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

[Ramkumar Ramachandra via llvm-commits]

================
@@ -87,6 +87,19 @@ class HashRecognize {
   LLVM_ABI static CRCTable genSarwateTable(const APInt &GenPoly,
                                            bool ByteOrderSwapped);
 
+  // Auxilary entry point after analysis to generate constants for a GF(2)
+  // Barrett Reduction.
+  // The first APInt is the constant used in the first clmul operation, which is
+  // mu = floor(x^(BW+TC) / P(x)); bit width is TC+1.
+  // The second APInt is the constant used in the second clmul operation, which
+  // is the generating polynomial P(x) with the implied x^BW term included; bit
+  // width is BW+1.
+  // Both constants are bit-reflected across their respective
+  // widths for bit-reflected CRCs.
----------------
artagnon wrote:

```suggestion
  // Returns a pair of Mu of bitwidth TC+1 and FullGenPoly of bitwidth BW+1.
  // Mu is used in the first clmul operation. Mu = floor(x^(BW+TC) / P(x)).
  // FullGenPoly is used in the second clmul operation, and is \p GenPoly
  // with the implied BW'th bit.
  // Endianness is accounted for using \p ByteOrderSwapped.
```

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