Arkady, Actually, we encountered these problems running back to back multi-threaded transaction tests on Windows while trying to get to uDAPL 100% common code across Windows and Linux. Difference in schedulers bought out some of these problems along with extended runs with the server running continuously. We also run dapltest with multiple clients, one server, so we are pushing this pretty hard.
-arlin ________________________________ From: arkady kanevsky [mailto:[email protected]] Sent: Sunday, April 19, 2009 6:24 AM To: Davis, Arlin R Cc: OpenIB Subject: Re: [ofa-general] [PATCH] uDAPL v2: dapltest, adjust next port number for number of threads to avoid duplication Arlin, did we encounter this issue at interop? I do not recall seeing it. For IB, does Wait period between assigning the same QP# handles it? Thanks, Arkady On Sun, Apr 19, 2009 at 4:06 AM, Davis, Arlin R <[email protected]<mailto:[email protected]>> wrote: To avoid duplicating port numbers between different tests, the next port number to use must increment based on the number of endpoints per thread * the number of threads. Signed-off-by: Sean Hefty <[email protected]<mailto:[email protected]>> --- test/dapltest/test/dapl_server.c | 3 +++ 1 files changed, 3 insertions(+), 0 deletions(-) diff --git a/test/dapltest/test/dapl_server.c b/test/dapltest/test/dapl_server.c index d589e7b..4d386fe 100644 --- a/test/dapltest/test/dapl_server.c +++ b/test/dapltest/test/dapl_server.c @@ -480,6 +480,9 @@ DT_cs_Server (Params_t * params_ptr) case TRANSACTION_TEST: { /* create a thread to handle this pt_ptr; */ + ps_ptr->NextPortNumber += + (pt_ptr->Params.u.Transaction_Cmd.eps_per_thread - 1) * + pt_ptr->Client_Info.total_threads; DT_Tdep_PT_Debug (1,(phead,"%s: Creating Transaction Test Thread\n", module)); pt_ptr->thread = DT_Thread_Create (pt_ptr, DT_Transaction_Test_Server, -- 1.5.2.5 _______________________________________________ general mailing list [email protected]<mailto:[email protected]> http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general -- Cheers, Arkady Kanevsky
_______________________________________________ general mailing list [email protected] http://lists.openfabrics.org/cgi-bin/mailman/listinfo/general To unsubscribe, please visit http://openib.org/mailman/listinfo/openib-general
