third_party
/
virglrenderer
1
0
Fork 0
Code Issues Pull Requests Projects Releases Wiki Activity

shader: move output interface info into struct

Browse Source 
Compress the structure so that it fits into a 64 bit value.

Signed-off-by: Gert Wollny <gert.wollny@collabora.com>
Reviewed-by: Rohan Garg <rohan.garg@collabora.com>
macos/master
Gert Wollny 4 years ago
parent 729a819274
commit 1cc53f715c
3 changed files with 32 additions and 35 deletions
Show all changes Ignore whitespace when comparing lines Ignore changes in amount of whitespace Ignore changes in whitespace at EOL
Show Stats Download Patch File Download Diff File
  1. 10
      src/vrend_renderer.c
  2. 34
      src/vrend_shader.c
  3. 23
      src/vrend_shader.h

10
src/vrend_renderer.c
Unescape View File

@ -3400,7 +3400,7 @@ static inline void vrend_fill_shader_key(struct vrend_sub_context *sub_ctx,
break; break;
} }
key->fs_prim_is_points = (fs_prim_mode == PIPE_PRIM_POINTS); key->fs_prim_is_points = (fs_prim_mode == PIPE_PRIM_POINTS);
key->coord_replace = sub_ctx->rs_state.point_quad_rasterization key->fs_coord_replace = sub_ctx->rs_state.point_quad_rasterization
&& key->fs_prim_is_points && key->fs_prim_is_points
? sub_ctx->rs_state.sprite_coord_enable ? sub_ctx->rs_state.sprite_coord_enable
: 0x0; : 0x0;
@ -3437,12 +3437,8 @@ static inline void vrend_fill_shader_key(struct vrend_sub_context *sub_ctx,
break; break;
} }
if (next_type != -1 && sub_ctx->shaders[next_type]) { if (next_type != -1 && sub_ctx->shaders[next_type])
key->next_stage_pervertex_in = sub_ctx->shaders[next_type]->sinfo.has_pervertex_in; key->output = sub_ctx->shaders[next_type]->sinfo.in;
key->num_indirect_generic_outputs = sub_ctx->shaders[next_type]->sinfo.num_indirect_generic_inputs;
key->num_indirect_patch_outputs = sub_ctx->shaders[next_type]->sinfo.num_indirect_patch_inputs;
key->generic_outputs_expected_mask = sub_ctx->shaders[next_type]->sinfo.generic_inputs_emitted_mask;
}
} }
static int vrend_shader_create(struct vrend_context *ctx, static int vrend_shader_create(struct vrend_context *ctx,

34
src/vrend_shader.c
Unescape View File

@ -1171,7 +1171,7 @@ iter_declaration(struct tgsi_iterate_context *iter,
case TGSI_SEMANTIC_PATCH: case TGSI_SEMANTIC_PATCH:
case TGSI_SEMANTIC_GENERIC: case TGSI_SEMANTIC_GENERIC:
if (iter->processor.Processor == TGSI_PROCESSOR_FRAGMENT) { if (iter->processor.Processor == TGSI_PROCESSOR_FRAGMENT) {
if (ctx->key->coord_replace & (1 << ctx->inputs[i].sid)) { if (ctx->key->fs_coord_replace & (1 << ctx->inputs[i].sid)) {
if (ctx->cfg->use_gles) if (ctx->cfg->use_gles)
name_prefix = "vec4(gl_PointCoord.x, mix(1.0 - gl_PointCoord.y, gl_PointCoord.y, clamp(winsys_adjust_y, 0.0, 1.0)), 0.0, 1.0)"; name_prefix = "vec4(gl_PointCoord.x, mix(1.0 - gl_PointCoord.y, gl_PointCoord.y, clamp(winsys_adjust_y, 0.0, 1.0)), 0.0, 1.0)";
else else
@ -4425,8 +4425,8 @@ void rewrite_io_ranged(struct dump_ctx *ctx)
} }
if ((ctx->info.indirect_files & (1 << TGSI_FILE_OUTPUT)) || if ((ctx->info.indirect_files & (1 << TGSI_FILE_OUTPUT)) ||
ctx->key->num_indirect_generic_outputs || ctx->key->output.num_indirect_generic ||
ctx->key->num_indirect_patch_outputs) { ctx->key->output.num_indirect_patch) {
for (uint i = 0; i < ctx->num_outputs; ++i) { for (uint i = 0; i < ctx->num_outputs; ++i) {
if (ctx->outputs[i].name == TGSI_SEMANTIC_PATCH) { if (ctx->outputs[i].name == TGSI_SEMANTIC_PATCH) {
@ -4763,7 +4763,7 @@ static void handle_io_arrays(struct dump_ctx *ctx)
if (ctx->num_inputs > 0) if (ctx->num_inputs > 0)
if (evaluate_layout_overlays(ctx->num_inputs, ctx->inputs, if (evaluate_layout_overlays(ctx->num_inputs, ctx->inputs,
get_stage_input_name_prefix(ctx, ctx->prog_type), get_stage_input_name_prefix(ctx, ctx->prog_type),
ctx->key->coord_replace)) { ctx->key->fs_coord_replace)) {
require_enhanced_layouts = true; require_enhanced_layouts = true;
} }
@ -4779,7 +4779,7 @@ static void handle_io_arrays(struct dump_ctx *ctx)
rewrite_io_ranged(ctx); rewrite_io_ranged(ctx);
rewrite_components(ctx->num_inputs, ctx->inputs, rewrite_components(ctx->num_inputs, ctx->inputs,
get_stage_input_name_prefix(ctx, ctx->prog_type), get_stage_input_name_prefix(ctx, ctx->prog_type),
ctx->key->coord_replace, true); ctx->key->fs_coord_replace, true);
rewrite_components(ctx->num_outputs, ctx->outputs, rewrite_components(ctx->num_outputs, ctx->outputs,
get_stage_output_name_prefix(ctx->prog_type), 0, true); get_stage_output_name_prefix(ctx->prog_type), 0, true);
@ -6084,11 +6084,13 @@ emit_ios_generic(const struct dump_ctx *ctx,
postfix); postfix);
if (io->name == TGSI_SEMANTIC_GENERIC) { if (io->name == TGSI_SEMANTIC_GENERIC) {
if (iot == io_in) assert(io->sid < 32);
if (iot == io_in) {
generic_ios->inputs_emitted_mask |= 1 << io->sid; generic_ios->inputs_emitted_mask |= 1 << io->sid;
else } else {
generic_ios->outputs_emitted_mask |= 1 << io->sid; generic_ios->outputs_emitted_mask |= 1 << io->sid;
} }
}
} else { } else {
if (prefer_generic_io_block(ctx, iot)) { if (prefer_generic_io_block(ctx, iot)) {
@ -6293,7 +6295,7 @@ static void emit_ios_vs(const struct dump_ctx *ctx,
if (ctx->key->clip_plane_enable) { if (ctx->key->clip_plane_enable) {
emit_hdr(glsl_strbufs, "uniform vec4 clipp[8];\n"); emit_hdr(glsl_strbufs, "uniform vec4 clipp[8];\n");
} }
if ((ctx->key->gs_present || ctx->key->tes_present) && ctx->key->next_stage_pervertex_in) { if ((ctx->key->gs_present || ctx->key->tes_present) && ctx->key->output.use_pervertex) {
emit_hdrf(glsl_strbufs, "out gl_PerVertex {\n vec4 gl_Position;\n %s%s};\n", clip_buf, cull_buf); emit_hdrf(glsl_strbufs, "out gl_PerVertex {\n vec4 gl_Position;\n %s%s};\n", clip_buf, cull_buf);
} else { } else {
emit_hdrf(glsl_strbufs, "%s%s", clip_buf, cull_buf); emit_hdrf(glsl_strbufs, "%s%s", clip_buf, cull_buf);
@ -6362,7 +6364,7 @@ static void emit_ios_fs(const struct dump_ctx *ctx,
} }
if (ctx->cfg->use_gles && !ctx->winsys_adjust_y_emitted && if (ctx->cfg->use_gles && !ctx->winsys_adjust_y_emitted &&
(ctx->key->coord_replace & (1 << ctx->inputs[i].sid))) { (ctx->key->fs_coord_replace & (1 << ctx->inputs[i].sid))) {
*winsys_adjust_y_emitted = true; *winsys_adjust_y_emitted = true;
emit_hdr(glsl_strbufs, "uniform float winsys_adjust_y;\n"); emit_hdr(glsl_strbufs, "uniform float winsys_adjust_y;\n");
} }
@ -6603,7 +6605,7 @@ static void emit_ios_tcs(const struct dump_ctx *ctx,
*has_pervertex = true; *has_pervertex = true;
emit_hdrf(glsl_strbufs, "in gl_PerVertex {\n vec4 gl_Position; \n %s%s} gl_in[];\n", clip_var, cull_var); emit_hdrf(glsl_strbufs, "in gl_PerVertex {\n vec4 gl_Position; \n %s%s} gl_in[];\n", clip_var, cull_var);
} }
if (ctx->num_clip_dist && ctx->key->next_stage_pervertex_in) { if (ctx->num_clip_dist && ctx->key->output.use_pervertex) {
emit_hdrf(glsl_strbufs, "out gl_PerVertex {\n vec4 gl_Position;\n float gl_ClipDistance[%d];\n} gl_out[];\n", ctx->num_clip_dist); emit_hdrf(glsl_strbufs, "out gl_PerVertex {\n vec4 gl_Position;\n float gl_ClipDistance[%d];\n} gl_out[];\n", ctx->num_clip_dist);
emit_hdr(glsl_strbufs, "vec4 clip_dist_temp[2];\n"); emit_hdr(glsl_strbufs, "vec4 clip_dist_temp[2];\n");
} }
@ -6662,7 +6664,7 @@ static void emit_ios_tes(const struct dump_ctx *ctx,
*has_pervertex = true; *has_pervertex = true;
emit_hdrf(glsl_strbufs, "in gl_PerVertex {\n vec4 gl_Position; \n %s%s} gl_in[];\n", clip_var, cull_var); emit_hdrf(glsl_strbufs, "in gl_PerVertex {\n vec4 gl_Position; \n %s%s} gl_in[];\n", clip_var, cull_var);
} }
if (ctx->num_clip_dist && ctx->key->next_stage_pervertex_in) { if (ctx->num_clip_dist && ctx->key->output.use_pervertex) {
emit_hdrf(glsl_strbufs, "out gl_PerVertex {\n vec4 gl_Position;\n float gl_ClipDistance[%d];\n} gl_out[];\n", ctx->num_clip_dist); emit_hdrf(glsl_strbufs, "out gl_PerVertex {\n vec4 gl_Position;\n float gl_ClipDistance[%d];\n} gl_out[];\n", ctx->num_clip_dist);
emit_hdr(glsl_strbufs, "vec4 clip_dist_temp[2];\n"); emit_hdr(glsl_strbufs, "vec4 clip_dist_temp[2];\n");
} }
@ -6818,7 +6820,7 @@ static boolean analyze_instruction(struct tgsi_iterate_context *iter,
static void fill_sinfo(const struct dump_ctx *ctx, struct vrend_shader_info *sinfo) static void fill_sinfo(const struct dump_ctx *ctx, struct vrend_shader_info *sinfo)
{ {
sinfo->num_ucp = ctx->key->clip_plane_enable ? 8 : 0; sinfo->num_ucp = ctx->key->clip_plane_enable ? 8 : 0;
sinfo->has_pervertex_in = ctx->has_pervertex; sinfo->in.use_pervertex = ctx->has_pervertex;
sinfo->fs_info.has_sample_input = ctx->has_sample_input; sinfo->fs_info.has_sample_input = ctx->has_sample_input;
bool has_prop = (ctx->num_clip_dist_prop + ctx->num_cull_dist_prop) > 0; bool has_prop = (ctx->num_clip_dist_prop + ctx->num_cull_dist_prop) > 0;
sinfo->out.num_clip = has_prop ? ctx->num_clip_dist_prop : (ctx->num_clip_dist ? ctx->num_clip_dist : 8); sinfo->out.num_clip = has_prop ? ctx->num_clip_dist_prop : (ctx->num_clip_dist ? ctx->num_clip_dist : 8);
@ -6833,9 +6835,9 @@ static void fill_sinfo(const struct dump_ctx *ctx, struct vrend_shader_info *sin
sinfo->ubo_indirect = !!(ctx->info.dimension_indirect_files & (1 << TGSI_FILE_CONSTANT)); sinfo->ubo_indirect = !!(ctx->info.dimension_indirect_files & (1 << TGSI_FILE_CONSTANT));
if (ctx->generic_ios.input_range.used) if (ctx->generic_ios.input_range.used)
sinfo->num_indirect_generic_inputs = ctx->generic_ios.input_range.io.last - ctx->generic_ios.input_range.io.sid + 1; sinfo->in.num_indirect_generic = ctx->generic_ios.input_range.io.last - ctx->generic_ios.input_range.io.sid + 1;
if (ctx->patch_ios.input_range.used) if (ctx->patch_ios.input_range.used)
sinfo->num_indirect_patch_inputs = ctx->patch_ios.input_range.io.last - ctx->patch_ios.input_range.io.sid + 1; sinfo->in.num_indirect_patch = ctx->patch_ios.input_range.io.last - ctx->patch_ios.input_range.io.sid + 1;
if (ctx->generic_ios.output_range.used) if (ctx->generic_ios.output_range.used)
sinfo->out.num_indirect_generic = ctx->generic_ios.output_range.io.last - ctx->generic_ios.output_range.io.sid + 1; sinfo->out.num_indirect_generic = ctx->generic_ios.output_range.io.last - ctx->generic_ios.output_range.io.sid + 1;
@ -6887,7 +6889,7 @@ static void fill_sinfo(const struct dump_ctx *ctx, struct vrend_shader_info *sin
free(sinfo->image_arrays); free(sinfo->image_arrays);
sinfo->image_arrays = ctx->image_arrays; sinfo->image_arrays = ctx->image_arrays;
sinfo->num_image_arrays = ctx->num_image_arrays; sinfo->num_image_arrays = ctx->num_image_arrays;
sinfo->generic_inputs_emitted_mask = ctx->generic_ios.inputs_emitted_mask; sinfo->in.generic_emitted_mask = ctx->generic_ios.inputs_emitted_mask;
for (unsigned i = 0; i < ctx->num_outputs; ++i) { for (unsigned i = 0; i < ctx->num_outputs; ++i) {
if (ctx->outputs[i].invariant) if (ctx->outputs[i].invariant)
@ -6972,7 +6974,7 @@ bool vrend_convert_shader(const struct vrend_context *rctx,
ctx.has_sample_input = false; ctx.has_sample_input = false;
ctx.req_local_mem = req_local_mem; ctx.req_local_mem = req_local_mem;
ctx.guest_sent_io_arrays = key->input.guest_sent_io_arrays; ctx.guest_sent_io_arrays = key->input.guest_sent_io_arrays;
ctx.generic_ios.outputs_expected_mask = key->generic_outputs_expected_mask; ctx.generic_ios.outputs_expected_mask = key->output.generic_emitted_mask;
tgsi_scan_shader(tokens, &ctx.info); tgsi_scan_shader(tokens, &ctx.info);
/* if we are in core profile mode we should use GLSL 1.40 */ /* if we are in core profile mode we should use GLSL 1.40 */

23
src/vrend_shader.h
Unescape View File

@ -89,9 +89,18 @@ struct vrend_shader_info_out {
uint64_t guest_sent_io_arrays : 1; uint64_t guest_sent_io_arrays : 1;
}; };
struct vrend_shader_info_in {
uint64_t generic_emitted_mask : 32;
uint64_t num_indirect_generic : 8;
uint64_t num_indirect_patch : 8;
uint64_t use_pervertex : 1;
};
struct vrend_shader_info { struct vrend_shader_info {
uint64_t invariant_outputs; uint64_t invariant_outputs;
struct vrend_shader_info_out out; struct vrend_shader_info_out out;
struct vrend_shader_info_in in;
struct vrend_layout_info generic_outputs_layout[64]; struct vrend_layout_info generic_outputs_layout[64];
struct vrend_array *sampler_arrays; struct vrend_array *sampler_arrays;
@ -104,7 +113,6 @@ struct vrend_shader_info {
uint32_t images_used_mask; uint32_t images_used_mask;
uint32_t ubo_used_mask; uint32_t ubo_used_mask;
uint32_t ssbo_used_mask; uint32_t ssbo_used_mask;
uint32_t generic_inputs_emitted_mask;
uint32_t shadow_samp_mask; uint32_t shadow_samp_mask;
uint32_t attrib_input_mask; uint32_t attrib_input_mask;
uint32_t fs_blend_equation_advanced; uint32_t fs_blend_equation_advanced;
@ -118,12 +126,6 @@ struct vrend_shader_info {
int num_sampler_arrays; int num_sampler_arrays;
int num_image_arrays; int num_image_arrays;
uint8_t num_indirect_generic_inputs;
uint8_t num_indirect_patch_inputs;
uint8_t has_pervertex_in : 1;
uint8_t ubo_indirect : 1; uint8_t ubo_indirect : 1;
uint8_t tes_point_mode : 1; uint8_t tes_point_mode : 1;
}; };
@ -133,14 +135,14 @@ struct vrend_shader_key {
struct vrend_fs_shader_info *fs_info; struct vrend_fs_shader_info *fs_info;
struct vrend_shader_info_out input; struct vrend_shader_info_out input;
struct vrend_shader_info_in output;
uint32_t coord_replace; uint32_t fs_coord_replace;
uint32_t cbufs_are_a8_bitmask; uint32_t cbufs_are_a8_bitmask;
uint32_t cbufs_signed_int_bitmask; uint32_t cbufs_signed_int_bitmask;
uint32_t cbufs_unsigned_int_bitmask; uint32_t cbufs_unsigned_int_bitmask;
uint32_t attrib_signed_int_bitmask; uint32_t attrib_signed_int_bitmask;
uint32_t attrib_unsigned_int_bitmask; uint32_t attrib_unsigned_int_bitmask;
uint32_t generic_outputs_expected_mask;
uint32_t compiled_fs_uid; uint32_t compiled_fs_uid;
uint32_t fs_prim_is_points : 1; uint32_t fs_prim_is_points : 1;
@ -153,13 +155,10 @@ struct vrend_shader_key {
uint32_t tes_present : 1; uint32_t tes_present : 1;
uint32_t flatshade : 1; uint32_t flatshade : 1;
uint32_t fs_logicop_enabled : 1; uint32_t fs_logicop_enabled : 1;
uint32_t next_stage_pervertex_in : 1;
uint32_t fs_logicop_func : 4; uint32_t fs_logicop_func : 4;
uint8_t alpha_test; uint8_t alpha_test;
uint8_t clip_plane_enable; uint8_t clip_plane_enable;
uint8_t num_indirect_generic_outputs;
uint8_t num_indirect_patch_outputs;
uint8_t fs_swizzle_output_rgb_to_bgr; uint8_t fs_swizzle_output_rgb_to_bgr;
uint8_t surface_component_bits[PIPE_MAX_COLOR_BUFS]; uint8_t surface_component_bits[PIPE_MAX_COLOR_BUFS];
struct vrend_layout_info prev_stage_generic_and_patch_outputs_layout[64]; struct vrend_layout_info prev_stage_generic_and_patch_outputs_layout[64];

Write Preview
Loading…
Cancel
Save