Skip to content

Commit 2c7d40b

Browse files
fhahnfhahn
committed
[VPlan] Generalize SCALAR-STEPS removal to any unroll factor.
Follow-up to dfca6c0 to extend isUnrolled handle any unrolled VPlan, which means there's a single UF, but it will be > 1 if unrolling took place.
1 parent 2d1517d commit 2c7d40b

File tree

107 files changed

+778
-1135
lines changed

Some content is hidden

Large Commits have some content hidden by default. Use the searchbox below for content that may be hidden.

107 files changed

+778
-1135
lines changed

‎llvm/lib/Transforms/Vectorize/VPlan.h

Lines changed: 3 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -3598,8 +3598,9 @@ class VPlan {
35983598
UFs.insert(UF);
35993599
}
36003600

3601-
/// Returns true if the VPlan already has been unrolled, i.e. it has UF = 1.
3602-
bool isUnrolled() const { return UFs.size() == 1 && UFs.back() == 1; }
3601+
/// Returns true if the VPlan already has been unrolled, i.e. it has a single
3602+
/// concrete UF.
3603+
bool isUnrolled() const { return UFs.size() == 1; }
36033604

36043605
/// Return a string with the name of the plan and the applicable VFs and UFs.
36053606
std::string getName() const;

‎llvm/test/Transforms/LoopVectorize/AArch64/call-costs.ll

Lines changed: 2 additions & 3 deletions
Original file line numberDiff line numberDiff line change
@@ -13,8 +13,7 @@ define void @fshl_operand_first_order_recurrence(ptr %dst, ptr noalias %src) {
1313
; CHECK: [[VECTOR_BODY]]:
1414
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
1515
; CHECK-NEXT: [[VECTOR_RECUR:%.*]] = phi <2 x i64> [ <i64 poison, i64 0>, %[[VECTOR_PH]] ], [ [[WIDE_LOAD1:%.*]], %[[VECTOR_BODY]] ]
16-
; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0
17-
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[SRC]], i64 [[TMP0]]
16+
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[SRC]], i64 [[INDEX]]
1817
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 0
1918
; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 2
2019
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <2 x i64>, ptr [[TMP4]], align 8
@@ -23,7 +22,7 @@ define void @fshl_operand_first_order_recurrence(ptr %dst, ptr noalias %src) {
2322
; CHECK-NEXT: [[TMP7:%.*]] = shufflevector <2 x i64> [[WIDE_LOAD]], <2 x i64> [[WIDE_LOAD1]], <2 x i32> <i32 1, i32 2>
2423
; CHECK-NEXT: [[TMP8:%.*]] = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> splat (i64 1), <2 x i64> [[TMP6]], <2 x i64> splat (i64 1))
2524
; CHECK-NEXT: [[TMP9:%.*]] = call <2 x i64> @llvm.fshl.v2i64(<2 x i64> splat (i64 1), <2 x i64> [[TMP7]], <2 x i64> splat (i64 1))
26-
; CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[DST]], i64 [[TMP0]]
25+
; CHECK-NEXT: [[TMP10:%.*]] = getelementptr inbounds i64, ptr [[DST]], i64 [[INDEX]]
2726
; CHECK-NEXT: [[TMP12:%.*]] = getelementptr inbounds i64, ptr [[TMP10]], i32 0
2827
; CHECK-NEXT: [[TMP13:%.*]] = getelementptr inbounds i64, ptr [[TMP10]], i32 2
2928
; CHECK-NEXT: store <2 x i64> [[TMP8]], ptr [[TMP12]], align 8

‎llvm/test/Transforms/LoopVectorize/AArch64/conditional-branches-cost.ll

Lines changed: 2 additions & 4 deletions
Original file line numberDiff line numberDiff line change
@@ -81,8 +81,7 @@ define void @loop_dependent_cond(ptr %src, ptr noalias %dst, i64 %N) {
8181
; DEFAULT-NEXT: br label %[[VECTOR_BODY:.*]]
8282
; DEFAULT: [[VECTOR_BODY]]:
8383
; DEFAULT-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[PRED_STORE_CONTINUE7:.*]] ]
84-
; DEFAULT-NEXT: [[TMP1:%.*]] = add i64 [[INDEX]], 0
85-
; DEFAULT-NEXT: [[TMP3:%.*]] = getelementptr double, ptr [[SRC]], i64 [[TMP1]]
84+
; DEFAULT-NEXT: [[TMP3:%.*]] = getelementptr double, ptr [[SRC]], i64 [[INDEX]]
8685
; DEFAULT-NEXT: [[TMP5:%.*]] = getelementptr double, ptr [[TMP3]], i32 0
8786
; DEFAULT-NEXT: [[TMP6:%.*]] = getelementptr double, ptr [[TMP3]], i32 2
8887
; DEFAULT-NEXT: [[WIDE_LOAD:%.*]] = load <2 x double>, ptr [[TMP5]], align 8
@@ -341,8 +340,7 @@ define void @latch_branch_cost(ptr %dst) {
341340
; DEFAULT-NEXT: br label %[[VECTOR_BODY:.*]]
342341
; DEFAULT: [[VECTOR_BODY]]:
343342
; DEFAULT-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
344-
; DEFAULT-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0
345-
; DEFAULT-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[DST]], i64 [[TMP0]]
343+
; DEFAULT-NEXT: [[TMP2:%.*]] = getelementptr i8, ptr [[DST]], i64 [[INDEX]]
346344
; DEFAULT-NEXT: [[TMP6:%.*]] = getelementptr i8, ptr [[TMP2]], i32 0
347345
; DEFAULT-NEXT: [[TMP5:%.*]] = getelementptr i8, ptr [[TMP2]], i32 16
348346
; DEFAULT-NEXT: store <16 x i8> zeroinitializer, ptr [[TMP6]], align 1

‎llvm/test/Transforms/LoopVectorize/AArch64/divs-with-scalable-vfs.ll

Lines changed: 1 addition & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -33,8 +33,7 @@ define void @sdiv_feeding_gep(ptr %dst, i32 %x, i64 %M, i64 %conv6, i64 %N) {
3333
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
3434
; CHECK: [[VECTOR_BODY]]:
3535
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, %[[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], %[[VECTOR_BODY]] ]
36-
; CHECK-NEXT: [[TMP12:%.*]] = add i64 [[INDEX]], 0
37-
; CHECK-NEXT: [[TMP24:%.*]] = sub i64 [[TMP12]], [[TMP22]]
36+
; CHECK-NEXT: [[TMP24:%.*]] = sub i64 [[INDEX]], [[TMP22]]
3837
; CHECK-NEXT: [[TMP26:%.*]] = trunc i64 [[TMP24]] to i32
3938
; CHECK-NEXT: [[TMP30:%.*]] = add i32 [[TMP28]], [[TMP26]]
4039
; CHECK-NEXT: [[TMP32:%.*]] = sext i32 [[TMP30]] to i64

‎llvm/test/Transforms/LoopVectorize/AArch64/epilog-vectorization-factors.ll

Lines changed: 9 additions & 12 deletions
Original file line numberDiff line numberDiff line change
@@ -18,8 +18,7 @@ define void @add_i8(ptr noalias nocapture noundef writeonly %A, ptr nocapture no
1818
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
1919
; CHECK: vector.body:
2020
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
21-
; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0
22-
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[B:%.*]], i64 [[TMP0]]
21+
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[B:%.*]], i64 [[INDEX]]
2322
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
2423
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 16
2524
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 32
@@ -28,7 +27,7 @@ define void @add_i8(ptr noalias nocapture noundef writeonly %A, ptr nocapture no
2827
; CHECK-NEXT: [[WIDE_LOAD2:%.*]] = load <16 x i8>, ptr [[TMP3]], align 1
2928
; CHECK-NEXT: [[WIDE_LOAD3:%.*]] = load <16 x i8>, ptr [[TMP4]], align 1
3029
; CHECK-NEXT: [[WIDE_LOAD4:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
31-
; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[C:%.*]], i64 [[TMP0]]
30+
; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[C:%.*]], i64 [[INDEX]]
3231
; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds i8, ptr [[TMP6]], i32 0
3332
; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds i8, ptr [[TMP6]], i32 16
3433
; CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds i8, ptr [[TMP6]], i32 32
@@ -41,7 +40,7 @@ define void @add_i8(ptr noalias nocapture noundef writeonly %A, ptr nocapture no
4140
; CHECK-NEXT: [[TMP12:%.*]] = add <16 x i8> [[WIDE_LOAD6]], [[WIDE_LOAD2]]
4241
; CHECK-NEXT: [[TMP13:%.*]] = add <16 x i8> [[WIDE_LOAD7]], [[WIDE_LOAD3]]
4342
; CHECK-NEXT: [[TMP14:%.*]] = add <16 x i8> [[WIDE_LOAD8]], [[WIDE_LOAD4]]
44-
; CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[TMP0]]
43+
; CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[INDEX]]
4544
; CHECK-NEXT: [[TMP16:%.*]] = getelementptr inbounds i8, ptr [[TMP15]], i32 0
4645
; CHECK-NEXT: [[TMP17:%.*]] = getelementptr inbounds i8, ptr [[TMP15]], i32 16
4746
; CHECK-NEXT: [[TMP18:%.*]] = getelementptr inbounds i8, ptr [[TMP15]], i32 32
@@ -135,8 +134,7 @@ define void @add_i16(ptr noalias nocapture noundef writeonly %A, ptr nocapture n
135134
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
136135
; CHECK: vector.body:
137136
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
138-
; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0
139-
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i16, ptr [[B:%.*]], i64 [[TMP0]]
137+
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i16, ptr [[B:%.*]], i64 [[INDEX]]
140138
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i16, ptr [[TMP1]], i32 0
141139
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i16, ptr [[TMP1]], i32 8
142140
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i16, ptr [[TMP1]], i32 16
@@ -145,7 +143,7 @@ define void @add_i16(ptr noalias nocapture noundef writeonly %A, ptr nocapture n
145143
; CHECK-NEXT: [[WIDE_LOAD2:%.*]] = load <8 x i16>, ptr [[TMP3]], align 1
146144
; CHECK-NEXT: [[WIDE_LOAD3:%.*]] = load <8 x i16>, ptr [[TMP4]], align 1
147145
; CHECK-NEXT: [[WIDE_LOAD4:%.*]] = load <8 x i16>, ptr [[TMP5]], align 1
148-
; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i16, ptr [[C:%.*]], i64 [[TMP0]]
146+
; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i16, ptr [[C:%.*]], i64 [[INDEX]]
149147
; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds i16, ptr [[TMP6]], i32 0
150148
; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds i16, ptr [[TMP6]], i32 8
151149
; CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds i16, ptr [[TMP6]], i32 16
@@ -158,7 +156,7 @@ define void @add_i16(ptr noalias nocapture noundef writeonly %A, ptr nocapture n
158156
; CHECK-NEXT: [[TMP12:%.*]] = add <8 x i16> [[WIDE_LOAD6]], [[WIDE_LOAD2]]
159157
; CHECK-NEXT: [[TMP13:%.*]] = add <8 x i16> [[WIDE_LOAD7]], [[WIDE_LOAD3]]
160158
; CHECK-NEXT: [[TMP14:%.*]] = add <8 x i16> [[WIDE_LOAD8]], [[WIDE_LOAD4]]
161-
; CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds i16, ptr [[A:%.*]], i64 [[TMP0]]
159+
; CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds i16, ptr [[A:%.*]], i64 [[INDEX]]
162160
; CHECK-NEXT: [[TMP16:%.*]] = getelementptr inbounds i16, ptr [[TMP15]], i32 0
163161
; CHECK-NEXT: [[TMP17:%.*]] = getelementptr inbounds i16, ptr [[TMP15]], i32 8
164162
; CHECK-NEXT: [[TMP18:%.*]] = getelementptr inbounds i16, ptr [[TMP15]], i32 16
@@ -252,8 +250,7 @@ define void @add_i32(ptr noalias nocapture noundef writeonly %A, ptr nocapture n
252250
; CHECK-NEXT: br label [[VECTOR_BODY:%.*]]
253251
; CHECK: vector.body:
254252
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
255-
; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0
256-
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[TMP0]]
253+
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i32, ptr [[B:%.*]], i64 [[INDEX]]
257254
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 0
258255
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 4
259256
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i32, ptr [[TMP1]], i32 8
@@ -262,7 +259,7 @@ define void @add_i32(ptr noalias nocapture noundef writeonly %A, ptr nocapture n
262259
; CHECK-NEXT: [[WIDE_LOAD2:%.*]] = load <4 x i32>, ptr [[TMP3]], align 1
263260
; CHECK-NEXT: [[WIDE_LOAD3:%.*]] = load <4 x i32>, ptr [[TMP4]], align 1
264261
; CHECK-NEXT: [[WIDE_LOAD4:%.*]] = load <4 x i32>, ptr [[TMP5]], align 1
265-
; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[C:%.*]], i64 [[TMP0]]
262+
; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i32, ptr [[C:%.*]], i64 [[INDEX]]
266263
; CHECK-NEXT: [[TMP7:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 0
267264
; CHECK-NEXT: [[TMP8:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 4
268265
; CHECK-NEXT: [[TMP9:%.*]] = getelementptr inbounds i32, ptr [[TMP6]], i32 8
@@ -275,7 +272,7 @@ define void @add_i32(ptr noalias nocapture noundef writeonly %A, ptr nocapture n
275272
; CHECK-NEXT: [[TMP12:%.*]] = add <4 x i32> [[WIDE_LOAD6]], [[WIDE_LOAD2]]
276273
; CHECK-NEXT: [[TMP13:%.*]] = add <4 x i32> [[WIDE_LOAD7]], [[WIDE_LOAD3]]
277274
; CHECK-NEXT: [[TMP14:%.*]] = add <4 x i32> [[WIDE_LOAD8]], [[WIDE_LOAD4]]
278-
; CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[TMP0]]
275+
; CHECK-NEXT: [[TMP15:%.*]] = getelementptr inbounds i32, ptr [[A:%.*]], i64 [[INDEX]]
279276
; CHECK-NEXT: [[TMP16:%.*]] = getelementptr inbounds i32, ptr [[TMP15]], i32 0
280277
; CHECK-NEXT: [[TMP17:%.*]] = getelementptr inbounds i32, ptr [[TMP15]], i32 4
281278
; CHECK-NEXT: [[TMP18:%.*]] = getelementptr inbounds i32, ptr [[TMP15]], i32 8

‎llvm/test/Transforms/LoopVectorize/AArch64/epilog-vectorization-widen-inductions.ll

Lines changed: 5 additions & 10 deletions
Original file line numberDiff line numberDiff line change
@@ -125,8 +125,7 @@ define void @test_widen_induction(ptr %A, i64 %N) {
125125
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
126126
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
127127
; CHECK-NEXT: [[STEP_ADD:%.*]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
128-
; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0
129-
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[TMP0]]
128+
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
130129
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
131130
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 2
132131
; CHECK-NEXT: store <2 x i64> [[VEC_IND]], ptr [[TMP2]], align 4
@@ -213,8 +212,7 @@ define void @test_widen_induction_variable_start(ptr %A, i64 %N, i64 %start) {
213212
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <2 x i64> [ [[INDUCTION]], [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
214213
; CHECK-NEXT: [[STEP_ADD:%.*]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
215214
; CHECK-NEXT: [[OFFSET_IDX:%.*]] = add i64 [[START]], [[INDEX]]
216-
; CHECK-NEXT: [[TMP1:%.*]] = add i64 [[OFFSET_IDX]], 0
217-
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[TMP1]]
215+
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[OFFSET_IDX]]
218216
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 0
219217
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP2]], i32 2
220218
; CHECK-NEXT: store <2 x i64> [[VEC_IND]], ptr [[TMP3]], align 4
@@ -299,8 +297,7 @@ define void @test_widen_induction_step_2(ptr %A, i64 %N, i32 %step) {
299297
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
300298
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <2 x i64> [ <i64 0, i64 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
301299
; CHECK-NEXT: [[STEP_ADD:%.*]] = add <2 x i64> [[VEC_IND]], splat (i64 2)
302-
; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0
303-
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[TMP0]]
300+
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i64, ptr [[A:%.*]], i64 [[INDEX]]
304301
; CHECK-NEXT: [[TMP2:%.*]] = add <2 x i64> [[VEC_IND]], splat (i64 10)
305302
; CHECK-NEXT: [[TMP3:%.*]] = add <2 x i64> [[STEP_ADD]], splat (i64 10)
306303
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i64, ptr [[TMP1]], i32 0
@@ -390,8 +387,7 @@ define void @test_widen_extended_induction(ptr %dst) {
390387
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <2 x i8> [ <i8 0, i8 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
391388
; CHECK-NEXT: [[STEP_ADD:%.*]] = add <2 x i8> [[VEC_IND]], splat (i8 2)
392389
; CHECK-NEXT: [[OFFSET_IDX:%.*]] = trunc i32 [[INDEX]] to i8
393-
; CHECK-NEXT: [[TMP0:%.*]] = add i8 [[OFFSET_IDX]], 0
394-
; CHECK-NEXT: [[TMP1:%.*]] = zext i8 [[TMP0]] to i64
390+
; CHECK-NEXT: [[TMP1:%.*]] = zext i8 [[OFFSET_IDX]] to i64
395391
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds [6 x i8], ptr [[DST:%.*]], i64 0, i64 [[TMP1]]
396392
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i32 0
397393
; CHECK-NEXT: [[TMP4:%.*]] = getelementptr inbounds i8, ptr [[TMP2]], i32 2
@@ -470,8 +466,7 @@ define void @test_widen_truncated_induction(ptr %A) {
470466
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
471467
; CHECK-NEXT: [[VEC_IND:%.*]] = phi <2 x i8> [ <i8 0, i8 1>, [[VECTOR_PH]] ], [ [[VEC_IND_NEXT:%.*]], [[VECTOR_BODY]] ]
472468
; CHECK-NEXT: [[STEP_ADD:%.*]] = add <2 x i8> [[VEC_IND]], splat (i8 2)
473-
; CHECK-NEXT: [[TMP0:%.*]] = add i64 [[INDEX]], 0
474-
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[TMP0]]
469+
; CHECK-NEXT: [[TMP1:%.*]] = getelementptr inbounds i8, ptr [[A:%.*]], i64 [[INDEX]]
475470
; CHECK-NEXT: [[TMP2:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 0
476471
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[TMP1]], i32 2
477472
; CHECK-NEXT: store <2 x i8> [[VEC_IND]], ptr [[TMP2]], align 1

‎llvm/test/Transforms/LoopVectorize/AArch64/fixed-order-recurrence.ll

Lines changed: 4 additions & 6 deletions
Original file line numberDiff line numberDiff line change
@@ -25,8 +25,7 @@ define void @firstorderrec(ptr nocapture noundef readonly %x, ptr noalias nocapt
2525
; CHECK-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
2626
; CHECK-NEXT: [[VECTOR_RECUR:%.*]] = phi <16 x i8> [ [[VECTOR_RECUR_INIT]], [[VECTOR_PH]] ], [ [[WIDE_LOAD1:%.*]], [[VECTOR_BODY]] ]
2727
; CHECK-NEXT: [[OFFSET_IDX:%.*]] = add i64 1, [[INDEX]]
28-
; CHECK-NEXT: [[TMP1:%.*]] = add i64 [[OFFSET_IDX]], 0
29-
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[X]], i64 [[TMP1]]
28+
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[X]], i64 [[OFFSET_IDX]]
3029
; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
3130
; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 16
3231
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
@@ -35,7 +34,7 @@ define void @firstorderrec(ptr nocapture noundef readonly %x, ptr noalias nocapt
3534
; CHECK-NEXT: [[TMP8:%.*]] = shufflevector <16 x i8> [[WIDE_LOAD]], <16 x i8> [[WIDE_LOAD1]], <16 x i32> <i32 15, i32 16, i32 17, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30>
3635
; CHECK-NEXT: [[TMP9:%.*]] = add <16 x i8> [[WIDE_LOAD]], [[TMP7]]
3736
; CHECK-NEXT: [[TMP10:%.*]] = add <16 x i8> [[WIDE_LOAD1]], [[TMP8]]
38-
; CHECK-NEXT: [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[Y:%.*]], i64 [[TMP1]]
37+
; CHECK-NEXT: [[TMP11:%.*]] = getelementptr inbounds i8, ptr [[Y:%.*]], i64 [[OFFSET_IDX]]
3938
; CHECK-NEXT: [[TMP13:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i32 0
4039
; CHECK-NEXT: [[TMP14:%.*]] = getelementptr inbounds i8, ptr [[TMP11]], i32 16
4140
; CHECK-NEXT: store <16 x i8> [[TMP9]], ptr [[TMP13]], align 1
@@ -121,8 +120,7 @@ define void @thirdorderrec(ptr nocapture noundef readonly %x, ptr noalias nocapt
121120
; CHECK-NEXT: [[VECTOR_RECUR2:%.*]] = phi <16 x i8> [ [[VECTOR_RECUR_INIT1]], [[VECTOR_PH]] ], [ [[TMP8:%.*]], [[VECTOR_BODY]] ]
122121
; CHECK-NEXT: [[VECTOR_RECUR4:%.*]] = phi <16 x i8> [ [[VECTOR_RECUR_INIT3]], [[VECTOR_PH]] ], [ [[TMP10:%.*]], [[VECTOR_BODY]] ]
123122
; CHECK-NEXT: [[OFFSET_IDX:%.*]] = add i64 3, [[INDEX]]
124-
; CHECK-NEXT: [[TMP1:%.*]] = add i64 [[OFFSET_IDX]], 0
125-
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[X]], i64 [[TMP1]]
123+
; CHECK-NEXT: [[TMP3:%.*]] = getelementptr inbounds i8, ptr [[X]], i64 [[OFFSET_IDX]]
126124
; CHECK-NEXT: [[TMP5:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 0
127125
; CHECK-NEXT: [[TMP6:%.*]] = getelementptr inbounds i8, ptr [[TMP3]], i32 16
128126
; CHECK-NEXT: [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP5]], align 1
@@ -139,7 +137,7 @@ define void @thirdorderrec(ptr nocapture noundef readonly %x, ptr noalias nocapt
139137
; CHECK-NEXT: [[TMP16:%.*]] = add <16 x i8> [[TMP14]], [[TMP8]]
140138
; CHECK-NEXT: [[TMP17:%.*]] = add <16 x i8> [[TMP15]], [[WIDE_LOAD]]
141139
; CHECK-NEXT: [[TMP18:%.*]] = add <16 x i8> [[TMP16]], [[WIDE_LOAD5]]
142-
; CHECK-NEXT: [[TMP19:%.*]] = getelementptr inbounds i8, ptr [[Y:%.*]], i64 [[TMP1]]
140+
; CHECK-NEXT: [[TMP19:%.*]] = getelementptr inbounds i8, ptr [[Y:%.*]], i64 [[OFFSET_IDX]]
143141
; CHECK-NEXT: [[TMP21:%.*]] = getelementptr inbounds i8, ptr [[TMP19]], i32 0
144142
; CHECK-NEXT: [[TMP22:%.*]] = getelementptr inbounds i8, ptr [[TMP19]], i32 16
145143
; CHECK-NEXT: store <16 x i8> [[TMP17]], ptr [[TMP21]], align 1

0 commit comments

Comments
 (0)