Performance/Floating point
From HaskellWiki
Isaac Dupree (Talk | contribs) m (Performance:Arrays -> new name Performance/Arrays) |
|||
Line 5: | Line 5: | ||
One time when <tt>Float</tt> might be a good idea is if you have a ''lot'' of them, say a giant array of <tt>Float</tt>s. An unboxed array of <tt>Float</tt> (see [[Performance/Arrays]]) takes up half the space in the heap compared to an unboxed array of <tt>Double</tt>. However, boxed <tt>Float</tt>s will only take up less space than boxed <tt>Double</tt>s if you are on a 32-bit machine (on a 64-bit machine, a <tt>Float</tt> still takes up 64 bits). | One time when <tt>Float</tt> might be a good idea is if you have a ''lot'' of them, say a giant array of <tt>Float</tt>s. An unboxed array of <tt>Float</tt> (see [[Performance/Arrays]]) takes up half the space in the heap compared to an unboxed array of <tt>Double</tt>. However, boxed <tt>Float</tt>s will only take up less space than boxed <tt>Double</tt>s if you are on a 32-bit machine (on a 64-bit machine, a <tt>Float</tt> still takes up 64 bits). | ||
+ | |||
+ | The speed claims may not be true due to Doubles not necessarily being | ||
+ | aligned as the machine wishes. We could do with some benchmarking on various platforms to see what's what. | ||
== GHC-specific advice == | == GHC-specific advice == | ||
On x86 (and other platforms with GHC prior to version 6.4.2), use the <tt>-fexcess-precision</tt> flag to improve performance of floating-point intensive code (up to 2x speedups have been seen). This will keep more intermediates in registers instead of memory, at the expense of occasional differences in results due to unpredictable rounding. See the [http://www.haskell.org/ghc/docs/latest/html/users_guide/options-optimise.html#options-f GHC documentation] for more details. Switching on GCCs ''-ffast-math'' and ''-O3'' can also help (use <tt>-optc-ffast-math</tt> and <tt>-optc-O3</tt>). | On x86 (and other platforms with GHC prior to version 6.4.2), use the <tt>-fexcess-precision</tt> flag to improve performance of floating-point intensive code (up to 2x speedups have been seen). This will keep more intermediates in registers instead of memory, at the expense of occasional differences in results due to unpredictable rounding. See the [http://www.haskell.org/ghc/docs/latest/html/users_guide/options-optimise.html#options-f GHC documentation] for more details. Switching on GCCs ''-ffast-math'' and ''-O3'' can also help (use <tt>-optc-ffast-math</tt> and <tt>-optc-O3</tt>). | ||
+ | |||
+ | Where available, the <tt>-optc-march=pentium4 -optc-mfpmath=sse</tt> flags may also help. | ||
+ | |||
+ | Not that the <tt>-fexcess-precision</tt> flag may make programs behave oddly, | ||
+ | e.g. after falling an <tt>if x < 0</tt> branch you may find that <tt>x</tt> is now not less than zero, as it has been written out to memory and thus some precision lost in the mean time. |
Revision as of 21:15, 31 October 2006
Haskell Performance Resource
Constructs: Techniques: |
1 Don't use Float
Floats (probably 32-bits) are almost always a bad idea, unless you Really Know What You Are Doing. Use Doubles. There's rarely a speed disadvantage—modern machines will use the same floating-point unit for both. With Doubles, you are much less likely to hang yourself with numerical errors.
One time when Float might be a good idea is if you have a lot of them, say a giant array of Floats. An unboxed array of Float (see Performance/Arrays) takes up half the space in the heap compared to an unboxed array of Double. However, boxed Floats will only take up less space than boxed Doubles if you are on a 32-bit machine (on a 64-bit machine, a Float still takes up 64 bits).
The speed claims may not be true due to Doubles not necessarily being aligned as the machine wishes. We could do with some benchmarking on various platforms to see what's what.
2 GHC-specific advice
On x86 (and other platforms with GHC prior to version 6.4.2), use the -fexcess-precision flag to improve performance of floating-point intensive code (up to 2x speedups have been seen). This will keep more intermediates in registers instead of memory, at the expense of occasional differences in results due to unpredictable rounding. See the GHC documentation for more details. Switching on GCCs -ffast-math and -O3 can also help (use -optc-ffast-math and -optc-O3).
Where available, the -optc-march=pentium4 -optc-mfpmath=sse flags may also help.
Not that the -fexcess-precision flag may make programs behave oddly, e.g. after falling an if x < 0 branch you may find that x is now not less than zero, as it has been written out to memory and thus some precision lost in the mean time.