[Mesa-dev] [PATCH 1/8] radeonsi: remove 8 bytes from si_shader_key
Marek Olšák
maraeo at gmail.com
Mon Jun 5 16:50:54 UTC 2017
From: Marek Olšák <marek.olsak at amd.com>
We can use a union in si_shader_key::mono.
---
src/gallium/drivers/radeonsi/si_shader.c | 14 +++++++-------
src/gallium/drivers/radeonsi/si_shader.h | 9 ++++++---
src/gallium/drivers/radeonsi/si_state_shaders.c | 8 ++++----
3 files changed, 17 insertions(+), 14 deletions(-)
diff --git a/src/gallium/drivers/radeonsi/si_shader.c b/src/gallium/drivers/radeonsi/si_shader.c
index 5c7deeb..2c92269 100644
--- a/src/gallium/drivers/radeonsi/si_shader.c
+++ b/src/gallium/drivers/radeonsi/si_shader.c
@@ -2485,21 +2485,21 @@ static void si_copy_tcs_inputs(struct lp_build_tgsi_context *bld_base)
invocation_id = unpack_param(ctx, ctx->param_tcs_rel_ids, 8, 5);
buffer = desc_from_addr_base64k(ctx, ctx->param_tcs_offchip_addr_base64k);
buffer_offset = LLVMGetParam(ctx->main_fn, ctx->param_tcs_offchip_offset);
lds_vertex_stride = unpack_param(ctx, ctx->param_vs_state_bits, 24, 8);
lds_vertex_offset = LLVMBuildMul(gallivm->builder, invocation_id,
lds_vertex_stride, "");
lds_base = get_tcs_in_current_patch_offset(ctx);
lds_base = LLVMBuildAdd(gallivm->builder, lds_base, lds_vertex_offset, "");
- inputs = ctx->shader->key.mono.ff_tcs_inputs_to_copy;
+ inputs = ctx->shader->key.mono.u.ff_tcs_inputs_to_copy;
while (inputs) {
unsigned i = u_bit_scan64(&inputs);
LLVMValueRef lds_ptr = LLVMBuildAdd(gallivm->builder, lds_base,
LLVMConstInt(ctx->i32, 4 * i, 0),
"");
LLVMValueRef buffer_addr = get_tcs_tes_buffer_address(ctx,
get_rel_patch_id(ctx),
invocation_id,
@@ -3012,21 +3012,21 @@ static void si_llvm_emit_vs_epilogue(struct lp_build_tgsi_context *bld_base)
"");
outputs[i].vertex_stream[j] =
(info->output_streams[i] >> (2 * j)) & 3;
}
}
if (ctx->shader->selector->so.num_outputs)
si_llvm_emit_streamout(ctx, outputs, i, 0);
/* Export PrimitiveID. */
- if (ctx->shader->key.mono.vs_export_prim_id) {
+ if (ctx->shader->key.mono.u.vs_export_prim_id) {
outputs[i].semantic_name = TGSI_SEMANTIC_PRIMID;
outputs[i].semantic_index = 0;
outputs[i].values[0] = bitcast(bld_base, TGSI_TYPE_FLOAT,
get_primitive_id(bld_base, 0));
for (j = 1; j < 4; j++)
outputs[i].values[j] = LLVMConstReal(ctx->f32, 0);
memset(outputs[i].vertex_stream, 0,
sizeof(outputs[i].vertex_stream));
i++;
@@ -5266,37 +5266,37 @@ static void si_dump_shader_key(unsigned processor, const struct si_shader *shade
const struct si_shader_key *key = &shader->key;
fprintf(f, "SHADER KEY\n");
switch (processor) {
case PIPE_SHADER_VERTEX:
si_dump_shader_key_vs(key, &key->part.vs.prolog,
"part.vs.prolog", f);
fprintf(f, " as_es = %u\n", key->as_es);
fprintf(f, " as_ls = %u\n", key->as_ls);
- fprintf(f, " mono.vs_export_prim_id = %u\n",
- key->mono.vs_export_prim_id);
+ fprintf(f, " mono.u.vs_export_prim_id = %u\n",
+ key->mono.u.vs_export_prim_id);
break;
case PIPE_SHADER_TESS_CTRL:
if (shader->selector->screen->b.chip_class >= GFX9) {
si_dump_shader_key_vs(key, &key->part.tcs.ls_prolog,
"part.tcs.ls_prolog", f);
}
fprintf(f, " part.tcs.epilog.prim_mode = %u\n", key->part.tcs.epilog.prim_mode);
- fprintf(f, " mono.ff_tcs_inputs_to_copy = 0x%"PRIx64"\n", key->mono.ff_tcs_inputs_to_copy);
+ fprintf(f, " mono.u.ff_tcs_inputs_to_copy = 0x%"PRIx64"\n", key->mono.u.ff_tcs_inputs_to_copy);
break;
case PIPE_SHADER_TESS_EVAL:
fprintf(f, " as_es = %u\n", key->as_es);
- fprintf(f, " mono.vs_export_prim_id = %u\n",
- key->mono.vs_export_prim_id);
+ fprintf(f, " mono.u.vs_export_prim_id = %u\n",
+ key->mono.u.vs_export_prim_id);
break;
case PIPE_SHADER_GEOMETRY:
if (shader->is_gs_copy_shader)
break;
if (shader->selector->screen->b.chip_class >= GFX9 &&
key->part.gs.es->type == PIPE_SHADER_VERTEX) {
si_dump_shader_key_vs(key, &key->part.gs.vs_prolog,
"part.gs.vs_prolog", f);
diff --git a/src/gallium/drivers/radeonsi/si_shader.h b/src/gallium/drivers/radeonsi/si_shader.h
index 7112621..f739769 100644
--- a/src/gallium/drivers/radeonsi/si_shader.h
+++ b/src/gallium/drivers/radeonsi/si_shader.h
@@ -480,23 +480,26 @@ struct si_shader_key {
/* These two are initially set according to the NEXT_SHADER property,
* or guessed if the property doesn't seem correct.
*/
unsigned as_es:1; /* export shader, which precedes GS */
unsigned as_ls:1; /* local shader, which precedes TCS */
/* Flags for monolithic compilation only. */
struct {
/* One byte for every input: SI_FIX_FETCH_* enums. */
uint8_t vs_fix_fetch[SI_MAX_ATTRIBS];
- uint64_t ff_tcs_inputs_to_copy; /* for fixed-func TCS */
- /* When PS needs PrimID and GS is disabled. */
- unsigned vs_export_prim_id:1;
+
+ union {
+ uint64_t ff_tcs_inputs_to_copy; /* for fixed-func TCS */
+ /* When PS needs PrimID and GS is disabled. */
+ unsigned vs_export_prim_id:1;
+ } u;
} mono;
/* Optimization flags for asynchronous compilation only. */
struct {
struct {
uint64_t kill_outputs; /* "get_unique_index" bits */
unsigned clip_disable:1;
} hw_vs; /* HW VS (it can be VS, TES, GS) */
/* For shaders where monolithic variants have better code.
diff --git a/src/gallium/drivers/radeonsi/si_state_shaders.c b/src/gallium/drivers/radeonsi/si_state_shaders.c
index 41f8bdf..7ed6722 100644
--- a/src/gallium/drivers/radeonsi/si_state_shaders.c
+++ b/src/gallium/drivers/radeonsi/si_state_shaders.c
@@ -836,21 +836,21 @@ static void si_shader_gs(struct si_screen *sscreen, struct si_shader *shader)
static void si_shader_vs(struct si_screen *sscreen, struct si_shader *shader,
struct si_shader_selector *gs)
{
struct si_pm4_state *pm4;
unsigned num_user_sgprs;
unsigned nparams, vgpr_comp_cnt;
uint64_t va;
unsigned oc_lds_en;
unsigned window_space =
shader->selector->info.properties[TGSI_PROPERTY_VS_WINDOW_SPACE_POSITION];
- bool enable_prim_id = shader->key.mono.vs_export_prim_id || shader->selector->info.uses_primid;
+ bool enable_prim_id = shader->key.mono.u.vs_export_prim_id || shader->selector->info.uses_primid;
pm4 = si_get_shader_pm4_state(shader);
if (!pm4)
return;
/* We always write VGT_GS_MODE in the VS state, because every switch
* between different shader pipelines involving a different GS or no
* GS at all involves a switch of the VS (different GS use different
* copy shaders). On the other hand, when the API switches from a GS to
* no GS and then back to the same GS used originally, the GS state is
@@ -1260,46 +1260,46 @@ static inline void si_shader_selector_key(struct pipe_context *ctx,
si_shader_selector_key_vs(sctx, sel, key, &key->part.vs.prolog);
if (sctx->tes_shader.cso)
key->as_ls = 1;
else if (sctx->gs_shader.cso)
key->as_es = 1;
else {
si_shader_selector_key_hw_vs(sctx, sel, key);
if (sctx->ps_shader.cso && sctx->ps_shader.cso->info.uses_primid)
- key->mono.vs_export_prim_id = 1;
+ key->mono.u.vs_export_prim_id = 1;
}
break;
case PIPE_SHADER_TESS_CTRL:
if (sctx->b.chip_class >= GFX9) {
si_shader_selector_key_vs(sctx, sctx->vs_shader.cso,
key, &key->part.tcs.ls_prolog);
key->part.tcs.ls = sctx->vs_shader.cso;
}
key->part.tcs.epilog.prim_mode =
sctx->tes_shader.cso->info.properties[TGSI_PROPERTY_TES_PRIM_MODE];
key->part.tcs.epilog.tes_reads_tess_factors =
sctx->tes_shader.cso->info.reads_tess_factors;
if (sel == sctx->fixed_func_tcs_shader.cso)
- key->mono.ff_tcs_inputs_to_copy = sctx->vs_shader.cso->outputs_written;
+ key->mono.u.ff_tcs_inputs_to_copy = sctx->vs_shader.cso->outputs_written;
break;
case PIPE_SHADER_TESS_EVAL:
if (sctx->gs_shader.cso)
key->as_es = 1;
else {
si_shader_selector_key_hw_vs(sctx, sel, key);
if (sctx->ps_shader.cso && sctx->ps_shader.cso->info.uses_primid)
- key->mono.vs_export_prim_id = 1;
+ key->mono.u.vs_export_prim_id = 1;
}
break;
case PIPE_SHADER_GEOMETRY:
if (sctx->b.chip_class >= GFX9) {
if (sctx->tes_shader.cso) {
key->part.gs.es = sctx->tes_shader.cso;
} else {
si_shader_selector_key_vs(sctx, sctx->vs_shader.cso,
key, &key->part.gs.vs_prolog);
key->part.gs.es = sctx->vs_shader.cso;
--
2.7.4
More information about the mesa-dev
mailing list