On 2/19/24 21:26, Alexandre Oliva wrote:
This backport for gcc-13 is required for pr90838.c to get the expected
count of andi instructions on riscv64-elf, rather than fail because of
two extra andi insns in functions where it is not necessary. (On
riscv32-elf, it passes). Regstrapped on x86_64-linux-gnu, along with
other backports, and tested manually on riscv64-elf. Ok to install?
From: Raphael Moreira Zinsly <rzin...@ventanamicro.com>
Changes since v1:
- Remove subreg from operand 1.
-- >8 --
We were not able to match the CTZ sign extend pattern on RISC-V
because it gets optimized to zero extend and/or to ANDI patterns.
For the ANDI case, combine scrambles the RTL and generates the
extension by using subregs.
gcc/ChangeLog:
PR target/106888
* config/riscv/bitmanip.md
(<bitmanip_optab>disi2): Match with any_extend.
(<bitmanip_optab>disi2_sext): New pattern to match
with sign extend using an ANDI instruction.
gcc/testsuite/ChangeLog:
PR target/106888
* gcc.target/riscv/pr106888.c: New test.
* gcc.target/riscv/zbbw.c: Check for ANDI.
In general, shouldn't backports be focused on correctness issues? It's
unclear what the motivation is for backporting this change into gcc-13.
Not objecting, trying understand at this stage.
Jeff
