This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
Re: patch: promoting complex modes (plus splitting complex args)
- From: Aldy Hernandez <aldyh at redhat dot com>
- To: Richard Kenner <kenner at vlsi1 dot ultra dot nyu dot edu>
- Cc: gcc-patches at gcc dot gnu dot org
- Date: Wed, 22 Jan 2003 11:08:04 -0800
- Subject: Re: patch: promoting complex modes (plus splitting complex args)
- References: <10301221823.AA17110@vlsi1.ultra.nyu.edu>
On Wed, Jan 22, 2003 at 01:23:55PM -0500, Richard Kenner wrote:
> The culprit is in the code expanding PARM_DECLs which falls through to
> the RESULT_DECL case, particularly because we only handle promotions
> of REG:
>
> My first question would be when did we start promoting FP modes and what
> sense does that make?
>
> The reason I added promoted modes to GCC was for a case where a machine did
> not have instructions in a smaller integer mode, such as a byte. That
> would mean, for example, that every compare instruction would require
> a sign- or zero-extension. Since you assign less than you reference
> a variable, it's more efficient to store the variable in the wider more and
> make sure the extension is done on every store.
>
> Please explain how this concept is useful for floating-point.
See the first message in this thread.
Aldy