This is the mail archive of the gcc-bugs@gcc.gnu.org mailing list for the GCC project.

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]

Re: c/859: sparc-specific pessimal code for simple loop

To: "David S. Miller" <davem at redhat dot com>
Subject: Re: c/859: sparc-specific pessimal code for simple loop
From: Geoff Keating <geoffk at geoffk dot org>
Date: 22 Nov 2000 13:19:09 -0800
CC: jakub at redhat dot com, gcc-gnats at gcc dot gnu dot org, gcc-bugs at gcc dot gnu dot org, bernds at redhat dot com, jbuck at synopsys dot com
CC: rth at redhat dot com
References: <20001121213559.26387.qmail@sourceware.cygnus.com> <20001121230119.A19367@redhat.com> <200011220702.XAA06961@pizda.ninka.net>

"David S. Miller" <davem@redhat.com> writes:

>    Date: Tue, 21 Nov 2000 23:01:19 -0800
>    From: Richard Henderson <rth@redhat.com>
> 
>    This is a STRICT_ALIGNMENT machine, therefore DFmode and DImode
>    operands must be 8-byte aligned, thus the EXTRA_CONSTRAINT check
>    ought to be redundant, no?
> 
> I remember there were some issues with var-args, which could make
> "long long" and other DF/DI mode values go into the stack only 4-byte
> aligned.  This case needed to be handled, and I am pretty sure this is
> where all the wacky constraint stuff comes from.

Then all that code didn't work, because I recently changed the sparc
varargs code to explicitly copy 8-byte values to a temporary because
multiple varargs testcases were failing.  So it can all probably go away.

-- 
- Geoffrey Keating <geoffk@geoffk.org>

References:
- c/859: sparc-specific pessimal code for simple loop
  - From: jbuck
- Re: c/859: sparc-specific pessimal code for simple loop
  - From: Richard Henderson
- Re: c/859: sparc-specific pessimal code for simple loop
  - From: David S. Miller

Index Nav:	[Date Index] [Subject Index] [Author Index] [Thread Index]
Message Nav:	[Date Prev] [Date Next]	[Thread Prev] [Thread Next]