Programming Language Performance

Chase · **Joined:**Mon Sep 22, 2008 9:51 am **Posts:**112

Quote:

Ahh you didn't rerun the fortran version.

I dumped the FORTRAN version. The print formatting was impossible to get right, and the other natively compiled code tests basically show what FORTRAN would do anyway.

Impossible, come now, you should know thats not true. I would rate it in the region of fairly to moderately difficult. Or at the very least, 'not easy'.

eddieringle · PostPosted:Fri Apr 10, 2009 1:02 pm

Any chance you can test GenPrime using Ruby 1.9, Tycho? I'd like to see the differences.

eddieringle · PostPosted:Tue Apr 14, 2009 6:32 pm

Nevermind, ran it on my laptop running Ubuntu 8.10.

Versions:

Code:

ruby -v
ruby 1.8.7 (2008-08-11 patchlevel 72) [i486-linux]

ruby1.9 -v
ruby 1.9.0 (2008-06-20 revision 17482) [i486-linux]

php -v
PHP 5.2.6-2ubuntu4.1 with Suhosin-Patch 0.9.6.2 (cli) (built: Feb 11 2009 20:38:24)

Results:
ruby 1.8.7:

Code:

ruby genprime.rb 250000 1000000
Found  250000 primes in  219.86265 seconds (last was  3497861)
Found  500000 primes in  617.49590 seconds (last was  7368787)
Found  750000 primes in 1132.28696 seconds (last was 11381621)
Found 1000000 primes in 1750.21173 seconds (last was 15485863)

ruby 1.9.0:

Code:

ruby1.9 genprime.rb 250000 1000000
Found  250000 primes in  45.06478 seconds (last was  3497861)
Found  500000 primes in 124.48929 seconds (last was  7368787)
Found  750000 primes in 227.51096 seconds (last was 11381621)
Found 1000000 primes in 350.35934 seconds (last was 15485863)

php 5.2.6:

Code:

Found  250000 primes in  84.97550 seconds (last was  3497861)
Found  500000 primes in 237.42517 seconds (last was  7368787)
Found  750000 primes in 435.71111 seconds (last was 11381621)
Found 1000000 primes in 665.24334 seconds (last was 15485863)

These are all Ubuntu versions, NOT the latest releases from each language's developers. (Especially the PHP version, 5.2.6 is old)

I'm surprised how much faster 1.9 is (and now I must switch to Ruby as my primary scripting language, due to a deal between ChaosR and I.)

Tycho · PostPosted:Tue Apr 14, 2009 7:55 pm

Impressive. I wonder what the differences are between the 1.8.x and 1.9.x series which so drastically changed performance?

ChaosR · PostPosted:Wed Apr 15, 2009 7:04 am

A bytecode interpreter instead of a syntax tree interpreter.

Tycho · PostPosted:Wed Apr 15, 2009 2:22 pm

At least Ruby now can be considered competition for the others...

eddieringle · PostPosted:Wed Nov 11, 2009 10:17 pm

With today's release of Go, I thought I would write a version of GenPrime for it, one because I wanted to see if it was really as speedy as it says and two to practice this new language.

First, the C version as a comparison:

Code:

eddie@ganglion:~/dev/genprime$ ./genprime-c 250000 1000000
Found   250000 primes in    5.75923 seconds (last was    3497861)
Found   500000 primes in   16.15777 seconds (last was    7368787)
Found   750000 primes in   29.94442 seconds (last was   11381621)
Found  1000000 primes in   46.37428 seconds (last was   15485863)

Now for Go:

Code:

eddie@ganglion:~/dev/genprime$ ./genprime-go 250000 1000000
Found   250000 primes in   4.253223 seconds (last was  3497861)
Found   500000 primes in  12.105688 seconds (last was  7368787)
Found   750000 primes in  22.779920 seconds (last was 11381621)
Found  1000000 primes in  34.426428 seconds (last was 15485863)

I compiled the Go version with the Plan 9 compiler (6g). Apparently there is a GCC port that is reported to produced more optimized code in some cases, so if anyone wants to test this code with the GCC port I'd be interested to see the results.

I'll be committing my code to my GitHub fork of GenPrime.

eddieringle · PostPosted:Wed Nov 11, 2009 10:21 pm

Assembly dump of isprime and genprime functions:

Code:

--- prog list "isprime" ---
(genprime.go:46) TEXT    isprime+0(SB),$40-16
(genprime.go:46) MOVL    x+0(FP),CX
(genprime.go:15) JMP     ,37
(genprime.go:15) JMP     ,42
(genprime.go:15) CMPL    CX,$2
(genprime.go:15) JGE     ,36
(genprime.go:16) MOVL    $0,.noname+8(FP)
(genprime.go:16) RET     ,
(genprime.go:15) JMP     ,42
(genprime.go:18) JMP     ,44
(genprime.go:18) JMP     ,49
(genprime.go:18) CMPL    CX,$4
(genprime.go:18) JGE     ,43
(genprime.go:19) MOVL    $1,.noname+8(FP)
(genprime.go:19) RET     ,
(genprime.go:18) JMP     ,49
(genprime.go:21) JMP     ,51
(genprime.go:21) JMP     ,56
(genprime.go:21) CMPL    CX,$5
(genprime.go:21) JNE     ,50
(genprime.go:22) MOVL    $1,.noname+8(FP)
(genprime.go:22) RET     ,
(genprime.go:21) JMP     ,56
(genprime.go:24) JMP     ,58
(genprime.go:24) JMP     ,71
(genprime.go:24) MOVL    CX,BX
(genprime.go:24) MOVL    CX,BP
(genprime.go:24) MOVL    CX,R8
(genprime.go:24) SARL    $31,R8
(genprime.go:24) SUBL    R8,BP
(genprime.go:24) SARL    $1,BP
(genprime.go:24) IMULL   $2,BP
(genprime.go:24) SUBL    BP,BX
(genprime.go:24) CMPL    BX,$0
(genprime.go:24) JNE     ,57
(genprime.go:25) MOVL    $0,.noname+8(FP)
(genprime.go:25) RET     ,
(genprime.go:24) JMP     ,71
(genprime.go:27) JMP     ,73
(genprime.go:27) JMP     ,88
(genprime.go:27) MOVL    CX,BX
(genprime.go:27) MOVL    CX,R8
(genprime.go:27) MOVL    $1717986919,AX
(genprime.go:27) IMULL   CX,
(genprime.go:27) SARL    $1,DX
(genprime.go:27) SARL    $31,R8
(genprime.go:27) SUBL    R8,DX
(genprime.go:27) MOVL    DX,BP
(genprime.go:27) IMULL   $5,BP
(genprime.go:27) SUBL    BP,BX
(genprime.go:27) CMPL    BX,$0
(genprime.go:27) JNE     ,72
(genprime.go:28) MOVL    $0,.noname+8(FP)
(genprime.go:28) RET     ,
(genprime.go:27) JMP     ,88
(genprime.go:30) JMP     ,90
(genprime.go:30) JMP     ,122
(genprime.go:30) MOVL    CX,BX
(genprime.go:30) INCL    ,BX
(genprime.go:30) MOVL    BX,R8
(genprime.go:30) MOVL    $715827883,AX
(genprime.go:30) IMULL   BX,
(genprime.go:30) SARL    $0,DX
(genprime.go:30) SARL    $31,R8
(genprime.go:30) SUBL    R8,DX
(genprime.go:30) MOVL    DX,BP
(genprime.go:30) IMULL   $6,BP
(genprime.go:30) SUBL    BP,BX
(genprime.go:30) CMPL    BX,$0
(genprime.go:30) JEQ     ,89
(genprime.go:31) JMP     ,105
(genprime.go:31) JMP     ,121
(genprime.go:31) MOVL    CX,BX
(genprime.go:31) DECL    ,BX
(genprime.go:31) MOVL    BX,R8
(genprime.go:31) MOVL    $715827883,AX
(genprime.go:31) IMULL   BX,
(genprime.go:31) SARL    $0,DX
(genprime.go:31) SARL    $31,R8
(genprime.go:31) SUBL    R8,DX
(genprime.go:31) MOVL    DX,BP
(genprime.go:31) IMULL   $6,BP
(genprime.go:31) SUBL    BP,BX
(genprime.go:31) CMPL    BX,$0
(genprime.go:31) JEQ     ,104
(genprime.go:32) MOVL    $0,.noname+8(FP)
(genprime.go:32) RET     ,
(genprime.go:31) JMP     ,121
(genprime.go:30) JMP     ,122
(genprime.go:35) MOVL    $0,match+-4(SP)
(genprime.go:36) CVTSL2SD CX,X1
(genprime.go:36) MOVSD   X1,(SP)
(genprime.go:36) CALL    ,math.Sqrt+0(SB)
(genprime.go:36) MOVL    match+-4(SP),SI
(genprime.go:36) MOVSD   8(SP),X0
(genprime.go:36) MOVSD   $(1.00000000000000000e+00),X1
(genprime.go:36) ADDSD   X1,X0
(genprime.go:36) MOVSD   X0,X5
(genprime.go:37) MOVL    $3,CX
(genprime.go:37) JMP     ,135
(genprime.go:37) JMP     ,150
(genprime.go:37) ADDL    $2,CX
(genprime.go:37) CVTSL2SD CX,X1
(genprime.go:37) MOVSD   X1,X0
(genprime.go:37) UCOMISD X1,X5
(genprime.go:37) JLS     ,133
(genprime.go:38) JMP     ,141
(genprime.go:38) JMP     ,149
(genprime.go:38) MOVL    x+0(FP),AX
(genprime.go:38) CDQ     ,
(genprime.go:38) IDIVL   CX,
(genprime.go:38) CMPL    DX,$0
(genprime.go:38) JNE     ,140
(genprime.go:39) MOVL    $1,SI
(genprime.go:40) JMP     ,133
(genprime.go:38) JMP     ,149
(genprime.go:37) JMP     ,134
(genprime.go:43) JMP     ,152
(genprime.go:43) JMP     ,157
(genprime.go:43) CMPL    SI,$1
(genprime.go:43) JNE     ,151
(genprime.go:44) MOVL    $0,.noname+8(FP)
(genprime.go:44) RET     ,
(genprime.go:43) JMP     ,157
(genprime.go:46) MOVL    $1,.noname+8(FP)
(genprime.go:46) RET     ,
(genprime.go:46) CALL    ,runtime.throwreturn+0(SB)
(genprime.go:46) RET     ,

--- prog list "genprime" ---
(genprime.go:56) TEXT    genprime+0(SB),$24-16
(genprime.go:50) MOVL    $1,AX
(genprime.go:51) MOVL    $0,CX
(genprime.go:51) JMP     ,167
(genprime.go:51) JMP     ,184
(genprime.go:51) INCL    ,AX
(genprime.go:51) MOVL    CX,count+-8(SP)
(genprime.go:51) MOVL    max+0(FP),BP
(genprime.go:51) CMPL    CX,BP
(genprime.go:51) JGE     ,165
(genprime.go:52) JMP     ,173
(genprime.go:52) JMP     ,183
(genprime.go:52) MOVL    AX,current+-4(SP)
(genprime.go:52) MOVL    AX,(SP)
(genprime.go:52) CALL    ,isprime+0(SB)
(genprime.go:52) MOVL    count+-8(SP),CX
(genprime.go:52) MOVL    current+-4(SP),AX
(genprime.go:52) MOVL    8(SP),BX
(genprime.go:52) CMPL    BX,$1
(genprime.go:52) JNE     ,172
(genprime.go:53) INCL    ,CX
(genprime.go:52) JMP     ,183
(genprime.go:51) JMP     ,166
(genprime.go:56) MOVL    AX,BX
(genprime.go:56) DECL    ,BX
(genprime.go:56) MOVL    BX,.noname+8(FP)
(genprime.go:56) RET     ,
(genprime.go:56) CALL    ,runtime.throwreturn+0(SB)
(genprime.go:56) RET     ,

Tycho · PostPosted:Thu Nov 12, 2009 12:12 am

Ah, you used floating-point math, while the other implementations use integer math.

Yes, you can get some speedups doing this, especially on Core 2 processors because a tight loop of integer divisions will always lose to a tight loop of floating-point divisions. You can do the same with the C version by using 'float' for prime_t and fmod() to replace the modulus operator.

I'll work up some code which demonstrates this.

eddieringle · PostPosted:Fri Nov 13, 2009 6:12 am

Quote:

I'll work up some code which demonstrates this.

Well?

Tycho · PostPosted:Fri Nov 13, 2009 1:50 pm

Quote:

I'll work up some code which demonstrates this.

Well?

Sorry, have a bit going on this week... I'll probably get to this on Sunday.

eddieringle · PostPosted:Thu Feb 18, 2010 3:15 pm

Quote:

I'll work up some code which demonstrates this.

Well?

Sorry, have a bit going on this week... I'll probably get to this on Sunday.

Did you ever get to this?

Tycho · PostPosted:Sat Feb 20, 2010 4:16 am

No, I can do so tomorrow if you remind me somehow. Email would do.

dodger · **Joined:**Sat Mar 13, 2010 9:18 am **Posts:**3

That Java vs. Python results in your first post made me suspicous, cos Python is not that slow. Second thought was that its cos you ran it on a multi-core, maybe? Did Java utilize both cores? Cos Python does not support multi-threading, rather multi-processing.
Oh, and whats about Delphi? Not sure if Delphi still counts, cos Ive heard it compiles with C now. Not sure. Long time ago that Ive used Delphi/Pascal.

Tycho · PostPosted:Mon Mar 15, 2010 10:39 am

Quote:

That Java vs. Python results in your first post made me suspicous, cos Python is not that slow. Second thought was that its cos you ran it on a multi-core, maybe? Did Java utilize both cores? Cos Python does not support multi-threading, rather multi-processing.

Python really is that slow. I have numerous examples to show it, too. If you or anyone else is interested, I can make these public. It's sad, too, because I'm a huge Python fan. It's a fantastic language, but the main implementation (CPython) is horrendously slow. Hopefully the Unladen Swallow guys finally get some optimization done and get their code merged into CPython.

And no, multi-core couldn't have made a difference. We didn't really do anything in these tests that could be done in parallel, nor did we explicitly try to do so.

Quote:

Oh, and whats about Delphi? Not sure if Delphi still counts, cos Ive heard it compiles with C now. Not sure. Long time ago that Ive used Delphi/Pascal.

It's increasingly difficult to find Delphi or Pascal compilers. But if you want to contribute the code, go ahead. I just won't be able to test it.

Programming Language Performance

Who is online