test/intern/cycles/util/optimization.h

/* SPDX-FileCopyrightText: 2011-2022 Blender Foundation
 *
 * SPDX-License-Identifier: Apache-2.0 */

#ifndef __UTIL_OPTIMIZATION_H__
#define __UTIL_OPTIMIZATION_H__

#ifndef __KERNEL_GPU__

/* x86
 *
 * Compile a regular, SSE2 and SSE3 kernel. */

#  if defined(i386) || defined(_M_IX86)

/* We require minimum SSE2 support on x86, so auto enable. */
#    define __KERNEL_SSE2__
#    ifdef WITH_KERNEL_SSE2
#      define WITH_CYCLES_OPTIMIZED_KERNEL_SSE2
#    endif

/* x86-64
 *
 * Compile a regular (includes SSE2), SSE3, SSE 4.1, AVX and AVX2 kernel. */

#  elif defined(__x86_64__) || defined(_M_X64)

/* SSE2 is always available on x86-64 CPUs, so auto enable */
#    define __KERNEL_SSE2__
/* no SSE2 kernel on x86-64, part of regular kernel */
#    ifdef WITH_KERNEL_SSE41
#      define WITH_CYCLES_OPTIMIZED_KERNEL_SSE41
#    endif
#    ifdef WITH_KERNEL_AVX2
#      define WITH_CYCLES_OPTIMIZED_KERNEL_AVX2
#    endif

/* Arm Neon
 *
 * Compile a SSE4 kernel emulated with Neon. Most code is shared with
 * SSE, some specializations for performance and compatibility are made
 * made testing for __KERNEL_NEON__. */

#  elif defined(__ARM_NEON) && defined(WITH_SSE2NEON)

#    define __KERNEL_NEON__
#    define __KERNEL_SSE__
#    define __KERNEL_SSE2__
#    define __KERNEL_SSE3__
#    define __KERNEL_SSE41__

#  endif

#endif

#endif /* __UTIL_OPTIMIZATION_H__ */
License headers: use SPDX-FileCopyrightText in intern/cycles 2023-06-14 16:52:36 +10:00			`/* SPDX-FileCopyrightText: 2011-2022 Blender Foundation`
			`*`
			`* SPDX-License-Identifier: Apache-2.0 */`
Cycles: Skip the compilation of the dedicated SSE2 kernel on x86-64, we can assume SSE2 here, so just re-use the regular one. Saves 500kb in the blender binary. Reviewed by: brecht Differential Revision: https://developer.blender.org/D199 2014-01-14 20:39:21 +01:00
Code cleanup: some reshuffling of SIMD defines moving more code to util_optimization.h. 2014-01-15 15:11:50 +01:00			`#ifndef __UTIL_OPTIMIZATION_H__`
			`#define __UTIL_OPTIMIZATION_H__`

			`#ifndef __KERNEL_GPU__`

			`/* x86`
			`*`
			`* Compile a regular, SSE2 and SSE3 kernel. */`

			`# if defined(i386) \|\| defined(_M_IX86)`

Fix T52368: Cycles OSL trace() failing on Windows 32 bit. 2017-09-20 19:36:24 +02:00			`/* We require minimum SSE2 support on x86, so auto enable. */`
			`# define __KERNEL_SSE2__`
Cycles: Minor cleanup, whitespace around keyword and preprocessor indent 2016-04-13 08:58:52 +02:00			`# ifdef WITH_KERNEL_SSE2`
			`# define WITH_CYCLES_OPTIMIZED_KERNEL_SSE2`
			`# endif`
Code cleanup: some reshuffling of SIMD defines moving more code to util_optimization.h. 2014-01-15 15:11:50 +01:00
			`/* x86-64`
			`*`
Cleanup: Avoid some warnings on OS X with clang and update comment. 2015-10-25 20:48:28 +01:00			`* Compile a regular (includes SSE2), SSE3, SSE 4.1, AVX and AVX2 kernel. */`
Code cleanup: some reshuffling of SIMD defines moving more code to util_optimization.h. 2014-01-15 15:11:50 +01:00
Cycles: add support for Arm Neon instructions using sse2neon Based on patch contributed by Apple and Stefan Werner. Ref D8237, T78710 2021-02-14 15:01:26 +01:00			`# elif defined(__x86_64__) \|\| defined(_M_X64)`
Cycles: Skip the compilation of the dedicated SSE2 kernel on x86-64, we can assume SSE2 here, so just re-use the regular one. Saves 500kb in the blender binary. Reviewed by: brecht Differential Revision: https://developer.blender.org/D199 2014-01-14 20:39:21 +01:00
Code cleanup: some reshuffling of SIMD defines moving more code to util_optimization.h. 2014-01-15 15:11:50 +01:00			`/* SSE2 is always available on x86-64 CPUs, so auto enable */`
Cycles: Minor cleanup, whitespace around keyword and preprocessor indent 2016-04-13 08:58:52 +02:00			`# define __KERNEL_SSE2__`
Cycles: Skip the compilation of the dedicated SSE2 kernel on x86-64, we can assume SSE2 here, so just re-use the regular one. Saves 500kb in the blender binary. Reviewed by: brecht Differential Revision: https://developer.blender.org/D199 2014-01-14 20:39:21 +01:00			`/* no SSE2 kernel on x86-64, part of regular kernel */`
Cycles: Minor cleanup, whitespace around keyword and preprocessor indent 2016-04-13 08:58:52 +02:00			`# ifdef WITH_KERNEL_SSE41`
			`# define WITH_CYCLES_OPTIMIZED_KERNEL_SSE41`
			`# endif`
			`# ifdef WITH_KERNEL_AVX2`
			`# define WITH_CYCLES_OPTIMIZED_KERNEL_AVX2`
			`# endif`
Cycles: Add an AVX2 CPU kernel. This kernel is compiled with AVX2, FMA3, and BMI compiler flags. At the moment only Intel Haswell benefits from this, but future AMD CPUs will have these instructions as well. Makes rendering on Haswell CPUs a few percent faster, only benchmarked with clang on OS X though. Part of my GSoC 2014. 2014-06-13 22:23:58 +02:00
Cycles: add support for Arm Neon instructions using sse2neon Based on patch contributed by Apple and Stefan Werner. Ref D8237, T78710 2021-02-14 15:01:26 +01:00			`/* Arm Neon`
			`*`
			`* Compile a SSE4 kernel emulated with Neon. Most code is shared with`
			`* SSE, some specializations for performance and compatibility are made`
			`* made testing for __KERNEL_NEON__. */`

			`# elif defined(__ARM_NEON) && defined(WITH_SSE2NEON)`

			`# define __KERNEL_NEON__`
			`# define __KERNEL_SSE__`
			`# define __KERNEL_SSE2__`
			`# define __KERNEL_SSE3__`
			`# define __KERNEL_SSE41__`

			`# endif`
Cycles: Skip the compilation of the dedicated SSE2 kernel on x86-64, we can assume SSE2 here, so just re-use the regular one. Saves 500kb in the blender binary. Reviewed by: brecht Differential Revision: https://developer.blender.org/D199 2014-01-14 20:39:21 +01:00
Code cleanup: some reshuffling of SIMD defines moving more code to util_optimization.h. 2014-01-15 15:11:50 +01:00			`#endif`

			`#endif /* __UTIL_OPTIMIZATION_H__ */`