From 7b6c2f6dad0bc4c892788e495c75f169c6917bd5 Mon Sep 17 00:00:00 2001 From: Wilson Fung Date: Wed, 7 Nov 2012 05:40:42 -0800 Subject: nvopencl_wrapper now works on machine with both AMD and NVIDIA GPUs. It will automatically compile the OpenCL kernels on the NVIDIA platform. [git-p4: depot-paths = "//depot/gpgpu_sim_research/fermi/distribution/": change = 14558] --- CHANGES | 5 ++++- 1 file changed, 4 insertions(+), 1 deletion(-) (limited to 'CHANGES') diff --git a/CHANGES b/CHANGES index 8119f29..d46de2b 100644 --- a/CHANGES +++ b/CHANGES @@ -25,7 +25,10 @@ Version 3.1.1+edits (development branch) versus 3.1.1 '=' and ';'. This allow options with a long string of sub-options (e.g. DRAM timing) to have a more friendly format. - Changed the DRAM timing option to a more reader-friendly format. -- Add support for the PTX v3 codes generated by new OpenCL drivers. +- Added support for the PTX version 3.0 codes generated by NVIDIA OpenCL + drivers newer than version 256.40. +- nvopencl_wrapper now works on machine with both AMD and NVIDIA GPUs. It will + automatically compile the OpenCL Kernels via the NVIDIA platform. - Bug Fixes: - Fixed a compile error that happens with newer gcc/g++ versions (4.7.1) - Fixed a bug with the association between cuobjdump output and cubin -- cgit v1.3