https://gcc.gnu.org/bugzilla/show_bug.cgi?id=127166
Bug ID: 127166
Summary: [17 Regression] Recent saturation/clipping match.pd
changes regressing on RISC-V
Product: gcc
Version: 17.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: tree-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: law at gcc dot gnu.org
Target Milestone: ---
I *think* this stuff was working in gcc-16, so this represents a regression
with gcc-17.
This change:
commit 3c5cc8819fe03b63cba9ff82569623d092fdf0f5 (HEAD)
Author: Kyrylo Tkachov <[email protected]>
Date: Wed Aug 19 11:54:14 2026 +0200
Match: Reject NARROW_CLIP saturating truncation at the minimum [PR126981]
[ ... ]
Is causing several failures on RISC-V:
Tests that now fail, but worked before (6 tests):
unix/-march=rv64gc_zba_zbb_zbs_zicond: gcc:
gcc.target/riscv/rvv/autovec/pr120378-1.c -O3 -ftree-vectorize
scan-assembler-times vnclipu\\.wi 2
unix/-march=rv64gc_zba_zbb_zbs_zicond: gcc:
gcc.target/riscv/rvv/autovec/pr120378-1.c -O3 -ftree-vectorize
scan-tree-dump-times optimized ".SAT_TRUNC " 1
unix/-march=rv64gc_zba_zbb_zbs_zicond: gcc:
gcc.target/riscv/rvv/autovec/pr120378-1.c -O3 -ftree-vectorize
scan-tree-dump-times optimized "MAX_EXPR " 1
unix/-march=rv64gc_zba_zbb_zbs_zicond: gcc:
gcc.target/riscv/rvv/autovec/pr120378-2.c -O3 -ftree-vectorize
scan-assembler-times vnclipu\\.wi 1
unix/-march=rv64gc_zba_zbb_zbs_zicond: gcc:
gcc.target/riscv/rvv/autovec/pr120378-3.c -O3 -ftree-vectorize
scan-assembler-times vnclipu\\.wi 3
unix/-march=rv64gc_zba_zbb_zbs_zicond: gcc:
gcc.target/riscv/rvv/autovec/pr120378-4.c -O3 -ftree-vectorize
scan-assembler-times vnclipu\\.wi 2
Just perusing the generated code for pr120378-1, the old code looks
considerably better.
I haven't done any debugging on this other than bisection.
