[llvm] [VPlan] Remove unused first mask op from VPBlendRecipe. (PR #87770)

[via llvm-commits]

================
@@ -0,0 +1,179 @@
+; NOTE: Assertions have been autogenerated by utils/update_test_checks.py UTC_ARGS: --version 4
+; RUN: opt -p loop-vectorize -force-vector-width=4 -force-vector-interleave=1 -S %s | FileCheck %s
+
+target datalayout = "e-m:e-p:64:64-i64:64-i128:128-n32:64-S128"
+
+; Test cases for https://github.com/llvm/llvm-project/issues/87410.
+define void @test_not_first_lane_only_constant(ptr %A, ptr noalias %B)  {
+; CHECK-LABEL: define void @test_not_first_lane_only_constant(
+; CHECK-SAME: ptr [[A:%.*]], ptr noalias [[B:%.*]]) {
+; CHECK-NEXT:  entry:
+; CHECK-NEXT:    br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
+; CHECK:       vector.ph:
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT3:%.*]] = insertelement <4 x ptr> poison, ptr [[B]], i64 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT4:%.*]] = shufflevector <4 x ptr> [[BROADCAST_SPLATINSERT3]], <4 x ptr> poison, <4 x i32> zeroinitializer
+; CHECK-NEXT:    br label [[VECTOR_BODY:%.*]]
+; CHECK:       vector.body:
+; CHECK-NEXT:    [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
+; CHECK-NEXT:    [[OFFSET_IDX:%.*]] = trunc i32 [[INDEX]] to i16
+; CHECK-NEXT:    [[TMP0:%.*]] = add i16 [[OFFSET_IDX]], 0
+; CHECK-NEXT:    [[TMP1:%.*]] = getelementptr inbounds i16, ptr [[A]], i16 [[TMP0]]
+; CHECK-NEXT:    [[PREDPHI:%.*]] = select <4 x i1> zeroinitializer, <4 x ptr> poison, <4 x ptr> [[BROADCAST_SPLAT4]]
+; CHECK-NEXT:    [[TMP12:%.*]] = extractelement <4 x ptr> [[PREDPHI]], i32 0
+; CHECK-NEXT:    [[TMP13:%.*]] = load i16, ptr [[TMP12]], align 2
+; CHECK-NEXT:    [[BROADCAST_SPLATINSERT5:%.*]] = insertelement <4 x i16> poison, i16 [[TMP13]], i64 0
+; CHECK-NEXT:    [[BROADCAST_SPLAT6:%.*]] = shufflevector <4 x i16> [[BROADCAST_SPLATINSERT5]], <4 x i16> poison, <4 x i32> zeroinitializer
+; CHECK-NEXT:    [[TMP2:%.*]] = getelementptr inbounds i16, ptr [[TMP1]], i32 0
+; CHECK-NEXT:    store <4 x i16> [[BROADCAST_SPLAT6]], ptr [[TMP2]], align 2
+; CHECK-NEXT:    [[INDEX_NEXT]] = add nuw i32 [[INDEX]], 4
+; CHECK-NEXT:    [[TMP14:%.*]] = icmp eq i32 [[INDEX_NEXT]], 1000
+; CHECK-NEXT:    br i1 [[TMP14]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
+; CHECK:       middle.block:
+; CHECK-NEXT:    br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
+; CHECK:       scalar.ph:
+; CHECK-NEXT:    [[BC_RESUME_VAL:%.*]] = phi i16 [ 1000, [[MIDDLE_BLOCK]] ], [ 0, [[ENTRY:%.*]] ]
+; CHECK-NEXT:    br label [[LOOP_HEADER:%.*]]
+; CHECK:       loop.header:
+; CHECK-NEXT:    [[IV:%.*]] = phi i16 [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], [[LOOP_LATCH:%.*]] ]
+; CHECK-NEXT:    [[GEP_A:%.*]] = getelementptr inbounds i16, ptr [[A]], i16 [[IV]]
+; CHECK-NEXT:    br i1 false, label [[LOOP_LATCH]], label [[ELSE_1:%.*]]
+; CHECK:       else.1:
+; CHECK-NEXT:    br i1 false, label [[THEN_2:%.*]], label [[ELSE_2:%.*]]
+; CHECK:       then.2:
+; CHECK-NEXT:    br label [[ELSE_2]]
+; CHECK:       else.2:
+; CHECK-NEXT:    br label [[LOOP_LATCH]]
+; CHECK:       loop.latch:
+; CHECK-NEXT:    [[MERGE:%.*]] = phi ptr [ [[B]], [[ELSE_2]] ], [ poison, [[LOOP_HEADER]] ]
+; CHECK-NEXT:    [[L:%.*]] = load i16, ptr [[MERGE]], align 2
+; CHECK-NEXT:    [[IV_NEXT]] = add i16 [[IV]], 1
+; CHECK-NEXT:    store i16 [[L]], ptr [[GEP_A]], align 2
+; CHECK-NEXT:    [[C_2:%.*]] = icmp eq i16 [[IV_NEXT]], 1000
+; CHECK-NEXT:    br i1 [[C_2]], label [[EXIT]], label [[LOOP_HEADER]], !llvm.loop [[LOOP3:![0-9]+]]
+; CHECK:       exit:
+; CHECK-NEXT:    ret void
+;
+entry:
+  br label %loop.header
+
+loop.header:
+  %iv = phi i16 [ 0, %entry ], [ %iv.next, %loop.latch ]
+  %gep.A = getelementptr inbounds i16, ptr %A, i16 %iv
+  br i1 false, label %loop.latch, label %else.1
+
+else.1:
+  br i1 false, label %then.2, label %else.2
+
+then.2:
+  br label %else.2
+
+else.2:
+  br label %loop.latch
+
+loop.latch:
+  %merge = phi ptr [ %B, %else.2 ], [ poison, %loop.header ]
----------------
ayalz wrote:

This still breaks if the operands are swapped:
`%merge = phi ptr [ poison, %loop.header ], [ %B, %else.2 ]`
so that the mask associated with the %loop.header predecessor gets dropped,
but the mask of the problematic %else.2 predecessor is still used to compute the blend.

https://github.com/llvm/llvm-project/pull/87770