Septembre 2006 - Gambit-list - mailman.iro.umontreal.ca

Gambit-C 4.0 beta 18
by Marc Feeley 25 Sep '06

25 Sep '06

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 Gambit-C 4.0 beta 18 is now available. The source tarball is here: http://www.iro.umontreal.ca/~gambit/gambc-4.0b18.tar.gz A prebuilt installation for Microsoft Windows is available here: http://www.iro.umontreal.ca/~gambit/gambc-4.0b18-win.zip Just unzip it in C: to get C:\Gambit-C\4.0b18, and then add C:\Gambit- C\4.0b18\bin to your path. Here are the highlights of this release: - Speculative inlining has been improved. This greatly improves the performance of compiled code when no declarations are given. The "no declaration" mode closely conforms to the R5RS semantics, where all global variables including car, +, etc are mutable (which is useful for debugging). Check http://www.iro.umontreal.ca/~gambit/ bench.html for a comparison with Chicken, MzScheme and Scheme48. It shows that when these systems are put in an R5RS conformant mode, Gambit is often an order of magnitude faster. Gambit is also quite fast in other situations (when predefined procedures are bound to immutable variables, when fixnum/flonum specific operations are used, and in unsafe mode). - The benchmark suite (misc/bench.tgz) has been updated to support Bigloo, Chicken, MzScheme and Scheme48. There is a script to easily generate HTML tables showing the benchmark results. - More predefined procedures are now inlined, including some recursive functions on lists: map, for-each, assq, and memq. - Renaming of fixnum and flonum specific functions to the R6RS name (e.g. fx+ replaces ##fixnum.+, and fl+ replaces ##flonum.+). The ##fixnum... and ##flonum... versions are still available in this release but they will be phased out. In safe mode, which is the default, the fixnum specific functions now raise an exception on overflow. There are now "wrapping" versions which do not raise an exception on overflow, that is (fxwrap+ (expt 2 28) (expt 2 28)) => - -536870912 . - Added an option to open-tcp-server to listen for connection requests on a given network interface (the default is to listen on all network interfaces). - The polling interval for process ports has been shortened to improve promptness. - The speed of memory moving operations, such as substring, has been improved. - Fixed a GC bug with tables and added table-copy . Fixed a bug with char-ready? and peek-char . Sorry to disappoint, but this release does not include a module system. That's something I'm still working on and I did not want to delay this release any further. Marc -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.3 (Darwin) iD8DBQFFEH1x//V9Zc2T/v4RAtdTAJsFds8HFzpaCLb0C9+kMgeCAO2+KwCeLcQt 2/U7r2tfw2BaX7NKU9rj7ps= =NI2L -----END PGP SIGNATURE-----

5 14

Building Gambit-C on a Sharp Zaurus
by Neil Bartlett 25 Sep '06

25 Sep '06

Hi all, I am trying to build Gambit (4.0 b7) on a Sharp Zaurus C3100. This is a handheld computer running Linux (2.4.20 kernel) on an ARM processor (Intel XScale at 415 Mhz) with 128Mb RAM. It should be perfectly capable of both building and running Gambit - for example I have previously built the full OCaml stack using GCC running onboard the device. Unfortunately I am having some issues. "configure" and "make" work without any problems, although I have to avoid the "--enable-single-host" option. However "make check" immediately fails with the following error: ------------ TEST 1 (debugging support) ../gsi/gsi -f debug.scm > test1.out *** ERROR IN check, "debug.scm"@82.29 -- (Argument 1) VECTOR expected (vector->list #\u0301) make[1]: *** [test1] Error 70 make[1]: Leaving directory `/home/zaurus/build/gambc40b17/tests' Can anybody offer me enlightenment? Many thanks Neil

2 1

Mrs.Mellor Dele
by centre.centre＠virgilio.it 24 Sep '06

24 Sep '06

Mrs.Mellor Dele THE MANAGING DIRECTOR INTERNATIONAL PROMOTIONS/PRICE AWARD DEPARTMENT.IN MADRID FREE LOTTO CASINO AWARD 2006 OUR REF: CL/99/06 YOU’RE REF:CO/578/34/06. BATCH No:44-694-88-ES LUCKY No: 20/35/49/70/90. NOTICE We are delighted to notify you of the result of the free lotto casino programs held on 23.September,2006 your email address were attached to one of the ten winning Ticket numbers (16/25/22/27)- with Game (Number 45-60-71.You have therefore been approved for a lump sum pay out of €150,000.00 (One Hundred Fifty Thousand Euro) CONGRATULATIONS! All participants were selected through a computer ballot system from our sponsors databases, including over 5,000 companies and 20,000 individual E-mail addresses and names submitted by our agents drawn from Asia, Africa, Europe, North And South America, Oceania and around the world. To file for your urgent claim, please contact our affiliate agent. CASINO AWARD AGENCY C/AIPERCINES 9,28830 SAN FEMANDO DE HENARES MADRID ESPANA Tel:+34-635-112-356 E-mail:casinoagency77@aim. com MORRIS CATALDO. It is part of our security protocol to avoid double claiming and to minimize identification difficulties advise you to keep this information confidential until your claim is processed successfully and to avoid unwarranted abuse of this program by some participants to please note that you will be required to pay for the issuance of your winning Certificate and all winnings must be claimed not later than 30.September,2006.after this date all unclaimed winnings will be null and void. In Order to avoid unnecessary delays and complications remember to quote your ticket number and game numbers in all correspondence. Furthermore, should there be any change of address do inform our agents as soon as possible. Congratulations once more and thank you for being part of our promotional program.Note: Anybody under the age of 18 is automatically disqualified. Yours faithfully, Mrs.Mellor Dele. Award Centre Co-coordinador.

1 0

Gamsock darcs repo available
by Jeff Read 20 Sep '06

20 Sep '06

Hi all, In the spirit of "release early, release often" I have elected to make a Darcs repository available which will allow you to track the progress of Gamsock. You can access it by doing: darcs get http://ii-0-ii.com/parodycheck/repos/gamsock Notes: * Documentation is nonexistent but you are welcome to peruse and bang on the code. * Actual socket calls (socket, bind, connect, send, recv, etc.) tend to bear their POSIX names. (Change this?) * Sockets and sockaddrs are opaque types. Sockaddrs have constructors for each of the different address families, which convert the parameters into the appropriate sockaddr data; and accessors to convert the other way. (Should I give people access to the raw sockaddr bits, as sb-bsd-sockets for SBCL does?) * Calls which on the C side take enums can take either numeric values or symbol constants on the Scheme side. E.g., you can do (socket PF_INET SOCK_STREAM 0) or (socket 'inet 'stream 0). (Should I favor one over the other?) * Likewise, calls which on the C side take the bitwise inclusive OR of a set of flags, can take a list of symbols instead on the Scheme side. * Currently there are no protocol names specified. For an inet socket you may specify 0 as the protocol, in which case the protocol will be selected by the socket type (SOCK_STREAM, SOCK_DGRAM, etc.) * Unix-domain and IPv4 sockets (TCP and UDP) should be supported. Only TCP IPv4 sockets have been thoroughly tested. One can initiate and accept connections, send, and receive. * IP addresses (for make-inet-sockaddr) take the form of a four-element u8vector. You may also pass in a string (e.g., "127.0.0.1"). This latter option is kind of arbitrary and may disappear, or be moved to a facility to do DNS lookups when it becomes available. Port numbers are fixnums. * You can only send u8vectors of data at this time. You also get a u8vector back when you do recv. * There will be a facility (not yet implemented) that will let you wrap stream sockets at least with Gambit ports, mimicking the facility of tcp-sockets already in place in Gambit. * This has been tested on 4.0b17. YMMV on any other version (including yesterday's release). * If there is an error, an os-exception will be raised with the error code of the exception set to errno. (Change this? Make the errors meaningful, platform-agnostic?) * For the time being, Gamsock works on Linux only. Currently on BSD, Gamsock compiles but when you try to bind a socket to an address, it raises an os-exception with code EINVAL. I think it has to do with the fact that OpenBSD sockaddrs have an additional sa_len element that Linux sockaddrs lack. More testing is necessary. I hope this is interesting or useful to some of you out there. Please share with me your thoughts and suggestions. --Jeff

1 0

open-tcp-client error
by Will Bendick 17 Sep '06

17 Sep '06

Hi, when I start gambit with: /usr/local/Gambit-C/bin/gsi things are fine, but if I start it with: /usr/local/Gambit-C/bin/gsi -:s /usr/pkg/share/slib/gambit.init - in order to get the slib stuff I have a problem with open-tcp-client > (open-tcp-client (list 'server-address: "zoo-crew.org" 'port-number: 80 'eol-encoding: 'cr-lf)) *** ERROR IN (console)(a)2.1 -- (Argument 1) Port settings expected (open-tcp-client '(server-address: "zoo-crew.org" port-number: 80 eol-encoding: cr-lf)) 1> Is this a known problem? Is there a known fix? I'm using Gambit Version 4.0 beta 17 |/|/ Bendick

2 1

Mailing list problem
by Marc Feeley 12 Sep '06

12 Sep '06

Hello. The Gambit mailing list server had some problems which prevented new subscriptions to the list. That problem should now be solved. Marc

1 0

A new socket interface for Gambit
by Jeff Read 12 Sep '06

12 Sep '06

Greetings, In light of Taylor's and Christian's concerns over socket support for Gambit -- concerns which I share -- I am currently working on Gamsock, a library that provides a convenient wrapper over the BSD socket interface for Gambit. The code is still not-quite-ready-for-prime-time, but currently I can bind, connect, and send to and receive from TCP sockets (and ostensibly Unix-domain and UDP sockets as well, but I haven't tested these fully). I anticipate doing a preliminary release soon (fixnum days, but probably less than 2^4 days). I hope that Gamsock fills a void that some of you have programming with Gambit. By the way, I seem unable to subscribe to gambit-list with my gmail account; it says "you need a valid email address". Is gmail not valid? Is it considered too spammy by Mailman? --Jeff

1 0

Re: [gambit-list] file read performance
by Marc Feeley 09 Sep '06

09 Sep '06

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On 4-Sep-06, at 4:35 AM, phil(a)phildawes.net wrote: > Hi Gambit List, > > What's the quickest way of reading a file from disk into memory? > > I tried 'read-all' with 'read-line', but unfortunately even with > compiling I get an order of magnitude worse performance than doing > the same in python. > I'm using the following code, compiled using gsc and gcc -O3, and I > get 500ms to read the 'bib' file supplied in the bench.tgz package. > > ---------------------- > > (define (f) > (call-with-input-file "bib" > (lambda (p) (read-all p read-line)))) > > (time (f)) > > ------------------------ > > In python, the same takes about 50ms > e.g. > for line in file("bib").readlines(): > # do something > > Is there anything I can do to speed this up? > (I've tried setting > #define ___MAX_CHR 0xff > ) > > Many thanks, > > Phil It depends what you mean by "reading into memory". You might want to slirp the content of the file into a single u8vector (as a sequence of bytes), or you might want to get the list of strings corresponding to the lines in the file (which requires a byte to character conversion). On my MacBook Pro, the former takes about 60 milliseconds for the "bib" file and the latter about 200 milliseconds (I also get about 50 ms for the Python code). I tried to improve the time with a specialized buffering but that only improved the timing by 20%. Using 1 byte characters instead of 4 byte characters helps, but it is still roughly 2.5 times slower than slirping the bytes into a u8vector. If you iterate using read-line (instead of using read- all) you save some more because the list of lines does not have to be kept in memory all the time. See the code below. I think the largest cost here is the allocation of the strings and conversion from bytes to characters. I'm not sure if the conversion to characters is a cost that Python is paying (can the text be encoded in something other than ISO-8859-1? that's something that you can easily change with Gambit's I/O system). Marc (declare (standard-bindings) (block) (fixnum) (not safe)) (define (read-all-file-v1 filename) (let* ((size (file-size filename)) (buffer (make-u8vector size))) (call-with-input-file filename (lambda (in) (let ((n (read-subu8vector buffer 0 size in))) buffer))))) (define (read-all-file-v2 filename) (call-with-input-file filename (lambda (in) (read-all in read-line)))) (define (read-all-file-v3 filename) (call-with-input-file filename (lambda (in) (let* ((size 4096) (buffer (make-string (+ size 1)))) (let loop1 ((rev-lines '()) (incomplete-line #f)) (let ((n (read-substring buffer 0 size in))) (string-set! buffer n #\newline) (if (= n 0) (reverse (if incomplete-line (cons incomplete-line rev-lines) rev-lines)) (let loop2 ((i 0) (rev-lines rev-lines) (incomplete-line incomplete-line)) (let loop3 ((j i)) (if (char=? (string-ref buffer j) #\newline) (if (= j n) (loop1 rev-lines (cond ((= i j) incomplete-line) (incomplete-line (string-append incomplete-line (substring buffer i j))) (else (substring buffer i j)))) (loop2 (+ j 1) (cons (if incomplete-line (string-append incomplete-line (substring buffer i j)) (substring buffer i j)) rev-lines) #f)) (loop3 (+ j 1)))))))))))) (define (read-all-file-v4 filename) (call-with-input-file filename (lambda (in) (let loop () (let ((line (read-line in))) (if (string? line) (begin ; do something with line (loop)))))))) (define filename "bench/src/bib") ; Execution time with character size = 4 bytes 1 byte (time (read-all-file-v1 filename)) ; 57 ms 57 ms (time (read-all-file-v2 filename)) ; 229 ms 170 ms (time (read-all-file-v3 filename)) ; 181 ms 146 ms (time (read-all-file-v4 filename)) ; 162 ms 130 ms -----BEGIN PGP SIGNATURE----- Version: GnuPG v1.4.3 (Darwin) iD8DBQFFAafz//V9Zc2T/v4RAr97AJ43/61utp7eFKE7M2f3AuaUBNs/aACgvw+j 7m+CVSFm+kqVjkRtIRgm5Lg= =+0j4 -----END PGP SIGNATURE-----

2 1

Register usage
by Bradley Lucier 07 Sep '06

07 Sep '06

Marc: The code for the main loop of this function is scheduled in 35 cycles when target gvm registers is 20 and target arg registers is 8; it's scheduled in 113 cycles when gvm registers is 5 and arg registers is 3 (the default). This is with the compile command gcc -mcpu=970 -m64 -no-cpp-precomp -Wall -W -Wno-unused -O1 -fno- math-errno -fschedule-insns2 -fno-trapping-math -fno-strict-aliasing - fwrapv -fomit-frame-pointer -fPIC -fno-common -bundle -flat_namespace -undefined suppress -I${GSC_CC_O_ARG1}include -D___DYNAMIC - D___SINGLE_HOST -o ${GSC_CC_O_ARG2} $* ${GSC_CC_O_ARG3} -save-temps - dR -freorder-blocks in gsc-cc-o on my PowerMac G5. This is a problem. You know that lambda-lifting and register allocation don't work well together in the compiler. This is a rather extreme example. It's hard to look for faster algorithms when the simple ones are compiled to something that takes three times as long as it should take. Brad (define (direct-fft-recursive a W-table) ;; This is a direcct complex fft, using a decimation-in-time algorithm ;; with inputs in natural order and outputs in bit-reversed order. ;; The table of "twiddle" factors is in bit-reversed order. ;; this is from page 66 of Chu and George (let ((W (##f64vector 0. 0.))) (define (make-W W-table W K) ;; look at Colin Percival's FFT notes for how to calculate this. ;; this works, but I'll be damned if I know why. (cond ((##fixnum.even? K) (let* ((K-index K) (K-index+1 (##fixnum.+ K-index 1)) (basic-real (##f64vector-ref W-table K-index)) (basic-imag (##f64vector-ref W-table K-index+1)) (result-real basic-real) (result-imag basic-imag)) (##f64vector-set! W 0 result-real) (##f64vector-set! W 1 result-imag))) (else (let* ((K-index (##fixnum.bitwise-and K -2)) (K-index+1 (##fixnum.+ K-index 1)) (basic-real (##f64vector-ref W-table K-index)) (basic-imag (##f64vector-ref W-table K-index+1)) (result-real (##flonum.- basic-imag)) (result-imag basic-real)) (##f64vector-set! W 0 result-real) (##f64vector-set! W 1 result-imag))))) (define (main-loop M N K SizeOfGroup) (##declare (not interrupts-enabled)) (let inner-loop ((K K) (JFirst M)) (if (##fixnum.< JFirst N) (let* ((JLast (##fixnum.+ JFirst SizeOfGroup))) (make-W W-table W K) (let J-loop ((J JFirst)) (if (##fixnum.< J JLast) (let ((J+1 (##fixnum.+ J 1 )) (J+SizeOfGroup (##fixnum.+ J SizeOfGroup)) (J+SizeOfGroup+1 (##fixnum.+ J 1 SizeOfGroup))) (let ((W-real (##f64vector-ref W 0)) (W-imag (##f64vector-ref W 1)) (a_J-real (##f64vector-ref a J )) (a_J-imag (##f64vector-ref a J+1)) (a_J+SizeOfGroup-real (##f64vector-ref a J+SizeOfGroup )) (a_J+SizeOfGroup-imag (##f64vector-ref a J+SizeOfGroup+1))) (let ((Temp-real (##flonum.- (##flonum.* W-real a_J+SizeOfGroup-real) (##flonum.* W-imag a_J+SizeOfGroup-imag))) (Temp-imag (##flonum.+ (##flonum.* W-real a_J+SizeOfGroup-imag) (##flonum.* W-imag a_J+SizeOfGroup-real)))) (let ((Result-J-real (##flonum.+ a_J-real Temp-real)) (Result-J-imag (##flonum.+ a_J-imag Temp-imag)) (Result-J+SizeOfGroup-real (##flonum.- a_J-real Temp-real)) (Result-J+SizeOfGroup-imag (##flonum.- a_J-imag Temp-imag))) (##f64vector-set! a J Result-J-real) (##f64vector-set! a J+1 Result-J-imag) (##f64vector-set! a J+SizeOfGroup Result-J+SizeOfGroup-real) (##f64vector-set! a J+SizeOfGroup+1 Result-J+SizeOfGroup-imag) (J-loop (##fixnum.+ J 2)))))) (inner-loop (##fixnum.+ K 1) (##fixnum.+ JLast SizeOfGroup)))))))) (define (recursive-bit M N K SizeOfGroup) (if (##fixnum.<= 2 SizeOfGroup) (begin (main-loop M N K SizeOfGroup) (if (##fixnum.< 2048 (##fixnum.- N M)) (begin (recursive-bit M (##fixnum.+ M SizeOfGroup) (##fixnum.* K 2) (##fixnum.quotient SizeOfGroup 2)) (recursive-bit (##fixnum.+ M SizeOfGroup) N (##fixnum.+ (##fixnum.* K 2) 1) (##fixnum.quotient SizeOfGroup 2))) (recursive-bit M N (##fixnum.* K 2) (##fixnum.quotient SizeOfGroup 2)))))) (recursive-bit 0 (##f64vector-length a) 0 (##fixnum.quotient (##f64vector-length a) 2))))

1 0

Re: [gambit-list] "Number of arguments exceeds implementation
by Bill Richter 06 Sep '06

06 Sep '06

>(define (filter-map/tail fn tail lis . morelis) [...] ) > >Christian, your code isn't tail-recursive, right? No it's not - it can't be if the order of the output should be the same as the input (and mutation isn't used). Thanks, Christian. That's a very helpful way to put it: functions like map can't be tail-recursive if they preserve the order unless we use mutation. (Unless you throw in some `reverse's) I had noted that mutation was used in various tail-recursive map-like functions, but I had only dimly noted the rule. I have my Sudoku code working now. It takes too long, but 1) quite likely that's because of beginner tail-recursion problems 2) The real problem was some "infinite loops" that I fixed.

2 2