296 lines
9.4 KiB
C++
296 lines
9.4 KiB
C++
//===- X86EvexToVex.cpp ---------------------------------------------------===//
|
|
// Compress EVEX instructions to VEX encoding when possible to reduce code size
|
|
//
|
|
// 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
|
|
//
|
|
//===----------------------------------------------------------------------===//
|
|
//
|
|
/// \file
|
|
/// This file defines the pass that goes over all AVX-512 instructions which
|
|
/// are encoded using the EVEX prefix and if possible replaces them by their
|
|
/// corresponding VEX encoding which is usually shorter by 2 bytes.
|
|
/// EVEX instructions may be encoded via the VEX prefix when the AVX-512
|
|
/// instruction has a corresponding AVX/AVX2 opcode, when vector length
|
|
/// accessed by instruction is less than 512 bits and when it does not use
|
|
// the xmm or the mask registers or xmm/ymm registers with indexes higher than 15.
|
|
/// The pass applies code reduction on the generated code for AVX-512 instrs.
|
|
//
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
#include "MCTargetDesc/X86BaseInfo.h"
|
|
#include "MCTargetDesc/X86InstComments.h"
|
|
#include "X86.h"
|
|
#include "X86InstrInfo.h"
|
|
#include "X86Subtarget.h"
|
|
#include "llvm/ADT/StringRef.h"
|
|
#include "llvm/CodeGen/MachineFunction.h"
|
|
#include "llvm/CodeGen/MachineFunctionPass.h"
|
|
#include "llvm/CodeGen/MachineInstr.h"
|
|
#include "llvm/CodeGen/MachineOperand.h"
|
|
#include "llvm/MC/MCInstrDesc.h"
|
|
#include "llvm/Pass.h"
|
|
#include <cassert>
|
|
#include <cstdint>
|
|
|
|
using namespace llvm;
|
|
|
|
// Including the generated EVEX2VEX tables.
|
|
struct X86EvexToVexCompressTableEntry {
|
|
uint16_t EvexOpcode;
|
|
uint16_t VexOpcode;
|
|
|
|
bool operator<(const X86EvexToVexCompressTableEntry &RHS) const {
|
|
return EvexOpcode < RHS.EvexOpcode;
|
|
}
|
|
|
|
friend bool operator<(const X86EvexToVexCompressTableEntry &TE,
|
|
unsigned Opc) {
|
|
return TE.EvexOpcode < Opc;
|
|
}
|
|
};
|
|
#include "X86GenEVEX2VEXTables.inc"
|
|
|
|
#define EVEX2VEX_DESC "Compressing EVEX instrs to VEX encoding when possible"
|
|
#define EVEX2VEX_NAME "x86-evex-to-vex-compress"
|
|
|
|
#define DEBUG_TYPE EVEX2VEX_NAME
|
|
|
|
namespace {
|
|
|
|
class EvexToVexInstPass : public MachineFunctionPass {
|
|
|
|
/// For EVEX instructions that can be encoded using VEX encoding, replace
|
|
/// them by the VEX encoding in order to reduce size.
|
|
bool CompressEvexToVexImpl(MachineInstr &MI) const;
|
|
|
|
public:
|
|
static char ID;
|
|
|
|
EvexToVexInstPass() : MachineFunctionPass(ID) { }
|
|
|
|
StringRef getPassName() const override { return EVEX2VEX_DESC; }
|
|
|
|
/// Loop over all of the basic blocks, replacing EVEX instructions
|
|
/// by equivalent VEX instructions when possible for reducing code size.
|
|
bool runOnMachineFunction(MachineFunction &MF) override;
|
|
|
|
// This pass runs after regalloc and doesn't support VReg operands.
|
|
MachineFunctionProperties getRequiredProperties() const override {
|
|
return MachineFunctionProperties().set(
|
|
MachineFunctionProperties::Property::NoVRegs);
|
|
}
|
|
|
|
private:
|
|
/// Machine instruction info used throughout the class.
|
|
const X86InstrInfo *TII = nullptr;
|
|
|
|
const X86Subtarget *ST = nullptr;
|
|
};
|
|
|
|
} // end anonymous namespace
|
|
|
|
char EvexToVexInstPass::ID = 0;
|
|
|
|
bool EvexToVexInstPass::runOnMachineFunction(MachineFunction &MF) {
|
|
TII = MF.getSubtarget<X86Subtarget>().getInstrInfo();
|
|
|
|
ST = &MF.getSubtarget<X86Subtarget>();
|
|
if (!ST->hasAVX512())
|
|
return false;
|
|
|
|
bool Changed = false;
|
|
|
|
/// Go over all basic blocks in function and replace
|
|
/// EVEX encoded instrs by VEX encoding when possible.
|
|
for (MachineBasicBlock &MBB : MF) {
|
|
|
|
// Traverse the basic block.
|
|
for (MachineInstr &MI : MBB)
|
|
Changed |= CompressEvexToVexImpl(MI);
|
|
}
|
|
|
|
return Changed;
|
|
}
|
|
|
|
static bool usesExtendedRegister(const MachineInstr &MI) {
|
|
auto isHiRegIdx = [](unsigned Reg) {
|
|
// Check for XMM register with indexes between 16 - 31.
|
|
if (Reg >= X86::XMM16 && Reg <= X86::XMM31)
|
|
return true;
|
|
|
|
// Check for YMM register with indexes between 16 - 31.
|
|
if (Reg >= X86::YMM16 && Reg <= X86::YMM31)
|
|
return true;
|
|
|
|
return false;
|
|
};
|
|
|
|
// Check that operands are not ZMM regs or
|
|
// XMM/YMM regs with hi indexes between 16 - 31.
|
|
for (const MachineOperand &MO : MI.explicit_operands()) {
|
|
if (!MO.isReg())
|
|
continue;
|
|
|
|
Register Reg = MO.getReg();
|
|
|
|
assert(!(Reg >= X86::ZMM0 && Reg <= X86::ZMM31) &&
|
|
"ZMM instructions should not be in the EVEX->VEX tables");
|
|
|
|
if (isHiRegIdx(Reg))
|
|
return true;
|
|
}
|
|
|
|
return false;
|
|
}
|
|
|
|
// Do any custom cleanup needed to finalize the conversion.
|
|
static bool performCustomAdjustments(MachineInstr &MI, unsigned NewOpc,
|
|
const X86Subtarget *ST) {
|
|
(void)NewOpc;
|
|
unsigned Opc = MI.getOpcode();
|
|
switch (Opc) {
|
|
case X86::VPDPBUSDSZ256m:
|
|
case X86::VPDPBUSDSZ256r:
|
|
case X86::VPDPBUSDSZ128m:
|
|
case X86::VPDPBUSDSZ128r:
|
|
case X86::VPDPBUSDZ256m:
|
|
case X86::VPDPBUSDZ256r:
|
|
case X86::VPDPBUSDZ128m:
|
|
case X86::VPDPBUSDZ128r:
|
|
case X86::VPDPWSSDSZ256m:
|
|
case X86::VPDPWSSDSZ256r:
|
|
case X86::VPDPWSSDSZ128m:
|
|
case X86::VPDPWSSDSZ128r:
|
|
case X86::VPDPWSSDZ256m:
|
|
case X86::VPDPWSSDZ256r:
|
|
case X86::VPDPWSSDZ128m:
|
|
case X86::VPDPWSSDZ128r:
|
|
// These can only VEX convert if AVXVNNI is enabled.
|
|
return ST->hasAVXVNNI();
|
|
case X86::VALIGNDZ128rri:
|
|
case X86::VALIGNDZ128rmi:
|
|
case X86::VALIGNQZ128rri:
|
|
case X86::VALIGNQZ128rmi: {
|
|
assert((NewOpc == X86::VPALIGNRrri || NewOpc == X86::VPALIGNRrmi) &&
|
|
"Unexpected new opcode!");
|
|
unsigned Scale = (Opc == X86::VALIGNQZ128rri ||
|
|
Opc == X86::VALIGNQZ128rmi) ? 8 : 4;
|
|
MachineOperand &Imm = MI.getOperand(MI.getNumExplicitOperands()-1);
|
|
Imm.setImm(Imm.getImm() * Scale);
|
|
break;
|
|
}
|
|
case X86::VSHUFF32X4Z256rmi:
|
|
case X86::VSHUFF32X4Z256rri:
|
|
case X86::VSHUFF64X2Z256rmi:
|
|
case X86::VSHUFF64X2Z256rri:
|
|
case X86::VSHUFI32X4Z256rmi:
|
|
case X86::VSHUFI32X4Z256rri:
|
|
case X86::VSHUFI64X2Z256rmi:
|
|
case X86::VSHUFI64X2Z256rri: {
|
|
assert((NewOpc == X86::VPERM2F128rr || NewOpc == X86::VPERM2I128rr ||
|
|
NewOpc == X86::VPERM2F128rm || NewOpc == X86::VPERM2I128rm) &&
|
|
"Unexpected new opcode!");
|
|
MachineOperand &Imm = MI.getOperand(MI.getNumExplicitOperands()-1);
|
|
int64_t ImmVal = Imm.getImm();
|
|
// Set bit 5, move bit 1 to bit 4, copy bit 0.
|
|
Imm.setImm(0x20 | ((ImmVal & 2) << 3) | (ImmVal & 1));
|
|
break;
|
|
}
|
|
case X86::VRNDSCALEPDZ128rri:
|
|
case X86::VRNDSCALEPDZ128rmi:
|
|
case X86::VRNDSCALEPSZ128rri:
|
|
case X86::VRNDSCALEPSZ128rmi:
|
|
case X86::VRNDSCALEPDZ256rri:
|
|
case X86::VRNDSCALEPDZ256rmi:
|
|
case X86::VRNDSCALEPSZ256rri:
|
|
case X86::VRNDSCALEPSZ256rmi:
|
|
case X86::VRNDSCALESDZr:
|
|
case X86::VRNDSCALESDZm:
|
|
case X86::VRNDSCALESSZr:
|
|
case X86::VRNDSCALESSZm:
|
|
case X86::VRNDSCALESDZr_Int:
|
|
case X86::VRNDSCALESDZm_Int:
|
|
case X86::VRNDSCALESSZr_Int:
|
|
case X86::VRNDSCALESSZm_Int:
|
|
const MachineOperand &Imm = MI.getOperand(MI.getNumExplicitOperands()-1);
|
|
int64_t ImmVal = Imm.getImm();
|
|
// Ensure that only bits 3:0 of the immediate are used.
|
|
if ((ImmVal & 0xf) != ImmVal)
|
|
return false;
|
|
break;
|
|
}
|
|
|
|
return true;
|
|
}
|
|
|
|
|
|
// For EVEX instructions that can be encoded using VEX encoding
|
|
// replace them by the VEX encoding in order to reduce size.
|
|
bool EvexToVexInstPass::CompressEvexToVexImpl(MachineInstr &MI) const {
|
|
// VEX format.
|
|
// # of bytes: 0,2,3 1 1 0,1 0,1,2,4 0,1
|
|
// [Prefixes] [VEX] OPCODE ModR/M [SIB] [DISP] [IMM]
|
|
//
|
|
// EVEX format.
|
|
// # of bytes: 4 1 1 1 4 / 1 1
|
|
// [Prefixes] EVEX Opcode ModR/M [SIB] [Disp32] / [Disp8*N] [Immediate]
|
|
|
|
const MCInstrDesc &Desc = MI.getDesc();
|
|
|
|
// Check for EVEX instructions only.
|
|
if ((Desc.TSFlags & X86II::EncodingMask) != X86II::EVEX)
|
|
return false;
|
|
|
|
// Check for EVEX instructions with mask or broadcast as in these cases
|
|
// the EVEX prefix is needed in order to carry this information
|
|
// thus preventing the transformation to VEX encoding.
|
|
if (Desc.TSFlags & (X86II::EVEX_K | X86II::EVEX_B))
|
|
return false;
|
|
|
|
// Check for EVEX instructions with L2 set. These instructions are 512-bits
|
|
// and can't be converted to VEX.
|
|
if (Desc.TSFlags & X86II::EVEX_L2)
|
|
return false;
|
|
|
|
#ifndef NDEBUG
|
|
// Make sure the tables are sorted.
|
|
static std::atomic<bool> TableChecked(false);
|
|
if (!TableChecked.load(std::memory_order_relaxed)) {
|
|
assert(llvm::is_sorted(X86EvexToVex128CompressTable) &&
|
|
"X86EvexToVex128CompressTable is not sorted!");
|
|
assert(llvm::is_sorted(X86EvexToVex256CompressTable) &&
|
|
"X86EvexToVex256CompressTable is not sorted!");
|
|
TableChecked.store(true, std::memory_order_relaxed);
|
|
}
|
|
#endif
|
|
|
|
// Use the VEX.L bit to select the 128 or 256-bit table.
|
|
ArrayRef<X86EvexToVexCompressTableEntry> Table =
|
|
(Desc.TSFlags & X86II::VEX_L) ? makeArrayRef(X86EvexToVex256CompressTable)
|
|
: makeArrayRef(X86EvexToVex128CompressTable);
|
|
|
|
const auto *I = llvm::lower_bound(Table, MI.getOpcode());
|
|
if (I == Table.end() || I->EvexOpcode != MI.getOpcode())
|
|
return false;
|
|
|
|
unsigned NewOpc = I->VexOpcode;
|
|
|
|
if (usesExtendedRegister(MI))
|
|
return false;
|
|
|
|
if (!performCustomAdjustments(MI, NewOpc, ST))
|
|
return false;
|
|
|
|
MI.setDesc(TII->get(NewOpc));
|
|
MI.setAsmPrinterFlag(X86::AC_EVEX_2_VEX);
|
|
return true;
|
|
}
|
|
|
|
INITIALIZE_PASS(EvexToVexInstPass, EVEX2VEX_NAME, EVEX2VEX_DESC, false, false)
|
|
|
|
FunctionPass *llvm::createX86EvexToVexInsts() {
|
|
return new EvexToVexInstPass();
|
|
}
|