#include "util/u_suballoc.h"
#include "util/u_transfer.h"
+#define ATI_VENDOR_ID 0x1002
+
#define R600_RESOURCE_FLAG_TRANSFER (PIPE_RESOURCE_FLAG_DRV_PRIV << 0)
#define R600_RESOURCE_FLAG_FLUSHED_DEPTH (PIPE_RESOURCE_FLAG_DRV_PRIV << 1)
#define R600_RESOURCE_FLAG_FORCE_TILING (PIPE_RESOURCE_FLAG_DRV_PRIV << 2)
-
-#define R600_QUERY_DRAW_CALLS (PIPE_QUERY_DRIVER_SPECIFIC + 0)
-#define R600_QUERY_REQUESTED_VRAM (PIPE_QUERY_DRIVER_SPECIFIC + 1)
-#define R600_QUERY_REQUESTED_GTT (PIPE_QUERY_DRIVER_SPECIFIC + 2)
-#define R600_QUERY_BUFFER_WAIT_TIME (PIPE_QUERY_DRIVER_SPECIFIC + 3)
-#define R600_QUERY_NUM_CS_FLUSHES (PIPE_QUERY_DRIVER_SPECIFIC + 4)
-#define R600_QUERY_NUM_BYTES_MOVED (PIPE_QUERY_DRIVER_SPECIFIC + 5)
-#define R600_QUERY_VRAM_USAGE (PIPE_QUERY_DRIVER_SPECIFIC + 6)
-#define R600_QUERY_GTT_USAGE (PIPE_QUERY_DRIVER_SPECIFIC + 7)
-#define R600_QUERY_GPU_TEMPERATURE (PIPE_QUERY_DRIVER_SPECIFIC + 8)
-#define R600_QUERY_CURRENT_GPU_SCLK (PIPE_QUERY_DRIVER_SPECIFIC + 9)
-#define R600_QUERY_CURRENT_GPU_MCLK (PIPE_QUERY_DRIVER_SPECIFIC + 10)
-#define R600_QUERY_GPU_LOAD (PIPE_QUERY_DRIVER_SPECIFIC + 11)
-#define R600_QUERY_NUM_COMPILATIONS (PIPE_QUERY_DRIVER_SPECIFIC + 12)
-#define R600_QUERY_NUM_SHADERS_CREATED (PIPE_QUERY_DRIVER_SPECIFIC + 13)
+#define R600_RESOURCE_FLAG_DISABLE_DCC (PIPE_RESOURCE_FLAG_DRV_PRIV << 3)
#define R600_CONTEXT_STREAMOUT_FLUSH (1u << 0)
-#define R600_CONTEXT_PRIVATE_FLAG (1u << 1)
+/* Pipeline & streamout query controls. */
+#define R600_CONTEXT_START_PIPELINE_STATS (1u << 1)
+#define R600_CONTEXT_STOP_PIPELINE_STATS (1u << 2)
+#define R600_CONTEXT_PRIVATE_FLAG (1u << 3)
/* special primitive types */
#define R600_PRIM_RECTANGLE_LIST PIPE_PRIM_MAX
/* Debug flags. */
/* logging */
#define DBG_TEX (1 << 0)
-#define DBG_TEXMIP (1 << 1)
+/* gap - reuse */
#define DBG_COMPUTE (1 << 2)
#define DBG_VM (1 << 3)
-#define DBG_TRACE_CS (1 << 4)
+/* gap - reuse */
/* shader logging */
#define DBG_FS (1 << 5)
#define DBG_VS (1 << 6)
#define DBG_NO_IR (1 << 12)
#define DBG_NO_TGSI (1 << 13)
#define DBG_NO_ASM (1 << 14)
+#define DBG_PREOPT_IR (1 << 15)
+/* gaps */
+#define DBG_TEST_DMA (1 << 20)
/* Bits 21-31 are reserved for the r600g driver. */
/* features */
#define DBG_NO_ASYNC_DMA (1llu << 32)
#define DBG_CHECK_VM (1llu << 42)
#define DBG_NO_DCC (1llu << 43)
#define DBG_NO_DCC_CLEAR (1llu << 44)
+#define DBG_NO_RB_PLUS (1llu << 45)
+#define DBG_SI_SCHED (1llu << 46)
+#define DBG_MONOLITHIC_SHADERS (1llu << 47)
+#define DBG_NO_CE (1llu << 48)
+#define DBG_UNSAFE_MATH (1llu << 49)
#define R600_MAP_BUFFER_ALIGNMENT 64
+#define R600_MAX_VIEWPORTS 16
+
+enum r600_coherency {
+ R600_COHERENCY_NONE, /* no cache flushes needed */
+ R600_COHERENCY_SHADER,
+ R600_COHERENCY_CB_META,
+};
+
+#ifdef PIPE_ARCH_BIG_ENDIAN
+#define R600_BIG_ENDIAN 1
+#else
+#define R600_BIG_ENDIAN 0
+#endif
struct r600_common_context;
+struct r600_perfcounters;
+struct tgsi_shader_info;
struct radeon_shader_reloc {
- char *name;
+ char name[32];
uint64_t offset;
};
char *disasm_string;
};
+void radeon_shader_binary_init(struct radeon_shader_binary *b);
+void radeon_shader_binary_clean(struct radeon_shader_binary *b);
+
+/* Only 32-bit buffer allocations are supported, gallium doesn't support more
+ * at the moment.
+ */
struct r600_resource {
struct u_resource b;
/* Winsys objects. */
struct pb_buffer *buf;
- struct radeon_winsys_cs_handle *cs_buf;
uint64_t gpu_address;
/* Resource state. */
* use TC L2.
*/
bool TC_L2_dirty;
+
+ /* Whether the resource has been exported via resource_get_handle. */
+ bool is_shared;
+ unsigned external_usage; /* PIPE_HANDLE_USAGE_* */
};
struct r600_transfer {
};
struct r600_fmask_info {
- unsigned offset;
- unsigned size;
+ uint64_t offset;
+ uint64_t size;
unsigned alignment;
- unsigned pitch;
+ unsigned pitch_in_pixels;
unsigned bank_height;
unsigned slice_tile_max;
unsigned tile_mode_index;
};
struct r600_cmask_info {
- unsigned offset;
- unsigned size;
+ uint64_t offset;
+ uint64_t size;
unsigned alignment;
+ unsigned pitch;
+ unsigned height;
+ unsigned xalign;
+ unsigned yalign;
unsigned slice_tile_max;
unsigned base_address_reg;
};
+struct r600_htile_info {
+ unsigned pitch;
+ unsigned height;
+ unsigned xalign;
+ unsigned yalign;
+};
+
struct r600_texture {
struct r600_resource resource;
- unsigned size;
- unsigned pitch_override;
+ uint64_t size;
+ unsigned num_level0_transfers;
bool is_depth;
unsigned dirty_level_mask; /* each bit says if that mipmap is compressed */
unsigned stencil_dirty_level_mask; /* each bit says if that mipmap is compressed */
struct r600_fmask_info fmask;
struct r600_cmask_info cmask;
struct r600_resource *cmask_buffer;
- struct r600_resource *dcc_buffer;
+ uint64_t dcc_offset; /* 0 = disabled */
unsigned cb_color_info; /* fast clear enable bit */
unsigned color_clear_value[2];
+ unsigned last_msaa_resolve_target_micro_mode;
/* Depth buffer compression and fast clear. */
+ struct r600_htile_info htile;
struct r600_resource *htile_buffer;
bool depth_cleared; /* if it was cleared at least once */
float depth_clear_value;
+ bool stencil_cleared; /* if it was cleared at least once */
+ uint8_t stencil_clear_value;
bool non_disp_tiling; /* R600-Cayman only */
+
+ /* Counter that should be non-zero if the texture is bound to a
+ * framebuffer. Implemented in radeonsi only.
+ */
+ uint32_t framebuffers_bound;
};
struct r600_surface {
struct pipe_surface base;
+ const struct radeon_surf_level *level_info;
bool color_initialized;
bool depth_initialized;
/* Misc. color flags. */
bool alphatest_bypass;
bool export_16bpc;
+ bool color_is_int8;
/* Color registers. */
unsigned cb_color_info;
unsigned cb_color_dim; /* EG only */
unsigned cb_color_pitch; /* EG and later */
unsigned cb_color_slice; /* EG and later */
- unsigned cb_dcc_base; /* VI and later */
unsigned cb_color_attrib; /* EG and later */
unsigned cb_dcc_control; /* VI and later */
unsigned cb_color_fmask; /* CB_COLORn_FMASK (EG and later) or CB_COLORn_FRAG (r600) */
unsigned cb_color_fmask_slice; /* EG and later */
unsigned cb_color_cmask; /* CB_COLORn_TILE (r600 only) */
unsigned cb_color_mask; /* R600 only */
+ unsigned spi_shader_col_format; /* SI+, no blending, no alpha-to-coverage. */
+ unsigned spi_shader_col_format_alpha; /* SI+, alpha-to-coverage */
+ unsigned spi_shader_col_format_blend; /* SI+, blending without alpha. */
+ unsigned spi_shader_col_format_blend_alpha; /* SI+, blending with alpha. */
struct r600_resource *cb_buffer_fmask; /* Used for FMASK relocations. R600 only */
struct r600_resource *cb_buffer_cmask; /* Used for CMASK relocations. R600 only */
unsigned pa_su_poly_offset_db_fmt_cntl;
};
-struct r600_tiling_info {
- unsigned num_channels;
- unsigned num_banks;
- unsigned group_bytes;
-};
-
struct r600_common_screen {
struct pipe_screen b;
struct radeon_winsys *ws;
enum radeon_family family;
enum chip_class chip_class;
struct radeon_info info;
- struct r600_tiling_info tiling_info;
uint64_t debug_flags;
bool has_cp_dma;
bool has_streamout;
+ /* Texture filter settings. */
+ int force_aniso; /* -1 = disabled */
+
/* Auxiliary context. Mainly used to initialize resources.
* It must be locked prior to using and flushed before unlocking. */
struct pipe_context *aux_context;
pipe_mutex aux_context_lock;
- struct r600_resource *trace_bo;
- uint32_t *trace_ptr;
- unsigned cs_count;
-
/* This must be in the screen, because UE4 uses one context for
* compilation and another one for rendering.
*/
volatile unsigned gpu_load_stop_thread; /* bool */
char renderer_string[64];
+
+ /* Performance counters. */
+ struct r600_perfcounters *perfcounters;
+
+ /* If pipe_screen wants to re-emit the framebuffer state of all
+ * contexts, it should atomically increment this. Each context will
+ * compare this with its own last known value of the counter before
+ * drawing and re-emit the framebuffer state accordingly.
+ */
+ unsigned dirty_fb_counter;
+
+ /* Atomically increment this counter when an existing texture's
+ * metadata is enabled or disabled in a way that requires changing
+ * contexts' compressed texture binding masks.
+ */
+ unsigned compressed_colortex_counter;
+
+ /* Atomically increment this counter when an existing texture's
+ * backing buffer or tile mode parameters have changed that requires
+ * recomputation of shader descriptors.
+ */
+ unsigned dirty_tex_descriptor_counter;
+
+ void (*query_opaque_metadata)(struct r600_common_screen *rscreen,
+ struct r600_texture *rtex,
+ struct radeon_bo_metadata *md);
+
+ void (*apply_opaque_metadata)(struct r600_common_screen *rscreen,
+ struct r600_texture *rtex,
+ struct radeon_bo_metadata *md);
};
/* This encapsulates a state or an operation which can emitted into the GPU
int num_prims_gen_queries;
};
+struct r600_signed_scissor {
+ int minx;
+ int miny;
+ int maxx;
+ int maxy;
+};
+
+struct r600_scissors {
+ struct r600_atom atom;
+ unsigned dirty_mask;
+ struct pipe_scissor_state states[R600_MAX_VIEWPORTS];
+};
+
+struct r600_viewports {
+ struct r600_atom atom;
+ unsigned dirty_mask;
+ struct pipe_viewport_state states[R600_MAX_VIEWPORTS];
+ struct r600_signed_scissor as_scissor[R600_MAX_VIEWPORTS];
+};
+
struct r600_ring {
struct radeon_winsys_cs *cs;
- bool flushing;
void (*flush)(void *ctx, unsigned flags,
struct pipe_fence_handle **fence);
};
-struct r600_rings {
- struct r600_ring gfx;
- struct r600_ring dma;
-};
-
struct r600_common_context {
struct pipe_context b; /* base class */
struct radeon_winsys_ctx *ctx;
enum radeon_family family;
enum chip_class chip_class;
- struct r600_rings rings;
+ struct r600_ring gfx;
+ struct r600_ring dma;
+ struct pipe_fence_handle *last_sdma_fence;
unsigned initial_gfx_cs_size;
unsigned gpu_reset_counter;
+ unsigned last_dirty_fb_counter;
+ unsigned last_compressed_colortex_counter;
+ unsigned last_dirty_tex_descriptor_counter;
struct u_upload_mgr *uploader;
- struct u_suballocator *allocator_so_filled_size;
+ struct u_suballocator *allocator_zeroed_memory;
struct util_slab_mempool pool_transfers;
/* Current unaccounted memory usage. */
/* States. */
struct r600_streamout streamout;
+ struct r600_scissors scissors;
+ struct r600_viewports viewports;
+ bool scissor_enabled;
+ bool vs_writes_viewport_index;
+ bool vs_disables_clipping_viewport;
/* Additional context states. */
unsigned flags; /* flush flags */
/* Queries. */
- /* The list of active queries. Only one query of each type can be active. */
+ /* Maintain the list of active queries for pausing between IBs. */
int num_occlusion_queries;
- /* Keep track of non-timer queries, because they should be suspended
- * during context flushing.
- * The timer queries (TIME_ELAPSED) shouldn't be suspended for blits,
- * but they should be suspended between IBs. */
- struct list_head active_nontimer_queries;
- struct list_head active_timer_queries;
- unsigned num_cs_dw_nontimer_queries_suspend;
- unsigned num_cs_dw_timer_queries_suspend;
- /* If queries have been suspended. */
- bool queries_suspended_for_flush;
+ int num_perfect_occlusion_queries;
+ struct list_head active_queries;
+ unsigned num_cs_dw_queries_suspend;
/* Additional hardware info. */
unsigned backend_mask;
unsigned max_db; /* for OQ */
/* Misc stats. */
unsigned num_draw_calls;
+ unsigned num_spill_draw_calls;
+ unsigned num_compute_calls;
+ unsigned num_spill_compute_calls;
+ unsigned num_dma_calls;
+ uint64_t num_alloc_tex_transfer_bytes;
/* Render condition. */
- struct pipe_query *current_render_cond;
- unsigned current_render_cond_mode;
- boolean current_render_cond_cond;
- boolean predicate_drawing;
- /* For context flushing. */
- struct pipe_query *saved_render_cond;
- boolean saved_render_cond_cond;
- unsigned saved_render_cond_mode;
+ struct r600_atom render_cond_atom;
+ struct pipe_query *render_cond;
+ unsigned render_cond_mode;
+ boolean render_cond_invert;
+ bool render_cond_force_off; /* for u_blitter */
/* MSAA sample locations.
* The first index is the sample index.
* the GPU addresses are updated. */
struct list_head texture_buffers;
+ struct pipe_debug_callback debug;
+
/* Copy one resource to another using async DMA. */
void (*dma_copy)(struct pipe_context *ctx,
struct pipe_resource *dst,
const struct pipe_box *src_box);
void (*clear_buffer)(struct pipe_context *ctx, struct pipe_resource *dst,
- unsigned offset, unsigned size, unsigned value,
- bool is_framebuffer);
+ uint64_t offset, uint64_t size, unsigned value,
+ enum r600_coherency coher);
void (*blit_decompress_depth)(struct pipe_context *ctx,
struct r600_texture *texture,
unsigned first_layer, unsigned last_layer,
unsigned first_sample, unsigned last_sample);
+ void (*decompress_dcc)(struct pipe_context *ctx,
+ struct r600_texture *rtex);
+
/* Reallocate the buffer and update all resource bindings where
* the buffer is bound, including all resource descriptors. */
void (*invalidate_buffer)(struct pipe_context *ctx, struct pipe_resource *buf);
/* r600_buffer.c */
boolean r600_rings_is_buffer_referenced(struct r600_common_context *ctx,
- struct radeon_winsys_cs_handle *buf,
+ struct pb_buffer *buf,
enum radeon_bo_usage usage);
void *r600_buffer_map_sync_with_rings(struct r600_common_context *ctx,
struct r600_resource *resource,
unsigned usage);
bool r600_init_resource(struct r600_common_screen *rscreen,
struct r600_resource *res,
- unsigned size, unsigned alignment,
- bool use_reusable_pool);
+ uint64_t size, unsigned alignment);
struct pipe_resource *r600_buffer_create(struct pipe_screen *screen,
const struct pipe_resource *templ,
unsigned alignment);
r600_buffer_from_user_memory(struct pipe_screen *screen,
const struct pipe_resource *templ,
void *user_memory);
+void
+r600_invalidate_resource(struct pipe_context *ctx,
+ struct pipe_resource *resource);
/* r600_common_pipe.c */
void r600_draw_rectangle(struct blitter_context *blitter,
void r600_common_context_cleanup(struct r600_common_context *rctx);
void r600_context_add_resource_size(struct pipe_context *ctx, struct pipe_resource *r);
bool r600_can_dump_shader(struct r600_common_screen *rscreen,
- const struct tgsi_token *tokens);
+ unsigned processor);
void r600_screen_clear_buffer(struct r600_common_screen *rscreen, struct pipe_resource *dst,
- unsigned offset, unsigned size, unsigned value,
- bool is_framebuffer);
+ uint64_t offset, uint64_t size, unsigned value,
+ enum r600_coherency coher);
struct pipe_resource *r600_resource_create_common(struct pipe_screen *screen,
const struct pipe_resource *templ);
const char *r600_get_llvm_processor_name(enum radeon_family family);
-void r600_need_dma_space(struct r600_common_context *ctx, unsigned num_dw);
+void r600_need_dma_space(struct r600_common_context *ctx, unsigned num_dw,
+ struct r600_resource *dst, struct r600_resource *src);
+void r600_dma_emit_wait_idle(struct r600_common_context *rctx);
/* r600_gpu_load.c */
void r600_gpu_load_kill_thread(struct r600_common_screen *rscreen);
uint64_t r600_gpu_load_begin(struct r600_common_screen *rscreen);
unsigned r600_gpu_load_end(struct r600_common_screen *rscreen, uint64_t begin);
+/* r600_perfcounters.c */
+void r600_perfcounters_destroy(struct r600_common_screen *rscreen);
+
/* r600_query.c */
+void r600_init_screen_query_functions(struct r600_common_screen *rscreen);
void r600_query_init(struct r600_common_context *rctx);
-void r600_suspend_nontimer_queries(struct r600_common_context *ctx);
-void r600_resume_nontimer_queries(struct r600_common_context *ctx);
-void r600_suspend_timer_queries(struct r600_common_context *ctx);
-void r600_resume_timer_queries(struct r600_common_context *ctx);
+void r600_suspend_queries(struct r600_common_context *ctx);
+void r600_resume_queries(struct r600_common_context *ctx);
void r600_query_init_backend_mask(struct r600_common_context *ctx);
/* r600_streamout.c */
unsigned type, int diff);
void r600_streamout_init(struct r600_common_context *rctx);
+/* r600_test_dma.c */
+void r600_test_dma(struct r600_common_screen *rscreen);
+
/* r600_texture.c */
+bool r600_prepare_for_dma_blit(struct r600_common_context *rctx,
+ struct r600_texture *rdst,
+ unsigned dst_level, unsigned dstx,
+ unsigned dsty, unsigned dstz,
+ struct r600_texture *rsrc,
+ unsigned src_level,
+ const struct pipe_box *src_box);
void r600_texture_get_fmask_info(struct r600_common_screen *rscreen,
struct r600_texture *rtex,
unsigned nr_samples,
bool r600_init_flushed_depth_texture(struct pipe_context *ctx,
struct pipe_resource *texture,
struct r600_texture **staging);
+void r600_print_texture_info(struct r600_texture *rtex, FILE *f);
struct pipe_resource *r600_texture_create(struct pipe_screen *screen,
const struct pipe_resource *templ);
struct pipe_surface *r600_create_surface_custom(struct pipe_context *pipe,
struct pipe_resource *texture,
const struct pipe_surface *templ,
unsigned width, unsigned height);
-unsigned r600_translate_colorswap(enum pipe_format format);
+unsigned r600_translate_colorswap(enum pipe_format format, bool do_endian_swap);
+void vi_dcc_clear_level(struct r600_common_context *rctx,
+ struct r600_texture *rtex,
+ unsigned level, unsigned clear_value);
void evergreen_do_fast_color_clear(struct r600_common_context *rctx,
struct pipe_framebuffer_state *fb,
struct r600_atom *fb_state,
unsigned *buffers, unsigned *dirty_cbufs,
const union pipe_color_union *color);
+bool r600_texture_disable_dcc(struct r600_common_screen *rscreen,
+ struct r600_texture *rtex);
void r600_init_screen_texture_functions(struct r600_common_screen *rscreen);
void r600_init_context_texture_functions(struct r600_common_context *rctx);
+/* r600_viewport.c */
+void evergreen_apply_scissor_bug_workaround(struct r600_common_context *rctx,
+ struct pipe_scissor_state *scissor);
+void r600_set_scissor_enable(struct r600_common_context *rctx, bool enable);
+void r600_update_vs_writes_viewport_index(struct r600_common_context *rctx,
+ struct tgsi_shader_info *info);
+void r600_init_viewport_functions(struct r600_common_context *rctx);
+
/* cayman_msaa.c */
extern const uint32_t eg_sample_locs_2x[4];
extern const unsigned eg_max_dist_2x;
(struct pipe_resource *)res);
}
+static inline bool r600_get_strmout_en(struct r600_common_context *rctx)
+{
+ return rctx->streamout.streamout_enabled ||
+ rctx->streamout.prims_gen_query_enabled;
+}
+
+#define SQ_TEX_XY_FILTER_POINT 0x00
+#define SQ_TEX_XY_FILTER_BILINEAR 0x01
+#define SQ_TEX_XY_FILTER_ANISO_POINT 0x02
+#define SQ_TEX_XY_FILTER_ANISO_BILINEAR 0x03
+
+static inline unsigned eg_tex_filter(unsigned filter, unsigned max_aniso)
+{
+ if (filter == PIPE_TEX_FILTER_LINEAR)
+ return max_aniso > 1 ? SQ_TEX_XY_FILTER_ANISO_BILINEAR
+ : SQ_TEX_XY_FILTER_BILINEAR;
+ else
+ return max_aniso > 1 ? SQ_TEX_XY_FILTER_ANISO_POINT
+ : SQ_TEX_XY_FILTER_POINT;
+}
+
static inline unsigned r600_tex_aniso_filter(unsigned filter)
{
- if (filter <= 1) return 0;
- if (filter <= 2) return 1;
- if (filter <= 4) return 2;
- if (filter <= 8) return 3;
- /* else */ return 4;
+ if (filter < 2)
+ return 0;
+ if (filter < 4)
+ return 1;
+ if (filter < 8)
+ return 2;
+ if (filter < 16)
+ return 3;
+ return 4;
}
static inline unsigned r600_wavefront_size(enum radeon_family family)
} while (0);
#define R600_ERR(fmt, args...) \
- fprintf(stderr, "EE %s:%d %s - "fmt, __FILE__, __LINE__, __func__, ##args)
+ fprintf(stderr, "EE %s:%d %s - " fmt, __FILE__, __LINE__, __func__, ##args)
/* For MSAA sample positions. */
#define FILL_SREG(s0x, s0y, s1x, s1y, s2x, s2y, s3x, s3y) \
- (((s0x) & 0xf) | (((s0y) & 0xf) << 4) | \
- (((s1x) & 0xf) << 8) | (((s1y) & 0xf) << 12) | \
- (((s2x) & 0xf) << 16) | (((s2y) & 0xf) << 20) | \
- (((s3x) & 0xf) << 24) | (((s3y) & 0xf) << 28))
+ (((s0x) & 0xf) | (((unsigned)(s0y) & 0xf) << 4) | \
+ (((unsigned)(s1x) & 0xf) << 8) | (((unsigned)(s1y) & 0xf) << 12) | \
+ (((unsigned)(s2x) & 0xf) << 16) | (((unsigned)(s2y) & 0xf) << 20) | \
+ (((unsigned)(s3x) & 0xf) << 24) | (((unsigned)(s3y) & 0xf) << 28))
#endif