Performance Evaluation of Strassen Matrix Multiplication algorithms

[SergeyKostrov]

*** Performance Evaluation of Strassen Matrix Multiplication algorithms ***

[SergeyKostrov]

[ *** MinGW C++ compiler v6.1.0 64-bit *** ] [ SHBC 32768x32768 DRP=3rd ] Test terminated / Disk-bound processing / More than 128 GB of Virtual Memory used

[SergeyKostrov]

[ *** MinGW C++ compiler v6.1.0 64-bit *** ] [ SHBC 32768x32768 DRP=4th ] Test terminated / Disk-bound processing / More than 192 GB of Virtual Memory used

[SergeyKostrov]

[ *** MinGW C++ compiler v6.1.0 64-bit *** ] [ SHBC 32768x32768 DRP=5th ] Test terminated / Disk-bound processing / More than 224 GB of Virtual Memory used

[SergeyKostrov]

[ *** MinGW C++ compiler v6.1.0 64-bit *** ] [ MKL 32768x32768 ] Cblas SGEMM Matrix Size : 32768 x 32768 Matrix Size Threshold : N/A Matrix Partitions : N/A Degree of Recursion : N/A Result Sets Reflection: N/A Calculating... Cblas SGEMM - Pass 01 - Completed: 428.51901 secs Cblas SGEMM - Pass 02 - Completed: 428.65903 secs Cblas SGEMM - Pass 03 - Completed: 428.97202 secs Cblas SGEMM - Pass 04 - Completed: 428.75408 secs Cblas SGEMM - Pass 05 - Completed: 428.35297 secs

[SergeyKostrov]

////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// // [ 32768x32768 / OpenMP Threads=4 / Release ] //////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////

[SergeyKostrov]

[ *** Intel C++ compiler v13.1.0 ( u149 ) 64-bit *** ] [ SHBI 32768x32768 DRP=1st ] Strassen HBI Matrix Size : 32768 x 32768 Matrix Size Threshold : 16384 x 16384 Matrix Partitions : 8 Degree of Recursion : 1 Result Sets Reflection: N/A Calculating... Strassen HBI - Pass 01 - Completed: 381.06299 secs Strassen HBI - Pass 02 - Completed: 379.42599 secs Strassen HBI - Pass 03 - Completed: 379.47198 secs Strassen HBI - Pass 04 - Completed: 379.51999 secs Strassen HBI - Pass 05 - Completed: 379.31601 secs

[SergeyKostrov]

[ *** Intel C++ compiler v13.1.0 ( u149 ) 64-bit *** ] [ SHBC 32768x32768 DRP=1st ] Strassen HBC Matrix Size : 32768 x 32768 Matrix Size Threshold : 16384 x 16384 Matrix Partitions : 8 Degree of Recursion : 1 Result Sets Reflection: Disabled Calculating... Strassen HBC - Pass 01 - Completed: 380.48700 secs Strassen HBC - Pass 02 - Completed: 379.78400 secs Strassen HBC - Pass 03 - Completed: 379.67499 secs Strassen HBC - Pass 04 - Completed: 379.39499 secs Strassen HBC - Pass 05 - Completed: 379.56601 secs

[SergeyKostrov]

[ *** Intel C++ compiler v13.1.0 ( u149 ) 64-bit *** ] [ SHBC 32768x32768 DRP=2nd ] Strassen HBC Matrix Size : 32768 x 32768 Matrix Size Threshold : 8192 x 8192 Matrix Partitions : 57 Degree of Recursion : 2 Result Sets Reflection: Enabled Calculating... Strassen HBC - Pass 01 - Completed: 1596.04700 secs Strassen HBC - Pass 02 - Completed: 2793.77295 secs Strassen HBC - Pass 03 - Completed: 2286.28589 secs Strassen HBC - Pass 04 - Completed: 2459.93367 secs Strassen HBC - Pass 05 - Completed: 2375.14252 secs

[SergeyKostrov]

[ *** Intel C++ compiler v13.1.0 ( u149 ) 64-bit *** ] [ SHBC 32768x32768 DRP=3rd ] Test terminated / Disk-bound processing / More than 128 GB of Virtual Memory used

[SergeyKostrov]

[ *** Intel C++ compiler v13.1.0 ( u149 ) 64-bit *** ] [ SHBC 32768x32768 DRP=4th ] Test terminated / Disk-bound processing / More than 192 GB of Virtual Memory used

[SergeyKostrov]

[ *** Intel C++ compiler v13.1.0 ( u149 ) 64-bit *** ] [ SHBC 32768x32768 DRP=5th ] Test terminated / Disk-bound processing / More than 224 GB of Virtual Memory used

[SergeyKostrov]

[ *** Intel C++ compiler v13.1.0 ( u149 ) 64-bit *** ] [ MKL 32768x32768 ] Cblas SGEMM Matrix Size : 32768 x 32768 Matrix Size Threshold : N/A Matrix Partitions : N/A Degree of Recursion : N/A Result Sets Reflection: N/A Calculating... Cblas SGEMM - Pass 01 - Completed: 428.29999 secs Cblas SGEMM - Pass 02 - Completed: 428.53500 secs Cblas SGEMM - Pass 03 - Completed: 428.58099 secs Cblas SGEMM - Pass 04 - Completed: 428.70599 secs Cblas SGEMM - Pass 05 - Completed: 428.62799 secs

[SergeyKostrov]

////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// // [ 40960x40960 / OpenMP Threads=4 / Release ] //////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////

[SergeyKostrov]

[ *** Microsoft C++ compiler ( VS2008 PE ) 64-bit *** ] [ SHBI 40960x40960 DRP=1st ] Matrix Size : 40960 x 40960 Matrix Size Threshold : 20480 x 20480 Matrix Partitions : 8 Degree of Recursion : 1 Result Sets Reflection: N/A Calculating... Strassen HBI - Pass 01 - Completed: 1954.64612 secs Strassen HBI - Pass 02 - Completed: 2748.12915 secs Strassen HBI - Pass 03 - Completed: 2752.48218 secs Strassen HBI - Pass 04 - Completed: 3227.16113 secs Strassen HBI - Pass 05 - Completed: 3175.94897 secs

[SergeyKostrov]

[ *** Microsoft C++ compiler ( VS2008 PE ) 64-bit *** ] [ SHBC 40960x40960 DRP=1st ] Strassen HBC Matrix Size : 40960 x 40960 Matrix Size Threshold : 20480 x 20480 Matrix Partitions : 8 Degree of Recursion : 1 Result Sets Reflection: Disabled Calculating... Strassen HBC - Pass 01 - Completed: 1890.52905 secs Strassen HBC - Pass 02 - Completed: 2849.40503 secs Strassen HBC - Pass 03 - Completed: 2978.29028 secs Strassen HBC - Pass 04 - Completed: 3034.83936 secs Strassen HBC - Pass 05 - Completed: 3552.54907 secs

[SergeyKostrov]

[ *** Microsoft C++ compiler ( VS2008 PE ) 64-bit *** ] [ SHBC 40960x40960 DRP=2nd ] Test terminated / Disk-bound processing / More than 128 GB of Virtual Memory used

[SergeyKostrov]

[ *** Microsoft C++ compiler ( VS2008 PE ) 64-bit *** ] [ SHBC 40960x40960 DRP=3rd ] Test terminated / Disk-bound processing / More than 128 GB of Virtual Memory used

[SergeyKostrov]

[ *** Microsoft C++ compiler ( VS2008 PE ) 64-bit *** ] [ SHBC 40960x40960 DRP=4th ] Test terminated / Disk-bound processing / More than 128 GB of Virtual Memory used

[SergeyKostrov]

[ *** Microsoft C++ compiler ( VS2008 PE ) 64-bit *** ] [ SHBC 40960x40960 DRP=5th ] Test terminated / Disk-bound processing / More than 128 GB of Virtual Memory used

[SergeyKostrov]

[ *** Microsoft C++ compiler ( VS2008 PE ) 64-bit *** ] [ MKL 40960x40960 ] Cblas SGEMM Matrix Size : 40960 x 40960 Matrix Size Threshold : N/A Matrix Partitions : N/A Degree of Recursion : N/A Result Sets Reflection: N/A Calculating... Cblas SGEMM - Pass 01 - Completed: 825.21405 secs Cblas SGEMM - Pass 02 - Completed: 823.57605 secs Cblas SGEMM - Pass 03 - Completed: 823.63806 secs Cblas SGEMM - Pass 04 - Completed: 823.47162 secs Cblas SGEMM - Pass 05 - Completed: 823.39475 secs

[SergeyKostrov]

////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////// // [ 40960x40960 / OpenMP Threads=4 / Release ] //////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////////