[PATCH] D53137: Scalable vector core instruction support + size queries

[Diana Picus via Phabricator via llvm-commits]

rovka accepted this revision.
rovka marked 2 inline comments as done.
rovka added a comment.
This revision is now accepted and ready to land.

This looks good to me, maybe wait a while to see if anyone else has any further comments.



================
Comment at: llvm/include/llvm/IR/DataLayout.h:454
+    auto BaseSize = getTypeSizeInBits(Ty);
+    return { (BaseSize.getKnownMinSize() + 7) / 8, BaseSize.isScalable() };
   }
----------------
huntergr wrote:
> rovka wrote:
> > We already overload operator /, why not overload + as well so we don't have to change the body of this method?
> Scaling a size with * or / has a clear meaning to me, since it's independent of vscale; getting a vector that's half the size or four times larger just works.
> 
> Using + (or -) on the other hand doesn't seem to be as clear; I wasn't sure if a standalone int should be automatically treated as being the same as the TypeSize, or always considered Fixed. If we try for the former I can imagine quite a few bugs arising.
> 
> I could add a roundBitsToNearestByteSize method to move the arithmetic elsewhere if that would be acceptable?
You're right, + on TypeSizes would be confusing. This looks ok as-is then, no need to fiddle with it more.


================
Comment at: llvm/include/llvm/IR/DataLayout.h:656
+                     getTypeSizeInBits(VTy->getElementType()).getKnownMinSize();
+    return ScalableSize(MinBits, EltCnt.Scalable);
   }
----------------
huntergr wrote:
> rovka wrote:
> > Maybe just return VTy->getElementCount() * getTypeSizeInBits(VTy->getElementType()).getFixedSize().
> There's no support for generating a TypeSize from an ElementCount in that way; is that an interface you feel is useful?
> 
> (I'll certainly change the `getKnownMinSize` to `getFixedSize` though, since we're just referring to a scalar)
Actually, no, now that I think about it a bit more it might be clearer to spell it out this way.


================
Comment at: llvm/include/llvm/Support/TypeSize.h:122
+  // Use in places where a scalable size doesn't make sense (e.g. non-vector
+  // types, or vectors in backends which don't support scalable vectors)
+  uint64_t getFixedSize() const {
----------------
Microscopic nit: punctuation.


================
Comment at: llvm/include/llvm/Support/TypeSize.h:143
+  // NOTE: This interface is obsolete and will be removed in a future version
+  // of LLVM in favour of calling getFixedSize() directly
+  operator uint64_t() const {
----------------
Ditto.


================
Comment at: llvm/include/llvm/Support/TypeSize.h:148
+
+  // Additional convenience operators needed to avoid ambiguous parses
+  // TODO: Make uint64_t the default operator?
----------------
Ditto.


Repository:
  rG LLVM Github Monorepo

CHANGES SINCE LAST ACTION
  https://reviews.llvm.org/D53137/new/

https://reviews.llvm.org/D53137