diff options
Diffstat (limited to 'intern/cycles/util/types_float8_impl.h')
-rw-r--r-- | intern/cycles/util/types_float8_impl.h | 60 |
1 files changed, 26 insertions, 34 deletions
diff --git a/intern/cycles/util/types_float8_impl.h b/intern/cycles/util/types_float8_impl.h index 19818976b50..e8576cdaf70 100644 --- a/intern/cycles/util/types_float8_impl.h +++ b/intern/cycles/util/types_float8_impl.h @@ -2,87 +2,79 @@ * Original code Copyright 2017, Intel Corporation * Modifications Copyright 2018-2022 Blender Foundation. */ -#ifndef __UTIL_TYPES_FLOAT8_IMPL_H__ -#define __UTIL_TYPES_FLOAT8_IMPL_H__ +#pragma once #ifndef __UTIL_TYPES_H__ # error "Do not include this file directly, include util/types.h instead." #endif -#ifndef __KERNEL_GPU__ -# include <cstdio> -#endif - CCL_NAMESPACE_BEGIN -#ifndef __KERNEL_GPU__ -# ifdef __KERNEL_AVX2__ -__forceinline float8::float8() +#ifdef __KERNEL_AVX2__ +__forceinline float8_t::float8_t() { } -__forceinline float8::float8(const float8 &f) : m256(f.m256) +__forceinline float8_t::float8_t(const float8_t &f) : m256(f.m256) { } -__forceinline float8::float8(const __m256 &f) : m256(f) +__forceinline float8_t::float8_t(const __m256 &f) : m256(f) { } -__forceinline float8::operator const __m256 &() const +__forceinline float8_t::operator const __m256 &() const { return m256; } -__forceinline float8::operator __m256 &() +__forceinline float8_t::operator __m256 &() { return m256; } -__forceinline float8 &float8::operator=(const float8 &f) +__forceinline float8_t &float8_t::operator=(const float8_t &f) { m256 = f.m256; return *this; } -# endif /* __KERNEL_AVX2__ */ +#endif /* __KERNEL_AVX2__ */ -__forceinline float float8::operator[](int i) const +#ifndef __KERNEL_GPU__ +__forceinline float float8_t::operator[](int i) const { util_assert(i >= 0); util_assert(i < 8); return *(&a + i); } -__forceinline float &float8::operator[](int i) +__forceinline float &float8_t::operator[](int i) { util_assert(i >= 0); util_assert(i < 8); return *(&a + i); } +#endif -ccl_device_inline float8 make_float8(float f) +ccl_device_inline float8_t make_float8_t(float f) { -# ifdef __KERNEL_AVX2__ - float8 r(_mm256_set1_ps(f)); -# else - float8 r = {f, f, f, f, f, f, f, f}; -# endif +#ifdef __KERNEL_AVX2__ + float8_t r(_mm256_set1_ps(f)); +#else + float8_t r = {f, f, f, f, f, f, f, f}; +#endif return r; } -ccl_device_inline float8 -make_float8(float a, float b, float c, float d, float e, float f, float g, float h) +ccl_device_inline float8_t +make_float8_t(float a, float b, float c, float d, float e, float f, float g, float h) { -# ifdef __KERNEL_AVX2__ - float8 r(_mm256_set_ps(a, b, c, d, e, f, g, h)); -# else - float8 r = {a, b, c, d, e, f, g, h}; -# endif +#ifdef __KERNEL_AVX2__ + float8_t r(_mm256_setr_ps(a, b, c, d, e, f, g, h)); +#else + float8_t r = {a, b, c, d, e, f, g, h}; +#endif return r; } -#endif /* __KERNEL_GPU__ */ - CCL_NAMESPACE_END - -#endif /* __UTIL_TYPES_FLOAT8_IMPL_H__ */ |