[clang] [clang][CIR] Add codegen coverage for SVE RINTP intrinsics (PR #241236)
Muneeb Khan via cfe-commits
cfe-commits at lists.llvm.org
Wed Nov 11 03:27:53 PST 2026
https://github.com/muneebkhan85 created https://github.com/llvm/llvm-project/pull/241236
Add CIR codegen coverage for the f16, f32, and f64 variants of svrintp in the zeroing, merging, and don't-care forms. Verify that they lower to aarch64.sve.frintp with the appropriate inactive value.
Move and adapt sve-intrinsics/acle_sve_rintp.c to sve/rintp.c while retaining conventional Clang, overloaded-form, C++, and streaming-mode coverage.
>From 200fe78d2064ad37a6f4aaca036ad50d33e44e96 Mon Sep 17 00:00:00 2001
From: Muneeb Khan <muneeb.khan at huawei.com>
Date: Wed, 11 Nov 2026 19:25:14 +0800
Subject: [PATCH] [clang][CIR] Add codegen coverage for SVE RINTP intrinsics
Add CIR codegen coverage for the f16, f32, and f64 variants of
svrintp in the zeroing, merging, and don't-care forms. Verify that
they lower to aarch64.sve.frintp with the appropriate inactive value.
Move and adapt sve-intrinsics/acle_sve_rintp.c to sve/rintp.c while
retaining conventional Clang, overloaded-form, C++, and streaming-mode
coverage.
---
.../AArch64/sve-intrinsics/acle_sve_rintp.c | 176 ----------------
clang/test/CodeGen/AArch64/sve/rintp.c | 191 ++++++++++++++++++
2 files changed, 191 insertions(+), 176 deletions(-)
delete mode 100644 clang/test/CodeGen/AArch64/sve-intrinsics/acle_sve_rintp.c
create mode 100644 clang/test/CodeGen/AArch64/sve/rintp.c
diff --git a/clang/test/CodeGen/AArch64/sve-intrinsics/acle_sve_rintp.c b/clang/test/CodeGen/AArch64/sve-intrinsics/acle_sve_rintp.c
deleted file mode 100644
index 2d3a888a8ba025..00000000000000
--- a/clang/test/CodeGen/AArch64/sve-intrinsics/acle_sve_rintp.c
+++ /dev/null
@@ -1,176 +0,0 @@
-// NOTE: Assertions have been autogenerated by utils/update_cc_test_checks.py
-// REQUIRES: aarch64-registered-target
-// RUN: %clang_cc1 -fclang-abi-compat=latest -triple aarch64 -target-feature +sve -disable-O0-optnone -Werror -Wall -emit-llvm -o - %s | opt -S -passes=mem2reg,tailcallelim | FileCheck %s
-// RUN: %clang_cc1 -fclang-abi-compat=latest -triple aarch64 -target-feature +sve -disable-O0-optnone -Werror -Wall -emit-llvm -o - -x c++ %s | opt -S -passes=mem2reg,tailcallelim | FileCheck %s -check-prefix=CPP-CHECK
-// RUN: %clang_cc1 -fclang-abi-compat=latest -DSVE_OVERLOADED_FORMS -triple aarch64 -target-feature +sve -disable-O0-optnone -Werror -Wall -emit-llvm -o - %s | opt -S -passes=mem2reg,tailcallelim | FileCheck %s
-// RUN: %clang_cc1 -fclang-abi-compat=latest -DSVE_OVERLOADED_FORMS -triple aarch64 -target-feature +sve -disable-O0-optnone -Werror -Wall -emit-llvm -o - -x c++ %s | opt -S -passes=mem2reg,tailcallelim | FileCheck %s -check-prefix=CPP-CHECK
-// RUN: %clang_cc1 -fclang-abi-compat=latest -triple aarch64 -target-feature +sve -S -disable-O0-optnone -Werror -Wall -o /dev/null %s
-// RUN: %clang_cc1 -fclang-abi-compat=latest -triple aarch64 -target-feature +sme -S -disable-O0-optnone -Werror -Wall -o /dev/null %s
-
-#include <arm_sve.h>
-
-#if defined __ARM_FEATURE_SME
-#define MODE_ATTR __arm_streaming
-#else
-#define MODE_ATTR
-#endif
-
-#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
-
-// CHECK-LABEL: @test_svrintp_f16_z(
-// CHECK-NEXT: entry:
-// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.frintp.nxv8f16(<vscale x 8 x half> zeroinitializer, <vscale x 8 x i1> [[TMP0]], <vscale x 8 x half> [[OP:%.*]])
-// CHECK-NEXT: ret <vscale x 8 x half> [[TMP1]]
-//
-// CPP-CHECK-LABEL: @_Z18test_svrintp_f16_zu10__SVBool_tu13__SVFloat16_t(
-// CPP-CHECK-NEXT: entry:
-// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.frintp.nxv8f16(<vscale x 8 x half> zeroinitializer, <vscale x 8 x i1> [[TMP0]], <vscale x 8 x half> [[OP:%.*]])
-// CPP-CHECK-NEXT: ret <vscale x 8 x half> [[TMP1]]
-//
-svfloat16_t test_svrintp_f16_z(svbool_t pg, svfloat16_t op) MODE_ATTR
-{
- return SVE_ACLE_FUNC(svrintp,_f16,_z,)(pg, op);
-}
-
-// CHECK-LABEL: @test_svrintp_f32_z(
-// CHECK-NEXT: entry:
-// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.frintp.nxv4f32(<vscale x 4 x float> zeroinitializer, <vscale x 4 x i1> [[TMP0]], <vscale x 4 x float> [[OP:%.*]])
-// CHECK-NEXT: ret <vscale x 4 x float> [[TMP1]]
-//
-// CPP-CHECK-LABEL: @_Z18test_svrintp_f32_zu10__SVBool_tu13__SVFloat32_t(
-// CPP-CHECK-NEXT: entry:
-// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.frintp.nxv4f32(<vscale x 4 x float> zeroinitializer, <vscale x 4 x i1> [[TMP0]], <vscale x 4 x float> [[OP:%.*]])
-// CPP-CHECK-NEXT: ret <vscale x 4 x float> [[TMP1]]
-//
-svfloat32_t test_svrintp_f32_z(svbool_t pg, svfloat32_t op) MODE_ATTR
-{
- return SVE_ACLE_FUNC(svrintp,_f32,_z,)(pg, op);
-}
-
-// CHECK-LABEL: @test_svrintp_f64_z(
-// CHECK-NEXT: entry:
-// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.frintp.nxv2f64(<vscale x 2 x double> zeroinitializer, <vscale x 2 x i1> [[TMP0]], <vscale x 2 x double> [[OP:%.*]])
-// CHECK-NEXT: ret <vscale x 2 x double> [[TMP1]]
-//
-// CPP-CHECK-LABEL: @_Z18test_svrintp_f64_zu10__SVBool_tu13__SVFloat64_t(
-// CPP-CHECK-NEXT: entry:
-// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.frintp.nxv2f64(<vscale x 2 x double> zeroinitializer, <vscale x 2 x i1> [[TMP0]], <vscale x 2 x double> [[OP:%.*]])
-// CPP-CHECK-NEXT: ret <vscale x 2 x double> [[TMP1]]
-//
-svfloat64_t test_svrintp_f64_z(svbool_t pg, svfloat64_t op) MODE_ATTR
-{
- return SVE_ACLE_FUNC(svrintp,_f64,_z,)(pg, op);
-}
-
-// CHECK-LABEL: @test_svrintp_f16_m(
-// CHECK-NEXT: entry:
-// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.frintp.nxv8f16(<vscale x 8 x half> [[INACTIVE:%.*]], <vscale x 8 x i1> [[TMP0]], <vscale x 8 x half> [[OP:%.*]])
-// CHECK-NEXT: ret <vscale x 8 x half> [[TMP1]]
-//
-// CPP-CHECK-LABEL: @_Z18test_svrintp_f16_mu13__SVFloat16_tu10__SVBool_tS_(
-// CPP-CHECK-NEXT: entry:
-// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.frintp.nxv8f16(<vscale x 8 x half> [[INACTIVE:%.*]], <vscale x 8 x i1> [[TMP0]], <vscale x 8 x half> [[OP:%.*]])
-// CPP-CHECK-NEXT: ret <vscale x 8 x half> [[TMP1]]
-//
-svfloat16_t test_svrintp_f16_m(svfloat16_t inactive, svbool_t pg, svfloat16_t op) MODE_ATTR
-{
- return SVE_ACLE_FUNC(svrintp,_f16,_m,)(inactive, pg, op);
-}
-
-// CHECK-LABEL: @test_svrintp_f32_m(
-// CHECK-NEXT: entry:
-// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.frintp.nxv4f32(<vscale x 4 x float> [[INACTIVE:%.*]], <vscale x 4 x i1> [[TMP0]], <vscale x 4 x float> [[OP:%.*]])
-// CHECK-NEXT: ret <vscale x 4 x float> [[TMP1]]
-//
-// CPP-CHECK-LABEL: @_Z18test_svrintp_f32_mu13__SVFloat32_tu10__SVBool_tS_(
-// CPP-CHECK-NEXT: entry:
-// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.frintp.nxv4f32(<vscale x 4 x float> [[INACTIVE:%.*]], <vscale x 4 x i1> [[TMP0]], <vscale x 4 x float> [[OP:%.*]])
-// CPP-CHECK-NEXT: ret <vscale x 4 x float> [[TMP1]]
-//
-svfloat32_t test_svrintp_f32_m(svfloat32_t inactive, svbool_t pg, svfloat32_t op) MODE_ATTR
-{
- return SVE_ACLE_FUNC(svrintp,_f32,_m,)(inactive, pg, op);
-}
-
-// CHECK-LABEL: @test_svrintp_f64_m(
-// CHECK-NEXT: entry:
-// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.frintp.nxv2f64(<vscale x 2 x double> [[INACTIVE:%.*]], <vscale x 2 x i1> [[TMP0]], <vscale x 2 x double> [[OP:%.*]])
-// CHECK-NEXT: ret <vscale x 2 x double> [[TMP1]]
-//
-// CPP-CHECK-LABEL: @_Z18test_svrintp_f64_mu13__SVFloat64_tu10__SVBool_tS_(
-// CPP-CHECK-NEXT: entry:
-// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.frintp.nxv2f64(<vscale x 2 x double> [[INACTIVE:%.*]], <vscale x 2 x i1> [[TMP0]], <vscale x 2 x double> [[OP:%.*]])
-// CPP-CHECK-NEXT: ret <vscale x 2 x double> [[TMP1]]
-//
-svfloat64_t test_svrintp_f64_m(svfloat64_t inactive, svbool_t pg, svfloat64_t op) MODE_ATTR
-{
- return SVE_ACLE_FUNC(svrintp,_f64,_m,)(inactive, pg, op);
-}
-
-// CHECK-LABEL: @test_svrintp_f16_x(
-// CHECK-NEXT: entry:
-// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.frintp.nxv8f16(<vscale x 8 x half> undef, <vscale x 8 x i1> [[TMP0]], <vscale x 8 x half> [[OP:%.*]])
-// CHECK-NEXT: ret <vscale x 8 x half> [[TMP1]]
-//
-// CPP-CHECK-LABEL: @_Z18test_svrintp_f16_xu10__SVBool_tu13__SVFloat16_t(
-// CPP-CHECK-NEXT: entry:
-// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.frintp.nxv8f16(<vscale x 8 x half> undef, <vscale x 8 x i1> [[TMP0]], <vscale x 8 x half> [[OP:%.*]])
-// CPP-CHECK-NEXT: ret <vscale x 8 x half> [[TMP1]]
-//
-svfloat16_t test_svrintp_f16_x(svbool_t pg, svfloat16_t op) MODE_ATTR
-{
- return SVE_ACLE_FUNC(svrintp,_f16,_x,)(pg, op);
-}
-
-// CHECK-LABEL: @test_svrintp_f32_x(
-// CHECK-NEXT: entry:
-// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.frintp.nxv4f32(<vscale x 4 x float> undef, <vscale x 4 x i1> [[TMP0]], <vscale x 4 x float> [[OP:%.*]])
-// CHECK-NEXT: ret <vscale x 4 x float> [[TMP1]]
-//
-// CPP-CHECK-LABEL: @_Z18test_svrintp_f32_xu10__SVBool_tu13__SVFloat32_t(
-// CPP-CHECK-NEXT: entry:
-// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.frintp.nxv4f32(<vscale x 4 x float> undef, <vscale x 4 x i1> [[TMP0]], <vscale x 4 x float> [[OP:%.*]])
-// CPP-CHECK-NEXT: ret <vscale x 4 x float> [[TMP1]]
-//
-svfloat32_t test_svrintp_f32_x(svbool_t pg, svfloat32_t op) MODE_ATTR
-{
- return SVE_ACLE_FUNC(svrintp,_f32,_x,)(pg, op);
-}
-
-// CHECK-LABEL: @test_svrintp_f64_x(
-// CHECK-NEXT: entry:
-// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
-// CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.frintp.nxv2f64(<vscale x 2 x double> undef, <vscale x 2 x i1> [[TMP0]], <vscale x 2 x double> [[OP:%.*]])
-// CHECK-NEXT: ret <vscale x 2 x double> [[TMP1]]
-//
-// CPP-CHECK-LABEL: @_Z18test_svrintp_f64_xu10__SVBool_tu13__SVFloat64_t(
-// CPP-CHECK-NEXT: entry:
-// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG:%.*]])
-// CPP-CHECK-NEXT: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.frintp.nxv2f64(<vscale x 2 x double> undef, <vscale x 2 x i1> [[TMP0]], <vscale x 2 x double> [[OP:%.*]])
-// CPP-CHECK-NEXT: ret <vscale x 2 x double> [[TMP1]]
-//
-svfloat64_t test_svrintp_f64_x(svbool_t pg, svfloat64_t op) MODE_ATTR
-{
- return SVE_ACLE_FUNC(svrintp,_f64,_x,)(pg, op);
-}
diff --git a/clang/test/CodeGen/AArch64/sve/rintp.c b/clang/test/CodeGen/AArch64/sve/rintp.c
new file mode 100644
index 00000000000000..6bbd2b62bf04eb
--- /dev/null
+++ b/clang/test/CodeGen/AArch64/sve/rintp.c
@@ -0,0 +1,191 @@
+// REQUIRES: aarch64-registered-target
+
+// DEFINE: %{optimize} = opt -passes=mem2reg,instcombine,tailcallelim -S
+
+// RUN: %if cir-enabled %{%clang_cc1_cg_arm64_sve -fclangir -emit-cir -disable-O0-optnone -o - %s | FileCheck %s --check-prefixes=C,CIR %}
+// RUN: %if cir-enabled %{%clang_cc1_cg_arm64_sve -DSVE_OVERLOADED_FORMS -fclangir -emit-cir -disable-O0-optnone -o - %s | FileCheck %s --check-prefixes=C,CIR %}
+
+// RUN: %if cir-enabled %{%clang_cc1_cg_arm64_sve -fclangir -emit-llvm -disable-O0-optnone -o - %s | %{optimize} | FileCheck %s --check-prefixes=C,LLVM %}
+// RUN: %if cir-enabled %{%clang_cc1_cg_arm64_sve -DSVE_OVERLOADED_FORMS -fclangir -emit-llvm -disable-O0-optnone -o - %s | %{optimize} | FileCheck %s --check-prefixes=C,LLVM %}
+// RUN: %if cir-enabled %{%clang_cc1_cg_arm64_sve -DSVE_OVERLOADED_FORMS -fclangir -emit-llvm -disable-O0-optnone -o - -x c++ %s | %{optimize} | FileCheck %s --check-prefixes=CPP,LLVM %}
+
+// RUN: %clang_cc1_cg_arm64_sve -emit-llvm -disable-O0-optnone -o - %s | %{optimize} | FileCheck %s --check-prefixes=C,LLVM
+// RUN: %clang_cc1_cg_arm64_sve -DSVE_OVERLOADED_FORMS -emit-llvm -disable-O0-optnone -o - %s | %{optimize} | FileCheck %s --check-prefixes=C,LLVM
+// RUN: %clang_cc1_cg_arm64_sve -DSVE_OVERLOADED_FORMS -emit-llvm -disable-O0-optnone -o - -x c++ %s | %{optimize} | FileCheck %s --check-prefixes=CPP,LLVM
+
+// RUN: %clang_cc1_cg_arm64_sme -S -disable-O0-optnone -Werror -Wall -o /dev/null %s
+
+//=============================================================================
+// NOTES
+//
+// Tests for SVE RINTP intrinsics
+//=============================================================================
+
+// TODO: Add CIR coverage for the SME2 streaming-only svrintp_f32_x2 and
+// svrintp_f32_x4 forms, which are tested in sme2-intrinsics/acle_sme2_frint.c,
+// once CIRGenBuiltinAArch64.cpp supports SVE tuple arguments and results.
+
+#include <arm_sve.h>
+
+#if defined __ARM_FEATURE_SME
+#define MODE_ATTR __arm_streaming
+#else
+#define MODE_ATTR
+#endif
+
+#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
+
+// C-LABEL: @test_svrintp_f16_z(
+// CPP-LABEL: @_Z18test_svrintp_f16_zu10__SVBool_tu13__SVFloat16_t(
+svfloat16_t test_svrintp_f16_z(svbool_t pg, svfloat16_t op) MODE_ATTR
+{
+// CIR: %[[ZERO:.*]] = cir.const #cir.zero : !cir.vector<[8] x !cir.f16>
+// CIR: %[[CONVERT_PG:.*]] = cir.call_llvm_intrinsic "aarch64.sve.convert.from.svbool" %{{.*}} :
+// CIR-SAME: -> !cir.vector<[8] x !cir.int<u, 1>>
+// CIR: cir.call_llvm_intrinsic "aarch64.sve.frintp" %[[ZERO]], %[[CONVERT_PG]], %{{.*}} :
+// CIR-SAME: -> !cir.vector<[8] x !cir.f16>
+
+// LLVM-SAME: <vscale x 16 x i1> [[PG:%.*]], <vscale x 8 x half> [[OP:%.*]])
+// LLVM: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG]])
+// LLVM: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.frintp.nxv8f16(<vscale x 8 x half> zeroinitializer, <vscale x 8 x i1> [[TMP0]], <vscale x 8 x half> [[OP]])
+// LLVM: ret <vscale x 8 x half> [[TMP1]]
+ return SVE_ACLE_FUNC(svrintp,_f16,_z,)(pg, op);
+}
+
+// C-LABEL: @test_svrintp_f32_z(
+// CPP-LABEL: @_Z18test_svrintp_f32_zu10__SVBool_tu13__SVFloat32_t(
+svfloat32_t test_svrintp_f32_z(svbool_t pg, svfloat32_t op) MODE_ATTR
+{
+// CIR: %[[ZERO:.*]] = cir.const #cir.zero : !cir.vector<[4] x !cir.float>
+// CIR: %[[CONVERT_PG:.*]] = cir.call_llvm_intrinsic "aarch64.sve.convert.from.svbool" %{{.*}} :
+// CIR-SAME: -> !cir.vector<[4] x !cir.int<u, 1>>
+// CIR: cir.call_llvm_intrinsic "aarch64.sve.frintp" %[[ZERO]], %[[CONVERT_PG]], %{{.*}} :
+// CIR-SAME: -> !cir.vector<[4] x !cir.float>
+
+// LLVM-SAME: <vscale x 16 x i1> [[PG:%.*]], <vscale x 4 x float> [[OP:%.*]])
+// LLVM: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG]])
+// LLVM: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.frintp.nxv4f32(<vscale x 4 x float> zeroinitializer, <vscale x 4 x i1> [[TMP0]], <vscale x 4 x float> [[OP]])
+// LLVM: ret <vscale x 4 x float> [[TMP1]]
+ return SVE_ACLE_FUNC(svrintp,_f32,_z,)(pg, op);
+}
+
+// C-LABEL: @test_svrintp_f64_z(
+// CPP-LABEL: @_Z18test_svrintp_f64_zu10__SVBool_tu13__SVFloat64_t(
+svfloat64_t test_svrintp_f64_z(svbool_t pg, svfloat64_t op) MODE_ATTR
+{
+// CIR: %[[ZERO:.*]] = cir.const #cir.zero : !cir.vector<[2] x !cir.double>
+// CIR: %[[CONVERT_PG:.*]] = cir.call_llvm_intrinsic "aarch64.sve.convert.from.svbool" %{{.*}} :
+// CIR-SAME: -> !cir.vector<[2] x !cir.int<u, 1>>
+// CIR: cir.call_llvm_intrinsic "aarch64.sve.frintp" %[[ZERO]], %[[CONVERT_PG]], %{{.*}} :
+// CIR-SAME: -> !cir.vector<[2] x !cir.double>
+
+// LLVM-SAME: <vscale x 16 x i1> [[PG:%.*]], <vscale x 2 x double> [[OP:%.*]])
+// LLVM: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG]])
+// LLVM: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.frintp.nxv2f64(<vscale x 2 x double> zeroinitializer, <vscale x 2 x i1> [[TMP0]], <vscale x 2 x double> [[OP]])
+// LLVM: ret <vscale x 2 x double> [[TMP1]]
+ return SVE_ACLE_FUNC(svrintp,_f64,_z,)(pg, op);
+}
+
+// C-LABEL: @test_svrintp_f16_m(
+// CPP-LABEL: @_Z18test_svrintp_f16_mu13__SVFloat16_tu10__SVBool_tS_(
+svfloat16_t test_svrintp_f16_m(svfloat16_t inactive, svbool_t pg, svfloat16_t op) MODE_ATTR
+{
+// CIR: %[[CONVERT_PG:.*]] = cir.call_llvm_intrinsic "aarch64.sve.convert.from.svbool" %{{.*}} :
+// CIR-SAME: -> !cir.vector<[8] x !cir.int<u, 1>>
+// CIR: cir.call_llvm_intrinsic "aarch64.sve.frintp" %{{.*}}, %[[CONVERT_PG]], %{{.*}} :
+// CIR-SAME: -> !cir.vector<[8] x !cir.f16>
+
+// LLVM-SAME: <vscale x 8 x half> [[INACTIVE:%.*]], <vscale x 16 x i1> [[PG:%.*]], <vscale x 8 x half> [[OP:%.*]])
+// LLVM: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG]])
+// LLVM: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.frintp.nxv8f16(<vscale x 8 x half> [[INACTIVE]], <vscale x 8 x i1> [[TMP0]], <vscale x 8 x half> [[OP]])
+// LLVM: ret <vscale x 8 x half> [[TMP1]]
+ return SVE_ACLE_FUNC(svrintp,_f16,_m,)(inactive, pg, op);
+}
+
+// C-LABEL: @test_svrintp_f32_m(
+// CPP-LABEL: @_Z18test_svrintp_f32_mu13__SVFloat32_tu10__SVBool_tS_(
+svfloat32_t test_svrintp_f32_m(svfloat32_t inactive, svbool_t pg, svfloat32_t op) MODE_ATTR
+{
+// CIR: %[[CONVERT_PG:.*]] = cir.call_llvm_intrinsic "aarch64.sve.convert.from.svbool" %{{.*}} :
+// CIR-SAME: -> !cir.vector<[4] x !cir.int<u, 1>>
+// CIR: cir.call_llvm_intrinsic "aarch64.sve.frintp" %{{.*}}, %[[CONVERT_PG]], %{{.*}} :
+// CIR-SAME: -> !cir.vector<[4] x !cir.float>
+
+// LLVM-SAME: <vscale x 4 x float> [[INACTIVE:%.*]], <vscale x 16 x i1> [[PG:%.*]], <vscale x 4 x float> [[OP:%.*]])
+// LLVM: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG]])
+// LLVM: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.frintp.nxv4f32(<vscale x 4 x float> [[INACTIVE]], <vscale x 4 x i1> [[TMP0]], <vscale x 4 x float> [[OP]])
+// LLVM: ret <vscale x 4 x float> [[TMP1]]
+ return SVE_ACLE_FUNC(svrintp,_f32,_m,)(inactive, pg, op);
+}
+
+// C-LABEL: @test_svrintp_f64_m(
+// CPP-LABEL: @_Z18test_svrintp_f64_mu13__SVFloat64_tu10__SVBool_tS_(
+svfloat64_t test_svrintp_f64_m(svfloat64_t inactive, svbool_t pg, svfloat64_t op) MODE_ATTR
+{
+// CIR: %[[CONVERT_PG:.*]] = cir.call_llvm_intrinsic "aarch64.sve.convert.from.svbool" %{{.*}} :
+// CIR-SAME: -> !cir.vector<[2] x !cir.int<u, 1>>
+// CIR: cir.call_llvm_intrinsic "aarch64.sve.frintp" %{{.*}}, %[[CONVERT_PG]], %{{.*}} :
+// CIR-SAME: -> !cir.vector<[2] x !cir.double>
+
+// LLVM-SAME: <vscale x 2 x double> [[INACTIVE:%.*]], <vscale x 16 x i1> [[PG:%.*]], <vscale x 2 x double> [[OP:%.*]])
+// LLVM: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG]])
+// LLVM: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.frintp.nxv2f64(<vscale x 2 x double> [[INACTIVE]], <vscale x 2 x i1> [[TMP0]], <vscale x 2 x double> [[OP]])
+// LLVM: ret <vscale x 2 x double> [[TMP1]]
+ return SVE_ACLE_FUNC(svrintp,_f64,_m,)(inactive, pg, op);
+}
+
+// C-LABEL: @test_svrintp_f16_x(
+// CPP-LABEL: @_Z18test_svrintp_f16_xu10__SVBool_tu13__SVFloat16_t(
+svfloat16_t test_svrintp_f16_x(svbool_t pg, svfloat16_t op) MODE_ATTR
+{
+// CIR: %[[UNDEF:.*]] = cir.const #cir.undef : !cir.vector<[8] x !cir.f16>
+// CIR: %[[CONVERT_PG:.*]] = cir.call_llvm_intrinsic "aarch64.sve.convert.from.svbool" %{{.*}} :
+// CIR-SAME: -> !cir.vector<[8] x !cir.int<u, 1>>
+// CIR: cir.call_llvm_intrinsic "aarch64.sve.frintp" %[[UNDEF]], %[[CONVERT_PG]], %{{.*}} :
+// CIR-SAME: -> !cir.vector<[8] x !cir.f16>
+
+// LLVM-SAME: <vscale x 16 x i1> [[PG:%.*]], <vscale x 8 x half> [[OP:%.*]])
+// LLVM: [[TMP0:%.*]] = tail call <vscale x 8 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv8i1(<vscale x 16 x i1> [[PG]])
+// LLVM: [[TMP1:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sve.frintp.nxv8f16(<vscale x 8 x half> undef, <vscale x 8 x i1> [[TMP0]], <vscale x 8 x half> [[OP]])
+// LLVM: ret <vscale x 8 x half> [[TMP1]]
+ return SVE_ACLE_FUNC(svrintp,_f16,_x,)(pg, op);
+}
+
+// C-LABEL: @test_svrintp_f32_x(
+// CPP-LABEL: @_Z18test_svrintp_f32_xu10__SVBool_tu13__SVFloat32_t(
+svfloat32_t test_svrintp_f32_x(svbool_t pg, svfloat32_t op) MODE_ATTR
+{
+// CIR: %[[UNDEF:.*]] = cir.const #cir.undef : !cir.vector<[4] x !cir.float>
+// CIR: %[[CONVERT_PG:.*]] = cir.call_llvm_intrinsic "aarch64.sve.convert.from.svbool" %{{.*}} :
+// CIR-SAME: -> !cir.vector<[4] x !cir.int<u, 1>>
+// CIR: cir.call_llvm_intrinsic "aarch64.sve.frintp" %[[UNDEF]], %[[CONVERT_PG]], %{{.*}} :
+// CIR-SAME: -> !cir.vector<[4] x !cir.float>
+
+// LLVM-SAME: <vscale x 16 x i1> [[PG:%.*]], <vscale x 4 x float> [[OP:%.*]])
+// LLVM: [[TMP0:%.*]] = tail call <vscale x 4 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv4i1(<vscale x 16 x i1> [[PG]])
+// LLVM: [[TMP1:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sve.frintp.nxv4f32(<vscale x 4 x float> undef, <vscale x 4 x i1> [[TMP0]], <vscale x 4 x float> [[OP]])
+// LLVM: ret <vscale x 4 x float> [[TMP1]]
+ return SVE_ACLE_FUNC(svrintp,_f32,_x,)(pg, op);
+}
+
+// C-LABEL: @test_svrintp_f64_x(
+// CPP-LABEL: @_Z18test_svrintp_f64_xu10__SVBool_tu13__SVFloat64_t(
+svfloat64_t test_svrintp_f64_x(svbool_t pg, svfloat64_t op) MODE_ATTR
+{
+// CIR: %[[UNDEF:.*]] = cir.const #cir.undef : !cir.vector<[2] x !cir.double>
+// CIR: %[[CONVERT_PG:.*]] = cir.call_llvm_intrinsic "aarch64.sve.convert.from.svbool" %{{.*}} :
+// CIR-SAME: -> !cir.vector<[2] x !cir.int<u, 1>>
+// CIR: cir.call_llvm_intrinsic "aarch64.sve.frintp" %[[UNDEF]], %[[CONVERT_PG]], %{{.*}} :
+// CIR-SAME: -> !cir.vector<[2] x !cir.double>
+
+// LLVM-SAME: <vscale x 16 x i1> [[PG:%.*]], <vscale x 2 x double> [[OP:%.*]])
+// LLVM: [[TMP0:%.*]] = tail call <vscale x 2 x i1> @llvm.aarch64.sve.convert.from.svbool.nxv2i1(<vscale x 16 x i1> [[PG]])
+// LLVM: [[TMP1:%.*]] = tail call <vscale x 2 x double> @llvm.aarch64.sve.frintp.nxv2f64(<vscale x 2 x double> undef, <vscale x 2 x i1> [[TMP0]], <vscale x 2 x double> [[OP]])
+// LLVM: ret <vscale x 2 x double> [[TMP1]]
+ return SVE_ACLE_FUNC(svrintp,_f64,_x,)(pg, op);
+}
More information about the cfe-commits
mailing list