[AArch64] Support for Ampere1 core

[philipp.tomsich]

Summary

Add support for the Ampere Computing Ampere1 core.
Ampere1 implements the AArch64 state and is compatible with ARMv8.6-A.

[dmgreen]

You could consider adding some scheduling tests, like those in llvm/test/tools/llvm-mca/AArch64/Cortex/A55-basic-instructions.s. In the past we have tended not to add them for schedules, so they are not necessary, but can be useful for checking the details are as you expect.

llvm/include/llvm/Support/AArch64TargetParser.def

288

I believe a lot of these (BF16, DotProd etc) are mandatory in armv8.6 and should already be included without needing to specify them here.

llvm/lib/Target/AArch64/AArch64.td

1191

The formatting here is a bit off, compared to all the others.

llvm/lib/Target/AArch64/AArch64SchedAmpere1.td

14

This line doesn't add a lot :)

16

an out-of-order micro-architecture

llvm/lib/Target/AArch64/AArch64SchedPredAmpere.td

17

I don't deal with these predicates a lot, but the formatting looks like it could be better. Is it using tabs?

llvm/lib/Target/AArch64/AArch64Subtarget.h

83

This list can be alphabetical.

[philipp.tomsich]

Update to reflect review results.

[dmgreen]

I don't know the details of the scheduling latencies for the core, but this looks perfectly sensible. There are a few comments inline, but other than those this patch LGTM.

llvm/include/llvm/Support/AArch64TargetParser.def

289

The formatting is a bit off here.

llvm/lib/Target/AArch64/AArch64.td

1053

Should this include SSBS and MTE too, if they are included in the target parsing?

llvm/lib/Target/AArch64/AArch64Subtarget.h

83

This still needs to be alphabetical

[philipp.tomsich]

• additional tab -> space conversion (hopefully, I have emacs configured correctly now)
• added SBSS and MTE to the SubtargetFeatures
• inserted AMpere1 in the subtarget enum in alphabetical order

[philipp.tomsich]

Adds two fixes for issues uncovered by the CI:

• The update of NumAArch64CPUArchs had been lost in a "rebase gone wrong" (i.e., another target was added and the previously increased-by-1 value now was the already present upstream). This adjusts the value of NumAArch64CPUArchs to again account for the addition of an extra target.
• Adjusts the ARMCPUTestParams for ampere1 to reflect that ARMv8.6 encodes the crypto extensions differently than lower architecture levels.

[philipp.tomsich]

Rebased to main.
The earlier base-regression was a precommitted test-case that caused
SimpleLoopUnswitch tests (not related to these changes) to fail.

clang/test/Misc/target-invalid-cpu-note.c

 // Use CHECK-NEXT instead of multiple CHECK-SAME to ensure we will fail if there is anything extra in the output.
 // RUN: not %clang_cc1 -triple armv5--- -target-cpu not-a-cpu -fsyntax-only %s 2>&1 | FileCheck %s --check-prefix ARM
 // ARM: error: unknown target CPU 'not-a-cpu'
 // ARM-NEXT: note: valid target CPU values are: arm8, arm810, strongarm, strongarm110, strongarm1100, strongarm1110, arm7tdmi, arm7tdmi-s, arm710t, arm720t, arm9, arm9tdmi, arm920, arm920t, arm922t, arm940t, ep9312, arm10tdmi, arm1020t, arm9e, arm946e-s, arm966e-s, arm968e-s, arm10e, arm1020e, arm1022e, arm926ej-s, arm1136j-s, arm1136jf-s, mpcore, mpcorenovfp, arm1176jz-s, arm1176jzf-s, arm1156t2-s, arm1156t2f-s, cortex-m0, cortex-m0plus, cortex-m1, sc000, cortex-a5, cortex-a7, cortex-a8, cortex-a9, cortex-a12, cortex-a15, cortex-a17, krait, cortex-r4, cortex-r4f, cortex-r5, cortex-r7, cortex-r8, cortex-r52, sc300, cortex-m3, cortex-m4, cortex-m7, cortex-m23, cortex-m33, cortex-m35p, cortex-m55, cortex-a32, cortex-a35, cortex-a53, cortex-a55, cortex-a57, cortex-a72, cortex-a73, cortex-a75, cortex-a76, cortex-a76ae, cortex-a77, cortex-a78, cortex-a78c, cortex-a710, cortex-x1, neoverse-n1, neoverse-n2, neoverse-v1, cyclone, exynos-m3, exynos-m4, exynos-m5, kryo, iwmmxt, xscale, swift{{$}}
 
 // RUN: not %clang_cc1 -triple arm64--- -target-cpu not-a-cpu -fsyntax-only %s 2>&1 | FileCheck %s --check-prefix AARCH64
 // AARCH64: error: unknown target CPU 'not-a-cpu'
-// AARCH64-NEXT: note: valid target CPU values are: cortex-a34, cortex-a35, cortex-a53, cortex-a55, cortex-a510, cortex-a57, cortex-a65, cortex-a65ae, cortex-a72, cortex-a73, cortex-a75, cortex-a76, cortex-a76ae, cortex-a77, cortex-a78, cortex-a78c, cortex-a710, cortex-r82, cortex-x1, cortex-x2, neoverse-e1, neoverse-n1, neoverse-n2, neoverse-512tvb, neoverse-v1, cyclone, apple-a7, apple-a8, apple-a9, apple-a10, apple-a11, apple-a12, apple-a13, apple-a14, apple-m1, apple-s4, apple-s5, exynos-m3, exynos-m4, exynos-m5, falkor, saphira, kryo, thunderx2t99, thunderx3t110, thunderx, thunderxt88, thunderxt81, thunderxt83, tsv110, a64fx, carmel{{$}}
+// AARCH64-NEXT: note: valid target CPU values are: cortex-a34, cortex-a35, cortex-a53, cortex-a55, cortex-a510, cortex-a57, cortex-a65, cortex-a65ae, cortex-a72, cortex-a73, cortex-a75, cortex-a76, cortex-a76ae, cortex-a77, cortex-a78, cortex-a78c, cortex-a710, cortex-r82, cortex-x1, cortex-x2, neoverse-e1, neoverse-n1, neoverse-n2, neoverse-512tvb, neoverse-v1, cyclone, apple-a7, apple-a8, apple-a9, apple-a10, apple-a11, apple-a12, apple-a13, apple-a14, apple-m1, apple-s4, apple-s5, exynos-m3, exynos-m4, exynos-m5, falkor, saphira, kryo, thunderx2t99, thunderx3t110, thunderx, thunderxt88, thunderxt81, thunderxt83, tsv110, a64fx, carmel, ampere1{{$}}
 
 // RUN: not %clang_cc1 -triple arm64--- -tune-cpu not-a-cpu -fsyntax-only %s 2>&1 | FileCheck %s --check-prefix TUNE_AARCH64
 // TUNE_AARCH64: error: unknown target CPU 'not-a-cpu'
-// TUNE_AARCH64-NEXT: note: valid target CPU values are: cortex-a34, cortex-a35, cortex-a53, cortex-a55, cortex-a510, cortex-a57, cortex-a65, cortex-a65ae, cortex-a72, cortex-a73, cortex-a75, cortex-a76, cortex-a76ae, cortex-a77, cortex-a78, cortex-a78c, cortex-a710, cortex-r82, cortex-x1, cortex-x2, neoverse-e1, neoverse-n1, neoverse-n2, neoverse-512tvb, neoverse-v1, cyclone, apple-a7, apple-a8, apple-a9, apple-a10, apple-a11, apple-a12, apple-a13, apple-a14, apple-m1, apple-s4, apple-s5, exynos-m3, exynos-m4, exynos-m5, falkor, saphira, kryo, thunderx2t99, thunderx3t110, thunderx, thunderxt88, thunderxt81, thunderxt83, tsv110, a64fx, carmel{{$}}
+// TUNE_AARCH64-NEXT: note: valid target CPU values are: cortex-a34, cortex-a35, cortex-a53, cortex-a55, cortex-a510, cortex-a57, cortex-a65, cortex-a65ae, cortex-a72, cortex-a73, cortex-a75, cortex-a76, cortex-a76ae, cortex-a77, cortex-a78, cortex-a78c, cortex-a710, cortex-r82, cortex-x1, cortex-x2, neoverse-e1, neoverse-n1, neoverse-n2, neoverse-512tvb, neoverse-v1, cyclone, apple-a7, apple-a8, apple-a9, apple-a10, apple-a11, apple-a12, apple-a13, apple-a14, apple-m1, apple-s4, apple-s5, exynos-m3, exynos-m4, exynos-m5, falkor, saphira, kryo, thunderx2t99, thunderx3t110, thunderx, thunderxt88, thunderxt81, thunderxt83, tsv110, a64fx, carmel, ampere1{{$}}
 
 // RUN: not %clang_cc1 -triple i386--- -target-cpu not-a-cpu -fsyntax-only %s 2>&1 | FileCheck %s --check-prefix X86
 // X86: error: unknown target CPU 'not-a-cpu'
 // X86-NEXT: note: valid target CPU values are: i386, i486, winchip-c6, winchip2, c3, i586, pentium, pentium-mmx, pentiumpro, i686, pentium2, pentium3, pentium3m, pentium-m, c3-2, yonah, pentium4, pentium4m, prescott, nocona, core2, penryn, bonnell, atom, silvermont, slm, goldmont, goldmont-plus, tremont, nehalem, corei7, westmere, sandybridge, corei7-avx, ivybridge, core-avx-i, haswell, core-avx2, broadwell, skylake, skylake-avx512, skx, cascadelake, cooperlake, cannonlake, icelake-client, rocketlake, icelake-server, tigerlake, sapphirerapids, alderlake, knl, knm, lakemont, k6, k6-2, k6-3, athlon, athlon-tbird, athlon-xp, athlon-mp, athlon-4, k8, athlon64, athlon-fx, opteron, k8-sse3, athlon64-sse3, opteron-sse3, amdfam10, barcelona, btver1, btver2, bdver1, bdver2, bdver3, bdver4, znver1, znver2, znver3, x86-64, x86-64-v2, x86-64-v3, x86-64-v4, geode{{$}}
 
 // RUN: not %clang_cc1 -triple x86_64--- -target-cpu not-a-cpu -fsyntax-only %s 2>&1 | FileCheck %s --check-prefix X86_64
 // X86_64: error: unknown target CPU 'not-a-cpu'
 // X86_64-NEXT: note: valid target CPU values are: nocona, core2, penryn, bonnell, atom, silvermont, slm, goldmont, goldmont-plus, tremont, nehalem, corei7, westmere, sandybridge, corei7-avx, ivybridge, core-avx-i, haswell, core-avx2, broadwell, skylake, skylake-avx512, skx, cascadelake, cooperlake, cannonlake, icelake-client, rocketlake, icelake-server, tigerlake, sapphirerapids, alderlake, knl, knm, k8, athlon64, athlon-fx, opteron, k8-sse3, athlon64-sse3, opteron-sse3, amdfam10, barcelona, btver1, btver2, bdver1, bdver2, bdver3, bdver4, znver1, znver2, znver3, x86-64, x86-64-v2, x86-64-v3, x86-64-v4{{$}}

llvm/include/llvm/Support/AArch64TargetParser.def

 AARCH64_CPU_NAME("tsv110", ARMV8_2A, FK_CRYPTO_NEON_FP_ARMV8, false,
                  (AArch64::AEK_DOTPROD |
                   AArch64::AEK_FP16 | AArch64::AEK_FP16FML |
                   AArch64::AEK_PROFILE))
 AARCH64_CPU_NAME("a64fx", ARMV8_2A, FK_CRYPTO_NEON_FP_ARMV8, false,
                  (AArch64::AEK_FP16 | AArch64::AEK_SVE))
 AARCH64_CPU_NAME("carmel", ARMV8_2A, FK_CRYPTO_NEON_FP_ARMV8, false,
                  AArch64::AEK_FP16)
+AARCH64_CPU_NAME("ampere1", ARMV8_6A, FK_CRYPTO_NEON_FP_ARMV8, false,
+                 (AArch64::AEK_BF16 | AArch64::AEK_DOTPROD | AArch64::AEK_FP16 |
+                  AArch64::AEK_I8MM | AArch64::AEK_MTE | AArch64::AEK_RAS |
+                  AArch64::AEK_RCPC | AArch64::AEK_SB | AArch64::AEK_SSBS))
 // Invalid CPU
 AARCH64_CPU_NAME("invalid", INVALID, FK_INVALID, true, AArch64::AEK_INVALID)
 #undef AARCH64_CPU_NAME

llvm/lib/Target/AArch64/AArch64.td

 //===----------------------------------------------------------------------===//
 // Instruction Descriptions
 //===----------------------------------------------------------------------===//
 
 include "AArch64Schedule.td"
 include "AArch64InstrInfo.td"
 include "AArch64SchedPredicates.td"
 include "AArch64SchedPredExynos.td"
+include "AArch64SchedPredAmpere.td"
 include "AArch64Combine.td"
 
 def AArch64InstrInfo : InstrInfo;
 
 //===----------------------------------------------------------------------===//
 // Named operands for MRS/MSR/TLBI/...
 //===----------------------------------------------------------------------===//
 

 include "AArch64SchedExynosM3.td"
 include "AArch64SchedExynosM4.td"
 include "AArch64SchedExynosM5.td"
 include "AArch64SchedThunderX.td"
 include "AArch64SchedThunderX2T99.td"
 include "AArch64SchedA64FX.td"
 include "AArch64SchedThunderX3T110.td"
 include "AArch64SchedTSV110.td"
+include "AArch64SchedAmpere1.td"
 
 def TuneA35     : SubtargetFeature<"a35", "ARMProcFamily", "CortexA35",
                                 "Cortex-A35 ARM processors">;
 
 def TuneA53     : SubtargetFeature<"a53", "ARMProcFamily", "CortexA53",
                                    "Cortex-A53 ARM processors", [
                                    FeatureFuseAES,
                                    FeatureBalanceFPOps,
def TuneThunderXT83 : SubtargetFeature<"thunderxt83", "ARMProcFamily",
                                        FeaturePredictableSelectIsExpensive]>;
 
 def TuneTSV110 : SubtargetFeature<"tsv110", "ARMProcFamily", "TSV110",
                                   "HiSilicon TS-V110 processors", [
                                   FeatureCustomCheapAsMoveHandling,
                                   FeatureFuseAES,
                                   FeaturePostRAScheduler]>;
 
+def TuneAmpere1 : SubtargetFeature<"ampere1", "ARMProcFamily", "Ampere1",
+           "Ampere Computing Ampere-1 processors", [
+           FeaturePostRAScheduler,
+           FeatureFuseAES,
+           FeatureLSLFast,
+           FeatureAggressiveFMA,
+           FeatureArithmeticBccFusion,
+           FeatureCmpBccFusion,
+           FeatureFuseAddress,
+           FeatureFuseLiterals]>;
 
 def ProcessorFeatures {
   list<SubtargetFeature> A53  = [HasV8_0aOps, FeatureCRC, FeatureCrypto,
                                  FeatureFPARMv8, FeatureNEON, FeaturePerfMon];
   list<SubtargetFeature> A55  = [HasV8_2aOps, FeatureCrypto, FeatureFPARMv8,
                                  FeatureNEON, FeatureFullFP16, FeatureDotProd,
                                  FeatureRCPC, FeaturePerfMon];
   list<SubtargetFeature> A510 = [HasV9_0aOps, FeatureNEON, FeaturePerfMon,
def ProcessorFeatures {
   list<SubtargetFeature> ThunderX2T99  = [HasV8_1aOps, FeatureCRC, FeatureCrypto,
                                           FeatureFPARMv8, FeatureNEON, FeatureLSE];
   list<SubtargetFeature> ThunderX3T110 = [HasV8_3aOps, FeatureCRC, FeatureCrypto,
                                           FeatureFPARMv8, FeatureNEON, FeatureLSE,
                                           FeaturePAuth, FeaturePerfMon];
   list<SubtargetFeature> TSV110 = [HasV8_2aOps, FeatureCrypto, FeatureFPARMv8,
                                    FeatureNEON, FeaturePerfMon, FeatureSPE,
                                    FeatureFullFP16, FeatureFP16FML, FeatureDotProd];
+  list<SubtargetFeature> Ampere1 = [HasV8_6aOps, FeatureNEON, FeaturePerfMon];
 
   // ETE and TRBE are future architecture extensions. We temporarily enable them
   // by default for users targeting generic AArch64. The extensions do not
   // affect code generated by the compiler and can be used only by explicitly
   // mentioning the new system register names in assembly.
   list<SubtargetFeature> Generic = [FeatureFPARMv8, FeatureNEON, FeaturePerfMon, FeatureETE];
 }
 

 // Fujitsu A64FX
 def : ProcessorModel<"a64fx", A64FXModel, ProcessorFeatures.A64FX,
                      [TuneA64FX]>;
 
 // Nvidia Carmel
 def : ProcessorModel<"carmel", NoSchedModel, ProcessorFeatures.Carmel,
                      [TuneCarmel]>;
 
+// Ampere Computing
+def : ProcessorModel<"ampere1", Ampere1Model, ProcessorFeatures.Ampere1,
+         [TuneAmpere1]>;
+
 //===----------------------------------------------------------------------===//
 // Assembly parser
 //===----------------------------------------------------------------------===//
 
 def GenericAsmParserVariant : AsmParserVariant {
   int Variant = 0;
   string Name = "generic";
   string BreakCharacters = ".";

llvm/lib/Target/AArch64/AArch64SchedAmpere1.td

+//=- AArch64SchedAmpere1.td - Ampere-1 scheduling def -----*- tablegen -*-=//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// This file defines the machine model for the Ampere Computing Ampere-1 to
+// support instruction scheduling and other instruction cost heuristics.
+//
+//===----------------------------------------------------------------------===//
+
+//===----------------------------------------------------------------------===//
+
+// The Ampere-1 core is anout-of-order micro-architecture.  The front
+// end has branch prediction, with a 10-cycle recovery time from a
+// mispredicted branch.  Instructions coming out of the front end are
+// decoded into internal micro-ops (uops).
+
+def Ampere1Model : SchedMachineModel {
+  let IssueWidth            =   4;  // 4-way decode and dispatch
+  let MicroOpBufferSize     = 174;  // micro-op re-order buffer size
+  let LoadLatency           =   4;  // Optimistic load latency
+  let MispredictPenalty     =  10;  // Branch mispredict penalty
+  let LoopMicroOpBufferSize =  32;  // Instruction queue size
+  let CompleteModel = 1;
+
+  list<Predicate> UnsupportedFeatures = !listconcat(SVEUnsupported.F,
+                SMEUnsupported.F);
+}
+
+let SchedModel = Ampere1Model in {
+
+//===----------------------------------------------------------------------===//
+// Define each kind of processor resource and number available on Ampere-1.
+// Ampere-1 has 12 pipelines that 8 independent scheduler (4 integer, 2 FP,
+// and 2 memory) issue into.  The integer and FP schedulers can each issue
+// one uop per cycle, while the memory schedulers can each issue one load
+// and one store address calculation per cycle.
+
+def Ampere1UnitA  : ProcResource<2>;  // integer single-cycle, branch, and flags r/w
+def Ampere1UnitB  : ProcResource<2>;  // integer single-cycle, and complex shifts
+def Ampere1UnitBS : ProcResource<1>;  // integer multi-cycle
+def Ampere1UnitL  : ProcResource<2>;  // load
+def Ampere1UnitS  : ProcResource<2>;  // store address calculation
+def Ampere1UnitX  : ProcResource<1>;  // FP and vector operations, and flag write
+def Ampere1UnitY  : ProcResource<1>;  // FP and vector operations, and crypto
+def Ampere1UnitZ  : ProcResource<1>;  // FP store data and FP-to-integer moves
+
+def Ampere1UnitAB : ProcResGroup<[Ampere1UnitA, Ampere1UnitB]>;
+def Ampere1UnitXY : ProcResGroup<[Ampere1UnitX, Ampere1UnitY]>;
+
+//===----------------------------------------------------------------------===//
+// Define customized scheduler read/write types specific to the Ampere-1.
+
+def Ampere1Write_1cyc_1A : SchedWriteRes<[Ampere1UnitA]> {
+  let Latency = 1;
+  let NumMicroOps = 1;
+}
+
+def Ampere1Write_1cyc_2A : SchedWriteRes<[Ampere1UnitA, Ampere1UnitA]> {
+  let Latency = 1;
+  let NumMicroOps = 2;
+}
+
+def Ampere1Write_1cyc_1B : SchedWriteRes<[Ampere1UnitB]> {
+  let Latency = 1;
+  let NumMicroOps = 1;
+}
+
+def Ampere1Write_1cyc_1AB : SchedWriteRes<[Ampere1UnitAB]> {
+  let Latency = 1;
+  let NumMicroOps = 1;
+}
+
+def Ampere1Write_1cyc_1L : SchedWriteRes<[Ampere1UnitL]> {
+  let Latency = 1;
+  let NumMicroOps = 1;
+}
+
+def Ampere1Write_1cyc_1S : SchedWriteRes<[Ampere1UnitS]> {
+  let Latency = 1;
+  let NumMicroOps = 1;
+}
+
+def Ampere1Write_1cyc_2S : SchedWriteRes<[Ampere1UnitS, Ampere1UnitS]> {
+  let Latency = 1;
+  let NumMicroOps = 2;
+}
+
+def Ampere1Write_2cyc_1Y : SchedWriteRes<[Ampere1UnitY]> {
+  let Latency = 2;
+  let NumMicroOps = 1;
+}
+
+def Ampere1Write_2cyc_2AB : SchedWriteRes<[Ampere1UnitAB, Ampere1UnitAB]> {
+  let Latency = 2;
+  let NumMicroOps = 2;
+}
+
+def Ampere1Write_2cyc_1B_1AB : SchedWriteRes<[Ampere1UnitB, Ampere1UnitAB]> {
+  let Latency = 2;
+  let NumMicroOps = 2;
+}
+
+def Ampere1Write_2cyc_1B_1A : SchedWriteRes<[Ampere1UnitB, Ampere1UnitA]> {
+  let Latency = 2;
+  let NumMicroOps = 2;
+}
+
+def Ampere1Write_2cyc_1AB_1A : SchedWriteRes<[Ampere1UnitAB, Ampere1UnitA]> {
+  let Latency = 2;
+  let NumMicroOps = 2;
+}
+
+def Ampere1Write_2cyc_1AB_1L : SchedWriteRes<[Ampere1UnitAB, Ampere1UnitL]> {
+  let Latency = 2;
+  let NumMicroOps = 2;
+}
+
+def Ampere1Write_2cyc_1AB_2S : SchedWriteRes<[Ampere1UnitAB, Ampere1UnitS,
+                   Ampere1UnitS]> {
+  let Latency = 2;
+  let NumMicroOps = 3;
+}
+
+def Ampere1Write_2cyc_1AB_1S_1Z : SchedWriteRes<[Ampere1UnitAB, Ampere1UnitS,
+                Ampere1UnitZ]> {
+  let Latency = 2;
+  let NumMicroOps = 3;
+}
+
+def Ampere1Write_2cyc_1B_1S : SchedWriteRes<[Ampere1UnitB, Ampere1UnitS]> {
+  let Latency = 2;
+  let NumMicroOps = 2;
+}
+
+def Ampere1Write_2cyc_1XY : SchedWriteRes<[Ampere1UnitXY]> {
+  let Latency = 2;
+  let NumMicroOps = 1;
+}
+
+def Ampere1Write_2cyc_1S_1Z : SchedWriteRes<[Ampere1UnitS, Ampere1UnitZ]> {
+  let Latency = 2;
+  let NumMicroOps = 2;
+}
+
+def Ampere1Write_3cyc_1BS : SchedWriteRes<[Ampere1UnitBS]> {
+  let Latency = 3;
+  let NumMicroOps = 1;
+}
+
+def Ampere1Write_3cyc_1XY : SchedWriteRes<[Ampere1UnitXY]> {
+  let Latency = 3;
+  let NumMicroOps = 1;
+}
+
+def Ampere1Write_3cyc_1B_1S_1AB : SchedWriteRes<[Ampere1UnitB, Ampere1UnitS,
+                     Ampere1UnitAB]> {
+  let Latency = 2;
+  let NumMicroOps = 3;
+}
+
+def Ampere1Write_3cyc_1S_2Z : SchedWriteRes<[Ampere1UnitS, Ampere1UnitZ, Ampere1UnitZ]> {
+  let Latency = 2;
+  let NumMicroOps = 3;
+}
+
+def Ampere1Write_3cyc_2S_2Z : SchedWriteRes<[Ampere1UnitS, Ampere1UnitS,
+                         Ampere1UnitZ, Ampere1UnitZ]> {
+  let Latency = 2;
+  let NumMicroOps = 4;
+}
+
+def Ampere1Write_4cyc_1BS : SchedWriteRes<[Ampere1UnitBS]> {
+  let Latency = 4;
+  let NumMicroOps = 1;
+}
+
+def Ampere1Write_4cyc_1L : SchedWriteRes<[Ampere1UnitL]> {
+  let Latency = 4;
+  let NumMicroOps = 1;
+}
+
+def Ampere1Write_4cyc_1X : SchedWriteRes<[Ampere1UnitX]> {
+  let Latency = 4;
+  let NumMicroOps = 1;
+}
+
+def Ampere1Write_4cyc_1Y : SchedWriteRes<[Ampere1UnitY]> {
+  let Latency = 4;
+  let NumMicroOps = 1;
+}
+
+def Ampere1Write_4cyc_1Z : SchedWriteRes<[Ampere1UnitZ]> {
+  let Latency = 4;
+  let NumMicroOps = 1;
+}
+
+def Ampere1Write_4cyc_2L : SchedWriteRes<[Ampere1UnitL, Ampere1UnitL]> {
+  let Latency = 4;
+  let NumMicroOps = 2;
+}
+
+def Ampere1Write_4cyc_1XY : SchedWriteRes<[Ampere1UnitXY]> {
+  let Latency = 4;
+  let NumMicroOps = 1;
+}
+
+def Ampere1Write_4cyc_2XY : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY]> {
+  let Latency = 4;
+  let NumMicroOps = 2;
+}
+
+def Ampere1Write_4cyc_1XY_1S_1Z : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitS, Ampere1UnitZ]> {
+  let Latency = 4;
+  let NumMicroOps = 3;
+}
+
+def Ampere1Write_4cyc_3S_3Z : SchedWriteRes<[Ampere1UnitS, Ampere1UnitS, Ampere1UnitS,
+                         Ampere1UnitZ, Ampere1UnitZ, Ampere1UnitZ]> {
+  let Latency = 4;
+  let NumMicroOps = 6;
+}
+
+def Ampere1Write_5cyc_1AB_1L : SchedWriteRes<[Ampere1UnitAB, Ampere1UnitL]> {
+  let Latency = 5;
+  let NumMicroOps = 2;
+}
+
+def Ampere1Write_5cyc_1BS : SchedWriteRes<[Ampere1UnitBS]> {
+  let Latency = 5;
+  let NumMicroOps = 1;
+}
+
+def Ampere1Write_5cyc_1X : SchedWriteRes<[Ampere1UnitX]> {
+  let Latency = 5;
+  let NumMicroOps = 1;
+}
+
+def Ampere1Write_5cyc_1L : SchedWriteRes<[Ampere1UnitL]> {
+  let Latency = 5;
+  let NumMicroOps = 1;
+}
+
+def Ampere1Write_5cyc_2L : SchedWriteRes<[Ampere1UnitL, Ampere1UnitL]> {
+  let Latency = 5;
+  let NumMicroOps = 2;
+}
+
+def Ampere1Write_5cyc_1L_1BS : SchedWriteRes<[Ampere1UnitL, Ampere1UnitBS]> {
+  let Latency = 5;
+  let NumMicroOps = 2;
+}
+
+def Ampere1Write_5cyc_1XY : SchedWriteRes<[Ampere1UnitXY]> {
+  let Latency = 5;
+  let NumMicroOps = 1;
+}
+
+def Ampere1Write_5cyc_2XY : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY]> {
+  let Latency = 5;
+  let NumMicroOps = 2;
+}
+
+def Ampere1Write_5cyc_4S_4Z : SchedWriteRes<[Ampere1UnitS, Ampere1UnitS,
+               Ampere1UnitS, Ampere1UnitS,
+                         Ampere1UnitZ, Ampere1UnitZ,
+               Ampere1UnitZ, Ampere1UnitZ]> {
+  let Latency = 5;
+  let NumMicroOps = 8;
+}
+
+def Ampere1Write_5cyc_2XY_2S_2Z : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY,
+               Ampere1UnitS, Ampere1UnitS,
+                 Ampere1UnitZ, Ampere1UnitZ]> {
+  let Latency = 5;
+  let NumMicroOps = 6;
+}
+
+def Ampere1Write_6cyc_2XY_2S_2Z : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY,
+               Ampere1UnitS, Ampere1UnitS,
+                 Ampere1UnitZ, Ampere1UnitZ]> {
+  let Latency = 6;
+  let NumMicroOps = 6;
+}
+
+def Ampere1Write_6cyc_3XY_3S_3Z : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY, Ampere1UnitXY,
+               Ampere1UnitS, Ampere1UnitS, Ampere1UnitS,
+                 Ampere1UnitZ, Ampere1UnitZ, Ampere1UnitZ]> {
+  let Latency = 6;
+  let NumMicroOps = 9;
+}
+
+def Ampere1Write_6cyc_1AB_1L : SchedWriteRes<[Ampere1UnitAB, Ampere1UnitL]> {
+  let Latency = 6;
+  let NumMicroOps = 2;
+}
+
+def Ampere1Write_6cyc_1XY : SchedWriteRes<[Ampere1UnitXY]> {
+  let Latency = 6;
+  let NumMicroOps = 1;
+}
+
+def Ampere1Write_6cyc_2XY : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY]> {
+  let Latency = 6;
+  let NumMicroOps = 2;
+}
+
+def Ampere1Write_6cyc_3XY : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY, Ampere1UnitXY]> {
+  let Latency = 6;
+  let NumMicroOps = 3;
+}
+
+def Ampere1Write_6cyc_3L : SchedWriteRes<[Ampere1UnitL, Ampere1UnitL, Ampere1UnitL]> {
+  let Latency = 6;
+  let NumMicroOps = 3;
+}
+
+def Ampere1Write_6cyc_4L : SchedWriteRes<[Ampere1UnitL, Ampere1UnitL,
+                            Ampere1UnitL, Ampere1UnitL]> {
+  let Latency = 6;
+  let NumMicroOps = 4;
+}
+
+def Ampere1Write_6cyc_1XY_1Z : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitZ]> {
+  let Latency = 6;
+  let NumMicroOps = 2;
+}
+
+def Ampere1Write_7cyc_1BS : SchedWriteRes<[Ampere1UnitBS]> {
+  let Latency = 7;
+  let NumMicroOps = 1;
+}
+
+def Ampere1Write_7cyc_1BS_1XY : SchedWriteRes<[Ampere1UnitBS, Ampere1UnitXY]> {
+  let Latency = 7;
+  let NumMicroOps = 2;
+}
+
+def Ampere1Write_7cyc_1L_1XY : SchedWriteRes<[Ampere1UnitL, Ampere1UnitXY]> {
+  let Latency = 7;
+  let NumMicroOps = 2;
+}
+
+def Ampere1Write_7cyc_2L_2XY : SchedWriteRes<[Ampere1UnitL, Ampere1UnitL,
+                Ampere1UnitXY, Ampere1UnitXY]> {
+  let Latency = 7;
+  let NumMicroOps = 4;
+}
+
+def Ampere1Write_7cyc_2XY : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY]> {
+  let Latency = 7;
+  let NumMicroOps = 2;
+}
+
+def Ampere1Write_7cyc_4XY_4S_4Z : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY,
+               Ampere1UnitXY, Ampere1UnitXY,
+               Ampere1UnitS, Ampere1UnitS,
+               Ampere1UnitS, Ampere1UnitS,
+             Ampere1UnitZ, Ampere1UnitZ,
+                 Ampere1UnitZ, Ampere1UnitZ]> {
+  let Latency = 7;
+  let NumMicroOps = 12;
+}
+
+def Ampere1Write_8cyc_1BS_1A : SchedWriteRes<[Ampere1UnitBS, Ampere1UnitA]> {
+  let Latency = 8;
+  let NumMicroOps = 2;
+}
+
+def Ampere1Write_8cyc_1BS_2A : SchedWriteRes<[Ampere1UnitBS, Ampere1UnitA,
+                             Ampere1UnitA]> {
+  let Latency = 8;
+  let NumMicroOps = 3;
+}
+
+def Ampere1Write_8cyc_2XY : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY]> {
+  let Latency = 8;
+  let NumMicroOps = 2;
+}
+
+def Ampere1Write_8cyc_4XY : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY,
+                     Ampere1UnitXY, Ampere1UnitXY]> {
+  let Latency = 8;
+  let NumMicroOps = 4;
+}
+
+def Ampere1Write_8cyc_3L_3XY : SchedWriteRes<[Ampere1UnitL, Ampere1UnitL, Ampere1UnitL,
+                Ampere1UnitXY, Ampere1UnitXY, Ampere1UnitXY]> {
+  let Latency = 8;
+  let NumMicroOps = 6;
+}
+
+def Ampere1Write_8cyc_4L_4XY : SchedWriteRes<[Ampere1UnitL, Ampere1UnitL,
+                          Ampere1UnitL, Ampere1UnitL,
+                Ampere1UnitXY, Ampere1UnitXY,
+                Ampere1UnitXY, Ampere1UnitXY]> {
+  let Latency = 8;
+  let NumMicroOps = 8;
+}
+
+def Ampere1Write_9cyc_3L_3XY : SchedWriteRes<[Ampere1UnitL, Ampere1UnitL, Ampere1UnitL,
+                Ampere1UnitXY, Ampere1UnitXY, Ampere1UnitXY]> {
+  let Latency = 9;
+  let NumMicroOps = 6;
+}
+
+def Ampere1Write_9cyc_4L_4XY : SchedWriteRes<[Ampere1UnitL, Ampere1UnitL,
+                          Ampere1UnitL, Ampere1UnitL,
+                Ampere1UnitXY, Ampere1UnitXY,
+                Ampere1UnitXY, Ampere1UnitXY]> {
+  let Latency = 9;
+  let NumMicroOps = 8;
+}
+
+def Ampere1Write_9cyc_3XY : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY, Ampere1UnitXY]> {
+  let Latency = 9;
+  let NumMicroOps = 3;
+}
+
+def Ampere1Write_9cyc_2L_3XY : SchedWriteRes<[Ampere1UnitL, Ampere1UnitL,
+                          Ampere1UnitXY, Ampere1UnitXY, Ampere1UnitXY]> {
+  let Latency = 9;
+  let NumMicroOps = 5;
+}
+
+def Ampere1Write_9cyc_6XY_4S_4Z : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY,
+               Ampere1UnitXY, Ampere1UnitXY,
+             Ampere1UnitXY, Ampere1UnitXY,
+               Ampere1UnitS, Ampere1UnitS,
+               Ampere1UnitS, Ampere1UnitS,
+             Ampere1UnitZ, Ampere1UnitZ,
+                 Ampere1UnitZ, Ampere1UnitZ]> {
+  let Latency = 9;
+  let NumMicroOps = 14;
+}
+
+def Ampere1Write_9cyc_8XY_4S_4Z : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY,
+               Ampere1UnitXY, Ampere1UnitXY,
+               Ampere1UnitXY, Ampere1UnitXY,
+             Ampere1UnitXY, Ampere1UnitXY,
+               Ampere1UnitS, Ampere1UnitS,
+               Ampere1UnitS, Ampere1UnitS,
+             Ampere1UnitZ, Ampere1UnitZ,
+                 Ampere1UnitZ, Ampere1UnitZ]> {
+  let Latency = 9;
+  let NumMicroOps = 16;
+}
+
+def Ampere1Write_10cyc_2XY : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY]> {
+  let Latency = 10;
+  let NumMicroOps = 2;
+}
+
+def Ampere1Write_10cyc_1XY_1Z : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitZ]> {
+  let Latency = 10;
+  let NumMicroOps = 2;
+}
+
+def Ampere1Write_10cyc_1X_1Z : SchedWriteRes<[Ampere1UnitX, Ampere1UnitZ]> {
+  let Latency = 10;
+  let NumMicroOps = 2;
+}
+
+def Ampere1Write_10cyc_3L_3XY : SchedWriteRes<[Ampere1UnitL, Ampere1UnitL, Ampere1UnitL,
+                 Ampere1UnitXY, Ampere1UnitXY, Ampere1UnitXY]> {
+  let Latency = 10;
+  let NumMicroOps = 6;
+}
+
+def Ampere1Write_10cyc_1A_1BS_1X : SchedWriteRes<[Ampere1UnitA, Ampere1UnitBS, Ampere1UnitX]> {
+  let Latency = 10;
+  let NumMicroOps = 3;
+}
+
+def Ampere1Write_10cyc_1A_1BS_1XY : SchedWriteRes<[Ampere1UnitA, Ampere1UnitBS, Ampere1UnitXY]> {
+  let Latency = 10;
+  let NumMicroOps = 3;
+}
+
+def Ampere1Write_11cyc_1BS_1L : SchedWriteRes<[Ampere1UnitBS, Ampere1UnitL]> {
+  let Latency = 11;
+  let NumMicroOps = 2;
+}
+
+def Ampere1Write_11cyc_1A_1BS_1X : SchedWriteRes<[Ampere1UnitA, Ampere1UnitBS, Ampere1UnitX]> {
+  let Latency = 11;
+  let NumMicroOps = 3;
+}
+
+def Ampere1Write_11cyc_1A_1BS_1XY : SchedWriteRes<[Ampere1UnitA, Ampere1UnitBS, Ampere1UnitXY]> {
+  let Latency = 11;
+  let NumMicroOps = 3;
+}
+
+def Ampere1Write_11cyc_4L_8XY : SchedWriteRes<[Ampere1UnitL, Ampere1UnitL,
+                           Ampere1UnitL, Ampere1UnitL,
+                 Ampere1UnitXY, Ampere1UnitXY,
+                 Ampere1UnitXY, Ampere1UnitXY,
+                 Ampere1UnitXY, Ampere1UnitXY,
+                 Ampere1UnitXY, Ampere1UnitXY]> {
+  let Latency = 11;
+  let NumMicroOps = 12;
+}
+
+def Ampere1Write_12cyc_4L_8XY : SchedWriteRes<[Ampere1UnitL, Ampere1UnitL,
+                           Ampere1UnitL, Ampere1UnitL,
+                 Ampere1UnitXY, Ampere1UnitXY,
+                 Ampere1UnitXY, Ampere1UnitXY,
+                 Ampere1UnitXY, Ampere1UnitXY,
+                 Ampere1UnitXY, Ampere1UnitXY]> {
+  let Latency = 12;
+  let NumMicroOps = 12;
+}
+
+def Ampere1Write_12cyc_3XY : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY, Ampere1UnitXY]> {
+  let Latency = 12;
+  let NumMicroOps = 3;
+}
+
+def Ampere1Write_12cyc_4XY : SchedWriteRes<[Ampere1UnitXY, Ampere1UnitXY,
+                      Ampere1UnitXY, Ampere1UnitXY]> {
+  let Latency = 12;
+  let NumMicroOps = 4;
+}
+
+def Ampere1Write_18cyc_1BS : SchedWriteRes<[Ampere1UnitBS]> {
+  let Latency = 18;
+  let NumMicroOps = 1;
+}
+
+def Ampere1Write_19cyc_1XY : SchedWriteRes<[Ampere1UnitXY]> {
+  let Latency = 19;
+  let NumMicroOps = 1;
+}
+
+def Ampere1Write_25cyc_1XY : SchedWriteRes<[Ampere1UnitXY]> {
+  let Latency = 25;
+  let NumMicroOps = 1;
+}
+
+def Ampere1Write_32cyc_1XY : SchedWriteRes<[Ampere1UnitXY]> {
+  let Latency = 32;
+  let NumMicroOps = 1;
+}
+
+def Ampere1Write_34cyc_1BS : SchedWriteRes<[Ampere1UnitBS]> {
+  let Latency = 34;
+  let NumMicroOps = 1;
+}
+
+def Ampere1Write_34cyc_1XY : SchedWriteRes<[Ampere1UnitXY]> {
+  let Latency = 34;
+  let NumMicroOps = 1;
+}
+
+def Ampere1Write_39cyc_1XY : SchedWriteRes<[Ampere1UnitXY]> {
+  let Latency = 39;
+  let NumMicroOps = 1;
+}
+
+def Ampere1Write_62cyc_1XY : SchedWriteRes<[Ampere1UnitXY]> {
+  let Latency = 62;
+  let NumMicroOps = 1;
+}
+
+// For basic arithmetic, we have more flexibility for short shifts (LSL shift <= 4),
+// which are a single uop, and for extended registers, which have full flexibility
+// across Unit A or B for both uops.
+def Ampere1Write_Arith : SchedWriteVariant<[
+        SchedVar<RegExtendedPred, [Ampere1Write_2cyc_2AB]>,
+        SchedVar<AmpereCheapLSL,  [Ampere1Write_1cyc_1AB]>,
+        SchedVar<NoSchedPred,     [Ampere1Write_2cyc_1B_1AB]>]>;
+
+def Ampere1Write_ArithFlagsetting : SchedWriteVariant<[
+        SchedVar<RegExtendedPred, [Ampere1Write_2cyc_1AB_1A]>,
+        SchedVar<AmpereCheapLSL,  [Ampere1Write_1cyc_1A]>,
+        SchedVar<NoSchedPred,     [Ampere1Write_2cyc_1B_1A]>]>;
+
+//===----------------------------------------------------------------------===//
+// Map the target-defined scheduler read/write resources and latencies for Ampere-1.
+// This provides a coarse model, which is then specialised below.
+
+def : WriteRes<WriteImm,   [Ampere1UnitAB]>;  // MOVN, MOVZ
+def : WriteRes<WriteI,     [Ampere1UnitAB]>;  // ALU
+def : WriteRes<WriteISReg, [Ampere1UnitB, Ampere1UnitA]> {
+  let Latency = 2;
+  let NumMicroOps = 2;
+}  // ALU of Shifted-Reg
+def : WriteRes<WriteIEReg, [Ampere1UnitAB, Ampere1UnitA]> {
+  let Latency = 2;
+  let NumMicroOps = 2;
+}  // ALU of Extended-Reg
+def : WriteRes<WriteExtr,  [Ampere1UnitB]>;  // EXTR shifts a reg pair
+def : WriteRes<WriteIS,    [Ampere1UnitB]>;  // Shift/Scale
+def : WriteRes<WriteID32,  [Ampere1UnitBS]> {
+  let Latency = 18;
+}  // 32-bit Divide
+def : WriteRes<WriteID64,  [Ampere1UnitBS]> {
+  let Latency = 34;
+}  // 64-bit Divide
+def : WriteRes<WriteIM32,  [Ampere1UnitBS]> {
+  let Latency = 3;
+}  // 32-bit Multiply
+def : WriteRes<WriteIM64,  [Ampere1UnitBS]> {
+  let Latency = 3;
+}  // 32-bit Multiply
+def : WriteRes<WriteBr,    [Ampere1UnitA]>;
+def : WriteRes<WriteBrReg, [Ampere1UnitA, Ampere1UnitA]>;
+def : WriteRes<WriteLD,    [Ampere1UnitL]> {
+  let Latency = 4;
+}  // Load from base addr plus immediate offset
+def : WriteRes<WriteST,    [Ampere1UnitS]> {
+  let Latency = 1;
+}  // Store to base addr plus immediate offset
+def : WriteRes<WriteSTP,   [Ampere1UnitS, Ampere1UnitS]> {
+  let Latency = 1;
+  let NumMicroOps = 2;
+}  // Store a register pair.
+def : WriteRes<WriteAdr,   [Ampere1UnitAB]>;
+def : WriteRes<WriteLDIdx, [Ampere1UnitAB, Ampere1UnitS]> {
+  let Latency = 5;
+  let NumMicroOps = 2;
+}  // Load from a register index (maybe scaled).
+def : WriteRes<WriteSTIdx, [Ampere1UnitS, Ampere1UnitS]> {
+  let Latency = 1;
+  let NumMicroOps = 2;
+}  // Store to a register index (maybe scaled).
+def : WriteRes<WriteF,  [Ampere1UnitXY]> {
+  let Latency = 2;
+}  // General floating-point ops.
+def : WriteRes<WriteFCmp,  [Ampere1UnitX]> {
+  let Latency = 5;
+}  // Floating-point compare.
+def : WriteRes<WriteFCvt,  [Ampere1UnitXY]> {
+  let Latency = 6;
+}  // Float conversion.
+def : WriteRes<WriteFCopy, [Ampere1UnitXY]> {
+}  // Float-int register copy.
+def : WriteRes<WriteFImm,  [Ampere1UnitXY]> {
+  let Latency = 2;
+}  // Float-int register copy.
+def : WriteRes<WriteFMul,  [Ampere1UnitXY]> {
+  let Latency = 5;
+}  // Floating-point multiply.
+def : WriteRes<WriteFDiv,  [Ampere1UnitXY]> {
+  let Latency = 34;
+}  // Floating-point division.
+def : WriteRes<WriteVd,    [Ampere1UnitXY]> {
+  let Latency = 3;
+}  // 64bit Vector D ops.
+def : WriteRes<WriteVq,    [Ampere1UnitXY]> {
+  let Latency = 3;
+}  // 128bit Vector Q ops.
+def : WriteRes<WriteVLD,   [Ampere1UnitL, Ampere1UnitL]> {
+  let Latency = 5;
+}  // Vector loads.
+def : WriteRes<WriteVST,   [Ampere1UnitS, Ampere1UnitZ]> {
+  let Latency = 2;
+}  // Vector stores.
+
+def : WriteRes<WriteAtomic,  []> { let Unsupported = 1; }
+
+def : WriteRes<WriteSys,     []> { let Latency = 1; }
+def : WriteRes<WriteBarrier, []> { let Latency = 1; }
+def : WriteRes<WriteHint,    []> { let Latency = 1; }
+
+def : WriteRes<WriteLDHi,    []> {
+  let Latency = 4;
+}  // The second register of a load-pair: LDP,LDPSW,LDNP,LDXP,LDAXP
+
+// Forwarding logic.
+def : ReadAdvance<ReadI,       0>;
+def : ReadAdvance<ReadISReg,   0>;
+def : ReadAdvance<ReadIEReg,   0>;
+def : ReadAdvance<ReadIM,      0>;
+def : ReadAdvance<ReadIMA,     1, [WriteIM32, WriteIM64]>;
+def : ReadAdvance<ReadID,      0>;
+def : ReadAdvance<ReadExtrHi,  0>;
+def : ReadAdvance<ReadST,      0>;
+def : ReadAdvance<ReadAdrBase, 0>;
+def : ReadAdvance<ReadVLD,     0>;
+
+//===----------------------------------------------------------------------===//
+// Specialising the scheduling model further for Ampere-1.
+
+def : InstRW<[Ampere1Write_1cyc_1AB], (instrs COPY)>;
+
+// Branch instructions
+def : InstRW<[Ampere1Write_1cyc_1A], (instrs Bcc, BL, RET)>;
+def : InstRW<[Ampere1Write_1cyc_1A],
+  (instrs CBZW, CBZX, CBNZW, CBNZX, TBZW, TBZX, TBNZW, TBNZX)>;
+def : InstRW<[Ampere1Write_1cyc_2A], (instrs BLR)>;
+
+// Cryptography instructions
+// -- AES encryption/decryption
+def : InstRW<[Ampere1Write_2cyc_1XY], (instregex "^AES[DE]")>;
+def : InstRW<[Ampere1Write_2cyc_1XY], (instregex "^AESI?MC")>;
+// -- Polynomial multiplication
+def : InstRW<[Ampere1Write_2cyc_1XY], (instregex "^PMUL", "^PMULL")>;
+// -- SHA-256 hash
+def : InstRW<[Ampere1Write_4cyc_1X], (instregex "^SHA256(H|H2)")>;
+// -- SHA-256 schedule update
+def : InstRW<[Ampere1Write_4cyc_1Y], (instregex "^SHA256SU[01]")>;
+// -- SHA-3 instructions
+def : InstRW<[Ampere1Write_2cyc_1XY],
+  (instregex "^BCAX", "^EOR3", "^RAX1", "^XAR")>;
+// -- SHA-512 hash
+def : InstRW<[Ampere1Write_4cyc_1X], (instregex "^SHA512(H|H2)")>;
+// -- SHA-512 schedule update
+def : InstRW<[Ampere1Write_4cyc_1Y], (instregex "^SHA512SU[01]")>;
+// -- SHA1 choose/majority/parity
+def : InstRW<[Ampere1Write_4cyc_1X], (instregex "^SHA1[CMP]")>;
+// -- SHA1 hash/schedule update
+def : InstRW<[Ampere1Write_2cyc_1Y], (instregex "^SHA1SU[01]")>;
+def : InstRW<[Ampere1Write_2cyc_1Y], (instregex "^SHA1H")>;
+
+// FP and vector load instructions
+// -- Load 1-element structure to one/all lanes
+// ---- all lanes
+def : InstRW<[Ampere1Write_7cyc_1L_1XY],
+  (instregex "^LD1Rv(8b|4h|2s|16b|8h|4s|2d)")>;
+// ---- one lane
+def : InstRW<[Ampere1Write_7cyc_1L_1XY],
+  (instregex "^LD1i(8|16|32|64)")>;
+// -- Load 1-element structure to one/all lanes, 1D size
+def : InstRW<[Ampere1Write_5cyc_1L],
+  (instregex "^LD1Rv1d")>;
+// -- Load 1-element structures to 1 register
+def : InstRW<[Ampere1Write_5cyc_1L],
+  (instregex "^LD1Onev(8b|4h|2s|1d|16b|8h|4s|2d)")>;
+// -- Load 1-element structures to 2 registers
+def : InstRW<[Ampere1Write_5cyc_2L],
+  (instregex "^LD1Twov(8b|4h|2s|1d|16b|8h|4s|2d)")>;
+// -- Load 1-element structures to 3 registers
+def : InstRW<[Ampere1Write_6cyc_3L],
+  (instregex "^LD1Threev(8b|4h|2s|1d|16b|8h|4s|2d)")>;
+// -- Load 1-element structures to 4 registers
+def : InstRW<[Ampere1Write_6cyc_4L],
+  (instregex "^LD1Fourv(8b|4h|2s|1d|16b|8h|4s|2d)")>;
+// -- Load 2-element structure to all lanes of 2 registers, 1D size
+def : InstRW<[Ampere1Write_5cyc_2L],
+  (instregex "^LD2Rv1d")>;
+// -- Load 2-element structure to all lanes of 2 registers, other sizes
+def : InstRW<[Ampere1Write_7cyc_2L_2XY],
+  (instregex "^LD2Rv(8b|4h|2s|16b|8h|4s|2d)")>;
+// -- Load 2-element structure to one lane of 2 registers
+def : InstRW<[Ampere1Write_7cyc_2L_2XY],
+  (instregex "^LD2i(8|16|32|64)")>;
+// -- Load 2-element structures to 2 registers, 16B/8H/4S/2D size
+def : InstRW<[Ampere1Write_7cyc_2L_2XY],
+  (instregex "^LD2Twov(16b|8h|4s|2d)")>;
+// -- Load 2-element structures to 2 registers, 8B/4H/2S size
+def : InstRW<[Ampere1Write_9cyc_2L_3XY],
+  (instregex "^LD2Twov(8b|4h|2s)")>;
+// -- Load 3-element structure to all lanes of 3 registers, 1D size
+def : InstRW<[Ampere1Write_6cyc_3L],
+  (instregex "^LD3Rv1d")>;
+// -- Load 3-element structure to all lanes of 3 registers, other sizes
+def : InstRW<[Ampere1Write_8cyc_3L_3XY],
+  (instregex "^LD3Rv(8b|4h|2s|16b|8h|4s|2d)")>;
+// -- Load 3-element structure to one lane of 3 registers
+def : InstRW<[Ampere1Write_8cyc_3L_3XY],
+  (instregex "^LD3i(8|16|32|64)")>;
+// -- Load 3-element structures to 3 registers, 16B/8H/4S sizes
+def : InstRW<[Ampere1Write_9cyc_3L_3XY],
+  (instregex "^LD3Threev(16b|8h|4s)")>;
+// -- Load 3-element structures to 3 registers, 2D size
+def : InstRW<[Ampere1Write_8cyc_3L_3XY],
+  (instregex "^LD3Threev2d")>;
+// -- Load 3-element structures to 3 registers, 8B/4H/2S sizes
+def : InstRW<[Ampere1Write_10cyc_3L_3XY],
+  (instregex "^LD3Threev(8b|4h|2s)")>;
+// -- Load 4-element structure to all lanes of 4 registers, 1D size
+def : InstRW<[Ampere1Write_6cyc_4L],
+  (instregex "^LD4Rv1d")>;
+// -- Load 4-element structure to all lanes of 4 registers, other sizes
+def : InstRW<[Ampere1Write_8cyc_4L_4XY],
+  (instregex "^LD4Rv(8b|4h|2s|16b|8h|4s|2d)")>;
+// -- Load 4-element structure to one lane of 4 registers
+def : InstRW<[Ampere1Write_6cyc_4L],
+  (instregex "^LD4i(8|16|32|64)")>;
+// -- Load 4-element structures to 4 registers, 2D size
+def : InstRW<[Ampere1Write_9cyc_4L_4XY],
+  (instregex "^LD4Fourv2d")>;
+// -- Load 4-element structures to 4 registers, 2S size
+def : InstRW<[Ampere1Write_12cyc_4L_8XY],
+  (instregex "^LD4Fourv2s")>;
+// -- Load 4-element structures to 4 registers, other sizes
+def : InstRW<[Ampere1Write_11cyc_4L_8XY],
+  (instregex "^LD4Fourv(8b|4h|16b|8h|4s)")>;
+// -- Load pair, Q-form
+def : InstRW<[Ampere1Write_5cyc_2L], (instregex "LDN?PQ")>;
+// -- Load pair, S/D-form
+def : InstRW<[Ampere1Write_5cyc_1L_1BS], (instregex "LDN?P(S|D)")>;
+// -- Load register
+def : InstRW<[Ampere1Write_5cyc_1L], (instregex "LDU?R[BHSDQ]i")>;
+// -- Load register, sign-extended register
+def : InstRW<[Ampere1Write_6cyc_1AB_1L], (instregex "LDR[BHSDQ]ro(W|X)")>;
+
+// FP and vector store instructions
+// -- Store 1-element structure from one lane of 1 register
+def : InstRW<[Ampere1Write_4cyc_1XY_1S_1Z],
+  (instregex "^ST1i(8|16|32|64)")>;
+// -- Store 1-element structures from 1 register
+def : InstRW<[Ampere1Write_2cyc_1S_1Z],
+  (instregex "^ST1Onev(8b|4h|2s|1d|16b|8h|4s|2d)")>;
+// -- Store 1-element structures from 2 registers
+def : InstRW<[Ampere1Write_3cyc_2S_2Z],
+  (instregex "^ST1Twov(8b|4h|2s|1d|16b|8h|4s|2d)")>;
+// -- Store 1-element structures from 3 registers
+def : InstRW<[Ampere1Write_4cyc_3S_3Z],
+  (instregex "^ST1Threev(8b|4h|2s|1d|16b|8h|4s|2d)")>;
+// -- Store 1-element structures from 4 registers
+def : InstRW<[Ampere1Write_5cyc_4S_4Z],
+  (instregex "^ST1Fourv(8b|4h|2s|1d|16b|8h|4s|2d)")>;
+// -- Store 2-element structure from one lane of 2 registers
+def : InstRW<[Ampere1Write_5cyc_2XY_2S_2Z],
+  (instregex "^ST2i(8|16|32|64)")>;
+// -- Store 2-element structures from 2 registers, 16B/8H/4S/2D sizes
+def : InstRW<[Ampere1Write_5cyc_2XY_2S_2Z],
+        (instregex "^ST2Twov(16b|8h|4s|2d)")>;
+// -- Store 2-element structures from 2 registers, 8B/4H/2S sizes
+def : InstRW<[Ampere1Write_6cyc_2XY_2S_2Z],
+        (instregex "^ST2Twov(8b|4h|2s)")>;
+// -- Store 3-element structure from one lane of 3 registers
+def : InstRW<[Ampere1Write_6cyc_3XY_3S_3Z],
+  (instregex "^ST3i(8|16|32|64)")>;
+// -- Store 3-element structures from 3 registers
+def : InstRW<[Ampere1Write_6cyc_3XY_3S_3Z],
+  (instregex "^ST3Threev(8b|4h|2s|1d|16b|8h|4s|2d)")>;
+// -- Store 4-element structure from one lane of 4 registers
+def : InstRW<[Ampere1Write_7cyc_4XY_4S_4Z],
+  (instregex "^ST4i(8|16|32|64)")>;
+// -- Store 4-element structures from 4 registers, 16B/8H/4S sizes
+def : InstRW<[Ampere1Write_9cyc_8XY_4S_4Z],
+  (instregex "^ST4Fourv(16b|8h|4s)")>;
+// -- Store 4-element structures from 4 registers, 2D sizes
+def : InstRW<[Ampere1Write_7cyc_4XY_4S_4Z],
+  (instregex "^ST4Fourv2d")>;
+// -- Store 4-element structures from 4 registers, 8B/4H/2S sizes
+def : InstRW<[Ampere1Write_9cyc_6XY_4S_4Z],
+  (instregex "^ST4Fourv(8b|4h|2s)")>;
+// -- Store pair, Q-form
+def : InstRW<[Ampere1Write_3cyc_2S_2Z], (instregex "^STN?PQ")>;
+// -- Store pair, S/D-form
+def : InstRW<[Ampere1Write_3cyc_1S_2Z],  (instregex "^STN?P[SD]")>;
+// -- Store register
+def : InstRW<[Ampere1Write_2cyc_1S_1Z],  (instregex "^STU?R[BHSDQ](ui|i)")>;
+// -- Store register, sign-extended register offset
+def : InstRW<[Ampere1Write_2cyc_1AB_1S_1Z], (instregex "^STR[BHSDQ]ro[XW]")>;
+
+// FP data processing, bfloat16 format
+def : InstRW<[Ampere1Write_5cyc_1XY], (instrs BFCVT)>;
+def : InstRW<[Ampere1Write_7cyc_2XY], (instrs BFCVTN, BFCVTN2)>;
+def : InstRW<[Ampere1Write_2cyc_1XY], (instregex "^BFDOTv", "^BF16DOT")>;
+def : InstRW<[Ampere1Write_4cyc_2XY], (instrs BFMMLA)>;
+def : InstRW<[Ampere1Write_5cyc_1XY], (instregex "^BFMLAL")>;
+
+// FP data processing, scalar/vector, half precision
+def : InstRW<[Ampere1Write_4cyc_1XY], (instregex "^F(ABD|ABS)v.[fi]16")>;
+def : InstRW<[Ampere1Write_4cyc_1XY],
+  (instregex "^F(ADD|ADDP|CADD|NEG|NMUL|SUB)v.[fi]16")>;
+def : InstRW<[Ampere1Write_4cyc_1XY],
+  (instregex "^F(AC|CM)(EQ|GE|GT|LE|LT)v.[fi]16")>;
+def : InstRW<[Ampere1Write_4cyc_1XY],
+  (instregex "^F(AC|CM)(EQ|GE|GT|LE|LT)16")>;
+def : InstRW<[Ampere1Write_4cyc_1X],
+        (instregex "^FCMPE?H")>;
+def : InstRW<[Ampere1Write_10cyc_1A_1BS_1X],
+  (instregex "^FCCMPE?H")>;
+def : InstRW<[Ampere1Write_10cyc_1A_1BS_1XY],
+  (instregex "^FCSELH")>;
+def : InstRW<[Ampere1Write_4cyc_1XY], (instregex "^FCVT[AMNPZ][SU]v.[if]16")>;
+def : InstRW<[Ampere1Write_4cyc_1XY], (instregex "^[SU]CVTFv.[fi]16")>;
+def : InstRW<[Ampere1Write_25cyc_1XY], (instregex "^FDIVv.[if]16", "FDIVH")>;
+def : InstRW<[Ampere1Write_4cyc_1XY], (instregex "^F(MAX|MIN)(NM)?P?v.[if]16")>;
+def : InstRW<[Ampere1Write_8cyc_2XY], (instregex "^F(MAX|MIN)(NM)?Vv4[if]16")>;
+def : InstRW<[Ampere1Write_12cyc_3XY], (instregex "^F(MAX|MIN)(NM)?Vv8[if]16")>;
+def : InstRW<[Ampere1Write_4cyc_1XY], (instregex "^FMULX?v.[if]16")>;
+def : InstRW<[Ampere1Write_4cyc_1XY], (instrs FMULX16)>;
+def : InstRW<[Ampere1Write_4cyc_1XY], (instregex "^FN?M(ADD|SUB)[H]rrr")>;
+def : InstRW<[Ampere1Write_4cyc_1XY], (instregex "^FML[AS]v.[if]16")>;
+def : InstRW<[Ampere1Write_4cyc_1XY], (instregex "^FRECPXv.[if]16")>;
+def : InstRW<[Ampere1Write_4cyc_1XY], (instregex "^F(RECP|RSQRT)S16")>;
+def : InstRW<[Ampere1Write_4cyc_1XY], (instregex "^FRINT[AIMNPXZ]v.[if]16")>;
+def : InstRW<[Ampere1Write_39cyc_1XY], (instregex "^FSQRTv.f16", "^FSQRTHr")>;
+
+// FP data processing, scalar/vector, single/double precision
+def : InstRW<[Ampere1Write_5cyc_1XY], (instregex "^F(ABD|ABS)v.[fi](32|64)")>;
+def : InstRW<[Ampere1Write_5cyc_1XY],
+  (instregex "^F(ADD|ADDP|CADD|NEG|NMUL|SUB)v.[fi](32|64)")>;
+def : InstRW<[Ampere1Write_5cyc_1XY],
+  (instregex "^F(AC|CM)(EQ|GE|GT|LE|LT)v.[fi](32|64)")>;
+def : InstRW<[Ampere1Write_5cyc_1XY],
+  (instregex "^F(AC|CM)(EQ|GE|GT|LE|LT)(32|64)")>;
+def : InstRW<[Ampere1Write_5cyc_1X],
+        (instregex "^FCMPE?(S|D)")>;
+def : InstRW<[Ampere1Write_11cyc_1A_1BS_1X],
+  (instregex "^FCCMPE?(S|D)")>;
+def : InstRW<[Ampere1Write_11cyc_1A_1BS_1XY],
+  (instregex "^FCSEL(S|D)")>;
+def : InstRW<[Ampere1Write_6cyc_1XY], (instregex "^FCVT[AMNPZ][SU]v.[if](32|64)")>;
+def : InstRW<[Ampere1Write_6cyc_1XY], (instregex "^[SU]CVTFv.[fi](32|64)")>;
+def : InstRW<[Ampere1Write_34cyc_1XY], (instregex "^FDIVv.[if](64)", "FDIVD")>;
+def : InstRW<[Ampere1Write_19cyc_1XY], (instregex "^FDIVv.[if](32)", "FDIVS")>;
+def : InstRW<[Ampere1Write_5cyc_1XY], (instregex "^F(MAX|MIN)(NM)?P?v.[if](32|64)")>;
+def : InstRW<[Ampere1Write_10cyc_2XY], (instregex "^F(MAX|MIN)(NM)?Vv.[if](32|64)")>;
+def : InstRW<[Ampere1Write_6cyc_1XY], (instregex "^FMULX?v.[if](32|64)")>;
+def : InstRW<[Ampere1Write_6cyc_1XY], (instrs FMULX32, FMULX64)>;
+def : InstRW<[Ampere1Write_5cyc_1XY], (instregex "^FN?M(ADD|SUB)[SD]rrr")>;
+def : InstRW<[Ampere1Write_5cyc_1XY], (instregex "^FML[AS]v.[if](32|64)")>;
+def : InstRW<[Ampere1Write_5cyc_1XY], (instregex "^FRECPXv.[if](32|64)")>;
+def : InstRW<[Ampere1Write_6cyc_1XY], (instregex "^F(RECP|RSQRT)S(32|64)")>;
+def : InstRW<[Ampere1Write_6cyc_1XY], (instregex "^FRINT[AIMNPXZ]v.[if](32|64)")>;
+def : InstRW<[Ampere1Write_6cyc_1XY], (instregex "^FRINT(32|64)")>;
+def : InstRW<[Ampere1Write_62cyc_1XY], (instregex "^FSQRTv.f64", "^FSQRTDr")>;
+def : InstRW<[Ampere1Write_32cyc_1XY], (instregex "^FSQRTv.f32", "^FSQRTSr")>;
+
+// FP miscellaneous instructions
+def : InstRW<[Ampere1Write_10cyc_1XY_1Z], (instregex "^FCVT[AMNPZ][SU][SU][XW][HSD]r")>;
+def : InstRW<[Ampere1Write_5cyc_1XY], (instregex "^FCVT[HSD]Hr")>;
+def : InstRW<[Ampere1Write_6cyc_1XY], (instregex "^FCVT[HSD][SD]r")>;
+def : InstRW<[Ampere1Write_6cyc_1XY], (instregex "^FCVTLv")>;
+def : InstRW<[Ampere1Write_8cyc_2XY], (instregex "^FCVT(N|XN)v")>;
+def : InstRW<[Ampere1Write_10cyc_1X_1Z], (instrs FJCVTZS)>;
+def : InstRW<[Ampere1Write_5cyc_1BS], (instregex "^FMOV[HSD][WX]r")>;
+def : InstRW<[Ampere1Write_7cyc_1BS_1XY], (instregex "^FMOVDXHighr")>;
+def : InstRW<[Ampere1Write_2cyc_1XY], (instregex "^FMOV[HSD][ri]")>;
+def : InstRW<[Ampere1Write_6cyc_1XY_1Z], (instregex "^FMOVXDHighr")>;
+def : InstRW<[Ampere1Write_4cyc_1Z], (instregex "^FMOV[WX][HSD]r")>;
+
+// Integer arithmetic and logical instructions
+def : InstRW<[Ampere1Write_1cyc_1A],
+  (instregex "ADC(W|X)r", "SBC(W|X)r")>;
+def : InstRW<[Ampere1Write_Arith],
+  (instregex "(ADD|AND|BIC|EON|EOR|ORN|ORR|SUB)(W|X)r")>;
+def : InstRW<[Ampere1Write_ArithFlagsetting],
+  (instregex "(ADD|AND|BIC|SUB)S(W|X)r")>;
+def : InstRW<[Ampere1Write_1cyc_1A],
+  (instregex "(ADC|SBC)S(W|X)r")>;
+def : InstRW<[Ampere1Write_1cyc_1A], (instrs RMIF)>;
+def : InstRW<[Ampere1Write_1cyc_1A],
+  (instregex "(CCMN|CCMP)(X|W)")>;
+def : InstRW<[Ampere1Write_1cyc_1A],
+  (instregex "(CSEL|CSINC|CSINV|CSNEG)(X|W)")>;
+def : InstRW<[Ampere1Write_18cyc_1BS], (instrs SDIVWr, UDIVWr)>;
+def : InstRW<[Ampere1Write_34cyc_1BS], (instrs SDIVXr, UDIVXr)>;
+def : InstRW<[Ampere1Write_3cyc_1BS],
+  (instregex "(S|U)MULHr")>;
+def : InstRW<[Ampere1Write_4cyc_1BS],
+  (instregex "(S|U)?M(ADD|SUB)L?r")>;
+
+// Integer load instructions
+def : InstRW<[Ampere1Write_4cyc_2L],
+  (instregex "(LDNP|LDP|LDPSW)(X|W)")>;
+def : InstRW<[Ampere1Write_4cyc_1L],
+  (instregex "LDR(B|D|H|Q|S)ui")>;
+def : InstRW<[Ampere1Write_4cyc_1L],
+  (instregex "LDR(D|Q|W|X)l")>;
+def : InstRW<[Ampere1Write_4cyc_1L],
+  (instregex "LDTR(B|H|W|X)i")>;
+def : InstRW<[Ampere1Write_4cyc_1L],
+  (instregex "LDTRS(BW|BX|HW|HX|W)i")>;
+def : InstRW<[Ampere1Write_4cyc_1L],
+  (instregex "LDUR(BB|HH|X|W)i")>;
+def : InstRW<[Ampere1Write_4cyc_1L],
+  (instregex "LDURS(BW|BX|HW|HX|W)i")>;
+def : InstRW<[Ampere1Write_5cyc_1AB_1L],
+  (instregex "LDR(HH|SHW|SHX|W|X)ro(W|X)")>;
+def : InstRW<[Ampere1Write_1cyc_1L],
+  (instrs PRFMl, PRFUMi, PRFUMi)>;
+def : InstRW<[Ampere1Write_2cyc_1AB_1L],
+  (instrs PRFMroW, PRFMroX)>;
+
+// Integer miscellaneous instructions
+def : InstRW<[Ampere1Write_1cyc_1A],  (instrs ADR, ADRP)>;
+def : InstRW<[Ampere1Write_1cyc_1B],  (instregex "EXTR(W|X)")>;
+def : InstRW<[Ampere1Write_1cyc_1B],  (instregex "(S|U)?BFM(W|X)")>;
+def : InstRW<[Ampere1Write_3cyc_1BS], (instregex "^CRC32C?[BHWX]")>;
+def : InstRW<[Ampere1Write_1cyc_1B],  (instregex "CLS(W|X)")>;
+def : InstRW<[Ampere1Write_1cyc_1A],  (instrs SETF8, SETF16)>;
+def : InstRW<[Ampere1Write_1cyc_1AB],
+  (instrs MOVKWi, MOVKXi, MOVNWi, MOVNXi, MOVZWi, MOVZXi)>;
+def : InstRW<[Ampere1Write_1cyc_1B],
+  (instregex "(RBIT|REV|REV16)(W|X)r", "REV32Xr")>;
+def : InstRW<[Ampere1Write_1cyc_1B],
+  (instregex "(ASR|LSL|LSR|ROR)V(W|X)r")>;
+
+// Integer store instructions
+def : InstRW<[Ampere1Write_1cyc_2S],  (instregex "STNP(X|W)i")>;
+def : InstRW<[Ampere1Write_2cyc_1B_1S],
+  (instrs STPWi, STPXi)>;
+def : InstRW<[Ampere1Write_3cyc_1B_1S_1AB],
+  (instregex "STP(W|X)(pre|post)")>;
+def : InstRW<[Ampere1Write_1cyc_1S],
+  (instrs STTRBi, STTRHi, STTRWi, STTRXi)>;
+def : InstRW<[Ampere1Write_1cyc_1S],
+  (instregex "STUR(BB|HH|X|W)i",
+       "STR(X|W)ui",
+       "STUR(BB|HH|X|W)i")>;
+def : InstRW<[Ampere1Write_1cyc_2S], (instrs STRWroX, STRXroX)>;
+def : InstRW<[Ampere1Write_2cyc_1AB_2S], (instrs STRWroW, STRXroW)>;
+
+// Pointer authentication
+//def : InstRW<[Ampere1Write_7cyc_1BS],
+//  (instrs AUTIAZ, AUTIBZ, AUTIASP, AUTIBSP, AUTIA1716, AUTIB1716)>;
+def : InstRW<[Ampere1Write_8cyc_1BS_1A],
+  (instregex "BRA(A|AZ|B|BZ)", "RETA(A|B)", "ERETA(A|B)")>;
+def : InstRW<[Ampere1Write_8cyc_1BS_2A],
+  (instrs BLRAA, BLRAAZ, BLRAB, BLRABZ)>;
+//def : InstRW<[Ampere1Write_7cyc_1BS],
+//  (instrs PACIAZ, PACIBZ, PACIASP, PACIBSP, PACIA1716, PACIB1716)>;
+def : InstRW<[Ampere1Write_11cyc_1BS_1L], (instregex "^LDRA(A|B)")>;
+def : InstRW<[Ampere1Write_7cyc_1BS], (instrs XPACD, XPACI)>;
+
+// Vector integer instructions
+// -- absolute difference
+def : InstRW<[Ampere1Write_3cyc_1XY],
+             (instregex "^SABAv", "^SABALv", "^SABDv", "^SABDLv",
+          "^UABAv", "^UABALv", "^UABDv", "^UABDLv")>;
+// -- arithmetic
+def : InstRW<[Ampere1Write_3cyc_1XY],
+        (instregex "^ABSv", "^(ADD|SUB)v", "^SADDLv", "^SADDW", "SHADD",
+       "SHSUB", "^SRHADD", "^URHADD", "SSUBL", "SSUBW",
+       "^UADDLv", "^UADDW", "UHADD", "UHSUB", "USUBL", "USUBW")>;
+// -- arithmetic, horizontal, 16B
+def : InstRW<[Ampere1Write_12cyc_4XY],
+            (instregex "^ADDVv16i8v", "^SADDLVv16i8v", "^UADDLVv16i8v")>;
+def : InstRW<[Ampere1Write_12cyc_4XY],
+            (instregex "^[SU](MIN|MAX)Vv16i8v")>;
+// -- arithmetic, horizontal, 4H/4S
+def : InstRW<[Ampere1Write_6cyc_2XY],
+            (instregex "^[SU]?ADDL?V(v8i8|v4i16|v2i32)v")>;
+def : InstRW<[Ampere1Write_6cyc_2XY],
+            (instregex "^[SU](MIN|MAX)V(v4i16|v4i32)v")>;
+// -- arithmetic, horizontal, 8B/8H
+def : InstRW<[Ampere1Write_9cyc_3XY],
+            (instregex "^[SU]?ADDL?V(v8i16|v4i32)v")>;
+def : InstRW<[Ampere1Write_9cyc_3XY],
+            (instregex "^[SU](MIN|MAX)V(v8i8|v8i16)v")>;
+// -- arithmetic, narrowing
+def : InstRW<[Ampere1Write_5cyc_2XY], (instregex "(ADD|SUB)HNv.*")>;
+def : InstRW<[Ampere1Write_5cyc_2XY], (instregex "(RADD|RSUB)HNv.*")>;
+// -- arithmetic, pairwise
+def : InstRW<[Ampere1Write_3cyc_1XY],
+  (instregex "^ADDPv", "^SADALP", "^UADALP", "^SADDLPv", "^UADDLPv")>;
+// -- arithmetic, saturating
+def : InstRW<[Ampere1Write_3cyc_1XY],
+  (instregex "^SQADD", "^SQSUB", "^SUQADD", "^UQADD", "^UQSUB", "^USQADD")>;
+// -- bit count
+def : InstRW<[Ampere1Write_2cyc_1XY],
+  (instregex "^(CLS|CLZ|CNT)v")>;
+// -- compare
+def : InstRW<[Ampere1Write_3cyc_1XY],
+        (instregex "^CMEQv", "^CMGEv", "^CMGTv", "^CMLEv", "^CMLTv",
+       "^CMHIv", "^CMHSv")>;
+// -- compare non-zero
+def : InstRW<[Ampere1Write_2cyc_1XY], (instregex "^CMTSTv")>;
+// -- dot product
+def : InstRW<[Ampere1Write_3cyc_1XY], (instregex "^(S|SU|U|US)DOTv")>;
+// -- fp reciprocal estimate
+def : InstRW<[Ampere1Write_5cyc_1XY], (instregex "^FRECPEv", "^FRSQRTEv")>;
+// -- integer reciprocal estimate
+def : InstRW<[Ampere1Write_5cyc_1XY], (instregex "^URECPEv", "^URSQRTEv")>;
+// -- logical
+def : InstRW<[Ampere1Write_2cyc_1XY],
+  (instregex "^ANDv", "^BICv", "^EORv", "^ORRv", "^ORNv", "^NOTv")>;
+// -- logical, narrowing
+def : InstRW<[Ampere1Write_5cyc_2XY],
+  (instregex "RSHRNv", 
+       "SHRNv", "SQSHRNv", "SQSHRUNv",
+       "UQXTNv")>;
+// -- matrix multiply
+def : InstRW<[Ampere1Write_6cyc_2XY],
+  (instrs SMMLA, UMMLA, USMMLA)>;
+// -- max/min
+def : InstRW<[Ampere1Write_3cyc_1XY],
+        (instregex "^SMAXv", "^SMINv", "^UMAXv", "^UMINv")>;
+def : InstRW<[Ampere1Write_3cyc_1XY],
+        (instregex "^SMAXPv", "^SMINPv", "^UMAXPv", "^UMINPv")>;
+// -- move immediate
+def : InstRW<[Ampere1Write_2cyc_1XY], (instregex "^MOVIv", "^MVNIv")>;
+// -- multiply
+def : InstRW<[Ampere1Write_3cyc_1XY],
+  (instregex "MULv", "SMULLv", "UMULLv", "SQDMUL(H|L)v", "SQRDMULHv")>;
+// -- multiply accumulate
+def : InstRW<[Ampere1Write_3cyc_1XY],
+  (instregex "MLAv", "MLSv", "(S|U|SQD)(MLAL|MLSL)v", "SQRDML(A|S)Hv")>;
+// -- negation, saturating
+def : InstRW<[Ampere1Write_2cyc_1XY], (instregex "^SQABS", "^SQNEG")>;
+// -- reverse bits/bytes
+def : InstRW<[Ampere1Write_2cyc_1XY],
+  (instregex "^RBITv", "^REV16v", "^REV32v", "^REV64v")>;
+// -- shift
+def : InstRW<[Ampere1Write_3cyc_1XY], (instregex "^[SU]SHL(v16i8|v8i16|v4i32|v2i64)")>;
+// -- shift and accumulate
+def : InstRW<[Ampere1Write_3cyc_1XY],
+  (instregex "SRSRAv", "SSRAv", "URSRAv", "USRAv")>;
+// -- shift, saturating
+def : InstRW<[Ampere1Write_3cyc_1XY],
+  (instregex "^SQRSHLv", "^SQRSHRNv", "^SQRSHRUNv", "^SQSHL", "^SQSHLU",
+       "^SQXTNv", "^SQXTUNv", "^UQSHRNv", "UQRSHRNv", "^UQRSHL",
+       "^UQSHL")>;
+
+// Vector miscellaneous instructions
+// -- duplicate element
+def : InstRW<[Ampere1Write_2cyc_1XY], (instregex "^DUPv.+lane")>;
+// -- duplicate from GPR
+def : InstRW<[Ampere1Write_5cyc_1BS], (instregex "^DUPv.+gpr")>;  
+// -- extract narrow
+def : InstRW<[Ampere1Write_2cyc_1XY], (instregex "^XTNv")>;
+// -- insert/extract element
+def : InstRW<[Ampere1Write_2cyc_1XY], (instregex "^EXTv", "^INSv.+lane")>;
+// -- move FP immediate
+def : InstRW<[Ampere1Write_2cyc_1XY], (instregex "^FMOVv")>;
+// -- move element to GPR
+def : InstRW<[Ampere1Write_6cyc_1XY_1Z], (instregex "(S|U)MOVv")>;
+// -- move from GPR to any element
+def : InstRW<[Ampere1Write_7cyc_1BS_1XY], (instregex "^INSv.+gpr")>;
+// -- table lookup
+def : InstRW<[Ampere1Write_2cyc_1XY],
+            (instrs TBLv8i8One, TBLv16i8One, TBXv8i8One, TBXv16i8One)>;
+def : InstRW<[Ampere1Write_4cyc_2XY],
+            (instrs TBLv8i8Two, TBLv16i8Two, TBXv8i8Two, TBXv16i8Two)>;
+def : InstRW<[Ampere1Write_6cyc_3XY],
+            (instrs TBLv8i8Three, TBLv16i8Three, TBXv8i8Three, TBXv16i8Three)>;
+def : InstRW<[Ampere1Write_8cyc_4XY],
+            (instrs TBLv8i8Four, TBLv16i8Four, TBXv8i8Four, TBXv16i8Four)>;
+// -- transpose
+def : InstRW<[Ampere1Write_2cyc_1XY],
+              (instregex "^TRN1v", "^TRN2v", "^UZP1v", "^UZP2v")>;
+// -- zip/unzip
+def : InstRW<[Ampere1Write_2cyc_1XY], (instregex "^ZIP1v", "^ZIP2v")>;
+
+} // SchedModel = Ampere1Model

llvm/lib/Target/AArch64/AArch64SchedPredAmpere.td

+//===- AArch64SchedPredAmpere.td - AArch64 Sched Preds -----*- tablegen -*-===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+// This file defines scheduling predicate definitions that are used by the
+// AArch64 Ampere Computing processors.
+//
+//===----------------------------------------------------------------------===//
+
+// Auxiliary predicates.
+
+// Check for a LSL shift <= 4
+def AmpereCheapLSL : MCSchedPredicate<
+        CheckAny<[CheckShiftBy0,
+                                 CheckAll<
+                                   [CheckShiftLSL,
+                                    CheckAny<
+                                      [CheckShiftBy1,
+                                       CheckShiftBy2,
+                                       CheckShiftBy3,
+               CheckShiftBy4]>]>]>>;

llvm/lib/Target/AArch64/AArch64SchedPredicates.td

let FunctionMapper = "AArch64_AM::getShiftType" in {
   def CheckShiftLSL                : CheckImmOperand_s<3, "AArch64_AM::LSL">;
   def CheckShiftLSR                : CheckImmOperand_s<3, "AArch64_AM::LSR">;
   def CheckShiftASR                : CheckImmOperand_s<3, "AArch64_AM::ASR">;
   def CheckShiftROR                : CheckImmOperand_s<3, "AArch64_AM::ROR">;
   def CheckShiftMSL                : CheckImmOperand_s<3, "AArch64_AM::MSL">;
 }
 
 // Check for shifting in arithmetic and logic instructions.
-foreach I = {0-3, 8} in {
+foreach I = {0-4, 8} in {
   let FunctionMapper = "AArch64_AM::getShiftValue" in
   def CheckShiftBy#I        : CheckImmOperand<3, I>;
 }
 
 // Generic predicates.
 
 // Identify whether an instruction is the 16-bit NEON form based on its result.
 def CheckHForm             : CheckAll<[CheckIsRegOperand<0>,

llvm/lib/Target/AArch64/AArch64Subtarget.h

 //===--- AArch64Subtarget.h - Define Subtarget for the AArch64 -*- C++ -*--===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
enum ARMProcFamilyEnum : uint8_t {
     NeoverseV1,
     Saphira,
     ThunderX2T99,
     ThunderX,
     ThunderXT81,
     ThunderXT83,
     ThunderXT88,
     ThunderX3T110,
-    TSV110
+    TSV110,
+    Ampere1
   };
 
 protected:
   /// ARMProcFamily - ARM processor family: Cortex-A53, Cortex-A57, and others.
   ARMProcFamilyEnum ARMProcFamily = Others;
 
   bool HasV8_0aOps = false;
   bool HasV8_1aOps = false;

llvm/lib/Target/AArch64/AArch64Subtarget.cpp

 //===-- AArch64Subtarget.cpp - AArch64 Subtarget Information ----*- C++ -*-===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 //
case ThunderX3T110:
     PrefLoopLogAlignment = 2;
     MaxInterleaveFactor = 4;
     PrefetchDistance = 128;
     MinPrefetchStride = 1024;
     MaxPrefetchIterationsAhead = 4;
     // FIXME: remove this to enable 64-bit SLP if performance looks good.
     MinVectorRegisterBitWidth = 128;
     break;
+  case Ampere1:
+    CacheLineSize = 64;
+    PrefFunctionLogAlignment = 6;
+    PrefLoopLogAlignment = 6;
+    MaxInterleaveFactor = 4;
+    break;
   }
 }
 
 AArch64Subtarget::AArch64Subtarget(const Triple &TT, const std::string &CPU,
                                    const std::string &TuneCPU,
                                    const std::string &FS,
                                    const TargetMachine &TM, bool LittleEndian,
                                    unsigned MinSVEVectorSizeInBitsOverride,

llvm/test/CodeGen/AArch64/cpus.ll

 ; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=falkor 2>&1 | FileCheck %s
 ; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=saphira 2>&1 | FileCheck %s
 ; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=kryo 2>&1 | FileCheck %s
 ; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=thunderx2t99 2>&1 | FileCheck %s
 ; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=thunderx3t110 2>&1 | FileCheck %s
 ; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=tsv110 2>&1 | FileCheck %s
 ; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=apple-latest 2>&1 | FileCheck %s
 ; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=a64fx 2>&1 | FileCheck %s
+; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=ampere1 2>&1 | FileCheck %s
 ; RUN: llc < %s -mtriple=arm64-unknown-unknown -mcpu=invalidcpu 2>&1 | FileCheck %s --check-prefix=INVALID
 
 ; CHECK-NOT: {{.*}}  is not a recognized processor for this target
 ; INVALID: {{.*}}  is not a recognized processor for this target
 
 define i32 @f(i64 %z) {
   ret i32 0
 }

llvm/test/CodeGen/AArch64/neon-dot-product.ll

 ; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py
 ; RUN: llc -mtriple aarch64-none-linux-gnu -mattr=+dotprod    < %s | FileCheck %s
 ; RUN: llc -mtriple aarch64-none-linux-gnu -mcpu=cortex-a65   < %s | FileCheck %s
 ; RUN: llc -mtriple aarch64-none-linux-gnu -mcpu=cortex-a65ae < %s | FileCheck %s
 ; RUN: llc -mtriple aarch64-none-linux-gnu -mcpu=neoverse-e1  < %s | FileCheck %s
 ; RUN: llc -mtriple aarch64-none-linux-gnu -mcpu=neoverse-n1  < %s | FileCheck %s
 ; RUN: llc -mtriple aarch64-none-linux-gnu -mcpu=neoverse-n2  < %s | FileCheck %s
+; RUN: llc -mtriple aarch64-none-linux-gnu -mcpu=ampere1      < %s | FileCheck %s
 
 declare <2 x i32> @llvm.aarch64.neon.udot.v2i32.v8i8(<2 x i32>, <8 x i8>, <8 x i8>)
 declare <4 x i32> @llvm.aarch64.neon.udot.v4i32.v16i8(<4 x i32>, <16 x i8>, <16 x i8>)
 declare <2 x i32> @llvm.aarch64.neon.sdot.v2i32.v8i8(<2 x i32>, <8 x i8>, <8 x i8>)
 declare <4 x i32> @llvm.aarch64.neon.sdot.v4i32.v16i8(<4 x i32>, <16 x i8>, <16 x i8>)
 
 define <2 x i32> @test_vdot_u32(<2 x i32> %a, <8 x i8> %b, <8 x i8> %c) #0 {
 ; CHECK-LABEL: test_vdot_u32:

llvm/test/CodeGen/AArch64/remat.ll

 ; RUN: llc -mtriple=aarch64-linux-gnuabi -mcpu=exynos-m5 -o - %s | FileCheck %s
 ; RUN: llc -mtriple=aarch64-linux-gnuabi -mcpu=falkor -o - %s | FileCheck %s
 ; RUN: llc -mtriple=aarch64-linux-gnuabi -mcpu=saphira -o - %s | FileCheck %s
 ; RUN: llc -mtriple=aarch64-linux-gnuabi -mcpu=kryo -o - %s | FileCheck %s
 ; RUN: llc -mtriple=aarch64-linux-gnuabi -mcpu=thunderx2t99 -o - %s | FileCheck %s
 ; RUN: llc -mtriple=aarch64-linux-gnuabi -mcpu=tsv110 -o - %s | FileCheck %s
 ; RUN: llc -mtriple=aarch64-linux-gnuabi -mattr=+custom-cheap-as-move -o - %s | FileCheck %s
 ; RUN: llc -mtriple=aarch64-linux-gnuabi -mcpu=thunderx3t110 -o - %s | FileCheck %s
+; RUN: llc -mtriple=aarch64-linux-gnuabi -mcpu=ampere1 -o - %s | FileCheck %s
 
 %X = type { i64, i64, i64 }
 declare void @f(%X*)
 define void @t() {
 entry:
   %tmp = alloca %X
   call void @f(%X* %tmp)
 ; CHECK: add x0, sp, #8
 ; CHECK-NOT: mov
 ; CHECK-NEXT: bl f
   call void @f(%X* %tmp)
 ; CHECK: add x0, sp, #8
 ; CHECK-NOT: mov
 ; CHECK-NEXT: bl f
   ret void
 }

llvm/test/MC/AArch64/armv8.2a-dotprod.s

 // RUN: llvm-mc -triple aarch64 -mattr=+dotprod -show-encoding < %s | FileCheck %s  --check-prefix=CHECK-DOTPROD
 // RUN: llvm-mc -triple aarch64 -mcpu=cortex-a55 -show-encoding < %s | FileCheck %s --check-prefix=CHECK-DOTPROD
 // RUN: llvm-mc -triple aarch64 -mcpu=cortex-a65 -show-encoding < %s| FileCheck %s --check-prefix=CHECK-DOTPROD
 // RUN: llvm-mc -triple aarch64 -mcpu=cortex-a65ae -show-encoding < %s| FileCheck %s --check-prefix=CHECK-DOTPROD
 // RUN: llvm-mc -triple aarch64 -mcpu=cortex-a75 -show-encoding < %s | FileCheck %s --check-prefix=CHECK-DOTPROD
 // RUN: llvm-mc -triple aarch64 -mcpu=cortex-a76 -show-encoding < %s | FileCheck %s --check-prefix=CHECK-DOTPROD
 // RUN: llvm-mc -triple aarch64 -mcpu=cortex-a77 -show-encoding < %s     | FileCheck %s --check-prefix=CHECK-DOTPROD
 // RUN: llvm-mc -triple aarch64 -mcpu=cortex-a78 -show-encoding < %s     | FileCheck %s --check-prefix=CHECK-DOTPROD
 // RUN: llvm-mc -triple aarch64 -mcpu=cortex-x1 -show-encoding < %s     | FileCheck %s --check-prefix=CHECK-DOTPROD
 // RUN: llvm-mc -triple aarch64 -mcpu=neoverse-e1 -show-encoding < %s| FileCheck %s --check-prefix=CHECK-DOTPROD
 // RUN: llvm-mc -triple aarch64 -mcpu=neoverse-n1 -show-encoding < %s| FileCheck %s --check-prefix=CHECK-DOTPROD
 // RUN: llvm-mc -triple aarch64 -mcpu=neoverse-n2 -show-encoding < %s| FileCheck %s --check-prefix=CHECK-DOTPROD
 // RUN: llvm-mc -triple aarch64 -mcpu=tsv110 -show-encoding < %s | FileCheck %s --check-prefix=CHECK-DOTPROD
 // RUN: llvm-mc -triple aarch64 -mcpu=cortex-r82 -show-encoding < %s | FileCheck %s --check-prefix=CHECK-DOTPROD
 // RUN: llvm-mc -triple aarch64 -mattr=+v8r -show-encoding < %s | FileCheck %s --check-prefix=CHECK-DOTPROD
+// RUN: llvm-mc -triple aarch64 -mcpu=ampere1 -show-encoding < %s | FileCheck %s --check-prefix=CHECK-DOTPROD
 
 // RUN: not llvm-mc -triple aarch64 -mattr=+v8.2a -show-encoding < %s 2> %t
 // RUN: FileCheck --check-prefix=CHECK-NO-DOTPROD < %t %s
 // RUN: not llvm-mc -triple aarch64 -mattr=+v8r,-dotprod -show-encoding < %s 2> %t
 // RUN: FileCheck --check-prefix=CHECK-NO-DOTPROD < %t %s
 // RUN: not llvm-mc -triple aarch64 -mcpu=cortex-r82 -mattr=-dotprod -show-encoding < %s 2> %t
 // RUN: FileCheck --check-prefix=CHECK-NO-DOTPROD < %t %s
 // RUN: not llvm-mc -triple aarch64 -mcpu=cortex-a65 -mattr=-dotprod -show-encoding < %s 2> %t
 // RUN: FileCheck --check-prefix=CHECK-NO-DOTPROD < %t %s
 // RUN: not llvm-mc -triple aarch64 -mcpu=cortex-a65ae -mattr=-dotprod -show-encoding < %s 2> %t
 // RUN: FileCheck --check-prefix=CHECK-NO-DOTPROD < %t %s
 // RUN: not llvm-mc -triple aarch64 -mcpu=neoverse-e1 -mattr=-dotprod -show-encoding < %s 2> %t
 // RUN: FileCheck --check-prefix=CHECK-NO-DOTPROD < %t %s
 // RUN: not llvm-mc -triple aarch64 -mcpu=cortex-a77 -mattr=-dotprod -show-encoding < %s 2> %t
 // RUN: FileCheck --check-prefix=CHECK-NO-DOTPROD < %t %s
 // RUN: not llvm-mc -triple aarch64 -mcpu=cortex-a78 -mattr=-dotprod -show-encoding < %s 2> %t
 // RUN: FileCheck --check-prefix=CHECK-NO-DOTPROD < %t %s
 // RUN: not llvm-mc -triple aarch64 -mcpu=cortex-x1 -mattr=-dotprod -show-encoding < %s 2> %t
 // RUN: FileCheck --check-prefix=CHECK-NO-DOTPROD < %t %s
 // RUN: not llvm-mc -triple aarch64 -mcpu=neoverse-n1 -mattr=-dotprod -show-encoding < %s 2> %t
 // RUN: FileCheck --check-prefix=CHECK-NO-DOTPROD < %t %s
 // RUN: not llvm-mc -triple aarch64 -mcpu=neoverse-n2 -mattr=-dotprod -show-encoding < %s 2> %t
 // RUN: FileCheck --check-prefix=CHECK-NO-DOTPROD < %t %s
+// RUN: not llvm-mc -triple aarch64 -mcpu=ampere1 -mattr=-dotprod -show-encoding < %s 2> %t
+// RUN: FileCheck --check-prefix=CHECK-NO-DOTPROD < %t %s
 
 udot v0.2s, v1.8b, v2.8b
 sdot v0.2s, v1.8b, v2.8b
 udot v0.4s, v1.16b, v2.16b
 sdot v0.4s, v1.16b, v2.16b
 udot v0.2s, v1.8b, v2.4b[0]
 sdot v0.2s, v1.8b, v2.4b[1]
 udot v0.4s, v1.16b, v2.4b[2]

llvm/test/MC/AArch64/armv8.3a-rcpc.s

 // RUN: llvm-mc -triple aarch64-none-linux-gnu -show-encoding -mattr=+v8.3a < %s 2>&1 | FileCheck %s
 // RUN: llvm-mc -triple aarch64-none-linux-gnu -show-encoding -mcpu=cortex-a65 < %s 2>&1 | FileCheck %s
 // RUN: llvm-mc -triple aarch64-none-linux-gnu -show-encoding -mcpu=cortex-a65ae < %s 2>&1 | FileCheck %s
 // RUN: llvm-mc -triple aarch64-none-linux-gnu -show-encoding -mcpu=cortex-a75 < %s 2>&1 | FileCheck %s
 // RUN: llvm-mc -triple aarch64-none-linux-gnu -show-encoding -mcpu=cortex-a55 < %s 2>&1 | FileCheck %s
 // RUN: llvm-mc -triple aarch64-none-linux-gnu -show-encoding -mcpu=neoverse-e1 < %s 2>&1 | FileCheck %s
 // RUN: llvm-mc -triple aarch64-none-linux-gnu -show-encoding -mcpu=neoverse-n1 < %s 2>&1 | FileCheck %s
 // RUN: llvm-mc -triple aarch64-none-linux-gnu -show-encoding -mcpu=neoverse-n2 < %s 2>&1 | FileCheck %s
+// RUN: llvm-mc -triple aarch64-none-linux-gnu -show-encoding -mcpu=ampere1 < %s 2>&1 | FileCheck %s
 // RUN: llvm-mc -triple aarch64-none-linux-gnu -show-encoding -mattr=+v8.2a -mattr=+rcpc < %s 2>&1 | FileCheck %s
 // RUN: not llvm-mc -triple aarch64-none-linux-gnu -mattr=+v8.2a < %s 2> %t
 // RUN: FileCheck --check-prefix=CHECK-REQ %s < %t
 
   ldaprb w0, [x0, #0]
   ldaprh w0, [x17, #0]
   ldapr w0, [x1, #0]
   ldapr x0, [x0, #0]

llvm/test/MC/Disassembler/AArch64/armv8.3a-rcpc.txt

 # RUN: llvm-mc -triple aarch64-none-linux-gnu -mattr=+v8.2a -mattr=+rcpc --disassemble < %s | FileCheck %s
 # RUN: llvm-mc -triple aarch64-none-linux-gnu -mattr=+v8.3a --disassemble < %s | FileCheck %s
 # RUN: llvm-mc -triple aarch64-none-linux-gnu -mattr=+v8r --disassemble < %s | FileCheck %s
 # RUN: llvm-mc -triple aarch64-none-linux-gnu -mcpu=cortex-r82 --disassemble < %s | FileCheck %s
 # RUN: llvm-mc -triple aarch64-none-linux-gnu -mcpu=cortex-a55 --disassemble < %s | FileCheck %s
 # RUN: llvm-mc -triple aarch64-none-linux-gnu -mcpu=cortex-a65 --disassemble < %s | FileCheck %s
 # RUN: llvm-mc -triple aarch64-none-linux-gnu -mcpu=cortex-a65ae --disassemble < %s | FileCheck %s
 # RUN: llvm-mc -triple aarch64-none-linux-gnu -mcpu=cortex-a75 --disassemble < %s | FileCheck %s
 # RUN: llvm-mc -triple aarch64-none-linux-gnu -mcpu=cortex-a77 --disassemble < %s | FileCheck %s
 # RUN: llvm-mc -triple aarch64-none-linux-gnu -mcpu=cortex-a78 --disassemble < %s | FileCheck %s
 # RUN: llvm-mc -triple aarch64-none-linux-gnu -mcpu=cortex-x1 --disassemble < %s | FileCheck %s
 # RUN: llvm-mc -triple aarch64-none-linux-gnu -mcpu=neoverse-e1 --disassemble < %s | FileCheck %s
 # RUN: llvm-mc -triple aarch64-none-linux-gnu -mcpu=neoverse-n1 --disassemble < %s | FileCheck %s
 # RUN: llvm-mc -triple aarch64-none-linux-gnu -mcpu=neoverse-n2 --disassemble < %s | FileCheck %s
+# RUN: llvm-mc -triple aarch64-none-linux-gnu -mcpu=ampere1 --disassemble < %s | FileCheck %s
 
 # CHECK: ldaprb w0, [x0]
 # CHECK: ldaprh w0, [x0]
 # CHECK: ldapr w0, [x0]
 # CHECK: ldapr x0, [x0]
 [0x00,0xc0,0xbf,0x38]
 [0x00,0xc0,0xbf,0x78]
 [0x00,0xc0,0xbf,0xb8]

llvm/unittests/Support/TargetParserTest.cpp

 //===----------- TargetParser.cpp - Target Parser -------------------------===//
 //
 // Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
 // See https://llvm.org/LICENSE.txt for license information.
 // SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
 //
 //===----------------------------------------------------------------------===//
 
::testing::Values(
                              AArch64::AEK_FP16 | AArch64::AEK_RAS |
                              AArch64::AEK_LSE | AArch64::AEK_SVE |
                              AArch64::AEK_DOTPROD | AArch64::AEK_RCPC |
                              AArch64::AEK_RDM | AArch64::AEK_MTE |
                              AArch64::AEK_SSBS | AArch64::AEK_SB |
                              AArch64::AEK_SVE2 | AArch64::AEK_SVE2BITPERM |
                              AArch64::AEK_BF16 | AArch64::AEK_I8MM,
                          "8.5-A"),
+        ARMCPUTestParams("ampere1", "armv8.6-a", "crypto-neon-fp-armv8",
+                         AArch64::AEK_CRC | AArch64::AEK_CRYPTO |
+                             AArch64::AEK_FP | AArch64::AEK_SIMD |
+                             AArch64::AEK_FP16 | AArch64::AEK_RAS |
+                             AArch64::AEK_LSE | 
+                             AArch64::AEK_DOTPROD | AArch64::AEK_RCPC |
+                             AArch64::AEK_RDM | AArch64::AEK_MTE |
+                             AArch64::AEK_SSBS | AArch64::AEK_SB |
+                             AArch64::AEK_BF16 | AArch64::AEK_I8MM,
+                         "8.6-A"),
         ARMCPUTestParams(
             "neoverse-512tvb", "armv8.4-a", "crypto-neon-fp-armv8",
             AArch64::AEK_RAS | AArch64::AEK_SVE | AArch64::AEK_SSBS |
                 AArch64::AEK_RCPC | AArch64::AEK_CRC | AArch64::AEK_FP |
                 AArch64::AEK_SIMD | AArch64::AEK_RAS | AArch64::AEK_LSE |
                 AArch64::AEK_RDM | AArch64::AEK_RCPC | AArch64::AEK_DOTPROD |
                 AArch64::AEK_CRYPTO | AArch64::AEK_FP16 | AArch64::AEK_BF16,
             "8.4-A"),
::testing::Values(
                          "8.2-A"),
         ARMCPUTestParams("carmel", "armv8.2-a", "crypto-neon-fp-armv8",
                          AArch64::AEK_CRC | AArch64::AEK_CRYPTO |
                              AArch64::AEK_FP | AArch64::AEK_SIMD |
                              AArch64::AEK_FP16 | AArch64::AEK_RAS |
                              AArch64::AEK_LSE | AArch64::AEK_RDM,
                          "8.2-A")));
 
-static constexpr unsigned NumAArch64CPUArchs = 52;
+static constexpr unsigned NumAArch64CPUArchs = 53;
 
 TEST(TargetParserTest, testAArch64CPUArchList) {
   SmallVector<StringRef, NumAArch64CPUArchs> List;
   AArch64::fillValidCPUArchList(List);
 
   // No list exists for these in this test suite, so ensure all are
   // valid, and match the expected 'magic' count.
   EXPECT_EQ(List.size(), NumAArch64CPUArchs);