{"id":7442,"date":"2026-07-22T12:33:54","date_gmt":"2026-07-22T04:33:54","guid":{"rendered":"https:\/\/i007.cc\/wordpress\/?p=7442"},"modified":"2026-07-22T12:33:54","modified_gmt":"2026-07-22T04:33:54","slug":"%e7%94%a8-google-benchmark-%e5%ae%9e%e6%b5%8b%e5%af%b9%e6%af%94%e6%a0%87%e9%87%8f-vs-%e8%87%aa%e5%8a%a8%e5%90%91%e9%87%8f%e5%8c%96-vs-%e6%89%8b%e5%8a%a8simd%e4%b8%89%e7%a7%8d%e5%ae%9e%e7%8e%b0","status":"publish","type":"post","link":"https:\/\/i007.cc\/wordpress\/archives\/7442","title":{"rendered":"\u7528 Google Benchmark \u5b9e\u6d4b\u5bf9\u6bd4”\u6807\u91cf vs \u81ea\u52a8\u5411\u91cf\u5316 vs \u624b\u52a8SIMD”\u4e09\u79cd\u5b9e\u73b0\u7684\u6027\u80fd\u5dee\u5f02\uff1f"},"content":{"rendered":"
\n
\n

\u6211\u6765\u5199\u4e00\u4e2a\u5b8c\u6574\u7684\u3001\u53ef\u76f4\u63a5\u7f16\u8bd1\u8fd0\u884c\u7684\u5bf9\u6bd4\u5b9e\u9a8c\u3002<\/p>\n<\/div>\n<\/div>\n

\n
\n
\n
\n
\n
\n
\u68c0\u67e5\u7f16\u8bd1\u5668\u548cCPU\u652f\u6301\u7684\u6307\u4ee4\u96c6<\/span><\/div>\n<\/div>\n

\u68c0\u67e5\u7f16\u8bd1\u5668\u548cCPU\u652f\u6301\u7684\u6307\u4ee4\u96c6<\/span><\/p>\n

\n
<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n
\n
\n
\n
\n
\n

\u597d\uff0c\u73af\u5883\u652f\u6301\u5230 AVX-512\u3002\u6211\u5148\u786e\u8ba4 Google Benchmark \u5e93\u662f\u5426\u53ef\u7528\uff0c\u6ca1\u6709\u7684\u8bdd\u7528\u7eaf chrono<\/code> \u591a\u6b21\u91c7\u6837\u7684\u65b9\u5f0f\u5b9e\u73b0\uff0c\u907f\u514d\u7f51\u7edc\u4f9d\u8d56\u95ee\u9898\u3002<\/p>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n

\n
\n
\n
\n
\n
\n
Ran 2 commands<\/span><\/div>\n<\/div>\n

Ran 2 commands<\/span><\/p>\n

\n
<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n
\n
\n
\n
\n
\n

\u6ca1\u6709\u7f51\u7edc\u8bbf\u95ee\u6743\u9650\uff0c\u6ca1\u6cd5\u5b89\u88c5 Google Benchmark \u5e93\u3002\u6211\u6539\u7528\u7eaf std::chrono<\/code> + \u591a\u6b21\u91cd\u590d\u91c7\u6837\u53d6\u4e2d\u4f4d\u6570<\/strong>\u7684\u65b9\u5f0f\u6765\u5b9e\u73b0\uff0c\u6548\u679c\u4e00\u6837\u80fd\u8bf4\u660e\u95ee\u9898\uff0c\u800c\u4e14\u4ee3\u7801\u66f4\u7b80\u6d01\u3001\u4e0d\u9700\u8981\u989d\u5916\u4f9d\u8d56\uff0c\u4f60\u76f4\u63a5\u590d\u5236\u5c31\u80fd\u7f16\u8bd1\u8fd0\u884c\u3002<\/p>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n

\n
\n
\n
\n
\n
\n
Created a file, ran a command<\/span><\/div>\n<\/div>\n

Created a file, ran a command<\/span><\/p>\n

\n
<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n
\n
\n
\n
\n
\n

\u4e24\u4e2a\u7248\u672c\u90fd\u7f16\u8bd1\u6210\u529f\u4e86\u3002\u5148\u8dd1”\u7eaf\u6807\u91cf\u57fa\u7ebf”\uff08\u5173\u95ed\u4e86\u81ea\u52a8\u5411\u91cf\u5316\uff0c\u8fd9\u662f\u6700\u516c\u5e73\u7684\u5bf9\u7167\u7ec4\uff09\uff1a<\/p>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n

\n
\n
\n
\n
\n
\n
\u8fd0\u884c\u7eaf\u6807\u91cf\u57fa\u7ebf\u6d4b\u8bd5\uff08\u5173\u95ed\u81ea\u52a8\u5411\u91cf\u5316\uff09<\/span><\/div>\n<\/div>\n

\u8fd0\u884c\u7eaf\u6807\u91cf\u57fa\u7ebf\u6d4b\u8bd5\uff08\u5173\u95ed\u81ea\u52a8\u5411\u91cf\u5316\uff09<\/span><\/p>\n

\n
<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n
\n
\n
\n
\n
\n

\u518d\u8dd1”\u5b8c\u6574\u4f18\u5316\u7248\u672c”\uff08-O3<\/code> \u8ba9\u7f16\u8bd1\u5668\u81ea\u5df1\u4e5f\u5c1d\u8bd5\u81ea\u52a8\u5411\u91cf\u5316\u6807\u91cf\u5faa\u73af\uff09\uff1a<\/p>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n

\n
\n
\n
\n
\n
\n
\u8fd0\u884cO3\u5b8c\u6574\u4f18\u5316\u6d4b\u8bd5\uff0c\u770b\u7f16\u8bd1\u5668\u81ea\u52a8\u5411\u91cf\u5316\u6548\u679c<\/span><\/div>\n<\/div>\n

\u8fd0\u884cO3\u5b8c\u6574\u4f18\u5316\u6d4b\u8bd5\uff0c\u770b\u7f16\u8bd1\u5668\u81ea\u52a8\u5411\u91cf\u5316\u6548\u679c<\/span><\/p>\n

\n
<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n
\n
\n
\n
\n
\n

\u8ba9\u6211\u53cd\u6c47\u7f16\u786e\u8ba4\u4e00\u4e0b\uff0c\u9a8c\u8bc1”\u6807\u91cf”\u7248\u672c\u5728 -O3<\/code> \u4e0b\u662f\u5426\u771f\u7684\u88ab\u7f16\u8bd1\u5668\u81ea\u52a8\u8f6c\u6210\u4e86 SIMD \u6307\u4ee4\u2014\u2014\u8fd9\u80fd\u89e3\u91ca\u4e3a\u4ec0\u4e48\u4e24\u4e2a\u7248\u672c\u5728 -O3<\/code> \u4e0b\u5dee\u8ddd\u5f88\u5c0f\uff1a<\/p>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n

\n
\n
\n
\n
\n
\n
Ran 3 commands<\/span><\/div>\n<\/div>\n

Ran 3 commands<\/span><\/p>\n

\n
<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n
\n
\n
\n

\u8bc1\u636e\u627e\u5230\u4e86\u2014\u2014\u8fd9\u6b63\u597d\u89e3\u91ca\u4e86\u5b9e\u9a8c\u7ed3\u679c\u3002<\/p>\n

\u5b9e\u6d4b\u7ed3\u679c<\/h3>\n
\n\n\n\n\n\n\n\n\n
\u7248\u672c<\/th>\n\u8017\u65f6\uff082000\u4e07\u4e2afloat\u76f8\u52a0\uff09<\/th>\nvs \u6807\u91cf\u57fa\u7ebf<\/th>\n<\/tr>\n<\/thead>\n
\u7eaf\u6807\u91cf\uff08-fno-tree-vectorize<\/code>\uff09<\/td>\n15.50 ms<\/td>\n1.0x<\/td>\n<\/tr>\n
\u624b\u52a8 AVX2 SIMD<\/td>\n10.64 ms<\/td>\n1.46x<\/strong><\/td>\n<\/tr>\n
-O3<\/code>\uff08\u7f16\u8bd1\u5668\u81ea\u52a8\u5411\u91cf\u5316\uff09<\/td>\n12.35 ms<\/td>\n1.25x<\/td>\n<\/tr>\n
-O3<\/code> \u4e0b\u7684\u624b\u52a8 AVX2<\/td>\n11.22 ms<\/td>\n1.38x<\/td>\n<\/tr>\n<\/tbody>\n<\/table>\n<\/div>\n

\u5173\u952e\u53d1\u73b0<\/h3>\n

1. \u9a8c\u8bc1\u4e86\u6587\u5b57\u8bf4\u7684\u4e8b\u60c5\u2014\u2014\u7f16\u8bd1\u5668\u786e\u5b9e”\u5077\u5077”\u505a\u4e86\u5411\u91cf\u5316<\/strong><\/p>\n

\u6211\u53cd\u6c47\u7f16\u4e86 -O3<\/code> \u4e0b”\u6807\u91cf\u7248\u672c”\u7684\u51fd\u6570\uff0c\u770b\u5230\u4e86\u8fd9\u6837\u7684\u6307\u4ee4\uff1a<\/p>\n

\n
\n
<\/div>\n<\/div>\n
asm<\/div>\n
\n
vmovups (%rdi,%rax), %ymm1      \u2190 \u52a0\u8f7d256\u4f4d\uff088\u4e2afloat\uff09\u5230 ymm \u5bc4\u5b58\u5668\r\nvaddps  (%rsi,%rax), %ymm1, %ymm0   \u2190 \u4e00\u6761\u6307\u4ee4\u5b8c\u62108\u4e2afloat\u7684\u52a0\u6cd5\r\nvmovups %ymm0, (%rdx,%rax)<\/pre>\n
 <\/p>\n<\/div>\n<\/div>\n
ymm<\/code> \u5c31\u662f AVX2 \u7684 256 \u4f4d\u5bc4\u5b58\u5668\uff0cvaddps<\/code> \u662f\u5411\u91cf\u52a0\u6cd5\u6307\u4ee4\u2014\u2014\u4f60\u5199\u7684\u662f\u666e\u901a for<\/code> \u5faa\u73af\uff0c\u7f16\u8bd1\u5668\u5728 -O3<\/code> \u4e0b\u81ea\u5df1\u628a\u5b83\u7ffb\u8bd1\u6210\u4e86\u548c\u6211\u624b\u5199 AVX2 \u51e0\u4e4e\u4e00\u6837\u7684\u6307\u4ee4<\/strong>\u3002\u8fd9\u5c31\u662f\u4e3a\u4ec0\u4e48 -O3<\/code> \u4e0b\u4e24\u8005\u5dee\u8ddd\u4ece 1.46x \u7f29\u5c0f\u5230\u53ea\u6709 1.10x\u3002<\/p>\n
2. \u52a0\u901f\u6bd4\u6ca1\u6709\u8fbe\u5230\u7406\u8bba\u7684 8 \u500d\uff0c\u539f\u56e0\u6b63\u662f\u4e0a\u4e00\u8282\u63d0\u5230\u7684\uff1a\u5185\u5b58\u5e26\u5bbd\u74f6\u9888<\/strong><\/p>\n
\u8fd9\u4e2a\u5b9e\u9a8c\u91cc\u6bcf\u4e2a\u6570\u7ec4 76MB\uff0c\u4e09\u4e2a\u6570\u7ec4\u5171 228MB\uff0c\u8fdc\u8d85 CPU \u7f13\u5b58\u5bb9\u91cf\uff08\u901a\u5e38\u51e0MB\u5230\u51e0\u5341MB\uff09\u3002\u6570\u636e\u8981\u4ece\u4e3b\u5185\u5b58\u8bfb\u53d6\uff0c\u8ba1\u7b97\u672c\u8eab\uff08\u52a0\u6cd5\uff09\u53cd\u800c\u4e0d\u662f\u74f6\u9888\u2014\u2014SIMD \u628a”\u8ba1\u7b97”\u52a0\u5feb\u4e86 8 \u500d\uff0c\u4f46”\u5185\u5b58\u642c\u8fd0”\u7684\u901f\u5ea6\u6ca1\u53d8\uff0c\u6728\u6876\u6548\u5e94\u5bfc\u81f4\u6574\u4f53\u53ea\u63d0\u5347\u4e86 1.4 \u500d\u5de6\u53f3<\/strong>\u3002<\/p>\n
3. \u6b63\u786e\u6027\u9a8c\u8bc1\u901a\u8fc7<\/strong>\u2014\u2014\u4e24\u79cd\u5b9e\u73b0\u9010\u5143\u7d20\u6bd4\u5bf9\u5b8c\u5168\u4e00\u81f4\uff0c\u8bc1\u660e\u624b\u5199 SIMD \u6ca1\u6709\u5f15\u5165\u7cbe\u5ea6\u6216\u903b\u8f91\u9519\u8bef\u3002<\/p>\n
\n
\u60f3\u9a8c\u8bc1\u4e00\u4e0b”\u5185\u5b58\u5e26\u5bbd\u74f6\u9888”\u7684\u731c\u60f3\u5417\uff1f\u6211\u53ef\u4ee5\u628a\u6570\u7ec4\u6539\u5c0f\uff08\u6bd4\u5982\u80fd\u653e\u8fdb L2 \u7f13\u5b58\u7684\u51e0\u767eKB\uff09\uff0c\u8fd9\u6837\u8ba1\u7b97\u4f1a\u91cd\u65b0\u6210\u4e3a\u74f6\u9888\uff0cSIMD \u7684\u52a0\u901f\u6bd4\u5e94\u8be5\u4f1a\u66f4\u63a5\u8fd1\u7406\u8bba\u503c\u2014\u2014\u8fd9\u4e5f\u662f\u5b9e\u8df5\u4e2d\u5224\u65ad”\u95ee\u9898\u51fa\u5728\u8ba1\u7b97\u8fd8\u662f\u5185\u5b58”\u7684\u6807\u51c6\u6392\u67e5\u65b9\u6cd5\u3002<\/p>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n<\/div>\n","protected":false},"excerpt":{"rendered":"
\u6211\u6765\u5199\u4e00\u4e2a\u5b8c\u6574\u7684\u3001\u53ef\u76f4\u63a5\u7f16\u8bd1\u8fd0\u884c\u7684\u5bf9\u6bd4\u5b9e<\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"colormag_page_container_layout":"default_layout","colormag_page_sidebar_layout":"default_layout","footnotes":""},"categories":[24],"tags":[],"class_list":["post-7442","post","type-post","status-publish","format-standard","hentry","category-value_docs"],"_links":{"self":[{"href":"https:\/\/i007.cc\/wordpress\/wp-json\/wp\/v2\/posts\/7442","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/i007.cc\/wordpress\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/i007.cc\/wordpress\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/i007.cc\/wordpress\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/i007.cc\/wordpress\/wp-json\/wp\/v2\/comments?post=7442"}],"version-history":[{"count":1,"href":"https:\/\/i007.cc\/wordpress\/wp-json\/wp\/v2\/posts\/7442\/revisions"}],"predecessor-version":[{"id":7443,"href":"https:\/\/i007.cc\/wordpress\/wp-json\/wp\/v2\/posts\/7442\/revisions\/7443"}],"wp:attachment":[{"href":"https:\/\/i007.cc\/wordpress\/wp-json\/wp\/v2\/media?parent=7442"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/i007.cc\/wordpress\/wp-json\/wp\/v2\/categories?post=7442"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/i007.cc\/wordpress\/wp-json\/wp\/v2\/tags?post=7442"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}