vader fix for 4937 (PPC wmb location) #4955
Conversation
|
Has this been put in master yet? |
|
Can lead to incorrect answers. Clear bug. |
|
I don't think master has the bug. The function names are a bit different in that branch, but I think it's right in master (didn't hit any bugs on PPC anyway with master). My change in this PR was partially based on just the logic behind what a function called set_header() is probably doing, but also partially based on comparing vs what master has. |
|
@markalle Ok. You are correct that this was the intended order. Its a typo that isn't caught on x86. |
|
@bwbarrett Is this something small enough that it can be included in v3.0.1 without another RC? |
|
Is this needed on v2.x or v3.1.x? |
|
I don't think it's needed anywhere else, it's a pretty recent regression. I browsed in branch "v3.1.x" to make sure and doesn't look like it needs to go there. |
I have some tests that failed on PPC with the recent vader changes.
I'm suspicious of the set_header() function where there's a wmb()
call that looks like it boils down to
set some data
set the header to indicate the data is available
wmb
and I think the wmb needs to go up a line.
More details here:
open-mpi#4937
with a copy of the "maxsoak.c" testcase at
https://gist.github.com/markalle/a1c203297cb6af22a3fb5c24e62b2ba3
Signed-off-by: Mark Allen <markalle@us.ibm.com>
|
Oh shoot, re-pushing with the commit signed |
I have some tests that failed on PPC with the recent vader changes.
I'm suspicious of the set_header() function where there's a wmb()
call that looks like it boils down to
set some data
set the header to indicate the data is available
wmb
and I think the wmb needs to go up a line.
More details here:
#4937
with a copy of the "maxsoak.c" testcase at
https://gist.github.com/markalle/a1c203297cb6af22a3fb5c24e62b2ba3
Fixes #4937