https://gcc.gnu.org/bugzilla/show_bug.cgi?id=127004
Bug ID: 127004
Summary: [17 Regression] Recent changes to lower long-multipy
chagnes to inline regress on avr-elf
Product: gcc
Version: 17.0
Status: UNCONFIRMED
Severity: normal
Priority: P3
Component: target
Assignee: unassigned at gcc dot gnu.org
Reporter: law at gcc dot gnu.org
Target Milestone: ---
This change:
commit aac02c9666184d19334a0527ec444b5e55fbd113
Author: Konstantinos Eleftheriou <[email protected]>
Date: Thu Jul 9 02:22:54 2026 -0700
widening_mul: Lower long-multiply chains to inline longhand
The forwprop long-multiply recognizer canonicalizes longhand
high-part multiplies into
lhs = (N) (((2N) op1 * (2N) op2) >> N)
[ ... ]
Is causing an avr-elf regression:
Tests that now fail, but worked before (5 tests):
avr-sim: gcc: gcc.c-torture/execute/930921-1.c -O2 (test for excess errors)
avr-sim: gcc: gcc.c-torture/execute/930921-1.c -O2 -flto
-fno-use-linker-plugin -flto-partition=none (test for excess errors)
avr-sim: gcc: gcc.c-torture/execute/930921-1.c -O3 -fomit-frame-pointer
-funroll-loops -fpeel-loops -ftracer -finline-functions (test for excess
errors)
avr-sim: gcc: gcc.c-torture/execute/930921-1.c -O3 -g (test for excess
errors)
avr-sim: gcc: gcc.c-torture/execute/930921-1.c -Os (test for excess errors)
It seems to be triggering an ICE in the AVR backend, so most likely it's a
target issue rather than something generic. THe good news is it's an ICE, so
it should be possible to reproduce/test without a full avr-elf environment.
I'm not currently working on this.
