/* * Copyright (C) 2014 Intel Corporation * * Permission is hereby granted, free of charge, to any person obtaining a * copy of this software and associated documentation files (the "Software"), * to deal in the Software without restriction, including without limitation * the rights to use, copy, modify, merge, publish, distribute, sublicense, * and/or sell copies of the Software, and to permit persons to whom the * Software is furnished to do so, subject to the following conditions: * * The above copyright notice and this permission notice (including the next * paragraph) shall be included in all copies or substantial portions of the * Software. * * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS * IN THE SOFTWARE. * * Authors: * Jason Ekstrand (jason@jlekstrand.net) */ #include #include "util/rounding.h" /* for _mesa_roundeven */ #include "util/half_float.h" #include "nir_constant_expressions.h" /** * Evaluate one component of packSnorm4x8. */ static uint8_t pack_snorm_1x8(float x) { /* From section 8.4 of the GLSL 4.30 spec: * * packSnorm4x8 * ------------ * The conversion for component c of v to fixed point is done as * follows: * * packSnorm4x8: round(clamp(c, -1, +1) * 127.0) * * We must first cast the float to an int, because casting a negative * float to a uint is undefined. */ return (uint8_t) (int) _mesa_roundevenf(CLAMP(x, -1.0f, +1.0f) * 127.0f); } /** * Evaluate one component of packSnorm2x16. */ static uint16_t pack_snorm_1x16(float x) { /* From section 8.4 of the GLSL ES 3.00 spec: * * packSnorm2x16 * ------------- * The conversion for component c of v to fixed point is done as * follows: * * packSnorm2x16: round(clamp(c, -1, +1) * 32767.0) * * We must first cast the float to an int, because casting a negative * float to a uint is undefined. */ return (uint16_t) (int) _mesa_roundevenf(CLAMP(x, -1.0f, +1.0f) * 32767.0f); } /** * Evaluate one component of unpackSnorm4x8. */ static float unpack_snorm_1x8(uint8_t u) { /* From section 8.4 of the GLSL 4.30 spec: * * unpackSnorm4x8 * -------------- * The conversion for unpacked fixed-point value f to floating point is * done as follows: * * unpackSnorm4x8: clamp(f / 127.0, -1, +1) */ return CLAMP((int8_t) u / 127.0f, -1.0f, +1.0f); } /** * Evaluate one component of unpackSnorm2x16. */ static float unpack_snorm_1x16(uint16_t u) { /* From section 8.4 of the GLSL ES 3.00 spec: * * unpackSnorm2x16 * --------------- * The conversion for unpacked fixed-point value f to floating point is * done as follows: * * unpackSnorm2x16: clamp(f / 32767.0, -1, +1) */ return CLAMP((int16_t) u / 32767.0f, -1.0f, +1.0f); } /** * Evaluate one component packUnorm4x8. */ static uint8_t pack_unorm_1x8(float x) { /* From section 8.4 of the GLSL 4.30 spec: * * packUnorm4x8 * ------------ * The conversion for component c of v to fixed point is done as * follows: * * packUnorm4x8: round(clamp(c, 0, +1) * 255.0) */ return (uint8_t) (int) _mesa_roundevenf(CLAMP(x, 0.0f, 1.0f) * 255.0f); } /** * Evaluate one component packUnorm2x16. */ static uint16_t pack_unorm_1x16(float x) { /* From section 8.4 of the GLSL ES 3.00 spec: * * packUnorm2x16 * ------------- * The conversion for component c of v to fixed point is done as * follows: * * packUnorm2x16: round(clamp(c, 0, +1) * 65535.0) */ return (uint16_t) (int) _mesa_roundevenf(CLAMP(x, 0.0f, 1.0f) * 65535.0f); } /** * Evaluate one component of unpackUnorm4x8. */ static float unpack_unorm_1x8(uint8_t u) { /* From section 8.4 of the GLSL 4.30 spec: * * unpackUnorm4x8 * -------------- * The conversion for unpacked fixed-point value f to floating point is * done as follows: * * unpackUnorm4x8: f / 255.0 */ return (float) u / 255.0f; } /** * Evaluate one component of unpackUnorm2x16. */ static float unpack_unorm_1x16(uint16_t u) { /* From section 8.4 of the GLSL ES 3.00 spec: * * unpackUnorm2x16 * --------------- * The conversion for unpacked fixed-point value f to floating point is * done as follows: * * unpackUnorm2x16: f / 65535.0 */ return (float) u / 65535.0f; } /** * Evaluate one component of packHalf2x16. */ static uint16_t pack_half_1x16(float x) { return _mesa_float_to_half(x); } /** * Evaluate one component of unpackHalf2x16. */ static float unpack_half_1x16(uint16_t u) { return _mesa_half_to_float(u); } /* Some typed vector structures to make things like src0.y work */ typedef float float16_t; typedef float float32_t; typedef double float64_t; typedef bool bool32_t; struct float16_vec { float16_t x; float16_t y; float16_t z; float16_t w; }; struct float32_vec { float32_t x; float32_t y; float32_t z; float32_t w; }; struct float64_vec { float64_t x; float64_t y; float64_t z; float64_t w; }; struct int8_vec { int8_t x; int8_t y; int8_t z; int8_t w; }; struct int16_vec { int16_t x; int16_t y; int16_t z; int16_t w; }; struct int32_vec { int32_t x; int32_t y; int32_t z; int32_t w; }; struct int64_vec { int64_t x; int64_t y; int64_t z; int64_t w; }; struct uint8_vec { uint8_t x; uint8_t y; uint8_t z; uint8_t w; }; struct uint16_vec { uint16_t x; uint16_t y; uint16_t z; uint16_t w; }; struct uint32_vec { uint32_t x; uint32_t y; uint32_t z; uint32_t w; }; struct uint64_vec { uint64_t x; uint64_t y; uint64_t z; uint64_t w; }; struct bool32_vec { bool x; bool y; bool z; bool w; }; static nir_const_value evaluate_b2f(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const bool src0 = _src[0].u32[_i] != 0; float16_t dst = src0 ? 1.0 : 0.0; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const bool src0 = _src[0].u32[_i] != 0; float32_t dst = src0 ? 1.0 : 0.0; _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const bool src0 = _src[0].u32[_i] != 0; float64_t dst = src0 ? 1.0 : 0.0; _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_b2i(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const bool src0 = _src[0].u32[_i] != 0; int8_t dst = src0 ? 1 : 0; _dst_val.i8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const bool src0 = _src[0].u32[_i] != 0; int16_t dst = src0 ? 1 : 0; _dst_val.i16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const bool src0 = _src[0].u32[_i] != 0; int32_t dst = src0 ? 1 : 0; _dst_val.i32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const bool src0 = _src[0].u32[_i] != 0; int64_t dst = src0 ? 1 : 0; _dst_val.i64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_ball_fequal2(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { const struct float16_vec src0 = { _mesa_half_to_float(_src[0].u16[0]), _mesa_half_to_float(_src[0].u16[1]), 0, 0, }; const struct float16_vec src1 = { _mesa_half_to_float(_src[1].u16[0]), _mesa_half_to_float(_src[1].u16[1]), 0, 0, }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x == src1.x) && (src0.y == src1.y)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } case 32: { const struct float32_vec src0 = { _src[0].f32[0], _src[0].f32[1], 0, 0, }; const struct float32_vec src1 = { _src[1].f32[0], _src[1].f32[1], 0, 0, }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x == src1.x) && (src0.y == src1.y)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } case 64: { const struct float64_vec src0 = { _src[0].f64[0], _src[0].f64[1], 0, 0, }; const struct float64_vec src1 = { _src[1].f64[0], _src[1].f64[1], 0, 0, }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x == src1.x) && (src0.y == src1.y)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_ball_fequal3(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { const struct float16_vec src0 = { _mesa_half_to_float(_src[0].u16[0]), _mesa_half_to_float(_src[0].u16[1]), _mesa_half_to_float(_src[0].u16[2]), 0, }; const struct float16_vec src1 = { _mesa_half_to_float(_src[1].u16[0]), _mesa_half_to_float(_src[1].u16[1]), _mesa_half_to_float(_src[1].u16[2]), 0, }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x == src1.x) && (src0.y == src1.y) && (src0.z == src1.z)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } case 32: { const struct float32_vec src0 = { _src[0].f32[0], _src[0].f32[1], _src[0].f32[2], 0, }; const struct float32_vec src1 = { _src[1].f32[0], _src[1].f32[1], _src[1].f32[2], 0, }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x == src1.x) && (src0.y == src1.y) && (src0.z == src1.z)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } case 64: { const struct float64_vec src0 = { _src[0].f64[0], _src[0].f64[1], _src[0].f64[2], 0, }; const struct float64_vec src1 = { _src[1].f64[0], _src[1].f64[1], _src[1].f64[2], 0, }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x == src1.x) && (src0.y == src1.y) && (src0.z == src1.z)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_ball_fequal4(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { const struct float16_vec src0 = { _mesa_half_to_float(_src[0].u16[0]), _mesa_half_to_float(_src[0].u16[1]), _mesa_half_to_float(_src[0].u16[2]), _mesa_half_to_float(_src[0].u16[3]), }; const struct float16_vec src1 = { _mesa_half_to_float(_src[1].u16[0]), _mesa_half_to_float(_src[1].u16[1]), _mesa_half_to_float(_src[1].u16[2]), _mesa_half_to_float(_src[1].u16[3]), }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x == src1.x) && (src0.y == src1.y) && (src0.z == src1.z) && (src0.w == src1.w)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } case 32: { const struct float32_vec src0 = { _src[0].f32[0], _src[0].f32[1], _src[0].f32[2], _src[0].f32[3], }; const struct float32_vec src1 = { _src[1].f32[0], _src[1].f32[1], _src[1].f32[2], _src[1].f32[3], }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x == src1.x) && (src0.y == src1.y) && (src0.z == src1.z) && (src0.w == src1.w)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } case 64: { const struct float64_vec src0 = { _src[0].f64[0], _src[0].f64[1], _src[0].f64[2], _src[0].f64[3], }; const struct float64_vec src1 = { _src[1].f64[0], _src[1].f64[1], _src[1].f64[2], _src[1].f64[3], }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x == src1.x) && (src0.y == src1.y) && (src0.z == src1.z) && (src0.w == src1.w)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_ball_iequal2(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { const struct int8_vec src0 = { _src[0].i8[0], _src[0].i8[1], 0, 0, }; const struct int8_vec src1 = { _src[1].i8[0], _src[1].i8[1], 0, 0, }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x == src1.x) && (src0.y == src1.y)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } case 16: { const struct int16_vec src0 = { _src[0].i16[0], _src[0].i16[1], 0, 0, }; const struct int16_vec src1 = { _src[1].i16[0], _src[1].i16[1], 0, 0, }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x == src1.x) && (src0.y == src1.y)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } case 32: { const struct int32_vec src0 = { _src[0].i32[0], _src[0].i32[1], 0, 0, }; const struct int32_vec src1 = { _src[1].i32[0], _src[1].i32[1], 0, 0, }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x == src1.x) && (src0.y == src1.y)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } case 64: { const struct int64_vec src0 = { _src[0].i64[0], _src[0].i64[1], 0, 0, }; const struct int64_vec src1 = { _src[1].i64[0], _src[1].i64[1], 0, 0, }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x == src1.x) && (src0.y == src1.y)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_ball_iequal3(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { const struct int8_vec src0 = { _src[0].i8[0], _src[0].i8[1], _src[0].i8[2], 0, }; const struct int8_vec src1 = { _src[1].i8[0], _src[1].i8[1], _src[1].i8[2], 0, }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x == src1.x) && (src0.y == src1.y) && (src0.z == src1.z)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } case 16: { const struct int16_vec src0 = { _src[0].i16[0], _src[0].i16[1], _src[0].i16[2], 0, }; const struct int16_vec src1 = { _src[1].i16[0], _src[1].i16[1], _src[1].i16[2], 0, }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x == src1.x) && (src0.y == src1.y) && (src0.z == src1.z)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } case 32: { const struct int32_vec src0 = { _src[0].i32[0], _src[0].i32[1], _src[0].i32[2], 0, }; const struct int32_vec src1 = { _src[1].i32[0], _src[1].i32[1], _src[1].i32[2], 0, }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x == src1.x) && (src0.y == src1.y) && (src0.z == src1.z)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } case 64: { const struct int64_vec src0 = { _src[0].i64[0], _src[0].i64[1], _src[0].i64[2], 0, }; const struct int64_vec src1 = { _src[1].i64[0], _src[1].i64[1], _src[1].i64[2], 0, }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x == src1.x) && (src0.y == src1.y) && (src0.z == src1.z)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_ball_iequal4(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { const struct int8_vec src0 = { _src[0].i8[0], _src[0].i8[1], _src[0].i8[2], _src[0].i8[3], }; const struct int8_vec src1 = { _src[1].i8[0], _src[1].i8[1], _src[1].i8[2], _src[1].i8[3], }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x == src1.x) && (src0.y == src1.y) && (src0.z == src1.z) && (src0.w == src1.w)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } case 16: { const struct int16_vec src0 = { _src[0].i16[0], _src[0].i16[1], _src[0].i16[2], _src[0].i16[3], }; const struct int16_vec src1 = { _src[1].i16[0], _src[1].i16[1], _src[1].i16[2], _src[1].i16[3], }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x == src1.x) && (src0.y == src1.y) && (src0.z == src1.z) && (src0.w == src1.w)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } case 32: { const struct int32_vec src0 = { _src[0].i32[0], _src[0].i32[1], _src[0].i32[2], _src[0].i32[3], }; const struct int32_vec src1 = { _src[1].i32[0], _src[1].i32[1], _src[1].i32[2], _src[1].i32[3], }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x == src1.x) && (src0.y == src1.y) && (src0.z == src1.z) && (src0.w == src1.w)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } case 64: { const struct int64_vec src0 = { _src[0].i64[0], _src[0].i64[1], _src[0].i64[2], _src[0].i64[3], }; const struct int64_vec src1 = { _src[1].i64[0], _src[1].i64[1], _src[1].i64[2], _src[1].i64[3], }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x == src1.x) && (src0.y == src1.y) && (src0.z == src1.z) && (src0.w == src1.w)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_bany_fnequal2(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { const struct float16_vec src0 = { _mesa_half_to_float(_src[0].u16[0]), _mesa_half_to_float(_src[0].u16[1]), 0, 0, }; const struct float16_vec src1 = { _mesa_half_to_float(_src[1].u16[0]), _mesa_half_to_float(_src[1].u16[1]), 0, 0, }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x != src1.x) || (src0.y != src1.y)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } case 32: { const struct float32_vec src0 = { _src[0].f32[0], _src[0].f32[1], 0, 0, }; const struct float32_vec src1 = { _src[1].f32[0], _src[1].f32[1], 0, 0, }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x != src1.x) || (src0.y != src1.y)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } case 64: { const struct float64_vec src0 = { _src[0].f64[0], _src[0].f64[1], 0, 0, }; const struct float64_vec src1 = { _src[1].f64[0], _src[1].f64[1], 0, 0, }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x != src1.x) || (src0.y != src1.y)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_bany_fnequal3(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { const struct float16_vec src0 = { _mesa_half_to_float(_src[0].u16[0]), _mesa_half_to_float(_src[0].u16[1]), _mesa_half_to_float(_src[0].u16[2]), 0, }; const struct float16_vec src1 = { _mesa_half_to_float(_src[1].u16[0]), _mesa_half_to_float(_src[1].u16[1]), _mesa_half_to_float(_src[1].u16[2]), 0, }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x != src1.x) || (src0.y != src1.y) || (src0.z != src1.z)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } case 32: { const struct float32_vec src0 = { _src[0].f32[0], _src[0].f32[1], _src[0].f32[2], 0, }; const struct float32_vec src1 = { _src[1].f32[0], _src[1].f32[1], _src[1].f32[2], 0, }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x != src1.x) || (src0.y != src1.y) || (src0.z != src1.z)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } case 64: { const struct float64_vec src0 = { _src[0].f64[0], _src[0].f64[1], _src[0].f64[2], 0, }; const struct float64_vec src1 = { _src[1].f64[0], _src[1].f64[1], _src[1].f64[2], 0, }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x != src1.x) || (src0.y != src1.y) || (src0.z != src1.z)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_bany_fnequal4(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { const struct float16_vec src0 = { _mesa_half_to_float(_src[0].u16[0]), _mesa_half_to_float(_src[0].u16[1]), _mesa_half_to_float(_src[0].u16[2]), _mesa_half_to_float(_src[0].u16[3]), }; const struct float16_vec src1 = { _mesa_half_to_float(_src[1].u16[0]), _mesa_half_to_float(_src[1].u16[1]), _mesa_half_to_float(_src[1].u16[2]), _mesa_half_to_float(_src[1].u16[3]), }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x != src1.x) || (src0.y != src1.y) || (src0.z != src1.z) || (src0.w != src1.w)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } case 32: { const struct float32_vec src0 = { _src[0].f32[0], _src[0].f32[1], _src[0].f32[2], _src[0].f32[3], }; const struct float32_vec src1 = { _src[1].f32[0], _src[1].f32[1], _src[1].f32[2], _src[1].f32[3], }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x != src1.x) || (src0.y != src1.y) || (src0.z != src1.z) || (src0.w != src1.w)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } case 64: { const struct float64_vec src0 = { _src[0].f64[0], _src[0].f64[1], _src[0].f64[2], _src[0].f64[3], }; const struct float64_vec src1 = { _src[1].f64[0], _src[1].f64[1], _src[1].f64[2], _src[1].f64[3], }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x != src1.x) || (src0.y != src1.y) || (src0.z != src1.z) || (src0.w != src1.w)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_bany_inequal2(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { const struct int8_vec src0 = { _src[0].i8[0], _src[0].i8[1], 0, 0, }; const struct int8_vec src1 = { _src[1].i8[0], _src[1].i8[1], 0, 0, }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x != src1.x) || (src0.y != src1.y)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } case 16: { const struct int16_vec src0 = { _src[0].i16[0], _src[0].i16[1], 0, 0, }; const struct int16_vec src1 = { _src[1].i16[0], _src[1].i16[1], 0, 0, }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x != src1.x) || (src0.y != src1.y)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } case 32: { const struct int32_vec src0 = { _src[0].i32[0], _src[0].i32[1], 0, 0, }; const struct int32_vec src1 = { _src[1].i32[0], _src[1].i32[1], 0, 0, }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x != src1.x) || (src0.y != src1.y)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } case 64: { const struct int64_vec src0 = { _src[0].i64[0], _src[0].i64[1], 0, 0, }; const struct int64_vec src1 = { _src[1].i64[0], _src[1].i64[1], 0, 0, }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x != src1.x) || (src0.y != src1.y)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_bany_inequal3(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { const struct int8_vec src0 = { _src[0].i8[0], _src[0].i8[1], _src[0].i8[2], 0, }; const struct int8_vec src1 = { _src[1].i8[0], _src[1].i8[1], _src[1].i8[2], 0, }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x != src1.x) || (src0.y != src1.y) || (src0.z != src1.z)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } case 16: { const struct int16_vec src0 = { _src[0].i16[0], _src[0].i16[1], _src[0].i16[2], 0, }; const struct int16_vec src1 = { _src[1].i16[0], _src[1].i16[1], _src[1].i16[2], 0, }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x != src1.x) || (src0.y != src1.y) || (src0.z != src1.z)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } case 32: { const struct int32_vec src0 = { _src[0].i32[0], _src[0].i32[1], _src[0].i32[2], 0, }; const struct int32_vec src1 = { _src[1].i32[0], _src[1].i32[1], _src[1].i32[2], 0, }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x != src1.x) || (src0.y != src1.y) || (src0.z != src1.z)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } case 64: { const struct int64_vec src0 = { _src[0].i64[0], _src[0].i64[1], _src[0].i64[2], 0, }; const struct int64_vec src1 = { _src[1].i64[0], _src[1].i64[1], _src[1].i64[2], 0, }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x != src1.x) || (src0.y != src1.y) || (src0.z != src1.z)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_bany_inequal4(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { const struct int8_vec src0 = { _src[0].i8[0], _src[0].i8[1], _src[0].i8[2], _src[0].i8[3], }; const struct int8_vec src1 = { _src[1].i8[0], _src[1].i8[1], _src[1].i8[2], _src[1].i8[3], }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x != src1.x) || (src0.y != src1.y) || (src0.z != src1.z) || (src0.w != src1.w)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } case 16: { const struct int16_vec src0 = { _src[0].i16[0], _src[0].i16[1], _src[0].i16[2], _src[0].i16[3], }; const struct int16_vec src1 = { _src[1].i16[0], _src[1].i16[1], _src[1].i16[2], _src[1].i16[3], }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x != src1.x) || (src0.y != src1.y) || (src0.z != src1.z) || (src0.w != src1.w)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } case 32: { const struct int32_vec src0 = { _src[0].i32[0], _src[0].i32[1], _src[0].i32[2], _src[0].i32[3], }; const struct int32_vec src1 = { _src[1].i32[0], _src[1].i32[1], _src[1].i32[2], _src[1].i32[3], }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x != src1.x) || (src0.y != src1.y) || (src0.z != src1.z) || (src0.w != src1.w)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } case 64: { const struct int64_vec src0 = { _src[0].i64[0], _src[0].i64[1], _src[0].i64[2], _src[0].i64[3], }; const struct int64_vec src1 = { _src[1].i64[0], _src[1].i64[1], _src[1].i64[2], _src[1].i64[3], }; struct bool32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x != src1.x) || (src0.y != src1.y) || (src0.z != src1.z) || (src0.w != src1.w)); _dst_val.u32[0] = dst.x ? NIR_TRUE : NIR_FALSE; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_bcsel(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const bool src0 = _src[0].u32[_i] != 0; const uint8_t src1 = _src[1].u8[_i]; const uint8_t src2 = _src[2].u8[_i]; uint8_t dst = src0 ? src1 : src2; _dst_val.u8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const bool src0 = _src[0].u32[_i] != 0; const uint16_t src1 = _src[1].u16[_i]; const uint16_t src2 = _src[2].u16[_i]; uint16_t dst = src0 ? src1 : src2; _dst_val.u16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const bool src0 = _src[0].u32[_i] != 0; const uint32_t src1 = _src[1].u32[_i]; const uint32_t src2 = _src[2].u32[_i]; uint32_t dst = src0 ? src1 : src2; _dst_val.u32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const bool src0 = _src[0].u32[_i] != 0; const uint64_t src1 = _src[1].u64[_i]; const uint64_t src2 = _src[2].u64[_i]; uint64_t dst = src0 ? src1 : src2; _dst_val.u64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_bfi(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; const uint32_t src1 = _src[1].u32[_i]; const uint32_t src2 = _src[2].u32[_i]; uint32_t dst; unsigned mask = src0, insert = src1, base = src2; if (mask == 0) { dst = base; } else { unsigned tmp = mask; while (!(tmp & 1)) { tmp >>= 1; insert <<= 1; } dst = (base & ~mask) | (insert & mask); } _dst_val.u32[_i] = dst; } return _dst_val; } static nir_const_value evaluate_bfm(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; const int32_t src1 = _src[1].i32[_i]; uint32_t dst; int bits = src0, offset = src1; if (offset < 0 || bits < 0 || offset > 31 || bits > 31 || offset + bits > 32) dst = 0; /* undefined */ else dst = ((1u << bits) - 1) << offset; _dst_val.u32[_i] = dst; } return _dst_val; } static nir_const_value evaluate_bit_count(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const uint8_t src0 = _src[0].u8[_i]; uint32_t dst; dst = 0; for (unsigned bit = 0; bit < bit_size; bit++) { if ((src0 >> bit) & 1) dst++; } _dst_val.u32[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const uint16_t src0 = _src[0].u16[_i]; uint32_t dst; dst = 0; for (unsigned bit = 0; bit < bit_size; bit++) { if ((src0 >> bit) & 1) dst++; } _dst_val.u32[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; uint32_t dst; dst = 0; for (unsigned bit = 0; bit < bit_size; bit++) { if ((src0 >> bit) & 1) dst++; } _dst_val.u32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const uint64_t src0 = _src[0].u64[_i]; uint32_t dst; dst = 0; for (unsigned bit = 0; bit < bit_size; bit++) { if ((src0 >> bit) & 1) dst++; } _dst_val.u32[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_bitfield_insert(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; const uint32_t src1 = _src[1].u32[_i]; const int32_t src2 = _src[2].i32[_i]; const int32_t src3 = _src[3].i32[_i]; uint32_t dst; unsigned base = src0, insert = src1; int offset = src2, bits = src3; if (bits == 0) { dst = base; } else if (offset < 0 || bits < 0 || bits + offset > 32) { dst = 0; } else { unsigned mask = ((1ull << bits) - 1) << offset; dst = (base & ~mask) | ((insert << offset) & mask); } _dst_val.u32[_i] = dst; } return _dst_val; } static nir_const_value evaluate_bitfield_reverse(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; uint32_t dst; /* we're not winning any awards for speed here, but that's ok */ dst = 0; for (unsigned bit = 0; bit < 32; bit++) dst |= ((src0 >> bit) & 1) << (31 - bit); _dst_val.u32[_i] = dst; } return _dst_val; } static nir_const_value evaluate_cube_face_coord(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; const struct float32_vec src0 = { _src[0].f32[0], _src[0].f32[1], _src[0].f32[2], 0, }; struct float32_vec dst; dst.x = dst.y = 0.0; float absX = fabs(src0.x); float absY = fabs(src0.y); float absZ = fabs(src0.z); if (src0.x >= 0 && absX >= absY && absX >= absZ) { dst.x = -src0.y; dst.y = -src0.z; } if (src0.x < 0 && absX >= absY && absX >= absZ) { dst.x = -src0.y; dst.y = src0.z; } if (src0.y >= 0 && absY >= absX && absY >= absZ) { dst.x = src0.z; dst.y = src0.x; } if (src0.y < 0 && absY >= absX && absY >= absZ) { dst.x = -src0.z; dst.y = src0.x; } if (src0.z >= 0 && absZ >= absX && absZ >= absY) { dst.x = -src0.y; dst.y = src0.x; } if (src0.z < 0 && absZ >= absX && absZ >= absY) { dst.x = -src0.y; dst.y = -src0.x; } _dst_val.f32[0] = dst.x; _dst_val.f32[1] = dst.y; return _dst_val; } static nir_const_value evaluate_cube_face_index(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; const struct float32_vec src0 = { _src[0].f32[0], _src[0].f32[1], _src[0].f32[2], 0, }; struct float32_vec dst; float absX = fabs(src0.x); float absY = fabs(src0.y); float absZ = fabs(src0.z); if (src0.x >= 0 && absX >= absY && absX >= absZ) dst.x = 0; if (src0.x < 0 && absX >= absY && absX >= absZ) dst.x = 1; if (src0.y >= 0 && absY >= absX && absY >= absZ) dst.x = 2; if (src0.y < 0 && absY >= absX && absY >= absZ) dst.x = 3; if (src0.z >= 0 && absZ >= absX && absZ >= absY) dst.x = 4; if (src0.z < 0 && absZ >= absX && absZ >= absY) dst.x = 5; _dst_val.f32[0] = dst.x; return _dst_val; } static nir_const_value evaluate_extract_i16(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; const int8_t src1 = _src[1].i8[_i]; int8_t dst = (int16_t)(src0 >> (src1 * 16)); _dst_val.i8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; const int16_t src1 = _src[1].i16[_i]; int16_t dst = (int16_t)(src0 >> (src1 * 16)); _dst_val.i16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; const int32_t src1 = _src[1].i32[_i]; int32_t dst = (int16_t)(src0 >> (src1 * 16)); _dst_val.i32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; const int64_t src1 = _src[1].i64[_i]; int64_t dst = (int16_t)(src0 >> (src1 * 16)); _dst_val.i64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_extract_i8(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; const int8_t src1 = _src[1].i8[_i]; int8_t dst = (int8_t)(src0 >> (src1 * 8)); _dst_val.i8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; const int16_t src1 = _src[1].i16[_i]; int16_t dst = (int8_t)(src0 >> (src1 * 8)); _dst_val.i16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; const int32_t src1 = _src[1].i32[_i]; int32_t dst = (int8_t)(src0 >> (src1 * 8)); _dst_val.i32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; const int64_t src1 = _src[1].i64[_i]; int64_t dst = (int8_t)(src0 >> (src1 * 8)); _dst_val.i64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_extract_u16(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const uint8_t src0 = _src[0].u8[_i]; const uint8_t src1 = _src[1].u8[_i]; uint8_t dst = (uint16_t)(src0 >> (src1 * 16)); _dst_val.u8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const uint16_t src0 = _src[0].u16[_i]; const uint16_t src1 = _src[1].u16[_i]; uint16_t dst = (uint16_t)(src0 >> (src1 * 16)); _dst_val.u16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; const uint32_t src1 = _src[1].u32[_i]; uint32_t dst = (uint16_t)(src0 >> (src1 * 16)); _dst_val.u32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const uint64_t src0 = _src[0].u64[_i]; const uint64_t src1 = _src[1].u64[_i]; uint64_t dst = (uint16_t)(src0 >> (src1 * 16)); _dst_val.u64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_extract_u8(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const uint8_t src0 = _src[0].u8[_i]; const uint8_t src1 = _src[1].u8[_i]; uint8_t dst = (uint8_t)(src0 >> (src1 * 8)); _dst_val.u8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const uint16_t src0 = _src[0].u16[_i]; const uint16_t src1 = _src[1].u16[_i]; uint16_t dst = (uint8_t)(src0 >> (src1 * 8)); _dst_val.u16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; const uint32_t src1 = _src[1].u32[_i]; uint32_t dst = (uint8_t)(src0 >> (src1 * 8)); _dst_val.u32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const uint64_t src0 = _src[0].u64[_i]; const uint64_t src1 = _src[1].u64[_i]; uint64_t dst = (uint8_t)(src0 >> (src1 * 8)); _dst_val.u64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_f2b(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); bool32_t dst = src0 != 0.0; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; bool32_t dst = src0 != 0.0; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; bool32_t dst = src0 != 0.0; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_f2f16(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); float16_t dst = src0; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; float16_t dst = src0; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; float16_t dst = src0; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_f2f16_rtne(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); float16_t dst = src0; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; float16_t dst = src0; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; float16_t dst = src0; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_f2f16_rtz(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); float16_t dst = src0; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; float16_t dst = src0; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; float16_t dst = src0; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_f2f32(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); float32_t dst = src0; _dst_val.f32[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; float32_t dst = src0; _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; float32_t dst = src0; _dst_val.f32[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_f2f64(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); float64_t dst = src0; _dst_val.f64[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; float64_t dst = src0; _dst_val.f64[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; float64_t dst = src0; _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_f2i16(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); int16_t dst = src0; _dst_val.i16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; int16_t dst = src0; _dst_val.i16[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; int16_t dst = src0; _dst_val.i16[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_f2i32(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); int32_t dst = src0; _dst_val.i32[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; int32_t dst = src0; _dst_val.i32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; int32_t dst = src0; _dst_val.i32[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_f2i64(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); int64_t dst = src0; _dst_val.i64[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; int64_t dst = src0; _dst_val.i64[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; int64_t dst = src0; _dst_val.i64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_f2i8(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); int8_t dst = src0; _dst_val.i8[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; int8_t dst = src0; _dst_val.i8[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; int8_t dst = src0; _dst_val.i8[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_f2u16(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); uint16_t dst = src0; _dst_val.u16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; uint16_t dst = src0; _dst_val.u16[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; uint16_t dst = src0; _dst_val.u16[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_f2u32(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); uint32_t dst = src0; _dst_val.u32[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; uint32_t dst = src0; _dst_val.u32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; uint32_t dst = src0; _dst_val.u32[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_f2u64(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); uint64_t dst = src0; _dst_val.u64[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; uint64_t dst = src0; _dst_val.u64[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; uint64_t dst = src0; _dst_val.u64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_f2u8(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); uint8_t dst = src0; _dst_val.u8[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; uint8_t dst = src0; _dst_val.u8[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; uint8_t dst = src0; _dst_val.u8[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fabs(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); float16_t dst = fabs(src0); _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; float32_t dst = fabs(src0); _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; float64_t dst = fabs(src0); _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fadd(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); const float src1 = _mesa_half_to_float(_src[1].u16[_i]); float16_t dst = src0 + src1; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; const float32_t src1 = _src[1].f32[_i]; float32_t dst = src0 + src1; _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; const float64_t src1 = _src[1].f64[_i]; float64_t dst = src0 + src1; _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fall_equal2(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; const struct float32_vec src0 = { _src[0].f32[0], _src[0].f32[1], 0, 0, }; const struct float32_vec src1 = { _src[1].f32[0], _src[1].f32[1], 0, 0, }; struct float32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x == src1.x) && (src0.y == src1.y)) ? 1.0f : 0.0f; _dst_val.f32[0] = dst.x; return _dst_val; } static nir_const_value evaluate_fall_equal3(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; const struct float32_vec src0 = { _src[0].f32[0], _src[0].f32[1], _src[0].f32[2], 0, }; const struct float32_vec src1 = { _src[1].f32[0], _src[1].f32[1], _src[1].f32[2], 0, }; struct float32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x == src1.x) && (src0.y == src1.y) && (src0.z == src1.z)) ? 1.0f : 0.0f; _dst_val.f32[0] = dst.x; return _dst_val; } static nir_const_value evaluate_fall_equal4(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; const struct float32_vec src0 = { _src[0].f32[0], _src[0].f32[1], _src[0].f32[2], _src[0].f32[3], }; const struct float32_vec src1 = { _src[1].f32[0], _src[1].f32[1], _src[1].f32[2], _src[1].f32[3], }; struct float32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x == src1.x) && (src0.y == src1.y) && (src0.z == src1.z) && (src0.w == src1.w)) ? 1.0f : 0.0f; _dst_val.f32[0] = dst.x; return _dst_val; } static nir_const_value evaluate_fand(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; const float32_t src1 = _src[1].f32[_i]; float32_t dst = ((src0 != 0.0f) && (src1 != 0.0f)) ? 1.0f : 0.0f; _dst_val.f32[_i] = dst; } return _dst_val; } static nir_const_value evaluate_fany_nequal2(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; const struct float32_vec src0 = { _src[0].f32[0], _src[0].f32[1], 0, 0, }; const struct float32_vec src1 = { _src[1].f32[0], _src[1].f32[1], 0, 0, }; struct float32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x != src1.x) || (src0.y != src1.y)) ? 1.0f : 0.0f; _dst_val.f32[0] = dst.x; return _dst_val; } static nir_const_value evaluate_fany_nequal3(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; const struct float32_vec src0 = { _src[0].f32[0], _src[0].f32[1], _src[0].f32[2], 0, }; const struct float32_vec src1 = { _src[1].f32[0], _src[1].f32[1], _src[1].f32[2], 0, }; struct float32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x != src1.x) || (src0.y != src1.y) || (src0.z != src1.z)) ? 1.0f : 0.0f; _dst_val.f32[0] = dst.x; return _dst_val; } static nir_const_value evaluate_fany_nequal4(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; const struct float32_vec src0 = { _src[0].f32[0], _src[0].f32[1], _src[0].f32[2], _src[0].f32[3], }; const struct float32_vec src1 = { _src[1].f32[0], _src[1].f32[1], _src[1].f32[2], _src[1].f32[3], }; struct float32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x != src1.x) || (src0.y != src1.y) || (src0.z != src1.z) || (src0.w != src1.w)) ? 1.0f : 0.0f; _dst_val.f32[0] = dst.x; return _dst_val; } static nir_const_value evaluate_fceil(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); float16_t dst = bit_size == 64 ? ceil(src0) : ceilf(src0); _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; float32_t dst = bit_size == 64 ? ceil(src0) : ceilf(src0); _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; float64_t dst = bit_size == 64 ? ceil(src0) : ceilf(src0); _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fcos(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); float16_t dst = bit_size == 64 ? cos(src0) : cosf(src0); _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; float32_t dst = bit_size == 64 ? cos(src0) : cosf(src0); _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; float64_t dst = bit_size == 64 ? cos(src0) : cosf(src0); _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fcsel(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; const float32_t src1 = _src[1].f32[_i]; const float32_t src2 = _src[2].f32[_i]; float32_t dst = (src0 != 0.0f) ? src1 : src2; _dst_val.f32[_i] = dst; } return _dst_val; } static nir_const_value evaluate_fddx(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { float16_t dst = 0.0; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { float32_t dst = 0.0; _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { float64_t dst = 0.0; _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fddx_coarse(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { float16_t dst = 0.0; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { float32_t dst = 0.0; _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { float64_t dst = 0.0; _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fddx_fine(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { float16_t dst = 0.0; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { float32_t dst = 0.0; _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { float64_t dst = 0.0; _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fddy(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { float16_t dst = 0.0; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { float32_t dst = 0.0; _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { float64_t dst = 0.0; _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fddy_coarse(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { float16_t dst = 0.0; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { float32_t dst = 0.0; _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { float64_t dst = 0.0; _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fddy_fine(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { float16_t dst = 0.0; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { float32_t dst = 0.0; _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { float64_t dst = 0.0; _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fdiv(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); const float src1 = _mesa_half_to_float(_src[1].u16[_i]); float16_t dst = src0 / src1; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; const float32_t src1 = _src[1].f32[_i]; float32_t dst = src0 / src1; _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; const float64_t src1 = _src[1].f64[_i]; float64_t dst = src0 / src1; _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fdot2(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { const struct float16_vec src0 = { _mesa_half_to_float(_src[0].u16[0]), _mesa_half_to_float(_src[0].u16[1]), 0, 0, }; const struct float16_vec src1 = { _mesa_half_to_float(_src[1].u16[0]), _mesa_half_to_float(_src[1].u16[1]), 0, 0, }; struct float16_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x * src1.x) + (src0.y * src1.y)); _dst_val.u16[0] = _mesa_float_to_half(dst.x); break; } case 32: { const struct float32_vec src0 = { _src[0].f32[0], _src[0].f32[1], 0, 0, }; const struct float32_vec src1 = { _src[1].f32[0], _src[1].f32[1], 0, 0, }; struct float32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x * src1.x) + (src0.y * src1.y)); _dst_val.f32[0] = dst.x; break; } case 64: { const struct float64_vec src0 = { _src[0].f64[0], _src[0].f64[1], 0, 0, }; const struct float64_vec src1 = { _src[1].f64[0], _src[1].f64[1], 0, 0, }; struct float64_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x * src1.x) + (src0.y * src1.y)); _dst_val.f64[0] = dst.x; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fdot3(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { const struct float16_vec src0 = { _mesa_half_to_float(_src[0].u16[0]), _mesa_half_to_float(_src[0].u16[1]), _mesa_half_to_float(_src[0].u16[2]), 0, }; const struct float16_vec src1 = { _mesa_half_to_float(_src[1].u16[0]), _mesa_half_to_float(_src[1].u16[1]), _mesa_half_to_float(_src[1].u16[2]), 0, }; struct float16_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x * src1.x) + (src0.y * src1.y) + (src0.z * src1.z)); _dst_val.u16[0] = _mesa_float_to_half(dst.x); break; } case 32: { const struct float32_vec src0 = { _src[0].f32[0], _src[0].f32[1], _src[0].f32[2], 0, }; const struct float32_vec src1 = { _src[1].f32[0], _src[1].f32[1], _src[1].f32[2], 0, }; struct float32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x * src1.x) + (src0.y * src1.y) + (src0.z * src1.z)); _dst_val.f32[0] = dst.x; break; } case 64: { const struct float64_vec src0 = { _src[0].f64[0], _src[0].f64[1], _src[0].f64[2], 0, }; const struct float64_vec src1 = { _src[1].f64[0], _src[1].f64[1], _src[1].f64[2], 0, }; struct float64_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x * src1.x) + (src0.y * src1.y) + (src0.z * src1.z)); _dst_val.f64[0] = dst.x; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fdot4(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { const struct float16_vec src0 = { _mesa_half_to_float(_src[0].u16[0]), _mesa_half_to_float(_src[0].u16[1]), _mesa_half_to_float(_src[0].u16[2]), _mesa_half_to_float(_src[0].u16[3]), }; const struct float16_vec src1 = { _mesa_half_to_float(_src[1].u16[0]), _mesa_half_to_float(_src[1].u16[1]), _mesa_half_to_float(_src[1].u16[2]), _mesa_half_to_float(_src[1].u16[3]), }; struct float16_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x * src1.x) + (src0.y * src1.y) + (src0.z * src1.z) + (src0.w * src1.w)); _dst_val.u16[0] = _mesa_float_to_half(dst.x); break; } case 32: { const struct float32_vec src0 = { _src[0].f32[0], _src[0].f32[1], _src[0].f32[2], _src[0].f32[3], }; const struct float32_vec src1 = { _src[1].f32[0], _src[1].f32[1], _src[1].f32[2], _src[1].f32[3], }; struct float32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x * src1.x) + (src0.y * src1.y) + (src0.z * src1.z) + (src0.w * src1.w)); _dst_val.f32[0] = dst.x; break; } case 64: { const struct float64_vec src0 = { _src[0].f64[0], _src[0].f64[1], _src[0].f64[2], _src[0].f64[3], }; const struct float64_vec src1 = { _src[1].f64[0], _src[1].f64[1], _src[1].f64[2], _src[1].f64[3], }; struct float64_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x * src1.x) + (src0.y * src1.y) + (src0.z * src1.z) + (src0.w * src1.w)); _dst_val.f64[0] = dst.x; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fdot_replicated2(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { const struct float16_vec src0 = { _mesa_half_to_float(_src[0].u16[0]), _mesa_half_to_float(_src[0].u16[1]), 0, 0, }; const struct float16_vec src1 = { _mesa_half_to_float(_src[1].u16[0]), _mesa_half_to_float(_src[1].u16[1]), 0, 0, }; struct float16_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x * src1.x) + (src0.y * src1.y)); _dst_val.u16[0] = _mesa_float_to_half(dst.x); _dst_val.u16[1] = _mesa_float_to_half(dst.y); _dst_val.u16[2] = _mesa_float_to_half(dst.z); _dst_val.u16[3] = _mesa_float_to_half(dst.w); break; } case 32: { const struct float32_vec src0 = { _src[0].f32[0], _src[0].f32[1], 0, 0, }; const struct float32_vec src1 = { _src[1].f32[0], _src[1].f32[1], 0, 0, }; struct float32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x * src1.x) + (src0.y * src1.y)); _dst_val.f32[0] = dst.x; _dst_val.f32[1] = dst.y; _dst_val.f32[2] = dst.z; _dst_val.f32[3] = dst.w; break; } case 64: { const struct float64_vec src0 = { _src[0].f64[0], _src[0].f64[1], 0, 0, }; const struct float64_vec src1 = { _src[1].f64[0], _src[1].f64[1], 0, 0, }; struct float64_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x * src1.x) + (src0.y * src1.y)); _dst_val.f64[0] = dst.x; _dst_val.f64[1] = dst.y; _dst_val.f64[2] = dst.z; _dst_val.f64[3] = dst.w; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fdot_replicated3(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { const struct float16_vec src0 = { _mesa_half_to_float(_src[0].u16[0]), _mesa_half_to_float(_src[0].u16[1]), _mesa_half_to_float(_src[0].u16[2]), 0, }; const struct float16_vec src1 = { _mesa_half_to_float(_src[1].u16[0]), _mesa_half_to_float(_src[1].u16[1]), _mesa_half_to_float(_src[1].u16[2]), 0, }; struct float16_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x * src1.x) + (src0.y * src1.y) + (src0.z * src1.z)); _dst_val.u16[0] = _mesa_float_to_half(dst.x); _dst_val.u16[1] = _mesa_float_to_half(dst.y); _dst_val.u16[2] = _mesa_float_to_half(dst.z); _dst_val.u16[3] = _mesa_float_to_half(dst.w); break; } case 32: { const struct float32_vec src0 = { _src[0].f32[0], _src[0].f32[1], _src[0].f32[2], 0, }; const struct float32_vec src1 = { _src[1].f32[0], _src[1].f32[1], _src[1].f32[2], 0, }; struct float32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x * src1.x) + (src0.y * src1.y) + (src0.z * src1.z)); _dst_val.f32[0] = dst.x; _dst_val.f32[1] = dst.y; _dst_val.f32[2] = dst.z; _dst_val.f32[3] = dst.w; break; } case 64: { const struct float64_vec src0 = { _src[0].f64[0], _src[0].f64[1], _src[0].f64[2], 0, }; const struct float64_vec src1 = { _src[1].f64[0], _src[1].f64[1], _src[1].f64[2], 0, }; struct float64_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x * src1.x) + (src0.y * src1.y) + (src0.z * src1.z)); _dst_val.f64[0] = dst.x; _dst_val.f64[1] = dst.y; _dst_val.f64[2] = dst.z; _dst_val.f64[3] = dst.w; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fdot_replicated4(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { const struct float16_vec src0 = { _mesa_half_to_float(_src[0].u16[0]), _mesa_half_to_float(_src[0].u16[1]), _mesa_half_to_float(_src[0].u16[2]), _mesa_half_to_float(_src[0].u16[3]), }; const struct float16_vec src1 = { _mesa_half_to_float(_src[1].u16[0]), _mesa_half_to_float(_src[1].u16[1]), _mesa_half_to_float(_src[1].u16[2]), _mesa_half_to_float(_src[1].u16[3]), }; struct float16_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x * src1.x) + (src0.y * src1.y) + (src0.z * src1.z) + (src0.w * src1.w)); _dst_val.u16[0] = _mesa_float_to_half(dst.x); _dst_val.u16[1] = _mesa_float_to_half(dst.y); _dst_val.u16[2] = _mesa_float_to_half(dst.z); _dst_val.u16[3] = _mesa_float_to_half(dst.w); break; } case 32: { const struct float32_vec src0 = { _src[0].f32[0], _src[0].f32[1], _src[0].f32[2], _src[0].f32[3], }; const struct float32_vec src1 = { _src[1].f32[0], _src[1].f32[1], _src[1].f32[2], _src[1].f32[3], }; struct float32_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x * src1.x) + (src0.y * src1.y) + (src0.z * src1.z) + (src0.w * src1.w)); _dst_val.f32[0] = dst.x; _dst_val.f32[1] = dst.y; _dst_val.f32[2] = dst.z; _dst_val.f32[3] = dst.w; break; } case 64: { const struct float64_vec src0 = { _src[0].f64[0], _src[0].f64[1], _src[0].f64[2], _src[0].f64[3], }; const struct float64_vec src1 = { _src[1].f64[0], _src[1].f64[1], _src[1].f64[2], _src[1].f64[3], }; struct float64_vec dst; dst.x = dst.y = dst.z = dst.w = ((src0.x * src1.x) + (src0.y * src1.y) + (src0.z * src1.z) + (src0.w * src1.w)); _dst_val.f64[0] = dst.x; _dst_val.f64[1] = dst.y; _dst_val.f64[2] = dst.z; _dst_val.f64[3] = dst.w; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fdph(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { const struct float16_vec src0 = { _mesa_half_to_float(_src[0].u16[0]), _mesa_half_to_float(_src[0].u16[1]), _mesa_half_to_float(_src[0].u16[2]), 0, }; const struct float16_vec src1 = { _mesa_half_to_float(_src[1].u16[0]), _mesa_half_to_float(_src[1].u16[1]), _mesa_half_to_float(_src[1].u16[2]), _mesa_half_to_float(_src[1].u16[3]), }; struct float16_vec dst; dst.x = dst.y = dst.z = dst.w = src0.x * src1.x + src0.y * src1.y + src0.z * src1.z + src1.w; _dst_val.u16[0] = _mesa_float_to_half(dst.x); break; } case 32: { const struct float32_vec src0 = { _src[0].f32[0], _src[0].f32[1], _src[0].f32[2], 0, }; const struct float32_vec src1 = { _src[1].f32[0], _src[1].f32[1], _src[1].f32[2], _src[1].f32[3], }; struct float32_vec dst; dst.x = dst.y = dst.z = dst.w = src0.x * src1.x + src0.y * src1.y + src0.z * src1.z + src1.w; _dst_val.f32[0] = dst.x; break; } case 64: { const struct float64_vec src0 = { _src[0].f64[0], _src[0].f64[1], _src[0].f64[2], 0, }; const struct float64_vec src1 = { _src[1].f64[0], _src[1].f64[1], _src[1].f64[2], _src[1].f64[3], }; struct float64_vec dst; dst.x = dst.y = dst.z = dst.w = src0.x * src1.x + src0.y * src1.y + src0.z * src1.z + src1.w; _dst_val.f64[0] = dst.x; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fdph_replicated(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { const struct float16_vec src0 = { _mesa_half_to_float(_src[0].u16[0]), _mesa_half_to_float(_src[0].u16[1]), _mesa_half_to_float(_src[0].u16[2]), 0, }; const struct float16_vec src1 = { _mesa_half_to_float(_src[1].u16[0]), _mesa_half_to_float(_src[1].u16[1]), _mesa_half_to_float(_src[1].u16[2]), _mesa_half_to_float(_src[1].u16[3]), }; struct float16_vec dst; dst.x = dst.y = dst.z = dst.w = src0.x * src1.x + src0.y * src1.y + src0.z * src1.z + src1.w; _dst_val.u16[0] = _mesa_float_to_half(dst.x); _dst_val.u16[1] = _mesa_float_to_half(dst.y); _dst_val.u16[2] = _mesa_float_to_half(dst.z); _dst_val.u16[3] = _mesa_float_to_half(dst.w); break; } case 32: { const struct float32_vec src0 = { _src[0].f32[0], _src[0].f32[1], _src[0].f32[2], 0, }; const struct float32_vec src1 = { _src[1].f32[0], _src[1].f32[1], _src[1].f32[2], _src[1].f32[3], }; struct float32_vec dst; dst.x = dst.y = dst.z = dst.w = src0.x * src1.x + src0.y * src1.y + src0.z * src1.z + src1.w; _dst_val.f32[0] = dst.x; _dst_val.f32[1] = dst.y; _dst_val.f32[2] = dst.z; _dst_val.f32[3] = dst.w; break; } case 64: { const struct float64_vec src0 = { _src[0].f64[0], _src[0].f64[1], _src[0].f64[2], 0, }; const struct float64_vec src1 = { _src[1].f64[0], _src[1].f64[1], _src[1].f64[2], _src[1].f64[3], }; struct float64_vec dst; dst.x = dst.y = dst.z = dst.w = src0.x * src1.x + src0.y * src1.y + src0.z * src1.z + src1.w; _dst_val.f64[0] = dst.x; _dst_val.f64[1] = dst.y; _dst_val.f64[2] = dst.z; _dst_val.f64[3] = dst.w; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_feq(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); const float src1 = _mesa_half_to_float(_src[1].u16[_i]); bool32_t dst = src0 == src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; const float32_t src1 = _src[1].f32[_i]; bool32_t dst = src0 == src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; const float64_t src1 = _src[1].f64[_i]; bool32_t dst = src0 == src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fexp2(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); float16_t dst = exp2f(src0); _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; float32_t dst = exp2f(src0); _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; float64_t dst = exp2f(src0); _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_ffloor(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); float16_t dst = bit_size == 64 ? floor(src0) : floorf(src0); _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; float32_t dst = bit_size == 64 ? floor(src0) : floorf(src0); _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; float64_t dst = bit_size == 64 ? floor(src0) : floorf(src0); _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_ffma(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); const float src1 = _mesa_half_to_float(_src[1].u16[_i]); const float src2 = _mesa_half_to_float(_src[2].u16[_i]); float16_t dst = src0 * src1 + src2; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; const float32_t src1 = _src[1].f32[_i]; const float32_t src2 = _src[2].f32[_i]; float32_t dst = src0 * src1 + src2; _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; const float64_t src1 = _src[1].f64[_i]; const float64_t src2 = _src[2].f64[_i]; float64_t dst = src0 * src1 + src2; _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_ffract(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); float16_t dst = src0 - (bit_size == 64 ? floor(src0) : floorf(src0)); _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; float32_t dst = src0 - (bit_size == 64 ? floor(src0) : floorf(src0)); _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; float64_t dst = src0 - (bit_size == 64 ? floor(src0) : floorf(src0)); _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fge(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); const float src1 = _mesa_half_to_float(_src[1].u16[_i]); bool32_t dst = src0 >= src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; const float32_t src1 = _src[1].f32[_i]; bool32_t dst = src0 >= src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; const float64_t src1 = _src[1].f64[_i]; bool32_t dst = src0 >= src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_find_lsb(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; int32_t dst; dst = -1; for (unsigned bit = 0; bit < bit_size; bit++) { if ((src0 >> bit) & 1) { dst = bit; break; } } _dst_val.i32[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; int32_t dst; dst = -1; for (unsigned bit = 0; bit < bit_size; bit++) { if ((src0 >> bit) & 1) { dst = bit; break; } } _dst_val.i32[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; int32_t dst; dst = -1; for (unsigned bit = 0; bit < bit_size; bit++) { if ((src0 >> bit) & 1) { dst = bit; break; } } _dst_val.i32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; int32_t dst; dst = -1; for (unsigned bit = 0; bit < bit_size; bit++) { if ((src0 >> bit) & 1) { dst = bit; break; } } _dst_val.i32[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_flog2(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); float16_t dst = log2f(src0); _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; float32_t dst = log2f(src0); _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; float64_t dst = log2f(src0); _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_flrp(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); const float src1 = _mesa_half_to_float(_src[1].u16[_i]); const float src2 = _mesa_half_to_float(_src[2].u16[_i]); float16_t dst = src0 * (1 - src2) + src1 * src2; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; const float32_t src1 = _src[1].f32[_i]; const float32_t src2 = _src[2].f32[_i]; float32_t dst = src0 * (1 - src2) + src1 * src2; _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; const float64_t src1 = _src[1].f64[_i]; const float64_t src2 = _src[2].f64[_i]; float64_t dst = src0 * (1 - src2) + src1 * src2; _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_flt(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); const float src1 = _mesa_half_to_float(_src[1].u16[_i]); bool32_t dst = src0 < src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; const float32_t src1 = _src[1].f32[_i]; bool32_t dst = src0 < src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; const float64_t src1 = _src[1].f64[_i]; bool32_t dst = src0 < src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fmax(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); const float src1 = _mesa_half_to_float(_src[1].u16[_i]); float16_t dst = fmaxf(src0, src1); _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; const float32_t src1 = _src[1].f32[_i]; float32_t dst = fmaxf(src0, src1); _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; const float64_t src1 = _src[1].f64[_i]; float64_t dst = fmaxf(src0, src1); _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fmax3(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); const float src1 = _mesa_half_to_float(_src[1].u16[_i]); const float src2 = _mesa_half_to_float(_src[2].u16[_i]); float16_t dst = fmaxf(src0, fmaxf(src1, src2)); _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; const float32_t src1 = _src[1].f32[_i]; const float32_t src2 = _src[2].f32[_i]; float32_t dst = fmaxf(src0, fmaxf(src1, src2)); _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; const float64_t src1 = _src[1].f64[_i]; const float64_t src2 = _src[2].f64[_i]; float64_t dst = fmaxf(src0, fmaxf(src1, src2)); _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fmed3(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); const float src1 = _mesa_half_to_float(_src[1].u16[_i]); const float src2 = _mesa_half_to_float(_src[2].u16[_i]); float16_t dst = fmaxf(fminf(fmaxf(src0, src1), src2), fminf(src0, src1)); _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; const float32_t src1 = _src[1].f32[_i]; const float32_t src2 = _src[2].f32[_i]; float32_t dst = fmaxf(fminf(fmaxf(src0, src1), src2), fminf(src0, src1)); _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; const float64_t src1 = _src[1].f64[_i]; const float64_t src2 = _src[2].f64[_i]; float64_t dst = fmaxf(fminf(fmaxf(src0, src1), src2), fminf(src0, src1)); _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fmin(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); const float src1 = _mesa_half_to_float(_src[1].u16[_i]); float16_t dst = fminf(src0, src1); _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; const float32_t src1 = _src[1].f32[_i]; float32_t dst = fminf(src0, src1); _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; const float64_t src1 = _src[1].f64[_i]; float64_t dst = fminf(src0, src1); _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fmin3(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); const float src1 = _mesa_half_to_float(_src[1].u16[_i]); const float src2 = _mesa_half_to_float(_src[2].u16[_i]); float16_t dst = fminf(src0, fminf(src1, src2)); _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; const float32_t src1 = _src[1].f32[_i]; const float32_t src2 = _src[2].f32[_i]; float32_t dst = fminf(src0, fminf(src1, src2)); _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; const float64_t src1 = _src[1].f64[_i]; const float64_t src2 = _src[2].f64[_i]; float64_t dst = fminf(src0, fminf(src1, src2)); _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fmod(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); const float src1 = _mesa_half_to_float(_src[1].u16[_i]); float16_t dst = src0 - src1 * floorf(src0 / src1); _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; const float32_t src1 = _src[1].f32[_i]; float32_t dst = src0 - src1 * floorf(src0 / src1); _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; const float64_t src1 = _src[1].f64[_i]; float64_t dst = src0 - src1 * floorf(src0 / src1); _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fmov(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); float16_t dst = src0; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; float32_t dst = src0; _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; float64_t dst = src0; _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fmul(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); const float src1 = _mesa_half_to_float(_src[1].u16[_i]); float16_t dst = src0 * src1; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; const float32_t src1 = _src[1].f32[_i]; float32_t dst = src0 * src1; _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; const float64_t src1 = _src[1].f64[_i]; float64_t dst = src0 * src1; _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fne(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); const float src1 = _mesa_half_to_float(_src[1].u16[_i]); bool32_t dst = src0 != src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; const float32_t src1 = _src[1].f32[_i]; bool32_t dst = src0 != src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; const float64_t src1 = _src[1].f64[_i]; bool32_t dst = src0 != src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fneg(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); float16_t dst = -src0; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; float32_t dst = -src0; _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; float64_t dst = -src0; _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fnoise1_1(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { struct float16_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.u16[0] = _mesa_float_to_half(dst.x); break; } case 32: { struct float32_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f32[0] = dst.x; break; } case 64: { struct float64_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f64[0] = dst.x; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fnoise1_2(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { struct float16_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.u16[0] = _mesa_float_to_half(dst.x); break; } case 32: { struct float32_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f32[0] = dst.x; break; } case 64: { struct float64_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f64[0] = dst.x; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fnoise1_3(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { struct float16_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.u16[0] = _mesa_float_to_half(dst.x); break; } case 32: { struct float32_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f32[0] = dst.x; break; } case 64: { struct float64_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f64[0] = dst.x; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fnoise1_4(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { struct float16_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.u16[0] = _mesa_float_to_half(dst.x); break; } case 32: { struct float32_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f32[0] = dst.x; break; } case 64: { struct float64_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f64[0] = dst.x; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fnoise2_1(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { struct float16_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.u16[0] = _mesa_float_to_half(dst.x); _dst_val.u16[1] = _mesa_float_to_half(dst.y); break; } case 32: { struct float32_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f32[0] = dst.x; _dst_val.f32[1] = dst.y; break; } case 64: { struct float64_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f64[0] = dst.x; _dst_val.f64[1] = dst.y; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fnoise2_2(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { struct float16_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.u16[0] = _mesa_float_to_half(dst.x); _dst_val.u16[1] = _mesa_float_to_half(dst.y); break; } case 32: { struct float32_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f32[0] = dst.x; _dst_val.f32[1] = dst.y; break; } case 64: { struct float64_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f64[0] = dst.x; _dst_val.f64[1] = dst.y; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fnoise2_3(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { struct float16_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.u16[0] = _mesa_float_to_half(dst.x); _dst_val.u16[1] = _mesa_float_to_half(dst.y); break; } case 32: { struct float32_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f32[0] = dst.x; _dst_val.f32[1] = dst.y; break; } case 64: { struct float64_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f64[0] = dst.x; _dst_val.f64[1] = dst.y; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fnoise2_4(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { struct float16_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.u16[0] = _mesa_float_to_half(dst.x); _dst_val.u16[1] = _mesa_float_to_half(dst.y); break; } case 32: { struct float32_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f32[0] = dst.x; _dst_val.f32[1] = dst.y; break; } case 64: { struct float64_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f64[0] = dst.x; _dst_val.f64[1] = dst.y; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fnoise3_1(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { struct float16_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.u16[0] = _mesa_float_to_half(dst.x); _dst_val.u16[1] = _mesa_float_to_half(dst.y); _dst_val.u16[2] = _mesa_float_to_half(dst.z); break; } case 32: { struct float32_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f32[0] = dst.x; _dst_val.f32[1] = dst.y; _dst_val.f32[2] = dst.z; break; } case 64: { struct float64_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f64[0] = dst.x; _dst_val.f64[1] = dst.y; _dst_val.f64[2] = dst.z; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fnoise3_2(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { struct float16_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.u16[0] = _mesa_float_to_half(dst.x); _dst_val.u16[1] = _mesa_float_to_half(dst.y); _dst_val.u16[2] = _mesa_float_to_half(dst.z); break; } case 32: { struct float32_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f32[0] = dst.x; _dst_val.f32[1] = dst.y; _dst_val.f32[2] = dst.z; break; } case 64: { struct float64_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f64[0] = dst.x; _dst_val.f64[1] = dst.y; _dst_val.f64[2] = dst.z; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fnoise3_3(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { struct float16_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.u16[0] = _mesa_float_to_half(dst.x); _dst_val.u16[1] = _mesa_float_to_half(dst.y); _dst_val.u16[2] = _mesa_float_to_half(dst.z); break; } case 32: { struct float32_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f32[0] = dst.x; _dst_val.f32[1] = dst.y; _dst_val.f32[2] = dst.z; break; } case 64: { struct float64_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f64[0] = dst.x; _dst_val.f64[1] = dst.y; _dst_val.f64[2] = dst.z; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fnoise3_4(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { struct float16_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.u16[0] = _mesa_float_to_half(dst.x); _dst_val.u16[1] = _mesa_float_to_half(dst.y); _dst_val.u16[2] = _mesa_float_to_half(dst.z); break; } case 32: { struct float32_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f32[0] = dst.x; _dst_val.f32[1] = dst.y; _dst_val.f32[2] = dst.z; break; } case 64: { struct float64_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f64[0] = dst.x; _dst_val.f64[1] = dst.y; _dst_val.f64[2] = dst.z; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fnoise4_1(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { struct float16_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.u16[0] = _mesa_float_to_half(dst.x); _dst_val.u16[1] = _mesa_float_to_half(dst.y); _dst_val.u16[2] = _mesa_float_to_half(dst.z); _dst_val.u16[3] = _mesa_float_to_half(dst.w); break; } case 32: { struct float32_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f32[0] = dst.x; _dst_val.f32[1] = dst.y; _dst_val.f32[2] = dst.z; _dst_val.f32[3] = dst.w; break; } case 64: { struct float64_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f64[0] = dst.x; _dst_val.f64[1] = dst.y; _dst_val.f64[2] = dst.z; _dst_val.f64[3] = dst.w; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fnoise4_2(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { struct float16_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.u16[0] = _mesa_float_to_half(dst.x); _dst_val.u16[1] = _mesa_float_to_half(dst.y); _dst_val.u16[2] = _mesa_float_to_half(dst.z); _dst_val.u16[3] = _mesa_float_to_half(dst.w); break; } case 32: { struct float32_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f32[0] = dst.x; _dst_val.f32[1] = dst.y; _dst_val.f32[2] = dst.z; _dst_val.f32[3] = dst.w; break; } case 64: { struct float64_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f64[0] = dst.x; _dst_val.f64[1] = dst.y; _dst_val.f64[2] = dst.z; _dst_val.f64[3] = dst.w; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fnoise4_3(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { struct float16_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.u16[0] = _mesa_float_to_half(dst.x); _dst_val.u16[1] = _mesa_float_to_half(dst.y); _dst_val.u16[2] = _mesa_float_to_half(dst.z); _dst_val.u16[3] = _mesa_float_to_half(dst.w); break; } case 32: { struct float32_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f32[0] = dst.x; _dst_val.f32[1] = dst.y; _dst_val.f32[2] = dst.z; _dst_val.f32[3] = dst.w; break; } case 64: { struct float64_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f64[0] = dst.x; _dst_val.f64[1] = dst.y; _dst_val.f64[2] = dst.z; _dst_val.f64[3] = dst.w; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fnoise4_4(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { struct float16_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.u16[0] = _mesa_float_to_half(dst.x); _dst_val.u16[1] = _mesa_float_to_half(dst.y); _dst_val.u16[2] = _mesa_float_to_half(dst.z); _dst_val.u16[3] = _mesa_float_to_half(dst.w); break; } case 32: { struct float32_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f32[0] = dst.x; _dst_val.f32[1] = dst.y; _dst_val.f32[2] = dst.z; _dst_val.f32[3] = dst.w; break; } case 64: { struct float64_vec dst; dst.x = dst.y = dst.z = dst.w = 0.0f; _dst_val.f64[0] = dst.x; _dst_val.f64[1] = dst.y; _dst_val.f64[2] = dst.z; _dst_val.f64[3] = dst.w; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fnot(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); float16_t dst = bit_size == 64 ? ((src0 == 0.0) ? 1.0 : 0.0f) : ((src0 == 0.0f) ? 1.0f : 0.0f); _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; float32_t dst = bit_size == 64 ? ((src0 == 0.0) ? 1.0 : 0.0f) : ((src0 == 0.0f) ? 1.0f : 0.0f); _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; float64_t dst = bit_size == 64 ? ((src0 == 0.0) ? 1.0 : 0.0f) : ((src0 == 0.0f) ? 1.0f : 0.0f); _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_for(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; const float32_t src1 = _src[1].f32[_i]; float32_t dst = ((src0 != 0.0f) || (src1 != 0.0f)) ? 1.0f : 0.0f; _dst_val.f32[_i] = dst; } return _dst_val; } static nir_const_value evaluate_fpow(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); const float src1 = _mesa_half_to_float(_src[1].u16[_i]); float16_t dst = bit_size == 64 ? powf(src0, src1) : pow(src0, src1); _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; const float32_t src1 = _src[1].f32[_i]; float32_t dst = bit_size == 64 ? powf(src0, src1) : pow(src0, src1); _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; const float64_t src1 = _src[1].f64[_i]; float64_t dst = bit_size == 64 ? powf(src0, src1) : pow(src0, src1); _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fquantize2f16(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); float16_t dst = (fabs(src0) < ldexpf(1.0, -14)) ? copysignf(0.0f, src0) : _mesa_half_to_float(_mesa_float_to_half(src0)); _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; float32_t dst = (fabs(src0) < ldexpf(1.0, -14)) ? copysignf(0.0f, src0) : _mesa_half_to_float(_mesa_float_to_half(src0)); _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; float64_t dst = (fabs(src0) < ldexpf(1.0, -14)) ? copysignf(0.0f, src0) : _mesa_half_to_float(_mesa_float_to_half(src0)); _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_frcp(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); float16_t dst = bit_size == 64 ? 1.0 / src0 : 1.0f / src0; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; float32_t dst = bit_size == 64 ? 1.0 / src0 : 1.0f / src0; _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; float64_t dst = bit_size == 64 ? 1.0 / src0 : 1.0f / src0; _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_frem(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); const float src1 = _mesa_half_to_float(_src[1].u16[_i]); float16_t dst = src0 - src1 * truncf(src0 / src1); _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; const float32_t src1 = _src[1].f32[_i]; float32_t dst = src0 - src1 * truncf(src0 / src1); _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; const float64_t src1 = _src[1].f64[_i]; float64_t dst = src0 - src1 * truncf(src0 / src1); _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_frexp_exp(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; int32_t dst; frexp(src0, &dst); _dst_val.i32[_i] = dst; } return _dst_val; } static nir_const_value evaluate_frexp_sig(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; float64_t dst; int n; dst = frexp(src0, &n); _dst_val.f64[_i] = dst; } return _dst_val; } static nir_const_value evaluate_fround_even(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); float16_t dst = bit_size == 64 ? _mesa_roundeven(src0) : _mesa_roundevenf(src0); _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; float32_t dst = bit_size == 64 ? _mesa_roundeven(src0) : _mesa_roundevenf(src0); _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; float64_t dst = bit_size == 64 ? _mesa_roundeven(src0) : _mesa_roundevenf(src0); _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_frsq(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); float16_t dst = bit_size == 64 ? 1.0 / sqrt(src0) : 1.0f / sqrtf(src0); _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; float32_t dst = bit_size == 64 ? 1.0 / sqrt(src0) : 1.0f / sqrtf(src0); _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; float64_t dst = bit_size == 64 ? 1.0 / sqrt(src0) : 1.0f / sqrtf(src0); _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fsat(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); float16_t dst = bit_size == 64 ? ((src0 > 1.0) ? 1.0 : ((src0 <= 0.0) ? 0.0 : src0)) : ((src0 > 1.0f) ? 1.0f : ((src0 <= 0.0f) ? 0.0f : src0)); _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; float32_t dst = bit_size == 64 ? ((src0 > 1.0) ? 1.0 : ((src0 <= 0.0) ? 0.0 : src0)) : ((src0 > 1.0f) ? 1.0f : ((src0 <= 0.0f) ? 0.0f : src0)); _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; float64_t dst = bit_size == 64 ? ((src0 > 1.0) ? 1.0 : ((src0 <= 0.0) ? 0.0 : src0)) : ((src0 > 1.0f) ? 1.0f : ((src0 <= 0.0f) ? 0.0f : src0)); _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fsign(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); float16_t dst = bit_size == 64 ? ((src0 == 0.0) ? 0.0 : ((src0 > 0.0) ? 1.0 : -1.0)) : ((src0 == 0.0f) ? 0.0f : ((src0 > 0.0f) ? 1.0f : -1.0f)); _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; float32_t dst = bit_size == 64 ? ((src0 == 0.0) ? 0.0 : ((src0 > 0.0) ? 1.0 : -1.0)) : ((src0 == 0.0f) ? 0.0f : ((src0 > 0.0f) ? 1.0f : -1.0f)); _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; float64_t dst = bit_size == 64 ? ((src0 == 0.0) ? 0.0 : ((src0 > 0.0) ? 1.0 : -1.0)) : ((src0 == 0.0f) ? 0.0f : ((src0 > 0.0f) ? 1.0f : -1.0f)); _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fsin(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); float16_t dst = bit_size == 64 ? sin(src0) : sinf(src0); _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; float32_t dst = bit_size == 64 ? sin(src0) : sinf(src0); _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; float64_t dst = bit_size == 64 ? sin(src0) : sinf(src0); _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fsqrt(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); float16_t dst = bit_size == 64 ? sqrt(src0) : sqrtf(src0); _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; float32_t dst = bit_size == 64 ? sqrt(src0) : sqrtf(src0); _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; float64_t dst = bit_size == 64 ? sqrt(src0) : sqrtf(src0); _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fsub(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); const float src1 = _mesa_half_to_float(_src[1].u16[_i]); float16_t dst = src0 - src1; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; const float32_t src1 = _src[1].f32[_i]; float32_t dst = src0 - src1; _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; const float64_t src1 = _src[1].f64[_i]; float64_t dst = src0 - src1; _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_ftrunc(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); float16_t dst = bit_size == 64 ? trunc(src0) : truncf(src0); _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; float32_t dst = bit_size == 64 ? trunc(src0) : truncf(src0); _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; float64_t dst = bit_size == 64 ? trunc(src0) : truncf(src0); _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_fxor(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; const float32_t src1 = _src[1].f32[_i]; float32_t dst = (src0 != 0.0f && src1 == 0.0f) || (src0 == 0.0f && src1 != 0.0f) ? 1.0f : 0.0f; _dst_val.f32[_i] = dst; } return _dst_val; } static nir_const_value evaluate_i2b(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; bool32_t dst = src0 != 0; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; bool32_t dst = src0 != 0; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; bool32_t dst = src0 != 0; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; bool32_t dst = src0 != 0; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_i2f16(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; float16_t dst = src0; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; float16_t dst = src0; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; float16_t dst = src0; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; float16_t dst = src0; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_i2f32(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; float32_t dst = src0; _dst_val.f32[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; float32_t dst = src0; _dst_val.f32[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; float32_t dst = src0; _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; float32_t dst = src0; _dst_val.f32[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_i2f64(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; float64_t dst = src0; _dst_val.f64[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; float64_t dst = src0; _dst_val.f64[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; float64_t dst = src0; _dst_val.f64[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; float64_t dst = src0; _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_i2i16(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; int16_t dst = src0; _dst_val.i16[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; int16_t dst = src0; _dst_val.i16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; int16_t dst = src0; _dst_val.i16[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; int16_t dst = src0; _dst_val.i16[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_i2i32(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; int32_t dst = src0; _dst_val.i32[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; int32_t dst = src0; _dst_val.i32[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; int32_t dst = src0; _dst_val.i32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; int32_t dst = src0; _dst_val.i32[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_i2i64(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; int64_t dst = src0; _dst_val.i64[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; int64_t dst = src0; _dst_val.i64[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; int64_t dst = src0; _dst_val.i64[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; int64_t dst = src0; _dst_val.i64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_i2i8(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; int8_t dst = src0; _dst_val.i8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; int8_t dst = src0; _dst_val.i8[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; int8_t dst = src0; _dst_val.i8[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; int8_t dst = src0; _dst_val.i8[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_iabs(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; int8_t dst = (src0 < 0) ? -src0 : src0; _dst_val.i8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; int16_t dst = (src0 < 0) ? -src0 : src0; _dst_val.i16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; int32_t dst = (src0 < 0) ? -src0 : src0; _dst_val.i32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; int64_t dst = (src0 < 0) ? -src0 : src0; _dst_val.i64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_iadd(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; const int8_t src1 = _src[1].i8[_i]; int8_t dst = src0 + src1; _dst_val.i8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; const int16_t src1 = _src[1].i16[_i]; int16_t dst = src0 + src1; _dst_val.i16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; const int32_t src1 = _src[1].i32[_i]; int32_t dst = src0 + src1; _dst_val.i32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; const int64_t src1 = _src[1].i64[_i]; int64_t dst = src0 + src1; _dst_val.i64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_iand(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const uint8_t src0 = _src[0].u8[_i]; const uint8_t src1 = _src[1].u8[_i]; uint8_t dst = src0 & src1; _dst_val.u8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const uint16_t src0 = _src[0].u16[_i]; const uint16_t src1 = _src[1].u16[_i]; uint16_t dst = src0 & src1; _dst_val.u16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; const uint32_t src1 = _src[1].u32[_i]; uint32_t dst = src0 & src1; _dst_val.u32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const uint64_t src0 = _src[0].u64[_i]; const uint64_t src1 = _src[1].u64[_i]; uint64_t dst = src0 & src1; _dst_val.u64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_ibfe(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; const int32_t src1 = _src[1].i32[_i]; const int32_t src2 = _src[2].i32[_i]; int32_t dst; int base = src0; int offset = src1, bits = src2; if (bits == 0) { dst = 0; } else if (bits < 0 || offset < 0) { dst = 0; /* undefined */ } else if (offset + bits < 32) { dst = (base << (32 - bits - offset)) >> (32 - bits); } else { dst = base >> offset; } _dst_val.i32[_i] = dst; } return _dst_val; } static nir_const_value evaluate_ibitfield_extract(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; const int32_t src1 = _src[1].i32[_i]; const int32_t src2 = _src[2].i32[_i]; int32_t dst; int base = src0; int offset = src1, bits = src2; if (bits == 0) { dst = 0; } else if (offset < 0 || bits < 0 || offset + bits > 32) { dst = 0; } else { dst = (base << (32 - offset - bits)) >> offset; /* use sign-extending shift */ } _dst_val.i32[_i] = dst; } return _dst_val; } static nir_const_value evaluate_idiv(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; const int8_t src1 = _src[1].i8[_i]; int8_t dst = src1 == 0 ? 0 : (src0 / src1); _dst_val.i8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; const int16_t src1 = _src[1].i16[_i]; int16_t dst = src1 == 0 ? 0 : (src0 / src1); _dst_val.i16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; const int32_t src1 = _src[1].i32[_i]; int32_t dst = src1 == 0 ? 0 : (src0 / src1); _dst_val.i32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; const int64_t src1 = _src[1].i64[_i]; int64_t dst = src1 == 0 ? 0 : (src0 / src1); _dst_val.i64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_ieq(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; const int8_t src1 = _src[1].i8[_i]; bool32_t dst = src0 == src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; const int16_t src1 = _src[1].i16[_i]; bool32_t dst = src0 == src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; const int32_t src1 = _src[1].i32[_i]; bool32_t dst = src0 == src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; const int64_t src1 = _src[1].i64[_i]; bool32_t dst = src0 == src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_ifind_msb(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; int32_t dst; dst = -1; for (int bit = 31; bit >= 0; bit--) { /* If src0 < 0, we're looking for the first 0 bit. * if src0 >= 0, we're looking for the first 1 bit. */ if ((((src0 >> bit) & 1) && (src0 >= 0)) || (!((src0 >> bit) & 1) && (src0 < 0))) { dst = bit; break; } } _dst_val.i32[_i] = dst; } return _dst_val; } static nir_const_value evaluate_ige(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; const int8_t src1 = _src[1].i8[_i]; bool32_t dst = src0 >= src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; const int16_t src1 = _src[1].i16[_i]; bool32_t dst = src0 >= src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; const int32_t src1 = _src[1].i32[_i]; bool32_t dst = src0 >= src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; const int64_t src1 = _src[1].i64[_i]; bool32_t dst = src0 >= src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_ilt(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; const int8_t src1 = _src[1].i8[_i]; bool32_t dst = src0 < src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; const int16_t src1 = _src[1].i16[_i]; bool32_t dst = src0 < src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; const int32_t src1 = _src[1].i32[_i]; bool32_t dst = src0 < src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; const int64_t src1 = _src[1].i64[_i]; bool32_t dst = src0 < src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_imax(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; const int8_t src1 = _src[1].i8[_i]; int8_t dst = src1 > src0 ? src1 : src0; _dst_val.i8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; const int16_t src1 = _src[1].i16[_i]; int16_t dst = src1 > src0 ? src1 : src0; _dst_val.i16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; const int32_t src1 = _src[1].i32[_i]; int32_t dst = src1 > src0 ? src1 : src0; _dst_val.i32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; const int64_t src1 = _src[1].i64[_i]; int64_t dst = src1 > src0 ? src1 : src0; _dst_val.i64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_imax3(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; const int8_t src1 = _src[1].i8[_i]; const int8_t src2 = _src[2].i8[_i]; int8_t dst = MAX2(src0, MAX2(src1, src2)); _dst_val.i8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; const int16_t src1 = _src[1].i16[_i]; const int16_t src2 = _src[2].i16[_i]; int16_t dst = MAX2(src0, MAX2(src1, src2)); _dst_val.i16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; const int32_t src1 = _src[1].i32[_i]; const int32_t src2 = _src[2].i32[_i]; int32_t dst = MAX2(src0, MAX2(src1, src2)); _dst_val.i32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; const int64_t src1 = _src[1].i64[_i]; const int64_t src2 = _src[2].i64[_i]; int64_t dst = MAX2(src0, MAX2(src1, src2)); _dst_val.i64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_imed3(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; const int8_t src1 = _src[1].i8[_i]; const int8_t src2 = _src[2].i8[_i]; int8_t dst = MAX2(MIN2(MAX2(src0, src1), src2), MIN2(src0, src1)); _dst_val.i8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; const int16_t src1 = _src[1].i16[_i]; const int16_t src2 = _src[2].i16[_i]; int16_t dst = MAX2(MIN2(MAX2(src0, src1), src2), MIN2(src0, src1)); _dst_val.i16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; const int32_t src1 = _src[1].i32[_i]; const int32_t src2 = _src[2].i32[_i]; int32_t dst = MAX2(MIN2(MAX2(src0, src1), src2), MIN2(src0, src1)); _dst_val.i32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; const int64_t src1 = _src[1].i64[_i]; const int64_t src2 = _src[2].i64[_i]; int64_t dst = MAX2(MIN2(MAX2(src0, src1), src2), MIN2(src0, src1)); _dst_val.i64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_imin(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; const int8_t src1 = _src[1].i8[_i]; int8_t dst = src1 > src0 ? src0 : src1; _dst_val.i8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; const int16_t src1 = _src[1].i16[_i]; int16_t dst = src1 > src0 ? src0 : src1; _dst_val.i16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; const int32_t src1 = _src[1].i32[_i]; int32_t dst = src1 > src0 ? src0 : src1; _dst_val.i32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; const int64_t src1 = _src[1].i64[_i]; int64_t dst = src1 > src0 ? src0 : src1; _dst_val.i64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_imin3(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; const int8_t src1 = _src[1].i8[_i]; const int8_t src2 = _src[2].i8[_i]; int8_t dst = MIN2(src0, MIN2(src1, src2)); _dst_val.i8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; const int16_t src1 = _src[1].i16[_i]; const int16_t src2 = _src[2].i16[_i]; int16_t dst = MIN2(src0, MIN2(src1, src2)); _dst_val.i16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; const int32_t src1 = _src[1].i32[_i]; const int32_t src2 = _src[2].i32[_i]; int32_t dst = MIN2(src0, MIN2(src1, src2)); _dst_val.i32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; const int64_t src1 = _src[1].i64[_i]; const int64_t src2 = _src[2].i64[_i]; int64_t dst = MIN2(src0, MIN2(src1, src2)); _dst_val.i64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_imod(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; const int8_t src1 = _src[1].i8[_i]; int8_t dst = src1 == 0 ? 0 : ((src0 % src1 == 0 || (src0 >= 0) == (src1 >= 0)) ? src0 % src1 : src0 % src1 + src1); _dst_val.i8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; const int16_t src1 = _src[1].i16[_i]; int16_t dst = src1 == 0 ? 0 : ((src0 % src1 == 0 || (src0 >= 0) == (src1 >= 0)) ? src0 % src1 : src0 % src1 + src1); _dst_val.i16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; const int32_t src1 = _src[1].i32[_i]; int32_t dst = src1 == 0 ? 0 : ((src0 % src1 == 0 || (src0 >= 0) == (src1 >= 0)) ? src0 % src1 : src0 % src1 + src1); _dst_val.i32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; const int64_t src1 = _src[1].i64[_i]; int64_t dst = src1 == 0 ? 0 : ((src0 % src1 == 0 || (src0 >= 0) == (src1 >= 0)) ? src0 % src1 : src0 % src1 + src1); _dst_val.i64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_imov(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; int8_t dst = src0; _dst_val.i8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; int16_t dst = src0; _dst_val.i16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; int32_t dst = src0; _dst_val.i32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; int64_t dst = src0; _dst_val.i64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_imul(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; const int8_t src1 = _src[1].i8[_i]; int8_t dst = src0 * src1; _dst_val.i8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; const int16_t src1 = _src[1].i16[_i]; int16_t dst = src0 * src1; _dst_val.i16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; const int32_t src1 = _src[1].i32[_i]; int32_t dst = src0 * src1; _dst_val.i32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; const int64_t src1 = _src[1].i64[_i]; int64_t dst = src0 * src1; _dst_val.i64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_imul_high(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; const int32_t src1 = _src[1].i32[_i]; int32_t dst = (int32_t)(((int64_t) src0 * (int64_t) src1) >> 32); _dst_val.i32[_i] = dst; } return _dst_val; } static nir_const_value evaluate_ine(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; const int8_t src1 = _src[1].i8[_i]; bool32_t dst = src0 != src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; const int16_t src1 = _src[1].i16[_i]; bool32_t dst = src0 != src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; const int32_t src1 = _src[1].i32[_i]; bool32_t dst = src0 != src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; const int64_t src1 = _src[1].i64[_i]; bool32_t dst = src0 != src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_ineg(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; int8_t dst = -src0; _dst_val.i8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; int16_t dst = -src0; _dst_val.i16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; int32_t dst = -src0; _dst_val.i32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; int64_t dst = -src0; _dst_val.i64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_inot(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; int8_t dst = ~src0; _dst_val.i8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; int16_t dst = ~src0; _dst_val.i16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; int32_t dst = ~src0; _dst_val.i32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; int64_t dst = ~src0; _dst_val.i64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_ior(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const uint8_t src0 = _src[0].u8[_i]; const uint8_t src1 = _src[1].u8[_i]; uint8_t dst = src0 | src1; _dst_val.u8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const uint16_t src0 = _src[0].u16[_i]; const uint16_t src1 = _src[1].u16[_i]; uint16_t dst = src0 | src1; _dst_val.u16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; const uint32_t src1 = _src[1].u32[_i]; uint32_t dst = src0 | src1; _dst_val.u32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const uint64_t src0 = _src[0].u64[_i]; const uint64_t src1 = _src[1].u64[_i]; uint64_t dst = src0 | src1; _dst_val.u64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_irem(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; const int8_t src1 = _src[1].i8[_i]; int8_t dst = src1 == 0 ? 0 : src0 % src1; _dst_val.i8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; const int16_t src1 = _src[1].i16[_i]; int16_t dst = src1 == 0 ? 0 : src0 % src1; _dst_val.i16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; const int32_t src1 = _src[1].i32[_i]; int32_t dst = src1 == 0 ? 0 : src0 % src1; _dst_val.i32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; const int64_t src1 = _src[1].i64[_i]; int64_t dst = src1 == 0 ? 0 : src0 % src1; _dst_val.i64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_ishl(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; const uint32_t src1 = _src[1].u32[_i]; int8_t dst = src0 << src1; _dst_val.i8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; const uint32_t src1 = _src[1].u32[_i]; int16_t dst = src0 << src1; _dst_val.i16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; const uint32_t src1 = _src[1].u32[_i]; int32_t dst = src0 << src1; _dst_val.i32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; const uint32_t src1 = _src[1].u32[_i]; int64_t dst = src0 << src1; _dst_val.i64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_ishr(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; const uint32_t src1 = _src[1].u32[_i]; int8_t dst = src0 >> src1; _dst_val.i8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; const uint32_t src1 = _src[1].u32[_i]; int16_t dst = src0 >> src1; _dst_val.i16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; const uint32_t src1 = _src[1].u32[_i]; int32_t dst = src0 >> src1; _dst_val.i32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; const uint32_t src1 = _src[1].u32[_i]; int64_t dst = src0 >> src1; _dst_val.i64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_isign(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; int8_t dst = (src0 == 0) ? 0 : ((src0 > 0) ? 1 : -1); _dst_val.i8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; int16_t dst = (src0 == 0) ? 0 : ((src0 > 0) ? 1 : -1); _dst_val.i16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; int32_t dst = (src0 == 0) ? 0 : ((src0 > 0) ? 1 : -1); _dst_val.i32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; int64_t dst = (src0 == 0) ? 0 : ((src0 > 0) ? 1 : -1); _dst_val.i64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_isub(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const int8_t src0 = _src[0].i8[_i]; const int8_t src1 = _src[1].i8[_i]; int8_t dst = src0 - src1; _dst_val.i8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const int16_t src0 = _src[0].i16[_i]; const int16_t src1 = _src[1].i16[_i]; int16_t dst = src0 - src1; _dst_val.i16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; const int32_t src1 = _src[1].i32[_i]; int32_t dst = src0 - src1; _dst_val.i32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const int64_t src0 = _src[0].i64[_i]; const int64_t src1 = _src[1].i64[_i]; int64_t dst = src0 - src1; _dst_val.i64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_ixor(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const uint8_t src0 = _src[0].u8[_i]; const uint8_t src1 = _src[1].u8[_i]; uint8_t dst = src0 ^ src1; _dst_val.u8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const uint16_t src0 = _src[0].u16[_i]; const uint16_t src1 = _src[1].u16[_i]; uint16_t dst = src0 ^ src1; _dst_val.u16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; const uint32_t src1 = _src[1].u32[_i]; uint32_t dst = src0 ^ src1; _dst_val.u32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const uint64_t src0 = _src[0].u64[_i]; const uint64_t src1 = _src[1].u64[_i]; uint64_t dst = src0 ^ src1; _dst_val.u64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_ldexp(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); const int32_t src1 = _src[1].i32[_i]; float16_t dst; dst = (bit_size == 64) ? ldexp(src0, src1) : ldexpf(src0, src1); /* flush denormals to zero. */ if (!isnormal(dst)) dst = copysignf(0.0f, src0); _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; const int32_t src1 = _src[1].i32[_i]; float32_t dst; dst = (bit_size == 64) ? ldexp(src0, src1) : ldexpf(src0, src1); /* flush denormals to zero. */ if (!isnormal(dst)) dst = copysignf(0.0f, src0); _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; const int32_t src1 = _src[1].i32[_i]; float64_t dst; dst = (bit_size == 64) ? ldexp(src0, src1) : ldexpf(src0, src1); /* flush denormals to zero. */ if (!isnormal(dst)) dst = copysignf(0.0f, src0); _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_pack_32_2x16(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; const struct uint16_vec src0 = { _src[0].u16[0], _src[0].u16[1], 0, 0, }; struct uint32_vec dst; dst.x = src0.x | ((uint32_t)src0.y << 16); _dst_val.u32[0] = dst.x; return _dst_val; } static nir_const_value evaluate_pack_32_2x16_split(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const uint16_t src0 = _src[0].u16[_i]; const uint16_t src1 = _src[1].u16[_i]; uint32_t dst = src0 | ((uint32_t)src1 << 16); _dst_val.u32[_i] = dst; } return _dst_val; } static nir_const_value evaluate_pack_64_2x32(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; const struct uint32_vec src0 = { _src[0].u32[0], _src[0].u32[1], 0, 0, }; struct uint64_vec dst; dst.x = src0.x | ((uint64_t)src0.y << 32); _dst_val.u64[0] = dst.x; return _dst_val; } static nir_const_value evaluate_pack_64_2x32_split(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; const uint32_t src1 = _src[1].u32[_i]; uint64_t dst = src0 | ((uint64_t)src1 << 32); _dst_val.u64[_i] = dst; } return _dst_val; } static nir_const_value evaluate_pack_64_4x16(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; const struct uint16_vec src0 = { _src[0].u16[0], _src[0].u16[1], _src[0].u16[2], _src[0].u16[3], }; struct uint64_vec dst; dst.x = src0.x | ((uint64_t)src0.y << 16) | ((uint64_t)src0.z << 32) | ((uint64_t)src0.w << 48); _dst_val.u64[0] = dst.x; return _dst_val; } static nir_const_value evaluate_pack_half_2x16(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; const struct float32_vec src0 = { _src[0].f32[0], _src[0].f32[1], 0, 0, }; struct uint32_vec dst; dst.x = (uint32_t) pack_half_1x16(src0.x); dst.x |= ((uint32_t) pack_half_1x16(src0.y)) << 16; _dst_val.u32[0] = dst.x; return _dst_val; } static nir_const_value evaluate_pack_half_2x16_split(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; const struct float32_vec src0 = { _src[0].f32[0], 0, 0, 0, }; const struct float32_vec src1 = { _src[1].f32[0], 0, 0, 0, }; struct uint32_vec dst; dst.x = dst.y = dst.z = dst.w = pack_half_1x16(src0.x) | (pack_half_1x16(src1.x) << 16); _dst_val.u32[0] = dst.x; return _dst_val; } static nir_const_value evaluate_pack_snorm_2x16(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; const struct float32_vec src0 = { _src[0].f32[0], _src[0].f32[1], 0, 0, }; struct uint32_vec dst; dst.x = (uint32_t) pack_snorm_1x16(src0.x); dst.x |= ((uint32_t) pack_snorm_1x16(src0.y)) << 16; _dst_val.u32[0] = dst.x; return _dst_val; } static nir_const_value evaluate_pack_snorm_4x8(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; const struct float32_vec src0 = { _src[0].f32[0], _src[0].f32[1], _src[0].f32[2], _src[0].f32[3], }; struct uint32_vec dst; dst.x = (uint32_t) pack_snorm_1x8(src0.x); dst.x |= ((uint32_t) pack_snorm_1x8(src0.y)) << 8; dst.x |= ((uint32_t) pack_snorm_1x8(src0.z)) << 16; dst.x |= ((uint32_t) pack_snorm_1x8(src0.w)) << 24; _dst_val.u32[0] = dst.x; return _dst_val; } static nir_const_value evaluate_pack_unorm_2x16(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; const struct float32_vec src0 = { _src[0].f32[0], _src[0].f32[1], 0, 0, }; struct uint32_vec dst; dst.x = (uint32_t) pack_unorm_1x16(src0.x); dst.x |= ((uint32_t) pack_unorm_1x16(src0.y)) << 16; _dst_val.u32[0] = dst.x; return _dst_val; } static nir_const_value evaluate_pack_unorm_4x8(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; const struct float32_vec src0 = { _src[0].f32[0], _src[0].f32[1], _src[0].f32[2], _src[0].f32[3], }; struct uint32_vec dst; dst.x = (uint32_t) pack_unorm_1x8(src0.x); dst.x |= ((uint32_t) pack_unorm_1x8(src0.y)) << 8; dst.x |= ((uint32_t) pack_unorm_1x8(src0.z)) << 16; dst.x |= ((uint32_t) pack_unorm_1x8(src0.w)) << 24; _dst_val.u32[0] = dst.x; return _dst_val; } static nir_const_value evaluate_pack_uvec2_to_uint(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; const struct uint32_vec src0 = { _src[0].u32[0], _src[0].u32[1], 0, 0, }; struct uint32_vec dst; dst.x = (src0.x & 0xffff) | (src0.y << 16); _dst_val.u32[0] = dst.x; return _dst_val; } static nir_const_value evaluate_pack_uvec4_to_uint(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; const struct uint32_vec src0 = { _src[0].u32[0], _src[0].u32[1], _src[0].u32[2], _src[0].u32[3], }; struct uint32_vec dst; dst.x = (src0.x << 0) | (src0.y << 8) | (src0.z << 16) | (src0.w << 24); _dst_val.u32[0] = dst.x; return _dst_val; } static nir_const_value evaluate_seq(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; const float32_t src1 = _src[1].f32[_i]; float32_t dst = (src0 == src1) ? 1.0f : 0.0f; _dst_val.f32[_i] = dst; } return _dst_val; } static nir_const_value evaluate_sge(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const float src0 = _mesa_half_to_float(_src[0].u16[_i]); const float src1 = _mesa_half_to_float(_src[1].u16[_i]); float16_t dst = (src0 >= src1) ? 1.0f : 0.0f; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; const float32_t src1 = _src[1].f32[_i]; float32_t dst = (src0 >= src1) ? 1.0f : 0.0f; _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const float64_t src0 = _src[0].f64[_i]; const float64_t src1 = _src[1].f64[_i]; float64_t dst = (src0 >= src1) ? 1.0f : 0.0f; _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_slt(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; const float32_t src1 = _src[1].f32[_i]; float32_t dst = (src0 < src1) ? 1.0f : 0.0f; _dst_val.f32[_i] = dst; } return _dst_val; } static nir_const_value evaluate_sne(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const float32_t src0 = _src[0].f32[_i]; const float32_t src1 = _src[1].f32[_i]; float32_t dst = (src0 != src1) ? 1.0f : 0.0f; _dst_val.f32[_i] = dst; } return _dst_val; } static nir_const_value evaluate_u2f16(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const uint8_t src0 = _src[0].u8[_i]; float16_t dst = src0; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const uint16_t src0 = _src[0].u16[_i]; float16_t dst = src0; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; float16_t dst = src0; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const uint64_t src0 = _src[0].u64[_i]; float16_t dst = src0; _dst_val.u16[_i] = _mesa_float_to_half(dst); } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_u2f32(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const uint8_t src0 = _src[0].u8[_i]; float32_t dst = src0; _dst_val.f32[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const uint16_t src0 = _src[0].u16[_i]; float32_t dst = src0; _dst_val.f32[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; float32_t dst = src0; _dst_val.f32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const uint64_t src0 = _src[0].u64[_i]; float32_t dst = src0; _dst_val.f32[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_u2f64(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const uint8_t src0 = _src[0].u8[_i]; float64_t dst = src0; _dst_val.f64[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const uint16_t src0 = _src[0].u16[_i]; float64_t dst = src0; _dst_val.f64[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; float64_t dst = src0; _dst_val.f64[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const uint64_t src0 = _src[0].u64[_i]; float64_t dst = src0; _dst_val.f64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_u2u16(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const uint8_t src0 = _src[0].u8[_i]; uint16_t dst = src0; _dst_val.u16[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const uint16_t src0 = _src[0].u16[_i]; uint16_t dst = src0; _dst_val.u16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; uint16_t dst = src0; _dst_val.u16[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const uint64_t src0 = _src[0].u64[_i]; uint16_t dst = src0; _dst_val.u16[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_u2u32(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const uint8_t src0 = _src[0].u8[_i]; uint32_t dst = src0; _dst_val.u32[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const uint16_t src0 = _src[0].u16[_i]; uint32_t dst = src0; _dst_val.u32[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; uint32_t dst = src0; _dst_val.u32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const uint64_t src0 = _src[0].u64[_i]; uint32_t dst = src0; _dst_val.u32[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_u2u64(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const uint8_t src0 = _src[0].u8[_i]; uint64_t dst = src0; _dst_val.u64[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const uint16_t src0 = _src[0].u16[_i]; uint64_t dst = src0; _dst_val.u64[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; uint64_t dst = src0; _dst_val.u64[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const uint64_t src0 = _src[0].u64[_i]; uint64_t dst = src0; _dst_val.u64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_u2u8(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const uint8_t src0 = _src[0].u8[_i]; uint8_t dst = src0; _dst_val.u8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const uint16_t src0 = _src[0].u16[_i]; uint8_t dst = src0; _dst_val.u8[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; uint8_t dst = src0; _dst_val.u8[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const uint64_t src0 = _src[0].u64[_i]; uint8_t dst = src0; _dst_val.u8[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_uadd_carry(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const uint8_t src0 = _src[0].u8[_i]; const uint8_t src1 = _src[1].u8[_i]; uint8_t dst = src0 + src1 < src0; _dst_val.u8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const uint16_t src0 = _src[0].u16[_i]; const uint16_t src1 = _src[1].u16[_i]; uint16_t dst = src0 + src1 < src0; _dst_val.u16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; const uint32_t src1 = _src[1].u32[_i]; uint32_t dst = src0 + src1 < src0; _dst_val.u32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const uint64_t src0 = _src[0].u64[_i]; const uint64_t src1 = _src[1].u64[_i]; uint64_t dst = src0 + src1 < src0; _dst_val.u64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_ubfe(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; const int32_t src1 = _src[1].i32[_i]; const int32_t src2 = _src[2].i32[_i]; uint32_t dst; unsigned base = src0; int offset = src1, bits = src2; if (bits == 0) { dst = 0; } else if (bits < 0 || offset < 0) { dst = 0; /* undefined */ } else if (offset + bits < 32) { dst = (base << (32 - bits - offset)) >> (32 - bits); } else { dst = base >> offset; } _dst_val.u32[_i] = dst; } return _dst_val; } static nir_const_value evaluate_ubitfield_extract(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; const int32_t src1 = _src[1].i32[_i]; const int32_t src2 = _src[2].i32[_i]; uint32_t dst; unsigned base = src0; int offset = src1, bits = src2; if (bits == 0) { dst = 0; } else if (bits < 0 || offset < 0 || offset + bits > 32) { dst = 0; /* undefined per the spec */ } else { dst = (base >> offset) & ((1ull << bits) - 1); } _dst_val.u32[_i] = dst; } return _dst_val; } static nir_const_value evaluate_udiv(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const uint8_t src0 = _src[0].u8[_i]; const uint8_t src1 = _src[1].u8[_i]; uint8_t dst = src1 == 0 ? 0 : (src0 / src1); _dst_val.u8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const uint16_t src0 = _src[0].u16[_i]; const uint16_t src1 = _src[1].u16[_i]; uint16_t dst = src1 == 0 ? 0 : (src0 / src1); _dst_val.u16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; const uint32_t src1 = _src[1].u32[_i]; uint32_t dst = src1 == 0 ? 0 : (src0 / src1); _dst_val.u32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const uint64_t src0 = _src[0].u64[_i]; const uint64_t src1 = _src[1].u64[_i]; uint64_t dst = src1 == 0 ? 0 : (src0 / src1); _dst_val.u64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_ufind_msb(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const uint8_t src0 = _src[0].u8[_i]; int32_t dst; dst = -1; for (int bit = bit_size - 1; bit >= 0; bit--) { if ((src0 >> bit) & 1) { dst = bit; break; } } _dst_val.i32[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const uint16_t src0 = _src[0].u16[_i]; int32_t dst; dst = -1; for (int bit = bit_size - 1; bit >= 0; bit--) { if ((src0 >> bit) & 1) { dst = bit; break; } } _dst_val.i32[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; int32_t dst; dst = -1; for (int bit = bit_size - 1; bit >= 0; bit--) { if ((src0 >> bit) & 1) { dst = bit; break; } } _dst_val.i32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const uint64_t src0 = _src[0].u64[_i]; int32_t dst; dst = -1; for (int bit = bit_size - 1; bit >= 0; bit--) { if ((src0 >> bit) & 1) { dst = bit; break; } } _dst_val.i32[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_uge(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const uint8_t src0 = _src[0].u8[_i]; const uint8_t src1 = _src[1].u8[_i]; bool32_t dst = src0 >= src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const uint16_t src0 = _src[0].u16[_i]; const uint16_t src1 = _src[1].u16[_i]; bool32_t dst = src0 >= src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; const uint32_t src1 = _src[1].u32[_i]; bool32_t dst = src0 >= src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const uint64_t src0 = _src[0].u64[_i]; const uint64_t src1 = _src[1].u64[_i]; bool32_t dst = src0 >= src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_ult(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const uint8_t src0 = _src[0].u8[_i]; const uint8_t src1 = _src[1].u8[_i]; bool32_t dst = src0 < src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const uint16_t src0 = _src[0].u16[_i]; const uint16_t src1 = _src[1].u16[_i]; bool32_t dst = src0 < src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; const uint32_t src1 = _src[1].u32[_i]; bool32_t dst = src0 < src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const uint64_t src0 = _src[0].u64[_i]; const uint64_t src1 = _src[1].u64[_i]; bool32_t dst = src0 < src1; _dst_val.u32[_i] = dst ? NIR_TRUE : NIR_FALSE; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_umax(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const uint8_t src0 = _src[0].u8[_i]; const uint8_t src1 = _src[1].u8[_i]; uint8_t dst = src1 > src0 ? src1 : src0; _dst_val.u8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const uint16_t src0 = _src[0].u16[_i]; const uint16_t src1 = _src[1].u16[_i]; uint16_t dst = src1 > src0 ? src1 : src0; _dst_val.u16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; const uint32_t src1 = _src[1].u32[_i]; uint32_t dst = src1 > src0 ? src1 : src0; _dst_val.u32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const uint64_t src0 = _src[0].u64[_i]; const uint64_t src1 = _src[1].u64[_i]; uint64_t dst = src1 > src0 ? src1 : src0; _dst_val.u64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_umax3(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const uint8_t src0 = _src[0].u8[_i]; const uint8_t src1 = _src[1].u8[_i]; const uint8_t src2 = _src[2].u8[_i]; uint8_t dst = MAX2(src0, MAX2(src1, src2)); _dst_val.u8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const uint16_t src0 = _src[0].u16[_i]; const uint16_t src1 = _src[1].u16[_i]; const uint16_t src2 = _src[2].u16[_i]; uint16_t dst = MAX2(src0, MAX2(src1, src2)); _dst_val.u16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; const uint32_t src1 = _src[1].u32[_i]; const uint32_t src2 = _src[2].u32[_i]; uint32_t dst = MAX2(src0, MAX2(src1, src2)); _dst_val.u32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const uint64_t src0 = _src[0].u64[_i]; const uint64_t src1 = _src[1].u64[_i]; const uint64_t src2 = _src[2].u64[_i]; uint64_t dst = MAX2(src0, MAX2(src1, src2)); _dst_val.u64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_umax_4x8(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; const int32_t src1 = _src[1].i32[_i]; int32_t dst; dst = 0; for (int i = 0; i < 32; i += 8) { dst |= MAX2((src0 >> i) & 0xff, (src1 >> i) & 0xff) << i; } _dst_val.i32[_i] = dst; } return _dst_val; } static nir_const_value evaluate_umed3(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const uint8_t src0 = _src[0].u8[_i]; const uint8_t src1 = _src[1].u8[_i]; const uint8_t src2 = _src[2].u8[_i]; uint8_t dst = MAX2(MIN2(MAX2(src0, src1), src2), MIN2(src0, src1)); _dst_val.u8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const uint16_t src0 = _src[0].u16[_i]; const uint16_t src1 = _src[1].u16[_i]; const uint16_t src2 = _src[2].u16[_i]; uint16_t dst = MAX2(MIN2(MAX2(src0, src1), src2), MIN2(src0, src1)); _dst_val.u16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; const uint32_t src1 = _src[1].u32[_i]; const uint32_t src2 = _src[2].u32[_i]; uint32_t dst = MAX2(MIN2(MAX2(src0, src1), src2), MIN2(src0, src1)); _dst_val.u32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const uint64_t src0 = _src[0].u64[_i]; const uint64_t src1 = _src[1].u64[_i]; const uint64_t src2 = _src[2].u64[_i]; uint64_t dst = MAX2(MIN2(MAX2(src0, src1), src2), MIN2(src0, src1)); _dst_val.u64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_umin(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const uint8_t src0 = _src[0].u8[_i]; const uint8_t src1 = _src[1].u8[_i]; uint8_t dst = src1 > src0 ? src0 : src1; _dst_val.u8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const uint16_t src0 = _src[0].u16[_i]; const uint16_t src1 = _src[1].u16[_i]; uint16_t dst = src1 > src0 ? src0 : src1; _dst_val.u16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; const uint32_t src1 = _src[1].u32[_i]; uint32_t dst = src1 > src0 ? src0 : src1; _dst_val.u32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const uint64_t src0 = _src[0].u64[_i]; const uint64_t src1 = _src[1].u64[_i]; uint64_t dst = src1 > src0 ? src0 : src1; _dst_val.u64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_umin3(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const uint8_t src0 = _src[0].u8[_i]; const uint8_t src1 = _src[1].u8[_i]; const uint8_t src2 = _src[2].u8[_i]; uint8_t dst = MIN2(src0, MIN2(src1, src2)); _dst_val.u8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const uint16_t src0 = _src[0].u16[_i]; const uint16_t src1 = _src[1].u16[_i]; const uint16_t src2 = _src[2].u16[_i]; uint16_t dst = MIN2(src0, MIN2(src1, src2)); _dst_val.u16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; const uint32_t src1 = _src[1].u32[_i]; const uint32_t src2 = _src[2].u32[_i]; uint32_t dst = MIN2(src0, MIN2(src1, src2)); _dst_val.u32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const uint64_t src0 = _src[0].u64[_i]; const uint64_t src1 = _src[1].u64[_i]; const uint64_t src2 = _src[2].u64[_i]; uint64_t dst = MIN2(src0, MIN2(src1, src2)); _dst_val.u64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_umin_4x8(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; const int32_t src1 = _src[1].i32[_i]; int32_t dst; dst = 0; for (int i = 0; i < 32; i += 8) { dst |= MIN2((src0 >> i) & 0xff, (src1 >> i) & 0xff) << i; } _dst_val.i32[_i] = dst; } return _dst_val; } static nir_const_value evaluate_umod(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const uint8_t src0 = _src[0].u8[_i]; const uint8_t src1 = _src[1].u8[_i]; uint8_t dst = src1 == 0 ? 0 : src0 % src1; _dst_val.u8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const uint16_t src0 = _src[0].u16[_i]; const uint16_t src1 = _src[1].u16[_i]; uint16_t dst = src1 == 0 ? 0 : src0 % src1; _dst_val.u16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; const uint32_t src1 = _src[1].u32[_i]; uint32_t dst = src1 == 0 ? 0 : src0 % src1; _dst_val.u32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const uint64_t src0 = _src[0].u64[_i]; const uint64_t src1 = _src[1].u64[_i]; uint64_t dst = src1 == 0 ? 0 : src0 % src1; _dst_val.u64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_umul_high(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; const uint32_t src1 = _src[1].u32[_i]; uint32_t dst = (uint32_t)(((uint64_t) src0 * (uint64_t) src1) >> 32); _dst_val.u32[_i] = dst; } return _dst_val; } static nir_const_value evaluate_umul_unorm_4x8(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; const int32_t src1 = _src[1].i32[_i]; int32_t dst; dst = 0; for (int i = 0; i < 32; i += 8) { int src0_chan = (src0 >> i) & 0xff; int src1_chan = (src1 >> i) & 0xff; dst |= ((src0_chan * src1_chan) / 255) << i; } _dst_val.i32[_i] = dst; } return _dst_val; } static nir_const_value evaluate_unpack_32_2x16(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; const struct uint32_vec src0 = { _src[0].u32[0], 0, 0, 0, }; struct uint16_vec dst; dst.x = src0.x; dst.y = src0.x >> 16; _dst_val.u16[0] = dst.x; _dst_val.u16[1] = dst.y; return _dst_val; } static nir_const_value evaluate_unpack_32_2x16_split_x(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; uint16_t dst = src0; _dst_val.u16[_i] = dst; } return _dst_val; } static nir_const_value evaluate_unpack_32_2x16_split_y(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; uint16_t dst = src0 >> 16; _dst_val.u16[_i] = dst; } return _dst_val; } static nir_const_value evaluate_unpack_64_2x32(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; const struct uint64_vec src0 = { _src[0].u64[0], 0, 0, 0, }; struct uint32_vec dst; dst.x = src0.x; dst.y = src0.x >> 32; _dst_val.u32[0] = dst.x; _dst_val.u32[1] = dst.y; return _dst_val; } static nir_const_value evaluate_unpack_64_2x32_split_x(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const uint64_t src0 = _src[0].u64[_i]; uint32_t dst = src0; _dst_val.u32[_i] = dst; } return _dst_val; } static nir_const_value evaluate_unpack_64_2x32_split_y(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const uint64_t src0 = _src[0].u64[_i]; uint32_t dst = src0 >> 32; _dst_val.u32[_i] = dst; } return _dst_val; } static nir_const_value evaluate_unpack_64_4x16(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; const struct uint64_vec src0 = { _src[0].u64[0], 0, 0, 0, }; struct uint16_vec dst; dst.x = src0.x; dst.y = src0.x >> 16; dst.z = src0.x >> 32; dst.w = src0.w >> 48; _dst_val.u16[0] = dst.x; _dst_val.u16[1] = dst.y; _dst_val.u16[2] = dst.z; _dst_val.u16[3] = dst.w; return _dst_val; } static nir_const_value evaluate_unpack_half_2x16(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; const struct uint32_vec src0 = { _src[0].u32[0], 0, 0, 0, }; struct float32_vec dst; dst.x = unpack_half_1x16((uint16_t)(src0.x & 0xffff)); dst.y = unpack_half_1x16((uint16_t)(src0.x << 16)); _dst_val.f32[0] = dst.x; _dst_val.f32[1] = dst.y; return _dst_val; } static nir_const_value evaluate_unpack_half_2x16_split_x(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; float32_t dst = unpack_half_1x16((uint16_t)(src0 & 0xffff)); _dst_val.f32[_i] = dst; } return _dst_val; } static nir_const_value evaluate_unpack_half_2x16_split_y(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; float32_t dst = unpack_half_1x16((uint16_t)(src0 >> 16)); _dst_val.f32[_i] = dst; } return _dst_val; } static nir_const_value evaluate_unpack_snorm_2x16(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; const struct uint32_vec src0 = { _src[0].u32[0], 0, 0, 0, }; struct float32_vec dst; dst.x = unpack_snorm_1x16((uint16_t)(src0.x & 0xffff)); dst.y = unpack_snorm_1x16((uint16_t)(src0.x << 16)); _dst_val.f32[0] = dst.x; _dst_val.f32[1] = dst.y; return _dst_val; } static nir_const_value evaluate_unpack_snorm_4x8(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; const struct uint32_vec src0 = { _src[0].u32[0], 0, 0, 0, }; struct float32_vec dst; dst.x = unpack_snorm_1x8((uint8_t)(src0.x & 0xff)); dst.y = unpack_snorm_1x8((uint8_t)((src0.x >> 8) & 0xff)); dst.z = unpack_snorm_1x8((uint8_t)((src0.x >> 16) & 0xff)); dst.w = unpack_snorm_1x8((uint8_t)(src0.x >> 24)); _dst_val.f32[0] = dst.x; _dst_val.f32[1] = dst.y; _dst_val.f32[2] = dst.z; _dst_val.f32[3] = dst.w; return _dst_val; } static nir_const_value evaluate_unpack_unorm_2x16(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; const struct uint32_vec src0 = { _src[0].u32[0], 0, 0, 0, }; struct float32_vec dst; dst.x = unpack_unorm_1x16((uint16_t)(src0.x & 0xffff)); dst.y = unpack_unorm_1x16((uint16_t)(src0.x << 16)); _dst_val.f32[0] = dst.x; _dst_val.f32[1] = dst.y; return _dst_val; } static nir_const_value evaluate_unpack_unorm_4x8(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; const struct uint32_vec src0 = { _src[0].u32[0], 0, 0, 0, }; struct float32_vec dst; dst.x = unpack_unorm_1x8((uint8_t)(src0.x & 0xff)); dst.y = unpack_unorm_1x8((uint8_t)((src0.x >> 8) & 0xff)); dst.z = unpack_unorm_1x8((uint8_t)((src0.x >> 16) & 0xff)); dst.w = unpack_unorm_1x8((uint8_t)(src0.x >> 24)); _dst_val.f32[0] = dst.x; _dst_val.f32[1] = dst.y; _dst_val.f32[2] = dst.z; _dst_val.f32[3] = dst.w; return _dst_val; } static nir_const_value evaluate_usadd_4x8(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; const int32_t src1 = _src[1].i32[_i]; int32_t dst; dst = 0; for (int i = 0; i < 32; i += 8) { dst |= MIN2(((src0 >> i) & 0xff) + ((src1 >> i) & 0xff), 0xff) << i; } _dst_val.i32[_i] = dst; } return _dst_val; } static nir_const_value evaluate_ushr(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const uint8_t src0 = _src[0].u8[_i]; const uint32_t src1 = _src[1].u32[_i]; uint8_t dst = src0 >> src1; _dst_val.u8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const uint16_t src0 = _src[0].u16[_i]; const uint32_t src1 = _src[1].u32[_i]; uint16_t dst = src0 >> src1; _dst_val.u16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; const uint32_t src1 = _src[1].u32[_i]; uint32_t dst = src0 >> src1; _dst_val.u32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const uint64_t src0 = _src[0].u64[_i]; const uint32_t src1 = _src[1].u32[_i]; uint64_t dst = src0 >> src1; _dst_val.u64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_ussub_4x8(MAYBE_UNUSED unsigned num_components, UNUSED unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; for (unsigned _i = 0; _i < num_components; _i++) { const int32_t src0 = _src[0].i32[_i]; const int32_t src1 = _src[1].i32[_i]; int32_t dst; dst = 0; for (int i = 0; i < 32; i += 8) { int src0_chan = (src0 >> i) & 0xff; int src1_chan = (src1 >> i) & 0xff; if (src0_chan > src1_chan) dst |= (src0_chan - src1_chan) << i; } _dst_val.i32[_i] = dst; } return _dst_val; } static nir_const_value evaluate_usub_borrow(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { for (unsigned _i = 0; _i < num_components; _i++) { const uint8_t src0 = _src[0].u8[_i]; const uint8_t src1 = _src[1].u8[_i]; uint8_t dst = src0 < src1; _dst_val.u8[_i] = dst; } break; } case 16: { for (unsigned _i = 0; _i < num_components; _i++) { const uint16_t src0 = _src[0].u16[_i]; const uint16_t src1 = _src[1].u16[_i]; uint16_t dst = src0 < src1; _dst_val.u16[_i] = dst; } break; } case 32: { for (unsigned _i = 0; _i < num_components; _i++) { const uint32_t src0 = _src[0].u32[_i]; const uint32_t src1 = _src[1].u32[_i]; uint32_t dst = src0 < src1; _dst_val.u32[_i] = dst; } break; } case 64: { for (unsigned _i = 0; _i < num_components; _i++) { const uint64_t src0 = _src[0].u64[_i]; const uint64_t src1 = _src[1].u64[_i]; uint64_t dst = src0 < src1; _dst_val.u64[_i] = dst; } break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_vec2(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { const struct uint8_vec src0 = { _src[0].u8[0], 0, 0, 0, }; const struct uint8_vec src1 = { _src[1].u8[0], 0, 0, 0, }; struct uint8_vec dst; dst.x = src0.x; dst.y = src1.x; _dst_val.u8[0] = dst.x; _dst_val.u8[1] = dst.y; break; } case 16: { const struct uint16_vec src0 = { _src[0].u16[0], 0, 0, 0, }; const struct uint16_vec src1 = { _src[1].u16[0], 0, 0, 0, }; struct uint16_vec dst; dst.x = src0.x; dst.y = src1.x; _dst_val.u16[0] = dst.x; _dst_val.u16[1] = dst.y; break; } case 32: { const struct uint32_vec src0 = { _src[0].u32[0], 0, 0, 0, }; const struct uint32_vec src1 = { _src[1].u32[0], 0, 0, 0, }; struct uint32_vec dst; dst.x = src0.x; dst.y = src1.x; _dst_val.u32[0] = dst.x; _dst_val.u32[1] = dst.y; break; } case 64: { const struct uint64_vec src0 = { _src[0].u64[0], 0, 0, 0, }; const struct uint64_vec src1 = { _src[1].u64[0], 0, 0, 0, }; struct uint64_vec dst; dst.x = src0.x; dst.y = src1.x; _dst_val.u64[0] = dst.x; _dst_val.u64[1] = dst.y; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_vec3(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { const struct uint8_vec src0 = { _src[0].u8[0], 0, 0, 0, }; const struct uint8_vec src1 = { _src[1].u8[0], 0, 0, 0, }; const struct uint8_vec src2 = { _src[2].u8[0], 0, 0, 0, }; struct uint8_vec dst; dst.x = src0.x; dst.y = src1.x; dst.z = src2.x; _dst_val.u8[0] = dst.x; _dst_val.u8[1] = dst.y; _dst_val.u8[2] = dst.z; break; } case 16: { const struct uint16_vec src0 = { _src[0].u16[0], 0, 0, 0, }; const struct uint16_vec src1 = { _src[1].u16[0], 0, 0, 0, }; const struct uint16_vec src2 = { _src[2].u16[0], 0, 0, 0, }; struct uint16_vec dst; dst.x = src0.x; dst.y = src1.x; dst.z = src2.x; _dst_val.u16[0] = dst.x; _dst_val.u16[1] = dst.y; _dst_val.u16[2] = dst.z; break; } case 32: { const struct uint32_vec src0 = { _src[0].u32[0], 0, 0, 0, }; const struct uint32_vec src1 = { _src[1].u32[0], 0, 0, 0, }; const struct uint32_vec src2 = { _src[2].u32[0], 0, 0, 0, }; struct uint32_vec dst; dst.x = src0.x; dst.y = src1.x; dst.z = src2.x; _dst_val.u32[0] = dst.x; _dst_val.u32[1] = dst.y; _dst_val.u32[2] = dst.z; break; } case 64: { const struct uint64_vec src0 = { _src[0].u64[0], 0, 0, 0, }; const struct uint64_vec src1 = { _src[1].u64[0], 0, 0, 0, }; const struct uint64_vec src2 = { _src[2].u64[0], 0, 0, 0, }; struct uint64_vec dst; dst.x = src0.x; dst.y = src1.x; dst.z = src2.x; _dst_val.u64[0] = dst.x; _dst_val.u64[1] = dst.y; _dst_val.u64[2] = dst.z; break; } default: unreachable("unknown bit width"); } return _dst_val; } static nir_const_value evaluate_vec4(MAYBE_UNUSED unsigned num_components, unsigned bit_size, MAYBE_UNUSED nir_const_value *_src) { nir_const_value _dst_val = { {0, } }; switch (bit_size) { case 8: { const struct uint8_vec src0 = { _src[0].u8[0], 0, 0, 0, }; const struct uint8_vec src1 = { _src[1].u8[0], 0, 0, 0, }; const struct uint8_vec src2 = { _src[2].u8[0], 0, 0, 0, }; const struct uint8_vec src3 = { _src[3].u8[0], 0, 0, 0, }; struct uint8_vec dst; dst.x = src0.x; dst.y = src1.x; dst.z = src2.x; dst.w = src3.x; _dst_val.u8[0] = dst.x; _dst_val.u8[1] = dst.y; _dst_val.u8[2] = dst.z; _dst_val.u8[3] = dst.w; break; } case 16: { const struct uint16_vec src0 = { _src[0].u16[0], 0, 0, 0, }; const struct uint16_vec src1 = { _src[1].u16[0], 0, 0, 0, }; const struct uint16_vec src2 = { _src[2].u16[0], 0, 0, 0, }; const struct uint16_vec src3 = { _src[3].u16[0], 0, 0, 0, }; struct uint16_vec dst; dst.x = src0.x; dst.y = src1.x; dst.z = src2.x; dst.w = src3.x; _dst_val.u16[0] = dst.x; _dst_val.u16[1] = dst.y; _dst_val.u16[2] = dst.z; _dst_val.u16[3] = dst.w; break; } case 32: { const struct uint32_vec src0 = { _src[0].u32[0], 0, 0, 0, }; const struct uint32_vec src1 = { _src[1].u32[0], 0, 0, 0, }; const struct uint32_vec src2 = { _src[2].u32[0], 0, 0, 0, }; const struct uint32_vec src3 = { _src[3].u32[0], 0, 0, 0, }; struct uint32_vec dst; dst.x = src0.x; dst.y = src1.x; dst.z = src2.x; dst.w = src3.x; _dst_val.u32[0] = dst.x; _dst_val.u32[1] = dst.y; _dst_val.u32[2] = dst.z; _dst_val.u32[3] = dst.w; break; } case 64: { const struct uint64_vec src0 = { _src[0].u64[0], 0, 0, 0, }; const struct uint64_vec src1 = { _src[1].u64[0], 0, 0, 0, }; const struct uint64_vec src2 = { _src[2].u64[0], 0, 0, 0, }; const struct uint64_vec src3 = { _src[3].u64[0], 0, 0, 0, }; struct uint64_vec dst; dst.x = src0.x; dst.y = src1.x; dst.z = src2.x; dst.w = src3.x; _dst_val.u64[0] = dst.x; _dst_val.u64[1] = dst.y; _dst_val.u64[2] = dst.z; _dst_val.u64[3] = dst.w; break; } default: unreachable("unknown bit width"); } return _dst_val; } nir_const_value nir_eval_const_opcode(nir_op op, unsigned num_components, unsigned bit_width, nir_const_value *src) { switch (op) { case nir_op_b2f: return evaluate_b2f(num_components, bit_width, src); case nir_op_b2i: return evaluate_b2i(num_components, bit_width, src); case nir_op_ball_fequal2: return evaluate_ball_fequal2(num_components, bit_width, src); case nir_op_ball_fequal3: return evaluate_ball_fequal3(num_components, bit_width, src); case nir_op_ball_fequal4: return evaluate_ball_fequal4(num_components, bit_width, src); case nir_op_ball_iequal2: return evaluate_ball_iequal2(num_components, bit_width, src); case nir_op_ball_iequal3: return evaluate_ball_iequal3(num_components, bit_width, src); case nir_op_ball_iequal4: return evaluate_ball_iequal4(num_components, bit_width, src); case nir_op_bany_fnequal2: return evaluate_bany_fnequal2(num_components, bit_width, src); case nir_op_bany_fnequal3: return evaluate_bany_fnequal3(num_components, bit_width, src); case nir_op_bany_fnequal4: return evaluate_bany_fnequal4(num_components, bit_width, src); case nir_op_bany_inequal2: return evaluate_bany_inequal2(num_components, bit_width, src); case nir_op_bany_inequal3: return evaluate_bany_inequal3(num_components, bit_width, src); case nir_op_bany_inequal4: return evaluate_bany_inequal4(num_components, bit_width, src); case nir_op_bcsel: return evaluate_bcsel(num_components, bit_width, src); case nir_op_bfi: return evaluate_bfi(num_components, bit_width, src); case nir_op_bfm: return evaluate_bfm(num_components, bit_width, src); case nir_op_bit_count: return evaluate_bit_count(num_components, bit_width, src); case nir_op_bitfield_insert: return evaluate_bitfield_insert(num_components, bit_width, src); case nir_op_bitfield_reverse: return evaluate_bitfield_reverse(num_components, bit_width, src); case nir_op_cube_face_coord: return evaluate_cube_face_coord(num_components, bit_width, src); case nir_op_cube_face_index: return evaluate_cube_face_index(num_components, bit_width, src); case nir_op_extract_i16: return evaluate_extract_i16(num_components, bit_width, src); case nir_op_extract_i8: return evaluate_extract_i8(num_components, bit_width, src); case nir_op_extract_u16: return evaluate_extract_u16(num_components, bit_width, src); case nir_op_extract_u8: return evaluate_extract_u8(num_components, bit_width, src); case nir_op_f2b: return evaluate_f2b(num_components, bit_width, src); case nir_op_f2f16: return evaluate_f2f16(num_components, bit_width, src); case nir_op_f2f16_rtne: return evaluate_f2f16_rtne(num_components, bit_width, src); case nir_op_f2f16_rtz: return evaluate_f2f16_rtz(num_components, bit_width, src); case nir_op_f2f32: return evaluate_f2f32(num_components, bit_width, src); case nir_op_f2f64: return evaluate_f2f64(num_components, bit_width, src); case nir_op_f2i16: return evaluate_f2i16(num_components, bit_width, src); case nir_op_f2i32: return evaluate_f2i32(num_components, bit_width, src); case nir_op_f2i64: return evaluate_f2i64(num_components, bit_width, src); case nir_op_f2i8: return evaluate_f2i8(num_components, bit_width, src); case nir_op_f2u16: return evaluate_f2u16(num_components, bit_width, src); case nir_op_f2u32: return evaluate_f2u32(num_components, bit_width, src); case nir_op_f2u64: return evaluate_f2u64(num_components, bit_width, src); case nir_op_f2u8: return evaluate_f2u8(num_components, bit_width, src); case nir_op_fabs: return evaluate_fabs(num_components, bit_width, src); case nir_op_fadd: return evaluate_fadd(num_components, bit_width, src); case nir_op_fall_equal2: return evaluate_fall_equal2(num_components, bit_width, src); case nir_op_fall_equal3: return evaluate_fall_equal3(num_components, bit_width, src); case nir_op_fall_equal4: return evaluate_fall_equal4(num_components, bit_width, src); case nir_op_fand: return evaluate_fand(num_components, bit_width, src); case nir_op_fany_nequal2: return evaluate_fany_nequal2(num_components, bit_width, src); case nir_op_fany_nequal3: return evaluate_fany_nequal3(num_components, bit_width, src); case nir_op_fany_nequal4: return evaluate_fany_nequal4(num_components, bit_width, src); case nir_op_fceil: return evaluate_fceil(num_components, bit_width, src); case nir_op_fcos: return evaluate_fcos(num_components, bit_width, src); case nir_op_fcsel: return evaluate_fcsel(num_components, bit_width, src); case nir_op_fddx: return evaluate_fddx(num_components, bit_width, src); case nir_op_fddx_coarse: return evaluate_fddx_coarse(num_components, bit_width, src); case nir_op_fddx_fine: return evaluate_fddx_fine(num_components, bit_width, src); case nir_op_fddy: return evaluate_fddy(num_components, bit_width, src); case nir_op_fddy_coarse: return evaluate_fddy_coarse(num_components, bit_width, src); case nir_op_fddy_fine: return evaluate_fddy_fine(num_components, bit_width, src); case nir_op_fdiv: return evaluate_fdiv(num_components, bit_width, src); case nir_op_fdot2: return evaluate_fdot2(num_components, bit_width, src); case nir_op_fdot3: return evaluate_fdot3(num_components, bit_width, src); case nir_op_fdot4: return evaluate_fdot4(num_components, bit_width, src); case nir_op_fdot_replicated2: return evaluate_fdot_replicated2(num_components, bit_width, src); case nir_op_fdot_replicated3: return evaluate_fdot_replicated3(num_components, bit_width, src); case nir_op_fdot_replicated4: return evaluate_fdot_replicated4(num_components, bit_width, src); case nir_op_fdph: return evaluate_fdph(num_components, bit_width, src); case nir_op_fdph_replicated: return evaluate_fdph_replicated(num_components, bit_width, src); case nir_op_feq: return evaluate_feq(num_components, bit_width, src); case nir_op_fexp2: return evaluate_fexp2(num_components, bit_width, src); case nir_op_ffloor: return evaluate_ffloor(num_components, bit_width, src); case nir_op_ffma: return evaluate_ffma(num_components, bit_width, src); case nir_op_ffract: return evaluate_ffract(num_components, bit_width, src); case nir_op_fge: return evaluate_fge(num_components, bit_width, src); case nir_op_find_lsb: return evaluate_find_lsb(num_components, bit_width, src); case nir_op_flog2: return evaluate_flog2(num_components, bit_width, src); case nir_op_flrp: return evaluate_flrp(num_components, bit_width, src); case nir_op_flt: return evaluate_flt(num_components, bit_width, src); case nir_op_fmax: return evaluate_fmax(num_components, bit_width, src); case nir_op_fmax3: return evaluate_fmax3(num_components, bit_width, src); case nir_op_fmed3: return evaluate_fmed3(num_components, bit_width, src); case nir_op_fmin: return evaluate_fmin(num_components, bit_width, src); case nir_op_fmin3: return evaluate_fmin3(num_components, bit_width, src); case nir_op_fmod: return evaluate_fmod(num_components, bit_width, src); case nir_op_fmov: return evaluate_fmov(num_components, bit_width, src); case nir_op_fmul: return evaluate_fmul(num_components, bit_width, src); case nir_op_fne: return evaluate_fne(num_components, bit_width, src); case nir_op_fneg: return evaluate_fneg(num_components, bit_width, src); case nir_op_fnoise1_1: return evaluate_fnoise1_1(num_components, bit_width, src); case nir_op_fnoise1_2: return evaluate_fnoise1_2(num_components, bit_width, src); case nir_op_fnoise1_3: return evaluate_fnoise1_3(num_components, bit_width, src); case nir_op_fnoise1_4: return evaluate_fnoise1_4(num_components, bit_width, src); case nir_op_fnoise2_1: return evaluate_fnoise2_1(num_components, bit_width, src); case nir_op_fnoise2_2: return evaluate_fnoise2_2(num_components, bit_width, src); case nir_op_fnoise2_3: return evaluate_fnoise2_3(num_components, bit_width, src); case nir_op_fnoise2_4: return evaluate_fnoise2_4(num_components, bit_width, src); case nir_op_fnoise3_1: return evaluate_fnoise3_1(num_components, bit_width, src); case nir_op_fnoise3_2: return evaluate_fnoise3_2(num_components, bit_width, src); case nir_op_fnoise3_3: return evaluate_fnoise3_3(num_components, bit_width, src); case nir_op_fnoise3_4: return evaluate_fnoise3_4(num_components, bit_width, src); case nir_op_fnoise4_1: return evaluate_fnoise4_1(num_components, bit_width, src); case nir_op_fnoise4_2: return evaluate_fnoise4_2(num_components, bit_width, src); case nir_op_fnoise4_3: return evaluate_fnoise4_3(num_components, bit_width, src); case nir_op_fnoise4_4: return evaluate_fnoise4_4(num_components, bit_width, src); case nir_op_fnot: return evaluate_fnot(num_components, bit_width, src); case nir_op_for: return evaluate_for(num_components, bit_width, src); case nir_op_fpow: return evaluate_fpow(num_components, bit_width, src); case nir_op_fquantize2f16: return evaluate_fquantize2f16(num_components, bit_width, src); case nir_op_frcp: return evaluate_frcp(num_components, bit_width, src); case nir_op_frem: return evaluate_frem(num_components, bit_width, src); case nir_op_frexp_exp: return evaluate_frexp_exp(num_components, bit_width, src); case nir_op_frexp_sig: return evaluate_frexp_sig(num_components, bit_width, src); case nir_op_fround_even: return evaluate_fround_even(num_components, bit_width, src); case nir_op_frsq: return evaluate_frsq(num_components, bit_width, src); case nir_op_fsat: return evaluate_fsat(num_components, bit_width, src); case nir_op_fsign: return evaluate_fsign(num_components, bit_width, src); case nir_op_fsin: return evaluate_fsin(num_components, bit_width, src); case nir_op_fsqrt: return evaluate_fsqrt(num_components, bit_width, src); case nir_op_fsub: return evaluate_fsub(num_components, bit_width, src); case nir_op_ftrunc: return evaluate_ftrunc(num_components, bit_width, src); case nir_op_fxor: return evaluate_fxor(num_components, bit_width, src); case nir_op_i2b: return evaluate_i2b(num_components, bit_width, src); case nir_op_i2f16: return evaluate_i2f16(num_components, bit_width, src); case nir_op_i2f32: return evaluate_i2f32(num_components, bit_width, src); case nir_op_i2f64: return evaluate_i2f64(num_components, bit_width, src); case nir_op_i2i16: return evaluate_i2i16(num_components, bit_width, src); case nir_op_i2i32: return evaluate_i2i32(num_components, bit_width, src); case nir_op_i2i64: return evaluate_i2i64(num_components, bit_width, src); case nir_op_i2i8: return evaluate_i2i8(num_components, bit_width, src); case nir_op_iabs: return evaluate_iabs(num_components, bit_width, src); case nir_op_iadd: return evaluate_iadd(num_components, bit_width, src); case nir_op_iand: return evaluate_iand(num_components, bit_width, src); case nir_op_ibfe: return evaluate_ibfe(num_components, bit_width, src); case nir_op_ibitfield_extract: return evaluate_ibitfield_extract(num_components, bit_width, src); case nir_op_idiv: return evaluate_idiv(num_components, bit_width, src); case nir_op_ieq: return evaluate_ieq(num_components, bit_width, src); case nir_op_ifind_msb: return evaluate_ifind_msb(num_components, bit_width, src); case nir_op_ige: return evaluate_ige(num_components, bit_width, src); case nir_op_ilt: return evaluate_ilt(num_components, bit_width, src); case nir_op_imax: return evaluate_imax(num_components, bit_width, src); case nir_op_imax3: return evaluate_imax3(num_components, bit_width, src); case nir_op_imed3: return evaluate_imed3(num_components, bit_width, src); case nir_op_imin: return evaluate_imin(num_components, bit_width, src); case nir_op_imin3: return evaluate_imin3(num_components, bit_width, src); case nir_op_imod: return evaluate_imod(num_components, bit_width, src); case nir_op_imov: return evaluate_imov(num_components, bit_width, src); case nir_op_imul: return evaluate_imul(num_components, bit_width, src); case nir_op_imul_high: return evaluate_imul_high(num_components, bit_width, src); case nir_op_ine: return evaluate_ine(num_components, bit_width, src); case nir_op_ineg: return evaluate_ineg(num_components, bit_width, src); case nir_op_inot: return evaluate_inot(num_components, bit_width, src); case nir_op_ior: return evaluate_ior(num_components, bit_width, src); case nir_op_irem: return evaluate_irem(num_components, bit_width, src); case nir_op_ishl: return evaluate_ishl(num_components, bit_width, src); case nir_op_ishr: return evaluate_ishr(num_components, bit_width, src); case nir_op_isign: return evaluate_isign(num_components, bit_width, src); case nir_op_isub: return evaluate_isub(num_components, bit_width, src); case nir_op_ixor: return evaluate_ixor(num_components, bit_width, src); case nir_op_ldexp: return evaluate_ldexp(num_components, bit_width, src); case nir_op_pack_32_2x16: return evaluate_pack_32_2x16(num_components, bit_width, src); case nir_op_pack_32_2x16_split: return evaluate_pack_32_2x16_split(num_components, bit_width, src); case nir_op_pack_64_2x32: return evaluate_pack_64_2x32(num_components, bit_width, src); case nir_op_pack_64_2x32_split: return evaluate_pack_64_2x32_split(num_components, bit_width, src); case nir_op_pack_64_4x16: return evaluate_pack_64_4x16(num_components, bit_width, src); case nir_op_pack_half_2x16: return evaluate_pack_half_2x16(num_components, bit_width, src); case nir_op_pack_half_2x16_split: return evaluate_pack_half_2x16_split(num_components, bit_width, src); case nir_op_pack_snorm_2x16: return evaluate_pack_snorm_2x16(num_components, bit_width, src); case nir_op_pack_snorm_4x8: return evaluate_pack_snorm_4x8(num_components, bit_width, src); case nir_op_pack_unorm_2x16: return evaluate_pack_unorm_2x16(num_components, bit_width, src); case nir_op_pack_unorm_4x8: return evaluate_pack_unorm_4x8(num_components, bit_width, src); case nir_op_pack_uvec2_to_uint: return evaluate_pack_uvec2_to_uint(num_components, bit_width, src); case nir_op_pack_uvec4_to_uint: return evaluate_pack_uvec4_to_uint(num_components, bit_width, src); case nir_op_seq: return evaluate_seq(num_components, bit_width, src); case nir_op_sge: return evaluate_sge(num_components, bit_width, src); case nir_op_slt: return evaluate_slt(num_components, bit_width, src); case nir_op_sne: return evaluate_sne(num_components, bit_width, src); case nir_op_u2f16: return evaluate_u2f16(num_components, bit_width, src); case nir_op_u2f32: return evaluate_u2f32(num_components, bit_width, src); case nir_op_u2f64: return evaluate_u2f64(num_components, bit_width, src); case nir_op_u2u16: return evaluate_u2u16(num_components, bit_width, src); case nir_op_u2u32: return evaluate_u2u32(num_components, bit_width, src); case nir_op_u2u64: return evaluate_u2u64(num_components, bit_width, src); case nir_op_u2u8: return evaluate_u2u8(num_components, bit_width, src); case nir_op_uadd_carry: return evaluate_uadd_carry(num_components, bit_width, src); case nir_op_ubfe: return evaluate_ubfe(num_components, bit_width, src); case nir_op_ubitfield_extract: return evaluate_ubitfield_extract(num_components, bit_width, src); case nir_op_udiv: return evaluate_udiv(num_components, bit_width, src); case nir_op_ufind_msb: return evaluate_ufind_msb(num_components, bit_width, src); case nir_op_uge: return evaluate_uge(num_components, bit_width, src); case nir_op_ult: return evaluate_ult(num_components, bit_width, src); case nir_op_umax: return evaluate_umax(num_components, bit_width, src); case nir_op_umax3: return evaluate_umax3(num_components, bit_width, src); case nir_op_umax_4x8: return evaluate_umax_4x8(num_components, bit_width, src); case nir_op_umed3: return evaluate_umed3(num_components, bit_width, src); case nir_op_umin: return evaluate_umin(num_components, bit_width, src); case nir_op_umin3: return evaluate_umin3(num_components, bit_width, src); case nir_op_umin_4x8: return evaluate_umin_4x8(num_components, bit_width, src); case nir_op_umod: return evaluate_umod(num_components, bit_width, src); case nir_op_umul_high: return evaluate_umul_high(num_components, bit_width, src); case nir_op_umul_unorm_4x8: return evaluate_umul_unorm_4x8(num_components, bit_width, src); case nir_op_unpack_32_2x16: return evaluate_unpack_32_2x16(num_components, bit_width, src); case nir_op_unpack_32_2x16_split_x: return evaluate_unpack_32_2x16_split_x(num_components, bit_width, src); case nir_op_unpack_32_2x16_split_y: return evaluate_unpack_32_2x16_split_y(num_components, bit_width, src); case nir_op_unpack_64_2x32: return evaluate_unpack_64_2x32(num_components, bit_width, src); case nir_op_unpack_64_2x32_split_x: return evaluate_unpack_64_2x32_split_x(num_components, bit_width, src); case nir_op_unpack_64_2x32_split_y: return evaluate_unpack_64_2x32_split_y(num_components, bit_width, src); case nir_op_unpack_64_4x16: return evaluate_unpack_64_4x16(num_components, bit_width, src); case nir_op_unpack_half_2x16: return evaluate_unpack_half_2x16(num_components, bit_width, src); case nir_op_unpack_half_2x16_split_x: return evaluate_unpack_half_2x16_split_x(num_components, bit_width, src); case nir_op_unpack_half_2x16_split_y: return evaluate_unpack_half_2x16_split_y(num_components, bit_width, src); case nir_op_unpack_snorm_2x16: return evaluate_unpack_snorm_2x16(num_components, bit_width, src); case nir_op_unpack_snorm_4x8: return evaluate_unpack_snorm_4x8(num_components, bit_width, src); case nir_op_unpack_unorm_2x16: return evaluate_unpack_unorm_2x16(num_components, bit_width, src); case nir_op_unpack_unorm_4x8: return evaluate_unpack_unorm_4x8(num_components, bit_width, src); case nir_op_usadd_4x8: return evaluate_usadd_4x8(num_components, bit_width, src); case nir_op_ushr: return evaluate_ushr(num_components, bit_width, src); case nir_op_ussub_4x8: return evaluate_ussub_4x8(num_components, bit_width, src); case nir_op_usub_borrow: return evaluate_usub_borrow(num_components, bit_width, src); case nir_op_vec2: return evaluate_vec2(num_components, bit_width, src); case nir_op_vec3: return evaluate_vec3(num_components, bit_width, src); case nir_op_vec4: return evaluate_vec4(num_components, bit_width, src); default: unreachable("shouldn't get here"); } }