performance of loongson 2f
Some friends ask me about the performance of loongson2f. They want to know whether the performance of loongson 2f can surpass Marvell Sheeva CPU. Well I can not just say it's better or worse without giving the benchmark data.
Since there is a benchmark result of Marvell Sheeva CPU, we can run the same benchmark program on loongson 2f. The benchmark program is nbench.
Machine: gdium
OS: Debian squeeze
Kernel: Linux
1. gcc-4.3.4
CFLAGS = -s -static -Wall -O3
| TEST | Iterations/sec. | Old Index | New Index |
| Pentium 90* | AMD K6/233* | ||
| NUMERIC SORT | 358.24. | 9.19 | 3.02 |
| STRING SORT | 33.041 | 14.76 | 2.29 |
| BITFIELD | 5.5164e+07 | 9.46 | 1.98 |
| FP EMULATION | 47.402 | 22.75 | 5.25 |
| FOURIER | 4721.1 | 5.37 | 3.02 |
| ASSIGNMENT | 7.0534 | 26.84 | 6.96 |
| IDEA | 1597.4 | 24.43 | 7.25 |
| HUFFMAN | 575.17 | 15.95 | 5.09 |
| NEURAL NET | 4.2065 | 6.76 | 2.84 |
| LU DECOMPOSITION | 107.28 | 5.56 | 4.01 |
==========ORIGINAL BYTEMARK RESULTS========
INTEGER INDEX : 16.297
FLOATING-POINT INDEX: 5.864
Baseline (MSDOS*) : Pentium* 90, 256 KB L2-cache, Watcom* compiler 10.0
==========LINUX DATA BELOW============
CPU :
L2 Cache :
OS : Linux 2.6.24-gdium-1
C compiler : gcc version 4.3.4 (Debian 4.3.4-5)
libc : libc-2.9.so
MEMORY INDEX : 3.156
INTEGER INDEX : 4.918
FLOATING-POINT INDEX: 3.252
Baseline (LINUX) : AMD K6/233*, 512 KB L2-cache, gcc 2.7.2.3, libc-5.4.38
* Trademarks are property of their respective holder.
2. gcc-4.4
CFLAGS = -s -static -Wall -O3 -fomit-frame-pointer -funroll-loops
CFLAGS += -march=loongson2f -mtune=loongson2f -mabi=n32
| TEST | Iterations/sec. | Old Index | New Index |
| Pentium 90* | AMD K6/233* | ||
| NUMERIC SORT | 366.08 | 9.39 | 3.08 |
| STRING SORT | 46.686 | 20.86 | 3.23 |
| BITFIELD | 4.764e+07 | 8.17 | 1.71 |
| FP EMULATION | 90.2 | 43.28 | 9.99 |
| FOURIER | 5171.9 | 5.88 | 3.30 |
| ASSIGNMENT | 11.094 | 42.21 | 10.95 |
| IDEA | 1726.9 | 26.41 | 7.84 |
| HUFFMAN | 605 | 16.78 | 5.36 |
| NEURAL NET | 9.761 | 15.68 | 6.60 |
| LU DECOMPOSITION | 215.64 | 11.17 | 8.07 |
==========ORIGINAL BYTEMARK RESULTS========
INTEGER INDEX : 20.035
FLOATING-POINT INDEX: 10.100
Baseline (MSDOS*) : Pentium* 90, 256 KB L2-cache, Watcom* compiler 10.0
==========LINUX DATA BELOW============
CPU :
L2 Cache :
OS : Linux 2.6.24-gdium-1
C compiler : gcc-4.4
libc : libc-2.9.so
MEMORY INDEX : 3.922
INTEGER INDEX : 5.997
FLOATING-POINT INDEX: 5.602
Baseline (LINUX) : AMD K6/233*, 512 KB L2-cache, gcc 2.7.2.3, libc-5.4.38
* Trademarks are property of their respective holder.
You should probably also link to the Sheevaplug benchmark for comparison: http://computingplugs.com/index.php/SheevaPlug_Performance