- 新着としてマーク
- ブックマーク
- 購読
- ミュート
- RSS フィードを購読する
- ハイライト
- 印刷
- 不適切なコンテンツを報告
I have a block of code where I initialize ~20 1-dimension arrays, each with 250 REAL(8) elements. IVF says it fusesthe block of array assignments and then auto-parallelizes the loop (/Qparallel). When I execute theprogram, it crashesdue to a stack error (the exe stack space is set to 200MB!). If It put a !DEC$NOPARALLEL before the block of array assignments, it no longer crashes so its obviously the offending "loop".Or if I set OMP_NUM_THREADS=1, it no longer crashes so it must have something to do with thread creation.
I have tried beaking out the code into a small test case, but for some reason IVF will no longer fuse the block of assigments into a loop. Any ideas what could be going on here?
Al Greynolds www.ruda.com
- 新着としてマーク
- ブックマーク
- 購読
- ミュート
- RSS フィードを購読する
- ハイライト
- 印刷
- 不適切なコンテンツを報告
I have a block of code where I initialize ~20 1-dimension arrays, each with 250 REAL(8) elements. IVF says it fusesthe block of array assignments and then auto-parallelizes the loop (/Qparallel). When I execute theprogram, it crashesdue to a stack error (the exe stack space is set to 200MB!). If It put a !DEC$NOPARALLEL before the block of array assignments, it no longer crashes so its obviously the offending "loop".Or if I set OMP_NUM_THREADS=1, it no longer crashes so it must have something to do with thread creation.
This looks like a problem with the defaults for KMP_STACKSIZE, which is discussed in the help file.
コピーされたリンク
- 新着としてマーク
- ブックマーク
- 購読
- ミュート
- RSS フィードを購読する
- ハイライト
- 印刷
- 不適切なコンテンツを報告
I have a block of code where I initialize ~20 1-dimension arrays, each with 250 REAL(8) elements. IVF says it fusesthe block of array assignments and then auto-parallelizes the loop (/Qparallel). When I execute theprogram, it crashesdue to a stack error (the exe stack space is set to 200MB!). If It put a !DEC$NOPARALLEL before the block of array assignments, it no longer crashes so its obviously the offending "loop".Or if I set OMP_NUM_THREADS=1, it no longer crashes so it must have something to do with thread creation.
This looks like a problem with the defaults for KMP_STACKSIZE, which is discussed in the help file.