4 TGSI, Tungsten Graphics Shader Infrastructure, is an intermediate language
5 for describing shaders. Since Gallium is inherently shaderful, shaders are
6 an important part of the API. TGSI is the only intermediate representation
12 All TGSI instructions, known as *opcodes*, operate on arbitrary-precision
13 floating-point four-component vectors. An opcode may have up to one
14 destination register, known as *dst*, and between zero and three source
15 registers, called *src0* through *src2*, or simply *src* if there is only
18 Some instructions, like :opcode:`I2F`, permit re-interpretation of vector
19 components as integers. Other instructions permit using registers as
20 two-component vectors with double precision; see :ref:`doubleopcodes`.
22 When an instruction has a scalar result, the result is usually copied into
23 each of the components of *dst*. When this happens, the result is said to be
24 *replicated* to *dst*. :opcode:`RCP` is one such instruction.
29 TGSI supports modifiers on inputs (as well as saturate modifier on instructions).
31 For inputs which have a floating point type, both absolute value and negation
32 modifiers are supported (with absolute value being applied first).
33 TGSI_OPCODE_MOV is considered to have float input type for applying modifiers.
35 For inputs which have signed or unsigned type only the negate modifier is
42 ^^^^^^^^^^^^^^^^^^^^^^^^^
44 These opcodes are guaranteed to be available regardless of the driver being
47 .. opcode:: ARL - Address Register Load
51 dst.x = \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
78 dst.y &= max(src.x, 0) \\
79 dst.z &= (src.x > 0) ? max(src.y, 0)^{clamp(src.w, -128, 128))} : 0 \\
83 .. opcode:: RCP - Reciprocal
85 This instruction replicates its result.
92 .. opcode:: RSQ - Reciprocal Square Root
94 This instruction replicates its result. The results are undefined for src <= 0.
98 dst = \frac{1}{\sqrt{src.x}}
101 .. opcode:: SQRT - Square Root
103 This instruction replicates its result. The results are undefined for src < 0.
110 .. opcode:: EXP - Approximate Exponential Base 2
114 dst.x &= 2^{\lfloor src.x\rfloor} \\
115 dst.y &= src.x - \lfloor src.x\rfloor \\
116 dst.z &= 2^{src.x} \\
120 .. opcode:: LOG - Approximate Logarithm Base 2
124 dst.x &= \lfloor\log_2{|src.x|}\rfloor \\
125 dst.y &= \frac{|src.x|}{2^{\lfloor\log_2{|src.x|}\rfloor}} \\
126 dst.z &= \log_2{|src.x|} \\
130 .. opcode:: MUL - Multiply
134 dst.x = src0.x \times src1.x
136 dst.y = src0.y \times src1.y
138 dst.z = src0.z \times src1.z
140 dst.w = src0.w \times src1.w
143 .. opcode:: ADD - Add
147 dst.x = src0.x + src1.x
149 dst.y = src0.y + src1.y
151 dst.z = src0.z + src1.z
153 dst.w = src0.w + src1.w
156 .. opcode:: DP3 - 3-component Dot Product
158 This instruction replicates its result.
162 dst = src0.x \times src1.x + src0.y \times src1.y + src0.z \times src1.z
165 .. opcode:: DP4 - 4-component Dot Product
167 This instruction replicates its result.
171 dst = src0.x \times src1.x + src0.y \times src1.y + src0.z \times src1.z + src0.w \times src1.w
174 .. opcode:: DST - Distance Vector
179 dst.y &= src0.y \times src1.y\\
184 .. opcode:: MIN - Minimum
188 dst.x = min(src0.x, src1.x)
190 dst.y = min(src0.y, src1.y)
192 dst.z = min(src0.z, src1.z)
194 dst.w = min(src0.w, src1.w)
197 .. opcode:: MAX - Maximum
201 dst.x = max(src0.x, src1.x)
203 dst.y = max(src0.y, src1.y)
205 dst.z = max(src0.z, src1.z)
207 dst.w = max(src0.w, src1.w)
210 .. opcode:: SLT - Set On Less Than
214 dst.x = (src0.x < src1.x) ? 1.0F : 0.0F
216 dst.y = (src0.y < src1.y) ? 1.0F : 0.0F
218 dst.z = (src0.z < src1.z) ? 1.0F : 0.0F
220 dst.w = (src0.w < src1.w) ? 1.0F : 0.0F
223 .. opcode:: SGE - Set On Greater Equal Than
227 dst.x = (src0.x >= src1.x) ? 1.0F : 0.0F
229 dst.y = (src0.y >= src1.y) ? 1.0F : 0.0F
231 dst.z = (src0.z >= src1.z) ? 1.0F : 0.0F
233 dst.w = (src0.w >= src1.w) ? 1.0F : 0.0F
236 .. opcode:: MAD - Multiply And Add
240 dst.x = src0.x \times src1.x + src2.x
242 dst.y = src0.y \times src1.y + src2.y
244 dst.z = src0.z \times src1.z + src2.z
246 dst.w = src0.w \times src1.w + src2.w
249 .. opcode:: SUB - Subtract
253 dst.x = src0.x - src1.x
255 dst.y = src0.y - src1.y
257 dst.z = src0.z - src1.z
259 dst.w = src0.w - src1.w
262 .. opcode:: LRP - Linear Interpolate
266 dst.x = src0.x \times src1.x + (1 - src0.x) \times src2.x
268 dst.y = src0.y \times src1.y + (1 - src0.y) \times src2.y
270 dst.z = src0.z \times src1.z + (1 - src0.z) \times src2.z
272 dst.w = src0.w \times src1.w + (1 - src0.w) \times src2.w
275 .. opcode:: DP2A - 2-component Dot Product And Add
279 dst.x = src0.x \times src1.x + src0.y \times src1.y + src2.x
281 dst.y = src0.x \times src1.x + src0.y \times src1.y + src2.x
283 dst.z = src0.x \times src1.x + src0.y \times src1.y + src2.x
285 dst.w = src0.x \times src1.x + src0.y \times src1.y + src2.x
288 .. opcode:: FRC - Fraction
292 dst.x = src.x - \lfloor src.x\rfloor
294 dst.y = src.y - \lfloor src.y\rfloor
296 dst.z = src.z - \lfloor src.z\rfloor
298 dst.w = src.w - \lfloor src.w\rfloor
301 .. opcode:: CLAMP - Clamp
305 dst.x = clamp(src0.x, src1.x, src2.x)
307 dst.y = clamp(src0.y, src1.y, src2.y)
309 dst.z = clamp(src0.z, src1.z, src2.z)
311 dst.w = clamp(src0.w, src1.w, src2.w)
314 .. opcode:: FLR - Floor
316 This is identical to :opcode:`ARL`.
320 dst.x = \lfloor src.x\rfloor
322 dst.y = \lfloor src.y\rfloor
324 dst.z = \lfloor src.z\rfloor
326 dst.w = \lfloor src.w\rfloor
329 .. opcode:: ROUND - Round
342 .. opcode:: EX2 - Exponential Base 2
344 This instruction replicates its result.
351 .. opcode:: LG2 - Logarithm Base 2
353 This instruction replicates its result.
360 .. opcode:: POW - Power
362 This instruction replicates its result.
366 dst = src0.x^{src1.x}
368 .. opcode:: XPD - Cross Product
372 dst.x = src0.y \times src1.z - src1.y \times src0.z
374 dst.y = src0.z \times src1.x - src1.z \times src0.x
376 dst.z = src0.x \times src1.y - src1.x \times src0.y
381 .. opcode:: ABS - Absolute
394 .. opcode:: DPH - Homogeneous Dot Product
396 This instruction replicates its result.
400 dst = src0.x \times src1.x + src0.y \times src1.y + src0.z \times src1.z + src1.w
403 .. opcode:: COS - Cosine
405 This instruction replicates its result.
412 .. opcode:: DDX, DDX_FINE - Derivative Relative To X
414 The fine variant is only used when ``PIPE_CAP_TGSI_FS_FINE_DERIVATIVE`` is
415 advertised. When it is, the fine version guarantees one derivative per row
416 while DDX is allowed to be the same for the entire 2x2 quad.
420 dst.x = partialx(src.x)
422 dst.y = partialx(src.y)
424 dst.z = partialx(src.z)
426 dst.w = partialx(src.w)
429 .. opcode:: DDY, DDY_FINE - Derivative Relative To Y
431 The fine variant is only used when ``PIPE_CAP_TGSI_FS_FINE_DERIVATIVE`` is
432 advertised. When it is, the fine version guarantees one derivative per column
433 while DDY is allowed to be the same for the entire 2x2 quad.
437 dst.x = partialy(src.x)
439 dst.y = partialy(src.y)
441 dst.z = partialy(src.z)
443 dst.w = partialy(src.w)
446 .. opcode:: PK2H - Pack Two 16-bit Floats
451 .. opcode:: PK2US - Pack Two Unsigned 16-bit Scalars
456 .. opcode:: PK4B - Pack Four Signed 8-bit Scalars
461 .. opcode:: PK4UB - Pack Four Unsigned 8-bit Scalars
466 .. opcode:: SEQ - Set On Equal
470 dst.x = (src0.x == src1.x) ? 1.0F : 0.0F
472 dst.y = (src0.y == src1.y) ? 1.0F : 0.0F
474 dst.z = (src0.z == src1.z) ? 1.0F : 0.0F
476 dst.w = (src0.w == src1.w) ? 1.0F : 0.0F
479 .. opcode:: SGT - Set On Greater Than
483 dst.x = (src0.x > src1.x) ? 1.0F : 0.0F
485 dst.y = (src0.y > src1.y) ? 1.0F : 0.0F
487 dst.z = (src0.z > src1.z) ? 1.0F : 0.0F
489 dst.w = (src0.w > src1.w) ? 1.0F : 0.0F
492 .. opcode:: SIN - Sine
494 This instruction replicates its result.
501 .. opcode:: SLE - Set On Less Equal Than
505 dst.x = (src0.x <= src1.x) ? 1.0F : 0.0F
507 dst.y = (src0.y <= src1.y) ? 1.0F : 0.0F
509 dst.z = (src0.z <= src1.z) ? 1.0F : 0.0F
511 dst.w = (src0.w <= src1.w) ? 1.0F : 0.0F
514 .. opcode:: SNE - Set On Not Equal
518 dst.x = (src0.x != src1.x) ? 1.0F : 0.0F
520 dst.y = (src0.y != src1.y) ? 1.0F : 0.0F
522 dst.z = (src0.z != src1.z) ? 1.0F : 0.0F
524 dst.w = (src0.w != src1.w) ? 1.0F : 0.0F
527 .. opcode:: TEX - Texture Lookup
529 for array textures src0.y contains the slice for 1D,
530 and src0.z contain the slice for 2D.
532 for shadow textures with no arrays (and not cube map),
533 src0.z contains the reference value.
535 for shadow textures with arrays, src0.z contains
536 the reference value for 1D arrays, and src0.w contains
537 the reference value for 2D arrays and cube maps.
539 for cube map array shadow textures, the reference value
540 cannot be passed in src0.w, and TEX2 must be used instead.
546 shadow_ref = src0.z or src0.w (optional)
550 dst = texture\_sample(unit, coord, shadow_ref)
553 .. opcode:: TEX2 - Texture Lookup (for shadow cube map arrays only)
555 this is the same as TEX, but uses another reg to encode the
566 dst = texture\_sample(unit, coord, shadow_ref)
571 .. opcode:: TXD - Texture Lookup with Derivatives
583 dst = texture\_sample\_deriv(unit, coord, ddx, ddy)
586 .. opcode:: TXP - Projective Texture Lookup
590 coord.x = src0.x / src0.w
592 coord.y = src0.y / src0.w
594 coord.z = src0.z / src0.w
600 dst = texture\_sample(unit, coord)
603 .. opcode:: UP2H - Unpack Two 16-Bit Floats
609 Considered for removal.
611 .. opcode:: UP2US - Unpack Two Unsigned 16-Bit Scalars
617 Considered for removal.
619 .. opcode:: UP4B - Unpack Four Signed 8-Bit Values
625 Considered for removal.
627 .. opcode:: UP4UB - Unpack Four Unsigned 8-Bit Scalars
633 Considered for removal.
636 .. opcode:: ARR - Address Register Load With Round
649 .. opcode:: SSG - Set Sign
653 dst.x = (src.x > 0) ? 1 : (src.x < 0) ? -1 : 0
655 dst.y = (src.y > 0) ? 1 : (src.y < 0) ? -1 : 0
657 dst.z = (src.z > 0) ? 1 : (src.z < 0) ? -1 : 0
659 dst.w = (src.w > 0) ? 1 : (src.w < 0) ? -1 : 0
662 .. opcode:: CMP - Compare
666 dst.x = (src0.x < 0) ? src1.x : src2.x
668 dst.y = (src0.y < 0) ? src1.y : src2.y
670 dst.z = (src0.z < 0) ? src1.z : src2.z
672 dst.w = (src0.w < 0) ? src1.w : src2.w
675 .. opcode:: KILL_IF - Conditional Discard
677 Conditional discard. Allowed in fragment shaders only.
681 if (src.x < 0 || src.y < 0 || src.z < 0 || src.w < 0)
686 .. opcode:: KILL - Discard
688 Unconditional discard. Allowed in fragment shaders only.
691 .. opcode:: SCS - Sine Cosine
704 .. opcode:: TXB - Texture Lookup With Bias
706 for cube map array textures and shadow cube maps, the bias value
707 cannot be passed in src0.w, and TXB2 must be used instead.
709 if the target is a shadow texture, the reference value is always
710 in src.z (this prevents shadow 3d and shadow 2d arrays from
711 using this instruction, but this is not needed).
727 dst = texture\_sample(unit, coord, bias)
730 .. opcode:: TXB2 - Texture Lookup With Bias (some cube maps only)
732 this is the same as TXB, but uses another reg to encode the
733 lod bias value for cube map arrays and shadow cube maps.
734 Presumably shadow 2d arrays and shadow 3d targets could use
735 this encoding too, but this is not legal.
737 shadow cube map arrays are neither possible nor required.
747 dst = texture\_sample(unit, coord, bias)
750 .. opcode:: DIV - Divide
754 dst.x = \frac{src0.x}{src1.x}
756 dst.y = \frac{src0.y}{src1.y}
758 dst.z = \frac{src0.z}{src1.z}
760 dst.w = \frac{src0.w}{src1.w}
763 .. opcode:: DP2 - 2-component Dot Product
765 This instruction replicates its result.
769 dst = src0.x \times src1.x + src0.y \times src1.y
772 .. opcode:: TXL - Texture Lookup With explicit LOD
774 for cube map array textures, the explicit lod value
775 cannot be passed in src0.w, and TXL2 must be used instead.
777 if the target is a shadow texture, the reference value is always
778 in src.z (this prevents shadow 3d / 2d array / cube targets from
779 using this instruction, but this is not needed).
795 dst = texture\_sample(unit, coord, lod)
798 .. opcode:: TXL2 - Texture Lookup With explicit LOD (for cube map arrays only)
800 this is the same as TXL, but uses another reg to encode the
802 Presumably shadow 3d / 2d array / cube targets could use
803 this encoding too, but this is not legal.
805 shadow cube map arrays are neither possible nor required.
815 dst = texture\_sample(unit, coord, lod)
818 .. opcode:: PUSHA - Push Address Register On Stack
827 Considered for cleanup.
831 Considered for removal.
833 .. opcode:: POPA - Pop Address Register From Stack
842 Considered for cleanup.
846 Considered for removal.
849 .. opcode:: CALLNZ - Subroutine Call If Not Zero
855 Considered for cleanup.
859 Considered for removal.
863 ^^^^^^^^^^^^^^^^^^^^^^^^
865 These opcodes are primarily provided for special-use computational shaders.
866 Support for these opcodes indicated by a special pipe capability bit (TBD).
868 XXX doesn't look like most of the opcodes really belong here.
870 .. opcode:: CEIL - Ceiling
874 dst.x = \lceil src.x\rceil
876 dst.y = \lceil src.y\rceil
878 dst.z = \lceil src.z\rceil
880 dst.w = \lceil src.w\rceil
883 .. opcode:: TRUNC - Truncate
896 .. opcode:: MOD - Modulus
900 dst.x = src0.x \bmod src1.x
902 dst.y = src0.y \bmod src1.y
904 dst.z = src0.z \bmod src1.z
906 dst.w = src0.w \bmod src1.w
909 .. opcode:: UARL - Integer Address Register Load
911 Moves the contents of the source register, assumed to be an integer, into the
912 destination register, which is assumed to be an address (ADDR) register.
915 .. opcode:: SAD - Sum Of Absolute Differences
919 dst.x = |src0.x - src1.x| + src2.x
921 dst.y = |src0.y - src1.y| + src2.y
923 dst.z = |src0.z - src1.z| + src2.z
925 dst.w = |src0.w - src1.w| + src2.w
928 .. opcode:: TXF - Texel Fetch
930 As per NV_gpu_shader4, extract a single texel from a specified texture
931 image. The source sampler may not be a CUBE or SHADOW. src 0 is a
932 four-component signed integer vector used to identify the single texel
933 accessed. 3 components + level. Just like texture instructions, an optional
934 offset vector is provided, which is subject to various driver restrictions
935 (regarding range, source of offsets).
936 TXF(uint_vec coord, int_vec offset).
939 .. opcode:: TXQ - Texture Size Query
941 As per NV_gpu_program4, retrieve the dimensions of the texture depending on
942 the target. For 1D (width), 2D/RECT/CUBE (width, height), 3D (width, height,
943 depth), 1D array (width, layers), 2D array (width, height, layers).
944 Also return the number of accessible levels (last_level - first_level + 1)
947 For components which don't return a resource dimension, their value
955 dst.x = texture\_width(unit, lod)
957 dst.y = texture\_height(unit, lod)
959 dst.z = texture\_depth(unit, lod)
961 dst.w = texture\_levels(unit)
963 .. opcode:: TG4 - Texture Gather
965 As per ARB_texture_gather, gathers the four texels to be used in a bi-linear
966 filtering operation and packs them into a single register. Only works with
967 2D, 2D array, cubemaps, and cubemaps arrays. For 2D textures, only the
968 addressing modes of the sampler and the top level of any mip pyramid are
969 used. Set W to zero. It behaves like the TEX instruction, but a filtered
970 sample is not generated. The four samples that contribute to filtering are
971 placed into xyzw in clockwise order, starting with the (u,v) texture
972 coordinate delta at the following locations (-, +), (+, +), (+, -), (-, -),
973 where the magnitude of the deltas are half a texel.
975 PIPE_CAP_TEXTURE_SM5 enhances this instruction to support shadow per-sample
976 depth compares, single component selection, and a non-constant offset. It
977 doesn't allow support for the GL independent offset to get i0,j0. This would
978 require another CAP is hw can do it natively. For now we lower that before
987 dst = texture\_gather4 (unit, coord, component)
989 (with SM5 - cube array shadow)
997 dst = texture\_gather (uint, coord, compare)
999 .. opcode:: LODQ - level of detail query
1001 Compute the LOD information that the texture pipe would use to access the
1002 texture. The Y component contains the computed LOD lambda_prime. The X
1003 component contains the LOD that will be accessed, based on min/max lod's
1010 dst.xy = lodq(uint, coord);
1013 ^^^^^^^^^^^^^^^^^^^^^^^^
1014 These opcodes are used for integer operations.
1015 Support for these opcodes indicated by PIPE_SHADER_CAP_INTEGERS (all of them?)
1018 .. opcode:: I2F - Signed Integer To Float
1020 Rounding is unspecified (round to nearest even suggested).
1024 dst.x = (float) src.x
1026 dst.y = (float) src.y
1028 dst.z = (float) src.z
1030 dst.w = (float) src.w
1033 .. opcode:: U2F - Unsigned Integer To Float
1035 Rounding is unspecified (round to nearest even suggested).
1039 dst.x = (float) src.x
1041 dst.y = (float) src.y
1043 dst.z = (float) src.z
1045 dst.w = (float) src.w
1048 .. opcode:: F2I - Float to Signed Integer
1050 Rounding is towards zero (truncate).
1051 Values outside signed range (including NaNs) produce undefined results.
1064 .. opcode:: F2U - Float to Unsigned Integer
1066 Rounding is towards zero (truncate).
1067 Values outside unsigned range (including NaNs) produce undefined results.
1071 dst.x = (unsigned) src.x
1073 dst.y = (unsigned) src.y
1075 dst.z = (unsigned) src.z
1077 dst.w = (unsigned) src.w
1080 .. opcode:: UADD - Integer Add
1082 This instruction works the same for signed and unsigned integers.
1083 The low 32bit of the result is returned.
1087 dst.x = src0.x + src1.x
1089 dst.y = src0.y + src1.y
1091 dst.z = src0.z + src1.z
1093 dst.w = src0.w + src1.w
1096 .. opcode:: UMAD - Integer Multiply And Add
1098 This instruction works the same for signed and unsigned integers.
1099 The multiplication returns the low 32bit (as does the result itself).
1103 dst.x = src0.x \times src1.x + src2.x
1105 dst.y = src0.y \times src1.y + src2.y
1107 dst.z = src0.z \times src1.z + src2.z
1109 dst.w = src0.w \times src1.w + src2.w
1112 .. opcode:: UMUL - Integer Multiply
1114 This instruction works the same for signed and unsigned integers.
1115 The low 32bit of the result is returned.
1119 dst.x = src0.x \times src1.x
1121 dst.y = src0.y \times src1.y
1123 dst.z = src0.z \times src1.z
1125 dst.w = src0.w \times src1.w
1128 .. opcode:: IMUL_HI - Signed Integer Multiply High Bits
1130 The high 32bits of the multiplication of 2 signed integers are returned.
1134 dst.x = (src0.x \times src1.x) >> 32
1136 dst.y = (src0.y \times src1.y) >> 32
1138 dst.z = (src0.z \times src1.z) >> 32
1140 dst.w = (src0.w \times src1.w) >> 32
1143 .. opcode:: UMUL_HI - Unsigned Integer Multiply High Bits
1145 The high 32bits of the multiplication of 2 unsigned integers are returned.
1149 dst.x = (src0.x \times src1.x) >> 32
1151 dst.y = (src0.y \times src1.y) >> 32
1153 dst.z = (src0.z \times src1.z) >> 32
1155 dst.w = (src0.w \times src1.w) >> 32
1158 .. opcode:: IDIV - Signed Integer Division
1160 TBD: behavior for division by zero.
1164 dst.x = src0.x \ src1.x
1166 dst.y = src0.y \ src1.y
1168 dst.z = src0.z \ src1.z
1170 dst.w = src0.w \ src1.w
1173 .. opcode:: UDIV - Unsigned Integer Division
1175 For division by zero, 0xffffffff is returned.
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:: UMOD - Unsigned Integer Remainder
1190 If second arg is zero, 0xffffffff is returned.
1194 dst.x = src0.x \ src1.x
1196 dst.y = src0.y \ src1.y
1198 dst.z = src0.z \ src1.z
1200 dst.w = src0.w \ src1.w
1203 .. opcode:: NOT - Bitwise Not
1216 .. opcode:: AND - Bitwise And
1220 dst.x = src0.x \& src1.x
1222 dst.y = src0.y \& src1.y
1224 dst.z = src0.z \& src1.z
1226 dst.w = src0.w \& src1.w
1229 .. opcode:: OR - Bitwise Or
1233 dst.x = src0.x | src1.x
1235 dst.y = src0.y | src1.y
1237 dst.z = src0.z | src1.z
1239 dst.w = src0.w | src1.w
1242 .. opcode:: XOR - Bitwise Xor
1246 dst.x = src0.x \oplus src1.x
1248 dst.y = src0.y \oplus src1.y
1250 dst.z = src0.z \oplus src1.z
1252 dst.w = src0.w \oplus src1.w
1255 .. opcode:: IMAX - Maximum of Signed Integers
1259 dst.x = max(src0.x, src1.x)
1261 dst.y = max(src0.y, src1.y)
1263 dst.z = max(src0.z, src1.z)
1265 dst.w = max(src0.w, src1.w)
1268 .. opcode:: UMAX - Maximum of Unsigned Integers
1272 dst.x = max(src0.x, src1.x)
1274 dst.y = max(src0.y, src1.y)
1276 dst.z = max(src0.z, src1.z)
1278 dst.w = max(src0.w, src1.w)
1281 .. opcode:: IMIN - Minimum of Signed Integers
1285 dst.x = min(src0.x, src1.x)
1287 dst.y = min(src0.y, src1.y)
1289 dst.z = min(src0.z, src1.z)
1291 dst.w = min(src0.w, src1.w)
1294 .. opcode:: UMIN - Minimum of Unsigned Integers
1298 dst.x = min(src0.x, src1.x)
1300 dst.y = min(src0.y, src1.y)
1302 dst.z = min(src0.z, src1.z)
1304 dst.w = min(src0.w, src1.w)
1307 .. opcode:: SHL - Shift Left
1309 The shift count is masked with 0x1f before the shift is applied.
1313 dst.x = src0.x << (0x1f \& src1.x)
1315 dst.y = src0.y << (0x1f \& src1.y)
1317 dst.z = src0.z << (0x1f \& src1.z)
1319 dst.w = src0.w << (0x1f \& src1.w)
1322 .. opcode:: ISHR - Arithmetic Shift Right (of Signed Integer)
1324 The shift count is masked with 0x1f before the shift is applied.
1328 dst.x = src0.x >> (0x1f \& src1.x)
1330 dst.y = src0.y >> (0x1f \& src1.y)
1332 dst.z = src0.z >> (0x1f \& src1.z)
1334 dst.w = src0.w >> (0x1f \& src1.w)
1337 .. opcode:: USHR - Logical Shift Right
1339 The shift count is masked with 0x1f before the shift is applied.
1343 dst.x = src0.x >> (unsigned) (0x1f \& src1.x)
1345 dst.y = src0.y >> (unsigned) (0x1f \& src1.y)
1347 dst.z = src0.z >> (unsigned) (0x1f \& src1.z)
1349 dst.w = src0.w >> (unsigned) (0x1f \& src1.w)
1352 .. opcode:: UCMP - Integer Conditional Move
1356 dst.x = src0.x ? src1.x : src2.x
1358 dst.y = src0.y ? src1.y : src2.y
1360 dst.z = src0.z ? src1.z : src2.z
1362 dst.w = src0.w ? src1.w : src2.w
1366 .. opcode:: ISSG - Integer Set Sign
1370 dst.x = (src0.x < 0) ? -1 : (src0.x > 0) ? 1 : 0
1372 dst.y = (src0.y < 0) ? -1 : (src0.y > 0) ? 1 : 0
1374 dst.z = (src0.z < 0) ? -1 : (src0.z > 0) ? 1 : 0
1376 dst.w = (src0.w < 0) ? -1 : (src0.w > 0) ? 1 : 0
1380 .. opcode:: FSLT - Float Set On Less Than (ordered)
1382 Same comparison as SLT but returns integer instead of 1.0/0.0 float
1386 dst.x = (src0.x < src1.x) ? \sim 0 : 0
1388 dst.y = (src0.y < src1.y) ? \sim 0 : 0
1390 dst.z = (src0.z < src1.z) ? \sim 0 : 0
1392 dst.w = (src0.w < src1.w) ? \sim 0 : 0
1395 .. opcode:: ISLT - Signed Integer Set On Less Than
1399 dst.x = (src0.x < src1.x) ? \sim 0 : 0
1401 dst.y = (src0.y < src1.y) ? \sim 0 : 0
1403 dst.z = (src0.z < src1.z) ? \sim 0 : 0
1405 dst.w = (src0.w < src1.w) ? \sim 0 : 0
1408 .. opcode:: USLT - Unsigned Integer Set On Less Than
1412 dst.x = (src0.x < src1.x) ? \sim 0 : 0
1414 dst.y = (src0.y < src1.y) ? \sim 0 : 0
1416 dst.z = (src0.z < src1.z) ? \sim 0 : 0
1418 dst.w = (src0.w < src1.w) ? \sim 0 : 0
1421 .. opcode:: FSGE - Float Set On Greater Equal Than (ordered)
1423 Same comparison as SGE but returns integer instead of 1.0/0.0 float
1427 dst.x = (src0.x >= src1.x) ? \sim 0 : 0
1429 dst.y = (src0.y >= src1.y) ? \sim 0 : 0
1431 dst.z = (src0.z >= src1.z) ? \sim 0 : 0
1433 dst.w = (src0.w >= src1.w) ? \sim 0 : 0
1436 .. opcode:: ISGE - Signed Integer Set On Greater Equal Than
1440 dst.x = (src0.x >= src1.x) ? \sim 0 : 0
1442 dst.y = (src0.y >= src1.y) ? \sim 0 : 0
1444 dst.z = (src0.z >= src1.z) ? \sim 0 : 0
1446 dst.w = (src0.w >= src1.w) ? \sim 0 : 0
1449 .. opcode:: USGE - Unsigned Integer Set On Greater Equal Than
1453 dst.x = (src0.x >= src1.x) ? \sim 0 : 0
1455 dst.y = (src0.y >= src1.y) ? \sim 0 : 0
1457 dst.z = (src0.z >= src1.z) ? \sim 0 : 0
1459 dst.w = (src0.w >= src1.w) ? \sim 0 : 0
1462 .. opcode:: FSEQ - Float Set On Equal (ordered)
1464 Same comparison as SEQ but returns integer instead of 1.0/0.0 float
1468 dst.x = (src0.x == src1.x) ? \sim 0 : 0
1470 dst.y = (src0.y == src1.y) ? \sim 0 : 0
1472 dst.z = (src0.z == src1.z) ? \sim 0 : 0
1474 dst.w = (src0.w == src1.w) ? \sim 0 : 0
1477 .. opcode:: USEQ - Integer Set On Equal
1481 dst.x = (src0.x == src1.x) ? \sim 0 : 0
1483 dst.y = (src0.y == src1.y) ? \sim 0 : 0
1485 dst.z = (src0.z == src1.z) ? \sim 0 : 0
1487 dst.w = (src0.w == src1.w) ? \sim 0 : 0
1490 .. opcode:: FSNE - Float Set On Not Equal (unordered)
1492 Same comparison as SNE but returns integer instead of 1.0/0.0 float
1496 dst.x = (src0.x != src1.x) ? \sim 0 : 0
1498 dst.y = (src0.y != src1.y) ? \sim 0 : 0
1500 dst.z = (src0.z != src1.z) ? \sim 0 : 0
1502 dst.w = (src0.w != src1.w) ? \sim 0 : 0
1505 .. opcode:: USNE - Integer Set On Not Equal
1509 dst.x = (src0.x != src1.x) ? \sim 0 : 0
1511 dst.y = (src0.y != src1.y) ? \sim 0 : 0
1513 dst.z = (src0.z != src1.z) ? \sim 0 : 0
1515 dst.w = (src0.w != src1.w) ? \sim 0 : 0
1518 .. opcode:: INEG - Integer Negate
1533 .. opcode:: IABS - Integer Absolute Value
1547 These opcodes are used for bit-level manipulation of integers.
1549 .. opcode:: IBFE - Signed Bitfield Extract
1551 See SM5 instruction of the same name. Extracts a set of bits from the input,
1552 and sign-extends them if the high bit of the extracted window is set.
1556 def ibfe(value, offset, bits):
1557 offset = offset & 0x1f
1559 if bits == 0: return 0
1560 # Note: >> sign-extends
1561 if width + offset < 32:
1562 return (value << (32 - offset - bits)) >> (32 - bits)
1564 return value >> offset
1566 .. opcode:: UBFE - Unsigned Bitfield Extract
1568 See SM5 instruction of the same name. Extracts a set of bits from the input,
1569 without any sign-extension.
1573 def ubfe(value, offset, bits):
1574 offset = offset & 0x1f
1576 if bits == 0: return 0
1577 # Note: >> does not sign-extend
1578 if width + offset < 32:
1579 return (value << (32 - offset - bits)) >> (32 - bits)
1581 return value >> offset
1583 .. opcode:: BFI - Bitfield Insert
1585 See SM5 instruction of the same name. Replaces a bit region of 'base' with
1586 the low bits of 'insert'.
1590 def bfi(base, insert, offset, bits):
1591 offset = offset & 0x1f
1593 mask = ((1 << bits) - 1) << offset
1594 return ((insert << offset) & mask) | (base & ~mask)
1596 .. opcode:: BREV - Bitfield Reverse
1598 See SM5 instruction BFREV. Reverses the bits of the argument.
1600 .. opcode:: POPC - Population Count
1602 See SM5 instruction COUNTBITS. Counts the number of set bits in the argument.
1604 .. opcode:: LSB - Index of lowest set bit
1606 See SM5 instruction FIRSTBIT_LO. Computes the 0-based index of the first set
1607 bit of the argument. Returns -1 if none are set.
1609 .. opcode:: IMSB - Index of highest non-sign bit
1611 See SM5 instruction FIRSTBIT_SHI. Computes the 0-based index of the highest
1612 non-sign bit of the argument (i.e. highest 0 bit for negative numbers,
1613 highest 1 bit for positive numbers). Returns -1 if all bits are the same
1614 (i.e. for inputs 0 and -1).
1616 .. opcode:: UMSB - Index of highest set bit
1618 See SM5 instruction FIRSTBIT_HI. Computes the 0-based index of the highest
1619 set bit of the argument. Returns -1 if none are set.
1622 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1624 These opcodes are only supported in geometry shaders; they have no meaning
1625 in any other type of shader.
1627 .. opcode:: EMIT - Emit
1629 Generate a new vertex for the current primitive into the specified vertex
1630 stream using the values in the output registers.
1633 .. opcode:: ENDPRIM - End Primitive
1635 Complete the current primitive in the specified vertex stream (consisting of
1636 the emitted vertices), and start a new one.
1642 These opcodes are part of :term:`GLSL`'s opcode set. Support for these
1643 opcodes is determined by a special capability bit, ``GLSL``.
1644 Some require glsl version 1.30 (UIF/BREAKC/SWITCH/CASE/DEFAULT/ENDSWITCH).
1646 .. opcode:: CAL - Subroutine Call
1652 .. opcode:: RET - Subroutine Call Return
1657 .. opcode:: CONT - Continue
1659 Unconditionally moves the point of execution to the instruction after the
1660 last bgnloop. The instruction must appear within a bgnloop/endloop.
1664 Support for CONT is determined by a special capability bit,
1665 ``TGSI_CONT_SUPPORTED``. See :ref:`Screen` for more information.
1668 .. opcode:: BGNLOOP - Begin a Loop
1670 Start a loop. Must have a matching endloop.
1673 .. opcode:: BGNSUB - Begin Subroutine
1675 Starts definition of a subroutine. Must have a matching endsub.
1678 .. opcode:: ENDLOOP - End a Loop
1680 End a loop started with bgnloop.
1683 .. opcode:: ENDSUB - End Subroutine
1685 Ends definition of a subroutine.
1688 .. opcode:: NOP - No Operation
1693 .. opcode:: BRK - Break
1695 Unconditionally moves the point of execution to the instruction after the
1696 next endloop or endswitch. The instruction must appear within a loop/endloop
1697 or switch/endswitch.
1700 .. opcode:: BREAKC - Break Conditional
1702 Conditionally moves the point of execution to the instruction after the
1703 next endloop or endswitch. The instruction must appear within a loop/endloop
1704 or switch/endswitch.
1705 Condition evaluates to true if src0.x != 0 where src0.x is interpreted
1706 as an integer register.
1710 Considered for removal as it's quite inconsistent wrt other opcodes
1711 (could emulate with UIF/BRK/ENDIF).
1714 .. opcode:: IF - Float If
1716 Start an IF ... ELSE .. ENDIF block. Condition evaluates to true if
1720 where src0.x is interpreted as a floating point register.
1723 .. opcode:: UIF - Bitwise If
1725 Start an UIF ... ELSE .. ENDIF block. Condition evaluates to true if
1729 where src0.x is interpreted as an integer register.
1732 .. opcode:: ELSE - Else
1734 Starts an else block, after an IF or UIF statement.
1737 .. opcode:: ENDIF - End If
1739 Ends an IF or UIF block.
1742 .. opcode:: SWITCH - Switch
1744 Starts a C-style switch expression. The switch consists of one or multiple
1745 CASE statements, and at most one DEFAULT statement. Execution of a statement
1746 ends when a BRK is hit, but just like in C falling through to other cases
1747 without a break is allowed. Similarly, DEFAULT label is allowed anywhere not
1748 just as last statement, and fallthrough is allowed into/from it.
1749 CASE src arguments are evaluated at bit level against the SWITCH src argument.
1755 (some instructions here)
1758 (some instructions here)
1761 (some instructions here)
1766 .. opcode:: CASE - Switch case
1768 This represents a switch case label. The src arg must be an integer immediate.
1771 .. opcode:: DEFAULT - Switch default
1773 This represents the default case in the switch, which is taken if no other
1777 .. opcode:: ENDSWITCH - End of switch
1779 Ends a switch expression.
1785 The interpolation instructions allow an input to be interpolated in a
1786 different way than its declaration. This corresponds to the GLSL 4.00
1787 interpolateAt* functions. The first argument of each of these must come from
1788 ``TGSI_FILE_INPUT``.
1790 .. opcode:: INTERP_CENTROID - Interpolate at the centroid
1792 Interpolates the varying specified by src0 at the centroid
1794 .. opcode:: INTERP_SAMPLE - Interpolate at the specified sample
1796 Interpolates the varying specified by src0 at the sample id specified by
1797 src1.x (interpreted as an integer)
1799 .. opcode:: INTERP_OFFSET - Interpolate at the specified offset
1801 Interpolates the varying specified by src0 at the offset src1.xy from the
1802 pixel center (interpreted as floats)
1810 The double-precision opcodes reinterpret four-component vectors into
1811 two-component vectors with doubled precision in each component.
1813 Support for these opcodes is XXX undecided. :T
1815 .. opcode:: DADD - Add
1819 dst.xy = src0.xy + src1.xy
1821 dst.zw = src0.zw + src1.zw
1824 .. opcode:: DDIV - Divide
1828 dst.xy = src0.xy / src1.xy
1830 dst.zw = src0.zw / src1.zw
1832 .. opcode:: DSEQ - Set on Equal
1836 dst.xy = src0.xy == src1.xy ? 1.0F : 0.0F
1838 dst.zw = src0.zw == src1.zw ? 1.0F : 0.0F
1840 .. opcode:: DSLT - Set on Less than
1844 dst.xy = src0.xy < src1.xy ? 1.0F : 0.0F
1846 dst.zw = src0.zw < src1.zw ? 1.0F : 0.0F
1848 .. opcode:: DFRAC - Fraction
1852 dst.xy = src.xy - \lfloor src.xy\rfloor
1854 dst.zw = src.zw - \lfloor src.zw\rfloor
1857 .. opcode:: DFRACEXP - Convert Number to Fractional and Integral Components
1859 Like the ``frexp()`` routine in many math libraries, this opcode stores the
1860 exponent of its source to ``dst0``, and the significand to ``dst1``, such that
1861 :math:`dst1 \times 2^{dst0} = src` .
1865 dst0.xy = exp(src.xy)
1867 dst1.xy = frac(src.xy)
1869 dst0.zw = exp(src.zw)
1871 dst1.zw = frac(src.zw)
1873 .. opcode:: DLDEXP - Multiply Number by Integral Power of 2
1875 This opcode is the inverse of :opcode:`DFRACEXP`.
1879 dst.xy = src0.xy \times 2^{src1.xy}
1881 dst.zw = src0.zw \times 2^{src1.zw}
1883 .. opcode:: DMIN - Minimum
1887 dst.xy = min(src0.xy, src1.xy)
1889 dst.zw = min(src0.zw, src1.zw)
1891 .. opcode:: DMAX - Maximum
1895 dst.xy = max(src0.xy, src1.xy)
1897 dst.zw = max(src0.zw, src1.zw)
1899 .. opcode:: DMUL - Multiply
1903 dst.xy = src0.xy \times src1.xy
1905 dst.zw = src0.zw \times src1.zw
1908 .. opcode:: DMAD - Multiply And Add
1912 dst.xy = src0.xy \times src1.xy + src2.xy
1914 dst.zw = src0.zw \times src1.zw + src2.zw
1917 .. opcode:: DRCP - Reciprocal
1921 dst.xy = \frac{1}{src.xy}
1923 dst.zw = \frac{1}{src.zw}
1925 .. opcode:: DSQRT - Square Root
1929 dst.xy = \sqrt{src.xy}
1931 dst.zw = \sqrt{src.zw}
1934 .. _samplingopcodes:
1936 Resource Sampling Opcodes
1937 ^^^^^^^^^^^^^^^^^^^^^^^^^
1939 Those opcodes follow very closely semantics of the respective Direct3D
1940 instructions. If in doubt double check Direct3D documentation.
1941 Note that the swizzle on SVIEW (src1) determines texel swizzling
1946 Using provided address, sample data from the specified texture using the
1947 filtering mode identified by the gven sampler. The source data may come from
1948 any resource type other than buffers.
1950 Syntax: ``SAMPLE dst, address, sampler_view, sampler``
1952 Example: ``SAMPLE TEMP[0], TEMP[1], SVIEW[0], SAMP[0]``
1954 .. opcode:: SAMPLE_I
1956 Simplified alternative to the SAMPLE instruction. Using the provided
1957 integer address, SAMPLE_I fetches data from the specified sampler view
1958 without any filtering. The source data may come from any resource type
1961 Syntax: ``SAMPLE_I dst, address, sampler_view``
1963 Example: ``SAMPLE_I TEMP[0], TEMP[1], SVIEW[0]``
1965 The 'address' is specified as unsigned integers. If the 'address' is out of
1966 range [0...(# texels - 1)] the result of the fetch is always 0 in all
1967 components. As such the instruction doesn't honor address wrap modes, in
1968 cases where that behavior is desirable 'SAMPLE' instruction should be used.
1969 address.w always provides an unsigned integer mipmap level. If the value is
1970 out of the range then the instruction always returns 0 in all components.
1971 address.yz are ignored for buffers and 1d textures. address.z is ignored
1972 for 1d texture arrays and 2d textures.
1974 For 1D texture arrays address.y provides the array index (also as unsigned
1975 integer). If the value is out of the range of available array indices
1976 [0... (array size - 1)] then the opcode always returns 0 in all components.
1977 For 2D texture arrays address.z provides the array index, otherwise it
1978 exhibits the same behavior as in the case for 1D texture arrays. The exact
1979 semantics of the source address are presented in the table below:
1981 +---------------------------+----+-----+-----+---------+
1982 | resource type | X | Y | Z | W |
1983 +===========================+====+=====+=====+=========+
1984 | ``PIPE_BUFFER`` | x | | | ignored |
1985 +---------------------------+----+-----+-----+---------+
1986 | ``PIPE_TEXTURE_1D`` | x | | | mpl |
1987 +---------------------------+----+-----+-----+---------+
1988 | ``PIPE_TEXTURE_2D`` | x | y | | mpl |
1989 +---------------------------+----+-----+-----+---------+
1990 | ``PIPE_TEXTURE_3D`` | x | y | z | mpl |
1991 +---------------------------+----+-----+-----+---------+
1992 | ``PIPE_TEXTURE_RECT`` | x | y | | mpl |
1993 +---------------------------+----+-----+-----+---------+
1994 | ``PIPE_TEXTURE_CUBE`` | not allowed as source |
1995 +---------------------------+----+-----+-----+---------+
1996 | ``PIPE_TEXTURE_1D_ARRAY`` | x | idx | | mpl |
1997 +---------------------------+----+-----+-----+---------+
1998 | ``PIPE_TEXTURE_2D_ARRAY`` | x | y | idx | mpl |
1999 +---------------------------+----+-----+-----+---------+
2001 Where 'mpl' is a mipmap level and 'idx' is the array index.
2003 .. opcode:: SAMPLE_I_MS
2005 Just like SAMPLE_I but allows fetch data from multi-sampled surfaces.
2007 Syntax: ``SAMPLE_I_MS dst, address, sampler_view, sample``
2009 .. opcode:: SAMPLE_B
2011 Just like the SAMPLE instruction with the exception that an additional bias
2012 is applied to the level of detail computed as part of the instruction
2015 Syntax: ``SAMPLE_B dst, address, sampler_view, sampler, lod_bias``
2017 Example: ``SAMPLE_B TEMP[0], TEMP[1], SVIEW[0], SAMP[0], TEMP[2].x``
2019 .. opcode:: SAMPLE_C
2021 Similar to the SAMPLE instruction but it performs a comparison filter. The
2022 operands to SAMPLE_C are identical to SAMPLE, except that there is an
2023 additional float32 operand, reference value, which must be a register with
2024 single-component, or a scalar literal. SAMPLE_C makes the hardware use the
2025 current samplers compare_func (in pipe_sampler_state) to compare reference
2026 value against the red component value for the surce resource at each texel
2027 that the currently configured texture filter covers based on the provided
2030 Syntax: ``SAMPLE_C dst, address, sampler_view.r, sampler, ref_value``
2032 Example: ``SAMPLE_C TEMP[0], TEMP[1], SVIEW[0].r, SAMP[0], TEMP[2].x``
2034 .. opcode:: SAMPLE_C_LZ
2036 Same as SAMPLE_C, but LOD is 0 and derivatives are ignored. The LZ stands
2039 Syntax: ``SAMPLE_C_LZ dst, address, sampler_view.r, sampler, ref_value``
2041 Example: ``SAMPLE_C_LZ TEMP[0], TEMP[1], SVIEW[0].r, SAMP[0], TEMP[2].x``
2044 .. opcode:: SAMPLE_D
2046 SAMPLE_D is identical to the SAMPLE opcode except that the derivatives for
2047 the source address in the x direction and the y direction are provided by
2050 Syntax: ``SAMPLE_D dst, address, sampler_view, sampler, der_x, der_y``
2052 Example: ``SAMPLE_D TEMP[0], TEMP[1], SVIEW[0], SAMP[0], TEMP[2], TEMP[3]``
2054 .. opcode:: SAMPLE_L
2056 SAMPLE_L is identical to the SAMPLE opcode except that the LOD is provided
2057 directly as a scalar value, representing no anisotropy.
2059 Syntax: ``SAMPLE_L dst, address, sampler_view, sampler, explicit_lod``
2061 Example: ``SAMPLE_L TEMP[0], TEMP[1], SVIEW[0], SAMP[0], TEMP[2].x``
2065 Gathers the four texels to be used in a bi-linear filtering operation and
2066 packs them into a single register. Only works with 2D, 2D array, cubemaps,
2067 and cubemaps arrays. For 2D textures, only the addressing modes of the
2068 sampler and the top level of any mip pyramid are used. Set W to zero. It
2069 behaves like the SAMPLE instruction, but a filtered sample is not
2070 generated. The four samples that contribute to filtering are placed into
2071 xyzw in counter-clockwise order, starting with the (u,v) texture coordinate
2072 delta at the following locations (-, +), (+, +), (+, -), (-, -), where the
2073 magnitude of the deltas are half a texel.
2076 .. opcode:: SVIEWINFO
2078 Query the dimensions of a given sampler view. dst receives width, height,
2079 depth or array size and number of mipmap levels as int4. The dst can have a
2080 writemask which will specify what info is the caller interested in.
2082 Syntax: ``SVIEWINFO dst, src_mip_level, sampler_view``
2084 Example: ``SVIEWINFO TEMP[0], TEMP[1].x, SVIEW[0]``
2086 src_mip_level is an unsigned integer scalar. If it's out of range then
2087 returns 0 for width, height and depth/array size but the total number of
2088 mipmap is still returned correctly for the given sampler view. The returned
2089 width, height and depth values are for the mipmap level selected by the
2090 src_mip_level and are in the number of texels. For 1d texture array width
2091 is in dst.x, array size is in dst.y and dst.z is 0. The number of mipmaps is
2092 still in dst.w. In contrast to d3d10 resinfo, there's no way in the tgsi
2093 instruction encoding to specify the return type (float/rcpfloat/uint), hence
2094 always using uint. Also, unlike the SAMPLE instructions, the swizzle on src1
2095 resinfo allowing swizzling dst values is ignored (due to the interaction
2096 with rcpfloat modifier which requires some swizzle handling in the state
2099 .. opcode:: SAMPLE_POS
2101 Query the position of a given sample. dst receives float4 (x, y, 0, 0)
2102 indicated where the sample is located. If the resource is not a multi-sample
2103 resource and not a render target, the result is 0.
2105 .. opcode:: SAMPLE_INFO
2107 dst receives number of samples in x. If the resource is not a multi-sample
2108 resource and not a render target, the result is 0.
2111 .. _resourceopcodes:
2113 Resource Access Opcodes
2114 ^^^^^^^^^^^^^^^^^^^^^^^
2116 .. opcode:: LOAD - Fetch data from a shader resource
2118 Syntax: ``LOAD dst, resource, address``
2120 Example: ``LOAD TEMP[0], RES[0], TEMP[1]``
2122 Using the provided integer address, LOAD fetches data
2123 from the specified buffer or texture without any
2126 The 'address' is specified as a vector of unsigned
2127 integers. If the 'address' is out of range the result
2130 Only the first mipmap level of a resource can be read
2131 from using this instruction.
2133 For 1D or 2D texture arrays, the array index is
2134 provided as an unsigned integer in address.y or
2135 address.z, respectively. address.yz are ignored for
2136 buffers and 1D textures. address.z is ignored for 1D
2137 texture arrays and 2D textures. address.w is always
2140 .. opcode:: STORE - Write data to a shader resource
2142 Syntax: ``STORE resource, address, src``
2144 Example: ``STORE RES[0], TEMP[0], TEMP[1]``
2146 Using the provided integer address, STORE writes data
2147 to the specified buffer or texture.
2149 The 'address' is specified as a vector of unsigned
2150 integers. If the 'address' is out of range the result
2153 Only the first mipmap level of a resource can be
2154 written to using this instruction.
2156 For 1D or 2D texture arrays, the array index is
2157 provided as an unsigned integer in address.y or
2158 address.z, respectively. address.yz are ignored for
2159 buffers and 1D textures. address.z is ignored for 1D
2160 texture arrays and 2D textures. address.w is always
2164 .. _threadsyncopcodes:
2166 Inter-thread synchronization opcodes
2167 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2169 These opcodes are intended for communication between threads running
2170 within the same compute grid. For now they're only valid in compute
2173 .. opcode:: MFENCE - Memory fence
2175 Syntax: ``MFENCE resource``
2177 Example: ``MFENCE RES[0]``
2179 This opcode forces strong ordering between any memory access
2180 operations that affect the specified resource. This means that
2181 previous loads and stores (and only those) will be performed and
2182 visible to other threads before the program execution continues.
2185 .. opcode:: LFENCE - Load memory fence
2187 Syntax: ``LFENCE resource``
2189 Example: ``LFENCE RES[0]``
2191 Similar to MFENCE, but it only affects the ordering of memory loads.
2194 .. opcode:: SFENCE - Store memory fence
2196 Syntax: ``SFENCE resource``
2198 Example: ``SFENCE RES[0]``
2200 Similar to MFENCE, but it only affects the ordering of memory stores.
2203 .. opcode:: BARRIER - Thread group barrier
2207 This opcode suspends the execution of the current thread until all
2208 the remaining threads in the working group reach the same point of
2209 the program. Results are unspecified if any of the remaining
2210 threads terminates or never reaches an executed BARRIER instruction.
2218 These opcodes provide atomic variants of some common arithmetic and
2219 logical operations. In this context atomicity means that another
2220 concurrent memory access operation that affects the same memory
2221 location is guaranteed to be performed strictly before or after the
2222 entire execution of the atomic operation.
2224 For the moment they're only valid in compute programs.
2226 .. opcode:: ATOMUADD - Atomic integer addition
2228 Syntax: ``ATOMUADD dst, resource, offset, src``
2230 Example: ``ATOMUADD TEMP[0], RES[0], TEMP[1], TEMP[2]``
2232 The following operation is performed atomically on each component:
2236 dst_i = resource[offset]_i
2238 resource[offset]_i = dst_i + src_i
2241 .. opcode:: ATOMXCHG - Atomic exchange
2243 Syntax: ``ATOMXCHG dst, resource, offset, src``
2245 Example: ``ATOMXCHG TEMP[0], RES[0], TEMP[1], TEMP[2]``
2247 The following operation is performed atomically on each component:
2251 dst_i = resource[offset]_i
2253 resource[offset]_i = src_i
2256 .. opcode:: ATOMCAS - Atomic compare-and-exchange
2258 Syntax: ``ATOMCAS dst, resource, offset, cmp, src``
2260 Example: ``ATOMCAS TEMP[0], RES[0], TEMP[1], TEMP[2], TEMP[3]``
2262 The following operation is performed atomically on each component:
2266 dst_i = resource[offset]_i
2268 resource[offset]_i = (dst_i == cmp_i ? src_i : dst_i)
2271 .. opcode:: ATOMAND - Atomic bitwise And
2273 Syntax: ``ATOMAND dst, resource, offset, src``
2275 Example: ``ATOMAND TEMP[0], RES[0], TEMP[1], TEMP[2]``
2277 The following operation is performed atomically on each component:
2281 dst_i = resource[offset]_i
2283 resource[offset]_i = dst_i \& src_i
2286 .. opcode:: ATOMOR - Atomic bitwise Or
2288 Syntax: ``ATOMOR dst, resource, offset, src``
2290 Example: ``ATOMOR TEMP[0], RES[0], TEMP[1], TEMP[2]``
2292 The following operation is performed atomically on each component:
2296 dst_i = resource[offset]_i
2298 resource[offset]_i = dst_i | src_i
2301 .. opcode:: ATOMXOR - Atomic bitwise Xor
2303 Syntax: ``ATOMXOR dst, resource, offset, src``
2305 Example: ``ATOMXOR TEMP[0], RES[0], TEMP[1], TEMP[2]``
2307 The following operation is performed atomically on each component:
2311 dst_i = resource[offset]_i
2313 resource[offset]_i = dst_i \oplus src_i
2316 .. opcode:: ATOMUMIN - Atomic unsigned minimum
2318 Syntax: ``ATOMUMIN dst, resource, offset, src``
2320 Example: ``ATOMUMIN TEMP[0], RES[0], TEMP[1], TEMP[2]``
2322 The following operation is performed atomically on each component:
2326 dst_i = resource[offset]_i
2328 resource[offset]_i = (dst_i < src_i ? dst_i : src_i)
2331 .. opcode:: ATOMUMAX - Atomic unsigned maximum
2333 Syntax: ``ATOMUMAX dst, resource, offset, src``
2335 Example: ``ATOMUMAX TEMP[0], RES[0], TEMP[1], TEMP[2]``
2337 The following operation is performed atomically on each component:
2341 dst_i = resource[offset]_i
2343 resource[offset]_i = (dst_i > src_i ? dst_i : src_i)
2346 .. opcode:: ATOMIMIN - Atomic signed minimum
2348 Syntax: ``ATOMIMIN dst, resource, offset, src``
2350 Example: ``ATOMIMIN TEMP[0], RES[0], TEMP[1], TEMP[2]``
2352 The following operation is performed atomically on each component:
2356 dst_i = resource[offset]_i
2358 resource[offset]_i = (dst_i < src_i ? dst_i : src_i)
2361 .. opcode:: ATOMIMAX - Atomic signed maximum
2363 Syntax: ``ATOMIMAX dst, resource, offset, src``
2365 Example: ``ATOMIMAX TEMP[0], RES[0], TEMP[1], TEMP[2]``
2367 The following operation is performed atomically on each component:
2371 dst_i = resource[offset]_i
2373 resource[offset]_i = (dst_i > src_i ? dst_i : src_i)
2377 Explanation of symbols used
2378 ------------------------------
2385 :math:`|x|` Absolute value of `x`.
2387 :math:`\lceil x \rceil` Ceiling of `x`.
2389 clamp(x,y,z) Clamp x between y and z.
2390 (x < y) ? y : (x > z) ? z : x
2392 :math:`\lfloor x\rfloor` Floor of `x`.
2394 :math:`\log_2{x}` Logarithm of `x`, base 2.
2396 max(x,y) Maximum of x and y.
2399 min(x,y) Minimum of x and y.
2402 partialx(x) Derivative of x relative to fragment's X.
2404 partialy(x) Derivative of x relative to fragment's Y.
2406 pop() Pop from stack.
2408 :math:`x^y` `x` to the power `y`.
2410 push(x) Push x on stack.
2414 trunc(x) Truncate x, i.e. drop the fraction bits.
2421 discard Discard fragment.
2425 target Label of target instruction.
2436 Declares a register that is will be referenced as an operand in Instruction
2439 File field contains register file that is being declared and is one
2442 UsageMask field specifies which of the register components can be accessed
2443 and is one of TGSI_WRITEMASK.
2445 The Local flag specifies that a given value isn't intended for
2446 subroutine parameter passing and, as a result, the implementation
2447 isn't required to give any guarantees of it being preserved across
2448 subroutine boundaries. As it's merely a compiler hint, the
2449 implementation is free to ignore it.
2451 If Dimension flag is set to 1, a Declaration Dimension token follows.
2453 If Semantic flag is set to 1, a Declaration Semantic token follows.
2455 If Interpolate flag is set to 1, a Declaration Interpolate token follows.
2457 If file is TGSI_FILE_RESOURCE, a Declaration Resource token follows.
2459 If Array flag is set to 1, a Declaration Array token follows.
2462 ^^^^^^^^^^^^^^^^^^^^^^^^
2464 Declarations can optional have an ArrayID attribute which can be referred by
2465 indirect addressing operands. An ArrayID of zero is reserved and treaded as
2466 if no ArrayID is specified.
2468 If an indirect addressing operand refers to a specific declaration by using
2469 an ArrayID only the registers in this declaration are guaranteed to be
2470 accessed, accessing any register outside this declaration results in undefined
2471 behavior. Note that for compatibility the effective index is zero-based and
2472 not relative to the specified declaration
2474 If no ArrayID is specified with an indirect addressing operand the whole
2475 register file might be accessed by this operand. This is strongly discouraged
2476 and will prevent packing of scalar/vec2 arrays and effective alias analysis.
2478 Declaration Semantic
2479 ^^^^^^^^^^^^^^^^^^^^^^^^
2481 Vertex and fragment shader input and output registers may be labeled
2482 with semantic information consisting of a name and index.
2484 Follows Declaration token if Semantic bit is set.
2486 Since its purpose is to link a shader with other stages of the pipeline,
2487 it is valid to follow only those Declaration tokens that declare a register
2488 either in INPUT or OUTPUT file.
2490 SemanticName field contains the semantic name of the register being declared.
2491 There is no default value.
2493 SemanticIndex is an optional subscript that can be used to distinguish
2494 different register declarations with the same semantic name. The default value
2497 The meanings of the individual semantic names are explained in the following
2500 TGSI_SEMANTIC_POSITION
2501 """"""""""""""""""""""
2503 For vertex shaders, TGSI_SEMANTIC_POSITION indicates the vertex shader
2504 output register which contains the homogeneous vertex position in the clip
2505 space coordinate system. After clipping, the X, Y and Z components of the
2506 vertex will be divided by the W value to get normalized device coordinates.
2508 For fragment shaders, TGSI_SEMANTIC_POSITION is used to indicate that
2509 fragment shader input contains the fragment's window position. The X
2510 component starts at zero and always increases from left to right.
2511 The Y component starts at zero and always increases but Y=0 may either
2512 indicate the top of the window or the bottom depending on the fragment
2513 coordinate origin convention (see TGSI_PROPERTY_FS_COORD_ORIGIN).
2514 The Z coordinate ranges from 0 to 1 to represent depth from the front
2515 to the back of the Z buffer. The W component contains the interpolated
2516 reciprocal of the vertex position W component (corresponding to gl_Fragcoord,
2517 but unlike d3d10 which interpolates the same 1/w but then gives back
2518 the reciprocal of the interpolated value).
2520 Fragment shaders may also declare an output register with
2521 TGSI_SEMANTIC_POSITION. Only the Z component is writable. This allows
2522 the fragment shader to change the fragment's Z position.
2529 For vertex shader outputs or fragment shader inputs/outputs, this
2530 label indicates that the resister contains an R,G,B,A color.
2532 Several shader inputs/outputs may contain colors so the semantic index
2533 is used to distinguish them. For example, color[0] may be the diffuse
2534 color while color[1] may be the specular color.
2536 This label is needed so that the flat/smooth shading can be applied
2537 to the right interpolants during rasterization.
2541 TGSI_SEMANTIC_BCOLOR
2542 """"""""""""""""""""
2544 Back-facing colors are only used for back-facing polygons, and are only valid
2545 in vertex shader outputs. After rasterization, all polygons are front-facing
2546 and COLOR and BCOLOR end up occupying the same slots in the fragment shader,
2547 so all BCOLORs effectively become regular COLORs in the fragment shader.
2553 Vertex shader inputs and outputs and fragment shader inputs may be
2554 labeled with TGSI_SEMANTIC_FOG to indicate that the register contains
2555 a fog coordinate. Typically, the fragment shader will use the fog coordinate
2556 to compute a fog blend factor which is used to blend the normal fragment color
2557 with a constant fog color. But fog coord really is just an ordinary vec4
2558 register like regular semantics.
2564 Vertex shader input and output registers may be labeled with
2565 TGIS_SEMANTIC_PSIZE to indicate that the register contains a point size
2566 in the form (S, 0, 0, 1). The point size controls the width or diameter
2567 of points for rasterization. This label cannot be used in fragment
2570 When using this semantic, be sure to set the appropriate state in the
2571 :ref:`rasterizer` first.
2574 TGSI_SEMANTIC_TEXCOORD
2575 """"""""""""""""""""""
2577 Only available if PIPE_CAP_TGSI_TEXCOORD is exposed !
2579 Vertex shader outputs and fragment shader inputs may be labeled with
2580 this semantic to make them replaceable by sprite coordinates via the
2581 sprite_coord_enable state in the :ref:`rasterizer`.
2582 The semantic index permitted with this semantic is limited to <= 7.
2584 If the driver does not support TEXCOORD, sprite coordinate replacement
2585 applies to inputs with the GENERIC semantic instead.
2587 The intended use case for this semantic is gl_TexCoord.
2590 TGSI_SEMANTIC_PCOORD
2591 """"""""""""""""""""
2593 Only available if PIPE_CAP_TGSI_TEXCOORD is exposed !
2595 Fragment shader inputs may be labeled with TGSI_SEMANTIC_PCOORD to indicate
2596 that the register contains sprite coordinates in the form (x, y, 0, 1), if
2597 the current primitive is a point and point sprites are enabled. Otherwise,
2598 the contents of the register are undefined.
2600 The intended use case for this semantic is gl_PointCoord.
2603 TGSI_SEMANTIC_GENERIC
2604 """""""""""""""""""""
2606 All vertex/fragment shader inputs/outputs not labeled with any other
2607 semantic label can be considered to be generic attributes. Typical
2608 uses of generic inputs/outputs are texcoords and user-defined values.
2611 TGSI_SEMANTIC_NORMAL
2612 """"""""""""""""""""
2614 Indicates that a vertex shader input is a normal vector. This is
2615 typically only used for legacy graphics APIs.
2621 This label applies to fragment shader inputs only and indicates that
2622 the register contains front/back-face information of the form (F, 0,
2623 0, 1). The first component will be positive when the fragment belongs
2624 to a front-facing polygon, and negative when the fragment belongs to a
2625 back-facing polygon.
2628 TGSI_SEMANTIC_EDGEFLAG
2629 """"""""""""""""""""""
2631 For vertex shaders, this sematic label indicates that an input or
2632 output is a boolean edge flag. The register layout is [F, x, x, x]
2633 where F is 0.0 or 1.0 and x = don't care. Normally, the vertex shader
2634 simply copies the edge flag input to the edgeflag output.
2636 Edge flags are used to control which lines or points are actually
2637 drawn when the polygon mode converts triangles/quads/polygons into
2641 TGSI_SEMANTIC_STENCIL
2642 """""""""""""""""""""
2644 For fragment shaders, this semantic label indicates that an output
2645 is a writable stencil reference value. Only the Y component is writable.
2646 This allows the fragment shader to change the fragments stencilref value.
2649 TGSI_SEMANTIC_VIEWPORT_INDEX
2650 """"""""""""""""""""""""""""
2652 For geometry shaders, this semantic label indicates that an output
2653 contains the index of the viewport (and scissor) to use.
2654 Only the X value is used.
2660 For geometry shaders, this semantic label indicates that an output
2661 contains the layer value to use for the color and depth/stencil surfaces.
2662 Only the X value is used. (Also known as rendertarget array index.)
2665 TGSI_SEMANTIC_CULLDIST
2666 """"""""""""""""""""""
2668 Used as distance to plane for performing application-defined culling
2669 of individual primitives against a plane. When components of vertex
2670 elements are given this label, these values are assumed to be a
2671 float32 signed distance to a plane. Primitives will be completely
2672 discarded if the plane distance for all of the vertices in the
2673 primitive are < 0. If a vertex has a cull distance of NaN, that
2674 vertex counts as "out" (as if its < 0);
2675 The limits on both clip and cull distances are bound
2676 by the PIPE_MAX_CLIP_OR_CULL_DISTANCE_COUNT define which defines
2677 the maximum number of components that can be used to hold the
2678 distances and by the PIPE_MAX_CLIP_OR_CULL_DISTANCE_ELEMENT_COUNT
2679 which specifies the maximum number of registers which can be
2680 annotated with those semantics.
2683 TGSI_SEMANTIC_CLIPDIST
2684 """"""""""""""""""""""
2686 When components of vertex elements are identified this way, these
2687 values are each assumed to be a float32 signed distance to a plane.
2688 Primitive setup only invokes rasterization on pixels for which
2689 the interpolated plane distances are >= 0. Multiple clip planes
2690 can be implemented simultaneously, by annotating multiple
2691 components of one or more vertex elements with the above specified
2692 semantic. The limits on both clip and cull distances are bound
2693 by the PIPE_MAX_CLIP_OR_CULL_DISTANCE_COUNT define which defines
2694 the maximum number of components that can be used to hold the
2695 distances and by the PIPE_MAX_CLIP_OR_CULL_DISTANCE_ELEMENT_COUNT
2696 which specifies the maximum number of registers which can be
2697 annotated with those semantics.
2699 TGSI_SEMANTIC_SAMPLEID
2700 """"""""""""""""""""""
2702 For fragment shaders, this semantic label indicates that a system value
2703 contains the current sample id (i.e. gl_SampleID). Only the X value is used.
2705 TGSI_SEMANTIC_SAMPLEPOS
2706 """""""""""""""""""""""
2708 For fragment shaders, this semantic label indicates that a system value
2709 contains the current sample's position (i.e. gl_SamplePosition). Only the X
2710 and Y values are used.
2712 TGSI_SEMANTIC_SAMPLEMASK
2713 """"""""""""""""""""""""
2715 For fragment shaders, this semantic label indicates that an output contains
2716 the sample mask used to disable further sample processing
2717 (i.e. gl_SampleMask). Only the X value is used, up to 32x MS.
2719 TGSI_SEMANTIC_INVOCATIONID
2720 """"""""""""""""""""""""""
2722 For geometry shaders, this semantic label indicates that a system value
2723 contains the current invocation id (i.e. gl_InvocationID). Only the X value is
2726 Declaration Interpolate
2727 ^^^^^^^^^^^^^^^^^^^^^^^
2729 This token is only valid for fragment shader INPUT declarations.
2731 The Interpolate field specifes the way input is being interpolated by
2732 the rasteriser and is one of TGSI_INTERPOLATE_*.
2734 The Location field specifies the location inside the pixel that the
2735 interpolation should be done at, one of ``TGSI_INTERPOLATE_LOC_*``. Note that
2736 when per-sample shading is enabled, the implementation may choose to
2737 interpolate at the sample irrespective of the Location field.
2739 The CylindricalWrap bitfield specifies which register components
2740 should be subject to cylindrical wrapping when interpolating by the
2741 rasteriser. If TGSI_CYLINDRICAL_WRAP_X is set to 1, the X component
2742 should be interpolated according to cylindrical wrapping rules.
2745 Declaration Sampler View
2746 ^^^^^^^^^^^^^^^^^^^^^^^^
2748 Follows Declaration token if file is TGSI_FILE_SAMPLER_VIEW.
2750 DCL SVIEW[#], resource, type(s)
2752 Declares a shader input sampler view and assigns it to a SVIEW[#]
2755 resource can be one of BUFFER, 1D, 2D, 3D, 1DArray and 2DArray.
2757 type must be 1 or 4 entries (if specifying on a per-component
2758 level) out of UNORM, SNORM, SINT, UINT and FLOAT.
2761 Declaration Resource
2762 ^^^^^^^^^^^^^^^^^^^^
2764 Follows Declaration token if file is TGSI_FILE_RESOURCE.
2766 DCL RES[#], resource [, WR] [, RAW]
2768 Declares a shader input resource and assigns it to a RES[#]
2771 resource can be one of BUFFER, 1D, 2D, 3D, CUBE, 1DArray and
2774 If the RAW keyword is not specified, the texture data will be
2775 subject to conversion, swizzling and scaling as required to yield
2776 the specified data type from the physical data format of the bound
2779 If the RAW keyword is specified, no channel conversion will be
2780 performed: the values read for each of the channels (X,Y,Z,W) will
2781 correspond to consecutive words in the same order and format
2782 they're found in memory. No element-to-address conversion will be
2783 performed either: the value of the provided X coordinate will be
2784 interpreted in byte units instead of texel units. The result of
2785 accessing a misaligned address is undefined.
2787 Usage of the STORE opcode is only allowed if the WR (writable) flag
2792 ^^^^^^^^^^^^^^^^^^^^^^^^
2794 Properties are general directives that apply to the whole TGSI program.
2799 Specifies the fragment shader TGSI_SEMANTIC_POSITION coordinate origin.
2800 The default value is UPPER_LEFT.
2802 If UPPER_LEFT, the position will be (0,0) at the upper left corner and
2803 increase downward and rightward.
2804 If LOWER_LEFT, the position will be (0,0) at the lower left corner and
2805 increase upward and rightward.
2807 OpenGL defaults to LOWER_LEFT, and is configurable with the
2808 GL_ARB_fragment_coord_conventions extension.
2810 DirectX 9/10 use UPPER_LEFT.
2812 FS_COORD_PIXEL_CENTER
2813 """""""""""""""""""""
2815 Specifies the fragment shader TGSI_SEMANTIC_POSITION pixel center convention.
2816 The default value is HALF_INTEGER.
2818 If HALF_INTEGER, the fractionary part of the position will be 0.5
2819 If INTEGER, the fractionary part of the position will be 0.0
2821 Note that this does not affect the set of fragments generated by
2822 rasterization, which is instead controlled by half_pixel_center in the
2825 OpenGL defaults to HALF_INTEGER, and is configurable with the
2826 GL_ARB_fragment_coord_conventions extension.
2828 DirectX 9 uses INTEGER.
2829 DirectX 10 uses HALF_INTEGER.
2831 FS_COLOR0_WRITES_ALL_CBUFS
2832 """"""""""""""""""""""""""
2833 Specifies that writes to the fragment shader color 0 are replicated to all
2834 bound cbufs. This facilitates OpenGL's fragColor output vs fragData[0] where
2835 fragData is directed to a single color buffer, but fragColor is broadcast.
2838 """"""""""""""""""""""""""
2839 If this property is set on the program bound to the shader stage before the
2840 fragment shader, user clip planes should have no effect (be disabled) even if
2841 that shader does not write to any clip distance outputs and the rasterizer's
2842 clip_plane_enable is non-zero.
2843 This property is only supported by drivers that also support shader clip
2845 This is useful for APIs that don't have UCPs and where clip distances written
2846 by a shader cannot be disabled.
2851 Specifies the number of times a geometry shader should be executed for each
2852 input primitive. Each invocation will have a different
2853 TGSI_SEMANTIC_INVOCATIONID system value set. If not specified, assumed to
2856 VS_WINDOW_SPACE_POSITION
2857 """"""""""""""""""""""""""
2858 If this property is set on the vertex shader, the TGSI_SEMANTIC_POSITION output
2859 is assumed to contain window space coordinates.
2860 Division of X,Y,Z by W and the viewport transformation are disabled, and 1/W is
2861 directly taken from the 4-th component of the shader output.
2862 Naturally, clipping is not performed on window coordinates either.
2863 The effect of this property is undefined if a geometry or tessellation shader
2866 Texture Sampling and Texture Formats
2867 ------------------------------------
2869 This table shows how texture image components are returned as (x,y,z,w) tuples
2870 by TGSI texture instructions, such as :opcode:`TEX`, :opcode:`TXD`, and
2871 :opcode:`TXP`. For reference, OpenGL and Direct3D conventions are shown as
2874 +--------------------+--------------+--------------------+--------------+
2875 | Texture Components | Gallium | OpenGL | Direct3D 9 |
2876 +====================+==============+====================+==============+
2877 | R | (r, 0, 0, 1) | (r, 0, 0, 1) | (r, 1, 1, 1) |
2878 +--------------------+--------------+--------------------+--------------+
2879 | RG | (r, g, 0, 1) | (r, g, 0, 1) | (r, g, 1, 1) |
2880 +--------------------+--------------+--------------------+--------------+
2881 | RGB | (r, g, b, 1) | (r, g, b, 1) | (r, g, b, 1) |
2882 +--------------------+--------------+--------------------+--------------+
2883 | RGBA | (r, g, b, a) | (r, g, b, a) | (r, g, b, a) |
2884 +--------------------+--------------+--------------------+--------------+
2885 | A | (0, 0, 0, a) | (0, 0, 0, a) | (0, 0, 0, a) |
2886 +--------------------+--------------+--------------------+--------------+
2887 | L | (l, l, l, 1) | (l, l, l, 1) | (l, l, l, 1) |
2888 +--------------------+--------------+--------------------+--------------+
2889 | LA | (l, l, l, a) | (l, l, l, a) | (l, l, l, a) |
2890 +--------------------+--------------+--------------------+--------------+
2891 | I | (i, i, i, i) | (i, i, i, i) | N/A |
2892 +--------------------+--------------+--------------------+--------------+
2893 | UV | XXX TBD | (0, 0, 0, 1) | (u, v, 1, 1) |
2894 | | | [#envmap-bumpmap]_ | |
2895 +--------------------+--------------+--------------------+--------------+
2896 | Z | XXX TBD | (z, z, z, 1) | (0, z, 0, 1) |
2897 | | | [#depth-tex-mode]_ | |
2898 +--------------------+--------------+--------------------+--------------+
2899 | S | (s, s, s, s) | unknown | unknown |
2900 +--------------------+--------------+--------------------+--------------+
2902 .. [#envmap-bumpmap] http://www.opengl.org/registry/specs/ATI/envmap_bumpmap.txt
2903 .. [#depth-tex-mode] the default is (z, z, z, 1) but may also be (0, 0, 0, z)
2904 or (z, z, z, z) depending on the value of GL_DEPTH_TEXTURE_MODE.