schedule_instructions(SCHEDULE_POST);
- if (last_scratch > 0)
+ if (last_scratch > 0) {
prog_data->total_scratch = brw_get_scratch_size(last_scratch);
+
+ if (devinfo->is_haswell && stage == MESA_SHADER_COMPUTE) {
+ /* According to the MEDIA_VFE_STATE's "Per Thread Scratch Space"
+ * field documentation, Haswell supports a minimum of 2kB of
+ * scratch space for compute shaders, unlike every other stage
+ * and platform.
+ */
+ prog_data->total_scratch = MAX2(prog_data->total_scratch, 2048);
+ }
+ }
}
bool
OUT_BATCH(MEDIA_VFE_STATE << 16 | (dwords - 2));
if (prog_data->total_scratch) {
- if (brw->gen >= 8)
+ if (brw->gen >= 8) {
+ /* Broadwell's Per Thread Scratch Space is in the range [0, 11]
+ * where 0 = 1k, 1 = 4k, 2 = 8k, ..., 11 = 2M.
+ */
OUT_RELOC64(stage_state->scratch_bo,
I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
ffs(prog_data->total_scratch) - 11);
- else
+ } else if (brw->is_haswell) {
+ /* Haswell's Per Thread Scratch Space is in the range [0, 10]
+ * where 0 = 2k, 1 = 4k, 2 = 8k, ..., 10 = 2M.
+ */
+ OUT_RELOC(stage_state->scratch_bo,
+ I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
+ ffs(prog_data->total_scratch) - 12);
+ } else {
+ /* This is wrong but we'll fix it later */
OUT_RELOC(stage_state->scratch_bo,
I915_GEM_DOMAIN_RENDER, I915_GEM_DOMAIN_RENDER,
ffs(prog_data->total_scratch) - 11);
+ }
} else {
OUT_BATCH(0);
if (brw->gen >= 8)