diff options
| author | Yuri de Wit <admin@rodlogic.net> | 2015-02-09 13:41:02 -0600 |
|---|---|---|
| committer | Austin Seipp <austin@well-typed.com> | 2015-02-09 21:07:25 -0600 |
| commit | 83efb985d632d3a351f69cb6ce9dc5232127d545 (patch) | |
| tree | cbc524c15a509ea299e1aebf78e3092757ecb4f5 /compiler/nativeGen | |
| parent | a741e69a230eb6d6e3373ad1fbe53c73b5f95077 (diff) | |
| download | haskell-83efb985d632d3a351f69cb6ce9dc5232127d545.tar.gz | |
Replace .lhs with .hs in compiler comments
Summary: It looks like during .lhs -> .hs switch the comments were not updated. So doing exactly that.
Reviewers: austin, jstolarek, hvr, goldfire
Reviewed By: austin, jstolarek
Subscribers: thomie, goldfire
Differential Revision: https://phabricator.haskell.org/D621
GHC Trac Issues: #9986
Diffstat (limited to 'compiler/nativeGen')
| -rw-r--r-- | compiler/nativeGen/PIC.hs | 2 | ||||
| -rw-r--r-- | compiler/nativeGen/PPC/CodeGen.hs | 2 | ||||
| -rw-r--r-- | compiler/nativeGen/RegAlloc/Graph/TrivColorable.hs | 4 |
3 files changed, 4 insertions, 4 deletions
diff --git a/compiler/nativeGen/PIC.hs b/compiler/nativeGen/PIC.hs index 6326a8bdaf..d4739cb7d9 100644 --- a/compiler/nativeGen/PIC.hs +++ b/compiler/nativeGen/PIC.hs @@ -745,7 +745,7 @@ initializePicBase_ppc _ _ _ _ -- call 1f -- 1: popl %picReg -- addl __GLOBAL_OFFSET_TABLE__+.-1b, %picReg --- (See PprMach.lhs) +-- (See PprMach.hs) initializePicBase_x86 :: Arch -> OS -> Reg diff --git a/compiler/nativeGen/PPC/CodeGen.hs b/compiler/nativeGen/PPC/CodeGen.hs index 63a7c18c7d..e547ab6c95 100644 --- a/compiler/nativeGen/PPC/CodeGen.hs +++ b/compiler/nativeGen/PPC/CodeGen.hs @@ -890,7 +890,7 @@ genCCall' But there are substantial differences: * The number of registers used for parameter passing and the exact set of - nonvolatile registers differs (see MachRegs.lhs). + nonvolatile registers differs (see MachRegs.hs). * On Darwin, stack space is always reserved for parameters, even if they are passed in registers. The called routine may choose to save parameters from registers to the corresponding space on the stack. diff --git a/compiler/nativeGen/RegAlloc/Graph/TrivColorable.hs b/compiler/nativeGen/RegAlloc/Graph/TrivColorable.hs index eba2e43149..756980aefe 100644 --- a/compiler/nativeGen/RegAlloc/Graph/TrivColorable.hs +++ b/compiler/nativeGen/RegAlloc/Graph/TrivColorable.hs @@ -70,7 +70,7 @@ accSqueeze count maxCount squeeze ufm = acc count (eltsUFM ufm) ~~~~~~~~~~~~~~~~~~~~ BL 2007/09 Doing a nice fold over the UniqSet makes trivColorable use -32% of total compile time and 42% of total alloc when compiling SHA1.lhs from darcs. +32% of total compile time and 42% of total alloc when compiling SHA1.hs from darcs. Therefore the UniqFM is made non-abstract and we use custom fold. MS 2010/04 @@ -78,7 +78,7 @@ When converting UniqFM to use Data.IntMap, the fold cannot use UniqFM internal representation any more. But it is imperative that the assSqueeze stops the folding if the count gets greater or equal to maxCount. We thus convert UniqFM to a (lazy) list, do the fold and stops if necessary, which was -the most efficient variant tried. Benchmark compiling 10-times SHA1.lhs follows. +the most efficient variant tried. Benchmark compiling 10-times SHA1.hs follows. (original = previous implementation, folding = fold of the whole UFM, lazyFold = the current implementation, hackFold = using internal representation of Data.IntMap) |