4 TGSI, Tungsten Graphics Shader Infrastructure, is an intermediate language
5 for describing shaders. Since Gallium is inherently shaderful, shaders are
6 an important part of the API. TGSI is the only intermediate representation
12 All TGSI instructions, known as *opcodes*, operate on arbitrary-precision
13 floating-point four-component vectors. An opcode may have up to one
14 destination register, known as *dst*, and between zero and three source
15 registers, called *src0* through *src2*, or simply *src* if there is only
18 Some instructions, like :opcode:`I2F`, permit re-interpretation of vector
19 components as integers. Other instructions permit using registers as
20 two-component vectors with double precision; see :ref:`doubleopcodes`.
22 When an instruction has a scalar result, the result is usually copied into
23 each of the components of *dst*. When this happens, the result is said to be
24 *replicated* to *dst*. :opcode:`RCP` is one such instruction.
29 TGSI supports modifiers on inputs (as well as saturate modifier on instructions).
31 For inputs which have a floating point type, both absolute value and negation
32 modifiers are supported (with absolute value being applied first).
33 TGSI_OPCODE_MOV is considered to have float input type for applying modifiers.
35 For inputs which have signed or unsigned type only the negate modifier is
42 ^^^^^^^^^^^^^^^^^^^^^^^^^
44 These opcodes are guaranteed to be available regardless of the driver being
47 .. opcode:: ARL - Address Register Load
51 dst.x = (int) \lfloor src.x\rfloor
53 dst.y = (int) \lfloor src.y\rfloor
55 dst.z = (int) \lfloor src.z\rfloor
57 dst.w = (int) \lfloor src.w\rfloor
60 .. opcode:: MOV - Move
73 .. opcode:: LIT - Light Coefficients
78 dst.y &= max(src.x, 0) \\
79 dst.z &= (src.x > 0) ? max(src.y, 0)^{clamp(src.w, -128, 128))} : 0 \\
83 .. opcode:: RCP - Reciprocal
85 This instruction replicates its result.
92 .. opcode:: RSQ - Reciprocal Square Root
94 This instruction replicates its result. The results are undefined for src <= 0.
98 dst = \frac{1}{\sqrt{src.x}}
101 .. opcode:: SQRT - Square Root
103 This instruction replicates its result. The results are undefined for src < 0.
110 .. opcode:: EXP - Approximate Exponential Base 2
114 dst.x &= 2^{\lfloor src.x\rfloor} \\
115 dst.y &= src.x - \lfloor src.x\rfloor \\
116 dst.z &= 2^{src.x} \\
120 .. opcode:: LOG - Approximate Logarithm Base 2
124 dst.x &= \lfloor\log_2{|src.x|}\rfloor \\
125 dst.y &= \frac{|src.x|}{2^{\lfloor\log_2{|src.x|}\rfloor}} \\
126 dst.z &= \log_2{|src.x|} \\
130 .. opcode:: MUL - Multiply
134 dst.x = src0.x \times src1.x
136 dst.y = src0.y \times src1.y
138 dst.z = src0.z \times src1.z
140 dst.w = src0.w \times src1.w
143 .. opcode:: ADD - Add
147 dst.x = src0.x + src1.x
149 dst.y = src0.y + src1.y
151 dst.z = src0.z + src1.z
153 dst.w = src0.w + src1.w
156 .. opcode:: DP3 - 3-component Dot Product
158 This instruction replicates its result.
162 dst = src0.x \times src1.x + src0.y \times src1.y + src0.z \times src1.z
165 .. opcode:: DP4 - 4-component Dot Product
167 This instruction replicates its result.
171 dst = src0.x \times src1.x + src0.y \times src1.y + src0.z \times src1.z + src0.w \times src1.w
174 .. opcode:: DST - Distance Vector
179 dst.y &= src0.y \times src1.y\\
184 .. opcode:: MIN - Minimum
188 dst.x = min(src0.x, src1.x)
190 dst.y = min(src0.y, src1.y)
192 dst.z = min(src0.z, src1.z)
194 dst.w = min(src0.w, src1.w)
197 .. opcode:: MAX - Maximum
201 dst.x = max(src0.x, src1.x)
203 dst.y = max(src0.y, src1.y)
205 dst.z = max(src0.z, src1.z)
207 dst.w = max(src0.w, src1.w)
210 .. opcode:: SLT - Set On Less Than
214 dst.x = (src0.x < src1.x) ? 1.0F : 0.0F
216 dst.y = (src0.y < src1.y) ? 1.0F : 0.0F
218 dst.z = (src0.z < src1.z) ? 1.0F : 0.0F
220 dst.w = (src0.w < src1.w) ? 1.0F : 0.0F
223 .. opcode:: SGE - Set On Greater Equal Than
227 dst.x = (src0.x >= src1.x) ? 1.0F : 0.0F
229 dst.y = (src0.y >= src1.y) ? 1.0F : 0.0F
231 dst.z = (src0.z >= src1.z) ? 1.0F : 0.0F
233 dst.w = (src0.w >= src1.w) ? 1.0F : 0.0F
236 .. opcode:: MAD - Multiply And Add
240 dst.x = src0.x \times src1.x + src2.x
242 dst.y = src0.y \times src1.y + src2.y
244 dst.z = src0.z \times src1.z + src2.z
246 dst.w = src0.w \times src1.w + src2.w
249 .. opcode:: LRP - Linear Interpolate
253 dst.x = src0.x \times src1.x + (1 - src0.x) \times src2.x
255 dst.y = src0.y \times src1.y + (1 - src0.y) \times src2.y
257 dst.z = src0.z \times src1.z + (1 - src0.z) \times src2.z
259 dst.w = src0.w \times src1.w + (1 - src0.w) \times src2.w
262 .. opcode:: FMA - Fused Multiply-Add
264 Perform a * b + c with no intermediate rounding step.
268 dst.x = src0.x \times src1.x + src2.x
270 dst.y = src0.y \times src1.y + src2.y
272 dst.z = src0.z \times src1.z + src2.z
274 dst.w = src0.w \times src1.w + src2.w
277 .. opcode:: DP2A - 2-component Dot Product And Add
281 dst.x = src0.x \times src1.x + src0.y \times src1.y + src2.x
283 dst.y = src0.x \times src1.x + src0.y \times src1.y + src2.x
285 dst.z = src0.x \times src1.x + src0.y \times src1.y + src2.x
287 dst.w = src0.x \times src1.x + src0.y \times src1.y + src2.x
290 .. opcode:: FRC - Fraction
294 dst.x = src.x - \lfloor src.x\rfloor
296 dst.y = src.y - \lfloor src.y\rfloor
298 dst.z = src.z - \lfloor src.z\rfloor
300 dst.w = src.w - \lfloor src.w\rfloor
303 .. opcode:: CLAMP - Clamp
307 dst.x = clamp(src0.x, src1.x, src2.x)
309 dst.y = clamp(src0.y, src1.y, src2.y)
311 dst.z = clamp(src0.z, src1.z, src2.z)
313 dst.w = clamp(src0.w, src1.w, src2.w)
316 .. opcode:: FLR - Floor
320 dst.x = \lfloor src.x\rfloor
322 dst.y = \lfloor src.y\rfloor
324 dst.z = \lfloor src.z\rfloor
326 dst.w = \lfloor src.w\rfloor
329 .. opcode:: ROUND - Round
342 .. opcode:: EX2 - Exponential Base 2
344 This instruction replicates its result.
351 .. opcode:: LG2 - Logarithm Base 2
353 This instruction replicates its result.
360 .. opcode:: POW - Power
362 This instruction replicates its result.
366 dst = src0.x^{src1.x}
368 .. opcode:: XPD - Cross Product
372 dst.x = src0.y \times src1.z - src1.y \times src0.z
374 dst.y = src0.z \times src1.x - src1.z \times src0.x
376 dst.z = src0.x \times src1.y - src1.x \times src0.y
381 .. opcode:: DPH - Homogeneous Dot Product
383 This instruction replicates its result.
387 dst = src0.x \times src1.x + src0.y \times src1.y + src0.z \times src1.z + src1.w
390 .. opcode:: COS - Cosine
392 This instruction replicates its result.
399 .. opcode:: DDX, DDX_FINE - Derivative Relative To X
401 The fine variant is only used when ``PIPE_CAP_TGSI_FS_FINE_DERIVATIVE`` is
402 advertised. When it is, the fine version guarantees one derivative per row
403 while DDX is allowed to be the same for the entire 2x2 quad.
407 dst.x = partialx(src.x)
409 dst.y = partialx(src.y)
411 dst.z = partialx(src.z)
413 dst.w = partialx(src.w)
416 .. opcode:: DDY, DDY_FINE - Derivative Relative To Y
418 The fine variant is only used when ``PIPE_CAP_TGSI_FS_FINE_DERIVATIVE`` is
419 advertised. When it is, the fine version guarantees one derivative per column
420 while DDY is allowed to be the same for the entire 2x2 quad.
424 dst.x = partialy(src.x)
426 dst.y = partialy(src.y)
428 dst.z = partialy(src.z)
430 dst.w = partialy(src.w)
433 .. opcode:: PK2H - Pack Two 16-bit Floats
435 This instruction replicates its result.
439 dst = f32\_to\_f16(src.x) | f32\_to\_f16(src.y) << 16
442 .. opcode:: PK2US - Pack Two Unsigned 16-bit Scalars
447 .. opcode:: PK4B - Pack Four Signed 8-bit Scalars
452 .. opcode:: PK4UB - Pack Four Unsigned 8-bit Scalars
457 .. opcode:: SEQ - Set On Equal
461 dst.x = (src0.x == src1.x) ? 1.0F : 0.0F
463 dst.y = (src0.y == src1.y) ? 1.0F : 0.0F
465 dst.z = (src0.z == src1.z) ? 1.0F : 0.0F
467 dst.w = (src0.w == src1.w) ? 1.0F : 0.0F
470 .. opcode:: SGT - Set On Greater Than
474 dst.x = (src0.x > src1.x) ? 1.0F : 0.0F
476 dst.y = (src0.y > src1.y) ? 1.0F : 0.0F
478 dst.z = (src0.z > src1.z) ? 1.0F : 0.0F
480 dst.w = (src0.w > src1.w) ? 1.0F : 0.0F
483 .. opcode:: SIN - Sine
485 This instruction replicates its result.
492 .. opcode:: SLE - Set On Less Equal Than
496 dst.x = (src0.x <= src1.x) ? 1.0F : 0.0F
498 dst.y = (src0.y <= src1.y) ? 1.0F : 0.0F
500 dst.z = (src0.z <= src1.z) ? 1.0F : 0.0F
502 dst.w = (src0.w <= src1.w) ? 1.0F : 0.0F
505 .. opcode:: SNE - Set On Not Equal
509 dst.x = (src0.x != src1.x) ? 1.0F : 0.0F
511 dst.y = (src0.y != src1.y) ? 1.0F : 0.0F
513 dst.z = (src0.z != src1.z) ? 1.0F : 0.0F
515 dst.w = (src0.w != src1.w) ? 1.0F : 0.0F
518 .. opcode:: TEX - Texture Lookup
520 for array textures src0.y contains the slice for 1D,
521 and src0.z contain the slice for 2D.
523 for shadow textures with no arrays (and not cube map),
524 src0.z contains the reference value.
526 for shadow textures with arrays, src0.z contains
527 the reference value for 1D arrays, and src0.w contains
528 the reference value for 2D arrays and cube maps.
530 for cube map array shadow textures, the reference value
531 cannot be passed in src0.w, and TEX2 must be used instead.
537 shadow_ref = src0.z or src0.w (optional)
541 dst = texture\_sample(unit, coord, shadow_ref)
544 .. opcode:: TEX2 - Texture Lookup (for shadow cube map arrays only)
546 this is the same as TEX, but uses another reg to encode the
557 dst = texture\_sample(unit, coord, shadow_ref)
562 .. opcode:: TXD - Texture Lookup with Derivatives
574 dst = texture\_sample\_deriv(unit, coord, ddx, ddy)
577 .. opcode:: TXP - Projective Texture Lookup
581 coord.x = src0.x / src0.w
583 coord.y = src0.y / src0.w
585 coord.z = src0.z / src0.w
591 dst = texture\_sample(unit, coord)
594 .. opcode:: UP2H - Unpack Two 16-Bit Floats
598 dst.x = f16\_to\_f32(src0.x \& 0xffff)
600 dst.y = f16\_to\_f32(src0.x >> 16)
602 dst.z = f16\_to\_f32(src0.x \& 0xffff)
604 dst.w = f16\_to\_f32(src0.x >> 16)
608 Considered for removal.
610 .. opcode:: UP2US - Unpack Two Unsigned 16-Bit Scalars
616 Considered for removal.
618 .. opcode:: UP4B - Unpack Four Signed 8-Bit Values
624 Considered for removal.
626 .. opcode:: UP4UB - Unpack Four Unsigned 8-Bit Scalars
632 Considered for removal.
635 .. opcode:: ARR - Address Register Load With Round
639 dst.x = (int) round(src.x)
641 dst.y = (int) round(src.y)
643 dst.z = (int) round(src.z)
645 dst.w = (int) round(src.w)
648 .. opcode:: SSG - Set Sign
652 dst.x = (src.x > 0) ? 1 : (src.x < 0) ? -1 : 0
654 dst.y = (src.y > 0) ? 1 : (src.y < 0) ? -1 : 0
656 dst.z = (src.z > 0) ? 1 : (src.z < 0) ? -1 : 0
658 dst.w = (src.w > 0) ? 1 : (src.w < 0) ? -1 : 0
661 .. opcode:: CMP - Compare
665 dst.x = (src0.x < 0) ? src1.x : src2.x
667 dst.y = (src0.y < 0) ? src1.y : src2.y
669 dst.z = (src0.z < 0) ? src1.z : src2.z
671 dst.w = (src0.w < 0) ? src1.w : src2.w
674 .. opcode:: KILL_IF - Conditional Discard
676 Conditional discard. Allowed in fragment shaders only.
680 if (src.x < 0 || src.y < 0 || src.z < 0 || src.w < 0)
685 .. opcode:: KILL - Discard
687 Unconditional discard. Allowed in fragment shaders only.
690 .. opcode:: SCS - Sine Cosine
703 .. opcode:: TXB - Texture Lookup With Bias
705 for cube map array textures and shadow cube maps, the bias value
706 cannot be passed in src0.w, and TXB2 must be used instead.
708 if the target is a shadow texture, the reference value is always
709 in src.z (this prevents shadow 3d and shadow 2d arrays from
710 using this instruction, but this is not needed).
726 dst = texture\_sample(unit, coord, bias)
729 .. opcode:: TXB2 - Texture Lookup With Bias (some cube maps only)
731 this is the same as TXB, but uses another reg to encode the
732 lod bias value for cube map arrays and shadow cube maps.
733 Presumably shadow 2d arrays and shadow 3d targets could use
734 this encoding too, but this is not legal.
736 shadow cube map arrays are neither possible nor required.
746 dst = texture\_sample(unit, coord, bias)
749 .. opcode:: DIV - Divide
753 dst.x = \frac{src0.x}{src1.x}
755 dst.y = \frac{src0.y}{src1.y}
757 dst.z = \frac{src0.z}{src1.z}
759 dst.w = \frac{src0.w}{src1.w}
762 .. opcode:: DP2 - 2-component Dot Product
764 This instruction replicates its result.
768 dst = src0.x \times src1.x + src0.y \times src1.y
771 .. opcode:: TXL - Texture Lookup With explicit LOD
773 for cube map array textures, the explicit lod value
774 cannot be passed in src0.w, and TXL2 must be used instead.
776 if the target is a shadow texture, the reference value is always
777 in src.z (this prevents shadow 3d / 2d array / cube targets from
778 using this instruction, but this is not needed).
794 dst = texture\_sample(unit, coord, lod)
797 .. opcode:: TXL2 - Texture Lookup With explicit LOD (for cube map arrays only)
799 this is the same as TXL, but uses another reg to encode the
801 Presumably shadow 3d / 2d array / cube targets could use
802 this encoding too, but this is not legal.
804 shadow cube map arrays are neither possible nor required.
814 dst = texture\_sample(unit, coord, lod)
817 .. opcode:: PUSHA - Push Address Register On Stack
826 Considered for cleanup.
830 Considered for removal.
832 .. opcode:: POPA - Pop Address Register From Stack
841 Considered for cleanup.
845 Considered for removal.
848 .. opcode:: CALLNZ - Subroutine Call If Not Zero
854 Considered for cleanup.
858 Considered for removal.
862 ^^^^^^^^^^^^^^^^^^^^^^^^
864 These opcodes are primarily provided for special-use computational shaders.
865 Support for these opcodes indicated by a special pipe capability bit (TBD).
867 XXX doesn't look like most of the opcodes really belong here.
869 .. opcode:: CEIL - Ceiling
873 dst.x = \lceil src.x\rceil
875 dst.y = \lceil src.y\rceil
877 dst.z = \lceil src.z\rceil
879 dst.w = \lceil src.w\rceil
882 .. opcode:: TRUNC - Truncate
895 .. opcode:: MOD - Modulus
899 dst.x = src0.x \bmod src1.x
901 dst.y = src0.y \bmod src1.y
903 dst.z = src0.z \bmod src1.z
905 dst.w = src0.w \bmod src1.w
908 .. opcode:: UARL - Integer Address Register Load
910 Moves the contents of the source register, assumed to be an integer, into the
911 destination register, which is assumed to be an address (ADDR) register.
914 .. opcode:: SAD - Sum Of Absolute Differences
918 dst.x = |src0.x - src1.x| + src2.x
920 dst.y = |src0.y - src1.y| + src2.y
922 dst.z = |src0.z - src1.z| + src2.z
924 dst.w = |src0.w - src1.w| + src2.w
927 .. opcode:: TXF - Texel Fetch
929 As per NV_gpu_shader4, extract a single texel from a specified texture
930 image. The source sampler may not be a CUBE or SHADOW. src 0 is a
931 four-component signed integer vector used to identify the single texel
932 accessed. 3 components + level. Just like texture instructions, an optional
933 offset vector is provided, which is subject to various driver restrictions
934 (regarding range, source of offsets).
935 TXF(uint_vec coord, int_vec offset).
938 .. opcode:: TXQ - Texture Size Query
940 As per NV_gpu_program4, retrieve the dimensions of the texture depending on
941 the target. For 1D (width), 2D/RECT/CUBE (width, height), 3D (width, height,
942 depth), 1D array (width, layers), 2D array (width, height, layers).
943 Also return the number of accessible levels (last_level - first_level + 1)
946 For components which don't return a resource dimension, their value
953 dst.x = texture\_width(unit, lod)
955 dst.y = texture\_height(unit, lod)
957 dst.z = texture\_depth(unit, lod)
959 dst.w = texture\_levels(unit)
962 .. opcode:: TXQS - Texture Samples Query
964 This retrieves the number of samples in the texture, and stores it
965 into the x component. The other components are undefined.
969 dst.x = texture\_samples(unit)
972 .. opcode:: TG4 - Texture Gather
974 As per ARB_texture_gather, gathers the four texels to be used in a bi-linear
975 filtering operation and packs them into a single register. Only works with
976 2D, 2D array, cubemaps, and cubemaps arrays. For 2D textures, only the
977 addressing modes of the sampler and the top level of any mip pyramid are
978 used. Set W to zero. It behaves like the TEX instruction, but a filtered
979 sample is not generated. The four samples that contribute to filtering are
980 placed into xyzw in clockwise order, starting with the (u,v) texture
981 coordinate delta at the following locations (-, +), (+, +), (+, -), (-, -),
982 where the magnitude of the deltas are half a texel.
984 PIPE_CAP_TEXTURE_SM5 enhances this instruction to support shadow per-sample
985 depth compares, single component selection, and a non-constant offset. It
986 doesn't allow support for the GL independent offset to get i0,j0. This would
987 require another CAP is hw can do it natively. For now we lower that before
996 dst = texture\_gather4 (unit, coord, component)
998 (with SM5 - cube array shadow)
1006 dst = texture\_gather (uint, coord, compare)
1008 .. opcode:: LODQ - level of detail query
1010 Compute the LOD information that the texture pipe would use to access the
1011 texture. The Y component contains the computed LOD lambda_prime. The X
1012 component contains the LOD that will be accessed, based on min/max lod's
1019 dst.xy = lodq(uint, coord);
1022 ^^^^^^^^^^^^^^^^^^^^^^^^
1023 These opcodes are used for integer operations.
1024 Support for these opcodes indicated by PIPE_SHADER_CAP_INTEGERS (all of them?)
1027 .. opcode:: I2F - Signed Integer To Float
1029 Rounding is unspecified (round to nearest even suggested).
1033 dst.x = (float) src.x
1035 dst.y = (float) src.y
1037 dst.z = (float) src.z
1039 dst.w = (float) src.w
1042 .. opcode:: U2F - Unsigned Integer To Float
1044 Rounding is unspecified (round to nearest even suggested).
1048 dst.x = (float) src.x
1050 dst.y = (float) src.y
1052 dst.z = (float) src.z
1054 dst.w = (float) src.w
1057 .. opcode:: F2I - Float to Signed Integer
1059 Rounding is towards zero (truncate).
1060 Values outside signed range (including NaNs) produce undefined results.
1073 .. opcode:: F2U - Float to Unsigned Integer
1075 Rounding is towards zero (truncate).
1076 Values outside unsigned range (including NaNs) produce undefined results.
1080 dst.x = (unsigned) src.x
1082 dst.y = (unsigned) src.y
1084 dst.z = (unsigned) src.z
1086 dst.w = (unsigned) src.w
1089 .. opcode:: UADD - Integer Add
1091 This instruction works the same for signed and unsigned integers.
1092 The low 32bit of the result is returned.
1096 dst.x = src0.x + src1.x
1098 dst.y = src0.y + src1.y
1100 dst.z = src0.z + src1.z
1102 dst.w = src0.w + src1.w
1105 .. opcode:: UMAD - Integer Multiply And Add
1107 This instruction works the same for signed and unsigned integers.
1108 The multiplication returns the low 32bit (as does the result itself).
1112 dst.x = src0.x \times src1.x + src2.x
1114 dst.y = src0.y \times src1.y + src2.y
1116 dst.z = src0.z \times src1.z + src2.z
1118 dst.w = src0.w \times src1.w + src2.w
1121 .. opcode:: UMUL - Integer Multiply
1123 This instruction works the same for signed and unsigned integers.
1124 The low 32bit of the result is returned.
1128 dst.x = src0.x \times src1.x
1130 dst.y = src0.y \times src1.y
1132 dst.z = src0.z \times src1.z
1134 dst.w = src0.w \times src1.w
1137 .. opcode:: IMUL_HI - Signed Integer Multiply High Bits
1139 The high 32bits of the multiplication of 2 signed integers are returned.
1143 dst.x = (src0.x \times src1.x) >> 32
1145 dst.y = (src0.y \times src1.y) >> 32
1147 dst.z = (src0.z \times src1.z) >> 32
1149 dst.w = (src0.w \times src1.w) >> 32
1152 .. opcode:: UMUL_HI - Unsigned Integer Multiply High Bits
1154 The high 32bits of the multiplication of 2 unsigned integers are returned.
1158 dst.x = (src0.x \times src1.x) >> 32
1160 dst.y = (src0.y \times src1.y) >> 32
1162 dst.z = (src0.z \times src1.z) >> 32
1164 dst.w = (src0.w \times src1.w) >> 32
1167 .. opcode:: IDIV - Signed Integer Division
1169 TBD: behavior for division by zero.
1173 dst.x = src0.x \ src1.x
1175 dst.y = src0.y \ src1.y
1177 dst.z = src0.z \ src1.z
1179 dst.w = src0.w \ src1.w
1182 .. opcode:: UDIV - Unsigned Integer Division
1184 For division by zero, 0xffffffff is returned.
1188 dst.x = src0.x \ src1.x
1190 dst.y = src0.y \ src1.y
1192 dst.z = src0.z \ src1.z
1194 dst.w = src0.w \ src1.w
1197 .. opcode:: UMOD - Unsigned Integer Remainder
1199 If second arg is zero, 0xffffffff is returned.
1203 dst.x = src0.x \ src1.x
1205 dst.y = src0.y \ src1.y
1207 dst.z = src0.z \ src1.z
1209 dst.w = src0.w \ src1.w
1212 .. opcode:: NOT - Bitwise Not
1225 .. opcode:: AND - Bitwise And
1229 dst.x = src0.x \& src1.x
1231 dst.y = src0.y \& src1.y
1233 dst.z = src0.z \& src1.z
1235 dst.w = src0.w \& src1.w
1238 .. opcode:: OR - Bitwise Or
1242 dst.x = src0.x | src1.x
1244 dst.y = src0.y | src1.y
1246 dst.z = src0.z | src1.z
1248 dst.w = src0.w | src1.w
1251 .. opcode:: XOR - Bitwise Xor
1255 dst.x = src0.x \oplus src1.x
1257 dst.y = src0.y \oplus src1.y
1259 dst.z = src0.z \oplus src1.z
1261 dst.w = src0.w \oplus src1.w
1264 .. opcode:: IMAX - Maximum of Signed Integers
1268 dst.x = max(src0.x, src1.x)
1270 dst.y = max(src0.y, src1.y)
1272 dst.z = max(src0.z, src1.z)
1274 dst.w = max(src0.w, src1.w)
1277 .. opcode:: UMAX - Maximum of Unsigned Integers
1281 dst.x = max(src0.x, src1.x)
1283 dst.y = max(src0.y, src1.y)
1285 dst.z = max(src0.z, src1.z)
1287 dst.w = max(src0.w, src1.w)
1290 .. opcode:: IMIN - Minimum of Signed Integers
1294 dst.x = min(src0.x, src1.x)
1296 dst.y = min(src0.y, src1.y)
1298 dst.z = min(src0.z, src1.z)
1300 dst.w = min(src0.w, src1.w)
1303 .. opcode:: UMIN - Minimum of Unsigned Integers
1307 dst.x = min(src0.x, src1.x)
1309 dst.y = min(src0.y, src1.y)
1311 dst.z = min(src0.z, src1.z)
1313 dst.w = min(src0.w, src1.w)
1316 .. opcode:: SHL - Shift Left
1318 The shift count is masked with 0x1f before the shift is applied.
1322 dst.x = src0.x << (0x1f \& src1.x)
1324 dst.y = src0.y << (0x1f \& src1.y)
1326 dst.z = src0.z << (0x1f \& src1.z)
1328 dst.w = src0.w << (0x1f \& src1.w)
1331 .. opcode:: ISHR - Arithmetic Shift Right (of Signed Integer)
1333 The shift count is masked with 0x1f before the shift is applied.
1337 dst.x = src0.x >> (0x1f \& src1.x)
1339 dst.y = src0.y >> (0x1f \& src1.y)
1341 dst.z = src0.z >> (0x1f \& src1.z)
1343 dst.w = src0.w >> (0x1f \& src1.w)
1346 .. opcode:: USHR - Logical Shift Right
1348 The shift count is masked with 0x1f before the shift is applied.
1352 dst.x = src0.x >> (unsigned) (0x1f \& src1.x)
1354 dst.y = src0.y >> (unsigned) (0x1f \& src1.y)
1356 dst.z = src0.z >> (unsigned) (0x1f \& src1.z)
1358 dst.w = src0.w >> (unsigned) (0x1f \& src1.w)
1361 .. opcode:: UCMP - Integer Conditional Move
1365 dst.x = src0.x ? src1.x : src2.x
1367 dst.y = src0.y ? src1.y : src2.y
1369 dst.z = src0.z ? src1.z : src2.z
1371 dst.w = src0.w ? src1.w : src2.w
1375 .. opcode:: ISSG - Integer Set Sign
1379 dst.x = (src0.x < 0) ? -1 : (src0.x > 0) ? 1 : 0
1381 dst.y = (src0.y < 0) ? -1 : (src0.y > 0) ? 1 : 0
1383 dst.z = (src0.z < 0) ? -1 : (src0.z > 0) ? 1 : 0
1385 dst.w = (src0.w < 0) ? -1 : (src0.w > 0) ? 1 : 0
1389 .. opcode:: FSLT - Float Set On Less Than (ordered)
1391 Same comparison as SLT but returns integer instead of 1.0/0.0 float
1395 dst.x = (src0.x < src1.x) ? \sim 0 : 0
1397 dst.y = (src0.y < src1.y) ? \sim 0 : 0
1399 dst.z = (src0.z < src1.z) ? \sim 0 : 0
1401 dst.w = (src0.w < src1.w) ? \sim 0 : 0
1404 .. opcode:: ISLT - Signed Integer Set On Less Than
1408 dst.x = (src0.x < src1.x) ? \sim 0 : 0
1410 dst.y = (src0.y < src1.y) ? \sim 0 : 0
1412 dst.z = (src0.z < src1.z) ? \sim 0 : 0
1414 dst.w = (src0.w < src1.w) ? \sim 0 : 0
1417 .. opcode:: USLT - Unsigned Integer Set On Less Than
1421 dst.x = (src0.x < src1.x) ? \sim 0 : 0
1423 dst.y = (src0.y < src1.y) ? \sim 0 : 0
1425 dst.z = (src0.z < src1.z) ? \sim 0 : 0
1427 dst.w = (src0.w < src1.w) ? \sim 0 : 0
1430 .. opcode:: FSGE - Float Set On Greater Equal Than (ordered)
1432 Same comparison as SGE but returns integer instead of 1.0/0.0 float
1436 dst.x = (src0.x >= src1.x) ? \sim 0 : 0
1438 dst.y = (src0.y >= src1.y) ? \sim 0 : 0
1440 dst.z = (src0.z >= src1.z) ? \sim 0 : 0
1442 dst.w = (src0.w >= src1.w) ? \sim 0 : 0
1445 .. opcode:: ISGE - Signed Integer Set On Greater Equal Than
1449 dst.x = (src0.x >= src1.x) ? \sim 0 : 0
1451 dst.y = (src0.y >= src1.y) ? \sim 0 : 0
1453 dst.z = (src0.z >= src1.z) ? \sim 0 : 0
1455 dst.w = (src0.w >= src1.w) ? \sim 0 : 0
1458 .. opcode:: USGE - Unsigned Integer Set On Greater Equal Than
1462 dst.x = (src0.x >= src1.x) ? \sim 0 : 0
1464 dst.y = (src0.y >= src1.y) ? \sim 0 : 0
1466 dst.z = (src0.z >= src1.z) ? \sim 0 : 0
1468 dst.w = (src0.w >= src1.w) ? \sim 0 : 0
1471 .. opcode:: FSEQ - Float Set On Equal (ordered)
1473 Same comparison as SEQ but returns integer instead of 1.0/0.0 float
1477 dst.x = (src0.x == src1.x) ? \sim 0 : 0
1479 dst.y = (src0.y == src1.y) ? \sim 0 : 0
1481 dst.z = (src0.z == src1.z) ? \sim 0 : 0
1483 dst.w = (src0.w == src1.w) ? \sim 0 : 0
1486 .. opcode:: USEQ - Integer Set On Equal
1490 dst.x = (src0.x == src1.x) ? \sim 0 : 0
1492 dst.y = (src0.y == src1.y) ? \sim 0 : 0
1494 dst.z = (src0.z == src1.z) ? \sim 0 : 0
1496 dst.w = (src0.w == src1.w) ? \sim 0 : 0
1499 .. opcode:: FSNE - Float Set On Not Equal (unordered)
1501 Same comparison as SNE but returns integer instead of 1.0/0.0 float
1505 dst.x = (src0.x != src1.x) ? \sim 0 : 0
1507 dst.y = (src0.y != src1.y) ? \sim 0 : 0
1509 dst.z = (src0.z != src1.z) ? \sim 0 : 0
1511 dst.w = (src0.w != src1.w) ? \sim 0 : 0
1514 .. opcode:: USNE - Integer Set On Not Equal
1518 dst.x = (src0.x != src1.x) ? \sim 0 : 0
1520 dst.y = (src0.y != src1.y) ? \sim 0 : 0
1522 dst.z = (src0.z != src1.z) ? \sim 0 : 0
1524 dst.w = (src0.w != src1.w) ? \sim 0 : 0
1527 .. opcode:: INEG - Integer Negate
1542 .. opcode:: IABS - Integer Absolute Value
1556 These opcodes are used for bit-level manipulation of integers.
1558 .. opcode:: IBFE - Signed Bitfield Extract
1560 Like GLSL bitfieldExtract. Extracts a set of bits from the input, and
1561 sign-extends them if the high bit of the extracted window is set.
1565 def ibfe(value, offset, bits):
1566 if offset < 0 or bits < 0 or offset + bits > 32:
1568 if bits == 0: return 0
1569 # Note: >> sign-extends
1570 return (value << (32 - offset - bits)) >> (32 - bits)
1572 .. opcode:: UBFE - Unsigned Bitfield Extract
1574 Like GLSL bitfieldExtract. Extracts a set of bits from the input, without
1579 def ubfe(value, offset, bits):
1580 if offset < 0 or bits < 0 or offset + bits > 32:
1582 if bits == 0: return 0
1583 # Note: >> does not sign-extend
1584 return (value << (32 - offset - bits)) >> (32 - bits)
1586 .. opcode:: BFI - Bitfield Insert
1588 Like GLSL bitfieldInsert. Replaces a bit region of 'base' with the low bits
1593 def bfi(base, insert, offset, bits):
1594 if offset < 0 or bits < 0 or offset + bits > 32:
1596 # << defined such that mask == ~0 when bits == 32, offset == 0
1597 mask = ((1 << bits) - 1) << offset
1598 return ((insert << offset) & mask) | (base & ~mask)
1600 .. opcode:: BREV - Bitfield Reverse
1602 See SM5 instruction BFREV. Reverses the bits of the argument.
1604 .. opcode:: POPC - Population Count
1606 See SM5 instruction COUNTBITS. Counts the number of set bits in the argument.
1608 .. opcode:: LSB - Index of lowest set bit
1610 See SM5 instruction FIRSTBIT_LO. Computes the 0-based index of the first set
1611 bit of the argument. Returns -1 if none are set.
1613 .. opcode:: IMSB - Index of highest non-sign bit
1615 See SM5 instruction FIRSTBIT_SHI. Computes the 0-based index of the highest
1616 non-sign bit of the argument (i.e. highest 0 bit for negative numbers,
1617 highest 1 bit for positive numbers). Returns -1 if all bits are the same
1618 (i.e. for inputs 0 and -1).
1620 .. opcode:: UMSB - Index of highest set bit
1622 See SM5 instruction FIRSTBIT_HI. Computes the 0-based index of the highest
1623 set bit of the argument. Returns -1 if none are set.
1626 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1628 These opcodes are only supported in geometry shaders; they have no meaning
1629 in any other type of shader.
1631 .. opcode:: EMIT - Emit
1633 Generate a new vertex for the current primitive into the specified vertex
1634 stream using the values in the output registers.
1637 .. opcode:: ENDPRIM - End Primitive
1639 Complete the current primitive in the specified vertex stream (consisting of
1640 the emitted vertices), and start a new one.
1646 These opcodes are part of :term:`GLSL`'s opcode set. Support for these
1647 opcodes is determined by a special capability bit, ``GLSL``.
1648 Some require glsl version 1.30 (UIF/BREAKC/SWITCH/CASE/DEFAULT/ENDSWITCH).
1650 .. opcode:: CAL - Subroutine Call
1656 .. opcode:: RET - Subroutine Call Return
1661 .. opcode:: CONT - Continue
1663 Unconditionally moves the point of execution to the instruction after the
1664 last bgnloop. The instruction must appear within a bgnloop/endloop.
1668 Support for CONT is determined by a special capability bit,
1669 ``TGSI_CONT_SUPPORTED``. See :ref:`Screen` for more information.
1672 .. opcode:: BGNLOOP - Begin a Loop
1674 Start a loop. Must have a matching endloop.
1677 .. opcode:: BGNSUB - Begin Subroutine
1679 Starts definition of a subroutine. Must have a matching endsub.
1682 .. opcode:: ENDLOOP - End a Loop
1684 End a loop started with bgnloop.
1687 .. opcode:: ENDSUB - End Subroutine
1689 Ends definition of a subroutine.
1692 .. opcode:: NOP - No Operation
1697 .. opcode:: BRK - Break
1699 Unconditionally moves the point of execution to the instruction after the
1700 next endloop or endswitch. The instruction must appear within a loop/endloop
1701 or switch/endswitch.
1704 .. opcode:: BREAKC - Break Conditional
1706 Conditionally moves the point of execution to the instruction after the
1707 next endloop or endswitch. The instruction must appear within a loop/endloop
1708 or switch/endswitch.
1709 Condition evaluates to true if src0.x != 0 where src0.x is interpreted
1710 as an integer register.
1714 Considered for removal as it's quite inconsistent wrt other opcodes
1715 (could emulate with UIF/BRK/ENDIF).
1718 .. opcode:: IF - Float If
1720 Start an IF ... ELSE .. ENDIF block. Condition evaluates to true if
1724 where src0.x is interpreted as a floating point register.
1727 .. opcode:: UIF - Bitwise If
1729 Start an UIF ... ELSE .. ENDIF block. Condition evaluates to true if
1733 where src0.x is interpreted as an integer register.
1736 .. opcode:: ELSE - Else
1738 Starts an else block, after an IF or UIF statement.
1741 .. opcode:: ENDIF - End If
1743 Ends an IF or UIF block.
1746 .. opcode:: SWITCH - Switch
1748 Starts a C-style switch expression. The switch consists of one or multiple
1749 CASE statements, and at most one DEFAULT statement. Execution of a statement
1750 ends when a BRK is hit, but just like in C falling through to other cases
1751 without a break is allowed. Similarly, DEFAULT label is allowed anywhere not
1752 just as last statement, and fallthrough is allowed into/from it.
1753 CASE src arguments are evaluated at bit level against the SWITCH src argument.
1759 (some instructions here)
1762 (some instructions here)
1765 (some instructions here)
1770 .. opcode:: CASE - Switch case
1772 This represents a switch case label. The src arg must be an integer immediate.
1775 .. opcode:: DEFAULT - Switch default
1777 This represents the default case in the switch, which is taken if no other
1781 .. opcode:: ENDSWITCH - End of switch
1783 Ends a switch expression.
1789 The interpolation instructions allow an input to be interpolated in a
1790 different way than its declaration. This corresponds to the GLSL 4.00
1791 interpolateAt* functions. The first argument of each of these must come from
1792 ``TGSI_FILE_INPUT``.
1794 .. opcode:: INTERP_CENTROID - Interpolate at the centroid
1796 Interpolates the varying specified by src0 at the centroid
1798 .. opcode:: INTERP_SAMPLE - Interpolate at the specified sample
1800 Interpolates the varying specified by src0 at the sample id specified by
1801 src1.x (interpreted as an integer)
1803 .. opcode:: INTERP_OFFSET - Interpolate at the specified offset
1805 Interpolates the varying specified by src0 at the offset src1.xy from the
1806 pixel center (interpreted as floats)
1814 The double-precision opcodes reinterpret four-component vectors into
1815 two-component vectors with doubled precision in each component.
1817 .. opcode:: DABS - Absolute
1822 .. opcode:: DADD - Add
1826 dst.xy = src0.xy + src1.xy
1828 dst.zw = src0.zw + src1.zw
1830 .. opcode:: DSEQ - Set on Equal
1834 dst.x = src0.xy == src1.xy ? \sim 0 : 0
1836 dst.z = src0.zw == src1.zw ? \sim 0 : 0
1838 .. opcode:: DSNE - Set on Equal
1842 dst.x = src0.xy != src1.xy ? \sim 0 : 0
1844 dst.z = src0.zw != src1.zw ? \sim 0 : 0
1846 .. opcode:: DSLT - Set on Less than
1850 dst.x = src0.xy < src1.xy ? \sim 0 : 0
1852 dst.z = src0.zw < src1.zw ? \sim 0 : 0
1854 .. opcode:: DSGE - Set on Greater equal
1858 dst.x = src0.xy >= src1.xy ? \sim 0 : 0
1860 dst.z = src0.zw >= src1.zw ? \sim 0 : 0
1862 .. opcode:: DFRAC - Fraction
1866 dst.xy = src.xy - \lfloor src.xy\rfloor
1868 dst.zw = src.zw - \lfloor src.zw\rfloor
1870 .. opcode:: DTRUNC - Truncate
1874 dst.xy = trunc(src.xy)
1876 dst.zw = trunc(src.zw)
1878 .. opcode:: DCEIL - Ceiling
1882 dst.xy = \lceil src.xy\rceil
1884 dst.zw = \lceil src.zw\rceil
1886 .. opcode:: DFLR - Floor
1890 dst.xy = \lfloor src.xy\rfloor
1892 dst.zw = \lfloor src.zw\rfloor
1894 .. opcode:: DROUND - Fraction
1898 dst.xy = round(src.xy)
1900 dst.zw = round(src.zw)
1902 .. opcode:: DSSG - Set Sign
1906 dst.xy = (src.xy > 0) ? 1.0 : (src.xy < 0) ? -1.0 : 0.0
1908 dst.zw = (src.zw > 0) ? 1.0 : (src.zw < 0) ? -1.0 : 0.0
1910 .. opcode:: DFRACEXP - Convert Number to Fractional and Integral Components
1912 Like the ``frexp()`` routine in many math libraries, this opcode stores the
1913 exponent of its source to ``dst0``, and the significand to ``dst1``, such that
1914 :math:`dst1 \times 2^{dst0} = src` .
1918 dst0.xy = exp(src.xy)
1920 dst1.xy = frac(src.xy)
1922 dst0.zw = exp(src.zw)
1924 dst1.zw = frac(src.zw)
1926 .. opcode:: DLDEXP - Multiply Number by Integral Power of 2
1928 This opcode is the inverse of :opcode:`DFRACEXP`. The second
1929 source is an integer.
1933 dst.xy = src0.xy \times 2^{src1.x}
1935 dst.zw = src0.zw \times 2^{src1.y}
1937 .. opcode:: DMIN - Minimum
1941 dst.xy = min(src0.xy, src1.xy)
1943 dst.zw = min(src0.zw, src1.zw)
1945 .. opcode:: DMAX - Maximum
1949 dst.xy = max(src0.xy, src1.xy)
1951 dst.zw = max(src0.zw, src1.zw)
1953 .. opcode:: DMUL - Multiply
1957 dst.xy = src0.xy \times src1.xy
1959 dst.zw = src0.zw \times src1.zw
1962 .. opcode:: DMAD - Multiply And Add
1966 dst.xy = src0.xy \times src1.xy + src2.xy
1968 dst.zw = src0.zw \times src1.zw + src2.zw
1971 .. opcode:: DFMA - Fused Multiply-Add
1973 Perform a * b + c with no intermediate rounding step.
1977 dst.xy = src0.xy \times src1.xy + src2.xy
1979 dst.zw = src0.zw \times src1.zw + src2.zw
1982 .. opcode:: DDIV - Divide
1986 dst.xy = \frac{src0.xy}{src1.xy}
1988 dst.zw = \frac{src0.zw}{src1.zw}
1991 .. opcode:: DRCP - Reciprocal
1995 dst.xy = \frac{1}{src.xy}
1997 dst.zw = \frac{1}{src.zw}
1999 .. opcode:: DSQRT - Square Root
2003 dst.xy = \sqrt{src.xy}
2005 dst.zw = \sqrt{src.zw}
2007 .. opcode:: DRSQ - Reciprocal Square Root
2011 dst.xy = \frac{1}{\sqrt{src.xy}}
2013 dst.zw = \frac{1}{\sqrt{src.zw}}
2015 .. opcode:: F2D - Float to Double
2019 dst.xy = double(src0.x)
2021 dst.zw = double(src0.y)
2023 .. opcode:: D2F - Double to Float
2027 dst.x = float(src0.xy)
2029 dst.y = float(src0.zw)
2031 .. opcode:: I2D - Int to Double
2035 dst.xy = double(src0.x)
2037 dst.zw = double(src0.y)
2039 .. opcode:: D2I - Double to Int
2043 dst.x = int(src0.xy)
2045 dst.y = int(src0.zw)
2047 .. opcode:: U2D - Unsigned Int to Double
2051 dst.xy = double(src0.x)
2053 dst.zw = double(src0.y)
2055 .. opcode:: D2U - Double to Unsigned Int
2059 dst.x = unsigned(src0.xy)
2061 dst.y = unsigned(src0.zw)
2066 The 64-bit integer opcodes reinterpret four-component vectors into
2067 two-component vectors with 64-bits in each component.
2069 .. opcode:: I64ABS - 64-bit Integer Absolute Value
2074 .. opcode:: I64NEG - 64-bit Integer Negate
2083 .. opcode:: I64SSG - 64-bit Integer Set Sign
2087 dst.xy = (src0.xy < 0) ? -1 : (src0.xy > 0) ? 1 : 0
2088 dst.zw = (src0.zw < 0) ? -1 : (src0.zw > 0) ? 1 : 0
2090 .. opcode:: U64ADD - 64-bit Integer Add
2094 dst.xy = src0.xy + src1.xy
2095 dst.zw = src0.zw + src1.zw
2097 .. opcode:: U64MUL - 64-bit Integer Multiply
2101 dst.xy = src0.xy * src1.xy
2102 dst.zw = src0.zw * src1.zw
2104 .. opcode:: U64SEQ - 64-bit Integer Set on Equal
2108 dst.x = src0.xy == src1.xy ? \sim 0 : 0
2109 dst.z = src0.zw == src1.zw ? \sim 0 : 0
2111 .. opcode:: U64SNE - 64-bit Integer Set on Not Equal
2115 dst.x = src0.xy != src1.xy ? \sim 0 : 0
2116 dst.z = src0.zw != src1.zw ? \sim 0 : 0
2118 .. opcode:: U64SLT - 64-bit Unsigned Integer Set on Less Than
2122 dst.x = src0.xy < src1.xy ? \sim 0 : 0
2123 dst.z = src0.zw < src1.zw ? \sim 0 : 0
2125 .. opcode:: U64SGE - 64-bit Unsigned Integer Set on Greater Equal
2129 dst.x = src0.xy >= src1.xy ? \sim 0 : 0
2130 dst.z = src0.zw >= src1.zw ? \sim 0 : 0
2132 .. opcode:: I64SLT - 64-bit Signed Integer Set on Less Than
2136 dst.x = src0.xy < src1.xy ? \sim 0 : 0
2137 dst.z = src0.zw < src1.zw ? \sim 0 : 0
2139 .. opcode:: I64SGE - 64-bit Signed Integer Set on Greater Equal
2143 dst.x = src0.xy >= src1.xy ? \sim 0 : 0
2144 dst.z = src0.zw >= src1.zw ? \sim 0 : 0
2146 .. opcode:: I64MIN - Minimum of 64-bit Signed Integers
2150 dst.xy = min(src0.xy, src1.xy)
2151 dst.zw = min(src0.zw, src1.zw)
2153 .. opcode:: U64MIN - Minimum of 64-bit Unsigned Integers
2157 dst.xy = min(src0.xy, src1.xy)
2158 dst.zw = min(src0.zw, src1.zw)
2160 .. opcode:: I64MAX - Maximum of 64-bit Signed Integers
2164 dst.xy = max(src0.xy, src1.xy)
2165 dst.zw = max(src0.zw, src1.zw)
2167 .. opcode:: U64MAX - Maximum of 64-bit Unsigned Integers
2171 dst.xy = max(src0.xy, src1.xy)
2172 dst.zw = max(src0.zw, src1.zw)
2174 .. opcode:: U64SHL - Shift Left 64-bit Unsigned Integer
2176 The shift count is masked with 0x3f before the shift is applied.
2180 dst.xy = src0.xy << (0x3f \& src1.x)
2181 dst.zw = src0.zw << (0x3f \& src1.y)
2183 .. opcode:: I64SHR - Arithmetic Shift Right (of 64-bit Signed Integer)
2185 The shift count is masked with 0x3f before the shift is applied.
2189 dst.xy = src0.xy >> (0x3f \& src1.x)
2190 dst.zw = src0.zw >> (0x3f \& src1.y)
2192 .. opcode:: U64SHR - Logical Shift Right (of 64-bit Unsigned Integer)
2194 The shift count is masked with 0x3f before the shift is applied.
2198 dst.xy = src0.xy >> (unsigned) (0x3f \& src1.x)
2199 dst.zw = src0.zw >> (unsigned) (0x3f \& src1.y)
2201 .. opcode:: I64DIV - 64-bit Signed Integer Division
2205 dst.xy = src0.xy \ src1.xy
2206 dst.zw = src0.zw \ src1.zw
2208 .. opcode:: U64DIV - 64-bit Unsigned Integer Division
2212 dst.xy = src0.xy \ src1.xy
2213 dst.zw = src0.zw \ src1.zw
2215 .. opcode:: U64MOD - 64-bit Unsigned Integer Remainder
2219 dst.xy = src0.xy \bmod src1.xy
2220 dst.zw = src0.zw \bmod src1.zw
2222 .. opcode:: I64MOD - 64-bit Signed Integer Remainder
2226 dst.xy = src0.xy \bmod src1.xy
2227 dst.zw = src0.zw \bmod src1.zw
2229 .. opcode:: F2U64 - Float to 64-bit Unsigned Int
2233 dst.xy = (uint64_t) src0.x
2234 dst.zw = (uint64_t) src0.y
2236 .. opcode:: F2I64 - Float to 64-bit Int
2240 dst.xy = (int64_t) src0.x
2241 dst.zw = (int64_t) src0.y
2243 .. opcode:: U2I64 - Unsigned Integer to 64-bit Integer
2245 This is a zero extension.
2249 dst.xy = (uint64_t) src0.x
2250 dst.zw = (uint64_t) src0.y
2252 .. opcode:: I2I64 - Signed Integer to 64-bit Integer
2254 This is a sign extension.
2258 dst.xy = (int64_t) src0.x
2259 dst.zw = (int64_t) src0.y
2261 .. opcode:: D2U64 - Double to 64-bit Unsigned Int
2265 dst.xy = (uint64_t) src0.xy
2266 dst.zw = (uint64_t) src0.zw
2268 .. opcode:: D2I64 - Double to 64-bit Int
2272 dst.xy = (int64_t) src0.xy
2273 dst.zw = (int64_t) src0.zw
2275 .. opcode:: U642F - 64-bit unsigned integer to float
2279 dst.x = (float) src0.xy
2280 dst.y = (float) src0.zw
2282 .. opcode:: I642F - 64-bit Int to Float
2286 dst.x = (float) src0.xy
2287 dst.y = (float) src0.zw
2289 .. opcode:: U642D - 64-bit unsigned integer to double
2293 dst.xy = (double) src0.xy
2294 dst.zw = (double) src0.zw
2296 .. opcode:: I642D - 64-bit Int to double
2300 dst.xy = (double) src0.xy
2301 dst.zw = (double) src0.zw
2303 .. _samplingopcodes:
2305 Resource Sampling Opcodes
2306 ^^^^^^^^^^^^^^^^^^^^^^^^^
2308 Those opcodes follow very closely semantics of the respective Direct3D
2309 instructions. If in doubt double check Direct3D documentation.
2310 Note that the swizzle on SVIEW (src1) determines texel swizzling
2315 Using provided address, sample data from the specified texture using the
2316 filtering mode identified by the given sampler. The source data may come from
2317 any resource type other than buffers.
2319 Syntax: ``SAMPLE dst, address, sampler_view, sampler``
2321 Example: ``SAMPLE TEMP[0], TEMP[1], SVIEW[0], SAMP[0]``
2323 .. opcode:: SAMPLE_I
2325 Simplified alternative to the SAMPLE instruction. Using the provided
2326 integer address, SAMPLE_I fetches data from the specified sampler view
2327 without any filtering. The source data may come from any resource type
2330 Syntax: ``SAMPLE_I dst, address, sampler_view``
2332 Example: ``SAMPLE_I TEMP[0], TEMP[1], SVIEW[0]``
2334 The 'address' is specified as unsigned integers. If the 'address' is out of
2335 range [0...(# texels - 1)] the result of the fetch is always 0 in all
2336 components. As such the instruction doesn't honor address wrap modes, in
2337 cases where that behavior is desirable 'SAMPLE' instruction should be used.
2338 address.w always provides an unsigned integer mipmap level. If the value is
2339 out of the range then the instruction always returns 0 in all components.
2340 address.yz are ignored for buffers and 1d textures. address.z is ignored
2341 for 1d texture arrays and 2d textures.
2343 For 1D texture arrays address.y provides the array index (also as unsigned
2344 integer). If the value is out of the range of available array indices
2345 [0... (array size - 1)] then the opcode always returns 0 in all components.
2346 For 2D texture arrays address.z provides the array index, otherwise it
2347 exhibits the same behavior as in the case for 1D texture arrays. The exact
2348 semantics of the source address are presented in the table below:
2350 +---------------------------+----+-----+-----+---------+
2351 | resource type | X | Y | Z | W |
2352 +===========================+====+=====+=====+=========+
2353 | ``PIPE_BUFFER`` | x | | | ignored |
2354 +---------------------------+----+-----+-----+---------+
2355 | ``PIPE_TEXTURE_1D`` | x | | | mpl |
2356 +---------------------------+----+-----+-----+---------+
2357 | ``PIPE_TEXTURE_2D`` | x | y | | mpl |
2358 +---------------------------+----+-----+-----+---------+
2359 | ``PIPE_TEXTURE_3D`` | x | y | z | mpl |
2360 +---------------------------+----+-----+-----+---------+
2361 | ``PIPE_TEXTURE_RECT`` | x | y | | mpl |
2362 +---------------------------+----+-----+-----+---------+
2363 | ``PIPE_TEXTURE_CUBE`` | not allowed as source |
2364 +---------------------------+----+-----+-----+---------+
2365 | ``PIPE_TEXTURE_1D_ARRAY`` | x | idx | | mpl |
2366 +---------------------------+----+-----+-----+---------+
2367 | ``PIPE_TEXTURE_2D_ARRAY`` | x | y | idx | mpl |
2368 +---------------------------+----+-----+-----+---------+
2370 Where 'mpl' is a mipmap level and 'idx' is the array index.
2372 .. opcode:: SAMPLE_I_MS
2374 Just like SAMPLE_I but allows fetch data from multi-sampled surfaces.
2376 Syntax: ``SAMPLE_I_MS dst, address, sampler_view, sample``
2378 .. opcode:: SAMPLE_B
2380 Just like the SAMPLE instruction with the exception that an additional bias
2381 is applied to the level of detail computed as part of the instruction
2384 Syntax: ``SAMPLE_B dst, address, sampler_view, sampler, lod_bias``
2386 Example: ``SAMPLE_B TEMP[0], TEMP[1], SVIEW[0], SAMP[0], TEMP[2].x``
2388 .. opcode:: SAMPLE_C
2390 Similar to the SAMPLE instruction but it performs a comparison filter. The
2391 operands to SAMPLE_C are identical to SAMPLE, except that there is an
2392 additional float32 operand, reference value, which must be a register with
2393 single-component, or a scalar literal. SAMPLE_C makes the hardware use the
2394 current samplers compare_func (in pipe_sampler_state) to compare reference
2395 value against the red component value for the surce resource at each texel
2396 that the currently configured texture filter covers based on the provided
2399 Syntax: ``SAMPLE_C dst, address, sampler_view.r, sampler, ref_value``
2401 Example: ``SAMPLE_C TEMP[0], TEMP[1], SVIEW[0].r, SAMP[0], TEMP[2].x``
2403 .. opcode:: SAMPLE_C_LZ
2405 Same as SAMPLE_C, but LOD is 0 and derivatives are ignored. The LZ stands
2408 Syntax: ``SAMPLE_C_LZ dst, address, sampler_view.r, sampler, ref_value``
2410 Example: ``SAMPLE_C_LZ TEMP[0], TEMP[1], SVIEW[0].r, SAMP[0], TEMP[2].x``
2413 .. opcode:: SAMPLE_D
2415 SAMPLE_D is identical to the SAMPLE opcode except that the derivatives for
2416 the source address in the x direction and the y direction are provided by
2419 Syntax: ``SAMPLE_D dst, address, sampler_view, sampler, der_x, der_y``
2421 Example: ``SAMPLE_D TEMP[0], TEMP[1], SVIEW[0], SAMP[0], TEMP[2], TEMP[3]``
2423 .. opcode:: SAMPLE_L
2425 SAMPLE_L is identical to the SAMPLE opcode except that the LOD is provided
2426 directly as a scalar value, representing no anisotropy.
2428 Syntax: ``SAMPLE_L dst, address, sampler_view, sampler, explicit_lod``
2430 Example: ``SAMPLE_L TEMP[0], TEMP[1], SVIEW[0], SAMP[0], TEMP[2].x``
2434 Gathers the four texels to be used in a bi-linear filtering operation and
2435 packs them into a single register. Only works with 2D, 2D array, cubemaps,
2436 and cubemaps arrays. For 2D textures, only the addressing modes of the
2437 sampler and the top level of any mip pyramid are used. Set W to zero. It
2438 behaves like the SAMPLE instruction, but a filtered sample is not
2439 generated. The four samples that contribute to filtering are placed into
2440 xyzw in counter-clockwise order, starting with the (u,v) texture coordinate
2441 delta at the following locations (-, +), (+, +), (+, -), (-, -), where the
2442 magnitude of the deltas are half a texel.
2445 .. opcode:: SVIEWINFO
2447 Query the dimensions of a given sampler view. dst receives width, height,
2448 depth or array size and number of mipmap levels as int4. The dst can have a
2449 writemask which will specify what info is the caller interested in.
2451 Syntax: ``SVIEWINFO dst, src_mip_level, sampler_view``
2453 Example: ``SVIEWINFO TEMP[0], TEMP[1].x, SVIEW[0]``
2455 src_mip_level is an unsigned integer scalar. If it's out of range then
2456 returns 0 for width, height and depth/array size but the total number of
2457 mipmap is still returned correctly for the given sampler view. The returned
2458 width, height and depth values are for the mipmap level selected by the
2459 src_mip_level and are in the number of texels. For 1d texture array width
2460 is in dst.x, array size is in dst.y and dst.z is 0. The number of mipmaps is
2461 still in dst.w. In contrast to d3d10 resinfo, there's no way in the tgsi
2462 instruction encoding to specify the return type (float/rcpfloat/uint), hence
2463 always using uint. Also, unlike the SAMPLE instructions, the swizzle on src1
2464 resinfo allowing swizzling dst values is ignored (due to the interaction
2465 with rcpfloat modifier which requires some swizzle handling in the state
2468 .. opcode:: SAMPLE_POS
2470 Query the position of a given sample. dst receives float4 (x, y, 0, 0)
2471 indicated where the sample is located. If the resource is not a multi-sample
2472 resource and not a render target, the result is 0.
2474 .. opcode:: SAMPLE_INFO
2476 dst receives number of samples in x. If the resource is not a multi-sample
2477 resource and not a render target, the result is 0.
2480 .. _resourceopcodes:
2482 Resource Access Opcodes
2483 ^^^^^^^^^^^^^^^^^^^^^^^
2485 .. opcode:: LOAD - Fetch data from a shader buffer or image
2487 Syntax: ``LOAD dst, resource, address``
2489 Example: ``LOAD TEMP[0], BUFFER[0], TEMP[1]``
2491 Using the provided integer address, LOAD fetches data
2492 from the specified buffer or texture without any
2495 The 'address' is specified as a vector of unsigned
2496 integers. If the 'address' is out of range the result
2499 Only the first mipmap level of a resource can be read
2500 from using this instruction.
2502 For 1D or 2D texture arrays, the array index is
2503 provided as an unsigned integer in address.y or
2504 address.z, respectively. address.yz are ignored for
2505 buffers and 1D textures. address.z is ignored for 1D
2506 texture arrays and 2D textures. address.w is always
2509 A swizzle suffix may be added to the resource argument
2510 this will cause the resource data to be swizzled accordingly.
2512 .. opcode:: STORE - Write data to a shader resource
2514 Syntax: ``STORE resource, address, src``
2516 Example: ``STORE BUFFER[0], TEMP[0], TEMP[1]``
2518 Using the provided integer address, STORE writes data
2519 to the specified buffer or texture.
2521 The 'address' is specified as a vector of unsigned
2522 integers. If the 'address' is out of range the result
2525 Only the first mipmap level of a resource can be
2526 written to using this instruction.
2528 For 1D or 2D texture arrays, the array index is
2529 provided as an unsigned integer in address.y or
2530 address.z, respectively. address.yz are ignored for
2531 buffers and 1D textures. address.z is ignored for 1D
2532 texture arrays and 2D textures. address.w is always
2535 .. opcode:: RESQ - Query information about a resource
2537 Syntax: ``RESQ dst, resource``
2539 Example: ``RESQ TEMP[0], BUFFER[0]``
2541 Returns information about the buffer or image resource. For buffer
2542 resources, the size (in bytes) is returned in the x component. For
2543 image resources, .xyz will contain the width/height/layers of the
2544 image, while .w will contain the number of samples for multi-sampled
2547 .. opcode:: FBFETCH - Load data from framebuffer
2549 Syntax: ``FBFETCH dst, output``
2551 Example: ``FBFETCH TEMP[0], OUT[0]``
2553 This is only valid on ``COLOR`` semantic outputs. Returns the color
2554 of the current position in the framebuffer from before this fragment
2555 shader invocation. May return the same value from multiple calls for
2556 a particular output within a single invocation. Note that result may
2557 be undefined if a fragment is drawn multiple times without a blend
2561 .. _threadsyncopcodes:
2563 Inter-thread synchronization opcodes
2564 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2566 These opcodes are intended for communication between threads running
2567 within the same compute grid. For now they're only valid in compute
2570 .. opcode:: MFENCE - Memory fence
2572 Syntax: ``MFENCE resource``
2574 Example: ``MFENCE RES[0]``
2576 This opcode forces strong ordering between any memory access
2577 operations that affect the specified resource. This means that
2578 previous loads and stores (and only those) will be performed and
2579 visible to other threads before the program execution continues.
2582 .. opcode:: LFENCE - Load memory fence
2584 Syntax: ``LFENCE resource``
2586 Example: ``LFENCE RES[0]``
2588 Similar to MFENCE, but it only affects the ordering of memory loads.
2591 .. opcode:: SFENCE - Store memory fence
2593 Syntax: ``SFENCE resource``
2595 Example: ``SFENCE RES[0]``
2597 Similar to MFENCE, but it only affects the ordering of memory stores.
2600 .. opcode:: BARRIER - Thread group barrier
2604 This opcode suspends the execution of the current thread until all
2605 the remaining threads in the working group reach the same point of
2606 the program. Results are unspecified if any of the remaining
2607 threads terminates or never reaches an executed BARRIER instruction.
2609 .. opcode:: MEMBAR - Memory barrier
2613 This opcode waits for the completion of all memory accesses based on
2614 the type passed in. The type is an immediate bitfield with the following
2617 Bit 0: Shader storage buffers
2618 Bit 1: Atomic buffers
2620 Bit 3: Shared memory
2623 These may be passed in in any combination. An implementation is free to not
2624 distinguish between these as it sees fit. However these map to all the
2625 possibilities made available by GLSL.
2632 These opcodes provide atomic variants of some common arithmetic and
2633 logical operations. In this context atomicity means that another
2634 concurrent memory access operation that affects the same memory
2635 location is guaranteed to be performed strictly before or after the
2636 entire execution of the atomic operation. The resource may be a buffer
2637 or an image. In the case of an image, the offset works the same as for
2638 ``LOAD`` and ``STORE``, specified above. These atomic operations may
2639 only be used with 32-bit integer image formats.
2641 .. opcode:: ATOMUADD - Atomic integer addition
2643 Syntax: ``ATOMUADD dst, resource, offset, src``
2645 Example: ``ATOMUADD TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2647 The following operation is performed atomically:
2651 dst_x = resource[offset]
2653 resource[offset] = dst_x + src_x
2656 .. opcode:: ATOMXCHG - Atomic exchange
2658 Syntax: ``ATOMXCHG dst, resource, offset, src``
2660 Example: ``ATOMXCHG TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2662 The following operation is performed atomically:
2666 dst_x = resource[offset]
2668 resource[offset] = src_x
2671 .. opcode:: ATOMCAS - Atomic compare-and-exchange
2673 Syntax: ``ATOMCAS dst, resource, offset, cmp, src``
2675 Example: ``ATOMCAS TEMP[0], BUFFER[0], TEMP[1], TEMP[2], TEMP[3]``
2677 The following operation is performed atomically:
2681 dst_x = resource[offset]
2683 resource[offset] = (dst_x == cmp_x ? src_x : dst_x)
2686 .. opcode:: ATOMAND - Atomic bitwise And
2688 Syntax: ``ATOMAND dst, resource, offset, src``
2690 Example: ``ATOMAND TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2692 The following operation is performed atomically:
2696 dst_x = resource[offset]
2698 resource[offset] = dst_x \& src_x
2701 .. opcode:: ATOMOR - Atomic bitwise Or
2703 Syntax: ``ATOMOR dst, resource, offset, src``
2705 Example: ``ATOMOR TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2707 The following operation is performed atomically:
2711 dst_x = resource[offset]
2713 resource[offset] = dst_x | src_x
2716 .. opcode:: ATOMXOR - Atomic bitwise Xor
2718 Syntax: ``ATOMXOR dst, resource, offset, src``
2720 Example: ``ATOMXOR TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2722 The following operation is performed atomically:
2726 dst_x = resource[offset]
2728 resource[offset] = dst_x \oplus src_x
2731 .. opcode:: ATOMUMIN - Atomic unsigned minimum
2733 Syntax: ``ATOMUMIN dst, resource, offset, src``
2735 Example: ``ATOMUMIN TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2737 The following operation is performed atomically:
2741 dst_x = resource[offset]
2743 resource[offset] = (dst_x < src_x ? dst_x : src_x)
2746 .. opcode:: ATOMUMAX - Atomic unsigned maximum
2748 Syntax: ``ATOMUMAX dst, resource, offset, src``
2750 Example: ``ATOMUMAX TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2752 The following operation is performed atomically:
2756 dst_x = resource[offset]
2758 resource[offset] = (dst_x > src_x ? dst_x : src_x)
2761 .. opcode:: ATOMIMIN - Atomic signed minimum
2763 Syntax: ``ATOMIMIN dst, resource, offset, src``
2765 Example: ``ATOMIMIN TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2767 The following operation is performed atomically:
2771 dst_x = resource[offset]
2773 resource[offset] = (dst_x < src_x ? dst_x : src_x)
2776 .. opcode:: ATOMIMAX - Atomic signed maximum
2778 Syntax: ``ATOMIMAX dst, resource, offset, src``
2780 Example: ``ATOMIMAX TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2782 The following operation is performed atomically:
2786 dst_x = resource[offset]
2788 resource[offset] = (dst_x > src_x ? dst_x : src_x)
2796 These opcodes compare the given value across the shader invocations
2797 running in the current SIMD group. The details of exactly which
2798 invocations get compared are implementation-defined, and it would be a
2799 correct implementation to only ever consider the current thread's
2800 value. (i.e. SIMD group of 1). The argument is treated as a boolean.
2802 .. opcode:: VOTE_ANY - Value is set in any of the current invocations
2804 .. opcode:: VOTE_ALL - Value is set in all of the current invocations
2806 .. opcode:: VOTE_EQ - Value is the same in all of the current invocations
2809 Explanation of symbols used
2810 ------------------------------
2817 :math:`|x|` Absolute value of `x`.
2819 :math:`\lceil x \rceil` Ceiling of `x`.
2821 clamp(x,y,z) Clamp x between y and z.
2822 (x < y) ? y : (x > z) ? z : x
2824 :math:`\lfloor x\rfloor` Floor of `x`.
2826 :math:`\log_2{x}` Logarithm of `x`, base 2.
2828 max(x,y) Maximum of x and y.
2831 min(x,y) Minimum of x and y.
2834 partialx(x) Derivative of x relative to fragment's X.
2836 partialy(x) Derivative of x relative to fragment's Y.
2838 pop() Pop from stack.
2840 :math:`x^y` `x` to the power `y`.
2842 push(x) Push x on stack.
2846 trunc(x) Truncate x, i.e. drop the fraction bits.
2853 discard Discard fragment.
2857 target Label of target instruction.
2868 Declares a register that is will be referenced as an operand in Instruction
2871 File field contains register file that is being declared and is one
2874 UsageMask field specifies which of the register components can be accessed
2875 and is one of TGSI_WRITEMASK.
2877 The Local flag specifies that a given value isn't intended for
2878 subroutine parameter passing and, as a result, the implementation
2879 isn't required to give any guarantees of it being preserved across
2880 subroutine boundaries. As it's merely a compiler hint, the
2881 implementation is free to ignore it.
2883 If Dimension flag is set to 1, a Declaration Dimension token follows.
2885 If Semantic flag is set to 1, a Declaration Semantic token follows.
2887 If Interpolate flag is set to 1, a Declaration Interpolate token follows.
2889 If file is TGSI_FILE_RESOURCE, a Declaration Resource token follows.
2891 If Array flag is set to 1, a Declaration Array token follows.
2894 ^^^^^^^^^^^^^^^^^^^^^^^^
2896 Declarations can optional have an ArrayID attribute which can be referred by
2897 indirect addressing operands. An ArrayID of zero is reserved and treated as
2898 if no ArrayID is specified.
2900 If an indirect addressing operand refers to a specific declaration by using
2901 an ArrayID only the registers in this declaration are guaranteed to be
2902 accessed, accessing any register outside this declaration results in undefined
2903 behavior. Note that for compatibility the effective index is zero-based and
2904 not relative to the specified declaration
2906 If no ArrayID is specified with an indirect addressing operand the whole
2907 register file might be accessed by this operand. This is strongly discouraged
2908 and will prevent packing of scalar/vec2 arrays and effective alias analysis.
2909 This is only legal for TEMP and CONST register files.
2911 Declaration Semantic
2912 ^^^^^^^^^^^^^^^^^^^^^^^^
2914 Vertex and fragment shader input and output registers may be labeled
2915 with semantic information consisting of a name and index.
2917 Follows Declaration token if Semantic bit is set.
2919 Since its purpose is to link a shader with other stages of the pipeline,
2920 it is valid to follow only those Declaration tokens that declare a register
2921 either in INPUT or OUTPUT file.
2923 SemanticName field contains the semantic name of the register being declared.
2924 There is no default value.
2926 SemanticIndex is an optional subscript that can be used to distinguish
2927 different register declarations with the same semantic name. The default value
2930 The meanings of the individual semantic names are explained in the following
2933 TGSI_SEMANTIC_POSITION
2934 """"""""""""""""""""""
2936 For vertex shaders, TGSI_SEMANTIC_POSITION indicates the vertex shader
2937 output register which contains the homogeneous vertex position in the clip
2938 space coordinate system. After clipping, the X, Y and Z components of the
2939 vertex will be divided by the W value to get normalized device coordinates.
2941 For fragment shaders, TGSI_SEMANTIC_POSITION is used to indicate that
2942 fragment shader input (or system value, depending on which one is
2943 supported by the driver) contains the fragment's window position. The X
2944 component starts at zero and always increases from left to right.
2945 The Y component starts at zero and always increases but Y=0 may either
2946 indicate the top of the window or the bottom depending on the fragment
2947 coordinate origin convention (see TGSI_PROPERTY_FS_COORD_ORIGIN).
2948 The Z coordinate ranges from 0 to 1 to represent depth from the front
2949 to the back of the Z buffer. The W component contains the interpolated
2950 reciprocal of the vertex position W component (corresponding to gl_Fragcoord,
2951 but unlike d3d10 which interpolates the same 1/w but then gives back
2952 the reciprocal of the interpolated value).
2954 Fragment shaders may also declare an output register with
2955 TGSI_SEMANTIC_POSITION. Only the Z component is writable. This allows
2956 the fragment shader to change the fragment's Z position.
2963 For vertex shader outputs or fragment shader inputs/outputs, this
2964 label indicates that the register contains an R,G,B,A color.
2966 Several shader inputs/outputs may contain colors so the semantic index
2967 is used to distinguish them. For example, color[0] may be the diffuse
2968 color while color[1] may be the specular color.
2970 This label is needed so that the flat/smooth shading can be applied
2971 to the right interpolants during rasterization.
2975 TGSI_SEMANTIC_BCOLOR
2976 """"""""""""""""""""
2978 Back-facing colors are only used for back-facing polygons, and are only valid
2979 in vertex shader outputs. After rasterization, all polygons are front-facing
2980 and COLOR and BCOLOR end up occupying the same slots in the fragment shader,
2981 so all BCOLORs effectively become regular COLORs in the fragment shader.
2987 Vertex shader inputs and outputs and fragment shader inputs may be
2988 labeled with TGSI_SEMANTIC_FOG to indicate that the register contains
2989 a fog coordinate. Typically, the fragment shader will use the fog coordinate
2990 to compute a fog blend factor which is used to blend the normal fragment color
2991 with a constant fog color. But fog coord really is just an ordinary vec4
2992 register like regular semantics.
2998 Vertex shader input and output registers may be labeled with
2999 TGIS_SEMANTIC_PSIZE to indicate that the register contains a point size
3000 in the form (S, 0, 0, 1). The point size controls the width or diameter
3001 of points for rasterization. This label cannot be used in fragment
3004 When using this semantic, be sure to set the appropriate state in the
3005 :ref:`rasterizer` first.
3008 TGSI_SEMANTIC_TEXCOORD
3009 """"""""""""""""""""""
3011 Only available if PIPE_CAP_TGSI_TEXCOORD is exposed !
3013 Vertex shader outputs and fragment shader inputs may be labeled with
3014 this semantic to make them replaceable by sprite coordinates via the
3015 sprite_coord_enable state in the :ref:`rasterizer`.
3016 The semantic index permitted with this semantic is limited to <= 7.
3018 If the driver does not support TEXCOORD, sprite coordinate replacement
3019 applies to inputs with the GENERIC semantic instead.
3021 The intended use case for this semantic is gl_TexCoord.
3024 TGSI_SEMANTIC_PCOORD
3025 """"""""""""""""""""
3027 Only available if PIPE_CAP_TGSI_TEXCOORD is exposed !
3029 Fragment shader inputs may be labeled with TGSI_SEMANTIC_PCOORD to indicate
3030 that the register contains sprite coordinates in the form (x, y, 0, 1), if
3031 the current primitive is a point and point sprites are enabled. Otherwise,
3032 the contents of the register are undefined.
3034 The intended use case for this semantic is gl_PointCoord.
3037 TGSI_SEMANTIC_GENERIC
3038 """""""""""""""""""""
3040 All vertex/fragment shader inputs/outputs not labeled with any other
3041 semantic label can be considered to be generic attributes. Typical
3042 uses of generic inputs/outputs are texcoords and user-defined values.
3045 TGSI_SEMANTIC_NORMAL
3046 """"""""""""""""""""
3048 Indicates that a vertex shader input is a normal vector. This is
3049 typically only used for legacy graphics APIs.
3055 This label applies to fragment shader inputs (or system values,
3056 depending on which one is supported by the driver) and indicates that
3057 the register contains front/back-face information.
3059 If it is an input, it will be a floating-point vector in the form (F, 0, 0, 1),
3060 where F will be positive when the fragment belongs to a front-facing polygon,
3061 and negative when the fragment belongs to a back-facing polygon.
3063 If it is a system value, it will be an integer vector in the form (F, 0, 0, 1),
3064 where F is 0xffffffff when the fragment belongs to a front-facing polygon and
3065 0 when the fragment belongs to a back-facing polygon.
3068 TGSI_SEMANTIC_EDGEFLAG
3069 """"""""""""""""""""""
3071 For vertex shaders, this sematic label indicates that an input or
3072 output is a boolean edge flag. The register layout is [F, x, x, x]
3073 where F is 0.0 or 1.0 and x = don't care. Normally, the vertex shader
3074 simply copies the edge flag input to the edgeflag output.
3076 Edge flags are used to control which lines or points are actually
3077 drawn when the polygon mode converts triangles/quads/polygons into
3081 TGSI_SEMANTIC_STENCIL
3082 """""""""""""""""""""
3084 For fragment shaders, this semantic label indicates that an output
3085 is a writable stencil reference value. Only the Y component is writable.
3086 This allows the fragment shader to change the fragments stencilref value.
3089 TGSI_SEMANTIC_VIEWPORT_INDEX
3090 """"""""""""""""""""""""""""
3092 For geometry shaders, this semantic label indicates that an output
3093 contains the index of the viewport (and scissor) to use.
3094 This is an integer value, and only the X component is used.
3100 For geometry shaders, this semantic label indicates that an output
3101 contains the layer value to use for the color and depth/stencil surfaces.
3102 This is an integer value, and only the X component is used.
3103 (Also known as rendertarget array index.)
3106 TGSI_SEMANTIC_CULLDIST
3107 """"""""""""""""""""""
3109 Used as distance to plane for performing application-defined culling
3110 of individual primitives against a plane. When components of vertex
3111 elements are given this label, these values are assumed to be a
3112 float32 signed distance to a plane. Primitives will be completely
3113 discarded if the plane distance for all of the vertices in the
3114 primitive are < 0. If a vertex has a cull distance of NaN, that
3115 vertex counts as "out" (as if its < 0);
3116 The limits on both clip and cull distances are bound
3117 by the PIPE_MAX_CLIP_OR_CULL_DISTANCE_COUNT define which defines
3118 the maximum number of components that can be used to hold the
3119 distances and by the PIPE_MAX_CLIP_OR_CULL_DISTANCE_ELEMENT_COUNT
3120 which specifies the maximum number of registers which can be
3121 annotated with those semantics.
3124 TGSI_SEMANTIC_CLIPDIST
3125 """"""""""""""""""""""
3127 Note this covers clipping and culling distances.
3129 When components of vertex elements are identified this way, these
3130 values are each assumed to be a float32 signed distance to a plane.
3133 Primitive setup only invokes rasterization on pixels for which
3134 the interpolated plane distances are >= 0.
3137 Primitives will be completely discarded if the plane distance
3138 for all of the vertices in the primitive are < 0.
3139 If a vertex has a cull distance of NaN, that vertex counts as "out"
3142 Multiple clip/cull planes can be implemented simultaneously, by
3143 annotating multiple components of one or more vertex elements with
3144 the above specified semantic.
3145 The limits on both clip and cull distances are bound
3146 by the PIPE_MAX_CLIP_OR_CULL_DISTANCE_COUNT define which defines
3147 the maximum number of components that can be used to hold the
3148 distances and by the PIPE_MAX_CLIP_OR_CULL_DISTANCE_ELEMENT_COUNT
3149 which specifies the maximum number of registers which can be
3150 annotated with those semantics.
3151 The properties NUM_CLIPDIST_ENABLED and NUM_CULLDIST_ENABLED
3152 are used to divide up the 2 x vec4 space between clipping and culling.
3154 TGSI_SEMANTIC_SAMPLEID
3155 """"""""""""""""""""""
3157 For fragment shaders, this semantic label indicates that a system value
3158 contains the current sample id (i.e. gl_SampleID).
3159 This is an integer value, and only the X component is used.
3161 TGSI_SEMANTIC_SAMPLEPOS
3162 """""""""""""""""""""""
3164 For fragment shaders, this semantic label indicates that a system value
3165 contains the current sample's position (i.e. gl_SamplePosition). Only the X
3166 and Y values are used.
3168 TGSI_SEMANTIC_SAMPLEMASK
3169 """"""""""""""""""""""""
3171 For fragment shaders, this semantic label indicates that an output contains
3172 the sample mask used to disable further sample processing
3173 (i.e. gl_SampleMask). Only the X value is used, up to 32x MS.
3175 TGSI_SEMANTIC_INVOCATIONID
3176 """"""""""""""""""""""""""
3178 For geometry shaders, this semantic label indicates that a system value
3179 contains the current invocation id (i.e. gl_InvocationID).
3180 This is an integer value, and only the X component is used.
3182 TGSI_SEMANTIC_INSTANCEID
3183 """"""""""""""""""""""""
3185 For vertex shaders, this semantic label indicates that a system value contains
3186 the current instance id (i.e. gl_InstanceID). It does not include the base
3187 instance. This is an integer value, and only the X component is used.
3189 TGSI_SEMANTIC_VERTEXID
3190 """"""""""""""""""""""
3192 For vertex shaders, this semantic label indicates that a system value contains
3193 the current vertex id (i.e. gl_VertexID). It does (unlike in d3d10) include the
3194 base vertex. This is an integer value, and only the X component is used.
3196 TGSI_SEMANTIC_VERTEXID_NOBASE
3197 """""""""""""""""""""""""""""""
3199 For vertex shaders, this semantic label indicates that a system value contains
3200 the current vertex id without including the base vertex (this corresponds to
3201 d3d10 vertex id, so TGSI_SEMANTIC_VERTEXID_NOBASE + TGSI_SEMANTIC_BASEVERTEX
3202 == TGSI_SEMANTIC_VERTEXID). This is an integer value, and only the X component
3205 TGSI_SEMANTIC_BASEVERTEX
3206 """"""""""""""""""""""""
3208 For vertex shaders, this semantic label indicates that a system value contains
3209 the base vertex (i.e. gl_BaseVertex). Note that for non-indexed draw calls,
3210 this contains the first (or start) value instead.
3211 This is an integer value, and only the X component is used.
3213 TGSI_SEMANTIC_PRIMID
3214 """"""""""""""""""""
3216 For geometry and fragment shaders, this semantic label indicates the value
3217 contains the primitive id (i.e. gl_PrimitiveID). This is an integer value,
3218 and only the X component is used.
3219 FIXME: This right now can be either a ordinary input or a system value...
3225 For tessellation evaluation/control shaders, this semantic label indicates a
3226 generic per-patch attribute. Such semantics will not implicitly be per-vertex
3229 TGSI_SEMANTIC_TESSCOORD
3230 """""""""""""""""""""""
3232 For tessellation evaluation shaders, this semantic label indicates the
3233 coordinates of the vertex being processed. This is available in XYZ; W is
3236 TGSI_SEMANTIC_TESSOUTER
3237 """""""""""""""""""""""
3239 For tessellation evaluation/control shaders, this semantic label indicates the
3240 outer tessellation levels of the patch. Isoline tessellation will only have XY
3241 defined, triangle will have XYZ and quads will have XYZW defined. This
3242 corresponds to gl_TessLevelOuter.
3244 TGSI_SEMANTIC_TESSINNER
3245 """""""""""""""""""""""
3247 For tessellation evaluation/control shaders, this semantic label indicates the
3248 inner tessellation levels of the patch. The X value is only defined for
3249 triangle tessellation, while quads will have XY defined. This is entirely
3250 undefined for isoline tessellation.
3252 TGSI_SEMANTIC_VERTICESIN
3253 """"""""""""""""""""""""
3255 For tessellation evaluation/control shaders, this semantic label indicates the
3256 number of vertices provided in the input patch. Only the X value is defined.
3258 TGSI_SEMANTIC_HELPER_INVOCATION
3259 """""""""""""""""""""""""""""""
3261 For fragment shaders, this semantic indicates whether the current
3262 invocation is covered or not. Helper invocations are created in order
3263 to properly compute derivatives, however it may be desirable to skip
3264 some of the logic in those cases. See ``gl_HelperInvocation`` documentation.
3266 TGSI_SEMANTIC_BASEINSTANCE
3267 """"""""""""""""""""""""""
3269 For vertex shaders, the base instance argument supplied for this
3270 draw. This is an integer value, and only the X component is used.
3272 TGSI_SEMANTIC_DRAWID
3273 """"""""""""""""""""
3275 For vertex shaders, the zero-based index of the current draw in a
3276 ``glMultiDraw*`` invocation. This is an integer value, and only the X
3280 TGSI_SEMANTIC_WORK_DIM
3281 """"""""""""""""""""""
3283 For compute shaders started via opencl this retrieves the work_dim
3284 parameter to the clEnqueueNDRangeKernel call with which the shader
3288 TGSI_SEMANTIC_GRID_SIZE
3289 """""""""""""""""""""""
3291 For compute shaders, this semantic indicates the maximum (x, y, z) dimensions
3292 of a grid of thread blocks.
3295 TGSI_SEMANTIC_BLOCK_ID
3296 """"""""""""""""""""""
3298 For compute shaders, this semantic indicates the (x, y, z) coordinates of the
3299 current block inside of the grid.
3302 TGSI_SEMANTIC_BLOCK_SIZE
3303 """"""""""""""""""""""""
3305 For compute shaders, this semantic indicates the maximum (x, y, z) dimensions
3306 of a block in threads.
3309 TGSI_SEMANTIC_THREAD_ID
3310 """""""""""""""""""""""
3312 For compute shaders, this semantic indicates the (x, y, z) coordinates of the
3313 current thread inside of the block.
3316 Declaration Interpolate
3317 ^^^^^^^^^^^^^^^^^^^^^^^
3319 This token is only valid for fragment shader INPUT declarations.
3321 The Interpolate field specifes the way input is being interpolated by
3322 the rasteriser and is one of TGSI_INTERPOLATE_*.
3324 The Location field specifies the location inside the pixel that the
3325 interpolation should be done at, one of ``TGSI_INTERPOLATE_LOC_*``. Note that
3326 when per-sample shading is enabled, the implementation may choose to
3327 interpolate at the sample irrespective of the Location field.
3329 The CylindricalWrap bitfield specifies which register components
3330 should be subject to cylindrical wrapping when interpolating by the
3331 rasteriser. If TGSI_CYLINDRICAL_WRAP_X is set to 1, the X component
3332 should be interpolated according to cylindrical wrapping rules.
3335 Declaration Sampler View
3336 ^^^^^^^^^^^^^^^^^^^^^^^^
3338 Follows Declaration token if file is TGSI_FILE_SAMPLER_VIEW.
3340 DCL SVIEW[#], resource, type(s)
3342 Declares a shader input sampler view and assigns it to a SVIEW[#]
3345 resource can be one of BUFFER, 1D, 2D, 3D, 1DArray and 2DArray.
3347 type must be 1 or 4 entries (if specifying on a per-component
3348 level) out of UNORM, SNORM, SINT, UINT and FLOAT.
3350 For TEX\* style texture sample opcodes (as opposed to SAMPLE\* opcodes
3351 which take an explicit SVIEW[#] source register), there may be optionally
3352 SVIEW[#] declarations. In this case, the SVIEW index is implied by the
3353 SAMP index, and there must be a corresponding SVIEW[#] declaration for
3354 each SAMP[#] declaration. Drivers are free to ignore this if they wish.
3355 But note in particular that some drivers need to know the sampler type
3356 (float/int/unsigned) in order to generate the correct code, so cases
3357 where integer textures are sampled, SVIEW[#] declarations should be
3360 NOTE: It is NOT legal to mix SAMPLE\* style opcodes and TEX\* opcodes
3363 Declaration Resource
3364 ^^^^^^^^^^^^^^^^^^^^
3366 Follows Declaration token if file is TGSI_FILE_RESOURCE.
3368 DCL RES[#], resource [, WR] [, RAW]
3370 Declares a shader input resource and assigns it to a RES[#]
3373 resource can be one of BUFFER, 1D, 2D, 3D, CUBE, 1DArray and
3376 If the RAW keyword is not specified, the texture data will be
3377 subject to conversion, swizzling and scaling as required to yield
3378 the specified data type from the physical data format of the bound
3381 If the RAW keyword is specified, no channel conversion will be
3382 performed: the values read for each of the channels (X,Y,Z,W) will
3383 correspond to consecutive words in the same order and format
3384 they're found in memory. No element-to-address conversion will be
3385 performed either: the value of the provided X coordinate will be
3386 interpreted in byte units instead of texel units. The result of
3387 accessing a misaligned address is undefined.
3389 Usage of the STORE opcode is only allowed if the WR (writable) flag
3394 ^^^^^^^^^^^^^^^^^^^^^^^^
3396 Properties are general directives that apply to the whole TGSI program.
3401 Specifies the fragment shader TGSI_SEMANTIC_POSITION coordinate origin.
3402 The default value is UPPER_LEFT.
3404 If UPPER_LEFT, the position will be (0,0) at the upper left corner and
3405 increase downward and rightward.
3406 If LOWER_LEFT, the position will be (0,0) at the lower left corner and
3407 increase upward and rightward.
3409 OpenGL defaults to LOWER_LEFT, and is configurable with the
3410 GL_ARB_fragment_coord_conventions extension.
3412 DirectX 9/10 use UPPER_LEFT.
3414 FS_COORD_PIXEL_CENTER
3415 """""""""""""""""""""
3417 Specifies the fragment shader TGSI_SEMANTIC_POSITION pixel center convention.
3418 The default value is HALF_INTEGER.
3420 If HALF_INTEGER, the fractionary part of the position will be 0.5
3421 If INTEGER, the fractionary part of the position will be 0.0
3423 Note that this does not affect the set of fragments generated by
3424 rasterization, which is instead controlled by half_pixel_center in the
3427 OpenGL defaults to HALF_INTEGER, and is configurable with the
3428 GL_ARB_fragment_coord_conventions extension.
3430 DirectX 9 uses INTEGER.
3431 DirectX 10 uses HALF_INTEGER.
3433 FS_COLOR0_WRITES_ALL_CBUFS
3434 """"""""""""""""""""""""""
3435 Specifies that writes to the fragment shader color 0 are replicated to all
3436 bound cbufs. This facilitates OpenGL's fragColor output vs fragData[0] where
3437 fragData is directed to a single color buffer, but fragColor is broadcast.
3440 """"""""""""""""""""""""""
3441 If this property is set on the program bound to the shader stage before the
3442 fragment shader, user clip planes should have no effect (be disabled) even if
3443 that shader does not write to any clip distance outputs and the rasterizer's
3444 clip_plane_enable is non-zero.
3445 This property is only supported by drivers that also support shader clip
3447 This is useful for APIs that don't have UCPs and where clip distances written
3448 by a shader cannot be disabled.
3453 Specifies the number of times a geometry shader should be executed for each
3454 input primitive. Each invocation will have a different
3455 TGSI_SEMANTIC_INVOCATIONID system value set. If not specified, assumed to
3458 VS_WINDOW_SPACE_POSITION
3459 """"""""""""""""""""""""""
3460 If this property is set on the vertex shader, the TGSI_SEMANTIC_POSITION output
3461 is assumed to contain window space coordinates.
3462 Division of X,Y,Z by W and the viewport transformation are disabled, and 1/W is
3463 directly taken from the 4-th component of the shader output.
3464 Naturally, clipping is not performed on window coordinates either.
3465 The effect of this property is undefined if a geometry or tessellation shader
3471 The number of vertices written by the tessellation control shader. This
3472 effectively defines the patch input size of the tessellation evaluation shader
3478 This sets the tessellation primitive mode, one of ``PIPE_PRIM_TRIANGLES``,
3479 ``PIPE_PRIM_QUADS``, or ``PIPE_PRIM_LINES``. (Unlike in GL, there is no
3480 separate isolines settings, the regular lines is assumed to mean isolines.)
3485 This sets the spacing mode of the tessellation generator, one of
3486 ``PIPE_TESS_SPACING_*``.
3491 This sets the vertex order to be clockwise if the value is 1, or
3492 counter-clockwise if set to 0.
3497 If set to a non-zero value, this turns on point mode for the tessellator,
3498 which means that points will be generated instead of primitives.
3500 NUM_CLIPDIST_ENABLED
3503 How many clip distance scalar outputs are enabled.
3505 NUM_CULLDIST_ENABLED
3508 How many cull distance scalar outputs are enabled.
3510 FS_EARLY_DEPTH_STENCIL
3511 """"""""""""""""""""""
3513 Whether depth test, stencil test, and occlusion query should run before
3514 the fragment shader (regardless of fragment shader side effects). Corresponds
3515 to GLSL early_fragment_tests.
3520 Which shader stage will MOST LIKELY follow after this shader when the shader
3521 is bound. This is only a hint to the driver and doesn't have to be precise.
3522 Only set for VS and TES.
3524 CS_FIXED_BLOCK_WIDTH / HEIGHT / DEPTH
3525 """""""""""""""""""""""""""""""""""""
3527 Threads per block in each dimension, if known at compile time. If the block size
3528 is known all three should be at least 1. If it is unknown they should all be set
3534 The MUL TGSI operation (FP32 multiplication) will return 0 if either
3535 of the operands are equal to 0. That means that 0 * Inf = 0. This
3536 should be set the same way for an entire pipeline. Note that this
3537 applies not only to the literal MUL TGSI opcode, but all FP32
3538 multiplications implied by other operations, such as MAD, FMA, DP2,
3539 DP3, DP4, DPH, DST, LOG, LRP, XPD, and possibly others. If there is a
3540 mismatch between shaders, then it is unspecified whether this behavior
3544 Texture Sampling and Texture Formats
3545 ------------------------------------
3547 This table shows how texture image components are returned as (x,y,z,w) tuples
3548 by TGSI texture instructions, such as :opcode:`TEX`, :opcode:`TXD`, and
3549 :opcode:`TXP`. For reference, OpenGL and Direct3D conventions are shown as
3552 +--------------------+--------------+--------------------+--------------+
3553 | Texture Components | Gallium | OpenGL | Direct3D 9 |
3554 +====================+==============+====================+==============+
3555 | R | (r, 0, 0, 1) | (r, 0, 0, 1) | (r, 1, 1, 1) |
3556 +--------------------+--------------+--------------------+--------------+
3557 | RG | (r, g, 0, 1) | (r, g, 0, 1) | (r, g, 1, 1) |
3558 +--------------------+--------------+--------------------+--------------+
3559 | RGB | (r, g, b, 1) | (r, g, b, 1) | (r, g, b, 1) |
3560 +--------------------+--------------+--------------------+--------------+
3561 | RGBA | (r, g, b, a) | (r, g, b, a) | (r, g, b, a) |
3562 +--------------------+--------------+--------------------+--------------+
3563 | A | (0, 0, 0, a) | (0, 0, 0, a) | (0, 0, 0, a) |
3564 +--------------------+--------------+--------------------+--------------+
3565 | L | (l, l, l, 1) | (l, l, l, 1) | (l, l, l, 1) |
3566 +--------------------+--------------+--------------------+--------------+
3567 | LA | (l, l, l, a) | (l, l, l, a) | (l, l, l, a) |
3568 +--------------------+--------------+--------------------+--------------+
3569 | I | (i, i, i, i) | (i, i, i, i) | N/A |
3570 +--------------------+--------------+--------------------+--------------+
3571 | UV | XXX TBD | (0, 0, 0, 1) | (u, v, 1, 1) |
3572 | | | [#envmap-bumpmap]_ | |
3573 +--------------------+--------------+--------------------+--------------+
3574 | Z | XXX TBD | (z, z, z, 1) | (0, z, 0, 1) |
3575 | | | [#depth-tex-mode]_ | |
3576 +--------------------+--------------+--------------------+--------------+
3577 | S | (s, s, s, s) | unknown | unknown |
3578 +--------------------+--------------+--------------------+--------------+
3580 .. [#envmap-bumpmap] http://www.opengl.org/registry/specs/ATI/envmap_bumpmap.txt
3581 .. [#depth-tex-mode] the default is (z, z, z, 1) but may also be (0, 0, 0, z)
3582 or (z, z, z, z) depending on the value of GL_DEPTH_TEXTURE_MODE.