OpenHarmony/third_party_mesa3d
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

radv: cleanup helpers that compute NGG info and GS info on GFX9+

Browse Source 
Before moving them to the shader info link step.

Signed-off-by: Samuel Pitoiset <samuel.pitoiset@gmail.com>
Reviewed-by: Timur Kristóf <timur.kristof@gmail.com>
Part-of: <https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/18278>
This commit is contained in:
Samuel Pitoiset
2022-08-26 11:30:57 +02:00
committed by Marge Bot
parent 5cfd5c10eb
commit 19308db39d
1 changed files with 80 additions and 99 deletions
Download Patch File Download Diff File Expand all files Collapse all files

179
src/amd/vulkan/radv_pipeline.c
View File

@@ -1894,19 +1894,17 @@ radv_pipeline_init_depth_stencil_state(struct radv_graphics_pipeline *pipeline,
} }
static void static void
gfx9_get_gs_info(const struct radv_pipeline *pipeline, struct radv_pipeline_stage *stages, gfx9_get_gs_info(const struct radv_device *device, struct radv_pipeline_stage *es_stage,
struct gfx9_gs_info *out) struct radv_pipeline_stage *gs_stage)
{ {
const struct radv_physical_device *pdevice = pipeline->device->physical_device; const enum amd_gfx_level gfx_level = device->physical_device->rad_info.gfx_level;
struct radv_shader_info *gs_info = &stages[MESA_SHADER_GEOMETRY].info; struct radv_shader_info *gs_info = &gs_stage->info;
struct radv_shader_info *es_info; struct radv_shader_info *es_info = &es_stage->info;
bool has_tess = !!stages[MESA_SHADER_TESS_CTRL].nir; struct gfx9_gs_info *out = &gs_stage->info.gs_ring_info;
es_info = has_tess ? &stages[MESA_SHADER_TESS_EVAL].info : &stages[MESA_SHADER_VERTEX].info; const unsigned gs_num_invocations = MAX2(gs_info->gs.invocations, 1);
const bool uses_adjacency = gs_info->gs.input_prim == SHADER_PRIM_LINES_ADJACENCY ||
unsigned gs_num_invocations = MAX2(gs_info->gs.invocations, 1); gs_info->gs.input_prim == SHADER_PRIM_TRIANGLES_ADJACENCY;
bool uses_adjacency = gs_info->gs.input_prim == SHADER_PRIM_LINES_ADJACENCY ||
gs_info->gs.input_prim == SHADER_PRIM_TRIANGLES_ADJACENCY;
/* All these are in dwords: */ /* All these are in dwords: */
/* We can't allow using the whole LDS, because GS waves compete with /* We can't allow using the whole LDS, because GS waves compete with
@@ -1984,10 +1982,10 @@ gfx9_get_gs_info(const struct radv_pipeline *pipeline, struct radv_pipeline_stag
*/ */
es_verts -= min_es_verts - 1; es_verts -= min_es_verts - 1;
uint32_t es_verts_per_subgroup = es_verts; const uint32_t es_verts_per_subgroup = es_verts;
uint32_t gs_prims_per_subgroup = gs_prims; const uint32_t gs_prims_per_subgroup = gs_prims;
uint32_t gs_inst_prims_in_subgroup = gs_prims * gs_num_invocations; const uint32_t gs_inst_prims_in_subgroup = gs_prims * gs_num_invocations;
uint32_t max_prims_per_subgroup = gs_inst_prims_in_subgroup * gs_info->gs.vertices_out; const uint32_t max_prims_per_subgroup = gs_inst_prims_in_subgroup * gs_info->gs.vertices_out;
out->lds_size = align(esgs_lds_size, 128) / 128; out->lds_size = align(esgs_lds_size, 128) / 128;
out->vgt_gs_onchip_cntl = S_028A44_ES_VERTS_PER_SUBGRP(es_verts_per_subgroup) | out->vgt_gs_onchip_cntl = S_028A44_ES_VERTS_PER_SUBGRP(es_verts_per_subgroup) |
S_028A44_GS_PRIMS_PER_SUBGRP(gs_prims_per_subgroup) | S_028A44_GS_PRIMS_PER_SUBGRP(gs_prims_per_subgroup) |
@@ -1996,12 +1994,10 @@ gfx9_get_gs_info(const struct radv_pipeline *pipeline, struct radv_pipeline_stag
out->vgt_esgs_ring_itemsize = esgs_itemsize; out->vgt_esgs_ring_itemsize = esgs_itemsize;
assert(max_prims_per_subgroup <= max_out_prims); assert(max_prims_per_subgroup <= max_out_prims);
gl_shader_stage es_stage = has_tess ? MESA_SHADER_TESS_EVAL : MESA_SHADER_VERTEX; unsigned workgroup_size = ac_compute_esgs_workgroup_size(gfx_level, es_info->wave_size,
unsigned workgroup_size = ac_compute_esgs_workgroup_size(
pdevice->rad_info.gfx_level, stages[es_stage].info.wave_size,
es_verts_per_subgroup, gs_inst_prims_in_subgroup); es_verts_per_subgroup, gs_inst_prims_in_subgroup);
stages[es_stage].info.workgroup_size = workgroup_size; es_info->workgroup_size = workgroup_size;
stages[MESA_SHADER_GEOMETRY].info.workgroup_size = workgroup_size; gs_info->workgroup_size = workgroup_size;
} }
static void static void
@@ -2015,20 +2011,17 @@ clamp_gsprims_to_esverts(unsigned *max_gsprims, unsigned max_esverts, unsigned m
} }
static unsigned static unsigned
radv_get_num_input_vertices(const struct radv_pipeline_stage *stages) radv_get_num_input_vertices(const struct radv_pipeline_stage *es_stage,
const struct radv_pipeline_stage *gs_stage)
{ {
if (stages[MESA_SHADER_GEOMETRY].nir) { if (gs_stage) {
nir_shader *gs = stages[MESA_SHADER_GEOMETRY].nir; return gs_stage->nir->info.gs.vertices_in;
return gs->info.gs.vertices_in;
} }
if (stages[MESA_SHADER_TESS_CTRL].nir) { if (es_stage->stage == MESA_SHADER_TESS_EVAL) {
nir_shader *tes = stages[MESA_SHADER_TESS_EVAL].nir; if (es_stage->nir->info.tess.point_mode)
if (tes->info.tess.point_mode)
return 1; return 1;
if (tes->info.tess._primitive_mode == TESS_PRIMITIVE_ISOLINES) if (es_stage->nir->info.tess._primitive_mode == TESS_PRIMITIVE_ISOLINES)
return 2; return 2;
return 3; return 3;
} }
@@ -2046,20 +2039,17 @@ gfx10_emit_ge_pc_alloc(struct radeon_cmdbuf *cs, enum amd_gfx_level gfx_level,
} }
static unsigned static unsigned
radv_get_pre_rast_input_topology(struct radv_pipeline_stage *stages) radv_get_pre_rast_input_topology(const struct radv_pipeline_stage *es_stage,
const struct radv_pipeline_stage *gs_stage)
{ {
if (stages[MESA_SHADER_GEOMETRY].nir) { if (gs_stage) {
struct radv_shader_info *gs_info = &stages[MESA_SHADER_GEOMETRY].info; return gs_stage->nir->info.gs.input_primitive;
return gs_info->gs.input_prim;
} }
if (stages[MESA_SHADER_TESS_CTRL].nir) { if (es_stage->stage == MESA_SHADER_TESS_EVAL) {
struct radv_shader_info *tes_info = &stages[MESA_SHADER_TESS_EVAL].info; if (es_stage->nir->info.tess.point_mode)
if (tes_info->tes.point_mode)
return SHADER_PRIM_POINTS; return SHADER_PRIM_POINTS;
if (tes_info->tes._primitive_mode == TESS_PRIMITIVE_ISOLINES) if (es_stage->nir->info.tess._primitive_mode == TESS_PRIMITIVE_ISOLINES)
return SHADER_PRIM_LINES; return SHADER_PRIM_LINES;
return SHADER_PRIM_TRIANGLES; return SHADER_PRIM_TRIANGLES;
} }
@@ -2068,20 +2058,19 @@ radv_get_pre_rast_input_topology(struct radv_pipeline_stage *stages)
} }
static void static void
gfx10_get_ngg_info(struct radv_pipeline *pipeline, struct radv_pipeline_stage *stages, gfx10_get_ngg_info(const struct radv_device *device, struct radv_pipeline_stage *es_stage,
struct gfx10_ngg_info *ngg) struct radv_pipeline_stage *gs_stage)
{ {
const struct radv_physical_device *pdevice = pipeline->device->physical_device; const enum amd_gfx_level gfx_level = device->physical_device->rad_info.gfx_level;
struct radv_shader_info *gs_info = &stages[MESA_SHADER_GEOMETRY].info; struct radv_shader_info *gs_info = gs_stage ? &gs_stage->info : NULL;
struct radv_shader_info *es_info = struct radv_shader_info *es_info = &es_stage->info;
stages[MESA_SHADER_TESS_CTRL].nir ? &stages[MESA_SHADER_TESS_EVAL].info const unsigned max_verts_per_prim = radv_get_num_input_vertices(es_stage, gs_stage);
: &stages[MESA_SHADER_VERTEX].info; const unsigned min_verts_per_prim = gs_stage ? max_verts_per_prim : 1;
unsigned gs_type = stages[MESA_SHADER_GEOMETRY].nir ? MESA_SHADER_GEOMETRY : MESA_SHADER_VERTEX; struct gfx10_ngg_info *out = gs_stage ? &gs_info->ngg_info : &es_info->ngg_info;
unsigned max_verts_per_prim = radv_get_num_input_vertices(stages);
unsigned min_verts_per_prim = gs_type == MESA_SHADER_GEOMETRY ? max_verts_per_prim : 1;
unsigned gs_num_invocations = stages[MESA_SHADER_GEOMETRY].nir ? MAX2(gs_info->gs.invocations, 1) : 1;
const unsigned input_prim = radv_get_pre_rast_input_topology(stages); const unsigned gs_num_invocations = gs_stage ? MAX2(gs_info->gs.invocations, 1) : 1;
const unsigned input_prim = radv_get_pre_rast_input_topology(es_stage, gs_stage);
const bool uses_adjacency = input_prim == SHADER_PRIM_LINES_ADJACENCY || const bool uses_adjacency = input_prim == SHADER_PRIM_LINES_ADJACENCY ||
input_prim == SHADER_PRIM_TRIANGLES_ADJACENCY; input_prim == SHADER_PRIM_TRIANGLES_ADJACENCY;
@@ -2099,7 +2088,7 @@ gfx10_get_ngg_info(struct radv_pipeline *pipeline, struct radv_pipeline_stage *s
unsigned gsprim_lds_size = 0; unsigned gsprim_lds_size = 0;
/* All these are per subgroup: */ /* All these are per subgroup: */
const unsigned min_esverts = pdevice->rad_info.gfx_level >= GFX10_3 ? 29 : 24; const unsigned min_esverts = gfx_level >= GFX10_3 ? 29 : 24;
bool max_vert_out_per_gs_instance = false; bool max_vert_out_per_gs_instance = false;
unsigned max_esverts_base = 128; unsigned max_esverts_base = 128;
unsigned max_gsprims_base = 128; /* default prim group size clamp */ unsigned max_gsprims_base = 128; /* default prim group size clamp */
@@ -2114,7 +2103,7 @@ gfx10_get_ngg_info(struct radv_pipeline *pipeline, struct radv_pipeline_stage *s
*/ */
max_esverts_base = MIN2(max_esverts_base, 251 + max_verts_per_prim - 1); max_esverts_base = MIN2(max_esverts_base, 251 + max_verts_per_prim - 1);
if (gs_type == MESA_SHADER_GEOMETRY) { if (gs_stage) {
unsigned max_out_verts_per_gsprim = gs_info->gs.vertices_out * gs_num_invocations; unsigned max_out_verts_per_gsprim = gs_info->gs.vertices_out * gs_num_invocations;
if (max_out_verts_per_gsprim <= 256) { if (max_out_verts_per_gsprim <= 256) {
@@ -2135,9 +2124,7 @@ gfx10_get_ngg_info(struct radv_pipeline *pipeline, struct radv_pipeline_stage *s
} else { } else {
/* VS and TES. */ /* VS and TES. */
/* LDS size for passing data from GS to ES. */ /* LDS size for passing data from GS to ES. */
struct radv_streamout_info *so_info = stages[MESA_SHADER_TESS_CTRL].nir struct radv_streamout_info *so_info = &es_info->so;
? &stages[MESA_SHADER_TESS_EVAL].info.so
: &stages[MESA_SHADER_VERTEX].info.so;
if (so_info->num_outputs) if (so_info->num_outputs)
esvert_lds_size = 4 * so_info->num_outputs + 1; esvert_lds_size = 4 * so_info->num_outputs + 1;
@@ -2146,7 +2133,7 @@ gfx10_get_ngg_info(struct radv_pipeline *pipeline, struct radv_pipeline_stage *s
* corresponding to the ES thread of the provoking vertex. All * corresponding to the ES thread of the provoking vertex. All
* ES threads load and export PrimitiveID for their thread. * ES threads load and export PrimitiveID for their thread.
*/ */
if (!stages[MESA_SHADER_TESS_CTRL].nir && stages[MESA_SHADER_VERTEX].info.outinfo.export_prim_id) if (es_stage->stage == MESA_SHADER_VERTEX && es_stage->info.outinfo.export_prim_id)
esvert_lds_size = MAX2(esvert_lds_size, 1); esvert_lds_size = MAX2(esvert_lds_size, 1);
} }
@@ -2187,11 +2174,10 @@ gfx10_get_ngg_info(struct radv_pipeline *pipeline, struct radv_pipeline_stage *s
unsigned orig_max_gsprims; unsigned orig_max_gsprims;
unsigned wavesize; unsigned wavesize;
if (gs_type == MESA_SHADER_GEOMETRY) { if (gs_stage) {
wavesize = gs_info->wave_size; wavesize = gs_info->wave_size;
} else { } else {
wavesize = stages[MESA_SHADER_TESS_CTRL].nir ? stages[MESA_SHADER_TESS_EVAL].info.wave_size wavesize = es_info->wave_size;
: stages[MESA_SHADER_VERTEX].info.wave_size;
} }
do { do {
@@ -2206,7 +2192,7 @@ gfx10_get_ngg_info(struct radv_pipeline *pipeline, struct radv_pipeline_stage *s
max_esverts = MIN2(max_esverts, max_gsprims * max_verts_per_prim); max_esverts = MIN2(max_esverts, max_gsprims * max_verts_per_prim);
/* Hardware restriction: minimum value of max_esverts */ /* Hardware restriction: minimum value of max_esverts */
if (pdevice->rad_info.gfx_level == GFX10) if (gfx_level == GFX10)
max_esverts = MAX2(max_esverts, min_esverts - 1 + max_verts_per_prim); max_esverts = MAX2(max_esverts, min_esverts - 1 + max_verts_per_prim);
else else
max_esverts = MAX2(max_esverts, min_esverts); max_esverts = MAX2(max_esverts, min_esverts);
@@ -2229,26 +2215,26 @@ gfx10_get_ngg_info(struct radv_pipeline *pipeline, struct radv_pipeline_stage *s
} while (orig_max_esverts != max_esverts || orig_max_gsprims != max_gsprims); } while (orig_max_esverts != max_esverts || orig_max_gsprims != max_gsprims);
/* Verify the restriction. */ /* Verify the restriction. */
if (pdevice->rad_info.gfx_level == GFX10) if (gfx_level == GFX10)
assert(max_esverts >= min_esverts - 1 + max_verts_per_prim); assert(max_esverts >= min_esverts - 1 + max_verts_per_prim);
else else
assert(max_esverts >= min_esverts); assert(max_esverts >= min_esverts);
} else { } else {
/* Hardware restriction: minimum value of max_esverts */ /* Hardware restriction: minimum value of max_esverts */
if (pdevice->rad_info.gfx_level == GFX10) if (gfx_level == GFX10)
max_esverts = MAX2(max_esverts, min_esverts - 1 + max_verts_per_prim); max_esverts = MAX2(max_esverts, min_esverts - 1 + max_verts_per_prim);
else else
max_esverts = MAX2(max_esverts, min_esverts); max_esverts = MAX2(max_esverts, min_esverts);
} }
unsigned max_out_vertices = max_vert_out_per_gs_instance ? gs_info->gs.vertices_out unsigned max_out_vertices = max_vert_out_per_gs_instance ? gs_info->gs.vertices_out
: gs_type == MESA_SHADER_GEOMETRY : gs_stage
? max_gsprims * gs_num_invocations * gs_info->gs.vertices_out ? max_gsprims * gs_num_invocations * gs_info->gs.vertices_out
: max_esverts; : max_esverts;
assert(max_out_vertices <= 256); assert(max_out_vertices <= 256);
unsigned prim_amp_factor = 1; unsigned prim_amp_factor = 1;
if (gs_type == MESA_SHADER_GEOMETRY) { if (gs_stage) {
/* Number of output primitives per GS input primitive after /* Number of output primitives per GS input primitive after
* GS instancing. */ * GS instancing. */
prim_amp_factor = gs_info->gs.vertices_out; prim_amp_factor = gs_info->gs.vertices_out;
@@ -2259,35 +2245,36 @@ gfx10_get_ngg_info(struct radv_pipeline *pipeline, struct radv_pipeline_stage *s
* whenever this check passes, there is enough space for a full * whenever this check passes, there is enough space for a full
* primitive without vertex reuse. * primitive without vertex reuse.
*/ */
if (pdevice->rad_info.gfx_level == GFX10) if (gfx_level == GFX10)
ngg->hw_max_esverts = max_esverts - max_verts_per_prim + 1; out->hw_max_esverts = max_esverts - max_verts_per_prim + 1;
else else
ngg->hw_max_esverts = max_esverts; out->hw_max_esverts = max_esverts;
ngg->max_gsprims = max_gsprims; out->max_gsprims = max_gsprims;
ngg->max_out_verts = max_out_vertices; out->max_out_verts = max_out_vertices;
ngg->prim_amp_factor = prim_amp_factor; out->prim_amp_factor = prim_amp_factor;
ngg->max_vert_out_per_gs_instance = max_vert_out_per_gs_instance; out->max_vert_out_per_gs_instance = max_vert_out_per_gs_instance;
ngg->ngg_emit_size = max_gsprims * gsprim_lds_size; out->ngg_emit_size = max_gsprims * gsprim_lds_size;
ngg->enable_vertex_grouping = true; out->enable_vertex_grouping = true;
/* Don't count unusable vertices. */ /* Don't count unusable vertices. */
ngg->esgs_ring_size = MIN2(max_esverts, max_gsprims * max_verts_per_prim) * esvert_lds_size * 4; out->esgs_ring_size = MIN2(max_esverts, max_gsprims * max_verts_per_prim) * esvert_lds_size * 4;
if (gs_type == MESA_SHADER_GEOMETRY) { if (gs_stage) {
ngg->vgt_esgs_ring_itemsize = es_info->esgs_itemsize / 4; out->vgt_esgs_ring_itemsize = es_info->esgs_itemsize / 4;
} else { } else {
ngg->vgt_esgs_ring_itemsize = 1; out->vgt_esgs_ring_itemsize = 1;
} }
assert(ngg->hw_max_esverts >= min_esverts); /* HW limitation */ assert(out->hw_max_esverts >= min_esverts); /* HW limitation */
gl_shader_stage es_stage = stages[MESA_SHADER_TESS_CTRL].nir ? MESA_SHADER_TESS_EVAL : MESA_SHADER_VERTEX;
unsigned workgroup_size = unsigned workgroup_size =
ac_compute_ngg_workgroup_size( ac_compute_ngg_workgroup_size(
max_esverts, max_gsprims * gs_num_invocations, max_out_vertices, prim_amp_factor); max_esverts, max_gsprims * gs_num_invocations, max_out_vertices, prim_amp_factor);
stages[MESA_SHADER_GEOMETRY].info.workgroup_size = workgroup_size; if (gs_stage) {
stages[es_stage].info.workgroup_size = workgroup_size; gs_info->workgroup_size = workgroup_size;
}
es_info->workgroup_size = workgroup_size;
} }
static void static void
@@ -3405,25 +3392,19 @@ radv_fill_shader_info(struct radv_pipeline *pipeline,
} }
if (pipeline_has_ngg) { if (pipeline_has_ngg) {
struct gfx10_ngg_info *ngg_info; if (last_vgt_api_stage != MESA_SHADER_MESH) {
struct radv_pipeline_stage *es_stage =
stages[MESA_SHADER_TESS_EVAL].nir ? &stages[MESA_SHADER_TESS_EVAL] : &stages[MESA_SHADER_VERTEX];
struct radv_pipeline_stage *gs_stage =
stages[MESA_SHADER_GEOMETRY].nir ? &stages[MESA_SHADER_GEOMETRY] : NULL;
if (stages[MESA_SHADER_GEOMETRY].nir) gfx10_get_ngg_info(device, es_stage, gs_stage);
ngg_info = &stages[MESA_SHADER_GEOMETRY].info.ngg_info; }
else if (stages[MESA_SHADER_TESS_CTRL].nir)
ngg_info = &stages[MESA_SHADER_TESS_EVAL].info.ngg_info;
else if (stages[MESA_SHADER_VERTEX].nir)
ngg_info = &stages[MESA_SHADER_VERTEX].info.ngg_info;
else if (stages[MESA_SHADER_MESH].nir)
ngg_info = &stages[MESA_SHADER_MESH].info.ngg_info;
else
unreachable("Missing NGG shader stage.");
if (last_vgt_api_stage != MESA_SHADER_MESH)
gfx10_get_ngg_info(pipeline, stages, ngg_info);
} else if (stages[MESA_SHADER_GEOMETRY].nir) { } else if (stages[MESA_SHADER_GEOMETRY].nir) {
struct gfx9_gs_info *gs_info = &stages[MESA_SHADER_GEOMETRY].info.gs_ring_info; struct radv_pipeline_stage *es_stage =
stages[MESA_SHADER_TESS_EVAL].nir ? &stages[MESA_SHADER_TESS_EVAL] : &stages[MESA_SHADER_VERTEX];
gfx9_get_gs_info(pipeline, stages, gs_info); gfx9_get_gs_info(device, es_stage, &stages[MESA_SHADER_GEOMETRY]);
} else { } else {
gl_shader_stage hw_vs_api_stage = gl_shader_stage hw_vs_api_stage =
stages[MESA_SHADER_TESS_EVAL].nir ? MESA_SHADER_TESS_EVAL : MESA_SHADER_VERTEX; stages[MESA_SHADER_TESS_EVAL].nir ? MESA_SHADER_TESS_EVAL : MESA_SHADER_VERTEX;