ast added inline comments.
================
Comment at: llvm/lib/Target/BPF/BPFMIChecking.cpp:199
+ unsigned newOpcode;
+ switch(MI.getOpcode()) {
+ case BPF::XFADDW32: newOpcode = BPF::XADDW32; break;
----------------
yonghong-song wrote:
> yonghong-song wrote:
> > ast wrote:
> > > With this logic in place Andrii has a point. There is no need for
> > > -mcpu=v4 flag.
> > > The builtins will produce new insns and it will be up to kernel to accept
> > > them or not.
> > will make the change (removing -mcpu=v4).
> There will be one user-level change:
> . if user has "ret = __sync_fetch_and_add(p, r);", today's llvm will
> generate a fatal error saying that invalid XADD since return value is used.
> . but removing -mcpu=v4. the compilation will be successful and the error
> will happen in kernel since atomic_fetch_add is not supported.
>
> I guess this is an okay change, right?
right. With new llvm compilation will succeed, but it will fail to load on
older kernel and will work as expected on newer kernel. There is a change where
the error will be seen, but that's acceptable.
Repository:
rG LLVM Github Monorepo
CHANGES SINCE LAST ACTION
https://reviews.llvm.org/D72184/new/
https://reviews.llvm.org/D72184
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
