#include "brw_draw.h"
#include "brw_state.h"
#include "intel_batchbuffer.h"
+#include "intel_buffer_objects.h"
#include "brw_defines.h"
static void
-brw_emit_gpgpu_walker(struct brw_context *brw, const GLuint *num_groups)
+brw_emit_gpgpu_walker(struct brw_context *brw,
+ const void *compute_param,
+ bool indirect)
{
const struct brw_cs_prog_data *prog_data = brw->cs.prog_data;
+ const GLuint *num_groups;
+ uint32_t indirect_flag;
+
+ if (!indirect) {
+ num_groups = (const GLuint *)compute_param;
+ indirect_flag = 0;
+ } else {
+ GLintptr indirect_offset = (GLintptr)compute_param;
+ static const GLuint indirect_group_counts[3] = { 0, 0, 0 };
+ num_groups = indirect_group_counts;
+
+ struct gl_buffer_object *indirect_buffer = brw->ctx.DispatchIndirectBuffer;
+ drm_intel_bo *bo = intel_bufferobj_buffer(brw,
+ intel_buffer_object(indirect_buffer),
+ indirect_offset, 3 * sizeof(GLuint));
+
+ indirect_flag = GEN7_GPGPU_INDIRECT_PARAMETER_ENABLE;
+
+ brw_load_register_mem(brw, GEN7_GPGPU_DISPATCHDIMX, bo,
+ I915_GEM_DOMAIN_VERTEX, 0,
+ indirect_offset + 0);
+ brw_load_register_mem(brw, GEN7_GPGPU_DISPATCHDIMY, bo,
+ I915_GEM_DOMAIN_VERTEX, 0,
+ indirect_offset + 4);
+ brw_load_register_mem(brw, GEN7_GPGPU_DISPATCHDIMZ, bo,
+ I915_GEM_DOMAIN_VERTEX, 0,
+ indirect_offset + 8);
+ }
+
const unsigned simd_size = prog_data->simd_size;
unsigned group_size = prog_data->local_size[0] *
prog_data->local_size[1] * prog_data->local_size[2];
uint32_t dwords = brw->gen < 8 ? 11 : 15;
BEGIN_BATCH(dwords);
- OUT_BATCH(GPGPU_WALKER << 16 | (dwords - 2));
+ OUT_BATCH(GPGPU_WALKER << 16 | (dwords - 2) | indirect_flag);
OUT_BATCH(0);
if (brw->gen >= 8) {
OUT_BATCH(0); /* Indirect Data Length */
static void
-brw_dispatch_compute(struct gl_context *ctx, const GLuint *num_groups)
+brw_dispatch_compute_common(struct gl_context *ctx,
+ const void *compute_param,
+ bool indirect)
{
struct brw_context *brw = brw_context(ctx);
int estimated_buffer_space_needed;
brw->no_batch_wrap = true;
brw_upload_compute_state(brw);
- brw_emit_gpgpu_walker(brw, num_groups);
+ brw_emit_gpgpu_walker(brw, compute_param, indirect);
brw->no_batch_wrap = false;
*/
}
+static void
+brw_dispatch_compute(struct gl_context *ctx, const GLuint *num_groups) {
+ brw_dispatch_compute_common(ctx,
+ num_groups,
+ false);
+}
+
+static void
+brw_dispatch_compute_indirect(struct gl_context *ctx, GLintptr indirect)
+{
+ brw_dispatch_compute_common(ctx,
+ (void *)indirect,
+ true);
+}
void
brw_init_compute_functions(struct dd_function_table *functions)
{
functions->DispatchCompute = brw_dispatch_compute;
+ functions->DispatchComputeIndirect = brw_dispatch_compute_indirect;
}