intel_translate_stencil_op(stencil->ZFailFunc[b]);
}
-#if GEN_GEN >= 9
+#if GEN_GEN <= 5 || GEN_GEN >= 9
ds->StencilReferenceValue = _mesa_get_stencil_ref(ctx, 0);
ds->BackfaceStencilReferenceValue = _mesa_get_stencil_ref(ctx, b);
#endif
brw_gs_prog_data(stage_prog_data);
#endif
-#if GEN_GEN < 7
+#if GEN_GEN == 6
brw_batch_emit(brw, GENX(3DSTATE_CONSTANT_GS), cgs) {
if (active && stage_state->push_const_size != 0) {
cgs.Buffer0Valid = true;
gen7_emit_cs_stall_flush(brw);
#endif
- if (active) {
- brw_batch_emit(brw, GENX(3DSTATE_GS), gs) {
+ brw_batch_emit(brw, GENX(3DSTATE_GS), gs) {
+ if (active) {
INIT_THREAD_DISPATCH_FIELDS(gs, Vertex);
#if GEN_GEN >= 7
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) {
- /* 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.
- */
- upload_gs_state_for_tf(brw);
+ } else if (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.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;
+ gs.SVBIPayloadEnable = true;
+ gs.SVBIPostIncrementEnable = true;
+ gs.SVBIPostIncrementValue =
+ brw->ff_gs.prog_data->svbi_postincrement_value;
+ gs.Enable = true;
#endif
- } else {
- brw_batch_emit(brw, GENX(3DSTATE_GS), gs) {
+ } else {
gs.StatisticsEnable = true;
#if GEN_GEN < 7
gs.RenderingEnabled = true;
#endif
}
}
-#if GEN_GEN < 7
+
+#if GEN_GEN == 6
brw->gs.enabled = active;
#endif
}
#define blend_factor(x) brw_translate_blend_factor(x)
#define blend_eqn(x) brw_translate_blend_equation(x)
+/**
+ * Modify blend function to force destination alpha to 1.0
+ *
+ * If \c function specifies a blend function that uses destination alpha,
+ * replace it with a function that hard-wires destination alpha to 1.0. This
+ * is used when rendering to xRGB targets.
+ */
+static GLenum
+brw_fix_xRGB_alpha(GLenum function)
+{
+ switch (function) {
+ case GL_DST_ALPHA:
+ return GL_ONE;
+
+ case GL_ONE_MINUS_DST_ALPHA:
+ case GL_SRC_ALPHA_SATURATE:
+ return GL_ZERO;
+ }
+
+ return function;
+}
+
+#if GEN_GEN >= 6
+typedef struct GENX(BLEND_STATE_ENTRY) BLEND_ENTRY_GENXML;
+#else
+typedef struct GENX(COLOR_CALC_STATE) BLEND_ENTRY_GENXML;
+#endif
+
+UNUSED static bool
+set_blend_entry_bits(struct brw_context *brw, BLEND_ENTRY_GENXML *entry, int i,
+ bool alpha_to_one)
+{
+ struct gl_context *ctx = &brw->ctx;
+
+ /* _NEW_BUFFERS */
+ const struct gl_renderbuffer *rb = ctx->DrawBuffer->_ColorDrawBuffers[i];
+
+ bool independent_alpha_blend = false;
+
+ /* Used for implementing the following bit of GL_EXT_texture_integer:
+ * "Per-fragment operations that require floating-point color
+ * components, including multisample alpha operations, alpha test,
+ * blending, and dithering, have no effect when the corresponding
+ * colors are written to an integer color buffer."
+ */
+ const bool integer = ctx->DrawBuffer->_IntegerBuffers & (0x1 << i);
+
+ const unsigned blend_enabled = GEN_GEN >= 6 ?
+ ctx->Color.BlendEnabled & (1 << i) : ctx->Color.BlendEnabled;
+
+ /* _NEW_COLOR */
+ if (ctx->Color.ColorLogicOpEnabled) {
+ GLenum rb_type = rb ? _mesa_get_format_datatype(rb->Format)
+ : GL_UNSIGNED_NORMALIZED;
+ WARN_ONCE(ctx->Color.LogicOp != GL_COPY &&
+ rb_type != GL_UNSIGNED_NORMALIZED &&
+ rb_type != GL_FLOAT, "Ignoring %s logic op on %s "
+ "renderbuffer\n",
+ _mesa_enum_to_string(ctx->Color.LogicOp),
+ _mesa_enum_to_string(rb_type));
+ if (GEN_GEN >= 8 || rb_type == GL_UNSIGNED_NORMALIZED) {
+ entry->LogicOpEnable = true;
+ entry->LogicOpFunction =
+ intel_translate_logic_op(ctx->Color.LogicOp);
+ }
+ } else if (blend_enabled && !ctx->Color._AdvancedBlendMode
+ && (GEN_GEN <= 5 || !integer)) {
+ GLenum eqRGB = ctx->Color.Blend[i].EquationRGB;
+ GLenum eqA = ctx->Color.Blend[i].EquationA;
+ GLenum srcRGB = ctx->Color.Blend[i].SrcRGB;
+ GLenum dstRGB = ctx->Color.Blend[i].DstRGB;
+ GLenum srcA = ctx->Color.Blend[i].SrcA;
+ GLenum dstA = ctx->Color.Blend[i].DstA;
+
+ if (eqRGB == GL_MIN || eqRGB == GL_MAX)
+ srcRGB = dstRGB = GL_ONE;
+
+ if (eqA == GL_MIN || eqA == GL_MAX)
+ srcA = dstA = GL_ONE;
+
+ /* Due to hardware limitations, the destination may have information
+ * in an alpha channel even when the format specifies no alpha
+ * channel. In order to avoid getting any incorrect blending due to
+ * that alpha channel, coerce the blend factors to values that will
+ * not read the alpha channel, but will instead use the correct
+ * implicit value for alpha.
+ */
+ if (rb && !_mesa_base_format_has_channel(rb->_BaseFormat,
+ GL_TEXTURE_ALPHA_TYPE)) {
+ srcRGB = brw_fix_xRGB_alpha(srcRGB);
+ srcA = brw_fix_xRGB_alpha(srcA);
+ dstRGB = brw_fix_xRGB_alpha(dstRGB);
+ dstA = brw_fix_xRGB_alpha(dstA);
+ }
+
+ /* From the BLEND_STATE docs, DWord 0, Bit 29 (AlphaToOne Enable):
+ * "If Dual Source Blending is enabled, this bit must be disabled."
+ *
+ * We override SRC1_ALPHA to ONE and ONE_MINUS_SRC1_ALPHA to ZERO,
+ * and leave it enabled anyway.
+ */
+ if (GEN_GEN >= 6 && ctx->Color.Blend[i]._UsesDualSrc && alpha_to_one) {
+ srcRGB = fix_dual_blend_alpha_to_one(srcRGB);
+ srcA = fix_dual_blend_alpha_to_one(srcA);
+ dstRGB = fix_dual_blend_alpha_to_one(dstRGB);
+ dstA = fix_dual_blend_alpha_to_one(dstA);
+ }
+
+ entry->ColorBufferBlendEnable = true;
+ entry->DestinationBlendFactor = blend_factor(dstRGB);
+ entry->SourceBlendFactor = blend_factor(srcRGB);
+ entry->DestinationAlphaBlendFactor = blend_factor(dstA);
+ entry->SourceAlphaBlendFactor = blend_factor(srcA);
+ entry->ColorBlendFunction = blend_eqn(eqRGB);
+ entry->AlphaBlendFunction = blend_eqn(eqA);
+
+ if (srcA != srcRGB || dstA != dstRGB || eqA != eqRGB)
+ independent_alpha_blend = true;
+ }
+
+ return independent_alpha_blend;
+}
+
#if GEN_GEN >= 6
static void
genX(upload_blend_state)(struct brw_context *brw)
#else
{
#endif
-
- /* _NEW_BUFFERS */
- struct gl_renderbuffer *rb = ctx->DrawBuffer->_ColorDrawBuffers[i];
-
- /* Used for implementing the following bit of GL_EXT_texture_integer:
- * "Per-fragment operations that require floating-point color
- * components, including multisample alpha operations, alpha test,
- * blending, and dithering, have no effect when the corresponding
- * colors are written to an integer color buffer."
- */
- bool integer = ctx->DrawBuffer->_IntegerBuffers & (0x1 << i);
-
- /* _NEW_COLOR */
- if (ctx->Color.ColorLogicOpEnabled) {
- GLenum rb_type = rb ? _mesa_get_format_datatype(rb->Format)
- : GL_UNSIGNED_NORMALIZED;
- WARN_ONCE(ctx->Color.LogicOp != GL_COPY &&
- rb_type != GL_UNSIGNED_NORMALIZED &&
- rb_type != GL_FLOAT, "Ignoring %s logic op on %s "
- "renderbuffer\n",
- _mesa_enum_to_string(ctx->Color.LogicOp),
- _mesa_enum_to_string(rb_type));
- if (GEN_GEN >= 8 || rb_type == GL_UNSIGNED_NORMALIZED) {
- entry.LogicOpEnable = true;
- entry.LogicOpFunction =
- intel_translate_logic_op(ctx->Color.LogicOp);
- }
- } else if (ctx->Color.BlendEnabled & (1 << i) && !integer &&
- !ctx->Color._AdvancedBlendMode) {
- GLenum eqRGB = ctx->Color.Blend[i].EquationRGB;
- GLenum eqA = ctx->Color.Blend[i].EquationA;
- GLenum srcRGB = ctx->Color.Blend[i].SrcRGB;
- GLenum dstRGB = ctx->Color.Blend[i].DstRGB;
- GLenum srcA = ctx->Color.Blend[i].SrcA;
- GLenum dstA = ctx->Color.Blend[i].DstA;
-
- if (eqRGB == GL_MIN || eqRGB == GL_MAX)
- srcRGB = dstRGB = GL_ONE;
-
- if (eqA == GL_MIN || eqA == GL_MAX)
- srcA = dstA = GL_ONE;
-
- /* Due to hardware limitations, the destination may have information
- * in an alpha channel even when the format specifies no alpha
- * channel. In order to avoid getting any incorrect blending due to
- * that alpha channel, coerce the blend factors to values that will
- * not read the alpha channel, but will instead use the correct
- * implicit value for alpha.
- */
- if (rb && !_mesa_base_format_has_channel(rb->_BaseFormat,
- GL_TEXTURE_ALPHA_TYPE)) {
- srcRGB = brw_fix_xRGB_alpha(srcRGB);
- srcA = brw_fix_xRGB_alpha(srcA);
- dstRGB = brw_fix_xRGB_alpha(dstRGB);
- dstA = brw_fix_xRGB_alpha(dstA);
- }
-
- /* From the BLEND_STATE docs, DWord 0, Bit 29 (AlphaToOne Enable):
- * "If Dual Source Blending is enabled, this bit must be disabled."
- *
- * We override SRC1_ALPHA to ONE and ONE_MINUS_SRC1_ALPHA to ZERO,
- * and leave it enabled anyway.
- */
- if (ctx->Color.Blend[i]._UsesDualSrc && blend.AlphaToOneEnable) {
- srcRGB = fix_dual_blend_alpha_to_one(srcRGB);
- srcA = fix_dual_blend_alpha_to_one(srcA);
- dstRGB = fix_dual_blend_alpha_to_one(dstRGB);
- dstA = fix_dual_blend_alpha_to_one(dstA);
- }
-
- entry.ColorBufferBlendEnable = true;
- entry.DestinationBlendFactor = blend_factor(dstRGB);
- entry.SourceBlendFactor = blend_factor(srcRGB);
- entry.DestinationAlphaBlendFactor = blend_factor(dstA);
- entry.SourceAlphaBlendFactor = blend_factor(srcA);
- entry.ColorBlendFunction = blend_eqn(eqRGB);
- entry.AlphaBlendFunction = blend_eqn(eqA);
-
- if (srcA != srcRGB || dstA != dstRGB || eqA != eqRGB)
- blend.IndependentAlphaBlendEnable = true;
- }
+ blend.IndependentAlphaBlendEnable =
+ set_blend_entry_bits(brw, &entry, i, blend.AlphaToOneEnable) ||
+ blend.IndependentAlphaBlendEnable;
/* See section 8.1.6 "Pre-Blend Color Clamping" of the
* SandyBridge PRM Volume 2 Part 1 for HW requirements.
/* ---------------------------------------------------------------------- */
-#if GEN_GEN >= 6
static void
genX(upload_color_calc_state)(struct brw_context *brw)
{
struct gl_context *ctx = &brw->ctx;
brw_state_emit(brw, GENX(COLOR_CALC_STATE), 64, &brw->cc.state_offset, cc) {
+#if GEN_GEN <= 5
+ cc.IndependentAlphaBlendEnable =
+ set_blend_entry_bits(brw, &cc, 0, false);
+ set_depth_stencil_bits(brw, &cc);
+
+ if (ctx->Color.AlphaEnabled &&
+ ctx->DrawBuffer->_NumColorDrawBuffers <= 1) {
+ cc.AlphaTestEnable = true;
+ cc.AlphaTestFunction =
+ intel_translate_compare_func(ctx->Color.AlphaFunc);
+ }
+
+ cc.ColorDitherEnable = ctx->Color.DitherFlag;
+
+ cc.StatisticsEnable = brw->stats_wm;
+
+ cc.CCViewportStatePointer =
+ instruction_ro_bo(brw->batch.bo, brw->cc.vp_offset);
+#else
/* _NEW_COLOR */
- cc.AlphaTestFormat = ALPHATEST_UNORM8;
- UNCLAMPED_FLOAT_TO_UBYTE(cc.AlphaReferenceValueAsUNORM8,
- ctx->Color.AlphaRef);
+ cc.BlendConstantColorRed = ctx->Color.BlendColorUnclamped[0];
+ cc.BlendConstantColorGreen = ctx->Color.BlendColorUnclamped[1];
+ cc.BlendConstantColorBlue = ctx->Color.BlendColorUnclamped[2];
+ cc.BlendConstantColorAlpha = ctx->Color.BlendColorUnclamped[3];
#if GEN_GEN < 9
/* _NEW_STENCIL */
cc.StencilReferenceValue = _mesa_get_stencil_ref(ctx, 0);
cc.BackfaceStencilReferenceValue =
_mesa_get_stencil_ref(ctx, ctx->Stencil._BackFace);
+#endif
+
#endif
/* _NEW_COLOR */
- cc.BlendConstantColorRed = ctx->Color.BlendColorUnclamped[0];
- cc.BlendConstantColorGreen = ctx->Color.BlendColorUnclamped[1];
- cc.BlendConstantColorBlue = ctx->Color.BlendColorUnclamped[2];
- cc.BlendConstantColorAlpha = ctx->Color.BlendColorUnclamped[3];
+ UNCLAMPED_FLOAT_TO_UBYTE(cc.AlphaReferenceValueAsUNORM8,
+ ctx->Color.AlphaRef);
}
+#if GEN_GEN >= 6
brw_batch_emit(brw, GENX(3DSTATE_CC_STATE_POINTERS), ptr) {
ptr.ColorCalcStatePointer = brw->cc.state_offset;
#if GEN_GEN != 7
ptr.ColorCalcStatePointerValid = true;
#endif
}
+#else
+ brw->ctx.NewDriverState |= BRW_NEW_GEN4_UNIT_STATE;
+#endif
}
static const struct brw_tracked_state genX(color_calc_state) = {
.dirty = {
.mesa = _NEW_COLOR |
- _NEW_STENCIL,
+ _NEW_STENCIL |
+ (GEN_GEN <= 5 ? _NEW_BUFFERS |
+ _NEW_DEPTH
+ : 0),
.brw = BRW_NEW_BATCH |
BRW_NEW_BLORP |
- BRW_NEW_CC_STATE |
- BRW_NEW_STATE_BASE_ADDRESS,
+ (GEN_GEN <= 5 ? BRW_NEW_CC_VP |
+ BRW_NEW_STATS_WM
+ : BRW_NEW_CC_STATE |
+ BRW_NEW_STATE_BASE_ADDRESS),
},
.emit = genX(upload_color_calc_state),
};
-#endif
/* ---------------------------------------------------------------------- */
/* ---------------------------------------------------------------------- */
+#if GEN_GEN <= 5
+
+static void genX(upload_blend_constant_color)(struct brw_context *brw)
+{
+ struct gl_context *ctx = &brw->ctx;
+
+ brw_batch_emit(brw, GENX(3DSTATE_CONSTANT_COLOR), blend_cc) {
+ blend_cc.BlendConstantColorRed = ctx->Color.BlendColorUnclamped[0];
+ blend_cc.BlendConstantColorGreen = ctx->Color.BlendColorUnclamped[1];
+ blend_cc.BlendConstantColorBlue = ctx->Color.BlendColorUnclamped[2];
+ blend_cc.BlendConstantColorAlpha = ctx->Color.BlendColorUnclamped[3];
+ }
+}
+
+static const struct brw_tracked_state genX(blend_constant_color) = {
+ .dirty = {
+ .mesa = _NEW_COLOR,
+ .brw = BRW_NEW_CONTEXT |
+ BRW_NEW_BLORP,
+ },
+ .emit = genX(upload_blend_constant_color)
+};
+#endif
+
+/* ---------------------------------------------------------------------- */
+
void
genX(init_atoms)(struct brw_context *brw)
{
&brw_recalculate_urb_fence,
&genX(cc_vp),
- &brw_cc_unit,
+ &genX(color_calc_state),
/* Surface state setup. Must come before the VS/WM unit. The binding
* table upload must be last.
&brw_invariant_state,
&brw_binding_table_pointers,
- &brw_blend_constant_color,
+ &genX(blend_constant_color),
&brw_depthbuffer,