diff options
author | Martijn Berger <martijn.berger@gmail.com> | 2013-11-22 17:16:47 +0400 |
---|---|---|
committer | Brecht Van Lommel <brechtvanlommel@gmail.com> | 2013-11-22 17:42:41 +0400 |
commit | e3a79258d17e6cdca26120eab7a2c48c7c4d4a0f (patch) | |
tree | 77d59694458125dd7525faf59ed56ce505533981 /intern/cycles/kernel | |
parent | 5feb0d2bfe8f6723bf48073b1760b732bc6a5ceb (diff) |
Cycles: test code for sse 4.1 kernel and alignment for some vector types.
This is mostly work towards enabling the __KERNEL_SSE__ option to start using
SIMD operations for vector math operations. This 4.1 kernel performes about 8%
faster with that option but overall is still slower than without the option.
WITH_CYCLES_OPTIMIZED_KERNEL_SSE41 is the cmake flag for testing this kernel.
Alignment of int3, int4, float3, float4 to 16 bytes seems to give a slight 1-2%
speedup on tested systems with the current kernel already, so is enabled now.
Diffstat (limited to 'intern/cycles/kernel')
-rw-r--r-- | intern/cycles/kernel/CMakeLists.txt | 9 | ||||
-rw-r--r-- | intern/cycles/kernel/kernel.h | 9 | ||||
-rw-r--r-- | intern/cycles/kernel/kernel_sse41.cpp | 76 |
3 files changed, 92 insertions, 2 deletions
diff --git a/intern/cycles/kernel/CMakeLists.txt b/intern/cycles/kernel/CMakeLists.txt index 56ba0e08743..39444b91131 100644 --- a/intern/cycles/kernel/CMakeLists.txt +++ b/intern/cycles/kernel/CMakeLists.txt @@ -190,13 +190,18 @@ endif() include_directories(${INC}) include_directories(SYSTEM ${INC_SYS}) -add_library(cycles_kernel ${SRC} ${SRC_HEADERS} ${SRC_CLOSURE_HEADERS} ${SRC_SVM_HEADERS}) - if(WITH_CYCLES_OPTIMIZED_KERNEL) set_source_files_properties(kernel_sse2.cpp PROPERTIES COMPILE_FLAGS "${CYCLES_SSE2_KERNEL_FLAGS}") set_source_files_properties(kernel_sse3.cpp PROPERTIES COMPILE_FLAGS "${CYCLES_SSE3_KERNEL_FLAGS}") endif() +if(WITH_CYCLES_OPTIMIZED_KERNEL_SSE41) + set_source_files_properties(kernel_sse41.cpp PROPERTIES COMPILE_FLAGS "${CYCLES_SSE41_KERNEL_FLAGS}") + list(APPEND SRC kernel_sse41.cpp) +endif() + +add_library(cycles_kernel ${SRC} ${SRC_HEADERS} ${SRC_CLOSURE_HEADERS} ${SRC_SVM_HEADERS}) + if(WITH_CYCLES_CUDA) add_dependencies(cycles_kernel cycles_kernel_cuda) endif() diff --git a/intern/cycles/kernel/kernel.h b/intern/cycles/kernel/kernel.h index 361f5b0856d..105a3887da0 100644 --- a/intern/cycles/kernel/kernel.h +++ b/intern/cycles/kernel/kernel.h @@ -61,6 +61,15 @@ void kernel_cpu_sse3_convert_to_half_float(KernelGlobals *kg, uchar4 *rgba, floa float sample_scale, int x, int y, int offset, int stride); void kernel_cpu_sse3_shader(KernelGlobals *kg, uint4 *input, float4 *output, int type, int i); + +void kernel_cpu_sse41_path_trace(KernelGlobals *kg, float *buffer, unsigned int *rng_state, + int sample, int x, int y, int offset, int stride); +void kernel_cpu_sse41_convert_to_byte(KernelGlobals *kg, uchar4 *rgba, float *buffer, + float sample_scale, int x, int y, int offset, int stride); +void kernel_cpu_sse41_convert_to_half_float(KernelGlobals *kg, uchar4 *rgba, float *buffer, + float sample_scale, int x, int y, int offset, int stride); +void kernel_cpu_sse41_shader(KernelGlobals *kg, uint4 *input, float4 *output, + int type, int i); #endif CCL_NAMESPACE_END diff --git a/intern/cycles/kernel/kernel_sse41.cpp b/intern/cycles/kernel/kernel_sse41.cpp new file mode 100644 index 00000000000..0c68fd3651b --- /dev/null +++ b/intern/cycles/kernel/kernel_sse41.cpp @@ -0,0 +1,76 @@ +/* + * Copyright 2011-2013 Blender Foundation + * + * Licensed under the Apache License, Version 2.0 (the "License"); + * you may not use this file except in compliance with the License. + * You may obtain a copy of the License at + * + * http://www.apache.org/licenses/LICENSE-2.0 + * + * Unless required by applicable law or agreed to in writing, software + * distributed under the License is distributed on an "AS IS" BASIS, + * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. + * See the License for the specific language governing permissions and + * limitations under the License + */ + +/* Optimized CPU kernel entry points. This file is compiled with SSE3/SSSE3 + * optimization flags and nearly all functions inlined, while kernel.cpp + * is compiled without for other CPU's. */ + +#ifdef WITH_OPTIMIZED_KERNEL + +/* SSE optimization disabled for now on 32 bit, see bug #36316 */ +#if !(defined(__GNUC__) && (defined(i386) || defined(_M_IX86))) +#define __KERNEL_SSE2__ +#define __KERNEL_SSE3__ +#define __KERNEL_SSSE3__ +#define __KERNEL_SSE41__ +#endif + +#include "kernel.h" +#include "kernel_compat_cpu.h" +#include "kernel_math.h" +#include "kernel_types.h" +#include "kernel_globals.h" +#include "kernel_film.h" +#include "kernel_path.h" +#include "kernel_displace.h" + +CCL_NAMESPACE_BEGIN + +/* Path Tracing */ + +void kernel_cpu_sse41_path_trace(KernelGlobals *kg, float *buffer, unsigned int *rng_state, int sample, int x, int y, int offset, int stride) +{ +#ifdef __BRANCHED_PATH__ + if(kernel_data.integrator.branched) + kernel_branched_path_trace(kg, buffer, rng_state, sample, x, y, offset, stride); + else +#endif + kernel_path_trace(kg, buffer, rng_state, sample, x, y, offset, stride); +} + +/* Film */ + +void kernel_cpu_sse41_convert_to_byte(KernelGlobals *kg, uchar4 *rgba, float *buffer, float sample_scale, int x, int y, int offset, int stride) +{ + kernel_film_convert_to_byte(kg, rgba, buffer, sample_scale, x, y, offset, stride); +} + +void kernel_cpu_sse41_convert_to_half_float(KernelGlobals *kg, uchar4 *rgba, float *buffer, float sample_scale, int x, int y, int offset, int stride) +{ + kernel_film_convert_to_half_float(kg, rgba, buffer, sample_scale, x, y, offset, stride); +} + +/* Shader Evaluate */ + +void kernel_cpu_sse41_shader(KernelGlobals *kg, uint4 *input, float4 *output, int type, int i) +{ + kernel_shader_evaluate(kg, input, output, (ShaderEvalType)type, i); +} + +CCL_NAMESPACE_END + +#endif + |