[llvm] [LoopIdiomRecognize] Enable clmul optimization for CRC loops (PR #203405)

[Ramkumar Ramachandra via llvm-commits]

================
@@ -1549,7 +1553,145 @@ bool LoopIdiomRecognize::avoidLIRForMultiBlockLoop(bool IsMemset,
   return false;
 }
 
-bool LoopIdiomRecognize::optimizeCRCLoop(const PolynomialInfo &Info) {
+bool LoopIdiomRecognize::optimizeCRCLoopUsingClmul(const PolynomialInfo &Info) {
+  Type *CRCTy = Info.LHS->getType();
+  LLVMContext &Ctx = CRCTy->getContext();
+  unsigned CRCBW = CRCTy->getIntegerBitWidth();
+  // The TripCount determines how many bits of data are processed, regardless of
+  // whether the actual data bit width matches (if auxiliary data is even used
+  // at all).
+  unsigned TC = Info.TripCount;
+  // The first clmul uses 2*TC bits, and the second clmul uses CRCBW+TC bits.
+  // For simplicity, have both operate on the same bit width.
+  unsigned ClmulBW = std::max(2 * TC, CRCBW + TC);
+  auto *ClmulTy = IntegerType::get(Ctx, ClmulBW);
+
+  // This optimization should only be applied if clmul for the required width is
+  // a fast operation on the target.
+  // TODO: If TC > CRCBW, then the data could probably be split into multiple
+  // chunks and processed in a loop.
+  if (!TTI->haveFastClmul(ClmulTy))
+    return false;
+
+  // First, generate the constants required for GF(2) Barrett reduction.
+  auto [Mu, FullGenPoly] =
+      HashRecognize::genBarrettConstants(Info.RHS, TC, Info.ByteOrderSwapped);
+  Value *MuExt = ConstantInt::get(Ctx, Mu.zext(ClmulBW));
+  Value *FullGenPolyExt = ConstantInt::get(Ctx, FullGenPoly.zext(ClmulBW));
+
+  IRBuilder<> Builder(CurLoop->getLoopPreheader()->getTerminator());
+
+  auto GetMostSignificantTCBits = [&, TC](Value *Op, unsigned BW,
+                                          const Twine &Name) {
+    unsigned OpBW = Op->getType()->getIntegerBitWidth();
+    assert(BW > TC && OpBW > TC);
+    if (Info.ByteOrderSwapped)
+      return Builder.CreateLShr(Op, BW - TC, Name + ".be.lshr");
+    auto *Mask = ConstantInt::get(Ctx, APInt::getLowBitsSet(OpBW, TC));
+    return Builder.CreateAnd(Op, Mask, Name + ".le.mask");
+  };
+
+  Value *LHS = Builder.CreateZExt(Info.LHS, ClmulTy, "crc.ext");
+
+  // For the big-endian case, align the leftmost bit of the CRC with the
+  // leftmost bit of the data which is used. For the little-endian case, align
+  // the rightmost bits (nothing to do).
+  Value *CRCAlignTC;
+  if (!Info.ByteOrderSwapped || CRCBW == TC)
+    CRCAlignTC = LHS;
+  else if (CRCBW > TC)
+    CRCAlignTC = Builder.CreateLShr(LHS, CRCBW - TC, "crc.be.lshr");
+  else
+    CRCAlignTC = Builder.CreateShl(LHS, TC - CRCBW, "crc.be.shl");
----------------
artagnon wrote:

TC can never exceed the maximum of (CRCBW, DataBW). From HashRecognize:

```cpp
  if (TC > (LHSAux ? LHSAux->getType()->getIntegerBitWidth()
                   : LHS->getType()->getIntegerBitWidth()))
    return "Loop iterations exceed bitwidth of data";
```

Perhaps you can use this constraint to simplify the code a bit?

https://github.com/llvm/llvm-project/pull/203405