Hi!
On Fri, Nov 22, 2019 at 01:26:39PM +0800, luoxhu wrote:
> Update the code as you wish, Thanks:
The point is to make this interface easy and clear to use. So please
tell me what *you* think about that, don't just do what I think may be
a good solution!
> * gcc.target/powerpc/pr72804.c: Split the store function to...
> * gcc.target/powerpc/pr72804-1.c: ... this one. New.
Ah, good idea.
> +++ b/gcc/testsuite/gcc.target/powerpc/pr72804-1.c
> +/* store generates difference instructions as below:
> + P9: mtvsrdd;xxlnot;stxv.
> + P8/P7/P6 LE: not;not;std;std.
> + P8 BE: mtvsrd;mtvsrd;xxpermdi;xxlnor;stxvd2x.
> + P7/P6 BE: std;std;addi;lxvd2x;xxlnor;stxvd2x. */
But it should generate just the first or second. So just document that,
and base the tests around that as well?
> +/* { dg-final { scan-assembler-times {\mmtvsrdd\M} 1 { target power9+ } } }
> */
> +/* { dg-final { scan-assembler-times {\mxxlnor\M} 1 { target power9+ } } } */
> +/* { dg-final { scan-assembler-times {\mstxv\M} 1 { target power9+ } } } */
> +
> +/* { dg-final { scan-assembler-times {\mnot\M} 2 { xfail {! { {! power9+} &&
> {le} } } } } } */
> +
> +/* { dg-final { scan-assembler-times {\mstd\M} 2 { xfail { { {power8} &&
> {be} } || {power9+} } } } } */
These shouldn't xfail for p9 and later: it is an *unexpected* failure there.
Maybe it is easier to duplicate this test? One for p9+ and one for ! p9+?
> +# Return 1 if we're generating code for only power8 platforms.
> +
> +proc check_effective_target_power8 { } {
> + return [check_no_compiler_messages_nocache power8 assembly {
> + #if !(!defined(_ARCH_PWR9) && defined(_ARCH_PWR8))
> + #error NO
> + #endif
> + } ""]
> +}
Do we need this?
> +# Return 1 if we're generating code for power9 and future platforms.
> +
> +proc check_effective_target_power9+ { } {
> + return [check_no_compiler_messages_nocache power9+ assembly {
> + #if !(defined(_ARCH_PWR9))
> + #error NO
> + #endif
> + } ""]
> +}
Maybe it is useful to have even shorter names, p9+, since room is scarce
where it is used (just like le and be). But maybe it doesn't matter?
Segher
