llvm-project/llvm/lib/Target/PowerPC/InstPrinter/PPCInstPrinter.cpp

366 lines
11 KiB
C++
Raw Normal View History

//===-- PPCInstPrinter.cpp - Convert PPC MCInst to assembly syntax --------===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// This class prints an PPC MCInst to a .s file.
//
//===----------------------------------------------------------------------===//
#include "PPCInstPrinter.h"
#include "MCTargetDesc/PPCMCTargetDesc.h"
#include "MCTargetDesc/PPCPredicates.h"
#include "llvm/MC/MCExpr.h"
#include "llvm/MC/MCInst.h"
#include "llvm/MC/MCInstrInfo.h"
#include "llvm/MC/MCSymbol.h"
#include "llvm/Support/CommandLine.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Target/TargetOpcodes.h"
using namespace llvm;
#define DEBUG_TYPE "asm-printer"
// FIXME: Once the integrated assembler supports full register names, tie this
// to the verbose-asm setting.
static cl::opt<bool>
FullRegNames("ppc-asm-full-reg-names", cl::Hidden, cl::init(false),
cl::desc("Use full register names when printing assembly"));
#include "PPCGenAsmWriter.inc"
void PPCInstPrinter::printRegName(raw_ostream &OS, unsigned RegNo) const {
OS << getRegisterName(RegNo);
}
void PPCInstPrinter::printInst(const MCInst *MI, raw_ostream &O,
StringRef Annot) {
// Check for slwi/srwi mnemonics.
if (MI->getOpcode() == PPC::RLWINM) {
unsigned char SH = MI->getOperand(2).getImm();
unsigned char MB = MI->getOperand(3).getImm();
unsigned char ME = MI->getOperand(4).getImm();
bool useSubstituteMnemonic = false;
if (SH <= 31 && MB == 0 && ME == (31-SH)) {
O << "\tslwi "; useSubstituteMnemonic = true;
}
if (SH <= 31 && MB == (32-SH) && ME == 31) {
O << "\tsrwi "; useSubstituteMnemonic = true;
SH = 32-SH;
}
if (useSubstituteMnemonic) {
printOperand(MI, 0, O);
O << ", ";
printOperand(MI, 1, O);
O << ", " << (unsigned int)SH;
printAnnotation(O, Annot);
return;
}
}
if ((MI->getOpcode() == PPC::OR || MI->getOpcode() == PPC::OR8) &&
MI->getOperand(1).getReg() == MI->getOperand(2).getReg()) {
O << "\tmr ";
printOperand(MI, 0, O);
O << ", ";
printOperand(MI, 1, O);
printAnnotation(O, Annot);
return;
}
if (MI->getOpcode() == PPC::RLDICR) {
unsigned char SH = MI->getOperand(2).getImm();
unsigned char ME = MI->getOperand(3).getImm();
// rldicr RA, RS, SH, 63-SH == sldi RA, RS, SH
if (63-SH == ME) {
O << "\tsldi ";
printOperand(MI, 0, O);
O << ", ";
printOperand(MI, 1, O);
O << ", " << (unsigned int)SH;
printAnnotation(O, Annot);
return;
}
}
// For fast-isel, a COPY_TO_REGCLASS may survive this long. This is
// used when converting a 32-bit float to a 64-bit float as part of
// conversion to an integer (see PPCFastISel.cpp:SelectFPToI()),
// as otherwise we have problems with incorrect register classes
// in machine instruction verification. For now, just avoid trying
// to print it as such an instruction has no effect (a 32-bit float
// in a register is already in 64-bit form, just with lower
// precision). FIXME: Is there a better solution?
if (MI->getOpcode() == TargetOpcode::COPY_TO_REGCLASS)
return;
printInstruction(MI, O);
printAnnotation(O, Annot);
}
void PPCInstPrinter::printPredicateOperand(const MCInst *MI, unsigned OpNo,
raw_ostream &O,
const char *Modifier) {
unsigned Code = MI->getOperand(OpNo).getImm();
if (StringRef(Modifier) == "cc") {
switch ((PPC::Predicate)Code) {
case PPC::PRED_LT_MINUS:
case PPC::PRED_LT_PLUS:
case PPC::PRED_LT:
O << "lt";
return;
case PPC::PRED_LE_MINUS:
case PPC::PRED_LE_PLUS:
case PPC::PRED_LE:
O << "le";
return;
case PPC::PRED_EQ_MINUS:
case PPC::PRED_EQ_PLUS:
case PPC::PRED_EQ:
O << "eq";
return;
case PPC::PRED_GE_MINUS:
case PPC::PRED_GE_PLUS:
case PPC::PRED_GE:
O << "ge";
return;
case PPC::PRED_GT_MINUS:
case PPC::PRED_GT_PLUS:
case PPC::PRED_GT:
O << "gt";
return;
case PPC::PRED_NE_MINUS:
case PPC::PRED_NE_PLUS:
case PPC::PRED_NE:
O << "ne";
return;
case PPC::PRED_UN_MINUS:
case PPC::PRED_UN_PLUS:
case PPC::PRED_UN:
O << "un";
return;
case PPC::PRED_NU_MINUS:
case PPC::PRED_NU_PLUS:
case PPC::PRED_NU:
O << "nu";
return;
Add CR-bit tracking to the PowerPC backend for i1 values This change enables tracking i1 values in the PowerPC backend using the condition register bits. These bits can be treated on PowerPC as separate registers; individual bit operations (and, or, xor, etc.) are supported. Tracking booleans in CR bits has several advantages: - Reduction in register pressure (because we no longer need GPRs to store boolean values). - Logical operations on booleans can be handled more efficiently; we used to have to move all results from comparisons into GPRs, perform promoted logical operations in GPRs, and then move the result back into condition register bits to be used by conditional branches. This can be very inefficient, because the throughput of these CR <-> GPR moves have high latency and low throughput (especially when other associated instructions are accounted for). - On the POWER7 and similar cores, we can increase total throughput by using the CR bits. CR bit operations have a dedicated functional unit. Most of this is more-or-less mechanical: Adjustments were needed in the calling-convention code, support was added for spilling/restoring individual condition-register bits, and conditional branch instruction definitions taking specific CR bits were added (plus patterns and code for generating bit-level operations). This is enabled by default when running at -O2 and higher. For -O0 and -O1, where the ability to debug is more important, this feature is disabled by default. Individual CR bits do not have assigned DWARF register numbers, and storing values in CR bits makes them invisible to the debugger. It is critical, however, that we don't move i1 values that have been promoted to larger values (such as those passed as function arguments) into bit registers only to quickly turn around and move the values back into GPRs (such as happens when values are returned by functions). A pair of target-specific DAG combines are added to remove the trunc/extends in: trunc(binary-ops(binary-ops(zext(x), zext(y)), ...) and: zext(binary-ops(binary-ops(trunc(x), trunc(y)), ...) In short, we only want to use CR bits where some of the i1 values come from comparisons or are used by conditional branches or selects. To put it another way, if we can do the entire i1 computation in GPRs, then we probably should (on the POWER7, the GPR-operation throughput is higher, and for all cores, the CR <-> GPR moves are expensive). POWER7 test-suite performance results (from 10 runs in each configuration): SingleSource/Benchmarks/Misc/mandel-2: 35% speedup MultiSource/Benchmarks/Prolangs-C++/city/city: 21% speedup MultiSource/Benchmarks/MiBench/automotive-susan: 23% speedup SingleSource/Benchmarks/CoyoteBench/huffbench: 13% speedup SingleSource/Benchmarks/Misc-C++/Large/sphereflake: 13% speedup SingleSource/Benchmarks/Misc-C++/mandel-text: 10% speedup SingleSource/Benchmarks/Misc-C++-EH/spirit: 10% slowdown MultiSource/Applications/lemon/lemon: 8% slowdown llvm-svn: 202451
2014-02-28 08:27:01 +08:00
case PPC::PRED_BIT_SET:
case PPC::PRED_BIT_UNSET:
llvm_unreachable("Invalid use of bit predicate code");
}
llvm_unreachable("Invalid predicate code");
}
if (StringRef(Modifier) == "pm") {
switch ((PPC::Predicate)Code) {
case PPC::PRED_LT:
case PPC::PRED_LE:
case PPC::PRED_EQ:
case PPC::PRED_GE:
case PPC::PRED_GT:
case PPC::PRED_NE:
case PPC::PRED_UN:
case PPC::PRED_NU:
return;
case PPC::PRED_LT_MINUS:
case PPC::PRED_LE_MINUS:
case PPC::PRED_EQ_MINUS:
case PPC::PRED_GE_MINUS:
case PPC::PRED_GT_MINUS:
case PPC::PRED_NE_MINUS:
case PPC::PRED_UN_MINUS:
case PPC::PRED_NU_MINUS:
O << "-";
return;
case PPC::PRED_LT_PLUS:
case PPC::PRED_LE_PLUS:
case PPC::PRED_EQ_PLUS:
case PPC::PRED_GE_PLUS:
case PPC::PRED_GT_PLUS:
case PPC::PRED_NE_PLUS:
case PPC::PRED_UN_PLUS:
case PPC::PRED_NU_PLUS:
O << "+";
return;
Add CR-bit tracking to the PowerPC backend for i1 values This change enables tracking i1 values in the PowerPC backend using the condition register bits. These bits can be treated on PowerPC as separate registers; individual bit operations (and, or, xor, etc.) are supported. Tracking booleans in CR bits has several advantages: - Reduction in register pressure (because we no longer need GPRs to store boolean values). - Logical operations on booleans can be handled more efficiently; we used to have to move all results from comparisons into GPRs, perform promoted logical operations in GPRs, and then move the result back into condition register bits to be used by conditional branches. This can be very inefficient, because the throughput of these CR <-> GPR moves have high latency and low throughput (especially when other associated instructions are accounted for). - On the POWER7 and similar cores, we can increase total throughput by using the CR bits. CR bit operations have a dedicated functional unit. Most of this is more-or-less mechanical: Adjustments were needed in the calling-convention code, support was added for spilling/restoring individual condition-register bits, and conditional branch instruction definitions taking specific CR bits were added (plus patterns and code for generating bit-level operations). This is enabled by default when running at -O2 and higher. For -O0 and -O1, where the ability to debug is more important, this feature is disabled by default. Individual CR bits do not have assigned DWARF register numbers, and storing values in CR bits makes them invisible to the debugger. It is critical, however, that we don't move i1 values that have been promoted to larger values (such as those passed as function arguments) into bit registers only to quickly turn around and move the values back into GPRs (such as happens when values are returned by functions). A pair of target-specific DAG combines are added to remove the trunc/extends in: trunc(binary-ops(binary-ops(zext(x), zext(y)), ...) and: zext(binary-ops(binary-ops(trunc(x), trunc(y)), ...) In short, we only want to use CR bits where some of the i1 values come from comparisons or are used by conditional branches or selects. To put it another way, if we can do the entire i1 computation in GPRs, then we probably should (on the POWER7, the GPR-operation throughput is higher, and for all cores, the CR <-> GPR moves are expensive). POWER7 test-suite performance results (from 10 runs in each configuration): SingleSource/Benchmarks/Misc/mandel-2: 35% speedup MultiSource/Benchmarks/Prolangs-C++/city/city: 21% speedup MultiSource/Benchmarks/MiBench/automotive-susan: 23% speedup SingleSource/Benchmarks/CoyoteBench/huffbench: 13% speedup SingleSource/Benchmarks/Misc-C++/Large/sphereflake: 13% speedup SingleSource/Benchmarks/Misc-C++/mandel-text: 10% speedup SingleSource/Benchmarks/Misc-C++-EH/spirit: 10% slowdown MultiSource/Applications/lemon/lemon: 8% slowdown llvm-svn: 202451
2014-02-28 08:27:01 +08:00
case PPC::PRED_BIT_SET:
case PPC::PRED_BIT_UNSET:
llvm_unreachable("Invalid use of bit predicate code");
}
llvm_unreachable("Invalid predicate code");
}
assert(StringRef(Modifier) == "reg" &&
"Need to specify 'cc', 'pm' or 'reg' as predicate op modifier!");
printOperand(MI, OpNo+1, O);
}
[PowerPC] Initial support for the VSX instruction set VSX is an ISA extension supported on the POWER7 and later cores that enhances floating-point vector and scalar capabilities. Among other things, this adds <2 x double> support and generally helps to reduce register pressure. The interesting part of this ISA feature is the register configuration: there are 64 new 128-bit vector registers, the 32 of which are super-registers of the existing 32 scalar floating-point registers, and the second 32 of which overlap with the 32 Altivec vector registers. This makes things like vector insertion and extraction tricky: this can be free but only if we force a restriction to the right register subclass when needed. A new "minipass" PPCVSXCopy takes care of this (although it could do a more-optimal job of it; see the comment about unnecessary copies below). Please note that, currently, VSX is not enabled by default when targeting anything because it is not yet ready for that. The assembler and disassembler are fully implemented and tested. However: - CodeGen support causes miscompiles; test-suite runtime failures: MultiSource/Benchmarks/FreeBench/distray/distray MultiSource/Benchmarks/McCat/08-main/main MultiSource/Benchmarks/Olden/voronoi/voronoi MultiSource/Benchmarks/mafft/pairlocalalign MultiSource/Benchmarks/tramp3d-v4/tramp3d-v4 SingleSource/Benchmarks/CoyoteBench/almabench SingleSource/Benchmarks/Misc/matmul_f64_4x4 - The lowering currently falls back to using Altivec instructions far more than it should. Worse, there are some things that are scalarized through the stack that shouldn't be. - A lot of unnecessary copies make it past the optimizers, and this needs to be fixed. - Many more regression tests are needed. Normally, I'd fix these things prior to committing, but there are some students and other contributors who would like to work this, and so it makes sense to move this development process upstream where it can be subject to the regular code-review procedures. llvm-svn: 203768
2014-03-13 15:58:58 +08:00
void PPCInstPrinter::printU2ImmOperand(const MCInst *MI, unsigned OpNo,
raw_ostream &O) {
unsigned int Value = MI->getOperand(OpNo).getImm();
assert(Value <= 3 && "Invalid u2imm argument!");
O << (unsigned int)Value;
}
void PPCInstPrinter::printS5ImmOperand(const MCInst *MI, unsigned OpNo,
raw_ostream &O) {
int Value = MI->getOperand(OpNo).getImm();
Value = SignExtend32<5>(Value);
O << (int)Value;
}
void PPCInstPrinter::printU5ImmOperand(const MCInst *MI, unsigned OpNo,
raw_ostream &O) {
unsigned int Value = MI->getOperand(OpNo).getImm();
assert(Value <= 31 && "Invalid u5imm argument!");
O << (unsigned int)Value;
}
void PPCInstPrinter::printU6ImmOperand(const MCInst *MI, unsigned OpNo,
raw_ostream &O) {
unsigned int Value = MI->getOperand(OpNo).getImm();
assert(Value <= 63 && "Invalid u6imm argument!");
O << (unsigned int)Value;
}
void PPCInstPrinter::printS16ImmOperand(const MCInst *MI, unsigned OpNo,
raw_ostream &O) {
[PowerPC] Clean up generation of ha16() / lo16() markers When targeting the Darwin assembler, we need to generate markers ha16() and lo16() to designate the high and low parts of a (symbolic) immediate. This is necessary not just for plain symbols, but also for certain symbolic expression, typically along the lines of ha16(A - B). The latter doesn't work when simply using VariantKind flags on the symbol reference. This is why the current back-end uses hacks (explicitly called out as such via multiple FIXMEs) in the symbolLo/symbolHi print methods. This patch uses target-defined MCExpr codes to represent the Darwin ha16/lo16 constructs, following along the lines of the equivalent solution used by the ARM back end to handle their :upper16: / :lower16: markers. This allows us to get rid of special handling both in the symbolLo/symbolHi print method and in the common code MCExpr::print routine. Instead, the ha16 / lo16 markers are printed simply in a custom print routine for the target MCExpr types. (As a result, the symbolLo/symbolHi print methods can now replaced by a single printS16ImmOperand routine that also handles symbolic operands.) The patch also provides a EvaluateAsRelocatableImpl routine to handle ha16/lo16 constructs. This is not actually used at the moment by any in-tree code, but is provided as it makes merging into David Fang's out-of-tree Mach-O object writer simpler. Since there is no longer any need to treat VK_PPC_GAS_HA16 and VK_PPC_DARWIN_HA16 differently, they are merged into a single VK_PPC_ADDR16_HA (and likewise for the _LO16 types). llvm-svn: 182616
2013-05-24 06:26:41 +08:00
if (MI->getOperand(OpNo).isImm())
O << (short)MI->getOperand(OpNo).getImm();
else
printOperand(MI, OpNo, O);
}
void PPCInstPrinter::printU16ImmOperand(const MCInst *MI, unsigned OpNo,
raw_ostream &O) {
if (MI->getOperand(OpNo).isImm())
O << (unsigned short)MI->getOperand(OpNo).getImm();
else
printOperand(MI, OpNo, O);
}
void PPCInstPrinter::printBranchOperand(const MCInst *MI, unsigned OpNo,
raw_ostream &O) {
if (!MI->getOperand(OpNo).isImm())
return printOperand(MI, OpNo, O);
// Branches can take an immediate operand. This is used by the branch
// selection pass to print .+8, an eight byte displacement from the PC.
O << ".+";
printAbsBranchOperand(MI, OpNo, O);
}
void PPCInstPrinter::printAbsBranchOperand(const MCInst *MI, unsigned OpNo,
raw_ostream &O) {
if (!MI->getOperand(OpNo).isImm())
return printOperand(MI, OpNo, O);
O << (int)MI->getOperand(OpNo).getImm()*4;
}
void PPCInstPrinter::printcrbitm(const MCInst *MI, unsigned OpNo,
raw_ostream &O) {
unsigned CCReg = MI->getOperand(OpNo).getReg();
unsigned RegNo;
switch (CCReg) {
default: llvm_unreachable("Unknown CR register");
case PPC::CR0: RegNo = 0; break;
case PPC::CR1: RegNo = 1; break;
case PPC::CR2: RegNo = 2; break;
case PPC::CR3: RegNo = 3; break;
case PPC::CR4: RegNo = 4; break;
case PPC::CR5: RegNo = 5; break;
case PPC::CR6: RegNo = 6; break;
case PPC::CR7: RegNo = 7; break;
}
O << (0x80 >> RegNo);
}
void PPCInstPrinter::printMemRegImm(const MCInst *MI, unsigned OpNo,
raw_ostream &O) {
[PowerPC] Clean up generation of ha16() / lo16() markers When targeting the Darwin assembler, we need to generate markers ha16() and lo16() to designate the high and low parts of a (symbolic) immediate. This is necessary not just for plain symbols, but also for certain symbolic expression, typically along the lines of ha16(A - B). The latter doesn't work when simply using VariantKind flags on the symbol reference. This is why the current back-end uses hacks (explicitly called out as such via multiple FIXMEs) in the symbolLo/symbolHi print methods. This patch uses target-defined MCExpr codes to represent the Darwin ha16/lo16 constructs, following along the lines of the equivalent solution used by the ARM back end to handle their :upper16: / :lower16: markers. This allows us to get rid of special handling both in the symbolLo/symbolHi print method and in the common code MCExpr::print routine. Instead, the ha16 / lo16 markers are printed simply in a custom print routine for the target MCExpr types. (As a result, the symbolLo/symbolHi print methods can now replaced by a single printS16ImmOperand routine that also handles symbolic operands.) The patch also provides a EvaluateAsRelocatableImpl routine to handle ha16/lo16 constructs. This is not actually used at the moment by any in-tree code, but is provided as it makes merging into David Fang's out-of-tree Mach-O object writer simpler. Since there is no longer any need to treat VK_PPC_GAS_HA16 and VK_PPC_DARWIN_HA16 differently, they are merged into a single VK_PPC_ADDR16_HA (and likewise for the _LO16 types). llvm-svn: 182616
2013-05-24 06:26:41 +08:00
printS16ImmOperand(MI, OpNo, O);
O << '(';
if (MI->getOperand(OpNo+1).getReg() == PPC::R0)
O << "0";
else
printOperand(MI, OpNo+1, O);
O << ')';
}
void PPCInstPrinter::printMemRegReg(const MCInst *MI, unsigned OpNo,
raw_ostream &O) {
// When used as the base register, r0 reads constant zero rather than
// the value contained in the register. For this reason, the darwin
// assembler requires that we print r0 as 0 (no r) when used as the base.
if (MI->getOperand(OpNo).getReg() == PPC::R0)
O << "0";
else
printOperand(MI, OpNo, O);
O << ", ";
printOperand(MI, OpNo+1, O);
}
void PPCInstPrinter::printTLSCall(const MCInst *MI, unsigned OpNo,
raw_ostream &O) {
// On PPC64, VariantKind is VK_None, but on PPC32, it's VK_PLT, and it must
// come at the _end_ of the expression.
const MCOperand &Op = MI->getOperand(OpNo);
const MCSymbolRefExpr &refExp = cast<MCSymbolRefExpr>(*Op.getExpr());
O << refExp.getSymbol().getName();
O << '(';
printOperand(MI, OpNo+1, O);
O << ')';
if (refExp.getKind() != MCSymbolRefExpr::VK_None)
O << '@' << MCSymbolRefExpr::getVariantKindName(refExp.getKind());
}
/// stripRegisterPrefix - This method strips the character prefix from a
/// register name so that only the number is left. Used by for linux asm.
2010-11-26 00:42:51 +08:00
static const char *stripRegisterPrefix(const char *RegName) {
if (FullRegNames)
return RegName;
switch (RegName[0]) {
case 'r':
case 'f':
[PowerPC] Initial support for the VSX instruction set VSX is an ISA extension supported on the POWER7 and later cores that enhances floating-point vector and scalar capabilities. Among other things, this adds <2 x double> support and generally helps to reduce register pressure. The interesting part of this ISA feature is the register configuration: there are 64 new 128-bit vector registers, the 32 of which are super-registers of the existing 32 scalar floating-point registers, and the second 32 of which overlap with the 32 Altivec vector registers. This makes things like vector insertion and extraction tricky: this can be free but only if we force a restriction to the right register subclass when needed. A new "minipass" PPCVSXCopy takes care of this (although it could do a more-optimal job of it; see the comment about unnecessary copies below). Please note that, currently, VSX is not enabled by default when targeting anything because it is not yet ready for that. The assembler and disassembler are fully implemented and tested. However: - CodeGen support causes miscompiles; test-suite runtime failures: MultiSource/Benchmarks/FreeBench/distray/distray MultiSource/Benchmarks/McCat/08-main/main MultiSource/Benchmarks/Olden/voronoi/voronoi MultiSource/Benchmarks/mafft/pairlocalalign MultiSource/Benchmarks/tramp3d-v4/tramp3d-v4 SingleSource/Benchmarks/CoyoteBench/almabench SingleSource/Benchmarks/Misc/matmul_f64_4x4 - The lowering currently falls back to using Altivec instructions far more than it should. Worse, there are some things that are scalarized through the stack that shouldn't be. - A lot of unnecessary copies make it past the optimizers, and this needs to be fixed. - Many more regression tests are needed. Normally, I'd fix these things prior to committing, but there are some students and other contributors who would like to work this, and so it makes sense to move this development process upstream where it can be subject to the regular code-review procedures. llvm-svn: 203768
2014-03-13 15:58:58 +08:00
case 'v':
if (RegName[1] == 's')
return RegName + 2;
return RegName + 1;
case 'c': if (RegName[1] == 'r') return RegName + 2;
}
return RegName;
}
void PPCInstPrinter::printOperand(const MCInst *MI, unsigned OpNo,
raw_ostream &O) {
const MCOperand &Op = MI->getOperand(OpNo);
if (Op.isReg()) {
const char *RegName = getRegisterName(Op.getReg());
// The linux and AIX assembler does not take register prefixes.
if (!isDarwinSyntax())
RegName = stripRegisterPrefix(RegName);
O << RegName;
return;
}
if (Op.isImm()) {
O << Op.getImm();
return;
}
assert(Op.isExpr() && "unknown operand kind in printOperand");
O << *Op.getExpr();
}