* Split constant handling out of fold_builtin_fma
@ 2015-10-26 9:58 Richard Sandiford
2015-10-26 10:45 ` Richard Biener
0 siblings, 1 reply; 2+ messages in thread
From: Richard Sandiford @ 2015-10-26 9:58 UTC (permalink / raw)
To: gcc-patches
Just makes an upcoming patch a bit clearer.
Tested on x86_64-linux-gnu, aarch64-linux-gnu and arm-linux-gnueabi.
OK to install?
Thanks,
Richard
gcc/
* builtins.c (fold_builtin_fma): Remove constant handling.
(fold_builtin_3): Handle constant fma arguments here.
diff --git a/gcc/builtins.c b/gcc/builtins.c
index a03dffc..6cd8879 100644
--- a/gcc/builtins.c
+++ b/gcc/builtins.c
@@ -7881,18 +7881,13 @@ fold_fma (location_t loc ATTRIBUTE_UNUSED,
static tree
fold_builtin_fma (location_t loc, tree arg0, tree arg1, tree arg2, tree type)
{
+ /* ??? Only expand to FMA_EXPR if it's directly supported. */
if (validate_arg (arg0, REAL_TYPE)
&& validate_arg (arg1, REAL_TYPE)
- && validate_arg (arg2, REAL_TYPE))
- {
- tree tem = fold_fma (loc, type, arg0, arg1, arg2);
- if (tem)
- return tem;
+ && validate_arg (arg2, REAL_TYPE)
+ && optab_handler (fma_optab, TYPE_MODE (type)) != CODE_FOR_nothing)
+ return fold_build3_loc (loc, FMA_EXPR, type, arg0, arg1, arg2);
- /* ??? Only expand to FMA_EXPR if it's directly supported. */
- if (optab_handler (fma_optab, TYPE_MODE (type)) != CODE_FOR_nothing)
- return fold_build3_loc (loc, FMA_EXPR, type, arg0, arg1, arg2);
- }
return NULL_TREE;
}
@@ -9307,8 +9302,9 @@ fold_builtin_3 (location_t loc, tree fndecl,
return fold_builtin_sincos (loc, arg0, arg1, arg2);
CASE_FLT_FN (BUILT_IN_FMA):
+ if (tree tem = fold_fma (loc, type, arg0, arg1, arg2))
+ return tem;
return fold_builtin_fma (loc, arg0, arg1, arg2, type);
- break;
CASE_FLT_FN (BUILT_IN_REMQUO):
if (validate_arg (arg0, REAL_TYPE)
^ permalink raw reply [flat|nested] 2+ messages in thread
* Re: Split constant handling out of fold_builtin_fma
2015-10-26 9:58 Split constant handling out of fold_builtin_fma Richard Sandiford
@ 2015-10-26 10:45 ` Richard Biener
0 siblings, 0 replies; 2+ messages in thread
From: Richard Biener @ 2015-10-26 10:45 UTC (permalink / raw)
To: GCC Patches, richard.sandiford
On Mon, Oct 26, 2015 at 10:57 AM, Richard Sandiford
<richard.sandiford@arm.com> wrote:
> Just makes an upcoming patch a bit clearer.
>
> Tested on x86_64-linux-gnu, aarch64-linux-gnu and arm-linux-gnueabi.
> OK to install?
Ok.
Thanks,
Richard.
> Thanks,
> Richard
>
>
> gcc/
> * builtins.c (fold_builtin_fma): Remove constant handling.
> (fold_builtin_3): Handle constant fma arguments here.
>
> diff --git a/gcc/builtins.c b/gcc/builtins.c
> index a03dffc..6cd8879 100644
> --- a/gcc/builtins.c
> +++ b/gcc/builtins.c
> @@ -7881,18 +7881,13 @@ fold_fma (location_t loc ATTRIBUTE_UNUSED,
> static tree
> fold_builtin_fma (location_t loc, tree arg0, tree arg1, tree arg2, tree type)
> {
> + /* ??? Only expand to FMA_EXPR if it's directly supported. */
> if (validate_arg (arg0, REAL_TYPE)
> && validate_arg (arg1, REAL_TYPE)
> - && validate_arg (arg2, REAL_TYPE))
> - {
> - tree tem = fold_fma (loc, type, arg0, arg1, arg2);
> - if (tem)
> - return tem;
> + && validate_arg (arg2, REAL_TYPE)
> + && optab_handler (fma_optab, TYPE_MODE (type)) != CODE_FOR_nothing)
> + return fold_build3_loc (loc, FMA_EXPR, type, arg0, arg1, arg2);
>
> - /* ??? Only expand to FMA_EXPR if it's directly supported. */
> - if (optab_handler (fma_optab, TYPE_MODE (type)) != CODE_FOR_nothing)
> - return fold_build3_loc (loc, FMA_EXPR, type, arg0, arg1, arg2);
> - }
> return NULL_TREE;
> }
>
> @@ -9307,8 +9302,9 @@ fold_builtin_3 (location_t loc, tree fndecl,
> return fold_builtin_sincos (loc, arg0, arg1, arg2);
>
> CASE_FLT_FN (BUILT_IN_FMA):
> + if (tree tem = fold_fma (loc, type, arg0, arg1, arg2))
> + return tem;
> return fold_builtin_fma (loc, arg0, arg1, arg2, type);
> - break;
>
> CASE_FLT_FN (BUILT_IN_REMQUO):
> if (validate_arg (arg0, REAL_TYPE)
>
^ permalink raw reply [flat|nested] 2+ messages in thread
end of thread, other threads:[~2015-10-26 10:42 UTC | newest]
Thread overview: 2+ messages (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2015-10-26 9:58 Split constant handling out of fold_builtin_fma Richard Sandiford
2015-10-26 10:45 ` Richard Biener
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).