[RISCVGatherScatterLowering] Remove restriction that shift must have constant operand

[reames]

Summary

This has been present from the original patch, and doesn't appear to be strongly justified. We do need to be careful of commutativity.

[reames]

shl is not commutative.

[craig.topper]

llvm/lib/Target/RISCV/RISCVGatherScatterLowering.cpp

250

This might have been a hacky way of ensuring that the shift amount was the loop invariant splat part? the code right after the switch assumes the instructions are commutable.

[reames]

Address the commutativity bug.

[craig.topper]

LGTM

llvm/lib/Target/RISCV/RISCVGatherScatterLowering.cpp

bool RISCVGatherScatterLowering::matchStridedRecurrence(Value *Index, Loop *L,
   case Instruction::Or:
     // We need to be able to treat Or as Add.
     if (!haveNoCommonBitsSet(BO->getOperand(0), BO->getOperand(1), *DL))
       return false;
     break;
   case Instruction::Add:
     break;
   case Instruction::Shl:
-    // Only support shift by constant.
-    if (!isa<Constant>(BO->getOperand(1)))
-      return false;
     break;
   case Instruction::Mul:
     break;
   }
 
   // We should have one operand in the loop and one splat.
   Value *OtherOp;
   if (isa<Instruction>(BO->getOperand(0)) &&

llvm/test/CodeGen/RISCV/rvv/fixed-vector-strided-load-store.ll

vector.body:                                      ; preds = %vector.body, %entry
   %vec.ind.next = add <8 x i64> %vec.ind, <i64 8, i64 8, i64 8, i64 8, i64 8, i64 8, i64 8, i64 8>
   %i6 = icmp eq i64 %index.next, 1024
   br i1 %i6, label %for.cond.cleanup, label %vector.body
 
 for.cond.cleanup:                                 ; preds = %vector.body
   ret void
 }
 
+define void @gather_unknown_pow2(ptr noalias nocapture %A, ptr noalias nocapture readonly %B, i64 %shift) {
+; CHECK-LABEL: @gather_unknown_pow2(
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    [[STEP:%.*]] = shl i64 8, [[SHIFT:%.*]]
+; CHECK-NEXT:    [[STRIDE:%.*]] = shl i64 1, [[SHIFT]]
+; CHECK-NEXT:    [[TMP0:%.*]] = mul i64 [[STRIDE]], 4
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i64 [ 0, [[ENTRY:%.*]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[VEC_IND_SCALAR:%.*]] = phi i64 [ 0, [[ENTRY]] ], [ [[VEC_IND_NEXT_SCALAR:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr i32, ptr [[B:%.*]], i64 [[VEC_IND_SCALAR]]
+; CHECK-NEXT:    [[WIDE_MASKED_GATHER:%.*]] = call <8 x i32> @llvm.riscv.masked.strided.load.v8i32.p0.i64(<8 x i32> undef, ptr [[TMP1]], i64 [[TMP0]], <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>)
+; CHECK-NEXT:    [[I2:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
+; CHECK-NEXT:    [[WIDE_LOAD:%.*]] = load <8 x i32>, ptr [[I2]], align 1
+; CHECK-NEXT:    [[I4:%.*]] = add <8 x i32> [[WIDE_LOAD]], [[WIDE_MASKED_GATHER]]
+; CHECK-NEXT:    store <8 x i32> [[I4]], ptr [[I2]], align 1
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 8
+; CHECK-NEXT:    [[VEC_IND_NEXT_SCALAR]] = add i64 [[VEC_IND_SCALAR]], [[STEP]]
+; CHECK-NEXT:    [[I6:%.*]] = icmp eq i64 [[INDEX_NEXT]], 1024
+; CHECK-NEXT:    br i1 [[I6]], label [[FOR_COND_CLEANUP:%.*]], label [[VECTOR_BODY]]
+; CHECK:       for.cond.cleanup:
+; CHECK-NEXT:    ret void
+;
+entry:
+  %.splatinsert = insertelement <8 x i64> poison, i64 %shift, i64 0
+  %.splat = shufflevector <8 x i64> %.splatinsert, <8 x i64> poison, <8 x i32> zeroinitializer
+  br label %vector.body
+
+vector.body:                                      ; preds = %vector.body, %entry
+  %index = phi i64 [ 0, %entry ], [ %index.next, %vector.body ]
+  %vec.ind = phi <8 x i64> [ <i64 0, i64 1, i64 2, i64 3, i64 4, i64 5, i64 6, i64 7>, %entry ], [ %vec.ind.next, %vector.body ]
+  %i = shl nsw <8 x i64> %vec.ind, %.splat
+  %i1 = getelementptr inbounds i32, ptr %B, <8 x i64> %i
+  %wide.masked.gather = call <8 x i32> @llvm.masked.gather.v8i32.v8p0(<8 x ptr> %i1, i32 4, <8 x i1> <i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true, i1 true>, <8 x i32> undef)
+  %i2 = getelementptr inbounds i32, ptr %A, i64 %index
+  %wide.load = load <8 x i32>, ptr %i2, align 1
+  %i4 = add <8 x i32> %wide.load, %wide.masked.gather
+  store <8 x i32> %i4, ptr %i2, align 1
+  %index.next = add nuw i64 %index, 8
+  %vec.ind.next = add <8 x i64> %vec.ind, <i64 8, i64 8, i64 8, i64 8, i64 8, i64 8, i64 8, i64 8>
+  %i6 = icmp eq i64 %index.next, 1024
+  br i1 %i6, label %for.cond.cleanup, label %vector.body
+
+for.cond.cleanup:                                 ; preds = %vector.body
+  ret void
+}
+
 ;void scatter_pow2(signed char * __restrict  A, signed char * __restrict B) {
 ;  for (int i = 0; i < 1024; ++i)
 ;      A[i * 4] += B[i];
 ;}
 define void @scatter_pow2(ptr noalias nocapture %A, ptr noalias nocapture readonly %B) {
 ; CHECK-LABEL: @scatter_pow2(
 ; CHECK-NEXT:  entry:
 ; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]