讓你的 PHP 7 更快 (GCC PGO)
我們一直致力于提升PHP7的性能, 上個(gè)月我們注意到GCC的PGO能在Wordpress上能帶來(lái)近10%的性能提升, 這個(gè)讓我們很激動(dòng).
然而, PGO正如名字所說(shuō)(Profile Guided Optimization 有興趣的可以Google), 他需要用一些用例來(lái)獲得反饋, 也就是說(shuō)這個(gè)優(yōu)化是需要和一個(gè)特定的場(chǎng)景綁定的.
你對(duì)一個(gè)場(chǎng)景的優(yōu)化, 也許在另外一個(gè)場(chǎng)景就事與愿違了. 它不是一個(gè)通用的優(yōu)化. 所以我們不能簡(jiǎn)單的就包含這些優(yōu)化, 也無(wú)法直接發(fā)布PGO編譯后的PHP7.
當(dāng)然, 我們正在嘗試從PGO找出一些共性的優(yōu)化, 然后手工Apply到PHP7上去, 但這個(gè)很明顯不能做到針對(duì)一個(gè)場(chǎng)景的特別優(yōu)化所能達(dá)到的效果, 所以我決定寫這篇文章簡(jiǎn)單介紹下怎么使用PGO來(lái)編譯PHP7, 讓你編譯的PHP7能特別的讓你自己的獨(dú)立的應(yīng)用變得更快.
首先, 要決定的就是拿什么場(chǎng)景去Feedback GCC, 我們一般都會(huì)選擇: 在你要優(yōu)化的場(chǎng)景中: 訪問(wèn)量最大的, 耗時(shí)最多的, 資源消耗最重的一個(gè)頁(yè)面.
拿Wordpress為例, 我們選擇Wordpress的首頁(yè)(因?yàn)槭醉?yè)往往是訪問(wèn)量最大的).
我們以我的機(jī)器為例:
- Intel(R) Xeon(R) CPU X5687 @ 3.60GHz X 16(超線程),
- 48G Memory
php-fpm 采用固定32個(gè)worker, opcache采用默認(rèn)的配置(一定要記得加載opcache)
以wordpress 4.1為優(yōu)化場(chǎng)景..
首先我們來(lái)測(cè)試下目前WP在PHP7的性能(ab -n 10000 -c 100):
- $ ab -n 10000 -c 100 http://inf-dev-maybach.weibo.com:8000/wordpress/
- This is ApacheBench, Version 2.3 <$Revision: 655654 $>
- Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
- Licensed to The Apache Software Foundation, http://www.apache.org/
- Benchmarking inf-dev-maybach.weibo.com (be patient)
- Completed 1000 requests
- Completed 2000 requests
- Completed 3000 requests
- Completed 4000 requests
- Completed 5000 requests
- Completed 6000 requests
- Completed 7000 requests
- Completed 8000 requests
- Completed 9000 requests
- Completed 10000 requests
- Finished 10000 requests
- Server Software: nginx/1.7.12
- Server Hostname: inf-dev-maybach.weibo.com
- Server Port: 8000
- Document Path: /wordpress/
- Document Length: 9048 bytes
- Concurrency Level: 100
- Time taken for tests: 8.957 seconds
- Complete requests: 10000
- Failed requests: 0
- Write errors: 0
- Total transferred: 92860000 bytes
- HTML transferred: 90480000 bytes
- Requests per second: 1116.48 [#/sec] (mean)
- Time per request: 89.567 [ms] (mean)
- Time per request: 0.896 [ms] (mean, across all concurrent requests)
- Transfer rate: 10124.65 [Kbytes/sec] received
可見(jiàn)Wordpress 4.1 目前在這個(gè)機(jī)器上, 首頁(yè)的QPS可以到1116.48. 也就是每秒鐘可以處理這么多個(gè)對(duì)首頁(yè)的請(qǐng)求,
現(xiàn)在, 讓我們開(kāi)始教GCC, 讓他編譯出跑Wordpress4.1更快的PHP7來(lái), 首先要求GCC 4.0以上的版本, 不過(guò)我建議大家使用GCC-4.8以上的版本(現(xiàn)在都GCC-5.1了).
第一步, 自然是下載PHP7的源代碼了, 然后做./configure. 這些都沒(méi)什么區(qū)別
接下來(lái)就是有區(qū)別的地方了, 我們要首先第一遍編譯PHP7, 讓它生成會(huì)產(chǎn)生profile數(shù)據(jù)的可執(zhí)行文件:
- $ make prof-gen
注意, 我們用到了prof-gen參數(shù)(這個(gè)是PHP7的Makefile特有的, 不要嘗試在其他項(xiàng)目上也這么搞哈 :) )
然后, 讓我們開(kāi)始訓(xùn)練GCC:
- $ sapi/cgi/php-cgi -T 100 /home/huixinchen/local/www/htdocs/wordpress/index.php >/dev/null
也就是讓php-cgi跑100遍wordpress的首頁(yè), 從而生成一些在這個(gè)過(guò)程中的profile信息.
然后, 我們開(kāi)始第二次編譯PHP7.
- $ make prof-clean
- $ make prof-use && make install
好的, 就這么簡(jiǎn)單, PGO編譯完成了, 現(xiàn)在我們看看PGO編譯以后的PHP7的性能:
- $ ab -n10000 -c 100 http://inf-dev-maybach.weibo.com:8000/wordpress/
- This is ApacheBench, Version 2.3 <$Revision: 655654 $>
- Copyright 1996 Adam Twiss, Zeus Technology Ltd, http://www.zeustech.net/
- Licensed to The Apache Software Foundation, http://www.apache.org/
- Benchmarking inf-dev-maybach.weibo.com (be patient)
- Completed 1000 requests
- Completed 2000 requests
- Completed 3000 requests
- Completed 4000 requests
- Completed 5000 requests
- Completed 6000 requests
- Completed 7000 requests
- Completed 8000 requests
- Completed 9000 requests
- Completed 10000 requests
- Finished 10000 requests
- Server Software: nginx/1.7.12
- Server Hostname: inf-dev-maybach.weibo.com
- Server Port: 8000
- Document Path: /wordpress/
- Document Length: 9048 bytes
- Concurrency Level: 100
- Time taken for tests: 8.391 seconds
- Complete requests: 10000
- Failed requests: 0
- Write errors: 0
- Total transferred: 92860000 bytes
- HTML transferred: 90480000 bytes
- Requests per second: 1191.78 [#/sec] (mean)
- Time per request: 83.908 [ms] (mean)
- Time per request: 0.839 [ms] (mean, across all concurrent requests)
- Transfer rate: 10807.45 [Kbytes/sec] received
現(xiàn)在每秒鐘可以處理1191.78個(gè)QPS了, 提升是~7%. 還不賴哈(咦, 你不是說(shuō)10%么? 怎么成7%了? 呵呵, 正如我之前說(shuō)過(guò), 我們嘗試分析PGO都做了些什么優(yōu)化, 然后把一些通用的優(yōu)化手工Apply到PHP7中. 所以也就是說(shuō), 那~3%的比較通用的優(yōu)化已經(jīng)包含到了PHP7里面了, 當(dāng)然這個(gè)工作還在繼續(xù)).
于是就這么簡(jiǎn)單, 大家可以用自己的產(chǎn)品的經(jīng)典場(chǎng)景來(lái)訓(xùn)練GCC, 簡(jiǎn)單幾步, 獲得提升, 何樂(lè)而不為呢
