llvm-6502/lib/Target/Mips/MipsTargetMachine.cpp

192 lines
6.7 KiB
C++
Raw Normal View History

//===-- MipsTargetMachine.cpp - Define TargetMachine for Mips -------------===//
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
//
// Implements the info about Mips target spec.
//
//===----------------------------------------------------------------------===//
#include "MipsTargetMachine.h"
#include "Mips.h"
#include "Mips16FrameLowering.h"
#include "Mips16HardFloat.h"
#include "Mips16ISelDAGToDAG.h"
#include "Mips16ISelLowering.h"
#include "Mips16InstrInfo.h"
#include "MipsFrameLowering.h"
#include "MipsInstrInfo.h"
This patch enables llvm to switch between compiling for mips32/mips64 and mips16 on a per function basis. Because this patch is somewhat involved I have provide an overview of the key pieces of it. The patch is written so as to not change the behavior of the non mixed mode. We have tested this a lot but it is something new to switch subtargets so we don't want any chance of regression in the mainline compiler until we have more confidence in this. Mips32/64 are very different from Mip16 as is the case of ARM vs Thumb1. For that reason there are derived versions of the register info, frame info, instruction info and instruction selection classes. Now we register three separate passes for instruction selection. One which is used to switch subtargets (MipsModuleISelDAGToDAG.cpp) and then one for each of the current subtargets (Mips16ISelDAGToDAG.cpp and MipsSEISelDAGToDAG.cpp). When the ModuleISel pass runs, it determines if there is a need to switch subtargets and if so, the owning pointers in MipsTargetMachine are appropriately changed. When 16Isel or SEIsel is run, they will return immediately without doing any work if the current subtarget mode does not apply to them. In addition, MipsAsmPrinter needs to be reset on a function basis. The pass BasicTargetTransformInfo is substituted with a null pass since the pass is immutable and really needs to be a function pass for it to be used with changing subtargets. This will be fixed in a follow on patch. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@179118 91177308-0d34-0410-b5e6-96231b3b80d8
2013-04-09 19:46:01 +00:00
#include "MipsModuleISelDAGToDAG.h"
This is for an experimental option -mips-os16. The idea is to compile all Mips32 code as Mips16 unless it can't be compiled as Mips 16. For now this would happen as long as floating point instructions are not needed. Probably it would also make sense to compile as mips32 if atomic operations are needed too. There may be other cases too. A module pass prescans the IR and adds the mips16 or nomips16 attribute to functions depending on the functions needs. Mips 16 mode can result in a 40% code compression by utililizing 16 bit encoding of many instructions. The hope is for this to replace the traditional gcc way of dealing with Mips16 code using floating point which involves essentially using soft float but with a library implemented using mips32 floating point. This gcc method also requires creating stubs so that Mips32 code can interact with these Mips 16 functions that have floating point needs. My conjecture is that in reality this traditional gcc method would never win over this new method. I will be implementing the traditional gcc method also. Some of it is already done but I needed to do the stubs to finish the work and those required this mips16/32 mixed mode capability. I have more ideas for to make this new method much better and I think the old method will just live in llvm for anyone that needs the backward compatibility but I don't for what reason that would be needed. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@179185 91177308-0d34-0410-b5e6-96231b3b80d8
2013-04-10 16:58:04 +00:00
#include "MipsOs16.h"
This patch enables llvm to switch between compiling for mips32/mips64 and mips16 on a per function basis. Because this patch is somewhat involved I have provide an overview of the key pieces of it. The patch is written so as to not change the behavior of the non mixed mode. We have tested this a lot but it is something new to switch subtargets so we don't want any chance of regression in the mainline compiler until we have more confidence in this. Mips32/64 are very different from Mip16 as is the case of ARM vs Thumb1. For that reason there are derived versions of the register info, frame info, instruction info and instruction selection classes. Now we register three separate passes for instruction selection. One which is used to switch subtargets (MipsModuleISelDAGToDAG.cpp) and then one for each of the current subtargets (Mips16ISelDAGToDAG.cpp and MipsSEISelDAGToDAG.cpp). When the ModuleISel pass runs, it determines if there is a need to switch subtargets and if so, the owning pointers in MipsTargetMachine are appropriately changed. When 16Isel or SEIsel is run, they will return immediately without doing any work if the current subtarget mode does not apply to them. In addition, MipsAsmPrinter needs to be reset on a function basis. The pass BasicTargetTransformInfo is substituted with a null pass since the pass is immutable and really needs to be a function pass for it to be used with changing subtargets. This will be fixed in a follow on patch. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@179118 91177308-0d34-0410-b5e6-96231b3b80d8
2013-04-09 19:46:01 +00:00
#include "MipsSEFrameLowering.h"
#include "MipsSEISelDAGToDAG.h"
#include "MipsSEISelLowering.h"
#include "MipsSEInstrInfo.h"
This patch enables llvm to switch between compiling for mips32/mips64 and mips16 on a per function basis. Because this patch is somewhat involved I have provide an overview of the key pieces of it. The patch is written so as to not change the behavior of the non mixed mode. We have tested this a lot but it is something new to switch subtargets so we don't want any chance of regression in the mainline compiler until we have more confidence in this. Mips32/64 are very different from Mip16 as is the case of ARM vs Thumb1. For that reason there are derived versions of the register info, frame info, instruction info and instruction selection classes. Now we register three separate passes for instruction selection. One which is used to switch subtargets (MipsModuleISelDAGToDAG.cpp) and then one for each of the current subtargets (Mips16ISelDAGToDAG.cpp and MipsSEISelDAGToDAG.cpp). When the ModuleISel pass runs, it determines if there is a need to switch subtargets and if so, the owning pointers in MipsTargetMachine are appropriately changed. When 16Isel or SEIsel is run, they will return immediately without doing any work if the current subtarget mode does not apply to them. In addition, MipsAsmPrinter needs to be reset on a function basis. The pass BasicTargetTransformInfo is substituted with a null pass since the pass is immutable and really needs to be a function pass for it to be used with changing subtargets. This will be fixed in a follow on patch. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@179118 91177308-0d34-0410-b5e6-96231b3b80d8
2013-04-09 19:46:01 +00:00
#include "llvm/Analysis/TargetTransformInfo.h"
#include "llvm/CodeGen/Passes.h"
#include "llvm/PassManager.h"
This patch enables llvm to switch between compiling for mips32/mips64 and mips16 on a per function basis. Because this patch is somewhat involved I have provide an overview of the key pieces of it. The patch is written so as to not change the behavior of the non mixed mode. We have tested this a lot but it is something new to switch subtargets so we don't want any chance of regression in the mainline compiler until we have more confidence in this. Mips32/64 are very different from Mip16 as is the case of ARM vs Thumb1. For that reason there are derived versions of the register info, frame info, instruction info and instruction selection classes. Now we register three separate passes for instruction selection. One which is used to switch subtargets (MipsModuleISelDAGToDAG.cpp) and then one for each of the current subtargets (Mips16ISelDAGToDAG.cpp and MipsSEISelDAGToDAG.cpp). When the ModuleISel pass runs, it determines if there is a need to switch subtargets and if so, the owning pointers in MipsTargetMachine are appropriately changed. When 16Isel or SEIsel is run, they will return immediately without doing any work if the current subtarget mode does not apply to them. In addition, MipsAsmPrinter needs to be reset on a function basis. The pass BasicTargetTransformInfo is substituted with a null pass since the pass is immutable and really needs to be a function pass for it to be used with changing subtargets. This will be fixed in a follow on patch. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@179118 91177308-0d34-0410-b5e6-96231b3b80d8
2013-04-09 19:46:01 +00:00
#include "llvm/Support/Debug.h"
#include "llvm/Support/TargetRegistry.h"
#include "llvm/Support/raw_ostream.h"
#include "llvm/Transforms/Scalar.h"
using namespace llvm;
[Modules] Make Support/Debug.h modular. This requires it to not change behavior based on other files defining DEBUG_TYPE, which means it cannot define DEBUG_TYPE at all. This is actually better IMO as it forces folks to define relevant DEBUG_TYPEs for their files. However, it requires all files that currently use DEBUG(...) to define a DEBUG_TYPE if they don't already. I've updated all such files in LLVM and will do the same for other upstream projects. This still leaves one important change in how LLVM uses the DEBUG_TYPE macro going forward: we need to only define the macro *after* header files have been #include-ed. Previously, this wasn't possible because Debug.h required the macro to be pre-defined. This commit removes that. By defining DEBUG_TYPE after the includes two things are fixed: - Header files that need to provide a DEBUG_TYPE for some inline code can do so by defining the macro before their inline code and undef-ing it afterward so the macro does not escape. - We no longer have rampant ODR violations due to including headers with different DEBUG_TYPE definitions. This may be mostly an academic violation today, but with modules these types of violations are easy to check for and potentially very relevant. Where necessary to suppor headers with DEBUG_TYPE, I have moved the definitions below the includes in this commit. I plan to move the rest of the DEBUG_TYPE macros in LLVM in subsequent commits; this one is big enough. The comments in Debug.h, which were hilariously out of date already, have been updated to reflect the recommended practice going forward. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@206822 91177308-0d34-0410-b5e6-96231b3b80d8
2014-04-21 22:55:11 +00:00
#define DEBUG_TYPE "mips"
extern "C" void LLVMInitializeMipsTarget() {
// Register the target.
RegisterTargetMachine<MipsebTargetMachine> X(TheMipsTarget);
RegisterTargetMachine<MipselTargetMachine> Y(TheMipselTarget);
RegisterTargetMachine<MipsebTargetMachine> A(TheMips64Target);
RegisterTargetMachine<MipselTargetMachine> B(TheMips64elTarget);
}
// On function prologue, the stack is created by decrementing
// its pointer. Once decremented, all references are done with positive
// offset from the stack/frame pointer, using StackGrowsUp enables
// an easier handling.
Several changes to Mips backend, experimental fp support being the most important. - Cleanup in the Subtarget info with addition of new features, not all support yet, but they allow the future inclusion of features easier. Among new features, we have : Arch family info (mips1, mips2, ...), ABI info (o32, eabi), 64-bit integer and float registers, allegrex vector FPU (VFPU), single float only support. - TargetMachine now detects allegrex core. - Added allegrex (Mips32r2) sext_inreg instructions. - *Added Float Point Instructions*, handling single float only, and aliased accesses for 32-bit FPUs. - Some cleanup in FP instruction formats and FP register classes. - Calling conventions improved to support mips 32-bit EABI. - Added Asm Printer support for fp cond codes. - Added support for sret copy to a return register. - EABI support added into LowerCALL and FORMAL_ARGS. - MipsFunctionInfo now keeps a virtual register per function to track the sret on function entry until function ret. - MipsInstrInfo FP support into methods (isMoveInstr, isLoadFromStackSlot, ...), FP cond codes mapping and initial FP Branch Analysis. - Two new Mips SDNode to handle fp branch and compare instructions : FPBrcond, FPCmp - MipsTargetLowering : handling different FP classes, Allegrex support, sret return copy, no homing location within EABI, non 32-bit stack objects arguments, and asm constraint for float. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@53146 91177308-0d34-0410-b5e6-96231b3b80d8
2008-07-05 19:05:21 +00:00
// Using CodeModel::Large enables different CALL behavior.
MipsTargetMachine::MipsTargetMachine(const Target &T, StringRef TT,
StringRef CPU, StringRef FS,
const TargetOptions &Options,
Reloc::Model RM, CodeModel::Model CM,
CodeGenOpt::Level OL, bool isLittle)
: LLVMTargetMachine(T, TT, CPU, FS, Options, RM, CM, OL),
Subtarget(nullptr), DefaultSubtarget(TT, CPU, FS, isLittle, this),
NoMips16Subtarget(TT, CPU, FS.empty() ? "-mips16" : FS.str() + ",-mips16",
isLittle, this),
Mips16Subtarget(TT, CPU, FS.empty() ? "+mips16" : FS.str() + ",+mips16",
isLittle, this) {
Subtarget = &DefaultSubtarget;
initAsmInfo();
This patch enables llvm to switch between compiling for mips32/mips64 and mips16 on a per function basis. Because this patch is somewhat involved I have provide an overview of the key pieces of it. The patch is written so as to not change the behavior of the non mixed mode. We have tested this a lot but it is something new to switch subtargets so we don't want any chance of regression in the mainline compiler until we have more confidence in this. Mips32/64 are very different from Mip16 as is the case of ARM vs Thumb1. For that reason there are derived versions of the register info, frame info, instruction info and instruction selection classes. Now we register three separate passes for instruction selection. One which is used to switch subtargets (MipsModuleISelDAGToDAG.cpp) and then one for each of the current subtargets (Mips16ISelDAGToDAG.cpp and MipsSEISelDAGToDAG.cpp). When the ModuleISel pass runs, it determines if there is a need to switch subtargets and if so, the owning pointers in MipsTargetMachine are appropriately changed. When 16Isel or SEIsel is run, they will return immediately without doing any work if the current subtarget mode does not apply to them. In addition, MipsAsmPrinter needs to be reset on a function basis. The pass BasicTargetTransformInfo is substituted with a null pass since the pass is immutable and really needs to be a function pass for it to be used with changing subtargets. This will be fixed in a follow on patch. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@179118 91177308-0d34-0410-b5e6-96231b3b80d8
2013-04-09 19:46:01 +00:00
}
void MipsebTargetMachine::anchor() { }
MipsebTargetMachine::
MipsebTargetMachine(const Target &T, StringRef TT,
StringRef CPU, StringRef FS, const TargetOptions &Options,
Reloc::Model RM, CodeModel::Model CM,
CodeGenOpt::Level OL)
: MipsTargetMachine(T, TT, CPU, FS, Options, RM, CM, OL, false) {}
void MipselTargetMachine::anchor() { }
MipselTargetMachine::
MipselTargetMachine(const Target &T, StringRef TT,
StringRef CPU, StringRef FS, const TargetOptions &Options,
Reloc::Model RM, CodeModel::Model CM,
CodeGenOpt::Level OL)
: MipsTargetMachine(T, TT, CPU, FS, Options, RM, CM, OL, true) {}
void MipsTargetMachine::resetSubtarget(MachineFunction *MF) {
DEBUG(dbgs() << "resetSubtarget\n");
AttributeSet FnAttrs = MF->getFunction()->getAttributes();
bool Mips16Attr = FnAttrs.hasAttribute(AttributeSet::FunctionIndex, "mips16");
bool NoMips16Attr =
FnAttrs.hasAttribute(AttributeSet::FunctionIndex, "nomips16");
assert(!(Mips16Attr && NoMips16Attr) &&
"mips16 and nomips16 specified on the same function");
if (Mips16Attr)
Subtarget = &Mips16Subtarget;
else if (NoMips16Attr)
Subtarget = &NoMips16Subtarget;
else
Subtarget = &DefaultSubtarget;
MF->setSubtarget(Subtarget);
return;
}
namespace {
/// Mips Code Generator Pass Configuration Options.
class MipsPassConfig : public TargetPassConfig {
public:
MipsPassConfig(MipsTargetMachine *TM, PassManagerBase &PM)
: TargetPassConfig(TM, PM) {
// The current implementation of long branch pass requires a scratch
// register ($at) to be available before branch instructions. Tail merging
// can break this requirement, so disable it when long branch pass is
// enabled.
EnableTailMerge = !getMipsSubtarget().enableLongBranchPass();
}
MipsTargetMachine &getMipsTargetMachine() const {
return getTM<MipsTargetMachine>();
}
const MipsSubtarget &getMipsSubtarget() const {
return *getMipsTargetMachine().getSubtargetImpl();
}
void addIRPasses() override;
bool addInstSelector() override;
void addMachineSSAOptimization() override;
bool addPreEmitPass() override;
bool addPreRegAlloc() override;
};
} // namespace
TargetPassConfig *MipsTargetMachine::createPassConfig(PassManagerBase &PM) {
return new MipsPassConfig(this, PM);
}
This is for an experimental option -mips-os16. The idea is to compile all Mips32 code as Mips16 unless it can't be compiled as Mips 16. For now this would happen as long as floating point instructions are not needed. Probably it would also make sense to compile as mips32 if atomic operations are needed too. There may be other cases too. A module pass prescans the IR and adds the mips16 or nomips16 attribute to functions depending on the functions needs. Mips 16 mode can result in a 40% code compression by utililizing 16 bit encoding of many instructions. The hope is for this to replace the traditional gcc way of dealing with Mips16 code using floating point which involves essentially using soft float but with a library implemented using mips32 floating point. This gcc method also requires creating stubs so that Mips32 code can interact with these Mips 16 functions that have floating point needs. My conjecture is that in reality this traditional gcc method would never win over this new method. I will be implementing the traditional gcc method also. Some of it is already done but I needed to do the stubs to finish the work and those required this mips16/32 mixed mode capability. I have more ideas for to make this new method much better and I think the old method will just live in llvm for anyone that needs the backward compatibility but I don't for what reason that would be needed. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@179185 91177308-0d34-0410-b5e6-96231b3b80d8
2013-04-10 16:58:04 +00:00
void MipsPassConfig::addIRPasses() {
TargetPassConfig::addIRPasses();
if (getMipsSubtarget().os16())
addPass(createMipsOs16(getMipsTargetMachine()));
if (getMipsSubtarget().inMips16HardFloat())
addPass(createMips16HardFloat(getMipsTargetMachine()));
This is for an experimental option -mips-os16. The idea is to compile all Mips32 code as Mips16 unless it can't be compiled as Mips 16. For now this would happen as long as floating point instructions are not needed. Probably it would also make sense to compile as mips32 if atomic operations are needed too. There may be other cases too. A module pass prescans the IR and adds the mips16 or nomips16 attribute to functions depending on the functions needs. Mips 16 mode can result in a 40% code compression by utililizing 16 bit encoding of many instructions. The hope is for this to replace the traditional gcc way of dealing with Mips16 code using floating point which involves essentially using soft float but with a library implemented using mips32 floating point. This gcc method also requires creating stubs so that Mips32 code can interact with these Mips 16 functions that have floating point needs. My conjecture is that in reality this traditional gcc method would never win over this new method. I will be implementing the traditional gcc method also. Some of it is already done but I needed to do the stubs to finish the work and those required this mips16/32 mixed mode capability. I have more ideas for to make this new method much better and I think the old method will just live in llvm for anyone that needs the backward compatibility but I don't for what reason that would be needed. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@179185 91177308-0d34-0410-b5e6-96231b3b80d8
2013-04-10 16:58:04 +00:00
}
// Install an instruction selector pass using
// the ISelDag to gen Mips code.
bool MipsPassConfig::addInstSelector() {
addPass(createMipsModuleISelDag(getMipsTargetMachine()));
addPass(createMips16ISelDag(getMipsTargetMachine()));
addPass(createMipsSEISelDag(getMipsTargetMachine()));
return false;
}
void MipsPassConfig::addMachineSSAOptimization() {
addPass(createMipsOptimizePICCallPass(getMipsTargetMachine()));
TargetPassConfig::addMachineSSAOptimization();
}
bool MipsPassConfig::addPreRegAlloc() {
if (getOptLevel() == CodeGenOpt::None) {
addPass(createMipsOptimizePICCallPass(getMipsTargetMachine()));
return true;
}
else
return false;
}
This patch enables llvm to switch between compiling for mips32/mips64 and mips16 on a per function basis. Because this patch is somewhat involved I have provide an overview of the key pieces of it. The patch is written so as to not change the behavior of the non mixed mode. We have tested this a lot but it is something new to switch subtargets so we don't want any chance of regression in the mainline compiler until we have more confidence in this. Mips32/64 are very different from Mip16 as is the case of ARM vs Thumb1. For that reason there are derived versions of the register info, frame info, instruction info and instruction selection classes. Now we register three separate passes for instruction selection. One which is used to switch subtargets (MipsModuleISelDAGToDAG.cpp) and then one for each of the current subtargets (Mips16ISelDAGToDAG.cpp and MipsSEISelDAGToDAG.cpp). When the ModuleISel pass runs, it determines if there is a need to switch subtargets and if so, the owning pointers in MipsTargetMachine are appropriately changed. When 16Isel or SEIsel is run, they will return immediately without doing any work if the current subtarget mode does not apply to them. In addition, MipsAsmPrinter needs to be reset on a function basis. The pass BasicTargetTransformInfo is substituted with a null pass since the pass is immutable and really needs to be a function pass for it to be used with changing subtargets. This will be fixed in a follow on patch. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@179118 91177308-0d34-0410-b5e6-96231b3b80d8
2013-04-09 19:46:01 +00:00
void MipsTargetMachine::addAnalysisPasses(PassManagerBase &PM) {
if (Subtarget->allowMixed16_32()) {
This patch enables llvm to switch between compiling for mips32/mips64 and mips16 on a per function basis. Because this patch is somewhat involved I have provide an overview of the key pieces of it. The patch is written so as to not change the behavior of the non mixed mode. We have tested this a lot but it is something new to switch subtargets so we don't want any chance of regression in the mainline compiler until we have more confidence in this. Mips32/64 are very different from Mip16 as is the case of ARM vs Thumb1. For that reason there are derived versions of the register info, frame info, instruction info and instruction selection classes. Now we register three separate passes for instruction selection. One which is used to switch subtargets (MipsModuleISelDAGToDAG.cpp) and then one for each of the current subtargets (Mips16ISelDAGToDAG.cpp and MipsSEISelDAGToDAG.cpp). When the ModuleISel pass runs, it determines if there is a need to switch subtargets and if so, the owning pointers in MipsTargetMachine are appropriately changed. When 16Isel or SEIsel is run, they will return immediately without doing any work if the current subtarget mode does not apply to them. In addition, MipsAsmPrinter needs to be reset on a function basis. The pass BasicTargetTransformInfo is substituted with a null pass since the pass is immutable and really needs to be a function pass for it to be used with changing subtargets. This will be fixed in a follow on patch. git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@179118 91177308-0d34-0410-b5e6-96231b3b80d8
2013-04-09 19:46:01 +00:00
DEBUG(errs() << "No ");
//FIXME: The Basic Target Transform Info
// pass needs to become a function pass instead of
// being an immutable pass and then this method as it exists now
// would be unnecessary.
PM.add(createNoTargetTransformInfoPass());
} else
LLVMTargetMachine::addAnalysisPasses(PM);
DEBUG(errs() << "Target Transform Info Pass Added\n");
}
// Implemented by targets that want to run passes immediately before
// machine code is emitted. return true if -print-machineinstrs should
// print out the code after the passes.
bool MipsPassConfig::addPreEmitPass() {
MipsTargetMachine &TM = getMipsTargetMachine();
addPass(createMipsDelaySlotFillerPass(TM));
addPass(createMipsLongBranchPass(TM));
addPass(createMipsConstantIslandPass(TM));
return true;
}