part of J8 is openssl libeay on windows, and I believe its either also
distributed or presumed on other platforms. It includes a BN library. As a
start,
I need some help with the code below. after loading (windows 64 j802),
a =. dec2bn 1231231144 NB. looks ok.
┌──┬───────────┬──────────┐
│10│_1181373152│1231231144│
└──┴───────────┴──────────┘
|. a. i. bn2dec a
(hopefully doesn't crash)
I don't really know what gets returned yet, but what is worrysome is that
repeated calls return slightly different values.
Some issues I don't know how to deal with in the code below,
BNnum_bytes is a macro, and error 2 0 occurs if I try to call it. Is there a
workaround?
BN_dec2bn (called above) actually has a pointer to pointer first argument
(**BIGNUM). Is there something special I have to do in the function signature,
or when calling? I'm pretending its just a pointer.
require 'dll'
sslp =: IFWIN pick '';'D:\OpenSSL-Win64\bin\'
sslp =: IFWIN pick ''; '/',~ jpath '~bin' NB. with J802. cut this line if you
wish to point to downloaded folder
NB. OPENSSL =: jpath '~system/ssleay32.dll '
NB.OPENSSL =: sslp , '\ssleay32.dll '
OPENSSL =: sslp , (IFIOS + (;: 'Win Linux Android Darwin') i. <UNAME_z_) pick
'libeay32.dll '; (2 $ <'libssl.so.1.0.0 '), (2 $ <'/usr/lib/libssl.dylib ')
NB.OPENSSL =: sslp , (IFIOS + (;: 'Win Linux Android Darwin') i. <UNAME_z_)
pick 'libeay32.dll '; (2 $ <'libssl.so '); (2 $ <'libssl.0.9.8.dylib ')
SSLE =: sslp , '\openssl'
ssl =: 1 : '(OPENSSL , m)&cd'
BNctxnew =: ' BN_CTX_new *i' ssl
BNnew =: ' BN_new *i' ssl NB. probably don't use... unamanaged memory
version.
BNmul =: ' BN_mul + l *x *x *x *x' ssl
NB. BIGNUM *BN_bin2bn(const unsigned char *s,int len,BIGNUM *ret)
BN2bn =: ' BN_bin2bn *x *c l *x' ssl
NB. int BN_bn2bin(const BIGNUM *a, unsigned char *to)
BN2bin =: ' BN_bn2bin l *i *c' ssl
BN_bn2hex=: ' BN_bn2bin l *i *c' ssl NB.char * BN_bn2hex(const BIGNUM *a);
BN_hex2bn =: ' BN_bn2bin l *i *c' ssl NB.(BIGNUM **a, const char *str);
BNnum_bytes=: ' BN_num_bytes i *i' ssl NB. doest work as its macro?
NB. int BN_num_bits(const BIGNUM *a)
BN_num_bits=: ' BN_num_bits i *i' ssl
num_bytes =: BN_num_bits
NB. char *BN_bn2dec(const BIGNUM *num)
BN2dec=: ' BN_bn2dec *c *i' ssl
NB. int BN_dec2bn(BIGNUM **num, const char *str)
dec2BN=: ' BN_dec2bn x *i *c' ssl
NB. BN_print(BIO *fp, const BIGNUM *a);
pD =: 1!:2&2
dec2bn=: 3 : 0
o =. >BNnew 0{.a.
NB. o =. 0{.a.
NB.pD i=. dec2BN (o);(a. {~ 256 #. inv y)
pD i=. dec2BN (o);(": y)
2 {. i
)
bn2dec=: 3 : 0
NB.len =. BNnum_bytes y
NB.o =. len # '0'
pD y
pD o =. BN2dec (, 1 { y)
NB.pD a =.BNnum_bytes ,<o
memr (0{::o),0,(0{::y),2
)
----- Original Message -----
From: aai <[email protected]>
To: [email protected]
Cc:
Sent: Wednesday, September 2, 2015 3:50 AM
Subject: Re: [Jprogramming] Comparing J speed
The main reason for speed difference is that big integer calculation in
Haskell is based on the GNU Multiple Precision Arithmetic Library
(/GMP/), much faster than J's extended precision number calculation.
Op 2-9-2015 om 02:32 schreef Jon Hough:
> In this talk https://www.youtube.com/watch?v=apBWkBDVlow
> the presenter attempts to show Haskell hasn't sacrificed speed for
> expressiveness by comparing a Java Fibonacci calculator to his Haskell
> one.(skip to the 18:00 mark).Essentially, to calculate the 475000th Fibonacci
> number, it took his Java program ~8 seconds, while the very terse Haskell
> program took ~6 seconds.
> So I tried to do the same in J. My first attempt, used a tacit, memoized verb
> fib1 =:1:`(($:@:<:) + ($:@:-&2))@.(2&<)M.
>
>
> However, this gives a stack error for large numbers (~100000). So I decided
> to make an imperative verb,
>
>
> fib2 =: 3 : 0 x1 =. x:1 x2 =. x:1 c =. 0 while. c < y do. tmp =. x1 x1 =. x2
> x2 =. tmp + x1 c=.c+1 end. x2
>
>
>
>
>
>
>
>
>
>
>
>
> )
>
>
> This gets there, I can calculate the 475000th Fibonacci number, but
>
>
> timespacex 'fib2 475000'
>
>
>
>
> 36.183 1.31558e6
>
>
> It takes 36 seconds (of course, my hardware is different to that in the
> presentation, but still...).
>
>
> Is there a speedier way to do this in J? Preferably a tacit one liner would
> also be good.
>
>
> Thanks,
> Jon
>
>
> ----------------------------------------------------------------------
> For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
----------------------------------------------------------------------
For information about J forums see http://www.jsoftware.com/forums.htm
