Re: It's 1997. Do you know where your scheduler is?

"D. J. Bernstein" <djb@cr.yp.to>
29 Dec 1997 21:05:06 -0500

          From comp.compilers

Related articles
It's 1997. Do you know where your scheduler is? djb@cr.yp.to (D. J. Bernstein) (1997-12-19)
Re: It's 1997. Do you know where your scheduler is? chase@world.std.com (David Chase) (1997-12-23)
Re: It's 1997. Do you know where your scheduler is? djb@cr.yp.to (D. J. Bernstein) (1997-12-29)
Re: It's 1997. Do you know where your scheduler is? greened@eecs.umich.edu (David Greene) (1998-01-04)
| List of all articles for this month |
From: "D. J. Bernstein" <djb@cr.yp.to>
Newsgroups: comp.compilers
Date: 29 Dec 1997 21:05:06 -0500
Organization: Compilers Central
Keywords: optimize, GCC

> which release of gcc,


It really doesn't matter. Here are some Pentium cycle counts for the
same hand-scheduled 256-point complex FFT code:


      23085 (best) gcc 2.7.2.1 -O1 -fo-f-p
      41913 gcc 2.7.2.1 -O6 -fo-f-p
      47258 egcs 1.00 -O6 -fo-f-p -mpentiumpro
      56383 egcs 1.00 -O6 -fo-f-p -mpentium
      56860 (worst) egcs 1.00 -O6 -fo-f-p


I wonder what percentage of today's typical Pentium is wasted by bad
instruction scheduling, given that so many programmers rely on the C
compiler for optimization. Intel's compiler is allegedly not too
incompetent, but how many programs are actually compiled with it?


> which release of Pentium.


``Pentium'' refers to a specific chip design. All Pentium releases
use the same number of cycles for in-cache operations.


The Pentium Pro, as you noted, is a very different chip. Scheduling
code badly for the Pentium Pro would take quite a bit of effort.


Apparently the Pentium MMX is similar to a Pentium for optimization
purposes, and the Pentium II is just like a Pentium Pro; except, of
course, for the extra instructions and larger L1 caches.


---Dan
--


Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.