bartvdbraak/blender
1
0
Fork 4
Code Issues 4 Pull Requests 1 Actions 1 Packages Projects Releases Wiki Activity
ba7c2b7b73
blender/intern/cycles/util/util_system.cpp

251 lines
5.5 KiB
C++
Raw Normal View History

Cycles render engine, initial commit. This is the engine itself, blender modifications and build instructions will follow later. Cycles uses code from some great open source projects, many thanks them: * BVH building and traversal code from NVidia's "Understanding the Efficiency of Ray Traversal on GPUs": http://code.google.com/p/understanding-the-efficiency-of-ray-traversal-on-gpus/ * Open Shading Language for a large part of the shading system: http://code.google.com/p/openshadinglanguage/ * Blender for procedural textures and a few other nodes. * Approximate Catmull Clark subdivision from NVidia Mesh tools: http://code.google.com/p/nvidia-mesh-tools/ * Sobol direction vectors from: http://web.maths.unsw.edu.au/~fkuo/sobol/ * Film response functions from: http://www.cs.columbia.edu/CAVE/software/softlib/dorf.php
2011-04-27 11:58:34 +00:00
/*
Cycles: relicense GNU GPL source code to Apache version 2.0. More information in this post: http://code.blender.org/ Thanks to all contributes for giving their permission!
2013-08-18 14:16:15 +00:00
* Copyright 2011-2013 Blender Foundation
Cycles render engine, initial commit. This is the engine itself, blender modifications and build instructions will follow later. Cycles uses code from some great open source projects, many thanks them: * BVH building and traversal code from NVidia's "Understanding the Efficiency of Ray Traversal on GPUs": http://code.google.com/p/understanding-the-efficiency-of-ray-traversal-on-gpus/ * Open Shading Language for a large part of the shading system: http://code.google.com/p/openshadinglanguage/ * Blender for procedural textures and a few other nodes. * Approximate Catmull Clark subdivision from NVidia Mesh tools: http://code.google.com/p/nvidia-mesh-tools/ * Sobol direction vectors from: http://web.maths.unsw.edu.au/~fkuo/sobol/ * Film response functions from: http://www.cs.columbia.edu/CAVE/software/softlib/dorf.php
2011-04-27 11:58:34 +00:00
*
Cycles: relicense GNU GPL source code to Apache version 2.0. More information in this post: http://code.blender.org/ Thanks to all contributes for giving their permission!
2013-08-18 14:16:15 +00:00
* 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
Cycles render engine, initial commit. This is the engine itself, blender modifications and build instructions will follow later. Cycles uses code from some great open source projects, many thanks them: * BVH building and traversal code from NVidia's "Understanding the Efficiency of Ray Traversal on GPUs": http://code.google.com/p/understanding-the-efficiency-of-ray-traversal-on-gpus/ * Open Shading Language for a large part of the shading system: http://code.google.com/p/openshadinglanguage/ * Blender for procedural textures and a few other nodes. * Approximate Catmull Clark subdivision from NVidia Mesh tools: http://code.google.com/p/nvidia-mesh-tools/ * Sobol direction vectors from: http://web.maths.unsw.edu.au/~fkuo/sobol/ * Film response functions from: http://www.cs.columbia.edu/CAVE/software/softlib/dorf.php
2011-04-27 11:58:34 +00:00
*
Cycles: relicense GNU GPL source code to Apache version 2.0. More information in this post: http://code.blender.org/ Thanks to all contributes for giving their permission!
2013-08-18 14:16:15 +00:00
* http://www.apache.org/licenses/LICENSE-2.0
Cycles render engine, initial commit. This is the engine itself, blender modifications and build instructions will follow later. Cycles uses code from some great open source projects, many thanks them: * BVH building and traversal code from NVidia's "Understanding the Efficiency of Ray Traversal on GPUs": http://code.google.com/p/understanding-the-efficiency-of-ray-traversal-on-gpus/ * Open Shading Language for a large part of the shading system: http://code.google.com/p/openshadinglanguage/ * Blender for procedural textures and a few other nodes. * Approximate Catmull Clark subdivision from NVidia Mesh tools: http://code.google.com/p/nvidia-mesh-tools/ * Sobol direction vectors from: http://web.maths.unsw.edu.au/~fkuo/sobol/ * Film response functions from: http://www.cs.columbia.edu/CAVE/software/softlib/dorf.php
2011-04-27 11:58:34 +00:00
*
Cycles: relicense GNU GPL source code to Apache version 2.0. More information in this post: http://code.blender.org/ Thanks to all contributes for giving their permission!
2013-08-18 14:16:15 +00:00
* 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
Cleanup: Fix Cycles Apache header. This was already mixed a bit, but the dot belongs there.
2014-12-25 01:50:24 +00:00
* limitations under the License.
Cycles render engine, initial commit. This is the engine itself, blender modifications and build instructions will follow later. Cycles uses code from some great open source projects, many thanks them: * BVH building and traversal code from NVidia's "Understanding the Efficiency of Ray Traversal on GPUs": http://code.google.com/p/understanding-the-efficiency-of-ray-traversal-on-gpus/ * Open Shading Language for a large part of the shading system: http://code.google.com/p/openshadinglanguage/ * Blender for procedural textures and a few other nodes. * Approximate Catmull Clark subdivision from NVidia Mesh tools: http://code.google.com/p/nvidia-mesh-tools/ * Sobol direction vectors from: http://web.maths.unsw.edu.au/~fkuo/sobol/ * Film response functions from: http://www.cs.columbia.edu/CAVE/software/softlib/dorf.php
2011-04-27 11:58:34 +00:00
*/
#include "util_system.h"
Cycles: Use special debug panel to fine-tune debug flags This panel is only visible when debug_value is set to 256 and has no affect in other cases. However, if debug value is not set to this value, environment variables will be used to control which features are enabled, so there's no visible changes to anyone in fact. There are some changes needed to prevent devices re-enumeration on every Cycles session create. Reviewers: juicyfruit, lukasstockner97, dingto, brecht Reviewed By: lukasstockner97, dingto Differential Revision: https://developer.blender.org/D1720
2016-01-12 11:00:48 +00:00
#include "util_debug.h"
Cycles render engine, initial commit. This is the engine itself, blender modifications and build instructions will follow later. Cycles uses code from some great open source projects, many thanks them: * BVH building and traversal code from NVidia's "Understanding the Efficiency of Ray Traversal on GPUs": http://code.google.com/p/understanding-the-efficiency-of-ray-traversal-on-gpus/ * Open Shading Language for a large part of the shading system: http://code.google.com/p/openshadinglanguage/ * Blender for procedural textures and a few other nodes. * Approximate Catmull Clark subdivision from NVidia Mesh tools: http://code.google.com/p/nvidia-mesh-tools/ * Sobol direction vectors from: http://web.maths.unsw.edu.au/~fkuo/sobol/ * Film response functions from: http://www.cs.columbia.edu/CAVE/software/softlib/dorf.php
2011-04-27 11:58:34 +00:00
#include "util_types.h"
Cycles: Remove TM / R and whitespace from OpenCL device names. Was already done for CPU devices, now we also do this for OpenCL.
2015-05-21 21:43:18 +00:00
#include "util_string.h"
Cycles render engine, initial commit. This is the engine itself, blender modifications and build instructions will follow later. Cycles uses code from some great open source projects, many thanks them: * BVH building and traversal code from NVidia's "Understanding the Efficiency of Ray Traversal on GPUs": http://code.google.com/p/understanding-the-efficiency-of-ray-traversal-on-gpus/ * Open Shading Language for a large part of the shading system: http://code.google.com/p/openshadinglanguage/ * Blender for procedural textures and a few other nodes. * Approximate Catmull Clark subdivision from NVidia Mesh tools: http://code.google.com/p/nvidia-mesh-tools/ * Sobol direction vectors from: http://web.maths.unsw.edu.au/~fkuo/sobol/ * Film response functions from: http://www.cs.columbia.edu/CAVE/software/softlib/dorf.php
2011-04-27 11:58:34 +00:00
#ifdef _WIN32
Cycles: Re-implement some utilities to avoid use of boost The title says it all actually, the idea is to make Cycles only requiring Boost via 3rd party dependencies like OIIO and OSL. So now there are only few places which still uses Boost: - Foreach, function bindings and threading primitives. Those we can easily get rid with C++11 bump (which seems inevitable sooner or later if we'll want ot use newer LLVM for OSL), - Networking devices There's no quick solution for those currently, but there are some patches around which improves serialization. Reviewers: juicyfruit, mont29, campbellbarton, brecht, dingto Reviewed By: brecht, dingto Differential Revision: https://developer.blender.org/D1764
2016-02-05 08:09:39 +00:00
# if(!defined(FREE_WINDOWS))
# include <intrin.h>
# endif
# include "util_windows.h"
Cycles render engine, initial commit. This is the engine itself, blender modifications and build instructions will follow later. Cycles uses code from some great open source projects, many thanks them: * BVH building and traversal code from NVidia's "Understanding the Efficiency of Ray Traversal on GPUs": http://code.google.com/p/understanding-the-efficiency-of-ray-traversal-on-gpus/ * Open Shading Language for a large part of the shading system: http://code.google.com/p/openshadinglanguage/ * Blender for procedural textures and a few other nodes. * Approximate Catmull Clark subdivision from NVidia Mesh tools: http://code.google.com/p/nvidia-mesh-tools/ * Sobol direction vectors from: http://web.maths.unsw.edu.au/~fkuo/sobol/ * Film response functions from: http://www.cs.columbia.edu/CAVE/software/softlib/dorf.php
2011-04-27 11:58:34 +00:00
#elif defined(__APPLE__)
Cycles: Re-implement some utilities to avoid use of boost The title says it all actually, the idea is to make Cycles only requiring Boost via 3rd party dependencies like OIIO and OSL. So now there are only few places which still uses Boost: - Foreach, function bindings and threading primitives. Those we can easily get rid with C++11 bump (which seems inevitable sooner or later if we'll want ot use newer LLVM for OSL), - Networking devices There's no quick solution for those currently, but there are some patches around which improves serialization. Reviewers: juicyfruit, mont29, campbellbarton, brecht, dingto Reviewed By: brecht, dingto Differential Revision: https://developer.blender.org/D1764
2016-02-05 08:09:39 +00:00
# include <sys/sysctl.h>
# include <sys/types.h>
Cycles render engine, initial commit. This is the engine itself, blender modifications and build instructions will follow later. Cycles uses code from some great open source projects, many thanks them: * BVH building and traversal code from NVidia's "Understanding the Efficiency of Ray Traversal on GPUs": http://code.google.com/p/understanding-the-efficiency-of-ray-traversal-on-gpus/ * Open Shading Language for a large part of the shading system: http://code.google.com/p/openshadinglanguage/ * Blender for procedural textures and a few other nodes. * Approximate Catmull Clark subdivision from NVidia Mesh tools: http://code.google.com/p/nvidia-mesh-tools/ * Sobol direction vectors from: http://web.maths.unsw.edu.au/~fkuo/sobol/ * Film response functions from: http://www.cs.columbia.edu/CAVE/software/softlib/dorf.php
2011-04-27 11:58:34 +00:00
#else
Cycles: Re-implement some utilities to avoid use of boost The title says it all actually, the idea is to make Cycles only requiring Boost via 3rd party dependencies like OIIO and OSL. So now there are only few places which still uses Boost: - Foreach, function bindings and threading primitives. Those we can easily get rid with C++11 bump (which seems inevitable sooner or later if we'll want ot use newer LLVM for OSL), - Networking devices There's no quick solution for those currently, but there are some patches around which improves serialization. Reviewers: juicyfruit, mont29, campbellbarton, brecht, dingto Reviewed By: brecht, dingto Differential Revision: https://developer.blender.org/D1764
2016-02-05 08:09:39 +00:00
# include <unistd.h>
Cycles render engine, initial commit. This is the engine itself, blender modifications and build instructions will follow later. Cycles uses code from some great open source projects, many thanks them: * BVH building and traversal code from NVidia's "Understanding the Efficiency of Ray Traversal on GPUs": http://code.google.com/p/understanding-the-efficiency-of-ray-traversal-on-gpus/ * Open Shading Language for a large part of the shading system: http://code.google.com/p/openshadinglanguage/ * Blender for procedural textures and a few other nodes. * Approximate Catmull Clark subdivision from NVidia Mesh tools: http://code.google.com/p/nvidia-mesh-tools/ * Sobol direction vectors from: http://web.maths.unsw.edu.au/~fkuo/sobol/ * Film response functions from: http://www.cs.columbia.edu/CAVE/software/softlib/dorf.php
2011-04-27 11:58:34 +00:00
#endif
CCL_NAMESPACE_BEGIN
int system_cpu_thread_count()
{
static uint count = 0;
if(count > 0)
return count;
#ifdef _WIN32
SYSTEM_INFO info;
GetSystemInfo(&info);
count = (uint)info.dwNumberOfProcessors;
#elif defined(__APPLE__)
size_t len = sizeof(count);
int mib[2] = { CTL_HW, HW_NCPU };
sysctl(mib, 2, &count, &len, NULL, 0);
#else
count = (uint)sysconf(_SC_NPROCESSORS_ONLN);
#endif
if(count < 1)
count = 1;
return count;
}
Compile fixes for MinGW, checked with basic initial setup, no CUDA, gcc 4.5.2. OpenImageIO still gives link-time errors, will try to make a lib for MinGW, see if it is fixed.
2011-09-04 15:39:09 +00:00
#if !defined(_WIN32) || defined(FREE_WINDOWS)
Cycles render engine, initial commit. This is the engine itself, blender modifications and build instructions will follow later. Cycles uses code from some great open source projects, many thanks them: * BVH building and traversal code from NVidia's "Understanding the Efficiency of Ray Traversal on GPUs": http://code.google.com/p/understanding-the-efficiency-of-ray-traversal-on-gpus/ * Open Shading Language for a large part of the shading system: http://code.google.com/p/openshadinglanguage/ * Blender for procedural textures and a few other nodes. * Approximate Catmull Clark subdivision from NVidia Mesh tools: http://code.google.com/p/nvidia-mesh-tools/ * Sobol direction vectors from: http://web.maths.unsw.edu.au/~fkuo/sobol/ * Film response functions from: http://www.cs.columbia.edu/CAVE/software/softlib/dorf.php
2011-04-27 11:58:34 +00:00
static void __cpuid(int data[4], int selector)
{
Cycles: two 32 bit fixes with help from IRC user Agiofws.
2011-05-05 13:51:33 +00:00
#ifdef __x86_64__
Cycles render engine, initial commit. This is the engine itself, blender modifications and build instructions will follow later. Cycles uses code from some great open source projects, many thanks them: * BVH building and traversal code from NVidia's "Understanding the Efficiency of Ray Traversal on GPUs": http://code.google.com/p/understanding-the-efficiency-of-ray-traversal-on-gpus/ * Open Shading Language for a large part of the shading system: http://code.google.com/p/openshadinglanguage/ * Blender for procedural textures and a few other nodes. * Approximate Catmull Clark subdivision from NVidia Mesh tools: http://code.google.com/p/nvidia-mesh-tools/ * Sobol direction vectors from: http://web.maths.unsw.edu.au/~fkuo/sobol/ * Film response functions from: http://www.cs.columbia.edu/CAVE/software/softlib/dorf.php
2011-04-27 11:58:34 +00:00
asm("cpuid" : "=a" (data[0]), "=b" (data[1]), "=c" (data[2]), "=d" (data[3]) : "a"(selector));
Cycles: two 32 bit fixes with help from IRC user Agiofws.
2011-05-05 13:51:33 +00:00
#else
#ifdef __i386__
asm("pushl %%ebx \n\t"
"cpuid \n\t"
"movl %%ebx, %1 \n\t"
"popl %%ebx \n\t" : "=a" (data[0]), "=r" (data[1]), "=c" (data[2]), "=d" (data[3]) : "a"(selector));
#else
data[0] = data[1] = data[2] = data[3] = 0;
#endif
#endif
Cycles render engine, initial commit. This is the engine itself, blender modifications and build instructions will follow later. Cycles uses code from some great open source projects, many thanks them: * BVH building and traversal code from NVidia's "Understanding the Efficiency of Ray Traversal on GPUs": http://code.google.com/p/understanding-the-efficiency-of-ray-traversal-on-gpus/ * Open Shading Language for a large part of the shading system: http://code.google.com/p/openshadinglanguage/ * Blender for procedural textures and a few other nodes. * Approximate Catmull Clark subdivision from NVidia Mesh tools: http://code.google.com/p/nvidia-mesh-tools/ * Sobol direction vectors from: http://web.maths.unsw.edu.au/~fkuo/sobol/ * Film response functions from: http://www.cs.columbia.edu/CAVE/software/softlib/dorf.php
2011-04-27 11:58:34 +00:00
}
#endif
string system_cpu_brand_string()
{
char buf[48];
int result[4];
__cpuid(result, 0x80000000);
if(result[0] >= (int)0x80000004) {
__cpuid((int*)(buf+0), 0x80000002);
__cpuid((int*)(buf+16), 0x80000003);
__cpuid((int*)(buf+32), 0x80000004);
string brand = buf;
/* make it a bit more presentable */
Cycles: Remove TM / R and whitespace from OpenCL device names. Was already done for CPU devices, now we also do this for OpenCL.
2015-05-21 21:43:18 +00:00
brand = string_remove_trademark(brand);
Cycles render engine, initial commit. This is the engine itself, blender modifications and build instructions will follow later. Cycles uses code from some great open source projects, many thanks them: * BVH building and traversal code from NVidia's "Understanding the Efficiency of Ray Traversal on GPUs": http://code.google.com/p/understanding-the-efficiency-of-ray-traversal-on-gpus/ * Open Shading Language for a large part of the shading system: http://code.google.com/p/openshadinglanguage/ * Blender for procedural textures and a few other nodes. * Approximate Catmull Clark subdivision from NVidia Mesh tools: http://code.google.com/p/nvidia-mesh-tools/ * Sobol direction vectors from: http://web.maths.unsw.edu.au/~fkuo/sobol/ * Film response functions from: http://www.cs.columbia.edu/CAVE/software/softlib/dorf.php
2011-04-27 11:58:34 +00:00
return brand;
}
return "Unknown CPU";
}
Cycles: some warning fixes, cpu device task tweaks, avoid unnecessary tonemap in non-viewport render, and some utility functions.
2011-09-08 18:58:07 +00:00
int system_cpu_bits()
{
return (sizeof(void*)*8);
}
Better fix for #36935 and 36316: * 32 bit GCC builds now have the SSE BVH optimizations turned off, but still compile with SSE flags for better performance. * White color when rendering on Windows seems to have been unrelated to SSE, rather it was a graphics driver not supporting half float textures, added a check for that now.
2013-10-05 19:56:34 +00:00
#if defined(__x86_64__) || defined(_M_X64) || defined(i386) || defined(_M_IX86)
Fix #29259: cycles issues on certain processors. Now two versions of the kernel are compiled, one SSE optimized and the other not, and it will choose between them at runtime.
2011-11-15 15:13:38 +00:00
struct CPUCapabilities {
bool x64;
bool mmx;
bool sse;
bool sse2;
bool sse3;
bool ssse3;
bool sse41;
bool sse42;
bool sse4a;
bool avx;
Cycles: check for F16C support with __cpuid, as we do for BMI and BMI2
2015-05-11 15:49:36 +00:00
bool f16c;
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 20:23:58 +00:00
bool avx2;
Fix #29259: cycles issues on certain processors. Now two versions of the kernel are compiled, one SSE optimized and the other not, and it will choose between them at runtime.
2011-11-15 15:13:38 +00:00
bool xop;
bool fma3;
bool fma4;
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 20:23:58 +00:00
bool bmi1;
bool bmi2;
Fix #29259: cycles issues on certain processors. Now two versions of the kernel are compiled, one SSE optimized and the other not, and it will choose between them at runtime.
2011-11-15 15:13:38 +00:00
};
Fix cycles intersection issue with overlapping faces on windows 32 bit and CPU without SSE3 support, due to 80 bit precision float register being used for one bounding box but not the one next to it.
2013-02-04 16:12:37 +00:00
static CPUCapabilities& system_cpu_capabilities()
Fix #29259: cycles issues on certain processors. Now two versions of the kernel are compiled, one SSE optimized and the other not, and it will choose between them at runtime.
2011-11-15 15:13:38 +00:00
{
static CPUCapabilities caps;
static bool caps_init = false;
if(!caps_init) {
Cleanup: Remove unused Cycles cpuid code.
2015-02-11 10:28:42 +00:00
int result[4], num;
Fix #29259: cycles issues on certain processors. Now two versions of the kernel are compiled, one SSE optimized and the other not, and it will choose between them at runtime.
2011-11-15 15:13:38 +00:00
memset(&caps, 0, sizeof(caps));
__cpuid(result, 0);
num = result[0];
style cleanup
2012-06-06 22:38:39 +00:00
if(num >= 1) {
Fix #29259: cycles issues on certain processors. Now two versions of the kernel are compiled, one SSE optimized and the other not, and it will choose between them at runtime.
2011-11-15 15:13:38 +00:00
__cpuid(result, 0x00000001);
caps.mmx = (result[3] & ((int)1 << 23)) != 0;
caps.sse = (result[3] & ((int)1 << 25)) != 0;
caps.sse2 = (result[3] & ((int)1 << 26)) != 0;
caps.sse3 = (result[2] & ((int)1 << 0)) != 0;
caps.ssse3 = (result[2] & ((int)1 << 9)) != 0;
caps.sse41 = (result[2] & ((int)1 << 19)) != 0;
caps.sse42 = (result[2] & ((int)1 << 20)) != 0;
caps.fma3 = (result[2] & ((int)1 << 12)) != 0;
Fix T38815 For AVX support we need to check both OS support and CPU support.
2014-02-25 16:57:05 +00:00
caps.avx = false;
Fix missing brackets in cpuid bitfield check. concern raised by lukas_t (rBef73d547cc7c663ad180721094c81b3c81482ac3)
2014-02-25 18:50:53 +00:00
bool os_uses_xsave_xrestore = (result[2] & ((int)1 << 27)) != 0;
Fix T38815 For AVX support we need to check both OS support and CPU support.
2014-02-25 16:57:05 +00:00
bool cpu_avx_support = (result[2] & ((int)1 << 28)) != 0;
if( os_uses_xsave_xrestore && cpu_avx_support) {
// Check if the OS will save the YMM registers
uint32_t xcr_feature_mask;
Style cleanup: indentation, braces
2014-05-04 16:19:08 +00:00
#if defined(__GNUC__)
int edx; /* not used */
/* actual opcode for xgetbv */
__asm__ (".byte 0x0f, 0x01, 0xd0" : "=a" (xcr_feature_mask) , "=d" (edx) : "c" (0) );
#elif defined(_MSC_VER) && defined(_XCR_XFEATURE_ENABLED_MASK)
xcr_feature_mask = (uint32_t)_xgetbv(_XCR_XFEATURE_ENABLED_MASK); /* min VS2010 SP1 compiler is required */
#else
xcr_feature_mask = 0;
#endif
Fix T38815 For AVX support we need to check both OS support and CPU support.
2014-02-25 16:57:05 +00:00
caps.avx = (xcr_feature_mask & 0x6) == 0x6;
}
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 20:23:58 +00:00
Cycles: check for F16C support with __cpuid, as we do for BMI and BMI2
2015-05-11 15:49:36 +00:00
caps.f16c = (result[2] & ((int)1 << 29)) != 0;
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 20:23:58 +00:00
__cpuid(result, 0x00000007);
caps.bmi1 = (result[1] & ((int)1 << 3)) != 0;
caps.bmi2 = (result[1] & ((int)1 << 8)) != 0;
caps.avx2 = (result[1] & ((int)1 << 5)) != 0;
Fix #29259: cycles issues on certain processors. Now two versions of the kernel are compiled, one SSE optimized and the other not, and it will choose between them at runtime.
2011-11-15 15:13:38 +00:00
}
caps_init = true;
}
Fix cycles intersection issue with overlapping faces on windows 32 bit and CPU without SSE3 support, due to 80 bit precision float register being used for one bounding box but not the one next to it.
2013-02-04 16:12:37 +00:00
return caps;
}
bool system_cpu_support_sse2()
{
CPUCapabilities& caps = system_cpu_capabilities();
Cycles: Use special debug panel to fine-tune debug flags This panel is only visible when debug_value is set to 256 and has no affect in other cases. However, if debug value is not set to this value, environment variables will be used to control which features are enabled, so there's no visible changes to anyone in fact. There are some changes needed to prevent devices re-enumeration on every Cycles session create. Reviewers: juicyfruit, lukasstockner97, dingto, brecht Reviewed By: lukasstockner97, dingto Differential Revision: https://developer.blender.org/D1720
2016-01-12 11:00:48 +00:00
return DebugFlags().cpu.sse2 && caps.sse && caps.sse2;
Fix cycles intersection issue with overlapping faces on windows 32 bit and CPU without SSE3 support, due to 80 bit precision float register being used for one bounding box but not the one next to it.
2013-02-04 16:12:37 +00:00
}
bool system_cpu_support_sse3()
{
CPUCapabilities& caps = system_cpu_capabilities();
Cycles: Use special debug panel to fine-tune debug flags This panel is only visible when debug_value is set to 256 and has no affect in other cases. However, if debug value is not set to this value, environment variables will be used to control which features are enabled, so there's no visible changes to anyone in fact. There are some changes needed to prevent devices re-enumeration on every Cycles session create. Reviewers: juicyfruit, lukasstockner97, dingto, brecht Reviewed By: lukasstockner97, dingto Differential Revision: https://developer.blender.org/D1720
2016-01-12 11:00:48 +00:00
return DebugFlags().cpu.sse3 &&
caps.sse && caps.sse2 && caps.sse3 && caps.ssse3;
Fix #29259: cycles issues on certain processors. Now two versions of the kernel are compiled, one SSE optimized and the other not, and it will choose between them at runtime.
2011-11-15 15:13:38 +00:00
}
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.
2013-11-22 13:16:47 +00:00
bool system_cpu_support_sse41()
{
CPUCapabilities& caps = system_cpu_capabilities();
Cycles: Use special debug panel to fine-tune debug flags This panel is only visible when debug_value is set to 256 and has no affect in other cases. However, if debug value is not set to this value, environment variables will be used to control which features are enabled, so there's no visible changes to anyone in fact. There are some changes needed to prevent devices re-enumeration on every Cycles session create. Reviewers: juicyfruit, lukasstockner97, dingto, brecht Reviewed By: lukasstockner97, dingto Differential Revision: https://developer.blender.org/D1720
2016-01-12 11:00:48 +00:00
return DebugFlags().cpu.sse41 &&
caps.sse && caps.sse2 && caps.sse3 && caps.ssse3 && caps.sse41;
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.
2013-11-22 13:16:47 +00:00
}
Cycles: Add an AVX kernel for CPU rendering. * AVX is available on Intel Sandy Bridge and newer and AMD Bulldozer and newer. * We don't use dedicated AVX intrinsics yet, but gcc auto vectorization gives a 3% performance improvement for Caminandes. Tested on an i5-3570, Linux x64. * No change for Windows yet, MSVC 2008 does not support AVX. Reviewed by: brecht Differential Revision: https://developer.blender.org/D216
2014-01-16 16:04:11 +00:00
bool system_cpu_support_avx()
{
CPUCapabilities& caps = system_cpu_capabilities();
Cycles: Use special debug panel to fine-tune debug flags This panel is only visible when debug_value is set to 256 and has no affect in other cases. However, if debug value is not set to this value, environment variables will be used to control which features are enabled, so there's no visible changes to anyone in fact. There are some changes needed to prevent devices re-enumeration on every Cycles session create. Reviewers: juicyfruit, lukasstockner97, dingto, brecht Reviewed By: lukasstockner97, dingto Differential Revision: https://developer.blender.org/D1720
2016-01-12 11:00:48 +00:00
return DebugFlags().cpu.avx &&
caps.sse && caps.sse2 && caps.sse3 && caps.ssse3 && caps.sse41 && caps.avx;
Cycles: Add an AVX kernel for CPU rendering. * AVX is available on Intel Sandy Bridge and newer and AMD Bulldozer and newer. * We don't use dedicated AVX intrinsics yet, but gcc auto vectorization gives a 3% performance improvement for Caminandes. Tested on an i5-3570, Linux x64. * No change for Windows yet, MSVC 2008 does not support AVX. Reviewed by: brecht Differential Revision: https://developer.blender.org/D216
2014-01-16 16:04:11 +00:00
}
Cycles: Use special debug panel to fine-tune debug flags This panel is only visible when debug_value is set to 256 and has no affect in other cases. However, if debug value is not set to this value, environment variables will be used to control which features are enabled, so there's no visible changes to anyone in fact. There are some changes needed to prevent devices re-enumeration on every Cycles session create. Reviewers: juicyfruit, lukasstockner97, dingto, brecht Reviewed By: lukasstockner97, dingto Differential Revision: https://developer.blender.org/D1720
2016-01-12 11:00:48 +00:00
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 20:23:58 +00:00
bool system_cpu_support_avx2()
{
CPUCapabilities& caps = system_cpu_capabilities();
Cycles: Use special debug panel to fine-tune debug flags This panel is only visible when debug_value is set to 256 and has no affect in other cases. However, if debug value is not set to this value, environment variables will be used to control which features are enabled, so there's no visible changes to anyone in fact. There are some changes needed to prevent devices re-enumeration on every Cycles session create. Reviewers: juicyfruit, lukasstockner97, dingto, brecht Reviewed By: lukasstockner97, dingto Differential Revision: https://developer.blender.org/D1720
2016-01-12 11:00:48 +00:00
return DebugFlags().cpu.avx2 &&
caps.sse && caps.sse2 && caps.sse3 && caps.ssse3 && caps.sse41 && caps.avx && caps.f16c && caps.avx2 && caps.fma3 && caps.bmi1 && caps.bmi2;
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 20:23:58 +00:00
}
Fix #29259: cycles issues on certain processors. Now two versions of the kernel are compiled, one SSE optimized and the other not, and it will choose between them at runtime.
2011-11-15 15:13:38 +00:00
#else
Fix build error on non-x86 architectures as pointed out by Jochen Schmitt.
2013-06-04 11:21:13 +00:00
bool system_cpu_support_sse2()
{
return false;
}
bool system_cpu_support_sse3()
Fix #29259: cycles issues on certain processors. Now two versions of the kernel are compiled, one SSE optimized and the other not, and it will choose between them at runtime.
2011-11-15 15:13:38 +00:00
{
return false;
}
Cycles: Add an AVX kernel for CPU rendering. * AVX is available on Intel Sandy Bridge and newer and AMD Bulldozer and newer. * We don't use dedicated AVX intrinsics yet, but gcc auto vectorization gives a 3% performance improvement for Caminandes. Tested on an i5-3570, Linux x64. * No change for Windows yet, MSVC 2008 does not support AVX. Reviewed by: brecht Differential Revision: https://developer.blender.org/D216
2014-01-16 16:04:11 +00:00
bool system_cpu_support_sse41()
{
return false;
}
bool system_cpu_support_avx()
{
return false;
}
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 20:23:58 +00:00
bool system_cpu_support_avx2()
{
return false;
}
Cycles: Add an AVX kernel for CPU rendering. * AVX is available on Intel Sandy Bridge and newer and AMD Bulldozer and newer. * We don't use dedicated AVX intrinsics yet, but gcc auto vectorization gives a 3% performance improvement for Caminandes. Tested on an i5-3570, Linux x64. * No change for Windows yet, MSVC 2008 does not support AVX. Reviewed by: brecht Differential Revision: https://developer.blender.org/D216
2014-01-16 16:04:11 +00:00
Fix #29259: cycles issues on certain processors. Now two versions of the kernel are compiled, one SSE optimized and the other not, and it will choose between them at runtime.
2011-11-15 15:13:38 +00:00
#endif
Cycles render engine, initial commit. This is the engine itself, blender modifications and build instructions will follow later. Cycles uses code from some great open source projects, many thanks them: * BVH building and traversal code from NVidia's "Understanding the Efficiency of Ray Traversal on GPUs": http://code.google.com/p/understanding-the-efficiency-of-ray-traversal-on-gpus/ * Open Shading Language for a large part of the shading system: http://code.google.com/p/openshadinglanguage/ * Blender for procedural textures and a few other nodes. * Approximate Catmull Clark subdivision from NVidia Mesh tools: http://code.google.com/p/nvidia-mesh-tools/ * Sobol direction vectors from: http://web.maths.unsw.edu.au/~fkuo/sobol/ * Film response functions from: http://www.cs.columbia.edu/CAVE/software/softlib/dorf.php
2011-04-27 11:58:34 +00:00
CCL_NAMESPACE_END
Reference in New Issue Copy Permalink