[llvm] [AMDGPU] Constant folding of llvm.amdgcn.trig.preop (PR #98562)

[Jay Foad via llvm-commits]

================
@@ -1102,6 +1102,84 @@ GCNTTIImpl::instCombineIntrinsic(InstCombiner &IC, IntrinsicInst &II) const {
 
     break;
   }
+  case Intrinsic::amdgcn_trig_preop: {
+    // The intrinsic is declared with name mangling, but currently the
+    // instruction only exists for f64
+    if (!II.getType()->isDoubleTy())
+      break;
+
+    Value *Src = II.getArgOperand(0);
+    Value *Segment = II.getArgOperand(1);
+    if (isa<PoisonValue>(Src))
+      return IC.replaceInstUsesWith(II, Src);
+
+    if (isa<UndefValue>(Src)) {
+      auto *QNaN = ConstantFP::get(
+          II.getType(), APFloat::getQNaN(II.getType()->getFltSemantics()));
+      return IC.replaceInstUsesWith(II, QNaN);
+    }
+
+    const ConstantFP *Csrc = dyn_cast<ConstantFP>(Src);
+    if (!Csrc)
+      break;
+
+    if (II.isStrictFP())
+      break;
+
+    const APFloat &Fsrc = Csrc->getValueAPF();
+    if (Fsrc.isNaN()) {
+      // FIXME: We just need to make the nan quiet here, but that's unavailable
+      // on APFloat, only IEEEfloat
+      auto *Quieted = ConstantFP::get(
+          II.getType(), scalbn(Fsrc, 0, APFloat::rmNearestTiesToEven));
+      return IC.replaceInstUsesWith(II, Quieted);
+    }
+
+    const ConstantInt *Cseg = dyn_cast<ConstantInt>(Segment);
+    if (!Cseg)
+      break;
+
+    APInt Exponent = Fsrc.bitcastToAPInt().ashr(52);
+    Exponent &= 0x7ff;
+    Exponent = Exponent.trunc(32);
+
+    unsigned SegmentVal = Cseg->getValue().trunc(5).getZExtValue();
+    unsigned Shift = SegmentVal * 53;
+    if (Exponent.sgt(1077))
+      Shift += Exponent.getZExtValue() - 1077;
+
+    // 2.0/PI table.
+    static const uint32_t TwoByPi[] = {
+        0xa2f9836e, 0x4e441529, 0xfc2757d1, 0xf534ddc0, 0xdb629599, 0x3c439041,
+        0xfe5163ab, 0xdebbc561, 0xb7246e3a, 0x424dd2e0, 0x06492eea, 0x09d1921c,
+        0xfe1deb1c, 0xb129a73e, 0xe88235f5, 0x2ebb4484, 0xe99c7026, 0xb45f7e41,
+        0x3991d639, 0x835339f4, 0x9c845f8b, 0xbdf9283b, 0x1ff897ff, 0xde05980f,
+        0xef2f118b, 0x5a0a6d1f, 0x6d367ecf, 0x27cb09b7, 0x4f463f66, 0x9e5fea2d,
+        0x7527bac7, 0xebe5f17b, 0x3d0739f7, 0x8a5292ea, 0x6bfb5fb1, 0x1f8d5d08,
+        0x56033046};
+
+    // Return 0 for outbound segment (hardware behavior).
+    unsigned Idx = Shift >> 5;
+    if (Idx + 2 >= std::size(TwoByPi)) {
+      APFloat Zero = APFloat::getZero(II.getType()->getFltSemantics());
+      return IC.replaceInstUsesWith(II, ConstantFP::get(Src->getType(), Zero));
+    }
+
+    unsigned BShift = Shift & 0x1f;
+    APInt Thi = APInt(64, Make_64(TwoByPi[Idx], TwoByPi[Idx + 1]), false);
+    APInt Tlo = APInt(64, Make_64(TwoByPi[Idx + 2], 0), false);
----------------
jayfoad wrote:

Use uint64_t instead of APInt?

https://github.com/llvm/llvm-project/pull/98562