* x86 fixuns_truncsfsi2 and expand_fix
@ 1999-04-09 20:47 John Wehle
1999-04-10 2:05 ` Jeffrey A Law
1999-04-30 23:15 ` John Wehle
0 siblings, 2 replies; 4+ messages in thread
From: John Wehle @ 1999-04-09 20:47 UTC (permalink / raw)
To: egcs
The x86 can convert from a float to a signed integer. It does not
support converting a float to an unsigned integer. The x86 port
currently defines expanders for fixuns_truncsfsi2 which work by
doing a signed conversion to DImode and then taking the lower half.
Why is it useful for a port to provide these expanders considering
that expand_fix in optabs.c is also capable of converting a float to
an unsigned integer in the same fashion without these expanders being
present?
-- John
-------------------------------------------------------------------------
| Feith Systems | Voice: 1-215-646-8000 | Email: john@feith.com |
| John Wehle | Fax: 1-215-540-5495 | |
-------------------------------------------------------------------------
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: x86 fixuns_truncsfsi2 and expand_fix
1999-04-09 20:47 x86 fixuns_truncsfsi2 and expand_fix John Wehle
@ 1999-04-10 2:05 ` Jeffrey A Law
1999-04-30 23:15 ` Jeffrey A Law
1999-04-30 23:15 ` John Wehle
1 sibling, 1 reply; 4+ messages in thread
From: Jeffrey A Law @ 1999-04-10 2:05 UTC (permalink / raw)
To: John Wehle; +Cc: egcs
In message < 199904100347.XAA03073@jwlab.FEITH.COM >you write:
> The x86 can convert from a float to a signed integer. It does not
> support converting a float to an unsigned integer. The x86 port
> currently defines expanders for fixuns_truncsfsi2 which work by
> doing a signed conversion to DImode and then taking the lower half.
> Why is it useful for a port to provide these expanders considering
> that expand_fix in optabs.c is also capable of converting a float to
> an unsigned integer in the same fashion without these expanders being
> present?
None that I'm aware of. It may be from a time when expand_fix couldn't handle
that case. I remember that code went through some major rewriting in the early
90s.
I recommend killing the expanders if expand_fix can generate code that is
just as efficient as what i386.md generates.
jeff
^ permalink raw reply [flat|nested] 4+ messages in thread
* x86 fixuns_truncsfsi2 and expand_fix
1999-04-09 20:47 x86 fixuns_truncsfsi2 and expand_fix John Wehle
1999-04-10 2:05 ` Jeffrey A Law
@ 1999-04-30 23:15 ` John Wehle
1 sibling, 0 replies; 4+ messages in thread
From: John Wehle @ 1999-04-30 23:15 UTC (permalink / raw)
To: egcs
The x86 can convert from a float to a signed integer. It does not
support converting a float to an unsigned integer. The x86 port
currently defines expanders for fixuns_truncsfsi2 which work by
doing a signed conversion to DImode and then taking the lower half.
Why is it useful for a port to provide these expanders considering
that expand_fix in optabs.c is also capable of converting a float to
an unsigned integer in the same fashion without these expanders being
present?
-- John
-------------------------------------------------------------------------
| Feith Systems | Voice: 1-215-646-8000 | Email: john@feith.com |
| John Wehle | Fax: 1-215-540-5495 | |
-------------------------------------------------------------------------
^ permalink raw reply [flat|nested] 4+ messages in thread
* Re: x86 fixuns_truncsfsi2 and expand_fix
1999-04-10 2:05 ` Jeffrey A Law
@ 1999-04-30 23:15 ` Jeffrey A Law
0 siblings, 0 replies; 4+ messages in thread
From: Jeffrey A Law @ 1999-04-30 23:15 UTC (permalink / raw)
To: John Wehle; +Cc: egcs
In message < 199904100347.XAA03073@jwlab.FEITH.COM >you write:
> The x86 can convert from a float to a signed integer. It does not
> support converting a float to an unsigned integer. The x86 port
> currently defines expanders for fixuns_truncsfsi2 which work by
> doing a signed conversion to DImode and then taking the lower half.
> Why is it useful for a port to provide these expanders considering
> that expand_fix in optabs.c is also capable of converting a float to
> an unsigned integer in the same fashion without these expanders being
> present?
None that I'm aware of. It may be from a time when expand_fix couldn't handle
that case. I remember that code went through some major rewriting in the early
90s.
I recommend killing the expanders if expand_fix can generate code that is
just as efficient as what i386.md generates.
jeff
^ permalink raw reply [flat|nested] 4+ messages in thread
end of thread, other threads:[~1999-04-30 23:15 UTC | newest]
Thread overview: 4+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
1999-04-09 20:47 x86 fixuns_truncsfsi2 and expand_fix John Wehle
1999-04-10 2:05 ` Jeffrey A Law
1999-04-30 23:15 ` Jeffrey A Law
1999-04-30 23:15 ` John Wehle
This is a public inbox, see mirroring instructions
for how to clone and mirror all data and code used for this inbox;
as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).