andykaylor wrote:
@bcardosolopes, I have some general concerns about the way that we're
representing bitfields and bitfield accesses. It seems fundamentally
inconsistent with the way we represent unions. For example:
```
union U {
char a;
int b;
long c;
};
```
becomes
`!cir.record<union "U" {!s8i, !s32i, !s64i}`
but
```
struct S {
int x;
unsigned a:2;
unsigned b:5;
unsigned c:7;
} s;
```
becomes
`!cir.record<struct "S" {!s32i, !u16i}`
And all three bitfield members are hidden away inside the second member,
leading to the need for two operations to access a bitfield member:
```
%3 = cir.get_member %2[1] {name = "b"} : !cir.ptr<!rec_S> ->
!cir.ptr<!u16i> loc(#loc11)
%4 = cir.get_bitfield(#bfi_b, %3 : !cir.ptr<!u16i>) -> !u32i loc(#loc12)
```
Why are we representing each member of a union individually but packing
bitfields away in integer fields?
What would you think of something like this?
`!cir.record<struct "S" {!s32i, !cir.bitfield<2>, !cir.bitfield<5>,
!cir.bitfield<7>}`
That puts a bit more burden on the `RecordType::getElementOffset`
implementation, but I think it would let us defer some of the details to
lowering.
https://github.com/llvm/llvm-project/pull/142041
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
