Provide assembly implementation of SHA-512 for aarch64 in libc
Provide a CE assembly implementation that can be used on hardware that
has the CE SHA-512 instructions. This is considerably faster than the
generic C implementation.
Largely borrowed from libcrypto.
With input from naddy@
ok naddy@ tb@
Provide assembly implementation of SHA-512 for amd64 in libc
Provide a generic assembly implementation for SHA-512 - this is
considerably faster than the C version.
Largely borrowed from libcrypto.
With input from naddy@
ok naddy@ tb@
Provide assembly implementation of SHA-256 for aarch64 in libc
Provide a CE assembly implementation that can be used on hardware that
has the CE SHA-256 instructions. This is considerably faster than the
generic C implementation.
Largely borrowed from libcrypto.
With input from naddy@
ok tb@
lang/gcc/16: fix ICE in gcc by compiling LLVM-22 at sparc64
Seems that sparc_legitimize_pic_address can place a LABEL_REF in
UNSPEC_MOVE_PIC and UNSPEC_MOVE_GOTDATA, while
sparc_delegitimize_address assumes that both contain a SYMBOL_REF.
So, it needs to accept LABEL_REF for the PIC cases and keep UNSPEC_TLSLE
restricted to SYMBOL_REF.
This fixes an ICE during RTL combine when compiling PIC code on sparc64.
sync from base: SparcAsmParser: fix off by one and allow %r31 to be used
rust contains a sparc64.s file with
.cfi_register %r15, %r31
This fails with 'src/arch/sparc64.s:59:26: error: invalid register name'
Changing RegNo < 31 to RegNo <= 31 fixes this.
ok kettenis kirill
SparcAsmParser: fix off by one and allow %r31 to be used
rust contains a sparc64.s file with
.cfi_register %r15, %r31
This fails with 'src/arch/sparc64.s:59:26: error: invalid register name'
Changing RegNo < 31 to RegNo <= 31 fixes this.
ok kettenis kirill
Do not use <br/> inside <pre> because it causes ugly formatting.
Bug reported long ago by Mohamed Akram <mohd.akram at outlook.com>.
Patch from Michael Stapelberg <stapelberg at debian.org>.