https://gcc.gnu.org/bugzilla/show_bug.cgi?id=80239
Bug ID: 80239
Summary: [7 regression] 9% regression on dhrystone when
targetting Cortex-M7
Product: gcc
Version: 7.0
Status: UNCONFIRMED
Keywords: missed-optimization
Severity: normal
Priority: P3
Component: rtl-optimization
Assignee: unassigned at gcc dot gnu.org
Reporter: thopre01 at gcc dot gnu.org
Target Milestone: ---
Target: arm-none-eabi
Hi,
Performance for dhrystone has regressed by 9% after commit r235184 when
targeting Cortex-M7 with -O3 -fno-inline. I've tracked this to a couple of
changes likes the following:
- add r0, sp, #24
+ add r5, sp, #16
str r1, [lr, #4] @ unaligned
str r2, [lr, #8] @ unaligned
- mov r5, r0
- str r0, [sp, #4]
Where [sp, #4] appears to never be loaded again. The mov itself is problematic
as well as removing all the extra str still made the code slower.
