[Clang][SVE2.1] Add builtins for 2-way svdot (vectors, indexed)

[CarolineConcatto]

Summary

As described in: https://github.com/ARM-software/acle/pull/257

Patch by: David Sherwood <[e-mail]>

[CarolineConcatto]

Rebase

[dtemirbulatov]

clang/test/Sema/aarch64-sve2p1-intrinsics/acle_sve2p1_imm.cpp

112

It looks like an incorrect error report?

[dtemirbulatov]

clang/test/Sema/aarch64-sve2p1-intrinsics/acle_sve2p1_imm.cpp

112

Oh, I see this is only syntax test. Resolved.

[dtemirbulatov]

LGTM.

[CarolineConcatto]

-Rebase

clang/include/clang/Basic/arm_sve.td

 def SVSTNT1B_VNUM_X2 : MInst<"svstnt1_vnum[_{2}_x2]", "v}pl2", "cUc", [IsStructStore], MemEltTyDefault, "aarch64_sve_stnt1_pn_x2">;
 def SVSTNT1H_VNUM_X2 : MInst<"svstnt1_vnum[_{2}_x2]", "v}pl2", "sUshb", [IsStructStore], MemEltTyDefault, "aarch64_sve_stnt1_pn_x2">;
 def SVSTNT1W_VNUM_X2 : MInst<"svstnt1_vnum[_{2}_x2]", "v}pl2", "iUif", [IsStructStore], MemEltTyDefault, "aarch64_sve_stnt1_pn_x2">;
 def SVSTNT1D_VNUM_X2 : MInst<"svstnt1_vnum[_{2}_x2]", "v}pl2", "lUld", [IsStructStore], MemEltTyDefault, "aarch64_sve_stnt1_pn_x2">;
 def SVSTNT1B_VNUM_X4 : MInst<"svstnt1_vnum[_{2}_x4]", "v}pl4", "cUc", [IsStructStore], MemEltTyDefault, "aarch64_sve_stnt1_pn_x4">;
 def SVSTNT1H_VNUM_X4 : MInst<"svstnt1_vnum[_{2}_x4]", "v}pl4", "sUshb", [IsStructStore], MemEltTyDefault, "aarch64_sve_stnt1_pn_x4">;
 def SVSTNT1W_VNUM_X4 : MInst<"svstnt1_vnum[_{2}_x4]", "v}pl4", "iUif", [IsStructStore], MemEltTyDefault, "aarch64_sve_stnt1_pn_x4">;
 def SVSTNT1D_VNUM_X4 : MInst<"svstnt1_vnum[_{2}_x4]", "v}pl4", "lUld", [IsStructStore], MemEltTyDefault, "aarch64_sve_stnt1_pn_x4">;
+
+def SVDOT_X2_S : SInst<"svdot[_{d}_{2}_{3}]", "ddhh", "i",  MergeNone, "aarch64_sve_sdot_x2", [], []>;
+def SVDOT_X2_U : SInst<"svdot[_{d}_{2}_{3}]", "ddhh", "Ui", MergeNone, "aarch64_sve_udot_x2", [], []>;
+def SVDOT_X2_F : SInst<"svdot[_{d}_{2}_{3}]", "ddhh", "f",  MergeNone, "aarch64_sve_fdot_x2", [], []>;
+def SVDOT_LANE_X2_S : SInst<"svdot_lane[_{d}_{2}_{3}]", "ddhhi", "i",  MergeNone, "aarch64_sve_sdot_lane_x2", [], [ImmCheck<3, ImmCheck0_3>]>;
+def SVDOT_LANE_X2_U : SInst<"svdot_lane[_{d}_{2}_{3}]", "ddhhi", "Ui", MergeNone, "aarch64_sve_udot_lane_x2", [], [ImmCheck<3, ImmCheck0_3>]>;
+def SVDOT_LANE_X2_F : SInst<"svdot_lane[_{d}_{2}_{3}]", "ddhhi", "f",  MergeNone, "aarch64_sve_fdot_lane_x2", [], [ImmCheck<3, ImmCheck0_3>]>;
 }
 
 let TargetGuard = "sve2p1" in {
 def SVSCLAMP : SInst<"svclamp[_{d}]", "dddd", "csil",     MergeNone, "aarch64_sve_sclamp", [], []>;
 def SVUCLAMP : SInst<"svclamp[_{d}]", "dddd", "UcUsUiUl", MergeNone, "aarch64_sve_uclamp", [], []>;
 
 def SVPSEL_B : SInst<"svpsel_lane_b8",  "PPPm", "Pc", MergeNone, "", [], []>;
 def SVPSEL_H : SInst<"svpsel_lane_b16", "PPPm", "Ps", MergeNone, "", [], []>;
 def SVPSEL_S : SInst<"svpsel_lane_b32", "PPPm", "Pi", MergeNone, "", [], []>;
 def SVPSEL_D : SInst<"svpsel_lane_b64", "PPPm", "Pl", MergeNone, "", [], []>;
 
 def SVCNTP_COUNT : SInst<"svcntp_{d}", "n}i", "QcQsQiQl", MergeNone, "aarch64_sve_cntp_{d}", [IsOverloadNone], [ImmCheck<1, ImmCheck2_4_Mul2>]>;
 }

clang/test/CodeGen/aarch64-sve2p1-intrinsics/acle_sve2p1_dot.c

+// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
+// REQUIRES: aarch64-registered-target
+// RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sve2p1 -S -disable-O0-optnone -Werror -Wall -emit-llvm -o - %s | opt -S -p mem2reg,instcombine,tailcallelim | FileCheck %s
+// RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sve2p1 -S -disable-O0-optnone -Werror -Wall -emit-llvm -o - -x c++ %s | opt -S -p mem2reg,instcombine,tailcallelim | FileCheck %s -check-prefix=CPP-CHECK
+// RUN: %clang_cc1 -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve2p1 -S -disable-O0-optnone -Werror -Wall -emit-llvm -o - %s | opt -S -p mem2reg,instcombine,tailcallelim | FileCheck %s
+// RUN: %clang_cc1 -DSVE_OVERLOADED_FORMS -triple aarch64-none-linux-gnu -target-feature +sve2p1 -S -disable-O0-optnone -Werror -Wall -emit-llvm -o - -x c++ %s | opt -S -p mem2reg,instcombine,tailcallelim | FileCheck %s -check-prefix=CPP-CHECK
+// RUN: %clang_cc1 -triple aarch64-none-linux-gnu -target-feature +sve2p1 -S -disable-O0-optnone -Werror -Wall -o /dev/null %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) A1##A3
+#else
+#define SVE_ACLE_FUNC(A1,A2,A3) A1##A2##A3
+#endif
+
+// CHECK-LABEL: @test_svdot_s32_x2(
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.sdot.x2.nxv4i32(<vscale x 4 x i32> [[OP1:%.*]], <vscale x 8 x i16> [[OP2:%.*]], <vscale x 8 x i16> [[OP3:%.*]])
+// CHECK-NEXT:    ret <vscale x 4 x i32> [[TMP0]]
+//
+// CPP-CHECK-LABEL: @_Z17test_svdot_s32_x2u11__SVInt32_tu11__SVInt16_tu11__SVInt16_t(
+// CPP-CHECK-NEXT:  entry:
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.sdot.x2.nxv4i32(<vscale x 4 x i32> [[OP1:%.*]], <vscale x 8 x i16> [[OP2:%.*]], <vscale x 8 x i16> [[OP3:%.*]])
+// CPP-CHECK-NEXT:    ret <vscale x 4 x i32> [[TMP0]]
+//
+svint32_t test_svdot_s32_x2(svint32_t op1, svint16_t op2, svint16_t op3)
+{
+  return SVE_ACLE_FUNC(svdot,_s32_s16_s16,)(op1, op2, op3);
+}
+
+// CHECK-LABEL: @test_svdot_u32_x2(
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.udot.x2.nxv4i32(<vscale x 4 x i32> [[OP1:%.*]], <vscale x 8 x i16> [[OP2:%.*]], <vscale x 8 x i16> [[OP3:%.*]])
+// CHECK-NEXT:    ret <vscale x 4 x i32> [[TMP0]]
+//
+// CPP-CHECK-LABEL: @_Z17test_svdot_u32_x2u12__SVUint32_tu12__SVUint16_tu12__SVUint16_t(
+// CPP-CHECK-NEXT:  entry:
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.udot.x2.nxv4i32(<vscale x 4 x i32> [[OP1:%.*]], <vscale x 8 x i16> [[OP2:%.*]], <vscale x 8 x i16> [[OP3:%.*]])
+// CPP-CHECK-NEXT:    ret <vscale x 4 x i32> [[TMP0]]
+//
+svuint32_t test_svdot_u32_x2(svuint32_t op1, svuint16_t op2, svuint16_t op3)
+{
+  return SVE_ACLE_FUNC(svdot,_u32_u16_u16,)(op1, op2, op3);
+}
+
+// CHECK-LABEL: @test_svdot_f32_x2(
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.fdot.x2.nxv4f32(<vscale x 4 x float> [[OP1:%.*]], <vscale x 8 x half> [[OP2:%.*]], <vscale x 8 x half> [[OP3:%.*]])
+// CHECK-NEXT:    ret <vscale x 4 x float> [[TMP0]]
+//
+// CPP-CHECK-LABEL: @_Z17test_svdot_f32_x2u13__SVFloat32_tu13__SVFloat16_tu13__SVFloat16_t(
+// CPP-CHECK-NEXT:  entry:
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.fdot.x2.nxv4f32(<vscale x 4 x float> [[OP1:%.*]], <vscale x 8 x half> [[OP2:%.*]], <vscale x 8 x half> [[OP3:%.*]])
+// CPP-CHECK-NEXT:    ret <vscale x 4 x float> [[TMP0]]
+//
+svfloat32_t test_svdot_f32_x2(svfloat32_t op1, svfloat16_t op2, svfloat16_t op3)
+{
+  return SVE_ACLE_FUNC(svdot,_f32_f16_f16,)(op1, op2, op3);
+}
+
+
+
+// CHECK-LABEL: @test_svdot_lane_s32_x2(
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.sdot.lane.x2.nxv4i32(<vscale x 4 x i32> [[OP1:%.*]], <vscale x 8 x i16> [[OP2:%.*]], <vscale x 8 x i16> [[OP3:%.*]], i32 3)
+// CHECK-NEXT:    ret <vscale x 4 x i32> [[TMP0]]
+//
+// CPP-CHECK-LABEL: @_Z22test_svdot_lane_s32_x2u11__SVInt32_tu11__SVInt16_tu11__SVInt16_t(
+// CPP-CHECK-NEXT:  entry:
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.sdot.lane.x2.nxv4i32(<vscale x 4 x i32> [[OP1:%.*]], <vscale x 8 x i16> [[OP2:%.*]], <vscale x 8 x i16> [[OP3:%.*]], i32 3)
+// CPP-CHECK-NEXT:    ret <vscale x 4 x i32> [[TMP0]]
+//
+svint32_t test_svdot_lane_s32_x2(svint32_t op1, svint16_t op2, svint16_t op3)
+{
+  return SVE_ACLE_FUNC(svdot_lane,_s32_s16_s16,)(op1, op2, op3, 3);
+}
+
+// CHECK-LABEL: @test_svdot_lane_u32_x2(
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.udot.lane.x2.nxv4i32(<vscale x 4 x i32> [[OP1:%.*]], <vscale x 8 x i16> [[OP2:%.*]], <vscale x 8 x i16> [[OP3:%.*]], i32 3)
+// CHECK-NEXT:    ret <vscale x 4 x i32> [[TMP0]]
+//
+// CPP-CHECK-LABEL: @_Z22test_svdot_lane_u32_x2u12__SVUint32_tu12__SVUint16_tu12__SVUint16_t(
+// CPP-CHECK-NEXT:  entry:
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sve.udot.lane.x2.nxv4i32(<vscale x 4 x i32> [[OP1:%.*]], <vscale x 8 x i16> [[OP2:%.*]], <vscale x 8 x i16> [[OP3:%.*]], i32 3)
+// CPP-CHECK-NEXT:    ret <vscale x 4 x i32> [[TMP0]]
+//
+svuint32_t test_svdot_lane_u32_x2(svuint32_t op1, svuint16_t op2, svuint16_t op3)
+{
+  return SVE_ACLE_FUNC(svdot_lane,_u32_u16_u16,)(op1, op2, op3, 3);
+}
+
+// CHECK-LABEL: @test_svdot_lane_f32_x2(
+// CHECK-NEXT:  entry:
+// CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.fdot.lane.x2.nxv4f32(<vscale x 4 x float> [[OP1:%.*]], <vscale x 8 x half> [[OP2:%.*]], <vscale x 8 x half> [[OP3:%.*]], i32 3)
+// CHECK-NEXT:    ret <vscale x 4 x float> [[TMP0]]
+//
+// CPP-CHECK-LABEL: @_Z22test_svdot_lane_f32_x2u13__SVFloat32_tu13__SVFloat16_tu13__SVFloat16_t(
+// CPP-CHECK-NEXT:  entry:
+// CPP-CHECK-NEXT:    [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.fdot.lane.x2.nxv4f32(<vscale x 4 x float> [[OP1:%.*]], <vscale x 8 x half> [[OP2:%.*]], <vscale x 8 x half> [[OP3:%.*]], i32 3)
+// CPP-CHECK-NEXT:    ret <vscale x 4 x float> [[TMP0]]
+//
+svfloat32_t test_svdot_lane_f32_x2(svfloat32_t op1, svfloat16_t op2, svfloat16_t op3)
+{
+  return SVE_ACLE_FUNC(svdot_lane,_f32_f16_f16,)(op1, op2, op3, 3);
+}

clang/test/Sema/aarch64-sve2p1-intrinsics/acle_sve2p1_imm.cpp

void test_cntp(svcount_t c) {
   svcntp_c14(c, 1); // expected-error {{argument value 1 is outside the valid range [2, 4]}}
 
   svcntp_c8(c, 3);  // expected-error {{argument should be a multiple of 2}}
   svcntp_c11(c, 3); // expected-error {{argument should be a multiple of 2}}
   svcntp_c32(c, 3); // expected-error {{argument should be a multiple of 2}}
   svcntp_c14(c, 3); // expected-error {{argument should be a multiple of 2}}
 }
 
+void test_svdot_lane_2way(svint32_t s32, svuint32_t u32, svint16_t s16, svuint16_t u16,
+                          svfloat32_t f32, svfloat16_t f16) {
+  svdot_lane_s32_s16_s16(s32, s16, s16, 1); // expected-error {{argument value 4 is outside the valid range [0, 3]}}
+  svdot_lane_u32_u16_u16(u32, u16, u16, 1); // expected-error {{argument value 4 is outside the valid range [0, 3]}}
+  svdot_lane_f32_f16_f16(f32, f16, f16, 1); // expected-error {{argument value 4 is outside the valid range [0, 3]}}
+  svdot_lane_s32_s16_s16(s32, s16, s16, 4); // expected-error {{argument value 4 is outside the valid range [0, 3]}}
+  svdot_lane_u32_u16_u16(u32, u16, u16, 4); // expected-error {{argument value 4 is outside the valid range [0, 3]}}
+  svdot_lane_f32_f16_f16(f32, f16, f16, 4); // expected-error {{argument value 4 is outside the valid range [0, 3]}}
+}