Thanks much for the fix Jim!!<div><br></div><div>Sreeram Potluri</div><div><br><div class="gmail_quote">On Mon, Dec 12, 2011 at 6:30 PM, Jim Dinan <span dir="ltr"><<a href="mailto:dinan@mcs.anl.gov">dinan@mcs.anl.gov</a>></span> wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">Hi Sreeram,<br>
<br>
I committed a fix to the MPICH2 trunk (r9335). I added the missing initializations and set local buffers to -1.0 to make the tests more robust when run on a single process. I also fixed those bad floating point comparisons.<br>
<br>
Thanks again for reporting this.<br>
<br>
Best wishes,<br>
~Jim.<div class="HOEnZb"><div class="h5"><br>
<br>
On 12/12/11 4:16 PM, Jim Dinan wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Hi Sreeram,<br>
<br>
Your understanding of the test case is spot on; I just didn't read it<br>
closely enough before responding. I think this bug is the culprit:<br>
<br>
if (actual - expected > 1e-10) {<br>
<br>
This should have an fabs(), somehow we've just been getting lucky:<br>
<br>
if (fabs(actual - expected) > 1e-10) {<br>
<br>
This affects all of those strided tests. I'll put together a fix and let<br>
you know as soon as it's checked in. Great catch and thanks for the bug<br>
report!<br>
<br>
Best,<br>
~Jim.<br>
<br>
<br>
On 12/12/11 3:47 PM, sreeram potluri wrote:<br>
<blockquote class="gmail_quote" style="margin:0 0 0 .8ex;border-left:1px #ccc solid;padding-left:1ex">
Jim,<br>
<br>
My understanding of the test is that only a part of the loc_buf is<br>
overwritten by the MPI_Get operation (8*256 rather than 8*1024).<br>
<br>
Now the test also tries to validate the rest of the loc_buf which is not<br>
affected by the MPI_Get operation. This check can fail or pass based on<br>
the original value of the loc_buf, which is never set.<br>
<br>
Am I missing something basic?<br>
<br>
Sreeram Potluri<br>
<br>
On Mon, Dec 12, 2011 at 3:11 PM, Jim Dinan <<a href="mailto:dinan@mcs.anl.gov" target="_blank">dinan@mcs.anl.gov</a><br>
<mailto:<a href="mailto:dinan@mcs.anl.gov" target="_blank">dinan@mcs.anl.gov</a>>> wrote:<br>
<br>
Hi Sreeram,<br>
<br>
Setting the initial value of loc_buf shouldn't influence whether or<br>
not this test passes; the entire contents of loc_buf should be<br>
overwritten later by the call to MPI_Get. As Dave pointed out,<br>
setting this to 1.0+rank also weakens the test when run on one<br>
process since loc_buf will already contain the correct data. This<br>
is not the case when run on more processes since loc_buf should<br>
contain 1.0+((rank+1)%nproc)).<br>
<br>
This test has been in the MPICH test suite for several months and<br>
hasn't given us any trouble. Any chance that it's exercising a bug<br>
on your end?<br>
<br>
Best,<br>
~Jim.<br>
<br>
<br>
On 12/12/11 1:27 PM, Dave Goodell wrote:<br>
<br>
Sreeram,<br>
<br>
I'm not very familiar with this test, but this looks like an<br>
incorrect change to me. By setting the loc_buf (the origin<br>
"receive" buffer) to the expected value in this way then you<br>
open up a testing coverage gap if MPI_Get is replaced with a<br>
no-op function.<br>
<br>
If anything, this test could probably be strengthened by adding<br>
the following line in the same location that you recommend<br>
making your change:<br>
<br>
----8<----<br>
*(loc_buf + i) = -1.0;<br>
----8<----<br>
<br>
This way you don't assume anything about the values already in<br>
the buffer returned by MPI_Alloc_mem.<br>
<br>
Perhaps Jim can comment, since he wrote the test originally.<br>
<br>
-Dave<br>
<br>
On Dec 12, 2011, at 6:53 AM CST, sreeram potluri wrote:<br>
<br>
Hi Team,<br>
<br>
We have seen a validation error with the strided_get_indexed<br>
test in our internal testing with MVAPICH2. This appears to<br>
be an initialization issue with the benchmark. The path<br>
below fixes it in one way. Please accept the change if<br>
appropriate.<br>
<br>
Index: mpi/rma/strided_get_indexed.c<br>
==============================<u></u>__============================<u></u>==__=======<br>
--- mpi/rma/strided_get_indexed.c (revision 5064)<br>
+++ mpi/rma/strided_get_indexed.c (working copy)<br>
@@ -48,8 +48,10 @@<br>
if (rank == 0)<br>
if (verbose) printf("MPI RMA Strided Get Test:\n");<br>
<br>
- for (i = 0; i< XDIM*YDIM; i++)<br>
+ for (i = 0; i< XDIM*YDIM; i++) {<br>
*(win_buf + i) = 1.0 + rank;<br>
+ *(loc_buf + i) = 1.0 + rank;<br>
+ }<br>
<br>
MPI_Win_create(win_buf, bufsize, 1, MPI_INFO_NULL,<br>
MPI_COMM_WORLD,&buf_win);<br>
<br>
Regards<br>
Sreeram Potluri<br>
______________________________<u></u>___________________<br>
mpich-discuss mailing list <a href="mailto:mpich-discuss@mcs.anl.gov" target="_blank">mpich-discuss@mcs.anl.gov</a><br>
<mailto:<a href="mailto:mpich-discuss@mcs.anl.gov" target="_blank">mpich-discuss@mcs.anl.<u></u>gov</a>><br>
To manage subscription options or unsubscribe:<br>
<a href="https://lists.mcs.anl.gov/__mailman/listinfo/mpich-discuss" target="_blank">https://lists.mcs.anl.gov/__<u></u>mailman/listinfo/mpich-discuss</a><br>
<<a href="https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss" target="_blank">https://lists.mcs.anl.gov/<u></u>mailman/listinfo/mpich-discuss</a><u></u>><br>
<br>
<br>
______________________________<u></u>___________________<br>
mpich-discuss mailing list <a href="mailto:mpich-discuss@mcs.anl.gov" target="_blank">mpich-discuss@mcs.anl.gov</a><br>
<mailto:<a href="mailto:mpich-discuss@mcs.anl.gov" target="_blank">mpich-discuss@mcs.anl.<u></u>gov</a>><br>
To manage subscription options or unsubscribe:<br>
<a href="https://lists.mcs.anl.gov/__mailman/listinfo/mpich-discuss" target="_blank">https://lists.mcs.anl.gov/__<u></u>mailman/listinfo/mpich-discuss</a><br>
<<a href="https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss" target="_blank">https://lists.mcs.anl.gov/<u></u>mailman/listinfo/mpich-discuss</a><u></u>><br>
<br>
<br>
<br>
<br>
______________________________<u></u>_________________<br>
mpich-discuss mailing list <a href="mailto:mpich-discuss@mcs.anl.gov" target="_blank">mpich-discuss@mcs.anl.gov</a><br>
To manage subscription options or unsubscribe:<br>
<a href="https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss" target="_blank">https://lists.mcs.anl.gov/<u></u>mailman/listinfo/mpich-discuss</a><br>
</blockquote>
______________________________<u></u>_________________<br>
mpich-discuss mailing list <a href="mailto:mpich-discuss@mcs.anl.gov" target="_blank">mpich-discuss@mcs.anl.gov</a><br>
To manage subscription options or unsubscribe:<br>
<a href="https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss" target="_blank">https://lists.mcs.anl.gov/<u></u>mailman/listinfo/mpich-discuss</a><br>
</blockquote>
______________________________<u></u>_________________<br>
mpich-discuss mailing list <a href="mailto:mpich-discuss@mcs.anl.gov" target="_blank">mpich-discuss@mcs.anl.gov</a><br>
To manage subscription options or unsubscribe:<br>
<a href="https://lists.mcs.anl.gov/mailman/listinfo/mpich-discuss" target="_blank">https://lists.mcs.anl.gov/<u></u>mailman/listinfo/mpich-discuss</a><br>
</div></div></blockquote></div><br></div>