#include "intel_batchbuffer.h"
static void
-prepare_wm_constants(struct brw_context *brw)
+gen6_upload_wm_push_constants(struct brw_context *brw)
{
struct intel_context *intel = &brw->intel;
struct gl_context *ctx = &intel->ctx;
+ /* BRW_NEW_FRAGMENT_PROGRAM */
const struct brw_fragment_program *fp =
brw_fragment_program_const(brw->fragment_program);
- drm_intel_bo_unreference(brw->wm.push_const_bo);
- brw->wm.push_const_bo = NULL;
-
- /* Updates the ParamaterValues[i] pointers for all parameters of the
+ /* Updates the ParameterValues[i] pointers for all parameters of the
* basic type of PROGRAM_STATE_VAR.
*/
/* XXX: Should this happen somewhere before to get our state flag set? */
_mesa_load_state_parameters(ctx, fp->program.Base.Parameters);
+ /* CACHE_NEW_WM_PROG */
if (brw->wm.prog_data->nr_params != 0) {
float *constants;
unsigned int i;
- brw->wm.push_const_bo = drm_intel_bo_alloc(intel->bufmgr,
- "WM constant_bo",
- brw->wm.prog_data->nr_params *
- sizeof(float),
- 4096);
- drm_intel_gem_bo_map_gtt(brw->wm.push_const_bo);
- constants = brw->wm.push_const_bo->virtual;
+ constants = brw_state_batch(brw, AUB_TRACE_WM_CONSTANTS,
+ brw->wm.prog_data->nr_params *
+ sizeof(float),
+ 32, &brw->wm.push_const_offset);
+
for (i = 0; i < brw->wm.prog_data->nr_params; i++) {
- constants[i] = convert_param(brw->wm.prog_data->param_convert[i],
- *brw->wm.prog_data->param[i]);
+ constants[i] = *brw->wm.prog_data->param[i];
}
if (0) {
printf("\n");
printf("\n");
}
-
- drm_intel_gem_bo_unmap_gtt(brw->wm.push_const_bo);
}
}
-const struct brw_tracked_state gen6_wm_constants = {
+const struct brw_tracked_state gen6_wm_push_constants = {
.dirty = {
.mesa = _NEW_PROGRAM_CONSTANTS,
- .brw = BRW_NEW_FRAGMENT_PROGRAM,
- .cache = 0,
+ .brw = (BRW_NEW_BATCH |
+ BRW_NEW_FRAGMENT_PROGRAM),
+ .cache = CACHE_NEW_WM_PROG,
},
- .prepare = prepare_wm_constants,
+ .emit = gen6_upload_wm_push_constants,
};
static void
brw_fragment_program_const(brw->fragment_program);
uint32_t dw2, dw4, dw5, dw6;
- /* CACHE_NEW_WM_PROG */
+ /* _NEW_BUFFERS */
+ bool multisampled_fbo = ctx->DrawBuffer->Visual.samples > 1;
+
+ /* CACHE_NEW_WM_PROG */
if (brw->wm.prog_data->nr_params == 0) {
/* Disable the push constant buffers. */
BEGIN_BATCH(5);
- OUT_BATCH(CMD_3D_CONSTANT_PS_STATE << 16 | (5 - 2));
+ OUT_BATCH(_3DSTATE_CONSTANT_PS << 16 | (5 - 2));
OUT_BATCH(0);
OUT_BATCH(0);
OUT_BATCH(0);
ADVANCE_BATCH();
} else {
BEGIN_BATCH(5);
- OUT_BATCH(CMD_3D_CONSTANT_PS_STATE << 16 |
+ OUT_BATCH(_3DSTATE_CONSTANT_PS << 16 |
GEN6_CONSTANT_BUFFER_0_ENABLE |
(5 - 2));
- OUT_RELOC(brw->wm.push_const_bo,
- I915_GEM_DOMAIN_RENDER, 0, /* XXX: bad domain */
+ /* Pointer to the WM constant buffer. Covered by the set of
+ * state flags from gen6_upload_wm_push_constants.
+ */
+ OUT_BATCH(brw->wm.push_const_offset +
ALIGN(brw->wm.prog_data->nr_params,
brw->wm.prog_data->dispatch_width) / 8 - 1);
OUT_BATCH(0);
dw5 |= GEN6_WM_LINE_AA_WIDTH_1_0;
dw5 |= GEN6_WM_LINE_END_CAP_AA_WIDTH_0_5;
- /* OpenGL non-ieee floating point mode */
- dw2 |= GEN6_WM_FLOATING_POINT_MODE_ALT;
-
- /* BRW_NEW_NR_WM_SURFACES */
- dw2 |= brw->wm.nr_surfaces << GEN6_WM_BINDING_TABLE_ENTRY_COUNT_SHIFT;
+ /* Use ALT floating point mode for ARB fragment programs, because they
+ * require 0^0 == 1. Even though _CurrentFragmentProgram is used for
+ * rendering, CurrentFragmentProgram is used for this check to
+ * differentiate between the GLSL and non-GLSL cases.
+ */
+ if (ctx->Shader.CurrentFragmentProgram == NULL)
+ dw2 |= GEN6_WM_FLOATING_POINT_MODE_ALT;
/* CACHE_NEW_SAMPLER */
- dw2 |= (ALIGN(brw->wm.sampler_count, 4) / 4) << GEN6_WM_SAMPLER_COUNT_SHIFT;
+ dw2 |= (ALIGN(brw->sampler.count, 4) / 4) << GEN6_WM_SAMPLER_COUNT_SHIFT;
dw4 |= (brw->wm.prog_data->first_curbe_grf <<
GEN6_WM_DISPATCH_START_GRF_SHIFT_0);
+ dw4 |= (brw->wm.prog_data->first_curbe_grf_16 <<
+ GEN6_WM_DISPATCH_START_GRF_SHIFT_2);
- dw5 |= (40 - 1) << GEN6_WM_MAX_THREADS_SHIFT;
+ dw5 |= (brw->max_wm_threads - 1) << GEN6_WM_MAX_THREADS_SHIFT;
/* CACHE_NEW_WM_PROG */
- if (brw->wm.prog_data->dispatch_width == 8)
- dw5 |= GEN6_WM_8_DISPATCH_ENABLE;
- else
+ dw5 |= GEN6_WM_8_DISPATCH_ENABLE;
+ if (brw->wm.prog_data->prog_offset_16)
dw5 |= GEN6_WM_16_DISPATCH_ENABLE;
+ /* CACHE_NEW_WM_PROG | _NEW_COLOR */
+ if (brw->wm.prog_data->dual_src_blend &&
+ (ctx->Color.BlendEnabled & 1) &&
+ ctx->Color.Blend[0]._UsesDualSrc) {
+ dw5 |= GEN6_WM_DUAL_SOURCE_BLEND_ENABLE;
+ }
+
/* _NEW_LINE */
if (ctx->Line.StippleFlag)
dw5 |= GEN6_WM_LINE_STIPPLE_ENABLE;
- /* _NEW_POLYGONSTIPPLE */
+ /* _NEW_POLYGON */
if (ctx->Polygon.StippleFlag)
dw5 |= GEN6_WM_POLYGON_STIPPLE_ENABLE;
/* BRW_NEW_FRAGMENT_PROGRAM */
- if (fp->program.Base.InputsRead & (1 << FRAG_ATTRIB_WPOS))
+ if (fp->program.Base.InputsRead & VARYING_BIT_POS)
dw5 |= GEN6_WM_USES_SOURCE_DEPTH | GEN6_WM_USES_SOURCE_W;
if (fp->program.Base.OutputsWritten & BITFIELD64_BIT(FRAG_RESULT_DEPTH))
dw5 |= GEN6_WM_COMPUTED_DEPTH;
+ /* CACHE_NEW_WM_PROG */
+ dw6 |= brw->wm.prog_data->barycentric_interp_modes <<
+ GEN6_WM_BARYCENTRIC_INTERPOLATION_MODE_SHIFT;
- /* _NEW_COLOR */
- if (fp->program.UsesKill || ctx->Color.AlphaEnabled)
+ /* _NEW_COLOR, _NEW_MULTISAMPLE */
+ if (fp->program.UsesKill || ctx->Color.AlphaEnabled ||
+ ctx->Multisample.SampleAlphaToCoverage)
dw5 |= GEN6_WM_KILL_ENABLE;
if (brw_color_buffer_write_enabled(brw) ||
dw5 |= GEN6_WM_DISPATCH_ENABLE;
}
- dw6 |= GEN6_WM_PERSPECTIVE_PIXEL_BARYCENTRIC;
-
- dw6 |= brw_count_bits(brw->fragment_program->Base.InputsRead) <<
+ dw6 |= _mesa_bitcount_64(brw->fragment_program->Base.InputsRead) <<
GEN6_WM_NUM_SF_OUTPUTS_SHIFT;
+ if (multisampled_fbo) {
+ /* _NEW_MULTISAMPLE */
+ if (ctx->Multisample.Enabled)
+ dw6 |= GEN6_WM_MSRAST_ON_PATTERN;
+ else
+ dw6 |= GEN6_WM_MSRAST_OFF_PIXEL;
+ dw6 |= GEN6_WM_MSDISPMODE_PERPIXEL;
+ } else {
+ dw6 |= GEN6_WM_MSRAST_OFF_PIXEL;
+ dw6 |= GEN6_WM_MSDISPMODE_PERSAMPLE;
+ }
BEGIN_BATCH(9);
- OUT_BATCH(CMD_3D_WM_STATE << 16 | (9 - 2));
- OUT_RELOC(brw->wm.prog_bo, I915_GEM_DOMAIN_INSTRUCTION, 0, 0);
+ OUT_BATCH(_3DSTATE_WM << 16 | (9 - 2));
+ OUT_BATCH(brw->wm.prog_offset);
OUT_BATCH(dw2);
- OUT_BATCH(0); /* scratch space base offset */
+ if (brw->wm.prog_data->total_scratch) {
+ OUT_RELOC(brw->wm.scratch_bo, I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
+ ffs(brw->wm.prog_data->total_scratch) - 11);
+ } else {
+ OUT_BATCH(0);
+ }
OUT_BATCH(dw4);
OUT_BATCH(dw5);
OUT_BATCH(dw6);
OUT_BATCH(0); /* kernel 1 pointer */
- OUT_BATCH(0); /* kernel 2 pointer */
+ /* kernel 2 pointer */
+ OUT_BATCH(brw->wm.prog_offset + brw->wm.prog_data->prog_offset_16);
ADVANCE_BATCH();
}
const struct brw_tracked_state gen6_wm_state = {
.dirty = {
- .mesa = (_NEW_LINE | _NEW_POLYGONSTIPPLE | _NEW_COLOR | _NEW_BUFFERS |
- _NEW_PROGRAM_CONSTANTS | _NEW_POLYGON),
- .brw = (BRW_NEW_CURBE_OFFSETS |
- BRW_NEW_FRAGMENT_PROGRAM |
- BRW_NEW_NR_WM_SURFACES |
- BRW_NEW_URB_FENCE |
+ .mesa = (_NEW_LINE |
+ _NEW_COLOR |
+ _NEW_BUFFERS |
+ _NEW_PROGRAM_CONSTANTS |
+ _NEW_POLYGON |
+ _NEW_MULTISAMPLE),
+ .brw = (BRW_NEW_FRAGMENT_PROGRAM |
BRW_NEW_BATCH),
.cache = (CACHE_NEW_SAMPLER |
CACHE_NEW_WM_PROG)
projects / mesa.git / blobdiff