Changes between Version 1 and Version 2 of SIMD

Nov 16, 2010 8:07:44 PM (4 years ago)



  • SIMD

    v1 v2  
    5050 * Not all processors provide these machine operations on packed datatypes and not all x86 processors provide all the SSE extensions.  I propose to add capability tests (`sse, sse2` etc.) to `System.Info` for the programmer to be able to determine whether to use these packed primOps.  Thus if Haskell code uses primOps which the CPU does not support, there will be a compile-time error. 
    5151 * It would be possible to add a Cmm pass that converts operations on packed datatypes to operations on a set of scalar types, however, it is cleaner and more efficient to not use packed datatypes on machines that do not provide those SIMD instructions. 
     52 * On the x86_64 architecture, there are 4 Float registers and 2 Double registers.  These actually use the `xmm` registers, so there are only two left to use as packed registers.  My current solution is to reduce the Float registers to 3 so that there are 3 `xmm` registers. 
     53 * SSE4 capable CPUs have 16 `xmm` registers, but I am unsure as to how to approach checking for these extra registers, since the registers to use are hard coded in C.