Date: | Thu, 27 Mar 2008 10:21:11 +0100 |
---|---|
From: | "Daniel Sánchez Pedreño" <sanatox@xxxxxxxxx> |
Subject: | [Gems-users] Protocol race in MOESI_CMP_directory |
Dear list, recently I've been testing MOESI_CMP_directory protocol in gems-2.1 for 32 processors. For some benchmarks from SPLASH-2, the simulation crash due to an invalid transition in L2. This trantisition occurs with a L1_PUTO event when the current state is SLS. I`m not very experiencied in L2 cache protocols but I think that I can solve this problem just issuing a Writeback NACK to L1. Would be that correct? By the way, I didn´t touch the protocol nor the simulator, and I only find problems when simulating 32 processors. Thank you. |
[← Prev in Thread] | Current Thread | [Next in Thread→] |
---|---|---|
|
Previous by Date: | Re: [Gems-users] DB2 install, Derek Hower |
---|---|
Next by Date: | [Gems-users] Question on MOESI_CMP_token, KS Chow |
Previous by Thread: | [Gems-users] Process bind, 王佐 |
Next by Thread: | Re: [Gems-users] Protocol race in MOESI_CMP_directory, Mike Marty |
Indexes: | [Date] [Thread] |