https://gcc.gnu.org/g:f2044e1ced5cb21ba2eb2ffa8ce70e3958b7f430
commit f2044e1ced5cb21ba2eb2ffa8ce70e3958b7f430
Author: Michael Meissner <meiss...@linux.ibm.com>
Date: Thu Aug 14 02:21:48 2025 -0400
Update ChangeLog.*
Diff:
---
gcc/ChangeLog.float | 24 ++++++++++++++++++++++++
1 file changed, 24 insertions(+)
diff --git a/gcc/ChangeLog.float b/gcc/ChangeLog.float
index 56cd7dfb23b3..0d08970aacd3 100644
--- a/gcc/ChangeLog.float
+++ b/gcc/ChangeLog.float
@@ -1,3 +1,27 @@
+==================== Branch work219-float, patch #302 ====================
+
+Pass _Float16 arguments and return values in GPRs
+
+2025-08-14 Michael Meissner <meiss...@linux.ibm.com>
+
+gcc/
+
+ PR target/121525
+ * config/rs6000/rs6000-call.cc (USE_FP_FOR_ARG_P): Don't pass _Float16
+ arguments in vector registers unless -mieee16-gpr-args.
+ (rs6000_return_in_memory): Remove previous code possibly returning
+ _Float16 values in memory and passing arguments in memory.
+ (rs6000_pass_by_reference): Likewise.
+ (rs6000_function_value): Return _Float16 values in GPRs unless
+ -mno-ieee16-gpr-args is used.
+ * config/rs6000/rs6000.cc (rs6000_function_value): Return _Float16
+ variables in GPRs unless -mno-ieee16-gpr-args.
+ * config/rs6000/rs6000.md (movhf_xxspltiw): Add support for loading
+ _Float16 constants in GPRs.
+ * config/rs6000/rs6000.opt (-mieee16-return-in-memory): Delete.
+ (-mieee16-pass-by-reference): Likewise.
+ (-mieee16-gpr-args): New options.
+
==================== Branch work219-float, patch #301 ====================
Pass _Float16 arguments and return values in memory
