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:: SUB - Subtract
253 dst.x = src0.x - src1.x
255 dst.y = src0.y - src1.y
257 dst.z = src0.z - src1.z
259 dst.w = src0.w - src1.w
262 .. opcode:: LRP - Linear Interpolate
266 dst.x = src0.x \times src1.x + (1 - src0.x) \times src2.x
268 dst.y = src0.y \times src1.y + (1 - src0.y) \times src2.y
270 dst.z = src0.z \times src1.z + (1 - src0.z) \times src2.z
272 dst.w = src0.w \times src1.w + (1 - src0.w) \times src2.w
275 .. opcode:: FMA - Fused Multiply-Add
277 Perform a * b + c with no intermediate rounding step.
281 dst.x = src0.x \times src1.x + src2.x
283 dst.y = src0.y \times src1.y + src2.y
285 dst.z = src0.z \times src1.z + src2.z
287 dst.w = src0.w \times src1.w + src2.w
290 .. opcode:: DP2A - 2-component Dot Product And Add
294 dst.x = src0.x \times src1.x + src0.y \times src1.y + src2.x
296 dst.y = src0.x \times src1.x + src0.y \times src1.y + src2.x
298 dst.z = src0.x \times src1.x + src0.y \times src1.y + src2.x
300 dst.w = src0.x \times src1.x + src0.y \times src1.y + src2.x
303 .. opcode:: FRC - Fraction
307 dst.x = src.x - \lfloor src.x\rfloor
309 dst.y = src.y - \lfloor src.y\rfloor
311 dst.z = src.z - \lfloor src.z\rfloor
313 dst.w = src.w - \lfloor src.w\rfloor
316 .. opcode:: CLAMP - Clamp
320 dst.x = clamp(src0.x, src1.x, src2.x)
322 dst.y = clamp(src0.y, src1.y, src2.y)
324 dst.z = clamp(src0.z, src1.z, src2.z)
326 dst.w = clamp(src0.w, src1.w, src2.w)
329 .. opcode:: FLR - Floor
333 dst.x = \lfloor src.x\rfloor
335 dst.y = \lfloor src.y\rfloor
337 dst.z = \lfloor src.z\rfloor
339 dst.w = \lfloor src.w\rfloor
342 .. opcode:: ROUND - Round
355 .. opcode:: EX2 - Exponential Base 2
357 This instruction replicates its result.
364 .. opcode:: LG2 - Logarithm Base 2
366 This instruction replicates its result.
373 .. opcode:: POW - Power
375 This instruction replicates its result.
379 dst = src0.x^{src1.x}
381 .. opcode:: XPD - Cross Product
385 dst.x = src0.y \times src1.z - src1.y \times src0.z
387 dst.y = src0.z \times src1.x - src1.z \times src0.x
389 dst.z = src0.x \times src1.y - src1.x \times src0.y
394 .. opcode:: ABS - Absolute
407 .. opcode:: DPH - Homogeneous Dot Product
409 This instruction replicates its result.
413 dst = src0.x \times src1.x + src0.y \times src1.y + src0.z \times src1.z + src1.w
416 .. opcode:: COS - Cosine
418 This instruction replicates its result.
425 .. opcode:: DDX, DDX_FINE - Derivative Relative To X
427 The fine variant is only used when ``PIPE_CAP_TGSI_FS_FINE_DERIVATIVE`` is
428 advertised. When it is, the fine version guarantees one derivative per row
429 while DDX is allowed to be the same for the entire 2x2 quad.
433 dst.x = partialx(src.x)
435 dst.y = partialx(src.y)
437 dst.z = partialx(src.z)
439 dst.w = partialx(src.w)
442 .. opcode:: DDY, DDY_FINE - Derivative Relative To Y
444 The fine variant is only used when ``PIPE_CAP_TGSI_FS_FINE_DERIVATIVE`` is
445 advertised. When it is, the fine version guarantees one derivative per column
446 while DDY is allowed to be the same for the entire 2x2 quad.
450 dst.x = partialy(src.x)
452 dst.y = partialy(src.y)
454 dst.z = partialy(src.z)
456 dst.w = partialy(src.w)
459 .. opcode:: PK2H - Pack Two 16-bit Floats
461 This instruction replicates its result.
465 dst = f32\_to\_f16(src.x) | f32\_to\_f16(src.y) << 16
468 .. opcode:: PK2US - Pack Two Unsigned 16-bit Scalars
473 .. opcode:: PK4B - Pack Four Signed 8-bit Scalars
478 .. opcode:: PK4UB - Pack Four Unsigned 8-bit Scalars
483 .. opcode:: SEQ - Set On Equal
487 dst.x = (src0.x == src1.x) ? 1.0F : 0.0F
489 dst.y = (src0.y == src1.y) ? 1.0F : 0.0F
491 dst.z = (src0.z == src1.z) ? 1.0F : 0.0F
493 dst.w = (src0.w == src1.w) ? 1.0F : 0.0F
496 .. opcode:: SGT - Set On Greater Than
500 dst.x = (src0.x > src1.x) ? 1.0F : 0.0F
502 dst.y = (src0.y > src1.y) ? 1.0F : 0.0F
504 dst.z = (src0.z > src1.z) ? 1.0F : 0.0F
506 dst.w = (src0.w > src1.w) ? 1.0F : 0.0F
509 .. opcode:: SIN - Sine
511 This instruction replicates its result.
518 .. opcode:: SLE - Set On Less Equal Than
522 dst.x = (src0.x <= src1.x) ? 1.0F : 0.0F
524 dst.y = (src0.y <= src1.y) ? 1.0F : 0.0F
526 dst.z = (src0.z <= src1.z) ? 1.0F : 0.0F
528 dst.w = (src0.w <= src1.w) ? 1.0F : 0.0F
531 .. opcode:: SNE - Set On Not Equal
535 dst.x = (src0.x != src1.x) ? 1.0F : 0.0F
537 dst.y = (src0.y != src1.y) ? 1.0F : 0.0F
539 dst.z = (src0.z != src1.z) ? 1.0F : 0.0F
541 dst.w = (src0.w != src1.w) ? 1.0F : 0.0F
544 .. opcode:: TEX - Texture Lookup
546 for array textures src0.y contains the slice for 1D,
547 and src0.z contain the slice for 2D.
549 for shadow textures with no arrays (and not cube map),
550 src0.z contains the reference value.
552 for shadow textures with arrays, src0.z contains
553 the reference value for 1D arrays, and src0.w contains
554 the reference value for 2D arrays and cube maps.
556 for cube map array shadow textures, the reference value
557 cannot be passed in src0.w, and TEX2 must be used instead.
563 shadow_ref = src0.z or src0.w (optional)
567 dst = texture\_sample(unit, coord, shadow_ref)
570 .. opcode:: TEX2 - Texture Lookup (for shadow cube map arrays only)
572 this is the same as TEX, but uses another reg to encode the
583 dst = texture\_sample(unit, coord, shadow_ref)
588 .. opcode:: TXD - Texture Lookup with Derivatives
600 dst = texture\_sample\_deriv(unit, coord, ddx, ddy)
603 .. opcode:: TXP - Projective Texture Lookup
607 coord.x = src0.x / src0.w
609 coord.y = src0.y / src0.w
611 coord.z = src0.z / src0.w
617 dst = texture\_sample(unit, coord)
620 .. opcode:: UP2H - Unpack Two 16-Bit Floats
624 dst.x = f16\_to\_f32(src0.x \& 0xffff)
626 dst.y = f16\_to\_f32(src0.x >> 16)
628 dst.z = f16\_to\_f32(src0.x \& 0xffff)
630 dst.w = f16\_to\_f32(src0.x >> 16)
634 Considered for removal.
636 .. opcode:: UP2US - Unpack Two Unsigned 16-Bit Scalars
642 Considered for removal.
644 .. opcode:: UP4B - Unpack Four Signed 8-Bit Values
650 Considered for removal.
652 .. opcode:: UP4UB - Unpack Four Unsigned 8-Bit Scalars
658 Considered for removal.
661 .. opcode:: ARR - Address Register Load With Round
665 dst.x = (int) round(src.x)
667 dst.y = (int) round(src.y)
669 dst.z = (int) round(src.z)
671 dst.w = (int) round(src.w)
674 .. opcode:: SSG - Set Sign
678 dst.x = (src.x > 0) ? 1 : (src.x < 0) ? -1 : 0
680 dst.y = (src.y > 0) ? 1 : (src.y < 0) ? -1 : 0
682 dst.z = (src.z > 0) ? 1 : (src.z < 0) ? -1 : 0
684 dst.w = (src.w > 0) ? 1 : (src.w < 0) ? -1 : 0
687 .. opcode:: CMP - Compare
691 dst.x = (src0.x < 0) ? src1.x : src2.x
693 dst.y = (src0.y < 0) ? src1.y : src2.y
695 dst.z = (src0.z < 0) ? src1.z : src2.z
697 dst.w = (src0.w < 0) ? src1.w : src2.w
700 .. opcode:: KILL_IF - Conditional Discard
702 Conditional discard. Allowed in fragment shaders only.
706 if (src.x < 0 || src.y < 0 || src.z < 0 || src.w < 0)
711 .. opcode:: KILL - Discard
713 Unconditional discard. Allowed in fragment shaders only.
716 .. opcode:: SCS - Sine Cosine
729 .. opcode:: TXB - Texture Lookup With Bias
731 for cube map array textures and shadow cube maps, the bias value
732 cannot be passed in src0.w, and TXB2 must be used instead.
734 if the target is a shadow texture, the reference value is always
735 in src.z (this prevents shadow 3d and shadow 2d arrays from
736 using this instruction, but this is not needed).
752 dst = texture\_sample(unit, coord, bias)
755 .. opcode:: TXB2 - Texture Lookup With Bias (some cube maps only)
757 this is the same as TXB, but uses another reg to encode the
758 lod bias value for cube map arrays and shadow cube maps.
759 Presumably shadow 2d arrays and shadow 3d targets could use
760 this encoding too, but this is not legal.
762 shadow cube map arrays are neither possible nor required.
772 dst = texture\_sample(unit, coord, bias)
775 .. opcode:: DIV - Divide
779 dst.x = \frac{src0.x}{src1.x}
781 dst.y = \frac{src0.y}{src1.y}
783 dst.z = \frac{src0.z}{src1.z}
785 dst.w = \frac{src0.w}{src1.w}
788 .. opcode:: DP2 - 2-component Dot Product
790 This instruction replicates its result.
794 dst = src0.x \times src1.x + src0.y \times src1.y
797 .. opcode:: TXL - Texture Lookup With explicit LOD
799 for cube map array textures, the explicit lod value
800 cannot be passed in src0.w, and TXL2 must be used instead.
802 if the target is a shadow texture, the reference value is always
803 in src.z (this prevents shadow 3d / 2d array / cube targets from
804 using this instruction, but this is not needed).
820 dst = texture\_sample(unit, coord, lod)
823 .. opcode:: TXL2 - Texture Lookup With explicit LOD (for cube map arrays only)
825 this is the same as TXL, but uses another reg to encode the
827 Presumably shadow 3d / 2d array / cube targets could use
828 this encoding too, but this is not legal.
830 shadow cube map arrays are neither possible nor required.
840 dst = texture\_sample(unit, coord, lod)
843 .. opcode:: PUSHA - Push Address Register On Stack
852 Considered for cleanup.
856 Considered for removal.
858 .. opcode:: POPA - Pop Address Register From Stack
867 Considered for cleanup.
871 Considered for removal.
874 .. opcode:: CALLNZ - Subroutine Call If Not Zero
880 Considered for cleanup.
884 Considered for removal.
888 ^^^^^^^^^^^^^^^^^^^^^^^^
890 These opcodes are primarily provided for special-use computational shaders.
891 Support for these opcodes indicated by a special pipe capability bit (TBD).
893 XXX doesn't look like most of the opcodes really belong here.
895 .. opcode:: CEIL - Ceiling
899 dst.x = \lceil src.x\rceil
901 dst.y = \lceil src.y\rceil
903 dst.z = \lceil src.z\rceil
905 dst.w = \lceil src.w\rceil
908 .. opcode:: TRUNC - Truncate
921 .. opcode:: MOD - Modulus
925 dst.x = src0.x \bmod src1.x
927 dst.y = src0.y \bmod src1.y
929 dst.z = src0.z \bmod src1.z
931 dst.w = src0.w \bmod src1.w
934 .. opcode:: UARL - Integer Address Register Load
936 Moves the contents of the source register, assumed to be an integer, into the
937 destination register, which is assumed to be an address (ADDR) register.
940 .. opcode:: SAD - Sum Of Absolute Differences
944 dst.x = |src0.x - src1.x| + src2.x
946 dst.y = |src0.y - src1.y| + src2.y
948 dst.z = |src0.z - src1.z| + src2.z
950 dst.w = |src0.w - src1.w| + src2.w
953 .. opcode:: TXF - Texel Fetch
955 As per NV_gpu_shader4, extract a single texel from a specified texture
956 image. The source sampler may not be a CUBE or SHADOW. src 0 is a
957 four-component signed integer vector used to identify the single texel
958 accessed. 3 components + level. Just like texture instructions, an optional
959 offset vector is provided, which is subject to various driver restrictions
960 (regarding range, source of offsets).
961 TXF(uint_vec coord, int_vec offset).
964 .. opcode:: TXQ - Texture Size Query
966 As per NV_gpu_program4, retrieve the dimensions of the texture depending on
967 the target. For 1D (width), 2D/RECT/CUBE (width, height), 3D (width, height,
968 depth), 1D array (width, layers), 2D array (width, height, layers).
969 Also return the number of accessible levels (last_level - first_level + 1)
972 For components which don't return a resource dimension, their value
979 dst.x = texture\_width(unit, lod)
981 dst.y = texture\_height(unit, lod)
983 dst.z = texture\_depth(unit, lod)
985 dst.w = texture\_levels(unit)
988 .. opcode:: TXQS - Texture Samples Query
990 This retrieves the number of samples in the texture, and stores it
991 into the x component. The other components are undefined.
995 dst.x = texture\_samples(unit)
998 .. opcode:: TG4 - Texture Gather
1000 As per ARB_texture_gather, gathers the four texels to be used in a bi-linear
1001 filtering operation and packs them into a single register. Only works with
1002 2D, 2D array, cubemaps, and cubemaps arrays. For 2D textures, only the
1003 addressing modes of the sampler and the top level of any mip pyramid are
1004 used. Set W to zero. It behaves like the TEX instruction, but a filtered
1005 sample is not generated. The four samples that contribute to filtering are
1006 placed into xyzw in clockwise order, starting with the (u,v) texture
1007 coordinate delta at the following locations (-, +), (+, +), (+, -), (-, -),
1008 where the magnitude of the deltas are half a texel.
1010 PIPE_CAP_TEXTURE_SM5 enhances this instruction to support shadow per-sample
1011 depth compares, single component selection, and a non-constant offset. It
1012 doesn't allow support for the GL independent offset to get i0,j0. This would
1013 require another CAP is hw can do it natively. For now we lower that before
1022 dst = texture\_gather4 (unit, coord, component)
1024 (with SM5 - cube array shadow)
1032 dst = texture\_gather (uint, coord, compare)
1034 .. opcode:: LODQ - level of detail query
1036 Compute the LOD information that the texture pipe would use to access the
1037 texture. The Y component contains the computed LOD lambda_prime. The X
1038 component contains the LOD that will be accessed, based on min/max lod's
1045 dst.xy = lodq(uint, coord);
1048 ^^^^^^^^^^^^^^^^^^^^^^^^
1049 These opcodes are used for integer operations.
1050 Support for these opcodes indicated by PIPE_SHADER_CAP_INTEGERS (all of them?)
1053 .. opcode:: I2F - Signed Integer To Float
1055 Rounding is unspecified (round to nearest even suggested).
1059 dst.x = (float) src.x
1061 dst.y = (float) src.y
1063 dst.z = (float) src.z
1065 dst.w = (float) src.w
1068 .. opcode:: U2F - Unsigned Integer To Float
1070 Rounding is unspecified (round to nearest even suggested).
1074 dst.x = (float) src.x
1076 dst.y = (float) src.y
1078 dst.z = (float) src.z
1080 dst.w = (float) src.w
1083 .. opcode:: F2I - Float to Signed Integer
1085 Rounding is towards zero (truncate).
1086 Values outside signed range (including NaNs) produce undefined results.
1099 .. opcode:: F2U - Float to Unsigned Integer
1101 Rounding is towards zero (truncate).
1102 Values outside unsigned range (including NaNs) produce undefined results.
1106 dst.x = (unsigned) src.x
1108 dst.y = (unsigned) src.y
1110 dst.z = (unsigned) src.z
1112 dst.w = (unsigned) src.w
1115 .. opcode:: UADD - Integer Add
1117 This instruction works the same for signed and unsigned integers.
1118 The low 32bit of the result is returned.
1122 dst.x = src0.x + src1.x
1124 dst.y = src0.y + src1.y
1126 dst.z = src0.z + src1.z
1128 dst.w = src0.w + src1.w
1131 .. opcode:: UMAD - Integer Multiply And Add
1133 This instruction works the same for signed and unsigned integers.
1134 The multiplication returns the low 32bit (as does the result itself).
1138 dst.x = src0.x \times src1.x + src2.x
1140 dst.y = src0.y \times src1.y + src2.y
1142 dst.z = src0.z \times src1.z + src2.z
1144 dst.w = src0.w \times src1.w + src2.w
1147 .. opcode:: UMUL - Integer Multiply
1149 This instruction works the same for signed and unsigned integers.
1150 The low 32bit of the result is returned.
1154 dst.x = src0.x \times src1.x
1156 dst.y = src0.y \times src1.y
1158 dst.z = src0.z \times src1.z
1160 dst.w = src0.w \times src1.w
1163 .. opcode:: IMUL_HI - Signed Integer Multiply High Bits
1165 The high 32bits of the multiplication of 2 signed integers are returned.
1169 dst.x = (src0.x \times src1.x) >> 32
1171 dst.y = (src0.y \times src1.y) >> 32
1173 dst.z = (src0.z \times src1.z) >> 32
1175 dst.w = (src0.w \times src1.w) >> 32
1178 .. opcode:: UMUL_HI - Unsigned Integer Multiply High Bits
1180 The high 32bits of the multiplication of 2 unsigned integers are returned.
1184 dst.x = (src0.x \times src1.x) >> 32
1186 dst.y = (src0.y \times src1.y) >> 32
1188 dst.z = (src0.z \times src1.z) >> 32
1190 dst.w = (src0.w \times src1.w) >> 32
1193 .. opcode:: IDIV - Signed Integer Division
1195 TBD: behavior for division by zero.
1199 dst.x = src0.x \ src1.x
1201 dst.y = src0.y \ src1.y
1203 dst.z = src0.z \ src1.z
1205 dst.w = src0.w \ src1.w
1208 .. opcode:: UDIV - Unsigned Integer Division
1210 For division by zero, 0xffffffff is returned.
1214 dst.x = src0.x \ src1.x
1216 dst.y = src0.y \ src1.y
1218 dst.z = src0.z \ src1.z
1220 dst.w = src0.w \ src1.w
1223 .. opcode:: UMOD - Unsigned Integer Remainder
1225 If second arg is zero, 0xffffffff is returned.
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:: NOT - Bitwise Not
1251 .. opcode:: AND - Bitwise And
1255 dst.x = src0.x \& src1.x
1257 dst.y = src0.y \& src1.y
1259 dst.z = src0.z \& src1.z
1261 dst.w = src0.w \& src1.w
1264 .. opcode:: OR - Bitwise Or
1268 dst.x = src0.x | src1.x
1270 dst.y = src0.y | src1.y
1272 dst.z = src0.z | src1.z
1274 dst.w = src0.w | src1.w
1277 .. opcode:: XOR - Bitwise Xor
1281 dst.x = src0.x \oplus src1.x
1283 dst.y = src0.y \oplus src1.y
1285 dst.z = src0.z \oplus src1.z
1287 dst.w = src0.w \oplus src1.w
1290 .. opcode:: IMAX - Maximum of Signed Integers
1294 dst.x = max(src0.x, src1.x)
1296 dst.y = max(src0.y, src1.y)
1298 dst.z = max(src0.z, src1.z)
1300 dst.w = max(src0.w, src1.w)
1303 .. opcode:: UMAX - Maximum of Unsigned Integers
1307 dst.x = max(src0.x, src1.x)
1309 dst.y = max(src0.y, src1.y)
1311 dst.z = max(src0.z, src1.z)
1313 dst.w = max(src0.w, src1.w)
1316 .. opcode:: IMIN - Minimum of Signed Integers
1320 dst.x = min(src0.x, src1.x)
1322 dst.y = min(src0.y, src1.y)
1324 dst.z = min(src0.z, src1.z)
1326 dst.w = min(src0.w, src1.w)
1329 .. opcode:: UMIN - Minimum of Unsigned Integers
1333 dst.x = min(src0.x, src1.x)
1335 dst.y = min(src0.y, src1.y)
1337 dst.z = min(src0.z, src1.z)
1339 dst.w = min(src0.w, src1.w)
1342 .. opcode:: SHL - Shift Left
1344 The shift count is masked with 0x1f before the shift is applied.
1348 dst.x = src0.x << (0x1f \& src1.x)
1350 dst.y = src0.y << (0x1f \& src1.y)
1352 dst.z = src0.z << (0x1f \& src1.z)
1354 dst.w = src0.w << (0x1f \& src1.w)
1357 .. opcode:: ISHR - Arithmetic Shift Right (of Signed Integer)
1359 The shift count is masked with 0x1f before the shift is applied.
1363 dst.x = src0.x >> (0x1f \& src1.x)
1365 dst.y = src0.y >> (0x1f \& src1.y)
1367 dst.z = src0.z >> (0x1f \& src1.z)
1369 dst.w = src0.w >> (0x1f \& src1.w)
1372 .. opcode:: USHR - Logical Shift Right
1374 The shift count is masked with 0x1f before the shift is applied.
1378 dst.x = src0.x >> (unsigned) (0x1f \& src1.x)
1380 dst.y = src0.y >> (unsigned) (0x1f \& src1.y)
1382 dst.z = src0.z >> (unsigned) (0x1f \& src1.z)
1384 dst.w = src0.w >> (unsigned) (0x1f \& src1.w)
1387 .. opcode:: UCMP - Integer Conditional Move
1391 dst.x = src0.x ? src1.x : src2.x
1393 dst.y = src0.y ? src1.y : src2.y
1395 dst.z = src0.z ? src1.z : src2.z
1397 dst.w = src0.w ? src1.w : src2.w
1401 .. opcode:: ISSG - Integer Set Sign
1405 dst.x = (src0.x < 0) ? -1 : (src0.x > 0) ? 1 : 0
1407 dst.y = (src0.y < 0) ? -1 : (src0.y > 0) ? 1 : 0
1409 dst.z = (src0.z < 0) ? -1 : (src0.z > 0) ? 1 : 0
1411 dst.w = (src0.w < 0) ? -1 : (src0.w > 0) ? 1 : 0
1415 .. opcode:: FSLT - Float Set On Less Than (ordered)
1417 Same comparison as SLT but returns integer instead of 1.0/0.0 float
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:: ISLT - Signed Integer Set On Less Than
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:: USLT - Unsigned Integer Set On Less 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:: FSGE - Float Set On Greater Equal Than (ordered)
1458 Same comparison as SGE but returns integer instead of 1.0/0.0 float
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:: ISGE - Signed Integer Set On Greater Equal Than
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:: USGE - Unsigned Integer Set On Greater Equal Than
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:: FSEQ - Float Set On Equal (ordered)
1499 Same comparison as SEQ 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:: USEQ - Integer Set On 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:: FSNE - Float Set On Not Equal (unordered)
1527 Same comparison as SNE but returns integer instead of 1.0/0.0 float
1531 dst.x = (src0.x != src1.x) ? \sim 0 : 0
1533 dst.y = (src0.y != src1.y) ? \sim 0 : 0
1535 dst.z = (src0.z != src1.z) ? \sim 0 : 0
1537 dst.w = (src0.w != src1.w) ? \sim 0 : 0
1540 .. opcode:: USNE - Integer Set On Not Equal
1544 dst.x = (src0.x != src1.x) ? \sim 0 : 0
1546 dst.y = (src0.y != src1.y) ? \sim 0 : 0
1548 dst.z = (src0.z != src1.z) ? \sim 0 : 0
1550 dst.w = (src0.w != src1.w) ? \sim 0 : 0
1553 .. opcode:: INEG - Integer Negate
1568 .. opcode:: IABS - Integer Absolute Value
1582 These opcodes are used for bit-level manipulation of integers.
1584 .. opcode:: IBFE - Signed Bitfield Extract
1586 Like GLSL bitfieldExtract. Extracts a set of bits from the input, and
1587 sign-extends them if the high bit of the extracted window is set.
1591 def ibfe(value, offset, bits):
1592 if offset < 0 or bits < 0 or offset + bits > 32:
1594 if bits == 0: return 0
1595 # Note: >> sign-extends
1596 return (value << (32 - offset - bits)) >> (32 - bits)
1598 .. opcode:: UBFE - Unsigned Bitfield Extract
1600 Like GLSL bitfieldExtract. Extracts a set of bits from the input, without
1605 def ubfe(value, offset, bits):
1606 if offset < 0 or bits < 0 or offset + bits > 32:
1608 if bits == 0: return 0
1609 # Note: >> does not sign-extend
1610 return (value << (32 - offset - bits)) >> (32 - bits)
1612 .. opcode:: BFI - Bitfield Insert
1614 Like GLSL bitfieldInsert. Replaces a bit region of 'base' with the low bits
1619 def bfi(base, insert, offset, bits):
1620 if offset < 0 or bits < 0 or offset + bits > 32:
1622 # << defined such that mask == ~0 when bits == 32, offset == 0
1623 mask = ((1 << bits) - 1) << offset
1624 return ((insert << offset) & mask) | (base & ~mask)
1626 .. opcode:: BREV - Bitfield Reverse
1628 See SM5 instruction BFREV. Reverses the bits of the argument.
1630 .. opcode:: POPC - Population Count
1632 See SM5 instruction COUNTBITS. Counts the number of set bits in the argument.
1634 .. opcode:: LSB - Index of lowest set bit
1636 See SM5 instruction FIRSTBIT_LO. Computes the 0-based index of the first set
1637 bit of the argument. Returns -1 if none are set.
1639 .. opcode:: IMSB - Index of highest non-sign bit
1641 See SM5 instruction FIRSTBIT_SHI. Computes the 0-based index of the highest
1642 non-sign bit of the argument (i.e. highest 0 bit for negative numbers,
1643 highest 1 bit for positive numbers). Returns -1 if all bits are the same
1644 (i.e. for inputs 0 and -1).
1646 .. opcode:: UMSB - Index of highest set bit
1648 See SM5 instruction FIRSTBIT_HI. Computes the 0-based index of the highest
1649 set bit of the argument. Returns -1 if none are set.
1652 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
1654 These opcodes are only supported in geometry shaders; they have no meaning
1655 in any other type of shader.
1657 .. opcode:: EMIT - Emit
1659 Generate a new vertex for the current primitive into the specified vertex
1660 stream using the values in the output registers.
1663 .. opcode:: ENDPRIM - End Primitive
1665 Complete the current primitive in the specified vertex stream (consisting of
1666 the emitted vertices), and start a new one.
1672 These opcodes are part of :term:`GLSL`'s opcode set. Support for these
1673 opcodes is determined by a special capability bit, ``GLSL``.
1674 Some require glsl version 1.30 (UIF/BREAKC/SWITCH/CASE/DEFAULT/ENDSWITCH).
1676 .. opcode:: CAL - Subroutine Call
1682 .. opcode:: RET - Subroutine Call Return
1687 .. opcode:: CONT - Continue
1689 Unconditionally moves the point of execution to the instruction after the
1690 last bgnloop. The instruction must appear within a bgnloop/endloop.
1694 Support for CONT is determined by a special capability bit,
1695 ``TGSI_CONT_SUPPORTED``. See :ref:`Screen` for more information.
1698 .. opcode:: BGNLOOP - Begin a Loop
1700 Start a loop. Must have a matching endloop.
1703 .. opcode:: BGNSUB - Begin Subroutine
1705 Starts definition of a subroutine. Must have a matching endsub.
1708 .. opcode:: ENDLOOP - End a Loop
1710 End a loop started with bgnloop.
1713 .. opcode:: ENDSUB - End Subroutine
1715 Ends definition of a subroutine.
1718 .. opcode:: NOP - No Operation
1723 .. opcode:: BRK - Break
1725 Unconditionally moves the point of execution to the instruction after the
1726 next endloop or endswitch. The instruction must appear within a loop/endloop
1727 or switch/endswitch.
1730 .. opcode:: BREAKC - Break Conditional
1732 Conditionally moves the point of execution to the instruction after the
1733 next endloop or endswitch. The instruction must appear within a loop/endloop
1734 or switch/endswitch.
1735 Condition evaluates to true if src0.x != 0 where src0.x is interpreted
1736 as an integer register.
1740 Considered for removal as it's quite inconsistent wrt other opcodes
1741 (could emulate with UIF/BRK/ENDIF).
1744 .. opcode:: IF - Float If
1746 Start an IF ... ELSE .. ENDIF block. Condition evaluates to true if
1750 where src0.x is interpreted as a floating point register.
1753 .. opcode:: UIF - Bitwise If
1755 Start an UIF ... ELSE .. ENDIF block. Condition evaluates to true if
1759 where src0.x is interpreted as an integer register.
1762 .. opcode:: ELSE - Else
1764 Starts an else block, after an IF or UIF statement.
1767 .. opcode:: ENDIF - End If
1769 Ends an IF or UIF block.
1772 .. opcode:: SWITCH - Switch
1774 Starts a C-style switch expression. The switch consists of one or multiple
1775 CASE statements, and at most one DEFAULT statement. Execution of a statement
1776 ends when a BRK is hit, but just like in C falling through to other cases
1777 without a break is allowed. Similarly, DEFAULT label is allowed anywhere not
1778 just as last statement, and fallthrough is allowed into/from it.
1779 CASE src arguments are evaluated at bit level against the SWITCH src argument.
1785 (some instructions here)
1788 (some instructions here)
1791 (some instructions here)
1796 .. opcode:: CASE - Switch case
1798 This represents a switch case label. The src arg must be an integer immediate.
1801 .. opcode:: DEFAULT - Switch default
1803 This represents the default case in the switch, which is taken if no other
1807 .. opcode:: ENDSWITCH - End of switch
1809 Ends a switch expression.
1815 The interpolation instructions allow an input to be interpolated in a
1816 different way than its declaration. This corresponds to the GLSL 4.00
1817 interpolateAt* functions. The first argument of each of these must come from
1818 ``TGSI_FILE_INPUT``.
1820 .. opcode:: INTERP_CENTROID - Interpolate at the centroid
1822 Interpolates the varying specified by src0 at the centroid
1824 .. opcode:: INTERP_SAMPLE - Interpolate at the specified sample
1826 Interpolates the varying specified by src0 at the sample id specified by
1827 src1.x (interpreted as an integer)
1829 .. opcode:: INTERP_OFFSET - Interpolate at the specified offset
1831 Interpolates the varying specified by src0 at the offset src1.xy from the
1832 pixel center (interpreted as floats)
1840 The double-precision opcodes reinterpret four-component vectors into
1841 two-component vectors with doubled precision in each component.
1843 .. opcode:: DABS - Absolute
1848 .. opcode:: DADD - Add
1852 dst.xy = src0.xy + src1.xy
1854 dst.zw = src0.zw + src1.zw
1856 .. opcode:: DSEQ - Set on Equal
1860 dst.x = src0.xy == src1.xy ? \sim 0 : 0
1862 dst.z = src0.zw == src1.zw ? \sim 0 : 0
1864 .. opcode:: DSNE - Set on Equal
1868 dst.x = src0.xy != src1.xy ? \sim 0 : 0
1870 dst.z = src0.zw != src1.zw ? \sim 0 : 0
1872 .. opcode:: DSLT - Set on Less than
1876 dst.x = src0.xy < src1.xy ? \sim 0 : 0
1878 dst.z = src0.zw < src1.zw ? \sim 0 : 0
1880 .. opcode:: DSGE - Set on Greater equal
1884 dst.x = src0.xy >= src1.xy ? \sim 0 : 0
1886 dst.z = src0.zw >= src1.zw ? \sim 0 : 0
1888 .. opcode:: DFRAC - Fraction
1892 dst.xy = src.xy - \lfloor src.xy\rfloor
1894 dst.zw = src.zw - \lfloor src.zw\rfloor
1896 .. opcode:: DTRUNC - Truncate
1900 dst.xy = trunc(src.xy)
1902 dst.zw = trunc(src.zw)
1904 .. opcode:: DCEIL - Ceiling
1908 dst.xy = \lceil src.xy\rceil
1910 dst.zw = \lceil src.zw\rceil
1912 .. opcode:: DFLR - Floor
1916 dst.xy = \lfloor src.xy\rfloor
1918 dst.zw = \lfloor src.zw\rfloor
1920 .. opcode:: DROUND - Fraction
1924 dst.xy = round(src.xy)
1926 dst.zw = round(src.zw)
1928 .. opcode:: DSSG - Set Sign
1932 dst.xy = (src.xy > 0) ? 1.0 : (src.xy < 0) ? -1.0 : 0.0
1934 dst.zw = (src.zw > 0) ? 1.0 : (src.zw < 0) ? -1.0 : 0.0
1936 .. opcode:: DFRACEXP - Convert Number to Fractional and Integral Components
1938 Like the ``frexp()`` routine in many math libraries, this opcode stores the
1939 exponent of its source to ``dst0``, and the significand to ``dst1``, such that
1940 :math:`dst1 \times 2^{dst0} = src` .
1944 dst0.xy = exp(src.xy)
1946 dst1.xy = frac(src.xy)
1948 dst0.zw = exp(src.zw)
1950 dst1.zw = frac(src.zw)
1952 .. opcode:: DLDEXP - Multiply Number by Integral Power of 2
1954 This opcode is the inverse of :opcode:`DFRACEXP`. The second
1955 source is an integer.
1959 dst.xy = src0.xy \times 2^{src1.x}
1961 dst.zw = src0.zw \times 2^{src1.y}
1963 .. opcode:: DMIN - Minimum
1967 dst.xy = min(src0.xy, src1.xy)
1969 dst.zw = min(src0.zw, src1.zw)
1971 .. opcode:: DMAX - Maximum
1975 dst.xy = max(src0.xy, src1.xy)
1977 dst.zw = max(src0.zw, src1.zw)
1979 .. opcode:: DMUL - Multiply
1983 dst.xy = src0.xy \times src1.xy
1985 dst.zw = src0.zw \times src1.zw
1988 .. opcode:: DMAD - Multiply And Add
1992 dst.xy = src0.xy \times src1.xy + src2.xy
1994 dst.zw = src0.zw \times src1.zw + src2.zw
1997 .. opcode:: DFMA - Fused Multiply-Add
1999 Perform a * b + c with no intermediate rounding step.
2003 dst.xy = src0.xy \times src1.xy + src2.xy
2005 dst.zw = src0.zw \times src1.zw + src2.zw
2008 .. opcode:: DDIV - Divide
2012 dst.xy = \frac{src0.xy}{src1.xy}
2014 dst.zw = \frac{src0.zw}{src1.zw}
2017 .. opcode:: DRCP - Reciprocal
2021 dst.xy = \frac{1}{src.xy}
2023 dst.zw = \frac{1}{src.zw}
2025 .. opcode:: DSQRT - Square Root
2029 dst.xy = \sqrt{src.xy}
2031 dst.zw = \sqrt{src.zw}
2033 .. opcode:: DRSQ - Reciprocal Square Root
2037 dst.xy = \frac{1}{\sqrt{src.xy}}
2039 dst.zw = \frac{1}{\sqrt{src.zw}}
2041 .. opcode:: F2D - Float to Double
2045 dst.xy = double(src0.x)
2047 dst.zw = double(src0.y)
2049 .. opcode:: D2F - Double to Float
2053 dst.x = float(src0.xy)
2055 dst.y = float(src0.zw)
2057 .. opcode:: I2D - Int to Double
2061 dst.xy = double(src0.x)
2063 dst.zw = double(src0.y)
2065 .. opcode:: D2I - Double to Int
2069 dst.x = int(src0.xy)
2071 dst.y = int(src0.zw)
2073 .. opcode:: U2D - Unsigned Int to Double
2077 dst.xy = double(src0.x)
2079 dst.zw = double(src0.y)
2081 .. opcode:: D2U - Double to Unsigned Int
2085 dst.x = unsigned(src0.xy)
2087 dst.y = unsigned(src0.zw)
2092 The 64-bit integer opcodes reinterpret four-component vectors into
2093 two-component vectors with 64-bits in each component.
2095 .. opcode:: I64ABS - 64-bit Integer Absolute Value
2100 .. opcode:: I64NEG - 64-bit Integer Negate
2109 .. opcode:: I64SSG - 64-bit Integer Set Sign
2113 dst.xy = (src0.xy < 0) ? -1 : (src0.xy > 0) ? 1 : 0
2114 dst.zw = (src0.zw < 0) ? -1 : (src0.zw > 0) ? 1 : 0
2116 .. opcode:: U64ADD - 64-bit Integer Add
2120 dst.xy = src0.xy + src1.xy
2121 dst.zw = src0.zw + src1.zw
2123 .. opcode:: U64MUL - 64-bit Integer Multiply
2127 dst.xy = src0.xy * src1.xy
2128 dst.zw = src0.zw * src1.zw
2130 .. opcode:: U64SEQ - 64-bit Integer Set on Equal
2134 dst.x = src0.xy == src1.xy ? \sim 0 : 0
2135 dst.z = src0.zw == src1.zw ? \sim 0 : 0
2137 .. opcode:: U64SNE - 64-bit Integer Set on Not Equal
2141 dst.x = src0.xy != src1.xy ? \sim 0 : 0
2142 dst.z = src0.zw != src1.zw ? \sim 0 : 0
2144 .. opcode:: U64SLT - 64-bit Unsigned Integer Set on Less Than
2148 dst.x = src0.xy < src1.xy ? \sim 0 : 0
2149 dst.z = src0.zw < src1.zw ? \sim 0 : 0
2151 .. opcode:: U64SGE - 64-bit Unsigned Integer Set on Greater Equal
2155 dst.x = src0.xy >= src1.xy ? \sim 0 : 0
2156 dst.z = src0.zw >= src1.zw ? \sim 0 : 0
2158 .. opcode:: I64SLT - 64-bit Signed Integer Set on Less Than
2162 dst.x = src0.xy < src1.xy ? \sim 0 : 0
2163 dst.z = src0.zw < src1.zw ? \sim 0 : 0
2165 .. opcode:: I64SGE - 64-bit Signed Integer Set on Greater Equal
2169 dst.x = src0.xy >= src1.xy ? \sim 0 : 0
2170 dst.z = src0.zw >= src1.zw ? \sim 0 : 0
2172 .. opcode:: I64MIN - Minimum of 64-bit Signed Integers
2176 dst.xy = min(src0.xy, src1.xy)
2177 dst.zw = min(src0.zw, src1.zw)
2179 .. opcode:: U64MIN - Minimum of 64-bit Unsigned Integers
2183 dst.xy = min(src0.xy, src1.xy)
2184 dst.zw = min(src0.zw, src1.zw)
2186 .. opcode:: I64MAX - Maximum of 64-bit Signed Integers
2190 dst.xy = max(src0.xy, src1.xy)
2191 dst.zw = max(src0.zw, src1.zw)
2193 .. opcode:: U64MAX - Maximum of 64-bit Unsigned Integers
2197 dst.xy = max(src0.xy, src1.xy)
2198 dst.zw = max(src0.zw, src1.zw)
2200 .. opcode:: U64SHL - Shift Left 64-bit Unsigned Integer
2202 The shift count is masked with 0x3f before the shift is applied.
2206 dst.xy = src0.xy << (0x3f \& src1.x)
2207 dst.zw = src0.zw << (0x3f \& src1.y)
2209 .. opcode:: I64SHR - Arithmetic Shift Right (of 64-bit Signed Integer)
2211 The shift count is masked with 0x3f before the shift is applied.
2215 dst.xy = src0.xy >> (0x3f \& src1.x)
2216 dst.zw = src0.zw >> (0x3f \& src1.y)
2218 .. opcode:: U64SHR - Logical Shift Right (of 64-bit Unsigned Integer)
2220 The shift count is masked with 0x3f before the shift is applied.
2224 dst.xy = src0.xy >> (unsigned) (0x3f \& src1.x)
2225 dst.zw = src0.zw >> (unsigned) (0x3f \& src1.y)
2227 .. opcode:: I64DIV - 64-bit Signed Integer Division
2231 dst.xy = src0.xy \ src1.xy
2232 dst.zw = src0.zw \ src1.zw
2234 .. opcode:: U64DIV - 64-bit Unsigned Integer Division
2238 dst.xy = src0.xy \ src1.xy
2239 dst.zw = src0.zw \ src1.zw
2241 .. opcode:: U64MOD - 64-bit Unsigned Integer Remainder
2245 dst.xy = src0.xy \bmod src1.xy
2246 dst.zw = src0.zw \bmod src1.zw
2248 .. opcode:: I64MOD - 64-bit Signed Integer Remainder
2252 dst.xy = src0.xy \bmod src1.xy
2253 dst.zw = src0.zw \bmod src1.zw
2255 .. opcode:: F2U64 - Float to 64-bit Unsigned Int
2259 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
2267 dst.zw = (int64_t) src0.y
2269 .. opcode:: U2I64 - Unsigned Integer to 64-bit Integer
2271 This is a zero extension.
2275 dst.xy = (uint64_t) src0.x
2276 dst.zw = (uint64_t) src0.y
2278 .. opcode:: I2I64 - Signed Integer to 64-bit Integer
2280 This is a sign extension.
2284 dst.xy = (int64_t) src0.x
2285 dst.zw = (int64_t) src0.y
2287 .. opcode:: D2U64 - Double to 64-bit Unsigned Int
2291 dst.xy = (uint64_t) src0.xy
2292 dst.zw = (uint64_t) src0.zw
2294 .. opcode:: D2I64 - Double to 64-bit Int
2298 dst.xy = (int64_t) src0.xy
2299 dst.zw = (int64_t) src0.zw
2301 .. opcode:: U642F - 64-bit unsigned integer to float
2305 dst.x = (float) src0.xy
2306 dst.y = (float) src0.zw
2308 .. opcode:: I642F - 64-bit Int to Float
2312 dst.x = (float) src0.xy
2313 dst.y = (float) src0.zw
2315 .. opcode:: U642D - 64-bit unsigned integer to double
2319 dst.xy = (double) src0.xy
2320 dst.zw = (double) src0.zw
2322 .. opcode:: I642D - 64-bit Int to double
2326 dst.xy = (double) src0.xy
2327 dst.zw = (double) src0.zw
2329 .. _samplingopcodes:
2331 Resource Sampling Opcodes
2332 ^^^^^^^^^^^^^^^^^^^^^^^^^
2334 Those opcodes follow very closely semantics of the respective Direct3D
2335 instructions. If in doubt double check Direct3D documentation.
2336 Note that the swizzle on SVIEW (src1) determines texel swizzling
2341 Using provided address, sample data from the specified texture using the
2342 filtering mode identified by the given sampler. The source data may come from
2343 any resource type other than buffers.
2345 Syntax: ``SAMPLE dst, address, sampler_view, sampler``
2347 Example: ``SAMPLE TEMP[0], TEMP[1], SVIEW[0], SAMP[0]``
2349 .. opcode:: SAMPLE_I
2351 Simplified alternative to the SAMPLE instruction. Using the provided
2352 integer address, SAMPLE_I fetches data from the specified sampler view
2353 without any filtering. The source data may come from any resource type
2356 Syntax: ``SAMPLE_I dst, address, sampler_view``
2358 Example: ``SAMPLE_I TEMP[0], TEMP[1], SVIEW[0]``
2360 The 'address' is specified as unsigned integers. If the 'address' is out of
2361 range [0...(# texels - 1)] the result of the fetch is always 0 in all
2362 components. As such the instruction doesn't honor address wrap modes, in
2363 cases where that behavior is desirable 'SAMPLE' instruction should be used.
2364 address.w always provides an unsigned integer mipmap level. If the value is
2365 out of the range then the instruction always returns 0 in all components.
2366 address.yz are ignored for buffers and 1d textures. address.z is ignored
2367 for 1d texture arrays and 2d textures.
2369 For 1D texture arrays address.y provides the array index (also as unsigned
2370 integer). If the value is out of the range of available array indices
2371 [0... (array size - 1)] then the opcode always returns 0 in all components.
2372 For 2D texture arrays address.z provides the array index, otherwise it
2373 exhibits the same behavior as in the case for 1D texture arrays. The exact
2374 semantics of the source address are presented in the table below:
2376 +---------------------------+----+-----+-----+---------+
2377 | resource type | X | Y | Z | W |
2378 +===========================+====+=====+=====+=========+
2379 | ``PIPE_BUFFER`` | x | | | ignored |
2380 +---------------------------+----+-----+-----+---------+
2381 | ``PIPE_TEXTURE_1D`` | x | | | mpl |
2382 +---------------------------+----+-----+-----+---------+
2383 | ``PIPE_TEXTURE_2D`` | x | y | | mpl |
2384 +---------------------------+----+-----+-----+---------+
2385 | ``PIPE_TEXTURE_3D`` | x | y | z | mpl |
2386 +---------------------------+----+-----+-----+---------+
2387 | ``PIPE_TEXTURE_RECT`` | x | y | | mpl |
2388 +---------------------------+----+-----+-----+---------+
2389 | ``PIPE_TEXTURE_CUBE`` | not allowed as source |
2390 +---------------------------+----+-----+-----+---------+
2391 | ``PIPE_TEXTURE_1D_ARRAY`` | x | idx | | mpl |
2392 +---------------------------+----+-----+-----+---------+
2393 | ``PIPE_TEXTURE_2D_ARRAY`` | x | y | idx | mpl |
2394 +---------------------------+----+-----+-----+---------+
2396 Where 'mpl' is a mipmap level and 'idx' is the array index.
2398 .. opcode:: SAMPLE_I_MS
2400 Just like SAMPLE_I but allows fetch data from multi-sampled surfaces.
2402 Syntax: ``SAMPLE_I_MS dst, address, sampler_view, sample``
2404 .. opcode:: SAMPLE_B
2406 Just like the SAMPLE instruction with the exception that an additional bias
2407 is applied to the level of detail computed as part of the instruction
2410 Syntax: ``SAMPLE_B dst, address, sampler_view, sampler, lod_bias``
2412 Example: ``SAMPLE_B TEMP[0], TEMP[1], SVIEW[0], SAMP[0], TEMP[2].x``
2414 .. opcode:: SAMPLE_C
2416 Similar to the SAMPLE instruction but it performs a comparison filter. The
2417 operands to SAMPLE_C are identical to SAMPLE, except that there is an
2418 additional float32 operand, reference value, which must be a register with
2419 single-component, or a scalar literal. SAMPLE_C makes the hardware use the
2420 current samplers compare_func (in pipe_sampler_state) to compare reference
2421 value against the red component value for the surce resource at each texel
2422 that the currently configured texture filter covers based on the provided
2425 Syntax: ``SAMPLE_C dst, address, sampler_view.r, sampler, ref_value``
2427 Example: ``SAMPLE_C TEMP[0], TEMP[1], SVIEW[0].r, SAMP[0], TEMP[2].x``
2429 .. opcode:: SAMPLE_C_LZ
2431 Same as SAMPLE_C, but LOD is 0 and derivatives are ignored. The LZ stands
2434 Syntax: ``SAMPLE_C_LZ dst, address, sampler_view.r, sampler, ref_value``
2436 Example: ``SAMPLE_C_LZ TEMP[0], TEMP[1], SVIEW[0].r, SAMP[0], TEMP[2].x``
2439 .. opcode:: SAMPLE_D
2441 SAMPLE_D is identical to the SAMPLE opcode except that the derivatives for
2442 the source address in the x direction and the y direction are provided by
2445 Syntax: ``SAMPLE_D dst, address, sampler_view, sampler, der_x, der_y``
2447 Example: ``SAMPLE_D TEMP[0], TEMP[1], SVIEW[0], SAMP[0], TEMP[2], TEMP[3]``
2449 .. opcode:: SAMPLE_L
2451 SAMPLE_L is identical to the SAMPLE opcode except that the LOD is provided
2452 directly as a scalar value, representing no anisotropy.
2454 Syntax: ``SAMPLE_L dst, address, sampler_view, sampler, explicit_lod``
2456 Example: ``SAMPLE_L TEMP[0], TEMP[1], SVIEW[0], SAMP[0], TEMP[2].x``
2460 Gathers the four texels to be used in a bi-linear filtering operation and
2461 packs them into a single register. Only works with 2D, 2D array, cubemaps,
2462 and cubemaps arrays. For 2D textures, only the addressing modes of the
2463 sampler and the top level of any mip pyramid are used. Set W to zero. It
2464 behaves like the SAMPLE instruction, but a filtered sample is not
2465 generated. The four samples that contribute to filtering are placed into
2466 xyzw in counter-clockwise order, starting with the (u,v) texture coordinate
2467 delta at the following locations (-, +), (+, +), (+, -), (-, -), where the
2468 magnitude of the deltas are half a texel.
2471 .. opcode:: SVIEWINFO
2473 Query the dimensions of a given sampler view. dst receives width, height,
2474 depth or array size and number of mipmap levels as int4. The dst can have a
2475 writemask which will specify what info is the caller interested in.
2477 Syntax: ``SVIEWINFO dst, src_mip_level, sampler_view``
2479 Example: ``SVIEWINFO TEMP[0], TEMP[1].x, SVIEW[0]``
2481 src_mip_level is an unsigned integer scalar. If it's out of range then
2482 returns 0 for width, height and depth/array size but the total number of
2483 mipmap is still returned correctly for the given sampler view. The returned
2484 width, height and depth values are for the mipmap level selected by the
2485 src_mip_level and are in the number of texels. For 1d texture array width
2486 is in dst.x, array size is in dst.y and dst.z is 0. The number of mipmaps is
2487 still in dst.w. In contrast to d3d10 resinfo, there's no way in the tgsi
2488 instruction encoding to specify the return type (float/rcpfloat/uint), hence
2489 always using uint. Also, unlike the SAMPLE instructions, the swizzle on src1
2490 resinfo allowing swizzling dst values is ignored (due to the interaction
2491 with rcpfloat modifier which requires some swizzle handling in the state
2494 .. opcode:: SAMPLE_POS
2496 Query the position of a given sample. dst receives float4 (x, y, 0, 0)
2497 indicated where the sample is located. If the resource is not a multi-sample
2498 resource and not a render target, the result is 0.
2500 .. opcode:: SAMPLE_INFO
2502 dst receives number of samples in x. If the resource is not a multi-sample
2503 resource and not a render target, the result is 0.
2506 .. _resourceopcodes:
2508 Resource Access Opcodes
2509 ^^^^^^^^^^^^^^^^^^^^^^^
2511 .. opcode:: LOAD - Fetch data from a shader buffer or image
2513 Syntax: ``LOAD dst, resource, address``
2515 Example: ``LOAD TEMP[0], BUFFER[0], TEMP[1]``
2517 Using the provided integer address, LOAD fetches data
2518 from the specified buffer or texture without any
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 read
2526 from 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 A swizzle suffix may be added to the resource argument
2536 this will cause the resource data to be swizzled accordingly.
2538 .. opcode:: STORE - Write data to a shader resource
2540 Syntax: ``STORE resource, address, src``
2542 Example: ``STORE BUFFER[0], TEMP[0], TEMP[1]``
2544 Using the provided integer address, STORE writes data
2545 to the specified buffer or texture.
2547 The 'address' is specified as a vector of unsigned
2548 integers. If the 'address' is out of range the result
2551 Only the first mipmap level of a resource can be
2552 written to using this instruction.
2554 For 1D or 2D texture arrays, the array index is
2555 provided as an unsigned integer in address.y or
2556 address.z, respectively. address.yz are ignored for
2557 buffers and 1D textures. address.z is ignored for 1D
2558 texture arrays and 2D textures. address.w is always
2561 .. opcode:: RESQ - Query information about a resource
2563 Syntax: ``RESQ dst, resource``
2565 Example: ``RESQ TEMP[0], BUFFER[0]``
2567 Returns information about the buffer or image resource. For buffer
2568 resources, the size (in bytes) is returned in the x component. For
2569 image resources, .xyz will contain the width/height/layers of the
2570 image, while .w will contain the number of samples for multi-sampled
2573 .. opcode:: FBFETCH - Load data from framebuffer
2575 Syntax: ``FBFETCH dst, output``
2577 Example: ``FBFETCH TEMP[0], OUT[0]``
2579 This is only valid on ``COLOR`` semantic outputs. Returns the color
2580 of the current position in the framebuffer from before this fragment
2581 shader invocation. May return the same value from multiple calls for
2582 a particular output within a single invocation. Note that result may
2583 be undefined if a fragment is drawn multiple times without a blend
2587 .. _threadsyncopcodes:
2589 Inter-thread synchronization opcodes
2590 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
2592 These opcodes are intended for communication between threads running
2593 within the same compute grid. For now they're only valid in compute
2596 .. opcode:: MFENCE - Memory fence
2598 Syntax: ``MFENCE resource``
2600 Example: ``MFENCE RES[0]``
2602 This opcode forces strong ordering between any memory access
2603 operations that affect the specified resource. This means that
2604 previous loads and stores (and only those) will be performed and
2605 visible to other threads before the program execution continues.
2608 .. opcode:: LFENCE - Load memory fence
2610 Syntax: ``LFENCE resource``
2612 Example: ``LFENCE RES[0]``
2614 Similar to MFENCE, but it only affects the ordering of memory loads.
2617 .. opcode:: SFENCE - Store memory fence
2619 Syntax: ``SFENCE resource``
2621 Example: ``SFENCE RES[0]``
2623 Similar to MFENCE, but it only affects the ordering of memory stores.
2626 .. opcode:: BARRIER - Thread group barrier
2630 This opcode suspends the execution of the current thread until all
2631 the remaining threads in the working group reach the same point of
2632 the program. Results are unspecified if any of the remaining
2633 threads terminates or never reaches an executed BARRIER instruction.
2635 .. opcode:: MEMBAR - Memory barrier
2639 This opcode waits for the completion of all memory accesses based on
2640 the type passed in. The type is an immediate bitfield with the following
2643 Bit 0: Shader storage buffers
2644 Bit 1: Atomic buffers
2646 Bit 3: Shared memory
2649 These may be passed in in any combination. An implementation is free to not
2650 distinguish between these as it sees fit. However these map to all the
2651 possibilities made available by GLSL.
2658 These opcodes provide atomic variants of some common arithmetic and
2659 logical operations. In this context atomicity means that another
2660 concurrent memory access operation that affects the same memory
2661 location is guaranteed to be performed strictly before or after the
2662 entire execution of the atomic operation. The resource may be a buffer
2663 or an image. In the case of an image, the offset works the same as for
2664 ``LOAD`` and ``STORE``, specified above. These atomic operations may
2665 only be used with 32-bit integer image formats.
2667 .. opcode:: ATOMUADD - Atomic integer addition
2669 Syntax: ``ATOMUADD dst, resource, offset, src``
2671 Example: ``ATOMUADD TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2673 The following operation is performed atomically:
2677 dst_x = resource[offset]
2679 resource[offset] = dst_x + src_x
2682 .. opcode:: ATOMXCHG - Atomic exchange
2684 Syntax: ``ATOMXCHG dst, resource, offset, src``
2686 Example: ``ATOMXCHG TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2688 The following operation is performed atomically:
2692 dst_x = resource[offset]
2694 resource[offset] = src_x
2697 .. opcode:: ATOMCAS - Atomic compare-and-exchange
2699 Syntax: ``ATOMCAS dst, resource, offset, cmp, src``
2701 Example: ``ATOMCAS TEMP[0], BUFFER[0], TEMP[1], TEMP[2], TEMP[3]``
2703 The following operation is performed atomically:
2707 dst_x = resource[offset]
2709 resource[offset] = (dst_x == cmp_x ? src_x : dst_x)
2712 .. opcode:: ATOMAND - Atomic bitwise And
2714 Syntax: ``ATOMAND dst, resource, offset, src``
2716 Example: ``ATOMAND TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2718 The following operation is performed atomically:
2722 dst_x = resource[offset]
2724 resource[offset] = dst_x \& src_x
2727 .. opcode:: ATOMOR - Atomic bitwise Or
2729 Syntax: ``ATOMOR dst, resource, offset, src``
2731 Example: ``ATOMOR TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2733 The following operation is performed atomically:
2737 dst_x = resource[offset]
2739 resource[offset] = dst_x | src_x
2742 .. opcode:: ATOMXOR - Atomic bitwise Xor
2744 Syntax: ``ATOMXOR dst, resource, offset, src``
2746 Example: ``ATOMXOR TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2748 The following operation is performed atomically:
2752 dst_x = resource[offset]
2754 resource[offset] = dst_x \oplus src_x
2757 .. opcode:: ATOMUMIN - Atomic unsigned minimum
2759 Syntax: ``ATOMUMIN dst, resource, offset, src``
2761 Example: ``ATOMUMIN TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2763 The following operation is performed atomically:
2767 dst_x = resource[offset]
2769 resource[offset] = (dst_x < src_x ? dst_x : src_x)
2772 .. opcode:: ATOMUMAX - Atomic unsigned maximum
2774 Syntax: ``ATOMUMAX dst, resource, offset, src``
2776 Example: ``ATOMUMAX TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2778 The following operation is performed atomically:
2782 dst_x = resource[offset]
2784 resource[offset] = (dst_x > src_x ? dst_x : src_x)
2787 .. opcode:: ATOMIMIN - Atomic signed minimum
2789 Syntax: ``ATOMIMIN dst, resource, offset, src``
2791 Example: ``ATOMIMIN TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2793 The following operation is performed atomically:
2797 dst_x = resource[offset]
2799 resource[offset] = (dst_x < src_x ? dst_x : src_x)
2802 .. opcode:: ATOMIMAX - Atomic signed maximum
2804 Syntax: ``ATOMIMAX dst, resource, offset, src``
2806 Example: ``ATOMIMAX TEMP[0], BUFFER[0], TEMP[1], TEMP[2]``
2808 The following operation is performed atomically:
2812 dst_x = resource[offset]
2814 resource[offset] = (dst_x > src_x ? dst_x : src_x)
2822 These opcodes compare the given value across the shader invocations
2823 running in the current SIMD group. The details of exactly which
2824 invocations get compared are implementation-defined, and it would be a
2825 correct implementation to only ever consider the current thread's
2826 value. (i.e. SIMD group of 1). The argument is treated as a boolean.
2828 .. opcode:: VOTE_ANY - Value is set in any of the current invocations
2830 .. opcode:: VOTE_ALL - Value is set in all of the current invocations
2832 .. opcode:: VOTE_EQ - Value is the same in all of the current invocations
2835 Explanation of symbols used
2836 ------------------------------
2843 :math:`|x|` Absolute value of `x`.
2845 :math:`\lceil x \rceil` Ceiling of `x`.
2847 clamp(x,y,z) Clamp x between y and z.
2848 (x < y) ? y : (x > z) ? z : x
2850 :math:`\lfloor x\rfloor` Floor of `x`.
2852 :math:`\log_2{x}` Logarithm of `x`, base 2.
2854 max(x,y) Maximum of x and y.
2857 min(x,y) Minimum of x and y.
2860 partialx(x) Derivative of x relative to fragment's X.
2862 partialy(x) Derivative of x relative to fragment's Y.
2864 pop() Pop from stack.
2866 :math:`x^y` `x` to the power `y`.
2868 push(x) Push x on stack.
2872 trunc(x) Truncate x, i.e. drop the fraction bits.
2879 discard Discard fragment.
2883 target Label of target instruction.
2894 Declares a register that is will be referenced as an operand in Instruction
2897 File field contains register file that is being declared and is one
2900 UsageMask field specifies which of the register components can be accessed
2901 and is one of TGSI_WRITEMASK.
2903 The Local flag specifies that a given value isn't intended for
2904 subroutine parameter passing and, as a result, the implementation
2905 isn't required to give any guarantees of it being preserved across
2906 subroutine boundaries. As it's merely a compiler hint, the
2907 implementation is free to ignore it.
2909 If Dimension flag is set to 1, a Declaration Dimension token follows.
2911 If Semantic flag is set to 1, a Declaration Semantic token follows.
2913 If Interpolate flag is set to 1, a Declaration Interpolate token follows.
2915 If file is TGSI_FILE_RESOURCE, a Declaration Resource token follows.
2917 If Array flag is set to 1, a Declaration Array token follows.
2920 ^^^^^^^^^^^^^^^^^^^^^^^^
2922 Declarations can optional have an ArrayID attribute which can be referred by
2923 indirect addressing operands. An ArrayID of zero is reserved and treated as
2924 if no ArrayID is specified.
2926 If an indirect addressing operand refers to a specific declaration by using
2927 an ArrayID only the registers in this declaration are guaranteed to be
2928 accessed, accessing any register outside this declaration results in undefined
2929 behavior. Note that for compatibility the effective index is zero-based and
2930 not relative to the specified declaration
2932 If no ArrayID is specified with an indirect addressing operand the whole
2933 register file might be accessed by this operand. This is strongly discouraged
2934 and will prevent packing of scalar/vec2 arrays and effective alias analysis.
2935 This is only legal for TEMP and CONST register files.
2937 Declaration Semantic
2938 ^^^^^^^^^^^^^^^^^^^^^^^^
2940 Vertex and fragment shader input and output registers may be labeled
2941 with semantic information consisting of a name and index.
2943 Follows Declaration token if Semantic bit is set.
2945 Since its purpose is to link a shader with other stages of the pipeline,
2946 it is valid to follow only those Declaration tokens that declare a register
2947 either in INPUT or OUTPUT file.
2949 SemanticName field contains the semantic name of the register being declared.
2950 There is no default value.
2952 SemanticIndex is an optional subscript that can be used to distinguish
2953 different register declarations with the same semantic name. The default value
2956 The meanings of the individual semantic names are explained in the following
2959 TGSI_SEMANTIC_POSITION
2960 """"""""""""""""""""""
2962 For vertex shaders, TGSI_SEMANTIC_POSITION indicates the vertex shader
2963 output register which contains the homogeneous vertex position in the clip
2964 space coordinate system. After clipping, the X, Y and Z components of the
2965 vertex will be divided by the W value to get normalized device coordinates.
2967 For fragment shaders, TGSI_SEMANTIC_POSITION is used to indicate that
2968 fragment shader input (or system value, depending on which one is
2969 supported by the driver) contains the fragment's window position. The X
2970 component starts at zero and always increases from left to right.
2971 The Y component starts at zero and always increases but Y=0 may either
2972 indicate the top of the window or the bottom depending on the fragment
2973 coordinate origin convention (see TGSI_PROPERTY_FS_COORD_ORIGIN).
2974 The Z coordinate ranges from 0 to 1 to represent depth from the front
2975 to the back of the Z buffer. The W component contains the interpolated
2976 reciprocal of the vertex position W component (corresponding to gl_Fragcoord,
2977 but unlike d3d10 which interpolates the same 1/w but then gives back
2978 the reciprocal of the interpolated value).
2980 Fragment shaders may also declare an output register with
2981 TGSI_SEMANTIC_POSITION. Only the Z component is writable. This allows
2982 the fragment shader to change the fragment's Z position.
2989 For vertex shader outputs or fragment shader inputs/outputs, this
2990 label indicates that the register contains an R,G,B,A color.
2992 Several shader inputs/outputs may contain colors so the semantic index
2993 is used to distinguish them. For example, color[0] may be the diffuse
2994 color while color[1] may be the specular color.
2996 This label is needed so that the flat/smooth shading can be applied
2997 to the right interpolants during rasterization.
3001 TGSI_SEMANTIC_BCOLOR
3002 """"""""""""""""""""
3004 Back-facing colors are only used for back-facing polygons, and are only valid
3005 in vertex shader outputs. After rasterization, all polygons are front-facing
3006 and COLOR and BCOLOR end up occupying the same slots in the fragment shader,
3007 so all BCOLORs effectively become regular COLORs in the fragment shader.
3013 Vertex shader inputs and outputs and fragment shader inputs may be
3014 labeled with TGSI_SEMANTIC_FOG to indicate that the register contains
3015 a fog coordinate. Typically, the fragment shader will use the fog coordinate
3016 to compute a fog blend factor which is used to blend the normal fragment color
3017 with a constant fog color. But fog coord really is just an ordinary vec4
3018 register like regular semantics.
3024 Vertex shader input and output registers may be labeled with
3025 TGIS_SEMANTIC_PSIZE to indicate that the register contains a point size
3026 in the form (S, 0, 0, 1). The point size controls the width or diameter
3027 of points for rasterization. This label cannot be used in fragment
3030 When using this semantic, be sure to set the appropriate state in the
3031 :ref:`rasterizer` first.
3034 TGSI_SEMANTIC_TEXCOORD
3035 """"""""""""""""""""""
3037 Only available if PIPE_CAP_TGSI_TEXCOORD is exposed !
3039 Vertex shader outputs and fragment shader inputs may be labeled with
3040 this semantic to make them replaceable by sprite coordinates via the
3041 sprite_coord_enable state in the :ref:`rasterizer`.
3042 The semantic index permitted with this semantic is limited to <= 7.
3044 If the driver does not support TEXCOORD, sprite coordinate replacement
3045 applies to inputs with the GENERIC semantic instead.
3047 The intended use case for this semantic is gl_TexCoord.
3050 TGSI_SEMANTIC_PCOORD
3051 """"""""""""""""""""
3053 Only available if PIPE_CAP_TGSI_TEXCOORD is exposed !
3055 Fragment shader inputs may be labeled with TGSI_SEMANTIC_PCOORD to indicate
3056 that the register contains sprite coordinates in the form (x, y, 0, 1), if
3057 the current primitive is a point and point sprites are enabled. Otherwise,
3058 the contents of the register are undefined.
3060 The intended use case for this semantic is gl_PointCoord.
3063 TGSI_SEMANTIC_GENERIC
3064 """""""""""""""""""""
3066 All vertex/fragment shader inputs/outputs not labeled with any other
3067 semantic label can be considered to be generic attributes. Typical
3068 uses of generic inputs/outputs are texcoords and user-defined values.
3071 TGSI_SEMANTIC_NORMAL
3072 """"""""""""""""""""
3074 Indicates that a vertex shader input is a normal vector. This is
3075 typically only used for legacy graphics APIs.
3081 This label applies to fragment shader inputs (or system values,
3082 depending on which one is supported by the driver) and indicates that
3083 the register contains front/back-face information.
3085 If it is an input, it will be a floating-point vector in the form (F, 0, 0, 1),
3086 where F will be positive when the fragment belongs to a front-facing polygon,
3087 and negative when the fragment belongs to a back-facing polygon.
3089 If it is a system value, it will be an integer vector in the form (F, 0, 0, 1),
3090 where F is 0xffffffff when the fragment belongs to a front-facing polygon and
3091 0 when the fragment belongs to a back-facing polygon.
3094 TGSI_SEMANTIC_EDGEFLAG
3095 """"""""""""""""""""""
3097 For vertex shaders, this sematic label indicates that an input or
3098 output is a boolean edge flag. The register layout is [F, x, x, x]
3099 where F is 0.0 or 1.0 and x = don't care. Normally, the vertex shader
3100 simply copies the edge flag input to the edgeflag output.
3102 Edge flags are used to control which lines or points are actually
3103 drawn when the polygon mode converts triangles/quads/polygons into
3107 TGSI_SEMANTIC_STENCIL
3108 """""""""""""""""""""
3110 For fragment shaders, this semantic label indicates that an output
3111 is a writable stencil reference value. Only the Y component is writable.
3112 This allows the fragment shader to change the fragments stencilref value.
3115 TGSI_SEMANTIC_VIEWPORT_INDEX
3116 """"""""""""""""""""""""""""
3118 For geometry shaders, this semantic label indicates that an output
3119 contains the index of the viewport (and scissor) to use.
3120 This is an integer value, and only the X component is used.
3126 For geometry shaders, this semantic label indicates that an output
3127 contains the layer value to use for the color and depth/stencil surfaces.
3128 This is an integer value, and only the X component is used.
3129 (Also known as rendertarget array index.)
3132 TGSI_SEMANTIC_CULLDIST
3133 """"""""""""""""""""""
3135 Used as distance to plane for performing application-defined culling
3136 of individual primitives against a plane. When components of vertex
3137 elements are given this label, these values are assumed to be a
3138 float32 signed distance to a plane. Primitives will be completely
3139 discarded if the plane distance for all of the vertices in the
3140 primitive are < 0. If a vertex has a cull distance of NaN, that
3141 vertex counts as "out" (as if its < 0);
3142 The limits on both clip and cull distances are bound
3143 by the PIPE_MAX_CLIP_OR_CULL_DISTANCE_COUNT define which defines
3144 the maximum number of components that can be used to hold the
3145 distances and by the PIPE_MAX_CLIP_OR_CULL_DISTANCE_ELEMENT_COUNT
3146 which specifies the maximum number of registers which can be
3147 annotated with those semantics.
3150 TGSI_SEMANTIC_CLIPDIST
3151 """"""""""""""""""""""
3153 Note this covers clipping and culling distances.
3155 When components of vertex elements are identified this way, these
3156 values are each assumed to be a float32 signed distance to a plane.
3159 Primitive setup only invokes rasterization on pixels for which
3160 the interpolated plane distances are >= 0.
3163 Primitives will be completely discarded if the plane distance
3164 for all of the vertices in the primitive are < 0.
3165 If a vertex has a cull distance of NaN, that vertex counts as "out"
3168 Multiple clip/cull planes can be implemented simultaneously, by
3169 annotating multiple components of one or more vertex elements with
3170 the above specified semantic.
3171 The limits on both clip and cull distances are bound
3172 by the PIPE_MAX_CLIP_OR_CULL_DISTANCE_COUNT define which defines
3173 the maximum number of components that can be used to hold the
3174 distances and by the PIPE_MAX_CLIP_OR_CULL_DISTANCE_ELEMENT_COUNT
3175 which specifies the maximum number of registers which can be
3176 annotated with those semantics.
3177 The properties NUM_CLIPDIST_ENABLED and NUM_CULLDIST_ENABLED
3178 are used to divide up the 2 x vec4 space between clipping and culling.
3180 TGSI_SEMANTIC_SAMPLEID
3181 """"""""""""""""""""""
3183 For fragment shaders, this semantic label indicates that a system value
3184 contains the current sample id (i.e. gl_SampleID).
3185 This is an integer value, and only the X component is used.
3187 TGSI_SEMANTIC_SAMPLEPOS
3188 """""""""""""""""""""""
3190 For fragment shaders, this semantic label indicates that a system value
3191 contains the current sample's position (i.e. gl_SamplePosition). Only the X
3192 and Y values are used.
3194 TGSI_SEMANTIC_SAMPLEMASK
3195 """"""""""""""""""""""""
3197 For fragment shaders, this semantic label indicates that an output contains
3198 the sample mask used to disable further sample processing
3199 (i.e. gl_SampleMask). Only the X value is used, up to 32x MS.
3201 TGSI_SEMANTIC_INVOCATIONID
3202 """"""""""""""""""""""""""
3204 For geometry shaders, this semantic label indicates that a system value
3205 contains the current invocation id (i.e. gl_InvocationID).
3206 This is an integer value, and only the X component is used.
3208 TGSI_SEMANTIC_INSTANCEID
3209 """"""""""""""""""""""""
3211 For vertex shaders, this semantic label indicates that a system value contains
3212 the current instance id (i.e. gl_InstanceID). It does not include the base
3213 instance. This is an integer value, and only the X component is used.
3215 TGSI_SEMANTIC_VERTEXID
3216 """"""""""""""""""""""
3218 For vertex shaders, this semantic label indicates that a system value contains
3219 the current vertex id (i.e. gl_VertexID). It does (unlike in d3d10) include the
3220 base vertex. This is an integer value, and only the X component is used.
3222 TGSI_SEMANTIC_VERTEXID_NOBASE
3223 """""""""""""""""""""""""""""""
3225 For vertex shaders, this semantic label indicates that a system value contains
3226 the current vertex id without including the base vertex (this corresponds to
3227 d3d10 vertex id, so TGSI_SEMANTIC_VERTEXID_NOBASE + TGSI_SEMANTIC_BASEVERTEX
3228 == TGSI_SEMANTIC_VERTEXID). This is an integer value, and only the X component
3231 TGSI_SEMANTIC_BASEVERTEX
3232 """"""""""""""""""""""""
3234 For vertex shaders, this semantic label indicates that a system value contains
3235 the base vertex (i.e. gl_BaseVertex). Note that for non-indexed draw calls,
3236 this contains the first (or start) value instead.
3237 This is an integer value, and only the X component is used.
3239 TGSI_SEMANTIC_PRIMID
3240 """"""""""""""""""""
3242 For geometry and fragment shaders, this semantic label indicates the value
3243 contains the primitive id (i.e. gl_PrimitiveID). This is an integer value,
3244 and only the X component is used.
3245 FIXME: This right now can be either a ordinary input or a system value...
3251 For tessellation evaluation/control shaders, this semantic label indicates a
3252 generic per-patch attribute. Such semantics will not implicitly be per-vertex
3255 TGSI_SEMANTIC_TESSCOORD
3256 """""""""""""""""""""""
3258 For tessellation evaluation shaders, this semantic label indicates the
3259 coordinates of the vertex being processed. This is available in XYZ; W is
3262 TGSI_SEMANTIC_TESSOUTER
3263 """""""""""""""""""""""
3265 For tessellation evaluation/control shaders, this semantic label indicates the
3266 outer tessellation levels of the patch. Isoline tessellation will only have XY
3267 defined, triangle will have XYZ and quads will have XYZW defined. This
3268 corresponds to gl_TessLevelOuter.
3270 TGSI_SEMANTIC_TESSINNER
3271 """""""""""""""""""""""
3273 For tessellation evaluation/control shaders, this semantic label indicates the
3274 inner tessellation levels of the patch. The X value is only defined for
3275 triangle tessellation, while quads will have XY defined. This is entirely
3276 undefined for isoline tessellation.
3278 TGSI_SEMANTIC_VERTICESIN
3279 """"""""""""""""""""""""
3281 For tessellation evaluation/control shaders, this semantic label indicates the
3282 number of vertices provided in the input patch. Only the X value is defined.
3284 TGSI_SEMANTIC_HELPER_INVOCATION
3285 """""""""""""""""""""""""""""""
3287 For fragment shaders, this semantic indicates whether the current
3288 invocation is covered or not. Helper invocations are created in order
3289 to properly compute derivatives, however it may be desirable to skip
3290 some of the logic in those cases. See ``gl_HelperInvocation`` documentation.
3292 TGSI_SEMANTIC_BASEINSTANCE
3293 """"""""""""""""""""""""""
3295 For vertex shaders, the base instance argument supplied for this
3296 draw. This is an integer value, and only the X component is used.
3298 TGSI_SEMANTIC_DRAWID
3299 """"""""""""""""""""
3301 For vertex shaders, the zero-based index of the current draw in a
3302 ``glMultiDraw*`` invocation. This is an integer value, and only the X
3306 TGSI_SEMANTIC_WORK_DIM
3307 """"""""""""""""""""""
3309 For compute shaders started via opencl this retrieves the work_dim
3310 parameter to the clEnqueueNDRangeKernel call with which the shader
3314 TGSI_SEMANTIC_GRID_SIZE
3315 """""""""""""""""""""""
3317 For compute shaders, this semantic indicates the maximum (x, y, z) dimensions
3318 of a grid of thread blocks.
3321 TGSI_SEMANTIC_BLOCK_ID
3322 """"""""""""""""""""""
3324 For compute shaders, this semantic indicates the (x, y, z) coordinates of the
3325 current block inside of the grid.
3328 TGSI_SEMANTIC_BLOCK_SIZE
3329 """"""""""""""""""""""""
3331 For compute shaders, this semantic indicates the maximum (x, y, z) dimensions
3332 of a block in threads.
3335 TGSI_SEMANTIC_THREAD_ID
3336 """""""""""""""""""""""
3338 For compute shaders, this semantic indicates the (x, y, z) coordinates of the
3339 current thread inside of the block.
3342 Declaration Interpolate
3343 ^^^^^^^^^^^^^^^^^^^^^^^
3345 This token is only valid for fragment shader INPUT declarations.
3347 The Interpolate field specifes the way input is being interpolated by
3348 the rasteriser and is one of TGSI_INTERPOLATE_*.
3350 The Location field specifies the location inside the pixel that the
3351 interpolation should be done at, one of ``TGSI_INTERPOLATE_LOC_*``. Note that
3352 when per-sample shading is enabled, the implementation may choose to
3353 interpolate at the sample irrespective of the Location field.
3355 The CylindricalWrap bitfield specifies which register components
3356 should be subject to cylindrical wrapping when interpolating by the
3357 rasteriser. If TGSI_CYLINDRICAL_WRAP_X is set to 1, the X component
3358 should be interpolated according to cylindrical wrapping rules.
3361 Declaration Sampler View
3362 ^^^^^^^^^^^^^^^^^^^^^^^^
3364 Follows Declaration token if file is TGSI_FILE_SAMPLER_VIEW.
3366 DCL SVIEW[#], resource, type(s)
3368 Declares a shader input sampler view and assigns it to a SVIEW[#]
3371 resource can be one of BUFFER, 1D, 2D, 3D, 1DArray and 2DArray.
3373 type must be 1 or 4 entries (if specifying on a per-component
3374 level) out of UNORM, SNORM, SINT, UINT and FLOAT.
3376 For TEX\* style texture sample opcodes (as opposed to SAMPLE\* opcodes
3377 which take an explicit SVIEW[#] source register), there may be optionally
3378 SVIEW[#] declarations. In this case, the SVIEW index is implied by the
3379 SAMP index, and there must be a corresponding SVIEW[#] declaration for
3380 each SAMP[#] declaration. Drivers are free to ignore this if they wish.
3381 But note in particular that some drivers need to know the sampler type
3382 (float/int/unsigned) in order to generate the correct code, so cases
3383 where integer textures are sampled, SVIEW[#] declarations should be
3386 NOTE: It is NOT legal to mix SAMPLE\* style opcodes and TEX\* opcodes
3389 Declaration Resource
3390 ^^^^^^^^^^^^^^^^^^^^
3392 Follows Declaration token if file is TGSI_FILE_RESOURCE.
3394 DCL RES[#], resource [, WR] [, RAW]
3396 Declares a shader input resource and assigns it to a RES[#]
3399 resource can be one of BUFFER, 1D, 2D, 3D, CUBE, 1DArray and
3402 If the RAW keyword is not specified, the texture data will be
3403 subject to conversion, swizzling and scaling as required to yield
3404 the specified data type from the physical data format of the bound
3407 If the RAW keyword is specified, no channel conversion will be
3408 performed: the values read for each of the channels (X,Y,Z,W) will
3409 correspond to consecutive words in the same order and format
3410 they're found in memory. No element-to-address conversion will be
3411 performed either: the value of the provided X coordinate will be
3412 interpreted in byte units instead of texel units. The result of
3413 accessing a misaligned address is undefined.
3415 Usage of the STORE opcode is only allowed if the WR (writable) flag
3420 ^^^^^^^^^^^^^^^^^^^^^^^^
3422 Properties are general directives that apply to the whole TGSI program.
3427 Specifies the fragment shader TGSI_SEMANTIC_POSITION coordinate origin.
3428 The default value is UPPER_LEFT.
3430 If UPPER_LEFT, the position will be (0,0) at the upper left corner and
3431 increase downward and rightward.
3432 If LOWER_LEFT, the position will be (0,0) at the lower left corner and
3433 increase upward and rightward.
3435 OpenGL defaults to LOWER_LEFT, and is configurable with the
3436 GL_ARB_fragment_coord_conventions extension.
3438 DirectX 9/10 use UPPER_LEFT.
3440 FS_COORD_PIXEL_CENTER
3441 """""""""""""""""""""
3443 Specifies the fragment shader TGSI_SEMANTIC_POSITION pixel center convention.
3444 The default value is HALF_INTEGER.
3446 If HALF_INTEGER, the fractionary part of the position will be 0.5
3447 If INTEGER, the fractionary part of the position will be 0.0
3449 Note that this does not affect the set of fragments generated by
3450 rasterization, which is instead controlled by half_pixel_center in the
3453 OpenGL defaults to HALF_INTEGER, and is configurable with the
3454 GL_ARB_fragment_coord_conventions extension.
3456 DirectX 9 uses INTEGER.
3457 DirectX 10 uses HALF_INTEGER.
3459 FS_COLOR0_WRITES_ALL_CBUFS
3460 """"""""""""""""""""""""""
3461 Specifies that writes to the fragment shader color 0 are replicated to all
3462 bound cbufs. This facilitates OpenGL's fragColor output vs fragData[0] where
3463 fragData is directed to a single color buffer, but fragColor is broadcast.
3466 """"""""""""""""""""""""""
3467 If this property is set on the program bound to the shader stage before the
3468 fragment shader, user clip planes should have no effect (be disabled) even if
3469 that shader does not write to any clip distance outputs and the rasterizer's
3470 clip_plane_enable is non-zero.
3471 This property is only supported by drivers that also support shader clip
3473 This is useful for APIs that don't have UCPs and where clip distances written
3474 by a shader cannot be disabled.
3479 Specifies the number of times a geometry shader should be executed for each
3480 input primitive. Each invocation will have a different
3481 TGSI_SEMANTIC_INVOCATIONID system value set. If not specified, assumed to
3484 VS_WINDOW_SPACE_POSITION
3485 """"""""""""""""""""""""""
3486 If this property is set on the vertex shader, the TGSI_SEMANTIC_POSITION output
3487 is assumed to contain window space coordinates.
3488 Division of X,Y,Z by W and the viewport transformation are disabled, and 1/W is
3489 directly taken from the 4-th component of the shader output.
3490 Naturally, clipping is not performed on window coordinates either.
3491 The effect of this property is undefined if a geometry or tessellation shader
3497 The number of vertices written by the tessellation control shader. This
3498 effectively defines the patch input size of the tessellation evaluation shader
3504 This sets the tessellation primitive mode, one of ``PIPE_PRIM_TRIANGLES``,
3505 ``PIPE_PRIM_QUADS``, or ``PIPE_PRIM_LINES``. (Unlike in GL, there is no
3506 separate isolines settings, the regular lines is assumed to mean isolines.)
3511 This sets the spacing mode of the tessellation generator, one of
3512 ``PIPE_TESS_SPACING_*``.
3517 This sets the vertex order to be clockwise if the value is 1, or
3518 counter-clockwise if set to 0.
3523 If set to a non-zero value, this turns on point mode for the tessellator,
3524 which means that points will be generated instead of primitives.
3526 NUM_CLIPDIST_ENABLED
3529 How many clip distance scalar outputs are enabled.
3531 NUM_CULLDIST_ENABLED
3534 How many cull distance scalar outputs are enabled.
3536 FS_EARLY_DEPTH_STENCIL
3537 """"""""""""""""""""""
3539 Whether depth test, stencil test, and occlusion query should run before
3540 the fragment shader (regardless of fragment shader side effects). Corresponds
3541 to GLSL early_fragment_tests.
3546 Which shader stage will MOST LIKELY follow after this shader when the shader
3547 is bound. This is only a hint to the driver and doesn't have to be precise.
3548 Only set for VS and TES.
3550 CS_FIXED_BLOCK_WIDTH / HEIGHT / DEPTH
3551 """""""""""""""""""""""""""""""""""""
3553 Threads per block in each dimension, if known at compile time. If the block size
3554 is known all three should be at least 1. If it is unknown they should all be set
3560 The MUL TGSI operation (FP32 multiplication) will return 0 if either
3561 of the operands are equal to 0. That means that 0 * Inf = 0. This
3562 should be set the same way for an entire pipeline. Note that this
3563 applies not only to the literal MUL TGSI opcode, but all FP32
3564 multiplications implied by other operations, such as MAD, FMA, DP2,
3565 DP3, DP4, DPH, DST, LOG, LRP, XPD, and possibly others. If there is a
3566 mismatch between shaders, then it is unspecified whether this behavior
3570 Texture Sampling and Texture Formats
3571 ------------------------------------
3573 This table shows how texture image components are returned as (x,y,z,w) tuples
3574 by TGSI texture instructions, such as :opcode:`TEX`, :opcode:`TXD`, and
3575 :opcode:`TXP`. For reference, OpenGL and Direct3D conventions are shown as
3578 +--------------------+--------------+--------------------+--------------+
3579 | Texture Components | Gallium | OpenGL | Direct3D 9 |
3580 +====================+==============+====================+==============+
3581 | R | (r, 0, 0, 1) | (r, 0, 0, 1) | (r, 1, 1, 1) |
3582 +--------------------+--------------+--------------------+--------------+
3583 | RG | (r, g, 0, 1) | (r, g, 0, 1) | (r, g, 1, 1) |
3584 +--------------------+--------------+--------------------+--------------+
3585 | RGB | (r, g, b, 1) | (r, g, b, 1) | (r, g, b, 1) |
3586 +--------------------+--------------+--------------------+--------------+
3587 | RGBA | (r, g, b, a) | (r, g, b, a) | (r, g, b, a) |
3588 +--------------------+--------------+--------------------+--------------+
3589 | A | (0, 0, 0, a) | (0, 0, 0, a) | (0, 0, 0, a) |
3590 +--------------------+--------------+--------------------+--------------+
3591 | L | (l, l, l, 1) | (l, l, l, 1) | (l, l, l, 1) |
3592 +--------------------+--------------+--------------------+--------------+
3593 | LA | (l, l, l, a) | (l, l, l, a) | (l, l, l, a) |
3594 +--------------------+--------------+--------------------+--------------+
3595 | I | (i, i, i, i) | (i, i, i, i) | N/A |
3596 +--------------------+--------------+--------------------+--------------+
3597 | UV | XXX TBD | (0, 0, 0, 1) | (u, v, 1, 1) |
3598 | | | [#envmap-bumpmap]_ | |
3599 +--------------------+--------------+--------------------+--------------+
3600 | Z | XXX TBD | (z, z, z, 1) | (0, z, 0, 1) |
3601 | | | [#depth-tex-mode]_ | |
3602 +--------------------+--------------+--------------------+--------------+
3603 | S | (s, s, s, s) | unknown | unknown |
3604 +--------------------+--------------+--------------------+--------------+
3606 .. [#envmap-bumpmap] http://www.opengl.org/registry/specs/ATI/envmap_bumpmap.txt
3607 .. [#depth-tex-mode] the default is (z, z, z, 1) but may also be (0, 0, 0, z)
3608 or (z, z, z, z) depending on the value of GL_DEPTH_TEXTURE_MODE.