Author Topic: NVIDIA CUDA Toolkit 8.0.61 February 2017  (Read 927 times)



0 Members and 1 Guest are viewing this topic.

Stefan

  • Global Moderator
  • Hero Member
  • *****
  • Posts: 3901
    • View Profile
NVIDIA CUDA Toolkit 8.0.61 February 2017
« on: February 10, 2017, 05:24:50 PM »
https://developer.nvidia.com/cuda-downloads

Quote
New Features
CUDA Tools

CUDA Compilers.
 The CUDA compiler now supports Xcode 8.2.1.

NVRTC.
 NVRTC is no longer considered a preview feature.
CUDA Libraries

cuBLAS.
 The cuBLAS library added a new function
cublasGemmEx(
), which is
an extension of
cublas<t/>gemm()
. It allows the user to specify the algorithm,
as well as the precision of the computation and of the input and output matrices.
The function can be used to perform matrix-matrix multiplication at lower
precision.
Resolved Issues
General CUDA

CUDA Installer.
 On some SLES or openSUSE system configurations, the
NVIDIA GL library package may need to be locked before the steps for a GL-less
installation are performed. The NVIDIA GL library package can be locked with
this command:
sudo zypper addlock nvidia-glG04

Unified memory.
 On GP10x systems, applications that use
cudaMallocManaged()
 and attempt to use
cuda-gdb
 will incur random
spurious MMU faults that will take down the application.

Unified memory.
 Functions
cudaMallocHost()
 and
cudaHostRegister()
don't work correctly on multi-GPU systems with the IOMMU enabled on
Linux. The only workaround is to disable unified memory support with the
CUDA_DISABLE_UNIFIED_MEMORY=1
 environment variable.

Unified memory.
 Fixed an issue where
cuda-gdb
 or
cuda-memcheck
 would
crash when used on an application that calls
cudaMemPrefetchAsync()
.

Unified memory.
 Fixed a potential issue that can cause an application to hang
when using
cudaMemPrefetchAsync()
.