source: Sophya/trunk/SophyaLib/Manual/perfmachine.txt@ 3251

Last change on this file since 3251 was 3251, checked in by ansari, 18 years ago

Ajout mesures detailles avec cpupower, et mesures sur G5 (ccsvx01), Reza 18/05/2007

File size: 20.6 KB
RevLine 
[3158]1-------------------------------------------------------------------------------------
2 Comparaison performance de differentes machines en compilation / execution (calcul)
3 -----------------------
4 Mesures effectuees en Janvier 2007 , R. Ansari / C. Magneville
5-------------------------------------------------------------------------------------
6
7(a) eros3 : Bipro-bicoeur Xeon@2.4 GHz Linux (xeon-lx-2.4GHz) , gcc 3.2
8(b) ccali : Bipro-bicoeur Xeon@2.8 GHz Linux (xeon-lx-2.8GHz) , icc 8.0 ou 9.0
[3244]9 Flags de compilation avec [-O -g]
[3179]10(c) sgsda: AMD Bipro AMD opteron 248 @ 2.2 GHz (amd-lx-
[3244]11 Flags de compilation avec [-O -g]
[3241]12(cc) grid-saclay: AMD opteron 275 Bipro-bicoeur @ 2.2 GHz (amd275-lx)
[3244]13 Flags de compilation avec [-O -g]
[3158]14
15(d) asc: bipro alpha (@ ~1 GHz) server DS20 OSF (osf1) , cxx 6.5 (osf-asc)
16(nouveau asc 420 MFLOPS, moins puissante que l'ancien asc 800 MFLOPS)
17(e) xp1000-dapnia: alpha xp1000 @ ~ 600 MHz ? OSF1 , cxx ? (osf-xp1000)
[3241]18(e') cool: alpha xp1000 @ ~ 667 MHz ? OSF1 5.1 , cxx 6.3
[3158]19(f) superosf-dapnia: multi-proc alphaServer ES80 6 procs EV7 @ 1 GHz (super-osf)
20
21(g) ccsvx01: XServe G5 bipro @~1GHz (Darwin/OSX) (G5-osx-1GHz) , gcc 3.3
22(h) PowerBook-Reza : Apple G4 @ 1.25 GHz (G4-osx-1.25GHz) , gcc 3.3
23(i) MacBook-Reza: Apple/ Core double-coeur Intel @ 1.83 GHz (core-osx-1.83GHz) gcc 4
24(j) MacPro-Grosdidier : Apple / Xeon 2 double-coeur @ 3 GHz gcc 4.0.1 , compil SOPHYA -O2 -g
25
[3187]26(p) IBM-AIX regatta , xlC , IBM eServer pSeries 655 , 8 proc power4 @ 1.1 GHz
[3241]27(q) IBM-AIX meso , AIX 5.3, xlC V8 , IBM Power5 , 8 proc bi-coeur P575 @ 1.9 GHz
[3158]28
[3241]29(s) SGI-IRIX64 magique, CC
[3158]30
31NOTES :
32- Sur les machines Xeon, il y a une interaction entre process / threads par rapport a
33l'occupation des CPU's. On perd un facteur 3 en performance multi-threads/multi-taches.
34La machine MacPro avec OSX se debrouille quand meme mieux.
35- Effet du systeme ou carte mere ???
36
[3244]37Flag de compilation
38- Flag de compilation par defaut [-O -g] en general
39- Sur eros3 (xeon-linux gcc 3.3) [-O -g] OU [-O3 -g]
40- Sur Darwin [-g] ou [-O2 -g] (ou [-tune G5] sur XServe G5)
41 Sur les mac (en particulier G4/G5), grande difference entre -g et -Ox -g
42 mais peu de difference entre -O -O2 -O3
43- Sur machine aix-meso [-O -g] ou [-O3 -g]
[3187]44
[3251]45X/ Performances brutes cpupower et donnees SPEC ((http//www.spec.org)
46----------------------------------------------------------------------
[3244]47
[3251]48(1) MFLOPS -> cpupower 2 (x/y : -O -g / -O3)
49SPECint2000 (3) / SPECfp2000 (2) (http//www.spec.org)
50
51csh> cpupower 0 3000000 5
52 3 10^6 operations doubles - sur memoire 3x3 10^6 doubles (~50 MO)
53 ===> ~ 24 MO / MFLOPS
54csh> cpupower 2
55 1.6 10^9 operations doubles - sur 3x20000 doubles (~0.5 MO)
56
57
58Compilation avec -O (optimisation)
59 (1) cpupower 0 : debit memoire en MO/s
60 (2) cpupower 0 , MFLOPS
61 (5) cpupower 2 , MFLOPS
62
63Compilation avec -g (debug / sans optimisation)
64 (3) cpupower 0 , MFLOPS
65 (6) cpupower 2 , MFLOPS
66
67Compilation avec -O3 ou -fast ...( optimisation poussee)
68 (4) cpupower 0 , MFLOPS
69 (7) cpupower 2 , MFLOPS
70
71
72(3) cc -O / cc -g1 -O
73(4) cc -O3
74
75csh> cpupower 2
76
77----------------------------------------------------------------------------------------------
78 MFLOPS |(1) MO/s| (2) (3) (4) | (5) (6) (7)
79----------------------------------------------------------------------------------------------
80(a)xeon-lx-2.4GHz | 1290 | 53 53 55 | 338 340 320
81(a)xeon-lx-2.8GHzicc | 1560 | 65 62 60 | 914 409 914
82(c)amd-lx | 2280 | 95 77 97 | 666 314 686
83(cc)amd2-lx | |
84
85(e')osf-cool | 768 | 32 15 32 | 630 150 660
86(f)superosf
87
88(g)G5-osx-1 GHz | 2100 | 88 68 88 | 1000 255 1073
89(f)G4-osx-1.25GHz | 600 | 25 16 25 | 417 93 430
90(i)core-osx-1.83GHz | |
91(j)xeon-osx
92
93(p)ibm-aix-regatta
94(p)ibm-aix-meso | 3600 | 150 75 150 | 1480 203 3600
95----------------------------------------------------------------------------------------------
96
97csh cpupower 2
[3187]98----------------------------------------------------------------------
99 MFLOPS(1) SPECfp SPECint
100----------------------------------------------------------------------
101(b)xeon-lx-2.8GHz 900 1400 1400
102(c)amd-lx 690 1600 1300
103(cc)amd2-lx 675 1600 1300
104
[3251]105(e)osf-xp1000 648 500 400
[3241]106(f)superosf 842 1100 700
[3187]107
[3244]108(f)G4-osx-1.25GHz 92/380 (-g)/(-O2 -g)
109(i)core-osx-1.83GHz 310/880 1400 1500 (-g)/(-O2 -g)
[3187]110(j)xeon-osx 2600 2900 -
111
[3244]112(p)ibm-aix-regatta 730/1750 1050 700 (-O -g)/(-O3)
113(p)ibm-aix-meso 1250/3600 (-O -g)/(-O3)
[3187]114----------------------------------------------------------------------
115
116
[3158]117A/ Compilation tout SOPHYA
118----------------------------
119csh> time make all (1)
120ou
121csh> time make -j 2 all (2)
122 Temps CPU
123 Indice de performance 100*(1000/TCPU)
124 Temps elapsed (vrai)
125 Temps vrai / TCPU
126
127
128----------------------------------------------------------------------
129 CPU(s) IndPerf TElapsed , TCPU/Elapsed %
130----------------------------------------------------------------------
131(a)xeon-lx-2.4GHz (2) 615 s 162 410 s 150%
132 avec -O3 -g (2) 1300 s 77 760 s 172%
133(b)xeon-lx-2.8GHz (2) 755 s 132 540 s 140%
[3241]134(c)amd-lx (2) 336 s 297 175 s 192%
[3158]135
[3241]136(d)osf-asc (1) 1920 s 52 2340 s 83% (??)
137(e)osf-xp1000 (1) 533 s 187 660 s 80%
138(f)superosf (1) 895 s 112 910 s 98%
[3158]139
[3251]140(g)G5-osx-1GHz (2) 453 s 221 250 s 182%
141 -tune=G5 1100 s 90
142 -g -O 740 s 380 s 195%
143(h)G4-osx-1.25GHz (1) 660 s 151 710 s 93% [-g]
[3244]144 1500 s 94% [-O2 -g]
[3158]145(i)core-osx-1.83GHz (2) 209 s 478 116 s 180%
[3159]146 -O2 (1) 367 s 272 381 96%
[3158]147(j)xeon-osx
148
149(p)ibm-aix
150----------------------------------------------------------------------
151
152Taille shared libs :
[3251]153(a)
154(c) 33 MO
[3158]155(f) = (e) = 57 MO
[3251]156(g) 80 MO
[3158]157(i) 83 MO
158
[3187]159B/ Calcul brut (Tableaux de SOPHYA) avec / sans threads
160--------------------------------------------------------
[3158]161
[3244]162B.1/ arr = c1*a1+c2*a2 ( ~ 3 10^6 op. double)
[3241]163(1) time cpupower 2 # compile avec -O3 (/ -O -g)
[3158]164(2) time zthr arr 1 1000 1 thread
165(3) time zthr arr 2 1000 2 thread
166(4) time zthr arr 4 1000 4 thread
167(5) time zthr arr 6 1000 6 thread
168(6) time zthr arr 8 1000 8 thread
169
170-----------------------------------------------------------------------------------
171 (1)MFLOPS (2)CPU/Elap/% (2)IndPerf (3)CPU/Elap/% (3)IndPerf
172-----------------------------------------------------------------------------------
173(a)xeon-lx-2.4GHz 1167 5.15/5.2/99% 11.4/5.8/196%
174 (4)36.6/9.28/394%
175 -O3 -g 4.9/5./99%
176(b)xeon-lx-2.8GHz 920 2.3/2.3/100% 6.2/3.1/198%
177 (4)26/6.6/396%
178(c)amd-lx 690 3.6/3.6/99% 6.8/4/171%
179 (4)13.5/7/193%
180 (5)20.3/10.23/198%
[3187]181(cc)amd2-lx 675 2/2/99% 4.15/2.1/197%
182 (4)8.25/4.15/198%
183 (5)13.6/4.6/292%
184 (6)19.8/6.5/300%
[3158]185
[3241]186(d)osf-asc 420 6.3s/6.5s/99% 16.9/8.8/192%
187 (4)29.9/15.7/191%
188(e)osf-xp1000 648 5.1/5.3/96.6% 11.4/11.4/99%
[3158]189 (4)25.2/25.5/99%
[3241]190(f)superosf 842 2.87/2.88/99.6% 6.25/4.1/153%
[3158]191 (4)11.6/3.06/379%
192
[3251]193(h)G4-osx-1.25GHz 92 44s/48s/91% 86.7/99.8/92% [-g]
[3244]194 380 12.2/12.9/95% 24/25.3/95% [-O2 -g]
[3251]195(g)G5-osx-1GHz 1151 20s/20s/99% 40s/23s/170%
[3158]196 (4) 80.8/45/180%
[3251]197 -O -g 4.5/4.9/91% 9.3/4.7/197%
198 (4) 18.3/9.4/197%
[3159]199 -tune=G5 3.35/3.8/88% 7.1/3.6/196%
[3251]200(h)G4-osx-1.25GHz 92 44s/48s/91% 86.7/99.8/92% [-g]
201 380 12.2/12.9/95% 24/25.3/95% [-O2 -g]
[3159]202 (4) 14/7.5/187%
[3244]203(i)core-osx-1.83GHz 855 11.5/11.5/100% 23/11.6/192% [-g]
[3158]204 (4) 46/23/199%
[3244]205 -O2 3.85/3.89/99% 7.7/3.9/198% [-O2 -g]
[3159]206 (4) 15.4/7.77/198%
207
[3158]208(j)xeon-osx 2600 2.5/2.5/100% 5.1/2.6/199%
209 (4) 11.5/3.2/362%
210 (5) 17.4/4.77/365%
211
[3241]212(p)ibm-aix-regatta 1750/730 6.8/6.9/98% 13.1/6.75/195%
[3158]213 (4) 26.3/11.7/225%
[3241]214(q)ibm-aix-meso 3600/1250 3.6/3.75/96% 7.35/3.7/197%
215 (4) 12.46/4.2/298%
216 (5) 219/6.7/280%
217 (6) 24/4.5/530%
[3158]218
[3241]219
220(s)sgi-magique 460 60/60/99%
[3158]221 -----------------------------------------------------------------------------------
222
223
224B.2/ Multiplication de matrices mtx = mtx1 * mtx2
[3244]225 ~ 2 10^9 op. double / thread
[3241]226(1) time cpupower 2 (-O3 / -O -g)
[3158]227(2) time zthr mtx 1 1000 1 thread
228(3) time zthr mtx 2 1000 2 thread
229(4) time zthr mtx 4 1000 4 thread
230(5) time zthr mtx 6 1000 6 thread
231(6) time zthr mtx 8 1000 8 thread
232
233-----------------------------------------------------------------------------------
234 (1)MFLOPS (2)CPU/Elap/% (2)IndPerf (3)CPU/Elap/% (3)IndPerf
235-----------------------------------------------------------------------------------
236(a)xeon-lx-2.4GHz 1167 6.5/6.5/100% 17.4/8.8/198%
237 (4) 80.5/20.3/397%
238 (5) 114.5/29.6/387%
239 (6) 160/40.3/387%
240
241(b)xeon-lx-2.8GHz 920 3.4/3.4/100% 12/6.1/199%
242 (4) 55.8/14/400%
243(c)amd-lx 690 6.98/6.98/100% 14.1/8.15/173%
244 (4) 27.7/14.23/194%
245 (5) 41.4/21.07/196%
246 (6) 55.4/27.9/198.7%
[3187]247(cc)amd2-lx 675 4.1/4.1/100% 9.55/4.8/198%
248 (4) 20/10.27/195%
249 (5) 32.8/11.16/294%
250 (6) 42.75/13.8/309%
[3158]251
[3187]252
[3241]253(d)osf-asc 420 13.5s/13.7s/98% 32/16.5/194%
254 (4) 67.5/34.4/196%
255(e)osf-xp1000 648 13/14.1/92% 27.1/27.4/99%
[3158]256 (4) 54/54.7/99.6%
257 (5) 80.6/81/99.6%
258 (6) 107.8/108.3/99.5%
[3241]259(f)superosf 842 6.1/7.24/84% 12.35/6.29/196%
[3158]260 (4) 24.3/6.31/385%
261 (5) 36.5/10.9/335%
262 (6) 50.1/18.15/276%
263
[3251]264(g)G5-osx-1GHz 1151 23/23.7/97% 46.5/27.5/170%
[3158]265 (4) 93.4/49.4/189%
[3251]266 -O -g 6.2/6.2/100% 14.2/7.2/197%
267 (4) 28.3/14.36/197%
[3159]268 -tune=G5 5.7/5.8/98% 13.3/6.8/197%
269 (4) 26.8/13.56/197%
270 (6) 53.8/27.25/197%
[3251]271(h)G4-osx-1.25GHz 333 23.5/24.5/96% [-O2]
[3158]272(i)core-osx-1.83GHz 855 12.6/12.7/100% 25.8/13.4/194%
273 (4) 51.6/26/199%
[3159]274 -O2 4.25/4.5/94% 10.6/5.36/198%
275 (4) 20.87/10.68/198%
276 -O2 2 jobs // 2 x 5/5.4/92%
[3158]277(j)xeon-osx 2600 2.8/2.8/99% 9.3/4.66/199%
278 (4) 31.4/8.6/364%
279 (5) 47.1/12.96/364%
280 (6) 62.8/17.38/362%
281
[3241]282(p)ibm-aix-regatta 1750/730 9.5/9.7/98% 18.3/16.0/114%
[3158]283 (4) 38.3/24.7/155%
[3241]284(p)ibm-aix-meso 3600/1250 2.3/2.3/99% 5.1/2.64/194% (compil avec -O3)
285 (4) 11.4/4.16/272%
286 (5) 20.2/5.85/344%
287 (6) 29.9/6.74/442%
[3179]288
[3241]289(s)sgi-magique 460 49/49/99% 101/56/181%
[3179]290
[3158]291 -----------------------------------------------------------------------------------
292
[3187]293C/ Calcul fft (FFTW , FFTPack )
294-------------------------------
[3158]295
296(1) time cpupower 2
297(2) time tfft 2000000 W d 0 0 (avec FFTW)
298(3) time tfft 2000000 P d 0 0 (avec FFTPack_Sophya)
299
300IndPerf=1000/TCPU
301
302
303-----------------------------------------------------------------------------------
304 (1)MFLOPS (2)CPU/Elap/% (2)IndPerf (3)CPU/Elap/% (3)IndPerf
305-----------------------------------------------------------------------------------
306(a)xeon-lx-2.4GHz 1167 5.5/5.6/97% 180 7.4/7.4/100% 135
307 -O3 -g 6.8/7.1/96% 147
308(b)xeon-lx-2.8GHz 920
[3241]309(c)amd-lx 690 3.2/3.75/86% 4.2/4.2/100% 238
[3159]310 ~2x 4.7/4.7/99%
[3187]311(cc)amd2-lx 675 2.8/2.8/99% 3.56/3.58/99%
[3158]312
[3241]313(d)osf-asc 420 13.3/13.9/95.7% 75 12.2/17.6/70% 82
314(e)osf-xp1000 648 9.9/10.2/97% 101 9.3/9.46/98.5% 107
315(e')cool 9.1/9.3/98%
316(f)superosf 842 6./6.22/96.7% 166 5.1/5.18/97.4% 196
[3158]317
[3251]318(g)G5-osx-1GHz 1151 8.5/8.7/97.5% 120 11.5/11.6/99% 87
319 -O -g 5.1/5.2/98% 190 5.4/5.5/99% 180
320 -tune=G5 5/5.1/98% 200 5.5/5.6/97% 180
321(h)G4-osx-1.25GHz 92 15.2/15.9/96% 66 23.8/34.1/70% 42 [-g]
[3244]322 380 8.8/10.3/86% 14.7/20.9/65% [-O2 -g]
[3158]323(i)core-osx-1.83GHz 855 4.6/4.75/97% 217 7/7.08/99% 142
[3159]324 -O2 3.2/3.2/99% 312 3.6/3.6/99% 277
325 -O2 2 jobs // 2 x 3.9/4.3/90% 2x 4.7/5/91% 200
[3158]326(j)xeon-osx 2600 2.6/2.6/98% 384
[3159]327 2 jobs // ~2x 3.6/4.1/87% 250
328 4 jobs //
[3158]329
[3241]330(p)ibm-aix-regatta 1750/730 6.25/18.9/33% 160 5.25/15.7/33% 190
331(q)ibm-aix-meso 3600/1250 3.95/4.3/91% 250 3.82/4./94% 260
332 2 jobs // ~2x 3.88/4.2/92% 250
[3158]333
[3241]334
335(s)sgi-magique 460 22/22/98% 42 24.5/25/99% 40
[3158]336 -----------------------------------------------------------------------------------
[3187]337
338
[3241]339D/ Calcul inversion par lapack
340-------------------------------
[3187]341
[3241]342lpk inverse 1000,1000 0
343---> temps de calcul inversion par lapack
344-------------------------------------------------------------------------------------------
345 CPU/Elap/% (1)
346-------------------------------------------------------------------------------------------
347(a)xeon-lx-2.4GHz 5.6/~100%
348(b)xeon-lx-2.8GHz
349(c)amd-lx 5.5/5.5/99%
[3187]350
[3241]351(d)osf-asc
352(e')cool 2.8/2.9/95%
353(f)superosf
354
[3244]355(f)G4-osx-1.25GHz 2.3/~100% [-O2 -g]
[3241]356(h)G5-osx-1GHz 0.8/~100%
[3251]357 -O -g 0.86/~100%
[3245]358(i)core-osx-1.83GHz 1.93/~100%
[3241]359 -O2
360(p)ibm-aix-regatta
361(q)ibm-aix-meso 0.55/~100%
362
363--------------------------------------------------------------------------------------------
364
365
366K/ Efficacite de gestion de lock (mutex) avec les threads et tableaux
[3187]367----------------------------------------------------------------------
[3241]368(32 threads - operant sur 2000 vecteurs ~ 64000 lock/unlock/wait/broadcast)
369
370(1) time zthr syncp 32 2000 4
371(2) time zthr sync 32 2000 4
372(1) time zthr syncp 4 15000 130
373(2) time zthr sync 4 15000 130
374-------------------------------------------------------------------------------------------
375 CPU/Elap/% (1) (2) (3) (4)
376-------------------------------------------------------------------------------------------
377(a)xeon-lx-2.4GHz 23.5/14/168% 4.3/1.2/365% 7.9/5.5/142% 4/2.15/190%
378 Avant ThSafeOp 17/178%
[3187]379 avec -O3 -g
380(b)xeon-lx-2.8GHz (2)
[3241]381(c)amd-lx 0.6/1/63% 0.6/1/60% 3.5/3.5/102% 2.6/2.7/98%
[3187]382
[3241]383(d)osf-asc 4.5/3.4/132% 3.35/2/170% 15.8/10.5/150% 13/8/163%
384 5.4/100%(NoThSafe)
385(e')cool 1.3/1.37/95% 1.35/1.5/89% 5.3/5.3/99% 5.2/5.2/99%
[3187]386(e)superosf (1)
[3251]387
388
389(g)G5-osx-1GHz (2) 2.6/130% (NoThSafe)
390 -O -g 2.6/1.7/150% 6.8/3.7/187% 4/2.75/142% 4.7/3/155%
391(h)G4-osx-1.25GHz (1) 40.5/42.6/95% 42.2/43.5/97% [-g]
[3244]392 3.9/4.3/89% 3.5/4/89% 3.8/4/95% 4.3/4.6/93%
[3187]393
[3245]394(i)core-osx-1.83GHz 7.7/7.1/108% 7.8/6.7/116% 30.2/29.6/102% 30.3/29.2/104% [-g]
395 -O2 2.7/1.8/152% 6/3.16/190% 3.4/2.4/142% 3.2/2.5/164% [-O2 -g]
[3241]396(j)xeon-osx
397 Avant ThSafeOp 2.55/143%
[3187]398
[3241]399(p)ibm-aix-regatta 4.7/111%
400(q)ibm-aix-meso 7.5/2.8/300% 17/3.8/450% 8.2/3.05/270% 4.85/2.43/200%
401
402--------------------------------------------------------------------------------------------
403
404
405
406L/ I/O et PPF
407-----------------
408Ecriture/lecture de n=10^7 lignes de int+6double, Total ~ 500 MO
409(1) time tstdtable w xx.ppf swap 10000000 1024 0
410(2) time tstdtable r xx.ppf swap 10000000 1024 0
411(3) time tstdtable w xx.ppf swap 50000000 1024 0
412(4) time tstdtable r xx.ppf swap 50000000 1024 0
413
414-------------------------------------------------------------------------------------------
415 CPU/Elap/% (1) (2) (3) (4)
416-------------------------------------------------------------------------------------------
417(a)xeon-lx-2.4GHz 17/26/63% 5.5/5.6/94%
418(b)xeon-lx-2.8GHz
419(c)amd-lx 5.9/6./97% 3.4/3.4/100% 30/32/93% 24/165/13% ?
420
421(d)osf-asc
422(e')cool 15/30/50% 13/13/99%
423(f)superosf
424
[3251]425(g)G5-osx-1GHz 14/14.2/98% 6/6.14/99%
426
427(h)G4-osx-1.25GHz 26/29.7/87% 15.7/38.6/41% [-O2 -g]
[3245]428(i)core-osx-1.83GHz 37/37.8/98% 22/48.4/45% [-g]
429 -O2 10.5/17.4/55% 11.2/41.2/27% [-O2 -g]
[3241]430(p)ibm-aix-regatta
431(q)ibm-aix-meso 5.5/16.8/38% 5.7/13.2/43% 32.7/85/39% 29/60/49%
432
433--------------------------------------------------------------------------------------------
Note: See TracBrowser for help on using the repository browser.