To: eliz AT is DOT elta DOT co DOT il, dj AT ctron DOT com
Cc: djgpp AT sun DOT soe DOT clarkson DOT edu
Subject: Re: DJGPP Speed 
Date: Fri, 25 Mar 1994 01:18:03 -0800
From: Darryl Okahata <darrylo AT sr DOT hp DOT com>

>   Is 200 lines per second reasonable for DJGCC 1.11?  I tested it on my
> 486DX2-66/16M with a SCSI disk and 2Meg SMARTDrv cache, compiling the
> makeinfo.c file (it has about 7500 lines, so the overhead of loading the
> multiple passes of the compiler would be insignificant; it took about 35
> seconds to compile [GCC -c only, without ld]).  I have QEMM 7.03 and QDPMI

     This sounds reasonable, for an "average" 486DX2/66.  Even on a 16MB
60MHz Pentium w/512KB cache, "gcc -c makeinfo.c" takes about 28 seconds
with smartdrv (2MB) write-caching disabled, and about 19 seconds with
write-caching enabled.  Real mode (Turbo C) "gcc" is being used outside
of Windows (i.e., Windows is not running).  However, this system has
"average" SCSI disks (1900kb/sec peak thoughput, who knows what MSDOS is
getting), and so the I/O subsystem could use a lot of improvement
(Seagate Barracudas would be nice ;-).

[ One's first reaction to the above is that something is wrong with the
  Pentium.  Well, that's what I thought, soon after getting this
  Pentium.  However, after running a variety of different tests and
  benchmarks, the Pentium processor is performing as expected (a little
  faster, even).  The I/O subsystem is now the bottleneck, and this just
  goes to show how important it is to have a well-rounded system.  I'm
  using an Adaptec VLB 2842, but the disks are old and slow (a SCSI-1
  300MB CDC/Imprimis 94241-7, and a SCSI-2 520MB Fujitsu 2624FA). ]

>   I don't have Unix workstation handy right now, but I remember that GCC was
> *much* faster on a 40Mhz Sparc2 with about the same amount of RAM.  Isn't 486
> about the same MIPS as Sparc2?

     Unix workstations also have a well-tuned filesystem and buffer
cache.  MSDOS/Windows are rather primitive when it comes to disk I/O,
and could use a lot of improvement.  Raw processor speed is not enough.

     -- Darryl Okahata
	Internet: darrylo AT sr DOT hp DOT com

DISCLAIMER: this message is the author's personal opinion and does not
constitute the support, opinion or policy of Hewlett-Packard or of the
little green men that have been following him all day.