Browse Source

Merge pull request #554 from wernsaar/develop

added benchmarks for zgeru and cgeru
tags/v0.2.15^2
Zhang Xianyi 11 years ago
parent
commit
e42652f772
8 changed files with 57 additions and 14 deletions
  1. +37
    -4
      benchmark/Makefile
  2. +8
    -1
      benchmark/ger.c
  3. +1
    -1
      kernel/x86_64/caxpy.c
  4. +4
    -2
      kernel/x86_64/cdot.c
  5. +2
    -2
      kernel/x86_64/daxpy.c
  6. +1
    -1
      kernel/x86_64/ddot.c
  7. +1
    -1
      kernel/x86_64/zaxpy.c
  8. +3
    -2
      kernel/x86_64/zdot.c

+ 37
- 4
benchmark/Makefile View File

@@ -39,7 +39,7 @@ goto :: slinpack.goto dlinpack.goto clinpack.goto zlinpack.goto \
strsm.goto dtrsm.goto ctrsm.goto ztrsm.goto \
ssyrk.goto dsyrk.goto csyrk.goto zsyrk.goto \
ssyr2k.goto dsyr2k.goto csyr2k.goto zsyr2k.goto \
sger.goto dger.goto \
sger.goto dger.goto cger.goto zger.goto \
sdot.goto ddot.goto cdot.goto zdot.goto \
saxpy.goto daxpy.goto caxpy.goto zaxpy.goto \
sscal.goto dscal.goto cscal.goto zscal.goto \
@@ -62,7 +62,7 @@ acml :: slinpack.acml dlinpack.acml clinpack.acml zlinpack.acml \
strsm.acml dtrsm.acml ctrsm.acml ztrsm.acml \
ssyrk.acml dsyrk.acml csyrk.acml zsyrk.acml \
ssyr2k.acml dsyr2k.acml csyr2k.acml zsyr2k.acml \
sger.acml dger.acml \
sger.acml dger.acml cger.acml zger.acml \
sdot.acml ddot.acml cdot.acml zdot.acml \
saxpy.acml daxpy.acml caxpy.acml zaxpy.acml \
sscal.acml dscal.acml cscal.acml zscal.acml \
@@ -85,7 +85,7 @@ atlas :: slinpack.atlas dlinpack.atlas clinpack.atlas zlinpack.atlas \
strsm.atlas dtrsm.atlas ctrsm.atlas ztrsm.atlas \
ssyrk.atlas dsyrk.atlas csyrk.atlas zsyrk.atlas \
ssyr2k.atlas dsyr2k.atlas csyr2k.atlas zsyr2k.atlas \
sger.atlas dger.atlas \
sger.atlas dger.atlas cger.atlas zger.atlas\
sdot.atlas ddot.atlas \
saxpy.atlas daxpy.atlas caxpy.atlas zaxpy.atlas \
sscal.atlas dscal.atlas cscal.atlas zscal.atlas \
@@ -109,7 +109,7 @@ mkl :: slinpack.mkl dlinpack.mkl clinpack.mkl zlinpack.mkl \
strsm.mkl dtrsm.mkl ctrsm.mkl ztrsm.mkl \
ssyrk.mkl dsyrk.mkl csyrk.mkl zsyrk.mkl \
ssyr2k.mkl dsyr2k.mkl csyr2k.mkl zsyr2k.mkl \
sger.mkl dger.mkl \
sger.mkl dger.mkl cger.mkl zger.mkl \
sdot.mkl ddot.mkl cdot.mkl zdot.mkl \
saxpy.mkl daxpy.mkl caxpy.mkl zaxpy.mkl \
sscal.mkl dscal.mkl cscal.mkl zscal.mkl \
@@ -736,6 +736,32 @@ dger.atlas : dger.$(SUFFIX)
dger.mkl : dger.$(SUFFIX)
-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)

##################################### Cger ####################################################
cger.goto : cger.$(SUFFIX) ../$(LIBNAME)
$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm

cger.acml : cger.$(SUFFIX)
-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)

cger.atlas : cger.$(SUFFIX)
-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBATLAS) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)

cger.mkl : cger.$(SUFFIX)
-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)

##################################### Zger ####################################################
zger.goto : zger.$(SUFFIX) ../$(LIBNAME)
$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm

zger.acml : zger.$(SUFFIX)
-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBACML) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)

zger.atlas : zger.$(SUFFIX)
-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBATLAS) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)

zger.mkl : zger.$(SUFFIX)
-$(CC) $(CFLAGS) -o $(@F) $^ $(LIBMKL) $(CEXTRALIB) $(EXTRALIB) $(FEXTRALIB)

##################################### Ssymv ####################################################
ssymv.goto : ssymv.$(SUFFIX) ../$(LIBNAME)
$(CC) $(CFLAGS) -o $(@F) $^ $(CEXTRALIB) $(EXTRALIB) -lm
@@ -1348,6 +1374,13 @@ sger.$(SUFFIX) : ger.c
dger.$(SUFFIX) : ger.c
$(CC) $(CFLAGS) -c -UCOMPLEX -DDOUBLE -o $(@F) $^

cger.$(SUFFIX) : ger.c
$(CC) $(CFLAGS) -c -DCOMPLEX -UDOUBLE -o $(@F) $^

zger.$(SUFFIX) : ger.c
$(CC) $(CFLAGS) -c -DCOMPLEX -DDOUBLE -o $(@F) $^


ssymv.$(SUFFIX) : symv.c
$(CC) $(CFLAGS) -c -UCOMPLEX -UDOUBLE -o $(@F) $^



+ 8
- 1
benchmark/ger.c View File

@@ -35,12 +35,19 @@ USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.

#undef GER


#ifdef COMPLEX
#ifdef DOUBLE
#define GER BLASFUNC(zgeru)
#else
#define GER BLASFUNC(cgeru)
#endif
#else
#ifdef DOUBLE
#define GER BLASFUNC(dger)
#else
#define GER BLASFUNC(sger)
#endif
#endif


#if defined(__WIN32__) || defined(__WIN64__)


+ 1
- 1
kernel/x86_64/caxpy.c View File

@@ -84,7 +84,7 @@ int CNAME(BLASLONG n, BLASLONG dummy0, BLASLONG dummy1, FLOAT da_r, FLOAT da_i,
if ( (inc_x == 1) && (inc_y == 1) )
{

int n1 = n & -32;
BLASLONG n1 = n & -32;

if ( n1 )
{


+ 4
- 2
kernel/x86_64/cdot.c View File

@@ -109,7 +109,7 @@ FLOAT _Complex CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLASLONG in
if ( (inc_x == 1) && (inc_y == 1) )
{

int n1 = n & -16;
BLASLONG n1 = n & -16;

if ( n1 )
{
@@ -119,8 +119,10 @@ FLOAT _Complex CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLASLONG in
dot[4] += dot[6];
dot[5] += dot[7];
}

i = n1;
int j = i * 2;
BLASLONG j = i * 2;

while( i < n )
{



+ 2
- 2
kernel/x86_64/daxpy.c View File

@@ -80,9 +80,9 @@ int CNAME(BLASLONG n, BLASLONG dummy0, BLASLONG dummy1, FLOAT da, FLOAT *x, BLAS
{

#if defined(SANDYBRIDGE)
int n1 = n & -32;
BLASLONG n1 = n & -32;
#else
int n1 = n & -16;
BLASLONG n1 = n & -16;
#endif

if ( n1 )


+ 1
- 1
kernel/x86_64/ddot.c View File

@@ -83,7 +83,7 @@ FLOAT CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLASLONG inc_y)
if ( (inc_x == 1) && (inc_y == 1) )
{

int n1 = n & -16;
BLASLONG n1 = n & -16;

if ( n1 )
ddot_kernel_8(n1, x, y , &dot );


+ 1
- 1
kernel/x86_64/zaxpy.c View File

@@ -84,7 +84,7 @@ int CNAME(BLASLONG n, BLASLONG dummy0, BLASLONG dummy1, FLOAT da_r, FLOAT da_i,
if ( (inc_x == 1) && (inc_y == 1) )
{

int n1 = n & -16;
BLASLONG n1 = n & -16;

if ( n1 )
{


+ 3
- 2
kernel/x86_64/zdot.c View File

@@ -105,13 +105,14 @@ FLOAT _Complex CNAME(BLASLONG n, FLOAT *x, BLASLONG inc_x, FLOAT *y, BLASLONG in
if ( (inc_x == 1) && (inc_y == 1) )
{

int n1 = n & -8;
BLASLONG n1 = n & -8;

if ( n1 )
zdot_kernel_8(n1, x, y , dot );

i = n1;
int j = i * 2;
BLASLONG j = i * 2;

while( i < n )
{



Loading…
Cancel
Save