- Cvs-hugs - Haskell.org

cvs commit: hugs98/src prelude.h static.c storage.c storage.h subst.c
by Sigbjorn Finne 22 Jan '03

22 Jan '03

sof 2003/01/22 11:15:23 PST Modified files: src prelude.h static.c storage.c storage.h subst.c Log: Stephen Milborrow's optimisations to whatIs() -- enabled by defining FAST_WHATIS to 1 (which it is now by default.) Here's the original e-mail describing the changes made (and their effect): From: "Stephen Milborrow" <milbo(a)icon.co.za> To: <hugs-users(a)haskell.org> Subject: A Faster whatIs Date: Sat, 30 Nov 2002 11:10:47 +0200 Hello everyone: I made some changes to the Hugs sources which give small runtime speed gains. The results are in the table below. The gains are minor, but I thought some people in the group might be interested. WHATIS WHATIS1 FHEAP Speed Increase (percent) gamteb 8 9 5 parser 6 8 gcfail prolog 9 9 2 queens 10 13 4 Extra BSS memory (kBytes) 24 24 0 Nbr source lines changed 50 144 41 ---Notes On The Table The table headings WHATIS, WHATIS1, and FHEAP refer to three different sets of changes. WHATIS and WHATIS1 are described below. FHEAP was an experiment that simply used a fixed size heap (an array instead of a malloc) and is impractical because it rules out use of the Hugs -h flag. The (motley) collection of test programs are from nofib. I timed runhugs for each program without command line flags, except that for queens I used -h1000000. Different programs would probably give different results: the usual caveats apply. These programs were just the nofib ones I had at hand. I compiled with MSC 5.0 Service Pack 3 using modified Nov 2002 Hugs sources. I ran the timing tests on a Windows 98 machine. ---Strategy I wanted to see what speed gains could be achieved by taking a worm's eye view of the code, with no changes to fundamental algorithms. I also wanted to make changes that would be limited to just a few places in the code -- a minimal force approach. To start off, I ran the MSC profiler. This confirmed that whatIs() is a candidate for optimization, as already noted in Mark's Gopher implementation document and in comments in storage.c. But the profiler showed a few other hotspots too. (An interesting thing to do is sort the profiler results on execution line-count, which immediately tells you which are the most executed lines in the program.) ---WHATIS Change To reduce the time spent in whatIs, I created a byte array whatCode of whatIs codes. I then created a whatIs macro which replaces the whatIs function: #define whatIs(c) (isPair(c)? \ (isTag(fst(c)) ? fst(c) : AP ) : \ whatCode[c]) Negative indexing into whatCode is prevented by the isPair. To keep the size of whatCode reasonable, I had to reduce the range of unboxed ints -- the bigger the range, the bigger the whatCode array. I settled on a range of 2048 i.e. ints between -1023 and 1024 are unboxed, all others are boxed. The extra boxing will actually slow down execution of certain Haskell programs, but as far as I can tell the majority of real Hugs programs would be unaffected. Increasing this range to, say, 10 000 would increase memory usage by 10 000 bytes -- nothing really when you consider that the memory footprint of winhugs is about 10 MBytes. This change yields the speed improvements under the WHATIS column in the table. ---WHATIS1 Change IsTag is defined as #define isTag(c) (TAGMIN<=(c) && (c)<SPECMIN) I reduced the cost of IsTag slightly by changing defines in storage.h so that only box-cell-tags are in the range 1 to 0x7f. (I shifted special cell values down to start at 0x80). I then defined variants for IsTag and friends: #define isTag1(c) (((c) & TAG_MASK) == 0) #define whatIs1(c) (isPair(c)? \ (isTag1(fst(c))? fst(c) : AP ) : \ whatCode[c]) #define isAp1(c) (isPair(c) && !isTag1(fst(c))) whatIs1 is a faster version of whatIs, with the "bug" that it will return the wrong value if fst(c) is NIL. When used in several places as a replacement for whatIs, it gains us a few speed percentage points as shown in the WHATIS1 column above. WHATIS1 sits on top of WHATIS so the gains attributable to WHATIS1 alone are the differences between the percentages in the two columns. ---Another Candidate Another candidate for this kind of optimization is the line in eval(): if (!isCfun(n) && (ar=name(n).arity)<=(sp-base)) {... This is one of the most executed lines in the entire program. If the stored arity of Cfun's was offset by a largish number, say 10000, then the above test against isCfun wouldn't be needed. This change would introduce inefficiencies elsewhere (we would have to un-offset the stored arity before using it elsewhere) but the net effect would probably be a speed gain. I shied away from this change because it would require ubiquitous (though easy) code changes. ---Final Comments Optimizing code written by Mark Jones is a challenge. It becomes a little less daunting if we change the rules by allowing ourselves to waste some memory in the pursuit of speed, and to introduce some ugliness into the code. Even so, the speed improvements I got were small. If anyone is interested, I would be happy to send the sources. They are modified Nov 2002 sources with all the changes demarcated by #define's. Six files changed. I would be interested to know what results these changes give on other machines. --- Stephen Milborrow Revision Changes Path 1.40 +7 -2 hugs98/src/prelude.h 1.134 +7 -2 hugs98/src/static.c 1.59 +38 -2 hugs98/src/storage.c 1.48 +86 -2 hugs98/src/storage.h 1.26 +6 -2 hugs98/src/subst.c

1 0

cvs commit: hugs98/libraries/Hugs ConcBase.hs
by ross＠glass.cse.ogi.edu 22 Jan '03

22 Jan '03

ross 2003/01/22 06:46:18 PST Modified files: libraries/Hugs ConcBase.hs Log: remove swapMVar and readMVar (defined in Control.Concurrent.MVar) Revision Changes Path 1.9 +1 -17 hugs98/libraries/Hugs/ConcBase.hs

1 0

cvs commit: hugs98/src builtin.c
by ross＠glass.cse.ogi.edu 21 Jan '03

21 Jan '03

ross 2003/01/21 02:41:31 PST Modified files: src builtin.c Log: delete unused declaration Revision Changes Path 1.40 +2 -3 hugs98/src/builtin.c

1 0

cvs commit: hugs98/src Makefile.in input.c static.c hugs98/src/bcc32 Makefile hugs98/src/djgpp2 Makefile hugs98/src/msc Makefile hugs98/src/winhugs Makefile
by ross＠glass.cse.ogi.edu 21 Jan '03

21 Jan '03

ross 2003/01/21 02:17:57 PST Modified files: src Makefile.in input.c static.c src/bcc32 Makefile src/djgpp2 Makefile src/msc Makefile src/winhugs Makefile Log: add #include "module.h" to input.c and static.c Revision Changes Path 1.32 +1 -1 hugs98/src/Makefile.in 1.58 +3 -2 hugs98/src/input.c 1.133 +3 -2 hugs98/src/static.c 1.8 +1 -1 hugs98/src/bcc32/Makefile 1.7 +1 -1 hugs98/src/djgpp2/Makefile 1.20 +1 -1 hugs98/src/msc/Makefile 1.10 +1 -1 hugs98/src/winhugs/Makefile

1 0

cvs commit: hugs98/src static.c
by ross＠glass.cse.ogi.edu 21 Jan '03

21 Jan '03

ross 2003/01/21 02:05:49 PST Modified files: src static.c Log: delete unused variable. Revision Changes Path 1.132 +2 -3 hugs98/src/static.c

1 0

cvs commit: hugs98/libraries/Hugs IOArray.hs
by ross＠glass.cse.ogi.edu 20 Jan '03

20 Jan '03

ross 2003/01/20 05:54:31 PST Modified files: libraries/Hugs IOArray.hs Log: deleted unused import. Revision Changes Path 1.4 +0 -1 hugs98/libraries/Hugs/IOArray.hs

1 0

cvs commit: hugs98/tests testScript hugs98/tests/static mod148.hs mod149.hs mod150.hs mod151.hs mod152.hs mod153.hs mod154.hs mod155.hs mod151.output mod150.output mod155.output mod153.output mod152.output mod156.hs
by Sigbjorn Finne 18 Jan '03

18 Jan '03

sof 2003/01/18 08:54:30 PST Modified files: tests testScript Added files: tests/static mod148.hs mod149.hs mod150.hs mod151.hs mod152.hs mod153.hs mod154.hs mod155.hs mod151.output mod150.output mod155.output mod153.output mod152.output mod156.hs Log: more module exportation tests Revision Changes Path 1.43 +9 -0 hugs98/tests/testScript

1 0

cvs commit: hugs98/tests/static mod142.output mod143.output
by Sigbjorn Finne 18 Jan '03

18 Jan '03

sof 2003/01/18 08:53:36 PST Modified files: tests/static mod142.output mod143.output Log: upd Revision Changes Path 1.2 +2 -3 hugs98/tests/static/mod142.output 1.2 +2 -3 hugs98/tests/static/mod143.output

1 0

cvs commit: hugs98/src module.c
by Sigbjorn Finne 18 Jan '03

18 Jan '03

sof 2003/01/18 08:52:09 PST Modified files: src module.c Log: Carry prev. commit to its conclusion -- never report ambiguity errors when processing 'module <Foo>' forms in export lists; only do this when entities appear directly. (Errors due to duplicate exports are still reported though.) Revision Changes Path 1.3 +3 -7 hugs98/src/module.c

1 0

cvs commit: hugs98/libraries/Hugs Weak.hs
by ross＠glass.cse.ogi.edu 17 Jan '03

17 Jan '03

ross 2003/01/17 06:53:33 PST Modified files: libraries/Hugs Weak.hs Log: delete some stuff that's in System.Mem.Weak Revision Changes Path 1.2 +1 -16 hugs98/libraries/Hugs/Weak.hs

1 0

cvs commit: hugs98/src iomonad.c
by ross＠glass.cse.ogi.edu 16 Jan '03

16 Jan '03

ross 2003/01/16 10:43:17 PST Modified files: src iomonad.c Log: fix arg number for touchForeignPtr Revision Changes Path 1.38 +3 -3 hugs98/src/iomonad.c

1 0

proposed major change: Hugs and Exceptions
by Ross Paterson 16 Jan '03

16 Jan '03

Hugs has data Exception = IOException IOError | HugsException HugsException where IOErrors are passed to a failure continuation and printed using Show. HugsExceptions are redexes that caused an error, triggering a longjmp. If not caught, they get printed by the internal printer, e.g. {primDivInt 1 0}. In my tree, I've switched over to the full Exception type (as in Control.Exception), with all Exceptions triggering a longjmp, caught by runAndShowError and printed using Show, e.g. "divide by zero" -- less information, but more friendly. Almost all of Control.Exception (all but throwTo/throwDynTo) then works with Hugs. The failure continuation in the IO type is then unused, though it's still there for now. A disadvantage is that the Exception types and Show instances move into Hugs.Prelude. It touches quite a few files, but it passes the test suite (appropriately modified), and I'm thinking of committing it. Anyone want to advise caution?

1 0

cvs commit: hugs98/libraries/Hugs ConcBase.hs
by ross＠glass.cse.ogi.edu 16 Jan '03

16 Jan '03

ross 2003/01/16 02:53:23 PST Modified files: libraries/Hugs ConcBase.hs Log: cosmetic changes. Revision Changes Path 1.8 +61 -67 hugs98/libraries/Hugs/ConcBase.hs

1 0

cvs commit: hugs98/tests/libs prodcon.hs
by ross＠glass.cse.ogi.edu 16 Jan '03

16 Jan '03

ross 2003/01/16 02:51:47 PST Modified files: tests/libs prodcon.hs Log: fix my earlier mistake: the replacement of CVar should be MVar. Revision Changes Path 1.4 +3 -3 hugs98/tests/libs/prodcon.hs

1 0

cvs commit: hugs98/libraries/Hugs ConcBase.hs
by ross＠glass.cse.ogi.edu 14 Jan '03

14 Jan '03

ross 2003/01/14 03:03:03 PST Modified files: libraries/Hugs ConcBase.hs Log: delete unused import. Revision Changes Path 1.7 +0 -1 hugs98/libraries/Hugs/ConcBase.hs

1 0

cvs commit: hugs98/src module.c
by Sigbjorn Finne 13 Jan '03

13 Jan '03

sof 2003/01/12 10:33:40 PST Modified files: src module.c Log: - moderate the reporting of name clashes in export lists. Note: Support for H98's interpretation of "module M where" isn't included in this commit - I don't see a good way of implementing it properly given how names are handled internally. Revision Changes Path 1.2 +22 -19 hugs98/src/module.c

2 1

cvs commit: hugs98/tests testScript
by Sigbjorn Finne 13 Jan '03

13 Jan '03

sof 2003/01/12 10:24:07 PST Modified files: tests testScript Log: add +X to STDOPTS; set HUGSDIR to something sane Revision Changes Path 1.42 +2 -2 hugs98/tests/testScript

2 1

cvs commit: hugs98/tests/static mod110.output
by Sigbjorn Finne 13 Jan '03

13 Jan '03

sof 2003/01/12 10:11:41 PST Modified files: tests/static mod110.output Log: updated Revision Changes Path 1.5 +1 -1 hugs98/tests/static/mod110.output

2 1

cvs commit: hugs98/src machdep.c
by Sigbjorn Finne 12 Jan '03

12 Jan '03

sof 2003/01/12 10:34:53 PST Modified files: src machdep.c Log: hugsdir(): under Win32, look for and honour HUGSDIR env variable before consulting GetModuleFileName() Revision Changes Path 1.74 +17 -11 hugs98/src/machdep.c

1 0

cvs commit: hugs98/libraries/Hugs ForeignPtr.hs IO.hs Int.hs Ptr.hs
by ross＠glass.cse.ogi.edu 09 Jan '03

09 Jan '03

ross 2003/01/09 03:01:39 PST Modified files: libraries/Hugs ForeignPtr.hs IO.hs Int.hs Ptr.hs Log: make imports more specific. Revision Changes Path 1.2 +2 -2 hugs98/libraries/Hugs/ForeignPtr.hs 1.6 +2 -1 hugs98/libraries/Hugs/IO.hs 1.9 +5 -1 hugs98/libraries/Hugs/Int.hs 1.2 +1 -1 hugs98/libraries/Hugs/Ptr.hs

1 0

cvs commit: hugs98/libraries/Hugs Storable.hs
by ross＠glass.cse.ogi.edu 09 Jan '03

09 Jan '03

ross 2003/01/09 02:59:52 PST Modified files: libraries/Hugs Storable.hs Log: cosmetic export tweak. Revision Changes Path 1.5 +1 -1 hugs98/libraries/Hugs/Storable.hs

1 0

cvs commit: hugs98/src/unix MachDeps.h
by ross＠glass.cse.ogi.edu 07 Jan '03

07 Jan '03

ross 2003/01/07 08:00:47 PST Modified files: src/unix MachDeps.h Log: fix size (and alignment) of HsDouble (bug reported by Ron Legere). HsDouble is double (cf HsFFI.h) regardless of USE_DOUBLE_PRECISION. Revision Changes Path 1.2 +0 -8 hugs98/src/unix/MachDeps.h

1 0

cvs commit: hugs98/libraries/Hugs IOExts.hs Memo.hs Trace.hs hugs98/src builtin.c hugs98/src/unix convert_libraries
by ross＠glass.cse.ogi.edu 06 Jan '03

06 Jan '03

ross 2003/01/06 06:31:36 PST Modified files: libraries/Hugs IOExts.hs Memo.hs src builtin.c src/unix convert_libraries Removed files: libraries/Hugs Trace.hs Log: Switch to the libraries version of trace. Note that that version adds a newline after the string. Revision Changes Path 1.7 +0 -4 hugs98/libraries/Hugs/IOExts.hs 1.2 +2 -1 hugs98/libraries/Hugs/Memo.hs 1.39 +2 -15 hugs98/src/builtin.c 1.48 +1 -1 hugs98/src/unix/convert_libraries

1 0

spurious ambiguous exports
by Ross Paterson 06 Jan '03

06 Jan '03

The current CVS version fails on tests/static/mod33.hs: module M where id x = x and also module M(module M) where id x = x In both cases the export should be unambiguous. (The latter was a late clarification of the Report.)

1 1

cvs commit: hugs98/tests/libs libs.hs
by ross＠glass.cse.ogi.edu 05 Jan '03

05 Jan '03

ross 2003/01/05 08:06:35 PST Modified files: tests/libs libs.hs Log: Merge is gone. Revision Changes Path 1.4 +0 -1 hugs98/tests/libs/libs.hs

1 0