//===----------- PPCVSXSwapRemoval.cpp - Remove VSX LE Swaps -------------===//
//
// 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 pass analyzes vector computations and removes unnecessary
// doubleword swaps (xxswapd instructions). This pass is performed
// only for little-endian VSX code generation.
//
// For this specific case, loads and stores of v4i32, v4f32, v2i64,
// and v2f64 vectors are inefficient. These are implemented using
// the lxvd2x and stxvd2x instructions, which invert the order of
// doublewords in a vector register. Thus code generation inserts
// an xxswapd after each such load, and prior to each such store.
//
// The extra xxswapd instructions reduce performance. The purpose
// of this pass is to reduce the number of xxswapd instructions
// required for correctness.
//
// The primary insight is that much code that operates on vectors
// does not care about the relative order of elements in a register,
// so long as the correct memory order is preserved. If we have a
// computation where all input values are provided by lxvd2x/xxswapd,
// all outputs are stored using xxswapd/lxvd2x, and all intermediate
// computations are lane-insensitive (independent of element order),
// then all the xxswapd instructions associated with the loads and
// stores may be removed without changing observable semantics.
//
// This pass uses standard equivalence class infrastructure to create
// maximal webs of computations fitting the above description. Each
// such web is then optimized by removing its unnecessary xxswapd
// instructions.
//
// There are some lane-sensitive operations for which we can still
// permit the optimization, provided we modify those operations
// accordingly. Such operations are identified as using "special
// handling" within this module.
//
//===---------------------------------------------------------------------===//
#include "PPC.h"
#include "PPCInstrBuilder.h"
#include "PPCInstrInfo.h"
#include "PPCTargetMachine.h"
#include "llvm/ADT/DenseMap.h"
#include "llvm/ADT/EquivalenceClasses.h"
#include "llvm/CodeGen/MachineFunctionPass.h"
#include "llvm/CodeGen/MachineInstrBuilder.h"
#include "llvm/CodeGen/MachineRegisterInfo.h"
#include "llvm/Config/llvm-config.h"
#include "llvm/Support/Debug.h"
#include "llvm/Support/Format.h"
#include "llvm/Support/raw_ostream.h"
using namespace llvm;
#define DEBUG_TYPE "ppc-vsx-swaps"
namespace {
// A PPCVSXSwapEntry is created for each machine instruction that
// is relevant to a vector computation.
struct PPCVSXSwapEntry {
// Pointer to the instruction.
MachineInstr *VSEMI;
// Unique ID (position in the swap vector).
int VSEId;
// Attributes of this node.
unsigned int IsLoad : 1;
unsigned int IsStore : 1;
unsigned int IsSwap : 1;
unsigned int MentionsPhysVR : 1;
unsigned int IsSwappable : 1;
unsigned int MentionsPartialVR : 1;
unsigned int SpecialHandling : 3;
unsigned int WebRejected : 1;
unsigned int WillRemove : 1;
};
enum SHValues {
SH_NONE = 0,
SH_EXTRACT,
SH_INSERT,
SH_NOSWAP_LD,
SH_NOSWAP_ST,
SH_SPLAT,
SH_XXPERMDI,
SH_COPYWIDEN
};
struct PPCVSXSwapRemoval : public MachineFunctionPass {
static char ID;
const PPCInstrInfo *TII;
MachineFunction *MF;
MachineRegisterInfo *MRI;
// Swap entries are allocated in a vector for better performance.
std::vector<PPCVSXSwapEntry> SwapVector;
// A mapping is maintained between machine instructions and
// their swap entries. The key is the address of the MI.
DenseMap<MachineInstr*, int> SwapMap;
// Equivalence classes are used to gather webs of related computation.
// Swap entries are represented by their VSEId fields.
EquivalenceClasses<int> *EC;
PPCVSXSwapRemoval() : MachineFunctionPass(ID) {
initializePPCVSXSwapRemovalPass(*PassRegistry::getPassRegistry());
}
private:
// Initialize data structures.
void initialize(MachineFunction &MFParm);
// Walk the machine instructions to gather vector usage information.
// Return true iff vector mentions are present.
bool gatherVectorInstructions();
// Add an entry to the swap vector and swap map.
int addSwapEntry(MachineInstr *MI, PPCVSXSwapEntry &SwapEntry);
// Hunt backwards through COPY and SUBREG_TO_REG chains for a
// source register. VecIdx indicates the swap vector entry to
// mark as mentioning a physical register if the search leads
// to one.
unsigned lookThruCopyLike(unsigned SrcReg, unsigned VecIdx);
// Generate equivalence classes for related computations (webs).
void formWebs();
// Analyze webs and determine those that cannot be optimized.
void recordUnoptimizableWebs();
// Record which swap instructions can be safely removed.
void markSwapsForRemoval();
// Remove swaps and update other instructions requiring special
// handling. Return true iff any changes are made.
bool removeSwaps();
// Insert a swap instruction from SrcReg to DstReg at the given
// InsertPoint.
void insertSwap(MachineInstr *MI, MachineBasicBlock::iterator InsertPoint,
unsigned DstReg, unsigned SrcReg);
// Update instructions requiring special handling.
void handleSpecialSwappables(int EntryIdx);
// Dump a description of the entries in the swap vector.
void dumpSwapVector();
// Return true iff the given register is in the given class.
bool isRegInClass(unsigned Reg, const TargetRegisterClass *RC) {
if (Register::isVirtualRegister(Reg))
return RC->hasSubClassEq(MRI->getRegClass(Reg));
return RC->contains(Reg);
}
// Return true iff the given register is a full vector register.
bool isVecReg(unsigned Reg) {
return (isRegInClass(Reg, &PPC::VSRCRegClass) ||
isRegInClass(Reg, &PPC::VRRCRegClass));
}
// Return true iff the given register is a partial vector register.
bool isScalarVecReg(unsigned Reg) {
return (isRegInClass(Reg, &PPC::VSFRCRegClass) ||
isRegInClass(Reg, &PPC::VSSRCRegClass));
}
// Return true iff the given register mentions all or part of a
// vector register. Also sets Partial to true if the mention
// is for just the floating-point register overlap of the register.
bool isAnyVecReg(unsigned Reg, bool &Partial) {
if (isScalarVecReg(Reg))
Partial = true;
return isScalarVecReg(Reg) || isVecReg(Reg);
}
public:
// Main entry point for this pass.
bool runOnMachineFunction(MachineFunction &MF) override {
if (skipFunction(MF.getFunction()))
return false;
// If we don't have VSX on the subtarget, don't do anything.
// Also, on Power 9 the load and store ops preserve element order and so
// the swaps are not required.
const PPCSubtarget &STI = MF.getSubtarget<PPCSubtarget>();
if (!STI.hasVSX() || !STI.needsSwapsForVSXMemOps())
return false;
bool Changed = false;
initialize(MF);
if (gatherVectorInstructions()) {
formWebs();
recordUnoptimizableWebs();
markSwapsForRemoval();
Changed = removeSwaps();
}
// FIXME: See the allocation of EC in initialize().
delete EC;
return Changed;
}
};
// Initialize data structures for this pass. In particular, clear the
// swap vector and allocate the equivalence class mapping before
// processing each function.
void PPCVSXSwapRemoval::initialize(MachineFunction &MFParm) {
MF = &MFParm;
MRI = &MF->getRegInfo();
TII = MF->getSubtarget<PPCSubtarget>().getInstrInfo();
// An initial vector size of 256 appears to work well in practice.
// Small/medium functions with vector content tend not to incur a
// reallocation at this size. Three of the vector tests in
// projects/test-suite reallocate, which seems like a reasonable rate.
const int InitialVectorSize(256);
SwapVector.clear();
SwapVector.reserve(InitialVectorSize);
// FIXME: Currently we allocate EC each time because we don't have
// access to the set representation on which to call clear(). Should
// consider adding a clear() method to the EquivalenceClasses class.
EC = new EquivalenceClasses<int>;
}
// Create an entry in the swap vector for each instruction that mentions
// a full vector register, recording various characteristics of the
// instructions there.
bool PPCVSXSwapRemoval::gatherVectorInstructions() {
bool RelevantFunction = false;
for (MachineBasicBlock &MBB : *MF) {
for (MachineInstr &MI : MBB) {
if (MI.isDebugInstr())
continue;
bool RelevantInstr = false;
bool Partial = false;
for (const MachineOperand &MO : MI.operands()) {
if (!MO.isReg())
continue;
Register Reg = MO.getReg();
if (isAnyVecReg(Reg, Partial)) {
RelevantInstr = true;
break;
}
}
if (!RelevantInstr)
continue;
RelevantFunction = true;
// Create a SwapEntry initialized to zeros, then fill in the
// instruction and ID fields before pushing it to the back
// of the swap vector.
PPCVSXSwapEntry SwapEntry{};
int VecIdx = addSwapEntry(&MI, SwapEntry);
switch(MI.getOpcode()) {
default:
// Unless noted otherwise, an instruction is considered
// safe for the optimization. There are a large number of
// such true-SIMD instructions (all vector math, logical,
// select, compare, etc.). However, if the instruction
// mentions a partial vector register and does not have
// special handling defined, it is not swappable.
if (Partial)
SwapVector[VecIdx].MentionsPartialVR = 1;
else
SwapVector[VecIdx].IsSwappable = 1;
break;
case PPC::XXPERMDI: {
// This is a swap if it is of the form XXPERMDI t, s, s, 2.
// Unfortunately, MachineCSE ignores COPY and SUBREG_TO_REG, so we
// can also see XXPERMDI t, SUBREG_TO_REG(s), SUBREG_TO_REG(s), 2,
// for example. We have to look through chains of COPY and
// SUBREG_TO_REG to find the real source value for comparison.
// If the real source value is a physical register, then mark the
// XXPERMDI as mentioning a physical register.
int immed = MI.getOperand(3).getImm();
if (immed == 2) {
unsigned trueReg1 = lookThruCopyLike(MI.getOperand(1).getReg(),
VecIdx);
unsigned trueReg2 = lookThruCopyLike(MI.getOperand(2).getReg(),
VecIdx);
if (trueReg1 == trueReg2)
SwapVector[VecIdx].IsSwap = 1;
else {
// We can still handle these if the two registers are not
// identical, by adjusting the form of the XXPERMDI.
SwapVector[VecIdx].IsSwappable = 1;
SwapVector[VecIdx].SpecialHandling = SHValues::SH_XXPERMDI;
}
// This is a doubleword splat if it is of the form
// XXPERMDI t, s, s, 0 or XXPERMDI t, s, s, 3. As above we
// must look through chains of copy-likes to find the source
// register. We turn off the marking for mention of a physical
// register, because splatting it is safe; the optimization
// will not swap the value in the physical register. Whether
// or not the two input registers are identical, we can handle
// these by adjusting the form of the XXPERMDI.
} else if (immed == 0 || immed == 3) {
SwapVector[VecIdx].IsSwappable = 1;
SwapVector[VecIdx].SpecialHandling = SHValues::SH_XXPERMDI;
unsigned trueReg1 = lookThruCopyLike(MI.getOperand(1).getReg(),
VecIdx);
unsigned trueReg2 = lookThruCopyLike(MI.getOperand(2).getReg(),
VecIdx);
if (trueReg1 == trueReg2)
SwapVector[VecIdx].MentionsPhysVR = 0;
} else {
// We can still handle these by adjusting the form of the XXPERMDI.
SwapVector[VecIdx].IsSwappable = 1;
SwapVector[VecIdx].SpecialHandling = SHValues::SH_XXPERMDI;
}
break;
}
case PPC::LVX:
// Non-permuting loads are currently unsafe. We can use special
// handling for this in the future. By not marking these as
// IsSwap, we ensure computations containing them will be rejected
// for now.
SwapVector[VecIdx].IsLoad = 1;
break;
case PPC::LXVD2X:
case PPC::LXVW4X:
// Permuting loads are marked as both load and swap, and are
// safe for optimization.
SwapVector[VecIdx].IsLoad = 1;
SwapVector[VecIdx].IsSwap = 1;
break;
case PPC::LXSDX:
case PPC::LXSSPX:
case PPC::XFLOADf64:
case PPC::XFLOADf32:
// A load of a floating-point value into the high-order half of
// a vector register is safe, provided that we introduce a swap
// following the load, which will be done by the SUBREG_TO_REG
// support. So just mark these as safe.
SwapVector[VecIdx].IsLoad = 1;
SwapVector[VecIdx].IsSwappable = 1;
break;
case PPC::STVX:
// Non-permuting stores are currently unsafe. We can use special
// handling for this in the future. By not marking these as
// IsSwap, we ensure computations containing them will be rejected
// for now.
SwapVector[VecIdx].IsStore = 1;
break;
case PPC::STXVD2X:
case PPC::STXVW4X:
// Permuting stores are marked as both store and swap, and are
// safe for optimization.
SwapVector[VecIdx].IsStore = 1;
SwapVector[VecIdx].IsSwap = 1;
break;
case PPC::COPY:
// These are fine provided they are moving between full vector
// register classes.
if (isVecReg(MI.getOperand(0).getReg()) &&
isVecReg(MI.getOperand(1).getReg()))
SwapVector[VecIdx].IsSwappable = 1;
// If we have a copy from one scalar floating-point register
// to another, we can accept this even if it is a physical
// register. The only way this gets involved is if it feeds
// a SUBREG_TO_REG, which is handled by introducing a swap.
else if (isScalarVecReg(MI.getOperand(0).getReg()) &&
isScalarVecReg(MI.getOperand(1).getReg()))
SwapVector[VecIdx].IsSwappable = 1;
break;
case PPC::SUBREG_TO_REG: {
// These are fine provided they are moving between full vector
// register classes. If they are moving from a scalar
// floating-point class to a vector class, we can handle those
// as well, provided we introduce a swap. It is generally the
// case that we will introduce fewer swaps than we remove, but
// (FIXME) a cost model could be used. However, introduced
// swaps could potentially be CSEd, so this is not trivial.
if (isVecReg(MI.getOperand(0).getReg()) &&
isVecReg(MI.getOperand(2).getReg()))
SwapVector[VecIdx].IsSwappable = 1;
else if (isVecReg(MI.getOperand(0).getReg()) &&
isScalarVecReg(MI.getOperand(2).getReg())) {
SwapVector[VecIdx].IsSwappable = 1;
SwapVector[VecIdx].SpecialHandling = SHValues::SH_COPYWIDEN;
}
break;
}
case PPC::VSPLTB:
case PPC::VSPLTH:
case PPC::VSPLTW:
case PPC::XXSPLTW:
// Splats are lane-sensitive, but we can use special handling
// to adjust the source lane for the splat.
SwapVector[VecIdx].IsSwappable = 1;
SwapVector[VecIdx].SpecialHandling = SHValues::SH_SPLAT;
break;
// The presence of the following lane-sensitive operations in a
// web will kill the optimization, at least for now. For these
// we do nothing, causing the optimization to fail.
// FIXME: Some of these could be permitted with special handling,
// and will be phased in as time permits.
// FIXME: There is no simple and maintainable way to express a set
// of opcodes having a common attribute in TableGen. Should this
// change, this is a prime candidate to use such a mechanism.
case PPC::INLINEASM:
case PPC::INLINEASM_BR:
case PPC::EXTRACT_SUBREG:
case PPC::INSERT_SUBREG:
case PPC::COPY_TO_REGCLASS:
case PPC::LVEBX:
case PPC::LVEHX:
case PPC::LVEWX:
case PPC::LVSL:
case PPC::LVSR:
case PPC::LVXL:
case PPC::STVEBX:
case PPC::STVEHX:
case PPC::STVEWX:
case PPC::STVXL:
// We can handle STXSDX and STXSSPX similarly to LXSDX and LXSSPX,
// by adding special handling for narrowing copies as well as
// widening ones. However, I've experimented with this, and in
// practice we currently do not appear to use STXSDX fed by
// a narrowing copy from a full vector register. Since I can't
// generate any useful test cases, I've left this alone for now.
case PPC::STXSDX:
case PPC::STXSSPX:
case PPC::VCIPHER:
case PPC::VCIPHERLAST:
case PPC::VMRGHB:
case PPC::VMRGHH:
case PPC::VMRGHW:
case PPC::VMRGLB:
case PPC::VMRGLH:
case PPC::VMRGLW:
case PPC::VMULESB:
case PPC::VMULESH:
case PPC::VMULESW:
case PPC::VMULEUB:
case PPC::VMULEUH:
case PPC::VMULEUW:
case PPC::VMULOSB:
case PPC::VMULOSH:
case PPC::VMULOSW:
case PPC::VMULOUB:
case PPC::VMULOUH:
case PPC::VMULOUW:
case PPC::VNCIPHER:
case PPC::VNCIPHERLAST:
case PPC::VPERM:
case PPC::VPERMXOR:
case PPC::VPKPX:
case PPC::VPKSHSS:
case PPC::VPKSHUS:
case PPC::VPKSDSS:
case PPC::VPKSDUS:
case PPC::VPKSWSS:
case PPC::VPKSWUS:
case PPC::VPKUDUM:
case PPC::VPKUDUS:
case PPC::VPKUHUM:
case PPC::VPKUHUS:
case PPC::VPKUWUM:
case PPC::VPKUWUS:
case PPC::VPMSUMB:
case PPC::VPMSUMD:
case PPC::VPMSUMH:
case PPC::VPMSUMW:
case PPC::VRLB:
case PPC::VRLD:
case PPC::VRLH:
case PPC::VRLW:
case PPC::VSBOX:
case PPC::VSHASIGMAD:
case PPC::VSHASIGMAW:
case PPC::VSL:
case PPC::VSLDOI:
case PPC::VSLO:
case PPC::VSR:
case PPC::VSRO:
case PPC::VSUM2SWS:
case PPC::VSUM4SBS:
case PPC::VSUM4SHS:
case PPC::VSUM4UBS:
case PPC::VSUMSWS:
case PPC::VUPKHPX:
case PPC::VUPKHSB:
case PPC::VUPKHSH:
case PPC::VUPKHSW:
case PPC::VUPKLPX:
case PPC::VUPKLSB:
case PPC::VUPKLSH:
case PPC::VUPKLSW:
case PPC::XXMRGHW:
case PPC::XXMRGLW:
// XXSLDWI could be replaced by a general permute with one of three
// permute control vectors (for shift values 1, 2, 3). However,
// VPERM has a more restrictive register class.
case PPC::XXSLDWI:
case PPC::XSCVDPSPN:
case PPC::XSCVSPDPN:
break;
}
}
}
if (RelevantFunction) {
LLVM_DEBUG(dbgs() << "Swap vector when first built\n\n");
LLVM_DEBUG(dumpSwapVector());
}
return RelevantFunction;
}
// Add an entry to the swap vector and swap map, and make a
// singleton equivalence class for the entry.
int PPCVSXSwapRemoval::addSwapEntry(MachineInstr *MI,
PPCVSXSwapEntry& SwapEntry) {
SwapEntry.VSEMI = MI;
SwapEntry.VSEId = SwapVector.size();
SwapVector.push_back(SwapEntry);
EC->insert(SwapEntry.VSEId);
SwapMap[MI] = SwapEntry.VSEId;
return SwapEntry.VSEId;
}
// This is used to find the "true" source register for an
// XXPERMDI instruction, since MachineCSE does not handle the
// "copy-like" operations (Copy and SubregToReg). Returns
// the original SrcReg unless it is the target of a copy-like
// operation, in which case we chain backwards through all
// such operations to the ultimate source register. If a
// physical register is encountered, we stop the search and
// flag the swap entry indicated by VecIdx (the original
// XXPERMDI) as mentioning a physical register.
unsigned PPCVSXSwapRemoval::lookThruCopyLike(unsigned SrcReg,
unsigned VecIdx) {
MachineInstr *MI = MRI->getVRegDef(SrcReg);
if (!MI->isCopyLike())
return SrcReg;
unsigned CopySrcReg;
if (MI->isCopy())
CopySrcReg = MI->getOperand(1).getReg();
else {
assert(MI->isSubregToReg() && "bad opcode for lookThruCopyLike");
CopySrcReg = MI->getOperand(2).getReg();
}
if (!Register::isVirtualRegister(CopySrcReg)) {
if (!isScalarVecReg(CopySrcReg))
SwapVector[VecIdx].MentionsPhysVR = 1;
return CopySrcReg;
}
return lookThruCopyLike(CopySrcReg, VecIdx);
}
// Generate equivalence classes for related computations (webs) by
// def-use relationships of virtual registers. Mention of a physical
// register terminates the generation of equivalence classes as this
// indicates a use of a parameter, definition of a return value, use
// of a value returned from a call, or definition of a parameter to a
// call. Computations with physical register mentions are flagged
// as such so their containing webs will not be optimized.
void PPCVSXSwapRemoval::formWebs() {
LLVM_DEBUG(dbgs() << "\n*** Forming webs for swap removal ***\n\n");
for (unsigned EntryIdx = 0; EntryIdx < SwapVector.size(); ++EntryIdx) {
MachineInstr *MI = SwapVector[EntryIdx].VSEMI;
LLVM_DEBUG(dbgs() << "\n" << SwapVector[EntryIdx].VSEId << " ");
LLVM_DEBUG(MI->dump());
// It's sufficient to walk vector uses and join them to their unique
// definitions. In addition, check full vector register operands
// for physical regs. We exclude partial-vector register operands
// because we can handle them if copied to a full vector.
for (const MachineOperand &MO : MI->operands()) {
if (!MO.isReg())
continue;
Register Reg = MO.getReg();
if (!isVecReg(Reg) && !isScalarVecReg(Reg))
continue;
if (!Register::isVirtualRegister(Reg)) {
if (!(MI->isCopy() && isScalarVecReg(Reg)))
SwapVector[EntryIdx].MentionsPhysVR = 1;
continue;
}
if (!MO.isUse())
continue;
MachineInstr* DefMI = MRI->getVRegDef(Reg);
assert(SwapMap.find(DefMI) != SwapMap.end() &&
"Inconsistency: def of vector reg not found in swap map!");
int DefIdx = SwapMap[DefMI];
(void)EC->unionSets(SwapVector[DefIdx].VSEId,
SwapVector[EntryIdx].VSEId);
LLVM_DEBUG(dbgs() << format("Unioning %d with %d\n",
SwapVector[DefIdx].VSEId,
SwapVector[EntryIdx].VSEId));
LLVM_DEBUG(dbgs() << " Def: ");
LLVM_DEBUG(DefMI->dump());
}
}
}
// Walk the swap vector entries looking for conditions that prevent their
// containing computations from being optimized. When such conditions are
// found, mark the representative of the computation's equivalence class
// as rejected.
void PPCVSXSwapRemoval::recordUnoptimizableWebs() {
LLVM_DEBUG(dbgs() << "\n*** Rejecting webs for swap removal ***\n\n");
for (unsigned EntryIdx = 0; EntryIdx < SwapVector.size(); ++EntryIdx) {
int Repr = EC->getLeaderValue(SwapVector[EntryIdx].VSEId);
// If representative is already rejected, don't waste further time.
if (SwapVector[Repr].WebRejected)
continue;
// Reject webs containing mentions of physical or partial registers, or
// containing operations that we don't know how to handle in a lane-
// permuted region.
if (SwapVector[EntryIdx].MentionsPhysVR ||
SwapVector[EntryIdx].MentionsPartialVR ||
!(SwapVector[EntryIdx].IsSwappable || SwapVector[EntryIdx].IsSwap)) {
SwapVector[Repr].WebRejected = 1;
LLVM_DEBUG(
dbgs() << format("Web %d rejected for physreg, partial reg, or not "
"swap[pable]\n",
Repr));
LLVM_DEBUG(dbgs() << " in " << EntryIdx << ": ");
LLVM_DEBUG(SwapVector[EntryIdx].VSEMI->dump());
LLVM_DEBUG(dbgs() << "\n");
}
// Reject webs than contain swapping loads that feed something other
// than a swap instruction.
else if (SwapVector[EntryIdx].IsLoad && SwapVector[EntryIdx].IsSwap) {
MachineInstr *MI = SwapVector[EntryIdx].VSEMI;
Register DefReg = MI->getOperand(0).getReg();
// We skip debug instructions in the analysis. (Note that debug
// location information is still maintained by this optimization
// because it remains on the LXVD2X and STXVD2X instructions after
// the XXPERMDIs are removed.)
for (MachineInstr &UseMI : MRI->use_nodbg_instructions(DefReg)) {
int UseIdx = SwapMap[&UseMI];
if (!SwapVector[UseIdx].IsSwap || SwapVector[UseIdx].IsLoad ||
SwapVector[UseIdx].IsStore) {
SwapVector[Repr].WebRejected = 1;
LLVM_DEBUG(dbgs() << format(
"Web %d rejected for load not feeding swap\n", Repr));
LLVM_DEBUG(dbgs() << " def " << EntryIdx << ": ");
LLVM_DEBUG(MI->dump());
LLVM_DEBUG(dbgs() << " use " << UseIdx << ": ");
LLVM_DEBUG(UseMI.dump());
LLVM_DEBUG(dbgs() << "\n");
}
}
// Reject webs that contain swapping stores that are fed by something
// other than a swap instruction.
} else if (SwapVector[EntryIdx].IsStore && SwapVector[EntryIdx].IsSwap) {
MachineInstr *MI = SwapVector[EntryIdx].VSEMI;
Register UseReg = MI->getOperand(0).getReg();
MachineInstr *DefMI = MRI->getVRegDef(UseReg);
Register DefReg = DefMI->getOperand(0).getReg();
int DefIdx = SwapMap[DefMI];
if (!SwapVector[DefIdx].IsSwap || SwapVector[DefIdx].IsLoad ||
SwapVector[DefIdx].IsStore) {
SwapVector[Repr].WebRejected = 1;
LLVM_DEBUG(dbgs() << format(
"Web %d rejected for store not fed by swap\n", Repr));
LLVM_DEBUG(dbgs() << " def " << DefIdx << ": ");
LLVM_DEBUG(DefMI->dump());
LLVM_DEBUG(dbgs() << " use " << EntryIdx << ": ");
LLVM_DEBUG(MI->dump());
LLVM_DEBUG(dbgs() << "\n");
}
// Ensure all uses of the register defined by DefMI feed store
// instructions
for (MachineInstr &UseMI : MRI->use_nodbg_instructions(DefReg)) {
int UseIdx = SwapMap[&UseMI];
if (SwapVector[UseIdx].VSEMI->getOpcode() != MI->getOpcode()) {
SwapVector[Repr].WebRejected = 1;
LLVM_DEBUG(
dbgs() << format(
"Web %d rejected for swap not feeding only stores\n", Repr));
LLVM_DEBUG(dbgs() << " def "
<< " : ");
LLVM_DEBUG(DefMI->dump());
LLVM_DEBUG(dbgs() << " use " << UseIdx << ": ");
LLVM_DEBUG(SwapVector[UseIdx].VSEMI->dump());
LLVM_DEBUG(dbgs() << "\n");
}
}
}
}
LLVM_DEBUG(dbgs() << "Swap vector after web analysis:\n\n");
LLVM_DEBUG(dumpSwapVector());
}
// Walk the swap vector entries looking for swaps fed by permuting loads
// and swaps that feed permuting stores. If the containing computation
// has not been marked rejected, mark each such swap for removal.
// (Removal is delayed in case optimization has disturbed the pattern,
// such that multiple loads feed the same swap, etc.)
void PPCVSXSwapRemoval::markSwapsForRemoval() {
LLVM_DEBUG(dbgs() << "\n*** Marking swaps for removal ***\n\n");
for (unsigned EntryIdx = 0; EntryIdx < SwapVector.size(); ++EntryIdx) {
if (SwapVector[EntryIdx].IsLoad && SwapVector[EntryIdx].IsSwap) {
int Repr = EC->getLeaderValue(SwapVector[EntryIdx].VSEId);
if (!SwapVector[Repr].WebRejected) {
MachineInstr *MI = SwapVector[EntryIdx].VSEMI;
Register DefReg = MI->getOperand(0).getReg();
for (MachineInstr &UseMI : MRI->use_nodbg_instructions(DefReg)) {
int UseIdx = SwapMap[&UseMI];
SwapVector[UseIdx].WillRemove = 1;
LLVM_DEBUG(dbgs() << "Marking swap fed by load for removal: ");
LLVM_DEBUG(UseMI.dump());
}
}
} else if (SwapVector[EntryIdx].IsStore && SwapVector[EntryIdx].IsSwap) {
int Repr = EC->getLeaderValue(SwapVector[EntryIdx].VSEId);
if (!SwapVector[Repr].WebRejected) {
MachineInstr *MI = SwapVector[EntryIdx].VSEMI;
Register UseReg = MI->getOperand(0).getReg();
MachineInstr *DefMI = MRI->getVRegDef(UseReg);
int DefIdx = SwapMap[DefMI];
SwapVector[DefIdx].WillRemove = 1;
LLVM_DEBUG(dbgs() << "Marking swap feeding store for removal: ");
LLVM_DEBUG(DefMI->dump());
}
} else if (SwapVector[EntryIdx].IsSwappable &&
SwapVector[EntryIdx].SpecialHandling != 0) {
int Repr = EC->getLeaderValue(SwapVector[EntryIdx].VSEId);
if (!SwapVector[Repr].WebRejected)
handleSpecialSwappables(EntryIdx);
}
}
}
// Create an xxswapd instruction and insert it prior to the given point.
// MI is used to determine basic block and debug loc information.
// FIXME: When inserting a swap, we should check whether SrcReg is
// defined by another swap: SrcReg = XXPERMDI Reg, Reg, 2; If so,
// then instead we should generate a copy from Reg to DstReg.
void PPCVSXSwapRemoval::insertSwap(MachineInstr *MI,
MachineBasicBlock::iterator InsertPoint,
unsigned DstReg, unsigned SrcReg) {
BuildMI(*MI->getParent(), InsertPoint, MI->getDebugLoc(),
TII->get(PPC::XXPERMDI), DstReg)
.addReg(SrcReg)
.addReg(SrcReg)
.addImm(2);
}
// The identified swap entry requires special handling to allow its
// containing computation to be optimized. Perform that handling
// here.
// FIXME: Additional opportunities will be phased in with subsequent
// patches.
void PPCVSXSwapRemoval::handleSpecialSwappables(int EntryIdx) {
switch (SwapVector[EntryIdx].SpecialHandling) {
default:
llvm_unreachable("Unexpected special handling type");
// For splats based on an index into a vector, add N/2 modulo N
// to the index, where N is the number of vector elements.
case SHValues::SH_SPLAT: {
MachineInstr *MI = SwapVector[EntryIdx].VSEMI;
unsigned NElts;
LLVM_DEBUG(dbgs() << "Changing splat: ");
LLVM_DEBUG(MI->dump());
switch (MI->getOpcode()) {
default:
llvm_unreachable("Unexpected splat opcode");
case PPC::VSPLTB: NElts = 16; break;
case PPC::VSPLTH: NElts = 8; break;
case PPC::VSPLTW:
case PPC::XXSPLTW: NElts = 4; break;
}
unsigned EltNo;
if (MI->getOpcode() == PPC::XXSPLTW)
EltNo = MI->getOperand(2).getImm();
else
EltNo = MI->getOperand(1).getImm();
EltNo = (EltNo + NElts / 2) % NElts;
if (MI->getOpcode() == PPC::XXSPLTW)
MI->getOperand(2).setImm(EltNo);
else
MI->getOperand(1).setImm(EltNo);
LLVM_DEBUG(dbgs() << " Into: ");
LLVM_DEBUG(MI->dump());
break;
}
// For an XXPERMDI that isn't handled otherwise, we need to
// reverse the order of the operands. If the selector operand
// has a value of 0 or 3, we need to change it to 3 or 0,
// respectively. Otherwise we should leave it alone. (This
// is equivalent to reversing the two bits of the selector
// operand and complementing the result.)
case SHValues::SH_XXPERMDI: {
MachineInstr *MI = SwapVector[EntryIdx].VSEMI;
LLVM_DEBUG(dbgs() << "Changing XXPERMDI: ");
LLVM_DEBUG(MI->dump());
unsigned Selector = MI->getOperand(3).getImm();
if (Selector == 0 || Selector == 3)
Selector = 3 - Selector;
MI->getOperand(3).setImm(Selector);
Register Reg1 = MI->getOperand(1).getReg();
Register Reg2 = MI->getOperand(2).getReg();
MI->getOperand(1).setReg(Reg2);
MI->getOperand(2).setReg(Reg1);
// We also need to swap kill flag associated with the register.
bool IsKill1 = MI->getOperand(1).isKill();
bool IsKill2 = MI->getOperand(2).isKill();
MI->getOperand(1).setIsKill(IsKill2);
MI->getOperand(2).setIsKill(IsKill1);
LLVM_DEBUG(dbgs() << " Into: ");
LLVM_DEBUG(MI->dump());
break;
}
// For a copy from a scalar floating-point register to a vector
// register, removing swaps will leave the copied value in the
// wrong lane. Insert a swap following the copy to fix this.
case SHValues::SH_COPYWIDEN: {
MachineInstr *MI = SwapVector[EntryIdx].VSEMI;
LLVM_DEBUG(dbgs() << "Changing SUBREG_TO_REG: ");
LLVM_DEBUG(MI->dump());
Register DstReg = MI->getOperand(0).getReg();
const TargetRegisterClass *DstRC = MRI->getRegClass(DstReg);
Register NewVReg = MRI->createVirtualRegister(DstRC);
MI->getOperand(0).setReg(NewVReg);
LLVM_DEBUG(dbgs() << " Into: ");
LLVM_DEBUG(MI->dump());
auto InsertPoint = ++MachineBasicBlock::iterator(MI);
// Note that an XXPERMDI requires a VSRC, so if the SUBREG_TO_REG
// is copying to a VRRC, we need to be careful to avoid a register
// assignment problem. In this case we must copy from VRRC to VSRC
// prior to the swap, and from VSRC to VRRC following the swap.
// Coalescing will usually remove all this mess.
if (DstRC == &PPC::VRRCRegClass) {
Register VSRCTmp1 = MRI->createVirtualRegister(&PPC::VSRCRegClass);
Register VSRCTmp2 = MRI->createVirtualRegister(&PPC::VSRCRegClass);
BuildMI(*MI->getParent(), InsertPoint, MI->getDebugLoc(),
TII->get(PPC::COPY), VSRCTmp1)
.addReg(NewVReg);
LLVM_DEBUG(std::prev(InsertPoint)->dump());
insertSwap(MI, InsertPoint, VSRCTmp2, VSRCTmp1);
LLVM_DEBUG(std::prev(InsertPoint)->dump());
BuildMI(*MI->getParent(), InsertPoint, MI->getDebugLoc(),
TII->get(PPC::COPY), DstReg)
.addReg(VSRCTmp2);
LLVM_DEBUG(std::prev(InsertPoint)->dump());
} else {
insertSwap(MI, InsertPoint, DstReg, NewVReg);
LLVM_DEBUG(std::prev(InsertPoint)->dump());
}
break;
}
}
}
// Walk the swap vector and replace each entry marked for removal with
// a copy operation.
bool PPCVSXSwapRemoval::removeSwaps() {
LLVM_DEBUG(dbgs() << "\n*** Removing swaps ***\n\n");
bool Changed = false;
for (unsigned EntryIdx = 0; EntryIdx < SwapVector.size(); ++EntryIdx) {
if (SwapVector[EntryIdx].WillRemove) {
Changed = true;
MachineInstr *MI = SwapVector[EntryIdx].VSEMI;
MachineBasicBlock *MBB = MI->getParent();
BuildMI(*MBB, MI, MI->getDebugLoc(), TII->get(TargetOpcode::COPY),
MI->getOperand(0).getReg())
.add(MI->getOperand(1));
LLVM_DEBUG(dbgs() << format("Replaced %d with copy: ",
SwapVector[EntryIdx].VSEId));
LLVM_DEBUG(MI->dump());
MI->eraseFromParent();
}
}
return Changed;
}
#if !defined(NDEBUG) || defined(LLVM_ENABLE_DUMP)
// For debug purposes, dump the contents of the swap vector.
LLVM_DUMP_METHOD void PPCVSXSwapRemoval::dumpSwapVector() {
for (unsigned EntryIdx = 0; EntryIdx < SwapVector.size(); ++EntryIdx) {
MachineInstr *MI = SwapVector[EntryIdx].VSEMI;
int ID = SwapVector[EntryIdx].VSEId;
dbgs() << format("%6d", ID);
dbgs() << format("%6d", EC->getLeaderValue(ID));
dbgs() << format(" %bb.%3d", MI->getParent()->getNumber());
dbgs() << format(" %14s ", TII->getName(MI->getOpcode()).str().c_str());
if (SwapVector[EntryIdx].IsLoad)
dbgs() << "load ";
if (SwapVector[EntryIdx].IsStore)
dbgs() << "store ";
if (SwapVector[EntryIdx].IsSwap)
dbgs() << "swap ";
if (SwapVector[EntryIdx].MentionsPhysVR)
dbgs() << "physreg ";
if (SwapVector[EntryIdx].MentionsPartialVR)
dbgs() << "partialreg ";
if (SwapVector[EntryIdx].IsSwappable) {
dbgs() << "swappable ";
switch(SwapVector[EntryIdx].SpecialHandling) {
default:
dbgs() << "special:**unknown**";
break;
case SH_NONE:
break;
case SH_EXTRACT:
dbgs() << "special:extract ";
break;
case SH_INSERT:
dbgs() << "special:insert ";
break;
case SH_NOSWAP_LD:
dbgs() << "special:load ";
break;
case SH_NOSWAP_ST:
dbgs() << "special:store ";
break;
case SH_SPLAT:
dbgs() << "special:splat ";
break;
case SH_XXPERMDI:
dbgs() << "special:xxpermdi ";
break;
case SH_COPYWIDEN:
dbgs() << "special:copywiden ";
break;
}
}
if (SwapVector[EntryIdx].WebRejected)
dbgs() << "rejected ";
if (SwapVector[EntryIdx].WillRemove)
dbgs() << "remove ";
dbgs() << "\n";
// For no-asserts builds.
(void)MI;
(void)ID;
}
dbgs() << "\n";
}
#endif
} // end default namespace
INITIALIZE_PASS_BEGIN(PPCVSXSwapRemoval, DEBUG_TYPE,
"PowerPC VSX Swap Removal", false, false)
INITIALIZE_PASS_END(PPCVSXSwapRemoval, DEBUG_TYPE,
"PowerPC VSX Swap Removal", false, false)
char PPCVSXSwapRemoval::ID = 0;
FunctionPass*
llvm::createPPCVSXSwapRemovalPass() { return new PPCVSXSwapRemoval(); }