#include "program/program.h"
#include "program/prog_parameter.h"
#include "program/prog_statevars.h"
+#include "main/framebuffer.h"
#include "intel_batchbuffer.h"
static void
/* BRW_NEW_FRAGMENT_PROGRAM */
const struct brw_fragment_program *fp =
brw_fragment_program_const(brw->fragment_program);
- /* CACHE_NEW_WM_PROG */
+ /* BRW_NEW_FS_PROG_DATA */
const struct brw_wm_prog_data *prog_data = brw->wm.prog_data;
gen6_upload_push_constants(brw, &fp->program.Base, &prog_data->base,
const struct brw_tracked_state gen6_wm_push_constants = {
.dirty = {
.mesa = _NEW_PROGRAM_CONSTANTS,
- .brw = (BRW_NEW_BATCH |
- BRW_NEW_FRAGMENT_PROGRAM |
- BRW_NEW_PUSH_CONSTANT_ALLOCATION),
- .cache = CACHE_NEW_WM_PROG,
+ .brw = BRW_NEW_BATCH |
+ BRW_NEW_FRAGMENT_PROGRAM |
+ BRW_NEW_FS_PROG_DATA |
+ BRW_NEW_PUSH_CONSTANT_ALLOCATION,
},
.emit = gen6_upload_wm_push_constants,
};
-static void
-upload_wm_state(struct brw_context *brw)
+void
+gen6_upload_wm_state(struct brw_context *brw,
+ const struct brw_fragment_program *fp,
+ const struct brw_wm_prog_data *prog_data,
+ const struct brw_stage_state *stage_state,
+ bool multisampled_fbo, int min_inv_per_frag,
+ bool dual_source_blend_enable, bool kill_enable,
+ bool color_buffer_write_enable, bool msaa_enabled,
+ bool line_stipple_enable, bool polygon_stipple_enable,
+ bool statistic_enable)
{
- struct gl_context *ctx = &brw->ctx;
- /* BRW_NEW_FRAGMENT_PROGRAM */
- const struct brw_fragment_program *fp =
- brw_fragment_program_const(brw->fragment_program);
- /* CACHE_NEW_WM_PROG */
- const struct brw_wm_prog_data *prog_data = brw->wm.prog_data;
uint32_t dw2, dw4, dw5, dw6, ksp0, ksp2;
- /* _NEW_BUFFERS */
- bool multisampled_fbo = ctx->DrawBuffer->Visual.samples > 1;
-
/* 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]):
/* Pointer to the WM constant buffer. Covered by the set of
* state flags from gen6_upload_wm_push_constants.
*/
- OUT_BATCH(brw->wm.base.push_const_offset +
- brw->wm.base.push_const_size - 1);
+ OUT_BATCH(stage_state->push_const_offset +
+ stage_state->push_const_size - 1);
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
}
dw2 = dw4 = dw5 = dw6 = ksp2 = 0;
- dw4 |= GEN6_WM_STATISTICS_ENABLE;
+
+ if (statistic_enable)
+ dw4 |= GEN6_WM_STATISTICS_ENABLE;
+
dw5 |= GEN6_WM_LINE_AA_WIDTH_1_0;
dw5 |= GEN6_WM_LINE_END_CAP_AA_WIDTH_0_5;
- /* Use ALT floating point mode for ARB fragment programs, because they
- * require 0^0 == 1. Even though _CurrentFragmentProgram is used for
- * rendering, CurrentProgram[MESA_SHADER_FRAGMENT] is used for this check
- * to differentiate between the GLSL and non-GLSL cases.
- */
- if (ctx->_Shader->CurrentProgram[MESA_SHADER_FRAGMENT] == NULL)
+ if (prog_data->base.use_alt_mode)
dw2 |= GEN6_WM_FLOATING_POINT_MODE_ALT;
- dw2 |= (ALIGN(brw->wm.base.sampler_count, 4) / 4) <<
+ dw2 |= (ALIGN(stage_state->sampler_count, 4) / 4) <<
GEN6_WM_SAMPLER_COUNT_SHIFT;
dw2 |= ((prog_data->base.binding_table.size_bytes / 4) <<
dw5 |= (brw->max_wm_threads - 1) << GEN6_WM_MAX_THREADS_SHIFT;
- /* In case of non 1x per sample shading, only one of SIMD8 and SIMD16
- * should be enabled. We do 'SIMD16 only' dispatch if a SIMD16 shader
- * is successfully compiled. In majority of the cases that bring us
- * better performance than 'SIMD8 only' dispatch.
- */
- int min_inv_per_frag =
- _mesa_get_min_invocations_per_fragment(ctx, brw->fragment_program, false);
assert(min_inv_per_frag >= 1);
- if (prog_data->prog_offset_16) {
+ if (prog_data->prog_offset_16 || prog_data->no_8) {
dw5 |= GEN6_WM_16_DISPATCH_ENABLE;
- if (min_inv_per_frag == 1) {
+ if (!prog_data->no_8 && min_inv_per_frag == 1) {
dw5 |= GEN6_WM_8_DISPATCH_ENABLE;
dw4 |= (prog_data->base.dispatch_grf_start_reg <<
GEN6_WM_DISPATCH_START_GRF_SHIFT_0);
dw4 |= (prog_data->dispatch_grf_start_reg_16 <<
GEN6_WM_DISPATCH_START_GRF_SHIFT_2);
- ksp0 = brw->wm.base.prog_offset;
- ksp2 = brw->wm.base.prog_offset + prog_data->prog_offset_16;
+ ksp0 = stage_state->prog_offset;
+ ksp2 = stage_state->prog_offset + prog_data->prog_offset_16;
} else {
dw4 |= (prog_data->dispatch_grf_start_reg_16 <<
GEN6_WM_DISPATCH_START_GRF_SHIFT_0);
- ksp0 = brw->wm.base.prog_offset + prog_data->prog_offset_16;
+ ksp0 = stage_state->prog_offset + prog_data->prog_offset_16;
}
}
else {
dw5 |= GEN6_WM_8_DISPATCH_ENABLE;
dw4 |= (prog_data->base.dispatch_grf_start_reg <<
GEN6_WM_DISPATCH_START_GRF_SHIFT_0);
- ksp0 = brw->wm.base.prog_offset;
+ ksp0 = stage_state->prog_offset;
}
- /* CACHE_NEW_WM_PROG | _NEW_COLOR */
- if (prog_data->dual_src_blend &&
- (ctx->Color.BlendEnabled & 1) &&
- ctx->Color.Blend[0]._UsesDualSrc) {
+ if (dual_source_blend_enable)
dw5 |= GEN6_WM_DUAL_SOURCE_BLEND_ENABLE;
- }
- /* _NEW_LINE */
- if (ctx->Line.StippleFlag)
+ if (line_stipple_enable)
dw5 |= GEN6_WM_LINE_STIPPLE_ENABLE;
- /* _NEW_POLYGON */
- if (ctx->Polygon.StippleFlag)
+ if (polygon_stipple_enable)
dw5 |= GEN6_WM_POLYGON_STIPPLE_ENABLE;
/* BRW_NEW_FRAGMENT_PROGRAM */
dw6 |= prog_data->barycentric_interp_modes <<
GEN6_WM_BARYCENTRIC_INTERPOLATION_MODE_SHIFT;
- /* _NEW_COLOR, _NEW_MULTISAMPLE */
- if (fp->program.UsesKill || ctx->Color.AlphaEnabled ||
- ctx->Multisample.SampleAlphaToCoverage ||
- prog_data->uses_omask)
+ if (kill_enable)
dw5 |= GEN6_WM_KILL_ENABLE;
- /* _NEW_BUFFERS | _NEW_COLOR */
- if (brw_color_buffer_write_enabled(brw) ||
- dw5 & (GEN6_WM_KILL_ENABLE | GEN6_WM_COMPUTED_DEPTH)) {
+ if (color_buffer_write_enable ||
+ dw5 & (GEN6_WM_KILL_ENABLE | GEN6_WM_COMPUTED_DEPTH))
dw5 |= GEN6_WM_DISPATCH_ENABLE;
- }
/* From the SNB PRM, volume 2 part 1, page 278:
* "This bit is inserted in the PS payload header and made available to
dw6 |= prog_data->num_varying_inputs <<
GEN6_WM_NUM_SF_OUTPUTS_SHIFT;
if (multisampled_fbo) {
- /* _NEW_MULTISAMPLE */
- if (ctx->Multisample.Enabled)
+ if (msaa_enabled)
dw6 |= GEN6_WM_MSRAST_ON_PATTERN;
else
dw6 |= GEN6_WM_MSRAST_OFF_PIXEL;
OUT_BATCH(ksp0);
OUT_BATCH(dw2);
if (prog_data->base.total_scratch) {
- OUT_RELOC(brw->wm.base.scratch_bo,
+ OUT_RELOC(stage_state->scratch_bo,
I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
ffs(prog_data->base.total_scratch) - 11);
} else {
ADVANCE_BATCH();
}
+static void
+upload_wm_state(struct brw_context *brw)
+{
+ struct gl_context *ctx = &brw->ctx;
+ /* BRW_NEW_FRAGMENT_PROGRAM */
+ const struct brw_fragment_program *fp =
+ brw_fragment_program_const(brw->fragment_program);
+ /* BRW_NEW_FS_PROG_DATA */
+ const struct brw_wm_prog_data *prog_data = brw->wm.prog_data;
+
+ /* _NEW_BUFFERS */
+ const bool multisampled_fbo = _mesa_geometric_samples(ctx->DrawBuffer) > 1;
+
+ /* In case of non 1x per sample shading, only one of SIMD8 and SIMD16
+ * should be enabled. We do 'SIMD16 only' dispatch if a SIMD16 shader
+ * is successfully compiled. In majority of the cases that bring us
+ * better performance than 'SIMD8 only' dispatch.
+ */
+ const int min_inv_per_frag = _mesa_get_min_invocations_per_fragment(
+ ctx, brw->fragment_program, false);
+
+ /* BRW_NEW_FS_PROG_DATA | _NEW_COLOR */
+ const bool dual_src_blend_enable = prog_data->dual_src_blend &&
+ (ctx->Color.BlendEnabled & 1) &&
+ ctx->Color.Blend[0]._UsesDualSrc;
+
+ /* _NEW_COLOR, _NEW_MULTISAMPLE */
+ const bool kill_enable = prog_data->uses_kill || ctx->Color.AlphaEnabled ||
+ ctx->Multisample.SampleAlphaToCoverage ||
+ prog_data->uses_omask;
+
+ /* Rendering against the gl-context is always taken into account. */
+ const bool statistic_enable = true;
+
+ /* _NEW_LINE | _NEW_POLYGON | _NEW_BUFFERS | _NEW_COLOR |
+ * _NEW_MULTISAMPLE
+ */
+ gen6_upload_wm_state(brw, fp, prog_data, &brw->wm.base,
+ multisampled_fbo, min_inv_per_frag,
+ dual_src_blend_enable, kill_enable,
+ brw_color_buffer_write_enabled(brw),
+ ctx->Multisample.Enabled,
+ ctx->Line.StippleFlag, ctx->Polygon.StippleFlag,
+ statistic_enable);
+}
+
const struct brw_tracked_state gen6_wm_state = {
.dirty = {
- .mesa = (_NEW_LINE |
- _NEW_COLOR |
- _NEW_BUFFERS |
- _NEW_PROGRAM_CONSTANTS |
- _NEW_POLYGON |
- _NEW_MULTISAMPLE),
- .brw = (BRW_NEW_FRAGMENT_PROGRAM |
- BRW_NEW_BATCH |
- BRW_NEW_PUSH_CONSTANT_ALLOCATION),
- .cache = (CACHE_NEW_WM_PROG)
+ .mesa = _NEW_BUFFERS |
+ _NEW_COLOR |
+ _NEW_LINE |
+ _NEW_MULTISAMPLE |
+ _NEW_POLYGON |
+ _NEW_PROGRAM_CONSTANTS,
+ .brw = BRW_NEW_BATCH |
+ BRW_NEW_FRAGMENT_PROGRAM |
+ BRW_NEW_FS_PROG_DATA |
+ BRW_NEW_PUSH_CONSTANT_ALLOCATION,
},
.emit = upload_wm_state,
};
projects / mesa.git / blobdiff