#include "radeon/r600_pipe_common.h"
#include "radeon/r600_cs.h"
-
-#include "r600_llvm.h"
#include "r600_public.h"
#include "util/u_suballoc.h"
#include "util/list.h"
#include "util/u_transfer.h"
+#include "util/u_memory.h"
-#define R600_NUM_ATOMS 73
+#include "tgsi/tgsi_scan.h"
-#define R600_MAX_VIEWPORTS 16
+#define R600_NUM_ATOMS 52
/* read caches */
#define R600_CONTEXT_INV_VERTEX_CACHE (R600_CONTEXT_PRIVATE_FLAG << 0)
#define R600_CONTEXT_WAIT_CP_DMA_IDLE (R600_CONTEXT_PRIVATE_FLAG << 10)
/* the number of CS dwords for flushing and drawing */
-#define R600_MAX_FLUSH_CS_DWORDS 16
-#define R600_MAX_DRAW_CS_DWORDS 47
-#define R600_TRACE_CS_DWORDS 7
+#define R600_MAX_FLUSH_CS_DWORDS 18
+#define R600_MAX_DRAW_CS_DWORDS 58
+#define R600_MAX_PFP_SYNC_ME_DWORDS 16
#define R600_MAX_USER_CONST_BUFFERS 13
#define R600_MAX_DRIVER_CONST_BUFFERS 3
#define R600_MAX_CONST_BUFFERS (R600_MAX_USER_CONST_BUFFERS + R600_MAX_DRIVER_CONST_BUFFERS)
/* start driver buffers after user buffers */
-#define R600_UCP_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS)
-#define R600_BUFFER_INFO_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS + 1)
+#define R600_BUFFER_INFO_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS)
+#define R600_UCP_SIZE (4*4*8)
+#define R600_BUFFER_INFO_OFFSET (R600_UCP_SIZE)
+
+#define R600_LDS_INFO_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS + 1)
+/*
+ * Note GS doesn't use a constant buffer binding, just a resource index,
+ * so it's fine to have it exist at index 16.
+ */
#define R600_GS_RING_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS + 2)
/* Currently R600_MAX_CONST_BUFFERS just fits on the hw, which has a limit
* of 16 const buffers.
* In order to support d3d 11 mandated minimum of 15 user const buffers
* we'd have to squash all use cases into one driver buffer.
*/
-#define R600_SAMPLE_POSITIONS_CONST_BUFFER (R600_MAX_USER_CONST_BUFFERS)
-
#define R600_MAX_CONST_BUFFER_SIZE (4096 * sizeof(float[4]))
-#ifdef PIPE_ARCH_BIG_ENDIAN
-#define R600_BIG_ENDIAN 1
-#else
-#define R600_BIG_ENDIAN 0
-#endif
+/* HW stages */
+#define R600_HW_STAGE_PS 0
+#define R600_HW_STAGE_VS 1
+#define R600_HW_STAGE_GS 2
+#define R600_HW_STAGE_ES 3
+#define EG_HW_STAGE_LS 4
+#define EG_HW_STAGE_HS 5
+
+#define R600_NUM_HW_STAGES 4
+#define EG_NUM_HW_STAGES 6
struct r600_context;
struct r600_bytecode;
-struct r600_shader_key;
+union r600_shader_key;
/* This is an atom containing GPU commands that never change.
* This is supposed to be copied directly into the CS. */
struct r600_db_misc_state {
struct r600_atom atom;
- bool occlusion_query_enabled;
+ bool occlusion_queries_disabled;
bool flush_depthstencil_through_cb;
- bool flush_depthstencil_in_place;
+ bool flush_depth_inplace;
+ bool flush_stencil_inplace;
bool copy_depth, copy_stencil;
unsigned copy_sample;
unsigned log_samples;
unsigned db_shader_control;
bool htile_clear;
+ uint8_t ps_conservative_z;
};
struct r600_cb_misc_state {
unsigned clip_plane_enable; /* from rasterizer */
unsigned clip_dist_write; /* from vertex shader */
boolean clip_disable; /* from vertex shader */
+ boolean vs_out_viewport; /* from vertex shader */
};
struct r600_alphatest_state {
struct r600_atom atom;
unsigned sq_gpr_resource_mgmt_1;
unsigned sq_gpr_resource_mgmt_2;
+ unsigned sq_gpr_resource_mgmt_3;
+ bool dyn_gpr_enabled;
};
struct r600_stencil_ref
struct pipe_stencil_ref pipe_state;
};
-struct r600_viewport_state {
- struct r600_atom atom;
- struct pipe_viewport_state state;
- int idx;
-};
-
struct r600_shader_stages_state {
struct r600_atom atom;
unsigned geom_enable;
/* This must start from 16. */
/* features */
-#define DBG_LLVM (1 << 29)
#define DBG_NO_CP_DMA (1 << 30)
/* shader backend */
#define DBG_NO_SB (1 << 21)
struct r600_resource *tex_resource;
uint32_t tex_resource_words[8];
bool skip_mip_address_reloc;
+ bool is_stencil_sampler;
};
struct r600_rasterizer_state {
float offset_units;
float offset_scale;
bool offset_enable;
+ bool offset_units_unscaled;
bool scissor_enable;
bool multisample_enable;
};
enum pipe_format zs_format;
float offset_units;
float offset_scale;
+ bool offset_units_unscaled;
};
struct r600_blend_state {
struct tgsi_token *tokens;
struct pipe_stream_output_info so;
+ struct tgsi_shader_info info;
unsigned num_shaders;
/* PIPE_SHADER_[VERTEX|FRAGMENT|...] */
unsigned type;
+ /* geometry shader properties */
+ unsigned gs_output_prim;
+ unsigned gs_max_out_vertices;
+ unsigned gs_num_invocations;
+
+ /* TCS/VS */
+ uint64_t lds_patch_outputs_written_mask;
+ uint64_t lds_outputs_written_mask;
unsigned nr_ps_max_color_exports;
};
struct r600_samplerview_state views;
struct r600_sampler_states states;
bool is_array_sampler[NUM_TEX_UNITS];
+};
- /* cube array txq workaround */
- uint32_t *txq_constants;
- /* buffer related workarounds */
- uint32_t *buffer_constants;
+struct r600_shader_driver_constants_info {
+ /* currently 128 bytes for UCP/samplepos + sampler buffer constants */
+ uint32_t *constants;
+ uint32_t alloc_size;
+ bool vs_ucp_dirty;
+ bool texture_const_dirty;
+ bool ps_sample_pos_dirty;
};
struct r600_constbuf_state
struct r600_command_buffer *cb;
};
-struct r600_scissor_state
-{
- struct r600_atom atom;
- struct pipe_scissor_state scissor;
- bool enable; /* r6xx only */
- int idx;
-};
-
struct r600_fetch_shader {
struct r600_resource *buffer;
unsigned offset;
/* Hardware info. */
boolean has_vertex_cache;
- boolean keep_tiling_flags;
- unsigned default_ps_gprs, default_vs_gprs;
+ unsigned default_gprs[EG_NUM_HW_STAGES];
+ unsigned current_gprs[EG_NUM_HW_STAGES];
unsigned r6xx_num_clause_temp_gprs;
/* Miscellaneous state objects. */
/* State binding slots are here. */
struct r600_atom *atoms[R600_NUM_ATOMS];
+ /* Dirty atom bitmask for fast tests */
+ uint64_t dirty_atoms;
/* States for CS initialization. */
struct r600_command_buffer start_cs_cmd; /* invariant state mostly */
/** Compute specific registers initializations. The start_cs_cmd atom
struct r600_poly_offset_state poly_offset_state;
struct r600_cso_state rasterizer_state;
struct r600_sample_mask sample_mask;
- struct r600_scissor_state scissor[R600_MAX_VIEWPORTS];
struct r600_seamless_cube_map seamless_cube_map;
struct r600_config_state config_state;
struct r600_stencil_ref_state stencil_ref;
struct r600_vgt_state vgt_state;
- struct r600_viewport_state viewport[R600_MAX_VIEWPORTS];
/* Shaders and shader resources. */
struct r600_cso_state vertex_fetch_shader;
- struct r600_shader_state vertex_shader;
- struct r600_shader_state pixel_shader;
- struct r600_shader_state geometry_shader;
- struct r600_shader_state export_shader;
+ struct r600_shader_state hw_shader_stages[EG_NUM_HW_STAGES];
struct r600_cs_shader_state cs_shader_state;
struct r600_shader_stages_state shader_stages;
struct r600_gs_rings_state gs_rings;
struct r600_constbuf_state constbuf_state[PIPE_SHADER_TYPES];
struct r600_textures_info samplers[PIPE_SHADER_TYPES];
+
+ struct r600_shader_driver_constants_info driver_consts[PIPE_SHADER_TYPES];
+
/** Vertex buffers for fetch shaders */
struct r600_vertexbuf_state vertex_buffer_state;
/** Vertex buffers for compute shaders */
struct r600_pipe_shader_selector *ps_shader;
struct r600_pipe_shader_selector *vs_shader;
struct r600_pipe_shader_selector *gs_shader;
+
+ struct r600_pipe_shader_selector *tcs_shader;
+ struct r600_pipe_shader_selector *tes_shader;
+
+ struct r600_pipe_shader_selector *fixed_func_tcs_shader;
+
struct r600_rasterizer_state *rasterizer;
bool alpha_to_one;
bool force_blend_disable;
void *sb_context;
struct r600_isa *isa;
+ float sample_positions[4 * 16];
+ float tess_state[8];
+ bool tess_state_dirty;
+ struct r600_pipe_shader_selector *last_ls;
+ struct r600_pipe_shader_selector *last_tcs;
+ unsigned last_num_tcs_input_cp;
+ unsigned lds_alloc;
};
static inline void r600_emit_command_buffer(struct radeon_winsys_cs *cs,
struct r600_command_buffer *cb)
{
- assert(cs->cdw + cb->num_dw <= cs->max_dw);
- memcpy(cs->buf + cs->cdw, cb->buf, 4 * cb->num_dw);
- cs->cdw += cb->num_dw;
+ assert(cs->current.cdw + cb->num_dw <= cs->current.max_dw);
+ memcpy(cs->current.buf + cs->current.cdw, cb->buf, 4 * cb->num_dw);
+ cs->current.cdw += cb->num_dw;
}
-void r600_trace_emit(struct r600_context *rctx);
+static inline void r600_set_atom_dirty(struct r600_context *rctx,
+ struct r600_atom *atom,
+ bool dirty)
+{
+ uint64_t mask;
+
+ assert(atom->id != 0);
+ assert(atom->id < sizeof(mask) * 8);
+ mask = 1ull << atom->id;
+ if (dirty)
+ rctx->dirty_atoms |= mask;
+ else
+ rctx->dirty_atoms &= ~mask;
+}
+
+static inline void r600_mark_atom_dirty(struct r600_context *rctx,
+ struct r600_atom *atom)
+{
+ r600_set_atom_dirty(rctx, atom, true);
+}
static inline void r600_emit_atom(struct r600_context *rctx, struct r600_atom *atom)
{
atom->emit(&rctx->b, atom);
- atom->dirty = false;
- if (rctx->screen->b.trace_bo) {
- r600_trace_emit(rctx);
- }
+ r600_set_atom_dirty(rctx, atom, false);
}
-static inline void r600_set_cso_state(struct r600_cso_state *state, void *cso)
+static inline void r600_set_cso_state(struct r600_context *rctx,
+ struct r600_cso_state *state, void *cso)
{
state->cso = cso;
- state->atom.dirty = cso != NULL;
+ r600_set_atom_dirty(rctx, &state->atom, cso != NULL);
}
-static inline void r600_set_cso_state_with_cb(struct r600_cso_state *state, void *cso,
+static inline void r600_set_cso_state_with_cb(struct r600_context *rctx,
+ struct r600_cso_state *state, void *cso,
struct r600_command_buffer *cb)
{
state->cb = cb;
state->atom.num_dw = cb ? cb->num_dw : 0;
- r600_set_cso_state(state, cso);
+ r600_set_cso_state(rctx, state, cso);
}
/* compute_memory_pool.c */
struct compute_memory_pool* compute_memory_pool_new(
struct r600_screen *rscreen);
-/* evergreen_compute.c */
-void evergreen_set_cs_sampler_view(struct pipe_context *ctx_,
- unsigned start_slot, unsigned count,
- struct pipe_sampler_view **views);
-
/* evergreen_state.c */
struct pipe_sampler_view *
evergreen_create_sampler_view_custom(struct pipe_context *ctx,
const struct pipe_sampler_view *state,
unsigned width0, unsigned height0,
unsigned force_level);
-void evergreen_init_common_regs(struct r600_command_buffer *cb,
+void evergreen_init_common_regs(struct r600_context *ctx,
+ struct r600_command_buffer *cb,
enum chip_class ctx_chip_class,
enum radeon_family ctx_family,
int ctx_drm_minor);
void evergreen_update_es_state(struct pipe_context *ctx, struct r600_pipe_shader *shader);
void evergreen_update_gs_state(struct pipe_context *ctx, struct r600_pipe_shader *shader);
void evergreen_update_vs_state(struct pipe_context *ctx, struct r600_pipe_shader *shader);
+void evergreen_update_ls_state(struct pipe_context *ctx, struct r600_pipe_shader *shader);
+void evergreen_update_hs_state(struct pipe_context *ctx, struct r600_pipe_shader *shader);
void *evergreen_create_db_flush_dsa(struct r600_context *rctx);
void *evergreen_create_resolve_blend(struct r600_context *rctx);
void *evergreen_create_decompress_blend(struct r600_context *rctx);
void evergreen_init_color_surface_rat(struct r600_context *rctx,
struct r600_surface *surf);
void evergreen_update_db_shader_control(struct r600_context * rctx);
-
+bool evergreen_adjust_gprs(struct r600_context *rctx);
/* r600_blit.c */
void r600_init_blit_functions(struct r600_context *rctx);
void r600_decompress_depth_textures(struct r600_context *rctx,
/* r600_shader.c */
int r600_pipe_shader_create(struct pipe_context *ctx,
struct r600_pipe_shader *shader,
- struct r600_shader_key key);
+ union r600_shader_key key);
void r600_pipe_shader_destroy(struct pipe_context *ctx, struct r600_pipe_shader *shader);
void r600_begin_new_cs(struct r600_context *ctx);
void r600_flush_emit(struct r600_context *ctx);
void r600_need_cs_space(struct r600_context *ctx, unsigned num_dw, boolean count_draw_in);
+void r600_emit_pfp_sync_me(struct r600_context *rctx);
void r600_cp_dma_copy_buffer(struct r600_context *rctx,
struct pipe_resource *dst, uint64_t dst_offset,
struct pipe_resource *src, uint64_t src_offset,
unsigned size);
void evergreen_cp_dma_clear_buffer(struct r600_context *rctx,
struct pipe_resource *dst, uint64_t offset,
- unsigned size, uint32_t clear_value);
+ unsigned size, uint32_t clear_value,
+ enum r600_coherency coher);
void r600_dma_copy_buffer(struct r600_context *rctx,
struct pipe_resource *dst,
struct pipe_resource *src,
uint64_t dst_offset,
uint64_t src_offset,
uint64_t size);
+void evergreen_setup_tess_constants(struct r600_context *rctx,
+ const struct pipe_draw_info *info,
+ unsigned *num_patches);
+uint32_t evergreen_get_ls_hs_config(struct r600_context *rctx,
+ const struct pipe_draw_info *info,
+ unsigned num_patches);
+void evergreen_set_ls_hs_config(struct r600_context *rctx,
+ struct radeon_winsys_cs *cs,
+ uint32_t ls_hs_config);
+void evergreen_set_lds_alloc(struct r600_context *rctx,
+ struct radeon_winsys_cs *cs,
+ uint32_t lds_alloc);
/* r600_state_common.c */
void r600_init_common_state_functions(struct r600_context *rctx);
void r600_emit_vgt_state(struct r600_context *rctx, struct r600_atom *atom);
void r600_emit_clip_misc_state(struct r600_context *rctx, struct r600_atom *atom);
void r600_emit_stencil_ref(struct r600_context *rctx, struct r600_atom *atom);
-void r600_emit_viewport_state(struct r600_context *rctx, struct r600_atom *atom);
void r600_emit_shader(struct r600_context *rctx, struct r600_atom *a);
+void r600_add_atom(struct r600_context *rctx, struct r600_atom *atom, unsigned id);
void r600_init_atom(struct r600_context *rctx, struct r600_atom *atom, unsigned id,
void (*emit)(struct r600_context *ctx, struct r600_atom *state),
unsigned num_dw);
uint32_t r600_translate_stencil_op(int s_op);
uint32_t r600_translate_fill(uint32_t func);
unsigned r600_tex_wrap(unsigned wrap);
-unsigned r600_tex_filter(unsigned filter);
unsigned r600_tex_mipfilter(unsigned filter);
unsigned r600_tex_compare(unsigned compare);
bool sampler_state_needs_border_color(const struct pipe_sampler_state *state);
boolean vtx);
uint32_t r600_translate_texformat(struct pipe_screen *screen, enum pipe_format format,
const unsigned char *swizzle_view,
- uint32_t *word4_p, uint32_t *yuv_format_p);
-uint32_t r600_translate_colorformat(enum chip_class chip, enum pipe_format format);
-uint32_t r600_colorformat_endian_swap(uint32_t colorformat);
+ uint32_t *word4_p, uint32_t *yuv_format_p,
+ bool do_endian_swap);
+uint32_t r600_translate_colorformat(enum chip_class chip, enum pipe_format format,
+ bool do_endian_swap);
+uint32_t r600_colorformat_endian_swap(uint32_t colorformat, bool do_endian_swap);
/* r600_uvd.c */
struct pipe_video_codec *r600_uvd_create_decoder(struct pipe_context *context,
#define R600_LOOP_CONST_OFFSET 0X0003E200
#define EG_LOOP_CONST_OFFSET 0x0003A200
-#define PKT_TYPE_S(x) (((x) & 0x3) << 30)
-#define PKT_COUNT_S(x) (((x) & 0x3FFF) << 16)
-#define PKT3_IT_OPCODE_S(x) (((x) & 0xFF) << 8)
+#define PKT_TYPE_S(x) (((unsigned)(x) & 0x3) << 30)
+#define PKT_COUNT_S(x) (((unsigned)(x) & 0x3FFF) << 16)
+#define PKT3_IT_OPCODE_S(x) (((unsigned)(x) & 0xFF) << 8)
#define PKT3_PREDICATE(x) (((x) >> 0) & 0x1)
#define PKT3(op, count, predicate) (PKT_TYPE_S(3) | PKT_COUNT_S(count) | PKT3_IT_OPCODE_S(op) | PKT3_PREDICATE(predicate))
void r600_init_command_buffer(struct r600_command_buffer *cb, unsigned num_dw);
void r600_release_command_buffer(struct r600_command_buffer *cb);
-static inline void r600_write_compute_context_reg_seq(struct radeon_winsys_cs *cs, unsigned reg, unsigned num)
+static inline void radeon_compute_set_context_reg_seq(struct radeon_winsys_cs *cs, unsigned reg, unsigned num)
{
- r600_write_context_reg_seq(cs, reg, num);
+ radeon_set_context_reg_seq(cs, reg, num);
/* Set the compute bit on the packet header */
- cs->buf[cs->cdw - 2] |= RADEON_CP_PACKET3_COMPUTE_MODE;
+ cs->current.buf[cs->current.cdw - 2] |= RADEON_CP_PACKET3_COMPUTE_MODE;
}
-static inline void r600_write_ctl_const_seq(struct radeon_winsys_cs *cs, unsigned reg, unsigned num)
+static inline void radeon_set_ctl_const_seq(struct radeon_winsys_cs *cs, unsigned reg, unsigned num)
{
assert(reg >= R600_CTL_CONST_OFFSET);
- assert(cs->cdw+2+num <= cs->max_dw);
- cs->buf[cs->cdw++] = PKT3(PKT3_SET_CTL_CONST, num, 0);
- cs->buf[cs->cdw++] = (reg - R600_CTL_CONST_OFFSET) >> 2;
+ assert(cs->current.cdw + 2 + num <= cs->current.max_dw);
+ radeon_emit(cs, PKT3(PKT3_SET_CTL_CONST, num, 0));
+ radeon_emit(cs, (reg - R600_CTL_CONST_OFFSET) >> 2);
}
-static inline void r600_write_compute_context_reg(struct radeon_winsys_cs *cs, unsigned reg, unsigned value)
+static inline void radeon_compute_set_context_reg(struct radeon_winsys_cs *cs, unsigned reg, unsigned value)
{
- r600_write_compute_context_reg_seq(cs, reg, 1);
+ radeon_compute_set_context_reg_seq(cs, reg, 1);
radeon_emit(cs, value);
}
-static inline void r600_write_context_reg_flag(struct radeon_winsys_cs *cs, unsigned reg, unsigned value, unsigned flag)
+static inline void radeon_set_context_reg_flag(struct radeon_winsys_cs *cs, unsigned reg, unsigned value, unsigned flag)
{
if (flag & RADEON_CP_PACKET3_COMPUTE_MODE) {
- r600_write_compute_context_reg(cs, reg, value);
+ radeon_compute_set_context_reg(cs, reg, value);
} else {
- r600_write_context_reg(cs, reg, value);
+ radeon_set_context_reg(cs, reg, value);
}
}
-static inline void r600_write_ctl_const(struct radeon_winsys_cs *cs, unsigned reg, unsigned value)
+static inline void radeon_set_ctl_const(struct radeon_winsys_cs *cs, unsigned reg, unsigned value)
{
- r600_write_ctl_const_seq(cs, reg, 1);
+ radeon_set_ctl_const_seq(cs, reg, 1);
radeon_emit(cs, value);
}
{
return value * (1 << frac_bits);
}
-#define ALIGN_DIVUP(x, y) (((x) + (y) - 1) / (y))
/* 12.4 fixed-point */
static inline unsigned r600_pack_float_12p4(float x)
rtex->resource.b.b.format == PIPE_FORMAT_Z32_FLOAT);
}
+static inline unsigned r600_get_flush_flags(enum r600_coherency coher)
+{
+ switch (coher) {
+ default:
+ case R600_COHERENCY_NONE:
+ return 0;
+ case R600_COHERENCY_SHADER:
+ return R600_CONTEXT_INV_CONST_CACHE |
+ R600_CONTEXT_INV_VERTEX_CACHE |
+ R600_CONTEXT_INV_TEX_CACHE |
+ R600_CONTEXT_STREAMOUT_FLUSH;
+ case R600_COHERENCY_CB_META:
+ return R600_CONTEXT_FLUSH_AND_INV_CB |
+ R600_CONTEXT_FLUSH_AND_INV_CB_META;
+ }
+}
+
#define V_028A6C_OUTPRIM_TYPE_POINTLIST 0
#define V_028A6C_OUTPRIM_TYPE_LINESTRIP 1
#define V_028A6C_OUTPRIM_TYPE_TRISTRIP 2
-static inline unsigned r600_conv_prim_to_gs_out(unsigned mode)
-{
- static const int prim_conv[] = {
- V_028A6C_OUTPRIM_TYPE_POINTLIST,
- V_028A6C_OUTPRIM_TYPE_LINESTRIP,
- V_028A6C_OUTPRIM_TYPE_LINESTRIP,
- V_028A6C_OUTPRIM_TYPE_LINESTRIP,
- V_028A6C_OUTPRIM_TYPE_TRISTRIP,
- V_028A6C_OUTPRIM_TYPE_TRISTRIP,
- V_028A6C_OUTPRIM_TYPE_TRISTRIP,
- V_028A6C_OUTPRIM_TYPE_TRISTRIP,
- V_028A6C_OUTPRIM_TYPE_TRISTRIP,
- V_028A6C_OUTPRIM_TYPE_TRISTRIP,
- V_028A6C_OUTPRIM_TYPE_LINESTRIP,
- V_028A6C_OUTPRIM_TYPE_LINESTRIP,
- V_028A6C_OUTPRIM_TYPE_TRISTRIP,
- V_028A6C_OUTPRIM_TYPE_TRISTRIP,
- V_028A6C_OUTPRIM_TYPE_TRISTRIP
- };
- assert(mode < Elements(prim_conv));
-
- return prim_conv[mode];
-}
-
+unsigned r600_conv_prim_to_gs_out(unsigned mode);
#endif
projects / mesa.git / blobdiff