Skip to content

Commit fc3a21b

Browse files
Mel-ChenMel-Chen
committed
update test case
1 parent d7584a3 commit fc3a21b

File tree

1 file changed

+48
-40
lines changed

1 file changed

+48
-40
lines changed

‎llvm/test/Transforms/LoopVectorize/RISCV/inloop-reduction.ll

Lines changed: 48 additions & 40 deletions
Original file line numberDiff line numberDiff line change
@@ -140,42 +140,46 @@ define i32 @add_i16_i32(ptr nocapture readonly %x, i32 %n) {
140140
; IF-EVL-OUTLOOP-NEXT: br label [[VECTOR_BODY:%.*]]
141141
; IF-EVL-OUTLOOP: vector.body:
142142
; IF-EVL-OUTLOOP-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
143-
; IF-EVL-OUTLOOP-NEXT: [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP14:%.*]], [[VECTOR_BODY]] ]
144-
; IF-EVL-OUTLOOP-NEXT: [[TMP7:%.*]] = add i32 [[INDEX]], 0
145-
; IF-EVL-OUTLOOP-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[INDEX]], i64 0
143+
; IF-EVL-OUTLOOP-NEXT: [[EVL_BASED_IV:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], [[VECTOR_BODY]] ]
144+
; IF-EVL-OUTLOOP-NEXT: [[VEC_PHI:%.*]] = phi <vscale x 4 x i32> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP16:%.*]], [[VECTOR_BODY]] ]
145+
; IF-EVL-OUTLOOP-NEXT: [[TMP7:%.*]] = sub i32 [[N]], [[EVL_BASED_IV]]
146+
; IF-EVL-OUTLOOP-NEXT: [[TMP8:%.*]] = call i32 @llvm.experimental.get.vector.length.i32(i32 [[TMP7]], i32 4, i1 true)
147+
; IF-EVL-OUTLOOP-NEXT: [[TMP9:%.*]] = add i32 [[EVL_BASED_IV]], 0
148+
; IF-EVL-OUTLOOP-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 4 x i32> poison, i32 [[EVL_BASED_IV]], i64 0
146149
; IF-EVL-OUTLOOP-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 4 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 4 x i32> poison, <vscale x 4 x i32> zeroinitializer
147-
; IF-EVL-OUTLOOP-NEXT: [[TMP8:%.*]] = call <vscale x 4 x i32> @llvm.experimental.stepvector.nxv4i32()
148-
; IF-EVL-OUTLOOP-NEXT: [[TMP9:%.*]] = add <vscale x 4 x i32> zeroinitializer, [[TMP8]]
149-
; IF-EVL-OUTLOOP-NEXT: [[VEC_IV:%.*]] = add <vscale x 4 x i32> [[BROADCAST_SPLAT]], [[TMP9]]
150-
; IF-EVL-OUTLOOP-NEXT: [[TMP10:%.*]] = icmp ule <vscale x 4 x i32> [[VEC_IV]], [[BROADCAST_SPLAT2]]
151-
; IF-EVL-OUTLOOP-NEXT: [[TMP11:%.*]] = getelementptr inbounds i16, ptr [[X:%.*]], i32 [[TMP7]]
152-
; IF-EVL-OUTLOOP-NEXT: [[TMP12:%.*]] = getelementptr inbounds i16, ptr [[TMP11]], i32 0
153-
; IF-EVL-OUTLOOP-NEXT: [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 4 x i16> @llvm.masked.load.nxv4i16.p0(ptr [[TMP12]], i32 2, <vscale x 4 x i1> [[TMP10]], <vscale x 4 x i16> poison)
154-
; IF-EVL-OUTLOOP-NEXT: [[TMP13:%.*]] = sext <vscale x 4 x i16> [[WIDE_MASKED_LOAD]] to <vscale x 4 x i32>
155-
; IF-EVL-OUTLOOP-NEXT: [[TMP14]] = add <vscale x 4 x i32> [[VEC_PHI]], [[TMP13]]
156-
; IF-EVL-OUTLOOP-NEXT: [[TMP15:%.*]] = select <vscale x 4 x i1> [[TMP10]], <vscale x 4 x i32> [[TMP14]], <vscale x 4 x i32> [[VEC_PHI]]
150+
; IF-EVL-OUTLOOP-NEXT: [[TMP10:%.*]] = call <vscale x 4 x i32> @llvm.experimental.stepvector.nxv4i32()
151+
; IF-EVL-OUTLOOP-NEXT: [[TMP11:%.*]] = add <vscale x 4 x i32> zeroinitializer, [[TMP10]]
152+
; IF-EVL-OUTLOOP-NEXT: [[VEC_IV:%.*]] = add <vscale x 4 x i32> [[BROADCAST_SPLAT]], [[TMP11]]
153+
; IF-EVL-OUTLOOP-NEXT: [[TMP12:%.*]] = icmp ule <vscale x 4 x i32> [[VEC_IV]], [[BROADCAST_SPLAT2]]
154+
; IF-EVL-OUTLOOP-NEXT: [[TMP13:%.*]] = getelementptr inbounds i16, ptr [[X:%.*]], i32 [[TMP9]]
155+
; IF-EVL-OUTLOOP-NEXT: [[TMP14:%.*]] = getelementptr inbounds i16, ptr [[TMP13]], i32 0
156+
; IF-EVL-OUTLOOP-NEXT: [[VP_OP_LOAD:%.*]] = call <vscale x 4 x i16> @llvm.vp.load.nxv4i16.p0(ptr align 2 [[TMP14]], <vscale x 4 x i1> shufflevector (<vscale x 4 x i1> insertelement (<vscale x 4 x i1> poison, i1 true, i64 0), <vscale x 4 x i1> poison, <vscale x 4 x i32> zeroinitializer), i32 [[TMP8]])
157+
; IF-EVL-OUTLOOP-NEXT: [[TMP15:%.*]] = sext <vscale x 4 x i16> [[VP_OP_LOAD]] to <vscale x 4 x i32>
158+
; IF-EVL-OUTLOOP-NEXT: [[TMP16]] = add <vscale x 4 x i32> [[VEC_PHI]], [[TMP15]]
159+
; IF-EVL-OUTLOOP-NEXT: [[TMP17:%.*]] = select <vscale x 4 x i1> [[TMP12]], <vscale x 4 x i32> [[TMP16]], <vscale x 4 x i32> [[VEC_PHI]]
160+
; IF-EVL-OUTLOOP-NEXT: [[INDEX_EVL_NEXT]] = add i32 [[TMP8]], [[EVL_BASED_IV]]
157161
; IF-EVL-OUTLOOP-NEXT: [[INDEX_NEXT]] = add i32 [[INDEX]], [[TMP6]]
158-
; IF-EVL-OUTLOOP-NEXT: [[TMP16:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
159-
; IF-EVL-OUTLOOP-NEXT: br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
162+
; IF-EVL-OUTLOOP-NEXT: [[TMP18:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
163+
; IF-EVL-OUTLOOP-NEXT: br i1 [[TMP18]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
160164
; IF-EVL-OUTLOOP: middle.block:
161-
; IF-EVL-OUTLOOP-NEXT: [[TMP17:%.*]] = call i32 @llvm.vector.reduce.add.nxv4i32(<vscale x 4 x i32> [[TMP15]])
165+
; IF-EVL-OUTLOOP-NEXT: [[TMP19:%.*]] = call i32 @llvm.vector.reduce.add.nxv4i32(<vscale x 4 x i32> [[TMP17]])
162166
; IF-EVL-OUTLOOP-NEXT: br i1 true, label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[SCALAR_PH]]
163167
; IF-EVL-OUTLOOP: scalar.ph:
164168
; IF-EVL-OUTLOOP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
165-
; IF-EVL-OUTLOOP-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[FOR_BODY_PREHEADER]] ], [ [[TMP17]], [[MIDDLE_BLOCK]] ]
169+
; IF-EVL-OUTLOOP-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[FOR_BODY_PREHEADER]] ], [ [[TMP19]], [[MIDDLE_BLOCK]] ]
166170
; IF-EVL-OUTLOOP-NEXT: br label [[FOR_BODY:%.*]]
167171
; IF-EVL-OUTLOOP: for.body:
168172
; IF-EVL-OUTLOOP-NEXT: [[I_08:%.*]] = phi i32 [ [[INC:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
169173
; IF-EVL-OUTLOOP-NEXT: [[R_07:%.*]] = phi i32 [ [[ADD:%.*]], [[FOR_BODY]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
170174
; IF-EVL-OUTLOOP-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i16, ptr [[X]], i32 [[I_08]]
171-
; IF-EVL-OUTLOOP-NEXT: [[TMP18:%.*]] = load i16, ptr [[ARRAYIDX]], align 2
172-
; IF-EVL-OUTLOOP-NEXT: [[CONV:%.*]] = sext i16 [[TMP18]] to i32
175+
; IF-EVL-OUTLOOP-NEXT: [[TMP20:%.*]] = load i16, ptr [[ARRAYIDX]], align 2
176+
; IF-EVL-OUTLOOP-NEXT: [[CONV:%.*]] = sext i16 [[TMP20]] to i32
173177
; IF-EVL-OUTLOOP-NEXT: [[ADD]] = add nsw i32 [[R_07]], [[CONV]]
174178
; IF-EVL-OUTLOOP-NEXT: [[INC]] = add nuw nsw i32 [[I_08]], 1
175179
; IF-EVL-OUTLOOP-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INC]], [[N]]
176180
; IF-EVL-OUTLOOP-NEXT: br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
177181
; IF-EVL-OUTLOOP: for.cond.cleanup.loopexit:
178-
; IF-EVL-OUTLOOP-NEXT: [[ADD_LCSSA:%.*]] = phi i32 [ [[ADD]], [[FOR_BODY]] ], [ [[TMP17]], [[MIDDLE_BLOCK]] ]
182+
; IF-EVL-OUTLOOP-NEXT: [[ADD_LCSSA:%.*]] = phi i32 [ [[ADD]], [[FOR_BODY]] ], [ [[TMP19]], [[MIDDLE_BLOCK]] ]
179183
; IF-EVL-OUTLOOP-NEXT: br label [[FOR_COND_CLEANUP]]
180184
; IF-EVL-OUTLOOP: for.cond.cleanup:
181185
; IF-EVL-OUTLOOP-NEXT: [[R_0_LCSSA:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[ADD_LCSSA]], [[FOR_COND_CLEANUP_LOOPEXIT]] ]
@@ -204,42 +208,46 @@ define i32 @add_i16_i32(ptr nocapture readonly %x, i32 %n) {
204208
; IF-EVL-INLOOP-NEXT: br label [[VECTOR_BODY:%.*]]
205209
; IF-EVL-INLOOP: vector.body:
206210
; IF-EVL-INLOOP-NEXT: [[INDEX:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
207-
; IF-EVL-INLOOP-NEXT: [[VEC_PHI:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP16:%.*]], [[VECTOR_BODY]] ]
208-
; IF-EVL-INLOOP-NEXT: [[TMP7:%.*]] = add i32 [[INDEX]], 0
209-
; IF-EVL-INLOOP-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 8 x i32> poison, i32 [[INDEX]], i64 0
211+
; IF-EVL-INLOOP-NEXT: [[EVL_BASED_IV:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[INDEX_EVL_NEXT:%.*]], [[VECTOR_BODY]] ]
212+
; IF-EVL-INLOOP-NEXT: [[VEC_PHI:%.*]] = phi i32 [ 0, [[VECTOR_PH]] ], [ [[TMP18:%.*]], [[VECTOR_BODY]] ]
213+
; IF-EVL-INLOOP-NEXT: [[TMP7:%.*]] = sub i32 [[N]], [[EVL_BASED_IV]]
214+
; IF-EVL-INLOOP-NEXT: [[TMP8:%.*]] = call i32 @llvm.experimental.get.vector.length.i32(i32 [[TMP7]], i32 8, i1 true)
215+
; IF-EVL-INLOOP-NEXT: [[TMP9:%.*]] = add i32 [[EVL_BASED_IV]], 0
216+
; IF-EVL-INLOOP-NEXT: [[BROADCAST_SPLATINSERT:%.*]] = insertelement <vscale x 8 x i32> poison, i32 [[EVL_BASED_IV]], i64 0
210217
; IF-EVL-INLOOP-NEXT: [[BROADCAST_SPLAT:%.*]] = shufflevector <vscale x 8 x i32> [[BROADCAST_SPLATINSERT]], <vscale x 8 x i32> poison, <vscale x 8 x i32> zeroinitializer
211-
; IF-EVL-INLOOP-NEXT: [[TMP8:%.*]] = call <vscale x 8 x i32> @llvm.experimental.stepvector.nxv8i32()
212-
; IF-EVL-INLOOP-NEXT: [[TMP9:%.*]] = add <vscale x 8 x i32> zeroinitializer, [[TMP8]]
213-
; IF-EVL-INLOOP-NEXT: [[VEC_IV:%.*]] = add <vscale x 8 x i32> [[BROADCAST_SPLAT]], [[TMP9]]
214-
; IF-EVL-INLOOP-NEXT: [[TMP10:%.*]] = icmp ule <vscale x 8 x i32> [[VEC_IV]], [[BROADCAST_SPLAT2]]
215-
; IF-EVL-INLOOP-NEXT: [[TMP11:%.*]] = getelementptr inbounds i16, ptr [[X:%.*]], i32 [[TMP7]]
216-
; IF-EVL-INLOOP-NEXT: [[TMP12:%.*]] = getelementptr inbounds i16, ptr [[TMP11]], i32 0
217-
; IF-EVL-INLOOP-NEXT: [[WIDE_MASKED_LOAD:%.*]] = call <vscale x 8 x i16> @llvm.masked.load.nxv8i16.p0(ptr [[TMP12]], i32 2, <vscale x 8 x i1> [[TMP10]], <vscale x 8 x i16> poison)
218-
; IF-EVL-INLOOP-NEXT: [[TMP13:%.*]] = sext <vscale x 8 x i16> [[WIDE_MASKED_LOAD]] to <vscale x 8 x i32>
219-
; IF-EVL-INLOOP-NEXT: [[TMP14:%.*]] = select <vscale x 8 x i1> [[TMP10]], <vscale x 8 x i32> [[TMP13]], <vscale x 8 x i32> zeroinitializer
220-
; IF-EVL-INLOOP-NEXT: [[TMP15:%.*]] = call i32 @llvm.vector.reduce.add.nxv8i32(<vscale x 8 x i32> [[TMP14]])
221-
; IF-EVL-INLOOP-NEXT: [[TMP16]] = add i32 [[TMP15]], [[VEC_PHI]]
218+
; IF-EVL-INLOOP-NEXT: [[TMP10:%.*]] = call <vscale x 8 x i32> @llvm.experimental.stepvector.nxv8i32()
219+
; IF-EVL-INLOOP-NEXT: [[TMP11:%.*]] = add <vscale x 8 x i32> zeroinitializer, [[TMP10]]
220+
; IF-EVL-INLOOP-NEXT: [[VEC_IV:%.*]] = add <vscale x 8 x i32> [[BROADCAST_SPLAT]], [[TMP11]]
221+
; IF-EVL-INLOOP-NEXT: [[TMP12:%.*]] = icmp ule <vscale x 8 x i32> [[VEC_IV]], [[BROADCAST_SPLAT2]]
222+
; IF-EVL-INLOOP-NEXT: [[TMP13:%.*]] = getelementptr inbounds i16, ptr [[X:%.*]], i32 [[TMP9]]
223+
; IF-EVL-INLOOP-NEXT: [[TMP14:%.*]] = getelementptr inbounds i16, ptr [[TMP13]], i32 0
224+
; IF-EVL-INLOOP-NEXT: [[VP_OP_LOAD:%.*]] = call <vscale x 8 x i16> @llvm.vp.load.nxv8i16.p0(ptr align 2 [[TMP14]], <vscale x 8 x i1> shufflevector (<vscale x 8 x i1> insertelement (<vscale x 8 x i1> poison, i1 true, i64 0), <vscale x 8 x i1> poison, <vscale x 8 x i32> zeroinitializer), i32 [[TMP8]])
225+
; IF-EVL-INLOOP-NEXT: [[TMP15:%.*]] = sext <vscale x 8 x i16> [[VP_OP_LOAD]] to <vscale x 8 x i32>
226+
; IF-EVL-INLOOP-NEXT: [[TMP16:%.*]] = select <vscale x 8 x i1> [[TMP12]], <vscale x 8 x i32> [[TMP15]], <vscale x 8 x i32> zeroinitializer
227+
; IF-EVL-INLOOP-NEXT: [[TMP17:%.*]] = call i32 @llvm.vector.reduce.add.nxv8i32(<vscale x 8 x i32> [[TMP16]])
228+
; IF-EVL-INLOOP-NEXT: [[TMP18]] = add i32 [[TMP17]], [[VEC_PHI]]
229+
; IF-EVL-INLOOP-NEXT: [[INDEX_EVL_NEXT]] = add i32 [[TMP8]], [[EVL_BASED_IV]]
222230
; IF-EVL-INLOOP-NEXT: [[INDEX_NEXT]] = add i32 [[INDEX]], [[TMP6]]
223-
; IF-EVL-INLOOP-NEXT: [[TMP17:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
224-
; IF-EVL-INLOOP-NEXT: br i1 [[TMP17]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
231+
; IF-EVL-INLOOP-NEXT: [[TMP19:%.*]] = icmp eq i32 [[INDEX_NEXT]], [[N_VEC]]
232+
; IF-EVL-INLOOP-NEXT: br i1 [[TMP19]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP0:![0-9]+]]
225233
; IF-EVL-INLOOP: middle.block:
226234
; IF-EVL-INLOOP-NEXT: br i1 true, label [[FOR_COND_CLEANUP_LOOPEXIT:%.*]], label [[SCALAR_PH]]
227235
; IF-EVL-INLOOP: scalar.ph:
228236
; IF-EVL-INLOOP-NEXT: [[BC_RESUME_VAL:%.*]] = phi i32 [ [[N_VEC]], [[MIDDLE_BLOCK]] ], [ 0, [[FOR_BODY_PREHEADER]] ]
229-
; IF-EVL-INLOOP-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[FOR_BODY_PREHEADER]] ], [ [[TMP16]], [[MIDDLE_BLOCK]] ]
237+
; IF-EVL-INLOOP-NEXT: [[BC_MERGE_RDX:%.*]] = phi i32 [ 0, [[FOR_BODY_PREHEADER]] ], [ [[TMP18]], [[MIDDLE_BLOCK]] ]
230238
; IF-EVL-INLOOP-NEXT: br label [[FOR_BODY:%.*]]
231239
; IF-EVL-INLOOP: for.body:
232240
; IF-EVL-INLOOP-NEXT: [[I_08:%.*]] = phi i32 [ [[INC:%.*]], [[FOR_BODY]] ], [ [[BC_RESUME_VAL]], [[SCALAR_PH]] ]
233241
; IF-EVL-INLOOP-NEXT: [[R_07:%.*]] = phi i32 [ [[ADD:%.*]], [[FOR_BODY]] ], [ [[BC_MERGE_RDX]], [[SCALAR_PH]] ]
234242
; IF-EVL-INLOOP-NEXT: [[ARRAYIDX:%.*]] = getelementptr inbounds i16, ptr [[X]], i32 [[I_08]]
235-
; IF-EVL-INLOOP-NEXT: [[TMP18:%.*]] = load i16, ptr [[ARRAYIDX]], align 2
236-
; IF-EVL-INLOOP-NEXT: [[CONV:%.*]] = sext i16 [[TMP18]] to i32
243+
; IF-EVL-INLOOP-NEXT: [[TMP20:%.*]] = load i16, ptr [[ARRAYIDX]], align 2
244+
; IF-EVL-INLOOP-NEXT: [[CONV:%.*]] = sext i16 [[TMP20]] to i32
237245
; IF-EVL-INLOOP-NEXT: [[ADD]] = add nsw i32 [[R_07]], [[CONV]]
238246
; IF-EVL-INLOOP-NEXT: [[INC]] = add nuw nsw i32 [[I_08]], 1
239247
; IF-EVL-INLOOP-NEXT: [[EXITCOND:%.*]] = icmp eq i32 [[INC]], [[N]]
240248
; IF-EVL-INLOOP-NEXT: br i1 [[EXITCOND]], label [[FOR_COND_CLEANUP_LOOPEXIT]], label [[FOR_BODY]], !llvm.loop [[LOOP3:![0-9]+]]
241249
; IF-EVL-INLOOP: for.cond.cleanup.loopexit:
242-
; IF-EVL-INLOOP-NEXT: [[ADD_LCSSA:%.*]] = phi i32 [ [[ADD]], [[FOR_BODY]] ], [ [[TMP16]], [[MIDDLE_BLOCK]] ]
250+
; IF-EVL-INLOOP-NEXT: [[ADD_LCSSA:%.*]] = phi i32 [ [[ADD]], [[FOR_BODY]] ], [ [[TMP18]], [[MIDDLE_BLOCK]] ]
243251
; IF-EVL-INLOOP-NEXT: br label [[FOR_COND_CLEANUP]]
244252
; IF-EVL-INLOOP: for.cond.cleanup:
245253
; IF-EVL-INLOOP-NEXT: [[R_0_LCSSA:%.*]] = phi i32 [ 0, [[ENTRY:%.*]] ], [ [[ADD_LCSSA]], [[FOR_COND_CLEANUP_LOOPEXIT]] ]

0 commit comments

Comments
 (0)