Advertisement
This is comment for paste
TEST_CPU Multithread Version of Polyhedron Benchmark
Not a member of Pastebin yet?
Sign Up,
it unlocks many cool features!
- Updated version of multithreading TEST_CPU Fortran.uk Polyhedron Fortran Benchmark : that single thread (https://openbenchmarking.org/test/pts/polyhedron&eval=111335e02dea1589bab7a9564f8efef17d907dcb#metrics) Implemented by: David Frank [email protected], original gauss Tim Prince [email protected], Crout James Van Buskirk [email protected], Jos Bergervoet [email protected]. NOW gauss single thread with looping repetition OpenMP, crout single thread with looping repetition OpenMP, crout multi thread, and implementing with vendor blas lapack instead building own. by default as same as phoroix is 1000 repetition dual inversing 250x250, and 2000x2000 dual inversing, and check the difference.
- Example: ryzen 3900x
- C:\Temp>test_fpu_gemini_mt5.exe
- Inside arraySub() allocating memory ...
- Memory allocation sufficient
- Benchmark running, hopefully as only ACTIVE task
- Number of OpenMP threads: 24
- Test1: Gauss mt- 1000 ( 250x 250) inverts in 0.723 seconds Err= 0.5971541960567150E-14
- Test2: Croutstmt- 1000 ( 250x 250) inverts in 0.865 seconds Err= 0.7541962840872367E-14
- Calling Croutmt- 1 Bigsize
- Calling Croutmt- 2 Bigsize
- Test3: Croutmt- 2 ( 2000x 2000) inverts in 3.016 seconds Err= 0.9010735252907198E-09
- Test4: DGETRF - 2 ( 2000x 2000) inverts in 0.294 seconds Err= 0.2360292208074361E-11
- Test5: DGETRF2- 2 ( 2000x 2000) inverts in 0.289 seconds Err= 0.2635024887233167E-11
- Test6: DGESV - 2 ( 2000x 2000) inverts in 0.199 seconds Err= 0.2345595889465859E-11
- Total = 5.4 sec
Advertisement
Add Comment
Please, Sign In to add comment
Advertisement