https://gcc.gnu.org/g:cb22745e0f7dcc5b2a990169392c5ddf67f178bd
commit cb22745e0f7dcc5b2a990169392c5ddf67f178bd Author: Michael Meissner <[email protected]> Date: Fri Sep 25 01:52:18 2026 -0400 Update ChangeLog.* Diff: --- gcc/ChangeLog.float | 37 +++++++++++++++++++++++++++++++++++++ 1 file changed, 37 insertions(+) diff --git a/gcc/ChangeLog.float b/gcc/ChangeLog.float index a0eb2bb73fff..f713160f7f58 100644 --- a/gcc/ChangeLog.float +++ b/gcc/ChangeLog.float @@ -1,3 +1,40 @@ +==================== Branch work264-float, patch #212 ==================== + +Add ifunc support to truncate 32/64-bit to 16-bit fp. + +This patch rewrites the __trunc{sf,df}{bf,hf}2 functions to have ifunc support +to generate appropriate code for conversions. + +Also add _gpr versions to allow doing a conversion with 16-bit fp passed as an +unsigned short. + +2026-09-25 Michael Meissner <[email protected]> + + * config/rs6000/extendbf-sfdf2.c: Update to make the df versions use the + same code as the sf versions. + * config/rs6000/extendhf-sfdf2.c: Likewise. + * config/rs6000/truncsfdf-bf2.c: New file. + * config/rs6000/truncsfdf-hf2.c: Likewise. + * config/rs6000/t-float16: Use truncsfdf-bf2 and truncsfdf-hf2 instead + of the native soft-fp version. + +==================== Branch work264-float, patch #211 ==================== + +Add ifunc support to extend 16-bit fp to 32/64-bit. + +This patch rewrites the __extend{bf,hf}{sf,df}2 functions to have ifunc support +to generate appropriate code for conversions. + +Also add _gpr versions to allow doing a conversion with 16-bit fp passed as an +unsigned short. + +2026-09-24 Michael Meissner <[email protected]> + + * config/rs6000/extendbf-sfdf2.c: New file. + * config/rs6000/extendhf-sfdf2.c: Likewise. + * config/rs6000/t-float16: Use extendbf-sff2 and extendhf-sfhf2 instead + of the native soft-fp version. + ==================== Branch work264-float, patch #210 ==================== Add conversions between 128-bit integers and 16-bit floating point
