[LV] Add abs/smin/smax/umin/umax intrinsics to isTriviallyVectorizable

[craig.topper]

Summary

This patch adds support for vectorizing these intrinsics.

[arsenm]

llvm/test/Transforms/LoopVectorize/intrinsic.ll

1250

Probably should check the immediate operand value

[craig.topper]

Check the immediate on the vectorized abs intrinsic

[spatel]

LGTM

llvm/lib/Analysis/VectorUtils.cpp

static cl::opt<unsigned> MaxInterleaveGroupFactor(
     cl::init(8));
 
 /// Return true if all of the intrinsic's arguments and return type are scalars
 /// for the scalar form of the intrinsic, and vectors for the vector form of the
 /// intrinsic (except operands that are marked as always being scalar by
 /// hasVectorInstrinsicScalarOpd).
 bool llvm::isTriviallyVectorizable(Intrinsic::ID ID) {
   switch (ID) {
-  case Intrinsic::bswap: // Begin integer bit-manipulation.
+  case Intrinsic::abs:   // Begin integer bit-manipulation.
+  case Intrinsic::bswap:
   case Intrinsic::bitreverse:
   case Intrinsic::ctpop:
   case Intrinsic::ctlz:
   case Intrinsic::cttz:
   case Intrinsic::fshl:
   case Intrinsic::fshr:
+  case Intrinsic::smax:
+  case Intrinsic::smin:
+  case Intrinsic::umax:
+  case Intrinsic::umin:
   case Intrinsic::sadd_sat:
   case Intrinsic::ssub_sat:
   case Intrinsic::uadd_sat:
   case Intrinsic::usub_sat:
   case Intrinsic::smul_fix:
   case Intrinsic::smul_fix_sat:
   case Intrinsic::umul_fix:
   case Intrinsic::umul_fix_sat:
default:
     return false;
   }
 }
 
 /// Identifies if the vector form of the intrinsic has a scalar operand.
 bool llvm::hasVectorInstrinsicScalarOpd(Intrinsic::ID ID,
                                         unsigned ScalarOpdIdx) {
   switch (ID) {
+  case Intrinsic::abs:
   case Intrinsic::ctlz:
   case Intrinsic::cttz:
   case Intrinsic::powi:
     return (ScalarOpdIdx == 1);
   case Intrinsic::smul_fix:
   case Intrinsic::smul_fix_sat:
   case Intrinsic::umul_fix:
   case Intrinsic::umul_fix_sat:

llvm/test/Transforms/LoopVectorize/intrinsic.ll

for.body:                                         ; preds = %entry, %for.body
   %lftr.wideiv = trunc i64 %indvars.iv.next to i32
   %exitcond = icmp eq i32 %lftr.wideiv, %n
   br i1 %exitcond, label %for.end, label %for.body
 
 for.end:                                          ; preds = %for.body, %entry
   ret void
 }
 
+declare i64 @llvm.abs.i64 (i64, i1) nounwind readnone
+
+;CHECK-LABEL: @abs_i64(
+;CHECK: llvm.abs.v4i64
+;CHECK: ret void
+define void @abs_i64(i32 %n, i64* noalias %y, i64* noalias %x) nounwind uwtable {
+entry:
+  %cmp9 = icmp sgt i32 %n, 0
+  br i1 %cmp9, label %for.body, label %for.end
+
+for.body:                                         ; preds = %entry, %for.body
+  %indvars.iv = phi i64 [ %indvars.iv.next, %for.body ], [ 0, %entry ]
+  %arrayidx = getelementptr inbounds i64, i64* %y, i64 %indvars.iv
+  %0 = load i64, i64* %arrayidx, align 8
+  %call = tail call i64 @llvm.abs.i64(i64 %0, i1 true) nounwind readnone
+  %arrayidx4 = getelementptr inbounds i64, i64* %x, i64 %indvars.iv
+  store i64 %call, i64* %arrayidx4, align 8
+  %indvars.iv.next = add i64 %indvars.iv, 1
+  %lftr.wideiv = trunc i64 %indvars.iv.next to i32
+  %exitcond = icmp eq i32 %lftr.wideiv, %n
+  br i1 %exitcond, label %for.end, label %for.body
+
+for.end:                                          ; preds = %for.body, %entry
+  ret void
+}
+
+declare i32 @llvm.smin.i32 (i32, i32)
+
+define void @smin_i32(i32 %n, i32* noalias %x, i32* noalias %y) {
+; CHECK-LABEL: @smin_i32(
+; CHECK:         call <4 x i32> @llvm.smin.v4i32(<4 x i32> [[WIDE_LOADX:%.*]], <4 x i32> [[WIDE_LOADY:%.*]])
+; CHECK:         ret void
+entry:
+  %cmp = icmp sgt i32 %n, 0
+  br i1 %cmp, label %loop, label %end
+
+loop:
+  %iv = phi i32 [ %iv.next, %loop ], [ 0, %entry ]
+  %xi = getelementptr inbounds i32, i32* %x, i32 %iv
+  %yi = getelementptr inbounds i32, i32* %y, i32 %iv
+  %xld = load i32, i32* %xi, align 4
+  %yld = load i32, i32* %yi, align 4
+  %call = tail call i32 @llvm.smin.i32(i32 %xld, i32 %yld)
+  store i32 %call, i32* %xi, align 4
+  %iv.next = add i32 %iv, 1
+  %exitcond = icmp eq i32 %iv.next, %n
+  br i1 %exitcond, label %end, label %loop
+
+end:
+  ret void
+}
+
+declare i32 @llvm.smax.i32 (i32, i32)
+
+define void @smax_i32(i32 %n, i32* noalias %x, i32* noalias %y) {
+; CHECK-LABEL: @smax_i32(
+; CHECK:         call <4 x i32> @llvm.smax.v4i32(<4 x i32> [[WIDE_LOADX:%.*]], <4 x i32> [[WIDE_LOADY:%.*]])
+; CHECK:         ret void
+entry:
+  %cmp = icmp sgt i32 %n, 0
+  br i1 %cmp, label %loop, label %end
+
+loop:
+  %iv = phi i32 [ %iv.next, %loop ], [ 0, %entry ]
+  %xi = getelementptr inbounds i32, i32* %x, i32 %iv
+  %yi = getelementptr inbounds i32, i32* %y, i32 %iv
+  %xld = load i32, i32* %xi, align 4
+  %yld = load i32, i32* %yi, align 4
+  %call = tail call i32 @llvm.smax.i32(i32 %xld, i32 %yld)
+  store i32 %call, i32* %xi, align 4
+  %iv.next = add i32 %iv, 1
+  %exitcond = icmp eq i32 %iv.next, %n
+  br i1 %exitcond, label %end, label %loop
+
+end:
+  ret void
+}
+
+declare i32 @llvm.umin.i32 (i32, i32)
+
+define void @umin_i32(i32 %n, i32* noalias %x, i32* noalias %y) {
+; CHECK-LABEL: @umin_i32(
+; CHECK:         call <4 x i32> @llvm.umin.v4i32(<4 x i32> [[WIDE_LOADX:%.*]], <4 x i32> [[WIDE_LOADY:%.*]])
+; CHECK:         ret void
+entry:
+  %cmp = icmp sgt i32 %n, 0
+  br i1 %cmp, label %loop, label %end
+
+loop:
+  %iv = phi i32 [ %iv.next, %loop ], [ 0, %entry ]
+  %xi = getelementptr inbounds i32, i32* %x, i32 %iv
+  %yi = getelementptr inbounds i32, i32* %y, i32 %iv
+  %xld = load i32, i32* %xi, align 4
+  %yld = load i32, i32* %yi, align 4
+  %call = tail call i32 @llvm.umin.i32(i32 %xld, i32 %yld)
+  store i32 %call, i32* %xi, align 4
+  %iv.next = add i32 %iv, 1
+  %exitcond = icmp eq i32 %iv.next, %n
+  br i1 %exitcond, label %end, label %loop
+
+end:
+  ret void
+}
+
+declare i32 @llvm.umax.i32 (i32, i32)
+
+define void @umax_i32(i32 %n, i32* noalias %x, i32* noalias %y) {
+; CHECK-LABEL: @umax_i32(
+; CHECK:         call <4 x i32> @llvm.umax.v4i32(<4 x i32> [[WIDE_LOADX:%.*]], <4 x i32> [[WIDE_LOADY:%.*]])
+; CHECK:         ret void
+entry:
+  %cmp = icmp sgt i32 %n, 0
+  br i1 %cmp, label %loop, label %end
+
+loop:
+  %iv = phi i32 [ %iv.next, %loop ], [ 0, %entry ]
+  %xi = getelementptr inbounds i32, i32* %x, i32 %iv
+  %yi = getelementptr inbounds i32, i32* %y, i32 %iv
+  %xld = load i32, i32* %xi, align 4
+  %yld = load i32, i32* %yi, align 4
+  %call = tail call i32 @llvm.umax.i32(i32 %xld, i32 %yld)
+  store i32 %call, i32* %xi, align 4
+  %iv.next = add i32 %iv, 1
+  %exitcond = icmp eq i32 %iv.next, %n
+  br i1 %exitcond, label %end, label %loop
+
+end:
+  ret void
+}
+
 declare i32 @llvm.fshl.i32 (i32, i32, i32)
 
 define void @fshl_i32(i32 %n, i32* noalias %x, i32* noalias %y, i32 %shAmt) {
 ; CHECK-LABEL: @fshl_i32(
 ; CHECK:         call <4 x i32> @llvm.fshl.v4i32(<4 x i32> [[WIDE_LOADX:%.*]], <4 x i32> [[WIDE_LOADY:%.*]], <4 x i32> [[SPLAT:%.*]])
 ; CHECK:         ret void
 entry:
   %cmp = icmp sgt i32 %n, 0