[PATCH] D79546: [VE] Implements minimum MC layer for VE (4/4)

Kazushi Marukawa via Phabricator via llvm-commits llvm-commits at lists.llvm.org
Thu May 28 18:43:35 PDT 2020


kaz7 marked 6 inline comments as done.
kaz7 added a comment.

Update inline comments.



================
Comment at: llvm/lib/Target/VE/AsmParser/VEAsmParser.cpp:622
+  switch (getLexer().getKind()) {
+  case AsmToken::LParen:
+    // FIXME: Parsing "(" + %vreg + ", " + %vreg + ")"
----------------
simoll wrote:
> kaz7 wrote:
> > simoll wrote:
> > > I suppose there should be a TODO here for the `(m)1` and `(m)0` immediates?
> > Thank you for taking a look.  The answer is no.  This todo is for a special operand of `vsld` and `vsrd` instructions.  These instructions are something like `vsld %vx, (%vy, %vz), %sy, %vm`.  I had not noticed this special operand until I try to run my regression tests through NEC assembler.  On the other hand, `(m)1` and `(m)0` immediate values will be implemented using individual `parseMImmOperand` function.
> Fair enough. What i meant is that we currently do not parse `(m)1`-style immediates and there should be an additional TODO in the code line(s) where we would call a hypothetical `parseMImmOperand` function.
I see.  Leaving comments to where I'm going to add a call of parseMImmOperand function is still difficult.  Because the call is automatically generated by tablegen.  It confuses me often also.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D79546/new/

https://reviews.llvm.org/D79546





More information about the llvm-commits mailing list