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 and precise modifier
32 For arithmetic instruction having a precise modifier certain optimizations
33 which may alter the result are disallowed. Example: *add(mul(a,b),c)* can't be
34 optimized to TGSI_OPCODE_MAD, because some hardware only supports the fused
37 For inputs which have a floating point type, both absolute value and
38 negation modifiers are supported (with absolute value being applied
39 first). The only source of TGSI_OPCODE_MOV and the second and third
40 sources of TGSI_OPCODE_UCMP are considered to have float type for
43 For inputs which have signed or unsigned type only the negate modifier is
50 ^^^^^^^^^^^^^^^^^^^^^^^^^
52 These opcodes are guaranteed to be available regardless of the driver being
55 .. opcode:: ARL - Address Register Load
59 dst.x = (int) \lfloor src.x\rfloor
61 dst.y = (int) \lfloor src.y\rfloor
63 dst.z = (int) \lfloor src.z\rfloor
65 dst.w = (int) \lfloor src.w\rfloor
68 .. opcode:: MOV - Move
81 .. opcode:: LIT - Light Coefficients
86 dst.y &= max(src.x, 0) \\
87 dst.z &= (src.x > 0) ? max(src.y, 0)^{clamp(src.w, -128, 128))} : 0 \\
91 .. opcode:: RCP - Reciprocal
93 This instruction replicates its result.
100 .. opcode:: RSQ - Reciprocal Square Root
102 This instruction replicates its result. The results are undefined for src <= 0.
106 dst = \frac{1}{\sqrt{src.x}}
109 .. opcode:: SQRT - Square Root
111 This instruction replicates its result. The results are undefined for src < 0.
118 .. opcode:: EXP - Approximate Exponential Base 2
122 dst.x &= 2^{\lfloor src.x\rfloor} \\
123 dst.y &= src.x - \lfloor src.x\rfloor \\
124 dst.z &= 2^{src.x} \\
128 .. opcode:: LOG - Approximate Logarithm Base 2
132 dst.x &= \lfloor\log_2{|src.x|}\rfloor \\
133 dst.y &= \frac{|src.x|}{2^{\lfloor\log_2{|src.x|}\rfloor}} \\
134 dst.z &= \log_2{|src.x|} \\
138 .. opcode:: MUL - Multiply
142 dst.x = src0.x \times src1.x
144 dst.y = src0.y \times src1.y
146 dst.z = src0.z \times src1.z
148 dst.w = src0.w \times src1.w
151 .. opcode:: ADD - Add
155 dst.x = src0.x + src1.x
157 dst.y = src0.y + src1.y
159 dst.z = src0.z + src1.z
161 dst.w = src0.w + src1.w
164 .. opcode:: DP3 - 3-component Dot Product
166 This instruction replicates its result.
170 dst = src0.x \times src1.x + src0.y \times src1.y + src0.z \times src1.z
173 .. opcode:: DP4 - 4-component Dot Product
175 This instruction replicates its result.
179 dst = src0.x \times src1.x + src0.y \times src1.y + src0.z \times src1.z + src0.w \times src1.w
182 .. opcode:: DST - Distance Vector
187 dst.y &= src0.y \times src1.y\\
192 .. opcode:: MIN - Minimum
196 dst.x = min(src0.x, src1.x)
198 dst.y = min(src0.y, src1.y)
200 dst.z = min(src0.z, src1.z)
202 dst.w = min(src0.w, src1.w)
205 .. opcode:: MAX - Maximum
209 dst.x = max(src0.x, src1.x)
211 dst.y = max(src0.y, src1.y)
213 dst.z = max(src0.z, src1.z)
215 dst.w = max(src0.w, src1.w)
218 .. opcode:: SLT - Set On Less Than
222 dst.x = (src0.x < src1.x) ? 1.0F : 0.0F
224 dst.y = (src0.y < src1.y) ? 1.0F : 0.0F
226 dst.z = (src0.z < src1.z) ? 1.0F : 0.0F
228 dst.w = (src0.w < src1.w) ? 1.0F : 0.0F
231 .. opcode:: SGE - Set On Greater Equal Than
235 dst.x = (src0.x >= src1.x) ? 1.0F : 0.0F
237 dst.y = (src0.y >= src1.y) ? 1.0F : 0.0F
239 dst.z = (src0.z >= src1.z) ? 1.0F : 0.0F
241 dst.w = (src0.w >= src1.w) ? 1.0F : 0.0F
244 .. opcode:: MAD - Multiply And Add
246 Perform a * b + c. The implementation is free to decide whether there is an
247 intermediate rounding step or not.
251 dst.x = src0.x \times src1.x + src2.x
253 dst.y = src0.y \times src1.y + src2.y
255 dst.z = src0.z \times src1.z + src2.z
257 dst.w = src0.w \times src1.w + src2.w
260 .. opcode:: LRP - Linear Interpolate
264 dst.x = src0.x \times src1.x + (1 - src0.x) \times src2.x
266 dst.y = src0.y \times src1.y + (1 - src0.y) \times src2.y
268 dst.z = src0.z \times src1.z + (1 - src0.z) \times src2.z
270 dst.w = src0.w \times src1.w + (1 - src0.w) \times src2.w
273 .. opcode:: FMA - Fused Multiply-Add
275 Perform a * b + c with no intermediate rounding step.
279 dst.x = src0.x \times src1.x + src2.x
281 dst.y = src0.y \times src1.y + src2.y
283 dst.z = src0.z \times src1.z + src2.z
285 dst.w = src0.w \times src1.w + src2.w
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:: FLR - Floor
305 dst.x = \lfloor src.x\rfloor
307 dst.y = \lfloor src.y\rfloor
309 dst.z = \lfloor src.z\rfloor
311 dst.w = \lfloor src.w\rfloor
314 .. opcode:: ROUND - Round
327 .. opcode:: EX2 - Exponential Base 2
329 This instruction replicates its result.
336 .. opcode:: LG2 - Logarithm Base 2
338 This instruction replicates its result.
345 .. opcode:: POW - Power
347 This instruction replicates its result.
351 dst = src0.x^{src1.x}
354 .. opcode:: COS - Cosine
356 This instruction replicates its result.
363 .. opcode:: DDX, DDX_FINE - Derivative Relative To X
365 The fine variant is only used when ``PIPE_CAP_TGSI_FS_FINE_DERIVATIVE`` is
366 advertised. When it is, the fine version guarantees one derivative per row
367 while DDX is allowed to be the same for the entire 2x2 quad.
371 dst.x = partialx(src.x)
373 dst.y = partialx(src.y)
375 dst.z = partialx(src.z)
377 dst.w = partialx(src.w)
380 .. opcode:: DDY, DDY_FINE - Derivative Relative To Y
382 The fine variant is only used when ``PIPE_CAP_TGSI_FS_FINE_DERIVATIVE`` is
383 advertised. When it is, the fine version guarantees one derivative per column
384 while DDY is allowed to be the same for the entire 2x2 quad.
388 dst.x = partialy(src.x)
390 dst.y = partialy(src.y)
392 dst.z = partialy(src.z)
394 dst.w = partialy(src.w)
397 .. opcode:: PK2H - Pack Two 16-bit Floats
399 This instruction replicates its result.
403 dst = f32\_to\_f16(src.x) | f32\_to\_f16(src.y) << 16
406 .. opcode:: PK2US - Pack Two Unsigned 16-bit Scalars
411 .. opcode:: PK4B - Pack Four Signed 8-bit Scalars
416 .. opcode:: PK4UB - Pack Four Unsigned 8-bit Scalars
421 .. opcode:: SEQ - Set On Equal
425 dst.x = (src0.x == src1.x) ? 1.0F : 0.0F
427 dst.y = (src0.y == src1.y) ? 1.0F : 0.0F
429 dst.z = (src0.z == src1.z) ? 1.0F : 0.0F
431 dst.w = (src0.w == src1.w) ? 1.0F : 0.0F
434 .. opcode:: SGT - Set On Greater Than
438 dst.x = (src0.x > src1.x) ? 1.0F : 0.0F
440 dst.y = (src0.y > src1.y) ? 1.0F : 0.0F
442 dst.z = (src0.z > src1.z) ? 1.0F : 0.0F
444 dst.w = (src0.w > src1.w) ? 1.0F : 0.0F
447 .. opcode:: SIN - Sine
449 This instruction replicates its result.
456 .. opcode:: SLE - Set On Less Equal Than
460 dst.x = (src0.x <= src1.x) ? 1.0F : 0.0F
462 dst.y = (src0.y <= src1.y) ? 1.0F : 0.0F
464 dst.z = (src0.z <= src1.z) ? 1.0F : 0.0F
466 dst.w = (src0.w <= src1.w) ? 1.0F : 0.0F
469 .. opcode:: SNE - Set On Not Equal
473 dst.x = (src0.x != src1.x) ? 1.0F : 0.0F
475 dst.y = (src0.y != src1.y) ? 1.0F : 0.0F
477 dst.z = (src0.z != src1.z) ? 1.0F : 0.0F
479 dst.w = (src0.w != src1.w) ? 1.0F : 0.0F
482 .. opcode:: TEX - Texture Lookup
484 for array textures src0.y contains the slice for 1D,
485 and src0.z contain the slice for 2D.
487 for shadow textures with no arrays (and not cube map),
488 src0.z contains the reference value.
490 for shadow textures with arrays, src0.z contains
491 the reference value for 1D arrays, and src0.w contains
492 the reference value for 2D arrays and cube maps.
494 for cube map array shadow textures, the reference value
495 cannot be passed in src0.w, and TEX2 must be used instead.
501 shadow_ref = src0.z or src0.w (optional)
505 dst = texture\_sample(unit, coord, shadow_ref)
508 .. opcode:: TEX2 - Texture Lookup (for shadow cube map arrays only)
510 this is the same as TEX, but uses another reg to encode the
521 dst = texture\_sample(unit, coord, shadow_ref)
526 .. opcode:: TXD - Texture Lookup with Derivatives
538 dst = texture\_sample\_deriv(unit, coord, ddx, ddy)
541 .. opcode:: TXP - Projective Texture Lookup
545 coord.x = src0.x / src0.w
547 coord.y = src0.y / src0.w
549 coord.z = src0.z / src0.w
555 dst = texture\_sample(unit, coord)
558 .. opcode:: UP2H - Unpack Two 16-Bit Floats
562 dst.x = f16\_to\_f32(src0.x \& 0xffff)
564 dst.y = f16\_to\_f32(src0.x >> 16)
566 dst.z = f16\_to\_f32(src0.x \& 0xffff)
568 dst.w = f16\_to\_f32(src0.x >> 16)
572 Considered for removal.
574 .. opcode:: UP2US - Unpack Two Unsigned 16-Bit Scalars
580 Considered for removal.
582 .. opcode:: UP4B - Unpack Four Signed 8-Bit Values
588 Considered for removal.
590 .. opcode:: UP4UB - Unpack Four Unsigned 8-Bit Scalars
596 Considered for removal.
599 .. opcode:: ARR - Address Register Load With Round
603 dst.x = (int) round(src.x)
605 dst.y = (int) round(src.y)
607 dst.z = (int) round(src.z)
609 dst.w = (int) round(src.w)
612 .. opcode:: SSG - Set Sign
616 dst.x = (src.x > 0) ? 1 : (src.x < 0) ? -1 : 0
618 dst.y = (src.y > 0) ? 1 : (src.y < 0) ? -1 : 0
620 dst.z = (src.z > 0) ? 1 : (src.z < 0) ? -1 : 0
622 dst.w = (src.w > 0) ? 1 : (src.w < 0) ? -1 : 0
625 .. opcode:: CMP - Compare
629 dst.x = (src0.x < 0) ? src1.x : src2.x
631 dst.y = (src0.y < 0) ? src1.y : src2.y
633 dst.z = (src0.z < 0) ? src1.z : src2.z
635 dst.w = (src0.w < 0) ? src1.w : src2.w
638 .. opcode:: KILL_IF - Conditional Discard
640 Conditional discard. Allowed in fragment shaders only.
644 if (src.x < 0 || src.y < 0 || src.z < 0 || src.w < 0)
649 .. opcode:: KILL - Discard
651 Unconditional discard. Allowed in fragment shaders only.
654 .. opcode:: SCS - Sine Cosine
667 .. opcode:: TXB - Texture Lookup With Bias
669 for cube map array textures and shadow cube maps, the bias value
670 cannot be passed in src0.w, and TXB2 must be used instead.
672 if the target is a shadow texture, the reference value is always
673 in src.z (this prevents shadow 3d and shadow 2d arrays from
674 using this instruction, but this is not needed).
690 dst = texture\_sample(unit, coord, bias)
693 .. opcode:: TXB2 - Texture Lookup With Bias (some cube maps only)
695 this is the same as TXB, but uses another reg to encode the
696 lod bias value for cube map arrays and shadow cube maps.
697 Presumably shadow 2d arrays and shadow 3d targets could use
698 this encoding too, but this is not legal.
700 shadow cube map arrays are neither possible nor required.
710 dst = texture\_sample(unit, coord, bias)
713 .. opcode:: DIV - Divide
717 dst.x = \frac{src0.x}{src1.x}
719 dst.y = \frac{src0.y}{src1.y}
721 dst.z = \frac{src0.z}{src1.z}
723 dst.w = \frac{src0.w}{src1.w}
726 .. opcode:: DP2 - 2-component Dot Product
728 This instruction replicates its result.
732 dst = src0.x \times src1.x + src0.y \times src1.y
735 .. opcode:: TEX_LZ - Texture Lookup With LOD = 0
737 This is the same as TXL with LOD = 0. Like every texture opcode, it obeys
738 pipe_sampler_view::u.tex.first_level and pipe_sampler_state::min_lod.
739 There is no way to override those two in shaders.
755 dst = texture\_sample(unit, coord, lod)
758 .. opcode:: TXL - Texture Lookup With explicit LOD
760 for cube map array textures, the explicit lod value
761 cannot be passed in src0.w, and TXL2 must be used instead.
763 if the target is a shadow texture, the reference value is always
764 in src.z (this prevents shadow 3d / 2d array / cube targets from
765 using this instruction, but this is not needed).
781 dst = texture\_sample(unit, coord, lod)
784 .. opcode:: TXL2 - Texture Lookup With explicit LOD (for cube map arrays only)
786 this is the same as TXL, but uses another reg to encode the
788 Presumably shadow 3d / 2d array / cube targets could use
789 this encoding too, but this is not legal.
791 shadow cube map arrays are neither possible nor required.
801 dst = texture\_sample(unit, coord, lod)
805 ^^^^^^^^^^^^^^^^^^^^^^^^
807 These opcodes are primarily provided for special-use computational shaders.
808 Support for these opcodes indicated by a special pipe capability bit (TBD).
810 XXX doesn't look like most of the opcodes really belong here.
812 .. opcode:: CEIL - Ceiling
816 dst.x = \lceil src.x\rceil
818 dst.y = \lceil src.y\rceil
820 dst.z = \lceil src.z\rceil
822 dst.w = \lceil src.w\rceil
825 .. opcode:: TRUNC - Truncate
838 .. opcode:: MOD - Modulus
842 dst.x = src0.x \bmod src1.x
844 dst.y = src0.y \bmod src1.y
846 dst.z = src0.z \bmod src1.z
848 dst.w = src0.w \bmod src1.w
851 .. opcode:: UARL - Integer Address Register Load
853 Moves the contents of the source register, assumed to be an integer, into the
854 destination register, which is assumed to be an address (ADDR) register.
857 .. opcode:: TXF - Texel Fetch
859 As per NV_gpu_shader4, extract a single texel from a specified texture
860 image or PIPE_BUFFER resource. The source sampler may not be a CUBE or
862 four-component signed integer vector used to identify the single texel
863 accessed. 3 components + level. If the texture is multisampled, then
864 the fourth component indicates the sample, not the mipmap level.
865 Just like texture instructions, an optional
866 offset vector is provided, which is subject to various driver restrictions
867 (regarding range, source of offsets). This instruction ignores the sampler
870 TXF(uint_vec coord, int_vec offset).
873 .. opcode:: TXQ - Texture Size Query
875 As per NV_gpu_program4, retrieve the dimensions of the texture depending on
876 the target. For 1D (width), 2D/RECT/CUBE (width, height), 3D (width, height,
877 depth), 1D array (width, layers), 2D array (width, height, layers).
878 Also return the number of accessible levels (last_level - first_level + 1)
881 For components which don't return a resource dimension, their value
888 dst.x = texture\_width(unit, lod)
890 dst.y = texture\_height(unit, lod)
892 dst.z = texture\_depth(unit, lod)
894 dst.w = texture\_levels(unit)
897 .. opcode:: TXQS - Texture Samples Query
899 This retrieves the number of samples in the texture, and stores it
900 into the x component as an unsigned integer. The other components are
901 undefined. If the texture is not multisampled, this function returns
902 (1, undef, undef, undef).
906 dst.x = texture\_samples(unit)
909 .. opcode:: TG4 - Texture Gather
911 As per ARB_texture_gather, gathers the four texels to be used in a bi-linear
912 filtering operation and packs them into a single register. Only works with
913 2D, 2D array, cubemaps, and cubemaps arrays. For 2D textures, only the
914 addressing modes of the sampler and the top level of any mip pyramid are
915 used. Set W to zero. It behaves like the TEX instruction, but a filtered
916 sample is not generated. The four samples that contribute to filtering are
917 placed into xyzw in clockwise order, starting with the (u,v) texture
918 coordinate delta at the following locations (-, +), (+, +), (+, -), (-, -),
919 where the magnitude of the deltas are half a texel.
921 PIPE_CAP_TEXTURE_SM5 enhances this instruction to support shadow per-sample
922 depth compares, single component selection, and a non-constant offset. It
923 doesn't allow support for the GL independent offset to get i0,j0. This would
924 require another CAP is hw can do it natively. For now we lower that before
933 dst = texture\_gather4 (unit, coord, component)
935 (with SM5 - cube array shadow)
943 dst = texture\_gather (uint, coord, compare)
945 .. opcode:: LODQ - level of detail query
947 Compute the LOD information that the texture pipe would use to access the
948 texture. The Y component contains the computed LOD lambda_prime. The X
949 component contains the LOD that will be accessed, based on min/max lod's
956 dst.xy = lodq(uint, coord);
958 .. opcode:: CLOCK - retrieve the current shader time
960 Invoking this instruction multiple times in the same shader should
961 cause monotonically increasing values to be returned. The values
962 are implicitly 64-bit, so if fewer than 64 bits of precision are
963 available, to provide expected wraparound semantics, the value
964 should be shifted up so that the most significant bit of the time
965 is the most significant bit of the 64-bit value.
973 ^^^^^^^^^^^^^^^^^^^^^^^^
974 These opcodes are used for integer operations.
975 Support for these opcodes indicated by PIPE_SHADER_CAP_INTEGERS (all of them?)
978 .. opcode:: I2F - Signed Integer To Float
980 Rounding is unspecified (round to nearest even suggested).
984 dst.x = (float) src.x
986 dst.y = (float) src.y
988 dst.z = (float) src.z
990 dst.w = (float) src.w
993 .. opcode:: U2F - Unsigned Integer To Float
995 Rounding is unspecified (round to nearest even suggested).
999 dst.x = (float) src.x
1001 dst.y = (float) src.y
1003 dst.z = (float) src.z
1005 dst.w = (float) src.w
1008 .. opcode:: F2I - Float to Signed Integer
1010 Rounding is towards zero (truncate).
1011 Values outside signed range (including NaNs) produce undefined results.
1024 .. opcode:: F2U - Float to Unsigned Integer
1026 Rounding is towards zero (truncate).
1027 Values outside unsigned range (including NaNs) produce undefined results.
1031 dst.x = (unsigned) src.x
1033 dst.y = (unsigned) src.y
1035 dst.z = (unsigned) src.z
1037 dst.w = (unsigned) src.w
1040 .. opcode:: UADD - Integer Add
1042 This instruction works the same for signed and unsigned integers.
1043 The low 32bit of the result is returned.
1047 dst.x = src0.x + src1.x
1049 dst.y = src0.y + src1.y
1051 dst.z = src0.z + src1.z
1053 dst.w = src0.w + src1.w
1056 .. opcode:: UMAD - Integer Multiply And Add
1058 This instruction works the same for signed and unsigned integers.
1059 The multiplication returns the low 32bit (as does the result itself).
1063 dst.x = src0.x \times src1.x + src2.x
1065 dst.y = src0.y \times src1.y + src2.y
1067 dst.z = src0.z \times src1.z + src2.z
1069 dst.w = src0.w \times src1.w + src2.w
1072 .. opcode:: UMUL - Integer Multiply
1074 This instruction works the same for signed and unsigned integers.
1075 The low 32bit of the result is returned.
1079 dst.x = src0.x \times src1.x
1081 dst.y = src0.y \times src1.y
1083 dst.z = src0.z \times src1.z
1085 dst.w = src0.w \times src1.w
1088 .. opcode:: IMUL_HI - Signed Integer Multiply High Bits
1090 The high 32bits of the multiplication of 2 signed integers are returned.
1094 dst.x = (src0.x \times src1.x) >> 32
1096 dst.y = (src0.y \times src1.y) >> 32
1098 dst.z = (src0.z \times src1.z) >> 32
1100 dst.w = (src0.w \times src1.w) >> 32
1103 .. opcode:: UMUL_HI - Unsigned Integer Multiply High Bits
1105 The high 32bits of the multiplication of 2 unsigned integers are returned.
1109 dst.x = (src0.x \times src1.x) >> 32
1111 dst.y = (src0.y \times src1.y) >> 32
1113 dst.z = (src0.z \times src1.z) >> 32
1115 dst.w = (src0.w \times src1.w) >> 32
1118 .. opcode:: IDIV - Signed Integer Division
1120 TBD: behavior for division by zero.
1124 dst.x = \frac{src0.x}{src1.x}
1126 dst.y = \frac{src0.y}{src1.y}
1128 dst.z = \frac{src0.z}{src1.z}
1130 dst.w = \frac{src0.w}{src1.w}
1133 .. opcode:: UDIV - Unsigned Integer Division
1135 For division by zero, 0xffffffff is returned.
1139 dst.x = \frac{src0.x}{src1.x}
1141 dst.y = \frac{src0.y}{src1.y}
1143 dst.z = \frac{src0.z}{src1.z}
1145 dst.w = \frac{src0.w}{src1.w}
1148 .. opcode:: UMOD - Unsigned Integer Remainder
1150 If second arg is zero, 0xffffffff is returned.
1154 dst.x = src0.x \bmod src1.x
1156 dst.y = src0.y \bmod src1.y
1158 dst.z = src0.z \bmod src1.z
1160 dst.w = src0.w \bmod src1.w
1163 .. opcode:: NOT - Bitwise Not
1176 .. opcode:: AND - Bitwise And
1180 dst.x = src0.x \& src1.x
1182 dst.y = src0.y \& src1.y
1184 dst.z = src0.z \& src1.z
1186 dst.w = src0.w \& src1.w
1189 .. opcode:: OR - Bitwise Or
1193 dst.x = src0.x | src1.x
1195 dst.y = src0.y | src1.y
1197 dst.z = src0.z | src1.z
1199 dst.w = src0.w | src1.w
1202 .. opcode:: XOR - Bitwise Xor
1206 dst.x = src0.x \oplus src1.x
1208 dst.y = src0.y \oplus src1.y
1210 dst.z = src0.z \oplus src1.z
1212 dst.w = src0.w \oplus src1.w
1215 .. opcode:: IMAX - Maximum of Signed Integers
1219 dst.x = max(src0.x, src1.x)
1221 dst.y = max(src0.y, src1.y)
1223 dst.z = max(src0.z, src1.z)
1225 dst.w = max(src0.w, src1.w)
1228 .. opcode:: UMAX - Maximum of Unsigned Integers
1232 dst.x = max(src0.x, src1.x)
1234 dst.y = max(src0.y, src1.y)
1236 dst.z = max(src0.z, src1.z)
1238 dst.w = max(src0.w, src1.w)
1241 .. opcode:: IMIN - Minimum of Signed Integers
1245 dst.x = min(src0.x, src1.x)
1247 dst.y = min(src0.y, src1.y)
1249 dst.z = min(src0.z, src1.z)
1251 dst.w = min(src0.w, src1.w)
1254 .. opcode:: UMIN - Minimum of Unsigned Integers
1258 dst.x = min(src0.x, src1.x)
1260 dst.y = min(src0.y, src1.y)
1262 dst.z = min(src0.z, src1.z)
1264 dst.w = min(src0.w, src1.w)
1267 .. opcode:: SHL - Shift Left
1269 The shift count is masked with 0x1f before the shift is applied.
1273 dst.x = src0.x << (0x1f \& src1.x)
1275 dst.y = src0.y << (0x1f \& src1.y)
1277 dst.z = src0.z << (0x1f \& src1.z)
1279 dst.w = src0.w << (0x1f \& src1.w)
1282 .. opcode:: ISHR - Arithmetic Shift Right (of Signed Integer)
1284 The shift count is masked with 0x1f before the shift is applied.
1288 dst.x = src0.x >> (0x1f \& src1.x)
1290 dst.y = src0.y >> (0x1f \& src1.y)
1292 dst.z = src0.z >> (0x1f \& src1.z)
1294 dst.w = src0.w >> (0x1f \& src1.w)
1297 .. opcode:: USHR - Logical Shift Right
1299 The shift count is masked with 0x1f before the shift is applied.
1303 dst.x = src0.x >> (unsigned) (0x1f \& src1.x)
1305 dst.y = src0.y >> (unsigned) (0x1f \& src1.y)
1307 dst.z = src0.z >> (unsigned) (0x1f \& src1.z)
1309 dst.w = src0.w >> (unsigned) (0x1f \& src1.w)
1312 .. opcode:: UCMP - Integer Conditional Move
1316 dst.x = src0.x ? src1.x : src2.x
1318 dst.y = src0.y ? src1.y : src2.y
1320 dst.z = src0.z ? src1.z : src2.z
1322 dst.w = src0.w ? src1.w : src2.w
1326 .. opcode:: ISSG - Integer Set Sign
1330 dst.x = (src0.x < 0) ? -1 : (src0.x > 0) ? 1 : 0
1332 dst.y = (src0.y < 0) ? -1 : (src0.y > 0) ? 1 : 0
1334 dst.z = (src0.z < 0) ? -1 : (src0.z > 0) ? 1 : 0
1336 dst.w = (src0.w < 0) ? -1 : (src0.w > 0) ? 1 : 0
1340 .. opcode:: FSLT - Float Set On Less Than (ordered)
1342 Same comparison as SLT but returns integer instead of 1.0/0.0 float
1346 dst.x = (src0.x < src1.x) ? \sim 0 : 0
1348 dst.y = (src0.y < src1.y) ? \sim 0 : 0
1350 dst.z = (src0.z < src1.z) ? \sim 0 : 0
1352 dst.w = (src0.w < src1.w) ? \sim 0 : 0
1355 .. opcode:: ISLT - Signed Integer Set On Less Than
1359 dst.x = (src0.x < src1.x) ? \sim 0 : 0
1361 dst.y = (src0.y < src1.y) ? \sim 0 : 0
1363 dst.z = (src0.z < src1.z) ? \sim 0 : 0
1365 dst.w = (src0.w < src1.w) ? \sim 0 : 0
1368 .. opcode:: USLT - Unsigned Integer Set On Less Than
1372 dst.x = (src0.x < src1.x) ? \sim 0 : 0
1374 dst.y = (src0.y < src1.y) ? \sim 0 : 0
1376 dst.z = (src0.z < src1.z) ? \sim 0 : 0
1378 dst.w = (src0.w < src1.w) ? \sim 0 : 0
1381 .. opcode:: FSGE - Float Set On Greater Equal Than (ordered)
1383 Same comparison as SGE but returns integer instead of 1.0/0.0 float
1387 dst.x = (src0.x >= src1.x) ? \sim 0 : 0
1389 dst.y = (src0.y >= src1.y) ? \sim 0 : 0
1391 dst.z = (src0.z >= src1.z) ? \sim 0 : 0
1393 dst.w = (src0.w >= src1.w) ? \sim 0 : 0
1396 .. opcode:: ISGE - Signed Integer Set On Greater Equal Than
1400 dst.x = (src0.x >= src1.x) ? \sim 0 : 0
1402 dst.y = (src0.y >= src1.y) ? \sim 0 : 0
1404 dst.z = (src0.z >= src1.z) ? \sim 0 : 0
1406 dst.w = (src0.w >= src1.w) ? \sim 0 : 0
1409 .. opcode:: USGE - Unsigned Integer Set On Greater Equal Than
1413 dst.x = (src0.x >= src1.x) ? \sim 0 : 0
1415 dst.y = (src0.y >= src1.y) ? \sim 0 : 0
1417 dst.z = (src0.z >= src1.z) ? \sim 0 : 0
1419 dst.w = (src0.w >= src1.w) ? \sim 0 : 0
1422 .. opcode:: FSEQ - Float Set On Equal (ordered)
1424 Same comparison as SEQ but returns integer instead of 1.0/0.0 float
1428 dst.x = (src0.x == src1.x) ? \sim 0 : 0
1430 dst.y = (src0.y == src1.y) ? \sim 0 : 0
1432 dst.z = (src0.z == src1.z) ? \sim 0 : 0
1434 dst.w = (src0.w == src1.w) ? \sim 0 : 0
1437 .. opcode:: USEQ - Integer Set On Equal
1441 dst.x = (src0.x == src1.x) ? \sim 0 : 0
1443 dst.y = (src0.y == src1.y) ? \sim 0 : 0
1445 dst.z = (src0.z == src1.z) ? \sim 0 : 0
1447 dst.w = (src0.w == src1.w) ? \sim 0 : 0
1450 .. opcode:: FSNE - Float Set On Not Equal (unordered)
1452 Same comparison as SNE but returns integer instead of 1.0/0.0 float
1456 dst.x = (src0.x != src1.x) ? \sim 0 : 0
1458 dst.y = (src0.y != src1.y) ? \sim 0 : 0
1460 dst.z = (src0.z != src1.z) ? \sim 0 : 0
1462 dst.w = (src0.w != src1.w) ? \sim 0 : 0
1465 .. opcode:: USNE - Integer Set On Not Equal
1469 dst.x = (src0.x != src1.x) ? \sim 0 : 0
1471 dst.y = (src0.y != src1.y) ? \sim 0 : 0
1473 dst.z = (src0.z != src1.z) ? \sim 0 : 0
1475 dst.w = (src0.w != src1.w) ? \sim 0 : 0
1478 .. opcode:: INEG - Integer Negate
1493 .. opcode:: IABS - Integer Absolute Value
1507 These opcodes are used for bit-level manipulation of integers.
1509 .. opcode:: IBFE - Signed Bitfield Extract
1511 Like GLSL bitfieldExtract. Extracts a set of bits from the input, and
1512 sign-extends them if the high bit of the extracted window is set.
1516 def ibfe(value, offset, bits):
1517 if offset < 0 or bits < 0 or offset + bits > 32:
1519 if bits == 0: return 0
1520 # Note: >> sign-extends
1521 return (value << (32 - offset - bits)) >> (32 - bits)
1523 .. opcode:: UBFE - Unsigned Bitfield Extract
1525 Like GLSL bitfieldExtract. Extracts a set of bits from the input, without
1530 def ubfe(value, offset, bits):
1531 if offset < 0 or bits < 0 or offset + bits > 32:
1533 if bits == 0: return 0
1534 # Note: >> does not sign-extend
1535 return (value << (32 - offset - bits)) >> (32 - bits)
1537 .. opcode:: BFI - Bitfield Insert
1539 Like GLSL bitfieldInsert. Replaces a bit region of 'base' with the low bits
1544 def bfi(base, insert, offset, bits):
1545 if offset < 0 or bits < 0 or offset + bits > 32:
1547 # << defined such that mask == ~0 when bits == 32, offset == 0
1548 mask = ((1 << bits) - 1) << offset
1549 return ((insert << offset) & mask) | (base & ~mask)
1551 .. opcode:: BREV - Bitfield Reverse
1553 See SM5 instruction BFREV. Reverses the bits of the argument.
1555 .. opcode:: POPC - Population Count
1557 See SM5 instruction COUNTBITS. Counts the number of set bits in the argument.
1559 .. opcode:: LSB - Index of lowest set bit
1561 See SM5 instruction FIRSTBIT_LO. Computes the 0-based index of the first set
1562 bit of the argument. Returns -1 if none are set.
1564 .. opcode:: IMSB - Index of highest non-sign bit
1566 See SM5 instruction FIRSTBIT_SHI. Computes the 0-based index of the highest
1567 non-sign bit of the argument (i.e. highest 0 bit for negative numbers,
1568 highest 1 bit for positive numbers). Returns -1 if all bits are the same
1569 (i.e. for inputs 0 and -1).
1571 .. opcode:: UMSB - Index of highest set bit
1573 See SM5 instruction FIRSTBIT_HI. Computes the 0-based index of the highest
1574 set bit of the argument. Returns -1 if none are set.
1577 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1579 These opcodes are only supported in geometry shaders; they have no meaning
1580 in any other type of shader.
1582 .. opcode:: EMIT - Emit
1584 Generate a new vertex for the current primitive into the specified vertex
1585 stream using the values in the output registers.
1588 .. opcode:: ENDPRIM - End Primitive
1590 Complete the current primitive in the specified vertex stream (consisting of
1591 the emitted vertices), and start a new one.
1597 These opcodes are part of :term:`GLSL`'s opcode set. Support for these
1598 opcodes is determined by a special capability bit, ``GLSL``.
1599 Some require glsl version 1.30 (UIF/SWITCH/CASE/DEFAULT/ENDSWITCH).
1601 .. opcode:: CAL - Subroutine Call
1607 .. opcode:: RET - Subroutine Call Return
1612 .. opcode:: CONT - Continue
1614 Unconditionally moves the point of execution to the instruction after the
1615 last bgnloop. The instruction must appear within a bgnloop/endloop.
1619 Support for CONT is determined by a special capability bit,
1620 ``TGSI_CONT_SUPPORTED``. See :ref:`Screen` for more information.
1623 .. opcode:: BGNLOOP - Begin a Loop
1625 Start a loop. Must have a matching endloop.
1628 .. opcode:: BGNSUB - Begin Subroutine
1630 Starts definition of a subroutine. Must have a matching endsub.
1633 .. opcode:: ENDLOOP - End a Loop
1635 End a loop started with bgnloop.
1638 .. opcode:: ENDSUB - End Subroutine
1640 Ends definition of a subroutine.
1643 .. opcode:: NOP - No Operation
1648 .. opcode:: BRK - Break
1650 Unconditionally moves the point of execution to the instruction after the
1651 next endloop or endswitch. The instruction must appear within a loop/endloop
1652 or switch/endswitch.
1655 .. opcode:: IF - Float If
1657 Start an IF ... ELSE .. ENDIF block. Condition evaluates to true if
1661 where src0.x is interpreted as a floating point register.
1664 .. opcode:: UIF - Bitwise If
1666 Start an UIF ... ELSE .. ENDIF block. Condition evaluates to true if
1670 where src0.x is interpreted as an integer register.
1673 .. opcode:: ELSE - Else
1675 Starts an else block, after an IF or UIF statement.
1678 .. opcode:: ENDIF - End If
1680 Ends an IF or UIF block.
1683 .. opcode:: SWITCH - Switch
1685 Starts a C-style switch expression. The switch consists of one or multiple
1686 CASE statements, and at most one DEFAULT statement. Execution of a statement
1687 ends when a BRK is hit, but just like in C falling through to other cases
1688 without a break is allowed. Similarly, DEFAULT label is allowed anywhere not
1689 just as last statement, and fallthrough is allowed into/from it.
1690 CASE src arguments are evaluated at bit level against the SWITCH src argument.
1696 (some instructions here)
1699 (some instructions here)
1702 (some instructions here)
1707 .. opcode:: CASE - Switch case
1709 This represents a switch case label. The src arg must be an integer immediate.
1712 .. opcode:: DEFAULT - Switch default
1714 This represents the default case in the switch, which is taken if no other
1718 .. opcode:: ENDSWITCH - End of switch
1720 Ends a switch expression.
1726 The interpolation instructions allow an input to be interpolated in a
1727 different way than its declaration. This corresponds to the GLSL 4.00
1728 interpolateAt* functions. The first argument of each of these must come from
1729 ``TGSI_FILE_INPUT``.
1731 .. opcode:: INTERP_CENTROID - Interpolate at the centroid
1733 Interpolates the varying specified by src0 at the centroid
1735 .. opcode:: INTERP_SAMPLE - Interpolate at the specified sample
1737 Interpolates the varying specified by src0 at the sample id specified by
1738 src1.x (interpreted as an integer)
1740 .. opcode:: INTERP_OFFSET - Interpolate at the specified offset
1742 Interpolates the varying specified by src0 at the offset src1.xy from the
1743 pixel center (interpreted as floats)
1751 The double-precision opcodes reinterpret four-component vectors into
1752 two-component vectors with doubled precision in each component.
1754 .. opcode:: DABS - Absolute
1762 .. opcode:: DADD - Add
1766 dst.xy = src0.xy + src1.xy
1768 dst.zw = src0.zw + src1.zw
1770 .. opcode:: DSEQ - Set on Equal
1774 dst.x = src0.xy == src1.xy ? \sim 0 : 0
1776 dst.z = src0.zw == src1.zw ? \sim 0 : 0
1778 .. opcode:: DSNE - Set on Equal
1782 dst.x = src0.xy != src1.xy ? \sim 0 : 0
1784 dst.z = src0.zw != src1.zw ? \sim 0 : 0
1786 .. opcode:: DSLT - Set on Less than
1790 dst.x = src0.xy < src1.xy ? \sim 0 : 0
1792 dst.z = src0.zw < src1.zw ? \sim 0 : 0
1794 .. opcode:: DSGE - Set on Greater equal
1798 dst.x = src0.xy >= src1.xy ? \sim 0 : 0
1800 dst.z = src0.zw >= src1.zw ? \sim 0 : 0
1802 .. opcode:: DFRAC - Fraction
1806 dst.xy = src.xy - \lfloor src.xy\rfloor
1808 dst.zw = src.zw - \lfloor src.zw\rfloor
1810 .. opcode:: DTRUNC - Truncate
1814 dst.xy = trunc(src.xy)
1816 dst.zw = trunc(src.zw)
1818 .. opcode:: DCEIL - Ceiling
1822 dst.xy = \lceil src.xy\rceil
1824 dst.zw = \lceil src.zw\rceil
1826 .. opcode:: DFLR - Floor
1830 dst.xy = \lfloor src.xy\rfloor
1832 dst.zw = \lfloor src.zw\rfloor
1834 .. opcode:: DROUND - Fraction
1838 dst.xy = round(src.xy)
1840 dst.zw = round(src.zw)
1842 .. opcode:: DSSG - Set Sign
1846 dst.xy = (src.xy > 0) ? 1.0 : (src.xy < 0) ? -1.0 : 0.0
1848 dst.zw = (src.zw > 0) ? 1.0 : (src.zw < 0) ? -1.0 : 0.0
1850 .. opcode:: DFRACEXP - Convert Number to Fractional and Integral Components
1852 Like the ``frexp()`` routine in many math libraries, this opcode stores the
1853 exponent of its source to ``dst0``, and the significand to ``dst1``, such that
1854 :math:`dst1 \times 2^{dst0} = src` .
1858 dst0.xy = exp(src.xy)
1860 dst1.xy = frac(src.xy)
1862 dst0.zw = exp(src.zw)
1864 dst1.zw = frac(src.zw)
1866 .. opcode:: DLDEXP - Multiply Number by Integral Power of 2
1868 This opcode is the inverse of :opcode:`DFRACEXP`. The second
1869 source is an integer.
1873 dst.xy = src0.xy \times 2^{src1.x}
1875 dst.zw = src0.zw \times 2^{src1.y}
1877 .. opcode:: DMIN - Minimum
1881 dst.xy = min(src0.xy, src1.xy)
1883 dst.zw = min(src0.zw, src1.zw)
1885 .. opcode:: DMAX - Maximum
1889 dst.xy = max(src0.xy, src1.xy)
1891 dst.zw = max(src0.zw, src1.zw)
1893 .. opcode:: DMUL - Multiply
1897 dst.xy = src0.xy \times src1.xy
1899 dst.zw = src0.zw \times src1.zw
1902 .. opcode:: DMAD - Multiply And Add
1906 dst.xy = src0.xy \times src1.xy + src2.xy
1908 dst.zw = src0.zw \times src1.zw + src2.zw
1911 .. opcode:: DFMA - Fused Multiply-Add
1913 Perform a * b + c with no intermediate rounding step.
1917 dst.xy = src0.xy \times src1.xy + src2.xy
1919 dst.zw = src0.zw \times src1.zw + src2.zw
1922 .. opcode:: DDIV - Divide
1926 dst.xy = \frac{src0.xy}{src1.xy}
1928 dst.zw = \frac{src0.zw}{src1.zw}
1931 .. opcode:: DRCP - Reciprocal
1935 dst.xy = \frac{1}{src.xy}
1937 dst.zw = \frac{1}{src.zw}
1939 .. opcode:: DSQRT - Square Root
1943 dst.xy = \sqrt{src.xy}
1945 dst.zw = \sqrt{src.zw}
1947 .. opcode:: DRSQ - Reciprocal Square Root
1951 dst.xy = \frac{1}{\sqrt{src.xy}}
1953 dst.zw = \frac{1}{\sqrt{src.zw}}
1955 .. opcode:: F2D - Float to Double
1959 dst.xy = double(src0.x)
1961 dst.zw = double(src0.y)
1963 .. opcode:: D2F - Double to Float
1967 dst.x = float(src0.xy)
1969 dst.y = float(src0.zw)
1971 .. opcode:: I2D - Int to Double
1975 dst.xy = double(src0.x)
1977 dst.zw = double(src0.y)
1979 .. opcode:: D2I - Double to Int
1983 dst.x = int(src0.xy)
1985 dst.y = int(src0.zw)
1987 .. opcode:: U2D - Unsigned Int to Double
1991 dst.xy = double(src0.x)
1993 dst.zw = double(src0.y)
1995 .. opcode:: D2U - Double to Unsigned Int
1999 dst.x = unsigned(src0.xy)
2001 dst.y = unsigned(src0.zw)
2006 The 64-bit integer opcodes reinterpret four-component vectors into
2007 two-component vectors with 64-bits in each component.
2009 .. opcode:: I64ABS - 64-bit Integer Absolute Value
2017 .. opcode:: I64NEG - 64-bit Integer Negate
2027 .. opcode:: I64SSG - 64-bit Integer Set Sign
2031 dst.xy = (src0.xy < 0) ? -1 : (src0.xy > 0) ? 1 : 0
2033 dst.zw = (src0.zw < 0) ? -1 : (src0.zw > 0) ? 1 : 0
2035 .. opcode:: U64ADD - 64-bit Integer Add
2039 dst.xy = src0.xy + src1.xy
2041 dst.zw = src0.zw + src1.zw
2043 .. opcode:: U64MUL - 64-bit Integer Multiply
2047 dst.xy = src0.xy * src1.xy
2049 dst.zw = src0.zw * src1.zw
2051 .. opcode:: U64SEQ - 64-bit Integer Set on Equal
2055 dst.x = src0.xy == src1.xy ? \sim 0 : 0
2057 dst.z = src0.zw == src1.zw ? \sim 0 : 0
2059 .. opcode:: U64SNE - 64-bit Integer Set on Not Equal
2063 dst.x = src0.xy != src1.xy ? \sim 0 : 0
2065 dst.z = src0.zw != src1.zw ? \sim 0 : 0
2067 .. opcode:: U64SLT - 64-bit Unsigned Integer Set on Less Than
2071 dst.x = src0.xy < src1.xy ? \sim 0 : 0
2073 dst.z = src0.zw < src1.zw ? \sim 0 : 0
2075 .. opcode:: U64SGE - 64-bit Unsigned Integer Set on Greater Equal
2079 dst.x = src0.xy >= src1.xy ? \sim 0 : 0
2081 dst.z = src0.zw >= src1.zw ? \sim 0 : 0
2083 .. opcode:: I64SLT - 64-bit Signed Integer Set on Less Than
2087 dst.x = src0.xy < src1.xy ? \sim 0 : 0
2089 dst.z = src0.zw < src1.zw ? \sim 0 : 0
2091 .. opcode:: I64SGE - 64-bit Signed Integer Set on Greater Equal
2095 dst.x = src0.xy >= src1.xy ? \sim 0 : 0
2097 dst.z = src0.zw >= src1.zw ? \sim 0 : 0
2099 .. opcode:: I64MIN - Minimum of 64-bit Signed Integers
2103 dst.xy = min(src0.xy, src1.xy)
2105 dst.zw = min(src0.zw, src1.zw)
2107 .. opcode:: U64MIN - Minimum of 64-bit Unsigned Integers
2111 dst.xy = min(src0.xy, src1.xy)
2113 dst.zw = min(src0.zw, src1.zw)
2115 .. opcode:: I64MAX - Maximum of 64-bit Signed Integers
2119 dst.xy = max(src0.xy, src1.xy)
2121 dst.zw = max(src0.zw, src1.zw)
2123 .. opcode:: U64MAX - Maximum of 64-bit Unsigned Integers
2127 dst.xy = max(src0.xy, src1.xy)
2129 dst.zw = max(src0.zw, src1.zw)
2131 .. opcode:: U64SHL - Shift Left 64-bit Unsigned Integer
2133 The shift count is masked with 0x3f before the shift is applied.
2137 dst.xy = src0.xy << (0x3f \& src1.x)
2139 dst.zw = src0.zw << (0x3f \& src1.y)
2141 .. opcode:: I64SHR - Arithmetic Shift Right (of 64-bit Signed Integer)
2143 The shift count is masked with 0x3f before the shift is applied.
2147 dst.xy = src0.xy >> (0x3f \& src1.x)
2149 dst.zw = src0.zw >> (0x3f \& src1.y)
2151 .. opcode:: U64SHR - Logical Shift Right (of 64-bit Unsigned Integer)
2153 The shift count is masked with 0x3f before the shift is applied.
2157 dst.xy = src0.xy >> (unsigned) (0x3f \& src1.x)
2159 dst.zw = src0.zw >> (unsigned) (0x3f \& src1.y)
2161 .. opcode:: I64DIV - 64-bit Signed Integer Division
2165 dst.xy = \frac{src0.xy}{src1.xy}
2167 dst.zw = \frac{src0.zw}{src1.zw}
2169 .. opcode:: U64DIV - 64-bit Unsigned Integer Division
2173 dst.xy = \frac{src0.xy}{src1.xy}
2175 dst.zw = \frac{src0.zw}{src1.zw}
2177 .. opcode:: U64MOD - 64-bit Unsigned Integer Remainder
2181 dst.xy = src0.xy \bmod src1.xy
2183 dst.zw = src0.zw \bmod src1.zw
2185 .. opcode:: I64MOD - 64-bit Signed Integer Remainder
2189 dst.xy = src0.xy \bmod src1.xy
2191 dst.zw = src0.zw \bmod src1.zw
2193 .. opcode:: F2U64 - Float to 64-bit Unsigned Int
2197 dst.xy = (uint64_t) src0.x
2199 dst.zw = (uint64_t) src0.y
2201 .. opcode:: F2I64 - Float to 64-bit Int
2205 dst.xy = (int64_t) src0.x
2207 dst.zw = (int64_t) src0.y
2209 .. opcode:: U2I64 - Unsigned Integer to 64-bit Integer
2211 This is a zero extension.
2215 dst.xy = (uint64_t) src0.x
2217 dst.zw = (uint64_t) src0.y
2219 .. opcode:: I2I64 - Signed Integer to 64-bit Integer
2221 This is a sign extension.
2225 dst.xy = (int64_t) src0.x
2227 dst.zw = (int64_t) src0.y
2229 .. opcode:: D2U64 - Double to 64-bit Unsigned Int
2233 dst.xy = (uint64_t) src0.xy
2235 dst.zw = (uint64_t) src0.zw
2237 .. opcode:: D2I64 - Double to 64-bit Int
2241 dst.xy = (int64_t) src0.xy
2243 dst.zw = (int64_t) src0.zw
2245 .. opcode:: U642F - 64-bit unsigned integer to float
2249 dst.x = (float) src0.xy
2251 dst.y = (float) src0.zw
2253 .. opcode:: I642F - 64-bit Int to Float
2257 dst.x = (float) src0.xy
2259 dst.y = (float) src0.zw
2261 .. opcode:: U642D - 64-bit unsigned integer to double
2265 dst.xy = (double) src0.xy
2267 dst.zw = (double) src0.zw
2269 .. opcode:: I642D - 64-bit Int to double
2273 dst.xy = (double) src0.xy
2275 dst.zw = (double) src0.zw
2277 .. _samplingopcodes:
2279 Resource Sampling Opcodes
2280 ^^^^^^^^^^^^^^^^^^^^^^^^^
2282 Those opcodes follow very closely semantics of the respective Direct3D
2283 instructions. If in doubt double check Direct3D documentation.
2284 Note that the swizzle on SVIEW (src1) determines texel swizzling
2289 Using provided address, sample data from the specified texture using the
2290 filtering mode identified by the given sampler. The source data may come from
2291 any resource type other than buffers.
2293 Syntax: ``SAMPLE dst, address, sampler_view, sampler``
2295 Example: ``SAMPLE TEMP[0], TEMP[1], SVIEW[0], SAMP[0]``
2297 .. opcode:: SAMPLE_I
2299 Simplified alternative to the SAMPLE instruction. Using the provided
2300 integer address, SAMPLE_I fetches data from the specified sampler view
2301 without any filtering. The source data may come from any resource type
2304 Syntax: ``SAMPLE_I dst, address, sampler_view``
2306 Example: ``SAMPLE_I TEMP[0], TEMP[1], SVIEW[0]``
2308 The 'address' is specified as unsigned integers. If the 'address' is out of
2309 range [0...(# texels - 1)] the result of the fetch is always 0 in all
2310 components. As such the instruction doesn't honor address wrap modes, in
2311 cases where that behavior is desirable 'SAMPLE' instruction should be used.
2312 address.w always provides an unsigned integer mipmap level. If the value is
2313 out of the range then the instruction always returns 0 in all components.
2314 address.yz are ignored for buffers and 1d textures. address.z is ignored
2315 for 1d texture arrays and 2d textures.
2317 For 1D texture arrays address.y provides the array index (also as unsigned
2318 integer). If the value is out of the range of available array indices
2319 [0... (array size - 1)] then the opcode always returns 0 in all components.
2320 For 2D texture arrays address.z provides the array index, otherwise it
2321 exhibits the same behavior as in the case for 1D texture arrays. The exact
2322 semantics of the source address are presented in the table below:
2324 +---------------------------+----+-----+-----+---------+
2325 | resource type | X | Y | Z | W |
2326 +===========================+====+=====+=====+=========+
2327 | ``PIPE_BUFFER`` | x | | | ignored |
2328 +---------------------------+----+-----+-----+---------+
2329 | ``PIPE_TEXTURE_1D`` | x | | | mpl |
2330 +---------------------------+----+-----+-----+---------+
2331 | ``PIPE_TEXTURE_2D`` | x | y | | mpl |
2332 +---------------------------+----+-----+-----+---------+
2333 | ``PIPE_TEXTURE_3D`` | x | y | z | mpl |
2334 +---------------------------+----+-----+-----+---------+
2335 | ``PIPE_TEXTURE_RECT`` | x | y | | mpl |
2336 +---------------------------+----+-----+-----+---------+
2337 | ``PIPE_TEXTURE_CUBE`` | not allowed as source |
2338 +---------------------------+----+-----+-----+---------+
2339 | ``PIPE_TEXTURE_1D_ARRAY`` | x | idx | | mpl |
2340 +---------------------------+----+-----+-----+---------+
2341 | ``PIPE_TEXTURE_2D_ARRAY`` | x | y | idx | mpl |
2342 +---------------------------+----+-----+-----+---------+
2344 Where 'mpl' is a mipmap level and 'idx' is the array index.
2346 .. opcode:: SAMPLE_I_MS
2348 Just like SAMPLE_I but allows fetch data from multi-sampled surfaces.
2350 Syntax: ``SAMPLE_I_MS dst, address, sampler_view, sample``
2352 .. opcode:: SAMPLE_B
2354 Just like the SAMPLE instruction with the exception that an additional bias
2355 is applied to the level of detail computed as part of the instruction
2358 Syntax: ``SAMPLE_B dst, address, sampler_view, sampler, lod_bias``
2360 Example: ``SAMPLE_B TEMP[0], TEMP[1], SVIEW[0], SAMP[0], TEMP[2].x``
2362 .. opcode:: SAMPLE_C
2364 Similar to the SAMPLE instruction but it performs a comparison filter. The
2365 operands to SAMPLE_C are identical to SAMPLE, except that there is an
2366 additional float32 operand, reference value, which must be a register with
2367 single-component, or a scalar literal. SAMPLE_C makes the hardware use the
2368 current samplers compare_func (in pipe_sampler_state) to compare reference
2369 value against the red component value for the surce resource at each texel
2370 that the currently configured texture filter covers based on the provided
2373 Syntax: ``SAMPLE_C dst, address, sampler_view.r, sampler, ref_value``
2375 Example: ``SAMPLE_C TEMP[0], TEMP[1], SVIEW[0].r, SAMP[0], TEMP[2].x``
2377 .. opcode:: SAMPLE_C_LZ
2379 Same as SAMPLE_C, but LOD is 0 and derivatives are ignored. The LZ stands
2382 Syntax: ``SAMPLE_C_LZ dst, address, sampler_view.r, sampler, ref_value``
2384 Example: ``SAMPLE_C_LZ TEMP[0], TEMP[1], SVIEW[0].r, SAMP[0], TEMP[2].x``
2387 .. opcode:: SAMPLE_D
2389 SAMPLE_D is identical to the SAMPLE opcode except that the derivatives for
2390 the source address in the x direction and the y direction are provided by
2393 Syntax: ``SAMPLE_D dst, address, sampler_view, sampler, der_x, der_y``
2395 Example: ``SAMPLE_D TEMP[0], TEMP[1], SVIEW[0], SAMP[0], TEMP[2], TEMP[3]``
2397 .. opcode:: SAMPLE_L
2399 SAMPLE_L is identical to the SAMPLE opcode except that the LOD is provided
2400 directly as a scalar value, representing no anisotropy.
2402 Syntax: ``SAMPLE_L dst, address, sampler_view, sampler, explicit_lod``
2404 Example: ``SAMPLE_L TEMP[0], TEMP[1], SVIEW[0], SAMP[0], TEMP[2].x``
2408 Gathers the four texels to be used in a bi-linear filtering operation and
2409 packs them into a single register. Only works with 2D, 2D array, cubemaps,
2410 and cubemaps arrays. For 2D textures, only the addressing modes of the
2411 sampler and the top level of any mip pyramid are used. Set W to zero. It
2412 behaves like the SAMPLE instruction, but a filtered sample is not
2413 generated. The four samples that contribute to filtering are placed into
2414 xyzw in counter-clockwise order, starting with the (u,v) texture coordinate
2415 delta at the following locations (-, +), (+, +), (+, -), (-, -), where the
2416 magnitude of the deltas are half a texel.
2419 .. opcode:: SVIEWINFO
2421 Query the dimensions of a given sampler view. dst receives width, height,
2422 depth or array size and number of mipmap levels as int4. The dst can have a
2423 writemask which will specify what info is the caller interested in.
2425 Syntax: ``SVIEWINFO dst, src_mip_level, sampler_view``
2427 Example: ``SVIEWINFO TEMP[0], TEMP[1].x, SVIEW[0]``
2429 src_mip_level is an unsigned integer scalar. If it's out of range then
2430 returns 0 for width, height and depth/array size but the total number of
2431 mipmap is still returned correctly for the given sampler view. The returned
2432 width, height and depth values are for the mipmap level selected by the
2433 src_mip_level and are in the number of texels. For 1d texture array width
2434 is in dst.x, array size is in dst.y and dst.z is 0. The number of mipmaps is
2435 still in dst.w. In contrast to d3d10 resinfo, there's no way in the tgsi
2436 instruction encoding to specify the return type (float/rcpfloat/uint), hence
2437 always using uint. Also, unlike the SAMPLE instructions, the swizzle on src1
2438 resinfo allowing swizzling dst values is ignored (due to the interaction
2439 with rcpfloat modifier which requires some swizzle handling in the state
2442 .. opcode:: SAMPLE_POS
2444 Query the position of a sample in the given resource or render target
2445 when per-sample fragment shading is in effect.
2447 Syntax: ``SAMPLE_POS dst, source, sample_index``
2449 dst receives float4 (x, y, undef, undef) indicated where the sample is
2450 located. Sample locations are in the range [0, 1] where 0.5 is the center
2453 source is either a sampler view (to indicate a shader resource) or temp
2454 register (to indicate the render target). The source register may have
2455 an optional swizzle to apply to the returned result
2457 sample_index is an integer scalar indicating which sample position is to
2460 If per-sample shading is not in effect or the source resource or render
2461 target is not multisampled, the result is (0.5, 0.5, undef, undef).
2463 NOTE: no driver has implemented this opcode yet (and no state tracker
2464 emits it). This information is subject to change.
2466 .. opcode:: SAMPLE_INFO
2468 Query the number of samples in a multisampled resource or render target.
2470 Syntax: ``SAMPLE_INFO dst, source``
2472 dst receives int4 (n, 0, 0, 0) where n is the number of samples in a
2473 resource or the render target.
2475 source is either a sampler view (to indicate a shader resource) or temp
2476 register (to indicate the render target). The source register may have
2477 an optional swizzle to apply to the returned result
2479 If per-sample shading is not in effect or the source resource or render
2480 target is not multisampled, the result is (1, 0, 0, 0).
2482 NOTE: no driver has implemented this opcode yet (and no state tracker
2483 emits it). This information is subject to change.
2485 .. _resourceopcodes:
2487 Resource Access Opcodes
2488 ^^^^^^^^^^^^^^^^^^^^^^^
2490 For these opcodes, the resource can be a BUFFER, IMAGE, or MEMORY.
2492 .. opcode:: LOAD - Fetch data from a shader buffer or image
2494 Syntax: ``LOAD dst, resource, address``
2496 Example: ``LOAD TEMP[0], BUFFER[0], TEMP[1]``
2498 Using the provided integer address, LOAD fetches data
2499 from the specified buffer or texture without any
2502 The 'address' is specified as a vector of unsigned
2503 integers. If the 'address' is out of range the result
2506 Only the first mipmap level of a resource can be read
2507 from using this instruction.
2509 For 1D or 2D texture arrays, the array index is
2510 provided as an unsigned integer in address.y or
2511 address.z, respectively. address.yz are ignored for
2512 buffers and 1D textures. address.z is ignored for 1D
2513 texture arrays and 2D textures. address.w is always
2516 A swizzle suffix may be added to the resource argument
2517 this will cause the resource data to be swizzled accordingly.
2519 .. opcode:: STORE - Write data to a shader resource
2521 Syntax: ``STORE resource, address, src``
2523 Example: ``STORE BUFFER[0], TEMP[0], TEMP[1]``
2525 Using the provided integer address, STORE writes data
2526 to the specified buffer or texture.
2528 The 'address' is specified as a vector of unsigned
2529 integers. If the 'address' is out of range the result
2532 Only the first mipmap level of a resource can be
2533 written to using this instruction.
2535 For 1D or 2D texture arrays, the array index is
2536 provided as an unsigned integer in address.y or
2537 address.z, respectively. address.yz are ignored for
2538 buffers and 1D textures. address.z is ignored for 1D
2539 texture arrays and 2D textures. address.w is always
2542 .. opcode:: RESQ - Query information about a resource
2544 Syntax: ``RESQ dst, resource``
2546 Example: ``RESQ TEMP[0], BUFFER[0]``
2548 Returns information about the buffer or image resource. For buffer
2549 resources, the size (in bytes) is returned in the x component. For
2550 image resources, .xyz will contain the width/height/layers of the
2551 image, while .w will contain the number of samples for multi-sampled
2554 .. opcode:: FBFETCH - Load data from framebuffer
2556 Syntax: ``FBFETCH dst, output``
2558 Example: ``FBFETCH TEMP[0], OUT[0]``
2560 This is only valid on ``COLOR`` semantic outputs. Returns the color
2561 of the current position in the framebuffer from before this fragment
2562 shader invocation. May return the same value from multiple calls for
2563 a particular output within a single invocation. Note that result may
2564 be undefined if a fragment is drawn multiple times without a blend
2568 .. _threadsyncopcodes:
2570 Inter-thread synchronization opcodes
2571 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2573 These opcodes are intended for communication between threads running
2574 within the same compute grid. For now they're only valid in compute
2577 .. opcode:: BARRIER - Thread group barrier
2581 This opcode suspends the execution of the current thread until all
2582 the remaining threads in the working group reach the same point of
2583 the program. Results are unspecified if any of the remaining
2584 threads terminates or never reaches an executed BARRIER instruction.
2586 .. opcode:: MEMBAR - Memory barrier
2590 This opcode waits for the completion of all memory accesses based on
2591 the type passed in. The type is an immediate bitfield with the following
2594 Bit 0: Shader storage buffers
2595 Bit 1: Atomic buffers
2597 Bit 3: Shared memory
2600 These may be passed in in any combination. An implementation is free to not
2601 distinguish between these as it sees fit. However these map to all the
2602 possibilities made available by GLSL.
2609 These opcodes provide atomic variants of some common arithmetic and
2610 logical operations. In this context atomicity means that another
2611 concurrent memory access operation that affects the same memory
2612 location is guaranteed to be performed strictly before or after the
2613 entire execution of the atomic operation. The resource may be a BUFFER,
2614 IMAGE, or MEMORY. In the case of an image, the offset works the same as for
2615 ``LOAD`` and ``STORE``, specified above. These atomic operations may
2616 only be used with 32-bit integer image formats.
2618 .. opcode:: ATOMUADD - Atomic integer addition
2620 Syntax: ``ATOMUADD dst, resource, offset, src``
2622 Example: ``ATOMUADD TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2624 The following operation is performed atomically:
2628 dst_x = resource[offset]
2630 resource[offset] = dst_x + src_x
2633 .. opcode:: ATOMXCHG - Atomic exchange
2635 Syntax: ``ATOMXCHG dst, resource, offset, src``
2637 Example: ``ATOMXCHG TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2639 The following operation is performed atomically:
2643 dst_x = resource[offset]
2645 resource[offset] = src_x
2648 .. opcode:: ATOMCAS - Atomic compare-and-exchange
2650 Syntax: ``ATOMCAS dst, resource, offset, cmp, src``
2652 Example: ``ATOMCAS TEMP[0], BUFFER[0], TEMP[1], TEMP[2], TEMP[3]``
2654 The following operation is performed atomically:
2658 dst_x = resource[offset]
2660 resource[offset] = (dst_x == cmp_x ? src_x : dst_x)
2663 .. opcode:: ATOMAND - Atomic bitwise And
2665 Syntax: ``ATOMAND dst, resource, offset, src``
2667 Example: ``ATOMAND TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2669 The following operation is performed atomically:
2673 dst_x = resource[offset]
2675 resource[offset] = dst_x \& src_x
2678 .. opcode:: ATOMOR - Atomic bitwise Or
2680 Syntax: ``ATOMOR dst, resource, offset, src``
2682 Example: ``ATOMOR TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2684 The following operation is performed atomically:
2688 dst_x = resource[offset]
2690 resource[offset] = dst_x | src_x
2693 .. opcode:: ATOMXOR - Atomic bitwise Xor
2695 Syntax: ``ATOMXOR dst, resource, offset, src``
2697 Example: ``ATOMXOR TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2699 The following operation is performed atomically:
2703 dst_x = resource[offset]
2705 resource[offset] = dst_x \oplus src_x
2708 .. opcode:: ATOMUMIN - Atomic unsigned minimum
2710 Syntax: ``ATOMUMIN dst, resource, offset, src``
2712 Example: ``ATOMUMIN TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2714 The following operation is performed atomically:
2718 dst_x = resource[offset]
2720 resource[offset] = (dst_x < src_x ? dst_x : src_x)
2723 .. opcode:: ATOMUMAX - Atomic unsigned maximum
2725 Syntax: ``ATOMUMAX dst, resource, offset, src``
2727 Example: ``ATOMUMAX TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2729 The following operation is performed atomically:
2733 dst_x = resource[offset]
2735 resource[offset] = (dst_x > src_x ? dst_x : src_x)
2738 .. opcode:: ATOMIMIN - Atomic signed minimum
2740 Syntax: ``ATOMIMIN dst, resource, offset, src``
2742 Example: ``ATOMIMIN TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2744 The following operation is performed atomically:
2748 dst_x = resource[offset]
2750 resource[offset] = (dst_x < src_x ? dst_x : src_x)
2753 .. opcode:: ATOMIMAX - Atomic signed maximum
2755 Syntax: ``ATOMIMAX dst, resource, offset, src``
2757 Example: ``ATOMIMAX TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2759 The following operation is performed atomically:
2763 dst_x = resource[offset]
2765 resource[offset] = (dst_x > src_x ? dst_x : src_x)
2768 .. _interlaneopcodes:
2773 These opcodes reduce the given value across the shader invocations
2774 running in the current SIMD group. Every thread in the subgroup will receive
2775 the same result. The BALLOT operations accept a single-channel argument that
2776 is treated as a boolean and produce a 64-bit value.
2778 .. opcode:: VOTE_ANY - Value is set in any of the active invocations
2780 Syntax: ``VOTE_ANY dst, value``
2782 Example: ``VOTE_ANY TEMP[0].x, TEMP[1].x``
2785 .. opcode:: VOTE_ALL - Value is set in all of the active invocations
2787 Syntax: ``VOTE_ALL dst, value``
2789 Example: ``VOTE_ALL TEMP[0].x, TEMP[1].x``
2792 .. opcode:: VOTE_EQ - Value is the same in all of the active invocations
2794 Syntax: ``VOTE_EQ dst, value``
2796 Example: ``VOTE_EQ TEMP[0].x, TEMP[1].x``
2799 .. opcode:: BALLOT - Lanemask of whether the value is set in each active
2802 Syntax: ``BALLOT dst, value``
2804 Example: ``BALLOT TEMP[0].xy, TEMP[1].x``
2806 When the argument is a constant true, this produces a bitmask of active
2807 invocations. In fragment shaders, this can include helper invocations
2808 (invocations whose outputs and writes to memory are discarded, but which
2809 are used to compute derivatives).
2812 .. opcode:: READ_FIRST - Broadcast the value from the first active
2813 invocation to all active lanes
2815 Syntax: ``READ_FIRST dst, value``
2817 Example: ``READ_FIRST TEMP[0], TEMP[1]``
2820 .. opcode:: READ_INVOC - Retrieve the value from the given invocation
2821 (need not be uniform)
2823 Syntax: ``READ_INVOC dst, value, invocation``
2825 Example: ``READ_INVOC TEMP[0].xy, TEMP[1].xy, TEMP[2].x``
2827 invocation.x controls the invocation number to read from for all channels.
2828 The invocation number must be the same across all active invocations in a
2829 sub-group; otherwise, the results are undefined.
2832 Explanation of symbols used
2833 ------------------------------
2840 :math:`|x|` Absolute value of `x`.
2842 :math:`\lceil x \rceil` Ceiling of `x`.
2844 clamp(x,y,z) Clamp x between y and z.
2845 (x < y) ? y : (x > z) ? z : x
2847 :math:`\lfloor x\rfloor` Floor of `x`.
2849 :math:`\log_2{x}` Logarithm of `x`, base 2.
2851 max(x,y) Maximum of x and y.
2854 min(x,y) Minimum of x and y.
2857 partialx(x) Derivative of x relative to fragment's X.
2859 partialy(x) Derivative of x relative to fragment's Y.
2861 pop() Pop from stack.
2863 :math:`x^y` `x` to the power `y`.
2865 push(x) Push x on stack.
2869 trunc(x) Truncate x, i.e. drop the fraction bits.
2876 discard Discard fragment.
2880 target Label of target instruction.
2891 Declares a register that is will be referenced as an operand in Instruction
2894 File field contains register file that is being declared and is one
2897 UsageMask field specifies which of the register components can be accessed
2898 and is one of TGSI_WRITEMASK.
2900 The Local flag specifies that a given value isn't intended for
2901 subroutine parameter passing and, as a result, the implementation
2902 isn't required to give any guarantees of it being preserved across
2903 subroutine boundaries. As it's merely a compiler hint, the
2904 implementation is free to ignore it.
2906 If Dimension flag is set to 1, a Declaration Dimension token follows.
2908 If Semantic flag is set to 1, a Declaration Semantic token follows.
2910 If Interpolate flag is set to 1, a Declaration Interpolate token follows.
2912 If file is TGSI_FILE_RESOURCE, a Declaration Resource token follows.
2914 If Array flag is set to 1, a Declaration Array token follows.
2917 ^^^^^^^^^^^^^^^^^^^^^^^^
2919 Declarations can optional have an ArrayID attribute which can be referred by
2920 indirect addressing operands. An ArrayID of zero is reserved and treated as
2921 if no ArrayID is specified.
2923 If an indirect addressing operand refers to a specific declaration by using
2924 an ArrayID only the registers in this declaration are guaranteed to be
2925 accessed, accessing any register outside this declaration results in undefined
2926 behavior. Note that for compatibility the effective index is zero-based and
2927 not relative to the specified declaration
2929 If no ArrayID is specified with an indirect addressing operand the whole
2930 register file might be accessed by this operand. This is strongly discouraged
2931 and will prevent packing of scalar/vec2 arrays and effective alias analysis.
2932 This is only legal for TEMP and CONST register files.
2934 Declaration Semantic
2935 ^^^^^^^^^^^^^^^^^^^^^^^^
2937 Vertex and fragment shader input and output registers may be labeled
2938 with semantic information consisting of a name and index.
2940 Follows Declaration token if Semantic bit is set.
2942 Since its purpose is to link a shader with other stages of the pipeline,
2943 it is valid to follow only those Declaration tokens that declare a register
2944 either in INPUT or OUTPUT file.
2946 SemanticName field contains the semantic name of the register being declared.
2947 There is no default value.
2949 SemanticIndex is an optional subscript that can be used to distinguish
2950 different register declarations with the same semantic name. The default value
2953 The meanings of the individual semantic names are explained in the following
2956 TGSI_SEMANTIC_POSITION
2957 """"""""""""""""""""""
2959 For vertex shaders, TGSI_SEMANTIC_POSITION indicates the vertex shader
2960 output register which contains the homogeneous vertex position in the clip
2961 space coordinate system. After clipping, the X, Y and Z components of the
2962 vertex will be divided by the W value to get normalized device coordinates.
2964 For fragment shaders, TGSI_SEMANTIC_POSITION is used to indicate that
2965 fragment shader input (or system value, depending on which one is
2966 supported by the driver) contains the fragment's window position. The X
2967 component starts at zero and always increases from left to right.
2968 The Y component starts at zero and always increases but Y=0 may either
2969 indicate the top of the window or the bottom depending on the fragment
2970 coordinate origin convention (see TGSI_PROPERTY_FS_COORD_ORIGIN).
2971 The Z coordinate ranges from 0 to 1 to represent depth from the front
2972 to the back of the Z buffer. The W component contains the interpolated
2973 reciprocal of the vertex position W component (corresponding to gl_Fragcoord,
2974 but unlike d3d10 which interpolates the same 1/w but then gives back
2975 the reciprocal of the interpolated value).
2977 Fragment shaders may also declare an output register with
2978 TGSI_SEMANTIC_POSITION. Only the Z component is writable. This allows
2979 the fragment shader to change the fragment's Z position.
2986 For vertex shader outputs or fragment shader inputs/outputs, this
2987 label indicates that the register contains an R,G,B,A color.
2989 Several shader inputs/outputs may contain colors so the semantic index
2990 is used to distinguish them. For example, color[0] may be the diffuse
2991 color while color[1] may be the specular color.
2993 This label is needed so that the flat/smooth shading can be applied
2994 to the right interpolants during rasterization.
2998 TGSI_SEMANTIC_BCOLOR
2999 """"""""""""""""""""
3001 Back-facing colors are only used for back-facing polygons, and are only valid
3002 in vertex shader outputs. After rasterization, all polygons are front-facing
3003 and COLOR and BCOLOR end up occupying the same slots in the fragment shader,
3004 so all BCOLORs effectively become regular COLORs in the fragment shader.
3010 Vertex shader inputs and outputs and fragment shader inputs may be
3011 labeled with TGSI_SEMANTIC_FOG to indicate that the register contains
3012 a fog coordinate. Typically, the fragment shader will use the fog coordinate
3013 to compute a fog blend factor which is used to blend the normal fragment color
3014 with a constant fog color. But fog coord really is just an ordinary vec4
3015 register like regular semantics.
3021 Vertex shader input and output registers may be labeled with
3022 TGIS_SEMANTIC_PSIZE to indicate that the register contains a point size
3023 in the form (S, 0, 0, 1). The point size controls the width or diameter
3024 of points for rasterization. This label cannot be used in fragment
3027 When using this semantic, be sure to set the appropriate state in the
3028 :ref:`rasterizer` first.
3031 TGSI_SEMANTIC_TEXCOORD
3032 """"""""""""""""""""""
3034 Only available if PIPE_CAP_TGSI_TEXCOORD is exposed !
3036 Vertex shader outputs and fragment shader inputs may be labeled with
3037 this semantic to make them replaceable by sprite coordinates via the
3038 sprite_coord_enable state in the :ref:`rasterizer`.
3039 The semantic index permitted with this semantic is limited to <= 7.
3041 If the driver does not support TEXCOORD, sprite coordinate replacement
3042 applies to inputs with the GENERIC semantic instead.
3044 The intended use case for this semantic is gl_TexCoord.
3047 TGSI_SEMANTIC_PCOORD
3048 """"""""""""""""""""
3050 Only available if PIPE_CAP_TGSI_TEXCOORD is exposed !
3052 Fragment shader inputs may be labeled with TGSI_SEMANTIC_PCOORD to indicate
3053 that the register contains sprite coordinates in the form (x, y, 0, 1), if
3054 the current primitive is a point and point sprites are enabled. Otherwise,
3055 the contents of the register are undefined.
3057 The intended use case for this semantic is gl_PointCoord.
3060 TGSI_SEMANTIC_GENERIC
3061 """""""""""""""""""""
3063 All vertex/fragment shader inputs/outputs not labeled with any other
3064 semantic label can be considered to be generic attributes. Typical
3065 uses of generic inputs/outputs are texcoords and user-defined values.
3068 TGSI_SEMANTIC_NORMAL
3069 """"""""""""""""""""
3071 Indicates that a vertex shader input is a normal vector. This is
3072 typically only used for legacy graphics APIs.
3078 This label applies to fragment shader inputs (or system values,
3079 depending on which one is supported by the driver) and indicates that
3080 the register contains front/back-face information.
3082 If it is an input, it will be a floating-point vector in the form (F, 0, 0, 1),
3083 where F will be positive when the fragment belongs to a front-facing polygon,
3084 and negative when the fragment belongs to a back-facing polygon.
3086 If it is a system value, it will be an integer vector in the form (F, 0, 0, 1),
3087 where F is 0xffffffff when the fragment belongs to a front-facing polygon and
3088 0 when the fragment belongs to a back-facing polygon.
3091 TGSI_SEMANTIC_EDGEFLAG
3092 """"""""""""""""""""""
3094 For vertex shaders, this sematic label indicates that an input or
3095 output is a boolean edge flag. The register layout is [F, x, x, x]
3096 where F is 0.0 or 1.0 and x = don't care. Normally, the vertex shader
3097 simply copies the edge flag input to the edgeflag output.
3099 Edge flags are used to control which lines or points are actually
3100 drawn when the polygon mode converts triangles/quads/polygons into
3104 TGSI_SEMANTIC_STENCIL
3105 """""""""""""""""""""
3107 For fragment shaders, this semantic label indicates that an output
3108 is a writable stencil reference value. Only the Y component is writable.
3109 This allows the fragment shader to change the fragments stencilref value.
3112 TGSI_SEMANTIC_VIEWPORT_INDEX
3113 """"""""""""""""""""""""""""
3115 For geometry shaders, this semantic label indicates that an output
3116 contains the index of the viewport (and scissor) to use.
3117 This is an integer value, and only the X component is used.
3119 If PIPE_CAP_TGSI_VS_LAYER_VIEWPORT or PIPE_CAP_TGSI_TES_LAYER_VIEWPORT is
3120 supported, then this semantic label can also be used in vertex or
3121 tessellation evaluation shaders, respectively. Only the value written in the
3122 last vertex processing stage is used.
3128 For geometry shaders, this semantic label indicates that an output
3129 contains the layer value to use for the color and depth/stencil surfaces.
3130 This is an integer value, and only the X component is used.
3131 (Also known as rendertarget array index.)
3133 If PIPE_CAP_TGSI_VS_LAYER_VIEWPORT or PIPE_CAP_TGSI_TES_LAYER_VIEWPORT is
3134 supported, then this semantic label can also be used in vertex or
3135 tessellation evaluation shaders, respectively. Only the value written in the
3136 last vertex processing stage is used.
3139 TGSI_SEMANTIC_CULLDIST
3140 """"""""""""""""""""""
3142 Used as distance to plane for performing application-defined culling
3143 of individual primitives against a plane. When components of vertex
3144 elements are given this label, these values are assumed to be a
3145 float32 signed distance to a plane. Primitives will be completely
3146 discarded if the plane distance for all of the vertices in the
3147 primitive are < 0. If a vertex has a cull distance of NaN, that
3148 vertex counts as "out" (as if its < 0);
3149 The limits on both clip and cull distances are bound
3150 by the PIPE_MAX_CLIP_OR_CULL_DISTANCE_COUNT define which defines
3151 the maximum number of components that can be used to hold the
3152 distances and by the PIPE_MAX_CLIP_OR_CULL_DISTANCE_ELEMENT_COUNT
3153 which specifies the maximum number of registers which can be
3154 annotated with those semantics.
3157 TGSI_SEMANTIC_CLIPDIST
3158 """"""""""""""""""""""
3160 Note this covers clipping and culling distances.
3162 When components of vertex elements are identified this way, these
3163 values are each assumed to be a float32 signed distance to a plane.
3166 Primitive setup only invokes rasterization on pixels for which
3167 the interpolated plane distances are >= 0.
3170 Primitives will be completely discarded if the plane distance
3171 for all of the vertices in the primitive are < 0.
3172 If a vertex has a cull distance of NaN, that vertex counts as "out"
3175 Multiple clip/cull planes can be implemented simultaneously, by
3176 annotating multiple components of one or more vertex elements with
3177 the above specified semantic.
3178 The limits on both clip and cull distances are bound
3179 by the PIPE_MAX_CLIP_OR_CULL_DISTANCE_COUNT define which defines
3180 the maximum number of components that can be used to hold the
3181 distances and by the PIPE_MAX_CLIP_OR_CULL_DISTANCE_ELEMENT_COUNT
3182 which specifies the maximum number of registers which can be
3183 annotated with those semantics.
3184 The properties NUM_CLIPDIST_ENABLED and NUM_CULLDIST_ENABLED
3185 are used to divide up the 2 x vec4 space between clipping and culling.
3187 TGSI_SEMANTIC_SAMPLEID
3188 """"""""""""""""""""""
3190 For fragment shaders, this semantic label indicates that a system value
3191 contains the current sample id (i.e. gl_SampleID) as an unsigned int.
3192 Only the X component is used. If per-sample shading is not enabled,
3193 the result is (0, undef, undef, undef).
3195 Note that if the fragment shader uses this system value, the fragment
3196 shader is automatically executed at per sample frequency.
3198 TGSI_SEMANTIC_SAMPLEPOS
3199 """""""""""""""""""""""
3201 For fragment shaders, this semantic label indicates that a system
3202 value contains the current sample's position as float4(x, y, undef, undef)
3203 in the render target (i.e. gl_SamplePosition) when per-fragment shading
3204 is in effect. Position values are in the range [0, 1] where 0.5 is
3205 the center of the fragment.
3207 Note that if the fragment shader uses this system value, the fragment
3208 shader is automatically executed at per sample frequency.
3210 TGSI_SEMANTIC_SAMPLEMASK
3211 """"""""""""""""""""""""
3213 For fragment shaders, this semantic label can be applied to either a
3214 shader system value input or output.
3216 For a system value, the sample mask indicates the set of samples covered by
3217 the current primitive. If MSAA is not enabled, the value is (1, 0, 0, 0).
3219 For an output, the sample mask is used to disable further sample processing.
3221 For both, the register type is uint[4] but only the X component is used
3222 (i.e. gl_SampleMask[0]). Each bit corresponds to one sample position (up
3223 to 32x MSAA is supported).
3225 TGSI_SEMANTIC_INVOCATIONID
3226 """"""""""""""""""""""""""
3228 For geometry shaders, this semantic label indicates that a system value
3229 contains the current invocation id (i.e. gl_InvocationID).
3230 This is an integer value, and only the X component is used.
3232 TGSI_SEMANTIC_INSTANCEID
3233 """"""""""""""""""""""""
3235 For vertex shaders, this semantic label indicates that a system value contains
3236 the current instance id (i.e. gl_InstanceID). It does not include the base
3237 instance. This is an integer value, and only the X component is used.
3239 TGSI_SEMANTIC_VERTEXID
3240 """"""""""""""""""""""
3242 For vertex shaders, this semantic label indicates that a system value contains
3243 the current vertex id (i.e. gl_VertexID). It does (unlike in d3d10) include the
3244 base vertex. This is an integer value, and only the X component is used.
3246 TGSI_SEMANTIC_VERTEXID_NOBASE
3247 """""""""""""""""""""""""""""""
3249 For vertex shaders, this semantic label indicates that a system value contains
3250 the current vertex id without including the base vertex (this corresponds to
3251 d3d10 vertex id, so TGSI_SEMANTIC_VERTEXID_NOBASE + TGSI_SEMANTIC_BASEVERTEX
3252 == TGSI_SEMANTIC_VERTEXID). This is an integer value, and only the X component
3255 TGSI_SEMANTIC_BASEVERTEX
3256 """"""""""""""""""""""""
3258 For vertex shaders, this semantic label indicates that a system value contains
3259 the base vertex (i.e. gl_BaseVertex). Note that for non-indexed draw calls,
3260 this contains the first (or start) value instead.
3261 This is an integer value, and only the X component is used.
3263 TGSI_SEMANTIC_PRIMID
3264 """"""""""""""""""""
3266 For geometry and fragment shaders, this semantic label indicates the value
3267 contains the primitive id (i.e. gl_PrimitiveID). This is an integer value,
3268 and only the X component is used.
3269 FIXME: This right now can be either a ordinary input or a system value...
3275 For tessellation evaluation/control shaders, this semantic label indicates a
3276 generic per-patch attribute. Such semantics will not implicitly be per-vertex
3279 TGSI_SEMANTIC_TESSCOORD
3280 """""""""""""""""""""""
3282 For tessellation evaluation shaders, this semantic label indicates the
3283 coordinates of the vertex being processed. This is available in XYZ; W is
3286 TGSI_SEMANTIC_TESSOUTER
3287 """""""""""""""""""""""
3289 For tessellation evaluation/control shaders, this semantic label indicates the
3290 outer tessellation levels of the patch. Isoline tessellation will only have XY
3291 defined, triangle will have XYZ and quads will have XYZW defined. This
3292 corresponds to gl_TessLevelOuter.
3294 TGSI_SEMANTIC_TESSINNER
3295 """""""""""""""""""""""
3297 For tessellation evaluation/control shaders, this semantic label indicates the
3298 inner tessellation levels of the patch. The X value is only defined for
3299 triangle tessellation, while quads will have XY defined. This is entirely
3300 undefined for isoline tessellation.
3302 TGSI_SEMANTIC_VERTICESIN
3303 """"""""""""""""""""""""
3305 For tessellation evaluation/control shaders, this semantic label indicates the
3306 number of vertices provided in the input patch. Only the X value is defined.
3308 TGSI_SEMANTIC_HELPER_INVOCATION
3309 """""""""""""""""""""""""""""""
3311 For fragment shaders, this semantic indicates whether the current
3312 invocation is covered or not. Helper invocations are created in order
3313 to properly compute derivatives, however it may be desirable to skip
3314 some of the logic in those cases. See ``gl_HelperInvocation`` documentation.
3316 TGSI_SEMANTIC_BASEINSTANCE
3317 """"""""""""""""""""""""""
3319 For vertex shaders, the base instance argument supplied for this
3320 draw. This is an integer value, and only the X component is used.
3322 TGSI_SEMANTIC_DRAWID
3323 """"""""""""""""""""
3325 For vertex shaders, the zero-based index of the current draw in a
3326 ``glMultiDraw*`` invocation. This is an integer value, and only the X
3330 TGSI_SEMANTIC_WORK_DIM
3331 """"""""""""""""""""""
3333 For compute shaders started via opencl this retrieves the work_dim
3334 parameter to the clEnqueueNDRangeKernel call with which the shader
3338 TGSI_SEMANTIC_GRID_SIZE
3339 """""""""""""""""""""""
3341 For compute shaders, this semantic indicates the maximum (x, y, z) dimensions
3342 of a grid of thread blocks.
3345 TGSI_SEMANTIC_BLOCK_ID
3346 """"""""""""""""""""""
3348 For compute shaders, this semantic indicates the (x, y, z) coordinates of the
3349 current block inside of the grid.
3352 TGSI_SEMANTIC_BLOCK_SIZE
3353 """"""""""""""""""""""""
3355 For compute shaders, this semantic indicates the maximum (x, y, z) dimensions
3356 of a block in threads.
3359 TGSI_SEMANTIC_THREAD_ID
3360 """""""""""""""""""""""
3362 For compute shaders, this semantic indicates the (x, y, z) coordinates of the
3363 current thread inside of the block.
3366 TGSI_SEMANTIC_SUBGROUP_SIZE
3367 """""""""""""""""""""""""""
3369 This semantic indicates the subgroup size for the current invocation. This is
3370 an integer of at most 64, as it indicates the width of lanemasks. It does not
3371 depend on the number of invocations that are active.
3374 TGSI_SEMANTIC_SUBGROUP_INVOCATION
3375 """""""""""""""""""""""""""""""""
3377 The index of the current invocation within its subgroup.
3380 TGSI_SEMANTIC_SUBGROUP_EQ_MASK
3381 """"""""""""""""""""""""""""""
3383 A bit mask of ``bit index == TGSI_SEMANTIC_SUBGROUP_INVOCATION``, i.e.
3384 ``1 << subgroup_invocation`` in arbitrary precision arithmetic.
3387 TGSI_SEMANTIC_SUBGROUP_GE_MASK
3388 """"""""""""""""""""""""""""""
3390 A bit mask of ``bit index >= TGSI_SEMANTIC_SUBGROUP_INVOCATION``, i.e.
3391 ``((1 << (subgroup_size - subgroup_invocation)) - 1) << subgroup_invocation``
3392 in arbitrary precision arithmetic.
3395 TGSI_SEMANTIC_SUBGROUP_GT_MASK
3396 """"""""""""""""""""""""""""""
3398 A bit mask of ``bit index > TGSI_SEMANTIC_SUBGROUP_INVOCATION``, i.e.
3399 ``((1 << (subgroup_size - subgroup_invocation - 1)) - 1) << (subgroup_invocation + 1)``
3400 in arbitrary precision arithmetic.
3403 TGSI_SEMANTIC_SUBGROUP_LE_MASK
3404 """"""""""""""""""""""""""""""
3406 A bit mask of ``bit index <= TGSI_SEMANTIC_SUBGROUP_INVOCATION``, i.e.
3407 ``(1 << (subgroup_invocation + 1)) - 1`` in arbitrary precision arithmetic.
3410 TGSI_SEMANTIC_SUBGROUP_LT_MASK
3411 """"""""""""""""""""""""""""""
3413 A bit mask of ``bit index > TGSI_SEMANTIC_SUBGROUP_INVOCATION``, i.e.
3414 ``(1 << subgroup_invocation) - 1`` in arbitrary precision arithmetic.
3417 Declaration Interpolate
3418 ^^^^^^^^^^^^^^^^^^^^^^^
3420 This token is only valid for fragment shader INPUT declarations.
3422 The Interpolate field specifes the way input is being interpolated by
3423 the rasteriser and is one of TGSI_INTERPOLATE_*.
3425 The Location field specifies the location inside the pixel that the
3426 interpolation should be done at, one of ``TGSI_INTERPOLATE_LOC_*``. Note that
3427 when per-sample shading is enabled, the implementation may choose to
3428 interpolate at the sample irrespective of the Location field.
3430 The CylindricalWrap bitfield specifies which register components
3431 should be subject to cylindrical wrapping when interpolating by the
3432 rasteriser. If TGSI_CYLINDRICAL_WRAP_X is set to 1, the X component
3433 should be interpolated according to cylindrical wrapping rules.
3436 Declaration Sampler View
3437 ^^^^^^^^^^^^^^^^^^^^^^^^
3439 Follows Declaration token if file is TGSI_FILE_SAMPLER_VIEW.
3441 DCL SVIEW[#], resource, type(s)
3443 Declares a shader input sampler view and assigns it to a SVIEW[#]
3446 resource can be one of BUFFER, 1D, 2D, 3D, 1DArray and 2DArray.
3448 type must be 1 or 4 entries (if specifying on a per-component
3449 level) out of UNORM, SNORM, SINT, UINT and FLOAT.
3451 For TEX\* style texture sample opcodes (as opposed to SAMPLE\* opcodes
3452 which take an explicit SVIEW[#] source register), there may be optionally
3453 SVIEW[#] declarations. In this case, the SVIEW index is implied by the
3454 SAMP index, and there must be a corresponding SVIEW[#] declaration for
3455 each SAMP[#] declaration. Drivers are free to ignore this if they wish.
3456 But note in particular that some drivers need to know the sampler type
3457 (float/int/unsigned) in order to generate the correct code, so cases
3458 where integer textures are sampled, SVIEW[#] declarations should be
3461 NOTE: It is NOT legal to mix SAMPLE\* style opcodes and TEX\* opcodes
3464 Declaration Resource
3465 ^^^^^^^^^^^^^^^^^^^^
3467 Follows Declaration token if file is TGSI_FILE_RESOURCE.
3469 DCL RES[#], resource [, WR] [, RAW]
3471 Declares a shader input resource and assigns it to a RES[#]
3474 resource can be one of BUFFER, 1D, 2D, 3D, CUBE, 1DArray and
3477 If the RAW keyword is not specified, the texture data will be
3478 subject to conversion, swizzling and scaling as required to yield
3479 the specified data type from the physical data format of the bound
3482 If the RAW keyword is specified, no channel conversion will be
3483 performed: the values read for each of the channels (X,Y,Z,W) will
3484 correspond to consecutive words in the same order and format
3485 they're found in memory. No element-to-address conversion will be
3486 performed either: the value of the provided X coordinate will be
3487 interpreted in byte units instead of texel units. The result of
3488 accessing a misaligned address is undefined.
3490 Usage of the STORE opcode is only allowed if the WR (writable) flag
3495 ^^^^^^^^^^^^^^^^^^^^^^^^
3497 Properties are general directives that apply to the whole TGSI program.
3502 Specifies the fragment shader TGSI_SEMANTIC_POSITION coordinate origin.
3503 The default value is UPPER_LEFT.
3505 If UPPER_LEFT, the position will be (0,0) at the upper left corner and
3506 increase downward and rightward.
3507 If LOWER_LEFT, the position will be (0,0) at the lower left corner and
3508 increase upward and rightward.
3510 OpenGL defaults to LOWER_LEFT, and is configurable with the
3511 GL_ARB_fragment_coord_conventions extension.
3513 DirectX 9/10 use UPPER_LEFT.
3515 FS_COORD_PIXEL_CENTER
3516 """""""""""""""""""""
3518 Specifies the fragment shader TGSI_SEMANTIC_POSITION pixel center convention.
3519 The default value is HALF_INTEGER.
3521 If HALF_INTEGER, the fractionary part of the position will be 0.5
3522 If INTEGER, the fractionary part of the position will be 0.0
3524 Note that this does not affect the set of fragments generated by
3525 rasterization, which is instead controlled by half_pixel_center in the
3528 OpenGL defaults to HALF_INTEGER, and is configurable with the
3529 GL_ARB_fragment_coord_conventions extension.
3531 DirectX 9 uses INTEGER.
3532 DirectX 10 uses HALF_INTEGER.
3534 FS_COLOR0_WRITES_ALL_CBUFS
3535 """"""""""""""""""""""""""
3536 Specifies that writes to the fragment shader color 0 are replicated to all
3537 bound cbufs. This facilitates OpenGL's fragColor output vs fragData[0] where
3538 fragData is directed to a single color buffer, but fragColor is broadcast.
3541 """"""""""""""""""""""""""
3542 If this property is set on the program bound to the shader stage before the
3543 fragment shader, user clip planes should have no effect (be disabled) even if
3544 that shader does not write to any clip distance outputs and the rasterizer's
3545 clip_plane_enable is non-zero.
3546 This property is only supported by drivers that also support shader clip
3548 This is useful for APIs that don't have UCPs and where clip distances written
3549 by a shader cannot be disabled.
3554 Specifies the number of times a geometry shader should be executed for each
3555 input primitive. Each invocation will have a different
3556 TGSI_SEMANTIC_INVOCATIONID system value set. If not specified, assumed to
3559 VS_WINDOW_SPACE_POSITION
3560 """"""""""""""""""""""""""
3561 If this property is set on the vertex shader, the TGSI_SEMANTIC_POSITION output
3562 is assumed to contain window space coordinates.
3563 Division of X,Y,Z by W and the viewport transformation are disabled, and 1/W is
3564 directly taken from the 4-th component of the shader output.
3565 Naturally, clipping is not performed on window coordinates either.
3566 The effect of this property is undefined if a geometry or tessellation shader
3572 The number of vertices written by the tessellation control shader. This
3573 effectively defines the patch input size of the tessellation evaluation shader
3579 This sets the tessellation primitive mode, one of ``PIPE_PRIM_TRIANGLES``,
3580 ``PIPE_PRIM_QUADS``, or ``PIPE_PRIM_LINES``. (Unlike in GL, there is no
3581 separate isolines settings, the regular lines is assumed to mean isolines.)
3586 This sets the spacing mode of the tessellation generator, one of
3587 ``PIPE_TESS_SPACING_*``.
3592 This sets the vertex order to be clockwise if the value is 1, or
3593 counter-clockwise if set to 0.
3598 If set to a non-zero value, this turns on point mode for the tessellator,
3599 which means that points will be generated instead of primitives.
3601 NUM_CLIPDIST_ENABLED
3602 """"""""""""""""""""
3604 How many clip distance scalar outputs are enabled.
3606 NUM_CULLDIST_ENABLED
3607 """"""""""""""""""""
3609 How many cull distance scalar outputs are enabled.
3611 FS_EARLY_DEPTH_STENCIL
3612 """"""""""""""""""""""
3614 Whether depth test, stencil test, and occlusion query should run before
3615 the fragment shader (regardless of fragment shader side effects). Corresponds
3616 to GLSL early_fragment_tests.
3621 Which shader stage will MOST LIKELY follow after this shader when the shader
3622 is bound. This is only a hint to the driver and doesn't have to be precise.
3623 Only set for VS and TES.
3625 CS_FIXED_BLOCK_WIDTH / HEIGHT / DEPTH
3626 """""""""""""""""""""""""""""""""""""
3628 Threads per block in each dimension, if known at compile time. If the block size
3629 is known all three should be at least 1. If it is unknown they should all be set
3635 The MUL TGSI operation (FP32 multiplication) will return 0 if either
3636 of the operands are equal to 0. That means that 0 * Inf = 0. This
3637 should be set the same way for an entire pipeline. Note that this
3638 applies not only to the literal MUL TGSI opcode, but all FP32
3639 multiplications implied by other operations, such as MAD, FMA, DP2,
3640 DP3, DP4, DST, LOG, LRP, and possibly others. If there is a
3641 mismatch between shaders, then it is unspecified whether this behavior
3644 FS_POST_DEPTH_COVERAGE
3645 """"""""""""""""""""""
3647 When enabled, the input for TGSI_SEMANTIC_SAMPLEMASK will exclude samples
3648 that have failed the depth/stencil tests. This is only valid when
3649 FS_EARLY_DEPTH_STENCIL is also specified.
3652 Texture Sampling and Texture Formats
3653 ------------------------------------
3655 This table shows how texture image components are returned as (x,y,z,w) tuples
3656 by TGSI texture instructions, such as :opcode:`TEX`, :opcode:`TXD`, and
3657 :opcode:`TXP`. For reference, OpenGL and Direct3D conventions are shown as
3660 +--------------------+--------------+--------------------+--------------+
3661 | Texture Components | Gallium | OpenGL | Direct3D 9 |
3662 +====================+==============+====================+==============+
3663 | R | (r, 0, 0, 1) | (r, 0, 0, 1) | (r, 1, 1, 1) |
3664 +--------------------+--------------+--------------------+--------------+
3665 | RG | (r, g, 0, 1) | (r, g, 0, 1) | (r, g, 1, 1) |
3666 +--------------------+--------------+--------------------+--------------+
3667 | RGB | (r, g, b, 1) | (r, g, b, 1) | (r, g, b, 1) |
3668 +--------------------+--------------+--------------------+--------------+
3669 | RGBA | (r, g, b, a) | (r, g, b, a) | (r, g, b, a) |
3670 +--------------------+--------------+--------------------+--------------+
3671 | A | (0, 0, 0, a) | (0, 0, 0, a) | (0, 0, 0, a) |
3672 +--------------------+--------------+--------------------+--------------+
3673 | L | (l, l, l, 1) | (l, l, l, 1) | (l, l, l, 1) |
3674 +--------------------+--------------+--------------------+--------------+
3675 | LA | (l, l, l, a) | (l, l, l, a) | (l, l, l, a) |
3676 +--------------------+--------------+--------------------+--------------+
3677 | I | (i, i, i, i) | (i, i, i, i) | N/A |
3678 +--------------------+--------------+--------------------+--------------+
3679 | UV | XXX TBD | (0, 0, 0, 1) | (u, v, 1, 1) |
3680 | | | [#envmap-bumpmap]_ | |
3681 +--------------------+--------------+--------------------+--------------+
3682 | Z | XXX TBD | (z, z, z, 1) | (0, z, 0, 1) |
3683 | | | [#depth-tex-mode]_ | |
3684 +--------------------+--------------+--------------------+--------------+
3685 | S | (s, s, s, s) | unknown | unknown |
3686 +--------------------+--------------+--------------------+--------------+
3688 .. [#envmap-bumpmap] http://www.opengl.org/registry/specs/ATI/envmap_bumpmap.txt
3689 .. [#depth-tex-mode] the default is (z, z, z, 1) but may also be (0, 0, 0, z)
3690 or (z, z, z, z) depending on the value of GL_DEPTH_TEXTURE_MODE.