[san-tech][03126] NVIDIA GPUDirect v2.0 (CUDA 4.0) 正式リリース (2011/05/26)

Date: Fri, 27 May 2011 16:59:52 +0900
--------------------------------------------------
GPUDirect v2.0、Unified Virtual Addressing (UVA) 等 Faster Multi-GPU
Programmingをサポートする機能を含む、CUDA 4.0が正式リリースされました：
※CUDA 4.0は他にも盛りだくさんです

"CUDA Toolkit 4.0 and Parallel Nsight 2.0 Now Production Release"
 May 26 2011
  http://developer.nvidia.com/content/cuda-toolkit-40-and-parallel-nsight-20-now-production-release

  "CUDA 4.0- Features include Unified Virtual Addressing (UVA),
   Thrust C++ Template Performance Primitives Libraries and
   GPUDirect 2.0 GPU peer-to-peer communication technology."



CUDA Toolkit 4.0
  http://developer.nvidia.com/cuda-toolkit-40

Faster Multi-GPU Programming
*)Unified Virtual Addressing
*)GPUDirect v2.0 support for Peer-to-Peer Communication

NVIDIA GPUDirect
  http://developer.nvidia.com/gpudirect
  "First released in June 2010, GPUDirect v1.0 is already supported by
   InfiniBand solutions such as those available from Mellanox and
   QLogic, and other vendors are adding support for GPUDirect in their
   hardware and software products now."
.....
  "GPUDirect v2.0 adds support for peer-to-peer (P2P) communication
   between GPUs over PCIe in the same system, and lays the foundation
   for P2P communication between GPUs and other devices in a future
   release."

How Do I Get GPUDirect?
  "GPUDirect v2.0 Peer-to-Peer Communication is supported natively by
   the CUDA Driver.  All you need is CUDA Toolkit v4.0 with R270 drivers
   (or later) and a system with two or more Fermi-architecture (or later)
   GPUs on the same PCIe bus."

GPUDirect v2.0、Unified Virtual Addressing (UVA): Single Address Space
については、以下の紹介ビデオをお勧めします：
"CUDA 4.0 Overview" (約 44分)
  http://developer.download.nvidia.com/CUDA/training/cuda4_0_feature_intro.mp4
※第 2セッションの Faster Multi-GPU Programming (13～22分位)

GPU Computing Webinars
  http://developer.nvidia.com/gpu-computing-webinars