#include "intel_batchbuffer.h"
static void
-gen6_prepare_vs_push_constants(struct brw_context *brw)
+gen6_upload_vs_push_constants(struct brw_context *brw)
{
- struct intel_context *intel = &brw->intel;
- struct gl_context *ctx = &intel->ctx;
+ struct brw_stage_state *stage_state = &brw->vs.base;
+
/* _BRW_NEW_VERTEX_PROGRAM */
const struct brw_vertex_program *vp =
brw_vertex_program_const(brw->vertex_program);
- unsigned int nr_params = brw->vs.prog_data->nr_params / 4;
-
- if (brw->vertex_program->IsNVProgram)
- _mesa_load_tracked_matrices(ctx);
+ /* BRW_NEW_VS_PROG_DATA */
+ const struct brw_stage_prog_data *prog_data = &brw->vs.prog_data->base.base;
- /* Updates the ParamaterValues[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, vp->program.Base.Parameters);
+ gen6_upload_push_constants(brw, &vp->program.Base, prog_data,
+ stage_state, AUB_TRACE_VS_CONSTANTS);
- /* CACHE_NEW_VS_PROG | _NEW_TRANSFORM */
- if (brw->vs.prog_data->nr_params == 0 && !ctx->Transform.ClipPlanesEnabled) {
- brw->vs.push_const_size = 0;
- } else {
- int params_uploaded = 0;
- float *param;
- int i;
+ if (brw->gen >= 7) {
+ if (brw->gen == 7 && !brw->is_haswell && !brw->is_baytrail)
+ gen7_emit_vs_workaround_flush(brw);
- param = brw_state_batch(brw, AUB_TRACE_VS_CONSTANTS,
- (MAX_CLIP_PLANES + nr_params) *
- 4 * sizeof(float),
- 32, &brw->vs.push_const_offset);
-
- /* This should be loaded like any other param, but it's ad-hoc
- * until we redo the VS backend.
- */
- for (i = 0; i < MAX_CLIP_PLANES; i++) {
- if (ctx->Transform.ClipPlanesEnabled & (1 << i)) {
- memcpy(param, ctx->Transform._ClipUserPlane[i], 4 * sizeof(float));
- param += 4;
- params_uploaded++;
- }
- }
- /* Align to a reg for convenience for brw_vs_emit.c */
- if (params_uploaded & 1) {
- param += 4;
- params_uploaded++;
- }
-
- if (brw->vs.prog_data->uses_new_param_layout) {
- for (i = 0; i < brw->vs.prog_data->nr_params; i++) {
- *param = convert_param(brw->vs.prog_data->param_convert[i],
- brw->vs.prog_data->param[i]);
- param++;
- }
- params_uploaded += brw->vs.prog_data->nr_params / 4;
- } else {
- for (i = 0; i < vp->program.Base.Parameters->NumParameters; i++) {
- if (brw->vs.constant_map[i] != -1) {
- memcpy(param + brw->vs.constant_map[i] * 4,
- vp->program.Base.Parameters->ParameterValues[i],
- 4 * sizeof(float));
- params_uploaded++;
- }
- }
- }
-
- if (0) {
- printf("VS constant buffer:\n");
- for (i = 0; i < params_uploaded; i++) {
- float *buf = param + i * 4;
- printf("%d: %f %f %f %f\n",
- i, buf[0], buf[1], buf[2], buf[3]);
- }
- }
-
- brw->vs.push_const_size = (params_uploaded + 1) / 2;
- /* We can only push 32 registers of constants at a time. */
- assert(brw->vs.push_const_size <= 32);
+ gen7_upload_constant_state(brw, stage_state, true /* active */,
+ _3DSTATE_CONSTANT_VS);
}
}
-const struct brw_tracked_state gen6_vs_constants = {
+const struct brw_tracked_state gen6_vs_push_constants = {
.dirty = {
- .mesa = _NEW_TRANSFORM | _NEW_PROGRAM_CONSTANTS,
- .brw = (BRW_NEW_BATCH |
- BRW_NEW_VERTEX_PROGRAM),
- .cache = CACHE_NEW_VS_PROG,
+ .mesa = _NEW_PROGRAM_CONSTANTS |
+ _NEW_TRANSFORM,
+ .brw = BRW_NEW_BATCH |
+ BRW_NEW_BLORP |
+ BRW_NEW_PUSH_CONSTANT_ALLOCATION |
+ BRW_NEW_VERTEX_PROGRAM |
+ BRW_NEW_VS_PROG_DATA,
},
- .prepare = gen6_prepare_vs_push_constants,
+ .emit = gen6_upload_vs_push_constants,
};
static void
upload_vs_state(struct brw_context *brw)
{
- struct intel_context *intel = &brw->intel;
+ const struct brw_stage_state *stage_state = &brw->vs.base;
+ uint32_t floating_point_mode = 0;
- if (brw->vs.push_const_size == 0) {
+ /* From the BSpec, 3D Pipeline > Geometry > Vertex Shader > State,
+ * 3DSTATE_VS, Dword 5.0 "VS Function Enable":
+ *
+ * [DevSNB] A pipeline flush must be programmed prior to a 3DSTATE_VS
+ * command that causes the VS Function Enable to toggle. Pipeline
+ * flush can be executed by sending a PIPE_CONTROL command with CS
+ * stall bit set and a post sync operation.
+ *
+ * We've already done such a flush at the start of state upload, so we
+ * don't need to do another one here.
+ */
+
+ if (stage_state->push_const_size == 0) {
/* Disable the push constant buffers. */
BEGIN_BATCH(5);
OUT_BATCH(_3DSTATE_CONSTANT_VS << 16 | (5 - 2));
GEN6_CONSTANT_BUFFER_0_ENABLE |
(5 - 2));
/* Pointer to the VS constant buffer. Covered by the set of
- * state flags from gen6_prepare_wm_constants
+ * state flags from gen6_upload_vs_constants
*/
- OUT_BATCH(brw->vs.push_const_offset +
- brw->vs.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);
ADVANCE_BATCH();
}
+ if (brw->vs.prog_data->base.base.use_alt_mode)
+ floating_point_mode = GEN6_VS_FLOATING_POINT_MODE_ALT;
+
BEGIN_BATCH(6);
OUT_BATCH(_3DSTATE_VS << 16 | (6 - 2));
- OUT_BATCH(brw->vs.prog_offset);
- OUT_BATCH((0 << GEN6_VS_SAMPLER_COUNT_SHIFT) |
- GEN6_VS_FLOATING_POINT_MODE_ALT |
- (brw->vs.nr_surfaces << GEN6_VS_BINDING_TABLE_ENTRY_COUNT_SHIFT));
-
- if (brw->vs.prog_data->total_scratch) {
- OUT_RELOC(brw->vs.scratch_bo,
+ OUT_BATCH(stage_state->prog_offset);
+ OUT_BATCH(floating_point_mode |
+ ((ALIGN(stage_state->sampler_count, 4)/4) << GEN6_VS_SAMPLER_COUNT_SHIFT) |
+ ((brw->vs.prog_data->base.base.binding_table.size_bytes / 4) <<
+ GEN6_VS_BINDING_TABLE_ENTRY_COUNT_SHIFT));
+
+ if (brw->vs.prog_data->base.base.total_scratch) {
+ OUT_RELOC(stage_state->scratch_bo,
I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
- ffs(brw->vs.prog_data->total_scratch) - 11);
+ ffs(brw->vs.prog_data->base.base.total_scratch) - 11);
} else {
OUT_BATCH(0);
}
- OUT_BATCH((1 << GEN6_VS_DISPATCH_START_GRF_SHIFT) |
- (brw->vs.prog_data->urb_read_length << GEN6_VS_URB_READ_LENGTH_SHIFT) |
+ OUT_BATCH((brw->vs.prog_data->base.base.dispatch_grf_start_reg <<
+ GEN6_VS_DISPATCH_START_GRF_SHIFT) |
+ (brw->vs.prog_data->base.urb_read_length << GEN6_VS_URB_READ_LENGTH_SHIFT) |
(0 << GEN6_VS_URB_ENTRY_READ_OFFSET_SHIFT));
- OUT_BATCH(((brw->vs_max_threads - 1) << GEN6_VS_MAX_THREADS_SHIFT) |
+ OUT_BATCH(((brw->max_vs_threads - 1) << GEN6_VS_MAX_THREADS_SHIFT) |
GEN6_VS_STATISTICS_ENABLE |
GEN6_VS_ENABLE);
ADVANCE_BATCH();
* bug reports that led to this workaround, and may be more than
* what is strictly required to avoid the issue.
*/
- BEGIN_BATCH(4);
- OUT_BATCH(_3DSTATE_PIPE_CONTROL);
- OUT_BATCH(PIPE_CONTROL_DEPTH_STALL |
- PIPE_CONTROL_INSTRUCTION_FLUSH |
- PIPE_CONTROL_STATE_CACHE_INVALIDATE);
- OUT_BATCH(0); /* address */
- OUT_BATCH(0); /* write data */
- ADVANCE_BATCH();
+ brw_emit_pipe_control_flush(brw,
+ PIPE_CONTROL_DEPTH_STALL |
+ PIPE_CONTROL_INSTRUCTION_INVALIDATE |
+ PIPE_CONTROL_STATE_CACHE_INVALIDATE);
}
const struct brw_tracked_state gen6_vs_state = {
.dirty = {
- .mesa = _NEW_TRANSFORM | _NEW_PROGRAM_CONSTANTS,
- .brw = (BRW_NEW_NR_VS_SURFACES |
- BRW_NEW_URB_FENCE |
- BRW_NEW_CONTEXT |
- BRW_NEW_VERTEX_PROGRAM |
- BRW_NEW_BATCH),
- .cache = CACHE_NEW_VS_PROG
+ .mesa = _NEW_PROGRAM_CONSTANTS |
+ _NEW_TRANSFORM,
+ .brw = BRW_NEW_BATCH |
+ BRW_NEW_BLORP |
+ BRW_NEW_CONTEXT |
+ BRW_NEW_PUSH_CONSTANT_ALLOCATION |
+ BRW_NEW_VERTEX_PROGRAM |
+ BRW_NEW_VS_PROG_DATA,
},
.emit = upload_vs_state,
};