radeonsi/gfx10: move s_sendmsg gs_alloc_req to the beginning of shaders
[mesa.git] / src / gallium / drivers / radeonsi / si_pipe.h
index e222de1e906f1aa79375e9881ac164af852c6a87..563a201ec3344fb4f473a764781ec54cf9b5de7f 100644 (file)
 #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
 {
@@ -139,7 +144,7 @@ enum {
        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,
 
@@ -316,7 +321,7 @@ struct si_texture {
        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 */
@@ -481,6 +486,7 @@ struct si_screen {
                        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;
@@ -580,7 +586,7 @@ struct si_screen {
         *
         * 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
@@ -856,6 +862,10 @@ struct si_sdma_upload {
        unsigned                size;
 };
 
+struct si_small_prim_cull_info {
+       float scale[2], translate[2];
+};
+
 struct si_context {
        struct pipe_context             b; /* base class */
 
@@ -995,16 +1005,12 @@ struct si_context {
 
        /* 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;
@@ -1031,11 +1037,16 @@ struct si_context {
        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;
@@ -1071,10 +1082,8 @@ struct si_context {
        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 */
@@ -1248,6 +1257,10 @@ void si_blitter_begin(struct si_context *sctx, enum si_blitter_op op);
 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,
@@ -1363,14 +1376,14 @@ void si_check_vm_faults(struct si_context *sctx,
                        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,
@@ -1432,7 +1445,10 @@ void si_dispatch_prim_discard_cs_and_draw(struct si_context *sctx,
                                          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);
@@ -1483,6 +1499,8 @@ struct pipe_video_buffer *si_video_buffer_create(struct pipe_context *pipe,
                                                 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);
 
@@ -1638,7 +1656,7 @@ static inline struct si_shader_ctx_state *si_get_vs(struct si_context *sctx)
        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);