[clang] 8186e15 - [SME2] Add LUTI2 and LUTI4 single Builtins and Intrinsics (#73304)
via cfe-commits
cfe-commits at lists.llvm.org
Wed Dec 6 08:36:02 PST 2023
Author: Matthew Devereau
Date: 2023-12-06T16:35:56Z
New Revision: 8186e1500b1d9709f10199b7c599274f8a6f8e3a
URL: https://github.com/llvm/llvm-project/commit/8186e1500b1d9709f10199b7c599274f8a6f8e3a
DIFF: https://github.com/llvm/llvm-project/commit/8186e1500b1d9709f10199b7c599274f8a6f8e3a.diff
LOG: [SME2] Add LUTI2 and LUTI4 single Builtins and Intrinsics (#73304)
See https://github.com/ARM-software/acle/pull/217
Patch by: Hassnaa Hamdi <hassnaa.hamdi at arm.com>
Added:
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_luti2_lane_zt.c
clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_luti4_lane_zt.c
llvm/test/CodeGen/AArch64/sme2-intrinsics-luti2-lane.ll
llvm/test/CodeGen/AArch64/sme2-intrinsics-luti4-lane.ll
Modified:
clang/include/clang/Basic/arm_sme.td
clang/test/Sema/aarch64-sme2-intrinsics/acle_sme2_imm.cpp
llvm/include/llvm/IR/IntrinsicsAArch64.td
llvm/lib/Target/AArch64/AArch64SMEInstrInfo.td
llvm/lib/Target/AArch64/SMEInstrFormats.td
Removed:
################################################################################
diff --git a/clang/include/clang/Basic/arm_sme.td b/clang/include/clang/Basic/arm_sme.td
index 1520e8ad8cebd..fcff6fe35b7ba 100644
--- a/clang/include/clang/Basic/arm_sme.td
+++ b/clang/include/clang/Basic/arm_sme.td
@@ -338,6 +338,14 @@ let TargetGuard = "sme2" in {
def SVLUTI4_LANE_ZT_X4 : Inst<"svluti4_lane_zt_{d}_x4", "4.di[i", "sUsiUibhf", MergeNone, "aarch64_sme_luti4_lane_zt_x4", [IsStreaming, IsSharedZA, IsPreservesZA], [ImmCheck<0, ImmCheck0_0>, ImmCheck<2, ImmCheck0_1>]>;
}
+//
+// lookup table expand one register
+//
+let TargetGuard = "sme2" in {
+ def SVLUTI2_LANE_ZT : Inst<"svluti2_lane_zt_{d}", "di[i", "cUcsUsiUibhf", MergeNone, "aarch64_sme_luti2_lane_zt", [IsStreaming, IsSharedZA, IsPreservesZA], [ImmCheck<0, ImmCheck0_0>, ImmCheck<2, ImmCheck0_15>]>;
+ def SVLUTI4_LANE_ZT : Inst<"svluti4_lane_zt_{d}", "di[i", "cUcsUsiUibhf", MergeNone, "aarch64_sme_luti4_lane_zt", [IsStreaming, IsSharedZA, IsPreservesZA], [ImmCheck<0, ImmCheck0_0>, ImmCheck<2, ImmCheck0_7>]>;
+}
+
//
// lookup table expand two contiguous registers
//
diff --git a/clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_luti2_lane_zt.c b/clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_luti2_lane_zt.c
new file mode 100644
index 0000000000000..cb34db3695b24
--- /dev/null
+++ b/clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_luti2_lane_zt.c
@@ -0,0 +1,138 @@
+// 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 +sve -target-feature +sme2 -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 +sve -target-feature +sme2 -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 +sve -target-feature +sme2 -S -disable-O0-optnone -Werror -Wall -o /dev/null %s
+
+#include <arm_sme_draft_spec_subject_to_change.h>
+
+
+// CHECK-LABEL: @test_svluti2_lane_zt_u8(
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sme.luti2.lane.zt.nxv16i8(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 15)
+// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
+//
+// CPP-CHECK-LABEL: @_Z23test_svluti2_lane_zt_u8u11__SVUint8_t(
+// CPP-CHECK-NEXT: entry:
+// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sme.luti2.lane.zt.nxv16i8(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 15)
+// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
+//
+svuint8_t test_svluti2_lane_zt_u8(svuint8_t zn) __arm_streaming __arm_shared_za __arm_preserves_za {
+ return svluti2_lane_zt_u8(0, zn, 15);
+}
+
+
+// CHECK-LABEL: @test_svluti2_lane_zt_s8(
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sme.luti2.lane.zt.nxv16i8(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 15)
+// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
+//
+// CPP-CHECK-LABEL: @_Z23test_svluti2_lane_zt_s8u11__SVUint8_t(
+// CPP-CHECK-NEXT: entry:
+// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sme.luti2.lane.zt.nxv16i8(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 15)
+// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
+//
+svint8_t test_svluti2_lane_zt_s8(svuint8_t zn) __arm_streaming __arm_shared_za __arm_preserves_za {
+ return svluti2_lane_zt_s8(0, zn, 15);
+}
+
+// CHECK-LABEL: @test_svluti2_lane_zt_u16(
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sme.luti2.lane.zt.nxv8i16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 15)
+// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
+//
+// CPP-CHECK-LABEL: @_Z24test_svluti2_lane_zt_u16u11__SVUint8_t(
+// CPP-CHECK-NEXT: entry:
+// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sme.luti2.lane.zt.nxv8i16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 15)
+// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
+//
+svuint16_t test_svluti2_lane_zt_u16(svuint8_t zn) __arm_streaming __arm_shared_za __arm_preserves_za {
+ return svluti2_lane_zt_u16(0, zn, 15);
+}
+
+
+// CHECK-LABEL: @test_svluti2_lane_zt_s16(
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sme.luti2.lane.zt.nxv8i16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 15)
+// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
+//
+// CPP-CHECK-LABEL: @_Z24test_svluti2_lane_zt_s16u11__SVUint8_t(
+// CPP-CHECK-NEXT: entry:
+// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sme.luti2.lane.zt.nxv8i16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 15)
+// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
+//
+svint16_t test_svluti2_lane_zt_s16(svuint8_t zn) __arm_streaming __arm_shared_za __arm_preserves_za {
+ return svluti2_lane_zt_s16(0, zn, 15);
+}
+
+// CHECK-LABEL: @test_svluti2_lane_zt_f16(
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sme.luti2.lane.zt.nxv8f16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 15)
+// CHECK-NEXT: ret <vscale x 8 x half> [[TMP0]]
+//
+// CPP-CHECK-LABEL: @_Z24test_svluti2_lane_zt_f16u11__SVUint8_t(
+// CPP-CHECK-NEXT: entry:
+// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sme.luti2.lane.zt.nxv8f16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 15)
+// CPP-CHECK-NEXT: ret <vscale x 8 x half> [[TMP0]]
+//
+svfloat16_t test_svluti2_lane_zt_f16(svuint8_t zn) __arm_streaming __arm_shared_za __arm_preserves_za {
+ return svluti2_lane_zt_f16(0, zn, 15);
+}
+
+// CHECK-LABEL: @test_svluti2_lane_zt_bf16(
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sme.luti2.lane.zt.nxv8bf16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 15)
+// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
+//
+// CPP-CHECK-LABEL: @_Z25test_svluti2_lane_zt_bf16u11__SVUint8_t(
+// CPP-CHECK-NEXT: entry:
+// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sme.luti2.lane.zt.nxv8bf16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 15)
+// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
+//
+svbfloat16_t test_svluti2_lane_zt_bf16(svuint8_t zn) __arm_streaming __arm_shared_za __arm_preserves_za {
+ return svluti2_lane_zt_bf16(0, zn, 15);
+}
+
+// CHECK-LABEL: @test_svluti2_lane_zt_u32(
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sme.luti2.lane.zt.nxv4i32(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 15)
+// CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
+//
+// CPP-CHECK-LABEL: @_Z24test_svluti2_lane_zt_u32u11__SVUint8_t(
+// CPP-CHECK-NEXT: entry:
+// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sme.luti2.lane.zt.nxv4i32(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 15)
+// CPP-CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
+//
+svuint32_t test_svluti2_lane_zt_u32(svuint8_t zn) __arm_streaming __arm_shared_za __arm_preserves_za {
+ return svluti2_lane_zt_u32(0, zn, 15);
+}
+
+// CHECK-LABEL: @test_svluti2_lane_zt_s32(
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sme.luti2.lane.zt.nxv4i32(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 15)
+// CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
+//
+// CPP-CHECK-LABEL: @_Z24test_svluti2_lane_zt_s32u11__SVUint8_t(
+// CPP-CHECK-NEXT: entry:
+// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sme.luti2.lane.zt.nxv4i32(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 15)
+// CPP-CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
+//
+svint32_t test_svluti2_lane_zt_s32(svuint8_t zn) __arm_streaming __arm_shared_za __arm_preserves_za {
+ return svluti2_lane_zt_s32(0, zn, 15);
+}
+
+// CHECK-LABEL: @test_svluti2_lane_zt_f32(
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sme.luti2.lane.zt.nxv4f32(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 15)
+// CHECK-NEXT: ret <vscale x 4 x float> [[TMP0]]
+//
+// CPP-CHECK-LABEL: @_Z24test_svluti2_lane_zt_f32u11__SVUint8_t(
+// CPP-CHECK-NEXT: entry:
+// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sme.luti2.lane.zt.nxv4f32(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 15)
+// CPP-CHECK-NEXT: ret <vscale x 4 x float> [[TMP0]]
+//
+svfloat32_t test_svluti2_lane_zt_f32(svuint8_t zn) __arm_streaming __arm_shared_za __arm_preserves_za {
+ return svluti2_lane_zt_f32(0, zn, 15);
+}
diff --git a/clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_luti4_lane_zt.c b/clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_luti4_lane_zt.c
new file mode 100644
index 0000000000000..9815b0e825b30
--- /dev/null
+++ b/clang/test/CodeGen/aarch64-sme2-intrinsics/acle_sme2_luti4_lane_zt.c
@@ -0,0 +1,137 @@
+// 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 +sve -target-feature +sme2 -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 +sve -target-feature +sme2 -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 +sve -target-feature +sme2 -S -disable-O0-optnone -Werror -Wall -o /dev/null %s
+
+#include <arm_sme_draft_spec_subject_to_change.h>
+
+
+// CHECK-LABEL: @test_svluti4_lane_zt_u8(
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sme.luti4.lane.zt.nxv16i8(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
+// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
+//
+// CPP-CHECK-LABEL: @_Z23test_svluti4_lane_zt_u8u11__SVUint8_t(
+// CPP-CHECK-NEXT: entry:
+// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sme.luti4.lane.zt.nxv16i8(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
+// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
+//
+svuint8_t test_svluti4_lane_zt_u8(svuint8_t zn) __arm_streaming __arm_shared_za __arm_preserves_za {
+ return svluti4_lane_zt_u8(0, zn, 7);
+}
+
+
+// CHECK-LABEL: @test_svluti4_lane_zt_s8(
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sme.luti4.lane.zt.nxv16i8(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
+// CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
+//
+// CPP-CHECK-LABEL: @_Z23test_svluti4_lane_zt_s8u11__SVUint8_t(
+// CPP-CHECK-NEXT: entry:
+// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 16 x i8> @llvm.aarch64.sme.luti4.lane.zt.nxv16i8(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
+// CPP-CHECK-NEXT: ret <vscale x 16 x i8> [[TMP0]]
+//
+svint8_t test_svluti4_lane_zt_s8(svuint8_t zn) __arm_streaming __arm_shared_za __arm_preserves_za {
+ return svluti4_lane_zt_s8(0, zn, 7);
+}
+
+// CHECK-LABEL: @test_svluti4_lane_zt_u16(
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sme.luti4.lane.zt.nxv8i16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
+// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
+//
+// CPP-CHECK-LABEL: @_Z24test_svluti4_lane_zt_u16u11__SVUint8_t(
+// CPP-CHECK-NEXT: entry:
+// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sme.luti4.lane.zt.nxv8i16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
+// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
+//
+svuint16_t test_svluti4_lane_zt_u16(svuint8_t zn) __arm_streaming __arm_shared_za __arm_preserves_za {
+ return svluti4_lane_zt_u16(0, zn, 7);
+}
+
+// CHECK-LABEL: @test_svluti4_lane_zt_s16(
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sme.luti4.lane.zt.nxv8i16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
+// CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
+//
+// CPP-CHECK-LABEL: @_Z24test_svluti4_lane_zt_s16u11__SVUint8_t(
+// CPP-CHECK-NEXT: entry:
+// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x i16> @llvm.aarch64.sme.luti4.lane.zt.nxv8i16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
+// CPP-CHECK-NEXT: ret <vscale x 8 x i16> [[TMP0]]
+//
+svint16_t test_svluti4_lane_zt_s16(svuint8_t zn) __arm_streaming __arm_shared_za __arm_preserves_za {
+ return svluti4_lane_zt_s16(0, zn, 7);
+}
+
+// CHECK-LABEL: @test_svluti4_lane_zt_f16(
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sme.luti4.lane.zt.nxv8f16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
+// CHECK-NEXT: ret <vscale x 8 x half> [[TMP0]]
+//
+// CPP-CHECK-LABEL: @_Z24test_svluti4_lane_zt_f16u11__SVUint8_t(
+// CPP-CHECK-NEXT: entry:
+// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x half> @llvm.aarch64.sme.luti4.lane.zt.nxv8f16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
+// CPP-CHECK-NEXT: ret <vscale x 8 x half> [[TMP0]]
+//
+svfloat16_t test_svluti4_lane_zt_f16(svuint8_t zn) __arm_streaming __arm_shared_za __arm_preserves_za {
+ return svluti4_lane_zt_f16(0, zn, 7);
+}
+
+// CHECK-LABEL: @test_svluti4_lane_zt_bf16(
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sme.luti4.lane.zt.nxv8bf16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
+// CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
+//
+// CPP-CHECK-LABEL: @_Z25test_svluti4_lane_zt_bf16u11__SVUint8_t(
+// CPP-CHECK-NEXT: entry:
+// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 8 x bfloat> @llvm.aarch64.sme.luti4.lane.zt.nxv8bf16(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
+// CPP-CHECK-NEXT: ret <vscale x 8 x bfloat> [[TMP0]]
+//
+svbfloat16_t test_svluti4_lane_zt_bf16(svuint8_t zn) __arm_streaming __arm_shared_za __arm_preserves_za {
+ return svluti4_lane_zt_bf16(0, zn, 7);
+}
+
+// CHECK-LABEL: @test_svluti4_lane_zt_u32(
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sme.luti4.lane.zt.nxv4i32(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
+// CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
+//
+// CPP-CHECK-LABEL: @_Z24test_svluti4_lane_zt_u32u11__SVUint8_t(
+// CPP-CHECK-NEXT: entry:
+// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sme.luti4.lane.zt.nxv4i32(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
+// CPP-CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
+//
+svuint32_t test_svluti4_lane_zt_u32(svuint8_t zn) __arm_streaming __arm_shared_za __arm_preserves_za {
+ return svluti4_lane_zt_u32(0, zn, 7);
+}
+
+// CHECK-LABEL: @test_svluti4_lane_zt_s32(
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sme.luti4.lane.zt.nxv4i32(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
+// CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
+//
+// CPP-CHECK-LABEL: @_Z24test_svluti4_lane_zt_s32u11__SVUint8_t(
+// CPP-CHECK-NEXT: entry:
+// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x i32> @llvm.aarch64.sme.luti4.lane.zt.nxv4i32(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
+// CPP-CHECK-NEXT: ret <vscale x 4 x i32> [[TMP0]]
+//
+svint32_t test_svluti4_lane_zt_s32(svuint8_t zn) __arm_streaming __arm_shared_za __arm_preserves_za {
+ return svluti4_lane_zt_s32(0, zn, 7);
+}
+
+// CHECK-LABEL: @test_svluti4_lane_zt_f32(
+// CHECK-NEXT: entry:
+// CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sme.luti4.lane.zt.nxv4f32(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
+// CHECK-NEXT: ret <vscale x 4 x float> [[TMP0]]
+//
+// CPP-CHECK-LABEL: @_Z24test_svluti4_lane_zt_f32u11__SVUint8_t(
+// CPP-CHECK-NEXT: entry:
+// CPP-CHECK-NEXT: [[TMP0:%.*]] = tail call <vscale x 4 x float> @llvm.aarch64.sme.luti4.lane.zt.nxv4f32(i32 0, <vscale x 16 x i8> [[ZN:%.*]], i32 7)
+// CPP-CHECK-NEXT: ret <vscale x 4 x float> [[TMP0]]
+//
+svfloat32_t test_svluti4_lane_zt_f32(svuint8_t zn) __arm_streaming __arm_shared_za __arm_preserves_za {
+ return svluti4_lane_zt_f32(0, zn, 7);
+}
diff --git a/clang/test/Sema/aarch64-sme2-intrinsics/acle_sme2_imm.cpp b/clang/test/Sema/aarch64-sme2-intrinsics/acle_sme2_imm.cpp
index 2f42435804aa3..3750045fd8dff 100644
--- a/clang/test/Sema/aarch64-sme2-intrinsics/acle_sme2_imm.cpp
+++ b/clang/test/Sema/aarch64-sme2-intrinsics/acle_sme2_imm.cpp
@@ -78,6 +78,60 @@ void test_svluti4_lane_zt_x4(svuint8_t zn) __arm_streaming __arm_shared_za __arm
svluti4_lane_zt_f32_x4(0, zn, 2); // expected-error {{argument value 2 is outside the valid range [0, 1]}}
}
+void test_svluti2_lane_zt(svuint8_t zn_u8) __arm_streaming __arm_shared_za __arm_preserves_za {
+ // Test Reg Offset
+ svluti2_lane_zt_u8(1, zn_u8, 2); // expected-error {{argument value 1 is outside the valid range [0, 0]}}
+ // Test index value range
+ svluti2_lane_zt_u8(0, zn_u8, 16); // expected-error {{argument value 16 is outside the valid range [0, 15]}}
+ // Test Reg Offset
+ svluti2_lane_zt_u16(1, zn_u8, 2); // expected-error {{argument value 1 is outside the valid range [0, 0]}}
+ // Test index value range
+ svluti2_lane_zt_u16(0, zn_u8, 16); // expected-error {{argument value 16 is outside the valid range [0, 15]}}
+ // Test Reg Offset
+ svluti2_lane_zt_f16(1, zn_u8, 2); // expected-error {{argument value 1 is outside the valid range [0, 0]}}
+ // Test index value range
+ svluti2_lane_zt_f16(0, zn_u8, 16); // expected-error {{argument value 16 is outside the valid range [0, 15]}}
+ // Test Reg Offset
+ svluti2_lane_zt_bf16(1, zn_u8, 2); // expected-error {{argument value 1 is outside the valid range [0, 0]}}
+ // Test index value range
+ svluti2_lane_zt_bf16(0, zn_u8, 16); // expected-error {{argument value 16 is outside the valid range [0, 15]}}
+ // Test Reg Offset
+ svluti2_lane_zt_u32(1, zn_u8, 2); // expected-error {{argument value 1 is outside the valid range [0, 0]}}
+ // Test index value range
+ svluti2_lane_zt_u32(0, zn_u8, 16); // expected-error {{argument value 16 is outside the valid range [0, 15]}}
+ // Test Reg Offset
+ svluti2_lane_zt_f32(1, zn_u8, 2); // expected-error {{argument value 1 is outside the valid range [0, 0]}}
+ // Test index value range
+ svluti2_lane_zt_f32(0, zn_u8, 16); // expected-error {{argument value 16 is outside the valid range [0, 15]}}
+}
+
+void test_svluti4_lane_zt(svuint8_t zn_u8) __arm_streaming __arm_shared_za __arm_preserves_za {
+ // Test Reg Offset
+ svluti4_lane_zt_u8(1, zn_u8, 2); // expected-error {{argument value 1 is outside the valid range [0, 0]}}
+ // Test index value range
+ svluti4_lane_zt_u8(0, zn_u8, 8); // expected-error {{argument value 8 is outside the valid range [0, 7]}}
+ // Test Reg Offset
+ svluti4_lane_zt_u16(1, zn_u8, 2); // expected-error {{argument value 1 is outside the valid range [0, 0]}}
+ // Test index value range
+ svluti4_lane_zt_u16(0, zn_u8, 8); // expected-error {{argument value 8 is outside the valid range [0, 7]}}
+ // Test Reg Offset
+ svluti4_lane_zt_f16(1, zn_u8, 2); // expected-error {{argument value 1 is outside the valid range [0, 0]}}
+ // Test index value range
+ svluti4_lane_zt_f16(0, zn_u8, 8); // expected-error {{argument value 8 is outside the valid range [0, 7]}}
+ // Test Reg Offset
+ svluti4_lane_zt_bf16(1, zn_u8, 2); // expected-error {{argument value 1 is outside the valid range [0, 0]}}
+ // Test index value range
+ svluti4_lane_zt_bf16(0, zn_u8, 8); // expected-error {{argument value 8 is outside the valid range [0, 7]}}
+ // Test Reg Offset
+ svluti4_lane_zt_u32(1, zn_u8, 2); // expected-error {{argument value 1 is outside the valid range [0, 0]}}
+ // Test index value range
+ svluti4_lane_zt_u32(0, zn_u8, 8); // expected-error {{argument value 8 is outside the valid range [0, 7]}}
+ // Test Reg Offset
+ svluti4_lane_zt_f32(1, zn_u8, 2); // expected-error {{argument value 1 is outside the valid range [0, 0]}}
+ // Test index value range
+ svluti4_lane_zt_f32(0, zn_u8, 8); // expected-error {{argument value 8 is outside the valid range [0, 7]}}
+}
+
void test_svluti2_lane_zt_x2(svuint8_t zn_u8) __arm_streaming __arm_shared_za __arm_preserves_za {
// Test Reg Offset
svluti2_lane_zt_u8_x2(1, zn_u8, 2); // expected-error {{argument value 1 is outside the valid range [0, 0]}}
diff --git a/llvm/include/llvm/IR/IntrinsicsAArch64.td b/llvm/include/llvm/IR/IntrinsicsAArch64.td
index 0e6bf76ef0847..83fc208e7f7ee 100644
--- a/llvm/include/llvm/IR/IntrinsicsAArch64.td
+++ b/llvm/include/llvm/IR/IntrinsicsAArch64.td
@@ -3550,18 +3550,15 @@ let TargetPrefix = "aarch64" in {
def int_aarch64_sme_zero_zt : DefaultAttrsIntrinsic<[], [llvm_i32_ty], [ImmArg<ArgIndex<0>>, IntrWriteMem]>;
//
- // Lookup table expand four registers
+ // Lookup table expand one register
//
- def int_aarch64_sme_luti2_lane_zt_x4
- : DefaultAttrsIntrinsic<[llvm_anyvector_ty, LLVMMatchType<0>, LLVMMatchType<0>, LLVMMatchType<0>],
- [llvm_i32_ty, llvm_nxv16i8_ty, llvm_i32_ty],
+ def int_aarch64_sme_luti2_lane_zt
+ : DefaultAttrsIntrinsic<[llvm_anyvector_ty], [llvm_i32_ty, llvm_nxv16i8_ty, llvm_i32_ty],
[ImmArg<ArgIndex<0>>, ImmArg<ArgIndex<2>>, IntrReadMem]>;
- def int_aarch64_sme_luti4_lane_zt_x4
- : DefaultAttrsIntrinsic<[llvm_anyvector_ty, LLVMMatchType<0>, LLVMMatchType<0>, LLVMMatchType<0>],
- [llvm_i32_ty, llvm_nxv16i8_ty, llvm_i32_ty],
+ def int_aarch64_sme_luti4_lane_zt
+ : DefaultAttrsIntrinsic<[llvm_anyvector_ty], [llvm_i32_ty, llvm_nxv16i8_ty, llvm_i32_ty],
[ImmArg<ArgIndex<0>>, ImmArg<ArgIndex<2>>, IntrReadMem]>;
-
- //
+
// Lookup table expand two registers
//
def int_aarch64_sme_luti2_lane_zt_x2
@@ -3570,7 +3567,18 @@ let TargetPrefix = "aarch64" in {
def int_aarch64_sme_luti4_lane_zt_x2
: DefaultAttrsIntrinsic<[llvm_anyvector_ty, LLVMMatchType<0>], [llvm_i32_ty, llvm_nxv16i8_ty, llvm_i32_ty],
[ImmArg<ArgIndex<0>>, ImmArg<ArgIndex<2>>, IntrReadMem]>;
-
+
+ //
+ // Lookup table expand four registers
+ //
+ def int_aarch64_sme_luti2_lane_zt_x4
+ : DefaultAttrsIntrinsic<[llvm_anyvector_ty, LLVMMatchType<0>, LLVMMatchType<0>, LLVMMatchType<0>],
+ [llvm_i32_ty, llvm_nxv16i8_ty, llvm_i32_ty],
+ [ImmArg<ArgIndex<0>>, ImmArg<ArgIndex<2>>, IntrReadMem]>;
+ def int_aarch64_sme_luti4_lane_zt_x4
+ : DefaultAttrsIntrinsic<[llvm_anyvector_ty, LLVMMatchType<0>, LLVMMatchType<0>, LLVMMatchType<0>],
+ [llvm_i32_ty, llvm_nxv16i8_ty, llvm_i32_ty],
+ [ImmArg<ArgIndex<0>>, ImmArg<ArgIndex<2>>, IntrReadMem]>;
}
// SVE2.1 - ZIPQ1, ZIPQ2, UZPQ1, UZPQ2
diff --git a/llvm/lib/Target/AArch64/AArch64SMEInstrInfo.td b/llvm/lib/Target/AArch64/AArch64SMEInstrInfo.td
index a58799116003d..738a52eebad2a 100644
--- a/llvm/lib/Target/AArch64/AArch64SMEInstrInfo.td
+++ b/llvm/lib/Target/AArch64/AArch64SMEInstrInfo.td
@@ -549,11 +549,11 @@ defm STR_TX : sme2_spill_fill_vector<"str", 0b11111100, int_aarch64_sme_str_zt>;
def MOVT_XTI : sme2_movt_zt_to_scalar<"movt", 0b0011111>;
def MOVT_TIX : sme2_movt_scalar_to_zt<"movt", 0b0011111>;
-defm LUTI2_ZTZI : sme2_luti2_vector_index<"luti2">;
+defm LUTI2_ZTZI : sme2_luti2_vector_index<"luti2", int_aarch64_sme_luti2_lane_zt>;
defm LUTI2_2ZTZI : sme2_luti2_vector_vg2_index<"luti2">;
defm LUTI2_4ZTZI : sme2_luti2_vector_vg4_index<"luti2">;
-defm LUTI4_ZTZI : sme2_luti4_vector_index<"luti4">;
+defm LUTI4_ZTZI : sme2_luti4_vector_index<"luti4", int_aarch64_sme_luti4_lane_zt>;
defm LUTI4_2ZTZI : sme2_luti4_vector_vg2_index<"luti4">;
defm LUTI4_4ZTZI : sme2_luti4_vector_vg4_index<"luti4">;
diff --git a/llvm/lib/Target/AArch64/SMEInstrFormats.td b/llvm/lib/Target/AArch64/SMEInstrFormats.td
index 374d6a3abc381..4f8917618ea40 100644
--- a/llvm/lib/Target/AArch64/SMEInstrFormats.td
+++ b/llvm/lib/Target/AArch64/SMEInstrFormats.td
@@ -3227,28 +3227,54 @@ class sme2_luti_vector_index<bits<2> sz, bits<7> opc, RegisterOperand vector_ty,
class sme2_luti2_vector_index<bits<2> sz, RegisterOperand vector_ty,
string mnemonic>
- : sme2_luti_vector_index<sz, {1,?,?,?,?,0,0}, vector_ty, VectorIndexB, mnemonic> {
+ : sme2_luti_vector_index<sz, {1,?,?,?,?,0,0}, vector_ty, VectorIndexB32b_timm, mnemonic> {
bits<4> i;
let Inst{17-14} = i;
}
-multiclass sme2_luti2_vector_index<string mnemonic> {
+multiclass sme2_luti2_vector_index<string mnemonic, SDPatternOperator intrinsic> {
def _B : sme2_luti2_vector_index<0b00, ZPR8, mnemonic>;
def _H : sme2_luti2_vector_index<0b01, ZPR16, mnemonic>;
def _S : sme2_luti2_vector_index<0b10, ZPR32, mnemonic>;
+
+ def : Pat<(nxv16i8 (intrinsic (imm_to_zt untyped:$zt), nxv16i8:$zn, (i32 VectorIndexB32b_timm:$imm))),
+ (!cast<Instruction>(NAME # _B) $zt, nxv16i8:$zn, (i32 VectorIndexB32b_timm:$imm))>;
+ def : Pat<(nxv8i16 (intrinsic (imm_to_zt untyped:$zt), nxv16i8:$zn, (i32 VectorIndexB32b_timm:$imm))),
+ (!cast<Instruction>(NAME # _H) $zt, nxv16i8:$zn, (i32 VectorIndexB32b_timm:$imm))>;
+ def : Pat<(nxv4i32 (intrinsic (imm_to_zt untyped:$zt), nxv16i8:$zn, (i32 VectorIndexB32b_timm:$imm))),
+ (!cast<Instruction>(NAME # _S) $zt, nxv16i8:$zn, (i32 VectorIndexB32b_timm:$imm))>;
+ def : Pat<(nxv8f16 (intrinsic (imm_to_zt untyped:$zt), nxv16i8:$zn, (i32 VectorIndexB32b_timm:$imm))),
+ (!cast<Instruction>(NAME # _H) $zt, nxv16i8:$zn, (i32 VectorIndexB32b_timm:$imm))>;
+ def : Pat<(nxv8bf16 (intrinsic (imm_to_zt untyped:$zt), nxv16i8:$zn, (i32 VectorIndexB32b_timm:$imm))),
+ (!cast<Instruction>(NAME # _H) $zt, nxv16i8:$zn, (i32 VectorIndexB32b_timm:$imm))>;
+ def : Pat<(nxv4f32 (intrinsic (imm_to_zt untyped:$zt), nxv16i8:$zn, (i32 VectorIndexB32b_timm:$imm))),
+ (!cast<Instruction>(NAME # _S) $zt, nxv16i8:$zn, (i32 VectorIndexB32b_timm:$imm))>;
}
class sme2_luti4_vector_index<bits<2> sz, RegisterOperand vector_ty,
string mnemonic>
- : sme2_luti_vector_index<sz, {0,1,?,?,?,0,0}, vector_ty, VectorIndexH, mnemonic> {
+ : sme2_luti_vector_index<sz, {0,1,?,?,?,0,0}, vector_ty, VectorIndexH32b_timm, mnemonic> {
bits<3> i;
let Inst{16-14} = i;
}
-multiclass sme2_luti4_vector_index<string mnemonic> {
+multiclass sme2_luti4_vector_index<string mnemonic, SDPatternOperator intrinsic> {
def _B : sme2_luti4_vector_index<0b00, ZPR8, mnemonic>;
def _H : sme2_luti4_vector_index<0b01, ZPR16, mnemonic>;
def _S : sme2_luti4_vector_index<0b10, ZPR32, mnemonic>;
+
+ def : Pat<(nxv16i8 (intrinsic (imm_to_zt untyped:$zt), nxv16i8:$zn, (i32 VectorIndexH32b_timm:$imm))),
+ (!cast<Instruction>(NAME # _B) $zt, nxv16i8:$zn, (i32 VectorIndexH32b_timm:$imm))>;
+ def : Pat<(nxv8i16 (intrinsic (imm_to_zt untyped:$zt), nxv16i8:$zn, (i32 VectorIndexH32b_timm:$imm))),
+ (!cast<Instruction>(NAME # _H) $zt, nxv16i8:$zn, (i32 VectorIndexH32b_timm:$imm))>;
+ def : Pat<(nxv4i32 (intrinsic (imm_to_zt untyped:$zt), nxv16i8:$zn, (i32 VectorIndexH32b_timm:$imm))),
+ (!cast<Instruction>(NAME # _S) $zt, nxv16i8:$zn, (i32 VectorIndexH32b_timm:$imm))>;
+ def : Pat<(nxv8f16 (intrinsic (imm_to_zt untyped:$zt), nxv16i8:$zn, (i32 VectorIndexH32b_timm:$imm))),
+ (!cast<Instruction>(NAME # _H) $zt, nxv16i8:$zn, (i32 VectorIndexH32b_timm:$imm))>;
+ def : Pat<(nxv8bf16 (intrinsic (imm_to_zt untyped:$zt), nxv16i8:$zn, (i32 VectorIndexH32b_timm:$imm))),
+ (!cast<Instruction>(NAME # _H) $zt, nxv16i8:$zn, (i32 VectorIndexH32b_timm:$imm))>;
+ def : Pat<(nxv4f32 (intrinsic (imm_to_zt untyped:$zt), nxv16i8:$zn, (i32 VectorIndexH32b_timm:$imm))),
+ (!cast<Instruction>(NAME # _S) $zt, nxv16i8:$zn, (i32 VectorIndexH32b_timm:$imm))>;
}
// SME2 lookup table expand two contiguous registers
diff --git a/llvm/test/CodeGen/AArch64/sme2-intrinsics-luti2-lane.ll b/llvm/test/CodeGen/AArch64/sme2-intrinsics-luti2-lane.ll
new file mode 100644
index 0000000000000..1516e301196a7
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/sme2-intrinsics-luti2-lane.ll
@@ -0,0 +1,65 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
+; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sme2 < %s | FileCheck %s
+
+; lookup table expand one register
+
+define <vscale x 16 x i8> @luti2_i8(<vscale x 16 x i8> %x) {
+; CHECK-LABEL: luti2_i8:
+; CHECK: // %bb.0:
+; CHECK-NEXT: luti2 z0.b, zt0, z0[15]
+; CHECK-NEXT: ret
+ %res = call <vscale x 16 x i8> @llvm.aarch64.sme.luti2.lane.zt.nxv16i8(i32 0, <vscale x 16 x i8> %x, i32 15)
+ ret <vscale x 16 x i8> %res
+}
+
+define <vscale x 8 x i16> @luti2_i16(<vscale x 16 x i8> %x) {
+; CHECK-LABEL: luti2_i16:
+; CHECK: // %bb.0:
+; CHECK-NEXT: luti2 z0.h, zt0, z0[15]
+; CHECK-NEXT: ret
+ %res = call <vscale x 8 x i16> @llvm.aarch64.sme.luti2.lane.zt.nxv8i16(i32 0, <vscale x 16 x i8> %x, i32 15)
+ ret <vscale x 8 x i16> %res
+}
+
+define <vscale x 4 x i32> @luti2_i32(<vscale x 16 x i8> %x) {
+; CHECK-LABEL: luti2_i32:
+; CHECK: // %bb.0:
+; CHECK-NEXT: luti2 z0.s, zt0, z0[15]
+; CHECK-NEXT: ret
+ %res = call <vscale x 4 x i32> @llvm.aarch64.sme.luti2.lane.zt.nxv4i32(i32 0, <vscale x 16 x i8> %x, i32 15)
+ ret <vscale x 4 x i32> %res
+}
+
+define <vscale x 8 x half> @luti2_f16(<vscale x 16 x i8> %x) {
+; CHECK-LABEL: luti2_f16:
+; CHECK: // %bb.0:
+; CHECK-NEXT: luti2 z0.h, zt0, z0[15]
+; CHECK-NEXT: ret
+ %res = call <vscale x 8 x half> @llvm.aarch64.sme.luti2.lane.zt.nxv8f16(i32 0, <vscale x 16 x i8> %x, i32 15)
+ ret <vscale x 8 x half> %res
+}
+
+define <vscale x 8 x bfloat> @luti2_bf16(<vscale x 16 x i8> %x) {
+; CHECK-LABEL: luti2_bf16:
+; CHECK: // %bb.0:
+; CHECK-NEXT: luti2 z0.h, zt0, z0[15]
+; CHECK-NEXT: ret
+ %res = call <vscale x 8 x bfloat> @llvm.aarch64.sme.luti2.lane.zt.nxv8bf16(i32 0, <vscale x 16 x i8> %x, i32 15)
+ ret <vscale x 8 x bfloat> %res
+}
+
+define <vscale x 4 x float> @luti2_f32(<vscale x 16 x i8> %x) {
+; CHECK-LABEL: luti2_f32:
+; CHECK: // %bb.0:
+; CHECK-NEXT: luti2 z0.s, zt0, z0[15]
+; CHECK-NEXT: ret
+ %res = call <vscale x 4 x float> @llvm.aarch64.sme.luti2.lane.zt.nxv4f32(i32 0, <vscale x 16 x i8> %x, i32 15)
+ ret <vscale x 4 x float> %res
+}
+
+declare <vscale x 16 x i8> @llvm.aarch64.sme.luti2.lane.zt.nxv16i8(i32, <vscale x 16 x i8>, i32)
+declare <vscale x 8 x i16> @llvm.aarch64.sme.luti2.lane.zt.nxv8i16(i32, <vscale x 16 x i8>, i32)
+declare <vscale x 4 x i32> @llvm.aarch64.sme.luti2.lane.zt.nxv4i32(i32, <vscale x 16 x i8>, i32)
+declare <vscale x 8 x half> @llvm.aarch64.sme.luti2.lane.zt.nxv8f16(i32, <vscale x 16 x i8>, i32)
+declare <vscale x 8 x bfloat> @llvm.aarch64.sme.luti2.lane.zt.nxv8bf16(i32, <vscale x 16 x i8>, i32)
+declare <vscale x 4 x float> @llvm.aarch64.sme.luti2.lane.zt.nxv4f32(i32, <vscale x 16 x i8>, i32)
diff --git a/llvm/test/CodeGen/AArch64/sme2-intrinsics-luti4-lane.ll b/llvm/test/CodeGen/AArch64/sme2-intrinsics-luti4-lane.ll
new file mode 100644
index 0000000000000..6202398b1534b
--- /dev/null
+++ b/llvm/test/CodeGen/AArch64/sme2-intrinsics-luti4-lane.ll
@@ -0,0 +1,65 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 2
+; RUN: llc -mtriple=aarch64-linux-gnu -mattr=+sme2 < %s | FileCheck %s
+
+; lookup table expand one register
+
+define <vscale x 16 x i8> @luti4_i8(<vscale x 16 x i8> %x) {
+; CHECK-LABEL: luti4_i8:
+; CHECK: // %bb.0:
+; CHECK-NEXT: luti4 z0.b, zt0, z0[7]
+; CHECK-NEXT: ret
+ %res = call <vscale x 16 x i8> @llvm.aarch64.sme.luti4.lane.zt.nxv16i8(i32 0, <vscale x 16 x i8> %x, i32 7)
+ ret <vscale x 16 x i8> %res
+}
+
+define <vscale x 8 x i16> @luti4_i16(<vscale x 16 x i8> %x) {
+; CHECK-LABEL: luti4_i16:
+; CHECK: // %bb.0:
+; CHECK-NEXT: luti4 z0.h, zt0, z0[7]
+; CHECK-NEXT: ret
+ %res = call <vscale x 8 x i16> @llvm.aarch64.sme.luti4.lane.zt.nxv8i16(i32 0, <vscale x 16 x i8> %x, i32 7)
+ ret <vscale x 8 x i16> %res
+}
+
+define <vscale x 4 x i32> @luti4_i32(<vscale x 16 x i8> %x) {
+; CHECK-LABEL: luti4_i32:
+; CHECK: // %bb.0:
+; CHECK-NEXT: luti4 z0.s, zt0, z0[7]
+; CHECK-NEXT: ret
+ %res = call <vscale x 4 x i32> @llvm.aarch64.sme.luti4.lane.zt.nxv4i32(i32 0, <vscale x 16 x i8> %x, i32 7)
+ ret <vscale x 4 x i32> %res
+}
+
+define <vscale x 8 x half> @luti4_f16(<vscale x 16 x i8> %x) {
+; CHECK-LABEL: luti4_f16:
+; CHECK: // %bb.0:
+; CHECK-NEXT: luti4 z0.h, zt0, z0[7]
+; CHECK-NEXT: ret
+ %res = call <vscale x 8 x half> @llvm.aarch64.sme.luti4.lane.zt.nxv8f16(i32 0, <vscale x 16 x i8> %x, i32 7)
+ ret <vscale x 8 x half> %res
+}
+
+define <vscale x 8 x bfloat> @luti4_bf16(<vscale x 16 x i8> %x) {
+; CHECK-LABEL: luti4_bf16:
+; CHECK: // %bb.0:
+; CHECK-NEXT: luti4 z0.h, zt0, z0[7]
+; CHECK-NEXT: ret
+ %res = call <vscale x 8 x bfloat> @llvm.aarch64.sme.luti4.lane.zt.nxv8bf16(i32 0, <vscale x 16 x i8> %x, i32 7)
+ ret <vscale x 8 x bfloat> %res
+}
+
+define <vscale x 4 x float> @luti4_f32(<vscale x 16 x i8> %x) {
+; CHECK-LABEL: luti4_f32:
+; CHECK: // %bb.0:
+; CHECK-NEXT: luti4 z0.s, zt0, z0[7]
+; CHECK-NEXT: ret
+ %res = call <vscale x 4 x float> @llvm.aarch64.sme.luti4.lane.zt.nxv4f32(i32 0, <vscale x 16 x i8> %x, i32 7)
+ ret <vscale x 4 x float> %res
+}
+
+declare <vscale x 16 x i8> @llvm.aarch64.sme.luti4.lane.zt.nxv16i8(i32, <vscale x 16 x i8>, i32)
+declare <vscale x 8 x i16> @llvm.aarch64.sme.luti4.lane.zt.nxv8i16(i32, <vscale x 16 x i8>, i32)
+declare <vscale x 4 x i32> @llvm.aarch64.sme.luti4.lane.zt.nxv4i32(i32, <vscale x 16 x i8>, i32)
+declare <vscale x 8 x half> @llvm.aarch64.sme.luti4.lane.zt.nxv8f16(i32, <vscale x 16 x i8>, i32)
+declare <vscale x 8 x bfloat> @llvm.aarch64.sme.luti4.lane.zt.nxv8bf16(i32, <vscale x 16 x i8>, i32)
+declare <vscale x 4 x float> @llvm.aarch64.sme.luti4.lane.zt.nxv4f32(i32, <vscale x 16 x i8>, i32)
More information about the cfe-commits
mailing list