| commit | d6be803f5dd491ddc507bdb1223433d7e1ebd67a | |
| author | amonakov <amonakov@138bc75d-0d04-0410-961f-82ee72b054a4> | |
| Mon, 11 Jan 2016 12:55:31 +0000 (11 12:55 +0000) | ||
| committer | amonakov <amonakov@138bc75d-0d04-0410-961f-82ee72b054a4> | |
| Mon, 11 Jan 2016 12:55:31 +0000 (11 12:55 +0000) | ||
| tree | 9e5efed6bfd5c03b33fb3b973ab51fdeae168032 | treesnapshot (tar.gz zip) |
| parent | 1ef0270e330da4938f5cb3371a45f2fb528eb4c4 | commitdiff |
nvptx plugin: do not force JIT target SM version
When link_ptx runs, a CUDA device is already bound to current thread, so the
driver library knows the target architecture. There isn't any benefit from
forcing a specific target here; on the contrary, hardcoding sm_30 breaks
offloading on later (Maxwell, sm_5x) devices.
* plugin/plugin-nvptx.c (link_ptx): Do not set CU_JIT_TARGET.
git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@232227 138bc75d-0d04-0410-961f-82ee72b054a4
When link_ptx runs, a CUDA device is already bound to current thread, so the
driver library knows the target architecture. There isn't any benefit from
forcing a specific target here; on the contrary, hardcoding sm_30 breaks
offloading on later (Maxwell, sm_5x) devices.
* plugin/plugin-nvptx.c (link_ptx): Do not set CU_JIT_TARGET.
git-svn-id: svn+ssh://gcc.gnu.org/svn/gcc/trunk@232227 138bc75d-0d04-0410-961f-82ee72b054a4
| libgomp/ChangeLog | diffblobblamehistory | |
| libgomp/plugin/plugin-nvptx.c | diffblobblamehistory |