https://github.com/eddyz87 requested changes to this pull request.
The change itself looks good to me, however it is a breaking change, e.g.:
```
$ cat test.c
struct foo { long a; long b; };
struct foo bar(int a, int b) { return (struct foo){a, b}; }
{llvm} [~/tmp]
$ clang -mcpu=v3 -O2 --target=bpf -c -o test.o test.c && disasm test.o
test.o: file format elf64-bpf
Disassembly of section .text:
0000000000000000 <bar>:
; bar():
0: w0 = w1
1: r0 <<= 0x20
2: r0 s>>= 0x20
3: w2 = w2
4: r2 <<= 0x20
5: r2 s>>= 0x20
6: exit
{llvm} [~/tmp]
$ clang -mcpu=v4 -O2 --target=bpf -c -o test.o test.c && disasm test.o
test.o: file format elf64-bpf
Disassembly of section .text:
0000000000000000 <bar>:
; bar():
0: w1 = w1
1: r0 = (s32)r1
2: w1 = w2
3: r2 = (s32)r1
4: exit
```
Should it be gated behind at-least cpuv4? (It appears that we don't want to add
v5, do we?).
https://github.com/llvm/llvm-project/pull/206876
_______________________________________________
cfe-commits mailing list
[email protected]
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
