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:: DP2A - 2-component Dot Product And Add
292 dst.x = src0.x \times src1.x + src0.y \times src1.y + src2.x
294 dst.y = src0.x \times src1.x + src0.y \times src1.y + src2.x
296 dst.z = src0.x \times src1.x + src0.y \times src1.y + src2.x
298 dst.w = src0.x \times src1.x + src0.y \times src1.y + src2.x
301 .. opcode:: FRC - Fraction
305 dst.x = src.x - \lfloor src.x\rfloor
307 dst.y = src.y - \lfloor src.y\rfloor
309 dst.z = src.z - \lfloor src.z\rfloor
311 dst.w = src.w - \lfloor src.w\rfloor
314 .. opcode:: FLR - Floor
318 dst.x = \lfloor src.x\rfloor
320 dst.y = \lfloor src.y\rfloor
322 dst.z = \lfloor src.z\rfloor
324 dst.w = \lfloor src.w\rfloor
327 .. opcode:: ROUND - Round
340 .. opcode:: EX2 - Exponential Base 2
342 This instruction replicates its result.
349 .. opcode:: LG2 - Logarithm Base 2
351 This instruction replicates its result.
358 .. opcode:: POW - Power
360 This instruction replicates its result.
364 dst = src0.x^{src1.x}
366 .. opcode:: XPD - Cross Product
370 dst.x = src0.y \times src1.z - src1.y \times src0.z
372 dst.y = src0.z \times src1.x - src1.z \times src0.x
374 dst.z = src0.x \times src1.y - src1.x \times src0.y
379 .. opcode:: DPH - Homogeneous Dot Product
381 This instruction replicates its result.
385 dst = src0.x \times src1.x + src0.y \times src1.y + src0.z \times src1.z + src1.w
388 .. opcode:: COS - Cosine
390 This instruction replicates its result.
397 .. opcode:: DDX, DDX_FINE - Derivative Relative To X
399 The fine variant is only used when ``PIPE_CAP_TGSI_FS_FINE_DERIVATIVE`` is
400 advertised. When it is, the fine version guarantees one derivative per row
401 while DDX is allowed to be the same for the entire 2x2 quad.
405 dst.x = partialx(src.x)
407 dst.y = partialx(src.y)
409 dst.z = partialx(src.z)
411 dst.w = partialx(src.w)
414 .. opcode:: DDY, DDY_FINE - Derivative Relative To Y
416 The fine variant is only used when ``PIPE_CAP_TGSI_FS_FINE_DERIVATIVE`` is
417 advertised. When it is, the fine version guarantees one derivative per column
418 while DDY is allowed to be the same for the entire 2x2 quad.
422 dst.x = partialy(src.x)
424 dst.y = partialy(src.y)
426 dst.z = partialy(src.z)
428 dst.w = partialy(src.w)
431 .. opcode:: PK2H - Pack Two 16-bit Floats
433 This instruction replicates its result.
437 dst = f32\_to\_f16(src.x) | f32\_to\_f16(src.y) << 16
440 .. opcode:: PK2US - Pack Two Unsigned 16-bit Scalars
445 .. opcode:: PK4B - Pack Four Signed 8-bit Scalars
450 .. opcode:: PK4UB - Pack Four Unsigned 8-bit Scalars
455 .. opcode:: SEQ - Set On Equal
459 dst.x = (src0.x == src1.x) ? 1.0F : 0.0F
461 dst.y = (src0.y == src1.y) ? 1.0F : 0.0F
463 dst.z = (src0.z == src1.z) ? 1.0F : 0.0F
465 dst.w = (src0.w == src1.w) ? 1.0F : 0.0F
468 .. opcode:: SGT - Set On Greater Than
472 dst.x = (src0.x > src1.x) ? 1.0F : 0.0F
474 dst.y = (src0.y > src1.y) ? 1.0F : 0.0F
476 dst.z = (src0.z > src1.z) ? 1.0F : 0.0F
478 dst.w = (src0.w > src1.w) ? 1.0F : 0.0F
481 .. opcode:: SIN - Sine
483 This instruction replicates its result.
490 .. opcode:: SLE - Set On Less Equal Than
494 dst.x = (src0.x <= src1.x) ? 1.0F : 0.0F
496 dst.y = (src0.y <= src1.y) ? 1.0F : 0.0F
498 dst.z = (src0.z <= src1.z) ? 1.0F : 0.0F
500 dst.w = (src0.w <= src1.w) ? 1.0F : 0.0F
503 .. opcode:: SNE - Set On Not Equal
507 dst.x = (src0.x != src1.x) ? 1.0F : 0.0F
509 dst.y = (src0.y != src1.y) ? 1.0F : 0.0F
511 dst.z = (src0.z != src1.z) ? 1.0F : 0.0F
513 dst.w = (src0.w != src1.w) ? 1.0F : 0.0F
516 .. opcode:: TEX - Texture Lookup
518 for array textures src0.y contains the slice for 1D,
519 and src0.z contain the slice for 2D.
521 for shadow textures with no arrays (and not cube map),
522 src0.z contains the reference value.
524 for shadow textures with arrays, src0.z contains
525 the reference value for 1D arrays, and src0.w contains
526 the reference value for 2D arrays and cube maps.
528 for cube map array shadow textures, the reference value
529 cannot be passed in src0.w, and TEX2 must be used instead.
535 shadow_ref = src0.z or src0.w (optional)
539 dst = texture\_sample(unit, coord, shadow_ref)
542 .. opcode:: TEX2 - Texture Lookup (for shadow cube map arrays only)
544 this is the same as TEX, but uses another reg to encode the
555 dst = texture\_sample(unit, coord, shadow_ref)
560 .. opcode:: TXD - Texture Lookup with Derivatives
572 dst = texture\_sample\_deriv(unit, coord, ddx, ddy)
575 .. opcode:: TXP - Projective Texture Lookup
579 coord.x = src0.x / src0.w
581 coord.y = src0.y / src0.w
583 coord.z = src0.z / src0.w
589 dst = texture\_sample(unit, coord)
592 .. opcode:: UP2H - Unpack Two 16-Bit Floats
596 dst.x = f16\_to\_f32(src0.x \& 0xffff)
598 dst.y = f16\_to\_f32(src0.x >> 16)
600 dst.z = f16\_to\_f32(src0.x \& 0xffff)
602 dst.w = f16\_to\_f32(src0.x >> 16)
606 Considered for removal.
608 .. opcode:: UP2US - Unpack Two Unsigned 16-Bit Scalars
614 Considered for removal.
616 .. opcode:: UP4B - Unpack Four Signed 8-Bit Values
622 Considered for removal.
624 .. opcode:: UP4UB - Unpack Four Unsigned 8-Bit Scalars
630 Considered for removal.
633 .. opcode:: ARR - Address Register Load With Round
637 dst.x = (int) round(src.x)
639 dst.y = (int) round(src.y)
641 dst.z = (int) round(src.z)
643 dst.w = (int) round(src.w)
646 .. opcode:: SSG - Set Sign
650 dst.x = (src.x > 0) ? 1 : (src.x < 0) ? -1 : 0
652 dst.y = (src.y > 0) ? 1 : (src.y < 0) ? -1 : 0
654 dst.z = (src.z > 0) ? 1 : (src.z < 0) ? -1 : 0
656 dst.w = (src.w > 0) ? 1 : (src.w < 0) ? -1 : 0
659 .. opcode:: CMP - Compare
663 dst.x = (src0.x < 0) ? src1.x : src2.x
665 dst.y = (src0.y < 0) ? src1.y : src2.y
667 dst.z = (src0.z < 0) ? src1.z : src2.z
669 dst.w = (src0.w < 0) ? src1.w : src2.w
672 .. opcode:: KILL_IF - Conditional Discard
674 Conditional discard. Allowed in fragment shaders only.
678 if (src.x < 0 || src.y < 0 || src.z < 0 || src.w < 0)
683 .. opcode:: KILL - Discard
685 Unconditional discard. Allowed in fragment shaders only.
688 .. opcode:: SCS - Sine Cosine
701 .. opcode:: TXB - Texture Lookup With Bias
703 for cube map array textures and shadow cube maps, the bias value
704 cannot be passed in src0.w, and TXB2 must be used instead.
706 if the target is a shadow texture, the reference value is always
707 in src.z (this prevents shadow 3d and shadow 2d arrays from
708 using this instruction, but this is not needed).
724 dst = texture\_sample(unit, coord, bias)
727 .. opcode:: TXB2 - Texture Lookup With Bias (some cube maps only)
729 this is the same as TXB, but uses another reg to encode the
730 lod bias value for cube map arrays and shadow cube maps.
731 Presumably shadow 2d arrays and shadow 3d targets could use
732 this encoding too, but this is not legal.
734 shadow cube map arrays are neither possible nor required.
744 dst = texture\_sample(unit, coord, bias)
747 .. opcode:: DIV - Divide
751 dst.x = \frac{src0.x}{src1.x}
753 dst.y = \frac{src0.y}{src1.y}
755 dst.z = \frac{src0.z}{src1.z}
757 dst.w = \frac{src0.w}{src1.w}
760 .. opcode:: DP2 - 2-component Dot Product
762 This instruction replicates its result.
766 dst = src0.x \times src1.x + src0.y \times src1.y
769 .. opcode:: TEX_LZ - Texture Lookup With LOD = 0
771 This is the same as TXL with LOD = 0. Like every texture opcode, it obeys
772 pipe_sampler_view::u.tex.first_level and pipe_sampler_state::min_lod.
773 There is no way to override those two in shaders.
789 dst = texture\_sample(unit, coord, lod)
792 .. opcode:: TXL - Texture Lookup With explicit LOD
794 for cube map array textures, the explicit lod value
795 cannot be passed in src0.w, and TXL2 must be used instead.
797 if the target is a shadow texture, the reference value is always
798 in src.z (this prevents shadow 3d / 2d array / cube targets from
799 using this instruction, but this is not needed).
815 dst = texture\_sample(unit, coord, lod)
818 .. opcode:: TXL2 - Texture Lookup With explicit LOD (for cube map arrays only)
820 this is the same as TXL, but uses another reg to encode the
822 Presumably shadow 3d / 2d array / cube targets could use
823 this encoding too, but this is not legal.
825 shadow cube map arrays are neither possible nor required.
835 dst = texture\_sample(unit, coord, lod)
838 .. opcode:: CALLNZ - Subroutine Call If Not Zero
844 Considered for cleanup.
848 Considered for removal.
852 ^^^^^^^^^^^^^^^^^^^^^^^^
854 These opcodes are primarily provided for special-use computational shaders.
855 Support for these opcodes indicated by a special pipe capability bit (TBD).
857 XXX doesn't look like most of the opcodes really belong here.
859 .. opcode:: CEIL - Ceiling
863 dst.x = \lceil src.x\rceil
865 dst.y = \lceil src.y\rceil
867 dst.z = \lceil src.z\rceil
869 dst.w = \lceil src.w\rceil
872 .. opcode:: TRUNC - Truncate
885 .. opcode:: MOD - Modulus
889 dst.x = src0.x \bmod src1.x
891 dst.y = src0.y \bmod src1.y
893 dst.z = src0.z \bmod src1.z
895 dst.w = src0.w \bmod src1.w
898 .. opcode:: UARL - Integer Address Register Load
900 Moves the contents of the source register, assumed to be an integer, into the
901 destination register, which is assumed to be an address (ADDR) register.
904 .. opcode:: TXF - Texel Fetch
906 As per NV_gpu_shader4, extract a single texel from a specified texture
907 image or PIPE_BUFFER resource. The source sampler may not be a CUBE or
909 four-component signed integer vector used to identify the single texel
910 accessed. 3 components + level. If the texture is multisampled, then
911 the fourth component indicates the sample, not the mipmap level.
912 Just like texture instructions, an optional
913 offset vector is provided, which is subject to various driver restrictions
914 (regarding range, source of offsets). This instruction ignores the sampler
917 TXF(uint_vec coord, int_vec offset).
920 .. opcode:: TXQ - Texture Size Query
922 As per NV_gpu_program4, retrieve the dimensions of the texture depending on
923 the target. For 1D (width), 2D/RECT/CUBE (width, height), 3D (width, height,
924 depth), 1D array (width, layers), 2D array (width, height, layers).
925 Also return the number of accessible levels (last_level - first_level + 1)
928 For components which don't return a resource dimension, their value
935 dst.x = texture\_width(unit, lod)
937 dst.y = texture\_height(unit, lod)
939 dst.z = texture\_depth(unit, lod)
941 dst.w = texture\_levels(unit)
944 .. opcode:: TXQS - Texture Samples Query
946 This retrieves the number of samples in the texture, and stores it
947 into the x component as an unsigned integer. The other components are
948 undefined. If the texture is not multisampled, this function returns
949 (1, undef, undef, undef).
953 dst.x = texture\_samples(unit)
956 .. opcode:: TG4 - Texture Gather
958 As per ARB_texture_gather, gathers the four texels to be used in a bi-linear
959 filtering operation and packs them into a single register. Only works with
960 2D, 2D array, cubemaps, and cubemaps arrays. For 2D textures, only the
961 addressing modes of the sampler and the top level of any mip pyramid are
962 used. Set W to zero. It behaves like the TEX instruction, but a filtered
963 sample is not generated. The four samples that contribute to filtering are
964 placed into xyzw in clockwise order, starting with the (u,v) texture
965 coordinate delta at the following locations (-, +), (+, +), (+, -), (-, -),
966 where the magnitude of the deltas are half a texel.
968 PIPE_CAP_TEXTURE_SM5 enhances this instruction to support shadow per-sample
969 depth compares, single component selection, and a non-constant offset. It
970 doesn't allow support for the GL independent offset to get i0,j0. This would
971 require another CAP is hw can do it natively. For now we lower that before
980 dst = texture\_gather4 (unit, coord, component)
982 (with SM5 - cube array shadow)
990 dst = texture\_gather (uint, coord, compare)
992 .. opcode:: LODQ - level of detail query
994 Compute the LOD information that the texture pipe would use to access the
995 texture. The Y component contains the computed LOD lambda_prime. The X
996 component contains the LOD that will be accessed, based on min/max lod's
1003 dst.xy = lodq(uint, coord);
1005 .. opcode:: CLOCK - retrieve the current shader time
1007 Invoking this instruction multiple times in the same shader should
1008 cause monotonically increasing values to be returned. The values
1009 are implicitly 64-bit, so if fewer than 64 bits of precision are
1010 available, to provide expected wraparound semantics, the value
1011 should be shifted up so that the most significant bit of the time
1012 is the most significant bit of the 64-bit value.
1020 ^^^^^^^^^^^^^^^^^^^^^^^^
1021 These opcodes are used for integer operations.
1022 Support for these opcodes indicated by PIPE_SHADER_CAP_INTEGERS (all of them?)
1025 .. opcode:: I2F - Signed Integer To Float
1027 Rounding is unspecified (round to nearest even suggested).
1031 dst.x = (float) src.x
1033 dst.y = (float) src.y
1035 dst.z = (float) src.z
1037 dst.w = (float) src.w
1040 .. opcode:: U2F - Unsigned Integer To Float
1042 Rounding is unspecified (round to nearest even suggested).
1046 dst.x = (float) src.x
1048 dst.y = (float) src.y
1050 dst.z = (float) src.z
1052 dst.w = (float) src.w
1055 .. opcode:: F2I - Float to Signed Integer
1057 Rounding is towards zero (truncate).
1058 Values outside signed range (including NaNs) produce undefined results.
1071 .. opcode:: F2U - Float to Unsigned Integer
1073 Rounding is towards zero (truncate).
1074 Values outside unsigned range (including NaNs) produce undefined results.
1078 dst.x = (unsigned) src.x
1080 dst.y = (unsigned) src.y
1082 dst.z = (unsigned) src.z
1084 dst.w = (unsigned) src.w
1087 .. opcode:: UADD - Integer Add
1089 This instruction works the same for signed and unsigned integers.
1090 The low 32bit of the result is returned.
1094 dst.x = src0.x + src1.x
1096 dst.y = src0.y + src1.y
1098 dst.z = src0.z + src1.z
1100 dst.w = src0.w + src1.w
1103 .. opcode:: UMAD - Integer Multiply And Add
1105 This instruction works the same for signed and unsigned integers.
1106 The multiplication returns the low 32bit (as does the result itself).
1110 dst.x = src0.x \times src1.x + src2.x
1112 dst.y = src0.y \times src1.y + src2.y
1114 dst.z = src0.z \times src1.z + src2.z
1116 dst.w = src0.w \times src1.w + src2.w
1119 .. opcode:: UMUL - Integer Multiply
1121 This instruction works the same for signed and unsigned integers.
1122 The low 32bit of the result is returned.
1126 dst.x = src0.x \times src1.x
1128 dst.y = src0.y \times src1.y
1130 dst.z = src0.z \times src1.z
1132 dst.w = src0.w \times src1.w
1135 .. opcode:: IMUL_HI - Signed Integer Multiply High Bits
1137 The high 32bits of the multiplication of 2 signed integers are returned.
1141 dst.x = (src0.x \times src1.x) >> 32
1143 dst.y = (src0.y \times src1.y) >> 32
1145 dst.z = (src0.z \times src1.z) >> 32
1147 dst.w = (src0.w \times src1.w) >> 32
1150 .. opcode:: UMUL_HI - Unsigned Integer Multiply High Bits
1152 The high 32bits of the multiplication of 2 unsigned integers are returned.
1156 dst.x = (src0.x \times src1.x) >> 32
1158 dst.y = (src0.y \times src1.y) >> 32
1160 dst.z = (src0.z \times src1.z) >> 32
1162 dst.w = (src0.w \times src1.w) >> 32
1165 .. opcode:: IDIV - Signed Integer Division
1167 TBD: behavior for division by zero.
1171 dst.x = \frac{src0.x}{src1.x}
1173 dst.y = \frac{src0.y}{src1.y}
1175 dst.z = \frac{src0.z}{src1.z}
1177 dst.w = \frac{src0.w}{src1.w}
1180 .. opcode:: UDIV - Unsigned Integer Division
1182 For division by zero, 0xffffffff is returned.
1186 dst.x = \frac{src0.x}{src1.x}
1188 dst.y = \frac{src0.y}{src1.y}
1190 dst.z = \frac{src0.z}{src1.z}
1192 dst.w = \frac{src0.w}{src1.w}
1195 .. opcode:: UMOD - Unsigned Integer Remainder
1197 If second arg is zero, 0xffffffff is returned.
1201 dst.x = src0.x \bmod src1.x
1203 dst.y = src0.y \bmod src1.y
1205 dst.z = src0.z \bmod src1.z
1207 dst.w = src0.w \bmod src1.w
1210 .. opcode:: NOT - Bitwise Not
1223 .. opcode:: AND - Bitwise And
1227 dst.x = src0.x \& src1.x
1229 dst.y = src0.y \& src1.y
1231 dst.z = src0.z \& src1.z
1233 dst.w = src0.w \& src1.w
1236 .. opcode:: OR - Bitwise Or
1240 dst.x = src0.x | src1.x
1242 dst.y = src0.y | src1.y
1244 dst.z = src0.z | src1.z
1246 dst.w = src0.w | src1.w
1249 .. opcode:: XOR - Bitwise Xor
1253 dst.x = src0.x \oplus src1.x
1255 dst.y = src0.y \oplus src1.y
1257 dst.z = src0.z \oplus src1.z
1259 dst.w = src0.w \oplus src1.w
1262 .. opcode:: IMAX - Maximum of Signed Integers
1266 dst.x = max(src0.x, src1.x)
1268 dst.y = max(src0.y, src1.y)
1270 dst.z = max(src0.z, src1.z)
1272 dst.w = max(src0.w, src1.w)
1275 .. opcode:: UMAX - Maximum of Unsigned Integers
1279 dst.x = max(src0.x, src1.x)
1281 dst.y = max(src0.y, src1.y)
1283 dst.z = max(src0.z, src1.z)
1285 dst.w = max(src0.w, src1.w)
1288 .. opcode:: IMIN - Minimum of Signed Integers
1292 dst.x = min(src0.x, src1.x)
1294 dst.y = min(src0.y, src1.y)
1296 dst.z = min(src0.z, src1.z)
1298 dst.w = min(src0.w, src1.w)
1301 .. opcode:: UMIN - Minimum of Unsigned Integers
1305 dst.x = min(src0.x, src1.x)
1307 dst.y = min(src0.y, src1.y)
1309 dst.z = min(src0.z, src1.z)
1311 dst.w = min(src0.w, src1.w)
1314 .. opcode:: SHL - Shift Left
1316 The shift count is masked with 0x1f before the shift is applied.
1320 dst.x = src0.x << (0x1f \& src1.x)
1322 dst.y = src0.y << (0x1f \& src1.y)
1324 dst.z = src0.z << (0x1f \& src1.z)
1326 dst.w = src0.w << (0x1f \& src1.w)
1329 .. opcode:: ISHR - Arithmetic Shift Right (of Signed Integer)
1331 The shift count is masked with 0x1f before the shift is applied.
1335 dst.x = src0.x >> (0x1f \& src1.x)
1337 dst.y = src0.y >> (0x1f \& src1.y)
1339 dst.z = src0.z >> (0x1f \& src1.z)
1341 dst.w = src0.w >> (0x1f \& src1.w)
1344 .. opcode:: USHR - Logical Shift Right
1346 The shift count is masked with 0x1f before the shift is applied.
1350 dst.x = src0.x >> (unsigned) (0x1f \& src1.x)
1352 dst.y = src0.y >> (unsigned) (0x1f \& src1.y)
1354 dst.z = src0.z >> (unsigned) (0x1f \& src1.z)
1356 dst.w = src0.w >> (unsigned) (0x1f \& src1.w)
1359 .. opcode:: UCMP - Integer Conditional Move
1363 dst.x = src0.x ? src1.x : src2.x
1365 dst.y = src0.y ? src1.y : src2.y
1367 dst.z = src0.z ? src1.z : src2.z
1369 dst.w = src0.w ? src1.w : src2.w
1373 .. opcode:: ISSG - Integer Set Sign
1377 dst.x = (src0.x < 0) ? -1 : (src0.x > 0) ? 1 : 0
1379 dst.y = (src0.y < 0) ? -1 : (src0.y > 0) ? 1 : 0
1381 dst.z = (src0.z < 0) ? -1 : (src0.z > 0) ? 1 : 0
1383 dst.w = (src0.w < 0) ? -1 : (src0.w > 0) ? 1 : 0
1387 .. opcode:: FSLT - Float Set On Less Than (ordered)
1389 Same comparison as SLT but returns integer instead of 1.0/0.0 float
1393 dst.x = (src0.x < src1.x) ? \sim 0 : 0
1395 dst.y = (src0.y < src1.y) ? \sim 0 : 0
1397 dst.z = (src0.z < src1.z) ? \sim 0 : 0
1399 dst.w = (src0.w < src1.w) ? \sim 0 : 0
1402 .. opcode:: ISLT - Signed Integer Set On Less Than
1406 dst.x = (src0.x < src1.x) ? \sim 0 : 0
1408 dst.y = (src0.y < src1.y) ? \sim 0 : 0
1410 dst.z = (src0.z < src1.z) ? \sim 0 : 0
1412 dst.w = (src0.w < src1.w) ? \sim 0 : 0
1415 .. opcode:: USLT - Unsigned Integer Set On Less Than
1419 dst.x = (src0.x < src1.x) ? \sim 0 : 0
1421 dst.y = (src0.y < src1.y) ? \sim 0 : 0
1423 dst.z = (src0.z < src1.z) ? \sim 0 : 0
1425 dst.w = (src0.w < src1.w) ? \sim 0 : 0
1428 .. opcode:: FSGE - Float Set On Greater Equal Than (ordered)
1430 Same comparison as SGE but returns integer instead of 1.0/0.0 float
1434 dst.x = (src0.x >= src1.x) ? \sim 0 : 0
1436 dst.y = (src0.y >= src1.y) ? \sim 0 : 0
1438 dst.z = (src0.z >= src1.z) ? \sim 0 : 0
1440 dst.w = (src0.w >= src1.w) ? \sim 0 : 0
1443 .. opcode:: ISGE - Signed Integer Set On Greater Equal Than
1447 dst.x = (src0.x >= src1.x) ? \sim 0 : 0
1449 dst.y = (src0.y >= src1.y) ? \sim 0 : 0
1451 dst.z = (src0.z >= src1.z) ? \sim 0 : 0
1453 dst.w = (src0.w >= src1.w) ? \sim 0 : 0
1456 .. opcode:: USGE - Unsigned Integer Set On Greater Equal Than
1460 dst.x = (src0.x >= src1.x) ? \sim 0 : 0
1462 dst.y = (src0.y >= src1.y) ? \sim 0 : 0
1464 dst.z = (src0.z >= src1.z) ? \sim 0 : 0
1466 dst.w = (src0.w >= src1.w) ? \sim 0 : 0
1469 .. opcode:: FSEQ - Float Set On Equal (ordered)
1471 Same comparison as SEQ but returns integer instead of 1.0/0.0 float
1475 dst.x = (src0.x == src1.x) ? \sim 0 : 0
1477 dst.y = (src0.y == src1.y) ? \sim 0 : 0
1479 dst.z = (src0.z == src1.z) ? \sim 0 : 0
1481 dst.w = (src0.w == src1.w) ? \sim 0 : 0
1484 .. opcode:: USEQ - Integer Set On Equal
1488 dst.x = (src0.x == src1.x) ? \sim 0 : 0
1490 dst.y = (src0.y == src1.y) ? \sim 0 : 0
1492 dst.z = (src0.z == src1.z) ? \sim 0 : 0
1494 dst.w = (src0.w == src1.w) ? \sim 0 : 0
1497 .. opcode:: FSNE - Float Set On Not Equal (unordered)
1499 Same comparison as SNE but returns integer instead of 1.0/0.0 float
1503 dst.x = (src0.x != src1.x) ? \sim 0 : 0
1505 dst.y = (src0.y != src1.y) ? \sim 0 : 0
1507 dst.z = (src0.z != src1.z) ? \sim 0 : 0
1509 dst.w = (src0.w != src1.w) ? \sim 0 : 0
1512 .. opcode:: USNE - Integer Set On Not Equal
1516 dst.x = (src0.x != src1.x) ? \sim 0 : 0
1518 dst.y = (src0.y != src1.y) ? \sim 0 : 0
1520 dst.z = (src0.z != src1.z) ? \sim 0 : 0
1522 dst.w = (src0.w != src1.w) ? \sim 0 : 0
1525 .. opcode:: INEG - Integer Negate
1540 .. opcode:: IABS - Integer Absolute Value
1554 These opcodes are used for bit-level manipulation of integers.
1556 .. opcode:: IBFE - Signed Bitfield Extract
1558 Like GLSL bitfieldExtract. Extracts a set of bits from the input, and
1559 sign-extends them if the high bit of the extracted window is set.
1563 def ibfe(value, offset, bits):
1564 if offset < 0 or bits < 0 or offset + bits > 32:
1566 if bits == 0: return 0
1567 # Note: >> sign-extends
1568 return (value << (32 - offset - bits)) >> (32 - bits)
1570 .. opcode:: UBFE - Unsigned Bitfield Extract
1572 Like GLSL bitfieldExtract. Extracts a set of bits from the input, without
1577 def ubfe(value, offset, bits):
1578 if offset < 0 or bits < 0 or offset + bits > 32:
1580 if bits == 0: return 0
1581 # Note: >> does not sign-extend
1582 return (value << (32 - offset - bits)) >> (32 - bits)
1584 .. opcode:: BFI - Bitfield Insert
1586 Like GLSL bitfieldInsert. Replaces a bit region of 'base' with the low bits
1591 def bfi(base, insert, offset, bits):
1592 if offset < 0 or bits < 0 or offset + bits > 32:
1594 # << defined such that mask == ~0 when bits == 32, offset == 0
1595 mask = ((1 << bits) - 1) << offset
1596 return ((insert << offset) & mask) | (base & ~mask)
1598 .. opcode:: BREV - Bitfield Reverse
1600 See SM5 instruction BFREV. Reverses the bits of the argument.
1602 .. opcode:: POPC - Population Count
1604 See SM5 instruction COUNTBITS. Counts the number of set bits in the argument.
1606 .. opcode:: LSB - Index of lowest set bit
1608 See SM5 instruction FIRSTBIT_LO. Computes the 0-based index of the first set
1609 bit of the argument. Returns -1 if none are set.
1611 .. opcode:: IMSB - Index of highest non-sign bit
1613 See SM5 instruction FIRSTBIT_SHI. Computes the 0-based index of the highest
1614 non-sign bit of the argument (i.e. highest 0 bit for negative numbers,
1615 highest 1 bit for positive numbers). Returns -1 if all bits are the same
1616 (i.e. for inputs 0 and -1).
1618 .. opcode:: UMSB - Index of highest set bit
1620 See SM5 instruction FIRSTBIT_HI. Computes the 0-based index of the highest
1621 set bit of the argument. Returns -1 if none are set.
1624 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1626 These opcodes are only supported in geometry shaders; they have no meaning
1627 in any other type of shader.
1629 .. opcode:: EMIT - Emit
1631 Generate a new vertex for the current primitive into the specified vertex
1632 stream using the values in the output registers.
1635 .. opcode:: ENDPRIM - End Primitive
1637 Complete the current primitive in the specified vertex stream (consisting of
1638 the emitted vertices), and start a new one.
1644 These opcodes are part of :term:`GLSL`'s opcode set. Support for these
1645 opcodes is determined by a special capability bit, ``GLSL``.
1646 Some require glsl version 1.30 (UIF/BREAKC/SWITCH/CASE/DEFAULT/ENDSWITCH).
1648 .. opcode:: CAL - Subroutine Call
1654 .. opcode:: RET - Subroutine Call Return
1659 .. opcode:: CONT - Continue
1661 Unconditionally moves the point of execution to the instruction after the
1662 last bgnloop. The instruction must appear within a bgnloop/endloop.
1666 Support for CONT is determined by a special capability bit,
1667 ``TGSI_CONT_SUPPORTED``. See :ref:`Screen` for more information.
1670 .. opcode:: BGNLOOP - Begin a Loop
1672 Start a loop. Must have a matching endloop.
1675 .. opcode:: BGNSUB - Begin Subroutine
1677 Starts definition of a subroutine. Must have a matching endsub.
1680 .. opcode:: ENDLOOP - End a Loop
1682 End a loop started with bgnloop.
1685 .. opcode:: ENDSUB - End Subroutine
1687 Ends definition of a subroutine.
1690 .. opcode:: NOP - No Operation
1695 .. opcode:: BRK - Break
1697 Unconditionally moves the point of execution to the instruction after the
1698 next endloop or endswitch. The instruction must appear within a loop/endloop
1699 or switch/endswitch.
1702 .. opcode:: BREAKC - Break Conditional
1704 Conditionally moves the point of execution to the instruction after the
1705 next endloop or endswitch. The instruction must appear within a loop/endloop
1706 or switch/endswitch.
1707 Condition evaluates to true if src0.x != 0 where src0.x is interpreted
1708 as an integer register.
1712 Considered for removal as it's quite inconsistent wrt other opcodes
1713 (could emulate with UIF/BRK/ENDIF).
1716 .. opcode:: IF - Float If
1718 Start an IF ... ELSE .. ENDIF block. Condition evaluates to true if
1722 where src0.x is interpreted as a floating point register.
1725 .. opcode:: UIF - Bitwise If
1727 Start an UIF ... ELSE .. ENDIF block. Condition evaluates to true if
1731 where src0.x is interpreted as an integer register.
1734 .. opcode:: ELSE - Else
1736 Starts an else block, after an IF or UIF statement.
1739 .. opcode:: ENDIF - End If
1741 Ends an IF or UIF block.
1744 .. opcode:: SWITCH - Switch
1746 Starts a C-style switch expression. The switch consists of one or multiple
1747 CASE statements, and at most one DEFAULT statement. Execution of a statement
1748 ends when a BRK is hit, but just like in C falling through to other cases
1749 without a break is allowed. Similarly, DEFAULT label is allowed anywhere not
1750 just as last statement, and fallthrough is allowed into/from it.
1751 CASE src arguments are evaluated at bit level against the SWITCH src argument.
1757 (some instructions here)
1760 (some instructions here)
1763 (some instructions here)
1768 .. opcode:: CASE - Switch case
1770 This represents a switch case label. The src arg must be an integer immediate.
1773 .. opcode:: DEFAULT - Switch default
1775 This represents the default case in the switch, which is taken if no other
1779 .. opcode:: ENDSWITCH - End of switch
1781 Ends a switch expression.
1787 The interpolation instructions allow an input to be interpolated in a
1788 different way than its declaration. This corresponds to the GLSL 4.00
1789 interpolateAt* functions. The first argument of each of these must come from
1790 ``TGSI_FILE_INPUT``.
1792 .. opcode:: INTERP_CENTROID - Interpolate at the centroid
1794 Interpolates the varying specified by src0 at the centroid
1796 .. opcode:: INTERP_SAMPLE - Interpolate at the specified sample
1798 Interpolates the varying specified by src0 at the sample id specified by
1799 src1.x (interpreted as an integer)
1801 .. opcode:: INTERP_OFFSET - Interpolate at the specified offset
1803 Interpolates the varying specified by src0 at the offset src1.xy from the
1804 pixel center (interpreted as floats)
1812 The double-precision opcodes reinterpret four-component vectors into
1813 two-component vectors with doubled precision in each component.
1815 .. opcode:: DABS - Absolute
1823 .. opcode:: DADD - Add
1827 dst.xy = src0.xy + src1.xy
1829 dst.zw = src0.zw + src1.zw
1831 .. opcode:: DSEQ - Set on Equal
1835 dst.x = src0.xy == src1.xy ? \sim 0 : 0
1837 dst.z = src0.zw == src1.zw ? \sim 0 : 0
1839 .. opcode:: DSNE - Set on Equal
1843 dst.x = src0.xy != src1.xy ? \sim 0 : 0
1845 dst.z = src0.zw != src1.zw ? \sim 0 : 0
1847 .. opcode:: DSLT - Set on Less than
1851 dst.x = src0.xy < src1.xy ? \sim 0 : 0
1853 dst.z = src0.zw < src1.zw ? \sim 0 : 0
1855 .. opcode:: DSGE - Set on Greater equal
1859 dst.x = src0.xy >= src1.xy ? \sim 0 : 0
1861 dst.z = src0.zw >= src1.zw ? \sim 0 : 0
1863 .. opcode:: DFRAC - Fraction
1867 dst.xy = src.xy - \lfloor src.xy\rfloor
1869 dst.zw = src.zw - \lfloor src.zw\rfloor
1871 .. opcode:: DTRUNC - Truncate
1875 dst.xy = trunc(src.xy)
1877 dst.zw = trunc(src.zw)
1879 .. opcode:: DCEIL - Ceiling
1883 dst.xy = \lceil src.xy\rceil
1885 dst.zw = \lceil src.zw\rceil
1887 .. opcode:: DFLR - Floor
1891 dst.xy = \lfloor src.xy\rfloor
1893 dst.zw = \lfloor src.zw\rfloor
1895 .. opcode:: DROUND - Fraction
1899 dst.xy = round(src.xy)
1901 dst.zw = round(src.zw)
1903 .. opcode:: DSSG - Set Sign
1907 dst.xy = (src.xy > 0) ? 1.0 : (src.xy < 0) ? -1.0 : 0.0
1909 dst.zw = (src.zw > 0) ? 1.0 : (src.zw < 0) ? -1.0 : 0.0
1911 .. opcode:: DFRACEXP - Convert Number to Fractional and Integral Components
1913 Like the ``frexp()`` routine in many math libraries, this opcode stores the
1914 exponent of its source to ``dst0``, and the significand to ``dst1``, such that
1915 :math:`dst1 \times 2^{dst0} = src` .
1919 dst0.xy = exp(src.xy)
1921 dst1.xy = frac(src.xy)
1923 dst0.zw = exp(src.zw)
1925 dst1.zw = frac(src.zw)
1927 .. opcode:: DLDEXP - Multiply Number by Integral Power of 2
1929 This opcode is the inverse of :opcode:`DFRACEXP`. The second
1930 source is an integer.
1934 dst.xy = src0.xy \times 2^{src1.x}
1936 dst.zw = src0.zw \times 2^{src1.y}
1938 .. opcode:: DMIN - Minimum
1942 dst.xy = min(src0.xy, src1.xy)
1944 dst.zw = min(src0.zw, src1.zw)
1946 .. opcode:: DMAX - Maximum
1950 dst.xy = max(src0.xy, src1.xy)
1952 dst.zw = max(src0.zw, src1.zw)
1954 .. opcode:: DMUL - Multiply
1958 dst.xy = src0.xy \times src1.xy
1960 dst.zw = src0.zw \times src1.zw
1963 .. opcode:: DMAD - Multiply And Add
1967 dst.xy = src0.xy \times src1.xy + src2.xy
1969 dst.zw = src0.zw \times src1.zw + src2.zw
1972 .. opcode:: DFMA - Fused Multiply-Add
1974 Perform a * b + c with no intermediate rounding step.
1978 dst.xy = src0.xy \times src1.xy + src2.xy
1980 dst.zw = src0.zw \times src1.zw + src2.zw
1983 .. opcode:: DDIV - Divide
1987 dst.xy = \frac{src0.xy}{src1.xy}
1989 dst.zw = \frac{src0.zw}{src1.zw}
1992 .. opcode:: DRCP - Reciprocal
1996 dst.xy = \frac{1}{src.xy}
1998 dst.zw = \frac{1}{src.zw}
2000 .. opcode:: DSQRT - Square Root
2004 dst.xy = \sqrt{src.xy}
2006 dst.zw = \sqrt{src.zw}
2008 .. opcode:: DRSQ - Reciprocal Square Root
2012 dst.xy = \frac{1}{\sqrt{src.xy}}
2014 dst.zw = \frac{1}{\sqrt{src.zw}}
2016 .. opcode:: F2D - Float to Double
2020 dst.xy = double(src0.x)
2022 dst.zw = double(src0.y)
2024 .. opcode:: D2F - Double to Float
2028 dst.x = float(src0.xy)
2030 dst.y = float(src0.zw)
2032 .. opcode:: I2D - Int to Double
2036 dst.xy = double(src0.x)
2038 dst.zw = double(src0.y)
2040 .. opcode:: D2I - Double to Int
2044 dst.x = int(src0.xy)
2046 dst.y = int(src0.zw)
2048 .. opcode:: U2D - Unsigned Int to Double
2052 dst.xy = double(src0.x)
2054 dst.zw = double(src0.y)
2056 .. opcode:: D2U - Double to Unsigned Int
2060 dst.x = unsigned(src0.xy)
2062 dst.y = unsigned(src0.zw)
2067 The 64-bit integer opcodes reinterpret four-component vectors into
2068 two-component vectors with 64-bits in each component.
2070 .. opcode:: I64ABS - 64-bit Integer Absolute Value
2078 .. opcode:: I64NEG - 64-bit Integer Negate
2088 .. opcode:: I64SSG - 64-bit Integer Set Sign
2092 dst.xy = (src0.xy < 0) ? -1 : (src0.xy > 0) ? 1 : 0
2094 dst.zw = (src0.zw < 0) ? -1 : (src0.zw > 0) ? 1 : 0
2096 .. opcode:: U64ADD - 64-bit Integer Add
2100 dst.xy = src0.xy + src1.xy
2102 dst.zw = src0.zw + src1.zw
2104 .. opcode:: U64MUL - 64-bit Integer Multiply
2108 dst.xy = src0.xy * src1.xy
2110 dst.zw = src0.zw * src1.zw
2112 .. opcode:: U64SEQ - 64-bit Integer Set on Equal
2116 dst.x = src0.xy == src1.xy ? \sim 0 : 0
2118 dst.z = src0.zw == src1.zw ? \sim 0 : 0
2120 .. opcode:: U64SNE - 64-bit Integer Set on Not Equal
2124 dst.x = src0.xy != src1.xy ? \sim 0 : 0
2126 dst.z = src0.zw != src1.zw ? \sim 0 : 0
2128 .. opcode:: U64SLT - 64-bit Unsigned Integer Set on Less Than
2132 dst.x = src0.xy < src1.xy ? \sim 0 : 0
2134 dst.z = src0.zw < src1.zw ? \sim 0 : 0
2136 .. opcode:: U64SGE - 64-bit Unsigned Integer Set on Greater Equal
2140 dst.x = src0.xy >= src1.xy ? \sim 0 : 0
2142 dst.z = src0.zw >= src1.zw ? \sim 0 : 0
2144 .. opcode:: I64SLT - 64-bit Signed Integer Set on Less Than
2148 dst.x = src0.xy < src1.xy ? \sim 0 : 0
2150 dst.z = src0.zw < src1.zw ? \sim 0 : 0
2152 .. opcode:: I64SGE - 64-bit Signed Integer Set on Greater Equal
2156 dst.x = src0.xy >= src1.xy ? \sim 0 : 0
2158 dst.z = src0.zw >= src1.zw ? \sim 0 : 0
2160 .. opcode:: I64MIN - Minimum of 64-bit Signed Integers
2164 dst.xy = min(src0.xy, src1.xy)
2166 dst.zw = min(src0.zw, src1.zw)
2168 .. opcode:: U64MIN - Minimum of 64-bit Unsigned Integers
2172 dst.xy = min(src0.xy, src1.xy)
2174 dst.zw = min(src0.zw, src1.zw)
2176 .. opcode:: I64MAX - Maximum of 64-bit Signed Integers
2180 dst.xy = max(src0.xy, src1.xy)
2182 dst.zw = max(src0.zw, src1.zw)
2184 .. opcode:: U64MAX - Maximum of 64-bit Unsigned Integers
2188 dst.xy = max(src0.xy, src1.xy)
2190 dst.zw = max(src0.zw, src1.zw)
2192 .. opcode:: U64SHL - Shift Left 64-bit Unsigned Integer
2194 The shift count is masked with 0x3f before the shift is applied.
2198 dst.xy = src0.xy << (0x3f \& src1.x)
2200 dst.zw = src0.zw << (0x3f \& src1.y)
2202 .. opcode:: I64SHR - Arithmetic Shift Right (of 64-bit Signed Integer)
2204 The shift count is masked with 0x3f before the shift is applied.
2208 dst.xy = src0.xy >> (0x3f \& src1.x)
2210 dst.zw = src0.zw >> (0x3f \& src1.y)
2212 .. opcode:: U64SHR - Logical Shift Right (of 64-bit Unsigned Integer)
2214 The shift count is masked with 0x3f before the shift is applied.
2218 dst.xy = src0.xy >> (unsigned) (0x3f \& src1.x)
2220 dst.zw = src0.zw >> (unsigned) (0x3f \& src1.y)
2222 .. opcode:: I64DIV - 64-bit Signed Integer Division
2226 dst.xy = \frac{src0.xy}{src1.xy}
2228 dst.zw = \frac{src0.zw}{src1.zw}
2230 .. opcode:: U64DIV - 64-bit Unsigned Integer Division
2234 dst.xy = \frac{src0.xy}{src1.xy}
2236 dst.zw = \frac{src0.zw}{src1.zw}
2238 .. opcode:: U64MOD - 64-bit Unsigned Integer Remainder
2242 dst.xy = src0.xy \bmod src1.xy
2244 dst.zw = src0.zw \bmod src1.zw
2246 .. opcode:: I64MOD - 64-bit Signed Integer Remainder
2250 dst.xy = src0.xy \bmod src1.xy
2252 dst.zw = src0.zw \bmod src1.zw
2254 .. opcode:: F2U64 - Float to 64-bit Unsigned Int
2258 dst.xy = (uint64_t) src0.x
2260 dst.zw = (uint64_t) src0.y
2262 .. opcode:: F2I64 - Float to 64-bit Int
2266 dst.xy = (int64_t) src0.x
2268 dst.zw = (int64_t) src0.y
2270 .. opcode:: U2I64 - Unsigned Integer to 64-bit Integer
2272 This is a zero extension.
2276 dst.xy = (uint64_t) src0.x
2278 dst.zw = (uint64_t) src0.y
2280 .. opcode:: I2I64 - Signed Integer to 64-bit Integer
2282 This is a sign extension.
2286 dst.xy = (int64_t) src0.x
2288 dst.zw = (int64_t) src0.y
2290 .. opcode:: D2U64 - Double to 64-bit Unsigned Int
2294 dst.xy = (uint64_t) src0.xy
2296 dst.zw = (uint64_t) src0.zw
2298 .. opcode:: D2I64 - Double to 64-bit Int
2302 dst.xy = (int64_t) src0.xy
2304 dst.zw = (int64_t) src0.zw
2306 .. opcode:: U642F - 64-bit unsigned integer to float
2310 dst.x = (float) src0.xy
2312 dst.y = (float) src0.zw
2314 .. opcode:: I642F - 64-bit Int to Float
2318 dst.x = (float) src0.xy
2320 dst.y = (float) src0.zw
2322 .. opcode:: U642D - 64-bit unsigned integer to double
2326 dst.xy = (double) src0.xy
2328 dst.zw = (double) src0.zw
2330 .. opcode:: I642D - 64-bit Int to double
2334 dst.xy = (double) src0.xy
2336 dst.zw = (double) src0.zw
2338 .. _samplingopcodes:
2340 Resource Sampling Opcodes
2341 ^^^^^^^^^^^^^^^^^^^^^^^^^
2343 Those opcodes follow very closely semantics of the respective Direct3D
2344 instructions. If in doubt double check Direct3D documentation.
2345 Note that the swizzle on SVIEW (src1) determines texel swizzling
2350 Using provided address, sample data from the specified texture using the
2351 filtering mode identified by the given sampler. The source data may come from
2352 any resource type other than buffers.
2354 Syntax: ``SAMPLE dst, address, sampler_view, sampler``
2356 Example: ``SAMPLE TEMP[0], TEMP[1], SVIEW[0], SAMP[0]``
2358 .. opcode:: SAMPLE_I
2360 Simplified alternative to the SAMPLE instruction. Using the provided
2361 integer address, SAMPLE_I fetches data from the specified sampler view
2362 without any filtering. The source data may come from any resource type
2365 Syntax: ``SAMPLE_I dst, address, sampler_view``
2367 Example: ``SAMPLE_I TEMP[0], TEMP[1], SVIEW[0]``
2369 The 'address' is specified as unsigned integers. If the 'address' is out of
2370 range [0...(# texels - 1)] the result of the fetch is always 0 in all
2371 components. As such the instruction doesn't honor address wrap modes, in
2372 cases where that behavior is desirable 'SAMPLE' instruction should be used.
2373 address.w always provides an unsigned integer mipmap level. If the value is
2374 out of the range then the instruction always returns 0 in all components.
2375 address.yz are ignored for buffers and 1d textures. address.z is ignored
2376 for 1d texture arrays and 2d textures.
2378 For 1D texture arrays address.y provides the array index (also as unsigned
2379 integer). If the value is out of the range of available array indices
2380 [0... (array size - 1)] then the opcode always returns 0 in all components.
2381 For 2D texture arrays address.z provides the array index, otherwise it
2382 exhibits the same behavior as in the case for 1D texture arrays. The exact
2383 semantics of the source address are presented in the table below:
2385 +---------------------------+----+-----+-----+---------+
2386 | resource type | X | Y | Z | W |
2387 +===========================+====+=====+=====+=========+
2388 | ``PIPE_BUFFER`` | x | | | ignored |
2389 +---------------------------+----+-----+-----+---------+
2390 | ``PIPE_TEXTURE_1D`` | x | | | mpl |
2391 +---------------------------+----+-----+-----+---------+
2392 | ``PIPE_TEXTURE_2D`` | x | y | | mpl |
2393 +---------------------------+----+-----+-----+---------+
2394 | ``PIPE_TEXTURE_3D`` | x | y | z | mpl |
2395 +---------------------------+----+-----+-----+---------+
2396 | ``PIPE_TEXTURE_RECT`` | x | y | | mpl |
2397 +---------------------------+----+-----+-----+---------+
2398 | ``PIPE_TEXTURE_CUBE`` | not allowed as source |
2399 +---------------------------+----+-----+-----+---------+
2400 | ``PIPE_TEXTURE_1D_ARRAY`` | x | idx | | mpl |
2401 +---------------------------+----+-----+-----+---------+
2402 | ``PIPE_TEXTURE_2D_ARRAY`` | x | y | idx | mpl |
2403 +---------------------------+----+-----+-----+---------+
2405 Where 'mpl' is a mipmap level and 'idx' is the array index.
2407 .. opcode:: SAMPLE_I_MS
2409 Just like SAMPLE_I but allows fetch data from multi-sampled surfaces.
2411 Syntax: ``SAMPLE_I_MS dst, address, sampler_view, sample``
2413 .. opcode:: SAMPLE_B
2415 Just like the SAMPLE instruction with the exception that an additional bias
2416 is applied to the level of detail computed as part of the instruction
2419 Syntax: ``SAMPLE_B dst, address, sampler_view, sampler, lod_bias``
2421 Example: ``SAMPLE_B TEMP[0], TEMP[1], SVIEW[0], SAMP[0], TEMP[2].x``
2423 .. opcode:: SAMPLE_C
2425 Similar to the SAMPLE instruction but it performs a comparison filter. The
2426 operands to SAMPLE_C are identical to SAMPLE, except that there is an
2427 additional float32 operand, reference value, which must be a register with
2428 single-component, or a scalar literal. SAMPLE_C makes the hardware use the
2429 current samplers compare_func (in pipe_sampler_state) to compare reference
2430 value against the red component value for the surce resource at each texel
2431 that the currently configured texture filter covers based on the provided
2434 Syntax: ``SAMPLE_C dst, address, sampler_view.r, sampler, ref_value``
2436 Example: ``SAMPLE_C TEMP[0], TEMP[1], SVIEW[0].r, SAMP[0], TEMP[2].x``
2438 .. opcode:: SAMPLE_C_LZ
2440 Same as SAMPLE_C, but LOD is 0 and derivatives are ignored. The LZ stands
2443 Syntax: ``SAMPLE_C_LZ dst, address, sampler_view.r, sampler, ref_value``
2445 Example: ``SAMPLE_C_LZ TEMP[0], TEMP[1], SVIEW[0].r, SAMP[0], TEMP[2].x``
2448 .. opcode:: SAMPLE_D
2450 SAMPLE_D is identical to the SAMPLE opcode except that the derivatives for
2451 the source address in the x direction and the y direction are provided by
2454 Syntax: ``SAMPLE_D dst, address, sampler_view, sampler, der_x, der_y``
2456 Example: ``SAMPLE_D TEMP[0], TEMP[1], SVIEW[0], SAMP[0], TEMP[2], TEMP[3]``
2458 .. opcode:: SAMPLE_L
2460 SAMPLE_L is identical to the SAMPLE opcode except that the LOD is provided
2461 directly as a scalar value, representing no anisotropy.
2463 Syntax: ``SAMPLE_L dst, address, sampler_view, sampler, explicit_lod``
2465 Example: ``SAMPLE_L TEMP[0], TEMP[1], SVIEW[0], SAMP[0], TEMP[2].x``
2469 Gathers the four texels to be used in a bi-linear filtering operation and
2470 packs them into a single register. Only works with 2D, 2D array, cubemaps,
2471 and cubemaps arrays. For 2D textures, only the addressing modes of the
2472 sampler and the top level of any mip pyramid are used. Set W to zero. It
2473 behaves like the SAMPLE instruction, but a filtered sample is not
2474 generated. The four samples that contribute to filtering are placed into
2475 xyzw in counter-clockwise order, starting with the (u,v) texture coordinate
2476 delta at the following locations (-, +), (+, +), (+, -), (-, -), where the
2477 magnitude of the deltas are half a texel.
2480 .. opcode:: SVIEWINFO
2482 Query the dimensions of a given sampler view. dst receives width, height,
2483 depth or array size and number of mipmap levels as int4. The dst can have a
2484 writemask which will specify what info is the caller interested in.
2486 Syntax: ``SVIEWINFO dst, src_mip_level, sampler_view``
2488 Example: ``SVIEWINFO TEMP[0], TEMP[1].x, SVIEW[0]``
2490 src_mip_level is an unsigned integer scalar. If it's out of range then
2491 returns 0 for width, height and depth/array size but the total number of
2492 mipmap is still returned correctly for the given sampler view. The returned
2493 width, height and depth values are for the mipmap level selected by the
2494 src_mip_level and are in the number of texels. For 1d texture array width
2495 is in dst.x, array size is in dst.y and dst.z is 0. The number of mipmaps is
2496 still in dst.w. In contrast to d3d10 resinfo, there's no way in the tgsi
2497 instruction encoding to specify the return type (float/rcpfloat/uint), hence
2498 always using uint. Also, unlike the SAMPLE instructions, the swizzle on src1
2499 resinfo allowing swizzling dst values is ignored (due to the interaction
2500 with rcpfloat modifier which requires some swizzle handling in the state
2503 .. opcode:: SAMPLE_POS
2505 Query the position of a sample in the given resource or render target
2506 when per-sample fragment shading is in effect.
2508 Syntax: ``SAMPLE_POS dst, source, sample_index``
2510 dst receives float4 (x, y, undef, undef) indicated where the sample is
2511 located. Sample locations are in the range [0, 1] where 0.5 is the center
2514 source is either a sampler view (to indicate a shader resource) or temp
2515 register (to indicate the render target). The source register may have
2516 an optional swizzle to apply to the returned result
2518 sample_index is an integer scalar indicating which sample position is to
2521 If per-sample shading is not in effect or the source resource or render
2522 target is not multisampled, the result is (0.5, 0.5, undef, undef).
2524 NOTE: no driver has implemented this opcode yet (and no state tracker
2525 emits it). This information is subject to change.
2527 .. opcode:: SAMPLE_INFO
2529 Query the number of samples in a multisampled resource or render target.
2531 Syntax: ``SAMPLE_INFO dst, source``
2533 dst receives int4 (n, 0, 0, 0) where n is the number of samples in a
2534 resource or the render target.
2536 source is either a sampler view (to indicate a shader resource) or temp
2537 register (to indicate the render target). The source register may have
2538 an optional swizzle to apply to the returned result
2540 If per-sample shading is not in effect or the source resource or render
2541 target is not multisampled, the result is (1, 0, 0, 0).
2543 NOTE: no driver has implemented this opcode yet (and no state tracker
2544 emits it). This information is subject to change.
2546 .. _resourceopcodes:
2548 Resource Access Opcodes
2549 ^^^^^^^^^^^^^^^^^^^^^^^
2551 For these opcodes, the resource can be a BUFFER, IMAGE, or MEMORY.
2553 .. opcode:: LOAD - Fetch data from a shader buffer or image
2555 Syntax: ``LOAD dst, resource, address``
2557 Example: ``LOAD TEMP[0], BUFFER[0], TEMP[1]``
2559 Using the provided integer address, LOAD fetches data
2560 from the specified buffer or texture without any
2563 The 'address' is specified as a vector of unsigned
2564 integers. If the 'address' is out of range the result
2567 Only the first mipmap level of a resource can be read
2568 from using this instruction.
2570 For 1D or 2D texture arrays, the array index is
2571 provided as an unsigned integer in address.y or
2572 address.z, respectively. address.yz are ignored for
2573 buffers and 1D textures. address.z is ignored for 1D
2574 texture arrays and 2D textures. address.w is always
2577 A swizzle suffix may be added to the resource argument
2578 this will cause the resource data to be swizzled accordingly.
2580 .. opcode:: STORE - Write data to a shader resource
2582 Syntax: ``STORE resource, address, src``
2584 Example: ``STORE BUFFER[0], TEMP[0], TEMP[1]``
2586 Using the provided integer address, STORE writes data
2587 to the specified buffer or texture.
2589 The 'address' is specified as a vector of unsigned
2590 integers. If the 'address' is out of range the result
2593 Only the first mipmap level of a resource can be
2594 written to using this instruction.
2596 For 1D or 2D texture arrays, the array index is
2597 provided as an unsigned integer in address.y or
2598 address.z, respectively. address.yz are ignored for
2599 buffers and 1D textures. address.z is ignored for 1D
2600 texture arrays and 2D textures. address.w is always
2603 .. opcode:: RESQ - Query information about a resource
2605 Syntax: ``RESQ dst, resource``
2607 Example: ``RESQ TEMP[0], BUFFER[0]``
2609 Returns information about the buffer or image resource. For buffer
2610 resources, the size (in bytes) is returned in the x component. For
2611 image resources, .xyz will contain the width/height/layers of the
2612 image, while .w will contain the number of samples for multi-sampled
2615 .. opcode:: FBFETCH - Load data from framebuffer
2617 Syntax: ``FBFETCH dst, output``
2619 Example: ``FBFETCH TEMP[0], OUT[0]``
2621 This is only valid on ``COLOR`` semantic outputs. Returns the color
2622 of the current position in the framebuffer from before this fragment
2623 shader invocation. May return the same value from multiple calls for
2624 a particular output within a single invocation. Note that result may
2625 be undefined if a fragment is drawn multiple times without a blend
2629 .. _threadsyncopcodes:
2631 Inter-thread synchronization opcodes
2632 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2634 These opcodes are intended for communication between threads running
2635 within the same compute grid. For now they're only valid in compute
2638 .. opcode:: MFENCE - Memory fence
2640 Syntax: ``MFENCE resource``
2642 Example: ``MFENCE RES[0]``
2644 This opcode forces strong ordering between any memory access
2645 operations that affect the specified resource. This means that
2646 previous loads and stores (and only those) will be performed and
2647 visible to other threads before the program execution continues.
2650 .. opcode:: LFENCE - Load memory fence
2652 Syntax: ``LFENCE resource``
2654 Example: ``LFENCE RES[0]``
2656 Similar to MFENCE, but it only affects the ordering of memory loads.
2659 .. opcode:: SFENCE - Store memory fence
2661 Syntax: ``SFENCE resource``
2663 Example: ``SFENCE RES[0]``
2665 Similar to MFENCE, but it only affects the ordering of memory stores.
2668 .. opcode:: BARRIER - Thread group barrier
2672 This opcode suspends the execution of the current thread until all
2673 the remaining threads in the working group reach the same point of
2674 the program. Results are unspecified if any of the remaining
2675 threads terminates or never reaches an executed BARRIER instruction.
2677 .. opcode:: MEMBAR - Memory barrier
2681 This opcode waits for the completion of all memory accesses based on
2682 the type passed in. The type is an immediate bitfield with the following
2685 Bit 0: Shader storage buffers
2686 Bit 1: Atomic buffers
2688 Bit 3: Shared memory
2691 These may be passed in in any combination. An implementation is free to not
2692 distinguish between these as it sees fit. However these map to all the
2693 possibilities made available by GLSL.
2700 These opcodes provide atomic variants of some common arithmetic and
2701 logical operations. In this context atomicity means that another
2702 concurrent memory access operation that affects the same memory
2703 location is guaranteed to be performed strictly before or after the
2704 entire execution of the atomic operation. The resource may be a BUFFER,
2705 IMAGE, or MEMORY. In the case of an image, the offset works the same as for
2706 ``LOAD`` and ``STORE``, specified above. These atomic operations may
2707 only be used with 32-bit integer image formats.
2709 .. opcode:: ATOMUADD - Atomic integer addition
2711 Syntax: ``ATOMUADD dst, resource, offset, src``
2713 Example: ``ATOMUADD TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2715 The following operation is performed atomically:
2719 dst_x = resource[offset]
2721 resource[offset] = dst_x + src_x
2724 .. opcode:: ATOMXCHG - Atomic exchange
2726 Syntax: ``ATOMXCHG dst, resource, offset, src``
2728 Example: ``ATOMXCHG TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2730 The following operation is performed atomically:
2734 dst_x = resource[offset]
2736 resource[offset] = src_x
2739 .. opcode:: ATOMCAS - Atomic compare-and-exchange
2741 Syntax: ``ATOMCAS dst, resource, offset, cmp, src``
2743 Example: ``ATOMCAS TEMP[0], BUFFER[0], TEMP[1], TEMP[2], TEMP[3]``
2745 The following operation is performed atomically:
2749 dst_x = resource[offset]
2751 resource[offset] = (dst_x == cmp_x ? src_x : dst_x)
2754 .. opcode:: ATOMAND - Atomic bitwise And
2756 Syntax: ``ATOMAND dst, resource, offset, src``
2758 Example: ``ATOMAND TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2760 The following operation is performed atomically:
2764 dst_x = resource[offset]
2766 resource[offset] = dst_x \& src_x
2769 .. opcode:: ATOMOR - Atomic bitwise Or
2771 Syntax: ``ATOMOR dst, resource, offset, src``
2773 Example: ``ATOMOR TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2775 The following operation is performed atomically:
2779 dst_x = resource[offset]
2781 resource[offset] = dst_x | src_x
2784 .. opcode:: ATOMXOR - Atomic bitwise Xor
2786 Syntax: ``ATOMXOR dst, resource, offset, src``
2788 Example: ``ATOMXOR TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2790 The following operation is performed atomically:
2794 dst_x = resource[offset]
2796 resource[offset] = dst_x \oplus src_x
2799 .. opcode:: ATOMUMIN - Atomic unsigned minimum
2801 Syntax: ``ATOMUMIN dst, resource, offset, src``
2803 Example: ``ATOMUMIN TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2805 The following operation is performed atomically:
2809 dst_x = resource[offset]
2811 resource[offset] = (dst_x < src_x ? dst_x : src_x)
2814 .. opcode:: ATOMUMAX - Atomic unsigned maximum
2816 Syntax: ``ATOMUMAX dst, resource, offset, src``
2818 Example: ``ATOMUMAX TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2820 The following operation is performed atomically:
2824 dst_x = resource[offset]
2826 resource[offset] = (dst_x > src_x ? dst_x : src_x)
2829 .. opcode:: ATOMIMIN - Atomic signed minimum
2831 Syntax: ``ATOMIMIN dst, resource, offset, src``
2833 Example: ``ATOMIMIN TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2835 The following operation is performed atomically:
2839 dst_x = resource[offset]
2841 resource[offset] = (dst_x < src_x ? dst_x : src_x)
2844 .. opcode:: ATOMIMAX - Atomic signed maximum
2846 Syntax: ``ATOMIMAX dst, resource, offset, src``
2848 Example: ``ATOMIMAX TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2850 The following operation is performed atomically:
2854 dst_x = resource[offset]
2856 resource[offset] = (dst_x > src_x ? dst_x : src_x)
2859 .. _interlaneopcodes:
2864 These opcodes reduce the given value across the shader invocations
2865 running in the current SIMD group. Every thread in the subgroup will receive
2866 the same result. The BALLOT operations accept a single-channel argument that
2867 is treated as a boolean and produce a 64-bit value.
2869 .. opcode:: VOTE_ANY - Value is set in any of the active invocations
2871 Syntax: ``VOTE_ANY dst, value``
2873 Example: ``VOTE_ANY TEMP[0].x, TEMP[1].x``
2876 .. opcode:: VOTE_ALL - Value is set in all of the active invocations
2878 Syntax: ``VOTE_ALL dst, value``
2880 Example: ``VOTE_ALL TEMP[0].x, TEMP[1].x``
2883 .. opcode:: VOTE_EQ - Value is the same in all of the active invocations
2885 Syntax: ``VOTE_EQ dst, value``
2887 Example: ``VOTE_EQ TEMP[0].x, TEMP[1].x``
2890 .. opcode:: BALLOT - Lanemask of whether the value is set in each active
2893 Syntax: ``BALLOT dst, value``
2895 Example: ``BALLOT TEMP[0].xy, TEMP[1].x``
2897 When the argument is a constant true, this produces a bitmask of active
2898 invocations. In fragment shaders, this can include helper invocations
2899 (invocations whose outputs and writes to memory are discarded, but which
2900 are used to compute derivatives).
2903 .. opcode:: READ_FIRST - Broadcast the value from the first active
2904 invocation to all active lanes
2906 Syntax: ``READ_FIRST dst, value``
2908 Example: ``READ_FIRST TEMP[0], TEMP[1]``
2911 .. opcode:: READ_INVOC - Retrieve the value from the given invocation
2912 (need not be uniform)
2914 Syntax: ``READ_INVOC dst, value, invocation``
2916 Example: ``READ_INVOC TEMP[0].xy, TEMP[1].xy, TEMP[2].x``
2918 invocation.x controls the invocation number to read from for all channels.
2919 The invocation number must be the same across all active invocations in a
2920 sub-group; otherwise, the results are undefined.
2923 Explanation of symbols used
2924 ------------------------------
2931 :math:`|x|` Absolute value of `x`.
2933 :math:`\lceil x \rceil` Ceiling of `x`.
2935 clamp(x,y,z) Clamp x between y and z.
2936 (x < y) ? y : (x > z) ? z : x
2938 :math:`\lfloor x\rfloor` Floor of `x`.
2940 :math:`\log_2{x}` Logarithm of `x`, base 2.
2942 max(x,y) Maximum of x and y.
2945 min(x,y) Minimum of x and y.
2948 partialx(x) Derivative of x relative to fragment's X.
2950 partialy(x) Derivative of x relative to fragment's Y.
2952 pop() Pop from stack.
2954 :math:`x^y` `x` to the power `y`.
2956 push(x) Push x on stack.
2960 trunc(x) Truncate x, i.e. drop the fraction bits.
2967 discard Discard fragment.
2971 target Label of target instruction.
2982 Declares a register that is will be referenced as an operand in Instruction
2985 File field contains register file that is being declared and is one
2988 UsageMask field specifies which of the register components can be accessed
2989 and is one of TGSI_WRITEMASK.
2991 The Local flag specifies that a given value isn't intended for
2992 subroutine parameter passing and, as a result, the implementation
2993 isn't required to give any guarantees of it being preserved across
2994 subroutine boundaries. As it's merely a compiler hint, the
2995 implementation is free to ignore it.
2997 If Dimension flag is set to 1, a Declaration Dimension token follows.
2999 If Semantic flag is set to 1, a Declaration Semantic token follows.
3001 If Interpolate flag is set to 1, a Declaration Interpolate token follows.
3003 If file is TGSI_FILE_RESOURCE, a Declaration Resource token follows.
3005 If Array flag is set to 1, a Declaration Array token follows.
3008 ^^^^^^^^^^^^^^^^^^^^^^^^
3010 Declarations can optional have an ArrayID attribute which can be referred by
3011 indirect addressing operands. An ArrayID of zero is reserved and treated as
3012 if no ArrayID is specified.
3014 If an indirect addressing operand refers to a specific declaration by using
3015 an ArrayID only the registers in this declaration are guaranteed to be
3016 accessed, accessing any register outside this declaration results in undefined
3017 behavior. Note that for compatibility the effective index is zero-based and
3018 not relative to the specified declaration
3020 If no ArrayID is specified with an indirect addressing operand the whole
3021 register file might be accessed by this operand. This is strongly discouraged
3022 and will prevent packing of scalar/vec2 arrays and effective alias analysis.
3023 This is only legal for TEMP and CONST register files.
3025 Declaration Semantic
3026 ^^^^^^^^^^^^^^^^^^^^^^^^
3028 Vertex and fragment shader input and output registers may be labeled
3029 with semantic information consisting of a name and index.
3031 Follows Declaration token if Semantic bit is set.
3033 Since its purpose is to link a shader with other stages of the pipeline,
3034 it is valid to follow only those Declaration tokens that declare a register
3035 either in INPUT or OUTPUT file.
3037 SemanticName field contains the semantic name of the register being declared.
3038 There is no default value.
3040 SemanticIndex is an optional subscript that can be used to distinguish
3041 different register declarations with the same semantic name. The default value
3044 The meanings of the individual semantic names are explained in the following
3047 TGSI_SEMANTIC_POSITION
3048 """"""""""""""""""""""
3050 For vertex shaders, TGSI_SEMANTIC_POSITION indicates the vertex shader
3051 output register which contains the homogeneous vertex position in the clip
3052 space coordinate system. After clipping, the X, Y and Z components of the
3053 vertex will be divided by the W value to get normalized device coordinates.
3055 For fragment shaders, TGSI_SEMANTIC_POSITION is used to indicate that
3056 fragment shader input (or system value, depending on which one is
3057 supported by the driver) contains the fragment's window position. The X
3058 component starts at zero and always increases from left to right.
3059 The Y component starts at zero and always increases but Y=0 may either
3060 indicate the top of the window or the bottom depending on the fragment
3061 coordinate origin convention (see TGSI_PROPERTY_FS_COORD_ORIGIN).
3062 The Z coordinate ranges from 0 to 1 to represent depth from the front
3063 to the back of the Z buffer. The W component contains the interpolated
3064 reciprocal of the vertex position W component (corresponding to gl_Fragcoord,
3065 but unlike d3d10 which interpolates the same 1/w but then gives back
3066 the reciprocal of the interpolated value).
3068 Fragment shaders may also declare an output register with
3069 TGSI_SEMANTIC_POSITION. Only the Z component is writable. This allows
3070 the fragment shader to change the fragment's Z position.
3077 For vertex shader outputs or fragment shader inputs/outputs, this
3078 label indicates that the register contains an R,G,B,A color.
3080 Several shader inputs/outputs may contain colors so the semantic index
3081 is used to distinguish them. For example, color[0] may be the diffuse
3082 color while color[1] may be the specular color.
3084 This label is needed so that the flat/smooth shading can be applied
3085 to the right interpolants during rasterization.
3089 TGSI_SEMANTIC_BCOLOR
3090 """"""""""""""""""""
3092 Back-facing colors are only used for back-facing polygons, and are only valid
3093 in vertex shader outputs. After rasterization, all polygons are front-facing
3094 and COLOR and BCOLOR end up occupying the same slots in the fragment shader,
3095 so all BCOLORs effectively become regular COLORs in the fragment shader.
3101 Vertex shader inputs and outputs and fragment shader inputs may be
3102 labeled with TGSI_SEMANTIC_FOG to indicate that the register contains
3103 a fog coordinate. Typically, the fragment shader will use the fog coordinate
3104 to compute a fog blend factor which is used to blend the normal fragment color
3105 with a constant fog color. But fog coord really is just an ordinary vec4
3106 register like regular semantics.
3112 Vertex shader input and output registers may be labeled with
3113 TGIS_SEMANTIC_PSIZE to indicate that the register contains a point size
3114 in the form (S, 0, 0, 1). The point size controls the width or diameter
3115 of points for rasterization. This label cannot be used in fragment
3118 When using this semantic, be sure to set the appropriate state in the
3119 :ref:`rasterizer` first.
3122 TGSI_SEMANTIC_TEXCOORD
3123 """"""""""""""""""""""
3125 Only available if PIPE_CAP_TGSI_TEXCOORD is exposed !
3127 Vertex shader outputs and fragment shader inputs may be labeled with
3128 this semantic to make them replaceable by sprite coordinates via the
3129 sprite_coord_enable state in the :ref:`rasterizer`.
3130 The semantic index permitted with this semantic is limited to <= 7.
3132 If the driver does not support TEXCOORD, sprite coordinate replacement
3133 applies to inputs with the GENERIC semantic instead.
3135 The intended use case for this semantic is gl_TexCoord.
3138 TGSI_SEMANTIC_PCOORD
3139 """"""""""""""""""""
3141 Only available if PIPE_CAP_TGSI_TEXCOORD is exposed !
3143 Fragment shader inputs may be labeled with TGSI_SEMANTIC_PCOORD to indicate
3144 that the register contains sprite coordinates in the form (x, y, 0, 1), if
3145 the current primitive is a point and point sprites are enabled. Otherwise,
3146 the contents of the register are undefined.
3148 The intended use case for this semantic is gl_PointCoord.
3151 TGSI_SEMANTIC_GENERIC
3152 """""""""""""""""""""
3154 All vertex/fragment shader inputs/outputs not labeled with any other
3155 semantic label can be considered to be generic attributes. Typical
3156 uses of generic inputs/outputs are texcoords and user-defined values.
3159 TGSI_SEMANTIC_NORMAL
3160 """"""""""""""""""""
3162 Indicates that a vertex shader input is a normal vector. This is
3163 typically only used for legacy graphics APIs.
3169 This label applies to fragment shader inputs (or system values,
3170 depending on which one is supported by the driver) and indicates that
3171 the register contains front/back-face information.
3173 If it is an input, it will be a floating-point vector in the form (F, 0, 0, 1),
3174 where F will be positive when the fragment belongs to a front-facing polygon,
3175 and negative when the fragment belongs to a back-facing polygon.
3177 If it is a system value, it will be an integer vector in the form (F, 0, 0, 1),
3178 where F is 0xffffffff when the fragment belongs to a front-facing polygon and
3179 0 when the fragment belongs to a back-facing polygon.
3182 TGSI_SEMANTIC_EDGEFLAG
3183 """"""""""""""""""""""
3185 For vertex shaders, this sematic label indicates that an input or
3186 output is a boolean edge flag. The register layout is [F, x, x, x]
3187 where F is 0.0 or 1.0 and x = don't care. Normally, the vertex shader
3188 simply copies the edge flag input to the edgeflag output.
3190 Edge flags are used to control which lines or points are actually
3191 drawn when the polygon mode converts triangles/quads/polygons into
3195 TGSI_SEMANTIC_STENCIL
3196 """""""""""""""""""""
3198 For fragment shaders, this semantic label indicates that an output
3199 is a writable stencil reference value. Only the Y component is writable.
3200 This allows the fragment shader to change the fragments stencilref value.
3203 TGSI_SEMANTIC_VIEWPORT_INDEX
3204 """"""""""""""""""""""""""""
3206 For geometry shaders, this semantic label indicates that an output
3207 contains the index of the viewport (and scissor) to use.
3208 This is an integer value, and only the X component is used.
3210 If PIPE_CAP_TGSI_VS_LAYER_VIEWPORT or PIPE_CAP_TGSI_TES_LAYER_VIEWPORT is
3211 supported, then this semantic label can also be used in vertex or
3212 tessellation evaluation shaders, respectively. Only the value written in the
3213 last vertex processing stage is used.
3219 For geometry shaders, this semantic label indicates that an output
3220 contains the layer value to use for the color and depth/stencil surfaces.
3221 This is an integer value, and only the X component is used.
3222 (Also known as rendertarget array index.)
3224 If PIPE_CAP_TGSI_VS_LAYER_VIEWPORT or PIPE_CAP_TGSI_TES_LAYER_VIEWPORT is
3225 supported, then this semantic label can also be used in vertex or
3226 tessellation evaluation shaders, respectively. Only the value written in the
3227 last vertex processing stage is used.
3230 TGSI_SEMANTIC_CULLDIST
3231 """"""""""""""""""""""
3233 Used as distance to plane for performing application-defined culling
3234 of individual primitives against a plane. When components of vertex
3235 elements are given this label, these values are assumed to be a
3236 float32 signed distance to a plane. Primitives will be completely
3237 discarded if the plane distance for all of the vertices in the
3238 primitive are < 0. If a vertex has a cull distance of NaN, that
3239 vertex counts as "out" (as if its < 0);
3240 The limits on both clip and cull distances are bound
3241 by the PIPE_MAX_CLIP_OR_CULL_DISTANCE_COUNT define which defines
3242 the maximum number of components that can be used to hold the
3243 distances and by the PIPE_MAX_CLIP_OR_CULL_DISTANCE_ELEMENT_COUNT
3244 which specifies the maximum number of registers which can be
3245 annotated with those semantics.
3248 TGSI_SEMANTIC_CLIPDIST
3249 """"""""""""""""""""""
3251 Note this covers clipping and culling distances.
3253 When components of vertex elements are identified this way, these
3254 values are each assumed to be a float32 signed distance to a plane.
3257 Primitive setup only invokes rasterization on pixels for which
3258 the interpolated plane distances are >= 0.
3261 Primitives will be completely discarded if the plane distance
3262 for all of the vertices in the primitive are < 0.
3263 If a vertex has a cull distance of NaN, that vertex counts as "out"
3266 Multiple clip/cull planes can be implemented simultaneously, by
3267 annotating multiple components of one or more vertex elements with
3268 the above specified semantic.
3269 The limits on both clip and cull distances are bound
3270 by the PIPE_MAX_CLIP_OR_CULL_DISTANCE_COUNT define which defines
3271 the maximum number of components that can be used to hold the
3272 distances and by the PIPE_MAX_CLIP_OR_CULL_DISTANCE_ELEMENT_COUNT
3273 which specifies the maximum number of registers which can be
3274 annotated with those semantics.
3275 The properties NUM_CLIPDIST_ENABLED and NUM_CULLDIST_ENABLED
3276 are used to divide up the 2 x vec4 space between clipping and culling.
3278 TGSI_SEMANTIC_SAMPLEID
3279 """"""""""""""""""""""
3281 For fragment shaders, this semantic label indicates that a system value
3282 contains the current sample id (i.e. gl_SampleID) as an unsigned int.
3283 Only the X component is used. If per-sample shading is not enabled,
3284 the result is (0, undef, undef, undef).
3286 Note that if the fragment shader uses this system value, the fragment
3287 shader is automatically executed at per sample frequency.
3289 TGSI_SEMANTIC_SAMPLEPOS
3290 """""""""""""""""""""""
3292 For fragment shaders, this semantic label indicates that a system
3293 value contains the current sample's position as float4(x, y, undef, undef)
3294 in the render target (i.e. gl_SamplePosition) when per-fragment shading
3295 is in effect. Position values are in the range [0, 1] where 0.5 is
3296 the center of the fragment.
3298 Note that if the fragment shader uses this system value, the fragment
3299 shader is automatically executed at per sample frequency.
3301 TGSI_SEMANTIC_SAMPLEMASK
3302 """"""""""""""""""""""""
3304 For fragment shaders, this semantic label can be applied to either a
3305 shader system value input or output.
3307 For a system value, the sample mask indicates the set of samples covered by
3308 the current primitive. If MSAA is not enabled, the value is (1, 0, 0, 0).
3310 For an output, the sample mask is used to disable further sample processing.
3312 For both, the register type is uint[4] but only the X component is used
3313 (i.e. gl_SampleMask[0]). Each bit corresponds to one sample position (up
3314 to 32x MSAA is supported).
3316 TGSI_SEMANTIC_INVOCATIONID
3317 """"""""""""""""""""""""""
3319 For geometry shaders, this semantic label indicates that a system value
3320 contains the current invocation id (i.e. gl_InvocationID).
3321 This is an integer value, and only the X component is used.
3323 TGSI_SEMANTIC_INSTANCEID
3324 """"""""""""""""""""""""
3326 For vertex shaders, this semantic label indicates that a system value contains
3327 the current instance id (i.e. gl_InstanceID). It does not include the base
3328 instance. This is an integer value, and only the X component is used.
3330 TGSI_SEMANTIC_VERTEXID
3331 """"""""""""""""""""""
3333 For vertex shaders, this semantic label indicates that a system value contains
3334 the current vertex id (i.e. gl_VertexID). It does (unlike in d3d10) include the
3335 base vertex. This is an integer value, and only the X component is used.
3337 TGSI_SEMANTIC_VERTEXID_NOBASE
3338 """""""""""""""""""""""""""""""
3340 For vertex shaders, this semantic label indicates that a system value contains
3341 the current vertex id without including the base vertex (this corresponds to
3342 d3d10 vertex id, so TGSI_SEMANTIC_VERTEXID_NOBASE + TGSI_SEMANTIC_BASEVERTEX
3343 == TGSI_SEMANTIC_VERTEXID). This is an integer value, and only the X component
3346 TGSI_SEMANTIC_BASEVERTEX
3347 """"""""""""""""""""""""
3349 For vertex shaders, this semantic label indicates that a system value contains
3350 the base vertex (i.e. gl_BaseVertex). Note that for non-indexed draw calls,
3351 this contains the first (or start) value instead.
3352 This is an integer value, and only the X component is used.
3354 TGSI_SEMANTIC_PRIMID
3355 """"""""""""""""""""
3357 For geometry and fragment shaders, this semantic label indicates the value
3358 contains the primitive id (i.e. gl_PrimitiveID). This is an integer value,
3359 and only the X component is used.
3360 FIXME: This right now can be either a ordinary input or a system value...
3366 For tessellation evaluation/control shaders, this semantic label indicates a
3367 generic per-patch attribute. Such semantics will not implicitly be per-vertex
3370 TGSI_SEMANTIC_TESSCOORD
3371 """""""""""""""""""""""
3373 For tessellation evaluation shaders, this semantic label indicates the
3374 coordinates of the vertex being processed. This is available in XYZ; W is
3377 TGSI_SEMANTIC_TESSOUTER
3378 """""""""""""""""""""""
3380 For tessellation evaluation/control shaders, this semantic label indicates the
3381 outer tessellation levels of the patch. Isoline tessellation will only have XY
3382 defined, triangle will have XYZ and quads will have XYZW defined. This
3383 corresponds to gl_TessLevelOuter.
3385 TGSI_SEMANTIC_TESSINNER
3386 """""""""""""""""""""""
3388 For tessellation evaluation/control shaders, this semantic label indicates the
3389 inner tessellation levels of the patch. The X value is only defined for
3390 triangle tessellation, while quads will have XY defined. This is entirely
3391 undefined for isoline tessellation.
3393 TGSI_SEMANTIC_VERTICESIN
3394 """"""""""""""""""""""""
3396 For tessellation evaluation/control shaders, this semantic label indicates the
3397 number of vertices provided in the input patch. Only the X value is defined.
3399 TGSI_SEMANTIC_HELPER_INVOCATION
3400 """""""""""""""""""""""""""""""
3402 For fragment shaders, this semantic indicates whether the current
3403 invocation is covered or not. Helper invocations are created in order
3404 to properly compute derivatives, however it may be desirable to skip
3405 some of the logic in those cases. See ``gl_HelperInvocation`` documentation.
3407 TGSI_SEMANTIC_BASEINSTANCE
3408 """"""""""""""""""""""""""
3410 For vertex shaders, the base instance argument supplied for this
3411 draw. This is an integer value, and only the X component is used.
3413 TGSI_SEMANTIC_DRAWID
3414 """"""""""""""""""""
3416 For vertex shaders, the zero-based index of the current draw in a
3417 ``glMultiDraw*`` invocation. This is an integer value, and only the X
3421 TGSI_SEMANTIC_WORK_DIM
3422 """"""""""""""""""""""
3424 For compute shaders started via opencl this retrieves the work_dim
3425 parameter to the clEnqueueNDRangeKernel call with which the shader
3429 TGSI_SEMANTIC_GRID_SIZE
3430 """""""""""""""""""""""
3432 For compute shaders, this semantic indicates the maximum (x, y, z) dimensions
3433 of a grid of thread blocks.
3436 TGSI_SEMANTIC_BLOCK_ID
3437 """"""""""""""""""""""
3439 For compute shaders, this semantic indicates the (x, y, z) coordinates of the
3440 current block inside of the grid.
3443 TGSI_SEMANTIC_BLOCK_SIZE
3444 """"""""""""""""""""""""
3446 For compute shaders, this semantic indicates the maximum (x, y, z) dimensions
3447 of a block in threads.
3450 TGSI_SEMANTIC_THREAD_ID
3451 """""""""""""""""""""""
3453 For compute shaders, this semantic indicates the (x, y, z) coordinates of the
3454 current thread inside of the block.
3457 TGSI_SEMANTIC_SUBGROUP_SIZE
3458 """""""""""""""""""""""""""
3460 This semantic indicates the subgroup size for the current invocation. This is
3461 an integer of at most 64, as it indicates the width of lanemasks. It does not
3462 depend on the number of invocations that are active.
3465 TGSI_SEMANTIC_SUBGROUP_INVOCATION
3466 """""""""""""""""""""""""""""""""
3468 The index of the current invocation within its subgroup.
3471 TGSI_SEMANTIC_SUBGROUP_EQ_MASK
3472 """"""""""""""""""""""""""""""
3474 A bit mask of ``bit index == TGSI_SEMANTIC_SUBGROUP_INVOCATION``, i.e.
3475 ``1 << subgroup_invocation`` in arbitrary precision arithmetic.
3478 TGSI_SEMANTIC_SUBGROUP_GE_MASK
3479 """"""""""""""""""""""""""""""
3481 A bit mask of ``bit index >= TGSI_SEMANTIC_SUBGROUP_INVOCATION``, i.e.
3482 ``((1 << (subgroup_size - subgroup_invocation)) - 1) << subgroup_invocation``
3483 in arbitrary precision arithmetic.
3486 TGSI_SEMANTIC_SUBGROUP_GT_MASK
3487 """"""""""""""""""""""""""""""
3489 A bit mask of ``bit index > TGSI_SEMANTIC_SUBGROUP_INVOCATION``, i.e.
3490 ``((1 << (subgroup_size - subgroup_invocation - 1)) - 1) << (subgroup_invocation + 1)``
3491 in arbitrary precision arithmetic.
3494 TGSI_SEMANTIC_SUBGROUP_LE_MASK
3495 """"""""""""""""""""""""""""""
3497 A bit mask of ``bit index <= TGSI_SEMANTIC_SUBGROUP_INVOCATION``, i.e.
3498 ``(1 << (subgroup_invocation + 1)) - 1`` in arbitrary precision arithmetic.
3501 TGSI_SEMANTIC_SUBGROUP_LT_MASK
3502 """"""""""""""""""""""""""""""
3504 A bit mask of ``bit index > TGSI_SEMANTIC_SUBGROUP_INVOCATION``, i.e.
3505 ``(1 << subgroup_invocation) - 1`` in arbitrary precision arithmetic.
3508 Declaration Interpolate
3509 ^^^^^^^^^^^^^^^^^^^^^^^
3511 This token is only valid for fragment shader INPUT declarations.
3513 The Interpolate field specifes the way input is being interpolated by
3514 the rasteriser and is one of TGSI_INTERPOLATE_*.
3516 The Location field specifies the location inside the pixel that the
3517 interpolation should be done at, one of ``TGSI_INTERPOLATE_LOC_*``. Note that
3518 when per-sample shading is enabled, the implementation may choose to
3519 interpolate at the sample irrespective of the Location field.
3521 The CylindricalWrap bitfield specifies which register components
3522 should be subject to cylindrical wrapping when interpolating by the
3523 rasteriser. If TGSI_CYLINDRICAL_WRAP_X is set to 1, the X component
3524 should be interpolated according to cylindrical wrapping rules.
3527 Declaration Sampler View
3528 ^^^^^^^^^^^^^^^^^^^^^^^^
3530 Follows Declaration token if file is TGSI_FILE_SAMPLER_VIEW.
3532 DCL SVIEW[#], resource, type(s)
3534 Declares a shader input sampler view and assigns it to a SVIEW[#]
3537 resource can be one of BUFFER, 1D, 2D, 3D, 1DArray and 2DArray.
3539 type must be 1 or 4 entries (if specifying on a per-component
3540 level) out of UNORM, SNORM, SINT, UINT and FLOAT.
3542 For TEX\* style texture sample opcodes (as opposed to SAMPLE\* opcodes
3543 which take an explicit SVIEW[#] source register), there may be optionally
3544 SVIEW[#] declarations. In this case, the SVIEW index is implied by the
3545 SAMP index, and there must be a corresponding SVIEW[#] declaration for
3546 each SAMP[#] declaration. Drivers are free to ignore this if they wish.
3547 But note in particular that some drivers need to know the sampler type
3548 (float/int/unsigned) in order to generate the correct code, so cases
3549 where integer textures are sampled, SVIEW[#] declarations should be
3552 NOTE: It is NOT legal to mix SAMPLE\* style opcodes and TEX\* opcodes
3555 Declaration Resource
3556 ^^^^^^^^^^^^^^^^^^^^
3558 Follows Declaration token if file is TGSI_FILE_RESOURCE.
3560 DCL RES[#], resource [, WR] [, RAW]
3562 Declares a shader input resource and assigns it to a RES[#]
3565 resource can be one of BUFFER, 1D, 2D, 3D, CUBE, 1DArray and
3568 If the RAW keyword is not specified, the texture data will be
3569 subject to conversion, swizzling and scaling as required to yield
3570 the specified data type from the physical data format of the bound
3573 If the RAW keyword is specified, no channel conversion will be
3574 performed: the values read for each of the channels (X,Y,Z,W) will
3575 correspond to consecutive words in the same order and format
3576 they're found in memory. No element-to-address conversion will be
3577 performed either: the value of the provided X coordinate will be
3578 interpreted in byte units instead of texel units. The result of
3579 accessing a misaligned address is undefined.
3581 Usage of the STORE opcode is only allowed if the WR (writable) flag
3586 ^^^^^^^^^^^^^^^^^^^^^^^^
3588 Properties are general directives that apply to the whole TGSI program.
3593 Specifies the fragment shader TGSI_SEMANTIC_POSITION coordinate origin.
3594 The default value is UPPER_LEFT.
3596 If UPPER_LEFT, the position will be (0,0) at the upper left corner and
3597 increase downward and rightward.
3598 If LOWER_LEFT, the position will be (0,0) at the lower left corner and
3599 increase upward and rightward.
3601 OpenGL defaults to LOWER_LEFT, and is configurable with the
3602 GL_ARB_fragment_coord_conventions extension.
3604 DirectX 9/10 use UPPER_LEFT.
3606 FS_COORD_PIXEL_CENTER
3607 """""""""""""""""""""
3609 Specifies the fragment shader TGSI_SEMANTIC_POSITION pixel center convention.
3610 The default value is HALF_INTEGER.
3612 If HALF_INTEGER, the fractionary part of the position will be 0.5
3613 If INTEGER, the fractionary part of the position will be 0.0
3615 Note that this does not affect the set of fragments generated by
3616 rasterization, which is instead controlled by half_pixel_center in the
3619 OpenGL defaults to HALF_INTEGER, and is configurable with the
3620 GL_ARB_fragment_coord_conventions extension.
3622 DirectX 9 uses INTEGER.
3623 DirectX 10 uses HALF_INTEGER.
3625 FS_COLOR0_WRITES_ALL_CBUFS
3626 """"""""""""""""""""""""""
3627 Specifies that writes to the fragment shader color 0 are replicated to all
3628 bound cbufs. This facilitates OpenGL's fragColor output vs fragData[0] where
3629 fragData is directed to a single color buffer, but fragColor is broadcast.
3632 """"""""""""""""""""""""""
3633 If this property is set on the program bound to the shader stage before the
3634 fragment shader, user clip planes should have no effect (be disabled) even if
3635 that shader does not write to any clip distance outputs and the rasterizer's
3636 clip_plane_enable is non-zero.
3637 This property is only supported by drivers that also support shader clip
3639 This is useful for APIs that don't have UCPs and where clip distances written
3640 by a shader cannot be disabled.
3645 Specifies the number of times a geometry shader should be executed for each
3646 input primitive. Each invocation will have a different
3647 TGSI_SEMANTIC_INVOCATIONID system value set. If not specified, assumed to
3650 VS_WINDOW_SPACE_POSITION
3651 """"""""""""""""""""""""""
3652 If this property is set on the vertex shader, the TGSI_SEMANTIC_POSITION output
3653 is assumed to contain window space coordinates.
3654 Division of X,Y,Z by W and the viewport transformation are disabled, and 1/W is
3655 directly taken from the 4-th component of the shader output.
3656 Naturally, clipping is not performed on window coordinates either.
3657 The effect of this property is undefined if a geometry or tessellation shader
3663 The number of vertices written by the tessellation control shader. This
3664 effectively defines the patch input size of the tessellation evaluation shader
3670 This sets the tessellation primitive mode, one of ``PIPE_PRIM_TRIANGLES``,
3671 ``PIPE_PRIM_QUADS``, or ``PIPE_PRIM_LINES``. (Unlike in GL, there is no
3672 separate isolines settings, the regular lines is assumed to mean isolines.)
3677 This sets the spacing mode of the tessellation generator, one of
3678 ``PIPE_TESS_SPACING_*``.
3683 This sets the vertex order to be clockwise if the value is 1, or
3684 counter-clockwise if set to 0.
3689 If set to a non-zero value, this turns on point mode for the tessellator,
3690 which means that points will be generated instead of primitives.
3692 NUM_CLIPDIST_ENABLED
3693 """"""""""""""""""""
3695 How many clip distance scalar outputs are enabled.
3697 NUM_CULLDIST_ENABLED
3698 """"""""""""""""""""
3700 How many cull distance scalar outputs are enabled.
3702 FS_EARLY_DEPTH_STENCIL
3703 """"""""""""""""""""""
3705 Whether depth test, stencil test, and occlusion query should run before
3706 the fragment shader (regardless of fragment shader side effects). Corresponds
3707 to GLSL early_fragment_tests.
3712 Which shader stage will MOST LIKELY follow after this shader when the shader
3713 is bound. This is only a hint to the driver and doesn't have to be precise.
3714 Only set for VS and TES.
3716 CS_FIXED_BLOCK_WIDTH / HEIGHT / DEPTH
3717 """""""""""""""""""""""""""""""""""""
3719 Threads per block in each dimension, if known at compile time. If the block size
3720 is known all three should be at least 1. If it is unknown they should all be set
3726 The MUL TGSI operation (FP32 multiplication) will return 0 if either
3727 of the operands are equal to 0. That means that 0 * Inf = 0. This
3728 should be set the same way for an entire pipeline. Note that this
3729 applies not only to the literal MUL TGSI opcode, but all FP32
3730 multiplications implied by other operations, such as MAD, FMA, DP2,
3731 DP3, DP4, DPH, DST, LOG, LRP, XPD, and possibly others. If there is a
3732 mismatch between shaders, then it is unspecified whether this behavior
3735 FS_POST_DEPTH_COVERAGE
3736 """"""""""""""""""""""
3738 When enabled, the input for TGSI_SEMANTIC_SAMPLEMASK will exclude samples
3739 that have failed the depth/stencil tests. This is only valid when
3740 FS_EARLY_DEPTH_STENCIL is also specified.
3743 Texture Sampling and Texture Formats
3744 ------------------------------------
3746 This table shows how texture image components are returned as (x,y,z,w) tuples
3747 by TGSI texture instructions, such as :opcode:`TEX`, :opcode:`TXD`, and
3748 :opcode:`TXP`. For reference, OpenGL and Direct3D conventions are shown as
3751 +--------------------+--------------+--------------------+--------------+
3752 | Texture Components | Gallium | OpenGL | Direct3D 9 |
3753 +====================+==============+====================+==============+
3754 | R | (r, 0, 0, 1) | (r, 0, 0, 1) | (r, 1, 1, 1) |
3755 +--------------------+--------------+--------------------+--------------+
3756 | RG | (r, g, 0, 1) | (r, g, 0, 1) | (r, g, 1, 1) |
3757 +--------------------+--------------+--------------------+--------------+
3758 | RGB | (r, g, b, 1) | (r, g, b, 1) | (r, g, b, 1) |
3759 +--------------------+--------------+--------------------+--------------+
3760 | RGBA | (r, g, b, a) | (r, g, b, a) | (r, g, b, a) |
3761 +--------------------+--------------+--------------------+--------------+
3762 | A | (0, 0, 0, a) | (0, 0, 0, a) | (0, 0, 0, a) |
3763 +--------------------+--------------+--------------------+--------------+
3764 | L | (l, l, l, 1) | (l, l, l, 1) | (l, l, l, 1) |
3765 +--------------------+--------------+--------------------+--------------+
3766 | LA | (l, l, l, a) | (l, l, l, a) | (l, l, l, a) |
3767 +--------------------+--------------+--------------------+--------------+
3768 | I | (i, i, i, i) | (i, i, i, i) | N/A |
3769 +--------------------+--------------+--------------------+--------------+
3770 | UV | XXX TBD | (0, 0, 0, 1) | (u, v, 1, 1) |
3771 | | | [#envmap-bumpmap]_ | |
3772 +--------------------+--------------+--------------------+--------------+
3773 | Z | XXX TBD | (z, z, z, 1) | (0, z, 0, 1) |
3774 | | | [#depth-tex-mode]_ | |
3775 +--------------------+--------------+--------------------+--------------+
3776 | S | (s, s, s, s) | unknown | unknown |
3777 +--------------------+--------------+--------------------+--------------+
3779 .. [#envmap-bumpmap] http://www.opengl.org/registry/specs/ATI/envmap_bumpmap.txt
3780 .. [#depth-tex-mode] the default is (z, z, z, 1) but may also be (0, 0, 0, z)
3781 or (z, z, z, z) depending on the value of GL_DEPTH_TEXTURE_MODE.