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.
30 ^^^^^^^^^^^^^^^^^^^^^^^^^
32 These opcodes are guaranteed to be available regardless of the driver being
35 .. opcode:: ARL - Address Register Load
39 dst.x = \lfloor src.x\rfloor
41 dst.y = \lfloor src.y\rfloor
43 dst.z = \lfloor src.z\rfloor
45 dst.w = \lfloor src.w\rfloor
48 .. opcode:: MOV - Move
61 .. opcode:: LIT - Light Coefficients
69 dst.z = (src.x > 0) ? max(src.y, 0)^{clamp(src.w, -128, 128))} : 0
74 .. opcode:: RCP - Reciprocal
76 This instruction replicates its result.
83 .. opcode:: RSQ - Reciprocal Square Root
85 This instruction replicates its result.
89 dst = \frac{1}{\sqrt{|src.x|}}
92 .. opcode:: EXP - Approximate Exponential Base 2
96 dst.x = 2^{\lfloor src.x\rfloor}
98 dst.y = src.x - \lfloor src.x\rfloor
105 .. opcode:: LOG - Approximate Logarithm Base 2
109 dst.x = \lfloor\log_2{|src.x|}\rfloor
111 dst.y = \frac{|src.x|}{2^{\lfloor\log_2{|src.x|}\rfloor}}
113 dst.z = \log_2{|src.x|}
118 .. opcode:: MUL - Multiply
122 dst.x = src0.x \times src1.x
124 dst.y = src0.y \times src1.y
126 dst.z = src0.z \times src1.z
128 dst.w = src0.w \times src1.w
131 .. opcode:: ADD - Add
135 dst.x = src0.x + src1.x
137 dst.y = src0.y + src1.y
139 dst.z = src0.z + src1.z
141 dst.w = src0.w + src1.w
144 .. opcode:: DP3 - 3-component Dot Product
146 This instruction replicates its result.
150 dst = src0.x \times src1.x + src0.y \times src1.y + src0.z \times src1.z
153 .. opcode:: DP4 - 4-component Dot Product
155 This instruction replicates its result.
159 dst = src0.x \times src1.x + src0.y \times src1.y + src0.z \times src1.z + src0.w \times src1.w
162 .. opcode:: DST - Distance Vector
168 dst.y = src0.y \times src1.y
175 .. opcode:: MIN - Minimum
179 dst.x = min(src0.x, src1.x)
181 dst.y = min(src0.y, src1.y)
183 dst.z = min(src0.z, src1.z)
185 dst.w = min(src0.w, src1.w)
188 .. opcode:: MAX - Maximum
192 dst.x = max(src0.x, src1.x)
194 dst.y = max(src0.y, src1.y)
196 dst.z = max(src0.z, src1.z)
198 dst.w = max(src0.w, src1.w)
201 .. opcode:: SLT - Set On Less Than
205 dst.x = (src0.x < src1.x) ? 1 : 0
207 dst.y = (src0.y < src1.y) ? 1 : 0
209 dst.z = (src0.z < src1.z) ? 1 : 0
211 dst.w = (src0.w < src1.w) ? 1 : 0
214 .. opcode:: SGE - Set On Greater Equal Than
218 dst.x = (src0.x >= src1.x) ? 1 : 0
220 dst.y = (src0.y >= src1.y) ? 1 : 0
222 dst.z = (src0.z >= src1.z) ? 1 : 0
224 dst.w = (src0.w >= src1.w) ? 1 : 0
227 .. opcode:: MAD - Multiply And Add
231 dst.x = src0.x \times src1.x + src2.x
233 dst.y = src0.y \times src1.y + src2.y
235 dst.z = src0.z \times src1.z + src2.z
237 dst.w = src0.w \times src1.w + src2.w
240 .. opcode:: SUB - Subtract
244 dst.x = src0.x - src1.x
246 dst.y = src0.y - src1.y
248 dst.z = src0.z - src1.z
250 dst.w = src0.w - src1.w
253 .. opcode:: LRP - Linear Interpolate
257 dst.x = src0.x \times src1.x + (1 - src0.x) \times src2.x
259 dst.y = src0.y \times src1.y + (1 - src0.y) \times src2.y
261 dst.z = src0.z \times src1.z + (1 - src0.z) \times src2.z
263 dst.w = src0.w \times src1.w + (1 - src0.w) \times src2.w
266 .. opcode:: CND - Condition
270 dst.x = (src2.x > 0.5) ? src0.x : src1.x
272 dst.y = (src2.y > 0.5) ? src0.y : src1.y
274 dst.z = (src2.z > 0.5) ? src0.z : src1.z
276 dst.w = (src2.w > 0.5) ? src0.w : src1.w
279 .. opcode:: DP2A - 2-component Dot Product And Add
283 dst.x = src0.x \times src1.x + src0.y \times src1.y + src2.x
285 dst.y = src0.x \times src1.x + src0.y \times src1.y + src2.x
287 dst.z = src0.x \times src1.x + src0.y \times src1.y + src2.x
289 dst.w = src0.x \times src1.x + src0.y \times src1.y + src2.x
292 .. opcode:: FRC - Fraction
296 dst.x = src.x - \lfloor src.x\rfloor
298 dst.y = src.y - \lfloor src.y\rfloor
300 dst.z = src.z - \lfloor src.z\rfloor
302 dst.w = src.w - \lfloor src.w\rfloor
305 .. opcode:: CLAMP - Clamp
309 dst.x = clamp(src0.x, src1.x, src2.x)
311 dst.y = clamp(src0.y, src1.y, src2.y)
313 dst.z = clamp(src0.z, src1.z, src2.z)
315 dst.w = clamp(src0.w, src1.w, src2.w)
318 .. opcode:: FLR - Floor
320 This is identical to :opcode:`ARL`.
324 dst.x = \lfloor src.x\rfloor
326 dst.y = \lfloor src.y\rfloor
328 dst.z = \lfloor src.z\rfloor
330 dst.w = \lfloor src.w\rfloor
333 .. opcode:: ROUND - Round
346 .. opcode:: EX2 - Exponential Base 2
348 This instruction replicates its result.
355 .. opcode:: LG2 - Logarithm Base 2
357 This instruction replicates its result.
364 .. opcode:: POW - Power
366 This instruction replicates its result.
370 dst = src0.x^{src1.x}
372 .. opcode:: XPD - Cross Product
376 dst.x = src0.y \times src1.z - src1.y \times src0.z
378 dst.y = src0.z \times src1.x - src1.z \times src0.x
380 dst.z = src0.x \times src1.y - src1.x \times src0.y
385 .. opcode:: ABS - Absolute
398 .. opcode:: RCC - Reciprocal Clamped
400 This instruction replicates its result.
402 XXX cleanup on aisle three
406 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)
409 .. opcode:: DPH - Homogeneous Dot Product
411 This instruction replicates its result.
415 dst = src0.x \times src1.x + src0.y \times src1.y + src0.z \times src1.z + src1.w
418 .. opcode:: COS - Cosine
420 This instruction replicates its result.
427 .. opcode:: DDX - Derivative Relative To X
431 dst.x = partialx(src.x)
433 dst.y = partialx(src.y)
435 dst.z = partialx(src.z)
437 dst.w = partialx(src.w)
440 .. opcode:: DDY - Derivative Relative To Y
444 dst.x = partialy(src.x)
446 dst.y = partialy(src.y)
448 dst.z = partialy(src.z)
450 dst.w = partialy(src.w)
453 .. opcode:: KILP - Predicated Discard
458 .. opcode:: PK2H - Pack Two 16-bit Floats
463 .. opcode:: PK2US - Pack Two Unsigned 16-bit Scalars
468 .. opcode:: PK4B - Pack Four Signed 8-bit Scalars
473 .. opcode:: PK4UB - Pack Four Unsigned 8-bit Scalars
478 .. opcode:: RFL - Reflection Vector
482 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
484 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
486 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
492 Considered for removal.
495 .. opcode:: SEQ - Set On Equal
499 dst.x = (src0.x == src1.x) ? 1 : 0
501 dst.y = (src0.y == src1.y) ? 1 : 0
503 dst.z = (src0.z == src1.z) ? 1 : 0
505 dst.w = (src0.w == src1.w) ? 1 : 0
508 .. opcode:: SFL - Set On False
510 This instruction replicates its result.
518 Considered for removal.
521 .. opcode:: SGT - Set On Greater Than
525 dst.x = (src0.x > src1.x) ? 1 : 0
527 dst.y = (src0.y > src1.y) ? 1 : 0
529 dst.z = (src0.z > src1.z) ? 1 : 0
531 dst.w = (src0.w > src1.w) ? 1 : 0
534 .. opcode:: SIN - Sine
536 This instruction replicates its result.
543 .. opcode:: SLE - Set On Less Equal Than
547 dst.x = (src0.x <= src1.x) ? 1 : 0
549 dst.y = (src0.y <= src1.y) ? 1 : 0
551 dst.z = (src0.z <= src1.z) ? 1 : 0
553 dst.w = (src0.w <= src1.w) ? 1 : 0
556 .. opcode:: SNE - Set On Not Equal
560 dst.x = (src0.x != src1.x) ? 1 : 0
562 dst.y = (src0.y != src1.y) ? 1 : 0
564 dst.z = (src0.z != src1.z) ? 1 : 0
566 dst.w = (src0.w != src1.w) ? 1 : 0
569 .. opcode:: STR - Set On True
571 This instruction replicates its result.
578 .. opcode:: TEX - Texture Lookup
586 dst = texture_sample(unit, coord, bias)
588 for array textures src0.y contains the slice for 1D,
589 and src0.z contain the slice for 2D.
590 for shadow textures with no arrays, src0.z contains
592 for shadow textures with arrays, src0.z contains
593 the reference value for 1D arrays, and src0.w contains
594 the reference value for 2D arrays.
595 There is no way to pass a bias in the .w value for
596 shadow arrays, and GLSL doesn't allow this.
597 GLSL does allow cube shadows maps to take a bias value,
598 and we have to determine how this will look in TGSI.
600 .. opcode:: TXD - Texture Lookup with Derivatives
612 dst = texture_sample_deriv(unit, coord, bias, ddx, ddy)
615 .. opcode:: TXP - Projective Texture Lookup
619 coord.x = src0.x / src.w
621 coord.y = src0.y / src.w
623 coord.z = src0.z / src.w
629 dst = texture_sample(unit, coord, bias)
632 .. opcode:: UP2H - Unpack Two 16-Bit Floats
638 Considered for removal.
640 .. opcode:: UP2US - Unpack Two Unsigned 16-Bit Scalars
646 Considered for removal.
648 .. opcode:: UP4B - Unpack Four Signed 8-Bit Values
654 Considered for removal.
656 .. opcode:: UP4UB - Unpack Four Unsigned 8-Bit Scalars
662 Considered for removal.
664 .. opcode:: X2D - 2D Coordinate Transformation
668 dst.x = src0.x + src1.x \times src2.x + src1.y \times src2.y
670 dst.y = src0.y + src1.x \times src2.z + src1.y \times src2.w
672 dst.z = src0.x + src1.x \times src2.x + src1.y \times src2.y
674 dst.w = src0.y + src1.x \times src2.z + src1.y \times src2.w
678 Considered for removal.
681 .. opcode:: ARA - Address Register Add
687 Considered for removal.
689 .. opcode:: ARR - Address Register Load With Round
702 .. opcode:: BRA - Branch
708 Considered for removal.
710 .. opcode:: CAL - Subroutine Call
716 .. opcode:: RET - Subroutine Call Return
721 .. opcode:: SSG - Set Sign
725 dst.x = (src.x > 0) ? 1 : (src.x < 0) ? -1 : 0
727 dst.y = (src.y > 0) ? 1 : (src.y < 0) ? -1 : 0
729 dst.z = (src.z > 0) ? 1 : (src.z < 0) ? -1 : 0
731 dst.w = (src.w > 0) ? 1 : (src.w < 0) ? -1 : 0
734 .. opcode:: CMP - Compare
738 dst.x = (src0.x < 0) ? src1.x : src2.x
740 dst.y = (src0.y < 0) ? src1.y : src2.y
742 dst.z = (src0.z < 0) ? src1.z : src2.z
744 dst.w = (src0.w < 0) ? src1.w : src2.w
747 .. opcode:: KIL - Conditional Discard
751 if (src.x < 0 || src.y < 0 || src.z < 0 || src.w < 0)
756 .. opcode:: SCS - Sine Cosine
769 .. opcode:: TXB - Texture Lookup With Bias
783 dst = texture_sample(unit, coord, bias)
786 .. opcode:: NRM - 3-component Vector Normalise
790 dst.x = src.x / (src.x \times src.x + src.y \times src.y + src.z \times src.z)
792 dst.y = src.y / (src.x \times src.x + src.y \times src.y + src.z \times src.z)
794 dst.z = src.z / (src.x \times src.x + src.y \times src.y + src.z \times src.z)
799 .. opcode:: DIV - Divide
803 dst.x = \frac{src0.x}{src1.x}
805 dst.y = \frac{src0.y}{src1.y}
807 dst.z = \frac{src0.z}{src1.z}
809 dst.w = \frac{src0.w}{src1.w}
812 .. opcode:: DP2 - 2-component Dot Product
814 This instruction replicates its result.
818 dst = src0.x \times src1.x + src0.y \times src1.y
821 .. opcode:: TXL - Texture Lookup With explicit LOD
835 dst = texture_sample(unit, coord, lod)
838 .. opcode:: BRK - Break
848 .. opcode:: ELSE - Else
853 .. opcode:: ENDIF - End If
858 .. opcode:: PUSHA - Push Address Register On Stack
867 Considered for cleanup.
871 Considered for removal.
873 .. opcode:: POPA - Pop Address Register From Stack
882 Considered for cleanup.
886 Considered for removal.
890 ^^^^^^^^^^^^^^^^^^^^^^^^
892 These opcodes are primarily provided for special-use computational shaders.
893 Support for these opcodes indicated by a special pipe capability bit (TBD).
895 XXX so let's discuss it, yeah?
897 .. opcode:: CEIL - Ceiling
901 dst.x = \lceil src.x\rceil
903 dst.y = \lceil src.y\rceil
905 dst.z = \lceil src.z\rceil
907 dst.w = \lceil src.w\rceil
910 .. opcode:: I2F - Integer To Float
914 dst.x = (float) src.x
916 dst.y = (float) src.y
918 dst.z = (float) src.z
920 dst.w = (float) src.w
923 .. opcode:: NOT - Bitwise Not
936 .. opcode:: TRUNC - Truncate
949 .. opcode:: SHL - Shift Left
953 dst.x = src0.x << src1.x
955 dst.y = src0.y << src1.x
957 dst.z = src0.z << src1.x
959 dst.w = src0.w << src1.x
962 .. opcode:: SHR - Shift Right
966 dst.x = src0.x >> src1.x
968 dst.y = src0.y >> src1.x
970 dst.z = src0.z >> src1.x
972 dst.w = src0.w >> src1.x
975 .. opcode:: AND - Bitwise And
979 dst.x = src0.x & src1.x
981 dst.y = src0.y & src1.y
983 dst.z = src0.z & src1.z
985 dst.w = src0.w & src1.w
988 .. opcode:: OR - Bitwise Or
992 dst.x = src0.x | src1.x
994 dst.y = src0.y | src1.y
996 dst.z = src0.z | src1.z
998 dst.w = src0.w | src1.w
1001 .. opcode:: MOD - Modulus
1005 dst.x = src0.x \bmod src1.x
1007 dst.y = src0.y \bmod src1.y
1009 dst.z = src0.z \bmod src1.z
1011 dst.w = src0.w \bmod src1.w
1014 .. opcode:: XOR - Bitwise Xor
1018 dst.x = src0.x \oplus src1.x
1020 dst.y = src0.y \oplus src1.y
1022 dst.z = src0.z \oplus src1.z
1024 dst.w = src0.w \oplus src1.w
1027 .. opcode:: UCMP - Integer Conditional Move
1031 dst.x = src0.x ? src1.x : src2.x
1033 dst.y = src0.y ? src1.y : src2.y
1035 dst.z = src0.z ? src1.z : src2.z
1037 dst.w = src0.w ? src1.w : src2.w
1040 .. opcode:: UARL - Integer Address Register Load
1042 Moves the contents of the source register, assumed to be an integer, into the
1043 destination register, which is assumed to be an address (ADDR) register.
1046 .. opcode:: IABS - Integer Absolute Value
1059 .. opcode:: SAD - Sum Of Absolute Differences
1063 dst.x = |src0.x - src1.x| + src2.x
1065 dst.y = |src0.y - src1.y| + src2.y
1067 dst.z = |src0.z - src1.z| + src2.z
1069 dst.w = |src0.w - src1.w| + src2.w
1072 .. opcode:: TXF - Texel Fetch (as per NV_gpu_shader4), extract a single texel
1073 from a specified texture image. The source sampler may
1074 not be a CUBE or SHADOW.
1075 src 0 is a four-component signed integer vector used to
1076 identify the single texel accessed. 3 components + level.
1077 src 1 is a 3 component constant signed integer vector,
1078 with each component only have a range of
1079 -8..+8 (hw only seems to deal with this range, interface
1080 allows for up to unsigned int).
1081 TXF(uint_vec coord, int_vec offset).
1084 .. opcode:: TXQ - Texture Size Query (as per NV_gpu_program4)
1085 retrieve the dimensions of the texture
1086 depending on the target. For 1D (width), 2D/RECT/CUBE
1087 (width, height), 3D (width, height, depth),
1088 1D array (width, layers), 2D array (width, height, layers)
1094 dst.x = texture_width(unit, lod)
1096 dst.y = texture_height(unit, lod)
1098 dst.z = texture_depth(unit, lod)
1101 .. opcode:: CONT - Continue
1107 Support for CONT is determined by a special capability bit,
1108 ``TGSI_CONT_SUPPORTED``. See :ref:`Screen` for more information.
1112 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1114 These opcodes are only supported in geometry shaders; they have no meaning
1115 in any other type of shader.
1117 .. opcode:: EMIT - Emit
1122 .. opcode:: ENDPRIM - End Primitive
1130 These opcodes are part of :term:`GLSL`'s opcode set. Support for these
1131 opcodes is determined by a special capability bit, ``GLSL``.
1133 .. opcode:: BGNLOOP - Begin a Loop
1138 .. opcode:: BGNSUB - Begin Subroutine
1143 .. opcode:: ENDLOOP - End a Loop
1148 .. opcode:: ENDSUB - End Subroutine
1153 .. opcode:: NOP - No Operation
1158 .. opcode:: NRM4 - 4-component Vector Normalise
1160 This instruction replicates its result.
1164 dst = \frac{src.x}{src.x \times src.x + src.y \times src.y + src.z \times src.z + src.w \times src.w}
1172 .. opcode:: CALLNZ - Subroutine Call If Not Zero
1177 .. opcode:: IFC - If
1182 .. opcode:: BREAKC - Break Conditional
1191 The double-precision opcodes reinterpret four-component vectors into
1192 two-component vectors with doubled precision in each component.
1194 Support for these opcodes is XXX undecided. :T
1196 .. opcode:: DADD - Add
1200 dst.xy = src0.xy + src1.xy
1202 dst.zw = src0.zw + src1.zw
1205 .. opcode:: DDIV - Divide
1209 dst.xy = src0.xy / src1.xy
1211 dst.zw = src0.zw / src1.zw
1213 .. opcode:: DSEQ - Set on Equal
1217 dst.xy = src0.xy == src1.xy ? 1.0F : 0.0F
1219 dst.zw = src0.zw == src1.zw ? 1.0F : 0.0F
1221 .. opcode:: DSLT - Set on Less than
1225 dst.xy = src0.xy < src1.xy ? 1.0F : 0.0F
1227 dst.zw = src0.zw < src1.zw ? 1.0F : 0.0F
1229 .. opcode:: DFRAC - Fraction
1233 dst.xy = src.xy - \lfloor src.xy\rfloor
1235 dst.zw = src.zw - \lfloor src.zw\rfloor
1238 .. opcode:: DFRACEXP - Convert Number to Fractional and Integral Components
1240 Like the ``frexp()`` routine in many math libraries, this opcode stores the
1241 exponent of its source to ``dst0``, and the significand to ``dst1``, such that
1242 :math:`dst1 \times 2^{dst0} = src` .
1246 dst0.xy = exp(src.xy)
1248 dst1.xy = frac(src.xy)
1250 dst0.zw = exp(src.zw)
1252 dst1.zw = frac(src.zw)
1254 .. opcode:: DLDEXP - Multiply Number by Integral Power of 2
1256 This opcode is the inverse of :opcode:`DFRACEXP`.
1260 dst.xy = src0.xy \times 2^{src1.xy}
1262 dst.zw = src0.zw \times 2^{src1.zw}
1264 .. opcode:: DMIN - Minimum
1268 dst.xy = min(src0.xy, src1.xy)
1270 dst.zw = min(src0.zw, src1.zw)
1272 .. opcode:: DMAX - Maximum
1276 dst.xy = max(src0.xy, src1.xy)
1278 dst.zw = max(src0.zw, src1.zw)
1280 .. opcode:: DMUL - Multiply
1284 dst.xy = src0.xy \times src1.xy
1286 dst.zw = src0.zw \times src1.zw
1289 .. opcode:: DMAD - Multiply And Add
1293 dst.xy = src0.xy \times src1.xy + src2.xy
1295 dst.zw = src0.zw \times src1.zw + src2.zw
1298 .. opcode:: DRCP - Reciprocal
1302 dst.xy = \frac{1}{src.xy}
1304 dst.zw = \frac{1}{src.zw}
1306 .. opcode:: DSQRT - Square Root
1310 dst.xy = \sqrt{src.xy}
1312 dst.zw = \sqrt{src.zw}
1315 .. _resourceopcodes:
1317 Resource Access Opcodes
1318 ^^^^^^^^^^^^^^^^^^^^^^^^
1320 Those opcodes follow very closely semantics of the respective Direct3D
1321 instructions. If in doubt double check Direct3D documentation.
1323 .. opcode:: LOAD - Simplified alternative to the "SAMPLE" instruction.
1324 Using the provided integer address, LOAD fetches data
1325 from the specified buffer/texture without any filtering.
1326 The source data may come from any resource type other
1328 LOAD dst, address, resource
1330 LOAD TEMP[0], TEMP[1], RES[0]
1331 The 'address' is specified as unsigned integers. If the
1332 'address' is out of range [0...(# texels - 1)] the
1333 result of the fetch is always 0 in all components.
1334 As such the instruction doesn't honor address wrap
1335 modes, in cases where that behavior is desirable
1336 'sample' instruction should be used.
1337 address.w always provides an unsigned integer mipmap
1338 level. If the value is out of the range then the
1339 instruction always returns 0 in all components.
1340 address.yz are ignored for buffers and 1d textures.
1341 address.z is ignored for 1d texture arrays and 2d
1343 For 1D texture arrays address.y provides the array
1344 index (also as unsigned integer). If the value is
1345 out of the range of available array indices
1346 [0... (array size - 1)] then the opcode always returns
1347 0 in all components.
1348 For 2D texture arrays address.z provides the array
1349 index, otherwise it exhibits the same behavior as in
1350 the case for 1D texture arrays.
1351 The exeact semantics of the source address are presented
1353 resource type X Y Z W
1354 ------------- ------------------------
1355 PIPE_BUFFER x ignored
1356 PIPE_TEXTURE_1D x mpl
1357 PIPE_TEXTURE_2D x y mpl
1358 PIPE_TEXTURE_3D x y z mpl
1359 PIPE_TEXTURE_RECT x y mpl
1360 PIPE_TEXTURE_CUBE not allowed as source
1361 PIPE_TEXTURE_1D_ARRAY x idx mpl
1362 PIPE_TEXTURE_2D_ARRAY x y idx mpl
1364 Where 'mpl' is a mipmap level and 'idx' is the
1368 .. opcode:: LOAD_MS - Just like LOAD but allows fetch data from
1369 multi-sampled surfaces.
1371 .. opcode:: SAMPLE - Using provided address, sample data from the
1372 specified texture using the filtering mode identified
1373 by the gven sampler. The source data may come from
1374 any resource type other than buffers.
1375 SAMPLE dst, address, resource, sampler
1377 SAMPLE TEMP[0], TEMP[1], RES[0], SAMP[0]
1379 .. opcode:: SAMPLE_B - Just like the SAMPLE instruction with the
1380 exception that an additiona bias is applied to the
1381 level of detail computed as part of the instruction
1383 SAMPLE_B dst, address, resource, sampler, lod_bias
1385 SAMPLE_B TEMP[0], TEMP[1], RES[0], SAMP[0], TEMP[2].x
1387 .. opcode:: SAMPLE_C - Similar to the SAMPLE instruction but it
1388 performs a comparison filter. The operands to SAMPLE_C
1389 are identical to SAMPLE, except that tere is an additional
1390 float32 operand, reference value, which must be a register
1391 with single-component, or a scalar literal.
1392 SAMPLE_C makes the hardware use the current samplers
1393 compare_func (in pipe_sampler_state) to compare
1394 reference value against the red component value for the
1395 surce resource at each texel that the currently configured
1396 texture filter covers based on the provided coordinates.
1397 SAMPLE_C dst, address, resource.r, sampler, ref_value
1399 SAMPLE_C TEMP[0], TEMP[1], RES[0].r, SAMP[0], TEMP[2].x
1401 .. opcode:: SAMPLE_C_LZ - Same as SAMPLE_C, but LOD is 0 and derivatives
1402 are ignored. The LZ stands for level-zero.
1403 SAMPLE_C_LZ dst, address, resource.r, sampler, ref_value
1405 SAMPLE_C_LZ TEMP[0], TEMP[1], RES[0].r, SAMP[0], TEMP[2].x
1408 .. opcode:: SAMPLE_D - SAMPLE_D is identical to the SAMPLE opcode except
1409 that the derivatives for the source address in the x
1410 direction and the y direction are provided by extra
1412 SAMPLE_D dst, address, resource, sampler, der_x, der_y
1414 SAMPLE_D TEMP[0], TEMP[1], RES[0], SAMP[0], TEMP[2], TEMP[3]
1416 .. opcode:: SAMPLE_L - SAMPLE_L is identical to the SAMPLE opcode except
1417 that the LOD is provided directly as a scalar value,
1418 representing no anisotropy. Source addresses A channel
1420 SAMPLE_L dst, address, resource, sampler
1422 SAMPLE_L TEMP[0], TEMP[1], RES[0], SAMP[0]
1425 .. opcode:: GATHER4 - Gathers the four texels to be used in a bi-linear
1426 filtering operation and packs them into a single register.
1427 Only works with 2D, 2D array, cubemaps, and cubemaps arrays.
1428 For 2D textures, only the addressing modes of the sampler and
1429 the top level of any mip pyramid are used. Set W to zero.
1430 It behaves like the SAMPLE instruction, but a filtered
1431 sample is not generated. The four samples that contribute
1432 to filtering are placed into xyzw in counter-clockwise order,
1433 starting with the (u,v) texture coordinate delta at the
1434 following locations (-, +), (+, +), (+, -), (-, -), where
1435 the magnitude of the deltas are half a texel.
1438 .. opcode:: RESINFO - query the dimensions of a given input buffer.
1439 dst receives width, height, depth or array size and
1440 number of mipmap levels. The dst can have a writemask
1441 which will specify what info is the caller interested
1443 RESINFO dst, src_mip_level, resource
1445 RESINFO TEMP[0], TEMP[1].x, RES[0]
1446 src_mip_level is an unsigned integer scalar. If it's
1447 out of range then returns 0 for width, height and
1448 depth/array size but the total number of mipmap is
1449 still returned correctly for the given resource.
1450 The returned width, height and depth values are for
1451 the mipmap level selected by the src_mip_level and
1452 are in the number of texels.
1453 For 1d texture array width is in dst.x, array size
1454 is in dst.y and dst.zw are always 0.
1456 .. opcode:: SAMPLE_POS - query the position of a given sample.
1457 dst receives float4 (x, y, 0, 0) indicated where the
1458 sample is located. If the resource is not a multi-sample
1459 resource and not a render target, the result is 0.
1461 .. opcode:: SAMPLE_INFO - dst receives number of samples in x.
1462 If the resource is not a multi-sample resource and
1463 not a render target, the result is 0.
1466 Explanation of symbols used
1467 ------------------------------
1474 :math:`|x|` Absolute value of `x`.
1476 :math:`\lceil x \rceil` Ceiling of `x`.
1478 clamp(x,y,z) Clamp x between y and z.
1479 (x < y) ? y : (x > z) ? z : x
1481 :math:`\lfloor x\rfloor` Floor of `x`.
1483 :math:`\log_2{x}` Logarithm of `x`, base 2.
1485 max(x,y) Maximum of x and y.
1488 min(x,y) Minimum of x and y.
1491 partialx(x) Derivative of x relative to fragment's X.
1493 partialy(x) Derivative of x relative to fragment's Y.
1495 pop() Pop from stack.
1497 :math:`x^y` `x` to the power `y`.
1499 push(x) Push x on stack.
1503 trunc(x) Truncate x, i.e. drop the fraction bits.
1510 discard Discard fragment.
1514 target Label of target instruction.
1525 Declares a register that is will be referenced as an operand in Instruction
1528 File field contains register file that is being declared and is one
1531 UsageMask field specifies which of the register components can be accessed
1532 and is one of TGSI_WRITEMASK.
1534 Interpolate field is only valid for fragment shader INPUT register files.
1535 It specifes the way input is being interpolated by the rasteriser and is one
1536 of TGSI_INTERPOLATE.
1538 If Dimension flag is set to 1, a Declaration Dimension token follows.
1540 If Semantic flag is set to 1, a Declaration Semantic token follows.
1542 CylindricalWrap bitfield is only valid for fragment shader INPUT register
1543 files. It specifies which register components should be subject to cylindrical
1544 wrapping when interpolating by the rasteriser. If TGSI_CYLINDRICAL_WRAP_X
1545 is set to 1, the X component should be interpolated according to cylindrical
1548 If file is TGSI_FILE_RESOURCE, a Declaration Resource token follows.
1551 Declaration Semantic
1552 ^^^^^^^^^^^^^^^^^^^^^^^^
1554 Vertex and fragment shader input and output registers may be labeled
1555 with semantic information consisting of a name and index.
1557 Follows Declaration token if Semantic bit is set.
1559 Since its purpose is to link a shader with other stages of the pipeline,
1560 it is valid to follow only those Declaration tokens that declare a register
1561 either in INPUT or OUTPUT file.
1563 SemanticName field contains the semantic name of the register being declared.
1564 There is no default value.
1566 SemanticIndex is an optional subscript that can be used to distinguish
1567 different register declarations with the same semantic name. The default value
1570 The meanings of the individual semantic names are explained in the following
1573 TGSI_SEMANTIC_POSITION
1574 """"""""""""""""""""""
1576 For vertex shaders, TGSI_SEMANTIC_POSITION indicates the vertex shader
1577 output register which contains the homogeneous vertex position in the clip
1578 space coordinate system. After clipping, the X, Y and Z components of the
1579 vertex will be divided by the W value to get normalized device coordinates.
1581 For fragment shaders, TGSI_SEMANTIC_POSITION is used to indicate that
1582 fragment shader input contains the fragment's window position. The X
1583 component starts at zero and always increases from left to right.
1584 The Y component starts at zero and always increases but Y=0 may either
1585 indicate the top of the window or the bottom depending on the fragment
1586 coordinate origin convention (see TGSI_PROPERTY_FS_COORD_ORIGIN).
1587 The Z coordinate ranges from 0 to 1 to represent depth from the front
1588 to the back of the Z buffer. The W component contains the reciprocol
1589 of the interpolated vertex position W component.
1591 Fragment shaders may also declare an output register with
1592 TGSI_SEMANTIC_POSITION. Only the Z component is writable. This allows
1593 the fragment shader to change the fragment's Z position.
1600 For vertex shader outputs or fragment shader inputs/outputs, this
1601 label indicates that the resister contains an R,G,B,A color.
1603 Several shader inputs/outputs may contain colors so the semantic index
1604 is used to distinguish them. For example, color[0] may be the diffuse
1605 color while color[1] may be the specular color.
1607 This label is needed so that the flat/smooth shading can be applied
1608 to the right interpolants during rasterization.
1612 TGSI_SEMANTIC_BCOLOR
1613 """"""""""""""""""""
1615 Back-facing colors are only used for back-facing polygons, and are only valid
1616 in vertex shader outputs. After rasterization, all polygons are front-facing
1617 and COLOR and BCOLOR end up occupying the same slots in the fragment shader,
1618 so all BCOLORs effectively become regular COLORs in the fragment shader.
1624 Vertex shader inputs and outputs and fragment shader inputs may be
1625 labeled with TGSI_SEMANTIC_FOG to indicate that the register contains
1626 a fog coordinate in the form (F, 0, 0, 1). Typically, the fragment
1627 shader will use the fog coordinate to compute a fog blend factor which
1628 is used to blend the normal fragment color with a constant fog color.
1630 Only the first component matters when writing from the vertex shader;
1631 the driver will ensure that the coordinate is in this format when used
1632 as a fragment shader input.
1638 Vertex shader input and output registers may be labeled with
1639 TGIS_SEMANTIC_PSIZE to indicate that the register contains a point size
1640 in the form (S, 0, 0, 1). The point size controls the width or diameter
1641 of points for rasterization. This label cannot be used in fragment
1644 When using this semantic, be sure to set the appropriate state in the
1645 :ref:`rasterizer` first.
1648 TGSI_SEMANTIC_GENERIC
1649 """""""""""""""""""""
1651 All vertex/fragment shader inputs/outputs not labeled with any other
1652 semantic label can be considered to be generic attributes. Typical
1653 uses of generic inputs/outputs are texcoords and user-defined values.
1656 TGSI_SEMANTIC_NORMAL
1657 """"""""""""""""""""
1659 Indicates that a vertex shader input is a normal vector. This is
1660 typically only used for legacy graphics APIs.
1666 This label applies to fragment shader inputs only and indicates that
1667 the register contains front/back-face information of the form (F, 0,
1668 0, 1). The first component will be positive when the fragment belongs
1669 to a front-facing polygon, and negative when the fragment belongs to a
1670 back-facing polygon.
1673 TGSI_SEMANTIC_EDGEFLAG
1674 """"""""""""""""""""""
1676 For vertex shaders, this sematic label indicates that an input or
1677 output is a boolean edge flag. The register layout is [F, x, x, x]
1678 where F is 0.0 or 1.0 and x = don't care. Normally, the vertex shader
1679 simply copies the edge flag input to the edgeflag output.
1681 Edge flags are used to control which lines or points are actually
1682 drawn when the polygon mode converts triangles/quads/polygons into
1685 TGSI_SEMANTIC_STENCIL
1686 """"""""""""""""""""""
1688 For fragment shaders, this semantic label indicates than an output
1689 is a writable stencil reference value. Only the Y component is writable.
1690 This allows the fragment shader to change the fragments stencilref value.
1693 Declaration Resource
1694 ^^^^^^^^^^^^^^^^^^^^^^^^
1696 Follows Declaration token if file is TGSI_FILE_RESOURCE.
1698 DCL RES[#], resource, type(s)
1700 Declares a shader input resource and assigns it to a RES[#]
1703 resource can be one of BUFFER, 1D, 2D, 3D, CUBE, 1DArray and
1706 type must be 1 or 4 entries (if specifying on a per-component
1707 level) out of UNORM, SNORM, SINT, UINT and FLOAT.
1711 ^^^^^^^^^^^^^^^^^^^^^^^^
1714 Properties are general directives that apply to the whole TGSI program.
1719 Specifies the fragment shader TGSI_SEMANTIC_POSITION coordinate origin.
1720 The default value is UPPER_LEFT.
1722 If UPPER_LEFT, the position will be (0,0) at the upper left corner and
1723 increase downward and rightward.
1724 If LOWER_LEFT, the position will be (0,0) at the lower left corner and
1725 increase upward and rightward.
1727 OpenGL defaults to LOWER_LEFT, and is configurable with the
1728 GL_ARB_fragment_coord_conventions extension.
1730 DirectX 9/10 use UPPER_LEFT.
1732 FS_COORD_PIXEL_CENTER
1733 """""""""""""""""""""
1735 Specifies the fragment shader TGSI_SEMANTIC_POSITION pixel center convention.
1736 The default value is HALF_INTEGER.
1738 If HALF_INTEGER, the fractionary part of the position will be 0.5
1739 If INTEGER, the fractionary part of the position will be 0.0
1741 Note that this does not affect the set of fragments generated by
1742 rasterization, which is instead controlled by gl_rasterization_rules in the
1745 OpenGL defaults to HALF_INTEGER, and is configurable with the
1746 GL_ARB_fragment_coord_conventions extension.
1748 DirectX 9 uses INTEGER.
1749 DirectX 10 uses HALF_INTEGER.
1751 FS_COLOR0_WRITES_ALL_CBUFS
1752 """"""""""""""""""""""""""
1753 Specifies that writes to the fragment shader color 0 are replicated to all
1754 bound cbufs. This facilitates OpenGL's fragColor output vs fragData[0] where
1755 fragData is directed to a single color buffer, but fragColor is broadcast.
1758 """"""""""""""""""""""""""
1759 If this property is set on the program bound to the shader stage before the
1760 fragment shader, user clip planes should have no effect (be disabled) even if
1761 that shader does not write to any clip distance outputs and the rasterizer's
1762 clip_plane_enable is non-zero.
1763 This property is only supported by drivers that also support shader clip
1765 This is useful for APIs that don't have UCPs and where clip distances written
1766 by a shader cannot be disabled.
1769 Texture Sampling and Texture Formats
1770 ------------------------------------
1772 This table shows how texture image components are returned as (x,y,z,w) tuples
1773 by TGSI texture instructions, such as :opcode:`TEX`, :opcode:`TXD`, and
1774 :opcode:`TXP`. For reference, OpenGL and Direct3D conventions are shown as
1777 +--------------------+--------------+--------------------+--------------+
1778 | Texture Components | Gallium | OpenGL | Direct3D 9 |
1779 +====================+==============+====================+==============+
1780 | R | (r, 0, 0, 1) | (r, 0, 0, 1) | (r, 1, 1, 1) |
1781 +--------------------+--------------+--------------------+--------------+
1782 | RG | (r, g, 0, 1) | (r, g, 0, 1) | (r, g, 1, 1) |
1783 +--------------------+--------------+--------------------+--------------+
1784 | RGB | (r, g, b, 1) | (r, g, b, 1) | (r, g, b, 1) |
1785 +--------------------+--------------+--------------------+--------------+
1786 | RGBA | (r, g, b, a) | (r, g, b, a) | (r, g, b, a) |
1787 +--------------------+--------------+--------------------+--------------+
1788 | A | (0, 0, 0, a) | (0, 0, 0, a) | (0, 0, 0, a) |
1789 +--------------------+--------------+--------------------+--------------+
1790 | L | (l, l, l, 1) | (l, l, l, 1) | (l, l, l, 1) |
1791 +--------------------+--------------+--------------------+--------------+
1792 | LA | (l, l, l, a) | (l, l, l, a) | (l, l, l, a) |
1793 +--------------------+--------------+--------------------+--------------+
1794 | I | (i, i, i, i) | (i, i, i, i) | N/A |
1795 +--------------------+--------------+--------------------+--------------+
1796 | UV | XXX TBD | (0, 0, 0, 1) | (u, v, 1, 1) |
1797 | | | [#envmap-bumpmap]_ | |
1798 +--------------------+--------------+--------------------+--------------+
1799 | Z | XXX TBD | (z, z, z, 1) | (0, z, 0, 1) |
1800 | | | [#depth-tex-mode]_ | |
1801 +--------------------+--------------+--------------------+--------------+
1802 | S | (s, s, s, s) | unknown | unknown |
1803 +--------------------+--------------+--------------------+--------------+
1805 .. [#envmap-bumpmap] http://www.opengl.org/registry/specs/ATI/envmap_bumpmap.txt
1806 .. [#depth-tex-mode] the default is (z, z, z, 1) but may also be (0, 0, 0, z)
1807 or (z, z, z, z) depending on the value of GL_DEPTH_TEXTURE_MODE.