ARM Cortex-A15 GCC Compiler Tuning Performance - Phoronix
ARM Cortex-A15 GCC Compiler Tuning Performance Written by Michael Larabel in GNU on 4 December 2012 at 09:41 AM EST. 3 Comments To complement the recent compiler benchmarking on the ARM Cortex-A15 as found in the Samsung Exynos 5 Dual with the Samsung Chromebook, here's some compiler tuning benchmark results from the speedy low-power ARM system. Uploaded to OpenBenchmarking.org this morning are a
marisa-trie ã® PopCount を改良ã—ã¾ã—㟠- ã‚„ãŸï¼ ã¯ã¦ãªæ—¥è¨˜
簡潔ビットベクトルã®åŸºæœ¬æ“作ã§ã‚ã‚‹ rank/select ã«ãŠã„ã¦é‡è¦ãªå½¹å‰²ã‚’æžœãŸã™ PopCount を改良ã™ã‚‹ã“ã¨ã«ã‚ˆã‚Šï¼Œmarisa-trie （Google Code Archive - Long-term storage for Google Code Project Hosting.） ã‚’å°‘ã—ã ã‘高速化ã™ã‚‹ã“ã¨ãŒã§ãã¾ã—ãŸï¼Žæˆæžœã¯ Subversion ã‹ã‚‰ã‚½ãƒ¼ã‚¹ã‚³ãƒ¼ãƒ‰ã‚’ダウンãƒãƒ¼ãƒ‰ã—ã¦ãƒ“ルドã™ã‚‹ã“ã¨ã«ã‚ˆã‚Šç¢ºèªã§ãã¾ã™ï¼Ž 一ã¤ç›®ã®æ”¹è‰¯ç‚¹ã¯ SSE4.2 ã§è¿½åŠ ã•ã‚ŒãŸ popcnt 命令ã§ã™ï¼Žãã®åã®é€šã‚Šï¼ŒPopCount 専用㮠CPU 命令ã§ã‚り，PopCount を高速ã«å®Ÿç¾ã§ãã¾ã™ï¼ŽãŸã ã—，ç¾çŠ¶ã§ã¯ä½¿ãˆã‚‹ CPU ãŒé™ã‚‰ã‚Œã¦ã„ã‚‹ãŸã‚，configure ã« --enable-popcnt を指定ã—ã¦ãƒ“ルドã—ãŸã¨ãã ã‘有効ã«ãªã‚‹ã‚ˆã†ã«ã—ã¦ã„ã¾ã™ï¼Ž 二ã¤ç›®ã®æ”¹è‰¯ç‚¹ã¯ pop
GCC -mtune 何やã£ã¦ã‚‹ã‹
æ¦‚è¦ gcc -mtune=hoge ã£ã¦ã‚ã‚“ã¾ã‚ˆãã‚ã‹ã‚‰ãšä»˜ã‘ã¦ã‚‹ã‘ã©æ„味ã‚ã‚“ã®ï¼Ÿ 効果 å‰ã®k10 vs coremaã¿ãŸã„ãªæ„Ÿã˜ã§parsecã‚’ã²ã¨ã¨ãŠã‚Šå‹•ã‹ã—㟠-O2 ã®æ™‚é–“ã‚’1.0ã¨ã—ãŸæ™‚ã®å‡¦ç†æ™‚é–“ã®æ¯” app |-O2 -mtune-native streamcluster | 1.016 canneal | 1.016 vips | 0.954 bodytrack | 0.966 x264 | 1.049 blackscholes | 0.986 swaptions | 0.981 ferret | 0.997 ãªã‚“らã‹ã®åŠ¹æžœãŒã‚る…よã†ã«è¦‹ãˆã‚‹â€¦ï¼Ÿ (ãƒã‚¿ãƒãƒ¬: -march=avxãŒä»˜ã„ã¦ã¦ã€ãã‚Œã®åŠ¹æžœãŒå‡ºã¦ã„ã‚‹) 超簡å˜ã«GCCã®ã‚³ãƒ¼ãƒ‰è§£èª¬ gcc/*.c アーã‚éžä¾å˜ã®ã‚³ãƒ¼ãƒ‰ gcc/config/i386/* i386ã®ã‚³ãƒ¼ãƒ‰ gcc/config/i386/i3
__builtin_expectã®åŠ¹åŠ› - åˆå¦è€…ã®ç®¸ç½®
RLogã‚’ã„ã˜ã£ã¦ã¦çŸ¥ã£ãŸ__builtin_expectを試ã—ã¦ã¿ãŸã€‚__builtin_expectã¯ã‚ã‚‹å¼ãŒã»ã¨ã‚“ã©ã®å ´åˆã«æ±ºã¾ã£ãŸå®šæ•°ã«ãªã‚‹ã€ã¨è¨€ã†å ´åˆã«åˆ†å²äºˆæ¸¬ã®ãƒ’ントãªã©ã‚’与ãˆã¦é«˜é€ŸåŒ–を計るãŸã‚ã®gccディレクティブã ãã†ãªã€‚ RLogã¯dormant(ä¼‘çœ )状態ã®ãƒã‚°ãƒ•ã‚¡ã‚·ãƒªãƒ†ã‚£ã«æœ€é©åŒ–ã—ã¦ã‚ã£ã¦ã€ãƒ—ãƒãƒ€ã‚¯ã‚·ãƒ§ãƒ³ã‚³ãƒ¼ãƒ‰ã«ãƒã‚°ã‚³ãƒ¼ãƒ‰ã‚’残ã—ã£ã±ãªã—ã§ã‚‚ã•ã»ã©ãƒ‡ã‚°ãƒ¬ã—ãªã„ã®ãŒå£²ã‚Šãªã‚“ã ã‘ã©ã€ãã®ã‚«ãƒ©ã‚¯ãƒªãŒã“れ。 ã§ã€RLog自身ã«ã¤ã„ã¦ã¯ã‚ã¨ã§æ›¸ã。 ã©ã†ã‹ã？ 具体的ã«ã¯ __builtin_expect(A,B)ã¨æ›¸ã„ãŸå ´åˆ AãŒå®šæ•° Bã§ã‚る事を期待ã™ã‚‹ã€ã¨ã„ã†ãƒ’ãƒ³ãƒˆæƒ…å ±ã«ãªã‚‹ã€‚ 例ãˆã°ã€æ¯”較演算åãŒã»ã¨ã‚“ã©ã®å ´åˆæˆã‚Šç«‹ãŸãªã„ã€ã¨è¨€ã†å ´åˆ __builtin_expect(, 0)　ã¨ãªã‚‹ã€‚ #include <stdio.h> #ifdef EXPECT #define E
CentOSã§GCCã®ãƒãƒ¼ã‚¸ãƒ§ãƒ³ã‚¢ãƒƒãƒ—ã‚’è¡Œã£ã¦ã¿ã‚‹Â | レンタルサーãƒãƒ¼ãƒ»è‡ªå®…サーãƒãƒ¼è¨å®šãƒ»æ§‹ç¯‰ã®ãƒ’ント
CentOSã§GCCã®ãƒãƒ¼ã‚¸ãƒ§ãƒ³ã‚¢ãƒƒãƒ—ã‚’è¡Œã£ã¦ã¿ã‚‹
[PDF] ã„ã¾ã©ãã˜ã‚ƒãªã„ アセンブラ入門 / ä¹…ä¿ç”°å±•è¡Œ
ã„ã¾ã©ãã˜ã‚ƒãªã„ アセンブラプãƒã‚°ãƒ©ãƒŸãƒ³ã‚° x86アセンブリ言語ã®åŸºç¤Žã‹ã‚‰SSEã¾ã§ 1 自己紹介 ï‚— ä¹…ä¿ç”°å±•è¡Œ â—¦ @nobu_k, id:nobu-q ï‚— 検索エンジンSedueを作ã£ã¦ã¾ã™ 2 本日ã®å†…容 ï‚— x86アセンブリ言語ã®åŸºç¤Ž ï‚— インラインアセンブラã§éŠã¶ ï‚— SSEを使ã£ã¦ã¿ã‚‹ 3 アセンブリ言語ã¨ã¯ ï‚— 機械語ã«è¿‘ã„プãƒã‚°ãƒ©ãƒŸãƒ³ã‚°è¨€èªž â—¦ 機械語ã¨ä¸€å¯¾ä¸€ã§å¯¾å¿œ 4 if (x < 0) { x = -x; } mov eax, [ebp + 8] cmp eax, 0 jge L1 neg eax L1: 8B 45 08 3D 00 00 00 00 7D 02 F7 D8 C言語 アセンブリ言語 機械語èªã‚€ã®ã¯åŽ³ã—ã„ ã¾ã マシ アセンブリ言語ã®ç‰¹å¾´ ï‚— 移æ¤æ€§ãŒä½Žã„ â—¦ CPUや処ç†ç³»ã«ã‚ˆã£ã¦ã™ã¹ã¦ãŒå¤‰ã‚ã‚‹ ï‚— èªã¿æ›¸ããŒå¤§å¤‰ â—¦ å¯èªæ€§ãŒã‚‚ã®ã™ã”ã低ㄠ
çŒ«ç§‘ç ”ç©¶æ‰€ - gcc option
MinGW上ã§é »ç¹ã«gccを利用ã—ã¦ã„ã‚‹ãã›ã«ã™ãã«å¿˜ã‚Œã‚‹ã®ã§ãƒ¡ãƒ¢ã€‚-D,-I,-i,-L,-lç‰ã®ä¸€èˆ¬çš„ã§åˆ†ã‹ã‚Šåˆ‡ã£ã¦ã‚‹ã‚‚ã®ã¯æ›¸ã‹ãªã„。éšæ™‚è¿½åŠ ã€‚ ãªãŠã€æ£ç¢ºã«ã¯GCCã®ã‚ªãƒ³ãƒ©ã‚¤ãƒ³ãƒžãƒ‹ãƒ¥ã‚¢ãƒ«ãŒå˜åœ¨ã™ã‚‹ã®ã§ãã¡ã‚‰ã‚’å‚ç…§ã™ã¹ã。オプションã®ä¸€è¦§ï¼ˆãƒªãƒ³ã‚¯å…ˆã¯4.5.xã®ã‚‚ã®ï¼‰ã‚‚ã‚る。ã ãŒã€ã‚ã¾ã‚Šã«è†¨å¤§ã™ãŽã‚‹ã®ã§ã“ã®è¨˜äº‹ã§ã¯ã€Œã‚ˆã使用ã™ã‚‹ãŒæ„味を忘れãŒã¡ãªã‚ªãƒ—ションã€ã«é–¢ã—ã¦ã®ãƒ¡ãƒ¢ã¨ã™ã‚‹ã€‚ -f(no-)strict-aliasing 厳密ãªåˆ¥åè¦ç´„(aliasing rule)ã«å‰‡ã£ã¦ã„ã‚‹ã¨ã¿ãªã™ã‹ã€å‰‡ã£ã¦ã„ãªã„ã¨ã¿ãªã™ã‹ã€‚intã®å¤‰æ•°ã«å¯¾ã—ã¦short*ã§ã‚¢ã‚¯ã‚»ã‚¹ã™ã‚‹ã‚ˆã†ãªè¡Œå„€ã®æ‚ªã„コードãŒç„¡ã„ã¨å®£è¨€ã§ãã‚‹ãªã‚‰-fstrict-aliasingã«ã™ã‚‹ã€‚ #include <stdio.h> int main(int argc, char* argv[]){ int x = 0; short*
yunoã®é›‘記帳 - gcc/optimize
gcc4.1ã®ã€-O3ã‚„-Osã§æœ‰åŠ¹ã«ãªã‚‰ãªã„最é©åŒ–オプションã®ãƒ¡ãƒ¢ã€‚ 概è¦ã ã‘ãªã®ã§ã€é©å®œãƒžãƒ‹ãƒ¥ã‚¢ãƒ«ã‚’å‚ç…§ã—ãŸæ–¹ãŒã„ã„ã§ã™ã€‚ -fforce-addr メモリアドレス定数を演算ã™ã‚‹å‰ã«ã€ãれらを強制的ã«ãƒ¬ã‚¸ã‚¹ã‚¿ã¸ã‚³ãƒ”ーã™ã‚‹ã€‚ -fmerge-all-constants åŒä¸€ã®å®šæ•°ã€åŒä¸€ã®å¤‰æ•°ã‚’マージã™ã‚‹ã€‚ -fmodulo-sched スイングモジュãƒã‚¹ã‚±ã‚¸ãƒ¥ãƒ¼ãƒªãƒ³ã‚°(ã£ã¦ãªã«?)ã‚’è¡Œã†ã€‚ -fgcse-sm ã‚°ãƒãƒ¼ãƒãƒ«å…±é€šéƒ¨åˆ†å¼å‰Šé™¤ã®å¾Œã«ã‚¹ãƒˆã‚¢ãƒ¢ãƒ¼ã‚·ãƒ§ãƒ³ãƒ‘スを実行ã™ã‚‹ã€‚ -fgcse-las ã‚°ãƒãƒ¼ãƒãƒ«å…±é€šéƒ¨åˆ†å¼å‰Šé™¤ã§å†—é•·ãªãƒ¡ãƒ¢ãƒªãƒãƒ¼ãƒ‰ã‚’削除ã™ã‚‹ã€‚ -floop-optimize2 æ–°ã—ã„ループオプティマイザを使ã£ã¦ãƒ«ãƒ¼ãƒ—最é©åŒ–を実行ã™ã‚‹ã€‚ -funsafe-loop-optimizations ループ変数ãŒã‚ªãƒ¼ãƒãƒ¼ãƒ•ãƒãƒ¼ã—ãªã„ã€è‡ªæ˜Žã§ãªã„終了æ¡ä»¶ã¯æœ‰é™ã€ãªã©ã®ä»®å®šã‚’è¨ã‘ã¦ãƒ«ãƒ¼ãƒ—を最é©
KMC Staff Blog:GCC LTO marge
2009å¹´09月30æ—¥ GCC LTO marge Googleã®Diego Novilloæ°ãŒã€GCCã®trunkã«å¤§é‡ã®ãƒ‘ッãƒã‚’投ã’ã¾ã—ãŸã€‚ [LTO merge][0/15] Description of the final 15 patches http://gcc.gnu.org/ml/gcc/2009-09/msg00578.html http://gcc.gnu.org/ml/gcc-patches/2009-09/ ã¤ã„ã«LTO(Link-Time Optimization)branchãŒGCC trunkã«mergeã•ã‚Œã‚‹ã‚ˆã†ã§ã™ã€‚ 具体的ã«ã¯ã€-fltoã‚„-fwhopr(Whole Program Optimization)ã¨ã„ã†ã‚ªãƒ—ションãŒå¢—ãˆã¾ã™ã€‚ã“ã®ã‚ªãƒ—ションã«ã‚ˆã‚Šã€GCCã¯é€šå¸¸ã®ã‚¢ã‚»ãƒ³ãƒ–リコードã®ä»£ã‚ã‚Šã«ã€æœ€é©åŒ–ã®ãŸã‚ã®æƒ…å ±ã‚’å«ã‚“ã GIMPLEã¨ã„ã†ä¸é–“表ç¾ã‚’ダ
))
整列ã—ãŸãƒ¡ãƒ¢ãƒªã‚’é™çš„ã«ç¢ºä¿ã™ã‚‹ attribute ((aligned(n)) - PS3 Linux Information Site / Cell/B.E.ã®ãƒ‘ワーを体験ã—よã†
æ供：PS3 Linux Information Site / Cell/B.E.ã®ãƒ‘ワーを体験ã—よㆠ基本的ãªaligned属性ã®ã¤ã‘ã‹ãŸ 変数を確ä¿ã™ã‚‹ã¨ã(Variable Attributes)ã‹ã€åž‹ã‚’宣言ã™ã‚‹ã¨ã(Type Attributes)ã« __attribute__((aligned(n))) ã‚’ã¤ã‘ã¦aligned属性を付与ã—ã¾ã™ã€‚ 両者ã¯ã»ã¼åŒã˜åŠ¹æžœã‚’æŒã¡ã¾ã™ãŒæ§‹é€ 体ã®å ´åˆã«ã¯ç•°ãªã‚‹åŠ¹æžœãŒã‚ã‚Šã¾ã™ã€‚ã“ã‚Œã«ã¤ã„ã¦ã¯å¾Œè¿°ã—ã¾ã™ã€‚ 以下ã®ä¾‹æ–‡ã§ã¯ n = 128 ã¨ã—ã¦128ãƒã‚¤ãƒˆå¢ƒç•Œã«æ•´åˆ—ã—ã¦ã„ã¾ã™ã€‚ Variable Attributes ã¾ãšå¤‰æ•°ã‚’確ä¿ã™ã‚‹ã¨ãã«aligned属性をã¤ã‘る例を紹介ã—ã¾ã™ã€‚ 組ã¿è¾¼ã¿åž‹ã‚„é…åˆ—ã‚„æ§‹é€ ä½“ã§ã‚ã£ã¦ã‚‚åŒã˜åŠ¹æžœã§ã€128ãƒã‚¤ãƒˆã«æ•´åˆ—ã—ãŸå¤‰æ•°ãŒå¾—られã¾ã™ã€‚ struct AAA; int a __attribute__((al
Auto-vectorization in GCC - GNU Project
The goal of this project was to develop a loop and basic block vectorizer in GCC, based on the tree-ssa framework. It has been completed and the functionality has been part of GCC for years. Table of Contents Latest News Contributing Using the Vectorizer Vectorizable Loops Unvectorizable Loops Previous News and Status References/Documentation High-Level Plan of Implementation Latest News 2011-10-2
ã‚ãã‚‹ã™ã®éžå¯¾ç§°æ—¥è¨˜:コンパイラã®æ€§èƒ½æ¯”較 - livedoor Blog（ブãƒã‚°ï¼‰
é™å®šçš„ãªãŒã‚‰ï¼Œå§«é‡Žãƒ™ãƒ³ãƒ(Sサイズ)ã§Cコンパイラã®æœ€é©åŒ–性能を 比較ã—ã¦ã¿ã¾ã—ãŸï¼Ž 比較ã«ä¾›ã—ãŸã‚³ãƒ³ãƒ‘イラã¯æ¬¡ã®ã¨ãŠã‚Šï¼Ž (A)gcc-4.1.2-14 (CentOS 5.1ã®æ¨™æº–コンパイラ) (B)gcc-4.3(20080403スナップショット) (C)Intel C compiler 10.1(Linuxフリー版) 最é©åŒ–オプションã¯æ¬¡ã®ã¨ãŠã‚Šï¼Ž (1)-m64 -Os -march=core2 -mfpmath=sse,387 -fomit-frame-pointer -DSMALL -ftree-vectorize -ftree-vectorizer-verbose=3 -ftracer -falign-loops -fpeel-loops -funroll-loops (2)-m64 -O3 -march=core2 -mfpmath=sse,387 -fomit-fram
リリースã€éšœå®³æƒ…å ±ãªã©ã®ã‚µãƒ¼ãƒ“スã®ãŠçŸ¥ã‚‰ã›
最新ã®äººæ°—エントリーã®é…ä¿¡
j次ã®ãƒ–ックマーク
kå‰ã®ãƒ–ックマーク
lã‚ã¨ã§èªã‚€
eコメント一覧を開ã
oページを開ã
Compiling with Address Sanitizer (ASAN) with CLANG and with GCC-4.8
GitHub - kennethreitz/osx-gcc-installer: GCC Installer for OSX! Without Xcode!
Compile error · Issue #5 · pezmaster31/bamtools
http://pkgs.org/centos-5-rhel-5/centos-rhel-x86_64/gcc44-c++-4.4.4-13.el5.x86_64.rpm.html
Adrian Tam :: Using SSE (SIMD extension) in GCC
IBM Developer
Vladimir Makarov - GCC-4.5.0 comparison with previous releases and LLVM-2.7 on SPEC2000 for
https://www.airs.com/dnovillo/Papers/rhs2006.pdf
{{#tags}}- {{label}}
{{/tags}}