VyacheslavLevytskyy wrote:
LGTM, just some observations.
There are a few __spirv_event_t cases that would be nice to cover in the tests.
1) A test for a non-default address space.
For example,
`__spirv_event_t __attribute__((address_space(7))) e;`
lowers to
`@e = addrspace(7) global target("spirv.Event") zeroinitializer, align 8`
but no test currently exercises __attribute__((address_space(...))) on
__spirv_event_t.
2) Given that address-space-qualified declarations are accepted, it would also
be good to cover the associated sizeof/alignof behavior.
Size and alignment are computed using the default address-space pointer
properties, independent of the declaration's address space. That's fine today
because the SPIR-V target uses a single pointer width/alignment for all address
spaces.
So, probably it would be nice to cover sizeof/alignof for an
address-space-qualified __spirv_event_t.
3) Arrays of __spirv_event_t (__spirv_event_t list[N]) also seem worth
covering, as a common way of representing event lists.
https://github.com/llvm/llvm-project/pull/207077
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
