CollectorMetadata and Collector are rejiggered to get along with

per-function collector model. Collector is now the factory for
CollectorMetadata, so the latter may be subclassed.

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@44827 91177308-0d34-0410-b5e6-96231b3b80d8
This commit is contained in:
Gordon Henriksen 2007-12-11 00:30:17 +00:00
parent 519452282b
commit ad93c4f936
7 changed files with 424 additions and 295 deletions

View File

@ -36,8 +36,10 @@
</ul>
</li>
<li><a href="#intrinsics">Collection intrinsics</a>
<li><a href="#core">Core support</a>
<ul>
<li><a href="#gcattr">Specifying GC code generation:
<tt>gc "..."</tt></a></li>
<li><a href="#gcroot">Identifying GC roots on the stack:
<tt>llvm.gcroot</tt></a></li>
<li><a href="#barriers">Reading and writing references in the heap</a>
@ -198,11 +200,12 @@ garbage collector implementations in two manners:</p>
<ul>
<li>Emitting compatible code, including initialization in the main
program.</li>
program if necessary.</li>
<li>Loading a compiler plugin if the collector is not statically linked with
your compiler. For <tt>llc</tt>, use the <tt>-load</tt> option.</li>
<li>Selecting the collection algorithm with <tt>llc -gc=</tt> or by setting
<tt>llvm::TheCollector</tt>.</li>
<li>Selecting the collection algorithm by applying the <tt>gc "..."</tt>
attribute to your garbage collected functions, or equivalently with
the <tt>setCollector</tt> method.</li>
<li>Linking your final executable with the garbage collector runtime.</li>
</ul>
@ -211,7 +214,7 @@ garbage collector implementations in two manners:</p>
<table>
<tr>
<th>Collector</th>
<th><tt>llc</tt> arguments</th>
<th><tt>gc</tt> attribute</th>
<th>Linkage</th>
<th><tt>gcroot</tt></th>
<th><tt>gcread</tt></th>
@ -219,7 +222,7 @@ garbage collector implementations in two manners:</p>
</tr>
<tr valign="baseline">
<td><a href="#semispace">SemiSpace</a></td>
<td><tt>-gc=shadow-stack</tt></td>
<td><tt>gc "shadow-stack"</tt></td>
<td>TODO FIXME</td>
<td>required</td>
<td>optional</td>
@ -227,7 +230,7 @@ garbage collector implementations in two manners:</p>
</tr>
<tr valign="baseline">
<td><a href="#ocaml">Ocaml</a></td>
<td><tt>-gc=ocaml</tt></td>
<td><tt>gc "ocaml"</tt></td>
<td><i>provided by ocamlopt</i></td>
<td>required</td>
<td>optional</td>
@ -252,11 +255,12 @@ collectors may require user programs to utilize.</p>
<div class="doc_text">
<p>The ShadowStack collector is invoked with <tt>llc -gc=shadow-stack</tt>.
<p>The ShadowStack backend is invoked with the <tt>gc "shadow-stack"</tt>
function attribute.
Unlike many collectors which rely on a cooperative code generator to generate
stack maps, this algorithm carefully maintains a linked list of stack root
descriptors [<a href="#henderson02">Henderson2002</a>]. This so-called "shadow
stack," mirrors the machine stack. Maintaining this data structure is slower
stack" mirrors the machine stack. Maintaining this data structure is slower
than using stack maps, but has a significant portability advantage because it
requires no special support from the target code generator.</p>
@ -264,7 +268,7 @@ requires no special support from the target code generator.</p>
program may use <tt>load</tt> and <tt>store</tt> instead of <tt>llvm.gcread</tt>
and <tt>llvm.gcwrite</tt>.</p>
<p>The ShadowStack collector is a compiler plugin only. It must be paired with a
<p>ShadowStack is a code generator plugin only. It must be paired with a
compatible runtime.</p>
</div>
@ -277,8 +281,7 @@ compatible runtime.</p>
<div class="doc_text">
<p>The SemiSpace runtime implements with the <a href="runtime">suggested
runtime interface</a> and is compatible the ShadowStack collector's code
generation.</p>
runtime interface</a> and is compatible the ShadowStack backend.</p>
<p>SemiSpace is a very simple copying collector. When it starts up, it
allocates two blocks of memory for the heap. It uses a simple bump-pointer
@ -302,7 +305,8 @@ Enhancements would be welcomed.</p>
<div class="doc_text">
<p>The ocaml collector is invoked with <tt>llc -gc=ocaml</tt>. It supports the
<p>The ocaml backend is invoked with the <tt>gc "ocaml"</tt> function attribute.
It supports the
<a href="http://caml.inria.fr/">Objective Caml</a> language runtime by emitting
a type-accurate stack map in the form of an ocaml 3.10.0-compatible frametable.
The linkage requirements are satisfied automatically by the <tt>ocamlopt</tt>
@ -317,7 +321,7 @@ may use <tt>load</tt> and <tt>store</tt> instead of <tt>llvm.gcread</tt> and
<!-- *********************************************************************** -->
<div class="doc_section">
<a name="intrinsics">Collection intrinsics</a>
<a name="core">Core support</a>
</div>
<!-- *********************************************************************** -->
@ -335,6 +339,27 @@ specified by the runtime.</p>
</div>
<!-- ======================================================================= -->
<div class="doc_subsection">
<a name="gcattr">Specifying GC code generation: <tt>gc "..."</tt></a>
</div>
<div class="doc_code"><tt>
define <i>ty</i> @<i>name</i>(...) <u>gc "<i>collector</i>"</u> { ...
</tt></div>
<div class="doc_text">
<p>The <tt>gc</tt> function attribute is used to specify the desired collector
algorithm to the compiler. It is equivalent to specify the collector name
programmatically using the <tt>setCollector</tt> method of
<tt>Function</tt>.</p>
<p>Specifying the collector on a per-function basis allows LLVM to link together
programs which use different garbage collection algorithms.</p>
</div>
<!-- ======================================================================= -->
<div class="doc_subsection">
<a name="gcroot">Identifying GC roots on the stack: <tt>llvm.gcroot</tt></a>
@ -591,6 +616,10 @@ TODO
<div class="doc_text">
<p>User code specifies which collector plugin to use with the <tt>gc</tt>
function attribute or, equivalently, with the <tt>setCollector</tt> method of
<tt>Function</tt>.</p>
<p>To implement a collector plugin, it is necessary to subclass
<tt>llvm::Collector</tt>, which can be accomplished in a few lines of
boilerplate code. LLVM's infrastructure provides access to several important
@ -616,7 +645,7 @@ namespace {
};
CollectorRegistry::Add&lt;MyCollector&gt;
X("mygc", "My custom garbage collector.");
X("mygc", "My bespoke garbage collector.");
}</pre></blockquote>
<p>Using the LLVM makefiles (like the <a
@ -632,20 +661,20 @@ LOADABLE_MODULE = 1
include $(LEVEL)/Makefile.common</pre></blockquote>
<blockquote><pre
></pre></blockquote>
<p>Once the plugin is compiled, user code may be compiled using <tt>llc
-load=<var>MyGC.so</var> -gc=mygc</tt> (though <var>MyGC.so</var> may have some
other platform-specific extension).</p>
<!-- BEGIN FIXME: Gross -->
<p>To use a collector in a tool other than <tt>llc</tt>, simply assign a
<tt>Collector</tt> to the <tt>llvm::TheCollector</tt> variable:</p>
<p>Once the plugin is compiled, code using it may be compiled using <tt>llc
-load=<var>MyGC.so</var></tt> (though <var>MyGC.so</var> may have some other
platform-specific extension):</p>
<blockquote><pre
>TheCollector = new MyGC();</pre></blockquote>
<!-- /FIXME GROSS -->
>$ cat sample.ll
define void @f() gc "mygc" {
entry:
ret void
}
$ llvm-as &lt; sample.ll | llc -load=MyGC.so</pre></blockquote>
<p>It is also possible to statically link the collector plugin into tools, such
as a language-specific compiler front-end.</p>
</div>
@ -956,15 +985,18 @@ interest.</p>
<div class="doc_text">
<blockquote><pre
>CollectorMetadata &amp;MD = ...;
unsigned FrameSize = MD.getFrameSize();
size_t RootCount = MD.roots_size();
>for (iterator I = begin(), E = end(); I != E; ++I) {
CollectorMetadata *MD = *I;
unsigned FrameSize = MD-&gt;getFrameSize();
size_t RootCount = MD-&gt;roots_size();
for (CollectorMetadata::roots_iterator RI = MD.roots_begin(),
RE = MD.roots_end(); RI != RE; ++RI) {
int RootNum = RI->Num;
int RootStackOffset = RI->StackOffset;
Constant *RootMetadata = RI->Metadata;
for (CollectorMetadata::roots_iterator RI = MD-&gt;roots_begin(),
RE = MD-&gt;roots_end();
RI != RE; ++RI) {
int RootNum = RI->Num;
int RootStackOffset = RI->StackOffset;
Constant *RootMetadata = RI->Metadata;
}
}</pre></blockquote>
<p>LLVM automatically computes a stack map. All a <tt>Collector</tt> needs to do
@ -1021,10 +1053,8 @@ public:
CustomWriteBarriers = true;
}
protected:
virtual Pass *createCustomLoweringPass() const {
return new MyGCLoweringFunctionPass();
}
virtual bool initializeCustomLowering(Module &amp;M);
virtual bool performCustomLowering(Function &amp;F);
};</pre></blockquote>
<p>If any of these flags are set, then LLVM suppresses its default lowering for
@ -1041,56 +1071,53 @@ pass specified by the collector.</p>
</ul>
<p>If <tt>CustomReadBarriers</tt> or <tt>CustomWriteBarriers</tt> are specified,
the custom lowering pass <strong>must</strong> eliminate the corresponding
barriers.</p>
then <tt>performCustomLowering</tt> <strong>must</strong> eliminate the
corresponding barriers.</p>
<p>This template can be used as a starting point for a lowering pass:</p>
<p><tt>performCustomLowering</tt>, must comply with the same restrictions as <a
href="WritingAnLLVMPass.html#runOnFunction"><tt>runOnFunction</tt></a>, and
that <tt>initializeCustomLowering</tt> has the same semantics as <a
href="WritingAnLLVMPass.html#doInitialization_mod"><tt>doInitialization(Module
&amp;)</tt></a>.</p>
<p>The following can be used as a template:</p>
<blockquote><pre
>#include "llvm/Function.h"
#include "llvm/Module.h"
>#include "llvm/Module.h"
#include "llvm/Instructions.h"
namespace {
class VISIBILITY_HIDDEN MyGCLoweringFunctionPass : public FunctionPass {
static char ID;
public:
MyGCLoweringFunctionPass() : FunctionPass(intptr_t(&amp;ID)) {}
const char *getPassName() const { return "Lower GC Intrinsics"; }
bool runOnFunction(Function &amp;F) {
Module *M = F.getParent();
Function *GCReadInt = M-&gt;getFunction("llvm.gcread"),
*GCWriteInt = M-&gt;getFunction("llvm.gcwrite"),
*GCRootInt = M-&gt;getFunction("llvm.gcroot");
bool MadeChange = false;
for (Function::iterator BB = F.begin(), E = F.end(); BB != E; ++BB)
for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E;)
if (CallInst *CI = dyn_cast&lt;CallInst&gt;(II++))
if (Function *F = CI->getCalledFunction())
if (F == GCWriteInt) {
// Handle llvm.gcwrite.
CI->eraseFromParent();
MadeChange = true;
} else if (F == GCReadInt) {
// Handle llvm.gcread.
CI->eraseFromParent();
MadeChange = true;
} else if (F == GCRootInt) {
// Handle llvm.gcroot.
CI->eraseFromParent();
MadeChange = true;
}
return MadeChange;
}
};
bool MyCollector::initializeCustomLowering(Module &amp;M) {
return false;
}
char MyGCLoweringFunctionPass::ID = 0;
bool MyCollector::performCustomLowering(Function &amp;F) {
const Module *M = F.getParent();
Function *GCReadInt = M-&gt;getFunction("llvm.gcread"),
*GCWriteInt = M-&gt;getFunction("llvm.gcwrite"),
*GCRootInt = M-&gt;getFunction("llvm.gcroot");
bool MadeChange = false;
for (Function::iterator BB = F.begin(), E = F.end(); BB != E; ++BB)
for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E;)
if (CallInst *CI = dyn_cast&lt;CallInst&gt;(II++))
if (Function *F = CI->getCalledFunction())
if (F == GCWriteInt) {
// Handle llvm.gcwrite.
CI->eraseFromParent();
MadeChange = true;
} else if (F == GCReadInt) {
// Handle llvm.gcread.
CI->eraseFromParent();
MadeChange = true;
} else if (F == GCRootInt) {
// Handle llvm.gcroot.
CI->eraseFromParent();
MadeChange = true;
}
return MadeChange;
}</pre></blockquote>
</div>
@ -1130,15 +1157,18 @@ namespace {
<p>It can then use the following routines to access safe points.</p>
<blockquote><pre>
CollectorMetadata &amp;MD = ...;
size_t PointCount = MD.size();
<blockquote><pre
>for (iterator I = begin(), E = end(); I != E; ++I) {
CollectorMetadata *MD = *I;
size_t PointCount = MD-&gt;size();
for (CollectorMetadata::iterator PI = MD.begin(),
PE = MD.end(); PI != PE; ++PI) {
GC::PointKind PointKind = PI-&gt;Kind;
unsigned PointNum = PI-&gt;Num;
}</pre></blockquote>
for (CollectorMetadata::iterator PI = MD-&gt;begin(),
PE = MD-&gt;end(); PI != PE; ++PI) {
GC::PointKind PointKind = PI-&gt;Kind;
unsigned PointNum = PI-&gt;Num;
}
}
</pre></blockquote>
<p>Almost every collector requires <tt>PostCall</tt> safe points, since these
correspond to the moments when the function is suspended during a call to a
@ -1167,40 +1197,45 @@ safe point (because only the topmost function has been patched).</p>
<p>LLVM allows a collector to print arbitrary assembly code before and after
the rest of a module's assembly code. From the latter callback, the collector
can print stack maps from <tt>CollectorModuleMetadata</tt> populated by the code
generator.</p>
can print stack maps built by the code generator.</p>
<p>Note that LLVM does not currently support garbage collection code generation
in the JIT, nor using the object writers.</p>
<p>Note that LLVM does not currently have analogous APIs to support code
generation in the JIT, nor using the object writers.</p>
<blockquote><pre
>class MyCollector : public Collector {
virtual void beginAssembly(Module &amp;M, std::ostream &amp;OS, AsmPrinter &amp;AP,
const TargetAsmInfo &amp;TAI) const;
public:
virtual void beginAssembly(std::ostream &amp;OS, AsmPrinter &amp;AP,
const TargetAsmInfo &amp;TAI);
virtual void finishAssembly(Module &amp;M, CollectorModuleMetadata &amp;MMD,
std::ostream &amp;OS, AsmPrinter &amp;AP,
const TargetAsmInfo &amp;TAI) const;
virtual void finishAssembly(std::ostream &amp;OS, AsmPrinter &amp;AP,
const TargetAsmInfo &amp;TAI);
}</pre></blockquote>
<p>The collector should use <tt>AsmPrinter</tt> and <tt>TargetAsmInfo</tt> to
print portable assembly code to the <tt>std::ostream</tt>. The collector may
access the stack maps for the entire module using the methods of
<tt>CollectorModuleMetadata</tt>. Here's a realistic example:</p>
print portable assembly code to the <tt>std::ostream</tt>. The collector itself
contains the stack map for the entire module, and may access the
<tt>CollectorMetadata</tt> using its own <tt>begin()</tt> and <tt>end()</tt>
methods. Here's a realistic example:</p>
<blockquote><pre
>#include "llvm/CodeGen/AsmPrinter.h"
#include "llvm/Function.h"
#include "llvm/Target/TargetMachine.h"
#include "llvm/Target/TargetData.h"
#include "llvm/Target/TargetAsmInfo.h"
void MyCollector::finishAssembly(Module &amp;M,
CollectorModuleMetadata &amp;MMD,
std::ostream &amp;OS, AsmPrinter &amp;AP,
const TargetAsmInfo &amp;TAI) const {
void MyCollector::beginAssembly(std::ostream &amp;OS, AsmPrinter &amp;AP,
const TargetAsmInfo &amp;TAI) {
// Nothing to do.
}
void MyCollector::finishAssembly(std::ostream &amp;OS, AsmPrinter &amp;AP,
const TargetAsmInfo &amp;TAI) {
// Set up for emitting addresses.
const char *AddressDirective;
int AddressAlignLog;
if (TAI.getAddressSize() == sizeof(int32_t)) {
if (AP.TM.getTargetData()->getPointerSize() == sizeof(int32_t)) {
AddressDirective = TAI.getData32bitsDirective();
AddressAlignLog = 2;
} else {
@ -1212,8 +1247,7 @@ void MyCollector::finishAssembly(Module &amp;M,
AP.SwitchToDataSection(TAI.getDataSection());
// For each function...
for (CollectorModuleMetadata::iterator FI = MMD.begin(),
FE = MMD.end(); FI != FE; ++FI) {
for (iterator FI = begin(), FE = end(); FI != FE; ++FI) {
CollectorMetadata &amp;MD = **FI;
// Emit this data structure:

View File

@ -7,7 +7,7 @@
//
//===----------------------------------------------------------------------===//
//
// GCInfo records sufficient information about a machine function to enable
// Collector records sufficient information about a machine function to enable
// accurate garbage collectors. Specifically:
//
// - Safe points
@ -25,8 +25,8 @@
// This generic information should used by ABI-specific passes to emit support
// tables for the runtime garbage collector.
//
// GCSafePointPass identifies the GC safe points in the machine code. (Roots are
// identified in SelectionDAGISel.)
// MachineCodeAnalysis identifies the GC safe points in the machine code. (Roots
// are identified in SelectionDAGISel.)
//
//===----------------------------------------------------------------------===//
@ -35,19 +35,25 @@
#include "llvm/CodeGen/CollectorMetadata.h"
#include <iosfwd>
#include <string>
namespace llvm {
class AsmPrinter;
class FunctionPassManager;
class PassManager;
class TargetAsmInfo;
/// Collector describes a garbage collector's code generation requirements,
/// and provides overridable hooks for those needs which cannot be abstractly
/// described.
class Collector {
public:
typedef std::vector<CollectorMetadata*> list_type;
typedef list_type::iterator iterator;
private:
friend class CollectorModuleMetadata;
const Module *M;
std::string Name;
list_type Functions;
protected:
unsigned NeededSafePoints; //< Bitmask of required safe points.
bool CustomReadBarriers; //< Default is to insert loads.
@ -55,16 +61,20 @@ namespace llvm {
bool CustomRoots; //< Default is to pass through to backend.
bool InitRoots; //< If set, roots are nulled during lowering.
/// If any of the actions are set to Custom, this is expected to be
/// overriden to create a transform to lower those actions to LLVM IR.
virtual Pass *createCustomLoweringPass() const;
public:
Collector();
virtual ~Collector();
/// getName - The name of the collector, for debugging.
///
const std::string &getName() const { return Name; }
/// getModule - The module upon which the collector is operating.
///
const Module &getModule() const { return *M; }
/// True if this collector requires safe points of any kind. By default,
/// none are recorded.
bool needsSafePoints() const { return NeededSafePoints != 0; }
@ -94,40 +104,30 @@ namespace llvm {
bool initializeRoots() const { return InitRoots; }
/// Adds LLVM IR transforms to handle collection intrinsics. By default,
/// read- and write barriers are replaced with direct memory accesses, and
/// roots are passed on to the code generator.
void addLoweringPasses(FunctionPassManager &PM) const;
/// Same as addLoweringPasses(FunctionPassManager &), except uses a
/// PassManager for compatibility with unusual backends (such as MSIL or
/// CBackend).
void addLoweringPasses(PassManager &PM) const;
/// Adds target-independent MachineFunction pass to mark safe points. This
/// is added very late during code generation, just prior to output, and
/// importantly after all CFG transformations (like branch folding).
void addGenericMachineCodePass(FunctionPassManager &PM,
const TargetMachine &TM, bool Fast) const;
/// beginAssembly/finishAssembly - Emit module metadata as assembly code.
virtual void beginAssembly(Module &M, std::ostream &OS, AsmPrinter &AP,
const TargetAsmInfo &TAI) const;
virtual void finishAssembly(Module &M, CollectorModuleMetadata &CMM,
std::ostream &OS, AsmPrinter &AP,
const TargetAsmInfo &TAI) const;
private:
bool NeedsDefaultLoweringPass() const;
bool NeedsCustomLoweringPass() const;
virtual void beginAssembly(std::ostream &OS, AsmPrinter &AP,
const TargetAsmInfo &TAI);
virtual void finishAssembly(std::ostream &OS, AsmPrinter &AP,
const TargetAsmInfo &TAI);
/// begin/end - Iterators for function metadata.
///
iterator begin() { return Functions.begin(); }
iterator end() { return Functions.end(); }
/// insertFunctionMetadata - Creates metadata for a function.
///
CollectorMetadata *insertFunctionMetadata(const Function &F);
/// initializeCustomLowering/performCustomLowering - If any of the actions
/// are set to custom, performCustomLowering must be overriden to create a
/// transform to lower those actions to LLVM IR. initializeCustomLowering
/// is optional to override. These are the only Collector methods through
/// which the LLVM IR can be modified.
virtual bool initializeCustomLowering(Module &F);
virtual bool performCustomLowering(Function &F);
};
/// If set, the code generator should generate garbage collection as specified
/// by the collector properties.
extern const Collector *TheCollector; // FIXME: Find a better home!
}
#endif

View File

@ -9,7 +9,7 @@
//
// This file declares the CollectorMetadata and CollectorModuleMetadata classes,
// which are used as a communication channel from the target code generator
// to the target garbage collector. This interface allows code generators and
// to the target garbage collectors. This interface allows code generators and
// garbage collectors to be developed independently.
//
// The CollectorMetadata class records the data necessary to build a type
@ -37,19 +37,14 @@
#include "llvm/Pass.h"
#include "llvm/ADT/DenseMap.h"
#include "llvm/ADT/StringMap.h"
namespace llvm {
class AsmPrinter;
class Collector;
class Constant;
/// Creates a pass to print collector metadata.
///
Pass *createCollectorMetadataPrinter(std::ostream &OS);
/// Creates a pass to destroy collector metadata.
///
Pass *createCollectorMetadataDeleter();
class TargetAsmInfo;
namespace GC {
@ -77,7 +72,7 @@ namespace llvm {
struct GCRoot {
int Num; //< Usually a frame index.
int StackOffset; //< Offset from the stack pointer.
Constant *Metadata; //< From the call to llvm.gcroot.
Constant *Metadata; //< Metadata straight from the call to llvm.gcroot.
GCRoot(int N, Constant *MD) : Num(N), StackOffset(-1), Metadata(MD) {}
};
@ -93,6 +88,7 @@ namespace llvm {
private:
const Function &F;
Collector &C;
uint64_t FrameSize;
std::vector<GCRoot> Roots;
std::vector<GCPoint> SafePoints;
@ -107,14 +103,18 @@ namespace llvm {
// The bit vector is the more compact representation where >3.2% of roots
// are live per safe point (1.5% on 64-bit hosts).
friend class CollectorModuleMetadata;
CollectorMetadata(const Function &F);
public:
CollectorMetadata(const Function &F, Collector &C);
~CollectorMetadata();
/// getFunction - Return the function to which this metadata applies.
///
const Function &getFunction() const { return F; }
/// getCollector - Return the collector for the function.
///
Collector &getCollector() { return C; }
/// addStackRoot - Registers a root that lives on the stack. Num is the
/// stack object ID for the alloca (if the code generator is using
/// MachineFrameInfo).
@ -157,37 +157,36 @@ namespace llvm {
/// CollectorModuleMetadata - Garbage collection metadata for a whole module.
///
class CollectorModuleMetadata : public ImmutablePass {
typedef std::vector<CollectorMetadata*> list_type;
typedef DenseMap<const Function*,CollectorMetadata*> map_type;
typedef StringMap<Collector*> collector_map_type;
typedef std::vector<Collector*> list_type;
typedef DenseMap<const Function*,CollectorMetadata*> function_map_type;
Module *Mod;
list_type Functions;
map_type Map;
collector_map_type NameMap;
list_type Collectors;
function_map_type Map;
Collector *getOrCreateCollector(const Module *M, const std::string &Name);
public:
typedef list_type::iterator iterator;
typedef list_type::const_iterator iterator;
static char ID;
CollectorModuleMetadata();
~CollectorModuleMetadata();
/// clear - Used to delete module metadata. Collector invokes this as
/// necessary.
/// clear - Used to delete module metadata. The metadata deleter pass calls
/// this.
void clear();
/// begin/end - Iterators for function metadata.
/// begin/end - Iterators for collectors.
///
iterator begin() { return Functions.begin(); }
iterator end() { return Functions.end(); }
iterator begin() const { return Collectors.begin(); }
iterator end() const { return Collectors.end(); }
/// insert - Creates metadata for a function.
/// get - Look up function metadata.
///
CollectorMetadata& insert(const Function *F);
/// get - Looks up existing function metadata.
///
CollectorMetadata* get(const Function *F) const;
CollectorMetadata &get(const Function &F);
};
}

View File

@ -25,6 +25,9 @@ namespace llvm {
///
typedef Registry<Collector> CollectorRegistry;
/// FIXME: Collector instances are not useful on their own. These no longer
/// serve any purpose except to link in the plugins.
/// Creates an ocaml-compatible garbage collector.
Collector *createOcamlCollector();

View File

@ -135,6 +135,24 @@ namespace llvm {
/// for the Sparc.
FunctionPass *getRegisterAllocator(TargetMachine &T);
/// IntrinsicLowering Pass - Performs target-independent LLVM IR
/// transformations for highly portable collectors.
FunctionPass *createGCLoweringPass();
/// MachineCodeAnalysis Pass - Target-independent pass to mark safe points in
/// machine code. Must be added very late during code generation, just prior
/// to output, and importantly after all CFG transformations (such as branch
/// folding).
FunctionPass *createGCMachineCodeAnalysisPass();
/// Deleter Pass - Releases collector metadata.
///
FunctionPass *createCollectorMetadataDeleter();
/// Creates a pass to print collector metadata.
///
FunctionPass *createCollectorMetadataPrinter(std::ostream &OS);
/// createMachineLICMPass - This pass performs LICM on machine instructions.
///
FunctionPass *createMachineLICMPass();

View File

@ -13,6 +13,7 @@
//===----------------------------------------------------------------------===//
#include "llvm/CodeGen/Collector.h"
#include "llvm/CodeGen/Passes.h"
#include "llvm/IntrinsicInst.h"
#include "llvm/Module.h"
#include "llvm/PassManager.h"
@ -29,38 +30,40 @@ using namespace llvm;
namespace {
/// This pass rewrites calls to the llvm.gcread or llvm.gcwrite intrinsics,
/// replacing them with simple loads and stores as directed by the Collector.
/// This is useful for most garbage collectors.
/// LowerIntrinsics - This pass rewrites calls to the llvm.gcread or
/// llvm.gcwrite intrinsics, replacing them with simple loads and stores as
/// directed by the Collector. It also performs automatic root initialization
/// and custom intrinsic lowering.
class VISIBILITY_HIDDEN LowerIntrinsics : public FunctionPass {
const Collector &Coll;
/// GCRootInt, GCReadInt, GCWriteInt - The function prototypes for the
/// llvm.gc* intrinsics.
Function *GCRootInt, *GCReadInt, *GCWriteInt;
static bool NeedsDefaultLoweringPass(const Collector &C);
static bool NeedsCustomLoweringPass(const Collector &C);
static bool CouldBecomeSafePoint(Instruction *I);
static void InsertRootInitializers(Function &F,
bool PerformDefaultLowering(Function &F, Collector &Coll);
static bool InsertRootInitializers(Function &F,
AllocaInst **Roots, unsigned Count);
public:
static char ID;
LowerIntrinsics(const Collector &GC);
LowerIntrinsics();
const char *getPassName() const;
void getAnalysisUsage(AnalysisUsage &AU) const;
bool doInitialization(Module &M);
bool runOnFunction(Function &F);
};
/// This is a target-independent pass over the machine function representation
/// to identify safe points for the garbage collector in the machine code. It
/// inserts labels at safe points and populates the GCInfo class.
/// MachineCodeAnalysis - This is a target-independent pass over the machine
/// function representation to identify safe points for the garbage collector
/// in the machine code. It inserts labels at safe points and populates a
/// CollectorMetadata record for each function.
class VISIBILITY_HIDDEN MachineCodeAnalysis : public MachineFunctionPass {
const Collector &Coll;
const TargetMachine &Targ;
const TargetMachine *TM;
CollectorMetadata *MD;
MachineModuleInfo *MMI;
const TargetInstrInfo *TII;
@ -76,7 +79,7 @@ namespace {
public:
static char ID;
MachineCodeAnalysis(const Collector &C, const TargetMachine &T);
MachineCodeAnalysis();
const char *getPassName() const;
void getAnalysisUsage(AnalysisUsage &AU) const;
@ -87,8 +90,6 @@ namespace {
// -----------------------------------------------------------------------------
const Collector *llvm::TheCollector = 0;
Collector::Collector() :
NeededSafePoints(0),
CustomReadBarriers(false),
@ -97,85 +98,85 @@ Collector::Collector() :
InitRoots(true)
{}
Collector::~Collector() {}
void Collector::addLoweringPasses(FunctionPassManager &PM) const {
if (NeedsDefaultLoweringPass())
PM.add(new LowerIntrinsics(*this));
if (NeedsCustomLoweringPass())
PM.add(createCustomLoweringPass());
Collector::~Collector() {
for (iterator I = begin(), E = end(); I != E; ++I)
delete *I;
Functions.clear();
}
void Collector::addLoweringPasses(PassManager &PM) const {
if (NeedsDefaultLoweringPass())
PM.add(new LowerIntrinsics(*this));
if (NeedsCustomLoweringPass())
PM.add(createCustomLoweringPass());
}
void Collector::addGenericMachineCodePass(FunctionPassManager &PM,
const TargetMachine &TM,
bool Fast) const {
if (needsSafePoints())
PM.add(new MachineCodeAnalysis(*this, TM));
}
bool Collector::NeedsDefaultLoweringPass() const {
// Default lowering is necessary only if read or write barriers have a default
// action. The default for roots is no action.
return !customWriteBarrier()
|| !customReadBarrier()
|| initializeRoots();
}
bool Collector::NeedsCustomLoweringPass() const {
// Custom lowering is only necessary if enabled for some action.
return customWriteBarrier()
|| customReadBarrier()
|| customRoots();
}
Pass *Collector::createCustomLoweringPass() const {
cerr << "Collector must override createCustomLoweringPass.\n";
bool Collector::initializeCustomLowering(Module &M) { return false; }
bool Collector::performCustomLowering(Function &F) {
cerr << "gc " << getName() << " must override performCustomLowering.\n";
abort();
return 0;
}
void Collector::beginAssembly(Module &M, std::ostream &OS, AsmPrinter &AP,
const TargetAsmInfo &TAI) const {
void Collector::beginAssembly(std::ostream &OS, AsmPrinter &AP,
const TargetAsmInfo &TAI) {
// Default is no action.
}
void Collector::finishAssembly(Module &M, CollectorModuleMetadata &CMM,
std::ostream &OS, AsmPrinter &AP,
const TargetAsmInfo &TAI) const {
void Collector::finishAssembly(std::ostream &OS, AsmPrinter &AP,
const TargetAsmInfo &TAI) {
// Default is no action.
}
CollectorMetadata *Collector::insertFunctionMetadata(const Function &F) {
CollectorMetadata *CM = new CollectorMetadata(F, *this);
Functions.push_back(CM);
return CM;
}
// -----------------------------------------------------------------------------
FunctionPass *llvm::createGCLoweringPass() {
return new LowerIntrinsics();
}
char LowerIntrinsics::ID = 0;
LowerIntrinsics::LowerIntrinsics(const Collector &C)
: FunctionPass((intptr_t)&ID), Coll(C),
LowerIntrinsics::LowerIntrinsics()
: FunctionPass((intptr_t)&ID),
GCRootInt(0), GCReadInt(0), GCWriteInt(0) {}
const char *LowerIntrinsics::getPassName() const {
return "Lower Garbage Collection Instructions";
}
/// doInitialization - If this module uses the GC intrinsics, find them now. If
/// not, this pass does not do anything.
void LowerIntrinsics::getAnalysisUsage(AnalysisUsage &AU) const {
FunctionPass::getAnalysisUsage(AU);
AU.addRequired<CollectorModuleMetadata>();
}
/// doInitialization - If this module uses the GC intrinsics, find them now.
bool LowerIntrinsics::doInitialization(Module &M) {
GCReadInt = M.getFunction("llvm.gcread");
GCWriteInt = M.getFunction("llvm.gcwrite");
GCRootInt = M.getFunction("llvm.gcroot");
return false;
// FIXME: This is rather antisocial in the context of a JIT since it performs
// work against the entire module. But this cannot be done at
// runFunction time (initializeCustomLowering likely needs to change
// the module).
CollectorModuleMetadata *CMM = getAnalysisToUpdate<CollectorModuleMetadata>();
assert(CMM && "LowerIntrinsics didn't require CollectorModuleMetadata!?");
for (Module::iterator I = M.begin(), E = M.end(); I != E; ++I)
if (I->hasCollector())
CMM->get(*I); // Instantiate the Collector.
bool MadeChange = false;
for (CollectorModuleMetadata::iterator I = CMM->begin(),
E = CMM->end(); I != E; ++I)
if (NeedsCustomLoweringPass(**I))
if ((*I)->initializeCustomLowering(M))
MadeChange = true;
return MadeChange;
}
void LowerIntrinsics::InsertRootInitializers(Function &F, AllocaInst **Roots,
bool LowerIntrinsics::InsertRootInitializers(Function &F, AllocaInst **Roots,
unsigned Count) {
// Scroll past alloca instructions.
BasicBlock::iterator IP = F.getEntryBlock().begin();
@ -190,11 +191,32 @@ void LowerIntrinsics::InsertRootInitializers(Function &F, AllocaInst **Roots,
InitedRoots.insert(AI);
// Add root initializers.
bool MadeChange = false;
for (AllocaInst **I = Roots, **E = Roots + Count; I != E; ++I)
if (!InitedRoots.count(*I))
if (!InitedRoots.count(*I)) {
new StoreInst(ConstantPointerNull::get(cast<PointerType>(
cast<PointerType>((*I)->getType())->getElementType())),
*I, IP);
MadeChange = true;
}
return MadeChange;
}
bool LowerIntrinsics::NeedsDefaultLoweringPass(const Collector &C) {
// Default lowering is necessary only if read or write barriers have a default
// action. The default for roots is no action.
return !C.customWriteBarrier()
|| !C.customReadBarrier()
|| C.initializeRoots();
}
bool LowerIntrinsics::NeedsCustomLoweringPass(const Collector &C) {
// Custom lowering is only necessary if enabled for some action.
return C.customWriteBarrier()
|| C.customReadBarrier()
|| C.customRoots();
}
/// CouldBecomeSafePoint - Predicate to conservatively determine whether the
@ -228,9 +250,24 @@ bool LowerIntrinsics::CouldBecomeSafePoint(Instruction *I) {
/// runOnFunction - Replace gcread/gcwrite intrinsics with loads and stores.
/// Leave gcroot intrinsics; the code generator needs to see those.
bool LowerIntrinsics::runOnFunction(Function &F) {
// Quick exit for programs that do not declare the intrinsics.
if (!GCReadInt && !GCWriteInt && !GCRootInt) return false;
// Quick exit for functions that do not use GC.
if (!F.hasCollector()) return false;
CollectorMetadata &MD = getAnalysis<CollectorModuleMetadata>().get(F);
Collector &Coll = MD.getCollector();
bool MadeChange = false;
if (NeedsDefaultLoweringPass(Coll))
MadeChange |= PerformDefaultLowering(F, Coll);
if (NeedsCustomLoweringPass(Coll))
MadeChange |= Coll.performCustomLowering(F);
return MadeChange;
}
bool LowerIntrinsics::PerformDefaultLowering(Function &F, Collector &Coll) {
bool LowerWr = !Coll.customWriteBarrier();
bool LowerRd = !Coll.customReadBarrier();
bool InitRoots = Coll.initializeRoots();
@ -268,17 +305,21 @@ bool LowerIntrinsics::runOnFunction(Function &F) {
}
if (Roots.size())
InsertRootInitializers(F, Roots.begin(), Roots.size());
MadeChange |= InsertRootInitializers(F, Roots.begin(), Roots.size());
return MadeChange;
}
// -----------------------------------------------------------------------------
FunctionPass *llvm::createGCMachineCodeAnalysisPass() {
return new MachineCodeAnalysis();
}
char MachineCodeAnalysis::ID = 0;
MachineCodeAnalysis::MachineCodeAnalysis(const Collector &C, const TargetMachine &T)
: MachineFunctionPass(intptr_t(&ID)), Coll(C), Targ(T) {}
MachineCodeAnalysis::MachineCodeAnalysis()
: MachineFunctionPass(intptr_t(&ID)) {}
const char *MachineCodeAnalysis::getPassName() const {
return "Analyze Machine Code For Garbage Collection";
@ -304,10 +345,10 @@ void MachineCodeAnalysis::VisitCallPoint(MachineBasicBlock::iterator CI) {
MachineBasicBlock::iterator RAI = CI;
++RAI;
if (Coll.needsSafePoint(GC::PreCall))
if (MD->getCollector().needsSafePoint(GC::PreCall))
MD->addSafePoint(GC::PreCall, InsertLabel(*CI->getParent(), CI));
if (Coll.needsSafePoint(GC::PostCall))
if (MD->getCollector().needsSafePoint(GC::PostCall))
MD->addSafePoint(GC::PostCall, InsertLabel(*CI->getParent(), RAI));
}
@ -323,7 +364,7 @@ void MachineCodeAnalysis::FindSafePoints(MachineFunction &MF) {
void MachineCodeAnalysis::FindStackOffsets(MachineFunction &MF) {
uint64_t StackSize = MFI->getStackSize();
uint64_t OffsetAdjustment = MFI->getOffsetAdjustment();
uint64_t OffsetOfLocalArea = Targ.getFrameInfo()->getOffsetOfLocalArea();
uint64_t OffsetOfLocalArea = TM->getFrameInfo()->getOffsetOfLocalArea();
for (CollectorMetadata::roots_iterator RI = MD->roots_begin(),
RE = MD->roots_end(); RI != RE; ++RI)
@ -332,12 +373,16 @@ void MachineCodeAnalysis::FindStackOffsets(MachineFunction &MF) {
}
bool MachineCodeAnalysis::runOnMachineFunction(MachineFunction &MF) {
if (!Coll.needsSafePoints())
// Quick exit for functions that do not use GC.
if (!MF.getFunction()->hasCollector()) return false;
MD = &getAnalysis<CollectorModuleMetadata>().get(*MF.getFunction());
if (!MD->getCollector().needsSafePoints())
return false;
MD = getAnalysis<CollectorModuleMetadata>().get(MF.getFunction());
TM = &MF.getTarget();
MMI = &getAnalysis<MachineModuleInfo>();
TII = MF.getTarget().getInstrInfo();
TII = TM->getInstrInfo();
MFI = MF.getFrameInfo();
// Find the size of the stack frame.

View File

@ -13,8 +13,11 @@
//===----------------------------------------------------------------------===//
#include "llvm/CodeGen/CollectorMetadata.h"
#include "llvm/CodeGen/Collector.h"
#include "llvm/CodeGen/Collectors.h"
#include "llvm/CodeGen/MachineFrameInfo.h"
#include "llvm/CodeGen/MachineFunctionPass.h"
#include "llvm/Pass.h"
#include "llvm/CodeGen/Passes.h"
#include "llvm/Function.h"
#include "llvm/Support/Compiler.h"
@ -22,7 +25,7 @@ using namespace llvm;
namespace {
class VISIBILITY_HIDDEN Printer : public MachineFunctionPass {
class VISIBILITY_HIDDEN Printer : public FunctionPass {
static char ID;
std::ostream &OS;
@ -32,10 +35,10 @@ namespace {
const char *getPassName() const;
void getAnalysisUsage(AnalysisUsage &AU) const;
bool runOnMachineFunction(MachineFunction &MF);
bool runOnFunction(Function &F);
};
class VISIBILITY_HIDDEN Deleter : public MachineFunctionPass {
class VISIBILITY_HIDDEN Deleter : public FunctionPass {
static char ID;
public:
@ -44,7 +47,7 @@ namespace {
const char *getPassName() const;
void getAnalysisUsage(AnalysisUsage &AU) const;
bool runOnMachineFunction(MachineFunction &MF);
bool runOnFunction(Function &F);
bool doFinalization(Module &M);
};
@ -55,8 +58,8 @@ namespace {
// -----------------------------------------------------------------------------
CollectorMetadata::CollectorMetadata(const Function &F)
: F(F), FrameSize(~0LL) {}
CollectorMetadata::CollectorMetadata(const Function &F, Collector &C)
: F(F), C(C), FrameSize(~0LL) {}
CollectorMetadata::~CollectorMetadata() {}
@ -71,46 +74,71 @@ CollectorModuleMetadata::~CollectorModuleMetadata() {
clear();
}
CollectorMetadata& CollectorModuleMetadata::insert(const Function *F) {
assert(Map.find(F) == Map.end() && "Function GC metadata already exists!");
CollectorMetadata *FMD = new CollectorMetadata(*F);
Functions.push_back(FMD);
Map[F] = FMD;
return *FMD;
Collector *CollectorModuleMetadata::
getOrCreateCollector(const Module *M, const std::string &Name) {
const char *Start = Name.c_str();
collector_map_type::iterator NMI = NameMap.find(Start, Start + Name.size());
if (NMI != NameMap.end())
return NMI->getValue();
for (CollectorRegistry::iterator I = CollectorRegistry::begin(),
E = CollectorRegistry::end(); I != E; ++I) {
if (strcmp(Start, I->getName()) == 0) {
Collector *C = I->instantiate();
C->M = M;
C->Name = Name;
NameMap.GetOrCreateValue(Start, Start + Name.size()).setValue(C);
Collectors.push_back(C);
return C;
}
}
cerr << "unsupported collector: " << Name << "\n";
abort();
}
CollectorMetadata* CollectorModuleMetadata::get(const Function *F) const {
map_type::iterator I = Map.find(F);
if (I == Map.end())
return 0;
return I->second;
CollectorMetadata &CollectorModuleMetadata::get(const Function &F) {
assert(F.hasCollector());
function_map_type::iterator I = Map.find(&F);
if (I != Map.end())
return *I->second;
Collector *C = getOrCreateCollector(F.getParent(), F.getCollector());
CollectorMetadata *MD = C->insertFunctionMetadata(F);
Map[&F] = MD;
return *MD;
}
void CollectorModuleMetadata::clear() {
Map.clear();
// TODO: StringMap should provide a clear method.
while (!NameMap.empty())
NameMap.erase(NameMap.begin());
for (iterator I = begin(), E = end(); I != E; ++I)
delete *I;
Functions.clear();
Map.clear();
Collectors.clear();
}
// -----------------------------------------------------------------------------
char Printer::ID = 0;
Pass *llvm::createCollectorMetadataPrinter(std::ostream &OS) {
FunctionPass *llvm::createCollectorMetadataPrinter(std::ostream &OS) {
return new Printer(OS);
}
Printer::Printer(std::ostream &OS)
: MachineFunctionPass(intptr_t(&ID)), OS(OS) {}
: FunctionPass(intptr_t(&ID)), OS(OS) {}
const char *Printer::getPassName() const {
return "Print Garbage Collector Information";
}
void Printer::getAnalysisUsage(AnalysisUsage &AU) const {
MachineFunctionPass::getAnalysisUsage(AU);
FunctionPass::getAnalysisUsage(AU);
AU.setPreservesAll();
AU.addRequired<CollectorModuleMetadata>();
}
@ -125,9 +153,9 @@ static const char *DescKind(GC::PointKind Kind) {
}
}
bool Printer::runOnMachineFunction(MachineFunction &MF) {
if (CollectorMetadata *FD =
getAnalysis<CollectorModuleMetadata>().get(MF.getFunction())) {
bool Printer::runOnFunction(Function &F) {
if (F.hasCollector()) {
CollectorMetadata *FD = &getAnalysis<CollectorModuleMetadata>().get(F);
OS << "GC roots for " << FD->getFunction().getNameStart() << ":\n";
for (CollectorMetadata::roots_iterator RI = FD->roots_begin(),
@ -160,11 +188,11 @@ bool Printer::runOnMachineFunction(MachineFunction &MF) {
char Deleter::ID = 0;
Pass *llvm::createCollectorMetadataDeleter() {
FunctionPass *llvm::createCollectorMetadataDeleter() {
return new Deleter();
}
Deleter::Deleter() : MachineFunctionPass(intptr_t(&ID)) {}
Deleter::Deleter() : FunctionPass(intptr_t(&ID)) {}
const char *Deleter::getPassName() const {
return "Delete Garbage Collector Information";
@ -175,11 +203,13 @@ void Deleter::getAnalysisUsage(AnalysisUsage &AU) const {
AU.addRequired<CollectorModuleMetadata>();
}
bool Deleter::runOnMachineFunction(MachineFunction &MF) {
bool Deleter::runOnFunction(Function &MF) {
return false;
}
bool Deleter::doFinalization(Module &M) {
getAnalysis<CollectorModuleMetadata>().clear();
CollectorModuleMetadata *CMM = getAnalysisToUpdate<CollectorModuleMetadata>();
assert(CMM && "Deleter didn't require CollectorModuleMetadata?!");
CMM->clear();
return false;
}