2008-08-17 20:44:35 +02:00
|
|
|
//===-- ShadowStackGC.cpp - GC support for uncooperative targets ----------===//
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
//
|
|
|
|
// The LLVM Compiler Infrastructure
|
|
|
|
//
|
|
|
|
// This file is distributed under the University of Illinois Open Source
|
|
|
|
// License. See LICENSE.TXT for details.
|
|
|
|
//
|
|
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
//
|
|
|
|
// This file implements lowering for the llvm.gc* intrinsics for targets that do
|
|
|
|
// not natively support them (which includes the C backend). Note that the code
|
2008-08-17 20:44:35 +02:00
|
|
|
// generated is not quite as efficient as algorithms which generate stack maps
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// to identify roots.
|
|
|
|
//
|
|
|
|
// This pass implements the code transformation described in this paper:
|
|
|
|
// "Accurate Garbage Collection in an Uncooperative Environment"
|
|
|
|
// Fergus Henderson, ISMM, 2002
|
|
|
|
//
|
|
|
|
// In runtime/GC/SemiSpace.cpp is a prototype runtime which is compatible with
|
2008-08-17 20:44:35 +02:00
|
|
|
// ShadowStackGC.
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
//
|
|
|
|
// In order to support this particular transformation, all stack roots are
|
|
|
|
// coallocated in the stack. This allows a fully target-independent stack map
|
|
|
|
// while introducing only minor runtime overhead.
|
|
|
|
//
|
|
|
|
//===----------------------------------------------------------------------===//
|
|
|
|
|
|
|
|
#define DEBUG_TYPE "shadowstackgc"
|
2008-08-17 14:56:54 +02:00
|
|
|
#include "llvm/CodeGen/GCs.h"
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
#include "llvm/ADT/StringExtras.h"
|
2008-08-17 14:56:54 +02:00
|
|
|
#include "llvm/CodeGen/GCStrategy.h"
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
#include "llvm/IntrinsicInst.h"
|
|
|
|
#include "llvm/Module.h"
|
2010-06-25 10:48:19 +02:00
|
|
|
#include "llvm/Support/CallSite.h"
|
2008-04-13 08:22:09 +02:00
|
|
|
#include "llvm/Support/IRBuilder.h"
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
|
|
|
|
using namespace llvm;
|
|
|
|
|
|
|
|
namespace {
|
2009-01-16 07:53:46 +01:00
|
|
|
|
2009-10-25 07:33:48 +01:00
|
|
|
class ShadowStackGC : public GCStrategy {
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
/// RootChain - This is the global linked-list that contains the chain of GC
|
|
|
|
/// roots.
|
|
|
|
GlobalVariable *Head;
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
/// StackEntryTy - Abstract type of a link in the shadow stack.
|
2009-01-16 07:53:46 +01:00
|
|
|
///
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
const StructType *StackEntryTy;
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
/// Roots - GC roots in the current function. Each is a pair of the
|
|
|
|
/// intrinsic call and its corresponding alloca.
|
|
|
|
std::vector<std::pair<CallInst*,AllocaInst*> > Roots;
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
public:
|
2008-08-17 20:44:35 +02:00
|
|
|
ShadowStackGC();
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
bool initializeCustomLowering(Module &M);
|
|
|
|
bool performCustomLowering(Function &F);
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
private:
|
|
|
|
bool IsNullValue(Value *V);
|
|
|
|
Constant *GetFrameMap(Function &F);
|
|
|
|
const Type* GetConcreteStackEntryType(Function &F);
|
|
|
|
void CollectRoots(Function &F);
|
2009-07-22 02:24:57 +02:00
|
|
|
static GetElementPtrInst *CreateGEP(LLVMContext &Context,
|
2009-07-15 01:09:55 +02:00
|
|
|
IRBuilder<> &B, Value *BasePtr,
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
int Idx1, const char *Name);
|
2009-07-22 02:24:57 +02:00
|
|
|
static GetElementPtrInst *CreateGEP(LLVMContext &Context,
|
2009-07-15 01:09:55 +02:00
|
|
|
IRBuilder<> &B, Value *BasePtr,
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
int Idx1, int Idx2, const char *Name);
|
|
|
|
};
|
2008-05-13 02:00:25 +02:00
|
|
|
|
|
|
|
}
|
2009-01-16 07:53:46 +01:00
|
|
|
|
2008-08-17 20:44:35 +02:00
|
|
|
static GCRegistry::Add<ShadowStackGC>
|
|
|
|
X("shadow-stack", "Very portable GC for uncooperative code generators");
|
2009-01-16 07:53:46 +01:00
|
|
|
|
2008-05-13 02:00:25 +02:00
|
|
|
namespace {
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
/// EscapeEnumerator - This is a little algorithm to find all escape points
|
|
|
|
/// from a function so that "finally"-style code can be inserted. In addition
|
|
|
|
/// to finding the existing return and unwind instructions, it also (if
|
|
|
|
/// necessary) transforms any call instructions into invokes and sends them to
|
|
|
|
/// a landing pad.
|
2009-01-16 07:53:46 +01:00
|
|
|
///
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
/// It's wrapped up in a state machine using the same transform C# uses for
|
|
|
|
/// 'yield return' enumerators, This transform allows it to be non-allocating.
|
2009-10-25 07:33:48 +01:00
|
|
|
class EscapeEnumerator {
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
Function &F;
|
|
|
|
const char *CleanupBBName;
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// State.
|
|
|
|
int State;
|
|
|
|
Function::iterator StateBB, StateE;
|
2008-08-08 21:39:37 +02:00
|
|
|
IRBuilder<> Builder;
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
public:
|
|
|
|
EscapeEnumerator(Function &F, const char *N = "cleanup")
|
2009-07-22 02:24:57 +02:00
|
|
|
: F(F), CleanupBBName(N), State(0), Builder(F.getContext()) {}
|
2009-01-16 07:53:46 +01:00
|
|
|
|
2008-08-08 21:39:37 +02:00
|
|
|
IRBuilder<> *Next() {
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
switch (State) {
|
|
|
|
default:
|
|
|
|
return 0;
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
case 0:
|
|
|
|
StateBB = F.begin();
|
|
|
|
StateE = F.end();
|
|
|
|
State = 1;
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
case 1:
|
|
|
|
// Find all 'return' and 'unwind' instructions.
|
|
|
|
while (StateBB != StateE) {
|
|
|
|
BasicBlock *CurBB = StateBB++;
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// Branches and invokes do not escape, only unwind and return do.
|
|
|
|
TerminatorInst *TI = CurBB->getTerminator();
|
|
|
|
if (!isa<UnwindInst>(TI) && !isa<ReturnInst>(TI))
|
|
|
|
continue;
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
Builder.SetInsertPoint(TI->getParent(), TI);
|
|
|
|
return &Builder;
|
|
|
|
}
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
State = 2;
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// Find all 'call' instructions.
|
|
|
|
SmallVector<Instruction*,16> Calls;
|
|
|
|
for (Function::iterator BB = F.begin(),
|
|
|
|
E = F.end(); BB != E; ++BB)
|
|
|
|
for (BasicBlock::iterator II = BB->begin(),
|
|
|
|
EE = BB->end(); II != EE; ++II)
|
|
|
|
if (CallInst *CI = dyn_cast<CallInst>(II))
|
|
|
|
if (!CI->getCalledFunction() ||
|
|
|
|
!CI->getCalledFunction()->getIntrinsicID())
|
|
|
|
Calls.push_back(CI);
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
if (Calls.empty())
|
|
|
|
return 0;
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// Create a cleanup block.
|
2009-08-13 23:58:54 +02:00
|
|
|
BasicBlock *CleanupBB = BasicBlock::Create(F.getContext(),
|
|
|
|
CleanupBBName, &F);
|
|
|
|
UnwindInst *UI = new UnwindInst(F.getContext(), CleanupBB);
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// Transform the 'call' instructions into 'invoke's branching to the
|
|
|
|
// cleanup block. Go in reverse order to make prettier BB names.
|
|
|
|
SmallVector<Value*,16> Args;
|
|
|
|
for (unsigned I = Calls.size(); I != 0; ) {
|
|
|
|
CallInst *CI = cast<CallInst>(Calls[--I]);
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// Split the basic block containing the function call.
|
|
|
|
BasicBlock *CallBB = CI->getParent();
|
|
|
|
BasicBlock *NewBB =
|
|
|
|
CallBB->splitBasicBlock(CI, CallBB->getName() + ".cont");
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// Remove the unconditional branch inserted at the end of CallBB.
|
|
|
|
CallBB->getInstList().pop_back();
|
|
|
|
NewBB->getInstList().remove(CI);
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// Create a new invoke instruction.
|
|
|
|
Args.clear();
|
2010-06-25 10:48:19 +02:00
|
|
|
CallSite CS(CI);
|
|
|
|
Args.append(CS.arg_begin(), CS.arg_end());
|
2009-01-16 07:53:46 +01:00
|
|
|
|
2010-04-20 15:13:04 +02:00
|
|
|
InvokeInst *II = InvokeInst::Create(CI->getCalledValue(),
|
2008-04-06 22:25:17 +02:00
|
|
|
NewBB, CleanupBB,
|
|
|
|
Args.begin(), Args.end(),
|
|
|
|
CI->getName(), CallBB);
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
II->setCallingConv(CI->getCallingConv());
|
2008-09-25 23:00:45 +02:00
|
|
|
II->setAttributes(CI->getAttributes());
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
CI->replaceAllUsesWith(II);
|
|
|
|
delete CI;
|
|
|
|
}
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
Builder.SetInsertPoint(UI->getParent(), UI);
|
|
|
|
return &Builder;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
};
|
|
|
|
}
|
|
|
|
|
|
|
|
// -----------------------------------------------------------------------------
|
|
|
|
|
2008-08-17 20:44:35 +02:00
|
|
|
void llvm::linkShadowStackGC() { }
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
|
2008-08-17 20:44:35 +02:00
|
|
|
ShadowStackGC::ShadowStackGC() : Head(0), StackEntryTy(0) {
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
InitRoots = true;
|
|
|
|
CustomRoots = true;
|
|
|
|
}
|
|
|
|
|
2008-08-17 20:44:35 +02:00
|
|
|
Constant *ShadowStackGC::GetFrameMap(Function &F) {
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// doInitialization creates the abstract type of this value.
|
2009-10-06 17:40:36 +02:00
|
|
|
const Type *VoidPtr = Type::getInt8PtrTy(F.getContext());
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// Truncate the ShadowStackDescriptor if some metadata is null.
|
|
|
|
unsigned NumMeta = 0;
|
|
|
|
SmallVector<Constant*,16> Metadata;
|
|
|
|
for (unsigned I = 0; I != Roots.size(); ++I) {
|
2010-06-25 10:48:19 +02:00
|
|
|
Constant *C = cast<Constant>(Roots[I].first->getArgOperand(1));
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
if (!C->isNullValue())
|
|
|
|
NumMeta = I + 1;
|
|
|
|
Metadata.push_back(ConstantExpr::getBitCast(C, VoidPtr));
|
|
|
|
}
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
Constant *BaseElts[] = {
|
2009-08-13 23:58:54 +02:00
|
|
|
ConstantInt::get(Type::getInt32Ty(F.getContext()), Roots.size(), false),
|
|
|
|
ConstantInt::get(Type::getInt32Ty(F.getContext()), NumMeta, false),
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
};
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
Constant *DescriptorElts[] = {
|
2009-09-19 22:30:26 +02:00
|
|
|
ConstantStruct::get(F.getContext(), BaseElts, 2, false),
|
2009-07-30 00:17:13 +02:00
|
|
|
ConstantArray::get(ArrayType::get(VoidPtr, NumMeta),
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
Metadata.begin(), NumMeta)
|
|
|
|
};
|
2009-01-16 07:53:46 +01:00
|
|
|
|
2009-09-19 22:30:26 +02:00
|
|
|
Constant *FrameMap = ConstantStruct::get(F.getContext(), DescriptorElts, 2,
|
|
|
|
false);
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
std::string TypeName("gc_map.");
|
|
|
|
TypeName += utostr(NumMeta);
|
|
|
|
F.getParent()->addTypeName(TypeName, FrameMap->getType());
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// FIXME: Is this actually dangerous as WritingAnLLVMPass.html claims? Seems
|
|
|
|
// that, short of multithreaded LLVM, it should be safe; all that is
|
|
|
|
// necessary is that a simple Module::iterator loop not be invalidated.
|
|
|
|
// Appending to the GlobalVariable list is safe in that sense.
|
2009-01-16 07:53:46 +01:00
|
|
|
//
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// All of the output passes emit globals last. The ExecutionEngine
|
|
|
|
// explicitly supports adding globals to the module after
|
|
|
|
// initialization.
|
2009-01-16 07:53:46 +01:00
|
|
|
//
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// Still, if it isn't deemed acceptable, then this transformation needs
|
|
|
|
// to be a ModulePass (which means it cannot be in the 'llc' pipeline
|
|
|
|
// (which uses a FunctionPassManager (which segfaults (not asserts) if
|
|
|
|
// provided a ModulePass))).
|
2009-07-08 21:03:57 +02:00
|
|
|
Constant *GV = new GlobalVariable(*F.getParent(), FrameMap->getType(), true,
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
GlobalVariable::InternalLinkage,
|
2009-07-08 21:03:57 +02:00
|
|
|
FrameMap, "__gc_" + F.getName());
|
2009-01-16 07:53:46 +01:00
|
|
|
|
2009-08-13 23:58:54 +02:00
|
|
|
Constant *GEPIndices[2] = {
|
|
|
|
ConstantInt::get(Type::getInt32Ty(F.getContext()), 0),
|
|
|
|
ConstantInt::get(Type::getInt32Ty(F.getContext()), 0)
|
|
|
|
};
|
2009-07-29 20:55:55 +02:00
|
|
|
return ConstantExpr::getGetElementPtr(GV, GEPIndices, 2);
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
}
|
|
|
|
|
2008-08-17 20:44:35 +02:00
|
|
|
const Type* ShadowStackGC::GetConcreteStackEntryType(Function &F) {
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// doInitialization creates the generic version of this type.
|
|
|
|
std::vector<const Type*> EltTys;
|
|
|
|
EltTys.push_back(StackEntryTy);
|
|
|
|
for (size_t I = 0; I != Roots.size(); I++)
|
|
|
|
EltTys.push_back(Roots[I].second->getAllocatedType());
|
2009-08-06 01:16:16 +02:00
|
|
|
Type *Ty = StructType::get(F.getContext(), EltTys);
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
std::string TypeName("gc_stackentry.");
|
|
|
|
TypeName += F.getName();
|
|
|
|
F.getParent()->addTypeName(TypeName, Ty);
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
return Ty;
|
|
|
|
}
|
|
|
|
|
|
|
|
/// doInitialization - If this module uses the GC intrinsics, find them now. If
|
|
|
|
/// not, exit fast.
|
2008-08-17 20:44:35 +02:00
|
|
|
bool ShadowStackGC::initializeCustomLowering(Module &M) {
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// struct FrameMap {
|
|
|
|
// int32_t NumRoots; // Number of roots in stack frame.
|
|
|
|
// int32_t NumMeta; // Number of metadata descriptors. May be < NumRoots.
|
|
|
|
// void *Meta[]; // May be absent for roots without metadata.
|
|
|
|
// };
|
|
|
|
std::vector<const Type*> EltTys;
|
2009-08-13 23:58:54 +02:00
|
|
|
// 32 bits is ok up to a 32GB stack frame. :)
|
|
|
|
EltTys.push_back(Type::getInt32Ty(M.getContext()));
|
|
|
|
// Specifies length of variable length array.
|
|
|
|
EltTys.push_back(Type::getInt32Ty(M.getContext()));
|
2009-08-06 01:16:16 +02:00
|
|
|
StructType *FrameMapTy = StructType::get(M.getContext(), EltTys);
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
M.addTypeName("gc_map", FrameMapTy);
|
|
|
|
PointerType *FrameMapPtrTy = PointerType::getUnqual(FrameMapTy);
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// struct StackEntry {
|
|
|
|
// ShadowStackEntry *Next; // Caller's stack entry.
|
|
|
|
// FrameMap *Map; // Pointer to constant FrameMap.
|
|
|
|
// void *Roots[]; // Stack roots (in-place array, so we pretend).
|
|
|
|
// };
|
2009-08-14 01:27:32 +02:00
|
|
|
OpaqueType *RecursiveTy = OpaqueType::get(M.getContext());
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
EltTys.clear();
|
|
|
|
EltTys.push_back(PointerType::getUnqual(RecursiveTy));
|
|
|
|
EltTys.push_back(FrameMapPtrTy);
|
2009-08-06 01:16:16 +02:00
|
|
|
PATypeHolder LinkTyH = StructType::get(M.getContext(), EltTys);
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
RecursiveTy->refineAbstractTypeTo(LinkTyH.get());
|
|
|
|
StackEntryTy = cast<StructType>(LinkTyH.get());
|
|
|
|
const PointerType *StackEntryPtrTy = PointerType::getUnqual(StackEntryTy);
|
|
|
|
M.addTypeName("gc_stackentry", LinkTyH.get()); // FIXME: Is this safe from
|
|
|
|
// a FunctionPass?
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// Get the root chain if it already exists.
|
|
|
|
Head = M.getGlobalVariable("llvm_gc_root_chain");
|
|
|
|
if (!Head) {
|
|
|
|
// If the root chain does not exist, insert a new one with linkonce
|
|
|
|
// linkage!
|
2009-07-08 21:03:57 +02:00
|
|
|
Head = new GlobalVariable(M, StackEntryPtrTy, false,
|
Introduce new linkage types linkonce_odr, weak_odr, common_odr
and extern_weak_odr. These are the same as the non-odr versions,
except that they indicate that the global will only be overridden
by an *equivalent* global. In C, a function with weak linkage can
be overridden by a function which behaves completely differently.
This means that IP passes have to skip weak functions, since any
deductions made from the function definition might be wrong, since
the definition could be replaced by something completely different
at link time. This is not allowed in C++, thanks to the ODR
(One-Definition-Rule): if a function is replaced by another at
link-time, then the new function must be the same as the original
function. If a language knows that a function or other global can
only be overridden by an equivalent global, it can give it the
weak_odr linkage type, and the optimizers will understand that it
is alright to make deductions based on the function body. The
code generators on the other hand map weak and weak_odr linkage
to the same thing.
llvm-svn: 66339
2009-03-07 16:45:40 +01:00
|
|
|
GlobalValue::LinkOnceAnyLinkage,
|
2009-07-31 22:28:14 +02:00
|
|
|
Constant::getNullValue(StackEntryPtrTy),
|
2009-07-08 21:03:57 +02:00
|
|
|
"llvm_gc_root_chain");
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
} else if (Head->hasExternalLinkage() && Head->isDeclaration()) {
|
2009-07-31 22:28:14 +02:00
|
|
|
Head->setInitializer(Constant::getNullValue(StackEntryPtrTy));
|
Introduce new linkage types linkonce_odr, weak_odr, common_odr
and extern_weak_odr. These are the same as the non-odr versions,
except that they indicate that the global will only be overridden
by an *equivalent* global. In C, a function with weak linkage can
be overridden by a function which behaves completely differently.
This means that IP passes have to skip weak functions, since any
deductions made from the function definition might be wrong, since
the definition could be replaced by something completely different
at link time. This is not allowed in C++, thanks to the ODR
(One-Definition-Rule): if a function is replaced by another at
link-time, then the new function must be the same as the original
function. If a language knows that a function or other global can
only be overridden by an equivalent global, it can give it the
weak_odr linkage type, and the optimizers will understand that it
is alright to make deductions based on the function body. The
code generators on the other hand map weak and weak_odr linkage
to the same thing.
llvm-svn: 66339
2009-03-07 16:45:40 +01:00
|
|
|
Head->setLinkage(GlobalValue::LinkOnceAnyLinkage);
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
}
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
return true;
|
|
|
|
}
|
|
|
|
|
2008-08-17 20:44:35 +02:00
|
|
|
bool ShadowStackGC::IsNullValue(Value *V) {
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
if (Constant *C = dyn_cast<Constant>(V))
|
|
|
|
return C->isNullValue();
|
|
|
|
return false;
|
|
|
|
}
|
|
|
|
|
2008-08-17 20:44:35 +02:00
|
|
|
void ShadowStackGC::CollectRoots(Function &F) {
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// FIXME: Account for original alignment. Could fragment the root array.
|
|
|
|
// Approach 1: Null initialize empty slots at runtime. Yuck.
|
|
|
|
// Approach 2: Emit a map of the array instead of just a count.
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
assert(Roots.empty() && "Not cleaned up?");
|
2009-01-16 07:53:46 +01:00
|
|
|
|
2010-06-25 10:48:19 +02:00
|
|
|
SmallVector<std::pair<CallInst*, AllocaInst*>, 16> MetaRoots;
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
for (Function::iterator BB = F.begin(), E = F.end(); BB != E; ++BB)
|
|
|
|
for (BasicBlock::iterator II = BB->begin(), E = BB->end(); II != E;)
|
|
|
|
if (IntrinsicInst *CI = dyn_cast<IntrinsicInst>(II++))
|
|
|
|
if (Function *F = CI->getCalledFunction())
|
|
|
|
if (F->getIntrinsicID() == Intrinsic::gcroot) {
|
2010-06-25 10:48:19 +02:00
|
|
|
std::pair<CallInst*, AllocaInst*> Pair = std::make_pair(
|
|
|
|
CI, cast<AllocaInst>(CI->getArgOperand(0)->stripPointerCasts()));
|
|
|
|
if (IsNullValue(CI->getArgOperand(1)))
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
Roots.push_back(Pair);
|
|
|
|
else
|
|
|
|
MetaRoots.push_back(Pair);
|
|
|
|
}
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// Number roots with metadata (usually empty) at the beginning, so that the
|
|
|
|
// FrameMap::Meta array can be elided.
|
|
|
|
Roots.insert(Roots.begin(), MetaRoots.begin(), MetaRoots.end());
|
|
|
|
}
|
|
|
|
|
|
|
|
GetElementPtrInst *
|
2009-07-22 02:24:57 +02:00
|
|
|
ShadowStackGC::CreateGEP(LLVMContext &Context, IRBuilder<> &B, Value *BasePtr,
|
2008-08-17 20:44:35 +02:00
|
|
|
int Idx, int Idx2, const char *Name) {
|
2009-08-13 23:58:54 +02:00
|
|
|
Value *Indices[] = { ConstantInt::get(Type::getInt32Ty(Context), 0),
|
|
|
|
ConstantInt::get(Type::getInt32Ty(Context), Idx),
|
|
|
|
ConstantInt::get(Type::getInt32Ty(Context), Idx2) };
|
2008-04-13 08:22:09 +02:00
|
|
|
Value* Val = B.CreateGEP(BasePtr, Indices, Indices + 3, Name);
|
2009-01-16 07:53:46 +01:00
|
|
|
|
2008-04-13 08:22:09 +02:00
|
|
|
assert(isa<GetElementPtrInst>(Val) && "Unexpected folded constant");
|
2009-01-16 07:53:46 +01:00
|
|
|
|
2008-04-13 08:22:09 +02:00
|
|
|
return dyn_cast<GetElementPtrInst>(Val);
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
GetElementPtrInst *
|
2009-07-22 02:24:57 +02:00
|
|
|
ShadowStackGC::CreateGEP(LLVMContext &Context, IRBuilder<> &B, Value *BasePtr,
|
2008-08-17 20:44:35 +02:00
|
|
|
int Idx, const char *Name) {
|
2009-08-13 23:58:54 +02:00
|
|
|
Value *Indices[] = { ConstantInt::get(Type::getInt32Ty(Context), 0),
|
|
|
|
ConstantInt::get(Type::getInt32Ty(Context), Idx) };
|
2008-04-13 08:22:09 +02:00
|
|
|
Value *Val = B.CreateGEP(BasePtr, Indices, Indices + 2, Name);
|
2009-01-16 07:53:46 +01:00
|
|
|
|
2008-04-13 08:22:09 +02:00
|
|
|
assert(isa<GetElementPtrInst>(Val) && "Unexpected folded constant");
|
|
|
|
|
|
|
|
return dyn_cast<GetElementPtrInst>(Val);
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
}
|
|
|
|
|
|
|
|
/// runOnFunction - Insert code to maintain the shadow stack.
|
2008-08-17 20:44:35 +02:00
|
|
|
bool ShadowStackGC::performCustomLowering(Function &F) {
|
2009-07-22 02:24:57 +02:00
|
|
|
LLVMContext &Context = F.getContext();
|
2009-07-15 01:09:55 +02:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// Find calls to llvm.gcroot.
|
|
|
|
CollectRoots(F);
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// If there are no roots in this function, then there is no need to add a
|
|
|
|
// stack map entry for it.
|
|
|
|
if (Roots.empty())
|
|
|
|
return false;
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// Build the constant map and figure the type of the shadow stack entry.
|
|
|
|
Value *FrameMap = GetFrameMap(F);
|
|
|
|
const Type *ConcreteStackEntryTy = GetConcreteStackEntryType(F);
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// Build the shadow stack entry at the very start of the function.
|
|
|
|
BasicBlock::iterator IP = F.getEntryBlock().begin();
|
2008-08-08 21:39:37 +02:00
|
|
|
IRBuilder<> AtEntry(IP->getParent(), IP);
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
Instruction *StackEntry = AtEntry.CreateAlloca(ConcreteStackEntryTy, 0,
|
|
|
|
"gc_frame");
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
while (isa<AllocaInst>(IP)) ++IP;
|
|
|
|
AtEntry.SetInsertPoint(IP->getParent(), IP);
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// Initialize the map pointer and load the current head of the shadow stack.
|
|
|
|
Instruction *CurrentHead = AtEntry.CreateLoad(Head, "gc_currhead");
|
2009-07-15 01:09:55 +02:00
|
|
|
Instruction *EntryMapPtr = CreateGEP(Context, AtEntry, StackEntry,
|
|
|
|
0,1,"gc_frame.map");
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
AtEntry.CreateStore(FrameMap, EntryMapPtr);
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// After all the allocas...
|
|
|
|
for (unsigned I = 0, E = Roots.size(); I != E; ++I) {
|
|
|
|
// For each root, find the corresponding slot in the aggregate...
|
2009-07-15 01:09:55 +02:00
|
|
|
Value *SlotPtr = CreateGEP(Context, AtEntry, StackEntry, 1 + I, "gc_root");
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// And use it in lieu of the alloca.
|
|
|
|
AllocaInst *OriginalAlloca = Roots[I].second;
|
|
|
|
SlotPtr->takeName(OriginalAlloca);
|
|
|
|
OriginalAlloca->replaceAllUsesWith(SlotPtr);
|
|
|
|
}
|
2009-01-16 07:53:46 +01:00
|
|
|
|
2008-08-17 20:44:35 +02:00
|
|
|
// Move past the original stores inserted by GCStrategy::InitRoots. This isn't
|
|
|
|
// really necessary (the collector would never see the intermediate state at
|
|
|
|
// runtime), but it's nicer not to push the half-initialized entry onto the
|
|
|
|
// shadow stack.
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
while (isa<StoreInst>(IP)) ++IP;
|
|
|
|
AtEntry.SetInsertPoint(IP->getParent(), IP);
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// Push the entry onto the shadow stack.
|
2009-07-15 01:09:55 +02:00
|
|
|
Instruction *EntryNextPtr = CreateGEP(Context, AtEntry,
|
|
|
|
StackEntry,0,0,"gc_frame.next");
|
|
|
|
Instruction *NewHeadVal = CreateGEP(Context, AtEntry,
|
|
|
|
StackEntry, 0, "gc_newhead");
|
|
|
|
AtEntry.CreateStore(CurrentHead, EntryNextPtr);
|
|
|
|
AtEntry.CreateStore(NewHeadVal, Head);
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// For each instruction that escapes...
|
|
|
|
EscapeEnumerator EE(F, "gc_cleanup");
|
2008-08-08 21:39:37 +02:00
|
|
|
while (IRBuilder<> *AtExit = EE.Next()) {
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// Pop the entry from the shadow stack. Don't reuse CurrentHead from
|
|
|
|
// AtEntry, since that would make the value live for the entire function.
|
2009-07-15 01:09:55 +02:00
|
|
|
Instruction *EntryNextPtr2 = CreateGEP(Context, *AtExit, StackEntry, 0, 0,
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
"gc_frame.next");
|
|
|
|
Value *SavedHead = AtExit->CreateLoad(EntryNextPtr2, "gc_savedhead");
|
|
|
|
AtExit->CreateStore(SavedHead, Head);
|
|
|
|
}
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
// Delete the original allocas (which are no longer used) and the intrinsic
|
|
|
|
// calls (which are no longer valid). Doing this last avoids invalidating
|
|
|
|
// iterators.
|
|
|
|
for (unsigned I = 0, E = Roots.size(); I != E; ++I) {
|
|
|
|
Roots[I].first->eraseFromParent();
|
|
|
|
Roots[I].second->eraseFromParent();
|
|
|
|
}
|
2009-01-16 07:53:46 +01:00
|
|
|
|
With this patch, the LowerGC transformation becomes the
ShadowStackCollector, which additionally has reduced overhead with
no sacrifice in portability.
Considering a function @fun with 8 loop-local roots,
ShadowStackCollector introduces the following overhead
(x86):
; shadowstack prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl $___gc_fun, 20(%esp)
movl $0, 24(%esp)
movl $0, 28(%esp)
movl $0, 32(%esp)
movl $0, 36(%esp)
movl $0, 40(%esp)
movl $0, 44(%esp)
movl $0, 48(%esp)
movl $0, 52(%esp)
movl %ecx, 16(%esp)
leal 16(%esp), %ecx
movl %ecx, (%eax)
; shadowstack loop overhead
(none)
; shadowstack epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; shadowstack metadata
.align 3
___gc_fun: # __gc_fun
.long 8
.space 4
In comparison to LowerGC:
; lowergc prologue
movl L_llvm_gc_root_chain$non_lazy_ptr, %eax
movl (%eax), %ecx
movl %ecx, 48(%esp)
movl $8, 52(%esp)
movl $0, 60(%esp)
movl $0, 56(%esp)
movl $0, 68(%esp)
movl $0, 64(%esp)
movl $0, 76(%esp)
movl $0, 72(%esp)
movl $0, 84(%esp)
movl $0, 80(%esp)
movl $0, 92(%esp)
movl $0, 88(%esp)
movl $0, 100(%esp)
movl $0, 96(%esp)
movl $0, 108(%esp)
movl $0, 104(%esp)
movl $0, 116(%esp)
movl $0, 112(%esp)
; lowergc loop overhead
leal 44(%esp), %eax
movl %eax, 56(%esp)
leal 40(%esp), %eax
movl %eax, 64(%esp)
leal 36(%esp), %eax
movl %eax, 72(%esp)
leal 32(%esp), %eax
movl %eax, 80(%esp)
leal 28(%esp), %eax
movl %eax, 88(%esp)
leal 24(%esp), %eax
movl %eax, 96(%esp)
leal 20(%esp), %eax
movl %eax, 104(%esp)
leal 16(%esp), %eax
movl %eax, 112(%esp)
; lowergc epilogue
movl 48(%esp), %edx
movl %edx, (%ecx)
; lowergc metadata
(none)
llvm-svn: 45670
2008-01-07 02:30:53 +01:00
|
|
|
Roots.clear();
|
|
|
|
return true;
|
|
|
|
}
|