On Thu, 21 Sep 2023 13:26:14 GMT, Michal Sobierski <d...@openjdk.org> wrote:
>> sun/security/rsa/SignedObjectChain.java is very slow when run with C1, I >> suspect because some crypto intrinsics are only implemented in C2. Commit >> contains changes made to parallelize it. >> >> Comparison of before and after parallelization: >> time make test TEST=jdk/sun/security/rsa/SignedObjectChain.java >> TEST_VM_OPTS="-XX:+UseParallelGC -XX:TieredStopAtLevel=1" >> before: 270.72s user 4.88s system 108% cpu 4:14.43 total >> after: 410.76s user 7.50s system 555% cpu 1:15.23 total >> after second commit: 375.46s user 4.59s system 539% cpu 1:10.41 total >> >> time make test TEST=jdk/sun/security/rsa/SignedObjectChain.java >> TEST_VM_OPTS="-XX:+UseParallelGC" >> before: 63.67s user 4.67s system 161% cpu 42.424 total >> after: 130.36s user 7.47s system 585% cpu 23.526 total >> after second commit: 67.31s user 4.48s system 417% cpu 17.183 total >> >> time make test TEST=jdk/sun/security/rsa/SignedObjectChain.java >> TEST_VM_OPTS="-XX:+UseShenandoahGC -XX:TieredStopAtLevel=1" >> before: 281.99s user 5.54s system 108% cpu 4:24.09 total >> after: 386.98s user 8.62s system 496% cpu 1:19.73 total >> after second commit: 413.51s user 5.08s system 613% cpu 1:08.25 total >> >> time make test TEST=jdk/sun/security/rsa/SignedObjectChain.java >> TEST_VM_OPTS="-XX:+UseShenandoahGC" >> before: 65.86s user 5.05s system 156% cpu 45.215 total >> after: 135.90s user 7.66s system 585% cpu 24.502 total >> after second commit: 83.25s user 4.82s system 469% cpu 18.741 total > > Michal Sobierski has updated the pull request incrementally with one > additional commit since the last revision: > > Better approach to parallelize sun/security/rsa/SignedObjectChain.java > > time make test TEST=jdk/sun/security/rsa/SignedObjectChain.java > TEST_VM_OPTS="-XX:+UseParallelGC -XX:TieredStopAtLevel=1" > before: 270.72s user 4.88s system 108% cpu 4:14.43 total > after: 375.46s user 4.59s system 539% cpu 1:10.41 total > > time make test TEST=jdk/sun/security/rsa/SignedObjectChain.java > TEST_VM_OPTS="-XX:+UseParallelGC" > before: 63.67s user 4.67s system 161% cpu 42.424 total > after: 67.31s user 4.48s system 417% cpu 17.183 total > > time make test TEST=jdk/sun/security/rsa/SignedObjectChain.java > TEST_VM_OPTS="-XX:+UseShenandoahGC -XX:TieredStopAtLevel=1" > before: 281.99s user 5.54s system 108% cpu 4:24.09 total > after: 413.51s user 5.08s system 613% cpu 1:08.25 total > > time make test TEST=jdk/sun/security/rsa/SignedObjectChain.java > TEST_VM_OPTS="-XX:+UseShenandoahGC" > before: 65.86s user 5.05s system 156% cpu 45.215 total > after: 83.25s user 4.82s system 469% cpu 18.741 total Shows a very good improvement here as well: % time CONF=linux-x86_64-server-fastdebug make test TEST=sun/security/rsa/SignedObjectChain.java TEST_VM_OPTS="-XX:TieredStopAtLevel=1" Baseline: CONF=linux-x86_64-server-fastdebug make test 333.34s user 5.04s system 111% cpu 5:02.33 total CONF=linux-x86_64-server-fastdebug make test 292.03s user 5.27s system 114% cpu 4:20.71 total CONF=linux-x86_64-server-fastdebug make test 330.51s user 5.23s system 111% cpu 5:00.18 total Patched: CONF=linux-x86_64-server-fastdebug make test 309.72s user 4.92s system 505% cpu 1:02.25 total CONF=linux-x86_64-server-fastdebug make test 275.63s user 5.02s system 540% cpu 51.82 total CONF=linux-x86_64-server-fastdebug make test 299.18s user 5.06s system 546% cpu 55.64 total ------------- PR Comment: https://git.openjdk.org/jdk/pull/15860#issuecomment-1729655487
