Performance Results: Speedup for DGEMM/SGEMM
Home
News
Publications
People
Performance Results
Speedup for DGEMM/SGEMM
DGETRF/SGETRF and DGESV/DSGESV on various architectures
Cell Performance
Presentations
LAPACK Code
dsgesv.f
dslaconvertged2s.f
dslaconvertges2d.f
CELL BE Code
LAPACK Code
LAPACK Code
dsgesv.f
dslaconvertged2s.f
dslaconvertges2d.f
Performance Results
Performance Results
Speedup for DGEMM/SGEMM
DGETRF/SGETRF and DGESV/DSGESV on various architectures
Cell Performance
Performance Results
Ratio of execution times (speedup) for DGEMM/SGEMM (m=n=k)
Architecture (BLAS)
n
DGEMM
/
SGEMM
DGETRF
/
SGETRF
DGESV
/
DSGESV
# iter
Intel Pentium III Coppermine (Goto)
3500
2.10
2.24
1.92
4
Intel Pentium III Katmai (Goto)
3000
2.12
2.11
1.79
4
Sun UltraSPARC IIe (Sunperf)
3000
1.45
1.79
1.58
4
Intel Pentium IV Prescott (Goto)
4000
2.00
1.86
1.57
5
Intel Pentium IV-M Northwood (Goto)
4000
2.02
1.98
1.54
5
AMD Opteron (Goto)
4000
1.98
1.93
1.53
5
Cray X1 (libsci)
4000
1.68
1.54
1.38
7
IBM Power PC G5 (2.7 GHz) (VecLib)
5000
2.29
2.05
1.24
5
Compaq Alpha EV6 (CXML)
3000
0.99
1.08
1.01
4
IBM SP Power3 (ESSL)
3000
1.03
1.13
1.00
3
SGI Octane (ATLAS)
2000
1.08
1.13
0.91
4
Intel Itanium 2 (Goto and ATLAS)
1500
0.71
Jun 29 2022
Admin Login