On 6/19/2017 6:08 AM, Matthieu Bouron wrote: > Avoids overriding v0 (which containins the result of the tested > function) in checkasm_call_checked. > > Also properly calls checkasm_call_checked. > --- > tests/checkasm/aarch64/checkasm.S | 8 ++++---- > tests/checkasm/checkasm.h | 2 ++ > 2 files changed, 6 insertions(+), 4 deletions(-) > > diff --git a/tests/checkasm/aarch64/checkasm.S > b/tests/checkasm/aarch64/checkasm.S > index 53a2a478dc..75a9a56143 100644 > --- a/tests/checkasm/aarch64/checkasm.S > +++ b/tests/checkasm/aarch64/checkasm.S > @@ -112,10 +112,10 @@ function checkasm_checked_call, export=1 > movi v3.8h, #0 > > .macro check_reg_neon reg1, reg2 > - ldr q0, [x9], #16 > - uzp1 v1.2d, v\reg1\().2d, v\reg2\().2d > - eor v0.16b, v0.16b, v1.16b > - orr v3.16b, v3.16b, v0.16b > + ldr q1, [x9], #16 > + uzp1 v2.2d, v\reg1\().2d, v\reg2\().2d > + eor v1.16b, v1.16b, v2.16b > + orr v3.16b, v3.16b, v1.16b > .endm > check_reg_neon 8, 9 > check_reg_neon 10, 11 > diff --git a/tests/checkasm/checkasm.h b/tests/checkasm/checkasm.h > index 695d871220..5249669fba 100644 > --- a/tests/checkasm/checkasm.h > +++ b/tests/checkasm/checkasm.h > @@ -145,6 +145,8 @@ void checkasm_stack_clobber(uint64_t clobber, ...); > void checkasm_checked_call(void *func, ...); > #define declare_new(ret, ...) ret (*checked_call)(void *, int, int, int, > int, int, int, int, __VA_ARGS__)\ > = (void *)checkasm_checked_call; > +#define declare_new_float(ret, ...) ret (*checked_call)(void *, int, int, > int, int, int, int, int, __VA_ARGS__)\ > + = (void *)checkasm_checked_call;
Isn't this doing the same as the generic "#define declare_new_float" about 15 lines below? If declare_new_float() is no different than declare_new() for a given target, then just let the aforementioned check handle it. > #define CLOB (UINT64_C(0xdeadbeefdeadbeef)) > #define call_new(...) > (checkasm_stack_clobber(CLOB,CLOB,CLOB,CLOB,CLOB,CLOB,CLOB,CLOB,CLOB,CLOB,CLOB,CLOB,\ > > CLOB,CLOB,CLOB,CLOB,CLOB,CLOB,CLOB,CLOB,CLOB,CLOB,CLOB),\ > Should be ok either way if tested (The only aarch64 FATE client using git head hasn't run in twenty days, so no way for me to check if this fixes the problem anyway). _______________________________________________ ffmpeg-devel mailing list ffmpeg-devel@ffmpeg.org http://ffmpeg.org/mailman/listinfo/ffmpeg-devel
