2 * Copyright © 2010 - 2015 Intel Corporation
4 * Permission is hereby granted, free of charge, to any person obtaining a
5 * copy of this software and associated documentation files (the "Software"),
6 * to deal in the Software without restriction, including without limitation
7 * the rights to use, copy, modify, merge, publish, distribute, sublicense,
8 * and/or sell copies of the Software, and to permit persons to whom the
9 * Software is furnished to do so, subject to the following conditions:
11 * The above copyright notice and this permission notice (including the next
12 * paragraph) shall be included in all copies or substantial portions of the
15 * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
16 * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
17 * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
18 * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
19 * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
20 * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
27 #include "brw_device_info.h"
28 #include "main/mtypes.h"
36 struct brw_geometry_program
;
37 union gl_constant_value
;
40 const struct brw_device_info
*devinfo
;
46 * Array of the ra classes for the unaligned contiguous register
52 * Mapping for register-allocated objects in *regs to the first
53 * GRF for that object.
55 uint8_t *ra_reg_to_grf
;
62 * Array of the ra classes for the unaligned contiguous register
63 * block sizes used, indexed by register size.
68 * Mapping from classes to ra_reg ranges. Each of the per-size
69 * classes corresponds to a range of ra_reg nodes. This array stores
70 * those ranges in the form of first ra_reg in each class and the
71 * total number of ra_reg elements in the last array element. This
72 * way the range of the i'th class is given by:
73 * [ class_to_ra_reg_range[i], class_to_ra_reg_range[i+1] )
75 int class_to_ra_reg_range
[17];
78 * Mapping for register-allocated objects in *regs to the first
79 * GRF for that object.
81 uint8_t *ra_reg_to_grf
;
84 * ra class for the aligned pairs we use for PLN, which doesn't
87 int aligned_pairs_class
;
90 void (*shader_debug_log
)(void *, const char *str
, ...) PRINTFLIKE(2, 3);
91 void (*shader_perf_log
)(void *, const char *str
, ...) PRINTFLIKE(2, 3);
93 bool scalar_stage
[MESA_SHADER_STAGES
];
94 struct gl_shader_compiler_options glsl_compiler_options
[MESA_SHADER_STAGES
];
99 * Program key structures.
101 * When drawing, we look for the currently bound shaders in the program
102 * cache. This is essentially a hash table lookup, and these are the keys.
104 * Sometimes OpenGL features specified as state need to be simulated via
105 * shader code, due to a mismatch between the API and the hardware. This
106 * is often referred to as "non-orthagonal state" or "NOS". We store NOS
107 * in the program key so it's considered when searching for a program. If
108 * we haven't seen a particular combination before, we have to recompile a
109 * new specialized version.
111 * Shader compilation should not look up state in gl_context directly, but
112 * instead use the copy in the program key. This guarantees recompiles will
118 enum PACKED gen6_gather_sampler_wa
{
119 WA_SIGN
= 1, /* whether we need to sign extend */
120 WA_8BIT
= 2, /* if we have an 8bit format needing wa */
121 WA_16BIT
= 4, /* if we have a 16bit format needing wa */
125 * Sampler information needed by VS, WM, and GS program cache keys.
127 struct brw_sampler_prog_key_data
{
129 * EXT_texture_swizzle and DEPTH_TEXTURE_MODE swizzles.
131 uint16_t swizzles
[MAX_SAMPLERS
];
133 uint32_t gl_clamp_mask
[3];
136 * For RG32F, gather4's channel select is broken.
138 uint32_t gather_channel_quirk_mask
;
141 * Whether this sampler uses the compressed multisample surface layout.
143 uint32_t compressed_multisample_layout_mask
;
146 * Whether this sampler is using 16x multisampling. If so fetching from
147 * this sampler will be handled with a different instruction, ld2dms_w
153 * For Sandybridge, which shader w/a we need for gather quirks.
155 enum gen6_gather_sampler_wa gen6_gather_wa
[MAX_SAMPLERS
];
159 /** The program key for Vertex Shaders. */
160 struct brw_vs_prog_key
{
161 unsigned program_string_id
;
164 * Per-attribute workaround flags
166 uint8_t gl_attrib_wa_flags
[VERT_ATTRIB_MAX
];
168 bool copy_edgeflag
:1;
170 bool clamp_vertex_color
:1;
173 * How many user clipping planes are being uploaded to the vertex shader as
176 * These are used for lowering legacy gl_ClipVertex/gl_Position clipping to
179 unsigned nr_userclip_plane_consts
:4;
182 * For pre-Gen6 hardware, a bitfield indicating which texture coordinates
183 * are going to be replaced with point coordinates (as a consequence of a
184 * call to glTexEnvi(GL_POINT_SPRITE, GL_COORD_REPLACE, GL_TRUE)). Because
185 * our SF thread requires exact matching between VS outputs and FS inputs,
186 * these texture coordinates will need to be unconditionally included in
187 * the VUE, even if they aren't written by the vertex shader.
189 uint8_t point_coord_replace
;
191 struct brw_sampler_prog_key_data tex
;
194 /** The program key for Tessellation Control Shaders. */
195 struct brw_tcs_prog_key
197 unsigned program_string_id
;
199 GLenum tes_primitive_mode
;
201 unsigned input_vertices
;
203 /** A bitfield of per-patch outputs written. */
204 uint32_t patch_outputs_written
;
206 /** A bitfield of per-vertex outputs written. */
207 uint64_t outputs_written
;
209 struct brw_sampler_prog_key_data tex
;
212 /** The program key for Tessellation Evaluation Shaders. */
213 struct brw_tes_prog_key
215 unsigned program_string_id
;
217 /** A bitfield of per-patch inputs read. */
218 uint32_t patch_inputs_read
;
220 /** A bitfield of per-vertex inputs read. */
221 uint64_t inputs_read
;
223 struct brw_sampler_prog_key_data tex
;
226 /** The program key for Geometry Shaders. */
227 struct brw_gs_prog_key
229 unsigned program_string_id
;
231 struct brw_sampler_prog_key_data tex
;
234 /** The program key for Fragment/Pixel Shaders. */
235 struct brw_wm_prog_key
{
239 bool persample_shading
:1;
241 unsigned nr_color_regions
:5;
242 bool replicate_alpha
:1;
243 bool render_to_fbo
:1;
244 bool clamp_fragment_color
:1;
245 bool compute_pos_offset
:1;
246 bool compute_sample_id
:1;
248 bool high_quality_derivatives
:1;
250 uint16_t drawable_height
;
251 uint64_t input_slots_valid
;
252 unsigned program_string_id
;
253 GLenum alpha_test_func
; /* < For Gen4/5 MRT alpha test */
254 float alpha_test_ref
;
256 struct brw_sampler_prog_key_data tex
;
259 struct brw_cs_prog_key
{
260 uint32_t program_string_id
;
261 struct brw_sampler_prog_key_data tex
;
265 * Image metadata structure as laid out in the shader parameter
266 * buffer. Entries have to be 16B-aligned for the vec4 back-end to be
267 * able to use them. That's okay because the padding and any unused
268 * entries [most of them except when we're doing untyped surface
269 * access] will be removed by the uniform packing pass.
271 #define BRW_IMAGE_PARAM_SURFACE_IDX_OFFSET 0
272 #define BRW_IMAGE_PARAM_OFFSET_OFFSET 4
273 #define BRW_IMAGE_PARAM_SIZE_OFFSET 8
274 #define BRW_IMAGE_PARAM_STRIDE_OFFSET 12
275 #define BRW_IMAGE_PARAM_TILING_OFFSET 16
276 #define BRW_IMAGE_PARAM_SWIZZLING_OFFSET 20
277 #define BRW_IMAGE_PARAM_SIZE 24
279 struct brw_image_param
{
280 /** Surface binding table index. */
281 uint32_t surface_idx
;
283 /** Offset applied to the X and Y surface coordinates. */
286 /** Surface X, Y and Z dimensions. */
289 /** X-stride in bytes, Y-stride in pixels, horizontal slice stride in
290 * pixels, vertical slice stride in pixels.
294 /** Log2 of the tiling modulus in the X, Y and Z dimension. */
298 * Right shift to apply for bit 6 address swizzling. Two different
299 * swizzles can be specified and will be applied one after the other. The
300 * resulting address will be:
302 * addr' = addr ^ ((1 << 6) & ((addr >> swizzling[0]) ^
303 * (addr >> swizzling[1])))
305 * Use \c 0xff if any of the swizzles is not required.
307 uint32_t swizzling
[2];
310 struct brw_stage_prog_data
{
312 /** size of our binding table. */
316 * surface indices for the various groups of surfaces
318 uint32_t pull_constants_start
;
319 uint32_t texture_start
;
320 uint32_t gather_texture_start
;
324 uint32_t image_start
;
325 uint32_t shader_time_start
;
329 GLuint nr_params
; /**< number of float params/constants */
330 GLuint nr_pull_params
;
331 unsigned nr_image_params
;
333 unsigned curb_read_length
;
334 unsigned total_scratch
;
335 unsigned total_shared
;
338 * Register where the thread expects to find input data from the URB
339 * (typically uniforms, followed by vertex or fragment attributes).
341 unsigned dispatch_grf_start_reg
;
343 bool use_alt_mode
; /**< Use ALT floating point mode? Otherwise, IEEE. */
345 /* Pointers to tracked values (only valid once
346 * _mesa_load_state_parameters has been called at runtime).
348 const union gl_constant_value
**param
;
349 const union gl_constant_value
**pull_param
;
351 /** Image metadata passed to the shader as uniforms. */
352 struct brw_image_param
*image_param
;
355 /* Data about a particular attempt to compile a program. Note that
356 * there can be many of these, each in a different GL state
357 * corresponding to a different brw_wm_prog_key struct, with different
360 struct brw_wm_prog_data
{
361 struct brw_stage_prog_data base
;
363 GLuint num_varying_inputs
;
365 GLuint dispatch_grf_start_reg_16
;
367 GLuint reg_blocks_16
;
371 * surface indices the WM-specific surfaces
373 uint32_t render_target_start
;
377 uint8_t computed_depth_mode
;
378 bool computed_stencil
;
380 bool early_fragment_tests
;
383 bool uses_pos_offset
;
387 uint32_t prog_offset_16
;
390 * Mask of which interpolation modes are required by the fragment shader.
391 * Used in hardware setup on gen6+.
393 uint32_t barycentric_interp_modes
;
396 * Map from gl_varying_slot to the position within the FS setup data
397 * payload where the varying's attribute vertex deltas should be delivered.
398 * For varying slots that are not used by the FS, the value is -1.
400 int urb_setup
[VARYING_SLOT_MAX
];
403 struct brw_cs_prog_data
{
404 struct brw_stage_prog_data base
;
406 GLuint dispatch_grf_start_reg_16
;
407 unsigned local_size
[3];
410 bool uses_num_work_groups
;
411 unsigned local_invocation_id_regs
;
415 * surface indices the CS-specific surfaces
417 uint32_t work_groups_start
;
423 * Enum representing the i965-specific vertex results that don't correspond
424 * exactly to any element of gl_varying_slot. The values of this enum are
425 * assigned such that they don't conflict with gl_varying_slot.
429 BRW_VARYING_SLOT_NDC
= VARYING_SLOT_MAX
,
430 BRW_VARYING_SLOT_PAD
,
432 * Technically this is not a varying but just a placeholder that
433 * compile_sf_prog() inserts into its VUE map to cause the gl_PointCoord
434 * builtin variable to be compiled correctly. see compile_sf_prog() for
437 BRW_VARYING_SLOT_PNTC
,
438 BRW_VARYING_SLOT_COUNT
442 * Data structure recording the relationship between the gl_varying_slot enum
443 * and "slots" within the vertex URB entry (VUE). A "slot" is defined as a
444 * single octaword within the VUE (128 bits).
446 * Note that each BRW register contains 256 bits (2 octawords), so when
447 * accessing the VUE in URB_NOSWIZZLE mode, each register corresponds to two
448 * consecutive VUE slots. When accessing the VUE in URB_INTERLEAVED mode (as
449 * in a vertex shader), each register corresponds to a single VUE slot, since
450 * it contains data for two separate vertices.
454 * Bitfield representing all varying slots that are (a) stored in this VUE
455 * map, and (b) actually written by the shader. Does not include any of
456 * the additional varying slots defined in brw_varying_slot.
458 GLbitfield64 slots_valid
;
461 * Is this VUE map for a separate shader pipeline?
463 * Separable programs (GL_ARB_separate_shader_objects) can be mixed and matched
464 * without the linker having a chance to dead code eliminate unused varyings.
466 * This means that we have to use a fixed slot layout, based on the output's
467 * location field, rather than assigning slots in a compact contiguous block.
472 * Map from gl_varying_slot value to VUE slot. For gl_varying_slots that are
473 * not stored in a slot (because they are not written, or because
474 * additional processing is applied before storing them in the VUE), the
477 signed char varying_to_slot
[VARYING_SLOT_TESS_MAX
];
480 * Map from VUE slot to gl_varying_slot value. For slots that do not
481 * directly correspond to a gl_varying_slot, the value comes from
484 * For slots that are not in use, the value is BRW_VARYING_SLOT_PAD.
486 signed char slot_to_varying
[VARYING_SLOT_TESS_MAX
];
489 * Total number of VUE slots in use
494 * Number of per-patch VUE slots. Only valid for tessellation control
495 * shader outputs and tessellation evaluation shader inputs.
497 int num_per_patch_slots
;
500 * Number of per-vertex VUE slots. Only valid for tessellation control
501 * shader outputs and tessellation evaluation shader inputs.
503 int num_per_vertex_slots
;
506 void brw_print_vue_map(FILE *fp
, const struct brw_vue_map
*vue_map
);
509 * Convert a VUE slot number into a byte offset within the VUE.
511 static inline GLuint
brw_vue_slot_to_offset(GLuint slot
)
517 * Convert a vertex output (brw_varying_slot) into a byte offset within the
521 GLuint
brw_varying_to_offset(const struct brw_vue_map
*vue_map
, GLuint varying
)
523 return brw_vue_slot_to_offset(vue_map
->varying_to_slot
[varying
]);
526 void brw_compute_vue_map(const struct brw_device_info
*devinfo
,
527 struct brw_vue_map
*vue_map
,
528 GLbitfield64 slots_valid
,
529 bool separate_shader
);
531 void brw_compute_tess_vue_map(struct brw_vue_map
*const vue_map
,
532 const GLbitfield64 slots_valid
,
533 const GLbitfield is_patch
);
535 enum shader_dispatch_mode
{
536 DISPATCH_MODE_4X1_SINGLE
= 0,
537 DISPATCH_MODE_4X2_DUAL_INSTANCE
= 1,
538 DISPATCH_MODE_4X2_DUAL_OBJECT
= 2,
539 DISPATCH_MODE_SIMD8
= 3,
543 * @defgroup Tessellator parameter enumerations.
545 * These correspond to the hardware values in 3DSTATE_TE, and are provided
546 * as part of the tessellation evaluation shader.
550 enum brw_tess_partitioning
{
551 BRW_TESS_PARTITIONING_INTEGER
= 0,
552 BRW_TESS_PARTITIONING_ODD_FRACTIONAL
= 1,
553 BRW_TESS_PARTITIONING_EVEN_FRACTIONAL
= 2,
556 enum brw_tess_output_topology
{
557 BRW_TESS_OUTPUT_TOPOLOGY_POINT
= 0,
558 BRW_TESS_OUTPUT_TOPOLOGY_LINE
= 1,
559 BRW_TESS_OUTPUT_TOPOLOGY_TRI_CW
= 2,
560 BRW_TESS_OUTPUT_TOPOLOGY_TRI_CCW
= 3,
563 enum brw_tess_domain
{
564 BRW_TESS_DOMAIN_QUAD
= 0,
565 BRW_TESS_DOMAIN_TRI
= 1,
566 BRW_TESS_DOMAIN_ISOLINE
= 2,
570 struct brw_vue_prog_data
{
571 struct brw_stage_prog_data base
;
572 struct brw_vue_map vue_map
;
574 /** Should the hardware deliver input VUE handles for URB pull loads? */
575 bool include_vue_handles
;
577 GLuint urb_read_length
;
580 /* Used for calculating urb partitions. In the VS, this is the size of the
581 * URB entry used for both input and output to the thread. In the GS, this
582 * is the size of the URB entry used for output.
584 GLuint urb_entry_size
;
586 enum shader_dispatch_mode dispatch_mode
;
589 struct brw_vs_prog_data
{
590 struct brw_vue_prog_data base
;
592 GLbitfield64 inputs_read
;
594 unsigned nr_attributes
;
597 bool uses_instanceid
;
598 bool uses_basevertex
;
599 bool uses_baseinstance
;
603 struct brw_tcs_prog_data
605 struct brw_vue_prog_data base
;
607 /** Number vertices in output patch */
612 struct brw_tes_prog_data
614 struct brw_vue_prog_data base
;
616 enum brw_tess_partitioning partitioning
;
617 enum brw_tess_output_topology output_topology
;
618 enum brw_tess_domain domain
;
621 struct brw_gs_prog_data
623 struct brw_vue_prog_data base
;
626 * Size of an output vertex, measured in HWORDS (32 bytes).
628 unsigned output_vertex_size_hwords
;
630 unsigned output_topology
;
633 * Size of the control data (cut bits or StreamID bits), in hwords (32
634 * bytes). 0 if there is no control data.
636 unsigned control_data_header_size_hwords
;
639 * Format of the control data (either GEN7_GS_CONTROL_DATA_FORMAT_GSCTL_SID
640 * if the control data is StreamID bits, or
641 * GEN7_GS_CONTROL_DATA_FORMAT_GSCTL_CUT if the control data is cut bits).
642 * Ignored if control_data_header_size is 0.
644 unsigned control_data_format
;
646 bool include_primitive_id
;
649 * The number of vertices emitted, if constant - otherwise -1.
651 int static_vertex_count
;
656 * Gen6 transform feedback enabled flag.
658 bool gen6_xfb_enabled
;
661 * Gen6: Provoking vertex convention for odd-numbered triangles
667 * Gen6: Number of varyings that are output to transform feedback.
669 GLuint num_transform_feedback_bindings
:7; /* 0-BRW_MAX_SOL_BINDINGS */
672 * Gen6: Map from the index of a transform feedback binding table entry to the
673 * gl_varying_slot that should be streamed out through that binding table
676 unsigned char transform_feedback_bindings
[64 /* BRW_MAX_SOL_BINDINGS */];
679 * Gen6: Map from the index of a transform feedback binding table entry to the
680 * swizzles that should be used when streaming out data through that
681 * binding table entry.
683 unsigned char transform_feedback_swizzles
[64 /* BRW_MAX_SOL_BINDINGS */];
690 * Compile a vertex shader.
692 * Returns the final assembly and the program's size.
695 brw_compile_vs(const struct brw_compiler
*compiler
, void *log_data
,
697 const struct brw_vs_prog_key
*key
,
698 struct brw_vs_prog_data
*prog_data
,
699 const struct nir_shader
*shader
,
700 gl_clip_plane
*clip_planes
,
701 bool use_legacy_snorm_formula
,
702 int shader_time_index
,
703 unsigned *final_assembly_size
,
707 * Compile a tessellation control shader.
709 * Returns the final assembly and the program's size.
712 brw_compile_tcs(const struct brw_compiler
*compiler
,
715 const struct brw_tcs_prog_key
*key
,
716 struct brw_tcs_prog_data
*prog_data
,
717 const struct nir_shader
*nir
,
718 int shader_time_index
,
719 unsigned *final_assembly_size
,
723 * Compile a tessellation evaluation shader.
725 * Returns the final assembly and the program's size.
728 brw_compile_tes(const struct brw_compiler
*compiler
, void *log_data
,
730 const struct brw_tes_prog_key
*key
,
731 struct brw_tes_prog_data
*prog_data
,
732 const struct nir_shader
*shader
,
733 struct gl_shader_program
*shader_prog
,
734 int shader_time_index
,
735 unsigned *final_assembly_size
,
739 * Compile a vertex shader.
741 * Returns the final assembly and the program's size.
744 brw_compile_gs(const struct brw_compiler
*compiler
, void *log_data
,
746 const struct brw_gs_prog_key
*key
,
747 struct brw_gs_prog_data
*prog_data
,
748 const struct nir_shader
*shader
,
749 struct gl_shader_program
*shader_prog
,
750 int shader_time_index
,
751 unsigned *final_assembly_size
,
755 * Compile a fragment shader.
757 * Returns the final assembly and the program's size.
760 brw_compile_fs(const struct brw_compiler
*compiler
, void *log_data
,
762 const struct brw_wm_prog_key
*key
,
763 struct brw_wm_prog_data
*prog_data
,
764 const struct nir_shader
*shader
,
765 struct gl_program
*prog
,
766 int shader_time_index8
,
767 int shader_time_index16
,
769 unsigned *final_assembly_size
,
773 * Compile a compute shader.
775 * Returns the final assembly and the program's size.
778 brw_compile_cs(const struct brw_compiler
*compiler
, void *log_data
,
780 const struct brw_cs_prog_key
*key
,
781 struct brw_cs_prog_data
*prog_data
,
782 const struct nir_shader
*shader
,
783 int shader_time_index
,
784 unsigned *final_assembly_size
,
788 * Fill out local id payload for compute shader according to cs_prog_data.
791 brw_cs_fill_local_id_payload(const struct brw_cs_prog_data
*cs_prog_data
,
792 void *buffer
, uint32_t threads
, uint32_t stride
);