#5598: Function quotRem is inefficient
---------------------------------+------------------------------------------
Reporter: boris | Owner:
Type: task | Status: new
Priority: normal | Component: Compiler
Version: 7.0.3 | Keywords: division, performance, primop
Testcase: | Blockedby:
Os: Unknown/Multiple | Blocking:
Architecture: x86 | Failure: Runtime performance bug
---------------------------------+------------------------------------------
Comment(by boris):
Correction: two divs exist only for NCG. LLVM optimiser replaces div by
smarter code based on shifts, but this code is still duplicated for quot
and rem.
Division is duplicated because in the LLVM assembly the dividend is loaded
twice and optimiser fails to merge divisions.
{{{
134 %ln1ot = load i64* %ls1ln
135 %ln1ou = srem i64 %ln1ot, 123
136 store i64 %ln1ou, i64* %ls1lW
137 %ln1ov = load i64* %ls1ln
138 %ln1ow = sdiv i64 %ln1ov, 123
}}}
When I replaced %ln1ov with %ln1ot, the native assembly code performed
division only once.
--
Ticket URL: <http://hackage.haskell.org/trac/ghc/ticket/5598#comment:1>
GHC <http://www.haskell.org/ghc/>
The Glasgow Haskell Compiler
_______________________________________________
Glasgow-haskell-bugs mailing list
[email protected]
http://www.haskell.org/mailman/listinfo/glasgow-haskell-bugs
