nikola closed this revision.
nikola added a comment.
r281298
https://reviews.llvm.org/D24193
___
cfe-commits mailing list
cfe-commits@lists.llvm.org
http://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
hans accepted this revision.
hans added a comment.
This revision is now accepted and ready to land.
lgtm
I'm a little bit worried that we've started down a slippery slope of deciding
what is and isn't allowed in naked functions (for example, some initializers
could be allowed if we're sure
nikola updated this revision to Diff 70985.
nikola added a comment.
This should address Hans' comments, as for the code get I have no idea. I was
hoping someone more knowledgeable would tell me if this made sense or not?
https://reviews.llvm.org/D24193
Files:
lib/Sema/SemaDecl.cpp
rnk added a comment.
Won't the mid-level optimizer turn these register variables into undefs, and
we'll end up with this kind of IR?
void f() {
int register var asm ("eax") ;
asm volatile ("add %%eax, %0\n\tret" : : "r"(var));
}
->
define void @"\01?f@@YAXXZ"()
hans added a comment.
I think this is reasonable. Just a few comments:
Comment at: lib/Sema/SemaDecl.cpp:11796
@@ +11795,3 @@
+if (auto *DS = dyn_cast(S)) {
+ if (DS->isSingleDecl()) {
+if (auto *Var = dyn_cast_or_null(DS->getSingleDecl())) {
nikola created this revision.
nikola added reviewers: hans, rnk, compnerd.
nikola added a subscriber: cfe-commits.
Herald added a subscriber: aemerson.
I think the current mode is too restrictive, it will emit error for any
statement inside a naked function. Code I'm trying to compile for ARM
