#include "tgsi/tgsi_scan.h"
#include "util/u_inlines.h"
#include "util/u_queue.h"
+#include "util/simple_mtx.h"
#include "ac_binary.h"
#include "ac_llvm_build.h"
/* Shader IO unique indices are supported for TGSI_SEMANTIC_GENERIC with an
* index smaller than this.
*/
-#define SI_MAX_IO_GENERIC 43
+#define SI_MAX_IO_GENERIC 32
/* SGPR user data indices */
enum {
#define S_VS_STATE_LS_OUT_VERTEX_SIZE(x) (((unsigned)(x) & 0xFF) << 24)
#define C_VS_STATE_LS_OUT_VERTEX_SIZE 0x00FFFFFF
-/* Driver-specific system values. */
-enum {
- /* Values from set_tess_state. */
- TGSI_SEMANTIC_DEFAULT_TESSOUTER_SI = TGSI_SEMANTIC_COUNT,
- TGSI_SEMANTIC_DEFAULT_TESSINNER_SI,
-
- /* Up to 4 dwords in user SGPRs for compute shaders. */
- TGSI_SEMANTIC_CS_USER_DATA,
-};
-
enum {
/* Use a property enum that CS wouldn't use. */
TGSI_PROPERTY_CS_LOCAL_SIZE = TGSI_PROPERTY_FS_COORD_ORIGIN,
- /* The number of used user data dwords in the range [1, 4]. */
- TGSI_PROPERTY_CS_USER_DATA_DWORDS = TGSI_PROPERTY_FS_COORD_PIXEL_CENTER,
-
- /* Use a property enum that VS wouldn't use. */
- TGSI_PROPERTY_VS_BLIT_SGPRS = TGSI_PROPERTY_FS_COORD_ORIGIN,
-
/* These represent the number of SGPRs the shader uses. */
SI_VS_BLIT_SGPRS_POS = 3,
SI_VS_BLIT_SGPRS_POS_COLOR = 7,
struct util_queue_fence ready;
struct si_compiler_ctx_state compiler_ctx_state;
- mtx_t mutex;
+ simple_mtx_t mutex;
struct si_shader *first_variant; /* immutable after the first variant */
struct si_shader *last_variant; /* mutable */
struct si_shader *main_shader_part_ls; /* as_ls is set in the key */
struct si_shader *main_shader_part_es; /* as_es is set in the key */
struct si_shader *main_shader_part_ngg; /* as_ngg is set in the key */
+ struct si_shader *main_shader_part_ngg_es; /* for Wave32 TES before legacy GS */
struct si_shader *gs_copy_shader;
struct tgsi_token *tokens;
struct nir_shader *nir;
+ void *nir_binary;
+ unsigned nir_size;
+
struct pipe_stream_output_info so;
struct tgsi_shader_info info;
struct tgsi_tessctrl_info tcs_info;
/* PIPE_SHADER_[VERTEX|FRAGMENT|...] */
- unsigned type;
+ enum pipe_shader_type type;
bool vs_needs_prolog;
bool force_correct_derivs_after_kill;
bool prim_discard_cs_allowed;
unsigned pa_cl_vs_out_cntl;
ubyte clipdist_mask;
ubyte culldist_mask;
+ unsigned rast_prim;
/* ES parameters. */
unsigned esgs_itemsize; /* vertex stride */
unsigned gsvs_vertex_size;
unsigned max_gsvs_emit_size;
unsigned enabled_streamout_buffer_mask;
+ bool tess_turns_off_ngg;
/* PS parameters. */
unsigned color_attr_index[2];
struct si_gs_prolog_bits states;
/* Prologs of monolithic shaders shouldn't set EXEC. */
unsigned is_monolithic:1;
+ unsigned as_ngg:1;
} gs_prolog;
struct {
struct si_ps_prolog_bits states;
unsigned vs_export_prim_id:1;
struct {
unsigned interpolate_at_sample_force_center:1;
- unsigned fbfetch_msaa;
- unsigned fbfetch_is_1D;
- unsigned fbfetch_layered;
+ unsigned fbfetch_msaa:1;
+ unsigned fbfetch_is_1D:1;
+ unsigned fbfetch_layered:1;
} ps;
} u;
} mono;
unsigned vgt_gs_onchip_cntl;
unsigned vgt_gs_instance_cnt;
unsigned vgt_esgs_ring_itemsize;
- unsigned vgt_reuse_off;
unsigned spi_vs_out_config;
unsigned spi_shader_idx_format;
unsigned spi_shader_pos_format;
/*For save precompute registers value */
unsigned vgt_tf_param; /* VGT_TF_PARAM */
unsigned vgt_vertex_reuse_block_cntl; /* VGT_VERTEX_REUSE_BLOCK_CNTL */
+ unsigned pa_cl_vs_out_cntl;
unsigned ge_cntl;
};
bool si_shader_binary_upload(struct si_screen *sscreen, struct si_shader *shader,
uint64_t scratch_va);
void si_shader_dump(struct si_screen *sscreen, struct si_shader *shader,
- struct pipe_debug_callback *debug, unsigned processor,
+ struct pipe_debug_callback *debug,
FILE *f, bool check_debug_option);
void si_shader_dump_stats_for_shader_db(struct si_screen *screen,
struct si_shader *shader,
struct pipe_debug_callback *debug);
void si_multiwave_lds_size_workaround(struct si_screen *sscreen,
unsigned *lds_size);
-const char *si_get_shader_name(const struct si_shader *shader, unsigned processor);
+const char *si_get_shader_name(const struct si_shader *shader);
void si_shader_binary_clean(struct si_shader_binary *binary);
/* si_shader_nir.c */
struct tgsi_shader_info *info);
void si_nir_scan_tess_ctrl(const struct nir_shader *nir,
struct tgsi_tessctrl_info *out);
-void si_lower_nir(struct si_shader_selector *sel);
-void si_nir_opts(struct nir_shader *nir);
+void si_nir_adjust_driver_locations(struct nir_shader *nir);
+void si_finalize_nir(struct pipe_screen *screen, void *nirptr, bool optimize);
/* si_state_shaders.c */
void gfx9_get_gs_info(struct si_shader_selector *es,
{
if (key->as_ls)
return &sel->main_shader_part_ls;
+ if (key->as_es && key->as_ngg)
+ return &sel->main_shader_part_ngg_es;
if (key->as_es)
return &sel->main_shader_part_es;
if (key->as_ngg)