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 type only the negate modifier is supported. This
36 includes instructions which are otherwise ignorant if the type is signed or
37 unsigned, such as TGSI_OPCODE_UADD.
39 For inputs with unsigned type no modifiers are allowed.
45 ^^^^^^^^^^^^^^^^^^^^^^^^^
47 These opcodes are guaranteed to be available regardless of the driver being
50 .. opcode:: ARL - Address Register Load
54 dst.x = \lfloor src.x\rfloor
56 dst.y = \lfloor src.y\rfloor
58 dst.z = \lfloor src.z\rfloor
60 dst.w = \lfloor src.w\rfloor
63 .. opcode:: MOV - Move
76 .. opcode:: LIT - Light Coefficients
84 dst.z = (src.x > 0) ? max(src.y, 0)^{clamp(src.w, -128, 128))} : 0
89 .. opcode:: RCP - Reciprocal
91 This instruction replicates its result.
98 .. opcode:: RSQ - Reciprocal Square Root
100 This instruction replicates its result.
104 dst = \frac{1}{\sqrt{|src.x|}}
107 .. opcode:: SQRT - Square Root
109 This instruction replicates its result.
116 .. opcode:: EXP - Approximate Exponential Base 2
120 dst.x = 2^{\lfloor src.x\rfloor}
122 dst.y = src.x - \lfloor src.x\rfloor
129 .. opcode:: LOG - Approximate Logarithm Base 2
133 dst.x = \lfloor\log_2{|src.x|}\rfloor
135 dst.y = \frac{|src.x|}{2^{\lfloor\log_2{|src.x|}\rfloor}}
137 dst.z = \log_2{|src.x|}
142 .. opcode:: MUL - Multiply
146 dst.x = src0.x \times src1.x
148 dst.y = src0.y \times src1.y
150 dst.z = src0.z \times src1.z
152 dst.w = src0.w \times src1.w
155 .. opcode:: ADD - Add
159 dst.x = src0.x + src1.x
161 dst.y = src0.y + src1.y
163 dst.z = src0.z + src1.z
165 dst.w = src0.w + src1.w
168 .. opcode:: DP3 - 3-component Dot Product
170 This instruction replicates its result.
174 dst = src0.x \times src1.x + src0.y \times src1.y + src0.z \times src1.z
177 .. opcode:: DP4 - 4-component Dot Product
179 This instruction replicates its result.
183 dst = src0.x \times src1.x + src0.y \times src1.y + src0.z \times src1.z + src0.w \times src1.w
186 .. opcode:: DST - Distance Vector
192 dst.y = src0.y \times src1.y
199 .. opcode:: MIN - Minimum
203 dst.x = min(src0.x, src1.x)
205 dst.y = min(src0.y, src1.y)
207 dst.z = min(src0.z, src1.z)
209 dst.w = min(src0.w, src1.w)
212 .. opcode:: MAX - Maximum
216 dst.x = max(src0.x, src1.x)
218 dst.y = max(src0.y, src1.y)
220 dst.z = max(src0.z, src1.z)
222 dst.w = max(src0.w, src1.w)
225 .. opcode:: SLT - Set On Less Than
229 dst.x = (src0.x < src1.x) ? 1 : 0
231 dst.y = (src0.y < src1.y) ? 1 : 0
233 dst.z = (src0.z < src1.z) ? 1 : 0
235 dst.w = (src0.w < src1.w) ? 1 : 0
238 .. opcode:: SGE - Set On Greater Equal Than
242 dst.x = (src0.x >= src1.x) ? 1 : 0
244 dst.y = (src0.y >= src1.y) ? 1 : 0
246 dst.z = (src0.z >= src1.z) ? 1 : 0
248 dst.w = (src0.w >= src1.w) ? 1 : 0
251 .. opcode:: MAD - Multiply And Add
255 dst.x = src0.x \times src1.x + src2.x
257 dst.y = src0.y \times src1.y + src2.y
259 dst.z = src0.z \times src1.z + src2.z
261 dst.w = src0.w \times src1.w + src2.w
264 .. opcode:: SUB - Subtract
268 dst.x = src0.x - src1.x
270 dst.y = src0.y - src1.y
272 dst.z = src0.z - src1.z
274 dst.w = src0.w - src1.w
277 .. opcode:: LRP - Linear Interpolate
281 dst.x = src0.x \times src1.x + (1 - src0.x) \times src2.x
283 dst.y = src0.y \times src1.y + (1 - src0.y) \times src2.y
285 dst.z = src0.z \times src1.z + (1 - src0.z) \times src2.z
287 dst.w = src0.w \times src1.w + (1 - src0.w) \times src2.w
290 .. opcode:: CND - Condition
294 dst.x = (src2.x > 0.5) ? src0.x : src1.x
296 dst.y = (src2.y > 0.5) ? src0.y : src1.y
298 dst.z = (src2.z > 0.5) ? src0.z : src1.z
300 dst.w = (src2.w > 0.5) ? src0.w : src1.w
303 .. opcode:: DP2A - 2-component Dot Product And Add
307 dst.x = src0.x \times src1.x + src0.y \times src1.y + src2.x
309 dst.y = src0.x \times src1.x + src0.y \times src1.y + src2.x
311 dst.z = src0.x \times src1.x + src0.y \times src1.y + src2.x
313 dst.w = src0.x \times src1.x + src0.y \times src1.y + src2.x
316 .. opcode:: FRC - Fraction
320 dst.x = src.x - \lfloor src.x\rfloor
322 dst.y = src.y - \lfloor src.y\rfloor
324 dst.z = src.z - \lfloor src.z\rfloor
326 dst.w = src.w - \lfloor src.w\rfloor
329 .. opcode:: CLAMP - Clamp
333 dst.x = clamp(src0.x, src1.x, src2.x)
335 dst.y = clamp(src0.y, src1.y, src2.y)
337 dst.z = clamp(src0.z, src1.z, src2.z)
339 dst.w = clamp(src0.w, src1.w, src2.w)
342 .. opcode:: FLR - Floor
344 This is identical to :opcode:`ARL`.
348 dst.x = \lfloor src.x\rfloor
350 dst.y = \lfloor src.y\rfloor
352 dst.z = \lfloor src.z\rfloor
354 dst.w = \lfloor src.w\rfloor
357 .. opcode:: ROUND - Round
370 .. opcode:: EX2 - Exponential Base 2
372 This instruction replicates its result.
379 .. opcode:: LG2 - Logarithm Base 2
381 This instruction replicates its result.
388 .. opcode:: POW - Power
390 This instruction replicates its result.
394 dst = src0.x^{src1.x}
396 .. opcode:: XPD - Cross Product
400 dst.x = src0.y \times src1.z - src1.y \times src0.z
402 dst.y = src0.z \times src1.x - src1.z \times src0.x
404 dst.z = src0.x \times src1.y - src1.x \times src0.y
409 .. opcode:: ABS - Absolute
422 .. opcode:: RCC - Reciprocal Clamped
424 This instruction replicates its result.
426 XXX cleanup on aisle three
430 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)
433 .. opcode:: DPH - Homogeneous Dot Product
435 This instruction replicates its result.
439 dst = src0.x \times src1.x + src0.y \times src1.y + src0.z \times src1.z + src1.w
442 .. opcode:: COS - Cosine
444 This instruction replicates its result.
451 .. opcode:: DDX - Derivative Relative To X
455 dst.x = partialx(src.x)
457 dst.y = partialx(src.y)
459 dst.z = partialx(src.z)
461 dst.w = partialx(src.w)
464 .. opcode:: DDY - Derivative Relative To Y
468 dst.x = partialy(src.x)
470 dst.y = partialy(src.y)
472 dst.z = partialy(src.z)
474 dst.w = partialy(src.w)
477 .. opcode:: KILP - Predicated Discard
482 .. opcode:: PK2H - Pack Two 16-bit Floats
487 .. opcode:: PK2US - Pack Two Unsigned 16-bit Scalars
492 .. opcode:: PK4B - Pack Four Signed 8-bit Scalars
497 .. opcode:: PK4UB - Pack Four Unsigned 8-bit Scalars
502 .. opcode:: RFL - Reflection Vector
506 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
508 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
510 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
516 Considered for removal.
519 .. opcode:: SEQ - Set On Equal
523 dst.x = (src0.x == src1.x) ? 1 : 0
525 dst.y = (src0.y == src1.y) ? 1 : 0
527 dst.z = (src0.z == src1.z) ? 1 : 0
529 dst.w = (src0.w == src1.w) ? 1 : 0
532 .. opcode:: SFL - Set On False
534 This instruction replicates its result.
542 Considered for removal.
545 .. opcode:: SGT - Set On Greater Than
549 dst.x = (src0.x > src1.x) ? 1 : 0
551 dst.y = (src0.y > src1.y) ? 1 : 0
553 dst.z = (src0.z > src1.z) ? 1 : 0
555 dst.w = (src0.w > src1.w) ? 1 : 0
558 .. opcode:: SIN - Sine
560 This instruction replicates its result.
567 .. opcode:: SLE - Set On Less Equal Than
571 dst.x = (src0.x <= src1.x) ? 1 : 0
573 dst.y = (src0.y <= src1.y) ? 1 : 0
575 dst.z = (src0.z <= src1.z) ? 1 : 0
577 dst.w = (src0.w <= src1.w) ? 1 : 0
580 .. opcode:: SNE - Set On Not Equal
584 dst.x = (src0.x != src1.x) ? 1 : 0
586 dst.y = (src0.y != src1.y) ? 1 : 0
588 dst.z = (src0.z != src1.z) ? 1 : 0
590 dst.w = (src0.w != src1.w) ? 1 : 0
593 .. opcode:: STR - Set On True
595 This instruction replicates its result.
602 .. opcode:: TEX - Texture Lookup
610 dst = texture_sample(unit, coord, bias)
612 for array textures src0.y contains the slice for 1D,
613 and src0.z contain the slice for 2D.
614 for shadow textures with no arrays, src0.z contains
616 for shadow textures with arrays, src0.z contains
617 the reference value for 1D arrays, and src0.w contains
618 the reference value for 2D arrays.
619 There is no way to pass a bias in the .w value for
620 shadow arrays, and GLSL doesn't allow this.
621 GLSL does allow cube shadows maps to take a bias value,
622 and we have to determine how this will look in TGSI.
624 .. opcode:: TXD - Texture Lookup with Derivatives
636 dst = texture_sample_deriv(unit, coord, bias, ddx, ddy)
639 .. opcode:: TXP - Projective Texture Lookup
643 coord.x = src0.x / src.w
645 coord.y = src0.y / src.w
647 coord.z = src0.z / src.w
653 dst = texture_sample(unit, coord, bias)
656 .. opcode:: UP2H - Unpack Two 16-Bit Floats
662 Considered for removal.
664 .. opcode:: UP2US - Unpack Two Unsigned 16-Bit Scalars
670 Considered for removal.
672 .. opcode:: UP4B - Unpack Four Signed 8-Bit Values
678 Considered for removal.
680 .. opcode:: UP4UB - Unpack Four Unsigned 8-Bit Scalars
686 Considered for removal.
688 .. opcode:: X2D - 2D Coordinate Transformation
692 dst.x = src0.x + src1.x \times src2.x + src1.y \times src2.y
694 dst.y = src0.y + src1.x \times src2.z + src1.y \times src2.w
696 dst.z = src0.x + src1.x \times src2.x + src1.y \times src2.y
698 dst.w = src0.y + src1.x \times src2.z + src1.y \times src2.w
702 Considered for removal.
705 .. opcode:: ARA - Address Register Add
711 Considered for removal.
713 .. opcode:: ARR - Address Register Load With Round
726 .. opcode:: BRA - Branch
732 Considered for removal.
734 .. opcode:: CAL - Subroutine Call
740 .. opcode:: RET - Subroutine Call Return
745 .. opcode:: SSG - Set Sign
749 dst.x = (src.x > 0) ? 1 : (src.x < 0) ? -1 : 0
751 dst.y = (src.y > 0) ? 1 : (src.y < 0) ? -1 : 0
753 dst.z = (src.z > 0) ? 1 : (src.z < 0) ? -1 : 0
755 dst.w = (src.w > 0) ? 1 : (src.w < 0) ? -1 : 0
758 .. opcode:: CMP - Compare
762 dst.x = (src0.x < 0) ? src1.x : src2.x
764 dst.y = (src0.y < 0) ? src1.y : src2.y
766 dst.z = (src0.z < 0) ? src1.z : src2.z
768 dst.w = (src0.w < 0) ? src1.w : src2.w
771 .. opcode:: KIL - Conditional Discard
775 if (src.x < 0 || src.y < 0 || src.z < 0 || src.w < 0)
780 .. opcode:: SCS - Sine Cosine
793 .. opcode:: TXB - Texture Lookup With Bias
807 dst = texture_sample(unit, coord, bias)
810 .. opcode:: NRM - 3-component Vector Normalise
814 dst.x = src.x / (src.x \times src.x + src.y \times src.y + src.z \times src.z)
816 dst.y = src.y / (src.x \times src.x + src.y \times src.y + src.z \times src.z)
818 dst.z = src.z / (src.x \times src.x + src.y \times src.y + src.z \times src.z)
823 .. opcode:: DIV - Divide
827 dst.x = \frac{src0.x}{src1.x}
829 dst.y = \frac{src0.y}{src1.y}
831 dst.z = \frac{src0.z}{src1.z}
833 dst.w = \frac{src0.w}{src1.w}
836 .. opcode:: DP2 - 2-component Dot Product
838 This instruction replicates its result.
842 dst = src0.x \times src1.x + src0.y \times src1.y
845 .. opcode:: TXL - Texture Lookup With explicit LOD
859 dst = texture_sample(unit, coord, lod)
862 .. opcode:: BRK - Break
872 .. opcode:: ELSE - Else
877 .. opcode:: ENDIF - End If
882 .. opcode:: PUSHA - Push Address Register On Stack
891 Considered for cleanup.
895 Considered for removal.
897 .. opcode:: POPA - Pop Address Register From Stack
906 Considered for cleanup.
910 Considered for removal.
914 ^^^^^^^^^^^^^^^^^^^^^^^^
916 These opcodes are primarily provided for special-use computational shaders.
917 Support for these opcodes indicated by a special pipe capability bit (TBD).
919 XXX so let's discuss it, yeah?
921 .. opcode:: CEIL - Ceiling
925 dst.x = \lceil src.x\rceil
927 dst.y = \lceil src.y\rceil
929 dst.z = \lceil src.z\rceil
931 dst.w = \lceil src.w\rceil
934 .. opcode:: I2F - Integer To Float
938 dst.x = (float) src.x
940 dst.y = (float) src.y
942 dst.z = (float) src.z
944 dst.w = (float) src.w
947 .. opcode:: NOT - Bitwise Not
960 .. opcode:: TRUNC - Truncate
973 .. opcode:: SHL - Shift Left
977 dst.x = src0.x << src1.x
979 dst.y = src0.y << src1.x
981 dst.z = src0.z << src1.x
983 dst.w = src0.w << src1.x
986 .. opcode:: SHR - Shift Right
990 dst.x = src0.x >> src1.x
992 dst.y = src0.y >> src1.x
994 dst.z = src0.z >> src1.x
996 dst.w = src0.w >> src1.x
999 .. opcode:: AND - Bitwise And
1003 dst.x = src0.x & src1.x
1005 dst.y = src0.y & src1.y
1007 dst.z = src0.z & src1.z
1009 dst.w = src0.w & src1.w
1012 .. opcode:: OR - Bitwise Or
1016 dst.x = src0.x | src1.x
1018 dst.y = src0.y | src1.y
1020 dst.z = src0.z | src1.z
1022 dst.w = src0.w | src1.w
1025 .. opcode:: MOD - Modulus
1029 dst.x = src0.x \bmod src1.x
1031 dst.y = src0.y \bmod src1.y
1033 dst.z = src0.z \bmod src1.z
1035 dst.w = src0.w \bmod src1.w
1038 .. opcode:: XOR - Bitwise Xor
1042 dst.x = src0.x \oplus src1.x
1044 dst.y = src0.y \oplus src1.y
1046 dst.z = src0.z \oplus src1.z
1048 dst.w = src0.w \oplus src1.w
1051 .. opcode:: UCMP - Integer Conditional Move
1055 dst.x = src0.x ? src1.x : src2.x
1057 dst.y = src0.y ? src1.y : src2.y
1059 dst.z = src0.z ? src1.z : src2.z
1061 dst.w = src0.w ? src1.w : src2.w
1064 .. opcode:: UARL - Integer Address Register Load
1066 Moves the contents of the source register, assumed to be an integer, into the
1067 destination register, which is assumed to be an address (ADDR) register.
1070 .. opcode:: IABS - Integer Absolute Value
1083 .. opcode:: SAD - Sum Of Absolute Differences
1087 dst.x = |src0.x - src1.x| + src2.x
1089 dst.y = |src0.y - src1.y| + src2.y
1091 dst.z = |src0.z - src1.z| + src2.z
1093 dst.w = |src0.w - src1.w| + src2.w
1096 .. opcode:: TXF - Texel Fetch (as per NV_gpu_shader4), extract a single texel
1097 from a specified texture image. The source sampler may
1098 not be a CUBE or SHADOW.
1099 src 0 is a four-component signed integer vector used to
1100 identify the single texel accessed. 3 components + level.
1101 src 1 is a 3 component constant signed integer vector,
1102 with each component only have a range of
1103 -8..+8 (hw only seems to deal with this range, interface
1104 allows for up to unsigned int).
1105 TXF(uint_vec coord, int_vec offset).
1108 .. opcode:: TXQ - Texture Size Query (as per NV_gpu_program4)
1109 retrieve the dimensions of the texture
1110 depending on the target. For 1D (width), 2D/RECT/CUBE
1111 (width, height), 3D (width, height, depth),
1112 1D array (width, layers), 2D array (width, height, layers)
1118 dst.x = texture_width(unit, lod)
1120 dst.y = texture_height(unit, lod)
1122 dst.z = texture_depth(unit, lod)
1125 .. opcode:: CONT - Continue
1131 Support for CONT is determined by a special capability bit,
1132 ``TGSI_CONT_SUPPORTED``. See :ref:`Screen` for more information.
1136 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1138 These opcodes are only supported in geometry shaders; they have no meaning
1139 in any other type of shader.
1141 .. opcode:: EMIT - Emit
1146 .. opcode:: ENDPRIM - End Primitive
1154 These opcodes are part of :term:`GLSL`'s opcode set. Support for these
1155 opcodes is determined by a special capability bit, ``GLSL``.
1157 .. opcode:: BGNLOOP - Begin a Loop
1162 .. opcode:: BGNSUB - Begin Subroutine
1167 .. opcode:: ENDLOOP - End a Loop
1172 .. opcode:: ENDSUB - End Subroutine
1177 .. opcode:: NOP - No Operation
1182 .. opcode:: NRM4 - 4-component Vector Normalise
1184 This instruction replicates its result.
1188 dst = \frac{src.x}{src.x \times src.x + src.y \times src.y + src.z \times src.z + src.w \times src.w}
1196 .. opcode:: CALLNZ - Subroutine Call If Not Zero
1201 .. opcode:: IFC - If
1206 .. opcode:: BREAKC - Break Conditional
1215 The double-precision opcodes reinterpret four-component vectors into
1216 two-component vectors with doubled precision in each component.
1218 Support for these opcodes is XXX undecided. :T
1220 .. opcode:: DADD - Add
1224 dst.xy = src0.xy + src1.xy
1226 dst.zw = src0.zw + src1.zw
1229 .. opcode:: DDIV - Divide
1233 dst.xy = src0.xy / src1.xy
1235 dst.zw = src0.zw / src1.zw
1237 .. opcode:: DSEQ - Set on Equal
1241 dst.xy = src0.xy == src1.xy ? 1.0F : 0.0F
1243 dst.zw = src0.zw == src1.zw ? 1.0F : 0.0F
1245 .. opcode:: DSLT - Set on Less than
1249 dst.xy = src0.xy < src1.xy ? 1.0F : 0.0F
1251 dst.zw = src0.zw < src1.zw ? 1.0F : 0.0F
1253 .. opcode:: DFRAC - Fraction
1257 dst.xy = src.xy - \lfloor src.xy\rfloor
1259 dst.zw = src.zw - \lfloor src.zw\rfloor
1262 .. opcode:: DFRACEXP - Convert Number to Fractional and Integral Components
1264 Like the ``frexp()`` routine in many math libraries, this opcode stores the
1265 exponent of its source to ``dst0``, and the significand to ``dst1``, such that
1266 :math:`dst1 \times 2^{dst0} = src` .
1270 dst0.xy = exp(src.xy)
1272 dst1.xy = frac(src.xy)
1274 dst0.zw = exp(src.zw)
1276 dst1.zw = frac(src.zw)
1278 .. opcode:: DLDEXP - Multiply Number by Integral Power of 2
1280 This opcode is the inverse of :opcode:`DFRACEXP`.
1284 dst.xy = src0.xy \times 2^{src1.xy}
1286 dst.zw = src0.zw \times 2^{src1.zw}
1288 .. opcode:: DMIN - Minimum
1292 dst.xy = min(src0.xy, src1.xy)
1294 dst.zw = min(src0.zw, src1.zw)
1296 .. opcode:: DMAX - Maximum
1300 dst.xy = max(src0.xy, src1.xy)
1302 dst.zw = max(src0.zw, src1.zw)
1304 .. opcode:: DMUL - Multiply
1308 dst.xy = src0.xy \times src1.xy
1310 dst.zw = src0.zw \times src1.zw
1313 .. opcode:: DMAD - Multiply And Add
1317 dst.xy = src0.xy \times src1.xy + src2.xy
1319 dst.zw = src0.zw \times src1.zw + src2.zw
1322 .. opcode:: DRCP - Reciprocal
1326 dst.xy = \frac{1}{src.xy}
1328 dst.zw = \frac{1}{src.zw}
1330 .. opcode:: DSQRT - Square Root
1334 dst.xy = \sqrt{src.xy}
1336 dst.zw = \sqrt{src.zw}
1339 .. _samplingopcodes:
1341 Resource Sampling Opcodes
1342 ^^^^^^^^^^^^^^^^^^^^^^^^^
1344 Those opcodes follow very closely semantics of the respective Direct3D
1345 instructions. If in doubt double check Direct3D documentation.
1347 .. opcode:: SAMPLE - Using provided address, sample data from the
1348 specified texture using the filtering mode identified
1349 by the gven sampler. The source data may come from
1350 any resource type other than buffers.
1351 SAMPLE dst, address, sampler_view, sampler
1353 SAMPLE TEMP[0], TEMP[1], SVIEW[0], SAMP[0]
1355 .. opcode:: SAMPLE_I - Simplified alternative to the SAMPLE instruction.
1356 Using the provided integer address, SAMPLE_I fetches data
1357 from the specified sampler view without any filtering.
1358 The source data may come from any resource type other
1360 SAMPLE_I dst, address, sampler_view
1362 SAMPLE_I TEMP[0], TEMP[1], SVIEW[0]
1363 The 'address' is specified as unsigned integers. If the
1364 'address' is out of range [0...(# texels - 1)] the
1365 result of the fetch is always 0 in all components.
1366 As such the instruction doesn't honor address wrap
1367 modes, in cases where that behavior is desirable
1368 'SAMPLE' instruction should be used.
1369 address.w always provides an unsigned integer mipmap
1370 level. If the value is out of the range then the
1371 instruction always returns 0 in all components.
1372 address.yz are ignored for buffers and 1d textures.
1373 address.z is ignored for 1d texture arrays and 2d
1375 For 1D texture arrays address.y provides the array
1376 index (also as unsigned integer). If the value is
1377 out of the range of available array indices
1378 [0... (array size - 1)] then the opcode always returns
1379 0 in all components.
1380 For 2D texture arrays address.z provides the array
1381 index, otherwise it exhibits the same behavior as in
1382 the case for 1D texture arrays.
1383 The exact semantics of the source address are presented
1385 resource type X Y Z W
1386 ------------- ------------------------
1387 PIPE_BUFFER x ignored
1388 PIPE_TEXTURE_1D x mpl
1389 PIPE_TEXTURE_2D x y mpl
1390 PIPE_TEXTURE_3D x y z mpl
1391 PIPE_TEXTURE_RECT x y mpl
1392 PIPE_TEXTURE_CUBE not allowed as source
1393 PIPE_TEXTURE_1D_ARRAY x idx mpl
1394 PIPE_TEXTURE_2D_ARRAY x y idx mpl
1396 Where 'mpl' is a mipmap level and 'idx' is the
1399 .. opcode:: SAMPLE_I_MS - Just like SAMPLE_I but allows fetch data from
1400 multi-sampled surfaces.
1401 SAMPLE_I_MS dst, address, sampler_view, sample
1403 .. opcode:: SAMPLE_B - Just like the SAMPLE instruction with the
1404 exception that an additional bias is applied to the
1405 level of detail computed as part of the instruction
1407 SAMPLE_B dst, address, sampler_view, sampler, lod_bias
1409 SAMPLE_B TEMP[0], TEMP[1], SVIEW[0], SAMP[0], TEMP[2].x
1411 .. opcode:: SAMPLE_C - Similar to the SAMPLE instruction but it
1412 performs a comparison filter. The operands to SAMPLE_C
1413 are identical to SAMPLE, except that there is an additional
1414 float32 operand, reference value, which must be a register
1415 with single-component, or a scalar literal.
1416 SAMPLE_C makes the hardware use the current samplers
1417 compare_func (in pipe_sampler_state) to compare
1418 reference value against the red component value for the
1419 surce resource at each texel that the currently configured
1420 texture filter covers based on the provided coordinates.
1421 SAMPLE_C dst, address, sampler_view.r, sampler, ref_value
1423 SAMPLE_C TEMP[0], TEMP[1], SVIEW[0].r, SAMP[0], TEMP[2].x
1425 .. opcode:: SAMPLE_C_LZ - Same as SAMPLE_C, but LOD is 0 and derivatives
1426 are ignored. The LZ stands for level-zero.
1427 SAMPLE_C_LZ dst, address, sampler_view.r, sampler, ref_value
1429 SAMPLE_C_LZ TEMP[0], TEMP[1], SVIEW[0].r, SAMP[0], TEMP[2].x
1432 .. opcode:: SAMPLE_D - SAMPLE_D is identical to the SAMPLE opcode except
1433 that the derivatives for the source address in the x
1434 direction and the y direction are provided by extra
1436 SAMPLE_D dst, address, sampler_view, sampler, der_x, der_y
1438 SAMPLE_D TEMP[0], TEMP[1], SVIEW[0], SAMP[0], TEMP[2], TEMP[3]
1440 .. opcode:: SAMPLE_L - SAMPLE_L is identical to the SAMPLE opcode except
1441 that the LOD is provided directly as a scalar value,
1442 representing no anisotropy.
1443 SAMPLE_L dst, address, sampler_view, sampler, explicit_lod
1445 SAMPLE_L TEMP[0], TEMP[1], SVIEW[0], SAMP[0], TEMP[2].x
1447 .. opcode:: GATHER4 - Gathers the four texels to be used in a bi-linear
1448 filtering operation and packs them into a single register.
1449 Only works with 2D, 2D array, cubemaps, and cubemaps arrays.
1450 For 2D textures, only the addressing modes of the sampler and
1451 the top level of any mip pyramid are used. Set W to zero.
1452 It behaves like the SAMPLE instruction, but a filtered
1453 sample is not generated. The four samples that contribute
1454 to filtering are placed into xyzw in counter-clockwise order,
1455 starting with the (u,v) texture coordinate delta at the
1456 following locations (-, +), (+, +), (+, -), (-, -), where
1457 the magnitude of the deltas are half a texel.
1460 .. opcode:: SVIEWINFO - query the dimensions of a given sampler view.
1461 dst receives width, height, depth or array size and
1462 number of mipmap levels as int4. The dst can have a writemask
1463 which will specify what info is the caller interested
1465 SVIEWINFO dst, src_mip_level, sampler_view
1467 SVIEWINFO TEMP[0], TEMP[1].x, SVIEW[0]
1468 src_mip_level is an unsigned integer scalar. If it's
1469 out of range then returns 0 for width, height and
1470 depth/array size but the total number of mipmap is
1471 still returned correctly for the given sampler view.
1472 The returned width, height and depth values are for
1473 the mipmap level selected by the src_mip_level and
1474 are in the number of texels.
1475 For 1d texture array width is in dst.x, array size
1476 is in dst.y and dst.zw are always 0.
1478 .. opcode:: SAMPLE_POS - query the position of a given sample.
1479 dst receives float4 (x, y, 0, 0) indicated where the
1480 sample is located. If the resource is not a multi-sample
1481 resource and not a render target, the result is 0.
1483 .. opcode:: SAMPLE_INFO - dst receives number of samples in x.
1484 If the resource is not a multi-sample resource and
1485 not a render target, the result is 0.
1488 .. _resourceopcodes:
1490 Resource Access Opcodes
1491 ^^^^^^^^^^^^^^^^^^^^^^^
1493 .. opcode:: LOAD - Fetch data from a shader resource
1495 Syntax: ``LOAD dst, resource, address``
1497 Example: ``LOAD TEMP[0], RES[0], TEMP[1]``
1499 Using the provided integer address, LOAD fetches data
1500 from the specified buffer or texture without any
1503 The 'address' is specified as a vector of unsigned
1504 integers. If the 'address' is out of range the result
1507 Only the first mipmap level of a resource can be read
1508 from using this instruction.
1510 For 1D or 2D texture arrays, the array index is
1511 provided as an unsigned integer in address.y or
1512 address.z, respectively. address.yz are ignored for
1513 buffers and 1D textures. address.z is ignored for 1D
1514 texture arrays and 2D textures. address.w is always
1517 .. opcode:: STORE - Write data to a shader resource
1519 Syntax: ``STORE resource, address, src``
1521 Example: ``STORE RES[0], TEMP[0], TEMP[1]``
1523 Using the provided integer address, STORE writes data
1524 to the specified buffer or texture.
1526 The 'address' is specified as a vector of unsigned
1527 integers. If the 'address' is out of range the result
1530 Only the first mipmap level of a resource can be
1531 written to using this instruction.
1533 For 1D or 2D texture arrays, the array index is
1534 provided as an unsigned integer in address.y or
1535 address.z, respectively. address.yz are ignored for
1536 buffers and 1D textures. address.z is ignored for 1D
1537 texture arrays and 2D textures. address.w is always
1541 .. _threadsyncopcodes:
1543 Inter-thread synchronization opcodes
1544 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1546 These opcodes are intended for communication between threads running
1547 within the same compute grid. For now they're only valid in compute
1550 .. opcode:: MFENCE - Memory fence
1552 Syntax: ``MFENCE resource``
1554 Example: ``MFENCE RES[0]``
1556 This opcode forces strong ordering between any memory access
1557 operations that affect the specified resource. This means that
1558 previous loads and stores (and only those) will be performed and
1559 visible to other threads before the program execution continues.
1562 .. opcode:: LFENCE - Load memory fence
1564 Syntax: ``LFENCE resource``
1566 Example: ``LFENCE RES[0]``
1568 Similar to MFENCE, but it only affects the ordering of memory loads.
1571 .. opcode:: SFENCE - Store memory fence
1573 Syntax: ``SFENCE resource``
1575 Example: ``SFENCE RES[0]``
1577 Similar to MFENCE, but it only affects the ordering of memory stores.
1580 .. opcode:: BARRIER - Thread group barrier
1584 This opcode suspends the execution of the current thread until all
1585 the remaining threads in the working group reach the same point of
1586 the program. Results are unspecified if any of the remaining
1587 threads terminates or never reaches an executed BARRIER instruction.
1595 These opcodes provide atomic variants of some common arithmetic and
1596 logical operations. In this context atomicity means that another
1597 concurrent memory access operation that affects the same memory
1598 location is guaranteed to be performed strictly before or after the
1599 entire execution of the atomic operation.
1601 For the moment they're only valid in compute programs.
1603 .. opcode:: ATOMUADD - Atomic integer addition
1605 Syntax: ``ATOMUADD dst, resource, offset, src``
1607 Example: ``ATOMUADD TEMP[0], RES[0], TEMP[1], TEMP[2]``
1609 The following operation is performed atomically on each component:
1613 dst_i = resource[offset]_i
1615 resource[offset]_i = dst_i + src_i
1618 .. opcode:: ATOMXCHG - Atomic exchange
1620 Syntax: ``ATOMXCHG dst, resource, offset, src``
1622 Example: ``ATOMXCHG TEMP[0], RES[0], TEMP[1], TEMP[2]``
1624 The following operation is performed atomically on each component:
1628 dst_i = resource[offset]_i
1630 resource[offset]_i = src_i
1633 .. opcode:: ATOMCAS - Atomic compare-and-exchange
1635 Syntax: ``ATOMCAS dst, resource, offset, cmp, src``
1637 Example: ``ATOMCAS TEMP[0], RES[0], TEMP[1], TEMP[2], TEMP[3]``
1639 The following operation is performed atomically on each component:
1643 dst_i = resource[offset]_i
1645 resource[offset]_i = (dst_i == cmp_i ? src_i : dst_i)
1648 .. opcode:: ATOMAND - Atomic bitwise And
1650 Syntax: ``ATOMAND dst, resource, offset, src``
1652 Example: ``ATOMAND TEMP[0], RES[0], TEMP[1], TEMP[2]``
1654 The following operation is performed atomically on each component:
1658 dst_i = resource[offset]_i
1660 resource[offset]_i = dst_i \& src_i
1663 .. opcode:: ATOMOR - Atomic bitwise Or
1665 Syntax: ``ATOMOR dst, resource, offset, src``
1667 Example: ``ATOMOR TEMP[0], RES[0], TEMP[1], TEMP[2]``
1669 The following operation is performed atomically on each component:
1673 dst_i = resource[offset]_i
1675 resource[offset]_i = dst_i | src_i
1678 .. opcode:: ATOMXOR - Atomic bitwise Xor
1680 Syntax: ``ATOMXOR dst, resource, offset, src``
1682 Example: ``ATOMXOR TEMP[0], RES[0], TEMP[1], TEMP[2]``
1684 The following operation is performed atomically on each component:
1688 dst_i = resource[offset]_i
1690 resource[offset]_i = dst_i \oplus src_i
1693 .. opcode:: ATOMUMIN - Atomic unsigned minimum
1695 Syntax: ``ATOMUMIN dst, resource, offset, src``
1697 Example: ``ATOMUMIN TEMP[0], RES[0], TEMP[1], TEMP[2]``
1699 The following operation is performed atomically on each component:
1703 dst_i = resource[offset]_i
1705 resource[offset]_i = (dst_i < src_i ? dst_i : src_i)
1708 .. opcode:: ATOMUMAX - Atomic unsigned maximum
1710 Syntax: ``ATOMUMAX dst, resource, offset, src``
1712 Example: ``ATOMUMAX TEMP[0], RES[0], TEMP[1], TEMP[2]``
1714 The following operation is performed atomically on each component:
1718 dst_i = resource[offset]_i
1720 resource[offset]_i = (dst_i > src_i ? dst_i : src_i)
1723 .. opcode:: ATOMIMIN - Atomic signed minimum
1725 Syntax: ``ATOMIMIN dst, resource, offset, src``
1727 Example: ``ATOMIMIN TEMP[0], RES[0], TEMP[1], TEMP[2]``
1729 The following operation is performed atomically on each component:
1733 dst_i = resource[offset]_i
1735 resource[offset]_i = (dst_i < src_i ? dst_i : src_i)
1738 .. opcode:: ATOMIMAX - Atomic signed maximum
1740 Syntax: ``ATOMIMAX dst, resource, offset, src``
1742 Example: ``ATOMIMAX TEMP[0], RES[0], TEMP[1], TEMP[2]``
1744 The following operation is performed atomically on each component:
1748 dst_i = resource[offset]_i
1750 resource[offset]_i = (dst_i > src_i ? dst_i : src_i)
1754 Explanation of symbols used
1755 ------------------------------
1762 :math:`|x|` Absolute value of `x`.
1764 :math:`\lceil x \rceil` Ceiling of `x`.
1766 clamp(x,y,z) Clamp x between y and z.
1767 (x < y) ? y : (x > z) ? z : x
1769 :math:`\lfloor x\rfloor` Floor of `x`.
1771 :math:`\log_2{x}` Logarithm of `x`, base 2.
1773 max(x,y) Maximum of x and y.
1776 min(x,y) Minimum of x and y.
1779 partialx(x) Derivative of x relative to fragment's X.
1781 partialy(x) Derivative of x relative to fragment's Y.
1783 pop() Pop from stack.
1785 :math:`x^y` `x` to the power `y`.
1787 push(x) Push x on stack.
1791 trunc(x) Truncate x, i.e. drop the fraction bits.
1798 discard Discard fragment.
1802 target Label of target instruction.
1813 Declares a register that is will be referenced as an operand in Instruction
1816 File field contains register file that is being declared and is one
1819 UsageMask field specifies which of the register components can be accessed
1820 and is one of TGSI_WRITEMASK.
1822 The Local flag specifies that a given value isn't intended for
1823 subroutine parameter passing and, as a result, the implementation
1824 isn't required to give any guarantees of it being preserved across
1825 subroutine boundaries. As it's merely a compiler hint, the
1826 implementation is free to ignore it.
1828 If Dimension flag is set to 1, a Declaration Dimension token follows.
1830 If Semantic flag is set to 1, a Declaration Semantic token follows.
1832 If Interpolate flag is set to 1, a Declaration Interpolate token follows.
1834 If file is TGSI_FILE_RESOURCE, a Declaration Resource token follows.
1836 If Array flag is set to 1, a Declaration Array token follows.
1839 ^^^^^^^^^^^^^^^^^^^^^^^^
1841 Declarations can optional have an ArrayID attribute which can be referred by
1842 indirect addressing operands. An ArrayID of zero is reserved and treaded as
1843 if no ArrayID is specified.
1845 If an indirect addressing operand refers to a specific declaration by using
1846 an ArrayID only the registers in this declaration are guaranteed to be
1847 accessed, accessing any register outside this declaration results in undefined
1848 behavior. Note that for compatibility the effective index is zero-based and
1849 not relative to the specified declaration
1851 If no ArrayID is specified with an indirect addressing operand the whole
1852 register file might be accessed by this operand. This is strongly discouraged
1853 and will prevent packing of scalar/vec2 arrays and effective alias analysis.
1855 Declaration Semantic
1856 ^^^^^^^^^^^^^^^^^^^^^^^^
1858 Vertex and fragment shader input and output registers may be labeled
1859 with semantic information consisting of a name and index.
1861 Follows Declaration token if Semantic bit is set.
1863 Since its purpose is to link a shader with other stages of the pipeline,
1864 it is valid to follow only those Declaration tokens that declare a register
1865 either in INPUT or OUTPUT file.
1867 SemanticName field contains the semantic name of the register being declared.
1868 There is no default value.
1870 SemanticIndex is an optional subscript that can be used to distinguish
1871 different register declarations with the same semantic name. The default value
1874 The meanings of the individual semantic names are explained in the following
1877 TGSI_SEMANTIC_POSITION
1878 """"""""""""""""""""""
1880 For vertex shaders, TGSI_SEMANTIC_POSITION indicates the vertex shader
1881 output register which contains the homogeneous vertex position in the clip
1882 space coordinate system. After clipping, the X, Y and Z components of the
1883 vertex will be divided by the W value to get normalized device coordinates.
1885 For fragment shaders, TGSI_SEMANTIC_POSITION is used to indicate that
1886 fragment shader input contains the fragment's window position. The X
1887 component starts at zero and always increases from left to right.
1888 The Y component starts at zero and always increases but Y=0 may either
1889 indicate the top of the window or the bottom depending on the fragment
1890 coordinate origin convention (see TGSI_PROPERTY_FS_COORD_ORIGIN).
1891 The Z coordinate ranges from 0 to 1 to represent depth from the front
1892 to the back of the Z buffer. The W component contains the reciprocol
1893 of the interpolated vertex position W component.
1895 Fragment shaders may also declare an output register with
1896 TGSI_SEMANTIC_POSITION. Only the Z component is writable. This allows
1897 the fragment shader to change the fragment's Z position.
1904 For vertex shader outputs or fragment shader inputs/outputs, this
1905 label indicates that the resister contains an R,G,B,A color.
1907 Several shader inputs/outputs may contain colors so the semantic index
1908 is used to distinguish them. For example, color[0] may be the diffuse
1909 color while color[1] may be the specular color.
1911 This label is needed so that the flat/smooth shading can be applied
1912 to the right interpolants during rasterization.
1916 TGSI_SEMANTIC_BCOLOR
1917 """"""""""""""""""""
1919 Back-facing colors are only used for back-facing polygons, and are only valid
1920 in vertex shader outputs. After rasterization, all polygons are front-facing
1921 and COLOR and BCOLOR end up occupying the same slots in the fragment shader,
1922 so all BCOLORs effectively become regular COLORs in the fragment shader.
1928 Vertex shader inputs and outputs and fragment shader inputs may be
1929 labeled with TGSI_SEMANTIC_FOG to indicate that the register contains
1930 a fog coordinate in the form (F, 0, 0, 1). Typically, the fragment
1931 shader will use the fog coordinate to compute a fog blend factor which
1932 is used to blend the normal fragment color with a constant fog color.
1934 Only the first component matters when writing from the vertex shader;
1935 the driver will ensure that the coordinate is in this format when used
1936 as a fragment shader input.
1942 Vertex shader input and output registers may be labeled with
1943 TGIS_SEMANTIC_PSIZE to indicate that the register contains a point size
1944 in the form (S, 0, 0, 1). The point size controls the width or diameter
1945 of points for rasterization. This label cannot be used in fragment
1948 When using this semantic, be sure to set the appropriate state in the
1949 :ref:`rasterizer` first.
1952 TGSI_SEMANTIC_GENERIC
1953 """""""""""""""""""""
1955 All vertex/fragment shader inputs/outputs not labeled with any other
1956 semantic label can be considered to be generic attributes. Typical
1957 uses of generic inputs/outputs are texcoords and user-defined values.
1960 TGSI_SEMANTIC_NORMAL
1961 """"""""""""""""""""
1963 Indicates that a vertex shader input is a normal vector. This is
1964 typically only used for legacy graphics APIs.
1970 This label applies to fragment shader inputs only and indicates that
1971 the register contains front/back-face information of the form (F, 0,
1972 0, 1). The first component will be positive when the fragment belongs
1973 to a front-facing polygon, and negative when the fragment belongs to a
1974 back-facing polygon.
1977 TGSI_SEMANTIC_EDGEFLAG
1978 """"""""""""""""""""""
1980 For vertex shaders, this sematic label indicates that an input or
1981 output is a boolean edge flag. The register layout is [F, x, x, x]
1982 where F is 0.0 or 1.0 and x = don't care. Normally, the vertex shader
1983 simply copies the edge flag input to the edgeflag output.
1985 Edge flags are used to control which lines or points are actually
1986 drawn when the polygon mode converts triangles/quads/polygons into
1989 TGSI_SEMANTIC_STENCIL
1990 """"""""""""""""""""""
1992 For fragment shaders, this semantic label indicates than an output
1993 is a writable stencil reference value. Only the Y component is writable.
1994 This allows the fragment shader to change the fragments stencilref value.
1997 Declaration Interpolate
1998 ^^^^^^^^^^^^^^^^^^^^^^^
2000 This token is only valid for fragment shader INPUT declarations.
2002 The Interpolate field specifes the way input is being interpolated by
2003 the rasteriser and is one of TGSI_INTERPOLATE_*.
2005 The CylindricalWrap bitfield specifies which register components
2006 should be subject to cylindrical wrapping when interpolating by the
2007 rasteriser. If TGSI_CYLINDRICAL_WRAP_X is set to 1, the X component
2008 should be interpolated according to cylindrical wrapping rules.
2011 Declaration Sampler View
2012 ^^^^^^^^^^^^^^^^^^^^^^^^
2014 Follows Declaration token if file is TGSI_FILE_SAMPLER_VIEW.
2016 DCL SVIEW[#], resource, type(s)
2018 Declares a shader input sampler view and assigns it to a SVIEW[#]
2021 resource can be one of BUFFER, 1D, 2D, 3D, 1DArray and 2DArray.
2023 type must be 1 or 4 entries (if specifying on a per-component
2024 level) out of UNORM, SNORM, SINT, UINT and FLOAT.
2027 Declaration Resource
2028 ^^^^^^^^^^^^^^^^^^^^
2030 Follows Declaration token if file is TGSI_FILE_RESOURCE.
2032 DCL RES[#], resource [, WR] [, RAW]
2034 Declares a shader input resource and assigns it to a RES[#]
2037 resource can be one of BUFFER, 1D, 2D, 3D, CUBE, 1DArray and
2040 If the RAW keyword is not specified, the texture data will be
2041 subject to conversion, swizzling and scaling as required to yield
2042 the specified data type from the physical data format of the bound
2045 If the RAW keyword is specified, no channel conversion will be
2046 performed: the values read for each of the channels (X,Y,Z,W) will
2047 correspond to consecutive words in the same order and format
2048 they're found in memory. No element-to-address conversion will be
2049 performed either: the value of the provided X coordinate will be
2050 interpreted in byte units instead of texel units. The result of
2051 accessing a misaligned address is undefined.
2053 Usage of the STORE opcode is only allowed if the WR (writable) flag
2058 ^^^^^^^^^^^^^^^^^^^^^^^^
2061 Properties are general directives that apply to the whole TGSI program.
2066 Specifies the fragment shader TGSI_SEMANTIC_POSITION coordinate origin.
2067 The default value is UPPER_LEFT.
2069 If UPPER_LEFT, the position will be (0,0) at the upper left corner and
2070 increase downward and rightward.
2071 If LOWER_LEFT, the position will be (0,0) at the lower left corner and
2072 increase upward and rightward.
2074 OpenGL defaults to LOWER_LEFT, and is configurable with the
2075 GL_ARB_fragment_coord_conventions extension.
2077 DirectX 9/10 use UPPER_LEFT.
2079 FS_COORD_PIXEL_CENTER
2080 """""""""""""""""""""
2082 Specifies the fragment shader TGSI_SEMANTIC_POSITION pixel center convention.
2083 The default value is HALF_INTEGER.
2085 If HALF_INTEGER, the fractionary part of the position will be 0.5
2086 If INTEGER, the fractionary part of the position will be 0.0
2088 Note that this does not affect the set of fragments generated by
2089 rasterization, which is instead controlled by gl_rasterization_rules in the
2092 OpenGL defaults to HALF_INTEGER, and is configurable with the
2093 GL_ARB_fragment_coord_conventions extension.
2095 DirectX 9 uses INTEGER.
2096 DirectX 10 uses HALF_INTEGER.
2098 FS_COLOR0_WRITES_ALL_CBUFS
2099 """"""""""""""""""""""""""
2100 Specifies that writes to the fragment shader color 0 are replicated to all
2101 bound cbufs. This facilitates OpenGL's fragColor output vs fragData[0] where
2102 fragData is directed to a single color buffer, but fragColor is broadcast.
2105 """"""""""""""""""""""""""
2106 If this property is set on the program bound to the shader stage before the
2107 fragment shader, user clip planes should have no effect (be disabled) even if
2108 that shader does not write to any clip distance outputs and the rasterizer's
2109 clip_plane_enable is non-zero.
2110 This property is only supported by drivers that also support shader clip
2112 This is useful for APIs that don't have UCPs and where clip distances written
2113 by a shader cannot be disabled.
2116 Texture Sampling and Texture Formats
2117 ------------------------------------
2119 This table shows how texture image components are returned as (x,y,z,w) tuples
2120 by TGSI texture instructions, such as :opcode:`TEX`, :opcode:`TXD`, and
2121 :opcode:`TXP`. For reference, OpenGL and Direct3D conventions are shown as
2124 +--------------------+--------------+--------------------+--------------+
2125 | Texture Components | Gallium | OpenGL | Direct3D 9 |
2126 +====================+==============+====================+==============+
2127 | R | (r, 0, 0, 1) | (r, 0, 0, 1) | (r, 1, 1, 1) |
2128 +--------------------+--------------+--------------------+--------------+
2129 | RG | (r, g, 0, 1) | (r, g, 0, 1) | (r, g, 1, 1) |
2130 +--------------------+--------------+--------------------+--------------+
2131 | RGB | (r, g, b, 1) | (r, g, b, 1) | (r, g, b, 1) |
2132 +--------------------+--------------+--------------------+--------------+
2133 | RGBA | (r, g, b, a) | (r, g, b, a) | (r, g, b, a) |
2134 +--------------------+--------------+--------------------+--------------+
2135 | A | (0, 0, 0, a) | (0, 0, 0, a) | (0, 0, 0, a) |
2136 +--------------------+--------------+--------------------+--------------+
2137 | L | (l, l, l, 1) | (l, l, l, 1) | (l, l, l, 1) |
2138 +--------------------+--------------+--------------------+--------------+
2139 | LA | (l, l, l, a) | (l, l, l, a) | (l, l, l, a) |
2140 +--------------------+--------------+--------------------+--------------+
2141 | I | (i, i, i, i) | (i, i, i, i) | N/A |
2142 +--------------------+--------------+--------------------+--------------+
2143 | UV | XXX TBD | (0, 0, 0, 1) | (u, v, 1, 1) |
2144 | | | [#envmap-bumpmap]_ | |
2145 +--------------------+--------------+--------------------+--------------+
2146 | Z | XXX TBD | (z, z, z, 1) | (0, z, 0, 1) |
2147 | | | [#depth-tex-mode]_ | |
2148 +--------------------+--------------+--------------------+--------------+
2149 | S | (s, s, s, s) | unknown | unknown |
2150 +--------------------+--------------+--------------------+--------------+
2152 .. [#envmap-bumpmap] http://www.opengl.org/registry/specs/ATI/envmap_bumpmap.txt
2153 .. [#depth-tex-mode] the default is (z, z, z, 1) but may also be (0, 0, 0, z)
2154 or (z, z, z, z) depending on the value of GL_DEPTH_TEXTURE_MODE.