Commit
This commit does not belong to any branch on this repository, and may belong to a fork outside of the repository.
[SYSTEMDS-3636] Ultra-sparse tsmm right w/ multi-threaded transpose
Further improvement of ultra-sparse tsmm right: w/ multi-threaded transpose. On a scenario of 10 times G %*% t(G) with G being germany_osm, the runtime changes as follows OLD: matrix mult: 13.346214013s matrix mult: 5.498598342s matrix mult: 5.11548485s matrix mult: 5.573473983s matrix mult: 5.673529942s matrix mult: 6.08607291s matrix mult: 6.244303553s matrix mult: 6.422722927s matrix mult: 4.995632087s matrix mult: 9.085500786s SystemDS Statistics: Total elapsed time: 71.007 sec. Total compilation time: 0.792 sec. Total execution time: 70.215 sec. Cache hits (Mem/Li/WB/FS/HDFS): 11/0/0/0/1. Cache writes (Li/WB/FS/HDFS): 0/11/0/0. Cache times (ACQr/m, RLS, EXP): 4.696/0.000/9.415/0.000 sec. HOP DAGs recompiled (PRED, SB): 0/0. HOP DAGs recompile time: 0.000 sec. Total JIT compile time: 5.987 sec. Total JVM GC count: 1. Total JVM GC time: 0.059 sec. Heavy hitter instructions: 1 tsmm 68.039 10 2 != 1.577 1 3 uak+ 0.585 1 4 + 0.027 22 5 print 0.008 12 6 mvvar 0.001 31 7 createvar 0.001 12 8 rmvar 0.000 45 9 time 0.000 20 10 - 0.000 10 NEW matrix mult: 12.17142539s matrix mult: 5.063393773s matrix mult: 4.764698928s matrix mult: 4.771695393s matrix mult: 5.434539822s matrix mult: 4.640708695s matrix mult: 4.967180443s matrix mult: 5.156199379s matrix mult: 5.472330144s matrix mult: 5.310449401s SystemDS Statistics: Total elapsed time: 60.405 sec. Total compilation time: 0.880 sec. Total execution time: 59.525 sec. Cache hits (Mem/Li/WB/FS/HDFS): 11/0/0/0/1. Cache writes (Li/WB/FS/HDFS): 0/11/0/0. Cache times (ACQr/m, RLS, EXP): 4.223/0.000/8.626/0.000 sec. HOP DAGs recompiled (PRED, SB): 0/0. HOP DAGs recompile time: 0.000 sec. Total JIT compile time: 6.956 sec. Total JVM GC count: 1. Total JVM GC time: 0.062 sec. Heavy hitter instructions: 1 tsmm 57.751 10 2 != 1.296 1 3 uak+ 0.465 1 4 + 0.029 22 5 print 0.008 12 6 mvvar 0.001 31 7 createvar 0.001 12 8 rmvar 0.000 45 9 time 0.000 20 10 - 0.000 10
- Loading branch information