Skip to content

Commit 5f4d3a3

Browse files
nikicLukacma
nikic
authored and
Lukacma
committed
[IR] Replace alignment argument with attribute on masked intrinsics (llvm#163802)
The `masked.load`, `masked.store`, `masked.gather` and `masked.scatter` intrinsics currently accept a separate alignment immarg. Replace this with an `align` attribute on the pointer / vector of pointers argument. This is the standard representation for alignment information on intrinsics, and is already used by all other memory intrinsics. This means the signatures now match llvm.expandload, llvm.vp.load, etc. (Things like llvm.memcpy used to have a separate alignment argument as well, but were already migrated a long time ago.) It's worth noting that the masked.gather and masked.scatter intrinsics previously accepted a zero alignment to indicate the ABI type alignment of the element type. This special case is gone now: If the align attribute is omitted, the implied alignment is 1, as usual. If ABI alignment is desired, it needs to be explicitly emitted (which the IRBuilder API already requires anyway).
1 parent e0d4b2b commit 5f4d3a3

File tree

301 files changed

+5657
-5709
lines changed

Some content is hidden

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

301 files changed

+5657
-5709
lines changed

‎clang/test/CodeGen/AArch64/sve-intrinsics/acle_sve_ld1.c‎

Lines changed: 52 additions & 52 deletions
Large diffs are not rendered by default.

‎clang/test/CodeGen/AArch64/sve-intrinsics/acle_sve_ld1sb.c‎

Lines changed: 24 additions & 24 deletions
Large diffs are not rendered by default.

‎clang/test/CodeGen/AArch64/sve-intrinsics/acle_sve_ld1sh.c‎

Lines changed: 16 additions & 16 deletions
Original file line numberDiff line numberDiff line change
@@ -25,14 +25,14 @@
2525
// CHECK-LABEL: @test_svld1sh_s32(
2626
// CHECK-NEXT: entry:
2727
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
28-
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i16> @llvm.masked.load.nxv4i16.p0(ptr [[BASE:%.*]], i32 1, <vscale x 4 x i1> [[TMP0]], <vscale x 4 x i16> zeroinitializer)
28+
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i16> @llvm.masked.load.nxv4i16.p0(ptr align 1 [[BASE:%.*]], <vscale x 4 x i1> [[TMP0]], <vscale x 4 x i16> zeroinitializer)
2929
// CHECK-NEXT: [[TMP2:%.*]] = sext <vscale x 4 x i16> [[TMP1]] to <vscale x 4 x i32>
3030
// CHECK-NEXT: ret <vscale x 4 x i32> [[TMP2]]
3131
//
3232
// CPP-CHECK-LABEL: @_Z16test_svld1sh_s32u10__SVBool_tPKs(
3333
// CPP-CHECK-NEXT: entry:
3434
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
35-
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i16> @llvm.masked.load.nxv4i16.p0(ptr [[BASE:%.*]], i32 1, <vscale x 4 x i1> [[TMP0]], <vscale x 4 x i16> zeroinitializer)
35+
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i16> @llvm.masked.load.nxv4i16.p0(ptr align 1 [[BASE:%.*]], <vscale x 4 x i1> [[TMP0]], <vscale x 4 x i16> zeroinitializer)
3636
// CPP-CHECK-NEXT: [[TMP2:%.*]] = sext <vscale x 4 x i16> [[TMP1]] to <vscale x 4 x i32>
3737
// CPP-CHECK-NEXT: ret <vscale x 4 x i32> [[TMP2]]
3838
//
@@ -44,14 +44,14 @@ svint32_t test_svld1sh_s32(svbool_t pg, const int16_t *base) MODE_ATTR
4444
// CHECK-LABEL: @test_svld1sh_s64(
4545
// CHECK-NEXT: entry:
4646
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
47-
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i16> @llvm.masked.load.nxv2i16.p0(ptr [[BASE:%.*]], i32 1, <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i16> zeroinitializer)
47+
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i16> @llvm.masked.load.nxv2i16.p0(ptr align 1 [[BASE:%.*]], <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i16> zeroinitializer)
4848
// CHECK-NEXT: [[TMP2:%.*]] = sext <vscale x 2 x i16> [[TMP1]] to <vscale x 2 x i64>
4949
// CHECK-NEXT: ret <vscale x 2 x i64> [[TMP2]]
5050
//
5151
// CPP-CHECK-LABEL: @_Z16test_svld1sh_s64u10__SVBool_tPKs(
5252
// CPP-CHECK-NEXT: entry:
5353
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
54-
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i16> @llvm.masked.load.nxv2i16.p0(ptr [[BASE:%.*]], i32 1, <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i16> zeroinitializer)
54+
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i16> @llvm.masked.load.nxv2i16.p0(ptr align 1 [[BASE:%.*]], <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i16> zeroinitializer)
5555
// CPP-CHECK-NEXT: [[TMP2:%.*]] = sext <vscale x 2 x i16> [[TMP1]] to <vscale x 2 x i64>
5656
// CPP-CHECK-NEXT: ret <vscale x 2 x i64> [[TMP2]]
5757
//
@@ -63,14 +63,14 @@ svint64_t test_svld1sh_s64(svbool_t pg, const int16_t *base) MODE_ATTR
6363
// CHECK-LABEL: @test_svld1sh_u32(
6464
// CHECK-NEXT: entry:
6565
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
66-
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i16> @llvm.masked.load.nxv4i16.p0(ptr [[BASE:%.*]], i32 1, <vscale x 4 x i1> [[TMP0]], <vscale x 4 x i16> zeroinitializer)
66+
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i16> @llvm.masked.load.nxv4i16.p0(ptr align 1 [[BASE:%.*]], <vscale x 4 x i1> [[TMP0]], <vscale x 4 x i16> zeroinitializer)
6767
// CHECK-NEXT: [[TMP2:%.*]] = sext <vscale x 4 x i16> [[TMP1]] to <vscale x 4 x i32>
6868
// CHECK-NEXT: ret <vscale x 4 x i32> [[TMP2]]
6969
//
7070
// CPP-CHECK-LABEL: @_Z16test_svld1sh_u32u10__SVBool_tPKs(
7171
// CPP-CHECK-NEXT: entry:
7272
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
73-
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i16> @llvm.masked.load.nxv4i16.p0(ptr [[BASE:%.*]], i32 1, <vscale x 4 x i1> [[TMP0]], <vscale x 4 x i16> zeroinitializer)
73+
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i16> @llvm.masked.load.nxv4i16.p0(ptr align 1 [[BASE:%.*]], <vscale x 4 x i1> [[TMP0]], <vscale x 4 x i16> zeroinitializer)
7474
// CPP-CHECK-NEXT: [[TMP2:%.*]] = sext <vscale x 4 x i16> [[TMP1]] to <vscale x 4 x i32>
7575
// CPP-CHECK-NEXT: ret <vscale x 4 x i32> [[TMP2]]
7676
//
@@ -82,14 +82,14 @@ svuint32_t test_svld1sh_u32(svbool_t pg, const int16_t *base) MODE_ATTR
8282
// CHECK-LABEL: @test_svld1sh_u64(
8383
// CHECK-NEXT: entry:
8484
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
85-
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i16> @llvm.masked.load.nxv2i16.p0(ptr [[BASE:%.*]], i32 1, <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i16> zeroinitializer)
85+
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i16> @llvm.masked.load.nxv2i16.p0(ptr align 1 [[BASE:%.*]], <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i16> zeroinitializer)
8686
// CHECK-NEXT: [[TMP2:%.*]] = sext <vscale x 2 x i16> [[TMP1]] to <vscale x 2 x i64>
8787
// CHECK-NEXT: ret <vscale x 2 x i64> [[TMP2]]
8888
//
8989
// CPP-CHECK-LABEL: @_Z16test_svld1sh_u64u10__SVBool_tPKs(
9090
// CPP-CHECK-NEXT: entry:
9191
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
92-
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i16> @llvm.masked.load.nxv2i16.p0(ptr [[BASE:%.*]], i32 1, <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i16> zeroinitializer)
92+
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i16> @llvm.masked.load.nxv2i16.p0(ptr align 1 [[BASE:%.*]], <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i16> zeroinitializer)
9393
// CPP-CHECK-NEXT: [[TMP2:%.*]] = sext <vscale x 2 x i16> [[TMP1]] to <vscale x 2 x i64>
9494
// CPP-CHECK-NEXT: ret <vscale x 2 x i64> [[TMP2]]
9595
//
@@ -105,7 +105,7 @@ svuint64_t test_svld1sh_u64(svbool_t pg, const int16_t *base) MODE_ATTR
105105
// CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 3
106106
// CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
107107
// CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
108-
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i16> @llvm.masked.load.nxv4i16.p0(ptr [[TMP3]], i32 1, <vscale x 4 x i1> [[TMP0]], <vscale x 4 x i16> zeroinitializer)
108+
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i16> @llvm.masked.load.nxv4i16.p0(ptr align 1 [[TMP3]], <vscale x 4 x i1> [[TMP0]], <vscale x 4 x i16> zeroinitializer)
109109
// CHECK-NEXT: [[TMP5:%.*]] = sext <vscale x 4 x i16> [[TMP4]] to <vscale x 4 x i32>
110110
// CHECK-NEXT: ret <vscale x 4 x i32> [[TMP5]]
111111
//
@@ -116,7 +116,7 @@ svuint64_t test_svld1sh_u64(svbool_t pg, const int16_t *base) MODE_ATTR
116116
// CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 3
117117
// CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
118118
// CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
119-
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i16> @llvm.masked.load.nxv4i16.p0(ptr [[TMP3]], i32 1, <vscale x 4 x i1> [[TMP0]], <vscale x 4 x i16> zeroinitializer)
119+
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i16> @llvm.masked.load.nxv4i16.p0(ptr align 1 [[TMP3]], <vscale x 4 x i1> [[TMP0]], <vscale x 4 x i16> zeroinitializer)
120120
// CPP-CHECK-NEXT: [[TMP5:%.*]] = sext <vscale x 4 x i16> [[TMP4]] to <vscale x 4 x i32>
121121
// CPP-CHECK-NEXT: ret <vscale x 4 x i32> [[TMP5]]
122122
//
@@ -132,7 +132,7 @@ svint32_t test_svld1sh_vnum_s32(svbool_t pg, const int16_t *base, int64_t vnum)
132132
// CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 2
133133
// CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
134134
// CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
135-
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i16> @llvm.masked.load.nxv2i16.p0(ptr [[TMP3]], i32 1, <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i16> zeroinitializer)
135+
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i16> @llvm.masked.load.nxv2i16.p0(ptr align 1 [[TMP3]], <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i16> zeroinitializer)
136136
// CHECK-NEXT: [[TMP5:%.*]] = sext <vscale x 2 x i16> [[TMP4]] to <vscale x 2 x i64>
137137
// CHECK-NEXT: ret <vscale x 2 x i64> [[TMP5]]
138138
//
@@ -143,7 +143,7 @@ svint32_t test_svld1sh_vnum_s32(svbool_t pg, const int16_t *base, int64_t vnum)
143143
// CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 2
144144
// CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
145145
// CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
146-
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i16> @llvm.masked.load.nxv2i16.p0(ptr [[TMP3]], i32 1, <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i16> zeroinitializer)
146+
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i16> @llvm.masked.load.nxv2i16.p0(ptr align 1 [[TMP3]], <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i16> zeroinitializer)
147147
// CPP-CHECK-NEXT: [[TMP5:%.*]] = sext <vscale x 2 x i16> [[TMP4]] to <vscale x 2 x i64>
148148
// CPP-CHECK-NEXT: ret <vscale x 2 x i64> [[TMP5]]
149149
//
@@ -159,7 +159,7 @@ svint64_t test_svld1sh_vnum_s64(svbool_t pg, const int16_t *base, int64_t vnum)
159159
// CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 3
160160
// CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
161161
// CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
162-
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i16> @llvm.masked.load.nxv4i16.p0(ptr [[TMP3]], i32 1, <vscale x 4 x i1> [[TMP0]], <vscale x 4 x i16> zeroinitializer)
162+
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i16> @llvm.masked.load.nxv4i16.p0(ptr align 1 [[TMP3]], <vscale x 4 x i1> [[TMP0]], <vscale x 4 x i16> zeroinitializer)
163163
// CHECK-NEXT: [[TMP5:%.*]] = sext <vscale x 4 x i16> [[TMP4]] to <vscale x 4 x i32>
164164
// CHECK-NEXT: ret <vscale x 4 x i32> [[TMP5]]
165165
//
@@ -170,7 +170,7 @@ svint64_t test_svld1sh_vnum_s64(svbool_t pg, const int16_t *base, int64_t vnum)
170170
// CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 3
171171
// CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
172172
// CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
173-
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i16> @llvm.masked.load.nxv4i16.p0(ptr [[TMP3]], i32 1, <vscale x 4 x i1> [[TMP0]], <vscale x 4 x i16> zeroinitializer)
173+
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 4 x i16> @llvm.masked.load.nxv4i16.p0(ptr align 1 [[TMP3]], <vscale x 4 x i1> [[TMP0]], <vscale x 4 x i16> zeroinitializer)
174174
// CPP-CHECK-NEXT: [[TMP5:%.*]] = sext <vscale x 4 x i16> [[TMP4]] to <vscale x 4 x i32>
175175
// CPP-CHECK-NEXT: ret <vscale x 4 x i32> [[TMP5]]
176176
//
@@ -186,7 +186,7 @@ svuint32_t test_svld1sh_vnum_u32(svbool_t pg, const int16_t *base, int64_t vnum)
186186
// CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 2
187187
// CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
188188
// CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
189-
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i16> @llvm.masked.load.nxv2i16.p0(ptr [[TMP3]], i32 1, <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i16> zeroinitializer)
189+
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i16> @llvm.masked.load.nxv2i16.p0(ptr align 1 [[TMP3]], <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i16> zeroinitializer)
190190
// CHECK-NEXT: [[TMP5:%.*]] = sext <vscale x 2 x i16> [[TMP4]] to <vscale x 2 x i64>
191191
// CHECK-NEXT: ret <vscale x 2 x i64> [[TMP5]]
192192
//
@@ -197,7 +197,7 @@ svuint32_t test_svld1sh_vnum_u32(svbool_t pg, const int16_t *base, int64_t vnum)
197197
// CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 2
198198
// CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
199199
// CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
200-
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i16> @llvm.masked.load.nxv2i16.p0(ptr [[TMP3]], i32 1, <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i16> zeroinitializer)
200+
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i16> @llvm.masked.load.nxv2i16.p0(ptr align 1 [[TMP3]], <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i16> zeroinitializer)
201201
// CPP-CHECK-NEXT: [[TMP5:%.*]] = sext <vscale x 2 x i16> [[TMP4]] to <vscale x 2 x i64>
202202
// CPP-CHECK-NEXT: ret <vscale x 2 x i64> [[TMP5]]
203203
//

‎clang/test/CodeGen/AArch64/sve-intrinsics/acle_sve_ld1sw.c‎

Lines changed: 8 additions & 8 deletions
Original file line numberDiff line numberDiff line change
@@ -25,14 +25,14 @@
2525
// CHECK-LABEL: @test_svld1sw_s64(
2626
// CHECK-NEXT: entry:
2727
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
28-
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i32> @llvm.masked.load.nxv2i32.p0(ptr [[BASE:%.*]], i32 1, <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i32> zeroinitializer)
28+
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i32> @llvm.masked.load.nxv2i32.p0(ptr align 1 [[BASE:%.*]], <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i32> zeroinitializer)
2929
// CHECK-NEXT: [[TMP2:%.*]] = sext <vscale x 2 x i32> [[TMP1]] to <vscale x 2 x i64>
3030
// CHECK-NEXT: ret <vscale x 2 x i64> [[TMP2]]
3131
//
3232
// CPP-CHECK-LABEL: @_Z16test_svld1sw_s64u10__SVBool_tPKi(
3333
// CPP-CHECK-NEXT: entry:
3434
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
35-
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i32> @llvm.masked.load.nxv2i32.p0(ptr [[BASE:%.*]], i32 1, <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i32> zeroinitializer)
35+
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i32> @llvm.masked.load.nxv2i32.p0(ptr align 1 [[BASE:%.*]], <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i32> zeroinitializer)
3636
// CPP-CHECK-NEXT: [[TMP2:%.*]] = sext <vscale x 2 x i32> [[TMP1]] to <vscale x 2 x i64>
3737
// CPP-CHECK-NEXT: ret <vscale x 2 x i64> [[TMP2]]
3838
//
@@ -44,14 +44,14 @@ svint64_t test_svld1sw_s64(svbool_t pg, const int32_t *base) MODE_ATTR
4444
// CHECK-LABEL: @test_svld1sw_u64(
4545
// CHECK-NEXT: entry:
4646
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
47-
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i32> @llvm.masked.load.nxv2i32.p0(ptr [[BASE:%.*]], i32 1, <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i32> zeroinitializer)
47+
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i32> @llvm.masked.load.nxv2i32.p0(ptr align 1 [[BASE:%.*]], <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i32> zeroinitializer)
4848
// CHECK-NEXT: [[TMP2:%.*]] = sext <vscale x 2 x i32> [[TMP1]] to <vscale x 2 x i64>
4949
// CHECK-NEXT: ret <vscale x 2 x i64> [[TMP2]]
5050
//
5151
// CPP-CHECK-LABEL: @_Z16test_svld1sw_u64u10__SVBool_tPKi(
5252
// CPP-CHECK-NEXT: entry:
5353
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
54-
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i32> @llvm.masked.load.nxv2i32.p0(ptr [[BASE:%.*]], i32 1, <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i32> zeroinitializer)
54+
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i32> @llvm.masked.load.nxv2i32.p0(ptr align 1 [[BASE:%.*]], <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i32> zeroinitializer)
5555
// CPP-CHECK-NEXT: [[TMP2:%.*]] = sext <vscale x 2 x i32> [[TMP1]] to <vscale x 2 x i64>
5656
// CPP-CHECK-NEXT: ret <vscale x 2 x i64> [[TMP2]]
5757
//
@@ -67,7 +67,7 @@ svuint64_t test_svld1sw_u64(svbool_t pg, const int32_t *base) MODE_ATTR
6767
// CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 3
6868
// CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
6969
// CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
70-
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i32> @llvm.masked.load.nxv2i32.p0(ptr [[TMP3]], i32 1, <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i32> zeroinitializer)
70+
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i32> @llvm.masked.load.nxv2i32.p0(ptr align 1 [[TMP3]], <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i32> zeroinitializer)
7171
// CHECK-NEXT: [[TMP5:%.*]] = sext <vscale x 2 x i32> [[TMP4]] to <vscale x 2 x i64>
7272
// CHECK-NEXT: ret <vscale x 2 x i64> [[TMP5]]
7373
//
@@ -78,7 +78,7 @@ svuint64_t test_svld1sw_u64(svbool_t pg, const int32_t *base) MODE_ATTR
7878
// CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 3
7979
// CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
8080
// CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
81-
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i32> @llvm.masked.load.nxv2i32.p0(ptr [[TMP3]], i32 1, <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i32> zeroinitializer)
81+
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i32> @llvm.masked.load.nxv2i32.p0(ptr align 1 [[TMP3]], <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i32> zeroinitializer)
8282
// CPP-CHECK-NEXT: [[TMP5:%.*]] = sext <vscale x 2 x i32> [[TMP4]] to <vscale x 2 x i64>
8383
// CPP-CHECK-NEXT: ret <vscale x 2 x i64> [[TMP5]]
8484
//
@@ -94,7 +94,7 @@ svint64_t test_svld1sw_vnum_s64(svbool_t pg, const int32_t *base, int64_t vnum)
9494
// CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 3
9595
// CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
9696
// CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
97-
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i32> @llvm.masked.load.nxv2i32.p0(ptr [[TMP3]], i32 1, <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i32> zeroinitializer)
97+
// CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i32> @llvm.masked.load.nxv2i32.p0(ptr align 1 [[TMP3]], <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i32> zeroinitializer)
9898
// CHECK-NEXT: [[TMP5:%.*]] = sext <vscale x 2 x i32> [[TMP4]] to <vscale x 2 x i64>
9999
// CHECK-NEXT: ret <vscale x 2 x i64> [[TMP5]]
100100
//
@@ -105,7 +105,7 @@ svint64_t test_svld1sw_vnum_s64(svbool_t pg, const int32_t *base, int64_t vnum)
105105
// CPP-CHECK-NEXT: [[TMP2:%.*]] = shl nuw nsw i64 [[TMP1]], 3
106106
// CPP-CHECK-NEXT: [[DOTIDX:%.*]] = mul i64 [[VNUM:%.*]], [[TMP2]]
107107
// CPP-CHECK-NEXT: [[TMP3:%.*]] = getelementptr i8, ptr [[BASE:%.*]], i64 [[DOTIDX]]
108-
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i32> @llvm.masked.load.nxv2i32.p0(ptr [[TMP3]], i32 1, <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i32> zeroinitializer)
108+
// CPP-CHECK-NEXT: [[TMP4:%.*]] = tail call <vscale x 2 x i32> @llvm.masked.load.nxv2i32.p0(ptr align 1 [[TMP3]], <vscale x 2 x i1> [[TMP0]], <vscale x 2 x i32> zeroinitializer)
109109
// CPP-CHECK-NEXT: [[TMP5:%.*]] = sext <vscale x 2 x i32> [[TMP4]] to <vscale x 2 x i64>
110110
// CPP-CHECK-NEXT: ret <vscale x 2 x i64> [[TMP5]]
111111
//

0 commit comments

Comments
 (0)