https://gcc.gnu.org/bugzilla/show_bug.cgi?id=92071
Jakub Jelinek <jakub at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
CC| |ebotcazou at gcc dot gnu.org
--- Comment #8 from Jakub Jelinek <jakub at gcc dot gnu.org> ---
The ICE started with r10-2840-g70cdb21e579191fe9f0f1d45e328908e59c0179e
but we generated silently wrong-code before that, I believe since
r7-5812-ga271e415611a80f1e86e625fd61360e193d04474
ldr r3, .L3
ldr r3, [r3]
ldm r3, {r0, r1}
b bar
is what we generate with that r244249 and up, which likely doesn't handle if a
is not 8 (or just 4) byte aligned.
Before that we emitted
push {r4, r5, lr}
ldr r3, .L4
ldr r3, [r3]
ldrb r0, [r3] @ zero_extendqisi2
ldrb r1, [r3, #5] @ zero_extendqisi2
ldrb ip, [r3, #1] @ zero_extendqisi2
ldrb r2, [r3, #4] @ zero_extendqisi2
ldrb r5, [r3, #2] @ zero_extendqisi2
ldrb r4, [r3, #6] @ zero_extendqisi2
orr r2, r2, r1, lsl #8
ldrb lr, [r3, #3] @ zero_extendqisi2
orr ip, r0, ip, lsl #8
ldrb r1, [r3, #7] @ zero_extendqisi2
orr r0, ip, r5, lsl #16
orr r2, r2, r4, lsl #16
orr r0, r0, lr, lsl #24
orr r1, r2, r1, lsl #24
pop {r4, r5, lr}
b bar
which clearly can handle unaligned load.
