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:`Double Opcodes`.
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 = \lfloor src.x\rfloor
53 dst.y = \lfloor src.y\rfloor
55 dst.z = \lfloor src.z\rfloor
57 dst.w = \lfloor src.w\rfloor
60 .. opcode:: MOV - Move
73 .. opcode:: LIT - Light Coefficients
81 dst.z = (src.x > 0) ? max(src.y, 0)^{clamp(src.w, -128, 128))} : 0
86 .. opcode:: RCP - Reciprocal
88 This instruction replicates its result.
95 .. opcode:: RSQ - Reciprocal Square Root
97 This instruction replicates its result.
101 dst = \frac{1}{\sqrt{|src.x|}}
104 .. opcode:: SQRT - Square Root
106 This instruction replicates its result.
113 .. opcode:: EXP - Approximate Exponential Base 2
117 dst.x = 2^{\lfloor src.x\rfloor}
119 dst.y = src.x - \lfloor src.x\rfloor
126 .. opcode:: LOG - Approximate Logarithm Base 2
130 dst.x = \lfloor\log_2{|src.x|}\rfloor
132 dst.y = \frac{|src.x|}{2^{\lfloor\log_2{|src.x|}\rfloor}}
134 dst.z = \log_2{|src.x|}
139 .. opcode:: MUL - Multiply
143 dst.x = src0.x \times src1.x
145 dst.y = src0.y \times src1.y
147 dst.z = src0.z \times src1.z
149 dst.w = src0.w \times src1.w
152 .. opcode:: ADD - Add
156 dst.x = src0.x + src1.x
158 dst.y = src0.y + src1.y
160 dst.z = src0.z + src1.z
162 dst.w = src0.w + src1.w
165 .. opcode:: DP3 - 3-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
174 .. opcode:: DP4 - 4-component Dot Product
176 This instruction replicates its result.
180 dst = src0.x \times src1.x + src0.y \times src1.y + src0.z \times src1.z + src0.w \times src1.w
183 .. opcode:: DST - Distance Vector
189 dst.y = src0.y \times src1.y
196 .. opcode:: MIN - Minimum
200 dst.x = min(src0.x, src1.x)
202 dst.y = min(src0.y, src1.y)
204 dst.z = min(src0.z, src1.z)
206 dst.w = min(src0.w, src1.w)
209 .. opcode:: MAX - Maximum
213 dst.x = max(src0.x, src1.x)
215 dst.y = max(src0.y, src1.y)
217 dst.z = max(src0.z, src1.z)
219 dst.w = max(src0.w, src1.w)
222 .. opcode:: SLT - Set On Less Than
226 dst.x = (src0.x < src1.x) ? 1 : 0
228 dst.y = (src0.y < src1.y) ? 1 : 0
230 dst.z = (src0.z < src1.z) ? 1 : 0
232 dst.w = (src0.w < src1.w) ? 1 : 0
235 .. opcode:: SGE - Set On Greater Equal Than
239 dst.x = (src0.x >= src1.x) ? 1 : 0
241 dst.y = (src0.y >= src1.y) ? 1 : 0
243 dst.z = (src0.z >= src1.z) ? 1 : 0
245 dst.w = (src0.w >= src1.w) ? 1 : 0
248 .. opcode:: MAD - Multiply And Add
252 dst.x = src0.x \times src1.x + src2.x
254 dst.y = src0.y \times src1.y + src2.y
256 dst.z = src0.z \times src1.z + src2.z
258 dst.w = src0.w \times src1.w + src2.w
261 .. opcode:: SUB - Subtract
265 dst.x = src0.x - src1.x
267 dst.y = src0.y - src1.y
269 dst.z = src0.z - src1.z
271 dst.w = src0.w - src1.w
274 .. opcode:: LRP - Linear Interpolate
278 dst.x = src0.x \times src1.x + (1 - src0.x) \times src2.x
280 dst.y = src0.y \times src1.y + (1 - src0.y) \times src2.y
282 dst.z = src0.z \times src1.z + (1 - src0.z) \times src2.z
284 dst.w = src0.w \times src1.w + (1 - src0.w) \times src2.w
287 .. opcode:: CND - Condition
291 dst.x = (src2.x > 0.5) ? src0.x : src1.x
293 dst.y = (src2.y > 0.5) ? src0.y : src1.y
295 dst.z = (src2.z > 0.5) ? src0.z : src1.z
297 dst.w = (src2.w > 0.5) ? src0.w : src1.w
300 .. opcode:: DP2A - 2-component Dot Product And Add
304 dst.x = src0.x \times src1.x + src0.y \times src1.y + src2.x
306 dst.y = src0.x \times src1.x + src0.y \times src1.y + src2.x
308 dst.z = src0.x \times src1.x + src0.y \times src1.y + src2.x
310 dst.w = src0.x \times src1.x + src0.y \times src1.y + src2.x
313 .. opcode:: FRC - Fraction
317 dst.x = src.x - \lfloor src.x\rfloor
319 dst.y = src.y - \lfloor src.y\rfloor
321 dst.z = src.z - \lfloor src.z\rfloor
323 dst.w = src.w - \lfloor src.w\rfloor
326 .. opcode:: CLAMP - Clamp
330 dst.x = clamp(src0.x, src1.x, src2.x)
332 dst.y = clamp(src0.y, src1.y, src2.y)
334 dst.z = clamp(src0.z, src1.z, src2.z)
336 dst.w = clamp(src0.w, src1.w, src2.w)
339 .. opcode:: FLR - Floor
341 This is identical to :opcode:`ARL`.
345 dst.x = \lfloor src.x\rfloor
347 dst.y = \lfloor src.y\rfloor
349 dst.z = \lfloor src.z\rfloor
351 dst.w = \lfloor src.w\rfloor
354 .. opcode:: ROUND - Round
367 .. opcode:: EX2 - Exponential Base 2
369 This instruction replicates its result.
376 .. opcode:: LG2 - Logarithm Base 2
378 This instruction replicates its result.
385 .. opcode:: POW - Power
387 This instruction replicates its result.
391 dst = src0.x^{src1.x}
393 .. opcode:: XPD - Cross Product
397 dst.x = src0.y \times src1.z - src1.y \times src0.z
399 dst.y = src0.z \times src1.x - src1.z \times src0.x
401 dst.z = src0.x \times src1.y - src1.x \times src0.y
406 .. opcode:: ABS - Absolute
419 .. opcode:: RCC - Reciprocal Clamped
421 This instruction replicates its result.
423 XXX cleanup on aisle three
427 dst = (1 / src.x) > 0 ? clamp(1 / src.x, 5.42101e-020, 1.884467e+019) : clamp(1 / src.x, -1.884467e+019, -5.42101e-020)
430 .. opcode:: DPH - Homogeneous Dot Product
432 This instruction replicates its result.
436 dst = src0.x \times src1.x + src0.y \times src1.y + src0.z \times src1.z + src1.w
439 .. opcode:: COS - Cosine
441 This instruction replicates its result.
448 .. opcode:: DDX - Derivative Relative To X
452 dst.x = partialx(src.x)
454 dst.y = partialx(src.y)
456 dst.z = partialx(src.z)
458 dst.w = partialx(src.w)
461 .. opcode:: DDY - Derivative Relative To Y
465 dst.x = partialy(src.x)
467 dst.y = partialy(src.y)
469 dst.z = partialy(src.z)
471 dst.w = partialy(src.w)
474 .. opcode:: KILP - Predicated Discard
476 Not really predicated, just unconditional discard
479 .. opcode:: PK2H - Pack Two 16-bit Floats
484 .. opcode:: PK2US - Pack Two Unsigned 16-bit Scalars
489 .. opcode:: PK4B - Pack Four Signed 8-bit Scalars
494 .. opcode:: PK4UB - Pack Four Unsigned 8-bit Scalars
499 .. opcode:: RFL - Reflection Vector
503 dst.x = 2 \times (src0.x \times src1.x + src0.y \times src1.y + src0.z \times src1.z) / (src0.x \times src0.x + src0.y \times src0.y + src0.z \times src0.z) \times src0.x - src1.x
505 dst.y = 2 \times (src0.x \times src1.x + src0.y \times src1.y + src0.z \times src1.z) / (src0.x \times src0.x + src0.y \times src0.y + src0.z \times src0.z) \times src0.y - src1.y
507 dst.z = 2 \times (src0.x \times src1.x + src0.y \times src1.y + src0.z \times src1.z) / (src0.x \times src0.x + src0.y \times src0.y + src0.z \times src0.z) \times src0.z - src1.z
513 Considered for removal.
516 .. opcode:: SEQ - Set On Equal
520 dst.x = (src0.x == src1.x) ? 1 : 0
522 dst.y = (src0.y == src1.y) ? 1 : 0
524 dst.z = (src0.z == src1.z) ? 1 : 0
526 dst.w = (src0.w == src1.w) ? 1 : 0
529 .. opcode:: SFL - Set On False
531 This instruction replicates its result.
539 Considered for removal.
542 .. opcode:: SGT - Set On Greater Than
546 dst.x = (src0.x > src1.x) ? 1 : 0
548 dst.y = (src0.y > src1.y) ? 1 : 0
550 dst.z = (src0.z > src1.z) ? 1 : 0
552 dst.w = (src0.w > src1.w) ? 1 : 0
555 .. opcode:: SIN - Sine
557 This instruction replicates its result.
564 .. opcode:: SLE - Set On Less Equal Than
568 dst.x = (src0.x <= src1.x) ? 1 : 0
570 dst.y = (src0.y <= src1.y) ? 1 : 0
572 dst.z = (src0.z <= src1.z) ? 1 : 0
574 dst.w = (src0.w <= src1.w) ? 1 : 0
577 .. opcode:: SNE - Set On Not Equal
581 dst.x = (src0.x != src1.x) ? 1 : 0
583 dst.y = (src0.y != src1.y) ? 1 : 0
585 dst.z = (src0.z != src1.z) ? 1 : 0
587 dst.w = (src0.w != src1.w) ? 1 : 0
590 .. opcode:: STR - Set On True
592 This instruction replicates its result.
599 .. opcode:: TEX - Texture Lookup
607 dst = texture_sample(unit, coord, bias)
609 for array textures src0.y contains the slice for 1D,
610 and src0.z contain the slice for 2D.
611 for shadow textures with no arrays, src0.z contains
613 for shadow textures with arrays, src0.z contains
614 the reference value for 1D arrays, and src0.w contains
615 the reference value for 2D arrays.
616 There is no way to pass a bias in the .w value for
617 shadow arrays, and GLSL doesn't allow this.
618 GLSL does allow cube shadows maps to take a bias value,
619 and we have to determine how this will look in TGSI.
621 .. opcode:: TXD - Texture Lookup with Derivatives
633 dst = texture_sample_deriv(unit, coord, bias, ddx, ddy)
636 .. opcode:: TXP - Projective Texture Lookup
640 coord.x = src0.x / src.w
642 coord.y = src0.y / src.w
644 coord.z = src0.z / src.w
650 dst = texture_sample(unit, coord, bias)
653 .. opcode:: UP2H - Unpack Two 16-Bit Floats
659 Considered for removal.
661 .. opcode:: UP2US - Unpack Two Unsigned 16-Bit Scalars
667 Considered for removal.
669 .. opcode:: UP4B - Unpack Four Signed 8-Bit Values
675 Considered for removal.
677 .. opcode:: UP4UB - Unpack Four Unsigned 8-Bit Scalars
683 Considered for removal.
685 .. opcode:: X2D - 2D Coordinate Transformation
689 dst.x = src0.x + src1.x \times src2.x + src1.y \times src2.y
691 dst.y = src0.y + src1.x \times src2.z + src1.y \times src2.w
693 dst.z = src0.x + src1.x \times src2.x + src1.y \times src2.y
695 dst.w = src0.y + src1.x \times src2.z + src1.y \times src2.w
699 Considered for removal.
702 .. opcode:: ARA - Address Register Add
708 Considered for removal.
710 .. opcode:: ARR - Address Register Load With Round
723 .. opcode:: SSG - Set Sign
727 dst.x = (src.x > 0) ? 1 : (src.x < 0) ? -1 : 0
729 dst.y = (src.y > 0) ? 1 : (src.y < 0) ? -1 : 0
731 dst.z = (src.z > 0) ? 1 : (src.z < 0) ? -1 : 0
733 dst.w = (src.w > 0) ? 1 : (src.w < 0) ? -1 : 0
736 .. opcode:: CMP - Compare
740 dst.x = (src0.x < 0) ? src1.x : src2.x
742 dst.y = (src0.y < 0) ? src1.y : src2.y
744 dst.z = (src0.z < 0) ? src1.z : src2.z
746 dst.w = (src0.w < 0) ? src1.w : src2.w
749 .. opcode:: KIL - Conditional Discard
753 if (src.x < 0 || src.y < 0 || src.z < 0 || src.w < 0)
758 .. opcode:: SCS - Sine Cosine
771 .. opcode:: TXB - Texture Lookup With Bias
785 dst = texture_sample(unit, coord, bias)
788 .. opcode:: NRM - 3-component Vector Normalise
792 dst.x = src.x / (src.x \times src.x + src.y \times src.y + src.z \times src.z)
794 dst.y = src.y / (src.x \times src.x + src.y \times src.y + src.z \times src.z)
796 dst.z = src.z / (src.x \times src.x + src.y \times src.y + src.z \times src.z)
801 .. opcode:: DIV - Divide
805 dst.x = \frac{src0.x}{src1.x}
807 dst.y = \frac{src0.y}{src1.y}
809 dst.z = \frac{src0.z}{src1.z}
811 dst.w = \frac{src0.w}{src1.w}
814 .. opcode:: DP2 - 2-component Dot Product
816 This instruction replicates its result.
820 dst = src0.x \times src1.x + src0.y \times src1.y
823 .. opcode:: TXL - Texture Lookup With explicit LOD
837 dst = texture_sample(unit, coord, lod)
840 .. opcode:: PUSHA - Push Address Register On Stack
849 Considered for cleanup.
853 Considered for removal.
855 .. opcode:: POPA - Pop Address Register From Stack
864 Considered for cleanup.
868 Considered for removal.
871 .. opcode:: BRA - Branch
877 Considered for removal.
880 .. opcode:: CALLNZ - Subroutine Call If Not Zero
886 Considered for cleanup.
890 Considered for removal.
894 ^^^^^^^^^^^^^^^^^^^^^^^^
896 These opcodes are primarily provided for special-use computational shaders.
897 Support for these opcodes indicated by a special pipe capability bit (TBD).
899 XXX doesn't look like most of the opcodes really belong here.
901 .. opcode:: CEIL - Ceiling
905 dst.x = \lceil src.x\rceil
907 dst.y = \lceil src.y\rceil
909 dst.z = \lceil src.z\rceil
911 dst.w = \lceil src.w\rceil
914 .. opcode:: TRUNC - Truncate
927 .. opcode:: MOD - Modulus
931 dst.x = src0.x \bmod src1.x
933 dst.y = src0.y \bmod src1.y
935 dst.z = src0.z \bmod src1.z
937 dst.w = src0.w \bmod src1.w
940 .. opcode:: UARL - Integer Address Register Load
942 Moves the contents of the source register, assumed to be an integer, into the
943 destination register, which is assumed to be an address (ADDR) register.
946 .. opcode:: SAD - Sum Of Absolute Differences
950 dst.x = |src0.x - src1.x| + src2.x
952 dst.y = |src0.y - src1.y| + src2.y
954 dst.z = |src0.z - src1.z| + src2.z
956 dst.w = |src0.w - src1.w| + src2.w
959 .. opcode:: TXF - Texel Fetch (as per NV_gpu_shader4), extract a single texel
960 from a specified texture image. The source sampler may
961 not be a CUBE or SHADOW.
962 src 0 is a four-component signed integer vector used to
963 identify the single texel accessed. 3 components + level.
964 src 1 is a 3 component constant signed integer vector,
965 with each component only have a range of
966 -8..+8 (hw only seems to deal with this range, interface
967 allows for up to unsigned int).
968 TXF(uint_vec coord, int_vec offset).
971 .. opcode:: TXQ - Texture Size Query (as per NV_gpu_program4)
972 retrieve the dimensions of the texture
973 depending on the target. For 1D (width), 2D/RECT/CUBE
974 (width, height), 3D (width, height, depth),
975 1D array (width, layers), 2D array (width, height, layers)
981 dst.x = texture_width(unit, lod)
983 dst.y = texture_height(unit, lod)
985 dst.z = texture_depth(unit, lod)
989 ^^^^^^^^^^^^^^^^^^^^^^^^
990 These opcodes are used for integer operations.
991 Support for these opcodes indicated by PIPE_SHADER_CAP_INTEGERS (all of them?)
994 .. opcode:: I2F - Signed Integer To Float
996 Rounding is unspecified (round to nearest even suggested).
1000 dst.x = (float) src.x
1002 dst.y = (float) src.y
1004 dst.z = (float) src.z
1006 dst.w = (float) src.w
1009 .. opcode:: U2F - Unsigned Integer To Float
1011 Rounding is unspecified (round to nearest even suggested).
1015 dst.x = (float) src.x
1017 dst.y = (float) src.y
1019 dst.z = (float) src.z
1021 dst.w = (float) src.w
1024 .. opcode:: F2I - Float to Signed Integer
1026 Rounding is towards zero (truncate).
1027 Values outside signed range (including NaNs) produce undefined results.
1040 .. opcode:: F2U - Float to Unsigned Integer
1042 Rounding is towards zero (truncate).
1043 Values outside unsigned range (including NaNs) produce undefined results.
1047 dst.x = (unsigned) src.x
1049 dst.y = (unsigned) src.y
1051 dst.z = (unsigned) src.z
1053 dst.w = (unsigned) src.w
1056 .. opcode:: UADD - Integer Add
1058 This instruction works the same for signed and unsigned integers.
1059 The low 32bit of the result is returned.
1063 dst.x = src0.x + src1.x
1065 dst.y = src0.y + src1.y
1067 dst.z = src0.z + src1.z
1069 dst.w = src0.w + src1.w
1072 .. opcode:: UMAD - Integer Multiply And Add
1074 This instruction works the same for signed and unsigned integers.
1075 The multiplication returns the low 32bit (as does the result itself).
1079 dst.x = src0.x \times src1.x + src2.x
1081 dst.y = src0.y \times src1.y + src2.y
1083 dst.z = src0.z \times src1.z + src2.z
1085 dst.w = src0.w \times src1.w + src2.w
1088 .. opcode:: UMUL - Integer Multiply
1090 This instruction works the same for signed and unsigned integers.
1091 The low 32bit of the result is returned.
1095 dst.x = src0.x \times src1.x
1097 dst.y = src0.y \times src1.y
1099 dst.z = src0.z \times src1.z
1101 dst.w = src0.w \times src1.w
1104 .. opcode:: IDIV - Signed Integer Division
1106 TBD: behavior for division by zero.
1110 dst.x = src0.x \ src1.x
1112 dst.y = src0.y \ src1.y
1114 dst.z = src0.z \ src1.z
1116 dst.w = src0.w \ src1.w
1119 .. opcode:: UDIV - Unsigned Integer Division
1121 For division by zero, 0xffffffff is returned.
1125 dst.x = src0.x \ src1.x
1127 dst.y = src0.y \ src1.y
1129 dst.z = src0.z \ src1.z
1131 dst.w = src0.w \ src1.w
1134 .. opcode:: UMOD - Unsigned Integer Remainder
1136 If second arg is zero, 0xffffffff is returned.
1140 dst.x = src0.x \ src1.x
1142 dst.y = src0.y \ src1.y
1144 dst.z = src0.z \ src1.z
1146 dst.w = src0.w \ src1.w
1149 .. opcode:: NOT - Bitwise Not
1162 .. opcode:: AND - Bitwise And
1166 dst.x = src0.x & src1.x
1168 dst.y = src0.y & src1.y
1170 dst.z = src0.z & src1.z
1172 dst.w = src0.w & src1.w
1175 .. opcode:: OR - Bitwise Or
1179 dst.x = src0.x | src1.x
1181 dst.y = src0.y | src1.y
1183 dst.z = src0.z | src1.z
1185 dst.w = src0.w | src1.w
1188 .. opcode:: XOR - Bitwise Xor
1192 dst.x = src0.x \oplus src1.x
1194 dst.y = src0.y \oplus src1.y
1196 dst.z = src0.z \oplus src1.z
1198 dst.w = src0.w \oplus src1.w
1201 .. opcode:: IMAX - Maximum of Signed Integers
1205 dst.x = max(src0.x, src1.x)
1207 dst.y = max(src0.y, src1.y)
1209 dst.z = max(src0.z, src1.z)
1211 dst.w = max(src0.w, src1.w)
1214 .. opcode:: UMAX - Maximum of Unsigned Integers
1218 dst.x = max(src0.x, src1.x)
1220 dst.y = max(src0.y, src1.y)
1222 dst.z = max(src0.z, src1.z)
1224 dst.w = max(src0.w, src1.w)
1227 .. opcode:: IMIN - Minimum of Signed Integers
1231 dst.x = min(src0.x, src1.x)
1233 dst.y = min(src0.y, src1.y)
1235 dst.z = min(src0.z, src1.z)
1237 dst.w = min(src0.w, src1.w)
1240 .. opcode:: UMIN - Minimum of Unsigned Integers
1244 dst.x = min(src0.x, src1.x)
1246 dst.y = min(src0.y, src1.y)
1248 dst.z = min(src0.z, src1.z)
1250 dst.w = min(src0.w, src1.w)
1253 .. opcode:: SHL - Shift Left
1257 dst.x = src0.x << src1.x
1259 dst.y = src0.y << src1.x
1261 dst.z = src0.z << src1.x
1263 dst.w = src0.w << src1.x
1266 .. opcode:: ISHR - Arithmetic Shift Right (of Signed Integer)
1270 dst.x = src0.x >> src1.x
1272 dst.y = src0.y >> src1.x
1274 dst.z = src0.z >> src1.x
1276 dst.w = src0.w >> src1.x
1279 .. opcode:: USHR - Logical Shift Right
1283 dst.x = src0.x >> (unsigned) src1.x
1285 dst.y = src0.y >> (unsigned) src1.x
1287 dst.z = src0.z >> (unsigned) src1.x
1289 dst.w = src0.w >> (unsigned) src1.x
1292 .. opcode:: UCMP - Integer Conditional Move
1296 dst.x = src0.x ? src1.x : src2.x
1298 dst.y = src0.y ? src1.y : src2.y
1300 dst.z = src0.z ? src1.z : src2.z
1302 dst.w = src0.w ? src1.w : src2.w
1306 .. opcode:: ISSG - Integer Set Sign
1310 dst.x = (src0.x < 0) ? -1 : (src0.x > 0) ? 1 : 0
1312 dst.y = (src0.y < 0) ? -1 : (src0.y > 0) ? 1 : 0
1314 dst.z = (src0.z < 0) ? -1 : (src0.z > 0) ? 1 : 0
1316 dst.w = (src0.w < 0) ? -1 : (src0.w > 0) ? 1 : 0
1320 .. opcode:: ISLT - Signed Integer Set On Less Than
1324 dst.x = (src0.x < src1.x) ? ~0 : 0
1326 dst.y = (src0.y < src1.y) ? ~0 : 0
1328 dst.z = (src0.z < src1.z) ? ~0 : 0
1330 dst.w = (src0.w < src1.w) ? ~0 : 0
1333 .. opcode:: USLT - Unsigned Integer Set On Less Than
1337 dst.x = (src0.x < src1.x) ? ~0 : 0
1339 dst.y = (src0.y < src1.y) ? ~0 : 0
1341 dst.z = (src0.z < src1.z) ? ~0 : 0
1343 dst.w = (src0.w < src1.w) ? ~0 : 0
1346 .. opcode:: ISGE - Signed Integer Set On Greater Equal Than
1350 dst.x = (src0.x >= src1.x) ? ~0 : 0
1352 dst.y = (src0.y >= src1.y) ? ~0 : 0
1354 dst.z = (src0.z >= src1.z) ? ~0 : 0
1356 dst.w = (src0.w >= src1.w) ? ~0 : 0
1359 .. opcode:: USGE - Unsigned Integer Set On Greater Equal Than
1363 dst.x = (src0.x >= src1.x) ? ~0 : 0
1365 dst.y = (src0.y >= src1.y) ? ~0 : 0
1367 dst.z = (src0.z >= src1.z) ? ~0 : 0
1369 dst.w = (src0.w >= src1.w) ? ~0 : 0
1372 .. opcode:: USEQ - Integer Set On Equal
1376 dst.x = (src0.x == src1.x) ? ~0 : 0
1378 dst.y = (src0.y == src1.y) ? ~0 : 0
1380 dst.z = (src0.z == src1.z) ? ~0 : 0
1382 dst.w = (src0.w == src1.w) ? ~0 : 0
1385 .. opcode:: USNE - Integer Set On Not Equal
1389 dst.x = (src0.x != src1.x) ? ~0 : 0
1391 dst.y = (src0.y != src1.y) ? ~0 : 0
1393 dst.z = (src0.z != src1.z) ? ~0 : 0
1395 dst.w = (src0.w != src1.w) ? ~0 : 0
1398 .. opcode:: INEG - Integer Negate
1413 .. opcode:: IABS - Integer Absolute Value
1427 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1429 These opcodes are only supported in geometry shaders; they have no meaning
1430 in any other type of shader.
1432 .. opcode:: EMIT - Emit
1434 Generate a new vertex for the current primitive using the values in the
1438 .. opcode:: ENDPRIM - End Primitive
1440 Complete the current primitive (consisting of the emitted vertices),
1441 and start a new one.
1447 These opcodes are part of :term:`GLSL`'s opcode set. Support for these
1448 opcodes is determined by a special capability bit, ``GLSL``.
1449 Some require glsl version 1.30 (UIF/BREAKC/SWITCH/CASE/DEFAULT/ENDSWITCH).
1451 .. opcode:: CAL - Subroutine Call
1457 .. opcode:: RET - Subroutine Call Return
1462 .. opcode:: CONT - Continue
1464 Unconditionally moves the point of execution to the instruction after the
1465 last bgnloop. The instruction must appear within a bgnloop/endloop.
1469 Support for CONT is determined by a special capability bit,
1470 ``TGSI_CONT_SUPPORTED``. See :ref:`Screen` for more information.
1473 .. opcode:: BGNLOOP - Begin a Loop
1475 Start a loop. Must have a matching endloop.
1478 .. opcode:: BGNSUB - Begin Subroutine
1480 Starts definition of a subroutine. Must have a matching endsub.
1483 .. opcode:: ENDLOOP - End a Loop
1485 End a loop started with bgnloop.
1488 .. opcode:: ENDSUB - End Subroutine
1490 Ends definition of a subroutine.
1493 .. opcode:: NOP - No Operation
1498 .. opcode:: BRK - Break
1500 Unconditionally moves the point of execution to the instruction after the
1501 next endloop or endswitch. The instruction must appear within a loop/endloop
1502 or switch/endswitch.
1505 .. opcode:: BREAKC - Break Conditional
1507 Conditionally moves the point of execution to the instruction after the
1508 next endloop or endswitch. The instruction must appear within a loop/endloop
1509 or switch/endswitch.
1510 Condition evaluates to true if src0.x != 0 where src0.x is interpreted
1511 as an integer register.
1515 Considered for removal as it's quite inconsistent wrt other opcodes
1516 (could emulate with UIF/BRK/ENDIF).
1519 .. opcode:: IF - Float If
1521 Start an IF ... ELSE .. ENDIF block. Condition evaluates to true if
1525 where src0.x is interpreted as a floating point register.
1528 .. opcode:: UIF - Bitwise If
1530 Start an UIF ... ELSE .. ENDIF block. Condition evaluates to true if
1534 where src0.x is interpreted as an integer register.
1537 .. opcode:: ELSE - Else
1539 Starts an else block, after an IF or UIF statement.
1542 .. opcode:: ENDIF - End If
1544 Ends an IF or UIF block.
1547 .. opcode:: SWITCH - Switch
1549 Starts a C-style switch expression. The switch consists of one or multiple
1550 CASE statements, and at most one DEFAULT statement. Execution of a statement
1551 ends when a BRK is hit, but just like in C falling through to other cases
1552 without a break is allowed. Similarly, DEFAULT label is allowed anywhere not
1553 just as last statement, and fallthrough is allowed into/from it.
1554 CASE src arguments are evaluated at bit level against the SWITCH src argument.
1559 (some instructions here)
1562 (some instructions here)
1565 (some instructions here)
1570 .. opcode:: CASE - Switch case
1572 This represents a switch case label. The src arg must be an integer immediate.
1575 .. opcode:: DEFAULT - Switch default
1577 This represents the default case in the switch, which is taken if no other
1581 .. opcode:: ENDSWITCH - End of switch
1583 Ends a switch expression.
1586 .. opcode:: NRM4 - 4-component Vector Normalise
1588 This instruction replicates its result.
1592 dst = \frac{src.x}{src.x \times src.x + src.y \times src.y + src.z \times src.z + src.w \times src.w}
1600 The double-precision opcodes reinterpret four-component vectors into
1601 two-component vectors with doubled precision in each component.
1603 Support for these opcodes is XXX undecided. :T
1605 .. opcode:: DADD - Add
1609 dst.xy = src0.xy + src1.xy
1611 dst.zw = src0.zw + src1.zw
1614 .. opcode:: DDIV - Divide
1618 dst.xy = src0.xy / src1.xy
1620 dst.zw = src0.zw / src1.zw
1622 .. opcode:: DSEQ - Set on Equal
1626 dst.xy = src0.xy == src1.xy ? 1.0F : 0.0F
1628 dst.zw = src0.zw == src1.zw ? 1.0F : 0.0F
1630 .. opcode:: DSLT - Set on Less than
1634 dst.xy = src0.xy < src1.xy ? 1.0F : 0.0F
1636 dst.zw = src0.zw < src1.zw ? 1.0F : 0.0F
1638 .. opcode:: DFRAC - Fraction
1642 dst.xy = src.xy - \lfloor src.xy\rfloor
1644 dst.zw = src.zw - \lfloor src.zw\rfloor
1647 .. opcode:: DFRACEXP - Convert Number to Fractional and Integral Components
1649 Like the ``frexp()`` routine in many math libraries, this opcode stores the
1650 exponent of its source to ``dst0``, and the significand to ``dst1``, such that
1651 :math:`dst1 \times 2^{dst0} = src` .
1655 dst0.xy = exp(src.xy)
1657 dst1.xy = frac(src.xy)
1659 dst0.zw = exp(src.zw)
1661 dst1.zw = frac(src.zw)
1663 .. opcode:: DLDEXP - Multiply Number by Integral Power of 2
1665 This opcode is the inverse of :opcode:`DFRACEXP`.
1669 dst.xy = src0.xy \times 2^{src1.xy}
1671 dst.zw = src0.zw \times 2^{src1.zw}
1673 .. opcode:: DMIN - Minimum
1677 dst.xy = min(src0.xy, src1.xy)
1679 dst.zw = min(src0.zw, src1.zw)
1681 .. opcode:: DMAX - Maximum
1685 dst.xy = max(src0.xy, src1.xy)
1687 dst.zw = max(src0.zw, src1.zw)
1689 .. opcode:: DMUL - Multiply
1693 dst.xy = src0.xy \times src1.xy
1695 dst.zw = src0.zw \times src1.zw
1698 .. opcode:: DMAD - Multiply And Add
1702 dst.xy = src0.xy \times src1.xy + src2.xy
1704 dst.zw = src0.zw \times src1.zw + src2.zw
1707 .. opcode:: DRCP - Reciprocal
1711 dst.xy = \frac{1}{src.xy}
1713 dst.zw = \frac{1}{src.zw}
1715 .. opcode:: DSQRT - Square Root
1719 dst.xy = \sqrt{src.xy}
1721 dst.zw = \sqrt{src.zw}
1724 .. _samplingopcodes:
1726 Resource Sampling Opcodes
1727 ^^^^^^^^^^^^^^^^^^^^^^^^^
1729 Those opcodes follow very closely semantics of the respective Direct3D
1730 instructions. If in doubt double check Direct3D documentation.
1732 .. opcode:: SAMPLE - Using provided address, sample data from the
1733 specified texture using the filtering mode identified
1734 by the gven sampler. The source data may come from
1735 any resource type other than buffers.
1736 SAMPLE dst, address, sampler_view, sampler
1738 SAMPLE TEMP[0], TEMP[1], SVIEW[0], SAMP[0]
1740 .. opcode:: SAMPLE_I - Simplified alternative to the SAMPLE instruction.
1741 Using the provided integer address, SAMPLE_I fetches data
1742 from the specified sampler view without any filtering.
1743 The source data may come from any resource type other
1745 SAMPLE_I dst, address, sampler_view
1747 SAMPLE_I TEMP[0], TEMP[1], SVIEW[0]
1748 The 'address' is specified as unsigned integers. If the
1749 'address' is out of range [0...(# texels - 1)] the
1750 result of the fetch is always 0 in all components.
1751 As such the instruction doesn't honor address wrap
1752 modes, in cases where that behavior is desirable
1753 'SAMPLE' instruction should be used.
1754 address.w always provides an unsigned integer mipmap
1755 level. If the value is out of the range then the
1756 instruction always returns 0 in all components.
1757 address.yz are ignored for buffers and 1d textures.
1758 address.z is ignored for 1d texture arrays and 2d
1760 For 1D texture arrays address.y provides the array
1761 index (also as unsigned integer). If the value is
1762 out of the range of available array indices
1763 [0... (array size - 1)] then the opcode always returns
1764 0 in all components.
1765 For 2D texture arrays address.z provides the array
1766 index, otherwise it exhibits the same behavior as in
1767 the case for 1D texture arrays.
1768 The exact semantics of the source address are presented
1770 resource type X Y Z W
1771 ------------- ------------------------
1772 PIPE_BUFFER x ignored
1773 PIPE_TEXTURE_1D x mpl
1774 PIPE_TEXTURE_2D x y mpl
1775 PIPE_TEXTURE_3D x y z mpl
1776 PIPE_TEXTURE_RECT x y mpl
1777 PIPE_TEXTURE_CUBE not allowed as source
1778 PIPE_TEXTURE_1D_ARRAY x idx mpl
1779 PIPE_TEXTURE_2D_ARRAY x y idx mpl
1781 Where 'mpl' is a mipmap level and 'idx' is the
1784 .. opcode:: SAMPLE_I_MS - Just like SAMPLE_I but allows fetch data from
1785 multi-sampled surfaces.
1786 SAMPLE_I_MS dst, address, sampler_view, sample
1788 .. opcode:: SAMPLE_B - Just like the SAMPLE instruction with the
1789 exception that an additional bias is applied to the
1790 level of detail computed as part of the instruction
1792 SAMPLE_B dst, address, sampler_view, sampler, lod_bias
1794 SAMPLE_B TEMP[0], TEMP[1], SVIEW[0], SAMP[0], TEMP[2].x
1796 .. opcode:: SAMPLE_C - Similar to the SAMPLE instruction but it
1797 performs a comparison filter. The operands to SAMPLE_C
1798 are identical to SAMPLE, except that there is an additional
1799 float32 operand, reference value, which must be a register
1800 with single-component, or a scalar literal.
1801 SAMPLE_C makes the hardware use the current samplers
1802 compare_func (in pipe_sampler_state) to compare
1803 reference value against the red component value for the
1804 surce resource at each texel that the currently configured
1805 texture filter covers based on the provided coordinates.
1806 SAMPLE_C dst, address, sampler_view.r, sampler, ref_value
1808 SAMPLE_C TEMP[0], TEMP[1], SVIEW[0].r, SAMP[0], TEMP[2].x
1810 .. opcode:: SAMPLE_C_LZ - Same as SAMPLE_C, but LOD is 0 and derivatives
1811 are ignored. The LZ stands for level-zero.
1812 SAMPLE_C_LZ dst, address, sampler_view.r, sampler, ref_value
1814 SAMPLE_C_LZ TEMP[0], TEMP[1], SVIEW[0].r, SAMP[0], TEMP[2].x
1817 .. opcode:: SAMPLE_D - SAMPLE_D is identical to the SAMPLE opcode except
1818 that the derivatives for the source address in the x
1819 direction and the y direction are provided by extra
1821 SAMPLE_D dst, address, sampler_view, sampler, der_x, der_y
1823 SAMPLE_D TEMP[0], TEMP[1], SVIEW[0], SAMP[0], TEMP[2], TEMP[3]
1825 .. opcode:: SAMPLE_L - SAMPLE_L is identical to the SAMPLE opcode except
1826 that the LOD is provided directly as a scalar value,
1827 representing no anisotropy.
1828 SAMPLE_L dst, address, sampler_view, sampler, explicit_lod
1830 SAMPLE_L TEMP[0], TEMP[1], SVIEW[0], SAMP[0], TEMP[2].x
1832 .. opcode:: GATHER4 - Gathers the four texels to be used in a bi-linear
1833 filtering operation and packs them into a single register.
1834 Only works with 2D, 2D array, cubemaps, and cubemaps arrays.
1835 For 2D textures, only the addressing modes of the sampler and
1836 the top level of any mip pyramid are used. Set W to zero.
1837 It behaves like the SAMPLE instruction, but a filtered
1838 sample is not generated. The four samples that contribute
1839 to filtering are placed into xyzw in counter-clockwise order,
1840 starting with the (u,v) texture coordinate delta at the
1841 following locations (-, +), (+, +), (+, -), (-, -), where
1842 the magnitude of the deltas are half a texel.
1845 .. opcode:: SVIEWINFO - query the dimensions of a given sampler view.
1846 dst receives width, height, depth or array size and
1847 number of mipmap levels as int4. The dst can have a writemask
1848 which will specify what info is the caller interested
1850 SVIEWINFO dst, src_mip_level, sampler_view
1852 SVIEWINFO TEMP[0], TEMP[1].x, SVIEW[0]
1853 src_mip_level is an unsigned integer scalar. If it's
1854 out of range then returns 0 for width, height and
1855 depth/array size but the total number of mipmap is
1856 still returned correctly for the given sampler view.
1857 The returned width, height and depth values are for
1858 the mipmap level selected by the src_mip_level and
1859 are in the number of texels.
1860 For 1d texture array width is in dst.x, array size
1861 is in dst.y and dst.zw are always 0.
1863 .. opcode:: SAMPLE_POS - query the position of a given sample.
1864 dst receives float4 (x, y, 0, 0) indicated where the
1865 sample is located. If the resource is not a multi-sample
1866 resource and not a render target, the result is 0.
1868 .. opcode:: SAMPLE_INFO - dst receives number of samples in x.
1869 If the resource is not a multi-sample resource and
1870 not a render target, the result is 0.
1873 .. _resourceopcodes:
1875 Resource Access Opcodes
1876 ^^^^^^^^^^^^^^^^^^^^^^^
1878 .. opcode:: LOAD - Fetch data from a shader resource
1880 Syntax: ``LOAD dst, resource, address``
1882 Example: ``LOAD TEMP[0], RES[0], TEMP[1]``
1884 Using the provided integer address, LOAD fetches data
1885 from the specified buffer or texture without any
1888 The 'address' is specified as a vector of unsigned
1889 integers. If the 'address' is out of range the result
1892 Only the first mipmap level of a resource can be read
1893 from using this instruction.
1895 For 1D or 2D texture arrays, the array index is
1896 provided as an unsigned integer in address.y or
1897 address.z, respectively. address.yz are ignored for
1898 buffers and 1D textures. address.z is ignored for 1D
1899 texture arrays and 2D textures. address.w is always
1902 .. opcode:: STORE - Write data to a shader resource
1904 Syntax: ``STORE resource, address, src``
1906 Example: ``STORE RES[0], TEMP[0], TEMP[1]``
1908 Using the provided integer address, STORE writes data
1909 to the specified buffer or texture.
1911 The 'address' is specified as a vector of unsigned
1912 integers. If the 'address' is out of range the result
1915 Only the first mipmap level of a resource can be
1916 written to using this instruction.
1918 For 1D or 2D texture arrays, the array index is
1919 provided as an unsigned integer in address.y or
1920 address.z, respectively. address.yz are ignored for
1921 buffers and 1D textures. address.z is ignored for 1D
1922 texture arrays and 2D textures. address.w is always
1926 .. _threadsyncopcodes:
1928 Inter-thread synchronization opcodes
1929 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1931 These opcodes are intended for communication between threads running
1932 within the same compute grid. For now they're only valid in compute
1935 .. opcode:: MFENCE - Memory fence
1937 Syntax: ``MFENCE resource``
1939 Example: ``MFENCE RES[0]``
1941 This opcode forces strong ordering between any memory access
1942 operations that affect the specified resource. This means that
1943 previous loads and stores (and only those) will be performed and
1944 visible to other threads before the program execution continues.
1947 .. opcode:: LFENCE - Load memory fence
1949 Syntax: ``LFENCE resource``
1951 Example: ``LFENCE RES[0]``
1953 Similar to MFENCE, but it only affects the ordering of memory loads.
1956 .. opcode:: SFENCE - Store memory fence
1958 Syntax: ``SFENCE resource``
1960 Example: ``SFENCE RES[0]``
1962 Similar to MFENCE, but it only affects the ordering of memory stores.
1965 .. opcode:: BARRIER - Thread group barrier
1969 This opcode suspends the execution of the current thread until all
1970 the remaining threads in the working group reach the same point of
1971 the program. Results are unspecified if any of the remaining
1972 threads terminates or never reaches an executed BARRIER instruction.
1980 These opcodes provide atomic variants of some common arithmetic and
1981 logical operations. In this context atomicity means that another
1982 concurrent memory access operation that affects the same memory
1983 location is guaranteed to be performed strictly before or after the
1984 entire execution of the atomic operation.
1986 For the moment they're only valid in compute programs.
1988 .. opcode:: ATOMUADD - Atomic integer addition
1990 Syntax: ``ATOMUADD dst, resource, offset, src``
1992 Example: ``ATOMUADD TEMP[0], RES[0], TEMP[1], TEMP[2]``
1994 The following operation is performed atomically on each component:
1998 dst_i = resource[offset]_i
2000 resource[offset]_i = dst_i + src_i
2003 .. opcode:: ATOMXCHG - Atomic exchange
2005 Syntax: ``ATOMXCHG dst, resource, offset, src``
2007 Example: ``ATOMXCHG TEMP[0], RES[0], TEMP[1], TEMP[2]``
2009 The following operation is performed atomically on each component:
2013 dst_i = resource[offset]_i
2015 resource[offset]_i = src_i
2018 .. opcode:: ATOMCAS - Atomic compare-and-exchange
2020 Syntax: ``ATOMCAS dst, resource, offset, cmp, src``
2022 Example: ``ATOMCAS TEMP[0], RES[0], TEMP[1], TEMP[2], TEMP[3]``
2024 The following operation is performed atomically on each component:
2028 dst_i = resource[offset]_i
2030 resource[offset]_i = (dst_i == cmp_i ? src_i : dst_i)
2033 .. opcode:: ATOMAND - Atomic bitwise And
2035 Syntax: ``ATOMAND dst, resource, offset, src``
2037 Example: ``ATOMAND TEMP[0], RES[0], TEMP[1], TEMP[2]``
2039 The following operation is performed atomically on each component:
2043 dst_i = resource[offset]_i
2045 resource[offset]_i = dst_i \& src_i
2048 .. opcode:: ATOMOR - Atomic bitwise Or
2050 Syntax: ``ATOMOR dst, resource, offset, src``
2052 Example: ``ATOMOR TEMP[0], RES[0], TEMP[1], TEMP[2]``
2054 The following operation is performed atomically on each component:
2058 dst_i = resource[offset]_i
2060 resource[offset]_i = dst_i | src_i
2063 .. opcode:: ATOMXOR - Atomic bitwise Xor
2065 Syntax: ``ATOMXOR dst, resource, offset, src``
2067 Example: ``ATOMXOR TEMP[0], RES[0], TEMP[1], TEMP[2]``
2069 The following operation is performed atomically on each component:
2073 dst_i = resource[offset]_i
2075 resource[offset]_i = dst_i \oplus src_i
2078 .. opcode:: ATOMUMIN - Atomic unsigned minimum
2080 Syntax: ``ATOMUMIN dst, resource, offset, src``
2082 Example: ``ATOMUMIN TEMP[0], RES[0], TEMP[1], TEMP[2]``
2084 The following operation is performed atomically on each component:
2088 dst_i = resource[offset]_i
2090 resource[offset]_i = (dst_i < src_i ? dst_i : src_i)
2093 .. opcode:: ATOMUMAX - Atomic unsigned maximum
2095 Syntax: ``ATOMUMAX dst, resource, offset, src``
2097 Example: ``ATOMUMAX TEMP[0], RES[0], TEMP[1], TEMP[2]``
2099 The following operation is performed atomically on each component:
2103 dst_i = resource[offset]_i
2105 resource[offset]_i = (dst_i > src_i ? dst_i : src_i)
2108 .. opcode:: ATOMIMIN - Atomic signed minimum
2110 Syntax: ``ATOMIMIN dst, resource, offset, src``
2112 Example: ``ATOMIMIN TEMP[0], RES[0], TEMP[1], TEMP[2]``
2114 The following operation is performed atomically on each component:
2118 dst_i = resource[offset]_i
2120 resource[offset]_i = (dst_i < src_i ? dst_i : src_i)
2123 .. opcode:: ATOMIMAX - Atomic signed maximum
2125 Syntax: ``ATOMIMAX dst, resource, offset, src``
2127 Example: ``ATOMIMAX TEMP[0], RES[0], TEMP[1], TEMP[2]``
2129 The following operation is performed atomically on each component:
2133 dst_i = resource[offset]_i
2135 resource[offset]_i = (dst_i > src_i ? dst_i : src_i)
2139 Explanation of symbols used
2140 ------------------------------
2147 :math:`|x|` Absolute value of `x`.
2149 :math:`\lceil x \rceil` Ceiling of `x`.
2151 clamp(x,y,z) Clamp x between y and z.
2152 (x < y) ? y : (x > z) ? z : x
2154 :math:`\lfloor x\rfloor` Floor of `x`.
2156 :math:`\log_2{x}` Logarithm of `x`, base 2.
2158 max(x,y) Maximum of x and y.
2161 min(x,y) Minimum of x and y.
2164 partialx(x) Derivative of x relative to fragment's X.
2166 partialy(x) Derivative of x relative to fragment's Y.
2168 pop() Pop from stack.
2170 :math:`x^y` `x` to the power `y`.
2172 push(x) Push x on stack.
2176 trunc(x) Truncate x, i.e. drop the fraction bits.
2183 discard Discard fragment.
2187 target Label of target instruction.
2198 Declares a register that is will be referenced as an operand in Instruction
2201 File field contains register file that is being declared and is one
2204 UsageMask field specifies which of the register components can be accessed
2205 and is one of TGSI_WRITEMASK.
2207 The Local flag specifies that a given value isn't intended for
2208 subroutine parameter passing and, as a result, the implementation
2209 isn't required to give any guarantees of it being preserved across
2210 subroutine boundaries. As it's merely a compiler hint, the
2211 implementation is free to ignore it.
2213 If Dimension flag is set to 1, a Declaration Dimension token follows.
2215 If Semantic flag is set to 1, a Declaration Semantic token follows.
2217 If Interpolate flag is set to 1, a Declaration Interpolate token follows.
2219 If file is TGSI_FILE_RESOURCE, a Declaration Resource token follows.
2221 If Array flag is set to 1, a Declaration Array token follows.
2224 ^^^^^^^^^^^^^^^^^^^^^^^^
2226 Declarations can optional have an ArrayID attribute which can be referred by
2227 indirect addressing operands. An ArrayID of zero is reserved and treaded as
2228 if no ArrayID is specified.
2230 If an indirect addressing operand refers to a specific declaration by using
2231 an ArrayID only the registers in this declaration are guaranteed to be
2232 accessed, accessing any register outside this declaration results in undefined
2233 behavior. Note that for compatibility the effective index is zero-based and
2234 not relative to the specified declaration
2236 If no ArrayID is specified with an indirect addressing operand the whole
2237 register file might be accessed by this operand. This is strongly discouraged
2238 and will prevent packing of scalar/vec2 arrays and effective alias analysis.
2240 Declaration Semantic
2241 ^^^^^^^^^^^^^^^^^^^^^^^^
2243 Vertex and fragment shader input and output registers may be labeled
2244 with semantic information consisting of a name and index.
2246 Follows Declaration token if Semantic bit is set.
2248 Since its purpose is to link a shader with other stages of the pipeline,
2249 it is valid to follow only those Declaration tokens that declare a register
2250 either in INPUT or OUTPUT file.
2252 SemanticName field contains the semantic name of the register being declared.
2253 There is no default value.
2255 SemanticIndex is an optional subscript that can be used to distinguish
2256 different register declarations with the same semantic name. The default value
2259 The meanings of the individual semantic names are explained in the following
2262 TGSI_SEMANTIC_POSITION
2263 """"""""""""""""""""""
2265 For vertex shaders, TGSI_SEMANTIC_POSITION indicates the vertex shader
2266 output register which contains the homogeneous vertex position in the clip
2267 space coordinate system. After clipping, the X, Y and Z components of the
2268 vertex will be divided by the W value to get normalized device coordinates.
2270 For fragment shaders, TGSI_SEMANTIC_POSITION is used to indicate that
2271 fragment shader input contains the fragment's window position. The X
2272 component starts at zero and always increases from left to right.
2273 The Y component starts at zero and always increases but Y=0 may either
2274 indicate the top of the window or the bottom depending on the fragment
2275 coordinate origin convention (see TGSI_PROPERTY_FS_COORD_ORIGIN).
2276 The Z coordinate ranges from 0 to 1 to represent depth from the front
2277 to the back of the Z buffer. The W component contains the reciprocol
2278 of the interpolated vertex position W component.
2280 Fragment shaders may also declare an output register with
2281 TGSI_SEMANTIC_POSITION. Only the Z component is writable. This allows
2282 the fragment shader to change the fragment's Z position.
2289 For vertex shader outputs or fragment shader inputs/outputs, this
2290 label indicates that the resister contains an R,G,B,A color.
2292 Several shader inputs/outputs may contain colors so the semantic index
2293 is used to distinguish them. For example, color[0] may be the diffuse
2294 color while color[1] may be the specular color.
2296 This label is needed so that the flat/smooth shading can be applied
2297 to the right interpolants during rasterization.
2301 TGSI_SEMANTIC_BCOLOR
2302 """"""""""""""""""""
2304 Back-facing colors are only used for back-facing polygons, and are only valid
2305 in vertex shader outputs. After rasterization, all polygons are front-facing
2306 and COLOR and BCOLOR end up occupying the same slots in the fragment shader,
2307 so all BCOLORs effectively become regular COLORs in the fragment shader.
2313 Vertex shader inputs and outputs and fragment shader inputs may be
2314 labeled with TGSI_SEMANTIC_FOG to indicate that the register contains
2315 a fog coordinate in the form (F, 0, 0, 1). Typically, the fragment
2316 shader will use the fog coordinate to compute a fog blend factor which
2317 is used to blend the normal fragment color with a constant fog color.
2319 Only the first component matters when writing from the vertex shader;
2320 the driver will ensure that the coordinate is in this format when used
2321 as a fragment shader input.
2327 Vertex shader input and output registers may be labeled with
2328 TGIS_SEMANTIC_PSIZE to indicate that the register contains a point size
2329 in the form (S, 0, 0, 1). The point size controls the width or diameter
2330 of points for rasterization. This label cannot be used in fragment
2333 When using this semantic, be sure to set the appropriate state in the
2334 :ref:`rasterizer` first.
2337 TGSI_SEMANTIC_TEXCOORD
2338 """"""""""""""""""""""
2340 Only available if PIPE_CAP_TGSI_TEXCOORD is exposed !
2342 Vertex shader outputs and fragment shader inputs may be labeled with
2343 this semantic to make them replaceable by sprite coordinates via the
2344 sprite_coord_enable state in the :ref:`rasterizer`.
2345 The semantic index permitted with this semantic is limited to <= 7.
2347 If the driver does not support TEXCOORD, sprite coordinate replacement
2348 applies to inputs with the GENERIC semantic instead.
2350 The intended use case for this semantic is gl_TexCoord.
2353 TGSI_SEMANTIC_PCOORD
2354 """"""""""""""""""""
2356 Only available if PIPE_CAP_TGSI_TEXCOORD is exposed !
2358 Fragment shader inputs may be labeled with TGSI_SEMANTIC_PCOORD to indicate
2359 that the register contains sprite coordinates in the form (x, y, 0, 1), if
2360 the current primitive is a point and point sprites are enabled. Otherwise,
2361 the contents of the register are undefined.
2363 The intended use case for this semantic is gl_PointCoord.
2366 TGSI_SEMANTIC_GENERIC
2367 """""""""""""""""""""
2369 All vertex/fragment shader inputs/outputs not labeled with any other
2370 semantic label can be considered to be generic attributes. Typical
2371 uses of generic inputs/outputs are texcoords and user-defined values.
2374 TGSI_SEMANTIC_NORMAL
2375 """"""""""""""""""""
2377 Indicates that a vertex shader input is a normal vector. This is
2378 typically only used for legacy graphics APIs.
2384 This label applies to fragment shader inputs only and indicates that
2385 the register contains front/back-face information of the form (F, 0,
2386 0, 1). The first component will be positive when the fragment belongs
2387 to a front-facing polygon, and negative when the fragment belongs to a
2388 back-facing polygon.
2391 TGSI_SEMANTIC_EDGEFLAG
2392 """"""""""""""""""""""
2394 For vertex shaders, this sematic label indicates that an input or
2395 output is a boolean edge flag. The register layout is [F, x, x, x]
2396 where F is 0.0 or 1.0 and x = don't care. Normally, the vertex shader
2397 simply copies the edge flag input to the edgeflag output.
2399 Edge flags are used to control which lines or points are actually
2400 drawn when the polygon mode converts triangles/quads/polygons into
2404 TGSI_SEMANTIC_STENCIL
2405 """""""""""""""""""""
2407 For fragment shaders, this semantic label indicates that an output
2408 is a writable stencil reference value. Only the Y component is writable.
2409 This allows the fragment shader to change the fragments stencilref value.
2412 TGSI_SEMANTIC_VIEWPORT_INDEX
2413 """"""""""""""""""""""""""""
2415 For geometry shaders, this semantic label indicates that an output
2416 contains the index of the viewport (and scissor) to use.
2417 Only the X value is used.
2423 For geometry shaders, this semantic label indicates that an output
2424 contains the layer value to use for the color and depth/stencil surfaces.
2425 Only the X value is used. (Also known as rendertarget array index.)
2429 Declaration Interpolate
2430 ^^^^^^^^^^^^^^^^^^^^^^^
2432 This token is only valid for fragment shader INPUT declarations.
2434 The Interpolate field specifes the way input is being interpolated by
2435 the rasteriser and is one of TGSI_INTERPOLATE_*.
2437 The CylindricalWrap bitfield specifies which register components
2438 should be subject to cylindrical wrapping when interpolating by the
2439 rasteriser. If TGSI_CYLINDRICAL_WRAP_X is set to 1, the X component
2440 should be interpolated according to cylindrical wrapping rules.
2443 Declaration Sampler View
2444 ^^^^^^^^^^^^^^^^^^^^^^^^
2446 Follows Declaration token if file is TGSI_FILE_SAMPLER_VIEW.
2448 DCL SVIEW[#], resource, type(s)
2450 Declares a shader input sampler view and assigns it to a SVIEW[#]
2453 resource can be one of BUFFER, 1D, 2D, 3D, 1DArray and 2DArray.
2455 type must be 1 or 4 entries (if specifying on a per-component
2456 level) out of UNORM, SNORM, SINT, UINT and FLOAT.
2459 Declaration Resource
2460 ^^^^^^^^^^^^^^^^^^^^
2462 Follows Declaration token if file is TGSI_FILE_RESOURCE.
2464 DCL RES[#], resource [, WR] [, RAW]
2466 Declares a shader input resource and assigns it to a RES[#]
2469 resource can be one of BUFFER, 1D, 2D, 3D, CUBE, 1DArray and
2472 If the RAW keyword is not specified, the texture data will be
2473 subject to conversion, swizzling and scaling as required to yield
2474 the specified data type from the physical data format of the bound
2477 If the RAW keyword is specified, no channel conversion will be
2478 performed: the values read for each of the channels (X,Y,Z,W) will
2479 correspond to consecutive words in the same order and format
2480 they're found in memory. No element-to-address conversion will be
2481 performed either: the value of the provided X coordinate will be
2482 interpreted in byte units instead of texel units. The result of
2483 accessing a misaligned address is undefined.
2485 Usage of the STORE opcode is only allowed if the WR (writable) flag
2490 ^^^^^^^^^^^^^^^^^^^^^^^^
2493 Properties are general directives that apply to the whole TGSI program.
2498 Specifies the fragment shader TGSI_SEMANTIC_POSITION coordinate origin.
2499 The default value is UPPER_LEFT.
2501 If UPPER_LEFT, the position will be (0,0) at the upper left corner and
2502 increase downward and rightward.
2503 If LOWER_LEFT, the position will be (0,0) at the lower left corner and
2504 increase upward and rightward.
2506 OpenGL defaults to LOWER_LEFT, and is configurable with the
2507 GL_ARB_fragment_coord_conventions extension.
2509 DirectX 9/10 use UPPER_LEFT.
2511 FS_COORD_PIXEL_CENTER
2512 """""""""""""""""""""
2514 Specifies the fragment shader TGSI_SEMANTIC_POSITION pixel center convention.
2515 The default value is HALF_INTEGER.
2517 If HALF_INTEGER, the fractionary part of the position will be 0.5
2518 If INTEGER, the fractionary part of the position will be 0.0
2520 Note that this does not affect the set of fragments generated by
2521 rasterization, which is instead controlled by half_pixel_center in the
2524 OpenGL defaults to HALF_INTEGER, and is configurable with the
2525 GL_ARB_fragment_coord_conventions extension.
2527 DirectX 9 uses INTEGER.
2528 DirectX 10 uses HALF_INTEGER.
2530 FS_COLOR0_WRITES_ALL_CBUFS
2531 """"""""""""""""""""""""""
2532 Specifies that writes to the fragment shader color 0 are replicated to all
2533 bound cbufs. This facilitates OpenGL's fragColor output vs fragData[0] where
2534 fragData is directed to a single color buffer, but fragColor is broadcast.
2537 """"""""""""""""""""""""""
2538 If this property is set on the program bound to the shader stage before the
2539 fragment shader, user clip planes should have no effect (be disabled) even if
2540 that shader does not write to any clip distance outputs and the rasterizer's
2541 clip_plane_enable is non-zero.
2542 This property is only supported by drivers that also support shader clip
2544 This is useful for APIs that don't have UCPs and where clip distances written
2545 by a shader cannot be disabled.
2548 Texture Sampling and Texture Formats
2549 ------------------------------------
2551 This table shows how texture image components are returned as (x,y,z,w) tuples
2552 by TGSI texture instructions, such as :opcode:`TEX`, :opcode:`TXD`, and
2553 :opcode:`TXP`. For reference, OpenGL and Direct3D conventions are shown as
2556 +--------------------+--------------+--------------------+--------------+
2557 | Texture Components | Gallium | OpenGL | Direct3D 9 |
2558 +====================+==============+====================+==============+
2559 | R | (r, 0, 0, 1) | (r, 0, 0, 1) | (r, 1, 1, 1) |
2560 +--------------------+--------------+--------------------+--------------+
2561 | RG | (r, g, 0, 1) | (r, g, 0, 1) | (r, g, 1, 1) |
2562 +--------------------+--------------+--------------------+--------------+
2563 | RGB | (r, g, b, 1) | (r, g, b, 1) | (r, g, b, 1) |
2564 +--------------------+--------------+--------------------+--------------+
2565 | RGBA | (r, g, b, a) | (r, g, b, a) | (r, g, b, a) |
2566 +--------------------+--------------+--------------------+--------------+
2567 | A | (0, 0, 0, a) | (0, 0, 0, a) | (0, 0, 0, a) |
2568 +--------------------+--------------+--------------------+--------------+
2569 | L | (l, l, l, 1) | (l, l, l, 1) | (l, l, l, 1) |
2570 +--------------------+--------------+--------------------+--------------+
2571 | LA | (l, l, l, a) | (l, l, l, a) | (l, l, l, a) |
2572 +--------------------+--------------+--------------------+--------------+
2573 | I | (i, i, i, i) | (i, i, i, i) | N/A |
2574 +--------------------+--------------+--------------------+--------------+
2575 | UV | XXX TBD | (0, 0, 0, 1) | (u, v, 1, 1) |
2576 | | | [#envmap-bumpmap]_ | |
2577 +--------------------+--------------+--------------------+--------------+
2578 | Z | XXX TBD | (z, z, z, 1) | (0, z, 0, 1) |
2579 | | | [#depth-tex-mode]_ | |
2580 +--------------------+--------------+--------------------+--------------+
2581 | S | (s, s, s, s) | unknown | unknown |
2582 +--------------------+--------------+--------------------+--------------+
2584 .. [#envmap-bumpmap] http://www.opengl.org/registry/specs/ATI/envmap_bumpmap.txt
2585 .. [#depth-tex-mode] the default is (z, z, z, 1) but may also be (0, 0, 0, z)
2586 or (z, z, z, z) depending on the value of GL_DEPTH_TEXTURE_MODE.