Skip to content

Commit 1ab9b92

Browse files
fhahnfhahn
committed
!fixup update tests.
1 parent 9d15429 commit 1ab9b92

File tree

2 files changed

+32
-73
lines changed

2 files changed

+32
-73
lines changed

‎llvm/test/Transforms/LoopVectorize/AArch64/interleave-with-gaps.ll‎

Lines changed: 0 additions & 51 deletions
Original file line numberDiff line numberDiff line change
@@ -288,56 +288,6 @@ define void @main_vector_loop_fixed_single_vector_iteration_with_runtime_checks(
288288
; CHECK-LABEL: define void @main_vector_loop_fixed_single_vector_iteration_with_runtime_checks(
289289
; CHECK-SAME: ptr noalias [[A:%.*]], ptr noalias [[B:%.*]], ptr noalias [[C:%.*]], ptr noalias [[D:%.*]], ptr noalias [[E:%.*]], ptr noalias [[F:%.*]], ptr noalias [[G:%.*]], ptr noalias [[H:%.*]], ptr noalias [[I:%.*]], ptr noalias [[J:%.*]], ptr noalias [[K:%.*]], ptr [[L:%.*]]) #[[ATTR1:[0-9]+]] {
290290
; CHECK-NEXT: [[ENTRY:.*]]:
291-
; CHECK-NEXT: br i1 true, label %[[SCALAR_PH:.*]], label %[[VECTOR_PH:.*]]
292-
; CHECK: [[VECTOR_PH]]:
293-
; CHECK-NEXT: br label %[[VECTOR_BODY:.*]]
294-
; CHECK: [[VECTOR_BODY]]:
295-
; CHECK-NEXT: [[TMP0:%.*]] = getelementptr i64, ptr [[J]], i64 0
296-
; CHECK-NEXT: [[WIDE_VEC:%.*]] = load <16 x i64>, ptr [[TMP0]], align 8
297-
; CHECK-NEXT: [[STRIDED_VEC:%.*]] = shufflevector <16 x i64> [[WIDE_VEC]], <16 x i64> poison, <8 x i32> <i32 0, i32 2, i32 4, i32 6, i32 8, i32 10, i32 12, i32 14>
298-
; CHECK-NEXT: [[TMP1:%.*]] = trunc <8 x i64> [[STRIDED_VEC]] to <8 x i16>
299-
; CHECK-NEXT: [[TMP6:%.*]] = getelementptr i16, ptr [[K]], i64 0
300-
; CHECK-NEXT: [[TMP7:%.*]] = getelementptr i16, ptr [[K]], i64 2
301-
; CHECK-NEXT: [[TMP8:%.*]] = getelementptr i16, ptr [[K]], i64 4
302-
; CHECK-NEXT: [[TMP9:%.*]] = getelementptr i16, ptr [[K]], i64 6
303-
; CHECK-NEXT: [[TMP10:%.*]] = getelementptr i16, ptr [[K]], i64 8
304-
; CHECK-NEXT: [[TMP11:%.*]] = getelementptr i16, ptr [[K]], i64 10
305-
; CHECK-NEXT: [[TMP12:%.*]] = getelementptr i16, ptr [[K]], i64 12
306-
; CHECK-NEXT: [[TMP13:%.*]] = getelementptr i16, ptr [[K]], i64 14
307-
; CHECK-NEXT: [[TMP14:%.*]] = extractelement <8 x i16> [[TMP1]], i32 0
308-
; CHECK-NEXT: store i16 [[TMP14]], ptr [[TMP6]], align 2
309-
; CHECK-NEXT: [[TMP15:%.*]] = extractelement <8 x i16> [[TMP1]], i32 1
310-
; CHECK-NEXT: store i16 [[TMP15]], ptr [[TMP7]], align 2
311-
; CHECK-NEXT: [[TMP16:%.*]] = extractelement <8 x i16> [[TMP1]], i32 2
312-
; CHECK-NEXT: store i16 [[TMP16]], ptr [[TMP8]], align 2
313-
; CHECK-NEXT: [[TMP17:%.*]] = extractelement <8 x i16> [[TMP1]], i32 3
314-
; CHECK-NEXT: store i16 [[TMP17]], ptr [[TMP9]], align 2
315-
; CHECK-NEXT: [[TMP18:%.*]] = extractelement <8 x i16> [[TMP1]], i32 4
316-
; CHECK-NEXT: store i16 [[TMP18]], ptr [[TMP10]], align 2
317-
; CHECK-NEXT: [[TMP19:%.*]] = extractelement <8 x i16> [[TMP1]], i32 5
318-
; CHECK-NEXT: store i16 [[TMP19]], ptr [[TMP11]], align 2
319-
; CHECK-NEXT: [[TMP20:%.*]] = extractelement <8 x i16> [[TMP1]], i32 6
320-
; CHECK-NEXT: store i16 [[TMP20]], ptr [[TMP12]], align 2
321-
; CHECK-NEXT: [[TMP21:%.*]] = extractelement <8 x i16> [[TMP1]], i32 7
322-
; CHECK-NEXT: store i16 [[TMP21]], ptr [[TMP13]], align 2
323-
; CHECK-NEXT: store i64 0, ptr [[A]], align 8
324-
; CHECK-NEXT: store i64 0, ptr [[B]], align 8
325-
; CHECK-NEXT: store i64 0, ptr [[C]], align 8
326-
; CHECK-NEXT: store i64 0, ptr [[D]], align 8
327-
; CHECK-NEXT: store i64 0, ptr [[E]], align 8
328-
; CHECK-NEXT: store i64 0, ptr [[F]], align 8
329-
; CHECK-NEXT: store i64 0, ptr [[G]], align 8
330-
; CHECK-NEXT: store i64 0, ptr [[H]], align 8
331-
; CHECK-NEXT: store i64 0, ptr [[I]], align 8
332-
; CHECK-NEXT: store i64 0, ptr [[L]], align 8
333-
; CHECK-NEXT: br label %[[MIDDLE_BLOCK:.*]]
334-
; CHECK: [[MIDDLE_BLOCK]]:
335-
; CHECK-NEXT: br label %[[SCALAR_PH]]
336-
; CHECK: [[SCALAR_PH]]:
337-
; CHECK-NEXT: [[BC_RESUME_VAL:%.*]] = phi i64 [ 0, %[[MIDDLE_BLOCK]] ], [ 0, %[[ENTRY]] ]
338-
; CHECK-NEXT: br label %[[LOOP:.*]]
339-
; CHECK: [[LOOP]]:
340-
; CHECK-NEXT: [[IV:%.*]] = phi i64 [ [[BC_RESUME_VAL]], %[[SCALAR_PH]] ], [ [[IV_NEXT:%.*]], %[[LOOP]] ]
341291
; CHECK-NEXT: [[GEP_J:%.*]] = getelementptr i64, ptr [[J]], i64 [[IV]]
342292
; CHECK-NEXT: [[L_J:%.*]] = load i64, ptr [[GEP_J]], align 8
343293
; CHECK-NEXT: [[L_TRUNC:%.*]] = trunc i64 [[L_J]] to i16
@@ -355,7 +305,6 @@ define void @main_vector_loop_fixed_single_vector_iteration_with_runtime_checks(
355305
; CHECK-NEXT: store i64 0, ptr [[L]], align 8
356306
; CHECK-NEXT: [[IV_NEXT]] = add i64 [[IV]], 2
357307
; CHECK-NEXT: [[EC:%.*]] = icmp ult i64 [[IV]], 14
358-
; CHECK-NEXT: br i1 [[EC]], label %[[LOOP]], label %[[EXIT:.*]], !llvm.loop [[LOOP8:![0-9]+]]
359308
; CHECK: [[EXIT]]:
360309
; CHECK-NEXT: ret void
361310
;

‎llvm/test/Transforms/LoopVectorize/AArch64/partial-reduce-dot-product.ll‎

Lines changed: 32 additions & 22 deletions
Original file line numberDiff line numberDiff line change
@@ -1816,13 +1816,12 @@ define i64 @dotp_cost_disagreement(ptr %a, ptr %b) #0 {
18161816
; CHECK-INTERLEAVE1-NEXT: entry:
18171817
; CHECK-INTERLEAVE1-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
18181818
; CHECK-INTERLEAVE1-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 2
1819-
; CHECK-INTERLEAVE1-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 16, [[TMP1]]
1820-
; CHECK-INTERLEAVE1-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
1819+
; CHECK-INTERLEAVE1-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
18211820
; CHECK-INTERLEAVE1: vector.ph:
18221821
; CHECK-INTERLEAVE1-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
18231822
; CHECK-INTERLEAVE1-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 2
1824-
; CHECK-INTERLEAVE1-NEXT: [[N_MOD_VF:%.*]] = urem i64 16, [[TMP3]]
1825-
; CHECK-INTERLEAVE1-NEXT: [[N_VEC:%.*]] = sub i64 16, [[N_MOD_VF]]
1823+
; CHECK-INTERLEAVE1-NEXT: [[N_MOD_VF:%.*]] = urem i64 41, [[TMP3]]
1824+
; CHECK-INTERLEAVE1-NEXT: [[N_VEC:%.*]] = sub i64 41, [[N_MOD_VF]]
18261825
; CHECK-INTERLEAVE1-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
18271826
; CHECK-INTERLEAVE1-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 2
18281827
; CHECK-INTERLEAVE1-NEXT: br label [[VECTOR_BODY:%.*]]
@@ -1845,7 +1844,7 @@ define i64 @dotp_cost_disagreement(ptr %a, ptr %b) #0 {
18451844
; CHECK-INTERLEAVE1-NEXT: br i1 [[TMP16]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
18461845
; CHECK-INTERLEAVE1: middle.block:
18471846
; CHECK-INTERLEAVE1-NEXT: [[TMP17:%.*]] = call i64 @llvm.vector.reduce.add.nxv2i64(<vscale x 2 x i64> [[TMP15]])
1848-
; CHECK-INTERLEAVE1-NEXT: [[CMP_N:%.*]] = icmp eq i64 16, [[N_VEC]]
1847+
; CHECK-INTERLEAVE1-NEXT: [[CMP_N:%.*]] = icmp eq i64 41, [[N_VEC]]
18491848
; CHECK-INTERLEAVE1-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
18501849
; CHECK-INTERLEAVE1: scalar.ph:
18511850
;
@@ -1854,13 +1853,13 @@ define i64 @dotp_cost_disagreement(ptr %a, ptr %b) #0 {
18541853
; CHECK-INTERLEAVED-NEXT: entry:
18551854
; CHECK-INTERLEAVED-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
18561855
; CHECK-INTERLEAVED-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 4
1857-
; CHECK-INTERLEAVED-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 16, [[TMP1]]
1856+
; CHECK-INTERLEAVED-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 41, [[TMP1]]
18581857
; CHECK-INTERLEAVED-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
18591858
; CHECK-INTERLEAVED: vector.ph:
18601859
; CHECK-INTERLEAVED-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
18611860
; CHECK-INTERLEAVED-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 4
1862-
; CHECK-INTERLEAVED-NEXT: [[N_MOD_VF:%.*]] = urem i64 16, [[TMP3]]
1863-
; CHECK-INTERLEAVED-NEXT: [[N_VEC:%.*]] = sub i64 16, [[N_MOD_VF]]
1861+
; CHECK-INTERLEAVED-NEXT: [[N_MOD_VF:%.*]] = urem i64 41, [[TMP3]]
1862+
; CHECK-INTERLEAVED-NEXT: [[N_VEC:%.*]] = sub i64 41, [[N_MOD_VF]]
18641863
; CHECK-INTERLEAVED-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
18651864
; CHECK-INTERLEAVED-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 4
18661865
; CHECK-INTERLEAVED-NEXT: br label [[VECTOR_BODY:%.*]]
@@ -1897,35 +1896,46 @@ define i64 @dotp_cost_disagreement(ptr %a, ptr %b) #0 {
18971896
; CHECK-INTERLEAVED: middle.block:
18981897
; CHECK-INTERLEAVED-NEXT: [[BIN_RDX:%.*]] = add <vscale x 2 x i64> [[TMP25]], [[TMP24]]
18991898
; CHECK-INTERLEAVED-NEXT: [[TMP27:%.*]] = call i64 @llvm.vector.reduce.add.nxv2i64(<vscale x 2 x i64> [[BIN_RDX]])
1900-
; CHECK-INTERLEAVED-NEXT: [[CMP_N:%.*]] = icmp eq i64 16, [[N_VEC]]
1899+
; CHECK-INTERLEAVED-NEXT: [[CMP_N:%.*]] = icmp eq i64 41, [[N_VEC]]
19011900
; CHECK-INTERLEAVED-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
19021901
; CHECK-INTERLEAVED: scalar.ph:
19031902
;
19041903
; CHECK-MAXBW-LABEL: define i64 @dotp_cost_disagreement(
19051904
; CHECK-MAXBW-SAME: ptr [[A:%.*]], ptr [[B:%.*]]) #[[ATTR0]] {
19061905
; CHECK-MAXBW-NEXT: entry:
1907-
; CHECK-MAXBW-NEXT: br i1 false, label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
1906+
; CHECK-MAXBW-NEXT: [[TMP0:%.*]] = call i64 @llvm.vscale.i64()
1907+
; CHECK-MAXBW-NEXT: [[TMP1:%.*]] = mul nuw i64 [[TMP0]], 8
1908+
; CHECK-MAXBW-NEXT: [[MIN_ITERS_CHECK:%.*]] = icmp ult i64 41, [[TMP1]]
1909+
; CHECK-MAXBW-NEXT: br i1 [[MIN_ITERS_CHECK]], label [[SCALAR_PH:%.*]], label [[VECTOR_PH:%.*]]
19081910
; CHECK-MAXBW: vector.ph:
1911+
; CHECK-MAXBW-NEXT: [[TMP2:%.*]] = call i64 @llvm.vscale.i64()
1912+
; CHECK-MAXBW-NEXT: [[TMP3:%.*]] = mul nuw i64 [[TMP2]], 8
1913+
; CHECK-MAXBW-NEXT: [[N_MOD_VF:%.*]] = urem i64 41, [[TMP3]]
1914+
; CHECK-MAXBW-NEXT: [[N_VEC:%.*]] = sub i64 41, [[N_MOD_VF]]
1915+
; CHECK-MAXBW-NEXT: [[TMP4:%.*]] = call i64 @llvm.vscale.i64()
1916+
; CHECK-MAXBW-NEXT: [[TMP5:%.*]] = mul nuw i64 [[TMP4]], 8
19091917
; CHECK-MAXBW-NEXT: br label [[VECTOR_BODY:%.*]]
19101918
; CHECK-MAXBW: vector.body:
19111919
; CHECK-MAXBW-NEXT: [[INDEX:%.*]] = phi i64 [ 0, [[VECTOR_PH]] ], [ [[INDEX_NEXT:%.*]], [[VECTOR_BODY]] ]
1912-
; CHECK-MAXBW-NEXT: [[VEC_PHI:%.*]] = phi <16 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP13:%.*]], [[VECTOR_BODY]] ]
1920+
; CHECK-MAXBW-NEXT: [[VEC_PHI:%.*]] = phi <vscale x 8 x i64> [ zeroinitializer, [[VECTOR_PH]] ], [ [[TMP14:%.*]], [[VECTOR_BODY]] ]
19131921
; CHECK-MAXBW-NEXT: [[TMP7:%.*]] = getelementptr inbounds nuw i8, ptr [[A]], i64 [[INDEX]]
19141922
; CHECK-MAXBW-NEXT: [[TMP8:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP7]], i32 0
1915-
; CHECK-MAXBW-NEXT: [[WIDE_LOAD:%.*]] = load <16 x i8>, ptr [[TMP8]], align 1
1916-
; CHECK-MAXBW-NEXT: [[TMP2:%.*]] = zext <16 x i8> [[WIDE_LOAD]] to <16 x i64>
1923+
; CHECK-MAXBW-NEXT: [[WIDE_LOAD:%.*]] = load <vscale x 8 x i8>, ptr [[TMP8]], align 1
1924+
; CHECK-MAXBW-NEXT: [[TMP9:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD]] to <vscale x 8 x i64>
19171925
; CHECK-MAXBW-NEXT: [[TMP10:%.*]] = add nuw nsw i64 [[INDEX]], 1
19181926
; CHECK-MAXBW-NEXT: [[TMP11:%.*]] = getelementptr inbounds nuw i8, ptr [[B]], i64 [[TMP10]]
19191927
; CHECK-MAXBW-NEXT: [[TMP12:%.*]] = getelementptr inbounds nuw i8, ptr [[TMP11]], i32 0
1920-
; CHECK-MAXBW-NEXT: [[WIDE_LOAD1:%.*]] = load <16 x i8>, ptr [[TMP12]], align 1
1921-
; CHECK-MAXBW-NEXT: [[TMP6:%.*]] = zext <16 x i8> [[WIDE_LOAD1]] to <16 x i64>
1922-
; CHECK-MAXBW-NEXT: [[TMP14:%.*]] = mul nuw nsw <16 x i64> [[TMP6]], [[TMP2]]
1923-
; CHECK-MAXBW-NEXT: [[TMP13]] = add <16 x i64> [[VEC_PHI]], [[TMP14]]
1924-
; CHECK-MAXBW-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], 16
1925-
; CHECK-MAXBW-NEXT: br i1 true, label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
1928+
; CHECK-MAXBW-NEXT: [[WIDE_LOAD1:%.*]] = load <vscale x 8 x i8>, ptr [[TMP12]], align 1
1929+
; CHECK-MAXBW-NEXT: [[TMP17:%.*]] = zext <vscale x 8 x i8> [[WIDE_LOAD1]] to <vscale x 8 x i64>
1930+
; CHECK-MAXBW-NEXT: [[TMP13:%.*]] = mul nuw nsw <vscale x 8 x i64> [[TMP17]], [[TMP9]]
1931+
; CHECK-MAXBW-NEXT: [[TMP14]] = add <vscale x 8 x i64> [[VEC_PHI]], [[TMP13]]
1932+
; CHECK-MAXBW-NEXT: [[INDEX_NEXT]] = add nuw i64 [[INDEX]], [[TMP5]]
1933+
; CHECK-MAXBW-NEXT: [[TMP15:%.*]] = icmp eq i64 [[INDEX_NEXT]], [[N_VEC]]
1934+
; CHECK-MAXBW-NEXT: br i1 [[TMP15]], label [[MIDDLE_BLOCK:%.*]], label [[VECTOR_BODY]], !llvm.loop [[LOOP21:![0-9]+]]
19261935
; CHECK-MAXBW: middle.block:
1927-
; CHECK-MAXBW-NEXT: [[TMP9:%.*]] = call i64 @llvm.vector.reduce.add.v16i64(<16 x i64> [[TMP13]])
1928-
; CHECK-MAXBW-NEXT: br i1 true, label [[EXIT:%.*]], label [[SCALAR_PH]]
1936+
; CHECK-MAXBW-NEXT: [[TMP16:%.*]] = call i64 @llvm.vector.reduce.add.nxv8i64(<vscale x 8 x i64> [[TMP14]])
1937+
; CHECK-MAXBW-NEXT: [[CMP_N:%.*]] = icmp eq i64 41, [[N_VEC]]
1938+
; CHECK-MAXBW-NEXT: br i1 [[CMP_N]], label [[EXIT:%.*]], label [[SCALAR_PH]]
19291939
; CHECK-MAXBW: scalar.ph:
19301940
;
19311941
entry:
@@ -1943,7 +1953,7 @@ for.body: ; preds = %entry, %for.body
19431953
%conv3 = zext i8 %1 to i64
19441954
%mul = mul nuw nsw i64 %conv3, %conv
19451955
%add = add i64 %sum, %mul
1946-
%exitcond.not = icmp eq i64 %i.iv.next, 16
1956+
%exitcond.not = icmp eq i64 %i.iv.next, 41
19471957
br i1 %exitcond.not, label %exit, label %for.body
19481958

19491959
exit: ; preds = %for.body

0 commit comments

Comments
 (0)