From e480b8ac999a7132ce003f102d5d5a80a776c2f6 Mon Sep 17 00:00:00 2001 From: Wilson Fung Date: Thu, 11 Jul 2013 13:19:38 -0800 Subject: Adding option to force global memory accesses to skip L1 data cache while still caching data from local memory space. [git-p4: depot-paths = "//depot/gpgpu_sim_research/fermi/distribution/": change = 16601] --- CHANGES | 3 +++ 1 file changed, 3 insertions(+) (limited to 'CHANGES') diff --git a/CHANGES b/CHANGES index d74f655..c840208 100644 --- a/CHANGES +++ b/CHANGES @@ -29,6 +29,9 @@ Version 3.2.1+edits (development branch) versus 3.2.1 - Implemented a Static Warp Limiting Scheduler similar described in Rogers et. al. (MICRO 2012). - Some whitespace cleanup. +- Adding an option to force global memory access to skip L1 data cache, while + local memory accesses can still be cached in L1 data cache. This feature can + be used to emulate the behavior of '-Xptxas -dlcm=cg'. - Bug Fixes: - Fixed the flit count sent to GPUWattch for atomic operations. - Fix for Bug 51 - Updated the function declaration of -- cgit v1.3