[Analysis] Add support for vscale in computeKnownBitsFromOperator

In ValueTracking.cpp we use a function called
computeKnownBitsFromOperator to determine the known bits of a value.
For the vscale intrinsic if the function contains the vscale_range
attribute we can use the maximum and minimum values of vscale to
determine some known zero and one bits. This should help to improve
code quality by allowing certain optimisations to take place.

Tests added here:

  Transforms/InstCombine/icmp-vscale.ll

Differential Revision: https://reviews.llvm.org/D109883

1 files changed, 26 insertions, 0 deletions

diff --git a/llvm/lib/Analysis/ValueTracking.cpp b/llvm/lib/Analysis/ValueTracking.cpp
index 8d18d133..57869e3 100644
--- a/llvm/lib/Analysis/ValueTracking.cpp
+++ b/llvm/lib/Analysis/ValueTracking.cpp
@@ -1689,6 +1689,32 @@ static void computeKnownBitsFromOperator(const Operator *I,
         if (BitWidth >= 32)
           Known.Zero.setBitsFrom(31);
         break;
+      case Intrinsic::vscale: {
+        if (!II->getFunction()->hasFnAttribute(Attribute::VScaleRange))
+          break;
+
+        auto VScaleRange = II->getFunction()
+                               ->getFnAttribute(Attribute::VScaleRange)
+                               .getVScaleRangeArgs();
+
+        if (VScaleRange.second == 0)
+          break;
+
+        // If vscale min = max then we know the exact value at compile time
+        // and hence we know the exact bits.
+        if (VScaleRange.first == VScaleRange.second) {
+          Known.One = VScaleRange.first;
+          Known.Zero = VScaleRange.first;
+          Known.Zero.flipAllBits();
+          break;
+        }
+
+        unsigned FirstZeroHighBit = 32 - countLeadingZeros(VScaleRange.second);
+        if (FirstZeroHighBit < BitWidth)
+          Known.Zero.setBitsFrom(FirstZeroHighBit);
+
+        break;
+      }
       }
     }
     break;