GCC 4.1.0 Performance Benchmark
A guy on kerneltrap posted a small performance benchmark on the recent GCC versions (3.3.5, 3.4.4, 4.0.1, 4.1.0).
GCC 3.3.5: 85.2 µsec (17% slower) GCC 3.4.4: 75.7 µsec (4% slower) GCC 4.0.1: 81.5 µsec (12% slower) GCC 4.1.0: 72.7 µsec (baseline)
Related posts:
- Java and C benchmark
- Linux Gets Automatic Tuning
- Linux Terminal Speed Performance Comparison
- Tomcat Performance: Linux faster than Windows
- MySQL Optimization and Performance Tips
- AMD DDR2 Linux Performance
- MySQL Performance Monitoring and Optimization Tools
- Optimize MySQL Performance With MySQLTuner
- Crack pdf password with pdfcrack
- Browse Faster Using a Local DNS Cache
Popular Related Items »
Incoming search terms
- gcc performance
- gcc benchmark
- gcc 3 vs 4
- benchmark gcc
- gcc optimization benchmark
- gcc arm performance
- arm gcc performance
- performance gcc
- gcc pipe performance
- gcc compiler performance
- gcc 4 1 performance
- gcc fpic performance
- gcc march performance
- vc vs gcc performance
- gcc vs vc benchmark
- gcc 4 performance
- gcc vc benchmark
- gcc -fpic performance
- gcc vs vc
- arm compiler benchmark
- gcc vs intel benchmark
- gcc performance tuning
- gcc optimization benchmarks
- mipsel gcc optimization
- gcc march performance gain
- benchmark gcc vs
- gcc3 vs gcc4
- linux howto benchmark gcc
- gcc -g performance
- march performance gcc
- linux gcc performance
- gcc windows performance
- gcc -pipe performance
- gcc optimization parameters
- gcc performance linux windows
- gcc benchmark versions
- gcc benchmark arm
- gcc 4 slower
- gcc 4 compiler performance
- gcc vc performance
- gcc arm benchmark
- gcc vs vc performance
- gcc march benchmark
- php gcc performance
- arm gcc benchmark
Tomasz Chmielewski said,
March 15, 2006 @ 11:34
Any more info on the benchmark? How was it tested?
My own “benchmark” (bzip -9 bash) on a mipsel platform shows that gcc-4.1.0 generated bzip2 binary was about 5% slower than bzip2 generated with gcc-3.4.4.
Claus Christensen said,
March 15, 2006 @ 15:29
I don’t now anymore about how he carried out the test, not even what gcc optimization parameters he used. I think I better do some testing myself.
What gcc parameters did you use?
Claus Christensen said,
March 15, 2006 @ 20:03
Here another GCC 4.1.0 benchmark from redeeman, that shows a performance gain:
http://www.linuxin.dk/artikler/index.php?id=7510
Towers of hanoi, gcc benchmark
Compile time:
3.4, c: 0.05s
4.0, c: 0.05s
4.1, c: 0.04s
3.4, c++: 0.10s
4.0, c++: 0.09s
4.1, c++: 0.05s
Conclusion: gcc 4.1 is faster to compile than previous versions
Binary Sizes(Unstripped):
3.4, c: 9161
4.0, c: 9377
4.1, c: 9161
3.4, c++: 9476
4.0, c++: 9508
4.1, c++: 9327
Conclusion: gcc 4.1 produces smaller binaries, most notably for c++
Binary Sizes(Stripped):
3.4, c: 4984
4.0, c: 4992
4.1, c: 4984
3.4, c++: 5224
4.0, c++: 5240
4.1, c++: 4984
Conclusion: Stripped binaries also smaller
Execution time (with 32 bricks):
3.4, c: 62.34s
4.0, c: 37.90s
4.1, c: 32.70s
3.4, c++: 52.62s
4.0, c++: 37.99s
4.1, c++: 33.63s
Conclusion: 4.1 is faster.