public inbox for gcc-cvs@sourceware.org
help / color / mirror / Atom feed
* [gcc(refs/users/meissner/heads/work107)] Update type for __float128.
@ 2023-01-27 4:00 Michael Meissner
0 siblings, 0 replies; only message in thread
From: Michael Meissner @ 2023-01-27 4:00 UTC (permalink / raw)
To: gcc-cvs
https://gcc.gnu.org/g:6e41a4798ebc78c62d53e0c0fa0f9fe2eb9bd836
commit 6e41a4798ebc78c62d53e0c0fa0f9fe2eb9bd836
Author: Michael Meissner <meissner@linux.ibm.com>
Date: Thu Jan 26 22:59:58 2023 -0500
Update type for __float128.
2022-01-26 Michael Meissner <meissner@linux.ibm.com>
gcc/
* config/rs6000/rs6000-builtin.cc (rs6000_init_builtins): Don't use long
double as the base type for __float128 if long double is IEEE 128-bit.
Use build_variant_type_copy to make a type based off of _Float128.
Diff:
---
gcc/config/rs6000/rs6000-builtin.cc | 34 +++++++++++++++-------------------
1 file changed, 15 insertions(+), 19 deletions(-)
diff --git a/gcc/config/rs6000/rs6000-builtin.cc b/gcc/config/rs6000/rs6000-builtin.cc
index 737a5c42bfb..873ae7ce875 100644
--- a/gcc/config/rs6000/rs6000-builtin.cc
+++ b/gcc/config/rs6000/rs6000-builtin.cc
@@ -730,25 +730,21 @@ rs6000_init_builtins (void)
if (TARGET_FLOAT128_TYPE)
{
- if (TARGET_IEEEQUAD && TARGET_LONG_DOUBLE_128)
- ieee128_float_type_node = long_double_type_node;
- else
- {
- /* For C we only need to register the __ieee128 name for
- it. For C++, we create a distinct type which will mangle
- differently (u9__ieee128) vs. _Float128 (DF128_) and behave
- backwards compatibly. */
- if (float128t_type_node == NULL_TREE)
- {
- float128t_type_node = make_node (REAL_TYPE);
- TYPE_PRECISION (float128t_type_node)
- = TYPE_PRECISION (float128_type_node);
- layout_type (float128t_type_node);
- SET_TYPE_MODE (float128t_type_node,
- TYPE_MODE (float128_type_node));
- }
- ieee128_float_type_node = float128t_type_node;
- }
+ /* For C we only need to register the __ieee128 name for it. For C++, we
+ create a distinct type which will mangle differently (u9__ieee128)
+ vs. _Float128 (DF128_) and behave backwards compatibly.
+
+ In the past we used the long double type if long double was IEEE
+ 128-bit. This leads to differences if _Float128 and __float128 are
+ not compatible types. In particular, it showed up in generating
+ signaling NaNs with __builtin_nansf128 and storing it into __float128,
+ which at the time was based on long double. Because they were
+ different types, the signaling part of the NaN was lost in the
+ conversion. */
+ if (float128t_type_node == NULL_TREE)
+ float128t_type_node = build_variant_type_copy (float128_type_node);
+
+ ieee128_float_type_node = float128t_type_node;
t = build_qualified_type (ieee128_float_type_node, TYPE_QUAL_CONST);
lang_hooks.types.register_builtin_type (ieee128_float_type_node,
"__ieee128");
^ permalink raw reply [flat|nested] only message in thread
only message in thread, other threads:[~2023-01-27 4:00 UTC | newest]
Thread overview: (only message) (download: mbox.gz / follow: Atom feed)
-- links below jump to the message on this page --
2023-01-27 4:00 [gcc(refs/users/meissner/heads/work107)] Update type for __float128 Michael Meissner
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).