From 07b32c3fb3c1a271bcc6e0a33ae6e134d10dd0b3 Mon Sep 17 00:00:00 2001 From: Tor Aamodt Date: Thu, 20 Jun 2013 11:46:19 -0800 Subject: cleaning up descriptions in CHANGES (new feature vs bug fix) [git-p4: depot-paths = "//depot/gpgpu_sim_research/fermi/distribution/": change = 16480] --- CHANGES | 29 ++++++++++++++--------------- 1 file changed, 14 insertions(+), 15 deletions(-) diff --git a/CHANGES b/CHANGES index a7ea699..cf651c9 100644 --- a/CHANGES +++ b/CHANGES @@ -6,6 +6,19 @@ Version 3.2.1+edits (development branch) versus 3.2.1 the tag_array to the cache access functions. Updated the cache_statistic printing - providing a more meaningful breakdown. Cleaned up power_stats.cc/h to reflect the changes in the cache statistics. +- Adding support for cudaFuncSetCacheConfig API, that allows changing the + L1 Cache and Shared Memory configurations across kernels. The support + enables the user to specify two more configurations (Preferred L1) or + (Preferred Shared Memory) besides the default config. If the + cudaFuncSetCacheConfig API is used to set the cache configuration + of a specific kernel to either of these configuration (cudaFuncCachePreferShared, + cudaFuncCachePreferL1), the simulator will change the cache configuration + at kernel launch accordingly, if there is no alternative configurations + provided to the simulator it will use the default configurations with a + warning message display +- Removed the use of gcc in compilation and the need for a bunch of unnecessary + 'extern "c"' statements. +- Removed print at kernel finish that dumped a warp issue histogram. - Bug Fixes: - Fixed the flit count sent to GPUWattch for atomic operations. - Fix for Bug 51 - Updated the function declaration of @@ -16,29 +29,15 @@ Version 3.2.1+edits (development branch) versus 3.2.1 - Fix for Bug 53 - Returning the CL_DEVICE_TYPE property in proper size from clGetDeviceInfo(...). - Fix for Bug 54 - Added code to automatically determine workgroup size. - - Adding support for cudaFuncSetCacheConfig API, that allows changing the - L1 Cache and Shared Memory configurations across kernels. The support - enable the user to specify two more configurations (Preferred L1) or - (Preferred Shared Memory) besides the default config. If the - cudaFuncSetCacheConfig API is used to set the cache configuration - of a specific kernel to either of these configuration (cudaFuncCachePreferShared, - cudaFuncCachePreferL1), the simulator will change the cache configuration - at kernel launch accordingly, if there is no alternative configurations - provided to the simulator it will use the default configurations with a - warning message display - - Removed a spamming print at kernel finish that dumped a warp issuing - histogram. - Fixing OpenCL functional simulation. Now the functional simulation routine is called instead of mistakenly printing that the functional simulation is not implemented. - - Removed the use of gcc in compilation and the need for a bunch of unnecessary - 'extern "c"' statements. - Fixed a mismatched new[] delete[] - 'Constant dynamic' power not included in average/min/max power values. - Average/min/max per-kernel powers not being reset at kernel boundaries causing incorrect per-kernel values. - Fixed a dependency error in the src/cuda-sim Makefile. - - Fixing a source of non-determinism in GPGPU-Sim (Bug 147). + - Fixing a source of non-determinism in GPGPU-Sim (Bug 147). Version 3.2.1 versus 3.2.0 - Added kernel name and launch uids to performance statistics log. -- cgit v1.3