maxjhandsome
/
llvm-project
forked from OSchip/llvm-project
1
0
Fork 0
Code Issues Pull Requests Packages Releases Wiki Activity
llvm-project/lldb/source/Expression/IRExecutionUnit.cpp

705 lines
22 KiB
C++
Raw Normal View History

Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
//===-- IRExecutionUnit.cpp -------------------------------------*- C++ -*-===//
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
//
// The LLVM Compiler Infrastructure
//
// This file is distributed under the University of Illinois Open Source
// License. See LICENSE.TXT for details.
//
//===----------------------------------------------------------------------===//
// C Includes
// C++ Includes
// Other libraries and framework includes
#include "llvm/ExecutionEngine/ExecutionEngine.h"
Fixed a problem where inline assembly errors caused LLDB to crash. <rdar://problem/13497915> llvm-svn: 178115
2013-03-27 11:09:55 +08:00
#include "llvm/IR/LLVMContext.h"
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
#include "llvm/IR/Module.h"
Fixed a problem where inline assembly errors caused LLDB to crash. <rdar://problem/13497915> llvm-svn: 178115
2013-03-27 11:09:55 +08:00
#include "llvm/Support/SourceMgr.h"
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
// Project includes
#include "lldb/Core/DataBufferHeap.h"
#include "lldb/Core/DataExtractor.h"
#include "lldb/Core/Disassembler.h"
#include "lldb/Core/Log.h"
#include "lldb/Expression/IRExecutionUnit.h"
#include "lldb/Target/ExecutionContext.h"
#include "lldb/Target/Target.h"
using namespace lldb_private;
After discussing with Chris Lattner, we require C++11, so lets get rid of the macros and just use C++11. llvm-svn: 179805
2013-04-19 06:45:39 +08:00
IRExecutionUnit::IRExecutionUnit (std::unique_ptr<llvm::LLVMContext> &context_ap,
std::unique_ptr<llvm::Module> &module_ap,
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
ConstString &name,
Audited the existing Materializer code to ensure that it works in the absence of a process. Codepaths in the Materializer now use the best execution context scope available to them. llvm-svn: 179539
2013-04-16 01:12:47 +08:00
const lldb::TargetSP &target_sp,
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
std::vector<std::string> &cpu_features) :
Audited the existing Materializer code to ensure that it works in the absence of a process. Codepaths in the Materializer now use the best execution context scope available to them. llvm-svn: 179539
2013-04-16 01:12:47 +08:00
IRMemoryMap(target_sp),
Since we use C++11, we should switch over to using std::unique_ptr when C++11 is being used. To do this, we follow what we have done for shared pointers and we define a STD_UNIQUE_PTR macro that can be used and it will "do the right thing". Due to some API differences in std::unique_ptr and due to the fact that we need to be able to compile without C++11, we can't use move semantics so some code needed to change so that it can compile with either C++. Anyone wanting to use a unique_ptr or auto_ptr should now use the "STD_UNIQUE_PTR(TYPE)" macro. llvm-svn: 179779
2013-04-19 02:10:51 +08:00
m_context_ap(context_ap.release()),
m_module_ap(module_ap.release()),
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
m_module(m_module_ap.get()),
Updated the IRExecutionUnit to keep local copies of the data it writes down into the process even if the process doesn't exist. This will allow the IR interpreter to access static data allocated on the expression's behalf. Also cleaned up object ownership in the IRExecutionUnit so that allocations are created into the allocations vector. This avoids needless data copies. <rdar://problem/13424594> llvm-svn: 177456
2013-03-20 07:03:21 +08:00
m_cpu_features(cpu_features),
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
m_name(name),
m_did_jit(false),
m_function_load_addr(LLDB_INVALID_ADDRESS),
m_function_end_load_addr(LLDB_INVALID_ADDRESS)
{
}
lldb::addr_t
IRExecutionUnit::WriteNow (const uint8_t *bytes,
size_t size,
Error &error)
{
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
lldb::addr_t allocation_process_addr = Malloc (size,
8,
lldb::ePermissionsWritable | lldb::ePermissionsReadable,
eAllocationPolicyMirror,
error);
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
if (!error.Success())
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
return LLDB_INVALID_ADDRESS;
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
WriteMemory(allocation_process_addr, bytes, size, error);
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
if (!error.Success())
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
{
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
Error err;
Free (allocation_process_addr, err);
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
return LLDB_INVALID_ADDRESS;
}
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
if (Log *log = lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_EXPRESSIONS))
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
{
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
DataBufferHeap my_buffer(size, 0);
Error err;
ReadMemory(my_buffer.GetBytes(), allocation_process_addr, size, err);
Updated the IRExecutionUnit to keep local copies of the data it writes down into the process even if the process doesn't exist. This will allow the IR interpreter to access static data allocated on the expression's behalf. Also cleaned up object ownership in the IRExecutionUnit so that allocations are created into the allocations vector. This avoids needless data copies. <rdar://problem/13424594> llvm-svn: 177456
2013-03-20 07:03:21 +08:00
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
if (err.Success())
{
DataExtractor my_extractor(my_buffer.GetBytes(), my_buffer.GetByteSize(), lldb::eByteOrderBig, 8);
StreamString ss;
my_extractor.Dump(&ss, 0, lldb::eFormatBytesWithASCII, 1, my_buffer.GetByteSize(), 32, allocation_process_addr, 0, 0);
log->PutCString(ss.GetData());
}
}
return allocation_process_addr;
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
}
void
IRExecutionUnit::FreeNow (lldb::addr_t allocation)
{
if (allocation == LLDB_INVALID_ADDRESS)
return;
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
Error err;
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
Free(allocation, err);
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
}
Error
IRExecutionUnit::DisassembleFunction (Stream &stream,
lldb::ProcessSP &process_wp)
{
<rdar://problem/13521159> LLDB is crashing when logging is enabled from lldb-perf-clang. This has to do with the global destructor chain as the process and its threads are being torn down. All logging channels now make one and only one instance that is kept in a global pointer which is never freed. This guarantees that logging can correctly continue as the process tears itself down. llvm-svn: 178191
2013-03-28 07:08:40 +08:00
Log *log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_EXPRESSIONS));
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
ExecutionContext exe_ctx(process_wp);
Error ret;
ret.Clear();
lldb::addr_t func_local_addr = LLDB_INVALID_ADDRESS;
lldb::addr_t func_remote_addr = LLDB_INVALID_ADDRESS;
for (JittedFunction &function : m_jitted_functions)
{
if (strstr(function.m_name.c_str(), m_name.AsCString()))
{
func_local_addr = function.m_local_addr;
func_remote_addr = function.m_remote_addr;
}
}
if (func_local_addr == LLDB_INVALID_ADDRESS)
{
ret.SetErrorToGenericError();
ret.SetErrorStringWithFormat("Couldn't find function %s for disassembly", m_name.AsCString());
return ret;
}
if (log)
log->Printf("Found function, has local address 0x%" PRIx64 " and remote address 0x%" PRIx64, (uint64_t)func_local_addr, (uint64_t)func_remote_addr);
std::pair <lldb::addr_t, lldb::addr_t> func_range;
func_range = GetRemoteRangeForLocal(func_local_addr);
if (func_range.first == 0 && func_range.second == 0)
{
ret.SetErrorToGenericError();
ret.SetErrorStringWithFormat("Couldn't find code range for function %s", m_name.AsCString());
return ret;
}
if (log)
log->Printf("Function's code range is [0x%" PRIx64 "+0x%" PRIx64 "]", func_range.first, func_range.second);
Target *target = exe_ctx.GetTargetPtr();
if (!target)
{
ret.SetErrorToGenericError();
ret.SetErrorString("Couldn't find the target");
return ret;
}
lldb::DataBufferSP buffer_sp(new DataBufferHeap(func_range.second, 0));
Process *process = exe_ctx.GetProcessPtr();
Error err;
process->ReadMemory(func_remote_addr, buffer_sp->GetBytes(), buffer_sp->GetByteSize(), err);
if (!err.Success())
{
ret.SetErrorToGenericError();
ret.SetErrorStringWithFormat("Couldn't read from process: %s", err.AsCString("unknown error"));
return ret;
}
ArchSpec arch(target->GetArchitecture());
const char *plugin_name = NULL;
const char *flavor_string = NULL;
The DisassemblerLLVMC has a retain cycle - the InstructionLLVMC's contained in its instruction list have a shared pointer back to their DisassemblerLLVMC. This checkin force clears the InstructionList in all the places we use the DisassemblerSP to stop the leaking for now. I'll go back and fix this for real when I have time to do so. <rdar://problem/14581918> llvm-svn: 187473
2013-07-31 10:19:15 +08:00
lldb::DisassemblerSP disassembler_sp = Disassembler::FindPlugin(arch, flavor_string, plugin_name);
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
The DisassemblerLLVMC has a retain cycle - the InstructionLLVMC's contained in its instruction list have a shared pointer back to their DisassemblerLLVMC. This checkin force clears the InstructionList in all the places we use the DisassemblerSP to stop the leaking for now. I'll go back and fix this for real when I have time to do so. <rdar://problem/14581918> llvm-svn: 187473
2013-07-31 10:19:15 +08:00
if (!disassembler_sp)
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
{
ret.SetErrorToGenericError();
ret.SetErrorStringWithFormat("Unable to find disassembler plug-in for %s architecture.", arch.GetArchitectureName());
return ret;
}
if (!process)
{
ret.SetErrorToGenericError();
ret.SetErrorString("Couldn't find the process");
return ret;
}
DataExtractor extractor(buffer_sp,
process->GetByteOrder(),
target->GetArchitecture().GetAddressByteSize());
if (log)
{
log->Printf("Function data has contents:");
<rdar://problem/13521159> LLDB is crashing when logging is enabled from lldb-perf-clang. This has to do with the global destructor chain as the process and its threads are being torn down. All logging channels now make one and only one instance that is kept in a global pointer which is never freed. This guarantees that logging can correctly continue as the process tears itself down. llvm-svn: 178191
2013-03-28 07:08:40 +08:00
extractor.PutToLog (log,
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
0,
extractor.GetByteSize(),
func_remote_addr,
16,
DataExtractor::TypeUInt8);
}
The DisassemblerLLVMC has a retain cycle - the InstructionLLVMC's contained in its instruction list have a shared pointer back to their DisassemblerLLVMC. This checkin force clears the InstructionList in all the places we use the DisassemblerSP to stop the leaking for now. I'll go back and fix this for real when I have time to do so. <rdar://problem/14581918> llvm-svn: 187473
2013-07-31 10:19:15 +08:00
disassembler_sp->DecodeInstructions (Address (func_remote_addr), extractor, 0, UINT32_MAX, false, false);
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
The DisassemblerLLVMC has a retain cycle - the InstructionLLVMC's contained in its instruction list have a shared pointer back to their DisassemblerLLVMC. This checkin force clears the InstructionList in all the places we use the DisassemblerSP to stop the leaking for now. I'll go back and fix this for real when I have time to do so. <rdar://problem/14581918> llvm-svn: 187473
2013-07-31 10:19:15 +08:00
InstructionList &instruction_list = disassembler_sp->GetInstructionList();
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
const uint32_t max_opcode_byte_size = instruction_list.GetMaxOpcocdeByteSize();
for (size_t instruction_index = 0, num_instructions = instruction_list.GetSize();
instruction_index < num_instructions;
++instruction_index)
{
Instruction *instruction = instruction_list.GetInstructionAtIndex(instruction_index).get();
instruction->Dump (&stream,
max_opcode_byte_size,
true,
true,
&exe_ctx);
stream.PutChar('\n');
}
The DisassemblerLLVMC has a retain cycle - the InstructionLLVMC's contained in its instruction list have a shared pointer back to their DisassemblerLLVMC. This checkin force clears the InstructionList in all the places we use the DisassemblerSP to stop the leaking for now. I'll go back and fix this for real when I have time to do so. <rdar://problem/14581918> llvm-svn: 187473
2013-07-31 10:19:15 +08:00
// FIXME: The DisassemblerLLVMC has a reference cycle and won't go away if it has any active instructions.
// I'll fix that but for now, just clear the list and it will go away nicely.
disassembler_sp->GetInstructionList().Clear();
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
return ret;
}
Fixed a problem where inline assembly errors caused LLDB to crash. <rdar://problem/13497915> llvm-svn: 178115
2013-03-27 11:09:55 +08:00
static void ReportInlineAsmError(const llvm::SMDiagnostic &diagnostic, void *Context, unsigned LocCookie)
{
Error *err = static_cast<Error*>(Context);
if (err && err->Success())
{
err->SetErrorToGenericError();
err->SetErrorStringWithFormat("Inline assembly error: %s", diagnostic.getMessage().str().c_str());
}
}
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
void
IRExecutionUnit::GetRunnableInfo(Error &error,
lldb::addr_t &func_addr,
lldb::addr_t &func_end)
{
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
lldb::ProcessSP process_sp(GetProcessWP().lock());
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
func_addr = LLDB_INVALID_ADDRESS;
func_end = LLDB_INVALID_ADDRESS;
if (!process_sp)
{
error.SetErrorToGenericError();
error.SetErrorString("Couldn't write the JIT compiled code into the process because the process is invalid");
return;
}
if (m_did_jit)
{
func_addr = m_function_load_addr;
func_end = m_function_end_load_addr;
return;
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
};
m_did_jit = true;
Log *log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_EXPRESSIONS));
std::string error_string;
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
if (log)
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
{
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
std::string s;
llvm::raw_string_ostream oss(s);
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
m_module->print(oss, NULL);
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
oss.flush();
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
log->Printf ("Module being sent to JIT: \n%s", s.c_str());
}
llvm::Triple triple(m_module->getTargetTriple());
llvm::Function *function = m_module->getFunction (m_name.AsCString());
llvm::Reloc::Model relocModel;
llvm::CodeModel::Model codeModel;
if (triple.isOSBinFormatELF())
{
relocModel = llvm::Reloc::Static;
// This will be small for 32-bit and large for 64-bit.
codeModel = llvm::CodeModel::JITDefault;
}
else
{
relocModel = llvm::Reloc::PIC_;
codeModel = llvm::CodeModel::Small;
}
m_module_ap->getContext().setInlineAsmDiagnosticHandler(ReportInlineAsmError, &error);
llvm::EngineBuilder builder(m_module_ap.get());
builder.setEngineKind(llvm::EngineKind::JIT)
.setErrorStr(&error_string)
.setRelocationModel(relocModel)
.setJITMemoryManager(new MemoryManager(*this))
.setOptLevel(llvm::CodeGenOpt::Less)
.setAllocateGVsWithCode(true)
.setCodeModel(codeModel)
.setUseMCJIT(true);
llvm::StringRef mArch;
llvm::StringRef mCPU;
llvm::SmallVector<std::string, 0> mAttrs;
for (std::string &feature : m_cpu_features)
mAttrs.push_back(feature);
llvm::TargetMachine *target_machine = builder.selectTarget(triple,
mArch,
mCPU,
mAttrs);
m_execution_engine_ap.reset(builder.create(target_machine));
if (!m_execution_engine_ap.get())
{
error.SetErrorToGenericError();
error.SetErrorStringWithFormat("Couldn't JIT the function: %s", error_string.c_str());
return;
}
else
{
m_module_ap.release(); // ownership was transferred
}
m_execution_engine_ap->DisableLazyCompilation();
// We don't actually need the function pointer here, this just forces it to get resolved.
void *fun_ptr = m_execution_engine_ap->getPointerToFunction(function);
if (!error.Success())
{
// We got an error through our callback!
return;
}
if (!function)
{
error.SetErrorToGenericError();
error.SetErrorStringWithFormat("Couldn't find '%s' in the JITted module", m_name.AsCString());
return;
}
if (!fun_ptr)
{
error.SetErrorToGenericError();
error.SetErrorStringWithFormat("'%s' was in the JITted module but wasn't lowered", m_name.AsCString());
return;
}
m_jitted_functions.push_back (JittedFunction(m_name.AsCString(), (lldb::addr_t)fun_ptr));
CommitAllocations(process_sp);
ReportAllocations(*m_execution_engine_ap);
WriteData(process_sp);
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
for (JittedFunction &jitted_function : m_jitted_functions)
{
jitted_function.m_remote_addr = GetRemoteAddressForLocal (jitted_function.m_local_addr);
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
if (!jitted_function.m_name.compare(m_name.AsCString()))
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
{
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
AddrRange func_range = GetRemoteRangeForLocal(jitted_function.m_local_addr);
m_function_end_load_addr = func_range.first + func_range.second;
m_function_load_addr = jitted_function.m_remote_addr;
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
}
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
}
if (log)
{
log->Printf("Code can be run in the target.");
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
StreamString disassembly_stream;
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
Error err = DisassembleFunction(disassembly_stream, process_sp);
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
if (!err.Success())
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
{
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
log->Printf("Couldn't disassemble function : %s", err.AsCString("unknown error"));
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
}
else
{
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
log->Printf("Function disassembly:\n%s", disassembly_stream.GetData());
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
}
}
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
func_addr = m_function_load_addr;
func_end = m_function_end_load_addr;
return;
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
}
IRExecutionUnit::~IRExecutionUnit ()
{
Flipped the big switch: LLDB now uses the new Materializer for all expressions that need to run in the target. This includes the following changes: - Removed a bunch of (de-)materialization code from ClangExpressionDeclMap and assumed the presence of a Materializer where we previously had a fallback. - Ensured that an IRMemoryMap is passed into ClangExpressionDeclMap::Materialize(). - Fixed object ownership on LLVMContext; it is now owned by the IRExecutionUnit, since the Module and the ExecutionEngine both depend on its existence. - Fixed a few bugs in IRMemoryMap and the Materializer that showed up during testing. llvm-svn: 179649
2013-04-17 07:25:35 +08:00
m_module_ap.reset();
m_execution_engine_ap.reset();
m_context_ap.reset();
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
}
IRExecutionUnit::MemoryManager::MemoryManager (IRExecutionUnit &parent) :
m_default_mm_ap (llvm::JITMemoryManager::CreateDefaultMemManager()),
m_parent (parent)
{
}
void
IRExecutionUnit::MemoryManager::setMemoryWritable ()
{
m_default_mm_ap->setMemoryWritable();
}
void
IRExecutionUnit::MemoryManager::setMemoryExecutable ()
{
m_default_mm_ap->setMemoryExecutable();
}
uint8_t *
IRExecutionUnit::MemoryManager::startFunctionBody(const llvm::Function *F,
uintptr_t &ActualSize)
{
return m_default_mm_ap->startFunctionBody(F, ActualSize);
}
uint8_t *
IRExecutionUnit::MemoryManager::allocateStub(const llvm::GlobalValue* F,
unsigned StubSize,
unsigned Alignment)
{
<rdar://problem/13521159> LLDB is crashing when logging is enabled from lldb-perf-clang. This has to do with the global destructor chain as the process and its threads are being torn down. All logging channels now make one and only one instance that is kept in a global pointer which is never freed. This guarantees that logging can correctly continue as the process tears itself down. llvm-svn: 178191
2013-03-28 07:08:40 +08:00
Log *log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_EXPRESSIONS));
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
uint8_t *return_value = m_default_mm_ap->allocateStub(F, StubSize, Alignment);
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
m_parent.m_records.push_back(AllocationRecord((uintptr_t)return_value,
lldb::ePermissionsReadable | lldb::ePermissionsWritable,
StubSize,
Alignment));
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
if (log)
{
log->Printf("IRExecutionUnit::allocateStub (F=%p, StubSize=%u, Alignment=%u) = %p",
F, StubSize, Alignment, return_value);
}
Updated the IRExecutionUnit to keep local copies of the data it writes down into the process even if the process doesn't exist. This will allow the IR interpreter to access static data allocated on the expression's behalf. Also cleaned up object ownership in the IRExecutionUnit so that allocations are created into the allocations vector. This avoids needless data copies. <rdar://problem/13424594> llvm-svn: 177456
2013-03-20 07:03:21 +08:00
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
return return_value;
}
void
IRExecutionUnit::MemoryManager::endFunctionBody(const llvm::Function *F,
uint8_t *FunctionStart,
uint8_t *FunctionEnd)
{
m_default_mm_ap->endFunctionBody(F, FunctionStart, FunctionEnd);
}
uint8_t *
IRExecutionUnit::MemoryManager::allocateSpace(intptr_t Size, unsigned Alignment)
{
<rdar://problem/13521159> LLDB is crashing when logging is enabled from lldb-perf-clang. This has to do with the global destructor chain as the process and its threads are being torn down. All logging channels now make one and only one instance that is kept in a global pointer which is never freed. This guarantees that logging can correctly continue as the process tears itself down. llvm-svn: 178191
2013-03-28 07:08:40 +08:00
Log *log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_EXPRESSIONS));
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
uint8_t *return_value = m_default_mm_ap->allocateSpace(Size, Alignment);
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
m_parent.m_records.push_back(AllocationRecord((uintptr_t)return_value,
lldb::ePermissionsReadable | lldb::ePermissionsWritable,
Size,
Alignment));
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
if (log)
{
log->Printf("IRExecutionUnit::allocateSpace(Size=%" PRIu64 ", Alignment=%u) = %p",
(uint64_t)Size, Alignment, return_value);
}
Updated the IRExecutionUnit to keep local copies of the data it writes down into the process even if the process doesn't exist. This will allow the IR interpreter to access static data allocated on the expression's behalf. Also cleaned up object ownership in the IRExecutionUnit so that allocations are created into the allocations vector. This avoids needless data copies. <rdar://problem/13424594> llvm-svn: 177456
2013-03-20 07:03:21 +08:00
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
return return_value;
}
uint8_t *
IRExecutionUnit::MemoryManager::allocateCodeSection(uintptr_t Size,
unsigned Alignment,
unsigned SectionID)
{
<rdar://problem/13521159> LLDB is crashing when logging is enabled from lldb-perf-clang. This has to do with the global destructor chain as the process and its threads are being torn down. All logging channels now make one and only one instance that is kept in a global pointer which is never freed. This guarantees that logging can correctly continue as the process tears itself down. llvm-svn: 178191
2013-03-28 07:08:40 +08:00
Log *log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_EXPRESSIONS));
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
uint8_t *return_value = m_default_mm_ap->allocateCodeSection(Size, Alignment, SectionID);
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
m_parent.m_records.push_back(AllocationRecord((uintptr_t)return_value,
Fixed the way we allocate executable memory on behalf of the JIT. We don't need it to be writable since we are using special APIs to write into it. <rdar://problem/13599185> llvm-svn: 179077
2013-04-09 09:13:08 +08:00
lldb::ePermissionsReadable | lldb::ePermissionsExecutable,
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
Size,
Alignment,
SectionID));
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
if (log)
{
log->Printf("IRExecutionUnit::allocateCodeSection(Size=0x%" PRIx64 ", Alignment=%u, SectionID=%u) = %p",
(uint64_t)Size, Alignment, SectionID, return_value);
}
Updated the IRExecutionUnit to keep local copies of the data it writes down into the process even if the process doesn't exist. This will allow the IR interpreter to access static data allocated on the expression's behalf. Also cleaned up object ownership in the IRExecutionUnit so that allocations are created into the allocations vector. This avoids needless data copies. <rdar://problem/13424594> llvm-svn: 177456
2013-03-20 07:03:21 +08:00
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
return return_value;
}
uint8_t *
IRExecutionUnit::MemoryManager::allocateDataSection(uintptr_t Size,
unsigned Alignment,
unsigned SectionID,
bool IsReadOnly)
{
<rdar://problem/13521159> LLDB is crashing when logging is enabled from lldb-perf-clang. This has to do with the global destructor chain as the process and its threads are being torn down. All logging channels now make one and only one instance that is kept in a global pointer which is never freed. This guarantees that logging can correctly continue as the process tears itself down. llvm-svn: 178191
2013-03-28 07:08:40 +08:00
Log *log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_EXPRESSIONS));
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
uint8_t *return_value = m_default_mm_ap->allocateDataSection(Size, Alignment, SectionID, IsReadOnly);
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
m_parent.m_records.push_back(AllocationRecord((uintptr_t)return_value,
lldb::ePermissionsReadable | lldb::ePermissionsWritable,
Size,
Alignment,
SectionID));
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
if (log)
{
log->Printf("IRExecutionUnit::allocateDataSection(Size=0x%" PRIx64 ", Alignment=%u, SectionID=%u) = %p",
(uint64_t)Size, Alignment, SectionID, return_value);
}
Updated the IRExecutionUnit to keep local copies of the data it writes down into the process even if the process doesn't exist. This will allow the IR interpreter to access static data allocated on the expression's behalf. Also cleaned up object ownership in the IRExecutionUnit so that allocations are created into the allocations vector. This avoids needless data copies. <rdar://problem/13424594> llvm-svn: 177456
2013-03-20 07:03:21 +08:00
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
return return_value;
}
uint8_t *
IRExecutionUnit::MemoryManager::allocateGlobal(uintptr_t Size,
unsigned Alignment)
{
<rdar://problem/13521159> LLDB is crashing when logging is enabled from lldb-perf-clang. This has to do with the global destructor chain as the process and its threads are being torn down. All logging channels now make one and only one instance that is kept in a global pointer which is never freed. This guarantees that logging can correctly continue as the process tears itself down. llvm-svn: 178191
2013-03-28 07:08:40 +08:00
Log *log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_EXPRESSIONS));
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
uint8_t *return_value = m_default_mm_ap->allocateGlobal(Size, Alignment);
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
m_parent.m_records.push_back(AllocationRecord((uintptr_t)return_value,
lldb::ePermissionsReadable | lldb::ePermissionsWritable,
Size,
Alignment));
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
if (log)
{
log->Printf("IRExecutionUnit::allocateGlobal(Size=0x%" PRIx64 ", Alignment=%u) = %p",
(uint64_t)Size, Alignment, return_value);
}
return return_value;
}
void
IRExecutionUnit::MemoryManager::deallocateFunctionBody(void *Body)
{
m_default_mm_ap->deallocateFunctionBody(Body);
}
lldb::addr_t
IRExecutionUnit::GetRemoteAddressForLocal (lldb::addr_t local_address)
{
Fixed IRExecutionUnit so that it looks up addresses correctly. We have been getting lucky since most expressions generate only one section (or the first code section contains all the code), but sometimes it actually matters. <rdar://problem/14180236> llvm-svn: 185054
2013-06-27 09:42:47 +08:00
Log *log(lldb_private::GetLogIfAllCategoriesSet (LIBLLDB_LOG_EXPRESSIONS));
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
for (AllocationRecord &record : m_records)
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
{
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
if (local_address >= record.m_host_address &&
local_address < record.m_host_address + record.m_size)
{
if (record.m_process_address == LLDB_INVALID_ADDRESS)
return LLDB_INVALID_ADDRESS;
Fixed IRExecutionUnit so that it looks up addresses correctly. We have been getting lucky since most expressions generate only one section (or the first code section contains all the code), but sometimes it actually matters. <rdar://problem/14180236> llvm-svn: 185054
2013-06-27 09:42:47 +08:00
lldb::addr_t ret = record.m_process_address + (local_address - record.m_host_address);
if (log)
{
log->Printf("IRExecutionUnit::GetRemoteAddressForLocal() found 0x%" PRIx64 " in [0x%" PRIx64 "..0x%" PRIx64 "], and returned 0x%" PRIx64 " from [0x%" PRIx64 "..0x%" PRIx64 "].",
local_address,
clean up about 22 warnings messages llvm-svn: 187900
2013-08-08 03:05:15 +08:00
(uint64_t)record.m_host_address,
(uint64_t)record.m_host_address + (uint64_t)record.m_size,
Fixed IRExecutionUnit so that it looks up addresses correctly. We have been getting lucky since most expressions generate only one section (or the first code section contains all the code), but sometimes it actually matters. <rdar://problem/14180236> llvm-svn: 185054
2013-06-27 09:42:47 +08:00
ret,
record.m_process_address,
record.m_process_address + record.m_size);
}
return ret;
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
}
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
}
return LLDB_INVALID_ADDRESS;
}
IRExecutionUnit::AddrRange
IRExecutionUnit::GetRemoteRangeForLocal (lldb::addr_t local_address)
{
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
for (AllocationRecord &record : m_records)
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
{
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
if (local_address >= record.m_host_address &&
local_address < record.m_host_address + record.m_size)
{
if (record.m_process_address == LLDB_INVALID_ADDRESS)
return AddrRange(0, 0);
return AddrRange(record.m_process_address, record.m_size);
}
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
}
return AddrRange (0, 0);
}
bool
IRExecutionUnit::CommitAllocations (lldb::ProcessSP &process_sp)
{
bool ret = true;
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
lldb_private::Error err;
for (AllocationRecord &record : m_records)
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
{
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
if (record.m_process_address != LLDB_INVALID_ADDRESS)
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
continue;
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
record.m_process_address = Malloc(record.m_size,
record.m_alignment,
record.m_permissions,
eAllocationPolicyProcessOnly,
err);
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
if (!err.Success())
{
ret = false;
break;
}
}
if (!ret)
{
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
for (AllocationRecord &record : m_records)
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
{
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
if (record.m_process_address != LLDB_INVALID_ADDRESS)
{
Free(record.m_process_address, err);
record.m_process_address = LLDB_INVALID_ADDRESS;
}
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
}
}
return ret;
}
void
IRExecutionUnit::ReportAllocations (llvm::ExecutionEngine &engine)
{
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
for (AllocationRecord &record : m_records)
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
{
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
if (record.m_process_address == LLDB_INVALID_ADDRESS)
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
continue;
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
if (record.m_section_id == eSectionIDInvalid)
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
continue;
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
engine.mapSectionAddress((void*)record.m_host_address, record.m_process_address);
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
}
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
// Trigger re-application of relocations.
engine.finalizeObject();
}
bool
IRExecutionUnit::WriteData (lldb::ProcessSP &process_sp)
{
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
for (AllocationRecord &record : m_records)
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
{
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
if (record.m_process_address == LLDB_INVALID_ADDRESS)
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
return false;
lldb_private::Error err;
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
WriteMemory (record.m_process_address, (uint8_t*)record.m_host_address, record.m_size, err);
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
}
return true;
}
void
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
IRExecutionUnit::AllocationRecord::dump (Log *log)
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
{
if (!log)
return;
log->Printf("[0x%llx+0x%llx]->0x%llx (alignment %d, section ID %d)",
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
(unsigned long long)m_host_address,
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
(unsigned long long)m_size,
Factored out memory access into the target process from IRExecutionUnit into a superclass called IRMemoryMap. IRMemoryMap handles all reading and writing, ensuring that areas are kept track of and memory is properly cached (and deleted). Also fixed several cases where we would simply leak binary data in the target process over time. Now the expression objects explicitly own their IRExecutionUnit and delete it when they go away. This is why I had to modify ClangUserExpression, ClangUtilityFunction, and ClangFunction. As a side effect of this, I am removing the JIT mutex for an IRMemoryMap. If it turns out that we need this mutex, I'll add it in then, but right now it's just adding complexity. This is part of a more general project to make expressions fully reusable. The next step is to make materialization and dematerialization use the IRMemoryMap API rather than writing and reading directly from the process's memory. This will allow the IR interpreter to use the same data, but in the host's memory, without having to use a different set of pointers. llvm-svn: 178832
2013-04-05 10:22:57 +08:00
(unsigned long long)m_process_address,
Refactored the expression parser so that the IR and the JITted code are managed by a standalone class that handles memory management itself. I have removed RecordingMemoryManager and ProcessDataAllocator, which filled similar roles and had confusing ownership, with a common class called IRExecutionUnit. The IRExecutionUnit manages all allocations ever made for an expression and frees them when it goes away. It also contains the code generator and can vend the Module for an expression to other clases. The end goal here is to make the output of the expression parser re-usable; that is, to avoid re-parsing when re-parsing isn't necessary. I've also cleaned up some code and used weak pointers in more places. Please let me know if you see any leaks; I checked myself as well but I might have missed a case. llvm-svn: 177364
2013-03-19 08:10:07 +08:00
(unsigned)m_alignment,
(unsigned)m_section_id);
}