https://gcc.gnu.org/g:10b6b78b32e3ba53254b52b7b5da8d5b6d890fd3
commit 10b6b78b32e3ba53254b52b7b5da8d5b6d890fd3 Author: Michael Meissner <[email protected]> Date: Sat Oct 17 01:39:03 2026 -0400 Update ChangeLog.* Diff: --- gcc/ChangeLog.float | 23 +++++++++++++++++++++++ 1 file changed, 23 insertions(+) diff --git a/gcc/ChangeLog.float b/gcc/ChangeLog.float index ee69516691e6..afd40e5278e1 100644 --- a/gcc/ChangeLog.float +++ b/gcc/ChangeLog.float @@ -1,3 +1,26 @@ +==================== Branch work266-float, patch #229 ==================== + +Add bfloat16 define_peephole2. + +2026-10-17 Michael Meissner <[email protected]> + + * config/rs6000/float16.md (bfloat16 define_peephole2): Add a peephole2 + to try to catch some cases where bfloat16 uses vector V4SFmode + arithmetic for some things and then converts the value to scalar SFmode + for other calculations. + * config/rs6000/rs6000.opt (-mfloat16-peephole2): New debug switch. + +==================== Branch work266-float, patch #228 ==================== + +Add -mdebug=extra1 through -mdebug=extra5 + +2026-10-15 Michael Meissner <[email protected]> + + * common/config/rs6000/rs6000rcommon.cc (rs6000_handle_option): Add + support for -mdebug=estra{1,2,3,4,5}. + * config/rs6000/rs6000.h (MASK_DEBUG_*): Likewise. + (TARGET_DEBUG_*): Likewise. + ==================== Branch work266-float, patch #227 ==================== Use xxspltib rather than 2 Altivec instructions.
