I see. By the way, when I tried test1 and test3 on my machine, the latter is only about 10% faster than the former.
It should be easy to have a version in ATS that beats Haskell: You first produce a non-linear thePrimes. Then you produce a linear thePrimes2 (where isPrime uses thePrimes). In this way, the memory foot print should be very small. And I bet the running time is much faster. Try 2^28 or 2^30 :) On Saturday, July 9, 2022 at 10:37:03 PM UTC-4 ice.r...@gmail.com wrote: > I will check ATS3 version, meanwhile GHC 8.10 produces much more optimised > code: > > ``` > [nix-shell:/data/devel/mobt2/haskell]$ ghc --version > The Glorious Glasgow Haskell Compilation System, version 8.10.7 > [nix-shell:/data/devel/mobt2/haskell]$ ghc -O2 --make app/Main.hs -o > haskell > [1 of 1] Compiling Main ( app/Main.hs, app/Main.o ) > Linking haskell ... > [nix-shell:/data/devel/mobt2/haskell]$ $(which time) ./haskell > 1077871 > 1.04user 0.00system 0:01.05elapsed 100%CPU (0avgtext+0avgdata > 72140maxresident)k > 0inputs+0outputs (0major+17298minor)pagefaults 0swaps > ``` > in comparison to ATS2 versions from repo above: > > ``` > [nix-shell:/data/devel/mobt2/ats2/src]$ make > patscc -O2 -DATS_MEMALLOC_LIBC -I"../libs/" -o test3 TEST/test3.dats > /run/current-system/sw/bin/time ./test3 > 1077871 > 1.11user 0.01system 0:01.12elapsed 100%CPU (0avgtext+0avgdata > 102456maxresident)k > 0inputs+0outputs (0major+25327minor)pagefaults 0swaps > patscc -O2 -DATS_MEMALLOC_LIBC -I"../libs/" -o test1 TEST/test1.dats > /run/current-system/sw/bin/time ./test1 > 1077871 > 2.11user 0.03system 0:02.14elapsed 99%CPU (0avgtext+0avgdata > 203448maxresident)k > 0inputs+0outputs (0major+50589minor)pagefaults 0swaps > ``` > ie, non-linear version (test1) is twice slower and using 2x memory of > test3 (which converts stream into stream_vt). But still, haskell version is > faster and using less memory :) > > вс, 10 июл. 2022 г. в 02:21, gmhwxi <gmh...@gmail.com>: > >> Here is a version I wrote in ATS3: >> >> >> https://github.com/githwxi/ATS-Xanadu/blob/master/xatslib/libcats/TEST/test02_isPrime.dats >> >> Currently, I can only compile the ATS3 code to JS. The generated JS code >> runs about 10 times slower >> than the C code generated from compiling a comparable ATS2 implementation: >> >> |thePrimes2| = 1077871 >> >> real 0m23.060s >> user 0m23.380s >> sys 0m0.188s >> On Saturday, July 9, 2022 at 5:33:43 PM UTC-4 gmhwxi wrote: >> >>> I took a look. Your ATS code looks very fine to me. >>> >>> I did some profiling on my own. In my trials, your ATS code is actually >>> a lot faster than >>> the Haskell code you posted. Note that my ghc version is very old: GHC >>> 7.10.3 >>> >>> ###### ATS ###### >>> >>> (* Your code using non-linear stream *) >>> hwxi@hongwei-t440p:/tmp$ patscc -O2 -o primes -DATS_MEMALLOC_LIBC >>> primes.dats >>> hwxi@hongwei-t440p:/tmp$ time ./primes >>> 1077871 >>> >>> real 0m3.118s >>> user 0m3.064s >>> sys 0m0.048s >>> >>> ###### Haskell ###### >>> (* Your haskell version *) >>> (* >>> Glasgow Haskell Compiler, Version 7.10.3, stage 2 booted by GHC version >>> 7.10.3 >>> *) >>> >>> hwxi@hongwei-t440p:/tmp$ ghc -O2 primes.hs >>> Linking primes ... >>> hwxi@hongwei-t440p:/tmp$ time ./primes >>> 1077871 >>> >>> real 0m8.195s >>> user 0m8.152s >>> sys 0m0.040s >>> >>> ################ >>> >>> ###### ATS ###### >>> (*My version using linear stream that is based on yours*) >>> >>> hwxi@hongwei-t440p:/tmp$ patscc -O2 -o primes2 -DATS_MEMALLOC_LIBC >>> primes2.dats >>> hwxi@hongwei-t440p:/tmp$ time ./primes2 >>> 1077871 >>> >>> real 0m2.120s >>> user 0m2.092s >>> sys 0m0.000s >>> >>> On Saturday, July 9, 2022 at 12:50:11 PM UTC-4 ice.r...@gmail.com wrote: >>> >>>> My initial motivation was this Haskell source code: >>>> https://github.com/dambaev/mobt2/blob/master/haskell/app/Main.hs which >>>> is using lazy list (of course) and recursive binding and I decided to >>>> check >>>> if it will be possible to get something similar >>>> >>>> ATS version using non-linear stream is here: >>>> https://github.com/dambaev/mobt2/blob/master/ats2/src/TEST/test1.dats >>>> , but it takes to much memory as `stream_take_while` duplicates data, as I >>>> have got, that datatype constructor can't be unfolded with `@stream_cons` >>>> pattern match >>>> >>>> There is another version, that generates primes with non-linear stream >>>> and then, converting it to linear stream >>>> https://github.com/dambaev/mobt2/blob/master/ats2/src/TEST/test3.dats >>>> . This is the closest version to haskell's one. but still is using more >>>> space and as twice as slow as Haskell, so I had started to think of how to >>>> eliminate intermediate data structure. >>>> >>>> So, not a production issue, hehe, just found an interesting topic to >>>> dig in :) >>>> >>>> сб, 9 июл. 2022 г. в 11:11, Hongwei Xi <gmh...@gmail.com>: >>>> >>>>> By looking at your first version, my simple answer is that a linear >>>>> stream cannot be used >>>>> in this way. The second version is possible but I am not sure what you >>>>> wanted to do exactly. >>>>> If you show me how to use a non-linear stream to do it, then I could >>>>> probably say more. >>>>> >>>>> On Sat, Jul 9, 2022 at 5:26 AM Dambaev Alexander <ice.r...@gmail.com> >>>>> wrote: >>>>> >>>>>> Hi, >>>>>> >>>>>> I had tried to implement function of type: >>>>>> >>>>>> ``` >>>>>> fun >>>>>> {a:t0p} >>>>>> isPrime >>>>>> ( xs: !stream_vt( int) >>>>>> , x: int >>>>>> ):<cloptr1> >>>>>> bool >>>>>> ``` >>>>>> >>>>>> Ie, I want to force evaluation of some portion of a stream, but I >>>>>> need to preserve it for a later use. >>>>>> >>>>>> I had tried to make a similar version: >>>>>> ``` >>>>>> fun >>>>>> {a:t0p} >>>>>> isPrime >>>>>> ( xs: stream_vt( int) >>>>>> , x: int >>>>>> ):<cloptr1> >>>>>> ( stream_vt( int), bool) >>>>>> ``` >>>>>> >>>>>> but failed as well, so I decided to ask for a direction if someone >>>>>> had tried to do similar stuff >>>>>> >>>>>> -- >>>>>> You received this message because you are subscribed to the Google >>>>>> Groups "ats-lang-users" group. >>>>>> To unsubscribe from this group and stop receiving emails from it, >>>>>> send an email to ats-lang-user...@googlegroups.com. >>>>>> To view this discussion on the web visit >>>>>> https://groups.google.com/d/msgid/ats-lang-users/CAHjn2KwFq7JH%2BiZE7bWCJ_L7oZ38K-kmGBFii7DZsdxWLDsGmg%40mail.gmail.com >>>>>> >>>>>> <https://groups.google.com/d/msgid/ats-lang-users/CAHjn2KwFq7JH%2BiZE7bWCJ_L7oZ38K-kmGBFii7DZsdxWLDsGmg%40mail.gmail.com?utm_medium=email&utm_source=footer> >>>>>> . >>>>>> >>>>> -- >>>>> You received this message because you are subscribed to the Google >>>>> Groups "ats-lang-users" group. >>>>> To unsubscribe from this group and stop receiving emails from it, send >>>>> an email to ats-lang-user...@googlegroups.com. >>>>> >>>> To view this discussion on the web visit >>>>> https://groups.google.com/d/msgid/ats-lang-users/CAPPSPLqp62MaoG8hugJ8h2mUt%2BsSAJ2eu6uRuJ%3D5nMOc4EbcfQ%40mail.gmail.com >>>>> >>>>> <https://groups.google.com/d/msgid/ats-lang-users/CAPPSPLqp62MaoG8hugJ8h2mUt%2BsSAJ2eu6uRuJ%3D5nMOc4EbcfQ%40mail.gmail.com?utm_medium=email&utm_source=footer> >>>>> . >>>>> >>>> -- >> You received this message because you are subscribed to the Google Groups >> "ats-lang-users" group. >> To unsubscribe from this group and stop receiving emails from it, send an >> email to ats-lang-user...@googlegroups.com. >> > To view this discussion on the web visit >> https://groups.google.com/d/msgid/ats-lang-users/d87d273b-c937-40b8-ae6a-8846a9fbb801n%40googlegroups.com >> >> <https://groups.google.com/d/msgid/ats-lang-users/d87d273b-c937-40b8-ae6a-8846a9fbb801n%40googlegroups.com?utm_medium=email&utm_source=footer> >> . >> > -- You received this message because you are subscribed to the Google Groups "ats-lang-users" group. To unsubscribe from this group and stop receiving emails from it, send an email to ats-lang-users+unsubscr...@googlegroups.com. To view this discussion on the web visit https://groups.google.com/d/msgid/ats-lang-users/69a5b6f5-69e9-48dd-9b1d-aa164f9b0566n%40googlegroups.com.
