perflab實驗報告69628

1、課程實驗報告 課 程 名 稱： 計算機組成與結構 實驗項目名稱： perflab-handout 專 業(yè) 班 級： 姓 名： 學 號： 指 導 教 師： 楊科華 完 成 時 間： 2016 年 5 月 27 日信息科學與工程學院實驗題目：perflab程序性能調優(yōu)實驗目的：理解編譯器，學習程序優(yōu)化，從優(yōu)化程序代碼和程序執(zhí)行速度兩方面著手。實驗要求：本次實驗，要求針對每個函數(shù)、每個人均至少寫出3種優(yōu)化版本、并根據driver報告的結果進行性能分析實驗環(huán)境: ubuntu-15.10、 x32系統(tǒng) 、VMware workstation實驗內容及操作步驟：   

2、60;     將下載下來的 kernels.c 中的 rotate、smooth 函數(shù)進行優(yōu)化。 rotate函數(shù)的作用是將圖像逆時針旋轉90°，smooth函數(shù)的作用是對于圖像中的每一個像素點，取它和周圍的像素點的平均值，讓圖片變得模糊。下面對代碼進行逐一優(yōu)化。源代碼的CPE測試：1.Naive_rotate 1)源代碼:char naive_rotate_descr = "naive_rotate: Naive baseline implementation"void naive_rotate(in

3、t dim, pixel *src, pixel *dst) int i, j; for (i = 0; i < dim; i+)for (j = 0; j < dim; j+) dstRIDX(dim-1-j, i, dim) = srcRIDX(i, j, dim);2）分析：這段代碼的作用就是用一個雙層循環(huán)將所有的像素進行行列調位、導致整幅圖畫進行了 90 度旋轉。然而分析一下代碼就能發(fā)現(xiàn)一個十分簡單的優(yōu)化方法：因為在最內層循環(huán)中，j的值每次都會改變，所以每執(zhí)行一次賦值就要計算一次dim-1-j，算多了自然就慢了。我們可以利用簡單的數(shù)學技巧改寫公式，將賦值語句改成dstRID

4、X(i, j, dim) = srcRIDX(j, dim-i-1, dim); 這樣就不用每次都計算了。3）優(yōu)化代碼1如下：char naive_rotate_descr2 = "naive_rotate2: only change the place of i and j"void naive_rotate2(int dim, pixel *src, pixel *dst) int i, j; for (i = 0; i < dim; i+) for (j = 0; j < dim; j+) dstRIDX(i, j, dim) = srcRIDX(j, d

5、im-i-1, dim);/i change less 優(yōu)化結果如下：這是一種最為簡單的優(yōu)化方案，由圖可知，速度提升不大，性能優(yōu)化結果也不是很好。再分析源代碼，從 cache 友好性來分析，這個代碼的效率機會很低， 所以按照 cache 的大小， 應在存儲的時候進行 32 個像素依次存儲 （列 存儲）。做到 cache 友好這樣就可以可以大幅度提高效率。4）優(yōu)化代碼2如下：char rotate_descr2 = "rotate2: version2 break into 4*4 blocks" void rotate2(int di

6、m, pixel *src, pixel*dst) int i, j,ii,jj; for(ii=0; ii < dim; ii+=4)for(jj=0; jj < dim; jj+=4)for(i=ii; i < ii+4; i+)for(j=jj; j < jj+4; j+) dstRIDX(dim-1-j,i,dim) = srcRIDX(i,j,dim); 優(yōu)化結果如下：用分塊的方式，進行優(yōu)化。將整個程序分成4*4的小塊，提高空間局部性5）優(yōu)化代碼3如下：char rotate_descr3 = "rotate3: version3 break int

7、o 32*32 blocks"void rotate3(int dim, pixel *src, pixel *dst) int i, j,ii,jj; for(ii=0; ii < dim; ii+=32)for(jj=0; jj < dim; jj+=32)for(i=ii; i < ii+32; i+)for(j=jj; j < jj+32; j+) dstRIDX(dim-1-j,i,dim) = srcRIDX(i,j,dim); 優(yōu)化結果如下：分成32*32塊，提高空間局部性6）優(yōu)化代碼4如下：char rotate_descr4 = "

8、rotate4: Current working version,using pointer rather than computing address"void rotate4(int dim, pixel *src, pixel *dst) int i; int j; int tmp1=dim*dim; int tmp2=dim *31; int tmp3=tmp1-dim; int tmp4=tmp1+32; int tmp5=dim+31; dst+=tmp3; for(i=0; i< dim; i+=32) for(j=0;j<dim;j+) *dst=*src

9、; dst+;src+=dim; *dst=*src; dst+;src+=dim; *dst=*src; dst+;src+=dim; *dst=*src; dst+;src+=dim; *dst=*src; dst+;src+=dim; *dst=*src; dst+;src+=dim; *dst=*src; dst+;src+=dim; *dst=*src; dst+;src+=dim; *dst=*src; dst+;src+=dim; *dst=*src; dst+;src+=dim; *dst=*src; dst+;src+=dim; *dst=*src; dst+;src+=di

10、m; *dst=*src; dst+;src+=dim; *dst=*src; dst+;src+=dim; *dst=*src; dst+;src+=dim; *dst=*src; dst+;src+=dim; *dst=*src; dst+;src+=dim; *dst=*src; dst+;src+=dim; *dst=*src; dst+;src+=dim; *dst=*src; dst+;src+=dim; *dst=*src; dst+;src+=dim; *dst=*src; dst+;src+=dim; *dst=*src; dst+;src+=dim; *dst=*src;

11、dst+;src+=dim; *dst=*src; dst+;src+=dim; *dst=*src; dst+;src+=dim; *dst=*src; dst+;src+=dim; *dst=*src; dst+;src+=dim; *dst=*src; dst+;src+=dim; *dst=*src; dst+;src+=dim; *dst=*src; dst+;src+=dim; *dst=*src; src+; src-=tmp2; dst-=tmp5; src+=tmp2; dst+=tmp4; 優(yōu)化結果如下：用循環(huán)展開，分成32路并行來寫2.Naive_smooth1)源代碼c

12、har naive_smooth_descr = "naive_smooth: Naive baseline implementation"void naive_smooth(int dim, pixel *src, pixel *dst) int i, j; for (i = 0; i < dim; i+)for (j = 0; j < dim; j+) dstRIDX(i, j, dim) = avg(dim, i, j, src);CPES性能如下：2）分析這段代碼很多次地調用 avg 函數(shù)，而 avg 函數(shù)內也頻繁調用 initialize_pixel_

13、sum 、 accumulate_sum、assign_sum_to_pixel 這幾個函數(shù)，且含有 2 層 for 循環(huán)。雖然會以損害程序的模塊性為代價，但消除函數(shù)調用的時間開銷，得到的代碼運行速度會快得多。所以，需要改寫代碼，不調用 avg 函數(shù)。 3）優(yōu)化代碼1如下：char smooth_descr1 = "smooth1: with less func call and grossly simplified calculation for central parts"void smooth1(int dim, pixel *src, pixel *dst) int

14、 i, j, ii, jj; pixel_sum sum; pixel current_pixel, cp; for (j = 0; j < dim; j+) dstRIDX(0, j, dim) = avg(dim, 0, j, src); dstRIDX(dim-1, j, dim) = avg(dim, dim-1, j, src); for (i = 0; i < dim; i+) dstRIDX(i, 0, dim) = avg(dim, i, 0, src); dstRIDX(i, dim-1, dim) = avg(dim, i, dim-1, src); for (

15、i = 1; i < dim-1; i+) for (j = 1; j < dim-1; j+) sum.red = sum.green = sum.blue = 0; for(ii = max(i-1, 0); ii <= min(i+1, dim-1); ii+) for(jj = max(j-1, 0); jj <= min(j+1, dim-1); jj+) cp=srcRIDX(ii,jj,dim); sum.red += cp.red; sum.green += cp.green; sum.blue += cp.blue; current_pixel.red

16、 = sum.red/9; current_pixel.green = sum.green/9; current_pixel.blue = sum.blue/9; dstRIDX(i, j, dim) = current_pixel; 優(yōu)化結果如下：4）優(yōu)化代碼2如下：char smooth_descr2 = "smooth2: test version"void smooth2(int dim, pixel *src, pixel *dst) int i,j; /no using avg() /corners dstRIDX(0,0,dim).red=(srcRIDX(0

17、,0,dim).red+srcRIDX(1,0,dim).red+srcRIDX(0,1,dim).red+srcRIDX(1,1,dim).red)>>2; dstRIDX(0,0,dim).blue=(srcRIDX(0,0,dim).blue+srcRIDX(1,0,dim).blue+srcRIDX(0,1,dim).blue+srcRIDX(1,1,dim).blue)>>2; dstRIDX(0,0,dim).green=(srcRIDX(0,0,dim).green+srcRIDX(1,0,dim).green+srcRIDX(0,1,dim).green

18、+srcRIDX(1,1,dim).green)>>2; dstRIDX(0,dim-1,dim).red=(srcRIDX(0,dim-1,dim).red+srcRIDX(1,dim-1,dim).red+srcRIDX(0,dim-2,dim).red+srcRIDX(1,dim-2,dim).red)>>2; dstRIDX(0,dim-1,dim).blue=(srcRIDX(0,dim-1,dim).blue+srcRIDX(1,dim-1,dim).blue+srcRIDX(0,dim-2,dim).blue+srcRIDX(1,dim-2,dim).bl

19、ue)>>2; dstRIDX(0,dim-1,dim).green=(srcRIDX(0,dim-1,dim).green+srcRIDX(1,dim-1,dim).green+srcRIDX(0,dim-2,dim).green+srcRIDX(1,dim-2,dim).green)>>2; dstRIDX(dim-1,0,dim).red=(srcRIDX(dim-1,0,dim).red+srcRIDX(dim-2,0,dim).red+srcRIDX(dim-1,1,dim).red+srcRIDX(dim-2,1,dim).red)>>2; ds

20、tRIDX(dim-1,0,dim).blue=(srcRIDX(dim-1,0,dim).blue+srcRIDX(dim-2,0,dim).blue+srcRIDX(dim-1,1,dim).blue+srcRIDX(dim-2,1,dim).blue)>>2; dstRIDX(dim-1,0,dim).green=(srcRIDX(dim-1,0,dim).green+srcRIDX(dim-2,0,dim).green+srcRIDX(dim-1,1,dim).green+srcRIDX(dim-2,1,dim).green)>>2; dstRIDX(dim-1

21、,dim-1,dim).red=(srcRIDX(dim-1,dim-1,dim).red+srcRIDX(dim-1,dim-2,dim).red+srcRIDX(dim-2,dim-1,dim).red+srcRIDX(dim-2,dim-2,dim).red)>>2; dstRIDX(dim-1,dim-1,dim).blue=(srcRIDX(dim-1,dim-1,dim).blue+srcRIDX(dim-1,dim-2,dim).blue+srcRIDX(dim-2,dim-1,dim).blue+srcRIDX(dim-2,dim-2,dim).blue)>&

22、gt;2; dstRIDX(dim-1,dim-1,dim).green=(srcRIDX(dim-1,dim-1,dim).green+srcRIDX(dim-1,dim-2,dim).green+srcRIDX(dim-2,dim-1,dim).green+srcRIDX(dim-2,dim-2,dim).green)>>2; /boarderfor(i=1;i<dim-1;i+) dstRIDX(i,0,dim).red=(srcRIDX(i,0,dim).red+srcRIDX(i-1,0,dim).red+srcRIDX(i-1,1,dim).red+srcRIDX

23、(i,1,dim).red+srcRIDX(i+1,0,dim).red+srcRIDX(i+1,1,dim).red)/6; dstRIDX(i,0,dim).blue=(srcRIDX(i,0,dim).blue+srcRIDX(i-1,0,dim).blue+srcRIDX(i-1,1,dim).blue+srcRIDX(i,1,dim).blue+srcRIDX(i+1,0,dim).blue+srcRIDX(i+1,1,dim).blue)/6; dstRIDX(i,0,dim).green=(srcRIDX(i,0,dim).green+srcRIDX(i-1,0,dim).gre

24、en+srcRIDX(i-1,1,dim).green+srcRIDX(i,1,dim).green+srcRIDX(i+1,0,dim).green+srcRIDX(i+1,1,dim).green)/6;for(i=1;i<dim-1;i+) dstRIDX(i,dim-1,dim).red=(srcRIDX(i,dim-1,dim).red+srcRIDX(i-1,dim-1,dim).red+srcRIDX(i-1,dim-2,dim).red+srcRIDX(i,dim-2,dim).red+srcRIDX(i+1,dim-1,dim).red+srcRIDX(i+1,dim-

25、2,dim).red)/6; dstRIDX(i,dim-1,dim).blue=(srcRIDX(i,dim-1,dim).blue+srcRIDX(i-1,dim-1,dim).blue+srcRIDX(i-1,dim-2,dim).blue+srcRIDX(i,dim-2,dim).blue+srcRIDX(i+1,dim-1,dim).blue+srcRIDX(i+1,dim-2,dim).blue)/6; dstRIDX(i,dim-1,dim).green=(srcRIDX(i,dim-1,dim).green+srcRIDX(i-1,dim-1,dim).green+srcRID

26、X(i-1,dim-2,dim).green+srcRIDX(i,dim-2,dim).green+srcRIDX(i+1,dim-1,dim).green+srcRIDX(i+1,dim-2,dim).green)/6; for(j=1;j<dim-1;j+) dstRIDX(0,j,dim).red=(srcRIDX(0,j,dim).red+srcRIDX(0,j-1,dim).red+srcRIDX(1,j-1,dim).red+srcRIDX(1,j,dim).red+srcRIDX(0,j+1,dim).red+srcRIDX(1,j+1,dim).red)/6; dstRI

27、DX(0,j,dim).blue=(srcRIDX(0,j,dim).blue+srcRIDX(0,j-1,dim).blue+srcRIDX(1,j-1,dim).blue+srcRIDX(1,j,dim).blue+srcRIDX(0,j+1,dim).blue+srcRIDX(1,j+1,dim).blue)/6; dstRIDX(0,j,dim).green=(srcRIDX(0,j,dim).green+srcRIDX(0,j-1,dim).green+srcRIDX(1,j-1,dim).green+srcRIDX(1,j,dim).green+srcRIDX(0,j+1,dim)

28、.green+srcRIDX(1,j+1,dim).green)/6; for(j=1;j<dim-1;j+) dstRIDX(dim-1,j,dim).red=(srcRIDX(dim-1,j,dim).red+srcRIDX(dim-1,j+1,dim).red+srcRIDX(dim-1,j-1,dim).red+srcRIDX(dim-2,j,dim).red+srcRIDX(dim-2,j+1,dim).red+srcRIDX(dim-2,j-1,dim).red)/6; dstRIDX(dim-1,j,dim).blue=(srcRIDX(dim-1,j,dim).blue+

29、srcRIDX(dim-1,j+1,dim).blue+srcRIDX(dim-1,j-1,dim).blue+srcRIDX(dim-2,j,dim).blue+srcRIDX(dim-2,j+1,dim).blue+srcRIDX(dim-2,j-1,dim).blue)/6; dstRIDX(dim-1,j,dim).green=(srcRIDX(dim-1,j,dim).green+srcRIDX(dim-1,j+1,dim).green+srcRIDX(dim-1,j-1,dim).green+srcRIDX(dim-2,j,dim).green+srcRIDX(dim-2,j+1,

30、dim).green+srcRIDX(dim-2,j-1,dim).green)/6; /common for(i=1;i<dim-1;i+) for(j=1;j<dim-1;j+) dstRIDX(i,j,dim).red=(srcRIDX(i,j,dim).red+srcRIDX(i+1,j,dim).red+srcRIDX(i-1,j,dim).red+srcRIDX(i,j-1,dim).red+srcRIDX(i+1,j-1,dim).red+srcRIDX(i-1,j-1,dim).red+srcRIDX(i,j+1,dim).red+srcRIDX(i+1,j+1,dim).red+srcRIDX(i-1,j+1,dim).red)/9; dstRIDX(i,j,dim).blue=(srcRIDX(i,j,dim).blue+srcRIDX(i+1,j,dim).blue+srcRIDX(i-1,j,dim).blue+srcRIDX(i,j-1,dim).blue+srcRIDX(i+1,j-1,dim).blue+srcRIDX(i-1,j-1,dim).blue+srcRIDX(i,j+