Skip to content

Commit f329f77

Browse files
committed
Merge pull request #425 from wernsaar/develop
added benchmark for lapack ?geev routines
2 parents 27af6e3 + 7c611a2 commit f329f77

4 files changed

Lines changed: 359 additions & 1 deletion

File tree

benchmark/Makefile

Lines changed: 72 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -40,6 +40,7 @@ goto :: slinpack.goto dlinpack.goto clinpack.goto zlinpack.goto \
4040
cherk.goto zherk.goto \
4141
cher2k.goto zher2k.goto \
4242
sgemv.goto dgemv.goto cgemv.goto zgemv.goto \
43+
sgeev.goto dgeev.goto cgeev.goto zgeev.goto \
4344
ssymm.goto dsymm.goto csymm.goto zsymm.goto
4445

4546
acml :: slinpack.acml dlinpack.acml clinpack.acml zlinpack.acml \
@@ -55,6 +56,7 @@ acml :: slinpack.acml dlinpack.acml clinpack.acml zlinpack.acml \
5556
cherk.acml zherk.acml \
5657
cher2k.acml zher2k.acml \
5758
sgemv.acml dgemv.acml cgemv.acml zgemv.acml \
59+
sgeev.acml dgeev.acml cgeev.acml zgeev.acml \
5860
ssymm.acml dsymm.acml csymm.acml zsymm.acml
5961

6062
atlas :: slinpack.atlas dlinpack.atlas clinpack.atlas zlinpack.atlas \
@@ -71,6 +73,7 @@ atlas :: slinpack.atlas dlinpack.atlas clinpack.atlas zlinpack.atlas \
7173
cherk.atlas zherk.atlas \
7274
cher2k.atlas zher2k.atlas \
7375
sgemv.atlas dgemv.atlas cgemv.atlas zgemv.atlas \
76+
sgeev.atlas dgeev.atlas cgeev.atlas zgeev.atlas \
7477
ssymm.atlas dsymm.atlas csymm.atlas zsymm.atlas
7578

7679
mkl :: slinpack.mkl dlinpack.mkl clinpack.mkl zlinpack.mkl \
@@ -86,6 +89,7 @@ mkl :: slinpack.mkl dlinpack.mkl clinpack.mkl zlinpack.mkl \
8689
cherk.mkl zherk.mkl \
8790
cher2k.mkl zher2k.mkl \
8891
sgemv.mkl dgemv.mkl cgemv.mkl zgemv.mkl \
92+
sgeev.mkl dgeev.mkl cgeev.mkl zgeev.mkl \
8993
ssymm.mkl dsymm.mkl csymm.mkl zsymm.mkl
9094

9195
all :: goto atlas acml mkl
@@ -720,6 +724,61 @@ dsymv.atlas : dsymv.$(SUFFIX)
720724
dsymv.mkl : dsymv.$(SUFFIX)
721725
-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
722726

727+
##################################### Sgeev ####################################################
728+
sgeev.goto : sgeev.$(SUFFIX) ../$(LIBNAME)
729+
$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
730+
731+
sgeev.acml : sgeev.$(SUFFIX)
732+
-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
733+
734+
sgeev.atlas : sgeev.$(SUFFIX)
735+
-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBATLAS) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
736+
737+
sgeev.mkl : sgeev.$(SUFFIX)
738+
-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
739+
740+
##################################### Dgeev ####################################################
741+
dgeev.goto : dgeev.$(SUFFIX) ../$(LIBNAME)
742+
$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
743+
744+
dgeev.acml : dgeev.$(SUFFIX)
745+
-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
746+
747+
dgeev.atlas : dgeev.$(SUFFIX)
748+
-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBATLAS) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
749+
750+
dgeev.mkl : dgeev.$(SUFFIX)
751+
-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
752+
753+
##################################### Cgeev ####################################################
754+
755+
cgeev.goto : cgeev.$(SUFFIX) ../$(LIBNAME)
756+
$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
757+
758+
cgeev.acml : cgeev.$(SUFFIX)
759+
-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
760+
761+
cgeev.atlas : cgeev.$(SUFFIX)
762+
-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBATLAS) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
763+
764+
cgeev.mkl : cgeev.$(SUFFIX)
765+
-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
766+
767+
##################################### Zgeev ####################################################
768+
769+
zgeev.goto : zgeev.$(SUFFIX) ../$(LIBNAME)
770+
$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
771+
772+
zgeev.acml : zgeev.$(SUFFIX)
773+
-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
774+
775+
zgeev.atlas : zgeev.$(SUFFIX)
776+
-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBATLAS) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
777+
778+
zgeev.mkl : zgeev.$(SUFFIX)
779+
-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)
780+
781+
723782

724783
###################################################################################################
725784

@@ -861,6 +920,19 @@ ssymv.$(SUFFIX) : symv.c
861920
dsymv.$(SUFFIX) : symv.c
862921
$(CC) $(CFLAGS) -c -UCOMPLEX -DDOUBLE -o $(@F) $^
863922

923+
sgeev.$(SUFFIX) : geev.c
924+
$(CC) $(CFLAGS) -c -UCOMPLEX -UDOUBLE -o $(@F) $^
925+
926+
dgeev.$(SUFFIX) : geev.c
927+
$(CC) $(CFLAGS) -c -UCOMPLEX -DDOUBLE -o $(@F) $^
928+
929+
cgeev.$(SUFFIX) : geev.c
930+
$(CC) $(CFLAGS) -c -DCOMPLEX -UDOUBLE -o $(@F) $^
931+
932+
zgeev.$(SUFFIX) : geev.c
933+
$(CC) $(CFLAGS) -c -DCOMPLEX -DDOUBLE -o $(@F) $^
934+
935+
864936

865937
clean ::
866938
@rm -f *.goto *.mkl *.acml *.atlas

benchmark/cholesky.c

Lines changed: 4 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -119,7 +119,11 @@ static __inline double getmflops(int ratio, int m, double secs){
119119

120120
int MAIN__(int argc, char *argv[]){
121121

122+
#ifndef COMPLEX
122123
char *trans[] = {"T", "N"};
124+
#else
125+
char *trans[] = {"C", "N"};
126+
#endif
123127
char *uplo[] = {"U", "L"};
124128
FLOAT alpha[] = {1.0, 0.0};
125129
FLOAT beta [] = {0.0, 0.0};

benchmark/geev.c

Lines changed: 260 additions & 0 deletions
Original file line numberDiff line numberDiff line change
@@ -0,0 +1,260 @@
1+
/*********************************************************************/
2+
/* Copyright 2009, 2010 The University of Texas at Austin. */
3+
/* All rights reserved. */
4+
/* */
5+
/* Redistribution and use in source and binary forms, with or */
6+
/* without modification, are permitted provided that the following */
7+
/* conditions are met: */
8+
/* */
9+
/* 1. Redistributions of source code must retain the above */
10+
/* copyright notice, this list of conditions and the following */
11+
/* disclaimer. */
12+
/* */
13+
/* 2. Redistributions in binary form must reproduce the above */
14+
/* copyright notice, this list of conditions and the following */
15+
/* disclaimer in the documentation and/or other materials */
16+
/* provided with the distribution. */
17+
/* */
18+
/* THIS SOFTWARE IS PROVIDED BY THE UNIVERSITY OF TEXAS AT */
19+
/* AUSTIN ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, */
20+
/* INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF */
21+
/* MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE */
22+
/* DISCLAIMED. IN NO EVENT SHALL THE UNIVERSITY OF TEXAS AT */
23+
/* AUSTIN OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, */
24+
/* INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES */
25+
/* (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE */
26+
/* GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR */
27+
/* BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF */
28+
/* LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT */
29+
/* (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT */
30+
/* OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE */
31+
/* POSSIBILITY OF SUCH DAMAGE. */
32+
/* */
33+
/* The views and conclusions contained in the software and */
34+
/* documentation are those of the authors and should not be */
35+
/* interpreted as representing official policies, either expressed */
36+
/* or implied, of The University of Texas at Austin. */
37+
/*********************************************************************/
38+
39+
#include <stdio.h>
40+
#include <stdlib.h>
41+
#ifdef __CYGWIN32__
42+
#include <sys/time.h>
43+
#endif
44+
#include "common.h"
45+
46+
47+
#undef GEEV
48+
49+
#ifndef COMPLEX
50+
#ifdef XDOUBLE
51+
#define GEEV BLASFUNC(qgeev)
52+
#elif defined(DOUBLE)
53+
#define GEEV BLASFUNC(dgeev)
54+
#else
55+
#define GEEV BLASFUNC(sgeev)
56+
#endif
57+
#else
58+
#ifdef XDOUBLE
59+
#define GEEV BLASFUNC(xgeev)
60+
#elif defined(DOUBLE)
61+
#define GEEV BLASFUNC(zgeev)
62+
#else
63+
#define GEEV BLASFUNC(cgeev)
64+
#endif
65+
#endif
66+
67+
#ifndef COMPLEX
68+
extern void GEEV( char* jobvl, char* jobvr, blasint* n, FLOAT* a,
69+
blasint* lda, FLOAT* wr, FLOAT* wi, FLOAT* vl, blasint* ldvl,
70+
FLOAT* vr, blasint* ldvr, FLOAT* work, blasint* lwork, blasint* info );
71+
#else
72+
extern void GEEV( char* jobvl, char* jobvr, blasint* n, FLOAT* a,
73+
blasint* lda, FLOAT* wr, FLOAT* vl, blasint* ldvl,
74+
FLOAT* vr, blasint* ldvr, FLOAT* work, blasint* lwork, FLOAT *rwork, blasint* info );
75+
#endif
76+
77+
#if defined(__WIN32__) || defined(__WIN64__)
78+
79+
#ifndef DELTA_EPOCH_IN_MICROSECS
80+
#define DELTA_EPOCH_IN_MICROSECS 11644473600000000ULL
81+
#endif
82+
83+
int gettimeofday(struct timeval *tv, void *tz){
84+
85+
FILETIME ft;
86+
unsigned __int64 tmpres = 0;
87+
static int tzflag;
88+
89+
if (NULL != tv)
90+
{
91+
GetSystemTimeAsFileTime(&ft);
92+
93+
tmpres |= ft.dwHighDateTime;
94+
tmpres <<= 32;
95+
tmpres |= ft.dwLowDateTime;
96+
97+
/*converting file time to unix epoch*/
98+
tmpres /= 10; /*convert into microseconds*/
99+
tmpres -= DELTA_EPOCH_IN_MICROSECS;
100+
tv->tv_sec = (long)(tmpres / 1000000UL);
101+
tv->tv_usec = (long)(tmpres % 1000000UL);
102+
}
103+
104+
return 0;
105+
}
106+
107+
#endif
108+
109+
#if !defined(__WIN32__) && !defined(__WIN64__) && !defined(__CYGWIN32__) && 0
110+
111+
static void *huge_malloc(BLASLONG size){
112+
int shmid;
113+
void *address;
114+
115+
#ifndef SHM_HUGETLB
116+
#define SHM_HUGETLB 04000
117+
#endif
118+
119+
if ((shmid =shmget(IPC_PRIVATE,
120+
(size + HUGE_PAGESIZE) & ~(HUGE_PAGESIZE - 1),
121+
SHM_HUGETLB | IPC_CREAT |0600)) < 0) {
122+
printf( "Memory allocation failed(shmget).\n");
123+
exit(1);
124+
}
125+
126+
address = shmat(shmid, NULL, SHM_RND);
127+
128+
if ((BLASLONG)address == -1){
129+
printf( "Memory allocation failed(shmat).\n");
130+
exit(1);
131+
}
132+
133+
shmctl(shmid, IPC_RMID, 0);
134+
135+
return address;
136+
}
137+
138+
#define malloc huge_malloc
139+
140+
#endif
141+
142+
int MAIN__(int argc, char *argv[]){
143+
144+
FLOAT *a,*vl,*vr,*wi,*wr,*work,*rwork;
145+
FLOAT wkopt[4];
146+
char job='V';
147+
char *p;
148+
149+
blasint m, i, j, info,lwork;
150+
double factor = 26.33;
151+
152+
int from = 1;
153+
int to = 200;
154+
int step = 1;
155+
156+
struct timeval start, stop;
157+
double time1;
158+
159+
argc--;argv++;
160+
161+
if (argc > 0) { from = atol(*argv); argc--; argv++;}
162+
if (argc > 0) { to = MAX(atol(*argv), from); argc--; argv++;}
163+
if (argc > 0) { step = atol(*argv); argc--; argv++;}
164+
165+
if ((p = getenv("OPENBLAS_JOB"))) job=*p;
166+
167+
if ( job == 'N' ) factor = 10.0;
168+
169+
fprintf(stderr, "From : %3d To : %3d Step = %3d Job=%c\n", from, to, step,job);
170+
171+
if (( a = (FLOAT *)malloc(sizeof(FLOAT) * to * to * COMPSIZE)) == NULL){
172+
fprintf(stderr,"Out of Memory!!\n");exit(1);
173+
}
174+
175+
if (( vl = (FLOAT *)malloc(sizeof(FLOAT) * to * to * COMPSIZE)) == NULL){
176+
fprintf(stderr,"Out of Memory!!\n");exit(1);
177+
}
178+
179+
if (( vr = (FLOAT *)malloc(sizeof(FLOAT) * to * to * COMPSIZE)) == NULL){
180+
fprintf(stderr,"Out of Memory!!\n");exit(1);
181+
}
182+
183+
if (( wr = (FLOAT *)malloc(sizeof(FLOAT) * to * COMPSIZE)) == NULL){
184+
fprintf(stderr,"Out of Memory!!\n");exit(1);
185+
}
186+
187+
if (( wi = (FLOAT *)malloc(sizeof(FLOAT) * to * COMPSIZE)) == NULL){
188+
fprintf(stderr,"Out of Memory!!\n");exit(1);
189+
}
190+
191+
if (( rwork = (FLOAT *)malloc(sizeof(FLOAT) * to * COMPSIZE)) == NULL){
192+
fprintf(stderr,"Out of Memory!!\n");exit(1);
193+
}
194+
195+
for(j = 0; j < to; j++){
196+
for(i = 0; i < to * COMPSIZE; i++){
197+
a[i + j * to * COMPSIZE] = ((FLOAT) rand() / (FLOAT) RAND_MAX) - 0.5;
198+
}
199+
}
200+
201+
202+
lwork = -1;
203+
m=to;
204+
#ifndef COMPLEX
205+
GEEV (&job, &job, &m, a, &m, wr, wi, vl, &m, vr, &m, wkopt, &lwork, &info);
206+
#else
207+
GEEV (&job, &job, &m, a, &m, wr, vl, &m, vr, &m, wkopt, &lwork,rwork, &info);
208+
#endif
209+
210+
lwork = (blasint)wkopt[0];
211+
if (( work = (FLOAT *)malloc(sizeof(FLOAT) * lwork * COMPSIZE)) == NULL){
212+
fprintf(stderr,"Out of Memory!!\n");exit(1);
213+
}
214+
215+
216+
#ifdef linux
217+
srandom(getpid());
218+
#endif
219+
220+
fprintf(stderr, " SIZE FLops Time Lwork\n");
221+
222+
for(m = from; m <= to; m += step){
223+
224+
fprintf(stderr, " %6d : ", (int)m);
225+
gettimeofday( &start, (struct timezone *)0);
226+
227+
lwork = -1;
228+
#ifndef COMPLEX
229+
GEEV (&job, &job, &m, a, &m, wr, wi, vl, &m, vr, &m, wkopt, &lwork, &info);
230+
#else
231+
GEEV (&job, &job, &m, a, &m, wr, vl, &m, vr, &m, wkopt, &lwork,rwork, &info);
232+
#endif
233+
234+
lwork = (blasint)wkopt[0];
235+
#ifndef COMPLEX
236+
GEEV (&job, &job, &m, a, &m, wr, wi, vl, &m, vr, &m, work, &lwork, &info);
237+
#else
238+
GEEV (&job, &job, &m, a, &m, wr, vl, &m, vr, &m, work, &lwork,rwork, &info);
239+
#endif
240+
241+
gettimeofday( &stop, (struct timezone *)0);
242+
243+
if (info) {
244+
fprintf(stderr, "failed to compute eigenvalues .. %d\n", info);
245+
exit(1);
246+
}
247+
248+
time1 = (double)(stop.tv_sec - start.tv_sec) + (double)((stop.tv_usec - start.tv_usec)) * 1.e-6;
249+
250+
fprintf(stderr,
251+
" %10.2f MFlops : %10.2f Sec : %d\n",
252+
COMPSIZE * COMPSIZE * factor * (double)m * (double)m * (double)m / time1 * 1.e-6,time1,lwork);
253+
254+
255+
}
256+
257+
return 0;
258+
}
259+
260+
void main(int argc, char *argv[]) __attribute__((weak, alias("MAIN__")));

0 commit comments

Comments
 (0)