[PATCH] D109421: [SDAG] Add intrinsic to convert between ppc_fp128 and fp128

[Eli Friedman via Phabricator via llvm-commits]

efriedma added a comment.

This could be a powerpc-specific intrinsic, maybe?  I guess that basically just shoving around the code a bit, but maybe clearer to readers.  The name is also a little ambiguous: it's not clear from the name that the target type must be fp128.

Long-term, we might want to consider replacing fpext/fptrunc instructions with something like fpconv.  We now have two pairs of floating-point types that have the same size, and if we ever add more floating-point types, the problems will just get more complicated.



================
Comment at: llvm/lib/IR/Function.cpp:1031
+  case IIT_PPCF128:
+    OutputTable.push_back(IITDescriptor::get(IITDescriptor::Quad, 0));
+    return;
----------------
Please define a new IITDescriptorKind for this.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D109421/new/

https://reviews.llvm.org/D109421