https://gcc.gnu.org/bugzilla/show_bug.cgi?id=122974
Bug ID: 122974
Summary: bfloat16_simd tests fail when forcing -mtune for an
M-class CPU
Product: gcc
Version: 16.0
URL: https://linaro.atlassian.net/browse/GNU-1673
Status: UNCONFIRMED
Keywords: testsuite-fail
Severity: normal
Priority: P3
Component: target
Assignee: unassigned at gcc dot gnu.org
Reporter: clyon at gcc dot gnu.org
Target Milestone: ---
Target: arm
When running the testsuite with
-mthumb/-march=armv6s-m/-mtune=cortex-m0/-mfloat-abi=soft/-mfpu=auto we have:
FAIL: gcc.target/arm/bfloat16_simd_1_2.c check-function-bodies stacktest1
FAIL: gcc.target/arm/bfloat16_simd_2_2.c check-function-bodies stacktest1
FAIL: gcc.target/arm/bfloat16_simd_3_2.c check-function-bodies stacktest1
For instance bfloat16_simd_1_2.c is compiled with
-mthumb -march=armv6s-m -mtune=cortex-m0 -mfloat-abi=soft -mfpu=auto
-march=armv8.2-a+bf16 -mfloat-abi=softfp -mfpu=auto
and the generated code is:
stacktest1:
sub sp, sp, #8
strh r0, [sp, #6] @ __bf16
ldrh r0, [sp, #6] @ __bf16
add r3, sp, #6
add sp, sp, #8
bx lr
so the test fails because we generate a store to sp + 6.
Adding -mtune=cortex-a57 generates:
stacktest1:
sub sp, sp, #8
add r3, sp, #6
strh r0, [r3] @ __bf16
ldrh r0, [sp, #6] @ __bf16
add sp, sp, #8
bx lr
It's not clear whether we'd want to update the test to accept a store to sp+6,
or if we want to force a suitable -mtune value.
In addition the code generated without -mtune=cortex-a57 contains a useless:
add r3, sp, #6
