Hi, I wasn't aware of that patch. The only difference between my patch and Andreas are that I have added tests to test encoding of doubles as well as float ones. Regards, Branislav ________________________________ From: Jan Beulich Sent: Thursday, June 15, 2023 2:09 PM To: Branislav Brzak Cc: Dragoslav Sicarov ; Djordje Todorovic ; binutils@sourceware.org Subject: Re: [PATCH 0/2] riscv: Fix gas when encoding BE floats/doubles On 15.06.2023 13:09, Branislav Brzak wrote: > RISCV gas currently does not take into consideration -mbig-endian > directive when encoding floats and doubles. The effect is that when > -mbig-endian is supplied floats and doubles are endian flipped only > on 2-byte boundaries, for example: > > .data: >       .float 1.1 > > Compiled as LE will be encoded as: 0xcdcc8c3f > And BE encoded will be: 0x8c3fcdcc > > The correct sequence is 0x3f8ccccd > > This patch introduces a fix in float/double endian encoding and > set of tests that verify validity of LE and BE float and double > encoding. > > Branislav Brzak (2): >       gas: Fix float/double BE encoding in RISCV. >       riscv: Add LE/BE float/double encoding tests. > > gas/config/tc-riscv.c > gas/testsuite/gas/riscv/double-be.d | 18 ++++++++++++++++++ > gas/testsuite/gas/riscv/double-le.d | 18 ++++++++++++++++++ > gas/testsuite/gas/riscv/double.s | 21 +++++++++++++++++++++ > gas/testsuite/gas/riscv/float-be.d | 13 +++++++++++++ > gas/testsuite/gas/riscv/float-le.d | 13 +++++++++++++ > gas/testsuite/gas/riscv/float.s | 21 +++++++++++++++++++++ > 7 files changed, 105 insertions(+), 1 deletion(-) Did you see https://sourceware.org/pipermail/binutils/2023-June/127845.html ? (Only the cover letter came through so far, so I can't tell yet whether you do anything different than Andreas does. Also, as an aside, it typically helps if you Cc arch maintainers for arch patches.) Jan