[Top][All Lists]
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
Re: [Devel] Question about optimizing FT_MulFix.
From: |
David Turner |
Subject: |
Re: [Devel] Question about optimizing FT_MulFix. |
Date: |
Mon, 30 Apr 2001 18:04:18 +0200 |
Nathan Hurst a écrit :
>
> As a followup:
>
> I'm surprised about the slowness of version 5 - the powerpc compiler
> writer's manual makes a big deal about avoiding branches when possible
> using the condition code registers(which aren't used by gcc) and the
> select instructions(a machine code equivalent of c=(condition
> register)?a:b). If I knew how to compile ppc asm under gcc I would
> implement 5 using these tools and see how it fares.
I agree. I too was expecting version 5 to be the best, at least on
my old Pentium, given that I remember was performed quite often by
compilers in the 386/486 days. How things change :-)
> Also worth trying
> would be to run the test on MacOS using the FixMul routine provided in the
> std libs which is presumably fairly carefully optimised as it is used
> lots in the quickdraw routines.
> It's of course questionable whether there is much point in spending any
> more time on this.
>
Agreed !
- David