| Chia-I Wu | d3e77a6 | 2014-08-18 14:39:31 +0800 | [diff] [blame] | 1 | /* |
| Courtney Goeltzenleuchter | 9cc421e | 2015-04-08 15:36:08 -0600 | [diff] [blame] | 2 | * Vulkan |
| Chia-I Wu | d3e77a6 | 2014-08-18 14:39:31 +0800 | [diff] [blame] | 3 | * |
| 4 | * Copyright (C) 2014 LunarG, Inc. |
| 5 | * |
| 6 | * Permission is hereby granted, free of charge, to any person obtaining a |
| 7 | * copy of this software and associated documentation files (the "Software"), |
| 8 | * to deal in the Software without restriction, including without limitation |
| 9 | * the rights to use, copy, modify, merge, publish, distribute, sublicense, |
| 10 | * and/or sell copies of the Software, and to permit persons to whom the |
| 11 | * Software is furnished to do so, subject to the following conditions: |
| 12 | * |
| 13 | * The above copyright notice and this permission notice shall be included |
| 14 | * in all copies or substantial portions of the Software. |
| 15 | * |
| 16 | * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR |
| 17 | * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, |
| 18 | * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL |
| 19 | * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER |
| 20 | * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING |
| 21 | * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER |
| 22 | * DEALINGS IN THE SOFTWARE. |
| Chia-I Wu | 44e4236 | 2014-09-02 08:32:09 +0800 | [diff] [blame] | 23 | * |
| 24 | * Authors: |
| 25 | * Chia-I Wu <olv@lunarg.com> |
| Chia-I Wu | d3e77a6 | 2014-08-18 14:39:31 +0800 | [diff] [blame] | 26 | */ |
| 27 | |
| 28 | #include "icd-utils.h" |
| 29 | |
| 30 | /* stolen from Mesa */ |
| 31 | uint16_t u_float_to_half(float f) |
| 32 | { |
| 33 | union fi { |
| 34 | float f; |
| 35 | uint32_t ui; |
| 36 | }; |
| 37 | |
| 38 | uint32_t sign_mask = 0x80000000; |
| 39 | uint32_t round_mask = ~0xfff; |
| 40 | uint32_t f32inf = 0xff << 23; |
| 41 | uint32_t f16inf = 0x1f << 23; |
| 42 | uint32_t sign; |
| 43 | union fi magic; |
| 44 | union fi f32; |
| 45 | uint16_t f16; |
| 46 | |
| 47 | magic.ui = 0xf << 23; |
| 48 | |
| 49 | f32.f = f; |
| 50 | |
| 51 | /* Sign */ |
| 52 | sign = f32.ui & sign_mask; |
| 53 | f32.ui ^= sign; |
| 54 | |
| 55 | if (f32.ui == f32inf) { |
| 56 | /* Inf */ |
| 57 | f16 = 0x7c00; |
| 58 | } else if (f32.ui > f32inf) { |
| 59 | /* NaN */ |
| 60 | f16 = 0x7e00; |
| 61 | } else { |
| 62 | /* Number */ |
| 63 | f32.ui &= round_mask; |
| 64 | f32.f *= magic.f; |
| 65 | f32.ui -= round_mask; |
| 66 | |
| 67 | /* |
| 68 | * Clamp to max finite value if overflowed. |
| 69 | * OpenGL has completely undefined rounding behavior for float to |
| 70 | * half-float conversions, and this matches what is mandated for float |
| 71 | * to fp11/fp10, which recommend round-to-nearest-finite too. |
| 72 | * (d3d10 is deeply unhappy about flushing such values to infinity, and |
| 73 | * while it also mandates round-to-zero it doesn't care nearly as much |
| 74 | * about that.) |
| 75 | */ |
| 76 | if (f32.ui > f16inf) |
| 77 | f32.ui = f16inf - 1; |
| 78 | |
| 79 | f16 = f32.ui >> 13; |
| 80 | } |
| 81 | |
| 82 | /* Sign */ |
| 83 | f16 |= sign >> 16; |
| 84 | |
| 85 | return f16; |
| 86 | } |