[PATCH] D38861: [CodeGen] Error on unsupported checked multiplies early

[Vedant Kumar via Phabricator via cfe-commits]

vsk added inline comments.


================
Comment at: lib/CodeGen/CGBuiltin.cpp:2263
+      }
+    }
+
----------------
joerg wrote:
> vsk wrote:
> > rjmccall wrote:
> > > Is there a reason this only fails on x86?  If LLVM doesn't have generic wide-operation lowering, this probably needs to be a target whitelist rather than a blacklist.
> > That makes sense. For the 128-bit operation, the whitelist is {x86-64, wasm64, mips64}. We don't support this operation for bit widths larger than 128 bits on any target. I'll update the patch accordingly.
> That sounds wrong. __int128_t should be supported by all 64bit architectures, not just those three.
I didn't mean to imply generic int128_t operations aren't supported. This patch just focuses on muloti4 (signed multiplication with overflow checking).


https://reviews.llvm.org/D38861