4 TGSI, Tungsten Graphics Shader Infrastructure, is an intermediate language
5 for describing shaders. Since Gallium is inherently shaderful, shaders are
6 an important part of the API. TGSI is the only intermediate representation
12 All TGSI instructions, known as *opcodes*, operate on arbitrary-precision
13 floating-point four-component vectors. An opcode may have up to one
14 destination register, known as *dst*, and between zero and three source
15 registers, called *src0* through *src2*, or simply *src* if there is only
18 Some instructions, like :opcode:`I2F`, permit re-interpretation of vector
19 components as integers. Other instructions permit using registers as
20 two-component vectors with double precision; see :ref:`doubleopcodes`.
22 When an instruction has a scalar result, the result is usually copied into
23 each of the components of *dst*. When this happens, the result is said to be
24 *replicated* to *dst*. :opcode:`RCP` is one such instruction.
29 TGSI supports modifiers on inputs (as well as saturate modifier on instructions).
31 For inputs which have a floating point type, both absolute value and negation
32 modifiers are supported (with absolute value being applied first).
33 TGSI_OPCODE_MOV is considered to have float input type for applying modifiers.
35 For inputs which have signed or unsigned type only the negate modifier is
42 ^^^^^^^^^^^^^^^^^^^^^^^^^
44 These opcodes are guaranteed to be available regardless of the driver being
47 .. opcode:: ARL - Address Register Load
51 dst.x = (int) \lfloor src.x\rfloor
53 dst.y = (int) \lfloor src.y\rfloor
55 dst.z = (int) \lfloor src.z\rfloor
57 dst.w = (int) \lfloor src.w\rfloor
60 .. opcode:: MOV - Move
73 .. opcode:: LIT - Light Coefficients
78 dst.y &= max(src.x, 0) \\
79 dst.z &= (src.x > 0) ? max(src.y, 0)^{clamp(src.w, -128, 128))} : 0 \\
83 .. opcode:: RCP - Reciprocal
85 This instruction replicates its result.
92 .. opcode:: RSQ - Reciprocal Square Root
94 This instruction replicates its result. The results are undefined for src <= 0.
98 dst = \frac{1}{\sqrt{src.x}}
101 .. opcode:: SQRT - Square Root
103 This instruction replicates its result. The results are undefined for src < 0.
110 .. opcode:: EXP - Approximate Exponential Base 2
114 dst.x &= 2^{\lfloor src.x\rfloor} \\
115 dst.y &= src.x - \lfloor src.x\rfloor \\
116 dst.z &= 2^{src.x} \\
120 .. opcode:: LOG - Approximate Logarithm Base 2
124 dst.x &= \lfloor\log_2{|src.x|}\rfloor \\
125 dst.y &= \frac{|src.x|}{2^{\lfloor\log_2{|src.x|}\rfloor}} \\
126 dst.z &= \log_2{|src.x|} \\
130 .. opcode:: MUL - Multiply
134 dst.x = src0.x \times src1.x
136 dst.y = src0.y \times src1.y
138 dst.z = src0.z \times src1.z
140 dst.w = src0.w \times src1.w
143 .. opcode:: ADD - Add
147 dst.x = src0.x + src1.x
149 dst.y = src0.y + src1.y
151 dst.z = src0.z + src1.z
153 dst.w = src0.w + src1.w
156 .. opcode:: DP3 - 3-component Dot Product
158 This instruction replicates its result.
162 dst = src0.x \times src1.x + src0.y \times src1.y + src0.z \times src1.z
165 .. opcode:: DP4 - 4-component Dot Product
167 This instruction replicates its result.
171 dst = src0.x \times src1.x + src0.y \times src1.y + src0.z \times src1.z + src0.w \times src1.w
174 .. opcode:: DST - Distance Vector
179 dst.y &= src0.y \times src1.y\\
184 .. opcode:: MIN - Minimum
188 dst.x = min(src0.x, src1.x)
190 dst.y = min(src0.y, src1.y)
192 dst.z = min(src0.z, src1.z)
194 dst.w = min(src0.w, src1.w)
197 .. opcode:: MAX - Maximum
201 dst.x = max(src0.x, src1.x)
203 dst.y = max(src0.y, src1.y)
205 dst.z = max(src0.z, src1.z)
207 dst.w = max(src0.w, src1.w)
210 .. opcode:: SLT - Set On Less Than
214 dst.x = (src0.x < src1.x) ? 1.0F : 0.0F
216 dst.y = (src0.y < src1.y) ? 1.0F : 0.0F
218 dst.z = (src0.z < src1.z) ? 1.0F : 0.0F
220 dst.w = (src0.w < src1.w) ? 1.0F : 0.0F
223 .. opcode:: SGE - Set On Greater Equal Than
227 dst.x = (src0.x >= src1.x) ? 1.0F : 0.0F
229 dst.y = (src0.y >= src1.y) ? 1.0F : 0.0F
231 dst.z = (src0.z >= src1.z) ? 1.0F : 0.0F
233 dst.w = (src0.w >= src1.w) ? 1.0F : 0.0F
236 .. opcode:: MAD - Multiply And Add
240 dst.x = src0.x \times src1.x + src2.x
242 dst.y = src0.y \times src1.y + src2.y
244 dst.z = src0.z \times src1.z + src2.z
246 dst.w = src0.w \times src1.w + src2.w
249 .. opcode:: LRP - Linear Interpolate
253 dst.x = src0.x \times src1.x + (1 - src0.x) \times src2.x
255 dst.y = src0.y \times src1.y + (1 - src0.y) \times src2.y
257 dst.z = src0.z \times src1.z + (1 - src0.z) \times src2.z
259 dst.w = src0.w \times src1.w + (1 - src0.w) \times src2.w
262 .. opcode:: FMA - Fused Multiply-Add
264 Perform a * b + c with no intermediate rounding step.
268 dst.x = src0.x \times src1.x + src2.x
270 dst.y = src0.y \times src1.y + src2.y
272 dst.z = src0.z \times src1.z + src2.z
274 dst.w = src0.w \times src1.w + src2.w
277 .. opcode:: DP2A - 2-component Dot Product And Add
281 dst.x = src0.x \times src1.x + src0.y \times src1.y + src2.x
283 dst.y = src0.x \times src1.x + src0.y \times src1.y + src2.x
285 dst.z = src0.x \times src1.x + src0.y \times src1.y + src2.x
287 dst.w = src0.x \times src1.x + src0.y \times src1.y + src2.x
290 .. opcode:: FRC - Fraction
294 dst.x = src.x - \lfloor src.x\rfloor
296 dst.y = src.y - \lfloor src.y\rfloor
298 dst.z = src.z - \lfloor src.z\rfloor
300 dst.w = src.w - \lfloor src.w\rfloor
303 .. opcode:: FLR - Floor
307 dst.x = \lfloor src.x\rfloor
309 dst.y = \lfloor src.y\rfloor
311 dst.z = \lfloor src.z\rfloor
313 dst.w = \lfloor src.w\rfloor
316 .. opcode:: ROUND - Round
329 .. opcode:: EX2 - Exponential Base 2
331 This instruction replicates its result.
338 .. opcode:: LG2 - Logarithm Base 2
340 This instruction replicates its result.
347 .. opcode:: POW - Power
349 This instruction replicates its result.
353 dst = src0.x^{src1.x}
355 .. opcode:: XPD - Cross Product
359 dst.x = src0.y \times src1.z - src1.y \times src0.z
361 dst.y = src0.z \times src1.x - src1.z \times src0.x
363 dst.z = src0.x \times src1.y - src1.x \times src0.y
368 .. opcode:: DPH - Homogeneous Dot Product
370 This instruction replicates its result.
374 dst = src0.x \times src1.x + src0.y \times src1.y + src0.z \times src1.z + src1.w
377 .. opcode:: COS - Cosine
379 This instruction replicates its result.
386 .. opcode:: DDX, DDX_FINE - Derivative Relative To X
388 The fine variant is only used when ``PIPE_CAP_TGSI_FS_FINE_DERIVATIVE`` is
389 advertised. When it is, the fine version guarantees one derivative per row
390 while DDX is allowed to be the same for the entire 2x2 quad.
394 dst.x = partialx(src.x)
396 dst.y = partialx(src.y)
398 dst.z = partialx(src.z)
400 dst.w = partialx(src.w)
403 .. opcode:: DDY, DDY_FINE - Derivative Relative To Y
405 The fine variant is only used when ``PIPE_CAP_TGSI_FS_FINE_DERIVATIVE`` is
406 advertised. When it is, the fine version guarantees one derivative per column
407 while DDY is allowed to be the same for the entire 2x2 quad.
411 dst.x = partialy(src.x)
413 dst.y = partialy(src.y)
415 dst.z = partialy(src.z)
417 dst.w = partialy(src.w)
420 .. opcode:: PK2H - Pack Two 16-bit Floats
422 This instruction replicates its result.
426 dst = f32\_to\_f16(src.x) | f32\_to\_f16(src.y) << 16
429 .. opcode:: PK2US - Pack Two Unsigned 16-bit Scalars
434 .. opcode:: PK4B - Pack Four Signed 8-bit Scalars
439 .. opcode:: PK4UB - Pack Four Unsigned 8-bit Scalars
444 .. opcode:: SEQ - Set On Equal
448 dst.x = (src0.x == src1.x) ? 1.0F : 0.0F
450 dst.y = (src0.y == src1.y) ? 1.0F : 0.0F
452 dst.z = (src0.z == src1.z) ? 1.0F : 0.0F
454 dst.w = (src0.w == src1.w) ? 1.0F : 0.0F
457 .. opcode:: SGT - Set On Greater Than
461 dst.x = (src0.x > src1.x) ? 1.0F : 0.0F
463 dst.y = (src0.y > src1.y) ? 1.0F : 0.0F
465 dst.z = (src0.z > src1.z) ? 1.0F : 0.0F
467 dst.w = (src0.w > src1.w) ? 1.0F : 0.0F
470 .. opcode:: SIN - Sine
472 This instruction replicates its result.
479 .. opcode:: SLE - Set On Less Equal Than
483 dst.x = (src0.x <= src1.x) ? 1.0F : 0.0F
485 dst.y = (src0.y <= src1.y) ? 1.0F : 0.0F
487 dst.z = (src0.z <= src1.z) ? 1.0F : 0.0F
489 dst.w = (src0.w <= src1.w) ? 1.0F : 0.0F
492 .. opcode:: SNE - Set On Not Equal
496 dst.x = (src0.x != src1.x) ? 1.0F : 0.0F
498 dst.y = (src0.y != src1.y) ? 1.0F : 0.0F
500 dst.z = (src0.z != src1.z) ? 1.0F : 0.0F
502 dst.w = (src0.w != src1.w) ? 1.0F : 0.0F
505 .. opcode:: TEX - Texture Lookup
507 for array textures src0.y contains the slice for 1D,
508 and src0.z contain the slice for 2D.
510 for shadow textures with no arrays (and not cube map),
511 src0.z contains the reference value.
513 for shadow textures with arrays, src0.z contains
514 the reference value for 1D arrays, and src0.w contains
515 the reference value for 2D arrays and cube maps.
517 for cube map array shadow textures, the reference value
518 cannot be passed in src0.w, and TEX2 must be used instead.
524 shadow_ref = src0.z or src0.w (optional)
528 dst = texture\_sample(unit, coord, shadow_ref)
531 .. opcode:: TEX2 - Texture Lookup (for shadow cube map arrays only)
533 this is the same as TEX, but uses another reg to encode the
544 dst = texture\_sample(unit, coord, shadow_ref)
549 .. opcode:: TXD - Texture Lookup with Derivatives
561 dst = texture\_sample\_deriv(unit, coord, ddx, ddy)
564 .. opcode:: TXP - Projective Texture Lookup
568 coord.x = src0.x / src0.w
570 coord.y = src0.y / src0.w
572 coord.z = src0.z / src0.w
578 dst = texture\_sample(unit, coord)
581 .. opcode:: UP2H - Unpack Two 16-Bit Floats
585 dst.x = f16\_to\_f32(src0.x \& 0xffff)
587 dst.y = f16\_to\_f32(src0.x >> 16)
589 dst.z = f16\_to\_f32(src0.x \& 0xffff)
591 dst.w = f16\_to\_f32(src0.x >> 16)
595 Considered for removal.
597 .. opcode:: UP2US - Unpack Two Unsigned 16-Bit Scalars
603 Considered for removal.
605 .. opcode:: UP4B - Unpack Four Signed 8-Bit Values
611 Considered for removal.
613 .. opcode:: UP4UB - Unpack Four Unsigned 8-Bit Scalars
619 Considered for removal.
622 .. opcode:: ARR - Address Register Load With Round
626 dst.x = (int) round(src.x)
628 dst.y = (int) round(src.y)
630 dst.z = (int) round(src.z)
632 dst.w = (int) round(src.w)
635 .. opcode:: SSG - Set Sign
639 dst.x = (src.x > 0) ? 1 : (src.x < 0) ? -1 : 0
641 dst.y = (src.y > 0) ? 1 : (src.y < 0) ? -1 : 0
643 dst.z = (src.z > 0) ? 1 : (src.z < 0) ? -1 : 0
645 dst.w = (src.w > 0) ? 1 : (src.w < 0) ? -1 : 0
648 .. opcode:: CMP - Compare
652 dst.x = (src0.x < 0) ? src1.x : src2.x
654 dst.y = (src0.y < 0) ? src1.y : src2.y
656 dst.z = (src0.z < 0) ? src1.z : src2.z
658 dst.w = (src0.w < 0) ? src1.w : src2.w
661 .. opcode:: KILL_IF - Conditional Discard
663 Conditional discard. Allowed in fragment shaders only.
667 if (src.x < 0 || src.y < 0 || src.z < 0 || src.w < 0)
672 .. opcode:: KILL - Discard
674 Unconditional discard. Allowed in fragment shaders only.
677 .. opcode:: SCS - Sine Cosine
690 .. opcode:: TXB - Texture Lookup With Bias
692 for cube map array textures and shadow cube maps, the bias value
693 cannot be passed in src0.w, and TXB2 must be used instead.
695 if the target is a shadow texture, the reference value is always
696 in src.z (this prevents shadow 3d and shadow 2d arrays from
697 using this instruction, but this is not needed).
713 dst = texture\_sample(unit, coord, bias)
716 .. opcode:: TXB2 - Texture Lookup With Bias (some cube maps only)
718 this is the same as TXB, but uses another reg to encode the
719 lod bias value for cube map arrays and shadow cube maps.
720 Presumably shadow 2d arrays and shadow 3d targets could use
721 this encoding too, but this is not legal.
723 shadow cube map arrays are neither possible nor required.
733 dst = texture\_sample(unit, coord, bias)
736 .. opcode:: DIV - Divide
740 dst.x = \frac{src0.x}{src1.x}
742 dst.y = \frac{src0.y}{src1.y}
744 dst.z = \frac{src0.z}{src1.z}
746 dst.w = \frac{src0.w}{src1.w}
749 .. opcode:: DP2 - 2-component Dot Product
751 This instruction replicates its result.
755 dst = src0.x \times src1.x + src0.y \times src1.y
758 .. opcode:: TXL - Texture Lookup With explicit LOD
760 for cube map array textures, the explicit lod value
761 cannot be passed in src0.w, and TXL2 must be used instead.
763 if the target is a shadow texture, the reference value is always
764 in src.z (this prevents shadow 3d / 2d array / cube targets from
765 using this instruction, but this is not needed).
781 dst = texture\_sample(unit, coord, lod)
784 .. opcode:: TXL2 - Texture Lookup With explicit LOD (for cube map arrays only)
786 this is the same as TXL, but uses another reg to encode the
788 Presumably shadow 3d / 2d array / cube targets could use
789 this encoding too, but this is not legal.
791 shadow cube map arrays are neither possible nor required.
801 dst = texture\_sample(unit, coord, lod)
804 .. opcode:: PUSHA - Push Address Register On Stack
813 Considered for cleanup.
817 Considered for removal.
819 .. opcode:: POPA - Pop Address Register From Stack
828 Considered for cleanup.
832 Considered for removal.
835 .. opcode:: CALLNZ - Subroutine Call If Not Zero
841 Considered for cleanup.
845 Considered for removal.
849 ^^^^^^^^^^^^^^^^^^^^^^^^
851 These opcodes are primarily provided for special-use computational shaders.
852 Support for these opcodes indicated by a special pipe capability bit (TBD).
854 XXX doesn't look like most of the opcodes really belong here.
856 .. opcode:: CEIL - Ceiling
860 dst.x = \lceil src.x\rceil
862 dst.y = \lceil src.y\rceil
864 dst.z = \lceil src.z\rceil
866 dst.w = \lceil src.w\rceil
869 .. opcode:: TRUNC - Truncate
882 .. opcode:: MOD - Modulus
886 dst.x = src0.x \bmod src1.x
888 dst.y = src0.y \bmod src1.y
890 dst.z = src0.z \bmod src1.z
892 dst.w = src0.w \bmod src1.w
895 .. opcode:: UARL - Integer Address Register Load
897 Moves the contents of the source register, assumed to be an integer, into the
898 destination register, which is assumed to be an address (ADDR) register.
901 .. opcode:: SAD - Sum Of Absolute Differences
905 dst.x = |src0.x - src1.x| + src2.x
907 dst.y = |src0.y - src1.y| + src2.y
909 dst.z = |src0.z - src1.z| + src2.z
911 dst.w = |src0.w - src1.w| + src2.w
914 .. opcode:: TXF - Texel Fetch
916 As per NV_gpu_shader4, extract a single texel from a specified texture
917 image. The source sampler may not be a CUBE or SHADOW. src 0 is a
918 four-component signed integer vector used to identify the single texel
919 accessed. 3 components + level. Just like texture instructions, an optional
920 offset vector is provided, which is subject to various driver restrictions
921 (regarding range, source of offsets).
922 TXF(uint_vec coord, int_vec offset).
925 .. opcode:: TXQ - Texture Size Query
927 As per NV_gpu_program4, retrieve the dimensions of the texture depending on
928 the target. For 1D (width), 2D/RECT/CUBE (width, height), 3D (width, height,
929 depth), 1D array (width, layers), 2D array (width, height, layers).
930 Also return the number of accessible levels (last_level - first_level + 1)
933 For components which don't return a resource dimension, their value
940 dst.x = texture\_width(unit, lod)
942 dst.y = texture\_height(unit, lod)
944 dst.z = texture\_depth(unit, lod)
946 dst.w = texture\_levels(unit)
949 .. opcode:: TXQS - Texture Samples Query
951 This retrieves the number of samples in the texture, and stores it
952 into the x component. The other components are undefined.
956 dst.x = texture\_samples(unit)
959 .. opcode:: TG4 - Texture Gather
961 As per ARB_texture_gather, gathers the four texels to be used in a bi-linear
962 filtering operation and packs them into a single register. Only works with
963 2D, 2D array, cubemaps, and cubemaps arrays. For 2D textures, only the
964 addressing modes of the sampler and the top level of any mip pyramid are
965 used. Set W to zero. It behaves like the TEX instruction, but a filtered
966 sample is not generated. The four samples that contribute to filtering are
967 placed into xyzw in clockwise order, starting with the (u,v) texture
968 coordinate delta at the following locations (-, +), (+, +), (+, -), (-, -),
969 where the magnitude of the deltas are half a texel.
971 PIPE_CAP_TEXTURE_SM5 enhances this instruction to support shadow per-sample
972 depth compares, single component selection, and a non-constant offset. It
973 doesn't allow support for the GL independent offset to get i0,j0. This would
974 require another CAP is hw can do it natively. For now we lower that before
983 dst = texture\_gather4 (unit, coord, component)
985 (with SM5 - cube array shadow)
993 dst = texture\_gather (uint, coord, compare)
995 .. opcode:: LODQ - level of detail query
997 Compute the LOD information that the texture pipe would use to access the
998 texture. The Y component contains the computed LOD lambda_prime. The X
999 component contains the LOD that will be accessed, based on min/max lod's
1006 dst.xy = lodq(uint, coord);
1009 ^^^^^^^^^^^^^^^^^^^^^^^^
1010 These opcodes are used for integer operations.
1011 Support for these opcodes indicated by PIPE_SHADER_CAP_INTEGERS (all of them?)
1014 .. opcode:: I2F - Signed Integer To Float
1016 Rounding is unspecified (round to nearest even suggested).
1020 dst.x = (float) src.x
1022 dst.y = (float) src.y
1024 dst.z = (float) src.z
1026 dst.w = (float) src.w
1029 .. opcode:: U2F - Unsigned Integer To Float
1031 Rounding is unspecified (round to nearest even suggested).
1035 dst.x = (float) src.x
1037 dst.y = (float) src.y
1039 dst.z = (float) src.z
1041 dst.w = (float) src.w
1044 .. opcode:: F2I - Float to Signed Integer
1046 Rounding is towards zero (truncate).
1047 Values outside signed range (including NaNs) produce undefined results.
1060 .. opcode:: F2U - Float to Unsigned Integer
1062 Rounding is towards zero (truncate).
1063 Values outside unsigned range (including NaNs) produce undefined results.
1067 dst.x = (unsigned) src.x
1069 dst.y = (unsigned) src.y
1071 dst.z = (unsigned) src.z
1073 dst.w = (unsigned) src.w
1076 .. opcode:: UADD - Integer Add
1078 This instruction works the same for signed and unsigned integers.
1079 The low 32bit of the result is returned.
1083 dst.x = src0.x + src1.x
1085 dst.y = src0.y + src1.y
1087 dst.z = src0.z + src1.z
1089 dst.w = src0.w + src1.w
1092 .. opcode:: UMAD - Integer Multiply And Add
1094 This instruction works the same for signed and unsigned integers.
1095 The multiplication returns the low 32bit (as does the result itself).
1099 dst.x = src0.x \times src1.x + src2.x
1101 dst.y = src0.y \times src1.y + src2.y
1103 dst.z = src0.z \times src1.z + src2.z
1105 dst.w = src0.w \times src1.w + src2.w
1108 .. opcode:: UMUL - Integer Multiply
1110 This instruction works the same for signed and unsigned integers.
1111 The low 32bit of the result is returned.
1115 dst.x = src0.x \times src1.x
1117 dst.y = src0.y \times src1.y
1119 dst.z = src0.z \times src1.z
1121 dst.w = src0.w \times src1.w
1124 .. opcode:: IMUL_HI - Signed Integer Multiply High Bits
1126 The high 32bits of the multiplication of 2 signed integers are returned.
1130 dst.x = (src0.x \times src1.x) >> 32
1132 dst.y = (src0.y \times src1.y) >> 32
1134 dst.z = (src0.z \times src1.z) >> 32
1136 dst.w = (src0.w \times src1.w) >> 32
1139 .. opcode:: UMUL_HI - Unsigned Integer Multiply High Bits
1141 The high 32bits of the multiplication of 2 unsigned integers are returned.
1145 dst.x = (src0.x \times src1.x) >> 32
1147 dst.y = (src0.y \times src1.y) >> 32
1149 dst.z = (src0.z \times src1.z) >> 32
1151 dst.w = (src0.w \times src1.w) >> 32
1154 .. opcode:: IDIV - Signed Integer Division
1156 TBD: behavior for division by zero.
1160 dst.x = src0.x \ src1.x
1162 dst.y = src0.y \ src1.y
1164 dst.z = src0.z \ src1.z
1166 dst.w = src0.w \ src1.w
1169 .. opcode:: UDIV - Unsigned Integer Division
1171 For division by zero, 0xffffffff is returned.
1175 dst.x = src0.x \ src1.x
1177 dst.y = src0.y \ src1.y
1179 dst.z = src0.z \ src1.z
1181 dst.w = src0.w \ src1.w
1184 .. opcode:: UMOD - Unsigned Integer Remainder
1186 If second arg is zero, 0xffffffff is returned.
1190 dst.x = src0.x \ src1.x
1192 dst.y = src0.y \ src1.y
1194 dst.z = src0.z \ src1.z
1196 dst.w = src0.w \ src1.w
1199 .. opcode:: NOT - Bitwise Not
1212 .. opcode:: AND - Bitwise And
1216 dst.x = src0.x \& src1.x
1218 dst.y = src0.y \& src1.y
1220 dst.z = src0.z \& src1.z
1222 dst.w = src0.w \& src1.w
1225 .. opcode:: OR - Bitwise Or
1229 dst.x = src0.x | src1.x
1231 dst.y = src0.y | src1.y
1233 dst.z = src0.z | src1.z
1235 dst.w = src0.w | src1.w
1238 .. opcode:: XOR - Bitwise Xor
1242 dst.x = src0.x \oplus src1.x
1244 dst.y = src0.y \oplus src1.y
1246 dst.z = src0.z \oplus src1.z
1248 dst.w = src0.w \oplus src1.w
1251 .. opcode:: IMAX - Maximum of Signed Integers
1255 dst.x = max(src0.x, src1.x)
1257 dst.y = max(src0.y, src1.y)
1259 dst.z = max(src0.z, src1.z)
1261 dst.w = max(src0.w, src1.w)
1264 .. opcode:: UMAX - Maximum of Unsigned Integers
1268 dst.x = max(src0.x, src1.x)
1270 dst.y = max(src0.y, src1.y)
1272 dst.z = max(src0.z, src1.z)
1274 dst.w = max(src0.w, src1.w)
1277 .. opcode:: IMIN - Minimum of Signed Integers
1281 dst.x = min(src0.x, src1.x)
1283 dst.y = min(src0.y, src1.y)
1285 dst.z = min(src0.z, src1.z)
1287 dst.w = min(src0.w, src1.w)
1290 .. opcode:: UMIN - Minimum of Unsigned Integers
1294 dst.x = min(src0.x, src1.x)
1296 dst.y = min(src0.y, src1.y)
1298 dst.z = min(src0.z, src1.z)
1300 dst.w = min(src0.w, src1.w)
1303 .. opcode:: SHL - Shift Left
1305 The shift count is masked with 0x1f before the shift is applied.
1309 dst.x = src0.x << (0x1f \& src1.x)
1311 dst.y = src0.y << (0x1f \& src1.y)
1313 dst.z = src0.z << (0x1f \& src1.z)
1315 dst.w = src0.w << (0x1f \& src1.w)
1318 .. opcode:: ISHR - Arithmetic Shift Right (of Signed Integer)
1320 The shift count is masked with 0x1f before the shift is applied.
1324 dst.x = src0.x >> (0x1f \& src1.x)
1326 dst.y = src0.y >> (0x1f \& src1.y)
1328 dst.z = src0.z >> (0x1f \& src1.z)
1330 dst.w = src0.w >> (0x1f \& src1.w)
1333 .. opcode:: USHR - Logical Shift Right
1335 The shift count is masked with 0x1f before the shift is applied.
1339 dst.x = src0.x >> (unsigned) (0x1f \& src1.x)
1341 dst.y = src0.y >> (unsigned) (0x1f \& src1.y)
1343 dst.z = src0.z >> (unsigned) (0x1f \& src1.z)
1345 dst.w = src0.w >> (unsigned) (0x1f \& src1.w)
1348 .. opcode:: UCMP - Integer Conditional Move
1352 dst.x = src0.x ? src1.x : src2.x
1354 dst.y = src0.y ? src1.y : src2.y
1356 dst.z = src0.z ? src1.z : src2.z
1358 dst.w = src0.w ? src1.w : src2.w
1362 .. opcode:: ISSG - Integer Set Sign
1366 dst.x = (src0.x < 0) ? -1 : (src0.x > 0) ? 1 : 0
1368 dst.y = (src0.y < 0) ? -1 : (src0.y > 0) ? 1 : 0
1370 dst.z = (src0.z < 0) ? -1 : (src0.z > 0) ? 1 : 0
1372 dst.w = (src0.w < 0) ? -1 : (src0.w > 0) ? 1 : 0
1376 .. opcode:: FSLT - Float Set On Less Than (ordered)
1378 Same comparison as SLT but returns integer instead of 1.0/0.0 float
1382 dst.x = (src0.x < src1.x) ? \sim 0 : 0
1384 dst.y = (src0.y < src1.y) ? \sim 0 : 0
1386 dst.z = (src0.z < src1.z) ? \sim 0 : 0
1388 dst.w = (src0.w < src1.w) ? \sim 0 : 0
1391 .. opcode:: ISLT - Signed Integer Set On Less Than
1395 dst.x = (src0.x < src1.x) ? \sim 0 : 0
1397 dst.y = (src0.y < src1.y) ? \sim 0 : 0
1399 dst.z = (src0.z < src1.z) ? \sim 0 : 0
1401 dst.w = (src0.w < src1.w) ? \sim 0 : 0
1404 .. opcode:: USLT - Unsigned Integer Set On Less Than
1408 dst.x = (src0.x < src1.x) ? \sim 0 : 0
1410 dst.y = (src0.y < src1.y) ? \sim 0 : 0
1412 dst.z = (src0.z < src1.z) ? \sim 0 : 0
1414 dst.w = (src0.w < src1.w) ? \sim 0 : 0
1417 .. opcode:: FSGE - Float Set On Greater Equal Than (ordered)
1419 Same comparison as SGE but returns integer instead of 1.0/0.0 float
1423 dst.x = (src0.x >= src1.x) ? \sim 0 : 0
1425 dst.y = (src0.y >= src1.y) ? \sim 0 : 0
1427 dst.z = (src0.z >= src1.z) ? \sim 0 : 0
1429 dst.w = (src0.w >= src1.w) ? \sim 0 : 0
1432 .. opcode:: ISGE - Signed Integer Set On Greater Equal Than
1436 dst.x = (src0.x >= src1.x) ? \sim 0 : 0
1438 dst.y = (src0.y >= src1.y) ? \sim 0 : 0
1440 dst.z = (src0.z >= src1.z) ? \sim 0 : 0
1442 dst.w = (src0.w >= src1.w) ? \sim 0 : 0
1445 .. opcode:: USGE - Unsigned Integer Set On Greater Equal Than
1449 dst.x = (src0.x >= src1.x) ? \sim 0 : 0
1451 dst.y = (src0.y >= src1.y) ? \sim 0 : 0
1453 dst.z = (src0.z >= src1.z) ? \sim 0 : 0
1455 dst.w = (src0.w >= src1.w) ? \sim 0 : 0
1458 .. opcode:: FSEQ - Float Set On Equal (ordered)
1460 Same comparison as SEQ but returns integer instead of 1.0/0.0 float
1464 dst.x = (src0.x == src1.x) ? \sim 0 : 0
1466 dst.y = (src0.y == src1.y) ? \sim 0 : 0
1468 dst.z = (src0.z == src1.z) ? \sim 0 : 0
1470 dst.w = (src0.w == src1.w) ? \sim 0 : 0
1473 .. opcode:: USEQ - Integer Set On Equal
1477 dst.x = (src0.x == src1.x) ? \sim 0 : 0
1479 dst.y = (src0.y == src1.y) ? \sim 0 : 0
1481 dst.z = (src0.z == src1.z) ? \sim 0 : 0
1483 dst.w = (src0.w == src1.w) ? \sim 0 : 0
1486 .. opcode:: FSNE - Float Set On Not Equal (unordered)
1488 Same comparison as SNE but returns integer instead of 1.0/0.0 float
1492 dst.x = (src0.x != src1.x) ? \sim 0 : 0
1494 dst.y = (src0.y != src1.y) ? \sim 0 : 0
1496 dst.z = (src0.z != src1.z) ? \sim 0 : 0
1498 dst.w = (src0.w != src1.w) ? \sim 0 : 0
1501 .. opcode:: USNE - Integer Set On Not Equal
1505 dst.x = (src0.x != src1.x) ? \sim 0 : 0
1507 dst.y = (src0.y != src1.y) ? \sim 0 : 0
1509 dst.z = (src0.z != src1.z) ? \sim 0 : 0
1511 dst.w = (src0.w != src1.w) ? \sim 0 : 0
1514 .. opcode:: INEG - Integer Negate
1529 .. opcode:: IABS - Integer Absolute Value
1543 These opcodes are used for bit-level manipulation of integers.
1545 .. opcode:: IBFE - Signed Bitfield Extract
1547 Like GLSL bitfieldExtract. Extracts a set of bits from the input, and
1548 sign-extends them if the high bit of the extracted window is set.
1552 def ibfe(value, offset, bits):
1553 if offset < 0 or bits < 0 or offset + bits > 32:
1555 if bits == 0: return 0
1556 # Note: >> sign-extends
1557 return (value << (32 - offset - bits)) >> (32 - bits)
1559 .. opcode:: UBFE - Unsigned Bitfield Extract
1561 Like GLSL bitfieldExtract. Extracts a set of bits from the input, without
1566 def ubfe(value, offset, bits):
1567 if offset < 0 or bits < 0 or offset + bits > 32:
1569 if bits == 0: return 0
1570 # Note: >> does not sign-extend
1571 return (value << (32 - offset - bits)) >> (32 - bits)
1573 .. opcode:: BFI - Bitfield Insert
1575 Like GLSL bitfieldInsert. Replaces a bit region of 'base' with the low bits
1580 def bfi(base, insert, offset, bits):
1581 if offset < 0 or bits < 0 or offset + bits > 32:
1583 # << defined such that mask == ~0 when bits == 32, offset == 0
1584 mask = ((1 << bits) - 1) << offset
1585 return ((insert << offset) & mask) | (base & ~mask)
1587 .. opcode:: BREV - Bitfield Reverse
1589 See SM5 instruction BFREV. Reverses the bits of the argument.
1591 .. opcode:: POPC - Population Count
1593 See SM5 instruction COUNTBITS. Counts the number of set bits in the argument.
1595 .. opcode:: LSB - Index of lowest set bit
1597 See SM5 instruction FIRSTBIT_LO. Computes the 0-based index of the first set
1598 bit of the argument. Returns -1 if none are set.
1600 .. opcode:: IMSB - Index of highest non-sign bit
1602 See SM5 instruction FIRSTBIT_SHI. Computes the 0-based index of the highest
1603 non-sign bit of the argument (i.e. highest 0 bit for negative numbers,
1604 highest 1 bit for positive numbers). Returns -1 if all bits are the same
1605 (i.e. for inputs 0 and -1).
1607 .. opcode:: UMSB - Index of highest set bit
1609 See SM5 instruction FIRSTBIT_HI. Computes the 0-based index of the highest
1610 set bit of the argument. Returns -1 if none are set.
1613 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1615 These opcodes are only supported in geometry shaders; they have no meaning
1616 in any other type of shader.
1618 .. opcode:: EMIT - Emit
1620 Generate a new vertex for the current primitive into the specified vertex
1621 stream using the values in the output registers.
1624 .. opcode:: ENDPRIM - End Primitive
1626 Complete the current primitive in the specified vertex stream (consisting of
1627 the emitted vertices), and start a new one.
1633 These opcodes are part of :term:`GLSL`'s opcode set. Support for these
1634 opcodes is determined by a special capability bit, ``GLSL``.
1635 Some require glsl version 1.30 (UIF/BREAKC/SWITCH/CASE/DEFAULT/ENDSWITCH).
1637 .. opcode:: CAL - Subroutine Call
1643 .. opcode:: RET - Subroutine Call Return
1648 .. opcode:: CONT - Continue
1650 Unconditionally moves the point of execution to the instruction after the
1651 last bgnloop. The instruction must appear within a bgnloop/endloop.
1655 Support for CONT is determined by a special capability bit,
1656 ``TGSI_CONT_SUPPORTED``. See :ref:`Screen` for more information.
1659 .. opcode:: BGNLOOP - Begin a Loop
1661 Start a loop. Must have a matching endloop.
1664 .. opcode:: BGNSUB - Begin Subroutine
1666 Starts definition of a subroutine. Must have a matching endsub.
1669 .. opcode:: ENDLOOP - End a Loop
1671 End a loop started with bgnloop.
1674 .. opcode:: ENDSUB - End Subroutine
1676 Ends definition of a subroutine.
1679 .. opcode:: NOP - No Operation
1684 .. opcode:: BRK - Break
1686 Unconditionally moves the point of execution to the instruction after the
1687 next endloop or endswitch. The instruction must appear within a loop/endloop
1688 or switch/endswitch.
1691 .. opcode:: BREAKC - Break Conditional
1693 Conditionally moves the point of execution to the instruction after the
1694 next endloop or endswitch. The instruction must appear within a loop/endloop
1695 or switch/endswitch.
1696 Condition evaluates to true if src0.x != 0 where src0.x is interpreted
1697 as an integer register.
1701 Considered for removal as it's quite inconsistent wrt other opcodes
1702 (could emulate with UIF/BRK/ENDIF).
1705 .. opcode:: IF - Float If
1707 Start an IF ... ELSE .. ENDIF block. Condition evaluates to true if
1711 where src0.x is interpreted as a floating point register.
1714 .. opcode:: UIF - Bitwise If
1716 Start an UIF ... ELSE .. ENDIF block. Condition evaluates to true if
1720 where src0.x is interpreted as an integer register.
1723 .. opcode:: ELSE - Else
1725 Starts an else block, after an IF or UIF statement.
1728 .. opcode:: ENDIF - End If
1730 Ends an IF or UIF block.
1733 .. opcode:: SWITCH - Switch
1735 Starts a C-style switch expression. The switch consists of one or multiple
1736 CASE statements, and at most one DEFAULT statement. Execution of a statement
1737 ends when a BRK is hit, but just like in C falling through to other cases
1738 without a break is allowed. Similarly, DEFAULT label is allowed anywhere not
1739 just as last statement, and fallthrough is allowed into/from it.
1740 CASE src arguments are evaluated at bit level against the SWITCH src argument.
1746 (some instructions here)
1749 (some instructions here)
1752 (some instructions here)
1757 .. opcode:: CASE - Switch case
1759 This represents a switch case label. The src arg must be an integer immediate.
1762 .. opcode:: DEFAULT - Switch default
1764 This represents the default case in the switch, which is taken if no other
1768 .. opcode:: ENDSWITCH - End of switch
1770 Ends a switch expression.
1776 The interpolation instructions allow an input to be interpolated in a
1777 different way than its declaration. This corresponds to the GLSL 4.00
1778 interpolateAt* functions. The first argument of each of these must come from
1779 ``TGSI_FILE_INPUT``.
1781 .. opcode:: INTERP_CENTROID - Interpolate at the centroid
1783 Interpolates the varying specified by src0 at the centroid
1785 .. opcode:: INTERP_SAMPLE - Interpolate at the specified sample
1787 Interpolates the varying specified by src0 at the sample id specified by
1788 src1.x (interpreted as an integer)
1790 .. opcode:: INTERP_OFFSET - Interpolate at the specified offset
1792 Interpolates the varying specified by src0 at the offset src1.xy from the
1793 pixel center (interpreted as floats)
1801 The double-precision opcodes reinterpret four-component vectors into
1802 two-component vectors with doubled precision in each component.
1804 .. opcode:: DABS - Absolute
1812 .. opcode:: DADD - Add
1816 dst.xy = src0.xy + src1.xy
1818 dst.zw = src0.zw + src1.zw
1820 .. opcode:: DSEQ - Set on Equal
1824 dst.x = src0.xy == src1.xy ? \sim 0 : 0
1826 dst.z = src0.zw == src1.zw ? \sim 0 : 0
1828 .. opcode:: DSNE - Set on Equal
1832 dst.x = src0.xy != src1.xy ? \sim 0 : 0
1834 dst.z = src0.zw != src1.zw ? \sim 0 : 0
1836 .. opcode:: DSLT - Set on Less than
1840 dst.x = src0.xy < src1.xy ? \sim 0 : 0
1842 dst.z = src0.zw < src1.zw ? \sim 0 : 0
1844 .. opcode:: DSGE - Set on Greater equal
1848 dst.x = src0.xy >= src1.xy ? \sim 0 : 0
1850 dst.z = src0.zw >= src1.zw ? \sim 0 : 0
1852 .. opcode:: DFRAC - Fraction
1856 dst.xy = src.xy - \lfloor src.xy\rfloor
1858 dst.zw = src.zw - \lfloor src.zw\rfloor
1860 .. opcode:: DTRUNC - Truncate
1864 dst.xy = trunc(src.xy)
1866 dst.zw = trunc(src.zw)
1868 .. opcode:: DCEIL - Ceiling
1872 dst.xy = \lceil src.xy\rceil
1874 dst.zw = \lceil src.zw\rceil
1876 .. opcode:: DFLR - Floor
1880 dst.xy = \lfloor src.xy\rfloor
1882 dst.zw = \lfloor src.zw\rfloor
1884 .. opcode:: DROUND - Fraction
1888 dst.xy = round(src.xy)
1890 dst.zw = round(src.zw)
1892 .. opcode:: DSSG - Set Sign
1896 dst.xy = (src.xy > 0) ? 1.0 : (src.xy < 0) ? -1.0 : 0.0
1898 dst.zw = (src.zw > 0) ? 1.0 : (src.zw < 0) ? -1.0 : 0.0
1900 .. opcode:: DFRACEXP - Convert Number to Fractional and Integral Components
1902 Like the ``frexp()`` routine in many math libraries, this opcode stores the
1903 exponent of its source to ``dst0``, and the significand to ``dst1``, such that
1904 :math:`dst1 \times 2^{dst0} = src` .
1908 dst0.xy = exp(src.xy)
1910 dst1.xy = frac(src.xy)
1912 dst0.zw = exp(src.zw)
1914 dst1.zw = frac(src.zw)
1916 .. opcode:: DLDEXP - Multiply Number by Integral Power of 2
1918 This opcode is the inverse of :opcode:`DFRACEXP`. The second
1919 source is an integer.
1923 dst.xy = src0.xy \times 2^{src1.x}
1925 dst.zw = src0.zw \times 2^{src1.y}
1927 .. opcode:: DMIN - Minimum
1931 dst.xy = min(src0.xy, src1.xy)
1933 dst.zw = min(src0.zw, src1.zw)
1935 .. opcode:: DMAX - Maximum
1939 dst.xy = max(src0.xy, src1.xy)
1941 dst.zw = max(src0.zw, src1.zw)
1943 .. opcode:: DMUL - Multiply
1947 dst.xy = src0.xy \times src1.xy
1949 dst.zw = src0.zw \times src1.zw
1952 .. opcode:: DMAD - Multiply And Add
1956 dst.xy = src0.xy \times src1.xy + src2.xy
1958 dst.zw = src0.zw \times src1.zw + src2.zw
1961 .. opcode:: DFMA - Fused Multiply-Add
1963 Perform a * b + c with no intermediate rounding step.
1967 dst.xy = src0.xy \times src1.xy + src2.xy
1969 dst.zw = src0.zw \times src1.zw + src2.zw
1972 .. opcode:: DDIV - Divide
1976 dst.xy = \frac{src0.xy}{src1.xy}
1978 dst.zw = \frac{src0.zw}{src1.zw}
1981 .. opcode:: DRCP - Reciprocal
1985 dst.xy = \frac{1}{src.xy}
1987 dst.zw = \frac{1}{src.zw}
1989 .. opcode:: DSQRT - Square Root
1993 dst.xy = \sqrt{src.xy}
1995 dst.zw = \sqrt{src.zw}
1997 .. opcode:: DRSQ - Reciprocal Square Root
2001 dst.xy = \frac{1}{\sqrt{src.xy}}
2003 dst.zw = \frac{1}{\sqrt{src.zw}}
2005 .. opcode:: F2D - Float to Double
2009 dst.xy = double(src0.x)
2011 dst.zw = double(src0.y)
2013 .. opcode:: D2F - Double to Float
2017 dst.x = float(src0.xy)
2019 dst.y = float(src0.zw)
2021 .. opcode:: I2D - Int to Double
2025 dst.xy = double(src0.x)
2027 dst.zw = double(src0.y)
2029 .. opcode:: D2I - Double to Int
2033 dst.x = int(src0.xy)
2035 dst.y = int(src0.zw)
2037 .. opcode:: U2D - Unsigned Int to Double
2041 dst.xy = double(src0.x)
2043 dst.zw = double(src0.y)
2045 .. opcode:: D2U - Double to Unsigned Int
2049 dst.x = unsigned(src0.xy)
2051 dst.y = unsigned(src0.zw)
2056 The 64-bit integer opcodes reinterpret four-component vectors into
2057 two-component vectors with 64-bits in each component.
2059 .. opcode:: I64ABS - 64-bit Integer Absolute Value
2067 .. opcode:: I64NEG - 64-bit Integer Negate
2077 .. opcode:: I64SSG - 64-bit Integer Set Sign
2081 dst.xy = (src0.xy < 0) ? -1 : (src0.xy > 0) ? 1 : 0
2083 dst.zw = (src0.zw < 0) ? -1 : (src0.zw > 0) ? 1 : 0
2085 .. opcode:: U64ADD - 64-bit Integer Add
2089 dst.xy = src0.xy + src1.xy
2091 dst.zw = src0.zw + src1.zw
2093 .. opcode:: U64MUL - 64-bit Integer Multiply
2097 dst.xy = src0.xy * src1.xy
2099 dst.zw = src0.zw * src1.zw
2101 .. opcode:: U64SEQ - 64-bit Integer Set on Equal
2105 dst.x = src0.xy == src1.xy ? \sim 0 : 0
2107 dst.z = src0.zw == src1.zw ? \sim 0 : 0
2109 .. opcode:: U64SNE - 64-bit Integer Set on Not Equal
2113 dst.x = src0.xy != src1.xy ? \sim 0 : 0
2115 dst.z = src0.zw != src1.zw ? \sim 0 : 0
2117 .. opcode:: U64SLT - 64-bit Unsigned Integer Set on Less Than
2121 dst.x = src0.xy < src1.xy ? \sim 0 : 0
2123 dst.z = src0.zw < src1.zw ? \sim 0 : 0
2125 .. opcode:: U64SGE - 64-bit Unsigned Integer Set on Greater Equal
2129 dst.x = src0.xy >= src1.xy ? \sim 0 : 0
2131 dst.z = src0.zw >= src1.zw ? \sim 0 : 0
2133 .. opcode:: I64SLT - 64-bit Signed Integer Set on Less Than
2137 dst.x = src0.xy < src1.xy ? \sim 0 : 0
2139 dst.z = src0.zw < src1.zw ? \sim 0 : 0
2141 .. opcode:: I64SGE - 64-bit Signed Integer Set on Greater Equal
2145 dst.x = src0.xy >= src1.xy ? \sim 0 : 0
2147 dst.z = src0.zw >= src1.zw ? \sim 0 : 0
2149 .. opcode:: I64MIN - Minimum of 64-bit Signed Integers
2153 dst.xy = min(src0.xy, src1.xy)
2155 dst.zw = min(src0.zw, src1.zw)
2157 .. opcode:: U64MIN - Minimum of 64-bit Unsigned Integers
2161 dst.xy = min(src0.xy, src1.xy)
2163 dst.zw = min(src0.zw, src1.zw)
2165 .. opcode:: I64MAX - Maximum of 64-bit Signed Integers
2169 dst.xy = max(src0.xy, src1.xy)
2171 dst.zw = max(src0.zw, src1.zw)
2173 .. opcode:: U64MAX - Maximum of 64-bit Unsigned Integers
2177 dst.xy = max(src0.xy, src1.xy)
2179 dst.zw = max(src0.zw, src1.zw)
2181 .. opcode:: U64SHL - Shift Left 64-bit Unsigned Integer
2183 The shift count is masked with 0x3f before the shift is applied.
2187 dst.xy = src0.xy << (0x3f \& src1.x)
2189 dst.zw = src0.zw << (0x3f \& src1.y)
2191 .. opcode:: I64SHR - Arithmetic Shift Right (of 64-bit Signed Integer)
2193 The shift count is masked with 0x3f before the shift is applied.
2197 dst.xy = src0.xy >> (0x3f \& src1.x)
2199 dst.zw = src0.zw >> (0x3f \& src1.y)
2201 .. opcode:: U64SHR - Logical Shift Right (of 64-bit Unsigned Integer)
2203 The shift count is masked with 0x3f before the shift is applied.
2207 dst.xy = src0.xy >> (unsigned) (0x3f \& src1.x)
2209 dst.zw = src0.zw >> (unsigned) (0x3f \& src1.y)
2211 .. opcode:: I64DIV - 64-bit Signed Integer Division
2215 dst.xy = src0.xy \ src1.xy
2217 dst.zw = src0.zw \ src1.zw
2219 .. opcode:: U64DIV - 64-bit Unsigned Integer Division
2223 dst.xy = src0.xy \ src1.xy
2225 dst.zw = src0.zw \ src1.zw
2227 .. opcode:: U64MOD - 64-bit Unsigned Integer Remainder
2231 dst.xy = src0.xy \bmod src1.xy
2233 dst.zw = src0.zw \bmod src1.zw
2235 .. opcode:: I64MOD - 64-bit Signed Integer Remainder
2239 dst.xy = src0.xy \bmod src1.xy
2241 dst.zw = src0.zw \bmod src1.zw
2243 .. opcode:: F2U64 - Float to 64-bit Unsigned Int
2247 dst.xy = (uint64_t) src0.x
2249 dst.zw = (uint64_t) src0.y
2251 .. opcode:: F2I64 - Float to 64-bit Int
2255 dst.xy = (int64_t) src0.x
2257 dst.zw = (int64_t) src0.y
2259 .. opcode:: U2I64 - Unsigned Integer to 64-bit Integer
2261 This is a zero extension.
2265 dst.xy = (uint64_t) src0.x
2267 dst.zw = (uint64_t) src0.y
2269 .. opcode:: I2I64 - Signed Integer to 64-bit Integer
2271 This is a sign extension.
2275 dst.xy = (int64_t) src0.x
2277 dst.zw = (int64_t) src0.y
2279 .. opcode:: D2U64 - Double to 64-bit Unsigned Int
2283 dst.xy = (uint64_t) src0.xy
2285 dst.zw = (uint64_t) src0.zw
2287 .. opcode:: D2I64 - Double to 64-bit Int
2291 dst.xy = (int64_t) src0.xy
2293 dst.zw = (int64_t) src0.zw
2295 .. opcode:: U642F - 64-bit unsigned integer to float
2299 dst.x = (float) src0.xy
2301 dst.y = (float) src0.zw
2303 .. opcode:: I642F - 64-bit Int to Float
2307 dst.x = (float) src0.xy
2309 dst.y = (float) src0.zw
2311 .. opcode:: U642D - 64-bit unsigned integer to double
2315 dst.xy = (double) src0.xy
2317 dst.zw = (double) src0.zw
2319 .. opcode:: I642D - 64-bit Int to double
2323 dst.xy = (double) src0.xy
2325 dst.zw = (double) src0.zw
2327 .. _samplingopcodes:
2329 Resource Sampling Opcodes
2330 ^^^^^^^^^^^^^^^^^^^^^^^^^
2332 Those opcodes follow very closely semantics of the respective Direct3D
2333 instructions. If in doubt double check Direct3D documentation.
2334 Note that the swizzle on SVIEW (src1) determines texel swizzling
2339 Using provided address, sample data from the specified texture using the
2340 filtering mode identified by the given sampler. The source data may come from
2341 any resource type other than buffers.
2343 Syntax: ``SAMPLE dst, address, sampler_view, sampler``
2345 Example: ``SAMPLE TEMP[0], TEMP[1], SVIEW[0], SAMP[0]``
2347 .. opcode:: SAMPLE_I
2349 Simplified alternative to the SAMPLE instruction. Using the provided
2350 integer address, SAMPLE_I fetches data from the specified sampler view
2351 without any filtering. The source data may come from any resource type
2354 Syntax: ``SAMPLE_I dst, address, sampler_view``
2356 Example: ``SAMPLE_I TEMP[0], TEMP[1], SVIEW[0]``
2358 The 'address' is specified as unsigned integers. If the 'address' is out of
2359 range [0...(# texels - 1)] the result of the fetch is always 0 in all
2360 components. As such the instruction doesn't honor address wrap modes, in
2361 cases where that behavior is desirable 'SAMPLE' instruction should be used.
2362 address.w always provides an unsigned integer mipmap level. If the value is
2363 out of the range then the instruction always returns 0 in all components.
2364 address.yz are ignored for buffers and 1d textures. address.z is ignored
2365 for 1d texture arrays and 2d textures.
2367 For 1D texture arrays address.y provides the array index (also as unsigned
2368 integer). If the value is out of the range of available array indices
2369 [0... (array size - 1)] then the opcode always returns 0 in all components.
2370 For 2D texture arrays address.z provides the array index, otherwise it
2371 exhibits the same behavior as in the case for 1D texture arrays. The exact
2372 semantics of the source address are presented in the table below:
2374 +---------------------------+----+-----+-----+---------+
2375 | resource type | X | Y | Z | W |
2376 +===========================+====+=====+=====+=========+
2377 | ``PIPE_BUFFER`` | x | | | ignored |
2378 +---------------------------+----+-----+-----+---------+
2379 | ``PIPE_TEXTURE_1D`` | x | | | mpl |
2380 +---------------------------+----+-----+-----+---------+
2381 | ``PIPE_TEXTURE_2D`` | x | y | | mpl |
2382 +---------------------------+----+-----+-----+---------+
2383 | ``PIPE_TEXTURE_3D`` | x | y | z | mpl |
2384 +---------------------------+----+-----+-----+---------+
2385 | ``PIPE_TEXTURE_RECT`` | x | y | | mpl |
2386 +---------------------------+----+-----+-----+---------+
2387 | ``PIPE_TEXTURE_CUBE`` | not allowed as source |
2388 +---------------------------+----+-----+-----+---------+
2389 | ``PIPE_TEXTURE_1D_ARRAY`` | x | idx | | mpl |
2390 +---------------------------+----+-----+-----+---------+
2391 | ``PIPE_TEXTURE_2D_ARRAY`` | x | y | idx | mpl |
2392 +---------------------------+----+-----+-----+---------+
2394 Where 'mpl' is a mipmap level and 'idx' is the array index.
2396 .. opcode:: SAMPLE_I_MS
2398 Just like SAMPLE_I but allows fetch data from multi-sampled surfaces.
2400 Syntax: ``SAMPLE_I_MS dst, address, sampler_view, sample``
2402 .. opcode:: SAMPLE_B
2404 Just like the SAMPLE instruction with the exception that an additional bias
2405 is applied to the level of detail computed as part of the instruction
2408 Syntax: ``SAMPLE_B dst, address, sampler_view, sampler, lod_bias``
2410 Example: ``SAMPLE_B TEMP[0], TEMP[1], SVIEW[0], SAMP[0], TEMP[2].x``
2412 .. opcode:: SAMPLE_C
2414 Similar to the SAMPLE instruction but it performs a comparison filter. The
2415 operands to SAMPLE_C are identical to SAMPLE, except that there is an
2416 additional float32 operand, reference value, which must be a register with
2417 single-component, or a scalar literal. SAMPLE_C makes the hardware use the
2418 current samplers compare_func (in pipe_sampler_state) to compare reference
2419 value against the red component value for the surce resource at each texel
2420 that the currently configured texture filter covers based on the provided
2423 Syntax: ``SAMPLE_C dst, address, sampler_view.r, sampler, ref_value``
2425 Example: ``SAMPLE_C TEMP[0], TEMP[1], SVIEW[0].r, SAMP[0], TEMP[2].x``
2427 .. opcode:: SAMPLE_C_LZ
2429 Same as SAMPLE_C, but LOD is 0 and derivatives are ignored. The LZ stands
2432 Syntax: ``SAMPLE_C_LZ dst, address, sampler_view.r, sampler, ref_value``
2434 Example: ``SAMPLE_C_LZ TEMP[0], TEMP[1], SVIEW[0].r, SAMP[0], TEMP[2].x``
2437 .. opcode:: SAMPLE_D
2439 SAMPLE_D is identical to the SAMPLE opcode except that the derivatives for
2440 the source address in the x direction and the y direction are provided by
2443 Syntax: ``SAMPLE_D dst, address, sampler_view, sampler, der_x, der_y``
2445 Example: ``SAMPLE_D TEMP[0], TEMP[1], SVIEW[0], SAMP[0], TEMP[2], TEMP[3]``
2447 .. opcode:: SAMPLE_L
2449 SAMPLE_L is identical to the SAMPLE opcode except that the LOD is provided
2450 directly as a scalar value, representing no anisotropy.
2452 Syntax: ``SAMPLE_L dst, address, sampler_view, sampler, explicit_lod``
2454 Example: ``SAMPLE_L TEMP[0], TEMP[1], SVIEW[0], SAMP[0], TEMP[2].x``
2458 Gathers the four texels to be used in a bi-linear filtering operation and
2459 packs them into a single register. Only works with 2D, 2D array, cubemaps,
2460 and cubemaps arrays. For 2D textures, only the addressing modes of the
2461 sampler and the top level of any mip pyramid are used. Set W to zero. It
2462 behaves like the SAMPLE instruction, but a filtered sample is not
2463 generated. The four samples that contribute to filtering are placed into
2464 xyzw in counter-clockwise order, starting with the (u,v) texture coordinate
2465 delta at the following locations (-, +), (+, +), (+, -), (-, -), where the
2466 magnitude of the deltas are half a texel.
2469 .. opcode:: SVIEWINFO
2471 Query the dimensions of a given sampler view. dst receives width, height,
2472 depth or array size and number of mipmap levels as int4. The dst can have a
2473 writemask which will specify what info is the caller interested in.
2475 Syntax: ``SVIEWINFO dst, src_mip_level, sampler_view``
2477 Example: ``SVIEWINFO TEMP[0], TEMP[1].x, SVIEW[0]``
2479 src_mip_level is an unsigned integer scalar. If it's out of range then
2480 returns 0 for width, height and depth/array size but the total number of
2481 mipmap is still returned correctly for the given sampler view. The returned
2482 width, height and depth values are for the mipmap level selected by the
2483 src_mip_level and are in the number of texels. For 1d texture array width
2484 is in dst.x, array size is in dst.y and dst.z is 0. The number of mipmaps is
2485 still in dst.w. In contrast to d3d10 resinfo, there's no way in the tgsi
2486 instruction encoding to specify the return type (float/rcpfloat/uint), hence
2487 always using uint. Also, unlike the SAMPLE instructions, the swizzle on src1
2488 resinfo allowing swizzling dst values is ignored (due to the interaction
2489 with rcpfloat modifier which requires some swizzle handling in the state
2492 .. opcode:: SAMPLE_POS
2494 Query the position of a given sample. dst receives float4 (x, y, 0, 0)
2495 indicated where the sample is located. If the resource is not a multi-sample
2496 resource and not a render target, the result is 0.
2498 .. opcode:: SAMPLE_INFO
2500 dst receives number of samples in x. If the resource is not a multi-sample
2501 resource and not a render target, the result is 0.
2504 .. _resourceopcodes:
2506 Resource Access Opcodes
2507 ^^^^^^^^^^^^^^^^^^^^^^^
2509 .. opcode:: LOAD - Fetch data from a shader buffer or image
2511 Syntax: ``LOAD dst, resource, address``
2513 Example: ``LOAD TEMP[0], BUFFER[0], TEMP[1]``
2515 Using the provided integer address, LOAD fetches data
2516 from the specified buffer or texture without any
2519 The 'address' is specified as a vector of unsigned
2520 integers. If the 'address' is out of range the result
2523 Only the first mipmap level of a resource can be read
2524 from using this instruction.
2526 For 1D or 2D texture arrays, the array index is
2527 provided as an unsigned integer in address.y or
2528 address.z, respectively. address.yz are ignored for
2529 buffers and 1D textures. address.z is ignored for 1D
2530 texture arrays and 2D textures. address.w is always
2533 A swizzle suffix may be added to the resource argument
2534 this will cause the resource data to be swizzled accordingly.
2536 .. opcode:: STORE - Write data to a shader resource
2538 Syntax: ``STORE resource, address, src``
2540 Example: ``STORE BUFFER[0], TEMP[0], TEMP[1]``
2542 Using the provided integer address, STORE writes data
2543 to the specified buffer or texture.
2545 The 'address' is specified as a vector of unsigned
2546 integers. If the 'address' is out of range the result
2549 Only the first mipmap level of a resource can be
2550 written to using this instruction.
2552 For 1D or 2D texture arrays, the array index is
2553 provided as an unsigned integer in address.y or
2554 address.z, respectively. address.yz are ignored for
2555 buffers and 1D textures. address.z is ignored for 1D
2556 texture arrays and 2D textures. address.w is always
2559 .. opcode:: RESQ - Query information about a resource
2561 Syntax: ``RESQ dst, resource``
2563 Example: ``RESQ TEMP[0], BUFFER[0]``
2565 Returns information about the buffer or image resource. For buffer
2566 resources, the size (in bytes) is returned in the x component. For
2567 image resources, .xyz will contain the width/height/layers of the
2568 image, while .w will contain the number of samples for multi-sampled
2571 .. opcode:: FBFETCH - Load data from framebuffer
2573 Syntax: ``FBFETCH dst, output``
2575 Example: ``FBFETCH TEMP[0], OUT[0]``
2577 This is only valid on ``COLOR`` semantic outputs. Returns the color
2578 of the current position in the framebuffer from before this fragment
2579 shader invocation. May return the same value from multiple calls for
2580 a particular output within a single invocation. Note that result may
2581 be undefined if a fragment is drawn multiple times without a blend
2585 .. _threadsyncopcodes:
2587 Inter-thread synchronization opcodes
2588 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2590 These opcodes are intended for communication between threads running
2591 within the same compute grid. For now they're only valid in compute
2594 .. opcode:: MFENCE - Memory fence
2596 Syntax: ``MFENCE resource``
2598 Example: ``MFENCE RES[0]``
2600 This opcode forces strong ordering between any memory access
2601 operations that affect the specified resource. This means that
2602 previous loads and stores (and only those) will be performed and
2603 visible to other threads before the program execution continues.
2606 .. opcode:: LFENCE - Load memory fence
2608 Syntax: ``LFENCE resource``
2610 Example: ``LFENCE RES[0]``
2612 Similar to MFENCE, but it only affects the ordering of memory loads.
2615 .. opcode:: SFENCE - Store memory fence
2617 Syntax: ``SFENCE resource``
2619 Example: ``SFENCE RES[0]``
2621 Similar to MFENCE, but it only affects the ordering of memory stores.
2624 .. opcode:: BARRIER - Thread group barrier
2628 This opcode suspends the execution of the current thread until all
2629 the remaining threads in the working group reach the same point of
2630 the program. Results are unspecified if any of the remaining
2631 threads terminates or never reaches an executed BARRIER instruction.
2633 .. opcode:: MEMBAR - Memory barrier
2637 This opcode waits for the completion of all memory accesses based on
2638 the type passed in. The type is an immediate bitfield with the following
2641 Bit 0: Shader storage buffers
2642 Bit 1: Atomic buffers
2644 Bit 3: Shared memory
2647 These may be passed in in any combination. An implementation is free to not
2648 distinguish between these as it sees fit. However these map to all the
2649 possibilities made available by GLSL.
2656 These opcodes provide atomic variants of some common arithmetic and
2657 logical operations. In this context atomicity means that another
2658 concurrent memory access operation that affects the same memory
2659 location is guaranteed to be performed strictly before or after the
2660 entire execution of the atomic operation. The resource may be a buffer
2661 or an image. In the case of an image, the offset works the same as for
2662 ``LOAD`` and ``STORE``, specified above. These atomic operations may
2663 only be used with 32-bit integer image formats.
2665 .. opcode:: ATOMUADD - Atomic integer addition
2667 Syntax: ``ATOMUADD dst, resource, offset, src``
2669 Example: ``ATOMUADD TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2671 The following operation is performed atomically:
2675 dst_x = resource[offset]
2677 resource[offset] = dst_x + src_x
2680 .. opcode:: ATOMXCHG - Atomic exchange
2682 Syntax: ``ATOMXCHG dst, resource, offset, src``
2684 Example: ``ATOMXCHG TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2686 The following operation is performed atomically:
2690 dst_x = resource[offset]
2692 resource[offset] = src_x
2695 .. opcode:: ATOMCAS - Atomic compare-and-exchange
2697 Syntax: ``ATOMCAS dst, resource, offset, cmp, src``
2699 Example: ``ATOMCAS TEMP[0], BUFFER[0], TEMP[1], TEMP[2], TEMP[3]``
2701 The following operation is performed atomically:
2705 dst_x = resource[offset]
2707 resource[offset] = (dst_x == cmp_x ? src_x : dst_x)
2710 .. opcode:: ATOMAND - Atomic bitwise And
2712 Syntax: ``ATOMAND dst, resource, offset, src``
2714 Example: ``ATOMAND TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2716 The following operation is performed atomically:
2720 dst_x = resource[offset]
2722 resource[offset] = dst_x \& src_x
2725 .. opcode:: ATOMOR - Atomic bitwise Or
2727 Syntax: ``ATOMOR dst, resource, offset, src``
2729 Example: ``ATOMOR TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2731 The following operation is performed atomically:
2735 dst_x = resource[offset]
2737 resource[offset] = dst_x | src_x
2740 .. opcode:: ATOMXOR - Atomic bitwise Xor
2742 Syntax: ``ATOMXOR dst, resource, offset, src``
2744 Example: ``ATOMXOR TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2746 The following operation is performed atomically:
2750 dst_x = resource[offset]
2752 resource[offset] = dst_x \oplus src_x
2755 .. opcode:: ATOMUMIN - Atomic unsigned minimum
2757 Syntax: ``ATOMUMIN dst, resource, offset, src``
2759 Example: ``ATOMUMIN TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2761 The following operation is performed atomically:
2765 dst_x = resource[offset]
2767 resource[offset] = (dst_x < src_x ? dst_x : src_x)
2770 .. opcode:: ATOMUMAX - Atomic unsigned maximum
2772 Syntax: ``ATOMUMAX dst, resource, offset, src``
2774 Example: ``ATOMUMAX TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2776 The following operation is performed atomically:
2780 dst_x = resource[offset]
2782 resource[offset] = (dst_x > src_x ? dst_x : src_x)
2785 .. opcode:: ATOMIMIN - Atomic signed minimum
2787 Syntax: ``ATOMIMIN dst, resource, offset, src``
2789 Example: ``ATOMIMIN TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2791 The following operation is performed atomically:
2795 dst_x = resource[offset]
2797 resource[offset] = (dst_x < src_x ? dst_x : src_x)
2800 .. opcode:: ATOMIMAX - Atomic signed maximum
2802 Syntax: ``ATOMIMAX dst, resource, offset, src``
2804 Example: ``ATOMIMAX TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2806 The following operation is performed atomically:
2810 dst_x = resource[offset]
2812 resource[offset] = (dst_x > src_x ? dst_x : src_x)
2820 These opcodes compare the given value across the shader invocations
2821 running in the current SIMD group. The details of exactly which
2822 invocations get compared are implementation-defined, and it would be a
2823 correct implementation to only ever consider the current thread's
2824 value. (i.e. SIMD group of 1). The argument is treated as a boolean.
2826 .. opcode:: VOTE_ANY - Value is set in any of the current invocations
2828 .. opcode:: VOTE_ALL - Value is set in all of the current invocations
2830 .. opcode:: VOTE_EQ - Value is the same in all of the current invocations
2833 Explanation of symbols used
2834 ------------------------------
2841 :math:`|x|` Absolute value of `x`.
2843 :math:`\lceil x \rceil` Ceiling of `x`.
2845 clamp(x,y,z) Clamp x between y and z.
2846 (x < y) ? y : (x > z) ? z : x
2848 :math:`\lfloor x\rfloor` Floor of `x`.
2850 :math:`\log_2{x}` Logarithm of `x`, base 2.
2852 max(x,y) Maximum of x and y.
2855 min(x,y) Minimum of x and y.
2858 partialx(x) Derivative of x relative to fragment's X.
2860 partialy(x) Derivative of x relative to fragment's Y.
2862 pop() Pop from stack.
2864 :math:`x^y` `x` to the power `y`.
2866 push(x) Push x on stack.
2870 trunc(x) Truncate x, i.e. drop the fraction bits.
2877 discard Discard fragment.
2881 target Label of target instruction.
2892 Declares a register that is will be referenced as an operand in Instruction
2895 File field contains register file that is being declared and is one
2898 UsageMask field specifies which of the register components can be accessed
2899 and is one of TGSI_WRITEMASK.
2901 The Local flag specifies that a given value isn't intended for
2902 subroutine parameter passing and, as a result, the implementation
2903 isn't required to give any guarantees of it being preserved across
2904 subroutine boundaries. As it's merely a compiler hint, the
2905 implementation is free to ignore it.
2907 If Dimension flag is set to 1, a Declaration Dimension token follows.
2909 If Semantic flag is set to 1, a Declaration Semantic token follows.
2911 If Interpolate flag is set to 1, a Declaration Interpolate token follows.
2913 If file is TGSI_FILE_RESOURCE, a Declaration Resource token follows.
2915 If Array flag is set to 1, a Declaration Array token follows.
2918 ^^^^^^^^^^^^^^^^^^^^^^^^
2920 Declarations can optional have an ArrayID attribute which can be referred by
2921 indirect addressing operands. An ArrayID of zero is reserved and treated as
2922 if no ArrayID is specified.
2924 If an indirect addressing operand refers to a specific declaration by using
2925 an ArrayID only the registers in this declaration are guaranteed to be
2926 accessed, accessing any register outside this declaration results in undefined
2927 behavior. Note that for compatibility the effective index is zero-based and
2928 not relative to the specified declaration
2930 If no ArrayID is specified with an indirect addressing operand the whole
2931 register file might be accessed by this operand. This is strongly discouraged
2932 and will prevent packing of scalar/vec2 arrays and effective alias analysis.
2933 This is only legal for TEMP and CONST register files.
2935 Declaration Semantic
2936 ^^^^^^^^^^^^^^^^^^^^^^^^
2938 Vertex and fragment shader input and output registers may be labeled
2939 with semantic information consisting of a name and index.
2941 Follows Declaration token if Semantic bit is set.
2943 Since its purpose is to link a shader with other stages of the pipeline,
2944 it is valid to follow only those Declaration tokens that declare a register
2945 either in INPUT or OUTPUT file.
2947 SemanticName field contains the semantic name of the register being declared.
2948 There is no default value.
2950 SemanticIndex is an optional subscript that can be used to distinguish
2951 different register declarations with the same semantic name. The default value
2954 The meanings of the individual semantic names are explained in the following
2957 TGSI_SEMANTIC_POSITION
2958 """"""""""""""""""""""
2960 For vertex shaders, TGSI_SEMANTIC_POSITION indicates the vertex shader
2961 output register which contains the homogeneous vertex position in the clip
2962 space coordinate system. After clipping, the X, Y and Z components of the
2963 vertex will be divided by the W value to get normalized device coordinates.
2965 For fragment shaders, TGSI_SEMANTIC_POSITION is used to indicate that
2966 fragment shader input (or system value, depending on which one is
2967 supported by the driver) contains the fragment's window position. The X
2968 component starts at zero and always increases from left to right.
2969 The Y component starts at zero and always increases but Y=0 may either
2970 indicate the top of the window or the bottom depending on the fragment
2971 coordinate origin convention (see TGSI_PROPERTY_FS_COORD_ORIGIN).
2972 The Z coordinate ranges from 0 to 1 to represent depth from the front
2973 to the back of the Z buffer. The W component contains the interpolated
2974 reciprocal of the vertex position W component (corresponding to gl_Fragcoord,
2975 but unlike d3d10 which interpolates the same 1/w but then gives back
2976 the reciprocal of the interpolated value).
2978 Fragment shaders may also declare an output register with
2979 TGSI_SEMANTIC_POSITION. Only the Z component is writable. This allows
2980 the fragment shader to change the fragment's Z position.
2987 For vertex shader outputs or fragment shader inputs/outputs, this
2988 label indicates that the register contains an R,G,B,A color.
2990 Several shader inputs/outputs may contain colors so the semantic index
2991 is used to distinguish them. For example, color[0] may be the diffuse
2992 color while color[1] may be the specular color.
2994 This label is needed so that the flat/smooth shading can be applied
2995 to the right interpolants during rasterization.
2999 TGSI_SEMANTIC_BCOLOR
3000 """"""""""""""""""""
3002 Back-facing colors are only used for back-facing polygons, and are only valid
3003 in vertex shader outputs. After rasterization, all polygons are front-facing
3004 and COLOR and BCOLOR end up occupying the same slots in the fragment shader,
3005 so all BCOLORs effectively become regular COLORs in the fragment shader.
3011 Vertex shader inputs and outputs and fragment shader inputs may be
3012 labeled with TGSI_SEMANTIC_FOG to indicate that the register contains
3013 a fog coordinate. Typically, the fragment shader will use the fog coordinate
3014 to compute a fog blend factor which is used to blend the normal fragment color
3015 with a constant fog color. But fog coord really is just an ordinary vec4
3016 register like regular semantics.
3022 Vertex shader input and output registers may be labeled with
3023 TGIS_SEMANTIC_PSIZE to indicate that the register contains a point size
3024 in the form (S, 0, 0, 1). The point size controls the width or diameter
3025 of points for rasterization. This label cannot be used in fragment
3028 When using this semantic, be sure to set the appropriate state in the
3029 :ref:`rasterizer` first.
3032 TGSI_SEMANTIC_TEXCOORD
3033 """"""""""""""""""""""
3035 Only available if PIPE_CAP_TGSI_TEXCOORD is exposed !
3037 Vertex shader outputs and fragment shader inputs may be labeled with
3038 this semantic to make them replaceable by sprite coordinates via the
3039 sprite_coord_enable state in the :ref:`rasterizer`.
3040 The semantic index permitted with this semantic is limited to <= 7.
3042 If the driver does not support TEXCOORD, sprite coordinate replacement
3043 applies to inputs with the GENERIC semantic instead.
3045 The intended use case for this semantic is gl_TexCoord.
3048 TGSI_SEMANTIC_PCOORD
3049 """"""""""""""""""""
3051 Only available if PIPE_CAP_TGSI_TEXCOORD is exposed !
3053 Fragment shader inputs may be labeled with TGSI_SEMANTIC_PCOORD to indicate
3054 that the register contains sprite coordinates in the form (x, y, 0, 1), if
3055 the current primitive is a point and point sprites are enabled. Otherwise,
3056 the contents of the register are undefined.
3058 The intended use case for this semantic is gl_PointCoord.
3061 TGSI_SEMANTIC_GENERIC
3062 """""""""""""""""""""
3064 All vertex/fragment shader inputs/outputs not labeled with any other
3065 semantic label can be considered to be generic attributes. Typical
3066 uses of generic inputs/outputs are texcoords and user-defined values.
3069 TGSI_SEMANTIC_NORMAL
3070 """"""""""""""""""""
3072 Indicates that a vertex shader input is a normal vector. This is
3073 typically only used for legacy graphics APIs.
3079 This label applies to fragment shader inputs (or system values,
3080 depending on which one is supported by the driver) and indicates that
3081 the register contains front/back-face information.
3083 If it is an input, it will be a floating-point vector in the form (F, 0, 0, 1),
3084 where F will be positive when the fragment belongs to a front-facing polygon,
3085 and negative when the fragment belongs to a back-facing polygon.
3087 If it is a system value, it will be an integer vector in the form (F, 0, 0, 1),
3088 where F is 0xffffffff when the fragment belongs to a front-facing polygon and
3089 0 when the fragment belongs to a back-facing polygon.
3092 TGSI_SEMANTIC_EDGEFLAG
3093 """"""""""""""""""""""
3095 For vertex shaders, this sematic label indicates that an input or
3096 output is a boolean edge flag. The register layout is [F, x, x, x]
3097 where F is 0.0 or 1.0 and x = don't care. Normally, the vertex shader
3098 simply copies the edge flag input to the edgeflag output.
3100 Edge flags are used to control which lines or points are actually
3101 drawn when the polygon mode converts triangles/quads/polygons into
3105 TGSI_SEMANTIC_STENCIL
3106 """""""""""""""""""""
3108 For fragment shaders, this semantic label indicates that an output
3109 is a writable stencil reference value. Only the Y component is writable.
3110 This allows the fragment shader to change the fragments stencilref value.
3113 TGSI_SEMANTIC_VIEWPORT_INDEX
3114 """"""""""""""""""""""""""""
3116 For geometry shaders, this semantic label indicates that an output
3117 contains the index of the viewport (and scissor) to use.
3118 This is an integer value, and only the X component is used.
3124 For geometry shaders, this semantic label indicates that an output
3125 contains the layer value to use for the color and depth/stencil surfaces.
3126 This is an integer value, and only the X component is used.
3127 (Also known as rendertarget array index.)
3130 TGSI_SEMANTIC_CULLDIST
3131 """"""""""""""""""""""
3133 Used as distance to plane for performing application-defined culling
3134 of individual primitives against a plane. When components of vertex
3135 elements are given this label, these values are assumed to be a
3136 float32 signed distance to a plane. Primitives will be completely
3137 discarded if the plane distance for all of the vertices in the
3138 primitive are < 0. If a vertex has a cull distance of NaN, that
3139 vertex counts as "out" (as if its < 0);
3140 The limits on both clip and cull distances are bound
3141 by the PIPE_MAX_CLIP_OR_CULL_DISTANCE_COUNT define which defines
3142 the maximum number of components that can be used to hold the
3143 distances and by the PIPE_MAX_CLIP_OR_CULL_DISTANCE_ELEMENT_COUNT
3144 which specifies the maximum number of registers which can be
3145 annotated with those semantics.
3148 TGSI_SEMANTIC_CLIPDIST
3149 """"""""""""""""""""""
3151 Note this covers clipping and culling distances.
3153 When components of vertex elements are identified this way, these
3154 values are each assumed to be a float32 signed distance to a plane.
3157 Primitive setup only invokes rasterization on pixels for which
3158 the interpolated plane distances are >= 0.
3161 Primitives will be completely discarded if the plane distance
3162 for all of the vertices in the primitive are < 0.
3163 If a vertex has a cull distance of NaN, that vertex counts as "out"
3166 Multiple clip/cull planes can be implemented simultaneously, by
3167 annotating multiple components of one or more vertex elements with
3168 the above specified semantic.
3169 The limits on both clip and cull distances are bound
3170 by the PIPE_MAX_CLIP_OR_CULL_DISTANCE_COUNT define which defines
3171 the maximum number of components that can be used to hold the
3172 distances and by the PIPE_MAX_CLIP_OR_CULL_DISTANCE_ELEMENT_COUNT
3173 which specifies the maximum number of registers which can be
3174 annotated with those semantics.
3175 The properties NUM_CLIPDIST_ENABLED and NUM_CULLDIST_ENABLED
3176 are used to divide up the 2 x vec4 space between clipping and culling.
3178 TGSI_SEMANTIC_SAMPLEID
3179 """"""""""""""""""""""
3181 For fragment shaders, this semantic label indicates that a system value
3182 contains the current sample id (i.e. gl_SampleID).
3183 This is an integer value, and only the X component is used.
3185 TGSI_SEMANTIC_SAMPLEPOS
3186 """""""""""""""""""""""
3188 For fragment shaders, this semantic label indicates that a system value
3189 contains the current sample's position (i.e. gl_SamplePosition). Only the X
3190 and Y values are used.
3192 TGSI_SEMANTIC_SAMPLEMASK
3193 """"""""""""""""""""""""
3195 For fragment shaders, this semantic label indicates that an output contains
3196 the sample mask used to disable further sample processing
3197 (i.e. gl_SampleMask). Only the X value is used, up to 32x MS.
3199 TGSI_SEMANTIC_INVOCATIONID
3200 """"""""""""""""""""""""""
3202 For geometry shaders, this semantic label indicates that a system value
3203 contains the current invocation id (i.e. gl_InvocationID).
3204 This is an integer value, and only the X component is used.
3206 TGSI_SEMANTIC_INSTANCEID
3207 """"""""""""""""""""""""
3209 For vertex shaders, this semantic label indicates that a system value contains
3210 the current instance id (i.e. gl_InstanceID). It does not include the base
3211 instance. This is an integer value, and only the X component is used.
3213 TGSI_SEMANTIC_VERTEXID
3214 """"""""""""""""""""""
3216 For vertex shaders, this semantic label indicates that a system value contains
3217 the current vertex id (i.e. gl_VertexID). It does (unlike in d3d10) include the
3218 base vertex. This is an integer value, and only the X component is used.
3220 TGSI_SEMANTIC_VERTEXID_NOBASE
3221 """""""""""""""""""""""""""""""
3223 For vertex shaders, this semantic label indicates that a system value contains
3224 the current vertex id without including the base vertex (this corresponds to
3225 d3d10 vertex id, so TGSI_SEMANTIC_VERTEXID_NOBASE + TGSI_SEMANTIC_BASEVERTEX
3226 == TGSI_SEMANTIC_VERTEXID). This is an integer value, and only the X component
3229 TGSI_SEMANTIC_BASEVERTEX
3230 """"""""""""""""""""""""
3232 For vertex shaders, this semantic label indicates that a system value contains
3233 the base vertex (i.e. gl_BaseVertex). Note that for non-indexed draw calls,
3234 this contains the first (or start) value instead.
3235 This is an integer value, and only the X component is used.
3237 TGSI_SEMANTIC_PRIMID
3238 """"""""""""""""""""
3240 For geometry and fragment shaders, this semantic label indicates the value
3241 contains the primitive id (i.e. gl_PrimitiveID). This is an integer value,
3242 and only the X component is used.
3243 FIXME: This right now can be either a ordinary input or a system value...
3249 For tessellation evaluation/control shaders, this semantic label indicates a
3250 generic per-patch attribute. Such semantics will not implicitly be per-vertex
3253 TGSI_SEMANTIC_TESSCOORD
3254 """""""""""""""""""""""
3256 For tessellation evaluation shaders, this semantic label indicates the
3257 coordinates of the vertex being processed. This is available in XYZ; W is
3260 TGSI_SEMANTIC_TESSOUTER
3261 """""""""""""""""""""""
3263 For tessellation evaluation/control shaders, this semantic label indicates the
3264 outer tessellation levels of the patch. Isoline tessellation will only have XY
3265 defined, triangle will have XYZ and quads will have XYZW defined. This
3266 corresponds to gl_TessLevelOuter.
3268 TGSI_SEMANTIC_TESSINNER
3269 """""""""""""""""""""""
3271 For tessellation evaluation/control shaders, this semantic label indicates the
3272 inner tessellation levels of the patch. The X value is only defined for
3273 triangle tessellation, while quads will have XY defined. This is entirely
3274 undefined for isoline tessellation.
3276 TGSI_SEMANTIC_VERTICESIN
3277 """"""""""""""""""""""""
3279 For tessellation evaluation/control shaders, this semantic label indicates the
3280 number of vertices provided in the input patch. Only the X value is defined.
3282 TGSI_SEMANTIC_HELPER_INVOCATION
3283 """""""""""""""""""""""""""""""
3285 For fragment shaders, this semantic indicates whether the current
3286 invocation is covered or not. Helper invocations are created in order
3287 to properly compute derivatives, however it may be desirable to skip
3288 some of the logic in those cases. See ``gl_HelperInvocation`` documentation.
3290 TGSI_SEMANTIC_BASEINSTANCE
3291 """"""""""""""""""""""""""
3293 For vertex shaders, the base instance argument supplied for this
3294 draw. This is an integer value, and only the X component is used.
3296 TGSI_SEMANTIC_DRAWID
3297 """"""""""""""""""""
3299 For vertex shaders, the zero-based index of the current draw in a
3300 ``glMultiDraw*`` invocation. This is an integer value, and only the X
3304 TGSI_SEMANTIC_WORK_DIM
3305 """"""""""""""""""""""
3307 For compute shaders started via opencl this retrieves the work_dim
3308 parameter to the clEnqueueNDRangeKernel call with which the shader
3312 TGSI_SEMANTIC_GRID_SIZE
3313 """""""""""""""""""""""
3315 For compute shaders, this semantic indicates the maximum (x, y, z) dimensions
3316 of a grid of thread blocks.
3319 TGSI_SEMANTIC_BLOCK_ID
3320 """"""""""""""""""""""
3322 For compute shaders, this semantic indicates the (x, y, z) coordinates of the
3323 current block inside of the grid.
3326 TGSI_SEMANTIC_BLOCK_SIZE
3327 """"""""""""""""""""""""
3329 For compute shaders, this semantic indicates the maximum (x, y, z) dimensions
3330 of a block in threads.
3333 TGSI_SEMANTIC_THREAD_ID
3334 """""""""""""""""""""""
3336 For compute shaders, this semantic indicates the (x, y, z) coordinates of the
3337 current thread inside of the block.
3340 Declaration Interpolate
3341 ^^^^^^^^^^^^^^^^^^^^^^^
3343 This token is only valid for fragment shader INPUT declarations.
3345 The Interpolate field specifes the way input is being interpolated by
3346 the rasteriser and is one of TGSI_INTERPOLATE_*.
3348 The Location field specifies the location inside the pixel that the
3349 interpolation should be done at, one of ``TGSI_INTERPOLATE_LOC_*``. Note that
3350 when per-sample shading is enabled, the implementation may choose to
3351 interpolate at the sample irrespective of the Location field.
3353 The CylindricalWrap bitfield specifies which register components
3354 should be subject to cylindrical wrapping when interpolating by the
3355 rasteriser. If TGSI_CYLINDRICAL_WRAP_X is set to 1, the X component
3356 should be interpolated according to cylindrical wrapping rules.
3359 Declaration Sampler View
3360 ^^^^^^^^^^^^^^^^^^^^^^^^
3362 Follows Declaration token if file is TGSI_FILE_SAMPLER_VIEW.
3364 DCL SVIEW[#], resource, type(s)
3366 Declares a shader input sampler view and assigns it to a SVIEW[#]
3369 resource can be one of BUFFER, 1D, 2D, 3D, 1DArray and 2DArray.
3371 type must be 1 or 4 entries (if specifying on a per-component
3372 level) out of UNORM, SNORM, SINT, UINT and FLOAT.
3374 For TEX\* style texture sample opcodes (as opposed to SAMPLE\* opcodes
3375 which take an explicit SVIEW[#] source register), there may be optionally
3376 SVIEW[#] declarations. In this case, the SVIEW index is implied by the
3377 SAMP index, and there must be a corresponding SVIEW[#] declaration for
3378 each SAMP[#] declaration. Drivers are free to ignore this if they wish.
3379 But note in particular that some drivers need to know the sampler type
3380 (float/int/unsigned) in order to generate the correct code, so cases
3381 where integer textures are sampled, SVIEW[#] declarations should be
3384 NOTE: It is NOT legal to mix SAMPLE\* style opcodes and TEX\* opcodes
3387 Declaration Resource
3388 ^^^^^^^^^^^^^^^^^^^^
3390 Follows Declaration token if file is TGSI_FILE_RESOURCE.
3392 DCL RES[#], resource [, WR] [, RAW]
3394 Declares a shader input resource and assigns it to a RES[#]
3397 resource can be one of BUFFER, 1D, 2D, 3D, CUBE, 1DArray and
3400 If the RAW keyword is not specified, the texture data will be
3401 subject to conversion, swizzling and scaling as required to yield
3402 the specified data type from the physical data format of the bound
3405 If the RAW keyword is specified, no channel conversion will be
3406 performed: the values read for each of the channels (X,Y,Z,W) will
3407 correspond to consecutive words in the same order and format
3408 they're found in memory. No element-to-address conversion will be
3409 performed either: the value of the provided X coordinate will be
3410 interpreted in byte units instead of texel units. The result of
3411 accessing a misaligned address is undefined.
3413 Usage of the STORE opcode is only allowed if the WR (writable) flag
3418 ^^^^^^^^^^^^^^^^^^^^^^^^
3420 Properties are general directives that apply to the whole TGSI program.
3425 Specifies the fragment shader TGSI_SEMANTIC_POSITION coordinate origin.
3426 The default value is UPPER_LEFT.
3428 If UPPER_LEFT, the position will be (0,0) at the upper left corner and
3429 increase downward and rightward.
3430 If LOWER_LEFT, the position will be (0,0) at the lower left corner and
3431 increase upward and rightward.
3433 OpenGL defaults to LOWER_LEFT, and is configurable with the
3434 GL_ARB_fragment_coord_conventions extension.
3436 DirectX 9/10 use UPPER_LEFT.
3438 FS_COORD_PIXEL_CENTER
3439 """""""""""""""""""""
3441 Specifies the fragment shader TGSI_SEMANTIC_POSITION pixel center convention.
3442 The default value is HALF_INTEGER.
3444 If HALF_INTEGER, the fractionary part of the position will be 0.5
3445 If INTEGER, the fractionary part of the position will be 0.0
3447 Note that this does not affect the set of fragments generated by
3448 rasterization, which is instead controlled by half_pixel_center in the
3451 OpenGL defaults to HALF_INTEGER, and is configurable with the
3452 GL_ARB_fragment_coord_conventions extension.
3454 DirectX 9 uses INTEGER.
3455 DirectX 10 uses HALF_INTEGER.
3457 FS_COLOR0_WRITES_ALL_CBUFS
3458 """"""""""""""""""""""""""
3459 Specifies that writes to the fragment shader color 0 are replicated to all
3460 bound cbufs. This facilitates OpenGL's fragColor output vs fragData[0] where
3461 fragData is directed to a single color buffer, but fragColor is broadcast.
3464 """"""""""""""""""""""""""
3465 If this property is set on the program bound to the shader stage before the
3466 fragment shader, user clip planes should have no effect (be disabled) even if
3467 that shader does not write to any clip distance outputs and the rasterizer's
3468 clip_plane_enable is non-zero.
3469 This property is only supported by drivers that also support shader clip
3471 This is useful for APIs that don't have UCPs and where clip distances written
3472 by a shader cannot be disabled.
3477 Specifies the number of times a geometry shader should be executed for each
3478 input primitive. Each invocation will have a different
3479 TGSI_SEMANTIC_INVOCATIONID system value set. If not specified, assumed to
3482 VS_WINDOW_SPACE_POSITION
3483 """"""""""""""""""""""""""
3484 If this property is set on the vertex shader, the TGSI_SEMANTIC_POSITION output
3485 is assumed to contain window space coordinates.
3486 Division of X,Y,Z by W and the viewport transformation are disabled, and 1/W is
3487 directly taken from the 4-th component of the shader output.
3488 Naturally, clipping is not performed on window coordinates either.
3489 The effect of this property is undefined if a geometry or tessellation shader
3495 The number of vertices written by the tessellation control shader. This
3496 effectively defines the patch input size of the tessellation evaluation shader
3502 This sets the tessellation primitive mode, one of ``PIPE_PRIM_TRIANGLES``,
3503 ``PIPE_PRIM_QUADS``, or ``PIPE_PRIM_LINES``. (Unlike in GL, there is no
3504 separate isolines settings, the regular lines is assumed to mean isolines.)
3509 This sets the spacing mode of the tessellation generator, one of
3510 ``PIPE_TESS_SPACING_*``.
3515 This sets the vertex order to be clockwise if the value is 1, or
3516 counter-clockwise if set to 0.
3521 If set to a non-zero value, this turns on point mode for the tessellator,
3522 which means that points will be generated instead of primitives.
3524 NUM_CLIPDIST_ENABLED
3525 """"""""""""""""""""
3527 How many clip distance scalar outputs are enabled.
3529 NUM_CULLDIST_ENABLED
3530 """"""""""""""""""""
3532 How many cull distance scalar outputs are enabled.
3534 FS_EARLY_DEPTH_STENCIL
3535 """"""""""""""""""""""
3537 Whether depth test, stencil test, and occlusion query should run before
3538 the fragment shader (regardless of fragment shader side effects). Corresponds
3539 to GLSL early_fragment_tests.
3544 Which shader stage will MOST LIKELY follow after this shader when the shader
3545 is bound. This is only a hint to the driver and doesn't have to be precise.
3546 Only set for VS and TES.
3548 CS_FIXED_BLOCK_WIDTH / HEIGHT / DEPTH
3549 """""""""""""""""""""""""""""""""""""
3551 Threads per block in each dimension, if known at compile time. If the block size
3552 is known all three should be at least 1. If it is unknown they should all be set
3558 The MUL TGSI operation (FP32 multiplication) will return 0 if either
3559 of the operands are equal to 0. That means that 0 * Inf = 0. This
3560 should be set the same way for an entire pipeline. Note that this
3561 applies not only to the literal MUL TGSI opcode, but all FP32
3562 multiplications implied by other operations, such as MAD, FMA, DP2,
3563 DP3, DP4, DPH, DST, LOG, LRP, XPD, and possibly others. If there is a
3564 mismatch between shaders, then it is unspecified whether this behavior
3568 Texture Sampling and Texture Formats
3569 ------------------------------------
3571 This table shows how texture image components are returned as (x,y,z,w) tuples
3572 by TGSI texture instructions, such as :opcode:`TEX`, :opcode:`TXD`, and
3573 :opcode:`TXP`. For reference, OpenGL and Direct3D conventions are shown as
3576 +--------------------+--------------+--------------------+--------------+
3577 | Texture Components | Gallium | OpenGL | Direct3D 9 |
3578 +====================+==============+====================+==============+
3579 | R | (r, 0, 0, 1) | (r, 0, 0, 1) | (r, 1, 1, 1) |
3580 +--------------------+--------------+--------------------+--------------+
3581 | RG | (r, g, 0, 1) | (r, g, 0, 1) | (r, g, 1, 1) |
3582 +--------------------+--------------+--------------------+--------------+
3583 | RGB | (r, g, b, 1) | (r, g, b, 1) | (r, g, b, 1) |
3584 +--------------------+--------------+--------------------+--------------+
3585 | RGBA | (r, g, b, a) | (r, g, b, a) | (r, g, b, a) |
3586 +--------------------+--------------+--------------------+--------------+
3587 | A | (0, 0, 0, a) | (0, 0, 0, a) | (0, 0, 0, a) |
3588 +--------------------+--------------+--------------------+--------------+
3589 | L | (l, l, l, 1) | (l, l, l, 1) | (l, l, l, 1) |
3590 +--------------------+--------------+--------------------+--------------+
3591 | LA | (l, l, l, a) | (l, l, l, a) | (l, l, l, a) |
3592 +--------------------+--------------+--------------------+--------------+
3593 | I | (i, i, i, i) | (i, i, i, i) | N/A |
3594 +--------------------+--------------+--------------------+--------------+
3595 | UV | XXX TBD | (0, 0, 0, 1) | (u, v, 1, 1) |
3596 | | | [#envmap-bumpmap]_ | |
3597 +--------------------+--------------+--------------------+--------------+
3598 | Z | XXX TBD | (z, z, z, 1) | (0, z, 0, 1) |
3599 | | | [#depth-tex-mode]_ | |
3600 +--------------------+--------------+--------------------+--------------+
3601 | S | (s, s, s, s) | unknown | unknown |
3602 +--------------------+--------------+--------------------+--------------+
3604 .. [#envmap-bumpmap] http://www.opengl.org/registry/specs/ATI/envmap_bumpmap.txt
3605 .. [#depth-tex-mode] the default is (z, z, z, 1) but may also be (0, 0, 0, z)
3606 or (z, z, z, z) depending on the value of GL_DEPTH_TEXTURE_MODE.