Skip to content

Commit 434575c

Browse files
paulwalker-armtstellar
paulwalker-arm
authored and
tstellar
committed
[SVE] Fix incorrect lowering of predicate permute builtins.
When lowering predicate permute builtins we incorrectly assume only the typically "active" bits for the specified element type play a role with all other bits zero'd. This is not the case because all bits are significant, with the element type specifying how they are grouped: b8 - permute using a block size of 1 bit b16 - permute using a block size of 2 bits b32 - permute using a block size of 4 bits b64 - permute using a block size of 8 bits The affected builtins are svrev, svtrn1, svtrn2, svuzp1, svuzp2, svzip1 and svzip2. This patch adds new intrinsics to support these operations and changes the builtin lowering code to emit them. The b8 case remains unchanged because for that operation the existing intrinsics work as required and their support for other predicate types has been maintained as useful if only as a way to test the correctness of their matching ISD nodes that code generation relies on. Differential Revision: https://reviews.llvm.org/D142065 (cherry picked from commit 26b79ca)
1 parent aa84ee5 commit 434575c

File tree

12 files changed

+474
-301
lines changed

12 files changed

+474
-301
lines changed

‎clang/include/clang/Basic/arm_sve.td

Lines changed: 31 additions & 10 deletions
Original file line numberDiff line numberDiff line change
@@ -1249,16 +1249,37 @@ def SVZIP1_BF16 : SInst<"svzip1[_{d}]", "ddd", "b", MergeNone, "aarch64_sve
12491249
def SVZIP2_BF16 : SInst<"svzip2[_{d}]", "ddd", "b", MergeNone, "aarch64_sve_zip2">;
12501250
}
12511251

1252-
def SVREV_B : SInst<"svrev_{d}", "PP", "PcPsPiPl", MergeNone, "aarch64_sve_rev">;
1253-
def SVSEL_B : SInst<"svsel[_b]", "PPPP", "Pc", MergeNone, "aarch64_sve_sel">;
1254-
def SVTRN1_B : SInst<"svtrn1_{d}", "PPP", "PcPsPiPl", MergeNone, "aarch64_sve_trn1">;
1255-
def SVTRN2_B : SInst<"svtrn2_{d}", "PPP", "PcPsPiPl", MergeNone, "aarch64_sve_trn2">;
1256-
def SVPUNPKHI : SInst<"svunpkhi[_b]", "PP", "Pc", MergeNone, "aarch64_sve_punpkhi">;
1257-
def SVPUNPKLO : SInst<"svunpklo[_b]", "PP", "Pc", MergeNone, "aarch64_sve_punpklo">;
1258-
def SVUZP1_B : SInst<"svuzp1_{d}", "PPP", "PcPsPiPl", MergeNone, "aarch64_sve_uzp1">;
1259-
def SVUZP2_B : SInst<"svuzp2_{d}", "PPP", "PcPsPiPl", MergeNone, "aarch64_sve_uzp2">;
1260-
def SVZIP1_B : SInst<"svzip1_{d}", "PPP", "PcPsPiPl", MergeNone, "aarch64_sve_zip1">;
1261-
def SVZIP2_B : SInst<"svzip2_{d}", "PPP", "PcPsPiPl", MergeNone, "aarch64_sve_zip2">;
1252+
def SVREV_B8 : SInst<"svrev_b8", "PP", "Pc", MergeNone, "aarch64_sve_rev">;
1253+
def SVREV_B16 : SInst<"svrev_b16", "PP", "Pc", MergeNone, "aarch64_sve_rev_b16", [IsOverloadNone]>;
1254+
def SVREV_B32 : SInst<"svrev_b32", "PP", "Pc", MergeNone, "aarch64_sve_rev_b32", [IsOverloadNone]>;
1255+
def SVREV_B64 : SInst<"svrev_b64", "PP", "Pc", MergeNone, "aarch64_sve_rev_b64", [IsOverloadNone]>;
1256+
def SVSEL_B : SInst<"svsel[_b]", "PPPP", "Pc", MergeNone, "aarch64_sve_sel">;
1257+
def SVTRN1_B8 : SInst<"svtrn1_b8", "PPP", "Pc", MergeNone, "aarch64_sve_trn1">;
1258+
def SVTRN1_B16 : SInst<"svtrn1_b16", "PPP", "Pc", MergeNone, "aarch64_sve_trn1_b16", [IsOverloadNone]>;
1259+
def SVTRN1_B32 : SInst<"svtrn1_b32", "PPP", "Pc", MergeNone, "aarch64_sve_trn1_b32", [IsOverloadNone]>;
1260+
def SVTRN1_B64 : SInst<"svtrn1_b64", "PPP", "Pc", MergeNone, "aarch64_sve_trn1_b64", [IsOverloadNone]>;
1261+
def SVTRN2_B8 : SInst<"svtrn2_b8", "PPP", "Pc", MergeNone, "aarch64_sve_trn2">;
1262+
def SVTRN2_B16 : SInst<"svtrn2_b16", "PPP", "Pc", MergeNone, "aarch64_sve_trn2_b16", [IsOverloadNone]>;
1263+
def SVTRN2_B32 : SInst<"svtrn2_b32", "PPP", "Pc", MergeNone, "aarch64_sve_trn2_b32", [IsOverloadNone]>;
1264+
def SVTRN2_B64 : SInst<"svtrn2_b64", "PPP", "Pc", MergeNone, "aarch64_sve_trn2_b64", [IsOverloadNone]>;
1265+
def SVPUNPKHI : SInst<"svunpkhi[_b]", "PP", "Pc", MergeNone, "aarch64_sve_punpkhi">;
1266+
def SVPUNPKLO : SInst<"svunpklo[_b]", "PP", "Pc", MergeNone, "aarch64_sve_punpklo">;
1267+
def SVUZP1_B8 : SInst<"svuzp1_b8", "PPP", "Pc", MergeNone, "aarch64_sve_uzp1">;
1268+
def SVUZP1_B16 : SInst<"svuzp1_b16", "PPP", "Pc", MergeNone, "aarch64_sve_uzp1_b16", [IsOverloadNone]>;
1269+
def SVUZP1_B32 : SInst<"svuzp1_b32", "PPP", "Pc", MergeNone, "aarch64_sve_uzp1_b32", [IsOverloadNone]>;
1270+
def SVUZP1_B64 : SInst<"svuzp1_b64", "PPP", "Pc", MergeNone, "aarch64_sve_uzp1_b64", [IsOverloadNone]>;
1271+
def SVUZP2_B8 : SInst<"svuzp2_b8", "PPP", "Pc", MergeNone, "aarch64_sve_uzp2">;
1272+
def SVUZP2_B16 : SInst<"svuzp2_b16", "PPP", "Pc", MergeNone, "aarch64_sve_uzp2_b16", [IsOverloadNone]>;
1273+
def SVUZP2_B32 : SInst<"svuzp2_b32", "PPP", "Pc", MergeNone, "aarch64_sve_uzp2_b32", [IsOverloadNone]>;
1274+
def SVUZP2_B64 : SInst<"svuzp2_b64", "PPP", "Pc", MergeNone, "aarch64_sve_uzp2_b64", [IsOverloadNone]>;
1275+
def SVZIP1_B8 : SInst<"svzip1_b8", "PPP", "Pc", MergeNone, "aarch64_sve_zip1">;
1276+
def SVZIP1_B16 : SInst<"svzip1_b16", "PPP", "Pc", MergeNone, "aarch64_sve_zip1_b16", [IsOverloadNone]>;
1277+
def SVZIP1_B32 : SInst<"svzip1_b32", "PPP", "Pc", MergeNone, "aarch64_sve_zip1_b32", [IsOverloadNone]>;
1278+
def SVZIP1_B64 : SInst<"svzip1_b64", "PPP", "Pc", MergeNone, "aarch64_sve_zip1_b64", [IsOverloadNone]>;
1279+
def SVZIP2_B : SInst<"svzip2_b8", "PPP", "Pc", MergeNone, "aarch64_sve_zip2">;
1280+
def SVZIP2_B16 : SInst<"svzip2_b16", "PPP", "Pc", MergeNone, "aarch64_sve_zip2_b16", [IsOverloadNone]>;
1281+
def SVZIP2_B32 : SInst<"svzip2_b32", "PPP", "Pc", MergeNone, "aarch64_sve_zip2_b32", [IsOverloadNone]>;
1282+
def SVZIP2_B64 : SInst<"svzip2_b64", "PPP", "Pc", MergeNone, "aarch64_sve_zip2_b64", [IsOverloadNone]>;
12621283

12631284
////////////////////////////////////////////////////////////////////////////////
12641285
// Predicate creation

‎clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_rev.c

Lines changed: 12 additions & 24 deletions
Original file line numberDiff line numberDiff line change
@@ -196,17 +196,13 @@ svbool_t test_svrev_b8(svbool_t op)
196196

197197
// CHECK-LABEL: @test_svrev_b16(
198198
// CHECK-NEXT: entry:
199-
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[OP:%.*]])
200-
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.rev.nxv8i1(<vscale x 8 x i1> [[TMP0]])
201-
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv8i1(<vscale x 8 x i1> [[TMP1]])
202-
// CHECK-NEXT: ret <vscale x 16 x i1> [[TMP2]]
199+
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.rev.b16(<vscale x 16 x i1> [[OP:%.*]])
200+
// CHECK-NEXT: ret <vscale x 16 x i1> [[TMP0]]
203201
//
204202
// CPP-CHECK-LABEL: @_Z14test_svrev_b16u10__SVBool_t(
205203
// CPP-CHECK-NEXT: entry:
206-
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[OP:%.*]])
207-
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.rev.nxv8i1(<vscale x 8 x i1> [[TMP0]])
208-
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv8i1(<vscale x 8 x i1> [[TMP1]])
209-
// CPP-CHECK-NEXT: ret <vscale x 16 x i1> [[TMP2]]
204+
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.rev.b16(<vscale x 16 x i1> [[OP:%.*]])
205+
// CPP-CHECK-NEXT: ret <vscale x 16 x i1> [[TMP0]]
210206
//
211207
svbool_t test_svrev_b16(svbool_t op)
212208
{
@@ -215,17 +211,13 @@ svbool_t test_svrev_b16(svbool_t op)
215211

216212
// CHECK-LABEL: @test_svrev_b32(
217213
// CHECK-NEXT: entry:
218-
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[OP:%.*]])
219-
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.rev.nxv4i1(<vscale x 4 x i1> [[TMP0]])
220-
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv4i1(<vscale x 4 x i1> [[TMP1]])
221-
// CHECK-NEXT: ret <vscale x 16 x i1> [[TMP2]]
214+
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.rev.b32(<vscale x 16 x i1> [[OP:%.*]])
215+
// CHECK-NEXT: ret <vscale x 16 x i1> [[TMP0]]
222216
//
223217
// CPP-CHECK-LABEL: @_Z14test_svrev_b32u10__SVBool_t(
224218
// CPP-CHECK-NEXT: entry:
225-
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[OP:%.*]])
226-
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.rev.nxv4i1(<vscale x 4 x i1> [[TMP0]])
227-
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv4i1(<vscale x 4 x i1> [[TMP1]])
228-
// CPP-CHECK-NEXT: ret <vscale x 16 x i1> [[TMP2]]
219+
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.rev.b32(<vscale x 16 x i1> [[OP:%.*]])
220+
// CPP-CHECK-NEXT: ret <vscale x 16 x i1> [[TMP0]]
229221
//
230222
svbool_t test_svrev_b32(svbool_t op)
231223
{
@@ -234,17 +226,13 @@ svbool_t test_svrev_b32(svbool_t op)
234226

235227
// CHECK-LABEL: @test_svrev_b64(
236228
// CHECK-NEXT: entry:
237-
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[OP:%.*]])
238-
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.rev.nxv2i1(<vscale x 2 x i1> [[TMP0]])
239-
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv2i1(<vscale x 2 x i1> [[TMP1]])
240-
// CHECK-NEXT: ret <vscale x 16 x i1> [[TMP2]]
229+
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.rev.b64(<vscale x 16 x i1> [[OP:%.*]])
230+
// CHECK-NEXT: ret <vscale x 16 x i1> [[TMP0]]
241231
//
242232
// CPP-CHECK-LABEL: @_Z14test_svrev_b64u10__SVBool_t(
243233
// CPP-CHECK-NEXT: entry:
244-
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[OP:%.*]])
245-
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.rev.nxv2i1(<vscale x 2 x i1> [[TMP0]])
246-
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv2i1(<vscale x 2 x i1> [[TMP1]])
247-
// CPP-CHECK-NEXT: ret <vscale x 16 x i1> [[TMP2]]
234+
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.rev.b64(<vscale x 16 x i1> [[OP:%.*]])
235+
// CPP-CHECK-NEXT: ret <vscale x 16 x i1> [[TMP0]]
248236
//
249237
svbool_t test_svrev_b64(svbool_t op)
250238
{

‎clang/test/CodeGen/aarch64-sve-intrinsics/acle_sve_trn1.c

Lines changed: 12 additions & 30 deletions
Original file line numberDiff line numberDiff line change
@@ -196,19 +196,13 @@ svbool_t test_svtrn1_b8(svbool_t op1, svbool_t op2)
196196

197197
// CHECK-LABEL: @test_svtrn1_b16(
198198
// CHECK-NEXT: entry:
199-
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[OP1:%.*]])
200-
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[OP2:%.*]])
201-
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.trn1.nxv8i1(<vscale x 8 x i1> [[TMP0]], <vscale x 8 x i1> [[TMP1]])
202-
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv8i1(<vscale x 8 x i1> [[TMP2]])
203-
// CHECK-NEXT: ret <vscale x 16 x i1> [[TMP3]]
199+
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.trn1.b16(<vscale x 16 x i1> [[OP1:%.*]], <vscale x 16 x i1> [[OP2:%.*]])
200+
// CHECK-NEXT: ret <vscale x 16 x i1> [[TMP0]]
204201
//
205202
// CPP-CHECK-LABEL: @_Z15test_svtrn1_b16u10__SVBool_tu10__SVBool_t(
206203
// CPP-CHECK-NEXT: entry:
207-
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[OP1:%.*]])
208-
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[OP2:%.*]])
209-
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.trn1.nxv8i1(<vscale x 8 x i1> [[TMP0]], <vscale x 8 x i1> [[TMP1]])
210-
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv8i1(<vscale x 8 x i1> [[TMP2]])
211-
// CPP-CHECK-NEXT: ret <vscale x 16 x i1> [[TMP3]]
204+
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.trn1.b16(<vscale x 16 x i1> [[OP1:%.*]], <vscale x 16 x i1> [[OP2:%.*]])
205+
// CPP-CHECK-NEXT: ret <vscale x 16 x i1> [[TMP0]]
212206
//
213207
svbool_t test_svtrn1_b16(svbool_t op1, svbool_t op2)
214208
{
@@ -217,19 +211,13 @@ svbool_t test_svtrn1_b16(svbool_t op1, svbool_t op2)
217211

218212
// CHECK-LABEL: @test_svtrn1_b32(
219213
// CHECK-NEXT: entry:
220-
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[OP1:%.*]])
221-
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[OP2:%.*]])
222-
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.trn1.nxv4i1(<vscale x 4 x i1> [[TMP0]], <vscale x 4 x i1> [[TMP1]])
223-
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv4i1(<vscale x 4 x i1> [[TMP2]])
224-
// CHECK-NEXT: ret <vscale x 16 x i1> [[TMP3]]
214+
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.trn1.b32(<vscale x 16 x i1> [[OP1:%.*]], <vscale x 16 x i1> [[OP2:%.*]])
215+
// CHECK-NEXT: ret <vscale x 16 x i1> [[TMP0]]
225216
//
226217
// CPP-CHECK-LABEL: @_Z15test_svtrn1_b32u10__SVBool_tu10__SVBool_t(
227218
// CPP-CHECK-NEXT: entry:
228-
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[OP1:%.*]])
229-
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[OP2:%.*]])
230-
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.trn1.nxv4i1(<vscale x 4 x i1> [[TMP0]], <vscale x 4 x i1> [[TMP1]])
231-
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv4i1(<vscale x 4 x i1> [[TMP2]])
232-
// CPP-CHECK-NEXT: ret <vscale x 16 x i1> [[TMP3]]
219+
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.trn1.b32(<vscale x 16 x i1> [[OP1:%.*]], <vscale x 16 x i1> [[OP2:%.*]])
220+
// CPP-CHECK-NEXT: ret <vscale x 16 x i1> [[TMP0]]
233221
//
234222
svbool_t test_svtrn1_b32(svbool_t op1, svbool_t op2)
235223
{
@@ -238,19 +226,13 @@ svbool_t test_svtrn1_b32(svbool_t op1, svbool_t op2)
238226

239227
// CHECK-LABEL: @test_svtrn1_b64(
240228
// CHECK-NEXT: entry:
241-
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[OP1:%.*]])
242-
// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[OP2:%.*]])
243-
// CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.trn1.nxv2i1(<vscale x 2 x i1> [[TMP0]], <vscale x 2 x i1> [[TMP1]])
244-
// CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv2i1(<vscale x 2 x i1> [[TMP2]])
245-
// CHECK-NEXT: ret <vscale x 16 x i1> [[TMP3]]
229+
// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.trn1.b64(<vscale x 16 x i1> [[OP1:%.*]], <vscale x 16 x i1> [[OP2:%.*]])
230+
// CHECK-NEXT: ret <vscale x 16 x i1> [[TMP0]]
246231
//
247232
// CPP-CHECK-LABEL: @_Z15test_svtrn1_b64u10__SVBool_tu10__SVBool_t(
248233
// CPP-CHECK-NEXT: entry:
249-
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[OP1:%.*]])
250-
// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[OP2:%.*]])
251-
// CPP-CHECK-NEXT: [[TMP2:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.trn1.nxv2i1(<vscale x 2 x i1> [[TMP0]], <vscale x 2 x i1> [[TMP1]])
252-
// CPP-CHECK-NEXT: [[TMP3:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.convert.to.svbool.nxv2i1(<vscale x 2 x i1> [[TMP2]])
253-
// CPP-CHECK-NEXT: ret <vscale x 16 x i1> [[TMP3]]
234+
// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i1> @llvm.aarch64.sve.trn1.b64(<vscale x 16 x i1> [[OP1:%.*]], <vscale x 16 x i1> [[OP2:%.*]])
235+
// CPP-CHECK-NEXT: ret <vscale x 16 x i1> [[TMP0]]
254236
//
255237
svbool_t test_svtrn1_b64(svbool_t op1, svbool_t op2)
256238
{

0 commit comments

Comments
 (0)