Mesa (master): radeonsi: split the shader key into 3 logical parts
Marek Olšák
mareko at kemper.freedesktop.org
Mon Nov 21 21:41:10 UTC 2016
Module: Mesa
Branch: master
Commit: 6d5c2a8b5c33122682d801f9b14a70e3d3d9a5cc
URL: http://cgit.freedesktop.org/mesa/mesa/commit/?id=6d5c2a8b5c33122682d801f9b14a70e3d3d9a5cc
Author: Marek Olšák <marek.olsak at amd.com>
Date: Sun Nov 13 03:17:46 2016 +0100
radeonsi: split the shader key into 3 logical parts
key->part.*: prolog and epilog flags only
key->as_{ls,es}: special flags
key->mono.*: flags for monolithic compilation only
Tested-by: Edmondo Tommasina <edmondo.tommasina at gmail.com>
Reviewed-by: Nicolai Hähnle <nicolai.haehnle at amd.com>
---
src/gallium/drivers/radeonsi/si_pipe.h | 4 +-
src/gallium/drivers/radeonsi/si_shader.c | 194 ++++++++++++------------
src/gallium/drivers/radeonsi/si_shader.h | 65 ++++----
src/gallium/drivers/radeonsi/si_state.c | 2 +-
src/gallium/drivers/radeonsi/si_state_shaders.c | 132 ++++++++--------
5 files changed, 203 insertions(+), 194 deletions(-)
diff --git a/src/gallium/drivers/radeonsi/si_pipe.h b/src/gallium/drivers/radeonsi/si_pipe.h
index df2f130..847281e 100644
--- a/src/gallium/drivers/radeonsi/si_pipe.h
+++ b/src/gallium/drivers/radeonsi/si_pipe.h
@@ -463,9 +463,9 @@ static inline struct si_shader* si_get_vs_state(struct si_context *sctx)
static inline bool si_vs_exports_prim_id(struct si_shader *shader)
{
if (shader->selector->type == PIPE_SHADER_VERTEX)
- return shader->key.vs.epilog.export_prim_id;
+ return shader->key.part.vs.epilog.export_prim_id;
else if (shader->selector->type == PIPE_SHADER_TESS_EVAL)
- return shader->key.tes.epilog.export_prim_id;
+ return shader->key.part.tes.epilog.export_prim_id;
else
return false;
}
diff --git a/src/gallium/drivers/radeonsi/si_shader.c b/src/gallium/drivers/radeonsi/si_shader.c
index 608cb72..a3388bd 100644
--- a/src/gallium/drivers/radeonsi/si_shader.c
+++ b/src/gallium/drivers/radeonsi/si_shader.c
@@ -68,7 +68,7 @@ static void si_llvm_emit_barrier(const struct lp_build_tgsi_action *action,
struct lp_build_tgsi_context *bld_base,
struct lp_build_emit_data *emit_data);
-static void si_dump_shader_key(unsigned shader, union si_shader_key *key,
+static void si_dump_shader_key(unsigned shader, struct si_shader_key *key,
FILE *f);
static void si_build_vs_prolog_function(struct si_shader_context *ctx,
@@ -416,7 +416,7 @@ static void declare_input_vs(
input, llvm_chan, "");
}
- fix_fetch = (ctx->shader->key.vs.fix_fetch >> (2 * input_index)) & 3;
+ fix_fetch = (ctx->shader->key.mono.vs.fix_fetch >> (2 * input_index)) & 3;
if (fix_fetch) {
/* The hardware returns an unsigned value; convert it to a
* signed one.
@@ -1252,7 +1252,7 @@ static void interp_fs_input(struct si_shader_context *ctx,
intr_name = interp_param ? "llvm.SI.fs.interp" : "llvm.SI.fs.constant";
if (semantic_name == TGSI_SEMANTIC_COLOR &&
- ctx->shader->key.ps.prolog.color_two_side) {
+ ctx->shader->key.part.ps.prolog.color_two_side) {
LLVMValueRef args[4];
LLVMValueRef is_face_positive;
LLVMValueRef back_attr_number;
@@ -1360,7 +1360,7 @@ static void declare_input_fs(
if (decl->Semantic.Name == TGSI_SEMANTIC_COLOR &&
decl->Interp.Interpolate == TGSI_INTERPOLATE_COLOR &&
- ctx->shader->key.ps.prolog.flatshade_colors)
+ ctx->shader->key.part.ps.prolog.flatshade_colors)
interp_param = NULL; /* load the constant color */
interp_fs_input(ctx, input_index, decl->Semantic.Name,
@@ -1832,13 +1832,13 @@ static void si_llvm_init_export_args(struct lp_build_tgsi_context *bld_base,
args[3] = lp_build_const_int32(base->gallivm, target);
if (ctx->type == PIPE_SHADER_FRAGMENT) {
- const union si_shader_key *key = &ctx->shader->key;
- unsigned col_formats = key->ps.epilog.spi_shader_col_format;
+ const struct si_shader_key *key = &ctx->shader->key;
+ unsigned col_formats = key->part.ps.epilog.spi_shader_col_format;
int cbuf = target - V_008DFC_SQ_EXP_MRT;
assert(cbuf >= 0 && cbuf < 8);
spi_shader_col_format = (col_formats >> (cbuf * 4)) & 0xf;
- is_int8 = (key->ps.epilog.color_is_int8 >> cbuf) & 0x1;
+ is_int8 = (key->part.ps.epilog.color_is_int8 >> cbuf) & 0x1;
}
args[4] = uint->zero; /* COMPR flag */
@@ -1991,13 +1991,13 @@ static void si_alpha_test(struct lp_build_tgsi_context *bld_base,
struct si_shader_context *ctx = si_shader_context(bld_base);
struct gallivm_state *gallivm = bld_base->base.gallivm;
- if (ctx->shader->key.ps.epilog.alpha_func != PIPE_FUNC_NEVER) {
+ if (ctx->shader->key.part.ps.epilog.alpha_func != PIPE_FUNC_NEVER) {
LLVMValueRef alpha_ref = LLVMGetParam(ctx->main_fn,
SI_PARAM_ALPHA_REF);
LLVMValueRef alpha_pass =
lp_build_cmp(&bld_base->base,
- ctx->shader->key.ps.epilog.alpha_func,
+ ctx->shader->key.part.ps.epilog.alpha_func,
alpha, alpha_ref);
LLVMValueRef arg =
lp_build_select(&bld_base->base,
@@ -2440,7 +2440,7 @@ static void si_copy_tcs_inputs(struct lp_build_tgsi_context *bld_base)
lds_base = get_tcs_in_current_patch_offset(ctx);
lds_base = LLVMBuildAdd(gallivm->builder, lds_base, lds_vertex_offset, "");
- inputs = ctx->shader->key.tcs.epilog.inputs_to_copy;
+ inputs = ctx->shader->key.mono.tcs.inputs_to_copy;
while (inputs) {
unsigned i = u_bit_scan64(&inputs);
@@ -2487,7 +2487,7 @@ static void si_write_tess_factors(struct lp_build_tgsi_context *bld_base,
invocation_id, bld_base->uint_bld.zero, ""));
/* Determine the layout of one tess factor element in the buffer. */
- switch (shader->key.tcs.epilog.prim_mode) {
+ switch (shader->key.part.tcs.epilog.prim_mode) {
case PIPE_PRIM_LINES:
stride = 2; /* 2 dwords, 1 vec2 store */
outer_comps = 2;
@@ -2874,31 +2874,31 @@ static void si_export_mrt_color(struct lp_build_tgsi_context *bld_base,
int i;
/* Clamp color */
- if (ctx->shader->key.ps.epilog.clamp_color)
+ if (ctx->shader->key.part.ps.epilog.clamp_color)
for (i = 0; i < 4; i++)
color[i] = si_llvm_saturate(bld_base, color[i]);
/* Alpha to one */
- if (ctx->shader->key.ps.epilog.alpha_to_one)
+ if (ctx->shader->key.part.ps.epilog.alpha_to_one)
color[3] = base->one;
/* Alpha test */
if (index == 0 &&
- ctx->shader->key.ps.epilog.alpha_func != PIPE_FUNC_ALWAYS)
+ ctx->shader->key.part.ps.epilog.alpha_func != PIPE_FUNC_ALWAYS)
si_alpha_test(bld_base, color[3]);
/* Line & polygon smoothing */
- if (ctx->shader->key.ps.epilog.poly_line_smoothing)
+ if (ctx->shader->key.part.ps.epilog.poly_line_smoothing)
color[3] = si_scale_alpha_by_sample_mask(bld_base, color[3],
samplemask_param);
/* If last_cbuf > 0, FS_COLOR0_WRITES_ALL_CBUFS is true. */
- if (ctx->shader->key.ps.epilog.last_cbuf > 0) {
+ if (ctx->shader->key.part.ps.epilog.last_cbuf > 0) {
LLVMValueRef args[8][9];
int c, last = -1;
/* Get the export arguments, also find out what the last one is. */
- for (c = 0; c <= ctx->shader->key.ps.epilog.last_cbuf; c++) {
+ for (c = 0; c <= ctx->shader->key.part.ps.epilog.last_cbuf; c++) {
si_llvm_init_export_args(bld_base, color,
V_008DFC_SQ_EXP_MRT + c, args[c]);
if (args[c][0] != bld_base->uint_bld.zero)
@@ -2906,7 +2906,7 @@ static void si_export_mrt_color(struct lp_build_tgsi_context *bld_base,
}
/* Emit all exports. */
- for (c = 0; c <= ctx->shader->key.ps.epilog.last_cbuf; c++) {
+ for (c = 0; c <= ctx->shader->key.part.ps.epilog.last_cbuf; c++) {
if (is_last && last == c) {
args[c][1] = bld_base->uint_bld.one; /* whether the EXEC mask is valid */
args[c][2] = bld_base->uint_bld.one; /* DONE bit */
@@ -5347,9 +5347,9 @@ static void create_function(struct si_shader_context *ctx)
params[SI_PARAM_DRAWID] = ctx->i32;
num_params = SI_PARAM_DRAWID+1;
- if (shader->key.vs.as_es) {
+ if (shader->key.as_es) {
params[ctx->param_es2gs_offset = num_params++] = ctx->i32;
- } else if (shader->key.vs.as_ls) {
+ } else if (shader->key.as_ls) {
params[SI_PARAM_LS_OUT_LAYOUT] = ctx->i32;
num_params = SI_PARAM_LS_OUT_LAYOUT+1;
} else {
@@ -5383,7 +5383,7 @@ static void create_function(struct si_shader_context *ctx)
num_prolog_vgprs += shader->selector->info.num_inputs;
/* PrimitiveID output. */
- if (!shader->key.vs.as_es && !shader->key.vs.as_ls)
+ if (!shader->key.as_es && !shader->key.as_ls)
for (i = 0; i <= VS_EPILOG_PRIMID_LOC; i++)
returns[num_returns++] = ctx->f32;
}
@@ -5417,7 +5417,7 @@ static void create_function(struct si_shader_context *ctx)
params[SI_PARAM_TCS_OFFCHIP_LAYOUT] = ctx->i32;
num_params = SI_PARAM_TCS_OFFCHIP_LAYOUT+1;
- if (shader->key.tes.as_es) {
+ if (shader->key.as_es) {
params[ctx->param_oc_lds = num_params++] = ctx->i32;
params[ctx->param_tess_offchip = num_params++] = ctx->i32;
params[ctx->param_es2gs_offset = num_params++] = ctx->i32;
@@ -5436,7 +5436,7 @@ static void create_function(struct si_shader_context *ctx)
params[ctx->param_tes_patch_id = num_params++] = ctx->i32;
/* PrimitiveID output. */
- if (!shader->key.tes.as_es)
+ if (!shader->key.as_es)
for (i = 0; i <= VS_EPILOG_PRIMID_LOC; i++)
returns[num_returns++] = ctx->f32;
break;
@@ -5590,7 +5590,7 @@ static void create_function(struct si_shader_context *ctx)
"ddxy_lds",
LOCAL_ADDR_SPACE);
- if ((ctx->type == PIPE_SHADER_VERTEX && shader->key.vs.as_ls) ||
+ if ((ctx->type == PIPE_SHADER_VERTEX && shader->key.as_ls) ||
ctx->type == PIPE_SHADER_TESS_CTRL ||
ctx->type == PIPE_SHADER_TESS_EVAL)
declare_tess_lds(ctx);
@@ -5609,9 +5609,9 @@ static void preload_ring_buffers(struct si_shader_context *ctx)
SI_PARAM_RW_BUFFERS);
if ((ctx->type == PIPE_SHADER_VERTEX &&
- ctx->shader->key.vs.as_es) ||
+ ctx->shader->key.as_es) ||
(ctx->type == PIPE_SHADER_TESS_EVAL &&
- ctx->shader->key.tes.as_es) ||
+ ctx->shader->key.as_es) ||
ctx->type == PIPE_SHADER_GEOMETRY) {
unsigned ring =
ctx->type == PIPE_SHADER_GEOMETRY ? SI_GS_RING_ESGS
@@ -5976,16 +5976,16 @@ static const char *si_get_shader_name(struct si_shader *shader,
{
switch (processor) {
case PIPE_SHADER_VERTEX:
- if (shader->key.vs.as_es)
+ if (shader->key.as_es)
return "Vertex Shader as ES";
- else if (shader->key.vs.as_ls)
+ else if (shader->key.as_ls)
return "Vertex Shader as LS";
else
return "Vertex Shader as VS";
case PIPE_SHADER_TESS_CTRL:
return "Tessellation Control Shader";
case PIPE_SHADER_TESS_EVAL:
- if (shader->key.tes.as_es)
+ if (shader->key.as_es)
return "Tessellation Evaluation Shader as ES";
else
return "Tessellation Evaluation Shader as VS";
@@ -6227,7 +6227,7 @@ si_generate_gs_copy_shader(struct si_screen *sscreen,
return shader;
}
-static void si_dump_shader_key(unsigned shader, union si_shader_key *key,
+static void si_dump_shader_key(unsigned shader, struct si_shader_key *key,
FILE *f)
{
int i;
@@ -6236,23 +6236,23 @@ static void si_dump_shader_key(unsigned shader, union si_shader_key *key,
switch (shader) {
case PIPE_SHADER_VERTEX:
- fprintf(f, " instance_divisors = {");
- for (i = 0; i < ARRAY_SIZE(key->vs.prolog.instance_divisors); i++)
+ fprintf(f, " part.vs.prolog.instance_divisors = {");
+ for (i = 0; i < ARRAY_SIZE(key->part.vs.prolog.instance_divisors); i++)
fprintf(f, !i ? "%u" : ", %u",
- key->vs.prolog.instance_divisors[i]);
+ key->part.vs.prolog.instance_divisors[i]);
fprintf(f, "}\n");
- fprintf(f, " as_es = %u\n", key->vs.as_es);
- fprintf(f, " as_ls = %u\n", key->vs.as_ls);
- fprintf(f, " export_prim_id = %u\n", key->vs.epilog.export_prim_id);
+ fprintf(f, " part.vs.epilog.export_prim_id = %u\n", key->part.vs.epilog.export_prim_id);
+ fprintf(f, " as_es = %u\n", key->as_es);
+ fprintf(f, " as_ls = %u\n", key->as_ls);
break;
case PIPE_SHADER_TESS_CTRL:
- fprintf(f, " prim_mode = %u\n", key->tcs.epilog.prim_mode);
+ fprintf(f, " part.tcs.epilog.prim_mode = %u\n", key->part.tcs.epilog.prim_mode);
break;
case PIPE_SHADER_TESS_EVAL:
- fprintf(f, " as_es = %u\n", key->tes.as_es);
- fprintf(f, " export_prim_id = %u\n", key->tes.epilog.export_prim_id);
+ fprintf(f, " part.tes.epilog.export_prim_id = %u\n", key->part.tes.epilog.export_prim_id);
+ fprintf(f, " as_es = %u\n", key->as_es);
break;
case PIPE_SHADER_GEOMETRY:
@@ -6260,22 +6260,22 @@ static void si_dump_shader_key(unsigned shader, union si_shader_key *key,
break;
case PIPE_SHADER_FRAGMENT:
- fprintf(f, " prolog.color_two_side = %u\n", key->ps.prolog.color_two_side);
- fprintf(f, " prolog.flatshade_colors = %u\n", key->ps.prolog.flatshade_colors);
- fprintf(f, " prolog.poly_stipple = %u\n", key->ps.prolog.poly_stipple);
- fprintf(f, " prolog.force_persp_sample_interp = %u\n", key->ps.prolog.force_persp_sample_interp);
- fprintf(f, " prolog.force_linear_sample_interp = %u\n", key->ps.prolog.force_linear_sample_interp);
- fprintf(f, " prolog.force_persp_center_interp = %u\n", key->ps.prolog.force_persp_center_interp);
- fprintf(f, " prolog.force_linear_center_interp = %u\n", key->ps.prolog.force_linear_center_interp);
- fprintf(f, " prolog.bc_optimize_for_persp = %u\n", key->ps.prolog.bc_optimize_for_persp);
- fprintf(f, " prolog.bc_optimize_for_linear = %u\n", key->ps.prolog.bc_optimize_for_linear);
- fprintf(f, " epilog.spi_shader_col_format = 0x%x\n", key->ps.epilog.spi_shader_col_format);
- fprintf(f, " epilog.color_is_int8 = 0x%X\n", key->ps.epilog.color_is_int8);
- fprintf(f, " epilog.last_cbuf = %u\n", key->ps.epilog.last_cbuf);
- fprintf(f, " epilog.alpha_func = %u\n", key->ps.epilog.alpha_func);
- fprintf(f, " epilog.alpha_to_one = %u\n", key->ps.epilog.alpha_to_one);
- fprintf(f, " epilog.poly_line_smoothing = %u\n", key->ps.epilog.poly_line_smoothing);
- fprintf(f, " epilog.clamp_color = %u\n", key->ps.epilog.clamp_color);
+ fprintf(f, " part.ps.prolog.color_two_side = %u\n", key->part.ps.prolog.color_two_side);
+ fprintf(f, " part.ps.prolog.flatshade_colors = %u\n", key->part.ps.prolog.flatshade_colors);
+ fprintf(f, " part.ps.prolog.poly_stipple = %u\n", key->part.ps.prolog.poly_stipple);
+ fprintf(f, " part.ps.prolog.force_persp_sample_interp = %u\n", key->part.ps.prolog.force_persp_sample_interp);
+ fprintf(f, " part.ps.prolog.force_linear_sample_interp = %u\n", key->part.ps.prolog.force_linear_sample_interp);
+ fprintf(f, " part.ps.prolog.force_persp_center_interp = %u\n", key->part.ps.prolog.force_persp_center_interp);
+ fprintf(f, " part.ps.prolog.force_linear_center_interp = %u\n", key->part.ps.prolog.force_linear_center_interp);
+ fprintf(f, " part.ps.prolog.bc_optimize_for_persp = %u\n", key->part.ps.prolog.bc_optimize_for_persp);
+ fprintf(f, " part.ps.prolog.bc_optimize_for_linear = %u\n", key->part.ps.prolog.bc_optimize_for_linear);
+ fprintf(f, " part.ps.epilog.spi_shader_col_format = 0x%x\n", key->part.ps.epilog.spi_shader_col_format);
+ fprintf(f, " part.ps.epilog.color_is_int8 = 0x%X\n", key->part.ps.epilog.color_is_int8);
+ fprintf(f, " part.ps.epilog.last_cbuf = %u\n", key->part.ps.epilog.last_cbuf);
+ fprintf(f, " part.ps.epilog.alpha_func = %u\n", key->part.ps.epilog.alpha_func);
+ fprintf(f, " part.ps.epilog.alpha_to_one = %u\n", key->part.ps.epilog.alpha_to_one);
+ fprintf(f, " part.ps.epilog.poly_line_smoothing = %u\n", key->part.ps.epilog.poly_line_smoothing);
+ fprintf(f, " part.ps.epilog.clamp_color = %u\n", key->part.ps.epilog.clamp_color);
break;
default:
@@ -6427,9 +6427,7 @@ static void si_eliminate_const_vs_outputs(struct si_shader_context *ctx)
exports.num = 0;
- if ((ctx->type == PIPE_SHADER_VERTEX &&
- (shader->key.vs.as_es || shader->key.vs.as_ls)) ||
- (ctx->type == PIPE_SHADER_TESS_EVAL && shader->key.tes.as_es))
+ if (shader->key.as_es || shader->key.as_ls)
return;
/* Process all LLVM instructions. */
@@ -6519,9 +6517,9 @@ static bool si_compile_tgsi_main(struct si_shader_context *ctx,
switch (ctx->type) {
case PIPE_SHADER_VERTEX:
ctx->load_input = declare_input_vs;
- if (shader->key.vs.as_ls)
+ if (shader->key.as_ls)
bld_base->emit_epilogue = si_llvm_emit_ls_epilogue;
- else if (shader->key.vs.as_es)
+ else if (shader->key.as_es)
bld_base->emit_epilogue = si_llvm_emit_es_epilogue;
else
bld_base->emit_epilogue = si_llvm_emit_vs_epilogue;
@@ -6534,7 +6532,7 @@ static bool si_compile_tgsi_main(struct si_shader_context *ctx,
break;
case PIPE_SHADER_TESS_EVAL:
bld_base->emit_fetch_funcs[TGSI_FILE_INPUT] = fetch_input_tes;
- if (shader->key.tes.as_es)
+ if (shader->key.as_es)
bld_base->emit_epilogue = si_llvm_emit_es_epilogue;
else
bld_base->emit_epilogue = si_llvm_emit_vs_epilogue;
@@ -6587,7 +6585,7 @@ static void si_get_vs_prolog_key(struct si_shader *shader,
struct tgsi_shader_info *info = &shader->selector->info;
memset(key, 0, sizeof(*key));
- key->vs_prolog.states = shader->key.vs.prolog;
+ key->vs_prolog.states = shader->key.part.vs.prolog;
key->vs_prolog.num_input_sgprs = shader->info.num_input_sgprs;
key->vs_prolog.last_input = MAX2(1, info->num_inputs) - 1;
@@ -6609,7 +6607,7 @@ static void si_get_vs_epilog_key(struct si_shader *shader,
key->vs_epilog.states = *states;
/* Set up the PrimitiveID output. */
- if (shader->key.vs.epilog.export_prim_id) {
+ if (shader->key.part.vs.epilog.export_prim_id) {
unsigned index = shader->selector->info.num_outputs;
unsigned offset = shader->info.nr_param_exports++;
@@ -6630,7 +6628,7 @@ static void si_get_ps_prolog_key(struct si_shader *shader,
struct tgsi_shader_info *info = &shader->selector->info;
memset(key, 0, sizeof(*key));
- key->ps_prolog.states = shader->key.ps.prolog;
+ key->ps_prolog.states = shader->key.part.ps.prolog;
key->ps_prolog.colors_read = info->colors_read;
key->ps_prolog.num_input_sgprs = shader->info.num_input_sgprs;
key->ps_prolog.num_input_vgprs = shader->info.num_input_vgprs;
@@ -6646,7 +6644,7 @@ static void si_get_ps_prolog_key(struct si_shader *shader,
if (info->colors_read) {
unsigned *color = shader->selector->color_attr_index;
- if (shader->key.ps.prolog.color_two_side) {
+ if (shader->key.part.ps.prolog.color_two_side) {
/* BCOLORs are stored after the last input. */
key->ps_prolog.num_interp_inputs = info->num_inputs;
key->ps_prolog.face_vgpr_index = shader->info.face_vgpr_index;
@@ -6662,7 +6660,7 @@ static void si_get_ps_prolog_key(struct si_shader *shader,
key->ps_prolog.color_attr_index[i] = color[i];
- if (shader->key.ps.prolog.flatshade_colors &&
+ if (shader->key.part.ps.prolog.flatshade_colors &&
interp == TGSI_INTERPOLATE_COLOR)
interp = TGSI_INTERPOLATE_CONSTANT;
@@ -6673,9 +6671,9 @@ static void si_get_ps_prolog_key(struct si_shader *shader,
case TGSI_INTERPOLATE_PERSPECTIVE:
case TGSI_INTERPOLATE_COLOR:
/* Force the interpolation location for colors here. */
- if (shader->key.ps.prolog.force_persp_sample_interp)
+ if (shader->key.part.ps.prolog.force_persp_sample_interp)
location = TGSI_INTERPOLATE_LOC_SAMPLE;
- if (shader->key.ps.prolog.force_persp_center_interp)
+ if (shader->key.part.ps.prolog.force_persp_center_interp)
location = TGSI_INTERPOLATE_LOC_CENTER;
switch (location) {
@@ -6700,9 +6698,9 @@ static void si_get_ps_prolog_key(struct si_shader *shader,
break;
case TGSI_INTERPOLATE_LINEAR:
/* Force the interpolation location for colors here. */
- if (shader->key.ps.prolog.force_linear_sample_interp)
+ if (shader->key.part.ps.prolog.force_linear_sample_interp)
location = TGSI_INTERPOLATE_LOC_SAMPLE;
- if (shader->key.ps.prolog.force_linear_center_interp)
+ if (shader->key.part.ps.prolog.force_linear_center_interp)
location = TGSI_INTERPOLATE_LOC_CENTER;
/* The VGPR assignment for non-monolithic shaders
@@ -6767,7 +6765,7 @@ static void si_get_ps_epilog_key(struct si_shader *shader,
key->ps_epilog.writes_z = info->writes_z;
key->ps_epilog.writes_stencil = info->writes_stencil;
key->ps_epilog.writes_samplemask = info->writes_samplemask;
- key->ps_epilog.states = shader->key.ps.epilog;
+ key->ps_epilog.states = shader->key.part.ps.epilog;
}
/**
@@ -7077,7 +7075,7 @@ int si_compile_tgsi_shader(struct si_screen *sscreen,
bool need_epilog;
need_prolog = sel->info.num_inputs;
- need_epilog = !shader->key.vs.as_es && !shader->key.vs.as_ls;
+ need_epilog = !shader->key.as_es && !shader->key.as_ls;
parts[need_prolog ? 1 : 0] = ctx.main_fn;
@@ -7090,7 +7088,7 @@ int si_compile_tgsi_shader(struct si_screen *sscreen,
if (need_epilog) {
union si_shader_part_key epilog_key;
- si_get_vs_epilog_key(shader, &shader->key.vs.epilog, &epilog_key);
+ si_get_vs_epilog_key(shader, &shader->key.part.vs.epilog, &epilog_key);
si_build_vs_epilog_function(&ctx, &epilog_key);
parts[need_prolog ? 2 : 1] = ctx.main_fn;
}
@@ -7104,19 +7102,19 @@ int si_compile_tgsi_shader(struct si_screen *sscreen,
parts[0] = ctx.main_fn;
memset(&epilog_key, 0, sizeof(epilog_key));
- epilog_key.tcs_epilog.states = shader->key.tcs.epilog;
+ epilog_key.tcs_epilog.states = shader->key.part.tcs.epilog;
si_build_tcs_epilog_function(&ctx, &epilog_key);
parts[1] = ctx.main_fn;
si_build_wrapper_function(&ctx, parts, 2, 0);
} else if (is_monolithic && ctx.type == PIPE_SHADER_TESS_EVAL &&
- !shader->key.tes.as_es) {
+ !shader->key.as_es) {
LLVMValueRef parts[2];
union si_shader_part_key epilog_key;
parts[0] = ctx.main_fn;
- si_get_vs_epilog_key(shader, &shader->key.tes.epilog, &epilog_key);
+ si_get_vs_epilog_key(shader, &shader->key.part.tes.epilog, &epilog_key);
si_build_vs_epilog_function(&ctx, &epilog_key);
parts[1] = ctx.main_fn;
@@ -7128,7 +7126,7 @@ int si_compile_tgsi_shader(struct si_screen *sscreen,
parts[1] = ctx.main_fn;
memset(&prolog_key, 0, sizeof(prolog_key));
- prolog_key.gs_prolog.states = shader->key.gs.prolog;
+ prolog_key.gs_prolog.states = shader->key.part.gs.prolog;
si_build_gs_prolog_function(&ctx, &prolog_key);
parts[0] = ctx.main_fn;
@@ -7319,16 +7317,16 @@ si_get_shader_part(struct si_screen *sscreen,
break;
case PIPE_SHADER_TESS_CTRL:
assert(!prolog);
- shader.key.tcs.epilog = key->tcs_epilog.states;
+ shader.key.part.tcs.epilog = key->tcs_epilog.states;
break;
case PIPE_SHADER_GEOMETRY:
assert(prolog);
break;
case PIPE_SHADER_FRAGMENT:
if (prolog)
- shader.key.ps.prolog = key->ps_prolog.states;
+ shader.key.part.ps.prolog = key->ps_prolog.states;
else
- shader.key.ps.epilog = key->ps_epilog.states;
+ shader.key.part.ps.epilog = key->ps_epilog.states;
break;
default:
unreachable("bad shader part");
@@ -7556,9 +7554,9 @@ static bool si_shader_select_vs_parts(struct si_screen *sscreen,
}
/* Get the epilog. */
- if (!shader->key.vs.as_es && !shader->key.vs.as_ls &&
+ if (!shader->key.as_es && !shader->key.as_ls &&
!si_get_vs_epilog(sscreen, tm, shader, debug,
- &shader->key.vs.epilog))
+ &shader->key.part.vs.epilog))
return false;
return true;
@@ -7572,12 +7570,12 @@ static bool si_shader_select_tes_parts(struct si_screen *sscreen,
struct si_shader *shader,
struct pipe_debug_callback *debug)
{
- if (shader->key.tes.as_es)
+ if (shader->key.as_es)
return true;
/* TES compiled as VS. */
return si_get_vs_epilog(sscreen, tm, shader, debug,
- &shader->key.tes.epilog);
+ &shader->key.part.tes.epilog);
}
/**
@@ -7637,7 +7635,7 @@ static bool si_shader_select_tcs_parts(struct si_screen *sscreen,
/* Get the epilog. */
memset(&epilog_key, 0, sizeof(epilog_key));
- epilog_key.tcs_epilog.states = shader->key.tcs.epilog;
+ epilog_key.tcs_epilog.states = shader->key.part.tcs.epilog;
shader->epilog = si_get_shader_part(sscreen, &sscreen->tcs_epilogs,
PIPE_SHADER_TESS_CTRL, false,
@@ -7657,11 +7655,11 @@ static bool si_shader_select_gs_parts(struct si_screen *sscreen,
{
union si_shader_part_key prolog_key;
- if (!shader->key.gs.prolog.tri_strip_adj_fix)
+ if (!shader->key.part.gs.prolog.tri_strip_adj_fix)
return true;
memset(&prolog_key, 0, sizeof(prolog_key));
- prolog_key.gs_prolog.states = shader->key.gs.prolog;
+ prolog_key.gs_prolog.states = shader->key.part.gs.prolog;
shader->prolog = si_get_shader_part(sscreen, &sscreen->gs_prologs,
PIPE_SHADER_GEOMETRY, true,
@@ -8057,34 +8055,34 @@ static bool si_shader_select_ps_parts(struct si_screen *sscreen,
return false;
/* Enable POS_FIXED_PT if polygon stippling is enabled. */
- if (shader->key.ps.prolog.poly_stipple) {
+ if (shader->key.part.ps.prolog.poly_stipple) {
shader->config.spi_ps_input_ena |= S_0286CC_POS_FIXED_PT_ENA(1);
assert(G_0286CC_POS_FIXED_PT_ENA(shader->config.spi_ps_input_addr));
}
/* Set up the enable bits for per-sample shading if needed. */
- if (shader->key.ps.prolog.force_persp_sample_interp &&
+ if (shader->key.part.ps.prolog.force_persp_sample_interp &&
(G_0286CC_PERSP_CENTER_ENA(shader->config.spi_ps_input_ena) ||
G_0286CC_PERSP_CENTROID_ENA(shader->config.spi_ps_input_ena))) {
shader->config.spi_ps_input_ena &= C_0286CC_PERSP_CENTER_ENA;
shader->config.spi_ps_input_ena &= C_0286CC_PERSP_CENTROID_ENA;
shader->config.spi_ps_input_ena |= S_0286CC_PERSP_SAMPLE_ENA(1);
}
- if (shader->key.ps.prolog.force_linear_sample_interp &&
+ if (shader->key.part.ps.prolog.force_linear_sample_interp &&
(G_0286CC_LINEAR_CENTER_ENA(shader->config.spi_ps_input_ena) ||
G_0286CC_LINEAR_CENTROID_ENA(shader->config.spi_ps_input_ena))) {
shader->config.spi_ps_input_ena &= C_0286CC_LINEAR_CENTER_ENA;
shader->config.spi_ps_input_ena &= C_0286CC_LINEAR_CENTROID_ENA;
shader->config.spi_ps_input_ena |= S_0286CC_LINEAR_SAMPLE_ENA(1);
}
- if (shader->key.ps.prolog.force_persp_center_interp &&
+ if (shader->key.part.ps.prolog.force_persp_center_interp &&
(G_0286CC_PERSP_SAMPLE_ENA(shader->config.spi_ps_input_ena) ||
G_0286CC_PERSP_CENTROID_ENA(shader->config.spi_ps_input_ena))) {
shader->config.spi_ps_input_ena &= C_0286CC_PERSP_SAMPLE_ENA;
shader->config.spi_ps_input_ena &= C_0286CC_PERSP_CENTROID_ENA;
shader->config.spi_ps_input_ena |= S_0286CC_PERSP_CENTER_ENA(1);
}
- if (shader->key.ps.prolog.force_linear_center_interp &&
+ if (shader->key.part.ps.prolog.force_linear_center_interp &&
(G_0286CC_LINEAR_SAMPLE_ENA(shader->config.spi_ps_input_ena) ||
G_0286CC_LINEAR_CENTROID_ENA(shader->config.spi_ps_input_ena))) {
shader->config.spi_ps_input_ena &= C_0286CC_LINEAR_SAMPLE_ENA;
@@ -8108,7 +8106,7 @@ static bool si_shader_select_ps_parts(struct si_screen *sscreen,
/* The sample mask input is always enabled, because the API shader always
* passes it through to the epilog. Disable it here if it's unused.
*/
- if (!shader->key.ps.epilog.poly_line_smoothing &&
+ if (!shader->key.part.ps.epilog.poly_line_smoothing &&
!shader->selector->info.reads_samplemask)
shader->config.spi_ps_input_ena &= C_0286CC_SAMPLE_COVERAGE_ENA;
@@ -8137,14 +8135,12 @@ int si_shader_create(struct si_screen *sscreen, LLVMTargetMachineRef tm,
* workaround must be applied.
*/
if (!mainp ||
+ shader->key.as_es != mainp->key.as_es ||
+ shader->key.as_ls != mainp->key.as_ls ||
(sel->type == PIPE_SHADER_VERTEX &&
- (shader->key.vs.as_es != mainp->key.vs.as_es ||
- shader->key.vs.as_ls != mainp->key.vs.as_ls ||
- shader->key.vs.fix_fetch)) ||
- (sel->type == PIPE_SHADER_TESS_EVAL &&
- shader->key.tes.as_es != mainp->key.tes.as_es) ||
+ shader->key.mono.vs.fix_fetch) ||
(sel->type == PIPE_SHADER_TESS_CTRL &&
- shader->key.tcs.epilog.inputs_to_copy) ||
+ shader->key.mono.tcs.inputs_to_copy) ||
sel->type == PIPE_SHADER_COMPUTE) {
/* Monolithic shader (compiled as a whole, has many variants,
* may take a long time to compile).
diff --git a/src/gallium/drivers/radeonsi/si_shader.h b/src/gallium/drivers/radeonsi/si_shader.h
index 59e7bfb..bed22c1 100644
--- a/src/gallium/drivers/radeonsi/si_shader.h
+++ b/src/gallium/drivers/radeonsi/si_shader.h
@@ -330,7 +330,6 @@ struct si_vs_epilog_bits {
/* Common TCS bits between the shader key and the epilog key. */
struct si_tcs_epilog_bits {
unsigned prim_mode:3;
- uint64_t inputs_to_copy;
};
struct si_gs_prolog_bits {
@@ -398,30 +397,44 @@ union si_shader_part_key {
} ps_epilog;
};
-union si_shader_key {
- struct {
- struct si_ps_prolog_bits prolog;
- struct si_ps_epilog_bits epilog;
- } ps;
- struct {
- struct si_vs_prolog_bits prolog;
- struct si_vs_epilog_bits epilog;
- unsigned as_es:1; /* export shader */
- unsigned as_ls:1; /* local shader */
-
- /* One pair of bits for every input: SI_FIX_FETCH_* enums. */
- uint32_t fix_fetch;
- } vs;
- struct {
- struct si_tcs_epilog_bits epilog;
- } tcs; /* tessellation control shader */
- struct {
- struct si_vs_epilog_bits epilog; /* same as VS */
- unsigned as_es:1; /* export shader */
- } tes; /* tessellation evaluation shader */
- struct {
- struct si_gs_prolog_bits prolog;
- } gs;
+struct si_shader_key {
+ /* Prolog and epilog flags. */
+ union {
+ struct {
+ struct si_ps_prolog_bits prolog;
+ struct si_ps_epilog_bits epilog;
+ } ps;
+ struct {
+ struct si_vs_prolog_bits prolog;
+ struct si_vs_epilog_bits epilog;
+ } vs;
+ struct {
+ struct si_tcs_epilog_bits epilog;
+ } tcs; /* tessellation control shader */
+ struct {
+ struct si_vs_epilog_bits epilog; /* same as VS */
+ } tes; /* tessellation evaluation shader */
+ struct {
+ struct si_gs_prolog_bits prolog;
+ } gs;
+ } part;
+
+ /* 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 */
+ unsigned as_ls:1; /* local shader */
+
+ /* Flags for monolithic compilation only. */
+ union {
+ struct {
+ /* One pair of bits for every input: SI_FIX_FETCH_* enums. */
+ uint32_t fix_fetch;
+ } vs;
+ struct {
+ uint64_t inputs_to_copy; /* for fixed-func TCS */
+ } tcs;
+ } mono;
};
struct si_shader_config {
@@ -470,7 +483,7 @@ struct si_shader {
struct si_pm4_state *pm4;
struct r600_resource *bo;
struct r600_resource *scratch_bo;
- union si_shader_key key;
+ struct si_shader_key key;
bool is_binary_shared;
bool is_gs_copy_shader;
diff --git a/src/gallium/drivers/radeonsi/si_state.c b/src/gallium/drivers/radeonsi/si_state.c
index 4a41aa8..096c641 100644
--- a/src/gallium/drivers/radeonsi/si_state.c
+++ b/src/gallium/drivers/radeonsi/si_state.c
@@ -119,7 +119,7 @@ static void si_emit_cb_render_state(struct si_context *sctx, struct r600_atom *a
if (sctx->b.family == CHIP_STONEY) {
unsigned spi_shader_col_format =
sctx->ps_shader.cso ?
- sctx->ps_shader.current->key.ps.epilog.spi_shader_col_format : 0;
+ sctx->ps_shader.current->key.part.ps.epilog.spi_shader_col_format : 0;
unsigned sx_ps_downconvert = 0;
unsigned sx_blend_opt_epsilon = 0;
unsigned sx_blend_opt_control = 0;
diff --git a/src/gallium/drivers/radeonsi/si_state_shaders.c b/src/gallium/drivers/radeonsi/si_state_shaders.c
index d0869e3..3323d3c 100644
--- a/src/gallium/drivers/radeonsi/si_state_shaders.c
+++ b/src/gallium/drivers/radeonsi/si_state_shaders.c
@@ -626,7 +626,7 @@ static unsigned si_get_ps_num_interp(struct si_shader *ps)
unsigned num_colors = !!(info->colors_read & 0x0f) +
!!(info->colors_read & 0xf0);
unsigned num_interp = ps->selector->info.num_inputs +
- (ps->key.ps.prolog.color_two_side ? num_colors : 0);
+ (ps->key.part.ps.prolog.color_two_side ? num_colors : 0);
assert(num_interp <= 32);
return MIN2(num_interp, 32);
@@ -634,7 +634,7 @@ static unsigned si_get_ps_num_interp(struct si_shader *ps)
static unsigned si_get_spi_shader_col_format(struct si_shader *shader)
{
- unsigned value = shader->key.ps.epilog.spi_shader_col_format;
+ unsigned value = shader->key.part.ps.epilog.spi_shader_col_format;
unsigned i, num_targets = (util_last_bit(value) + 3) / 4;
/* If the i-th target format is set, all previous target formats must
@@ -705,30 +705,30 @@ static void si_shader_ps(struct si_shader *shader)
G_0286CC_PERSP_PULL_MODEL_ENA(input_ena));
/* Validate interpolation optimization flags (read as implications). */
- assert(!shader->key.ps.prolog.bc_optimize_for_persp ||
+ assert(!shader->key.part.ps.prolog.bc_optimize_for_persp ||
(G_0286CC_PERSP_CENTER_ENA(input_ena) &&
G_0286CC_PERSP_CENTROID_ENA(input_ena)));
- assert(!shader->key.ps.prolog.bc_optimize_for_linear ||
+ assert(!shader->key.part.ps.prolog.bc_optimize_for_linear ||
(G_0286CC_LINEAR_CENTER_ENA(input_ena) &&
G_0286CC_LINEAR_CENTROID_ENA(input_ena)));
- assert(!shader->key.ps.prolog.force_persp_center_interp ||
+ assert(!shader->key.part.ps.prolog.force_persp_center_interp ||
(!G_0286CC_PERSP_SAMPLE_ENA(input_ena) &&
!G_0286CC_PERSP_CENTROID_ENA(input_ena)));
- assert(!shader->key.ps.prolog.force_linear_center_interp ||
+ assert(!shader->key.part.ps.prolog.force_linear_center_interp ||
(!G_0286CC_LINEAR_SAMPLE_ENA(input_ena) &&
!G_0286CC_LINEAR_CENTROID_ENA(input_ena)));
- assert(!shader->key.ps.prolog.force_persp_sample_interp ||
+ assert(!shader->key.part.ps.prolog.force_persp_sample_interp ||
(!G_0286CC_PERSP_CENTER_ENA(input_ena) &&
!G_0286CC_PERSP_CENTROID_ENA(input_ena)));
- assert(!shader->key.ps.prolog.force_linear_sample_interp ||
+ assert(!shader->key.part.ps.prolog.force_linear_sample_interp ||
(!G_0286CC_LINEAR_CENTER_ENA(input_ena) &&
!G_0286CC_LINEAR_CENTROID_ENA(input_ena)));
/* Validate cases when the optimizations are off (read as implications). */
- assert(shader->key.ps.prolog.bc_optimize_for_persp ||
+ assert(shader->key.part.ps.prolog.bc_optimize_for_persp ||
!G_0286CC_PERSP_CENTER_ENA(input_ena) ||
!G_0286CC_PERSP_CENTROID_ENA(input_ena));
- assert(shader->key.ps.prolog.bc_optimize_for_linear ||
+ assert(shader->key.part.ps.prolog.bc_optimize_for_linear ||
!G_0286CC_LINEAR_CENTER_ENA(input_ena) ||
!G_0286CC_LINEAR_CENTROID_ENA(input_ena));
@@ -818,9 +818,9 @@ static void si_shader_init_pm4_state(struct si_screen *sscreen,
{
switch (shader->selector->type) {
case PIPE_SHADER_VERTEX:
- if (shader->key.vs.as_ls)
+ if (shader->key.as_ls)
si_shader_ls(shader);
- else if (shader->key.vs.as_es)
+ else if (shader->key.as_es)
si_shader_es(sscreen, shader);
else
si_shader_vs(sscreen, shader, NULL);
@@ -829,7 +829,7 @@ static void si_shader_init_pm4_state(struct si_screen *sscreen,
si_shader_hs(shader);
break;
case PIPE_SHADER_TESS_EVAL:
- if (shader->key.tes.as_es)
+ if (shader->key.as_es)
si_shader_es(sscreen, shader);
else
si_shader_vs(sscreen, shader, NULL);
@@ -857,7 +857,7 @@ static unsigned si_get_alpha_test_func(struct si_context *sctx)
/* Compute the key for the hw shader variant */
static inline void si_shader_selector_key(struct pipe_context *ctx,
struct si_shader_selector *sel,
- union si_shader_key *key)
+ struct si_shader_key *key)
{
struct si_context *sctx = (struct si_context *)ctx;
unsigned i;
@@ -870,37 +870,37 @@ static inline void si_shader_selector_key(struct pipe_context *ctx,
unsigned count = MIN2(sel->info.num_inputs,
sctx->vertex_elements->count);
for (i = 0; i < count; ++i)
- key->vs.prolog.instance_divisors[i] =
+ key->part.vs.prolog.instance_divisors[i] =
sctx->vertex_elements->elements[i].instance_divisor;
- key->vs.fix_fetch =
+ key->mono.vs.fix_fetch =
sctx->vertex_elements->fix_fetch &
u_bit_consecutive(0, 2 * count);
}
if (sctx->tes_shader.cso)
- key->vs.as_ls = 1;
+ key->as_ls = 1;
else if (sctx->gs_shader.cso)
- key->vs.as_es = 1;
+ key->as_es = 1;
if (!sctx->gs_shader.cso && sctx->ps_shader.cso &&
sctx->ps_shader.cso->info.uses_primid)
- key->vs.epilog.export_prim_id = 1;
+ key->part.vs.epilog.export_prim_id = 1;
break;
case PIPE_SHADER_TESS_CTRL:
- key->tcs.epilog.prim_mode =
+ key->part.tcs.epilog.prim_mode =
sctx->tes_shader.cso->info.properties[TGSI_PROPERTY_TES_PRIM_MODE];
if (sel == sctx->fixed_func_tcs_shader.cso)
- key->tcs.epilog.inputs_to_copy = sctx->vs_shader.cso->outputs_written;
+ key->mono.tcs.inputs_to_copy = sctx->vs_shader.cso->outputs_written;
break;
case PIPE_SHADER_TESS_EVAL:
if (sctx->gs_shader.cso)
- key->tes.as_es = 1;
+ key->as_es = 1;
else if (sctx->ps_shader.cso && sctx->ps_shader.cso->info.uses_primid)
- key->tes.epilog.export_prim_id = 1;
+ key->part.tes.epilog.export_prim_id = 1;
break;
case PIPE_SHADER_GEOMETRY:
- key->gs.prolog.tri_strip_adj_fix = sctx->gs_tri_strip_adj_fix;
+ key->part.gs.prolog.tri_strip_adj_fix = sctx->gs_tri_strip_adj_fix;
break;
case PIPE_SHADER_FRAGMENT: {
struct si_state_rasterizer *rs = sctx->queued.named.rasterizer;
@@ -908,13 +908,13 @@ static inline void si_shader_selector_key(struct pipe_context *ctx,
if (sel->info.properties[TGSI_PROPERTY_FS_COLOR0_WRITES_ALL_CBUFS] &&
sel->info.colors_written == 0x1)
- key->ps.epilog.last_cbuf = MAX2(sctx->framebuffer.state.nr_cbufs, 1) - 1;
+ key->part.ps.epilog.last_cbuf = MAX2(sctx->framebuffer.state.nr_cbufs, 1) - 1;
if (blend) {
/* Select the shader color format based on whether
* blending or alpha are needed.
*/
- key->ps.epilog.spi_shader_col_format =
+ key->part.ps.epilog.spi_shader_col_format =
(blend->blend_enable_4bit & blend->need_src_alpha_4bit &
sctx->framebuffer.spi_shader_col_format_blend_alpha) |
(blend->blend_enable_4bit & ~blend->need_src_alpha_4bit &
@@ -928,29 +928,29 @@ static inline void si_shader_selector_key(struct pipe_context *ctx,
* the same format as the first output.
*/
if (blend->dual_src_blend)
- key->ps.epilog.spi_shader_col_format |=
- (key->ps.epilog.spi_shader_col_format & 0xf) << 4;
+ key->part.ps.epilog.spi_shader_col_format |=
+ (key->part.ps.epilog.spi_shader_col_format & 0xf) << 4;
} else
- key->ps.epilog.spi_shader_col_format = sctx->framebuffer.spi_shader_col_format;
+ key->part.ps.epilog.spi_shader_col_format = sctx->framebuffer.spi_shader_col_format;
/* If alpha-to-coverage is enabled, we have to export alpha
* even if there is no color buffer.
*/
- if (!(key->ps.epilog.spi_shader_col_format & 0xf) &&
+ if (!(key->part.ps.epilog.spi_shader_col_format & 0xf) &&
blend && blend->alpha_to_coverage)
- key->ps.epilog.spi_shader_col_format |= V_028710_SPI_SHADER_32_AR;
+ key->part.ps.epilog.spi_shader_col_format |= V_028710_SPI_SHADER_32_AR;
/* On SI and CIK except Hawaii, the CB doesn't clamp outputs
* to the range supported by the type if a channel has less
* than 16 bits and the export format is 16_ABGR.
*/
if (sctx->b.chip_class <= CIK && sctx->b.family != CHIP_HAWAII)
- key->ps.epilog.color_is_int8 = sctx->framebuffer.color_is_int8;
+ key->part.ps.epilog.color_is_int8 = sctx->framebuffer.color_is_int8;
/* Disable unwritten outputs (if WRITE_ALL_CBUFS isn't enabled). */
- if (!key->ps.epilog.last_cbuf) {
- key->ps.epilog.spi_shader_col_format &= sel->colors_written_4bit;
- key->ps.epilog.color_is_int8 &= sel->info.colors_written;
+ if (!key->part.ps.epilog.last_cbuf) {
+ key->part.ps.epilog.spi_shader_col_format &= sel->colors_written_4bit;
+ key->part.ps.epilog.color_is_int8 &= sel->info.colors_written;
}
if (rs) {
@@ -959,55 +959,55 @@ static inline void si_shader_selector_key(struct pipe_context *ctx,
sctx->current_rast_prim >= PIPE_PRIM_TRIANGLES_ADJACENCY;
bool is_line = !is_poly && sctx->current_rast_prim != PIPE_PRIM_POINTS;
- key->ps.prolog.color_two_side = rs->two_side && sel->info.colors_read;
- key->ps.prolog.flatshade_colors = rs->flatshade && sel->info.colors_read;
+ key->part.ps.prolog.color_two_side = rs->two_side && sel->info.colors_read;
+ key->part.ps.prolog.flatshade_colors = rs->flatshade && sel->info.colors_read;
if (sctx->queued.named.blend) {
- key->ps.epilog.alpha_to_one = sctx->queued.named.blend->alpha_to_one &&
+ key->part.ps.epilog.alpha_to_one = sctx->queued.named.blend->alpha_to_one &&
rs->multisample_enable;
}
- key->ps.prolog.poly_stipple = rs->poly_stipple_enable && is_poly;
- key->ps.epilog.poly_line_smoothing = ((is_poly && rs->poly_smooth) ||
+ key->part.ps.prolog.poly_stipple = rs->poly_stipple_enable && is_poly;
+ key->part.ps.epilog.poly_line_smoothing = ((is_poly && rs->poly_smooth) ||
(is_line && rs->line_smooth)) &&
sctx->framebuffer.nr_samples <= 1;
- key->ps.epilog.clamp_color = rs->clamp_fragment_color;
+ key->part.ps.epilog.clamp_color = rs->clamp_fragment_color;
if (rs->force_persample_interp &&
rs->multisample_enable &&
sctx->framebuffer.nr_samples > 1 &&
sctx->ps_iter_samples > 1) {
- key->ps.prolog.force_persp_sample_interp =
+ key->part.ps.prolog.force_persp_sample_interp =
sel->info.uses_persp_center ||
sel->info.uses_persp_centroid;
- key->ps.prolog.force_linear_sample_interp =
+ key->part.ps.prolog.force_linear_sample_interp =
sel->info.uses_linear_center ||
sel->info.uses_linear_centroid;
} else if (rs->multisample_enable &&
sctx->framebuffer.nr_samples > 1) {
- key->ps.prolog.bc_optimize_for_persp =
+ key->part.ps.prolog.bc_optimize_for_persp =
sel->info.uses_persp_center &&
sel->info.uses_persp_centroid;
- key->ps.prolog.bc_optimize_for_linear =
+ key->part.ps.prolog.bc_optimize_for_linear =
sel->info.uses_linear_center &&
sel->info.uses_linear_centroid;
} else {
/* Make sure SPI doesn't compute more than 1 pair
* of (i,j), which is the optimization here. */
- key->ps.prolog.force_persp_center_interp =
+ key->part.ps.prolog.force_persp_center_interp =
sel->info.uses_persp_center +
sel->info.uses_persp_centroid +
sel->info.uses_persp_sample > 1;
- key->ps.prolog.force_linear_center_interp =
+ key->part.ps.prolog.force_linear_center_interp =
sel->info.uses_linear_center +
sel->info.uses_linear_centroid +
sel->info.uses_linear_sample > 1;
}
}
- key->ps.epilog.alpha_func = si_get_alpha_test_func(sctx);
+ key->part.ps.epilog.alpha_func = si_get_alpha_test_func(sctx);
break;
}
default:
@@ -1018,7 +1018,7 @@ static inline void si_shader_selector_key(struct pipe_context *ctx,
/* Select the hw shader variant depending on the current state. */
static int si_shader_select_with_key(struct si_screen *sscreen,
struct si_shader_ctx_state *state,
- union si_shader_key *key,
+ struct si_shader_key *key,
LLVMTargetMachineRef tm,
struct pipe_debug_callback *debug,
bool wait,
@@ -1101,7 +1101,7 @@ static int si_shader_select(struct pipe_context *ctx,
struct si_shader_ctx_state *state)
{
struct si_context *sctx = (struct si_context *)ctx;
- union si_shader_key key;
+ struct si_shader_key key;
si_shader_selector_key(ctx, state->cso, &key);
return si_shader_select_with_key(sctx->screen, state, &key,
@@ -1110,7 +1110,7 @@ static int si_shader_select(struct pipe_context *ctx,
}
static void si_parse_next_shader_property(const struct tgsi_shader_info *info,
- union si_shader_key *key)
+ struct si_shader_key *key)
{
unsigned next_shader = info->properties[TGSI_PROPERTY_NEXT_SHADER];
@@ -1118,11 +1118,11 @@ static void si_parse_next_shader_property(const struct tgsi_shader_info *info,
case PIPE_SHADER_VERTEX:
switch (next_shader) {
case PIPE_SHADER_GEOMETRY:
- key->vs.as_es = 1;
+ key->as_es = 1;
break;
case PIPE_SHADER_TESS_CTRL:
case PIPE_SHADER_TESS_EVAL:
- key->vs.as_ls = 1;
+ key->as_ls = 1;
break;
default:
/* If POSITION isn't written, it can't be a HW VS.
@@ -1136,7 +1136,7 @@ static void si_parse_next_shader_property(const struct tgsi_shader_info *info,
case PIPE_SHADER_TESS_EVAL:
if (next_shader == PIPE_SHADER_GEOMETRY)
- key->tes.as_es = 1;
+ key->as_es = 1;
break;
}
}
@@ -1214,7 +1214,7 @@ void si_init_shader_selector_async(void *job, int thread_index)
/* Pre-compilation. */
if (sscreen->b.debug_flags & DBG_PRECOMPILE) {
struct si_shader_ctx_state state = {sel};
- union si_shader_key key;
+ struct si_shader_key key;
memset(&key, 0, sizeof(key));
si_parse_next_shader_property(&sel->info, &key);
@@ -1224,19 +1224,19 @@ void si_init_shader_selector_async(void *job, int thread_index)
*/
switch (sel->type) {
case PIPE_SHADER_TESS_CTRL:
- key.tcs.epilog.prim_mode = PIPE_PRIM_TRIANGLES;
+ key.part.tcs.epilog.prim_mode = PIPE_PRIM_TRIANGLES;
break;
case PIPE_SHADER_FRAGMENT:
- key.ps.prolog.bc_optimize_for_persp =
+ key.part.ps.prolog.bc_optimize_for_persp =
sel->info.uses_persp_center &&
sel->info.uses_persp_centroid;
- key.ps.prolog.bc_optimize_for_linear =
+ key.part.ps.prolog.bc_optimize_for_linear =
sel->info.uses_linear_center &&
sel->info.uses_linear_centroid;
- key.ps.epilog.alpha_func = PIPE_FUNC_ALWAYS;
+ key.part.ps.epilog.alpha_func = PIPE_FUNC_ALWAYS;
for (i = 0; i < 8; i++)
if (sel->info.colors_written & (1 << i))
- key.ps.epilog.spi_shader_col_format |=
+ key.part.ps.epilog.spi_shader_col_format |=
V_028710_SPI_SHADER_FP16_ABGR << (i * 4);
break;
}
@@ -1521,9 +1521,9 @@ static void si_delete_shader(struct si_context *sctx, struct si_shader *shader)
if (shader->pm4) {
switch (shader->selector->type) {
case PIPE_SHADER_VERTEX:
- if (shader->key.vs.as_ls)
+ if (shader->key.as_ls)
si_pm4_delete_state(sctx, ls, shader->pm4);
- else if (shader->key.vs.as_es)
+ else if (shader->key.as_es)
si_pm4_delete_state(sctx, es, shader->pm4);
else
si_pm4_delete_state(sctx, vs, shader->pm4);
@@ -1532,7 +1532,7 @@ static void si_delete_shader(struct si_context *sctx, struct si_shader *shader)
si_pm4_delete_state(sctx, hs, shader->pm4);
break;
case PIPE_SHADER_TESS_EVAL:
- if (shader->key.tes.as_es)
+ if (shader->key.as_es)
si_pm4_delete_state(sctx, es, shader->pm4);
else
si_pm4_delete_state(sctx, vs, shader->pm4);
@@ -1673,7 +1673,7 @@ static void si_emit_spi_map(struct si_context *sctx, struct r600_atom *atom)
}
}
- if (ps->key.ps.prolog.color_two_side) {
+ if (ps->key.part.ps.prolog.color_two_side) {
unsigned bcol = TGSI_SEMANTIC_BCOLOR;
for (i = 0; i < 2; i++) {
@@ -2281,8 +2281,8 @@ bool si_update_shaders(struct si_context *sctx)
si_mark_atom_dirty(sctx, &sctx->db_render_state);
}
- if (sctx->smoothing_enabled != sctx->ps_shader.current->key.ps.epilog.poly_line_smoothing) {
- sctx->smoothing_enabled = sctx->ps_shader.current->key.ps.epilog.poly_line_smoothing;
+ if (sctx->smoothing_enabled != sctx->ps_shader.current->key.part.ps.epilog.poly_line_smoothing) {
+ sctx->smoothing_enabled = sctx->ps_shader.current->key.part.ps.epilog.poly_line_smoothing;
si_mark_atom_dirty(sctx, &sctx->msaa_config);
if (sctx->b.chip_class == SI)
More information about the mesa-commit
mailing list