static void
i915LogicOp(struct gl_context * ctx, GLenum opcode)
{
struct i915_context *i915 = I915_CONTEXT(ctx);
int tmp = intel_translate_logic_op(opcode);
DBG("%s\n", __FUNCTION__);
I915_STATECHANGE(i915, I915_UPLOAD_CTX);
i915->state.Ctx[I915_CTXREG_STATE4] &= ~LOGICOP_MASK;
i915->state.Ctx[I915_CTXREG_STATE4] |= LOGIC_OP_FUNC(tmp);
}
static void
i830LogicOp(struct gl_context * ctx, GLenum opcode)
{
struct i830_context *i830 = i830_context(ctx);
int tmp = intel_translate_logic_op(opcode);
DBG("%s\n", __FUNCTION__);
I830_STATECHANGE(i830, I830_UPLOAD_CTX);
i830->state.Ctx[I830_CTXREG_STATE4] &= ~LOGICOP_MASK;
i830->state.Ctx[I830_CTXREG_STATE4] |= LOGIC_OP_FUNC(tmp);
}
static void i830LogicOp(GLcontext *ctx, GLenum opcode)
{
i830ContextPtr i830 = I830_CONTEXT(ctx);
int tmp = intel_translate_logic_op( opcode );
if (INTEL_DEBUG&DEBUG_DRI)
fprintf(stderr, "%s\n", __FUNCTION__);
I830_STATECHANGE(i830, I830_UPLOAD_CTX);
i830->state.Ctx[I830_CTXREG_STATE4] &= ~LOGICOP_MASK;
i830->state.Ctx[I830_CTXREG_STATE4] |= LOGIC_OP_FUNC(tmp);
}
static void
prepare_blend_state(struct brw_context *brw)
{
struct gl_context *ctx = &brw->intel.ctx;
struct gen6_blend_state *blend;
int b;
int nr_draw_buffers = ctx->DrawBuffer->_NumColorDrawBuffers;
int size;
/* We need at least one BLEND_STATE written, because we might do
* thread dispatch even if _NumColorDrawBuffers is 0 (for example
* for computed depth or alpha test), which will do an FB write
* with render target 0, which will reference BLEND_STATE[0] for
* alpha test enable.
*/
if (nr_draw_buffers == 0 && ctx->Color.AlphaEnabled)
nr_draw_buffers = 1;
size = sizeof(*blend) * nr_draw_buffers;
blend = brw_state_batch(brw, AUB_TRACE_BLEND_STATE,
size, 64, &brw->cc.blend_state_offset);
memset(blend, 0, size);
for (b = 0; b < nr_draw_buffers; b++) {
/* _NEW_COLOR */
if (ctx->Color._LogicOpEnabled) {
struct gl_renderbuffer *rb = ctx->DrawBuffer->_ColorDrawBuffers[b];
/* _NEW_BUFFERS */
/* Floating point RTs should have no effect from LogicOp,
* except for disabling of blending
*/
if (_mesa_get_format_datatype(rb->Format) != GL_FLOAT) {
blend[b].blend1.logic_op_enable = 1;
blend[b].blend1.logic_op_func =
intel_translate_logic_op(ctx->Color.LogicOp);
}
} else if (ctx->Color.BlendEnabled & (1 << b)) {
GLenum eqRGB = ctx->Color.Blend[0].EquationRGB;
GLenum eqA = ctx->Color.Blend[0].EquationA;
GLenum srcRGB = ctx->Color.Blend[0].SrcRGB;
GLenum dstRGB = ctx->Color.Blend[0].DstRGB;
GLenum srcA = ctx->Color.Blend[0].SrcA;
GLenum dstA = ctx->Color.Blend[0].DstA;
if (eqRGB == GL_MIN || eqRGB == GL_MAX) {
srcRGB = dstRGB = GL_ONE;
}
if (eqA == GL_MIN || eqA == GL_MAX) {
srcA = dstA = GL_ONE;
}
blend[b].blend0.dest_blend_factor = brw_translate_blend_factor(dstRGB);
blend[b].blend0.source_blend_factor = brw_translate_blend_factor(srcRGB);
blend[b].blend0.blend_func = brw_translate_blend_equation(eqRGB);
blend[b].blend0.ia_dest_blend_factor = brw_translate_blend_factor(dstA);
blend[b].blend0.ia_source_blend_factor = brw_translate_blend_factor(srcA);
blend[b].blend0.ia_blend_func = brw_translate_blend_equation(eqA);
blend[b].blend0.blend_enable = 1;
blend[b].blend0.ia_blend_enable = (srcA != srcRGB ||
dstA != dstRGB ||
eqA != eqRGB);
}
/* _NEW_COLOR */
if (ctx->Color.AlphaEnabled) {
blend[b].blend1.alpha_test_enable = 1;
blend[b].blend1.alpha_test_func =
intel_translate_compare_func(ctx->Color.AlphaFunc);
}
/* _NEW_COLOR */
if (ctx->Color.DitherFlag) {
blend[b].blend1.dither_enable = 1;
blend[b].blend1.y_dither_offset = 0;
blend[b].blend1.x_dither_offset = 0;
}
blend[b].blend1.write_disable_r = !ctx->Color.ColorMask[b][0];
blend[b].blend1.write_disable_g = !ctx->Color.ColorMask[b][1];
blend[b].blend1.write_disable_b = !ctx->Color.ColorMask[b][2];
blend[b].blend1.write_disable_a = !ctx->Color.ColorMask[b][3];
}
brw->state.dirty.cache |= CACHE_NEW_BLEND_STATE;
}
static void
gen6_upload_blend_state(struct brw_context *brw)
{
bool is_buffer_zero_integer_format = false;
struct gl_context *ctx = &brw->intel.ctx;
struct gen6_blend_state *blend;
int b;
int nr_draw_buffers = ctx->DrawBuffer->_NumColorDrawBuffers;
int size;
/* We need at least one BLEND_STATE written, because we might do
* thread dispatch even if _NumColorDrawBuffers is 0 (for example
* for computed depth or alpha test), which will do an FB write
* with render target 0, which will reference BLEND_STATE[0] for
* alpha test enable.
*/
if (nr_draw_buffers == 0 && ctx->Color.AlphaEnabled)
nr_draw_buffers = 1;
size = sizeof(*blend) * nr_draw_buffers;
blend = brw_state_batch(brw, AUB_TRACE_BLEND_STATE,
size, 64, &brw->cc.blend_state_offset);
memset(blend, 0, size);
for (b = 0; b < nr_draw_buffers; b++) {
/* _NEW_BUFFERS */
struct gl_renderbuffer *rb = ctx->DrawBuffer->_ColorDrawBuffers[b];
GLenum rb_type;
bool integer;
if (rb)
rb_type = _mesa_get_format_datatype(rb->Format);
else
rb_type = GL_UNSIGNED_NORMALIZED;
/* 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."
*/
integer = (rb_type == GL_INT || rb_type == GL_UNSIGNED_INT);
if(b == 0 && integer)
is_buffer_zero_integer_format = true;
/* _NEW_COLOR */
if (ctx->Color.ColorLogicOpEnabled) {
/* Floating point RTs should have no effect from LogicOp,
* except for disabling of blending, but other types should.
*
* However, from the Sandy Bridge PRM, Vol 2 Par 1, Section 8.1.11,
* "Logic Ops",
*
* "Logic Ops are only supported on *_UNORM surfaces (excluding
* _SRGB variants), otherwise Logic Ops must be DISABLED."
*/
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_lookup_enum_by_nr(ctx->Color.LogicOp),
_mesa_lookup_enum_by_nr(rb_type));
if (rb_type == GL_UNSIGNED_NORMALIZED) {
blend[b].blend1.logic_op_enable = 1;
blend[b].blend1.logic_op_func =
intel_translate_logic_op(ctx->Color.LogicOp);
}
} else if (ctx->Color.BlendEnabled & (1 << b) && !integer) {
GLenum eqRGB = ctx->Color.Blend[b].EquationRGB;
GLenum eqA = ctx->Color.Blend[b].EquationA;
GLenum srcRGB = ctx->Color.Blend[b].SrcRGB;
GLenum dstRGB = ctx->Color.Blend[b].DstRGB;
GLenum srcA = ctx->Color.Blend[b].SrcA;
GLenum dstA = ctx->Color.Blend[b].DstA;
if (eqRGB == GL_MIN || eqRGB == GL_MAX) {
srcRGB = dstRGB = GL_ONE;
}
if (eqA == GL_MIN || eqA == GL_MAX) {
srcA = dstA = GL_ONE;
}
blend[b].blend0.dest_blend_factor = brw_translate_blend_factor(dstRGB);
blend[b].blend0.source_blend_factor = brw_translate_blend_factor(srcRGB);
blend[b].blend0.blend_func = brw_translate_blend_equation(eqRGB);
blend[b].blend0.ia_dest_blend_factor = brw_translate_blend_factor(dstA);
blend[b].blend0.ia_source_blend_factor = brw_translate_blend_factor(srcA);
blend[b].blend0.ia_blend_func = brw_translate_blend_equation(eqA);
blend[b].blend0.blend_enable = 1;
blend[b].blend0.ia_blend_enable = (srcA != srcRGB ||
dstA != dstRGB ||
eqA != eqRGB);
}
/* See section 8.1.6 "Pre-Blend Color Clamping" of the
* SandyBridge PRM Volume 2 Part 1 for HW requirements.
*
* We do our ARB_color_buffer_float CLAMP_FRAGMENT_COLOR
* clamping in the fragment shader. For its clamping of
* blending, the spec says:
*
* "RESOLVED: For fixed-point color buffers, the inputs and
* the result of the blending equation are clamped. For
* floating-point color buffers, no clamping occurs."
*
* So, generally, we want clamping to the render target's range.
* And, good news, the hardware tables for both pre- and
* post-blend color clamping are either ignored, or any are
* allowed, or clamping is required but RT range clamping is a
* valid option.
*/
blend[b].blend1.pre_blend_clamp_enable = 1;
blend[b].blend1.post_blend_clamp_enable = 1;
blend[b].blend1.clamp_range = BRW_RENDERTARGET_CLAMPRANGE_FORMAT;
/* _NEW_COLOR */
if (ctx->Color.AlphaEnabled && !integer) {
blend[b].blend1.alpha_test_enable = 1;
blend[b].blend1.alpha_test_func =
intel_translate_compare_func(ctx->Color.AlphaFunc);
}
/* _NEW_COLOR */
if (ctx->Color.DitherFlag && !integer) {
blend[b].blend1.dither_enable = 1;
blend[b].blend1.y_dither_offset = 0;
blend[b].blend1.x_dither_offset = 0;
}
blend[b].blend1.write_disable_r = !ctx->Color.ColorMask[b][0];
blend[b].blend1.write_disable_g = !ctx->Color.ColorMask[b][1];
blend[b].blend1.write_disable_b = !ctx->Color.ColorMask[b][2];
blend[b].blend1.write_disable_a = !ctx->Color.ColorMask[b][3];
/* OpenGL specification 3.3 (page 196), section 4.1.3 says:
* "If drawbuffer zero is not NONE and the buffer it references has an
* integer format, the SAMPLE_ALPHA_TO_COVERAGE and SAMPLE_ALPHA_TO_ONE
* operations are skipped."
*/
if(!is_buffer_zero_integer_format) {
/* _NEW_MULTISAMPLE */
blend[b].blend1.alpha_to_coverage =
ctx->Multisample._Enabled && ctx->Multisample.SampleAlphaToCoverage;
/* From SandyBridge PRM, volume 2 Part 1, section 8.2.3, BLEND_STATE:
* DWord 1, Bit 30 (AlphaToOne Enable):
* "If Dual Source Blending is enabled, this bit must be disabled"
*/
WARN_ONCE(ctx->Color.Blend[b]._UsesDualSrc &&
ctx->Multisample._Enabled &&
ctx->Multisample.SampleAlphaToOne,
"HW workaround: disabling alpha to one with dual src "
"blending\n");
if (ctx->Color.Blend[b]._UsesDualSrc)
blend[b].blend1.alpha_to_one = false;
else
blend[b].blend1.alpha_to_one =
ctx->Multisample._Enabled && ctx->Multisample.SampleAlphaToOne;
blend[b].blend1.alpha_to_coverage_dither = (brw->intel.gen >= 7);
}
else {
blend[b].blend1.alpha_to_coverage = false;
blend[b].blend1.alpha_to_one = false;
}
}
brw->state.dirty.cache |= CACHE_NEW_BLEND_STATE;
}
/**
* Creates the state cache entry for the given CC unit key.
*/
static void upload_cc_unit(struct brw_context *brw)
{
struct gl_context *ctx = &brw->ctx;
struct brw_cc_unit_state *cc;
cc = brw_state_batch(brw, AUB_TRACE_CC_STATE,
sizeof(*cc), 64, &brw->cc.state_offset);
memset(cc, 0, sizeof(*cc));
/* _NEW_STENCIL | _NEW_BUFFERS */
if (ctx->Stencil._Enabled) {
const unsigned back = ctx->Stencil._BackFace;
cc->cc0.stencil_enable = 1;
cc->cc0.stencil_func =
intel_translate_compare_func(ctx->Stencil.Function[0]);
cc->cc0.stencil_fail_op =
intel_translate_stencil_op(ctx->Stencil.FailFunc[0]);
cc->cc0.stencil_pass_depth_fail_op =
intel_translate_stencil_op(ctx->Stencil.ZFailFunc[0]);
cc->cc0.stencil_pass_depth_pass_op =
intel_translate_stencil_op(ctx->Stencil.ZPassFunc[0]);
cc->cc1.stencil_ref = _mesa_get_stencil_ref(ctx, 0);
cc->cc1.stencil_write_mask = ctx->Stencil.WriteMask[0];
cc->cc1.stencil_test_mask = ctx->Stencil.ValueMask[0];
if (ctx->Stencil._TestTwoSide) {
cc->cc0.bf_stencil_enable = 1;
cc->cc0.bf_stencil_func =
intel_translate_compare_func(ctx->Stencil.Function[back]);
cc->cc0.bf_stencil_fail_op =
intel_translate_stencil_op(ctx->Stencil.FailFunc[back]);
cc->cc0.bf_stencil_pass_depth_fail_op =
intel_translate_stencil_op(ctx->Stencil.ZFailFunc[back]);
cc->cc0.bf_stencil_pass_depth_pass_op =
intel_translate_stencil_op(ctx->Stencil.ZPassFunc[back]);
cc->cc1.bf_stencil_ref = _mesa_get_stencil_ref(ctx, back);
cc->cc2.bf_stencil_write_mask = ctx->Stencil.WriteMask[back];
cc->cc2.bf_stencil_test_mask = ctx->Stencil.ValueMask[back];
}
/* Not really sure about this:
*/
if (ctx->Stencil.WriteMask[0] ||
(ctx->Stencil._TestTwoSide && ctx->Stencil.WriteMask[back]))
cc->cc0.stencil_write_enable = 1;
}
/* _NEW_COLOR */
if (ctx->Color.ColorLogicOpEnabled && ctx->Color.LogicOp != GL_COPY) {
cc->cc2.logicop_enable = 1;
cc->cc5.logicop_func = intel_translate_logic_op(ctx->Color.LogicOp);
} else if (ctx->Color.BlendEnabled) {
GLenum eqRGB = ctx->Color.Blend[0].EquationRGB;
GLenum eqA = ctx->Color.Blend[0].EquationA;
GLenum srcRGB = ctx->Color.Blend[0].SrcRGB;
GLenum dstRGB = ctx->Color.Blend[0].DstRGB;
GLenum srcA = ctx->Color.Blend[0].SrcA;
GLenum dstA = ctx->Color.Blend[0].DstA;
/* If the renderbuffer is XRGB, we have to frob the blend function to
* force the destination alpha to 1.0. This means replacing GL_DST_ALPHA
* with GL_ONE and GL_ONE_MINUS_DST_ALPHA with GL_ZERO.
*/
if (ctx->DrawBuffer->Visual.alphaBits == 0) {
srcRGB = brw_fix_xRGB_alpha(srcRGB);
srcA = brw_fix_xRGB_alpha(srcA);
dstRGB = brw_fix_xRGB_alpha(dstRGB);
dstA = brw_fix_xRGB_alpha(dstA);
}
if (eqRGB == GL_MIN || eqRGB == GL_MAX) {
srcRGB = dstRGB = GL_ONE;
}
if (eqA == GL_MIN || eqA == GL_MAX) {
srcA = dstA = GL_ONE;
}
cc->cc6.dest_blend_factor = brw_translate_blend_factor(dstRGB);
cc->cc6.src_blend_factor = brw_translate_blend_factor(srcRGB);
cc->cc6.blend_function = brw_translate_blend_equation(eqRGB);
cc->cc5.ia_dest_blend_factor = brw_translate_blend_factor(dstA);
cc->cc5.ia_src_blend_factor = brw_translate_blend_factor(srcA);
cc->cc5.ia_blend_function = brw_translate_blend_equation(eqA);
cc->cc3.blend_enable = 1;
cc->cc3.ia_blend_enable = (srcA != srcRGB ||
dstA != dstRGB ||
eqA != eqRGB);
}
/* _NEW_BUFFERS */
if (ctx->Color.AlphaEnabled && ctx->DrawBuffer->_NumColorDrawBuffers <= 1) {
cc->cc3.alpha_test = 1;
cc->cc3.alpha_test_func =
intel_translate_compare_func(ctx->Color.AlphaFunc);
cc->cc3.alpha_test_format = BRW_ALPHATEST_FORMAT_UNORM8;
UNCLAMPED_FLOAT_TO_UBYTE(cc->cc7.alpha_ref.ub[0], ctx->Color.AlphaRef);
}
if (ctx->Color.DitherFlag) {
cc->cc5.dither_enable = 1;
cc->cc6.y_dither_offset = 0;
cc->cc6.x_dither_offset = 0;
}
/* _NEW_DEPTH */
if (ctx->Depth.Test) {
cc->cc2.depth_test = 1;
cc->cc2.depth_test_function =
intel_translate_compare_func(ctx->Depth.Func);
cc->cc2.depth_write_enable = ctx->Depth.Mask;
}
if (brw->stats_wm || unlikely(INTEL_DEBUG & DEBUG_STATS))
cc->cc5.statistics_enable = 1;
/* CACHE_NEW_CC_VP */
cc->cc4.cc_viewport_state_offset = (brw->batch.bo->offset +
brw->cc.vp_offset) >> 5; /* reloc */
brw->state.dirty.cache |= CACHE_NEW_CC_UNIT;
/* Emit CC viewport relocation */
drm_intel_bo_emit_reloc(brw->batch.bo,
(brw->cc.state_offset +
offsetof(struct brw_cc_unit_state, cc4)),
brw->batch.bo, brw->cc.vp_offset,
I915_GEM_DOMAIN_INSTRUCTION, 0);
}
/**
* Creates the state cache entry for the given CC unit key.
*/
static dri_bo *
cc_unit_create_from_key(struct brw_context *brw, struct brw_cc_unit_key *key)
{
struct brw_cc_unit_state cc;
dri_bo *bo;
memset(&cc, 0, sizeof(cc));
/* _NEW_STENCIL */
if (key->stencil) {
cc.cc0.stencil_enable = 1;
cc.cc0.stencil_func =
intel_translate_compare_func(key->stencil_func[0]);
cc.cc0.stencil_fail_op =
intel_translate_stencil_op(key->stencil_fail_op[0]);
cc.cc0.stencil_pass_depth_fail_op =
intel_translate_stencil_op(key->stencil_pass_depth_fail_op[0]);
cc.cc0.stencil_pass_depth_pass_op =
intel_translate_stencil_op(key->stencil_pass_depth_pass_op[0]);
cc.cc1.stencil_ref = key->stencil_ref[0];
cc.cc1.stencil_write_mask = key->stencil_write_mask[0];
cc.cc1.stencil_test_mask = key->stencil_test_mask[0];
if (key->stencil_two_side) {
cc.cc0.bf_stencil_enable = 1;
cc.cc0.bf_stencil_func =
intel_translate_compare_func(key->stencil_func[1]);
cc.cc0.bf_stencil_fail_op =
intel_translate_stencil_op(key->stencil_fail_op[1]);
cc.cc0.bf_stencil_pass_depth_fail_op =
intel_translate_stencil_op(key->stencil_pass_depth_fail_op[1]);
cc.cc0.bf_stencil_pass_depth_pass_op =
intel_translate_stencil_op(key->stencil_pass_depth_pass_op[1]);
cc.cc1.bf_stencil_ref = key->stencil_ref[1];
cc.cc2.bf_stencil_write_mask = key->stencil_write_mask[1];
cc.cc2.bf_stencil_test_mask = key->stencil_test_mask[1];
}
/* Not really sure about this:
*/
if (key->stencil_write_mask[0] ||
(key->stencil_two_side && key->stencil_write_mask[1]))
cc.cc0.stencil_write_enable = 1;
}
/* _NEW_COLOR */
if (key->logic_op != GL_COPY) {
cc.cc2.logicop_enable = 1;
cc.cc5.logicop_func = intel_translate_logic_op(key->logic_op);
} else if (key->color_blend) {
GLenum eqRGB = key->blend_eq_rgb;
GLenum eqA = key->blend_eq_a;
GLenum srcRGB = key->blend_src_rgb;
GLenum dstRGB = key->blend_dst_rgb;
GLenum srcA = key->blend_src_a;
GLenum dstA = key->blend_dst_a;
if (eqRGB == GL_MIN || eqRGB == GL_MAX) {
srcRGB = dstRGB = GL_ONE;
}
if (eqA == GL_MIN || eqA == GL_MAX) {
srcA = dstA = GL_ONE;
}
cc.cc6.dest_blend_factor = brw_translate_blend_factor(dstRGB);
cc.cc6.src_blend_factor = brw_translate_blend_factor(srcRGB);
cc.cc6.blend_function = brw_translate_blend_equation(eqRGB);
cc.cc5.ia_dest_blend_factor = brw_translate_blend_factor(dstA);
cc.cc5.ia_src_blend_factor = brw_translate_blend_factor(srcA);
cc.cc5.ia_blend_function = brw_translate_blend_equation(eqA);
cc.cc3.blend_enable = 1;
cc.cc3.ia_blend_enable = (srcA != srcRGB ||
dstA != dstRGB ||
eqA != eqRGB);
}
if (key->alpha_enabled) {
cc.cc3.alpha_test = 1;
cc.cc3.alpha_test_func = intel_translate_compare_func(key->alpha_func);
cc.cc3.alpha_test_format = BRW_ALPHATEST_FORMAT_UNORM8;
UNCLAMPED_FLOAT_TO_UBYTE(cc.cc7.alpha_ref.ub[0], key->alpha_ref);
}
if (key->dither) {
cc.cc5.dither_enable = 1;
cc.cc6.y_dither_offset = 0;
cc.cc6.x_dither_offset = 0;
}
/* _NEW_DEPTH */
if (key->depth_test) {
cc.cc2.depth_test = 1;
cc.cc2.depth_test_function = intel_translate_compare_func(key->depth_func);
cc.cc2.depth_write_enable = key->depth_write;
}
/* CACHE_NEW_CC_VP */
cc.cc4.cc_viewport_state_offset = brw->cc.vp_bo->offset >> 5; /* reloc */
if (INTEL_DEBUG & DEBUG_STATS)
cc.cc5.statistics_enable = 1;
bo = brw_upload_cache(&brw->cache, BRW_CC_UNIT,
key, sizeof(*key),
&brw->cc.vp_bo, 1,
&cc, sizeof(cc),
NULL, NULL);
/* Emit CC viewport relocation */
dri_bo_emit_reloc(bo,
I915_GEM_DOMAIN_INSTRUCTION,
0,
0,
offsetof(struct brw_cc_unit_state, cc4),
brw->cc.vp_bo);
return bo;
}