#define SI_RESOURCE_FLAG_CLEAR (PIPE_RESOURCE_FLAG_DRV_PRIV << 7)
/* For const_uploader, upload data via GTT and copy to VRAM on context flush via SDMA. */
#define SI_RESOURCE_FLAG_UPLOAD_FLUSH_EXPLICIT_VIA_SDMA (PIPE_RESOURCE_FLAG_DRV_PRIV << 8)
+/* Set a micro tile mode: */
+#define SI_RESOURCE_FLAG_FORCE_MICRO_TILE_MODE (PIPE_RESOURCE_FLAG_DRV_PRIV << 9)
+#define SI_RESOURCE_FLAG_MICRO_TILE_MODE_SHIFT (util_logbase2(PIPE_RESOURCE_FLAG_DRV_PRIV) + 10)
+#define SI_RESOURCE_FLAG_MICRO_TILE_MODE_SET(x) (((x) & 0x3) << SI_RESOURCE_FLAG_MICRO_TILE_MODE_SHIFT)
+#define SI_RESOURCE_FLAG_MICRO_TILE_MODE_GET(x) (((x) >> SI_RESOURCE_FLAG_MICRO_TILE_MODE_SHIFT) & 0x3)
enum si_clear_code
{
DBG_TES = PIPE_SHADER_TESS_EVAL,
DBG_CS = PIPE_SHADER_COMPUTE,
DBG_NO_IR,
- DBG_NO_TGSI,
+ DBG_NO_NIR,
DBG_NO_ASM,
DBG_PREOPT_IR,
uint16_t stencil_dirty_level_mask; /* each bit says if that mipmap is compressed */
enum pipe_format db_render_format:16;
uint8_t stencil_clear_value;
- bool fmask_is_not_identity:1;
+ bool fmask_is_identity:1;
bool tc_compatible_htile:1;
bool htile_stencil_disabled:1;
bool depth_cleared:1; /* if it was cleared at least once */
uint32_t *state,
uint32_t *fmask_state);
+ unsigned num_vbos_in_user_sgprs;
unsigned pa_sc_raster_config;
unsigned pa_sc_raster_config_1;
unsigned se_tile_repeat;
*
* Design & limitations:
* - The shader cache is per screen (= per process), never saved to
- * disk, and skips redundant shader compilations from TGSI to bytecode.
+ * disk, and skips redundant shader compilations from NIR to bytecode.
* - It can only be used with one-variant-per-shader support, in which
* case only the main (typically middle) part of shaders is cached.
* - Only VS, TCS, TES, PS are cached, out of which only the hw VS
unsigned size;
};
+struct si_small_prim_cull_info {
+ float scale[2], translate[2];
+};
+
struct si_context {
struct pipe_context b; /* base class */
/* shader information */
struct si_vertex_elements *vertex_elements;
+ unsigned num_vertex_elements;
unsigned sprite_coord_enable;
unsigned cs_max_waves_per_sh;
bool flatshade;
bool do_update_shaders;
- /* vertex buffer descriptors */
- uint32_t *vb_descriptors_gpu_list;
- struct si_resource *vb_descriptors_buffer;
- unsigned vb_descriptors_offset;
-
/* shader descriptors */
struct si_descriptors descriptors[SI_NUM_DESCS];
unsigned descriptors_dirty;
uint32_t vs_blit_sh_data[SI_VS_BLIT_SGPRS_POS_TEXCOORD];
uint32_t cs_user_data[4];
- /* Vertex and index buffers. */
+ /* Vertex buffers. */
bool vertex_buffers_dirty;
bool vertex_buffer_pointer_dirty;
+ bool vertex_buffer_user_sgprs_dirty;
struct pipe_vertex_buffer vertex_buffer[SI_NUM_VERTEX_BUFFERS];
uint16_t vertex_buffer_unaligned; /* bitmask of not dword-aligned buffers */
+ uint32_t *vb_descriptors_gpu_list;
+ struct si_resource *vb_descriptors_buffer;
+ unsigned vb_descriptors_offset;
+ unsigned vb_descriptor_user_sgprs[5*4];
/* MSAA config state. */
int ps_iter_samples;
int last_restart_index;
int last_prim;
int last_multi_vgt_param;
- int last_rast_prim;
- int last_flatshade_first;
+ int last_gs_out_prim;
int last_binning_enabled;
- unsigned last_sc_line_stipple;
unsigned current_vs_state;
unsigned last_vs_state;
enum pipe_prim_type current_rast_prim; /* primitive type after TES, GS */
void si_blitter_end(struct si_context *sctx);
void si_init_blit_functions(struct si_context *sctx);
void si_decompress_textures(struct si_context *sctx, unsigned shader_mask);
+void si_decompress_subresource(struct pipe_context *ctx,
+ struct pipe_resource *tex,
+ unsigned planes, unsigned level,
+ unsigned first_layer, unsigned last_layer);
void si_resource_copy_region(struct pipe_context *ctx,
struct pipe_resource *dst,
unsigned dst_level,
struct radeon_saved_cs *saved, enum ring_type ring);
bool si_replace_shader(unsigned num, struct si_shader_binary *binary);
-/* si_dma.c */
-void si_init_dma_functions(struct si_context *sctx);
-
/* si_dma_cs.c */
void si_dma_emit_timestamp(struct si_context *sctx, struct si_resource *dst,
uint64_t offset);
void si_sdma_clear_buffer(struct si_context *sctx, struct pipe_resource *dst,
uint64_t offset, uint64_t size, unsigned clear_value);
+void si_sdma_copy_buffer(struct si_context *sctx, struct pipe_resource *dst,
+ struct pipe_resource *src, uint64_t dst_offset,
+ uint64_t src_offset, uint64_t size);
void si_need_dma_space(struct si_context *ctx, unsigned num_dw,
struct si_resource *dst, struct si_resource *src);
void si_flush_dma_cs(struct si_context *ctx, unsigned flags,
unsigned base_vertex,
uint64_t input_indexbuf_va,
unsigned input_indexbuf_max_elements);
-void si_initialize_prim_discard_tunables(struct si_context *sctx);
+void si_initialize_prim_discard_tunables(struct si_screen *sscreen,
+ bool is_aux_context,
+ unsigned *prim_discard_vertex_count_threshold,
+ unsigned *index_ring_size_per_ib);
/* si_pipe.c */
void si_init_compiler(struct si_screen *sscreen, struct ac_llvm_compiler *compiler);
const struct pipe_video_buffer *tmpl);
/* si_viewport.c */
+void si_get_small_prim_cull_info(struct si_context *sctx,
+ struct si_small_prim_cull_info *out);
void si_update_vs_viewport_state(struct si_context *ctx);
void si_init_viewport_functions(struct si_context *ctx);
return &sctx->vs_shader;
}
-static inline struct tgsi_shader_info *si_get_vs_info(struct si_context *sctx)
+static inline struct si_shader_info *si_get_vs_info(struct si_context *sctx)
{
struct si_shader_ctx_state *vs = si_get_vs(sctx);