
#endif
+#if GEN_GEN >= 6
brw_batch_emit(brw, GENX(3DSTATE_GS), gs) {
+#else
+ ctx->NewDriverState |= BRW_NEW_GEN4_UNIT_STATE;
+ brw_state_emit(brw, GENX(GS_STATE), 32, &brw->ff_gs.state_offset, gs) {
+#endif
+
+#if GEN_GEN >= 6
if (active) {
INIT_THREAD_DISPATCH_FIELDS(gs, Vertex);
#if GEN_GEN < 7
gs.SOStatisticsEnable = true;
- gs.RenderingEnabled = 1;
- if (brw->geometry_program->info.has_transform_feedback_varyings)
+ if (gs_prog->info.has_transform_feedback_varyings)
gs.SVBIPayloadEnable = true;
/* GEN6_GS_SPF_MODE and GEN6_GS_VECTOR_MASK_ENABLE are enabled as it
gs.VertexURBEntryOutputReadOffset = urb_entry_write_offset;
gs.VertexURBEntryOutputLength = MAX2(urb_entry_output_length, 1);
#endif
-#if GEN_GEN < 7
- } else if (brw->ff_gs.prog_active) {
+ }
+#endif
+
+#if GEN_GEN <= 6
+ if (!active && brw->ff_gs.prog_active) {
/* In gen6, transform feedback for the VS stage is done with an
* ad-hoc GS program. This function provides the needed 3DSTATE_GS
* for this.
*/
gs.KernelStartPointer = KSP(brw, brw->ff_gs.prog_offset);
gs.SingleProgramFlow = true;
- gs.VectorMaskEnable = true;
- gs.DispatchGRFStartRegisterForURBData = 2;
+ gs.DispatchGRFStartRegisterForURBData = GEN_GEN == 6 ? 2 : 1;
gs.VertexURBEntryReadLength = brw->ff_gs.prog_data->urb_read_length;
- gs.MaximumNumberofThreads = devinfo->max_gs_threads - 1;
- gs.StatisticsEnable = true;
- gs.SOStatisticsEnable = true;
- gs.RenderingEnabled = true;
+
+#if GEN_GEN <= 5
+ gs.GRFRegisterCount =
+ DIV_ROUND_UP(brw->ff_gs.prog_data->total_grf, 16) - 1;
+ /* BRW_NEW_URB_FENCE */
+ gs.NumberofURBEntries = brw->urb.nr_gs_entries;
+ gs.URBEntryAllocationSize = brw->urb.vsize - 1;
+ gs.MaximumNumberofThreads = brw->urb.nr_gs_entries >= 8 ? 1 : 0;
+ gs.FloatingPointMode = FLOATING_POINT_MODE_Alternate;
+#else
+ gs.Enable = true;
+ gs.VectorMaskEnable = true;
gs.SVBIPayloadEnable = true;
gs.SVBIPostIncrementEnable = true;
gs.SVBIPostIncrementValue =
brw->ff_gs.prog_data->svbi_postincrement_value;
- gs.Enable = true;
+ gs.SOStatisticsEnable = true;
+ gs.MaximumNumberofThreads = devinfo->max_gs_threads - 1;
#endif
- } else {
- gs.StatisticsEnable = true;
-#if GEN_GEN < 7
- gs.RenderingEnabled = true;
+ }
#endif
-
+ if (!active && !brw->ff_gs.prog_active) {
#if GEN_GEN < 8
gs.DispatchGRFStartRegisterForURBData = 1;
#if GEN_GEN >= 7
#endif
#endif
}
+
+#if GEN_GEN >= 6
+ gs.StatisticsEnable = true;
+#endif
+#if GEN_GEN == 5 || GEN_GEN == 6
+ gs.RenderingEnabled = true;
+#endif
+#if GEN_GEN <= 5
+ gs.MaximumVPIndex = brw->clip.viewport_count - 1;
+#endif
}
#if GEN_GEN == 6
static const struct brw_tracked_state genX(gs_state) = {
.dirty = {
- .mesa = (GEN_GEN < 7 ? _NEW_PROGRAM_CONSTANTS : 0),
+ .mesa = (GEN_GEN == 6 ? _NEW_PROGRAM_CONSTANTS : 0),
.brw = BRW_NEW_BATCH |
BRW_NEW_BLORP |
- BRW_NEW_CONTEXT |
- BRW_NEW_GEOMETRY_PROGRAM |
- BRW_NEW_GS_PROG_DATA |
+ (GEN_GEN <= 5 ? BRW_NEW_PUSH_CONSTANT_ALLOCATION |
+ BRW_NEW_PROGRAM_CACHE |
+ BRW_NEW_URB_FENCE |
+ BRW_NEW_VIEWPORT_COUNT
+ : 0) |
+ (GEN_GEN >= 6 ? BRW_NEW_CONTEXT |
+ BRW_NEW_GEOMETRY_PROGRAM |
+ BRW_NEW_GS_PROG_DATA
+ : 0) |
(GEN_GEN < 7 ? BRW_NEW_FF_GS_PROG_DATA : 0),
},
.emit = genX(upload_gs_state),
};
-#endif
/* ---------------------------------------------------------------------- */
};
static void
-upload_constant_state(struct brw_context *brw,
- struct brw_stage_state *stage_state,
- bool active, uint32_t stage)
+genX(upload_push_constant_packets)(struct brw_context *brw)
{
+ const struct gen_device_info *devinfo = &brw->screen->devinfo;
+ struct gl_context *ctx = &brw->ctx;
+
UNUSED uint32_t mocs = GEN_GEN < 8 ? GEN7_MOCS_L3 : 0;
- active = active && stage_state->push_const_size != 0;
- brw_batch_emit(brw, GENX(3DSTATE_CONSTANT_VS), pkt) {
- pkt._3DCommandSubOpcode = push_constant_opcodes[stage];
- if (active) {
+ struct brw_stage_state *stage_states[] = {
+ &brw->vs.base,
+ &brw->tcs.base,
+ &brw->tes.base,
+ &brw->gs.base,
+ &brw->wm.base,
+ };
+
+ if (GEN_GEN == 7 && !GEN_IS_HASWELL && !devinfo->is_baytrail &&
+ stage_states[MESA_SHADER_VERTEX]->push_constants_dirty)
+ gen7_emit_vs_workaround_flush(brw);
+
+ for (int stage = 0; stage <= MESA_SHADER_FRAGMENT; stage++) {
+ struct brw_stage_state *stage_state = stage_states[stage];
+ UNUSED struct gl_program *prog = ctx->_Shader->CurrentProgram[stage];
+
+ if (!stage_state->push_constants_dirty)
+ continue;
+
+ brw_batch_emit(brw, GENX(3DSTATE_CONSTANT_VS), pkt) {
+ pkt._3DCommandSubOpcode = push_constant_opcodes[stage];
+ if (stage_state->prog_data) {
#if GEN_GEN >= 8 || GEN_IS_HASWELL
- pkt.ConstantBody.ReadLength[2] = stage_state->push_const_size;
- pkt.ConstantBody.Buffer[2] =
- render_ro_bo(brw->curbe.curbe_bo, stage_state->push_const_offset);
+ /* The Skylake PRM contains the following restriction:
+ *
+ * "The driver must ensure The following case does not occur
+ * without a flush to the 3D engine: 3DSTATE_CONSTANT_* with
+ * buffer 3 read length equal to zero committed followed by a
+ * 3DSTATE_CONSTANT_* with buffer 0 read length not equal to
+ * zero committed."
+ *
+ * To avoid this, we program the buffers in the highest slots.
+ * This way, slot 0 is only used if slot 3 is also used.
+ */
+ int n = 3;
+
+ for (int i = 3; i >= 0; i--) {
+ const struct brw_ubo_range *range =
+ &stage_state->prog_data->ubo_ranges[i];
+
+ if (range->length == 0)
+ continue;
+
+ const struct gl_uniform_block *block =
+ prog->sh.UniformBlocks[range->block];
+ const struct gl_buffer_binding *binding =
+ &ctx->UniformBufferBindings[block->Binding];
+
+ if (binding->BufferObject == ctx->Shared->NullBufferObj) {
+ static unsigned msg_id = 0;
+ _mesa_gl_debug(ctx, &msg_id, MESA_DEBUG_SOURCE_API,
+ MESA_DEBUG_TYPE_UNDEFINED,
+ MESA_DEBUG_SEVERITY_HIGH,
+ "UBO %d unbound, %s shader uniform data "
+ "will be undefined.",
+ range->block,
+ _mesa_shader_stage_to_string(stage));
+ continue;
+ }
+
+ assert(binding->Offset % 32 == 0);
+
+ struct brw_bo *bo = intel_bufferobj_buffer(brw,
+ intel_buffer_object(binding->BufferObject),
+ binding->Offset, range->length * 32, false);
+
+ pkt.ConstantBody.ReadLength[n] = range->length;
+ pkt.ConstantBody.Buffer[n] =
+ ro_bo(bo, range->start * 32 + binding->Offset);
+ n--;
+ }
+
+ if (stage_state->push_const_size > 0) {
+ assert(n >= 0);
+ pkt.ConstantBody.ReadLength[n] = stage_state->push_const_size;
+ pkt.ConstantBody.Buffer[n] =
+ ro_bo(stage_state->push_const_bo,
+ stage_state->push_const_offset);
+ }
#else
- pkt.ConstantBody.ReadLength[0] = stage_state->push_const_size;
- pkt.ConstantBody.Buffer[0].offset =
- stage_state->push_const_offset | mocs;
+ pkt.ConstantBody.ReadLength[0] = stage_state->push_const_size;
+ pkt.ConstantBody.Buffer[0].offset =
+ stage_state->push_const_offset | mocs;
#endif
+ }
}
- }
- brw->ctx.NewDriverState |= GEN_GEN >= 9 ? BRW_NEW_SURFACES : 0;
+ stage_state->push_constants_dirty = false;
+ brw->ctx.NewDriverState |= GEN_GEN >= 9 ? BRW_NEW_SURFACES : 0;
+ }
}
+
+const struct brw_tracked_state genX(push_constant_packets) = {
+ .dirty = {
+ .mesa = 0,
+ .brw = BRW_NEW_DRAW_CALL,
+ },
+ .emit = genX(upload_push_constant_packets),
+};
#endif
#if GEN_GEN >= 6
{
struct brw_stage_state *stage_state = &brw->vs.base;
- /* _BRW_NEW_VERTEX_PROGRAM */
- const struct brw_program *vp = brw_program_const(brw->vertex_program);
+ /* BRW_NEW_VERTEX_PROGRAM */
+ const struct gl_program *vp = brw->programs[MESA_SHADER_VERTEX];
/* BRW_NEW_VS_PROG_DATA */
const struct brw_stage_prog_data *prog_data = brw->vs.base.prog_data;
- _mesa_shader_write_subroutine_indices(&brw->ctx, MESA_SHADER_VERTEX);
- gen6_upload_push_constants(brw, &vp->program, prog_data, stage_state);
-
-#if GEN_GEN >= 7
- if (GEN_GEN == 7 && !GEN_IS_HASWELL && !brw->is_baytrail)
- gen7_emit_vs_workaround_flush(brw);
-
- upload_constant_state(brw, stage_state, true /* active */,
- MESA_SHADER_VERTEX);
-#endif
+ gen6_upload_push_constants(brw, vp, prog_data, stage_state);
}
static const struct brw_tracked_state genX(vs_push_constants) = {
_NEW_TRANSFORM,
.brw = BRW_NEW_BATCH |
BRW_NEW_BLORP |
- BRW_NEW_PUSH_CONSTANT_ALLOCATION |
BRW_NEW_VERTEX_PROGRAM |
BRW_NEW_VS_PROG_DATA,
},
struct brw_stage_state *stage_state = &brw->gs.base;
/* BRW_NEW_GEOMETRY_PROGRAM */
- const struct brw_program *gp = brw_program_const(brw->geometry_program);
-
- if (gp) {
- /* BRW_NEW_GS_PROG_DATA */
- struct brw_stage_prog_data *prog_data = brw->gs.base.prog_data;
+ const struct gl_program *gp = brw->programs[MESA_SHADER_GEOMETRY];
- _mesa_shader_write_subroutine_indices(&brw->ctx, MESA_SHADER_GEOMETRY);
- gen6_upload_push_constants(brw, &gp->program, prog_data, stage_state);
- }
+ /* BRW_NEW_GS_PROG_DATA */
+ struct brw_stage_prog_data *prog_data = brw->gs.base.prog_data;
-#if GEN_GEN >= 7
- upload_constant_state(brw, stage_state, gp, MESA_SHADER_GEOMETRY);
-#endif
+ gen6_upload_push_constants(brw, gp, prog_data, stage_state);
}
static const struct brw_tracked_state genX(gs_push_constants) = {
.brw = BRW_NEW_BATCH |
BRW_NEW_BLORP |
BRW_NEW_GEOMETRY_PROGRAM |
- BRW_NEW_GS_PROG_DATA |
- BRW_NEW_PUSH_CONSTANT_ALLOCATION,
+ BRW_NEW_GS_PROG_DATA,
},
.emit = genX(upload_gs_push_constants),
};
{
struct brw_stage_state *stage_state = &brw->wm.base;
/* BRW_NEW_FRAGMENT_PROGRAM */
- const struct brw_program *fp = brw_program_const(brw->fragment_program);
+ const struct gl_program *fp = brw->programs[MESA_SHADER_FRAGMENT];
/* BRW_NEW_FS_PROG_DATA */
const struct brw_stage_prog_data *prog_data = brw->wm.base.prog_data;
- _mesa_shader_write_subroutine_indices(&brw->ctx, MESA_SHADER_FRAGMENT);
-
- gen6_upload_push_constants(brw, &fp->program, prog_data, stage_state);
-
-#if GEN_GEN >= 7
- upload_constant_state(brw, stage_state, true, MESA_SHADER_FRAGMENT);
-#endif
+ gen6_upload_push_constants(brw, fp, prog_data, stage_state);
}
static const struct brw_tracked_state genX(wm_push_constants) = {
.brw = BRW_NEW_BATCH |
BRW_NEW_BLORP |
BRW_NEW_FRAGMENT_PROGRAM |
- BRW_NEW_FS_PROG_DATA |
- BRW_NEW_PUSH_CONSTANT_ALLOCATION,
+ BRW_NEW_FS_PROG_DATA,
},
.emit = genX(upload_wm_push_constants),
};
genX(emit_3dstate_multisample2)(struct brw_context *brw,
unsigned num_samples)
{
- assert(brw->num_samples <= 16);
-
- unsigned log2_samples = ffs(MAX2(num_samples, 1)) - 1;
+ unsigned log2_samples = ffs(num_samples) - 1;
brw_batch_emit(brw, GENX(3DSTATE_MULTISAMPLE), multi) {
multi.PixelLocation = CENTER;
static void
genX(upload_multisample_state)(struct brw_context *brw)
{
+ assert(brw->num_samples > 0 && brw->num_samples <= 16);
+
genX(emit_3dstate_multisample2)(brw, brw->num_samples);
brw_batch_emit(brw, GENX(3DSTATE_SAMPLE_MASK), sm) {
static const struct brw_tracked_state genX(multisample_state) = {
.dirty = {
- .mesa = _NEW_MULTISAMPLE,
+ .mesa = _NEW_MULTISAMPLE |
+ (GEN_GEN == 10 ? _NEW_BUFFERS : 0),
.brw = BRW_NEW_BLORP |
BRW_NEW_CONTEXT |
BRW_NEW_NUM_SAMPLES,
cc.StatisticsEnable = brw->stats_wm;
cc.CCViewportStatePointer =
- instruction_ro_bo(brw->batch.bo, brw->cc.vp_offset);
+ ro_bo(brw->batch.state.bo, brw->cc.vp_offset);
#else
/* _NEW_COLOR */
cc.BlendConstantColorRed = ctx->Color.BlendColorUnclamped[0];
genX(upload_sbe)(struct brw_context *brw)
{
struct gl_context *ctx = &brw->ctx;
+ /* BRW_NEW_FRAGMENT_PROGRAM */
+ UNUSED const struct gl_program *fp = brw->programs[MESA_SHADER_FRAGMENT];
/* BRW_NEW_FS_PROG_DATA */
const struct brw_wm_prog_data *wm_prog_data =
brw_wm_prog_data(brw->wm.base.prog_data);
#if GEN_GEN >= 9
/* prepare the active component dwords */
- int input_index = 0;
- for (int attr = 0; attr < VARYING_SLOT_MAX; attr++) {
- if (!(brw->fragment_program->info.inputs_read &
- BITFIELD64_BIT(attr))) {
- continue;
- }
-
- assert(input_index < 32);
-
+ const int num_inputs = urb_entry_read_length * 2;
+ for (int input_index = 0; input_index < num_inputs; input_index++) {
sbe.AttributeActiveComponentFormat[input_index] = ACTIVE_COMPONENT_XYZW;
- ++input_index;
}
#endif
}
assert(start % 4 == 0);
uint32_t end = ALIGN(start + xfb_obj->Size[i], 4);
struct brw_bo *bo =
- intel_bufferobj_buffer(brw, bufferobj, start, end - start);
+ intel_bufferobj_buffer(brw, bufferobj, start, end - start, true);
assert(end <= bo->size);
brw_batch_emit(brw, GENX(3DSTATE_SO_BUFFER), sob) {
sob.SOBufferIndex = i;
- sob.SurfaceBaseAddress = render_bo(bo, start);
+ sob.SurfaceBaseAddress = rw_bo(bo, start);
#if GEN_GEN < 8
sob.SurfacePitch = linked_xfb_info->Buffers[i].Stride * 4;
- sob.SurfaceEndAddress = render_bo(bo, end);
+ sob.SurfaceEndAddress = rw_bo(bo, end);
#else
sob.SOBufferEnable = true;
sob.StreamOffsetWriteEnable = true;
sob.SurfaceSize = MAX2(xfb_obj->Size[i] / 4, 1) - 1;
sob.StreamOutputBufferOffsetAddress =
- instruction_bo(brw_obj->offset_bo, i * sizeof(uint32_t));
+ rw_bo(brw_obj->offset_bo, i * sizeof(uint32_t));
if (brw_obj->zero_offsets) {
/* Zero out the offset and write that to offset_bo */
#endif
}
-static inline bool
+static bool
query_active(struct gl_query_object *q)
{
return q && q->Active;
sos.RenderingDisable = true;
} else {
perf_debug("Rasterizer discard with a GL_PRIMITIVES_GENERATED "
- "query active relies on the clipper.");
+ "query active relies on the clipper.\n");
}
}
ps.MaximumNumberofThreads = devinfo->max_wm_threads - 1;
#endif
- if (prog_data->base.nr_params > 0)
+ if (prog_data->base.nr_params > 0 ||
+ prog_data->base.ubo_ranges[0].length > 0)
ps.PushConstantEnable = true;
#if GEN_GEN < 8
else
ps.PositionXYOffsetSelect = POSOFFSET_NONE;
- ps.RenderTargetFastClearEnable = brw->wm.fast_clear_op;
ps._8PixelDispatchEnable = prog_data->dispatch_8;
ps._16PixelDispatchEnable = prog_data->dispatch_16;
ps.DispatchGRFStartRegisterForConstantSetupData0 =
if (prog_data->base.total_scratch) {
ps.ScratchSpaceBasePointer =
- render_bo(stage_state->scratch_bo,
- ffs(stage_state->per_thread_scratch) - 11);
+ rw_bo(stage_state->scratch_bo,
+ ffs(stage_state->per_thread_scratch) - 11);
}
}
}
upload_te_state(struct brw_context *brw)
{
/* BRW_NEW_TESS_PROGRAMS */
- bool active = brw->tess_eval_program;
+ bool active = brw->programs[MESA_SHADER_TESS_EVAL];
/* BRW_NEW_TES_PROG_DATA */
const struct brw_tes_prog_data *tes_prog_data =
{
struct brw_stage_state *stage_state = &brw->tes.base;
/* BRW_NEW_TESS_PROGRAMS */
- const struct brw_program *tep = brw_program_const(brw->tess_eval_program);
+ const struct gl_program *tep = brw->programs[MESA_SHADER_TESS_EVAL];
- if (tep) {
- /* BRW_NEW_TES_PROG_DATA */
- const struct brw_stage_prog_data *prog_data = brw->tes.base.prog_data;
- _mesa_shader_write_subroutine_indices(&brw->ctx, MESA_SHADER_TESS_EVAL);
- gen6_upload_push_constants(brw, &tep->program, prog_data, stage_state);
- }
-
- upload_constant_state(brw, stage_state, tep, MESA_SHADER_TESS_EVAL);
+ /* BRW_NEW_TES_PROG_DATA */
+ const struct brw_stage_prog_data *prog_data = brw->tes.base.prog_data;
+ gen6_upload_push_constants(brw, tep, prog_data, stage_state);
}
static const struct brw_tracked_state genX(tes_push_constants) = {
.mesa = _NEW_PROGRAM_CONSTANTS,
.brw = BRW_NEW_BATCH |
BRW_NEW_BLORP |
- BRW_NEW_PUSH_CONSTANT_ALLOCATION |
BRW_NEW_TESS_PROGRAMS |
BRW_NEW_TES_PROG_DATA,
},
{
struct brw_stage_state *stage_state = &brw->tcs.base;
/* BRW_NEW_TESS_PROGRAMS */
- const struct brw_program *tcp = brw_program_const(brw->tess_ctrl_program);
- bool active = brw->tess_eval_program;
-
- if (active) {
- /* BRW_NEW_TCS_PROG_DATA */
- const struct brw_stage_prog_data *prog_data = brw->tcs.base.prog_data;
+ const struct gl_program *tcp = brw->programs[MESA_SHADER_TESS_CTRL];
- _mesa_shader_write_subroutine_indices(&brw->ctx, MESA_SHADER_TESS_CTRL);
- gen6_upload_push_constants(brw, &tcp->program, prog_data, stage_state);
- }
+ /* BRW_NEW_TCS_PROG_DATA */
+ const struct brw_stage_prog_data *prog_data = brw->tcs.base.prog_data;
- upload_constant_state(brw, stage_state, active, MESA_SHADER_TESS_CTRL);
+ gen6_upload_push_constants(brw, tcp, prog_data, stage_state);
}
static const struct brw_tracked_state genX(tcs_push_constants) = {
.brw = BRW_NEW_BATCH |
BRW_NEW_BLORP |
BRW_NEW_DEFAULT_TESS_LEVELS |
- BRW_NEW_PUSH_CONSTANT_ALLOCATION |
BRW_NEW_TESS_PROGRAMS |
BRW_NEW_TCS_PROG_DATA,
},
/* ---------------------------------------------------------------------- */
#if GEN_GEN >= 7
+static void
+genX(upload_cs_push_constants)(struct brw_context *brw)
+{
+ struct brw_stage_state *stage_state = &brw->cs.base;
+
+ /* BRW_NEW_COMPUTE_PROGRAM */
+ const struct gl_program *cp = brw->programs[MESA_SHADER_COMPUTE];
+
+ if (cp) {
+ /* BRW_NEW_CS_PROG_DATA */
+ struct brw_cs_prog_data *cs_prog_data =
+ brw_cs_prog_data(brw->cs.base.prog_data);
+
+ _mesa_shader_write_subroutine_indices(&brw->ctx, MESA_SHADER_COMPUTE);
+ brw_upload_cs_push_constants(brw, cp, cs_prog_data, stage_state);
+ }
+}
+
+const struct brw_tracked_state genX(cs_push_constants) = {
+ .dirty = {
+ .mesa = _NEW_PROGRAM_CONSTANTS,
+ .brw = BRW_NEW_BATCH |
+ BRW_NEW_BLORP |
+ BRW_NEW_COMPUTE_PROGRAM |
+ BRW_NEW_CS_PROG_DATA,
+ },
+ .emit = genX(upload_cs_push_constants),
+};
+
+/**
+ * Creates a new CS constant buffer reflecting the current CS program's
+ * constants, if needed by the CS program.
+ */
+static void
+genX(upload_cs_pull_constants)(struct brw_context *brw)
+{
+ struct brw_stage_state *stage_state = &brw->cs.base;
+
+ /* BRW_NEW_COMPUTE_PROGRAM */
+ struct brw_program *cp =
+ (struct brw_program *) brw->programs[MESA_SHADER_COMPUTE];
+
+ /* BRW_NEW_CS_PROG_DATA */
+ const struct brw_stage_prog_data *prog_data = brw->cs.base.prog_data;
+
+ _mesa_shader_write_subroutine_indices(&brw->ctx, MESA_SHADER_COMPUTE);
+ /* _NEW_PROGRAM_CONSTANTS */
+ brw_upload_pull_constants(brw, BRW_NEW_SURFACES, &cp->program,
+ stage_state, prog_data);
+}
+
+const struct brw_tracked_state genX(cs_pull_constants) = {
+ .dirty = {
+ .mesa = _NEW_PROGRAM_CONSTANTS,
+ .brw = BRW_NEW_BATCH |
+ BRW_NEW_BLORP |
+ BRW_NEW_COMPUTE_PROGRAM |
+ BRW_NEW_CS_PROG_DATA,
+ },
+ .emit = genX(upload_cs_pull_constants),
+};
+
static void
genX(upload_cs_state)(struct brw_context *brw)
{
brw, &stage_state->surf_offset[
prog_data->binding_table.shader_time_start],
brw->shader_time.bo, 0, ISL_FORMAT_RAW,
- brw->shader_time.bo->size, 1, true);
+ brw->shader_time.bo->size, 1,
+ RELOC_WRITE);
}
uint32_t *bind = brw_state_batch(brw, prog_data->binding_table.size_bytes,
32, &stage_state->bind_bo_offset);
+ /* The MEDIA_VFE_STATE documentation for Gen8+ says:
+ *
+ * "A stalling PIPE_CONTROL is required before MEDIA_VFE_STATE unless
+ * the only bits that are changed are scoreboard related: Scoreboard
+ * Enable, Scoreboard Type, Scoreboard Mask, Scoreboard * Delta. For
+ * these scoreboard related states, a MEDIA_STATE_FLUSH is sufficient."
+ *
+ * Earlier generations say "MI_FLUSH" instead of "stalling PIPE_CONTROL",
+ * but MI_FLUSH isn't really a thing, so we assume they meant PIPE_CONTROL.
+ */
+ brw_emit_pipe_control_flush(brw, PIPE_CONTROL_CS_STALL);
+
brw_batch_emit(brw, GENX(MEDIA_VFE_STATE), vfe) {
if (prog_data->total_scratch) {
- uint32_t bo_offset;
+ uint32_t per_thread_scratch_value;
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.
*/
- bo_offset = ffs(stage_state->per_thread_scratch) - 11;
+ per_thread_scratch_value = ffs(stage_state->per_thread_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.
*/
- bo_offset = ffs(stage_state->per_thread_scratch) - 12;
+ per_thread_scratch_value = ffs(stage_state->per_thread_scratch) - 12;
} else {
/* Earlier platforms use the range [0, 11] to mean [1kB, 12kB]
* where 0 = 1kB, 1 = 2kB, 2 = 3kB, ..., 11 = 12kB.
*/
- bo_offset = stage_state->per_thread_scratch / 1024 - 1;
+ per_thread_scratch_value = stage_state->per_thread_scratch / 1024 - 1;
}
- vfe.ScratchSpaceBasePointer =
- render_bo(stage_state->scratch_bo, bo_offset);
+ vfe.ScratchSpaceBasePointer = rw_bo(stage_state->scratch_bo, 0);
+ vfe.PerThreadScratchSpace = per_thread_scratch_value;
}
+ /* If brw->screen->subslice_total is greater than one, then
+ * devinfo->max_cs_threads stores number of threads per sub-slice;
+ * thus we need to multiply by that number by subslices to get
+ * the actual maximum number of threads; the -1 is because the HW
+ * has a bias of 1 (would not make sense to say the maximum number
+ * of threads is 0).
+ */
const uint32_t subslices = MAX2(brw->screen->subslice_total, 1);
vfe.MaximumNumberofThreads = devinfo->max_cs_threads * subslices - 1;
vfe.NumberofURBEntries = GEN_GEN >= 8 ? 2 : 0;
const struct GENX(INTERFACE_DESCRIPTOR_DATA) idd = {
.KernelStartPointer = brw->cs.base.prog_offset,
.SamplerStatePointer = stage_state->sampler_offset,
- .SamplerCount = DIV_ROUND_UP(stage_state->sampler_count, 4) >> 2,
+ .SamplerCount = DIV_ROUND_UP(CLAMP(stage_state->sampler_count, 0, 16), 4),
.BindingTablePointer = stage_state->bind_bo_offset,
.ConstantURBEntryReadLength = cs_prog_data->push.per_thread.regs,
.NumberofThreadsinGPGPUThreadGroup = cs_prog_data->threads,
- .SharedLocalMemorySize = encode_slm_size(devinfo->gen,
+ .SharedLocalMemorySize = encode_slm_size(GEN_GEN,
prog_data->total_shared),
.BarrierEnable = cs_prog_data->uses_barrier,
#if GEN_GEN >= 8 || GEN_IS_HASWELL
static void
genX(upload_raster)(struct brw_context *brw)
{
- struct gl_context *ctx = &brw->ctx;
+ const struct gl_context *ctx = &brw->ctx;
/* _NEW_BUFFERS */
- bool render_to_fbo = _mesa_is_user_fbo(ctx->DrawBuffer);
+ const bool render_to_fbo = _mesa_is_user_fbo(ctx->DrawBuffer);
/* _NEW_POLYGON */
- struct gl_polygon_attrib *polygon = &ctx->Polygon;
+ const struct gl_polygon_attrib *polygon = &ctx->Polygon;
/* _NEW_POINT */
- struct gl_point_attrib *point = &ctx->Point;
+ const struct gl_point_attrib *point = &ctx->Point;
brw_batch_emit(brw, GENX(3DSTATE_RASTER), raster) {
if (brw->polygon_front_bit == render_to_fbo)
raster.CullMode = CULLMODE_NONE;
}
- point->SmoothFlag = raster.SmoothPointEnable;
+ raster.SmoothPointEnable = point->SmoothFlag;
raster.DXMultisampleRasterizationEnable =
_mesa_is_multisample_enabled(ctx);
/* _NEW_LINE */
raster.AntialiasingEnable = ctx->Line.SmoothFlag;
+#if GEN_GEN == 10
+ /* _NEW_BUFFERS
+ * Antialiasing Enable bit MUST not be set when NUM_MULTISAMPLES > 1.
+ */
+ const bool multisampled_fbo =
+ _mesa_geometric_samples(ctx->DrawBuffer) > 1;
+ if (multisampled_fbo)
+ raster.AntialiasingEnable = false;
+#endif
+
/* _NEW_SCISSOR */
raster.ScissorRectangleEnable = ctx->Scissor.EnableFlags;
uint32_t report_id)
{
brw_batch_emit(brw, GENX(MI_REPORT_PERF_COUNT), mi_rpc) {
- mi_rpc.MemoryAddress = instruction_bo(bo, offset_in_bytes);
+ mi_rpc.MemoryAddress = ggtt_bo(bo, offset_in_bytes);
mi_rpc.ReportID = report_id;
}
}
color.ui[3] = float_as_int(1.0);
int alignment = 32;
- if (brw->gen >= 8) {
+ if (GEN_GEN >= 8) {
alignment = 64;
- } else if (brw->is_haswell && (is_integer_format || is_stencil_sampling)) {
+ } else if (GEN_IS_HASWELL && (is_integer_format || is_stencil_sampling)) {
alignment = 512;
}
texObj->StencilSampling,
&border_color_offset);
}
-
- samp_st.BorderColorPointer = border_color_offset;
-
- if (GEN_GEN < 6) {
- samp_st.BorderColorPointer += brw->batch.bo->offset64; /* reloc */
- brw_emit_reloc(&brw->batch, batch_offset_for_sampler_state + 8,
- brw->batch.bo, border_color_offset,
- I915_GEM_DOMAIN_SAMPLER, 0);
- }
+#if GEN_GEN < 6
+ samp_st.BorderColorPointer =
+ ro_bo(brw->batch.state.bo, border_color_offset);
+#else
+ samp_st.BorderColorPointer = border_color_offset;
+#endif
#if GEN_GEN >= 8
samp_st.LODPreClampMode = CLAMP_MODE_OGL;
genX(upload_fs_samplers)(struct brw_context *brw)
{
/* BRW_NEW_FRAGMENT_PROGRAM */
- struct gl_program *fs = (struct gl_program *) brw->fragment_program;
+ struct gl_program *fs = brw->programs[MESA_SHADER_FRAGMENT];
genX(upload_sampler_state_table)(brw, fs, &brw->wm.base);
}
genX(upload_vs_samplers)(struct brw_context *brw)
{
/* BRW_NEW_VERTEX_PROGRAM */
- struct gl_program *vs = (struct gl_program *) brw->vertex_program;
+ struct gl_program *vs = brw->programs[MESA_SHADER_VERTEX];
genX(upload_sampler_state_table)(brw, vs, &brw->vs.base);
}
genX(upload_gs_samplers)(struct brw_context *brw)
{
/* BRW_NEW_GEOMETRY_PROGRAM */
- struct gl_program *gs = (struct gl_program *) brw->geometry_program;
+ struct gl_program *gs = brw->programs[MESA_SHADER_GEOMETRY];
if (!gs)
return;
genX(upload_tcs_samplers)(struct brw_context *brw)
{
/* BRW_NEW_TESS_PROGRAMS */
- struct gl_program *tcs = (struct gl_program *) brw->tess_ctrl_program;
+ struct gl_program *tcs = brw->programs[MESA_SHADER_TESS_CTRL];
if (!tcs)
return;
genX(upload_tes_samplers)(struct brw_context *brw)
{
/* BRW_NEW_TESS_PROGRAMS */
- struct gl_program *tes = (struct gl_program *) brw->tess_eval_program;
+ struct gl_program *tes = brw->programs[MESA_SHADER_TESS_EVAL];
if (!tes)
return;
genX(upload_cs_samplers)(struct brw_context *brw)
{
/* BRW_NEW_COMPUTE_PROGRAM */
- struct gl_program *cs = (struct gl_program *) brw->compute_program;
+ struct gl_program *cs = brw->programs[MESA_SHADER_COMPUTE];
if (!cs)
return;
&genX(vs_samplers),
/* These set up state for brw_psp_urb_cbs */
- &brw_wm_unit,
+ &genX(wm_state),
&genX(sf_clip_viewport),
&genX(sf_state),
&genX(vs_state), /* always required, enabled or not */
- &brw_clip_unit,
- &brw_gs_unit,
+ &genX(clip_state),
+ &genX(gs_state),
/* Command packets:
*/
- &brw_invariant_state,
-
&brw_binding_table_pointers,
&genX(blend_constant_color),
*/
&brw_vs_pull_constants,
&brw_vs_ubo_surfaces,
- &brw_vs_abo_surfaces,
&brw_tcs_pull_constants,
&brw_tcs_ubo_surfaces,
- &brw_tcs_abo_surfaces,
&brw_tes_pull_constants,
&brw_tes_ubo_surfaces,
- &brw_tes_abo_surfaces,
&brw_gs_pull_constants,
&brw_gs_ubo_surfaces,
- &brw_gs_abo_surfaces,
&brw_wm_pull_constants,
&brw_wm_ubo_surfaces,
- &brw_wm_abo_surfaces,
&gen6_renderbuffer_surfaces,
&brw_renderbuffer_read_surfaces,
&brw_texture_surfaces,
+
+ &genX(push_constant_packets),
+
&brw_vs_binding_table,
&brw_tcs_binding_table,
&brw_tes_binding_table,
*/
&brw_vs_pull_constants,
&brw_vs_ubo_surfaces,
- &brw_vs_abo_surfaces,
&brw_tcs_pull_constants,
&brw_tcs_ubo_surfaces,
- &brw_tcs_abo_surfaces,
&brw_tes_pull_constants,
&brw_tes_ubo_surfaces,
- &brw_tes_abo_surfaces,
&brw_gs_pull_constants,
&brw_gs_ubo_surfaces,
- &brw_gs_abo_surfaces,
&brw_wm_pull_constants,
&brw_wm_ubo_surfaces,
- &brw_wm_abo_surfaces,
&gen6_renderbuffer_surfaces,
&brw_renderbuffer_read_surfaces,
&brw_texture_surfaces,
+
+ &genX(push_constant_packets),
+
&brw_vs_binding_table,
&brw_tcs_binding_table,
&brw_tes_binding_table,
{
&gen7_l3_state,
&brw_cs_image_surfaces,
- &gen7_cs_push_constants,
- &brw_cs_pull_constants,
+ &genX(cs_push_constants),
+ &genX(cs_pull_constants),
&brw_cs_ubo_surfaces,
- &brw_cs_abo_surfaces,
&brw_cs_texture_surfaces,
&brw_cs_work_groups_surface,
&genX(cs_samplers),