On 8/16/23 02:40, yanzhang.wang--- via Gcc-patches wrote:
From: Yanzhang Wang <yanzhang.w...@intel.com>
The pattern is enabled for scalar but not for vector. The patch try to
make it consistent and will convert below code,
shortcut_for_riscv_vrsub_case_1_32:
vl1re32.v v1,0(a1)
vsetvli zero,a2,e32,m1,ta,ma
vrsub.vi v1,v1,-1
vs1r.v v1,0(a0)
ret
to,
shortcut_for_riscv_vrsub_case_1_32:
vl1re32.v v1,0(a1)
vsetvli zero,a2,e32,m1,ta,ma
vnot.v v1,v1
vs1r.v v1,0(a0)
ret
gcc/ChangeLog:
* simplify-rtx.cc (simplify_context::simplify_binary_operation_1):
Get -1 with mode.
gcc/testsuite/ChangeLog:
* gcc.target/riscv/rvv/base/simplify-vrsub.c: New test.
Just a note. It is customary to indicate what testing you did for each
patch. A patch which changes target independent code should be
bootstrapped and regression tested on at least one major target (most
folks use x86_64 or aarch64).
If you change target code it is customary to run the testsuite on that
target. Ideally that would include a bootstrap and regression test, but
that's not always possible (cross compilers) in which case you just
build the toolchain and run the cross tests.
I went ahead and bootstrapped & regression tested this on
x86_64-linux-gnu where it passed without regressions.
I'll push this to the trunk.
Thanks,
jeff