Re: Permutation Tensor

From	anton@mips.complang.tuwien.ac.at (Anton Ertl)
Newsgroups	comp.lang.forth
Subject	Re: Permutation Tensor
Date	2012-07-11 13:50 +0000
Organization	Institut fuer Computersprachen, Technische Universitaet Wien
Message-ID	<2012Jul11.155058@mips.complang.tuwien.ac.at> (permalink)
References	<2012Jul10.121108@mips.complang.tuwien.ac.at> <62661493968435@frunobulax.edu>

Show all headers | View raw

mhx@iae.nl (Marcel Hendrix) writes:
>anton@mips.complang.tuwien.ac.at (Anton Ertl) writes Re: Permutation Tensor
>
>> m.a.m.hendrix@tue.nl writes:
>[..]
>>>It is possible to see that branching is indeed slow by swapping the test cases in lcsymbol2:
>[..]
>> Hmm, sounds like perfectly predictable branches.  Then it's probably
>> something else.  You can check the number of mispredictions with
>> performance counters.
[...]
>$0124B592  jne           $0124B5A4 offset NEAR
>$0124B598  mov           rbx, 1 d#
>$0124B59F  jmp           $0124B615 offset NEAR
>$0124B5A4  cmp           rbx, #54 b#
>$0124B5A8  jne           $0124B5BA offset NEAR
>$0124B5AE  mov           rbx, 1 d#
>$0124B5B5  jmp           $0124B615 offset NEAR
>$0124B5BA  cmp           rbx, #30 b#
>$0124B5BE  jne           $0124B5D0 offset NEAR
>$0124B5C4  mov           rbx, -1 d#
>$0124B5CB  jmp           $0124B615 offset NEAR
>$0124B5D0  cmp           rbx, #39 b#
>$0124B5D4  jne           $0124B5E6 offset NEAR
>$0124B5DA  mov           rbx, -1 d#
>$0124B5E1  jmp           $0124B615 offset NEAR
>$0124B5E6  cmp           rbx, #57 b#
>$0124B5EA  jne           $0124B5FC offset NEAR
>$0124B5F0  mov           rbx, -1 d#
>$0124B5F7  jmp           $0124B615 offset NEAR
>$0124B5FC  cmp           rbx, #27 b#
>$0124B600  jne           $0124B612 offset NEAR
>$0124B606  mov           rbx, 1 d#
>$0124B60D  jmp           $0124B615 offset NEAR
[...]

Hmm, I remember something about the number of branch prediction table
entries per cache line or something being limited.  Maybe that's the
reason for the low performance.

If I wanted to verify this, I would first use performance counters to
check if there is a high number of branch mispredictions, and if so,
insert some padding after the "jmp"s to see if it helps.

- anton
-- 
M. Anton Ertl  http://www.complang.tuwien.ac.at/anton/home.html
comp.lang.forth FAQs: http://www.complang.tuwien.ac.at/forth/faq/toc.html
     New standard: http://www.forth200x.org/forth200x.html
   EuroForth 2012: http://www.euroforth.org/ef12/

Back to comp.lang.forth | Previous | Next — Previous in thread | Next in thread | Find similar

Thread

Permutation Tensor Arnold Doray <invalid@invalid.com> - 2012-07-05 07:13 +0000
  Re: Permutation Tensor "A. K." <akk@nospam.org> - 2012-07-05 09:25 +0200
    Re: Permutation Tensor Arnold Doray <invalid@invalid.com> - 2012-07-05 08:42 +0000
      Re: Permutation Tensor "A. K." <akk@nospam.org> - 2012-07-05 11:25 +0200
        Re: Permutation Tensor Krishna Myneni <krishna.myneni@ccreweb.org> - 2012-07-06 18:49 -0700
          Re: Permutation Tensor mhx@iae.nl (Marcel Hendrix) - 2012-07-07 08:32 +0200
            Re: Permutation Tensor "A. K." <akk@nospam.org> - 2012-07-07 09:49 +0200
            Re: Permutation Tensor Krishna Myneni <krishna.myneni@ccreweb.org> - 2012-07-07 05:35 -0700
               Re: Permutation Tensor mhx@iae.nl (Marcel Hendrix) - 2012-07-07 15:45 +0200
                Re: Permutation Tensor Krishna Myneni <krishna.myneni@ccreweb.org> - 2012-07-07 07:32 -0700
                Re: Permutation Tensor mhx@iae.nl (Marcel Hendrix) - 2012-07-07 17:53 +0200
                Re: Permutation Tensor Krishna Myneni <krishna.myneni@ccreweb.org> - 2012-07-07 09:10 -0700
                Re: Permutation Tensor mhx@iae.nl (Marcel Hendrix) - 2012-07-07 18:41 +0200
                Re: Permutation Tensor Krishna Myneni <krishna.myneni@ccreweb.org> - 2012-07-07 12:24 -0700
                Re: Permutation Tensor Gerry Jackson <gerry@jackson9000.fsnet.co.uk> - 2012-07-08 07:48 +0100
                Re: Permutation Tensor Gerry Jackson <gerry@jackson9000.fsnet.co.uk> - 2012-07-08 08:09 +0100
                Re: Permutation Tensor mhx@iae.nl (Marcel Hendrix) - 2012-07-08 09:23 +0200
                Re: Permutation Tensor Krishna Myneni <krishna.myneni@ccreweb.org> - 2012-07-08 05:55 -0700
                Re: Permutation Tensor Gerry Jackson <gerry@jackson9000.fsnet.co.uk> - 2012-07-08 20:14 +0100
                Re: Permutation Tensor Krishna Myneni <krishna.myneni@ccreweb.org> - 2012-07-08 13:49 -0700
                Re: Permutation Tensor Paul Rubin <no.email@nospam.invalid> - 2012-07-07 13:03 -0700
                Re: Permutation Tensor mhx@iae.nl (Marcel Hendrix) - 2012-07-07 23:54 +0200
                Re: Permutation Tensor Paul Rubin <no.email@nospam.invalid> - 2012-07-07 16:43 -0700
                Re: Permutation Tensor mhx@iae.nl (Marcel Hendrix) - 2012-07-08 09:07 +0200
                Re: Permutation Tensor Albert van der Horst <albert@spenarnc.xs4all.nl> - 2012-07-08 13:03 +0000
                Re: Permutation Tensor anton@mips.complang.tuwien.ac.at (Anton Ertl) - 2012-07-09 14:50 +0000
                Re: Permutation Tensor m.a.m.hendrix@tue.nl - 2012-07-10 00:28 -0700
                Re: Permutation Tensor anton@mips.complang.tuwien.ac.at (Anton Ertl) - 2012-07-10 10:11 +0000
                Re: Permutation Tensor mhx@iae.nl (Marcel Hendrix) - 2012-07-10 21:37 +0200
                Re: Permutation Tensor anton@mips.complang.tuwien.ac.at (Anton Ertl) - 2012-07-11 13:50 +0000
                Re: Permutation Tensor Krishna Myneni <krishna.myneni@ccreweb.org> - 2012-07-10 05:10 -0700
        Re: Permutation Tensor Paul Rubin <no.email@nospam.invalid> - 2012-07-06 20:46 -0700
          Re: Permutation Tensor Krishna Myneni <krishna.myneni@ccreweb.org> - 2012-07-07 05:36 -0700
  Re: Permutation Tensor Krishna Myneni <krishna.myneni@ccreweb.org> - 2012-07-06 18:18 -0700
    Re: Permutation Tensor Krishna Myneni <krishna.myneni@ccreweb.org> - 2012-07-06 18:33 -0700
    Re: Permutation Tensor Krishna Myneni <krishna.myneni@ccreweb.org> - 2012-07-07 07:28 -0700
      Re: Permutation Tensor Krishna Myneni <krishna.myneni@ccreweb.org> - 2012-07-07 08:02 -0700
  Re: Permutation Tensor Albert van der Horst <albert@spenarnc.xs4all.nl> - 2012-07-07 19:58 +0000
    Re: Permutation Tensor Krishna Myneni <krishna.myneni@ccreweb.org> - 2012-07-08 05:45 -0700
      Re: Permutation Tensor Albert van der Horst <albert@spenarnc.xs4all.nl> - 2012-07-08 20:05 +0000
  Re: Permutation Tensor awegel@arcor.de (Alex Wegel) - 2012-07-08 14:24 +0200
    Re: Permutation Tensor awegel@arcor.de (Alex Wegel) - 2012-07-11 07:09 +0200
      Re: Permutation Tensor awegel@arcor.de (Alex Wegel) - 2012-07-11 18:42 +0200
        Re: Permutation Tensor mhx@iae.nl (Marcel Hendrix) - 2012-07-11 20:18 +0200
          Re: Permutation Tensor awegel@arcor.de (Alex Wegel) - 2012-07-12 12:43 +0200
  Re: Permutation Tensor awegel@arcor.de (Alex Wegel) - 2012-07-08 23:44 +0200
    Re: Permutation Tensor Krishna Myneni <krishna.myneni@ccreweb.org> - 2012-07-10 17:06 -0700
      Re: Permutation Tensor awegel@arcor.de (Alex Wegel) - 2012-07-11 07:09 +0200

csiph-web