sdesmalen created this revision.
sdesmalen added reviewers: SjoerdMeijer, efriedma.
Herald added a subscriber: tschuett.
Herald added a reviewer: rengolin.
This patch adds builtins for:
- sveorbt
- sveortb
- svpmul
- svpmullb, svpmullb_pair
- svpmullt, svpmullt_pair
The svpmullb and svpmullt builtins are expressed using the svpmullb_pair
and svpmullt_pair LLVM IR intrinsics, respectively.
https://reviews.llvm.org/D79480
Files:
clang/include/clang/Basic/arm_sve.td
clang/lib/CodeGen/CGBuiltin.cpp
clang/lib/CodeGen/CodeGenFunction.h
clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_eorbt.c
clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_eortb.c
clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_pmul.c
clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_pmullb.c
clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_pmullt.c
clang/utils/TableGen/SveEmitter.cpp
Index: clang/utils/TableGen/SveEmitter.cpp
===================================================================
--- clang/utils/TableGen/SveEmitter.cpp
+++ clang/utils/TableGen/SveEmitter.cpp
@@ -528,6 +528,10 @@
Bitwidth = ElementBitwidth;
NumVectors = 0;
break;
+ case 'R':
+ ElementBitwidth /= 2;
+ NumVectors = 0;
+ break;
case 'K':
Signed = true;
Float = false;
Index: clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_pmullt.c
===================================================================
--- /dev/null
+++ clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_pmullt.c
@@ -0,0 +1,101 @@
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -D__ARM_FEATURE_SVE2 -triple aarch64-none-linux-gnu -target-feature +sve2 -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -D__ARM_FEATURE_SVE2 -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve2 -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -fsyntax-only -verify -verify-ignore-unexpected=error %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -fsyntax-only -verify=overload -verify-ignore-unexpected=error %s
+
+#include <arm_sve.h>
+
+#ifdef SVE_OVERLOADED_FORMS
+// A simple used,unused... macro, long enough to represent any SVE builtin.
+#define SVE_ACLE_FUNC(A1,A2_UNUSED,A3,A4_UNUSED) A1##A3
+#else
+#define SVE_ACLE_FUNC(A1,A2,A3,A4) A1##A2##A3##A4
+#endif
+
+svuint8_t test_svpmullt_pair_u8(svuint8_t op1, svuint8_t op2)
+{
+ // CHECK-LABEL: test_svpmullt_pair_u8
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.pmullt.pair.nxv16i8(<vscale x 16 x i8> %op1, <vscale x 16 x i8> %op2)
+ // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'svpmullt_pair'}}
+ // expected-warning@+1 {{implicit declaration of function 'svpmullt_pair_u8'}}
+ return SVE_ACLE_FUNC(svpmullt_pair,_u8,,)(op1, op2);
+}
+
+svuint32_t test_svpmullt_pair_u32(svuint32_t op1, svuint32_t op2)
+{
+ // CHECK-LABEL: test_svpmullt_pair_u32
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.pmullt.pair.nxv4i32(<vscale x 4 x i32> %op1, <vscale x 4 x i32> %op2)
+ // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'svpmullt_pair'}}
+ // expected-warning@+1 {{implicit declaration of function 'svpmullt_pair_u32'}}
+ return SVE_ACLE_FUNC(svpmullt_pair,_u32,,)(op1, op2);
+}
+
+svuint8_t test_svpmullt_pair_n_u8(svuint8_t op1, uint8_t op2)
+{
+ // CHECK-LABEL: test_svpmullt_pair_n_u8
+ // CHECK: %[[DUP:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.dup.x.nxv16i8(i8 %op2)
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.pmullt.pair.nxv16i8(<vscale x 16 x i8> %op1, <vscale x 16 x i8> %[[DUP]])
+ // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'svpmullt_pair'}}
+ // expected-warning@+1 {{implicit declaration of function 'svpmullt_pair_n_u8'}}
+ return SVE_ACLE_FUNC(svpmullt_pair,_n_u8,,)(op1, op2);
+}
+
+svuint32_t test_svpmullt_pair_n_u32(svuint32_t op1, uint32_t op2)
+{
+ // CHECK-LABEL: test_svpmullt_pair_n_u32
+ // CHECK: %[[DUP:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.dup.x.nxv4i32(i32 %op2)
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.pmullt.pair.nxv4i32(<vscale x 4 x i32> %op1, <vscale x 4 x i32> %[[DUP]])
+ // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'svpmullt_pair'}}
+ // expected-warning@+1 {{implicit declaration of function 'svpmullt_pair_n_u32'}}
+ return SVE_ACLE_FUNC(svpmullt_pair,_n_u32,,)(op1, op2);
+}
+
+svuint16_t test_svpmullt_u16(svuint8_t op1, svuint8_t op2)
+{
+ // CHECK-LABEL: test_svpmullt_u16
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.pmullt.pair.nxv16i8(<vscale x 16 x i8> %op1, <vscale x 16 x i8> %op2)
+ // CHECK: %[[BITCAST:.*]] = bitcast <vscale x 16 x i8> %[[INTRINSIC]] to <vscale x 8 x i16>
+ // CHECK: ret <vscale x 8 x i16> %[[BITCAST]]
+ // overload-warning@+2 {{implicit declaration of function 'svpmullt'}}
+ // expected-warning@+1 {{implicit declaration of function 'svpmullt_u16'}}
+ return SVE_ACLE_FUNC(svpmullt,_u16,,)(op1, op2);
+}
+
+svuint64_t test_svpmullt_u64(svuint32_t op1, svuint32_t op2)
+{
+ // CHECK-LABEL: test_svpmullt_u64
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.pmullt.pair.nxv4i32(<vscale x 4 x i32> %op1, <vscale x 4 x i32> %op2)
+ // CHECK: %[[BITCAST:.*]] = bitcast <vscale x 4 x i32> %[[INTRINSIC]] to <vscale x 2 x i64>
+ // CHECK: ret <vscale x 2 x i64> %[[BITCAST]]
+ // overload-warning@+2 {{implicit declaration of function 'svpmullt'}}
+ // expected-warning@+1 {{implicit declaration of function 'svpmullt_u64'}}
+ return SVE_ACLE_FUNC(svpmullt,_u64,,)(op1, op2);
+}
+
+svuint16_t test_svpmullt_n_u16(svuint8_t op1, uint8_t op2)
+{
+ // CHECK-LABEL: test_svpmullt_n_u16
+ // CHECK: %[[DUP:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.dup.x.nxv16i8(i8 %op2)
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.pmullt.pair.nxv16i8(<vscale x 16 x i8> %op1, <vscale x 16 x i8> %[[DUP]])
+ // CHECK: %[[BITCAST:.*]] = bitcast <vscale x 16 x i8> %[[INTRINSIC]] to <vscale x 8 x i16>
+ // CHECK: ret <vscale x 8 x i16> %[[BITCAST]]
+ // overload-warning@+2 {{implicit declaration of function 'svpmullt'}}
+ // expected-warning@+1 {{implicit declaration of function 'svpmullt_n_u16'}}
+ return SVE_ACLE_FUNC(svpmullt,_n_u16,,)(op1, op2);
+}
+
+svuint64_t test_svpmullt_n_u64(svuint32_t op1, uint32_t op2)
+{
+ // CHECK-LABEL: test_svpmullt_n_u64
+ // CHECK: %[[DUP:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.dup.x.nxv4i32(i32 %op2)
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.pmullt.pair.nxv4i32(<vscale x 4 x i32> %op1, <vscale x 4 x i32> %[[DUP]])
+ // CHECK: %[[BITCAST:.*]] = bitcast <vscale x 4 x i32> %[[INTRINSIC]] to <vscale x 2 x i64>
+ // CHECK: ret <vscale x 2 x i64> %[[BITCAST]]
+ // overload-warning@+2 {{implicit declaration of function 'svpmullt'}}
+ // expected-warning@+1 {{implicit declaration of function 'svpmullt_n_u64'}}
+ return SVE_ACLE_FUNC(svpmullt,_n_u64,,)(op1, op2);
+}
Index: clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_pmullb.c
===================================================================
--- /dev/null
+++ clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_pmullb.c
@@ -0,0 +1,101 @@
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -D__ARM_FEATURE_SVE2 -triple aarch64-none-linux-gnu -target-feature +sve2 -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -D__ARM_FEATURE_SVE2 -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve2 -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -fsyntax-only -verify -verify-ignore-unexpected=error %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -fsyntax-only -verify=overload -verify-ignore-unexpected=error %s
+
+#include <arm_sve.h>
+
+#ifdef SVE_OVERLOADED_FORMS
+// A simple used,unused... macro, long enough to represent any SVE builtin.
+#define SVE_ACLE_FUNC(A1,A2_UNUSED,A3,A4_UNUSED) A1##A3
+#else
+#define SVE_ACLE_FUNC(A1,A2,A3,A4) A1##A2##A3##A4
+#endif
+
+svuint8_t test_svpmullb_pair_u8(svuint8_t op1, svuint8_t op2)
+{
+ // CHECK-LABEL: test_svpmullb_pair_u8
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.pmullb.pair.nxv16i8(<vscale x 16 x i8> %op1, <vscale x 16 x i8> %op2)
+ // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'svpmullb_pair'}}
+ // expected-warning@+1 {{implicit declaration of function 'svpmullb_pair_u8'}}
+ return SVE_ACLE_FUNC(svpmullb_pair,_u8,,)(op1, op2);
+}
+
+svuint32_t test_svpmullb_pair_u32(svuint32_t op1, svuint32_t op2)
+{
+ // CHECK-LABEL: test_svpmullb_pair_u32
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.pmullb.pair.nxv4i32(<vscale x 4 x i32> %op1, <vscale x 4 x i32> %op2)
+ // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'svpmullb_pair'}}
+ // expected-warning@+1 {{implicit declaration of function 'svpmullb_pair_u32'}}
+ return SVE_ACLE_FUNC(svpmullb_pair,_u32,,)(op1, op2);
+}
+
+svuint8_t test_svpmullb_pair_n_u8(svuint8_t op1, uint8_t op2)
+{
+ // CHECK-LABEL: test_svpmullb_pair_n_u8
+ // CHECK: %[[DUP:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.dup.x.nxv16i8(i8 %op2)
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.pmullb.pair.nxv16i8(<vscale x 16 x i8> %op1, <vscale x 16 x i8> %[[DUP]])
+ // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'svpmullb_pair'}}
+ // expected-warning@+1 {{implicit declaration of function 'svpmullb_pair_n_u8'}}
+ return SVE_ACLE_FUNC(svpmullb_pair,_n_u8,,)(op1, op2);
+}
+
+svuint32_t test_svpmullb_pair_n_u32(svuint32_t op1, uint32_t op2)
+{
+ // CHECK-LABEL: test_svpmullb_pair_n_u32
+ // CHECK: %[[DUP:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.dup.x.nxv4i32(i32 %op2)
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.pmullb.pair.nxv4i32(<vscale x 4 x i32> %op1, <vscale x 4 x i32> %[[DUP]])
+ // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'svpmullb_pair'}}
+ // expected-warning@+1 {{implicit declaration of function 'svpmullb_pair_n_u32'}}
+ return SVE_ACLE_FUNC(svpmullb_pair,_n_u32,,)(op1, op2);
+}
+
+svuint16_t test_svpmullb_u16(svuint8_t op1, svuint8_t op2)
+{
+ // CHECK-LABEL: test_svpmullb_u16
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.pmullb.pair.nxv16i8(<vscale x 16 x i8> %op1, <vscale x 16 x i8> %op2)
+ // CHECK: %[[BITCAST:.*]] = bitcast <vscale x 16 x i8> %[[INTRINSIC]] to <vscale x 8 x i16>
+ // CHECK: ret <vscale x 8 x i16> %[[BITCAST]]
+ // overload-warning@+2 {{implicit declaration of function 'svpmullb'}}
+ // expected-warning@+1 {{implicit declaration of function 'svpmullb_u16'}}
+ return SVE_ACLE_FUNC(svpmullb,_u16,,)(op1, op2);
+}
+
+svuint64_t test_svpmullb_u64(svuint32_t op1, svuint32_t op2)
+{
+ // CHECK-LABEL: test_svpmullb_u64
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.pmullb.pair.nxv4i32(<vscale x 4 x i32> %op1, <vscale x 4 x i32> %op2)
+ // CHECK: %[[BITCAST:.*]] = bitcast <vscale x 4 x i32> %[[INTRINSIC]] to <vscale x 2 x i64>
+ // CHECK: ret <vscale x 2 x i64> %[[BITCAST]]
+ // overload-warning@+2 {{implicit declaration of function 'svpmullb'}}
+ // expected-warning@+1 {{implicit declaration of function 'svpmullb_u64'}}
+ return SVE_ACLE_FUNC(svpmullb,_u64,,)(op1, op2);
+}
+
+svuint16_t test_svpmullb_n_u16(svuint8_t op1, uint8_t op2)
+{
+ // CHECK-LABEL: test_svpmullb_n_u16
+ // CHECK: %[[DUP:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.dup.x.nxv16i8(i8 %op2)
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.pmullb.pair.nxv16i8(<vscale x 16 x i8> %op1, <vscale x 16 x i8> %[[DUP]])
+ // CHECK: %[[BITCAST:.*]] = bitcast <vscale x 16 x i8> %[[INTRINSIC]] to <vscale x 8 x i16>
+ // CHECK: ret <vscale x 8 x i16> %[[BITCAST]]
+ // overload-warning@+2 {{implicit declaration of function 'svpmullb'}}
+ // expected-warning@+1 {{implicit declaration of function 'svpmullb_n_u16'}}
+ return SVE_ACLE_FUNC(svpmullb,_n_u16,,)(op1, op2);
+}
+
+svuint64_t test_svpmullb_n_u64(svuint32_t op1, uint32_t op2)
+{
+ // CHECK-LABEL: test_svpmullb_n_u64
+ // CHECK: %[[DUP:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.dup.x.nxv4i32(i32 %op2)
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.pmullb.pair.nxv4i32(<vscale x 4 x i32> %op1, <vscale x 4 x i32> %[[DUP]])
+ // CHECK: %[[BITCAST:.*]] = bitcast <vscale x 4 x i32> %[[INTRINSIC]] to <vscale x 2 x i64>
+ // CHECK: ret <vscale x 2 x i64> %[[BITCAST]]
+ // overload-warning@+2 {{implicit declaration of function 'svpmullb'}}
+ // expected-warning@+1 {{implicit declaration of function 'svpmullb_n_u64'}}
+ return SVE_ACLE_FUNC(svpmullb,_n_u64,,)(op1, op2);
+}
Index: clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_pmul.c
===================================================================
--- /dev/null
+++ clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_pmul.c
@@ -0,0 +1,34 @@
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -D__ARM_FEATURE_SVE2 -triple aarch64-none-linux-gnu -target-feature +sve2 -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -D__ARM_FEATURE_SVE2 -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve2 -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -fsyntax-only -verify -verify-ignore-unexpected=error %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -fsyntax-only -verify=overload -verify-ignore-unexpected=error %s
+
+#include <arm_sve.h>
+
+#ifdef SVE_OVERLOADED_FORMS
+// A simple used,unused... macro, long enough to represent any SVE builtin.
+#define SVE_ACLE_FUNC(A1,A2_UNUSED,A3,A4_UNUSED) A1##A3
+#else
+#define SVE_ACLE_FUNC(A1,A2,A3,A4) A1##A2##A3##A4
+#endif
+
+svuint8_t test_svpmul_u8(svuint8_t op1, svuint8_t op2)
+{
+ // CHECK-LABEL: test_svpmul_u8
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.pmul.nxv16i8(<vscale x 16 x i8> %op1, <vscale x 16 x i8> %op2)
+ // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'svpmul'}}
+ // expected-warning@+1 {{implicit declaration of function 'svpmul_u8'}}
+ return SVE_ACLE_FUNC(svpmul,_u8,,)(op1, op2);
+}
+
+svuint8_t test_svpmul_n_u8(svuint8_t op1, uint8_t op2)
+{
+ // CHECK-LABEL: test_svpmul_n_u8
+ // CHECK: %[[DUP:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.dup.x.nxv16i8(i8 %op2)
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.pmul.nxv16i8(<vscale x 16 x i8> %op1, <vscale x 16 x i8> %[[DUP]])
+ // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'svpmul'}}
+ // expected-warning@+1 {{implicit declaration of function 'svpmul_n_u8'}}
+ return SVE_ACLE_FUNC(svpmul,_n_u8,,)(op1, op2);
+}
Index: clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_eortb.c
===================================================================
--- /dev/null
+++ clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_eortb.c
@@ -0,0 +1,181 @@
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -D__ARM_FEATURE_SVE2 -triple aarch64-none-linux-gnu -target-feature +sve2 -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -D__ARM_FEATURE_SVE2 -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve2 -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -fsyntax-only -verify -verify-ignore-unexpected=error %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -fsyntax-only -verify=overload -verify-ignore-unexpected=error %s
+
+#include <arm_sve.h>
+
+#ifdef SVE_OVERLOADED_FORMS
+// A simple used,unused... macro, long enough to represent any SVE builtin.
+#define SVE_ACLE_FUNC(A1,A2_UNUSED,A3,A4_UNUSED) A1##A3
+#else
+#define SVE_ACLE_FUNC(A1,A2,A3,A4) A1##A2##A3##A4
+#endif
+
+svint8_t test_sveortb_s8(svint8_t op1, svint8_t op2, svint8_t op3)
+{
+ // CHECK-LABEL: test_sveortb_s8
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.eortb.nxv16i8(<vscale x 16 x i8> %op1, <vscale x 16 x i8> %op2, <vscale x 16 x i8> %op3)
+ // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveortb'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveortb_s8'}}
+ return SVE_ACLE_FUNC(sveortb,_s8,,)(op1, op2, op3);
+}
+
+svint16_t test_sveortb_s16(svint16_t op1, svint16_t op2, svint16_t op3)
+{
+ // CHECK-LABEL: test_sveortb_s16
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.eortb.nxv8i16(<vscale x 8 x i16> %op1, <vscale x 8 x i16> %op2, <vscale x 8 x i16> %op3)
+ // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveortb'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveortb_s16'}}
+ return SVE_ACLE_FUNC(sveortb,_s16,,)(op1, op2, op3);
+}
+
+svint32_t test_sveortb_s32(svint32_t op1, svint32_t op2, svint32_t op3)
+{
+ // CHECK-LABEL: test_sveortb_s32
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.eortb.nxv4i32(<vscale x 4 x i32> %op1, <vscale x 4 x i32> %op2, <vscale x 4 x i32> %op3)
+ // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveortb'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveortb_s32'}}
+ return SVE_ACLE_FUNC(sveortb,_s32,,)(op1, op2, op3);
+}
+
+svint64_t test_sveortb_s64(svint64_t op1, svint64_t op2, svint64_t op3)
+{
+ // CHECK-LABEL: test_sveortb_s64
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.eortb.nxv2i64(<vscale x 2 x i64> %op1, <vscale x 2 x i64> %op2, <vscale x 2 x i64> %op3)
+ // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveortb'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveortb_s64'}}
+ return SVE_ACLE_FUNC(sveortb,_s64,,)(op1, op2, op3);
+}
+
+svuint8_t test_sveortb_u8(svuint8_t op1, svuint8_t op2, svuint8_t op3)
+{
+ // CHECK-LABEL: test_sveortb_u8
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.eortb.nxv16i8(<vscale x 16 x i8> %op1, <vscale x 16 x i8> %op2, <vscale x 16 x i8> %op3)
+ // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveortb'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveortb_u8'}}
+ return SVE_ACLE_FUNC(sveortb,_u8,,)(op1, op2, op3);
+}
+
+svuint16_t test_sveortb_u16(svuint16_t op1, svuint16_t op2, svuint16_t op3)
+{
+ // CHECK-LABEL: test_sveortb_u16
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.eortb.nxv8i16(<vscale x 8 x i16> %op1, <vscale x 8 x i16> %op2, <vscale x 8 x i16> %op3)
+ // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveortb'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveortb_u16'}}
+ return SVE_ACLE_FUNC(sveortb,_u16,,)(op1, op2, op3);
+}
+
+svuint32_t test_sveortb_u32(svuint32_t op1, svuint32_t op2, svuint32_t op3)
+{
+ // CHECK-LABEL: test_sveortb_u32
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.eortb.nxv4i32(<vscale x 4 x i32> %op1, <vscale x 4 x i32> %op2, <vscale x 4 x i32> %op3)
+ // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveortb'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveortb_u32'}}
+ return SVE_ACLE_FUNC(sveortb,_u32,,)(op1, op2, op3);
+}
+
+svuint64_t test_sveortb_u64(svuint64_t op1, svuint64_t op2, svuint64_t op3)
+{
+ // CHECK-LABEL: test_sveortb_u64
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.eortb.nxv2i64(<vscale x 2 x i64> %op1, <vscale x 2 x i64> %op2, <vscale x 2 x i64> %op3)
+ // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveortb'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveortb_u64'}}
+ return SVE_ACLE_FUNC(sveortb,_u64,,)(op1, op2, op3);
+}
+
+svint8_t test_sveortb_n_s8(svint8_t op1, svint8_t op2, int8_t op3)
+{
+ // CHECK-LABEL: test_sveortb_n_s8
+ // CHECK: %[[DUP:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.dup.x.nxv16i8(i8 %op3)
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.eortb.nxv16i8(<vscale x 16 x i8> %op1, <vscale x 16 x i8> %op2, <vscale x 16 x i8> %[[DUP]])
+ // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveortb'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveortb_n_s8'}}
+ return SVE_ACLE_FUNC(sveortb,_n_s8,,)(op1, op2, op3);
+}
+
+svint16_t test_sveortb_n_s16(svint16_t op1, svint16_t op2, int16_t op3)
+{
+ // CHECK-LABEL: test_sveortb_n_s16
+ // CHECK: %[[DUP:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.dup.x.nxv8i16(i16 %op3)
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.eortb.nxv8i16(<vscale x 8 x i16> %op1, <vscale x 8 x i16> %op2, <vscale x 8 x i16> %[[DUP]])
+ // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveortb'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveortb_n_s16'}}
+ return SVE_ACLE_FUNC(sveortb,_n_s16,,)(op1, op2, op3);
+}
+
+svint32_t test_sveortb_n_s32(svint32_t op1, svint32_t op2, int32_t op3)
+{
+ // CHECK-LABEL: test_sveortb_n_s32
+ // CHECK: %[[DUP:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.dup.x.nxv4i32(i32 %op3)
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.eortb.nxv4i32(<vscale x 4 x i32> %op1, <vscale x 4 x i32> %op2, <vscale x 4 x i32> %[[DUP]])
+ // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveortb'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveortb_n_s32'}}
+ return SVE_ACLE_FUNC(sveortb,_n_s32,,)(op1, op2, op3);
+}
+
+svint64_t test_sveortb_n_s64(svint64_t op1, svint64_t op2, int64_t op3)
+{
+ // CHECK-LABEL: test_sveortb_n_s64
+ // CHECK: %[[DUP:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.dup.x.nxv2i64(i64 %op3)
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.eortb.nxv2i64(<vscale x 2 x i64> %op1, <vscale x 2 x i64> %op2, <vscale x 2 x i64> %[[DUP]])
+ // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveortb'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveortb_n_s64'}}
+ return SVE_ACLE_FUNC(sveortb,_n_s64,,)(op1, op2, op3);
+}
+
+svuint8_t test_sveortb_n_u8(svuint8_t op1, svuint8_t op2, uint8_t op3)
+{
+ // CHECK-LABEL: test_sveortb_n_u8
+ // CHECK: %[[DUP:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.dup.x.nxv16i8(i8 %op3)
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.eortb.nxv16i8(<vscale x 16 x i8> %op1, <vscale x 16 x i8> %op2, <vscale x 16 x i8> %[[DUP]])
+ // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveortb'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveortb_n_u8'}}
+ return SVE_ACLE_FUNC(sveortb,_n_u8,,)(op1, op2, op3);
+}
+
+svuint16_t test_sveortb_n_u16(svuint16_t op1, svuint16_t op2, uint16_t op3)
+{
+ // CHECK-LABEL: test_sveortb_n_u16
+ // CHECK: %[[DUP:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.dup.x.nxv8i16(i16 %op3)
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.eortb.nxv8i16(<vscale x 8 x i16> %op1, <vscale x 8 x i16> %op2, <vscale x 8 x i16> %[[DUP]])
+ // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveortb'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveortb_n_u16'}}
+ return SVE_ACLE_FUNC(sveortb,_n_u16,,)(op1, op2, op3);
+}
+
+svuint32_t test_sveortb_n_u32(svuint32_t op1, svuint32_t op2, uint32_t op3)
+{
+ // CHECK-LABEL: test_sveortb_n_u32
+ // CHECK: %[[DUP:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.dup.x.nxv4i32(i32 %op3)
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.eortb.nxv4i32(<vscale x 4 x i32> %op1, <vscale x 4 x i32> %op2, <vscale x 4 x i32> %[[DUP]])
+ // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveortb'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveortb_n_u32'}}
+ return SVE_ACLE_FUNC(sveortb,_n_u32,,)(op1, op2, op3);
+}
+
+svuint64_t test_sveortb_n_u64(svuint64_t op1, svuint64_t op2, uint64_t op3)
+{
+ // CHECK-LABEL: test_sveortb_n_u64
+ // CHECK: %[[DUP:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.dup.x.nxv2i64(i64 %op3)
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.eortb.nxv2i64(<vscale x 2 x i64> %op1, <vscale x 2 x i64> %op2, <vscale x 2 x i64> %[[DUP]])
+ // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveortb'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveortb_n_u64'}}
+ return SVE_ACLE_FUNC(sveortb,_n_u64,,)(op1, op2, op3);
+}
Index: clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_eorbt.c
===================================================================
--- /dev/null
+++ clang/test/CodeGen/aarch64-sve2-intrinsics/acle_sve2_eorbt.c
@@ -0,0 +1,181 @@
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -D__ARM_FEATURE_SVE2 -triple aarch64-none-linux-gnu -target-feature +sve2 -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -D__ARM_FEATURE_SVE2 -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve2 -fallow-half-arguments-and-returns -S -O1 -Werror -Wall -emit-llvm -o - %s | FileCheck %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -fsyntax-only -verify -verify-ignore-unexpected=error %s
+// RUN: %clang_cc1 -D__ARM_FEATURE_SVE -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve -fallow-half-arguments-and-returns -fsyntax-only -verify=overload -verify-ignore-unexpected=error %s
+
+#include <arm_sve.h>
+
+#ifdef SVE_OVERLOADED_FORMS
+// A simple used,unused... macro, long enough to represent any SVE builtin.
+#define SVE_ACLE_FUNC(A1,A2_UNUSED,A3,A4_UNUSED) A1##A3
+#else
+#define SVE_ACLE_FUNC(A1,A2,A3,A4) A1##A2##A3##A4
+#endif
+
+svint8_t test_sveorbt_s8(svint8_t op1, svint8_t op2, svint8_t op3)
+{
+ // CHECK-LABEL: test_sveorbt_s8
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.eorbt.nxv16i8(<vscale x 16 x i8> %op1, <vscale x 16 x i8> %op2, <vscale x 16 x i8> %op3)
+ // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveorbt'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveorbt_s8'}}
+ return SVE_ACLE_FUNC(sveorbt,_s8,,)(op1, op2, op3);
+}
+
+svint16_t test_sveorbt_s16(svint16_t op1, svint16_t op2, svint16_t op3)
+{
+ // CHECK-LABEL: test_sveorbt_s16
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.eorbt.nxv8i16(<vscale x 8 x i16> %op1, <vscale x 8 x i16> %op2, <vscale x 8 x i16> %op3)
+ // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveorbt'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveorbt_s16'}}
+ return SVE_ACLE_FUNC(sveorbt,_s16,,)(op1, op2, op3);
+}
+
+svint32_t test_sveorbt_s32(svint32_t op1, svint32_t op2, svint32_t op3)
+{
+ // CHECK-LABEL: test_sveorbt_s32
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.eorbt.nxv4i32(<vscale x 4 x i32> %op1, <vscale x 4 x i32> %op2, <vscale x 4 x i32> %op3)
+ // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveorbt'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveorbt_s32'}}
+ return SVE_ACLE_FUNC(sveorbt,_s32,,)(op1, op2, op3);
+}
+
+svint64_t test_sveorbt_s64(svint64_t op1, svint64_t op2, svint64_t op3)
+{
+ // CHECK-LABEL: test_sveorbt_s64
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.eorbt.nxv2i64(<vscale x 2 x i64> %op1, <vscale x 2 x i64> %op2, <vscale x 2 x i64> %op3)
+ // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveorbt'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveorbt_s64'}}
+ return SVE_ACLE_FUNC(sveorbt,_s64,,)(op1, op2, op3);
+}
+
+svuint8_t test_sveorbt_u8(svuint8_t op1, svuint8_t op2, svuint8_t op3)
+{
+ // CHECK-LABEL: test_sveorbt_u8
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.eorbt.nxv16i8(<vscale x 16 x i8> %op1, <vscale x 16 x i8> %op2, <vscale x 16 x i8> %op3)
+ // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveorbt'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveorbt_u8'}}
+ return SVE_ACLE_FUNC(sveorbt,_u8,,)(op1, op2, op3);
+}
+
+svuint16_t test_sveorbt_u16(svuint16_t op1, svuint16_t op2, svuint16_t op3)
+{
+ // CHECK-LABEL: test_sveorbt_u16
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.eorbt.nxv8i16(<vscale x 8 x i16> %op1, <vscale x 8 x i16> %op2, <vscale x 8 x i16> %op3)
+ // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveorbt'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveorbt_u16'}}
+ return SVE_ACLE_FUNC(sveorbt,_u16,,)(op1, op2, op3);
+}
+
+svuint32_t test_sveorbt_u32(svuint32_t op1, svuint32_t op2, svuint32_t op3)
+{
+ // CHECK-LABEL: test_sveorbt_u32
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.eorbt.nxv4i32(<vscale x 4 x i32> %op1, <vscale x 4 x i32> %op2, <vscale x 4 x i32> %op3)
+ // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveorbt'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveorbt_u32'}}
+ return SVE_ACLE_FUNC(sveorbt,_u32,,)(op1, op2, op3);
+}
+
+svuint64_t test_sveorbt_u64(svuint64_t op1, svuint64_t op2, svuint64_t op3)
+{
+ // CHECK-LABEL: test_sveorbt_u64
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.eorbt.nxv2i64(<vscale x 2 x i64> %op1, <vscale x 2 x i64> %op2, <vscale x 2 x i64> %op3)
+ // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveorbt'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveorbt_u64'}}
+ return SVE_ACLE_FUNC(sveorbt,_u64,,)(op1, op2, op3);
+}
+
+svint8_t test_sveorbt_n_s8(svint8_t op1, svint8_t op2, int8_t op3)
+{
+ // CHECK-LABEL: test_sveorbt_n_s8
+ // CHECK: %[[DUP:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.dup.x.nxv16i8(i8 %op3)
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.eorbt.nxv16i8(<vscale x 16 x i8> %op1, <vscale x 16 x i8> %op2, <vscale x 16 x i8> %[[DUP]])
+ // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveorbt'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveorbt_n_s8'}}
+ return SVE_ACLE_FUNC(sveorbt,_n_s8,,)(op1, op2, op3);
+}
+
+svint16_t test_sveorbt_n_s16(svint16_t op1, svint16_t op2, int16_t op3)
+{
+ // CHECK-LABEL: test_sveorbt_n_s16
+ // CHECK: %[[DUP:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.dup.x.nxv8i16(i16 %op3)
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.eorbt.nxv8i16(<vscale x 8 x i16> %op1, <vscale x 8 x i16> %op2, <vscale x 8 x i16> %[[DUP]])
+ // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveorbt'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveorbt_n_s16'}}
+ return SVE_ACLE_FUNC(sveorbt,_n_s16,,)(op1, op2, op3);
+}
+
+svint32_t test_sveorbt_n_s32(svint32_t op1, svint32_t op2, int32_t op3)
+{
+ // CHECK-LABEL: test_sveorbt_n_s32
+ // CHECK: %[[DUP:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.dup.x.nxv4i32(i32 %op3)
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.eorbt.nxv4i32(<vscale x 4 x i32> %op1, <vscale x 4 x i32> %op2, <vscale x 4 x i32> %[[DUP]])
+ // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveorbt'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveorbt_n_s32'}}
+ return SVE_ACLE_FUNC(sveorbt,_n_s32,,)(op1, op2, op3);
+}
+
+svint64_t test_sveorbt_n_s64(svint64_t op1, svint64_t op2, int64_t op3)
+{
+ // CHECK-LABEL: test_sveorbt_n_s64
+ // CHECK: %[[DUP:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.dup.x.nxv2i64(i64 %op3)
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.eorbt.nxv2i64(<vscale x 2 x i64> %op1, <vscale x 2 x i64> %op2, <vscale x 2 x i64> %[[DUP]])
+ // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveorbt'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveorbt_n_s64'}}
+ return SVE_ACLE_FUNC(sveorbt,_n_s64,,)(op1, op2, op3);
+}
+
+svuint8_t test_sveorbt_n_u8(svuint8_t op1, svuint8_t op2, uint8_t op3)
+{
+ // CHECK-LABEL: test_sveorbt_n_u8
+ // CHECK: %[[DUP:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.dup.x.nxv16i8(i8 %op3)
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 16 x i8> @llvm.aarch64.sve.eorbt.nxv16i8(<vscale x 16 x i8> %op1, <vscale x 16 x i8> %op2, <vscale x 16 x i8> %[[DUP]])
+ // CHECK: ret <vscale x 16 x i8> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveorbt'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveorbt_n_u8'}}
+ return SVE_ACLE_FUNC(sveorbt,_n_u8,,)(op1, op2, op3);
+}
+
+svuint16_t test_sveorbt_n_u16(svuint16_t op1, svuint16_t op2, uint16_t op3)
+{
+ // CHECK-LABEL: test_sveorbt_n_u16
+ // CHECK: %[[DUP:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.dup.x.nxv8i16(i16 %op3)
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 8 x i16> @llvm.aarch64.sve.eorbt.nxv8i16(<vscale x 8 x i16> %op1, <vscale x 8 x i16> %op2, <vscale x 8 x i16> %[[DUP]])
+ // CHECK: ret <vscale x 8 x i16> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveorbt'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveorbt_n_u16'}}
+ return SVE_ACLE_FUNC(sveorbt,_n_u16,,)(op1, op2, op3);
+}
+
+svuint32_t test_sveorbt_n_u32(svuint32_t op1, svuint32_t op2, uint32_t op3)
+{
+ // CHECK-LABEL: test_sveorbt_n_u32
+ // CHECK: %[[DUP:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.dup.x.nxv4i32(i32 %op3)
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 4 x i32> @llvm.aarch64.sve.eorbt.nxv4i32(<vscale x 4 x i32> %op1, <vscale x 4 x i32> %op2, <vscale x 4 x i32> %[[DUP]])
+ // CHECK: ret <vscale x 4 x i32> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveorbt'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveorbt_n_u32'}}
+ return SVE_ACLE_FUNC(sveorbt,_n_u32,,)(op1, op2, op3);
+}
+
+svuint64_t test_sveorbt_n_u64(svuint64_t op1, svuint64_t op2, uint64_t op3)
+{
+ // CHECK-LABEL: test_sveorbt_n_u64
+ // CHECK: %[[DUP:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.dup.x.nxv2i64(i64 %op3)
+ // CHECK: %[[INTRINSIC:.*]] = call <vscale x 2 x i64> @llvm.aarch64.sve.eorbt.nxv2i64(<vscale x 2 x i64> %op1, <vscale x 2 x i64> %op2, <vscale x 2 x i64> %[[DUP]])
+ // CHECK: ret <vscale x 2 x i64> %[[INTRINSIC]]
+ // overload-warning@+2 {{implicit declaration of function 'sveorbt'}}
+ // expected-warning@+1 {{implicit declaration of function 'sveorbt_n_u64'}}
+ return SVE_ACLE_FUNC(sveorbt,_n_u64,,)(op1, op2, op3);
+}
Index: clang/lib/CodeGen/CodeGenFunction.h
===================================================================
--- clang/lib/CodeGen/CodeGenFunction.h
+++ clang/lib/CodeGen/CodeGenFunction.h
@@ -3920,6 +3920,9 @@
llvm::ScalableVectorType *getSVEPredType(SVETypeFlags TypeFlags);
llvm::Value *EmitSVEAllTruePred(SVETypeFlags TypeFlags);
llvm::Value *EmitSVEDupX(llvm::Value *Scalar);
+ llvm::Value *EmitSVEPMull(SVETypeFlags TypeFlags,
+ llvm::SmallVectorImpl<llvm::Value *> &Ops,
+ unsigned BuiltinID);
llvm::Value *EmitSVEPredicateCast(llvm::Value *Pred,
llvm::ScalableVectorType *VTy);
llvm::Value *EmitSVEGatherLoad(SVETypeFlags TypeFlags,
Index: clang/lib/CodeGen/CGBuiltin.cpp
===================================================================
--- clang/lib/CodeGen/CGBuiltin.cpp
+++ clang/lib/CodeGen/CGBuiltin.cpp
@@ -7792,6 +7792,27 @@
return Builder.CreateCall(F, Ops);
}
+// SVE2's svpmullb and svpmullt builtins are similar to the svpmullb_pair and
+// svpmullt_pair intrinsics, with the exception that their results are bitcast
+// to a wider type.
+Value *CodeGenFunction::EmitSVEPMull(SVETypeFlags TypeFlags,
+ SmallVectorImpl<Value *> &Ops,
+ unsigned BuiltinID) {
+ // Splat scalar operand to vector (intrinsics with _n infix)
+ if (TypeFlags.hasSplatOperand()) {
+ unsigned OpNo = TypeFlags.getSplatOperand();
+ Ops[OpNo] = EmitSVEDupX(Ops[OpNo]);
+ }
+
+ // The pair-wise function has a narrower overloaded type.
+ Function *F = CGM.getIntrinsic(BuiltinID, Ops[0]->getType());
+ Value *Call = Builder.CreateCall(F, {Ops[0], Ops[1]});
+
+ // Now bitcast to the wider result type.
+ llvm::ScalableVectorType *Ty = getSVEType(TypeFlags);
+ return Builder.CreateBitCast(Call, Ty);
+}
+
Value *CodeGenFunction::EmitSVEPrefetchLoad(SVETypeFlags TypeFlags,
SmallVectorImpl<Value *> &Ops,
unsigned BuiltinID) {
@@ -8034,6 +8055,18 @@
return Builder.CreateCall(F, {Ops[0], Ops[1], Ops[0]});
}
+ case SVE::BI__builtin_sve_svpmullt_u16:
+ case SVE::BI__builtin_sve_svpmullt_u64:
+ case SVE::BI__builtin_sve_svpmullt_n_u16:
+ case SVE::BI__builtin_sve_svpmullt_n_u64:
+ return EmitSVEPMull(TypeFlags, Ops, Intrinsic::aarch64_sve_pmullt_pair);
+
+ case SVE::BI__builtin_sve_svpmullb_u16:
+ case SVE::BI__builtin_sve_svpmullb_u64:
+ case SVE::BI__builtin_sve_svpmullb_n_u16:
+ case SVE::BI__builtin_sve_svpmullb_n_u64:
+ return EmitSVEPMull(TypeFlags, Ops, Intrinsic::aarch64_sve_pmullb_pair);
+
case SVE::BI__builtin_sve_svdupq_n_b8:
case SVE::BI__builtin_sve_svdupq_n_b16:
case SVE::BI__builtin_sve_svdupq_n_b32:
Index: clang/include/clang/Basic/arm_sve.td
===================================================================
--- clang/include/clang/Basic/arm_sve.td
+++ clang/include/clang/Basic/arm_sve.td
@@ -66,6 +66,7 @@
// P: predicate type
// s: scalar of element type
// a: scalar of element type (splat to vector type)
+// R: scalar of 1/2 width element type (splat to vector type)
// e: 1/2 width unsigned elements, 2x element count
// h: 1/2 width elements, 2x element count
// q: 1/4 width elements, 4x element count
@@ -1319,6 +1320,26 @@
def SVSTNT1W_SCATTER_INDEX_S : MInst<"svstnt1w_scatter[_{2}base]_index[_{d}]", "vPuld", "lUl", [IsScatterStore], MemEltTyInt32, "aarch64_sve_stnt1_scatter_scalar_offset">;
}
+////////////////////////////////////////////////////////////////////////////////
+// SVE2 - Polynomial arithmetic
+
+let ArchGuard = "defined(__ARM_FEATURE_SVE2)" in {
+def SVEORBT : SInst<"sveorbt[_{d}]", "dddd", "csilUcUsUiUl", MergeNone, "aarch64_sve_eorbt">;
+def SVEORBT_N : SInst<"sveorbt[_n_{d}]", "ddda", "csilUcUsUiUl", MergeNone, "aarch64_sve_eorbt">;
+def SVEORTB : SInst<"sveortb[_{d}]", "dddd", "csilUcUsUiUl", MergeNone, "aarch64_sve_eortb">;
+def SVEORTB_N : SInst<"sveortb[_n_{d}]", "ddda", "csilUcUsUiUl", MergeNone, "aarch64_sve_eortb">;
+def SVPMUL : SInst<"svpmul[_{d}]", "ddd", "Uc", MergeNone, "aarch64_sve_pmul">;
+def SVPMUL_N : SInst<"svpmul[_n_{d}]", "dda", "Uc", MergeNone, "aarch64_sve_pmul">;
+def SVPMULLB : SInst<"svpmullb[_{d}]", "dhh", "UsUl", MergeNone>;
+def SVPMULLB_N : SInst<"svpmullb[_n_{d}]", "dhR", "UsUl", MergeNone>;
+def SVPMULLB_PAIR : SInst<"svpmullb_pair[_{d}]", "ddd", "UcUi", MergeNone, "aarch64_sve_pmullb_pair">;
+def SVPMULLB_PAIR_N : SInst<"svpmullb_pair[_n_{d}]", "dda", "UcUi", MergeNone, "aarch64_sve_pmullb_pair">;
+def SVPMULLT : SInst<"svpmullt[_{d}]", "dhh", "UsUl", MergeNone>;
+def SVPMULLT_N : SInst<"svpmullt[_n_{d}]", "dhR", "UsUl", MergeNone>;
+def SVPMULLT_PAIR : SInst<"svpmullt_pair[_{d}]", "ddd", "UcUi", MergeNone, "aarch64_sve_pmullt_pair">;
+def SVPMULLT_PAIR_N : SInst<"svpmullt_pair[_n_{d}]", "dda", "UcUi", MergeNone, "aarch64_sve_pmullt_pair">;
+}
+
////////////////////////////////////////////////////////////////////////////////
// SVE2 - Contiguous conflict detection
let ArchGuard = "defined(__ARM_FEATURE_SVE2)" in {
_______________________________________________
cfe-commits mailing list
cfe-commits@lists.llvm.org
https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
