projects / ghc-hetmet.git / commit
? search:
summary | shortlog | log | commit | commitdiff | tree
(parent: 25cc1d1) | patch
floating-point fix for x86_64
authorSimon Marlow <simonmar@microsoft.com>
Wed, 22 Feb 2006 14:07:19 +0000 (14:07 +0000)
committerSimon Marlow <simonmar@microsoft.com>
Wed, 22 Feb 2006 14:07:19 +0000 (14:07 +0000)
commit3034a6c8cfb50e2b5af4ef57c419986039b53a94
treee823e87d708c7e84ab6e0abc9af366100a61a6bctree | snapshot
parent25cc1d1f2d9bf9a83725e49884bd83ab35000e3ecommit | diff
floating-point fix for x86_64

For 32-bit float primtives like sinFloat#, we currently call the
double versions of the C library functions (sin(), cos() etc.).  It
seems more correct to call the float versions (sinf(), cosf() etc.).
This makes a difference on x86_64, I'm not entirely sure why, but this
way at least generates more consistent results and avoids extra
promotion/demotion instructions when calling these primitives.
ghc/compiler/cmm/PprC.hs diff | blob | history
ghc/compiler/nativeGen/MachCodeGen.hs diff | blob | history
Glasgow Haskell Compiler with heterogeneous metaprogramming (generalized arrows and modal types)