Changes between Version 1 and Version 2 of SIMD/Design
- Apr 16, 2012 1:23:56 PM (2 years ago)
v1 v2 1 1 [[PageOutline]] 2 2 3 = Using SIMD instructions in GHC = 4 5 '''Goal''': improve program running times by taking advantage of CPU's SIMD vector instructions. 6 7 '''How''': by extending GHC to generate code using SIMD vector instructions and by modifying libraries as necessary. 8 9 This page describes the issues involved and a design for implementing SIMD vector support in GHC. 10 11 Related pages: 12 * Notes on the [wiki:SIMDPlan current implementation plan] 3 This is the design document for SIMD support in GHC that resulted from the October 11, 2011 meeting at GHC HQ. Please see the [wiki:SIMD top-level GHC SIMD] page for further details. 13 4 14 5 == Introduction == … … 599 590 600 591 * [http://perilsofparallel.blogspot.com/2008/09/larrabee-vs-nvidia-mimd-vs-simd.html Blog article about Larrabee and Nvidia, MIMD vs. SIMD] 601 * [wiki:S imdLlvm SIMD LLVM] A previous (LLVM-specific) iteration of this SIMD proposal. 602 * [wiki: VectorComputing VectorComputing] A previous proposal to make use of x86 SSE in GHC. 592 * [wiki:S] A previous (LLVM-specific) iteration of this SIMD proposal. 593 * [wiki:] A previous proposal to make use of x86 SSE in GHC.