[llvm] [AArch64] Add Ampere1B scheduling/pipeline model (PR #81341)

Philipp Tomsich via llvm-commits llvm-commits at lists.llvm.org
Fri Feb 9 16:19:33 PST 2024


https://github.com/ptomsich created https://github.com/llvm/llvm-project/pull/81341

The Ampere1B core is enabled with a new scheduling/pipeline model, as it provides significant updates over the Ampere1 core; it reduces latencies on many instructions, has some micro-ops reassigned between the XY and X units, and provides modelling for the instructions added since Ampere1 and Ampere1A.

This is identical to the reverted #81338, but has the missing ',' added and was retested after a 'ninja clean'.

>From 0bf9a4b2e058f6ce34e02a40ed45d1c26c50efa2 Mon Sep 17 00:00:00 2001
From: Philipp Tomsich <philipp.tomsich at vrull.eu>
Date: Fri, 9 Feb 2024 15:48:46 -0800
Subject: [PATCH] [AArch64] Add Ampere1B scheduling/pipeline model

The Ampere1B core is enabled with a new scheduling/pipeline model, as it
provides significant updates over the Ampere1 core; it reduces latencies
on many instructions, has some micro-ops reassigned between the XY and X
units, and provides modelling for the instructions added since Ampere1
and Ampere1A.
---
 llvm/lib/Target/AArch64/AArch64.td            |    3 +-
 .../Target/AArch64/AArch64SchedAmpere1B.td    | 1063 +++++++++++++++++
 2 files changed, 1065 insertions(+), 1 deletion(-)
 create mode 100644 llvm/lib/Target/AArch64/AArch64SchedAmpere1B.td

diff --git a/llvm/lib/Target/AArch64/AArch64.td b/llvm/lib/Target/AArch64/AArch64.td
index e76204f5522515..156c48e02abf9c 100644
--- a/llvm/lib/Target/AArch64/AArch64.td
+++ b/llvm/lib/Target/AArch64/AArch64.td
@@ -837,6 +837,7 @@ include "AArch64SchedA64FX.td"
 include "AArch64SchedThunderX3T110.td"
 include "AArch64SchedTSV110.td"
 include "AArch64SchedAmpere1.td"
+include "AArch64SchedAmpere1B.td"
 include "AArch64SchedNeoverseN1.td"
 include "AArch64SchedNeoverseN2.td"
 include "AArch64SchedNeoverseV1.td"
@@ -1722,7 +1723,7 @@ def : ProcessorModel<"ampere1", Ampere1Model, ProcessorFeatures.Ampere1,
 def : ProcessorModel<"ampere1a", Ampere1Model, ProcessorFeatures.Ampere1A,
                      [TuneAmpere1A]>;
 
-def : ProcessorModel<"ampere1b", Ampere1Model, ProcessorFeatures.Ampere1B,
+def : ProcessorModel<"ampere1b", Ampere1BModel, ProcessorFeatures.Ampere1B,
                      [TuneAmpere1B]>;
 
 //===----------------------------------------------------------------------===//
diff --git a/llvm/lib/Target/AArch64/AArch64SchedAmpere1B.td b/llvm/lib/Target/AArch64/AArch64SchedAmpere1B.td
new file mode 100644
index 00000000000000..28c93c2b06aa11
--- /dev/null
+++ b/llvm/lib/Target/AArch64/AArch64SchedAmpere1B.td
@@ -0,0 +1,1063 @@
+//=- AArch64SchedAmpere1B.td - Ampere-1B 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-1B to
+// support instruction scheduling and other instruction cost heuristics.
+//
+//===----------------------------------------------------------------------===//
+
+// The Ampere-1 core is an out-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 Ampere1BModel : SchedMachineModel {
+  let IssueWidth            =   4;  // 4-way decode and dispatch
+  let MicroOpBufferSize     = 192;  // micro-op re-order buffer size
+  let LoadLatency           =   3;  // Optimistic load latency
+  let MispredictPenalty     =  10;  // Branch mispredict penalty
+  let LoopMicroOpBufferSize =  32;  // Instruction queue size
+  let CompleteModel = 0;
+
+  list<Predicate> UnsupportedFeatures = !listconcat(SVEUnsupported.F,
+                                                    SMEUnsupported.F,
+                                                    PAUnsupported.F);
+}
+
+let SchedModel = Ampere1BModel 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 Ampere1BUnitA  : ProcResource<2>;  // integer single-cycle, branch, and flags r/w
+def Ampere1BUnitB  : ProcResource<2>;  // integer single-cycle, and complex shifts
+def Ampere1BUnitBS : ProcResource<1>;  // integer multi-cycle
+def Ampere1BUnitL  : ProcResource<2>;  // load
+def Ampere1BUnitS  : ProcResource<2>;  // store address calculation
+def Ampere1BUnitX  : ProcResource<1>;  // FP and vector operations, and flag write
+def Ampere1BUnitY  : ProcResource<1>;  // FP and vector operations, and crypto
+def Ampere1BUnitZ  : ProcResource<1>;  // FP store data and FP-to-integer moves
+
+def Ampere1BUnitAB : ProcResGroup<[Ampere1BUnitA, Ampere1BUnitB]>;
+def Ampere1BUnitXY : ProcResGroup<[Ampere1BUnitX, Ampere1BUnitY]>;
+
+//===----------------------------------------------------------------------===//
+// Define customized scheduler read/write types specific to the Ampere-1.
+
+def Ampere1BWrite_1cyc_1A : SchedWriteRes<[Ampere1BUnitA]> {
+  let Latency = 1;
+  let NumMicroOps = 1;
+}
+
+def Ampere1BWrite_1cyc_2A : SchedWriteRes<[Ampere1BUnitA, Ampere1BUnitA]> {
+  let Latency = 1;
+  let NumMicroOps = 2;
+}
+
+def Ampere1BWrite_1cyc_1B : SchedWriteRes<[Ampere1BUnitB]> {
+  let Latency = 1;
+  let NumMicroOps = 1;
+}
+
+def Ampere1BWrite_1cyc_1BS : SchedWriteRes<[Ampere1BUnitBS]> {
+  let Latency = 1;
+  let NumMicroOps = 1;
+}
+
+def Ampere1BWrite_1cyc_1BS_1B : SchedWriteRes<[Ampere1BUnitBS, Ampere1BUnitB]> {
+  let Latency = 1;
+  let NumMicroOps = 2;
+}
+
+def Ampere1BWrite_1cyc_1AB : SchedWriteRes<[Ampere1BUnitAB]> {
+  let Latency = 1;
+  let NumMicroOps = 1;
+}
+
+def Ampere1BWrite_1cyc_1AB_1A : SchedWriteRes<[Ampere1BUnitAB, Ampere1BUnitA]> {
+  let Latency = 1;
+  let NumMicroOps = 2;
+}
+
+def Ampere1BWrite_1cyc_1L : SchedWriteRes<[Ampere1BUnitL]> {
+  let Latency = 1;
+  let NumMicroOps = 1;
+}
+
+def Ampere1BWrite_1cyc_1S : SchedWriteRes<[Ampere1BUnitS]> {
+  let Latency = 1;
+  let NumMicroOps = 1;
+}
+
+def Ampere1BWrite_1cyc_2S : SchedWriteRes<[Ampere1BUnitS, Ampere1BUnitS]> {
+  let Latency = 1;
+  let NumMicroOps = 2;
+}
+
+def Ampere1BWrite_2cyc_1Y : SchedWriteRes<[Ampere1BUnitY]> {
+  let Latency = 2;
+  let NumMicroOps = 1;
+}
+
+def Ampere1BWrite_2cyc_2AB : SchedWriteRes<[Ampere1BUnitAB, Ampere1BUnitAB]> {
+  let Latency = 2;
+  let NumMicroOps = 2;
+}
+
+def Ampere1BWrite_2cyc_1B_1AB : SchedWriteRes<[Ampere1BUnitB, Ampere1BUnitAB]> {
+  let Latency = 2;
+  let NumMicroOps = 2;
+}
+
+def Ampere1BWrite_2cyc_1B_1S : SchedWriteRes<[Ampere1BUnitB, Ampere1BUnitS]> {
+  let Latency = 2;
+  let NumMicroOps = 2;
+}
+
+def Ampere1BWrite_2cyc_1B_1S_1AB : SchedWriteRes<[Ampere1BUnitB,
+                                                  Ampere1BUnitS,
+                                                  Ampere1BUnitAB]> {
+  let Latency = 2;
+  let NumMicroOps = 3;
+}
+
+def Ampere1BWrite_2cyc_1XY : SchedWriteRes<[Ampere1BUnitXY]> {
+  let Latency = 2;
+  let NumMicroOps = 1;
+}
+
+def Ampere1BWrite_2cyc_1S_1Z : SchedWriteRes<[Ampere1BUnitS, Ampere1BUnitZ]> {
+  let Latency = 2;
+  let NumMicroOps = 2;
+}
+
+def Ampere1BWrite_3cyc_1BS : SchedWriteRes<[Ampere1BUnitBS]> {
+  let Latency = 3;
+  let NumMicroOps = 1;
+}
+
+def Ampere1BWrite_3cyc_1L : SchedWriteRes<[Ampere1BUnitL]> {
+  let Latency = 3;
+  let NumMicroOps = 1;
+}
+
+def Ampere1BWrite_3cyc_1X : SchedWriteRes<[Ampere1BUnitX]> {
+  let Latency = 3;
+  let NumMicroOps = 1;
+}
+
+def Ampere1BWrite_3cyc_1XY : SchedWriteRes<[Ampere1BUnitXY]> {
+  let Latency = 3;
+  let NumMicroOps = 1;
+}
+
+def Ampere1BWrite_3cyc_1Z : SchedWriteRes<[Ampere1BUnitZ]> {
+  let Latency = 3;
+  let NumMicroOps = 1;
+}
+
+def Ampere1BWrite_3cyc_1S_1Z : SchedWriteRes<[Ampere1BUnitS,
+                                              Ampere1BUnitZ]> {
+  let Latency = 3;
+  let NumMicroOps = 2;
+}
+
+def Ampere1BWrite_3cyc_1S_2Z : SchedWriteRes<[Ampere1BUnitS,
+                                              Ampere1BUnitZ, Ampere1BUnitZ]> {
+  let Latency = 3;
+  let NumMicroOps = 3;
+}
+
+def Ampere1BWrite_3cyc_2S_2Z : SchedWriteRes<[Ampere1BUnitS, Ampere1BUnitS,
+                                              Ampere1BUnitZ, Ampere1BUnitZ]> {
+  let Latency = 3;
+  let NumMicroOps = 4;
+}
+
+def Ampere1BWrite_4cyc_1BS_1AB : SchedWriteRes<[Ampere1BUnitBS, Ampere1BUnitAB]> {
+  let Latency = 4;
+  let NumMicroOps = 2;
+}
+
+def Ampere1BWrite_4cyc_1L : SchedWriteRes<[Ampere1BUnitL]> {
+  let Latency = 4;
+  let NumMicroOps = 1;
+}
+
+def Ampere1BWrite_4cyc_1L_1B : SchedWriteRes<[Ampere1BUnitL, Ampere1BUnitB]> {
+  let Latency = 4;
+  let NumMicroOps = 2;
+}
+
+def Ampere1BWrite_4cyc_1X : SchedWriteRes<[Ampere1BUnitX]> {
+  let Latency = 4;
+  let NumMicroOps = 1;
+}
+
+def Ampere1BWrite_4cyc_1XY : SchedWriteRes<[Ampere1BUnitXY]> {
+  let Latency = 4;
+  let NumMicroOps = 1;
+}
+
+def Ampere1BWrite_4cyc_2XY : SchedWriteRes<[Ampere1BUnitXY, Ampere1BUnitXY]> {
+  let Latency = 4;
+  let NumMicroOps = 2;
+}
+
+def Ampere1BWrite_5cyc_1BS : SchedWriteRes<[Ampere1BUnitBS]> {
+  let Latency = 5;
+  let NumMicroOps = 1;
+}
+
+def Ampere1BWrite_4cyc_1XY_1S_1Z : SchedWriteRes<[Ampere1BUnitXY,
+                                                  Ampere1BUnitS,
+                                                  Ampere1BUnitZ]> {
+  let Latency = 4;
+  let NumMicroOps = 3;
+}
+
+def Ampere1BWrite_5cyc_1BS : SchedWriteRes<[Ampere1BUnitBS]> {
+  let Latency = 5;
+  let NumMicroOps = 1;
+}
+
+def Ampere1BWrite_5cyc_4S_4Z : SchedWriteRes<[Ampere1BUnitS, Ampere1BUnitS,
+                                              Ampere1BUnitS, Ampere1BUnitS,
+                                              Ampere1BUnitZ, Ampere1BUnitZ,
+                                              Ampere1BUnitZ, Ampere1BUnitZ]> {
+  let Latency = 5;
+  let NumMicroOps = 8;
+}
+
+def Ampere1BWrite_5cyc_1L_1BS : SchedWriteRes<[Ampere1BUnitL,
+                                               Ampere1BUnitBS]> {
+  let Latency = 5;
+  let NumMicroOps = 2;
+}
+
+def Ampere1BWrite_5cyc_3L : SchedWriteRes<[Ampere1BUnitL,
+                                           Ampere1BUnitL,
+                                           Ampere1BUnitL]> {
+  let Latency = 5;
+  let NumMicroOps = 3;
+}
+
+def Ampere1BWrite_5cyc_4L : SchedWriteRes<[Ampere1BUnitL,
+                                           Ampere1BUnitL,
+                                           Ampere1BUnitL,
+                                           Ampere1BUnitL]> {
+  let Latency = 5;
+  let NumMicroOps = 4;
+}
+
+def Ampere1BWrite_5cyc_1X : SchedWriteRes<[Ampere1BUnitX]> {
+  let Latency = 5;
+  let NumMicroOps = 1;
+}
+
+def Ampere1BWrite_5cyc_2XY_2S_2Z : SchedWriteRes<[Ampere1BUnitXY, Ampere1BUnitXY,
+                                                  Ampere1BUnitS,  Ampere1BUnitS,
+                                                  Ampere1BUnitZ,  Ampere1BUnitZ]> {
+  let Latency = 5;
+  let NumMicroOps = 6;
+}
+
+def Ampere1BWrite_6cyc_1BS_1A : SchedWriteRes<[Ampere1BUnitBS, Ampere1BUnitA]> {
+  let Latency = 6;
+  let NumMicroOps = 2;
+}
+
+def Ampere1BWrite_6cyc_1BS_2A : SchedWriteRes<[Ampere1BUnitBS, Ampere1BUnitA,
+                                                               Ampere1BUnitA]> {
+  let Latency = 6;
+  let NumMicroOps = 3;
+}
+
+def Ampere1BWrite_6cyc_1X : SchedWriteRes<[Ampere1BUnitX]> {
+  let Latency = 6;
+  let NumMicroOps = 2;
+}
+
+def Ampere1BWrite_6cyc_2XY : SchedWriteRes<[Ampere1BUnitXY, Ampere1BUnitXY]> {
+  let Latency = 6;
+  let NumMicroOps = 2;
+}
+
+def Ampere1BWrite_6cyc_3XY : SchedWriteRes<[Ampere1BUnitXY, Ampere1BUnitXY,
+                                            Ampere1BUnitXY]> {
+  let Latency = 6;
+  let NumMicroOps = 3;
+}
+
+def Ampere1BWrite_6cyc_2XY_2S_2Z : SchedWriteRes<[Ampere1BUnitXY, Ampere1BUnitXY,
+                                                  Ampere1BUnitS,  Ampere1BUnitS,
+                                                  Ampere1BUnitZ,  Ampere1BUnitZ]> {
+  let Latency = 6;
+  let NumMicroOps = 6;
+}
+
+def Ampere1BWrite_6cyc_3XY_3S_3Z : SchedWriteRes<[Ampere1BUnitXY, Ampere1BUnitXY, Ampere1BUnitXY,
+                                                  Ampere1BUnitS,  Ampere1BUnitS,  Ampere1BUnitS,
+                                                  Ampere1BUnitZ,  Ampere1BUnitZ,  Ampere1BUnitZ]> {
+  let Latency = 6;
+  let NumMicroOps = 9;
+}
+
+def Ampere1BWrite_7cyc_1BS_1XY : SchedWriteRes<[Ampere1BUnitBS, Ampere1BUnitXY]> {
+  let Latency = 7;
+  let NumMicroOps = 2;
+}
+
+def Ampere1BWrite_7cyc_1XY_1Z : SchedWriteRes<[Ampere1BUnitXY, Ampere1BUnitZ]> {
+  let Latency = 7;
+  let NumMicroOps = 2;
+}
+
+def Ampere1BWrite_7cyc_4XY_4S_4Z : SchedWriteRes<[Ampere1BUnitXY, Ampere1BUnitXY,
+                                                  Ampere1BUnitXY, Ampere1BUnitXY,
+                                                  Ampere1BUnitS,  Ampere1BUnitS,
+                                                  Ampere1BUnitS,  Ampere1BUnitS,
+                                                  Ampere1BUnitZ,  Ampere1BUnitZ,
+                                                  Ampere1BUnitZ,  Ampere1BUnitZ]> {
+  let Latency = 7;
+  let NumMicroOps = 12;
+}
+
+def Ampere1BWrite_8cyc_1BS_1L : SchedWriteRes<[Ampere1BUnitBS, Ampere1BUnitL]> {
+  let Latency = 8;
+  let NumMicroOps = 2;
+}
+
+def Ampere1BWrite_8cyc_1BS_1XY : SchedWriteRes<[Ampere1BUnitBS, Ampere1BUnitXY]> {
+  let Latency = 8;
+  let NumMicroOps = 2;
+}
+
+def Ampere1BWrite_8cyc_2XY : SchedWriteRes<[Ampere1BUnitXY, Ampere1BUnitXY]> {
+  let Latency = 8;
+  let NumMicroOps = 2;
+}
+
+def Ampere1BWrite_8cyc_4XY : SchedWriteRes<[Ampere1BUnitXY, Ampere1BUnitXY,
+                                            Ampere1BUnitXY, Ampere1BUnitXY]> {
+  let Latency = 8;
+  let NumMicroOps = 4;
+}
+
+def Ampere1BWrite_9cyc_6XY_4S_4Z : SchedWriteRes<[Ampere1BUnitXY, Ampere1BUnitXY,
+                                                  Ampere1BUnitXY, Ampere1BUnitXY,
+                                                  Ampere1BUnitXY, Ampere1BUnitXY,
+                                                  Ampere1BUnitS,  Ampere1BUnitS,
+                                                  Ampere1BUnitS,  Ampere1BUnitS,
+                                                  Ampere1BUnitZ,  Ampere1BUnitZ,
+                                                  Ampere1BUnitZ,  Ampere1BUnitZ]> {
+  let Latency = 9;
+  let NumMicroOps = 14;
+}
+
+def Ampere1BWrite_9cyc_1A_1BS_1X : SchedWriteRes<[Ampere1BUnitA, Ampere1BUnitBS, Ampere1BUnitX]> {
+  let Latency = 9;
+  let NumMicroOps = 3;
+}
+
+def Ampere1BWrite_9cyc_1A_1BS_1XY : SchedWriteRes<[Ampere1BUnitA, Ampere1BUnitBS, Ampere1BUnitXY]> {
+  let Latency = 9;
+  let NumMicroOps = 3;
+}
+
+def Ampere1BWrite_9cyc_1X : SchedWriteRes<[Ampere1BUnitX]> {
+  let Latency = 9;
+  let NumMicroOps = 1;
+}
+
+def Ampere1BWrite_9cyc_3XY : SchedWriteRes<[Ampere1BUnitXY, Ampere1BUnitXY, Ampere1BUnitXY]> {
+  let Latency = 9;
+  let NumMicroOps = 3;
+}
+
+def Ampere1BWrite_11cyc_1BS_2XY : SchedWriteRes<[Ampere1BUnitBS, Ampere1BUnitXY, Ampere1BUnitXY]> {
+  let Latency = 11;
+  let NumMicroOps = 3;
+}
+
+def Ampere1BWrite_12cyc_1X : SchedWriteRes<[Ampere1BUnitX]> {
+  let Latency = 12;
+  let NumMicroOps = 1;
+}
+
+def Ampere1BWrite_13cyc_1BS_1X : SchedWriteRes<[Ampere1BUnitBS, Ampere1BUnitX]> {
+  let Latency = 13;
+  let NumMicroOps = 2;
+}
+
+def Ampere1BWrite_17cyc_1X : SchedWriteRes<[Ampere1BUnitX]> {
+  let Latency = 17;
+  let NumMicroOps = 1;
+}
+
+def Ampere1BWrite_19cyc_2BS_1X : SchedWriteRes<[Ampere1BUnitBS,
+                                                Ampere1BUnitBS,
+                                                Ampere1BUnitX]> {
+  let Latency = 13;
+  let NumMicroOps = 3;
+}
+
+def Ampere1BWrite_19cyc_1X : SchedWriteRes<[Ampere1BUnitX]> {
+  let Latency = 19;
+  let NumMicroOps = 1;
+}
+
+def Ampere1BWrite_21cyc_1X : SchedWriteRes<[Ampere1BUnitX]> {
+  let Latency = 21;
+  let NumMicroOps = 1;
+}
+
+def Ampere1BWrite_33cyc_1X : SchedWriteRes<[Ampere1BUnitX]> {
+  let Latency = 33;
+  let NumMicroOps = 1;
+}
+
+def Ampere1BWrite_39cyc_1X : SchedWriteRes<[Ampere1BUnitX]> {
+  let Latency = 39;
+  let NumMicroOps = 1;
+}
+
+def Ampere1BWrite_63cyc_1X : SchedWriteRes<[Ampere1BUnitX]> {
+  let Latency = 63;
+  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 Ampere1BWrite_Arith : SchedWriteVariant<[
+                                SchedVar<RegExtendedPred, [Ampere1Write_2cyc_2AB]>,
+                                SchedVar<IsCheapLSL,      [Ampere1Write_1cyc_1AB]>,
+                                SchedVar<NoSchedPred,     [Ampere1Write_2cyc_1B_1AB]>]>;
+
+def Ampere1BWrite_ArithFlagsetting : SchedWriteVariant<[
+                                SchedVar<RegExtendedPred, [Ampere1Write_2cyc_2AB]>,
+                                SchedVar<IsCheapLSL,      [Ampere1Write_1cyc_1AB]>,
+                                SchedVar<NoSchedPred,     [Ampere1Write_2cyc_1B_1AB]>]>;
+
+//===----------------------------------------------------------------------===//
+// 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,   [Ampere1BUnitAB]>;  // MOVN, MOVZ
+def : WriteRes<WriteI,     [Ampere1BUnitAB]>;  // ALU
+def : WriteRes<WriteISReg, [Ampere1BUnitB, Ampere1BUnitAB]> {
+  let Latency = 2;
+  let NumMicroOps = 2;
+}  // ALU of Shifted-Reg
+def : WriteRes<WriteIEReg, [Ampere1BUnitAB, Ampere1BUnitAB]> {
+  let Latency = 2;
+  let NumMicroOps = 2;
+}  // ALU of Extended-Reg
+def : WriteRes<WriteExtr,  [Ampere1BUnitB]>;  // EXTR shifts a reg pair
+def : WriteRes<WriteIS,    [Ampere1BUnitB]>;  // Shift/Scale
+def : WriteRes<WriteID32,  [Ampere1BUnitBS, Ampere1BUnitX]> {
+  let Latency = 13;
+}  // 32-bit Divide
+def : WriteRes<WriteID64,  [Ampere1BUnitBS, Ampere1BUnitX]> {
+  let Latency = 19;
+}  // 64-bit Divide
+def : WriteRes<WriteIM32,  [Ampere1BUnitBS]> {
+  let Latency = 3;
+}  // 32-bit Multiply
+def : WriteRes<WriteIM64,  [Ampere1BUnitBS, Ampere1UnitAB]> {
+  let Latency = 3;
+}  // 64-bit Multiply
+def : WriteRes<WriteBr,    [Ampere1BUnitA]>;
+def : WriteRes<WriteBrReg, [Ampere1BUnitA, Ampere1UnitA]>;
+def : WriteRes<WriteLD,    [Ampere1BUnitL]> {
+  let Latency = 3;
+}  // Load from base addr plus immediate offset
+def : WriteRes<WriteST,    [Ampere1BUnitS]> {
+  let Latency = 1;
+}  // Store to base addr plus immediate offset
+def : WriteRes<WriteSTP,   [Ampere1BUnitS, Ampere1BUnitS]> {
+  let Latency = 1;
+  let NumMicroOps = 1;
+}  // Store a register pair.
+def : WriteRes<WriteAdr,   [Ampere1BUnitAB]>;
+def : WriteRes<WriteLDIdx, [Ampere1BUnitAB, Ampere1BUnitS]> {
+  let Latency = 3;
+  let NumMicroOps = 1;
+}  // Load from a register index (maybe scaled).
+def : WriteRes<WriteSTIdx, [Ampere1BUnitS, Ampere1BUnitS]> {
+  let Latency = 1;
+  let NumMicroOps = 2;
+}  // Store to a register index (maybe scaled).
+def : WriteRes<WriteF,  [Ampere1BUnitXY]> {
+  let Latency = 2;
+}  // General floating-point ops.
+def : WriteRes<WriteFCmp,  [Ampere1BUnitX]> {
+  let Latency = 3;
+}  // Floating-point compare.
+def : WriteRes<WriteFCvt,  [Ampere1BUnitXY]> {
+  let Latency = 3;
+}  // Float conversion.
+def : WriteRes<WriteFCopy, [Ampere1BUnitXY]> {
+}  // Float-int register copy.
+def : WriteRes<WriteFImm,  [Ampere1BUnitXY]> {
+  let Latency = 2;
+}  // Float-int register copy.
+def : WriteRes<WriteFMul,  [Ampere1BUnitXY]> {
+  let Latency = 4;
+}  // Floating-point multiply.
+def : WriteRes<WriteFDiv,  [Ampere1BUnitXY]> {
+  let Latency = 19;
+}  // Floating-point division.
+def : WriteRes<WriteVd,    [Ampere1BUnitXY]> {
+  let Latency = 3;
+}  // 64bit Vector D ops.
+def : WriteRes<WriteVq,    [Ampere1BUnitXY]> {
+  let Latency = 3;
+}  // 128bit Vector Q ops.
+def : WriteRes<WriteVLD,   [Ampere1BUnitL, Ampere1BUnitL]> {
+  let Latency = 4;
+}  // Vector loads.
+def : WriteRes<WriteVST,   [Ampere1BUnitS, Ampere1BUnitZ]> {
+  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 = 3;
+}  // 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-1B.
+
+def : InstRW<[Ampere1BWrite_1cyc_1AB], (instrs COPY)>;
+
+// Branch instructions
+def : InstRW<[Ampere1BWrite_1cyc_1A], (instrs Bcc, BL, RET)>;
+def : InstRW<[Ampere1BWrite_1cyc_1A],
+        (instrs CBZW, CBZX, CBNZW, CBNZX, TBZW, TBZX, TBNZW, TBNZX)>;
+def : InstRW<[Ampere1BWrite_1cyc_2A], (instrs BLR)>;
+
+// Common Short Sequence Compression (CSSC)
+def : InstRW<[Ampere1BWrite_1cyc_1AB], (instrs ABS)>;
+def : InstRW<[Ampere1BWrite_1cyc_1BS], (instrs CNT)>;
+def : InstRW<[Ampere1BWrite_1cyc_1AB_1A], (instrs SMAX, SMIN)>;
+def : InstRW<[Ampere1BWrite_1cyc_1B], (instrs CTZ)>;
+def : InstRW<[Ampere1BWrite_1cyc_1AB_1A], (instrs UMAX, USMIN)>;
+
+// Cryptography instructions
+// -- AES encryption/decryption
+def : InstRW<[Ampere1BWrite_2cyc_1XY], (instregex "^AES[DE]")>;
+def : InstRW<[Ampere1BWrite_2cyc_1XY], (instregex "^AESI?MC")>;
+// -- Polynomial multiplication
+def : InstRW<[Ampere1BWrite_2cyc_1XY], (instregex "^PMUL", "^PMULL")>;
+// -- SHA-256 hash
+def : InstRW<[Ampere1BWrite_4cyc_1X], (instregex "^SHA256(H|H2)")>;
+// -- SHA-256 schedule update
+def : InstRW<[Ampere1BWrite_2cyc_1Y], (instregex "^SHA256SU[01]")>;
+// -- SHA-3 instructions
+def : InstRW<[Ampere1BWrite_2cyc_1XY],
+        (instregex "^BCAX", "^EOR3", "^RAX1", "^XAR")>;
+// -- SHA-512 hash
+def : InstRW<[Ampere1BWrite_4cyc_1X], (instregex "^SHA512(H|H2)")>;
+// -- SHA-512 schedule update
+def : InstRW<[Ampere1BWrite_2cyc_1Y], (instregex "^SHA512SU[01]")>;
+// -- SHA1 choose/majority/parity
+def : InstRW<[Ampere1BWrite_4cyc_1X], (instregex "^SHA1[CMP]")>;
+// -- SHA1 hash/schedule update
+def : InstRW<[Ampere1BWrite_2cyc_1Y], (instregex "^SHA1SU[01]")>;
+def : InstRW<[Ampere1BWrite_2cyc_1Y], (instregex "^SHA1H")>;
+// -- SM3 hash
+def : InstRW<[Ampere1BWrite_2cyc_1XY],
+    (instregex "^SM3PARTW[12]$", "^SM3SS1$", "^SM3TT[12][AB]$"0)>;
+def : InstRW<[Ampere1BWrite_4cyc_1X], (instrs SM4E, SM4ENCKEY)>;
+
+// FP and vector load instructions
+// -- Load 1-element structure to one/all lanes
+// ---- all lanes
+def : InstRW<[Ampere1BWrite_6cyc_1L_1XY],
+        (instregex "^LD1Rv(8b|4h|2s|16b|8h|4s|2d)")>;
+// ---- one lane
+def : InstRW<[Ampere1BWrite_6cyc_1L_1XY],
+        (instregex "^LD1i(8|16|32|64)")>;
+// -- Load 1-element structure to one/all lanes, 1D size
+def : InstRW<[Ampere1BWrite_4cyc_1L],
+        (instregex "^LD1Rv1d")>;
+// -- Load 1-element structures to 1 register
+def : InstRW<[Ampere1BWrite_4cyc_1L],
+        (instregex "^LD1Onev(8b|4h|2s|1d|16b|8h|4s|2d)")>;
+// -- Load 1-element structures to 2 registers
+def : InstRW<[Ampere1BWrite_4cyc_2L],
+        (instregex "^LD1Twov(8b|4h|2s|1d|16b|8h|4s|2d)")>;
+// -- Load 1-element structures to 3 registers
+def : InstRW<[Ampere1BWrite_5cyc_3L],
+        (instregex "^LD1Threev(8b|4h|2s|1d|16b|8h|4s|2d)")>;
+// -- Load 1-element structures to 4 registers
+def : InstRW<[Ampere1BWrite_5cyc_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<[Ampere1BWrite_4cyc_2L],
+        (instregex "^LD2Rv1d")>;
+// -- Load 2-element structure to all lanes of 2 registers, other sizes
+def : InstRW<[Ampere1BWrite_6cyc_2L_2XY],
+        (instregex "^LD2Rv(8b|4h|2s|16b|8h|4s|2d)")>;
+// -- Load 2-element structure to one lane of 2 registers
+def : InstRW<[Ampere1BWrite_6cyc_2L_2XY],
+        (instregex "^LD2i(8|16|32|64)")>;
+// -- Load 2-element structures to 2 registers, 16B/8H/4S/2D size
+def : InstRW<[Ampere1BWrite_6cyc_2L_2XY],
+        (instregex "^LD2Twov(16b|8h|4s|2d)")>;
+// -- Load 2-element structures to 2 registers, 8B/4H/2S size
+def : InstRW<[Ampere1BWrite_8cyc_2L_3XY],
+        (instregex "^LD2Twov(8b|4h|2s)")>;
+// -- Load 3-element structure to all lanes of 3 registers, 1D size
+def : InstRW<[Ampere1BWrite_5cyc_3L],
+        (instregex "^LD3Rv1d")>;
+// -- Load 3-element structure to all lanes of 3 registers, other sizes
+def : InstRW<[Ampere1BWrite_7cyc_3L_3XY],
+        (instregex "^LD3Rv(8b|4h|2s|16b|8h|4s|2d)")>;
+// -- Load 3-element structure to one lane of 3 registers
+def : InstRW<[Ampere1BWrite_7cyc_3L_3XY],
+        (instregex "^LD3i(8|16|32|64)")>;
+// -- Load 3-element structures to 3 registers, 16B/8H/4S sizes
+def : InstRW<[Ampere1BWrite_8cyc_3L_3XY],
+        (instregex "^LD3Threev(16b|8h|4s)")>;
+// -- Load 3-element structures to 3 registers, 2D size
+def : InstRW<[Ampere1BWrite_7cyc_3L_3XY],
+        (instregex "^LD3Threev2d")>;
+// -- Load 3-element structures to 3 registers, 8B/4H/2S sizes
+def : InstRW<[Ampere1BWrite_9cyc_3L_3XY],
+        (instregex "^LD3Threev(8b|4h|2s)")>;
+// -- Load 4-element structure to all lanes of 4 registers, 1D size
+def : InstRW<[Ampere1BWrite_5cyc_4L],
+        (instregex "^LD4Rv1d")>;
+// -- Load 4-element structure to all lanes of 4 registers, other sizes
+def : InstRW<[Ampere1BWrite_7cyc_4L_4XY],
+        (instregex "^LD4Rv(8b|4h|2s|16b|8h|4s|2d)")>;
+// -- Load 4-element structure to one lane of 4 registers
+def : InstRW<[Ampere1BWrite_7cyc_4L_4XY],
+        (instregex "^LD4i(8|16|32|64)")>;
+// -- Load 4-element structures to 4 registers, 2D size
+def : InstRW<[Ampere1BWrite_8cyc_4L_4XY],
+        (instregex "^LD4Fourv2d")>;
+// -- Load 4-element structures to 4 registers, 2S size
+def : InstRW<[Ampere1BWrite_11cyc_4L_8XY],
+        (instregex "^LD4Fourv2s")>;
+// -- Load 4-element structures to 4 registers, other sizes
+def : InstRW<[Ampere1BWrite_10cyc_4L_8XY],
+        (instregex "^LD4Fourv(8b|4h|16b|8h|4s)")>;
+// -- Load pair, Q-form
+def : InstRW<[Ampere1BWrite_4cyc_2L], (instregex "LDN?PQ")>;
+// -- Load pair, S/D-form
+def : InstRW<[Ampere1BWrite_5cyc_1L_1BS], (instregex "LDN?P(S|D)")>;
+// -- Load register
+def : InstRW<[Ampere1BWrite_4cyc_1L], (instregex "LDU?R[BHSDQ]i")>;
+// -- Load register, sign-extended register
+def : InstRW<[Ampere1BWrite_4cyc_1L], (instregex "LDR[BHSDQ]ro(W|X)")>;
+
+// FP and vector store instructions
+// -- Store 1-element structure from one lane of 1 register
+def : InstRW<[Ampere1BWrite_4cyc_1XY_1S_1Z],
+        (instregex "^ST1i(8|16|32|64)")>;
+// -- Store 1-element structures from 1 register
+def : InstRW<[Ampere1BWrite_2cyc_1S_1Z],
+        (instregex "^ST1Onev(8b|4h|2s|1d|16b|8h|4s|2d)")>;
+// -- Store 1-element structures from 2 registers
+def : InstRW<[Ampere1BWrite_3cyc_2S_2Z],
+        (instregex "^ST1Twov(8b|4h|2s|1d|16b|8h|4s|2d)")>;
+// -- Store 1-element structures from 3 registers
+def : InstRW<[Ampere1BWrite_4cyc_3S_3Z],
+        (instregex "^ST1Threev(8b|4h|2s|1d|16b|8h|4s|2d)")>;
+// -- Store 1-element structures from 4 registers
+def : InstRW<[Ampere1BWrite_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<[Ampere1BWrite_5cyc_2XY_2S_2Z],
+        (instregex "^ST2i(8|16|32|64)")>;
+// -- Store 2-element structures from 2 registers, 16B/8H/4S/2D sizes
+def : InstRW<[Ampere1BWrite_5cyc_2XY_2S_2Z],
+        (instregex "^ST2Twov(16b|8h|4s|2d)")>;
+// -- Store 2-element structures from 2 registers, 8B/4H/2S sizes
+def : InstRW<[Ampere1BWrite_6cyc_2XY_2S_2Z],
+        (instregex "^ST2Twov(8b|4h|2s)")>;
+// -- Store 3-element structure from one lane of 3 registers
+def : InstRW<[Ampere1BWrite_6cyc_3XY_3S_3Z],
+        (instregex "^ST3i(8|16|32|64)")>;
+// -- Store 3-element structures from 3 registers
+def : InstRW<[Ampere1BWrite_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<[Ampere1BWrite_7cyc_4XY_4S_4Z],
+        (instregex "^ST4i(8|16|32|64)")>;
+// -- Store 4-element structures from 4 registers, 16B/8H/4S sizes
+def : InstRW<[Ampere1BWrite_7cyc_4XY_4S_4Z],
+        (instregex "^ST4Fourv(16b|8h|4s)")>;
+// -- Store 4-element structures from 4 registers, 2D sizes
+def : InstRW<[Ampere1BWrite_7cyc_4XY_4S_4Z],
+        (instregex "^ST4Fourv2d")>;
+// -- Store 4-element structures from 4 registers, 8B/4H/2S sizes
+def : InstRW<[Ampere1BWrite_9cyc_6XY_4S_4Z],
+        (instregex "^ST4Fourv(8b|4h|2s)")>;
+// -- Store pair, Q-form
+def : InstRW<[Ampere1BWrite_3cyc_2S_2Z], (instregex "^STN?PQ")>;
+// -- Store pair, S/D-form
+def : InstRW<[Ampere1BWrite_3cyc_2S_2Z], (instregex "^STN?P[SD]")>;
+// -- Store register
+def : InstRW<[Ampere1BWrite_2cyc_1S_2Z], (instregex "^STU?R[BHSDQ](ui|i)")>;
+// -- Store register, sign-extended register offset
+def : InstRW<[Ampere1BWrite_2cyc_1S_1Z], (instregex "^STR[BHSDQ]ro[XW]")>;
+
+// FP data processing, bfloat16 format
+def : InstRW<[Ampere1BWrite_3cyc_1XY], (instrs BFCVT)>;
+def : InstRW<[Ampere1BWrite_8cyc_2XY], (instrs BFCVTN, BFCVTN2)>;
+def : InstRW<[Ampere1BWrite_2cyc_1XY], (instregex "^BFDOTv", "^BF16DOT")>;
+def : InstRW<[Ampere1BWrite_3cyc_1XY], (instrs BFMMLA)>;
+def : InstRW<[Ampere1BWrite_4cyc_1XY], (instregex "^BFMLAL")>;
+
+// FP data processing, scalar/vector, half precision
+def : InstRW<[Ampere1BWrite_3cyc_1XY], (instregex "^F(ABD|ABS)v.[fi]16")>;
+def : InstRW<[Ampere1BWrite_3cyc_1XY],
+        (instregex "^F(ADD|ADDP|CADD|NEG|NMUL|SUB)v.[fi]16")>;
+def : InstRW<[Ampere1BWrite_3cyc_1XY],
+        (instregex "^F(AC|CM)(EQ|GE|GT|LE|LT)v.[fi]16")>;
+def : InstRW<[Ampere1BWrite_3cyc_1XY],
+        (instregex "^F(AC|CM)(EQ|GE|GT|LE|LT)16")>;
+def : InstRW<[Ampere1BWrite_3cyc_1X],
+        (instregex "^FCMPE?H")>;
+def : InstRW<[Ampere1BWrite_9cyc_1A_1BS_1X],
+        (instregex "^FCCMPE?H")>;
+def : InstRW<[Ampere1BWrite_9cyc_1A_1BS_1XY],
+        (instregex "^FCSELH")>;
+def : InstRW<[Ampere1BWrite_3cyc_1XY], (instregex "^FCVT[AMNPZ][SU]v.[if]16")>;
+// Convert FP to integer, H-form
+def : InstRW<[Ampere1BWrite_3cyc_1XY], (instregex "^[SUd]CVTFv.[fi]16")>;
+// Convert to FP from GPR, H-form
+def : InstRW<[Ampere1BWrite_8cyc_1BS_1XY], (instregex "^[SU]CVTF_ZPmZ_[DSH]toH$")>;
+// Convert to FP from GPR, fixed-point, H-form
+def : InstRW<[Ampere1BWrite_11cyc_1BS_2XY], (instregex "^[SU]CVTF[SU][WX]Hri$")>;
+def : InstRW<[Ampere1BWrite_9cyc_1X], (instrs FDIVHrr)>;
+def : InstRW<[Ampere1BWrite_17cyc_1X], (instregex "^FDIVv.[if]16")>;
+def : InstRW<[Ampere1BWrite_3cyc_1XY], (instregex "^F(MAX|MIN)(NM)?P?v.[if]16")>;
+def : InstRW<[Ampere1BWrite_6cyc_2XY], (instregex "^F(MAX|MIN)(NM)?Vv4[if]16")>;
+def : InstRW<[Ampere1BWrite_9cyc_3XY], (instregex "^F(MAX|MIN)(NM)?Vv8[if]16")>;
+def : InstRW<[Ampere1BWrite_4cyc_1XY], (instregex "^FMULX?v.[if]16")>;
+def : InstRW<[Ampere1BWrite_4cyc_1XY], (instrs FMULX16)>;
+def : InstRW<[Ampere1BWrite_4cyc_1XY], (instregex "^FN?M(ADD|SUB)[H]rrr")>;
+def : InstRW<[Ampere1BWrite_4cyc_1XY], (instregex "^FML[AS]v.[if]16")>;
+def : InstRW<[Ampere1BWrite_3cyc_1XY], (instregex "^FRECPXv.[if]16")>;
+def : InstRW<[Ampere1BWrite_4cyc_1XY], (instregex "^F(RECP|RSQRT)S16")>;
+def : InstRW<[Ampere1BWrite_3cyc_1XY], (instregex "^FRINT[AIMNPXZ]v.[if]16")>;
+// FP square root, H-form
+def : InstRW<[Ampere1BWrite_21cyc_1X], (instrs FSQRTHr)>;
+// FP square root, vector-form, F16
+def : InstRW<[Ampere1BWrite_39cyc_1X], (instregex "^FSQRTv.f16")>;
+
+// FP data processing, scalar/vector, single/double precision
+def : InstRW<[Ampere1BWrite_3cyc_1XY], (instregex "^F(ABD|ABS)v.[fi](32|64)")>;
+def : InstRW<[Ampere1BWrite_3cyc_1XY],
+        (instregex "^F(ADD|ADDP|CADD|NEG|NMUL|SUB)v.[fi](32|64)")>;
+def : InstRW<[Ampere1BWrite_3cyc_1XY],
+        (instregex "^F(AC|CM)(EQ|GE|GT|LE|LT)v.[fi](32|64)")>;
+def : InstRW<[Ampere1BWrite_3cyc_1XY],
+        (instregex "^F(AC|CM)(EQ|GE|GT|LE|LT)(32|64)")>;
+def : InstRW<[Ampere1BWrite_3cyc_1X],
+        (instregex "^FCMPE?(S|D)")>;
+def : InstRW<[Ampere1BWrite_9cyc_1A_1BS_1X],
+        (instregex "^FCCMPE?(S|D)")>;
+def : InstRW<[Ampere1BWrite_9cyc_1A_1BS_1XY],
+        (instregex "^FCSEL(S|D)")>;
+def : InstRW<[Ampere1BWrite_3cyc_1XY], (instregex "^FCVT[AMNPZ][SU]v.[if](32|64)")>;
+// Convert FP to integer, S/D-form
+def : InstRW<[Ampere1BWrite_3cyc_1XY], (instregex "^[SUd]CVTFv.[fi](32|64)")>;
+// Convert to FP from GPR, S/D-form
+def : InstRW<[Ampere1BWrite_8cyc_1BS_1XY], (instregex "^[SU]CVTF_ZPmZ_[DSH]to[DS]$")>;
+// Convert to FP from GPR, fixed-point, S/D-form
+def : InstRW<[Ampere1BWrite_11cyc_1BS_2XY], (instregex "^[SU]CVTF[SU][WX][SD]ri$")>;
+def : InstRW<[Ampere1BWrite_19cyc_1X], (instregex "^FDIVv.[if](64)", "FDIVD")>;
+def : InstRW<[Ampere1BWrite_12cyc_1X], (instregex "^FDIVv.[if](32)", "FDIVS")>;
+def : InstRW<[Ampere1BWrite_3cyc_1XY], (instregex "^F(MAX|MIN)(NM)?P?v.[if](32|64)")>;
+def : InstRW<[Ampere1BWrite_6cyc_2XY], (instregex "^F(MAX|MIN)(NM)?Vv.[if](32|64)")>;
+def : InstRW<[Ampere1BWrite_4cyc_1XY], (instregex "^FMULX?v.[if](32|64)")>;
+def : InstRW<[Ampere1BWrite_4cyc_1XY], (instrs FMULX32, FMULX64)>;
+def : InstRW<[Ampere1BWrite_4cyc_1XY], (instregex "^FN?MUL")>;
+def : InstRW<[Ampere1BWrite_4cyc_1XY], (instregex "^FN?M(ADD|SUB)[SD]rrr")>;
+def : InstRW<[Ampere1BWrite_4cyc_1XY], (instregex "^FML[AS]v.[if](32|64)")>;
+def : InstRW<[Ampere1BWrite_3cyc_1XY], (instregex "^FRECPXv.[if](32|64)")>;
+def : InstRW<[Ampere1BWrite_3cyc_1XY], (instregex "^F(RECP|RSQRT)S(32|64)")>;
+def : InstRW<[Ampere1BWrite_3cyc_1XY], (instregex "^FRINT[AIMNPXZ]v.[if](32|64)")>;
+def : InstRW<[Ampere1BWrite_3cyc_1XY], (instregex "^FRINT(32|64)")>;
+def : InstRW<[Ampere1BWrite_63cyc_1X], (instregex "^FSQRTv.f64", "^FSQRTDr")>;
+def : InstRW<[Ampere1BWrite_33cyc_1X], (instregex "^FSQRTv.f32", "^FSQRTSr")>;
+
+// FP miscellaneous instructions
+def : InstRW<[Ampere1BWrite_7cyc_1XY_1Z], (instregex "^FCVT[AMNPZ][SU][SU][XW][HSD]r")>;
+def : InstRW<[Ampere1BWrite_3cyc_1XY], (instregex "^FCVT[HSD]Hr")>;
+def : InstRW<[Ampere1BWrite_3cyc_1XY], (instregex "^FCVT[HSD][SD]r")>;
+def : InstRW<[Ampere1BWrite_3cyc_1XY], (instregex "^FCVTLv")>;
+def : InstRW<[Ampere1BWrite_3cyc_1XY], (instregex "^FCVT(N|XN)v")>;
+def : InstRW<[Ampere1BWrite_7cyc_1X_1Z], (instrs FJCVTZS)>;
+def : InstRW<[Ampere1BWrite_5cyc_1BS], (instregex "^FMOV[HSD][WX]r")>;
+def : InstRW<[Ampere1BWrite_7cyc_1BS_1XY], (instregex "^FMOVDXHighr")>;
+def : InstRW<[Ampere1BWrite_2cyc_1XY], (instregex "^FMOV[HSD][ri]")>;
+def : InstRW<[Ampere1BWrite_5cyc_1X], (instregex "^FMOVXDHighr")>;
+def : InstRW<[Ampere1BWrite_3cyc_1Z], (instregex "^FMOV[WX][HSD]r")>;
+
+// Integer arithmetic and logical instructions
+def : InstRW<[Ampere1BWrite_1cyc_1A],
+        (instregex "ADC(W|X)r", "SBC(W|X)r")>;
+def : InstRW<[Ampere1BWrite_Arith],
+        (instregex "(ADD|AND|BIC|EON|EOR|ORN|ORR|SUB)[WX]r[sx]")>;
+def : InstRW<[Ampere1BWrite_1cyc_1AB],
+        (instregex "(ADD|AND|BIC|EON|EOR|ORN|ORR|SUB)[WX]r[ri]")>;
+def : InstRW<[Ampere1BWrite_ArithFlagsetting],
+        (instregex "(ADD|AND|BIC|SUB)S[WX]r[sx]")>;
+def : InstRW<[Ampere1BWrite_1cyc_1A],
+        (instregex "(ADD|AND|BIC|SUB)S[WX]r[ri]")>;
+def : InstRW<[Ampere1BWrite_1cyc_1A],
+        (instregex "(ADC|SBC)S[WX]r")>;
+def : InstRW<[Ampere1BWrite_1cyc_1A], (instrs RMIF)>;
+def : InstRW<[Ampere1BWrite_1cyc_1A],
+        (instregex "(CCMN|CCMP)(X|W)")>;
+def : InstRW<[Ampere1BWrite_1cyc_1A],
+        (instregex "(CSEL|CSINC|CSINV|CSNEG)(X|W)")>;
+def : InstRW<[Ampere1BWrite_13cyc_1BS_1X], (instrs SDIVWr, UDIVWr)>;
+def : InstRW<[Ampere1BWrite_19cyc_2BS_1X], (instrs SDIVXr, UDIVXr)>;
+def : InstRW<[Ampere1BWrite_3cyc_1BS],
+        (instregex "(S|U)MULHr")>;
+def : InstRW<[Ampere1BWrite_4cyc_1BS_1AB],
+        (instregex "(S|U)?M(ADD|SUB)L?r")>;
+
+// Integer load instructions
+def : InstRW<[Ampere1BWrite_3cyc_1L],
+        (instregex "(LDNP|LDP|LDPSW)(X|W)")>;
+def : InstRW<[Ampere1BWrite_3cyc_1L],
+        (instregex "LDR(B|D|H|Q|S)ui")>;
+def : InstRW<[Ampere1BWrite_3cyc_1L],
+        (instregex "LDR(D|Q|W|X)l")>;
+def : InstRW<[Ampere1BWrite_3cyc_1L],
+        (instregex "LDTR(B|H|W|X)i")>;
+def : InstRW<[Ampere1BWrite_3cyc_1L],
+        (instregex "LDTRS(BW|BX|HW|HX|W)i")>;
+def : InstRW<[Ampere1BWrite_3cyc_1L],
+        (instregex "LDUR(BB|HH|X|W)i")>;
+def : InstRW<[Ampere1BWrite_3cyc_1L],
+        (instregex "LDURS(BW|BX|HW|HX|W)i")>;
+def : InstRW<[Ampere1BWrite_3cyc_1L],
+        (instregex "LDR(HH|SHW|SHX|W|X)ro(W|X)")>;
+def : InstRW<[Ampere1BWrite_1cyc_1L],
+        (instrs PRFMl, PRFUMi, PRFUMi)>;
+def : InstRW<[Ampere1BWrite_1cyc_1L],
+        (instrs PRFMroW, PRFMroX)>;
+
+// Integer miscellaneous instructions
+def : InstRW<[Ampere1BWrite_1cyc_1A],  (instrs ADR, ADRP)>;
+def : InstRW<[Ampere1BWrite_1cyc_1B],  (instregex "EXTR(W|X)")>;
+def : InstRW<[Ampere1BWrite_1cyc_1B],  (instregex "(S|U)?BFM(W|X)")>;
+def : InstRW<[Ampere1BWrite_3cyc_1BS], (instregex "^CRC32C?[BHWX]")>;
+def : InstRW<[Ampere1BWrite_1cyc_1B],  (instregex "CLS(W|X)")>;
+def : InstRW<[Ampere1BWrite_1cyc_1A],  (instrs SETF8, SETF16)>;
+def : InstRW<[Ampere1BWrite_1cyc_1AB],
+        (instrs MOVKWi, MOVKXi, MOVNWi, MOVNXi, MOVZWi, MOVZXi)>;
+def : InstRW<[Ampere1BWrite_1cyc_1B],
+        (instregex "(RBIT|REV|REV16)(W|X)r", "REV32Xr")>;
+def : InstRW<[Ampere1BWrite_1cyc_1B],
+        (instregex "(ASR|LSL|LSR|ROR)V(W|X)r")>;
+
+// Integer store instructions
+def : InstRW<[Ampere1BWrite_1cyc_2S],        (instregex "STNP(X|W)i")>;
+def : InstRW<[Ampere1BWrite_1cyc_2S],        (instrs STPXi)>;
+def : InstRW<[Ampere1BWrite_2cyc_1B_1S],     (instrs STPWi)>;
+def : InstRW<[Ampere1BWrite_2cyc_1B_1S_1AB], (instregex "STP(W|X)(pre|post)")>;
+def : InstRW<[Ampere1BWrite_1cyc_1S],        (instrs STTRBi, STTRHi, STTRWi, STTRXi)>;
+def : InstRW<[Ampere1BWrite_1cyc_1S],        (instregex "STUR(BB|HH|X|W)i",
+                                                        "STR(X|W)ui",
+                                                        "STUR(BB|HH|X|W)i")>;
+def : InstRW<[Ampere1BWrite_1cyc_2S],        (instrs STRWroX, STRXroX)>;
+def : InstRW<[Ampere1BWrite_1cyc_2S],        (instrs STRWroW, STRXroW)>;
+
+// Memory tagging
+
+// Insert Random Tags
+def : InstRW<[Ampere1BWrite_1cyc_1BS_1B], (instrs IRG, IRGstack)>;
+// Load allocation tag
+def : InstRW<[Ampere1BWrite_4cyc_1L_1B], (instrs LDG, LDGM)>;
+// Store allocation tags
+def : InstRW<[Ampere1BWrite_1cyc_1S],
+    (instrs STGi, STGM, STGPreIndex, STGPostIndex)>;
+// Store allocation tags and pair of registers
+def : InstRW<[Ampere1BWrite_1cyc_2S],
+    (instrs STGPi, STGPpre, STGPpost)>;
+// Store allocation tags and zero data
+def : InstRW<[Ampere1BWrite_1cyc_1S],
+    (instrs STZGi, STZGM, STZGPreIndex, STZGPostIndex)>;
+// Store two tags
+def : InstRW<[Ampere1BWrite_1cyc_2S],
+    (instrs ST2Gi, ST2GPreIndex, ST2GPostIndex)>;
+// Store two tags and zero data
+def : InstRW<[Ampere1BWrite_1cyc_2S],
+    (instrs STZ2Gi, STZ2GPreIndex, STZ2GPostIndex)>;
+// Subtract Pointer
+def : InstRW<[Ampere1BWrite_1cyc_1AB], (instrs SUBP)>;
+// Subtract Pointer, flagset
+def : InstRW<[Ampere1BWrite_1cyc_1AB], (instrs SUBPS)>;
+// Insert Tag Mask
+def : InstRW<[Ampere1BWrite_1cyc_1AB], (instrs GMI)>;
+// Arithmetic, immediate to logical address tag
+def : InstRW<[Ampere1BWrite_1cyc_B], (instrs ADDG, SUBG)>;
+
+// Pointer authentication
+def : InstRW<[Ampere1BWrite_5cyc_1BS], (instregex "^AUT")>;
+def : InstRW<[Ampere1BWrite_6cyc_1BS_1A],
+        (instregex "BRA(A|AZ|B|BZ)", "RETA(A|B)", "ERETA(A|B)")>;
+def : InstRW<[Ampere1BWrite_6cyc_1BS_2A],
+        (instrs BLRAA, BLRAAZ, BLRAB, BLRABZ)>;
+def : InstRW<[Ampere1BWrite_5cyc_1BS], (instregex "^PAC")>;
+def : InstRW<[Ampere1BWrite_8cyc_1BS_1L], (instregex "^LDRA(A|B)")>;
+def : InstRW<[Ampere1BWrite_1cyc_1B], (instrs XPACD, XPACI)>;
+
+// Vector integer instructions
+// -- absolute difference
+def : InstRW<[Ampere1BWrite_2cyc_1XY],
+             (instregex "^SABAv", "^SABALv", "^SABDv", "^SABDLv",
+                        "^UABAv", "^UABALv", "^UABDv", "^UABDLv")>;
+// -- arithmetic
+def : InstRW<[Ampere1BWrite_2cyc_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<[Ampere1BWrite_8cyc_4XY],
+            (instregex "^ADDVv16i8v", "^SADDLVv16i8v", "^UADDLVv16i8v")>;
+def : InstRW<[Ampere1BWrite_8cyc_4XY],
+            (instregex "^[SU](MIN|MAX)Vv16i8v")>;
+// -- arithmetic, horizontal, 4H/4S
+def : InstRW<[Ampere1BWrite_4cyc_2XY],
+            (instregex "^[SU]?ADDL?V(v8i8|v4i16|v2i32)v")>;
+def : InstRW<[Ampere1BWrite_4cyc_2XY],
+            (instregex "^[SU](MIN|MAX)V(v4i16|v4i32)v")>;
+// -- arithmetic, horizontal, 8B/8H
+def : InstRW<[Ampere1BWrite_6cyc_3XY],
+            (instregex "^[SU]?ADDL?V(v8i16|v4i32)v")>;
+def : InstRW<[Ampere1BWrite_6cyc_3XY],
+            (instregex "^[SU](MIN|MAX)V(v8i8|v8i16)v")>;
+// -- arithmetic, narrowing
+def : InstRW<[Ampere1BWrite_6cyc_2XY], (instregex "(ADD|SUB)HNv.*")>;
+def : InstRW<[Ampere1BWrite_6cyc_2XY], (instregex "(RADD|RSUB)HNv.*")>;
+// -- arithmetic, pairwise
+def : InstRW<[Ampere1BWrite_2cyc_1XY],
+        (instregex "^ADDPv", "^SADALP", "^UADALP", "^SADDLPv", "^UADDLPv")>;
+// -- arithmetic, saturating
+def : InstRW<[Ampere1BWrite_2cyc_1XY],
+        (instregex "^SQADD", "^SQSUB", "^SUQADD", "^UQADD", "^UQSUB", "^USQADD")>;
+// -- bit count
+def : InstRW<[Ampere1BWrite_2cyc_1XY],
+        (instregex "^(CLS|CLZ|CNT)v")>;
+// -- compare
+def : InstRW<[Ampere1BWrite_2cyc_1XY],
+        (instregex "^CMEQv", "^CMGEv", "^CMGTv", "^CMLEv", "^CMLTv",
+                   "^CMHIv", "^CMHSv")>;
+// -- compare non-zero
+def : InstRW<[Ampere1BWrite_2cyc_1XY], (instregex "^CMTSTv")>;
+// -- dot product
+def : InstRW<[Ampere1BWrite_3cyc_1XY], (instregex "^(S|SU|U|US)DOTv")>;
+// -- fp reciprocal estimate
+def : InstRW<[Ampere1BWrite_6cyc_1X], (instregex "^FRECPEv", "^FRSQRTEv")>;
+// -- integer reciprocal estimate
+def : InstRW<[Ampere1BWrite_2cyc_1XY], (instregex "^URECPEv", "^URSQRTEv")>;
+// -- logical
+def : InstRW<[Ampere1BWrite_2cyc_1XY],
+        (instregex "^ANDv", "^BICv", "^EORv", "^ORRv", "^ORNv", "^NOTv")>;
+// -- logical, narrowing
+def : InstRW<[Ampere1BWrite_6cyc_2XY],
+        (instregex "RSHRNv",
+                   "SHRNv", "SQSHRNv", "SQSHRUNv",
+                   "UQXTNv")>;
+// -- matrix multiply
+def : InstRW<[Ampere1BWrite_3cyc_1XY],
+        (instrs SMMLA, UMMLA, USMMLA)>;
+// -- max/min
+def : InstRW<[Ampere1Write_2cyc_1XY],
+        (instregex "^SMAXv", "^SMINv", "^UMAXv", "^UMINv")>;
+def : InstRW<[Ampere1Write_2cyc_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_2cyc_1XY], (instregex "^[SU]SHL(v16i8|v8i16|v4i32|v2i64)")>;
+// -- shift and accumulate
+def : InstRW<[Ampere1Write_2cyc_1XY],
+        (instregex "SRSRAv", "SSRAv", "URSRAv", "USRAv")>;
+// -- shift, saturating
+def : InstRW<[Ampere1Write_2cyc_1XY],
+        (instregex "^SQRSHLv", "^SQRSHRNv", "^SQRSHRUNv", "^SQSHL", "^SQSHLU",
+                   "^SQXTNv", "^SQXTUNv", "^UQSHRNv", "UQRSHRNv", "^UQRSHL",
+                   "^UQSHL")>;
+
+// Vector miscellaneous instructions
+// -- duplicate element
+def : InstRW<[Ampere1BWrite_2cyc_1XY], (instregex "^DUPv.+lane")>;
+// -- duplicate from GPR
+def : InstRW<[Ampere1BWrite_5cyc_1BS], (instregex "^DUPv.+gpr")>;
+// -- extract narrow
+def : InstRW<[Ampere1BWrite_2cyc_1XY], (instregex "^XTNv")>;
+// -- insert/extract element
+def : InstRW<[Ampere1BWrite_2cyc_1XY], (instregex "^EXTv", "^INSv.+lane")>;
+// -- move FP immediate
+def : InstRW<[Ampere1BWrite_2cyc_1XY], (instregex "^FMOVv")>;
+// -- move element to GPR
+def : InstRW<[Ampere1BWrite_5cyc_1X], (instregex "(S|U)MOVv")>;
+// -- move from GPR to any element
+def : InstRW<[Ampere1BWrite_7cyc_1BS_1XY], (instregex "^INSv.+gpr")>;
+// -- table lookup
+def : InstRW<[Ampere1BWrite_2cyc_1XY],
+            (instrs TBLv8i8One, TBLv16i8One, TBXv8i8One, TBXv16i8One)>;
+def : InstRW<[Ampere1BWrite_4cyc_2XY],
+            (instrs TBLv8i8Two, TBLv16i8Two, TBXv8i8Two, TBXv16i8Two)>;
+def : InstRW<[Ampere1BWrite_6cyc_3XY],
+            (instrs TBLv8i8Three, TBLv16i8Three, TBXv8i8Three, TBXv16i8Three)>;
+def : InstRW<[Ampere1BWrite_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 = Ampere1BModel



More information about the llvm-commits mailing list