2016-08-04 02:17:35 +08:00
|
|
|
; RUN: llc -verify-machineinstrs < %s | FileCheck %s
|
2014-08-15 21:51:57 +08:00
|
|
|
target datalayout = "e-m:e-i64:64-n32:64"
|
|
|
|
target triple = "powerpc64le-unknown-linux-gnu"
|
|
|
|
|
|
|
|
; This test checks for misordering of a TOC restore instruction relative
|
|
|
|
; to subsequent uses of the TOC register. Previously this test broke
|
|
|
|
; because there was no TOC register dependency between the instructions,
|
|
|
|
; and the usual stack-adjust instructions that held the TOC restore in
|
|
|
|
; place were optimized away.
|
|
|
|
|
|
|
|
%"class.llvm::Module" = type { %"class.llvm::LLVMContext"*, %"class.llvm::iplist", %"class.llvm::iplist.0", %"class.llvm::iplist.9", %"struct.llvm::ilist", %"class.std::basic_string", %"class.llvm::ValueSymbolTable"*, %"class.llvm::StringMap", %"class.std::unique_ptr", %"class.std::basic_string", %"class.std::basic_string", i8*, %"class.llvm::RandomNumberGenerator"*, %"class.std::basic_string", %"class.llvm::DataLayout" }
|
|
|
|
%"class.llvm::iplist" = type { %"struct.llvm::ilist_traits", %"class.llvm::GlobalVariable"* }
|
|
|
|
%"struct.llvm::ilist_traits" = type { %"class.llvm::ilist_node" }
|
|
|
|
%"class.llvm::ilist_node" = type { %"class.llvm::ilist_half_node", %"class.llvm::GlobalVariable"* }
|
|
|
|
%"class.llvm::ilist_half_node" = type { %"class.llvm::GlobalVariable"* }
|
|
|
|
%"class.llvm::GlobalVariable" = type { %"class.llvm::GlobalObject", %"class.llvm::ilist_node", i8 }
|
|
|
|
%"class.llvm::GlobalObject" = type { %"class.llvm::GlobalValue", %"class.std::basic_string", %"class.llvm::Comdat"* }
|
|
|
|
%"class.llvm::GlobalValue" = type { %"class.llvm::Constant", i32, %"class.llvm::Module"* }
|
|
|
|
%"class.llvm::Constant" = type { %"class.llvm::User" }
|
|
|
|
%"class.llvm::User" = type { %"class.llvm::Value.base", i32, %"class.llvm::Use"* }
|
|
|
|
%"class.llvm::Value.base" = type <{ i32 (...)**, %"class.llvm::Type"*, %"class.llvm::Use"*, %"class.llvm::StringMapEntry"*, i8, i8, i16 }>
|
|
|
|
%"class.llvm::Type" = type { %"class.llvm::LLVMContext"*, i32, i32, %"class.llvm::Type"** }
|
|
|
|
%"class.llvm::StringMapEntry" = type opaque
|
|
|
|
%"class.llvm::Use" = type { %"class.llvm::Value"*, %"class.llvm::Use"*, %"class.llvm::PointerIntPair" }
|
|
|
|
%"class.llvm::Value" = type { i32 (...)**, %"class.llvm::Type"*, %"class.llvm::Use"*, %"class.llvm::StringMapEntry"*, i8, i8, i16 }
|
|
|
|
%"class.llvm::PointerIntPair" = type { i64 }
|
|
|
|
%"class.llvm::Comdat" = type { %"class.llvm::StringMapEntry.43"*, i32 }
|
|
|
|
%"class.llvm::StringMapEntry.43" = type opaque
|
|
|
|
%"class.llvm::iplist.0" = type { %"struct.llvm::ilist_traits.1", %"class.llvm::Function"* }
|
|
|
|
%"struct.llvm::ilist_traits.1" = type { %"class.llvm::ilist_node.7" }
|
|
|
|
%"class.llvm::ilist_node.7" = type { %"class.llvm::ilist_half_node.8", %"class.llvm::Function"* }
|
|
|
|
%"class.llvm::ilist_half_node.8" = type { %"class.llvm::Function"* }
|
|
|
|
%"class.llvm::Function" = type { %"class.llvm::GlobalObject", %"class.llvm::ilist_node.7", %"class.llvm::iplist.44", %"class.llvm::iplist.52", %"class.llvm::ValueSymbolTable"*, %"class.llvm::AttributeSet" }
|
|
|
|
%"class.llvm::iplist.44" = type { %"struct.llvm::ilist_traits.45", %"class.llvm::BasicBlock"* }
|
|
|
|
%"struct.llvm::ilist_traits.45" = type { %"class.llvm::ilist_half_node.51" }
|
|
|
|
%"class.llvm::ilist_half_node.51" = type { %"class.llvm::BasicBlock"* }
|
|
|
|
%"class.llvm::BasicBlock" = type { %"class.llvm::Value.base", %"class.llvm::ilist_node.61", %"class.llvm::iplist.62", %"class.llvm::Function"* }
|
|
|
|
%"class.llvm::ilist_node.61" = type { %"class.llvm::ilist_half_node.51", %"class.llvm::BasicBlock"* }
|
|
|
|
%"class.llvm::iplist.62" = type { %"struct.llvm::ilist_traits.63", %"class.llvm::Instruction"* }
|
|
|
|
%"struct.llvm::ilist_traits.63" = type { %"class.llvm::ilist_half_node.69" }
|
|
|
|
%"class.llvm::ilist_half_node.69" = type { %"class.llvm::Instruction"* }
|
|
|
|
%"class.llvm::Instruction" = type { %"class.llvm::User", %"class.llvm::ilist_node.70", %"class.llvm::BasicBlock"*, %"class.llvm::DebugLoc" }
|
|
|
|
%"class.llvm::ilist_node.70" = type { %"class.llvm::ilist_half_node.69", %"class.llvm::Instruction"* }
|
|
|
|
%"class.llvm::DebugLoc" = type { i32, i32 }
|
|
|
|
%"class.llvm::iplist.52" = type { %"struct.llvm::ilist_traits.53", %"class.llvm::Argument"* }
|
|
|
|
%"struct.llvm::ilist_traits.53" = type { %"class.llvm::ilist_half_node.59" }
|
|
|
|
%"class.llvm::ilist_half_node.59" = type { %"class.llvm::Argument"* }
|
|
|
|
%"class.llvm::Argument" = type { %"class.llvm::Value.base", %"class.llvm::ilist_node.60", %"class.llvm::Function"* }
|
|
|
|
%"class.llvm::ilist_node.60" = type { %"class.llvm::ilist_half_node.59", %"class.llvm::Argument"* }
|
|
|
|
%"class.llvm::AttributeSet" = type { %"class.llvm::AttributeSetImpl"* }
|
|
|
|
%"class.llvm::AttributeSetImpl" = type opaque
|
|
|
|
%"class.llvm::iplist.9" = type { %"struct.llvm::ilist_traits.10", %"class.llvm::GlobalAlias"* }
|
|
|
|
%"struct.llvm::ilist_traits.10" = type { %"class.llvm::ilist_node.16" }
|
|
|
|
%"class.llvm::ilist_node.16" = type { %"class.llvm::ilist_half_node.17", %"class.llvm::GlobalAlias"* }
|
|
|
|
%"class.llvm::ilist_half_node.17" = type { %"class.llvm::GlobalAlias"* }
|
|
|
|
%"class.llvm::GlobalAlias" = type { %"class.llvm::GlobalValue", %"class.llvm::ilist_node.16" }
|
|
|
|
%"struct.llvm::ilist" = type { %"class.llvm::iplist.18" }
|
|
|
|
%"class.llvm::iplist.18" = type { %"struct.llvm::ilist_traits.19", %"class.llvm::NamedMDNode"* }
|
|
|
|
%"struct.llvm::ilist_traits.19" = type { %"class.llvm::ilist_node.24" }
|
|
|
|
%"class.llvm::ilist_node.24" = type { %"class.llvm::ilist_half_node.25", %"class.llvm::NamedMDNode"* }
|
|
|
|
%"class.llvm::ilist_half_node.25" = type { %"class.llvm::NamedMDNode"* }
|
|
|
|
%"class.llvm::NamedMDNode" = type { %"class.llvm::ilist_node.24", %"class.std::basic_string", %"class.llvm::Module"*, i8* }
|
|
|
|
%"class.llvm::ValueSymbolTable" = type opaque
|
|
|
|
%"class.llvm::StringMap" = type { %"class.llvm::StringMapImpl", %"class.llvm::MallocAllocator" }
|
|
|
|
%"class.llvm::StringMapImpl" = type { %"class.llvm::StringMapEntryBase"**, i32, i32, i32, i32 }
|
|
|
|
%"class.llvm::StringMapEntryBase" = type { i32 }
|
|
|
|
%"class.llvm::MallocAllocator" = type { i8 }
|
|
|
|
%"class.std::unique_ptr" = type { %"class.std::tuple" }
|
|
|
|
%"class.std::tuple" = type { %"struct.std::_Tuple_impl" }
|
|
|
|
%"struct.std::_Tuple_impl" = type { %"struct.std::_Head_base.28" }
|
|
|
|
%"struct.std::_Head_base.28" = type { %"class.llvm::GVMaterializer"* }
|
|
|
|
%"class.llvm::GVMaterializer" = type opaque
|
|
|
|
%"class.llvm::RandomNumberGenerator" = type opaque
|
|
|
|
%"class.std::basic_string" = type { %"struct.std::basic_string<char, std::char_traits<char>, std::allocator<char> >::_Alloc_hider" }
|
|
|
|
%"struct.std::basic_string<char, std::char_traits<char>, std::allocator<char> >::_Alloc_hider" = type { i8* }
|
|
|
|
%"class.llvm::DataLayout" = type { i8, i32, i32, [4 x i8], %"class.llvm::SmallVector", %"class.llvm::SmallVector.29", %"class.llvm::SmallVector.36", i8* }
|
|
|
|
%"class.llvm::SmallVector" = type { %"class.llvm::SmallVectorImpl.base", %"struct.llvm::SmallVectorStorage" }
|
|
|
|
%"class.llvm::SmallVectorImpl.base" = type { %"class.llvm::SmallVectorTemplateBase.base" }
|
|
|
|
%"class.llvm::SmallVectorTemplateBase.base" = type { %"class.llvm::SmallVectorTemplateCommon.base" }
|
|
|
|
%"class.llvm::SmallVectorTemplateCommon.base" = type <{ %"class.llvm::SmallVectorBase", %"struct.llvm::AlignedCharArrayUnion" }>
|
|
|
|
%"class.llvm::SmallVectorBase" = type { i8*, i8*, i8* }
|
|
|
|
%"struct.llvm::AlignedCharArrayUnion" = type { %"struct.llvm::AlignedCharArray" }
|
|
|
|
%"struct.llvm::AlignedCharArray" = type { [1 x i8] }
|
|
|
|
%"struct.llvm::SmallVectorStorage" = type { [7 x %"struct.llvm::AlignedCharArrayUnion"] }
|
|
|
|
%"class.llvm::SmallVector.29" = type { %"class.llvm::SmallVectorImpl.30", %"struct.llvm::SmallVectorStorage.35" }
|
|
|
|
%"class.llvm::SmallVectorImpl.30" = type { %"class.llvm::SmallVectorTemplateBase.31" }
|
|
|
|
%"class.llvm::SmallVectorTemplateBase.31" = type { %"class.llvm::SmallVectorTemplateCommon.32" }
|
|
|
|
%"class.llvm::SmallVectorTemplateCommon.32" = type { %"class.llvm::SmallVectorBase", %"struct.llvm::AlignedCharArrayUnion.33" }
|
|
|
|
%"struct.llvm::AlignedCharArrayUnion.33" = type { %"struct.llvm::AlignedCharArray.34" }
|
|
|
|
%"struct.llvm::AlignedCharArray.34" = type { [8 x i8] }
|
|
|
|
%"struct.llvm::SmallVectorStorage.35" = type { [15 x %"struct.llvm::AlignedCharArrayUnion.33"] }
|
|
|
|
%"class.llvm::SmallVector.36" = type { %"class.llvm::SmallVectorImpl.37", %"struct.llvm::SmallVectorStorage.42" }
|
|
|
|
%"class.llvm::SmallVectorImpl.37" = type { %"class.llvm::SmallVectorTemplateBase.38" }
|
|
|
|
%"class.llvm::SmallVectorTemplateBase.38" = type { %"class.llvm::SmallVectorTemplateCommon.39" }
|
|
|
|
%"class.llvm::SmallVectorTemplateCommon.39" = type { %"class.llvm::SmallVectorBase", %"struct.llvm::AlignedCharArrayUnion.40" }
|
|
|
|
%"struct.llvm::AlignedCharArrayUnion.40" = type { %"struct.llvm::AlignedCharArray.41" }
|
|
|
|
%"struct.llvm::AlignedCharArray.41" = type { [16 x i8] }
|
|
|
|
%"struct.llvm::SmallVectorStorage.42" = type { [7 x %"struct.llvm::AlignedCharArrayUnion.40"] }
|
|
|
|
%"class.llvm::SMDiagnostic" = type { %"class.llvm::SourceMgr"*, %"class.llvm::SMLoc", %"class.std::basic_string", i32, i32, i32, %"class.std::basic_string", %"class.std::basic_string", %"class.std::vector.79", %"class.llvm::SmallVector.84" }
|
|
|
|
%"class.llvm::SourceMgr" = type { %"class.std::vector", %"class.std::vector.74", i8*, void (%"class.llvm::SMDiagnostic"*, i8*)*, i8* }
|
|
|
|
%"class.std::vector" = type { %"struct.std::_Vector_base" }
|
|
|
|
%"struct.std::_Vector_base" = type { %"struct.std::_Vector_base<llvm::SourceMgr::SrcBuffer, std::allocator<llvm::SourceMgr::SrcBuffer> >::_Vector_impl" }
|
|
|
|
%"struct.std::_Vector_base<llvm::SourceMgr::SrcBuffer, std::allocator<llvm::SourceMgr::SrcBuffer> >::_Vector_impl" = type { %"struct.llvm::SourceMgr::SrcBuffer"*, %"struct.llvm::SourceMgr::SrcBuffer"*, %"struct.llvm::SourceMgr::SrcBuffer"* }
|
|
|
|
%"struct.llvm::SourceMgr::SrcBuffer" = type { %"class.llvm::MemoryBuffer"*, %"class.llvm::SMLoc" }
|
|
|
|
%"class.llvm::MemoryBuffer" = type { i32 (...)**, i8*, i8* }
|
|
|
|
%"class.std::vector.74" = type { %"struct.std::_Vector_base.75" }
|
|
|
|
%"struct.std::_Vector_base.75" = type { %"struct.std::_Vector_base<std::basic_string<char>, std::allocator<std::basic_string<char> > >::_Vector_impl" }
|
|
|
|
%"struct.std::_Vector_base<std::basic_string<char>, std::allocator<std::basic_string<char> > >::_Vector_impl" = type { %"class.std::basic_string"*, %"class.std::basic_string"*, %"class.std::basic_string"* }
|
|
|
|
%"class.llvm::SMLoc" = type { i8* }
|
|
|
|
%"class.std::vector.79" = type { %"struct.std::_Vector_base.80" }
|
|
|
|
%"struct.std::_Vector_base.80" = type { %"struct.std::_Vector_base<std::pair<unsigned int, unsigned int>, std::allocator<std::pair<unsigned int, unsigned int> > >::_Vector_impl" }
|
|
|
|
%"struct.std::_Vector_base<std::pair<unsigned int, unsigned int>, std::allocator<std::pair<unsigned int, unsigned int> > >::_Vector_impl" = type { %"struct.std::pair"*, %"struct.std::pair"*, %"struct.std::pair"* }
|
|
|
|
%"struct.std::pair" = type { i32, i32 }
|
|
|
|
%"class.llvm::SmallVector.84" = type { %"class.llvm::SmallVectorImpl.85", %"struct.llvm::SmallVectorStorage.90" }
|
|
|
|
%"class.llvm::SmallVectorImpl.85" = type { %"class.llvm::SmallVectorTemplateBase.86" }
|
|
|
|
%"class.llvm::SmallVectorTemplateBase.86" = type { %"class.llvm::SmallVectorTemplateCommon.87" }
|
|
|
|
%"class.llvm::SmallVectorTemplateCommon.87" = type { %"class.llvm::SmallVectorBase", %"struct.llvm::AlignedCharArrayUnion.88" }
|
|
|
|
%"struct.llvm::AlignedCharArrayUnion.88" = type { %"struct.llvm::AlignedCharArray.89" }
|
|
|
|
%"struct.llvm::AlignedCharArray.89" = type { [24 x i8] }
|
|
|
|
%"struct.llvm::SmallVectorStorage.90" = type { [3 x %"struct.llvm::AlignedCharArrayUnion.88"] }
|
|
|
|
%"class.llvm::LLVMContext" = type { %"class.llvm::LLVMContextImpl"* }
|
|
|
|
%"class.llvm::LLVMContextImpl" = type opaque
|
|
|
|
%"class.std::allocator" = type { i8 }
|
|
|
|
%"class.llvm::ErrorOr.109" = type { %union.anon.110, i8, [7 x i8] }
|
|
|
|
%union.anon.110 = type { %"struct.llvm::AlignedCharArrayUnion.93" }
|
|
|
|
%"struct.llvm::AlignedCharArrayUnion.93" = type { %"struct.llvm::AlignedCharArray.94" }
|
|
|
|
%"struct.llvm::AlignedCharArray.94" = type { [16 x i8] }
|
|
|
|
%"class.llvm::ErrorOr" = type { %union.anon, i8, [7 x i8] }
|
|
|
|
%union.anon = type { %"struct.llvm::AlignedCharArrayUnion.93" }
|
|
|
|
%"class.std::error_category" = type { i32 (...)** }
|
|
|
|
%"struct.std::basic_string<char, std::char_traits<char>, std::allocator<char> >::_Rep" = type { %"struct.std::basic_string<char, std::char_traits<char>, std::allocator<char> >::_Rep_base" }
|
|
|
|
%"struct.std::basic_string<char, std::char_traits<char>, std::allocator<char> >::_Rep_base" = type { i64, i64, i32 }
|
|
|
|
%"class.llvm::SMFixIt" = type { %"class.llvm::SMRange", %"class.std::basic_string" }
|
|
|
|
%"class.llvm::SMRange" = type { %"class.llvm::SMLoc", %"class.llvm::SMLoc" }
|
|
|
|
%"struct.llvm::NamedRegionTimer" = type { %"class.llvm::TimeRegion" }
|
|
|
|
%"class.llvm::TimeRegion" = type { %"class.llvm::Timer"* }
|
|
|
|
%"class.llvm::Timer" = type { %"class.llvm::TimeRecord", %"class.std::basic_string", i8, %"class.llvm::TimerGroup"*, %"class.llvm::Timer"**, %"class.llvm::Timer"* }
|
|
|
|
%"class.llvm::TimeRecord" = type { double, double, double, i64 }
|
|
|
|
%"class.llvm::TimerGroup" = type { %"class.std::basic_string", %"class.llvm::Timer"*, %"class.std::vector.103", %"class.llvm::TimerGroup"**, %"class.llvm::TimerGroup"* }
|
|
|
|
%"class.std::vector.103" = type { %"struct.std::_Vector_base.104" }
|
|
|
|
%"struct.std::_Vector_base.104" = type { %"struct.std::_Vector_base<std::pair<llvm::TimeRecord, std::basic_string<char> >, std::allocator<std::pair<llvm::TimeRecord, std::basic_string<char> > > >::_Vector_impl" }
|
|
|
|
%"struct.std::_Vector_base<std::pair<llvm::TimeRecord, std::basic_string<char> >, std::allocator<std::pair<llvm::TimeRecord, std::basic_string<char> > > >::_Vector_impl" = type { %"struct.std::pair.108"*, %"struct.std::pair.108"*, %"struct.std::pair.108"* }
|
|
|
|
%"struct.std::pair.108" = type opaque
|
|
|
|
%struct.LLVMOpaqueContext = type opaque
|
|
|
|
%struct.LLVMOpaqueMemoryBuffer = type opaque
|
|
|
|
%struct.LLVMOpaqueModule = type opaque
|
|
|
|
%"class.llvm::raw_string_ostream" = type { %"class.llvm::raw_ostream.base", %"class.std::basic_string"* }
|
|
|
|
%"class.llvm::raw_ostream.base" = type <{ i32 (...)**, i8*, i8*, i8*, i32 }>
|
|
|
|
%"class.llvm::raw_ostream" = type { i32 (...)**, i8*, i8*, i8*, i32 }
|
|
|
|
|
|
|
|
@.str = private unnamed_addr constant [28 x i8] c"Could not open input file: \00", align 1
|
|
|
|
@.str1 = private unnamed_addr constant [54 x i8] c"!HasError && \22Cannot get value when an error exists!\22\00", align 1
|
|
|
|
@.str2 = private unnamed_addr constant [61 x i8] c"/home/wschmidt/llvm/llvm-test/include/llvm/Support/ErrorOr.h\00", align 1
|
|
|
|
@__PRETTY_FUNCTION__._ZN4llvm7ErrorOrISt10unique_ptrINS_12MemoryBufferESt14default_deleteIS2_EEE10getStorageEv = private unnamed_addr constant [206 x i8] c"storage_type *llvm::ErrorOr<std::unique_ptr<llvm::MemoryBuffer, std::default_delete<llvm::MemoryBuffer> > >::getStorage() [T = std::unique_ptr<llvm::MemoryBuffer, std::default_delete<llvm::MemoryBuffer> >]\00", align 1
|
|
|
|
@_ZNSs4_Rep20_S_empty_rep_storageE = external global [0 x i64]
|
|
|
|
|
|
|
|
declare void @_ZN4llvm12MemoryBuffer14getFileOrSTDINENS_9StringRefEl(%"class.llvm::ErrorOr"* sret, [2 x i64], i64) #1
|
|
|
|
|
|
|
|
declare void @_ZN4llvm16NamedRegionTimerC1ENS_9StringRefES1_b(%"struct.llvm::NamedRegionTimer"*, [2 x i64], [2 x i64], i1 zeroext) #1
|
|
|
|
|
|
|
|
; Function Attrs: nounwind
|
|
|
|
define %"class.llvm::Module"* @_ZN4llvm11ParseIRFileERKSsRNS_12SMDiagnosticERNS_11LLVMContextE(%"class.std::basic_string"* nocapture readonly dereferenceable(8) %Filename, %"class.llvm::SMDiagnostic"* dereferenceable(200) %Err, %"class.llvm::LLVMContext"* dereferenceable(8) %Context) #0 {
|
|
|
|
entry:
|
|
|
|
; CHECK: .globl _ZN4llvm11ParseIRFileERKSsRNS_12SMDiagnosticERNS_11LLVMContextE
|
|
|
|
; CHECK: bctrl
|
|
|
|
; CHECK: ld 2, 24(1)
|
|
|
|
; CHECK: addis [[REG:[0-9]+]], 2, .L.str@toc@ha
|
|
|
|
; CHECK: addi {{[0-9]+}}, [[REG]], .L.str@toc@l
|
|
|
|
; CHECK: bl _ZNSs6insertEmPKcm
|
|
|
|
%.atomicdst.i.i.i.i.i46 = alloca i32, align 4
|
|
|
|
%ref.tmp.i.i47 = alloca %"class.std::allocator", align 1
|
|
|
|
%.atomicdst.i.i.i.i.i = alloca i32, align 4
|
|
|
|
%ref.tmp.i.i = alloca %"class.std::allocator", align 1
|
|
|
|
%ref.tmp.i.i2.i = alloca %"class.std::allocator", align 1
|
|
|
|
%ref.tmp.i.i.i = alloca %"class.std::allocator", align 1
|
|
|
|
%FileOrErr = alloca %"class.llvm::ErrorOr", align 8
|
|
|
|
%ref.tmp = alloca %"class.llvm::SMDiagnostic", align 8
|
|
|
|
%ref.tmp5 = alloca %"class.std::basic_string", align 8
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%_M_p.i.i.i = getelementptr inbounds %"class.std::basic_string", %"class.std::basic_string"* %Filename, i64 0, i32 0, i32 0
|
2015-02-28 05:17:42 +08:00
|
|
|
%0 = load i8*, i8** %_M_p.i.i.i, align 8, !tbaa !1
|
2014-08-15 21:51:57 +08:00
|
|
|
%1 = ptrtoint i8* %0 to i64
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%arrayidx.i.i.i = getelementptr inbounds i8, i8* %0, i64 -24
|
2014-08-15 21:51:57 +08:00
|
|
|
%_M_length.i.i = bitcast i8* %arrayidx.i.i.i to i64*
|
2015-02-28 05:17:42 +08:00
|
|
|
%2 = load i64, i64* %_M_length.i.i, align 8, !tbaa !7
|
2014-08-15 21:51:57 +08:00
|
|
|
%.fca.0.insert18 = insertvalue [2 x i64] undef, i64 %1, 0
|
|
|
|
%.fca.1.insert21 = insertvalue [2 x i64] %.fca.0.insert18, i64 %2, 1
|
|
|
|
call void @_ZN4llvm12MemoryBuffer14getFileOrSTDINENS_9StringRefEl(%"class.llvm::ErrorOr"* sret %FileOrErr, [2 x i64] %.fca.1.insert21, i64 -1) #3
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%HasError.i24 = getelementptr inbounds %"class.llvm::ErrorOr", %"class.llvm::ErrorOr"* %FileOrErr, i64 0, i32 1
|
2015-02-28 05:17:42 +08:00
|
|
|
%bf.load.i25 = load i8, i8* %HasError.i24, align 8
|
2014-08-15 21:51:57 +08:00
|
|
|
%3 = and i8 %bf.load.i25, 1
|
|
|
|
%bf.cast.i26 = icmp eq i8 %3, 0
|
|
|
|
br i1 %bf.cast.i26, label %_ZN4llvm7ErrorOrISt10unique_ptrINS_12MemoryBufferESt14default_deleteIS2_EEE3getEv.exit, label %_ZNK4llvm7ErrorOrISt10unique_ptrINS_12MemoryBufferESt14default_deleteIS2_EEE8getErrorEv.exit
|
|
|
|
|
|
|
|
_ZNK4llvm7ErrorOrISt10unique_ptrINS_12MemoryBufferESt14default_deleteIS2_EEE8getErrorEv.exit: ; preds = %entry
|
|
|
|
%retval.sroa.0.0..sroa_cast.i = bitcast %"class.llvm::ErrorOr"* %FileOrErr to i64*
|
2015-02-28 05:17:42 +08:00
|
|
|
%retval.sroa.0.0.copyload.i = load i64, i64* %retval.sroa.0.0..sroa_cast.i, align 8
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%retval.sroa.3.0..sroa_idx.i = getelementptr inbounds %"class.llvm::ErrorOr", %"class.llvm::ErrorOr"* %FileOrErr, i64 0, i32 0, i32 0, i32 0, i32 0, i64 8
|
2014-08-15 21:51:57 +08:00
|
|
|
%retval.sroa.3.0..sroa_cast.i = bitcast i8* %retval.sroa.3.0..sroa_idx.i to i64*
|
2015-02-28 05:17:42 +08:00
|
|
|
%retval.sroa.3.0.copyload.i = load i64, i64* %retval.sroa.3.0..sroa_cast.i, align 8
|
2014-08-15 21:51:57 +08:00
|
|
|
%phitmp = trunc i64 %retval.sroa.0.0.copyload.i to i32
|
|
|
|
%cmp.i = icmp eq i32 %phitmp, 0
|
|
|
|
br i1 %cmp.i, label %cond.false.i.i, label %if.then
|
|
|
|
|
|
|
|
if.then: ; preds = %_ZNK4llvm7ErrorOrISt10unique_ptrINS_12MemoryBufferESt14default_deleteIS2_EEE8getErrorEv.exit
|
|
|
|
%.c = inttoptr i64 %retval.sroa.3.0.copyload.i to %"class.std::error_category"*
|
2015-02-28 05:17:42 +08:00
|
|
|
%4 = load i8*, i8** %_M_p.i.i.i, align 8, !tbaa !1
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%arrayidx.i.i.i30 = getelementptr inbounds i8, i8* %4, i64 -24
|
2014-08-15 21:51:57 +08:00
|
|
|
%_M_length.i.i31 = bitcast i8* %arrayidx.i.i.i30 to i64*
|
2015-02-28 05:17:42 +08:00
|
|
|
%5 = load i64, i64* %_M_length.i.i31, align 8, !tbaa !7
|
2014-08-15 21:51:57 +08:00
|
|
|
%6 = inttoptr i64 %retval.sroa.3.0.copyload.i to void (%"class.std::basic_string"*, %"class.std::error_category"*, i32)***
|
2015-02-28 05:17:42 +08:00
|
|
|
%vtable.i = load void (%"class.std::basic_string"*, %"class.std::error_category"*, i32)**, void (%"class.std::basic_string"*, %"class.std::error_category"*, i32)*** %6, align 8, !tbaa !11
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%vfn.i = getelementptr inbounds void (%"class.std::basic_string"*, %"class.std::error_category"*, i32)*, void (%"class.std::basic_string"*, %"class.std::error_category"*, i32)** %vtable.i, i64 3
|
2015-02-28 05:17:42 +08:00
|
|
|
%7 = load void (%"class.std::basic_string"*, %"class.std::error_category"*, i32)*, void (%"class.std::basic_string"*, %"class.std::error_category"*, i32)** %vfn.i, align 8
|
2014-08-15 21:51:57 +08:00
|
|
|
call void %7(%"class.std::basic_string"* sret %ref.tmp5, %"class.std::error_category"* %.c, i32 signext %phitmp) #3
|
2015-03-14 02:20:45 +08:00
|
|
|
%call2.i.i = call dereferenceable(8) %"class.std::basic_string"* @_ZNSs6insertEmPKcm(%"class.std::basic_string"* %ref.tmp5, i64 0, i8* getelementptr inbounds ([28 x i8], [28 x i8]* @.str, i64 0, i64 0), i64 27) #3
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%_M_p2.i.i.i.i = getelementptr inbounds %"class.std::basic_string", %"class.std::basic_string"* %call2.i.i, i64 0, i32 0, i32 0
|
2015-02-28 05:17:42 +08:00
|
|
|
%8 = load i8*, i8** %_M_p2.i.i.i.i, align 8, !tbaa !13
|
2015-03-14 02:20:45 +08:00
|
|
|
store i8* bitcast (i64* getelementptr inbounds ([0 x i64], [0 x i64]* @_ZNSs4_Rep20_S_empty_rep_storageE, i64 0, i64 3) to i8*), i8** %_M_p2.i.i.i.i, align 8, !tbaa !1
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%arrayidx.i.i.i36 = getelementptr inbounds i8, i8* %8, i64 -24
|
2014-08-15 21:51:57 +08:00
|
|
|
%_M_length.i.i37 = bitcast i8* %arrayidx.i.i.i36 to i64*
|
2015-02-28 05:17:42 +08:00
|
|
|
%9 = load i64, i64* %_M_length.i.i37, align 8, !tbaa !7
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%Filename.i = getelementptr inbounds %"class.llvm::SMDiagnostic", %"class.llvm::SMDiagnostic"* %ref.tmp, i64 0, i32 2
|
|
|
|
%10 = getelementptr inbounds %"class.std::allocator", %"class.std::allocator"* %ref.tmp.i.i2.i, i64 0, i32 0
|
2014-08-15 21:51:57 +08:00
|
|
|
%11 = bitcast %"class.llvm::SMDiagnostic"* %ref.tmp to i8*
|
Remove alignment argument from memcpy/memmove/memset in favour of alignment attributes (Step 1)
Summary:
This is a resurrection of work first proposed and discussed in Aug 2015:
http://lists.llvm.org/pipermail/llvm-dev/2015-August/089384.html
and initially landed (but then backed out) in Nov 2015:
http://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20151109/312083.html
The @llvm.memcpy/memmove/memset intrinsics currently have an explicit argument
which is required to be a constant integer. It represents the alignment of the
dest (and source), and so must be the minimum of the actual alignment of the
two.
This change is the first in a series that allows source and dest to each
have their own alignments by using the alignment attribute on their arguments.
In this change we:
1) Remove the alignment argument.
2) Add alignment attributes to the source & dest arguments. We, temporarily,
require that the alignments for source & dest be equal.
For example, code which used to read:
call void @llvm.memcpy.p0i8.p0i8.i32(i8* %dest, i8* %src, i32 100, i32 4, i1 false)
will now read
call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %dest, i8* align 4 %src, i32 100, i1 false)
Downstream users may have to update their lit tests that check for
@llvm.memcpy/memmove/memset call/declaration patterns. The following extended sed script
may help with updating the majority of your tests, but it does not catch all possible
patterns so some manual checking and updating will be required.
s~declare void @llvm\.mem(set|cpy|move)\.p([^(]*)\((.*), i32, i1\)~declare void @llvm.mem\1.p\2(\3, i1)~g
s~call void @llvm\.memset\.p([^(]*)i8\(i8([^*]*)\* (.*), i8 (.*), i8 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i8(i8\2* \3, i8 \4, i8 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i16\(i8([^*]*)\* (.*), i8 (.*), i16 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i16(i8\2* \3, i8 \4, i16 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i32\(i8([^*]*)\* (.*), i8 (.*), i32 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i32(i8\2* \3, i8 \4, i32 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i64\(i8([^*]*)\* (.*), i8 (.*), i64 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i64(i8\2* \3, i8 \4, i64 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i128\(i8([^*]*)\* (.*), i8 (.*), i128 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i128(i8\2* \3, i8 \4, i128 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i8\(i8([^*]*)\* (.*), i8 (.*), i8 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i8(i8\2* align \6 \3, i8 \4, i8 \5, i1 \7)~g
s~call void @llvm\.memset\.p([^(]*)i16\(i8([^*]*)\* (.*), i8 (.*), i16 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i16(i8\2* align \6 \3, i8 \4, i16 \5, i1 \7)~g
s~call void @llvm\.memset\.p([^(]*)i32\(i8([^*]*)\* (.*), i8 (.*), i32 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i32(i8\2* align \6 \3, i8 \4, i32 \5, i1 \7)~g
s~call void @llvm\.memset\.p([^(]*)i64\(i8([^*]*)\* (.*), i8 (.*), i64 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i64(i8\2* align \6 \3, i8 \4, i64 \5, i1 \7)~g
s~call void @llvm\.memset\.p([^(]*)i128\(i8([^*]*)\* (.*), i8 (.*), i128 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i128(i8\2* align \6 \3, i8 \4, i128 \5, i1 \7)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i8\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i8 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i8(i8\3* \4, i8\5* \6, i8 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i16\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i16 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i16(i8\3* \4, i8\5* \6, i16 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i32\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i32 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i32(i8\3* \4, i8\5* \6, i32 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i64\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i64 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i64(i8\3* \4, i8\5* \6, i64 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i128\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i128 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i128(i8\3* \4, i8\5* \6, i128 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i8\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i8 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i8(i8\3* align \8 \4, i8\5* align \8 \6, i8 \7, i1 \9)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i16\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i16 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i16(i8\3* align \8 \4, i8\5* align \8 \6, i16 \7, i1 \9)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i32\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i32 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i32(i8\3* align \8 \4, i8\5* align \8 \6, i32 \7, i1 \9)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i64\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i64 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i64(i8\3* align \8 \4, i8\5* align \8 \6, i64 \7, i1 \9)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i128\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i128 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i128(i8\3* align \8 \4, i8\5* align \8 \6, i128 \7, i1 \9)~g
The remaining changes in the series will:
Step 2) Expand the IRBuilder API to allow creation of memcpy/memmove with differing
source and dest alignments.
Step 3) Update Clang to use the new IRBuilder API.
Step 4) Update Polly to use the new IRBuilder API.
Step 5) Update LLVM passes that create memcpy/memmove calls to use the new IRBuilder API,
and those that use use MemIntrinsicInst::[get|set]Alignment() to use
getDestAlignment() and getSourceAlignment() instead.
Step 6) Remove the single-alignment IRBuilder API for memcpy/memmove, and the
MemIntrinsicInst::[get|set]Alignment() methods.
Reviewers: pete, hfinkel, lhames, reames, bollu
Reviewed By: reames
Subscribers: niosHD, reames, jholewinski, qcolombet, jfb, sanjoy, arsenm, dschuff, dylanmckay, mehdi_amini, sdardis, nemanjai, david2050, nhaehnle, javed.absar, sbc100, jgravelle-google, eraman, aheejin, kbarton, JDevlieghere, asb, rbar, johnrusso, simoncook, jordy.potman.lists, apazos, sabuasal, llvm-commits
Differential Revision: https://reviews.llvm.org/D41675
llvm-svn: 322965
2018-01-20 01:13:12 +08:00
|
|
|
call void @llvm.memset.p0i8.i64(i8* align 8 %11, i8 0, i64 16, i1 false) #3
|
2017-04-11 04:18:21 +08:00
|
|
|
call void @llvm.lifetime.start.p0i8(i64 1, i8* %10) #3
|
2014-08-15 21:51:57 +08:00
|
|
|
%tobool.i.i4.i = icmp eq i8* %4, null
|
|
|
|
br i1 %tobool.i.i4.i, label %if.then.i.i6.i, label %if.end.i.i8.i
|
|
|
|
|
|
|
|
if.then.i.i6.i: ; preds = %if.then
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%_M_p.i.i.i.i.i.i5.i = getelementptr inbounds %"class.std::basic_string", %"class.std::basic_string"* %Filename.i, i64 0, i32 0, i32 0
|
2015-03-14 02:20:45 +08:00
|
|
|
store i8* bitcast (i64* getelementptr inbounds ([0 x i64], [0 x i64]* @_ZNSs4_Rep20_S_empty_rep_storageE, i64 0, i64 3) to i8*), i8** %_M_p.i.i.i.i.i.i5.i, align 8, !tbaa !13
|
2014-08-15 21:51:57 +08:00
|
|
|
br label %_ZNK4llvm9StringRefcvSsEv.exit9.i
|
|
|
|
|
|
|
|
if.end.i.i8.i: ; preds = %if.then
|
|
|
|
call void @_ZNSsC1EPKcmRKSaIcE(%"class.std::basic_string"* %Filename.i, i8* %4, i64 %5, %"class.std::allocator"* dereferenceable(1) %ref.tmp.i.i2.i) #3
|
|
|
|
br label %_ZNK4llvm9StringRefcvSsEv.exit9.i
|
|
|
|
|
|
|
|
_ZNK4llvm9StringRefcvSsEv.exit9.i: ; preds = %if.end.i.i8.i, %if.then.i.i6.i
|
2017-04-11 04:18:21 +08:00
|
|
|
call void @llvm.lifetime.end.p0i8(i64 1, i8* %10) #3
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%LineNo.i = getelementptr inbounds %"class.llvm::SMDiagnostic", %"class.llvm::SMDiagnostic"* %ref.tmp, i64 0, i32 3
|
2014-08-15 21:51:57 +08:00
|
|
|
store i32 -1, i32* %LineNo.i, align 8, !tbaa !14
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%ColumnNo.i = getelementptr inbounds %"class.llvm::SMDiagnostic", %"class.llvm::SMDiagnostic"* %ref.tmp, i64 0, i32 4
|
2014-08-15 21:51:57 +08:00
|
|
|
store i32 -1, i32* %ColumnNo.i, align 4, !tbaa !21
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%Kind.i = getelementptr inbounds %"class.llvm::SMDiagnostic", %"class.llvm::SMDiagnostic"* %ref.tmp, i64 0, i32 5
|
2014-08-15 21:51:57 +08:00
|
|
|
store i32 0, i32* %Kind.i, align 8, !tbaa !22
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%Message.i = getelementptr inbounds %"class.llvm::SMDiagnostic", %"class.llvm::SMDiagnostic"* %ref.tmp, i64 0, i32 6
|
|
|
|
%12 = getelementptr inbounds %"class.std::allocator", %"class.std::allocator"* %ref.tmp.i.i.i, i64 0, i32 0
|
2017-04-11 04:18:21 +08:00
|
|
|
call void @llvm.lifetime.start.p0i8(i64 1, i8* %12) #3
|
2014-08-15 21:51:57 +08:00
|
|
|
%tobool.i.i.i = icmp eq i8* %8, null
|
|
|
|
br i1 %tobool.i.i.i, label %if.then.i.i.i, label %if.end.i.i.i
|
|
|
|
|
|
|
|
if.then.i.i.i: ; preds = %_ZNK4llvm9StringRefcvSsEv.exit9.i
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%_M_p.i.i.i.i.i.i.i = getelementptr inbounds %"class.std::basic_string", %"class.std::basic_string"* %Message.i, i64 0, i32 0, i32 0
|
2015-03-14 02:20:45 +08:00
|
|
|
store i8* bitcast (i64* getelementptr inbounds ([0 x i64], [0 x i64]* @_ZNSs4_Rep20_S_empty_rep_storageE, i64 0, i64 3) to i8*), i8** %_M_p.i.i.i.i.i.i.i, align 8, !tbaa !13
|
2014-08-15 21:51:57 +08:00
|
|
|
br label %_ZN4llvm12SMDiagnosticC2ENS_9StringRefENS_9SourceMgr8DiagKindES1_.exit
|
|
|
|
|
|
|
|
if.end.i.i.i: ; preds = %_ZNK4llvm9StringRefcvSsEv.exit9.i
|
|
|
|
call void @_ZNSsC1EPKcmRKSaIcE(%"class.std::basic_string"* %Message.i, i8* %8, i64 %9, %"class.std::allocator"* dereferenceable(1) %ref.tmp.i.i.i) #3
|
|
|
|
br label %_ZN4llvm12SMDiagnosticC2ENS_9StringRefENS_9SourceMgr8DiagKindES1_.exit
|
|
|
|
|
|
|
|
_ZN4llvm12SMDiagnosticC2ENS_9StringRefENS_9SourceMgr8DiagKindES1_.exit: ; preds = %if.then.i.i.i, %if.end.i.i.i
|
2017-04-11 04:18:21 +08:00
|
|
|
call void @llvm.lifetime.end.p0i8(i64 1, i8* %12) #3
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%_M_p.i.i.i.i.i = getelementptr inbounds %"class.llvm::SMDiagnostic", %"class.llvm::SMDiagnostic"* %ref.tmp, i64 0, i32 7, i32 0, i32 0
|
2015-03-14 02:20:45 +08:00
|
|
|
store i8* bitcast (i64* getelementptr inbounds ([0 x i64], [0 x i64]* @_ZNSs4_Rep20_S_empty_rep_storageE, i64 0, i64 3) to i8*), i8** %_M_p.i.i.i.i.i, align 8, !tbaa !13
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%Ranges.i = getelementptr inbounds %"class.llvm::SMDiagnostic", %"class.llvm::SMDiagnostic"* %ref.tmp, i64 0, i32 8
|
2014-08-15 21:51:57 +08:00
|
|
|
%13 = bitcast %"class.std::vector.79"* %Ranges.i to i8*
|
Remove alignment argument from memcpy/memmove/memset in favour of alignment attributes (Step 1)
Summary:
This is a resurrection of work first proposed and discussed in Aug 2015:
http://lists.llvm.org/pipermail/llvm-dev/2015-August/089384.html
and initially landed (but then backed out) in Nov 2015:
http://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20151109/312083.html
The @llvm.memcpy/memmove/memset intrinsics currently have an explicit argument
which is required to be a constant integer. It represents the alignment of the
dest (and source), and so must be the minimum of the actual alignment of the
two.
This change is the first in a series that allows source and dest to each
have their own alignments by using the alignment attribute on their arguments.
In this change we:
1) Remove the alignment argument.
2) Add alignment attributes to the source & dest arguments. We, temporarily,
require that the alignments for source & dest be equal.
For example, code which used to read:
call void @llvm.memcpy.p0i8.p0i8.i32(i8* %dest, i8* %src, i32 100, i32 4, i1 false)
will now read
call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %dest, i8* align 4 %src, i32 100, i1 false)
Downstream users may have to update their lit tests that check for
@llvm.memcpy/memmove/memset call/declaration patterns. The following extended sed script
may help with updating the majority of your tests, but it does not catch all possible
patterns so some manual checking and updating will be required.
s~declare void @llvm\.mem(set|cpy|move)\.p([^(]*)\((.*), i32, i1\)~declare void @llvm.mem\1.p\2(\3, i1)~g
s~call void @llvm\.memset\.p([^(]*)i8\(i8([^*]*)\* (.*), i8 (.*), i8 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i8(i8\2* \3, i8 \4, i8 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i16\(i8([^*]*)\* (.*), i8 (.*), i16 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i16(i8\2* \3, i8 \4, i16 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i32\(i8([^*]*)\* (.*), i8 (.*), i32 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i32(i8\2* \3, i8 \4, i32 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i64\(i8([^*]*)\* (.*), i8 (.*), i64 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i64(i8\2* \3, i8 \4, i64 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i128\(i8([^*]*)\* (.*), i8 (.*), i128 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i128(i8\2* \3, i8 \4, i128 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i8\(i8([^*]*)\* (.*), i8 (.*), i8 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i8(i8\2* align \6 \3, i8 \4, i8 \5, i1 \7)~g
s~call void @llvm\.memset\.p([^(]*)i16\(i8([^*]*)\* (.*), i8 (.*), i16 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i16(i8\2* align \6 \3, i8 \4, i16 \5, i1 \7)~g
s~call void @llvm\.memset\.p([^(]*)i32\(i8([^*]*)\* (.*), i8 (.*), i32 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i32(i8\2* align \6 \3, i8 \4, i32 \5, i1 \7)~g
s~call void @llvm\.memset\.p([^(]*)i64\(i8([^*]*)\* (.*), i8 (.*), i64 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i64(i8\2* align \6 \3, i8 \4, i64 \5, i1 \7)~g
s~call void @llvm\.memset\.p([^(]*)i128\(i8([^*]*)\* (.*), i8 (.*), i128 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i128(i8\2* align \6 \3, i8 \4, i128 \5, i1 \7)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i8\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i8 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i8(i8\3* \4, i8\5* \6, i8 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i16\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i16 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i16(i8\3* \4, i8\5* \6, i16 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i32\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i32 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i32(i8\3* \4, i8\5* \6, i32 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i64\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i64 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i64(i8\3* \4, i8\5* \6, i64 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i128\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i128 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i128(i8\3* \4, i8\5* \6, i128 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i8\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i8 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i8(i8\3* align \8 \4, i8\5* align \8 \6, i8 \7, i1 \9)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i16\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i16 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i16(i8\3* align \8 \4, i8\5* align \8 \6, i16 \7, i1 \9)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i32\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i32 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i32(i8\3* align \8 \4, i8\5* align \8 \6, i32 \7, i1 \9)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i64\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i64 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i64(i8\3* align \8 \4, i8\5* align \8 \6, i64 \7, i1 \9)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i128\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i128 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i128(i8\3* align \8 \4, i8\5* align \8 \6, i128 \7, i1 \9)~g
The remaining changes in the series will:
Step 2) Expand the IRBuilder API to allow creation of memcpy/memmove with differing
source and dest alignments.
Step 3) Update Clang to use the new IRBuilder API.
Step 4) Update Polly to use the new IRBuilder API.
Step 5) Update LLVM passes that create memcpy/memmove calls to use the new IRBuilder API,
and those that use use MemIntrinsicInst::[get|set]Alignment() to use
getDestAlignment() and getSourceAlignment() instead.
Step 6) Remove the single-alignment IRBuilder API for memcpy/memmove, and the
MemIntrinsicInst::[get|set]Alignment() methods.
Reviewers: pete, hfinkel, lhames, reames, bollu
Reviewed By: reames
Subscribers: niosHD, reames, jholewinski, qcolombet, jfb, sanjoy, arsenm, dschuff, dylanmckay, mehdi_amini, sdardis, nemanjai, david2050, nhaehnle, javed.absar, sbc100, jgravelle-google, eraman, aheejin, kbarton, JDevlieghere, asb, rbar, johnrusso, simoncook, jordy.potman.lists, apazos, sabuasal, llvm-commits
Differential Revision: https://reviews.llvm.org/D41675
llvm-svn: 322965
2018-01-20 01:13:12 +08:00
|
|
|
call void @llvm.memset.p0i8.i64(i8* align 8 %13, i8 0, i64 24, i1 false) #3
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%14 = getelementptr inbounds %"class.llvm::SMDiagnostic", %"class.llvm::SMDiagnostic"* %ref.tmp, i64 0, i32 9, i32 0, i32 0, i32 0, i32 1, i32 0, i32 0, i64 0
|
|
|
|
%BeginX.i.i.i.i.i.i = getelementptr inbounds %"class.llvm::SMDiagnostic", %"class.llvm::SMDiagnostic"* %ref.tmp, i64 0, i32 9, i32 0, i32 0, i32 0, i32 0, i32 0
|
2014-08-15 21:51:57 +08:00
|
|
|
store i8* %14, i8** %BeginX.i.i.i.i.i.i, align 8, !tbaa !23
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%EndX.i.i.i.i.i.i = getelementptr inbounds %"class.llvm::SMDiagnostic", %"class.llvm::SMDiagnostic"* %ref.tmp, i64 0, i32 9, i32 0, i32 0, i32 0, i32 0, i32 1
|
2014-08-15 21:51:57 +08:00
|
|
|
store i8* %14, i8** %EndX.i.i.i.i.i.i, align 8, !tbaa !25
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%CapacityX.i.i.i.i.i.i = getelementptr inbounds %"class.llvm::SMDiagnostic", %"class.llvm::SMDiagnostic"* %ref.tmp, i64 0, i32 9, i32 0, i32 0, i32 0, i32 0, i32 2
|
|
|
|
%add.ptr.i.i.i.i.i.i = getelementptr inbounds %"class.llvm::SMDiagnostic", %"class.llvm::SMDiagnostic"* %ref.tmp, i64 0, i32 9, i32 0, i32 0, i32 0, i32 1, i32 0, i32 0, i64 96
|
2014-08-15 21:51:57 +08:00
|
|
|
store i8* %add.ptr.i.i.i.i.i.i, i8** %CapacityX.i.i.i.i.i.i, align 8, !tbaa !26
|
|
|
|
%15 = bitcast %"class.llvm::SMDiagnostic"* %Err to i8*
|
Remove alignment argument from memcpy/memmove/memset in favour of alignment attributes (Step 1)
Summary:
This is a resurrection of work first proposed and discussed in Aug 2015:
http://lists.llvm.org/pipermail/llvm-dev/2015-August/089384.html
and initially landed (but then backed out) in Nov 2015:
http://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20151109/312083.html
The @llvm.memcpy/memmove/memset intrinsics currently have an explicit argument
which is required to be a constant integer. It represents the alignment of the
dest (and source), and so must be the minimum of the actual alignment of the
two.
This change is the first in a series that allows source and dest to each
have their own alignments by using the alignment attribute on their arguments.
In this change we:
1) Remove the alignment argument.
2) Add alignment attributes to the source & dest arguments. We, temporarily,
require that the alignments for source & dest be equal.
For example, code which used to read:
call void @llvm.memcpy.p0i8.p0i8.i32(i8* %dest, i8* %src, i32 100, i32 4, i1 false)
will now read
call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %dest, i8* align 4 %src, i32 100, i1 false)
Downstream users may have to update their lit tests that check for
@llvm.memcpy/memmove/memset call/declaration patterns. The following extended sed script
may help with updating the majority of your tests, but it does not catch all possible
patterns so some manual checking and updating will be required.
s~declare void @llvm\.mem(set|cpy|move)\.p([^(]*)\((.*), i32, i1\)~declare void @llvm.mem\1.p\2(\3, i1)~g
s~call void @llvm\.memset\.p([^(]*)i8\(i8([^*]*)\* (.*), i8 (.*), i8 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i8(i8\2* \3, i8 \4, i8 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i16\(i8([^*]*)\* (.*), i8 (.*), i16 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i16(i8\2* \3, i8 \4, i16 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i32\(i8([^*]*)\* (.*), i8 (.*), i32 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i32(i8\2* \3, i8 \4, i32 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i64\(i8([^*]*)\* (.*), i8 (.*), i64 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i64(i8\2* \3, i8 \4, i64 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i128\(i8([^*]*)\* (.*), i8 (.*), i128 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i128(i8\2* \3, i8 \4, i128 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i8\(i8([^*]*)\* (.*), i8 (.*), i8 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i8(i8\2* align \6 \3, i8 \4, i8 \5, i1 \7)~g
s~call void @llvm\.memset\.p([^(]*)i16\(i8([^*]*)\* (.*), i8 (.*), i16 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i16(i8\2* align \6 \3, i8 \4, i16 \5, i1 \7)~g
s~call void @llvm\.memset\.p([^(]*)i32\(i8([^*]*)\* (.*), i8 (.*), i32 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i32(i8\2* align \6 \3, i8 \4, i32 \5, i1 \7)~g
s~call void @llvm\.memset\.p([^(]*)i64\(i8([^*]*)\* (.*), i8 (.*), i64 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i64(i8\2* align \6 \3, i8 \4, i64 \5, i1 \7)~g
s~call void @llvm\.memset\.p([^(]*)i128\(i8([^*]*)\* (.*), i8 (.*), i128 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i128(i8\2* align \6 \3, i8 \4, i128 \5, i1 \7)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i8\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i8 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i8(i8\3* \4, i8\5* \6, i8 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i16\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i16 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i16(i8\3* \4, i8\5* \6, i16 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i32\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i32 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i32(i8\3* \4, i8\5* \6, i32 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i64\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i64 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i64(i8\3* \4, i8\5* \6, i64 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i128\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i128 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i128(i8\3* \4, i8\5* \6, i128 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i8\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i8 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i8(i8\3* align \8 \4, i8\5* align \8 \6, i8 \7, i1 \9)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i16\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i16 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i16(i8\3* align \8 \4, i8\5* align \8 \6, i16 \7, i1 \9)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i32\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i32 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i32(i8\3* align \8 \4, i8\5* align \8 \6, i32 \7, i1 \9)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i64\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i64 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i64(i8\3* align \8 \4, i8\5* align \8 \6, i64 \7, i1 \9)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i128\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i128 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i128(i8\3* align \8 \4, i8\5* align \8 \6, i128 \7, i1 \9)~g
The remaining changes in the series will:
Step 2) Expand the IRBuilder API to allow creation of memcpy/memmove with differing
source and dest alignments.
Step 3) Update Clang to use the new IRBuilder API.
Step 4) Update Polly to use the new IRBuilder API.
Step 5) Update LLVM passes that create memcpy/memmove calls to use the new IRBuilder API,
and those that use use MemIntrinsicInst::[get|set]Alignment() to use
getDestAlignment() and getSourceAlignment() instead.
Step 6) Remove the single-alignment IRBuilder API for memcpy/memmove, and the
MemIntrinsicInst::[get|set]Alignment() methods.
Reviewers: pete, hfinkel, lhames, reames, bollu
Reviewed By: reames
Subscribers: niosHD, reames, jholewinski, qcolombet, jfb, sanjoy, arsenm, dschuff, dylanmckay, mehdi_amini, sdardis, nemanjai, david2050, nhaehnle, javed.absar, sbc100, jgravelle-google, eraman, aheejin, kbarton, JDevlieghere, asb, rbar, johnrusso, simoncook, jordy.potman.lists, apazos, sabuasal, llvm-commits
Differential Revision: https://reviews.llvm.org/D41675
llvm-svn: 322965
2018-01-20 01:13:12 +08:00
|
|
|
call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 8 %15, i8* align 8 %11, i64 16, i1 false) #3
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%Filename.i38 = getelementptr inbounds %"class.llvm::SMDiagnostic", %"class.llvm::SMDiagnostic"* %Err, i64 0, i32 2
|
2014-08-15 21:51:57 +08:00
|
|
|
call void @_ZNSs4swapERSs(%"class.std::basic_string"* %Filename.i38, %"class.std::basic_string"* dereferenceable(8) %Filename.i) #3
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%LineNo.i39 = getelementptr inbounds %"class.llvm::SMDiagnostic", %"class.llvm::SMDiagnostic"* %Err, i64 0, i32 3
|
2014-08-15 21:51:57 +08:00
|
|
|
%16 = bitcast i32* %LineNo.i39 to i8*
|
|
|
|
%17 = bitcast i32* %LineNo.i to i8*
|
Remove alignment argument from memcpy/memmove/memset in favour of alignment attributes (Step 1)
Summary:
This is a resurrection of work first proposed and discussed in Aug 2015:
http://lists.llvm.org/pipermail/llvm-dev/2015-August/089384.html
and initially landed (but then backed out) in Nov 2015:
http://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20151109/312083.html
The @llvm.memcpy/memmove/memset intrinsics currently have an explicit argument
which is required to be a constant integer. It represents the alignment of the
dest (and source), and so must be the minimum of the actual alignment of the
two.
This change is the first in a series that allows source and dest to each
have their own alignments by using the alignment attribute on their arguments.
In this change we:
1) Remove the alignment argument.
2) Add alignment attributes to the source & dest arguments. We, temporarily,
require that the alignments for source & dest be equal.
For example, code which used to read:
call void @llvm.memcpy.p0i8.p0i8.i32(i8* %dest, i8* %src, i32 100, i32 4, i1 false)
will now read
call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %dest, i8* align 4 %src, i32 100, i1 false)
Downstream users may have to update their lit tests that check for
@llvm.memcpy/memmove/memset call/declaration patterns. The following extended sed script
may help with updating the majority of your tests, but it does not catch all possible
patterns so some manual checking and updating will be required.
s~declare void @llvm\.mem(set|cpy|move)\.p([^(]*)\((.*), i32, i1\)~declare void @llvm.mem\1.p\2(\3, i1)~g
s~call void @llvm\.memset\.p([^(]*)i8\(i8([^*]*)\* (.*), i8 (.*), i8 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i8(i8\2* \3, i8 \4, i8 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i16\(i8([^*]*)\* (.*), i8 (.*), i16 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i16(i8\2* \3, i8 \4, i16 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i32\(i8([^*]*)\* (.*), i8 (.*), i32 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i32(i8\2* \3, i8 \4, i32 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i64\(i8([^*]*)\* (.*), i8 (.*), i64 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i64(i8\2* \3, i8 \4, i64 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i128\(i8([^*]*)\* (.*), i8 (.*), i128 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i128(i8\2* \3, i8 \4, i128 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i8\(i8([^*]*)\* (.*), i8 (.*), i8 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i8(i8\2* align \6 \3, i8 \4, i8 \5, i1 \7)~g
s~call void @llvm\.memset\.p([^(]*)i16\(i8([^*]*)\* (.*), i8 (.*), i16 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i16(i8\2* align \6 \3, i8 \4, i16 \5, i1 \7)~g
s~call void @llvm\.memset\.p([^(]*)i32\(i8([^*]*)\* (.*), i8 (.*), i32 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i32(i8\2* align \6 \3, i8 \4, i32 \5, i1 \7)~g
s~call void @llvm\.memset\.p([^(]*)i64\(i8([^*]*)\* (.*), i8 (.*), i64 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i64(i8\2* align \6 \3, i8 \4, i64 \5, i1 \7)~g
s~call void @llvm\.memset\.p([^(]*)i128\(i8([^*]*)\* (.*), i8 (.*), i128 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i128(i8\2* align \6 \3, i8 \4, i128 \5, i1 \7)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i8\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i8 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i8(i8\3* \4, i8\5* \6, i8 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i16\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i16 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i16(i8\3* \4, i8\5* \6, i16 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i32\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i32 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i32(i8\3* \4, i8\5* \6, i32 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i64\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i64 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i64(i8\3* \4, i8\5* \6, i64 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i128\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i128 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i128(i8\3* \4, i8\5* \6, i128 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i8\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i8 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i8(i8\3* align \8 \4, i8\5* align \8 \6, i8 \7, i1 \9)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i16\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i16 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i16(i8\3* align \8 \4, i8\5* align \8 \6, i16 \7, i1 \9)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i32\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i32 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i32(i8\3* align \8 \4, i8\5* align \8 \6, i32 \7, i1 \9)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i64\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i64 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i64(i8\3* align \8 \4, i8\5* align \8 \6, i64 \7, i1 \9)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i128\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i128 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i128(i8\3* align \8 \4, i8\5* align \8 \6, i128 \7, i1 \9)~g
The remaining changes in the series will:
Step 2) Expand the IRBuilder API to allow creation of memcpy/memmove with differing
source and dest alignments.
Step 3) Update Clang to use the new IRBuilder API.
Step 4) Update Polly to use the new IRBuilder API.
Step 5) Update LLVM passes that create memcpy/memmove calls to use the new IRBuilder API,
and those that use use MemIntrinsicInst::[get|set]Alignment() to use
getDestAlignment() and getSourceAlignment() instead.
Step 6) Remove the single-alignment IRBuilder API for memcpy/memmove, and the
MemIntrinsicInst::[get|set]Alignment() methods.
Reviewers: pete, hfinkel, lhames, reames, bollu
Reviewed By: reames
Subscribers: niosHD, reames, jholewinski, qcolombet, jfb, sanjoy, arsenm, dschuff, dylanmckay, mehdi_amini, sdardis, nemanjai, david2050, nhaehnle, javed.absar, sbc100, jgravelle-google, eraman, aheejin, kbarton, JDevlieghere, asb, rbar, johnrusso, simoncook, jordy.potman.lists, apazos, sabuasal, llvm-commits
Differential Revision: https://reviews.llvm.org/D41675
llvm-svn: 322965
2018-01-20 01:13:12 +08:00
|
|
|
call void @llvm.memcpy.p0i8.p0i8.i64(i8* align 4 %16, i8* align 4 %17, i64 12, i1 false) #3
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%Message.i40 = getelementptr inbounds %"class.llvm::SMDiagnostic", %"class.llvm::SMDiagnostic"* %Err, i64 0, i32 6
|
2014-08-15 21:51:57 +08:00
|
|
|
call void @_ZNSs4swapERSs(%"class.std::basic_string"* %Message.i40, %"class.std::basic_string"* dereferenceable(8) %Message.i) #3
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%LineContents.i = getelementptr inbounds %"class.llvm::SMDiagnostic", %"class.llvm::SMDiagnostic"* %Err, i64 0, i32 7
|
|
|
|
%LineContents7.i = getelementptr inbounds %"class.llvm::SMDiagnostic", %"class.llvm::SMDiagnostic"* %ref.tmp, i64 0, i32 7
|
2014-08-15 21:51:57 +08:00
|
|
|
call void @_ZNSs4swapERSs(%"class.std::basic_string"* %LineContents.i, %"class.std::basic_string"* dereferenceable(8) %LineContents7.i) #3
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%Ranges.i41 = getelementptr inbounds %"class.llvm::SMDiagnostic", %"class.llvm::SMDiagnostic"* %Err, i64 0, i32 8
|
|
|
|
%_M_start.i7.i.i.i = getelementptr inbounds %"class.std::vector.79", %"class.std::vector.79"* %Ranges.i41, i64 0, i32 0, i32 0, i32 0
|
2015-02-28 05:17:42 +08:00
|
|
|
%18 = load %"struct.std::pair"*, %"struct.std::pair"** %_M_start.i7.i.i.i, align 8, !tbaa !27
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%_M_finish.i9.i.i.i = getelementptr inbounds %"class.llvm::SMDiagnostic", %"class.llvm::SMDiagnostic"* %Err, i64 0, i32 8, i32 0, i32 0, i32 1
|
|
|
|
%_M_end_of_storage.i11.i.i.i = getelementptr inbounds %"class.llvm::SMDiagnostic", %"class.llvm::SMDiagnostic"* %Err, i64 0, i32 8, i32 0, i32 0, i32 2
|
|
|
|
%_M_start2.i.i.i.i = getelementptr inbounds %"class.llvm::SMDiagnostic", %"class.llvm::SMDiagnostic"* %ref.tmp, i64 0, i32 8, i32 0, i32 0, i32 0
|
2014-08-15 21:51:57 +08:00
|
|
|
%19 = bitcast %"class.std::vector.79"* %Ranges.i41 to i8*
|
Remove alignment argument from memcpy/memmove/memset in favour of alignment attributes (Step 1)
Summary:
This is a resurrection of work first proposed and discussed in Aug 2015:
http://lists.llvm.org/pipermail/llvm-dev/2015-August/089384.html
and initially landed (but then backed out) in Nov 2015:
http://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20151109/312083.html
The @llvm.memcpy/memmove/memset intrinsics currently have an explicit argument
which is required to be a constant integer. It represents the alignment of the
dest (and source), and so must be the minimum of the actual alignment of the
two.
This change is the first in a series that allows source and dest to each
have their own alignments by using the alignment attribute on their arguments.
In this change we:
1) Remove the alignment argument.
2) Add alignment attributes to the source & dest arguments. We, temporarily,
require that the alignments for source & dest be equal.
For example, code which used to read:
call void @llvm.memcpy.p0i8.p0i8.i32(i8* %dest, i8* %src, i32 100, i32 4, i1 false)
will now read
call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %dest, i8* align 4 %src, i32 100, i1 false)
Downstream users may have to update their lit tests that check for
@llvm.memcpy/memmove/memset call/declaration patterns. The following extended sed script
may help with updating the majority of your tests, but it does not catch all possible
patterns so some manual checking and updating will be required.
s~declare void @llvm\.mem(set|cpy|move)\.p([^(]*)\((.*), i32, i1\)~declare void @llvm.mem\1.p\2(\3, i1)~g
s~call void @llvm\.memset\.p([^(]*)i8\(i8([^*]*)\* (.*), i8 (.*), i8 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i8(i8\2* \3, i8 \4, i8 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i16\(i8([^*]*)\* (.*), i8 (.*), i16 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i16(i8\2* \3, i8 \4, i16 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i32\(i8([^*]*)\* (.*), i8 (.*), i32 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i32(i8\2* \3, i8 \4, i32 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i64\(i8([^*]*)\* (.*), i8 (.*), i64 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i64(i8\2* \3, i8 \4, i64 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i128\(i8([^*]*)\* (.*), i8 (.*), i128 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i128(i8\2* \3, i8 \4, i128 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i8\(i8([^*]*)\* (.*), i8 (.*), i8 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i8(i8\2* align \6 \3, i8 \4, i8 \5, i1 \7)~g
s~call void @llvm\.memset\.p([^(]*)i16\(i8([^*]*)\* (.*), i8 (.*), i16 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i16(i8\2* align \6 \3, i8 \4, i16 \5, i1 \7)~g
s~call void @llvm\.memset\.p([^(]*)i32\(i8([^*]*)\* (.*), i8 (.*), i32 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i32(i8\2* align \6 \3, i8 \4, i32 \5, i1 \7)~g
s~call void @llvm\.memset\.p([^(]*)i64\(i8([^*]*)\* (.*), i8 (.*), i64 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i64(i8\2* align \6 \3, i8 \4, i64 \5, i1 \7)~g
s~call void @llvm\.memset\.p([^(]*)i128\(i8([^*]*)\* (.*), i8 (.*), i128 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i128(i8\2* align \6 \3, i8 \4, i128 \5, i1 \7)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i8\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i8 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i8(i8\3* \4, i8\5* \6, i8 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i16\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i16 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i16(i8\3* \4, i8\5* \6, i16 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i32\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i32 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i32(i8\3* \4, i8\5* \6, i32 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i64\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i64 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i64(i8\3* \4, i8\5* \6, i64 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i128\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i128 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i128(i8\3* \4, i8\5* \6, i128 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i8\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i8 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i8(i8\3* align \8 \4, i8\5* align \8 \6, i8 \7, i1 \9)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i16\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i16 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i16(i8\3* align \8 \4, i8\5* align \8 \6, i16 \7, i1 \9)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i32\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i32 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i32(i8\3* align \8 \4, i8\5* align \8 \6, i32 \7, i1 \9)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i64\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i64 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i64(i8\3* align \8 \4, i8\5* align \8 \6, i64 \7, i1 \9)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i128\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i128 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i128(i8\3* align \8 \4, i8\5* align \8 \6, i128 \7, i1 \9)~g
The remaining changes in the series will:
Step 2) Expand the IRBuilder API to allow creation of memcpy/memmove with differing
source and dest alignments.
Step 3) Update Clang to use the new IRBuilder API.
Step 4) Update Polly to use the new IRBuilder API.
Step 5) Update LLVM passes that create memcpy/memmove calls to use the new IRBuilder API,
and those that use use MemIntrinsicInst::[get|set]Alignment() to use
getDestAlignment() and getSourceAlignment() instead.
Step 6) Remove the single-alignment IRBuilder API for memcpy/memmove, and the
MemIntrinsicInst::[get|set]Alignment() methods.
Reviewers: pete, hfinkel, lhames, reames, bollu
Reviewed By: reames
Subscribers: niosHD, reames, jholewinski, qcolombet, jfb, sanjoy, arsenm, dschuff, dylanmckay, mehdi_amini, sdardis, nemanjai, david2050, nhaehnle, javed.absar, sbc100, jgravelle-google, eraman, aheejin, kbarton, JDevlieghere, asb, rbar, johnrusso, simoncook, jordy.potman.lists, apazos, sabuasal, llvm-commits
Differential Revision: https://reviews.llvm.org/D41675
llvm-svn: 322965
2018-01-20 01:13:12 +08:00
|
|
|
call void @llvm.memset.p0i8.i64(i8* align 8 %19, i8 0, i64 16, i1 false) #3
|
2015-02-28 05:17:42 +08:00
|
|
|
%20 = load %"struct.std::pair"*, %"struct.std::pair"** %_M_start2.i.i.i.i, align 8, !tbaa !27
|
2014-08-15 21:51:57 +08:00
|
|
|
store %"struct.std::pair"* %20, %"struct.std::pair"** %_M_start.i7.i.i.i, align 8, !tbaa !27
|
|
|
|
store %"struct.std::pair"* null, %"struct.std::pair"** %_M_start2.i.i.i.i, align 8, !tbaa !27
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%_M_finish3.i.i.i.i = getelementptr inbounds %"class.llvm::SMDiagnostic", %"class.llvm::SMDiagnostic"* %ref.tmp, i64 0, i32 8, i32 0, i32 0, i32 1
|
2015-02-28 05:17:42 +08:00
|
|
|
%21 = load %"struct.std::pair"*, %"struct.std::pair"** %_M_finish3.i.i.i.i, align 8, !tbaa !27
|
2014-08-15 21:51:57 +08:00
|
|
|
store %"struct.std::pair"* %21, %"struct.std::pair"** %_M_finish.i9.i.i.i, align 8, !tbaa !27
|
|
|
|
store %"struct.std::pair"* null, %"struct.std::pair"** %_M_finish3.i.i.i.i, align 8, !tbaa !27
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%_M_end_of_storage4.i.i.i.i = getelementptr inbounds %"class.llvm::SMDiagnostic", %"class.llvm::SMDiagnostic"* %ref.tmp, i64 0, i32 8, i32 0, i32 0, i32 2
|
2015-02-28 05:17:42 +08:00
|
|
|
%22 = load %"struct.std::pair"*, %"struct.std::pair"** %_M_end_of_storage4.i.i.i.i, align 8, !tbaa !27
|
2014-08-15 21:51:57 +08:00
|
|
|
store %"struct.std::pair"* %22, %"struct.std::pair"** %_M_end_of_storage.i11.i.i.i, align 8, !tbaa !27
|
|
|
|
store %"struct.std::pair"* null, %"struct.std::pair"** %_M_end_of_storage4.i.i.i.i, align 8, !tbaa !27
|
|
|
|
%tobool.i.i.i.i.i.i = icmp eq %"struct.std::pair"* %18, null
|
|
|
|
br i1 %tobool.i.i.i.i.i.i, label %_ZN4llvm12SMDiagnosticaSEOS0_.exit, label %if.then.i.i.i.i.i.i
|
|
|
|
|
|
|
|
if.then.i.i.i.i.i.i: ; preds = %_ZN4llvm12SMDiagnosticC2ENS_9StringRefENS_9SourceMgr8DiagKindES1_.exit
|
|
|
|
%23 = bitcast %"struct.std::pair"* %18 to i8*
|
|
|
|
call void @_ZdlPv(i8* %23) #3
|
|
|
|
br label %_ZN4llvm12SMDiagnosticaSEOS0_.exit
|
|
|
|
|
|
|
|
_ZN4llvm12SMDiagnosticaSEOS0_.exit: ; preds = %_ZN4llvm12SMDiagnosticC2ENS_9StringRefENS_9SourceMgr8DiagKindES1_.exit, %if.then.i.i.i.i.i.i
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%24 = getelementptr inbounds %"class.llvm::SMDiagnostic", %"class.llvm::SMDiagnostic"* %Err, i64 0, i32 9, i32 0
|
|
|
|
%25 = getelementptr inbounds %"class.llvm::SMDiagnostic", %"class.llvm::SMDiagnostic"* %ref.tmp, i64 0, i32 9, i32 0
|
2014-08-15 21:51:57 +08:00
|
|
|
%call2.i.i42 = call dereferenceable(48) %"class.llvm::SmallVectorImpl.85"* @_ZN4llvm15SmallVectorImplINS_7SMFixItEEaSEOS2_(%"class.llvm::SmallVectorImpl.85"* %24, %"class.llvm::SmallVectorImpl.85"* dereferenceable(48) %25) #3
|
|
|
|
call void @_ZN4llvm12SMDiagnosticD2Ev(%"class.llvm::SMDiagnostic"* %ref.tmp) #3
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%26 = getelementptr inbounds %"class.std::allocator", %"class.std::allocator"* %ref.tmp.i.i, i64 0, i32 0
|
2017-04-11 04:18:21 +08:00
|
|
|
call void @llvm.lifetime.start.p0i8(i64 1, i8* %26) #3
|
2014-08-15 21:51:57 +08:00
|
|
|
%27 = bitcast i8* %arrayidx.i.i.i36 to %"struct.std::basic_string<char, std::char_traits<char>, std::allocator<char> >::_Rep"*
|
|
|
|
%cmp.i.i.i = icmp eq i8* %arrayidx.i.i.i36, bitcast ([0 x i64]* @_ZNSs4_Rep20_S_empty_rep_storageE to i8*)
|
|
|
|
br i1 %cmp.i.i.i, label %_ZNSsD1Ev.exit, label %if.then.i.i.i45, !prof !28
|
|
|
|
|
|
|
|
if.then.i.i.i45: ; preds = %_ZN4llvm12SMDiagnosticaSEOS0_.exit
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%_M_refcount.i.i.i = getelementptr inbounds i8, i8* %8, i64 -8
|
2014-08-15 21:51:57 +08:00
|
|
|
%28 = bitcast i8* %_M_refcount.i.i.i to i32*
|
|
|
|
br i1 icmp ne (i8* bitcast (i32 (i32*, void (i8*)*)* @__pthread_key_create to i8*), i8* null), label %if.then.i.i.i.i, label %if.else.i.i.i.i
|
|
|
|
|
|
|
|
if.then.i.i.i.i: ; preds = %if.then.i.i.i45
|
|
|
|
%.atomicdst.i.i.i.i.i.0..atomicdst.i.i.i.i.0..atomicdst.i.i.i.0..atomicdst.i.i.0..atomicdst.i.0..sroa_cast = bitcast i32* %.atomicdst.i.i.i.i.i to i8*
|
2017-04-11 04:18:21 +08:00
|
|
|
call void @llvm.lifetime.start.p0i8(i64 4, i8* %.atomicdst.i.i.i.i.i.0..atomicdst.i.i.i.i.0..atomicdst.i.i.i.0..atomicdst.i.i.0..atomicdst.i.0..sroa_cast)
|
2014-08-15 21:51:57 +08:00
|
|
|
%29 = atomicrmw volatile add i32* %28, i32 -1 acq_rel
|
|
|
|
store i32 %29, i32* %.atomicdst.i.i.i.i.i, align 4
|
2015-02-28 05:17:42 +08:00
|
|
|
%.atomicdst.i.i.i.i.i.0..atomicdst.i.i.i.i.0..atomicdst.i.i.i.0..atomicdst.i.i.0..atomicdst.i.0..atomicdst.0..atomicdst.0..i.i.i.i.i = load volatile i32, i32* %.atomicdst.i.i.i.i.i, align 4
|
2017-04-11 04:18:21 +08:00
|
|
|
call void @llvm.lifetime.end.p0i8(i64 4, i8* %.atomicdst.i.i.i.i.i.0..atomicdst.i.i.i.i.0..atomicdst.i.i.i.0..atomicdst.i.i.0..atomicdst.i.0..sroa_cast)
|
2014-08-15 21:51:57 +08:00
|
|
|
br label %_ZN9__gnu_cxxL27__exchange_and_add_dispatchEPii.exit.i.i.i
|
|
|
|
|
|
|
|
if.else.i.i.i.i: ; preds = %if.then.i.i.i45
|
2015-02-28 05:17:42 +08:00
|
|
|
%30 = load i32, i32* %28, align 4, !tbaa !29
|
2014-08-15 21:51:57 +08:00
|
|
|
%add.i.i.i.i.i = add nsw i32 %30, -1
|
|
|
|
store i32 %add.i.i.i.i.i, i32* %28, align 4, !tbaa !29
|
|
|
|
br label %_ZN9__gnu_cxxL27__exchange_and_add_dispatchEPii.exit.i.i.i
|
|
|
|
|
|
|
|
_ZN9__gnu_cxxL27__exchange_and_add_dispatchEPii.exit.i.i.i: ; preds = %if.else.i.i.i.i, %if.then.i.i.i.i
|
|
|
|
%retval.0.i.i.i.i = phi i32 [ %.atomicdst.i.i.i.i.i.0..atomicdst.i.i.i.i.0..atomicdst.i.i.i.0..atomicdst.i.i.0..atomicdst.i.0..atomicdst.0..atomicdst.0..i.i.i.i.i, %if.then.i.i.i.i ], [ %30, %if.else.i.i.i.i ]
|
|
|
|
%cmp3.i.i.i = icmp slt i32 %retval.0.i.i.i.i, 1
|
|
|
|
br i1 %cmp3.i.i.i, label %if.then4.i.i.i, label %_ZNSsD1Ev.exit
|
|
|
|
|
|
|
|
if.then4.i.i.i: ; preds = %_ZN9__gnu_cxxL27__exchange_and_add_dispatchEPii.exit.i.i.i
|
|
|
|
call void @_ZNSs4_Rep10_M_destroyERKSaIcE(%"struct.std::basic_string<char, std::char_traits<char>, std::allocator<char> >::_Rep"* %27, %"class.std::allocator"* dereferenceable(1) %ref.tmp.i.i) #3
|
|
|
|
br label %_ZNSsD1Ev.exit
|
|
|
|
|
|
|
|
_ZNSsD1Ev.exit: ; preds = %_ZN4llvm12SMDiagnosticaSEOS0_.exit, %_ZN9__gnu_cxxL27__exchange_and_add_dispatchEPii.exit.i.i.i, %if.then4.i.i.i
|
2017-04-11 04:18:21 +08:00
|
|
|
call void @llvm.lifetime.end.p0i8(i64 1, i8* %26) #3
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%31 = getelementptr inbounds %"class.std::allocator", %"class.std::allocator"* %ref.tmp.i.i47, i64 0, i32 0
|
2017-04-11 04:18:21 +08:00
|
|
|
call void @llvm.lifetime.start.p0i8(i64 1, i8* %31) #3
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%_M_p.i.i.i.i48 = getelementptr inbounds %"class.std::basic_string", %"class.std::basic_string"* %ref.tmp5, i64 0, i32 0, i32 0
|
2015-02-28 05:17:42 +08:00
|
|
|
%32 = load i8*, i8** %_M_p.i.i.i.i48, align 8, !tbaa !1
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%arrayidx.i.i.i49 = getelementptr inbounds i8, i8* %32, i64 -24
|
2014-08-15 21:51:57 +08:00
|
|
|
%33 = bitcast i8* %arrayidx.i.i.i49 to %"struct.std::basic_string<char, std::char_traits<char>, std::allocator<char> >::_Rep"*
|
|
|
|
%cmp.i.i.i50 = icmp eq i8* %arrayidx.i.i.i49, bitcast ([0 x i64]* @_ZNSs4_Rep20_S_empty_rep_storageE to i8*)
|
|
|
|
br i1 %cmp.i.i.i50, label %_ZNSsD1Ev.exit62, label %if.then.i.i.i52, !prof !28
|
|
|
|
|
|
|
|
if.then.i.i.i52: ; preds = %_ZNSsD1Ev.exit
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%_M_refcount.i.i.i51 = getelementptr inbounds i8, i8* %32, i64 -8
|
2014-08-15 21:51:57 +08:00
|
|
|
%34 = bitcast i8* %_M_refcount.i.i.i51 to i32*
|
|
|
|
br i1 icmp ne (i8* bitcast (i32 (i32*, void (i8*)*)* @__pthread_key_create to i8*), i8* null), label %if.then.i.i.i.i55, label %if.else.i.i.i.i57
|
|
|
|
|
|
|
|
if.then.i.i.i.i55: ; preds = %if.then.i.i.i52
|
|
|
|
%.atomicdst.i.i.i.i.i46.0..atomicdst.i.i.i.i.0..atomicdst.i.i.i.0..atomicdst.i.i.0..atomicdst.i.0..sroa_cast = bitcast i32* %.atomicdst.i.i.i.i.i46 to i8*
|
2017-04-11 04:18:21 +08:00
|
|
|
call void @llvm.lifetime.start.p0i8(i64 4, i8* %.atomicdst.i.i.i.i.i46.0..atomicdst.i.i.i.i.0..atomicdst.i.i.i.0..atomicdst.i.i.0..atomicdst.i.0..sroa_cast)
|
2014-08-15 21:51:57 +08:00
|
|
|
%35 = atomicrmw volatile add i32* %34, i32 -1 acq_rel
|
|
|
|
store i32 %35, i32* %.atomicdst.i.i.i.i.i46, align 4
|
2015-02-28 05:17:42 +08:00
|
|
|
%.atomicdst.i.i.i.i.i46.0..atomicdst.i.i.i.i.0..atomicdst.i.i.i.0..atomicdst.i.i.0..atomicdst.i.0..atomicdst.0..atomicdst.0..i.i.i.i.i54 = load volatile i32, i32* %.atomicdst.i.i.i.i.i46, align 4
|
2017-04-11 04:18:21 +08:00
|
|
|
call void @llvm.lifetime.end.p0i8(i64 4, i8* %.atomicdst.i.i.i.i.i46.0..atomicdst.i.i.i.i.0..atomicdst.i.i.i.0..atomicdst.i.i.0..atomicdst.i.0..sroa_cast)
|
2014-08-15 21:51:57 +08:00
|
|
|
br label %_ZN9__gnu_cxxL27__exchange_and_add_dispatchEPii.exit.i.i.i60
|
|
|
|
|
|
|
|
if.else.i.i.i.i57: ; preds = %if.then.i.i.i52
|
2015-02-28 05:17:42 +08:00
|
|
|
%36 = load i32, i32* %34, align 4, !tbaa !29
|
2014-08-15 21:51:57 +08:00
|
|
|
%add.i.i.i.i.i56 = add nsw i32 %36, -1
|
|
|
|
store i32 %add.i.i.i.i.i56, i32* %34, align 4, !tbaa !29
|
|
|
|
br label %_ZN9__gnu_cxxL27__exchange_and_add_dispatchEPii.exit.i.i.i60
|
|
|
|
|
|
|
|
_ZN9__gnu_cxxL27__exchange_and_add_dispatchEPii.exit.i.i.i60: ; preds = %if.else.i.i.i.i57, %if.then.i.i.i.i55
|
|
|
|
%retval.0.i.i.i.i58 = phi i32 [ %.atomicdst.i.i.i.i.i46.0..atomicdst.i.i.i.i.0..atomicdst.i.i.i.0..atomicdst.i.i.0..atomicdst.i.0..atomicdst.0..atomicdst.0..i.i.i.i.i54, %if.then.i.i.i.i55 ], [ %36, %if.else.i.i.i.i57 ]
|
|
|
|
%cmp3.i.i.i59 = icmp slt i32 %retval.0.i.i.i.i58, 1
|
|
|
|
br i1 %cmp3.i.i.i59, label %if.then4.i.i.i61, label %_ZNSsD1Ev.exit62
|
|
|
|
|
|
|
|
if.then4.i.i.i61: ; preds = %_ZN9__gnu_cxxL27__exchange_and_add_dispatchEPii.exit.i.i.i60
|
|
|
|
call void @_ZNSs4_Rep10_M_destroyERKSaIcE(%"struct.std::basic_string<char, std::char_traits<char>, std::allocator<char> >::_Rep"* %33, %"class.std::allocator"* dereferenceable(1) %ref.tmp.i.i47) #3
|
|
|
|
br label %_ZNSsD1Ev.exit62
|
|
|
|
|
|
|
|
_ZNSsD1Ev.exit62: ; preds = %_ZNSsD1Ev.exit, %_ZN9__gnu_cxxL27__exchange_and_add_dispatchEPii.exit.i.i.i60, %if.then4.i.i.i61
|
2017-04-11 04:18:21 +08:00
|
|
|
call void @llvm.lifetime.end.p0i8(i64 1, i8* %31) #3
|
2014-08-15 21:51:57 +08:00
|
|
|
br label %cleanup
|
|
|
|
|
|
|
|
cond.false.i.i: ; preds = %_ZNK4llvm7ErrorOrISt10unique_ptrINS_12MemoryBufferESt14default_deleteIS2_EEE8getErrorEv.exit
|
2015-03-14 02:20:45 +08:00
|
|
|
call void @__assert_fail(i8* getelementptr inbounds ([54 x i8], [54 x i8]* @.str1, i64 0, i64 0), i8* getelementptr inbounds ([61 x i8], [61 x i8]* @.str2, i64 0, i64 0), i32 zeroext 242, i8* getelementptr inbounds ([206 x i8], [206 x i8]* @__PRETTY_FUNCTION__._ZN4llvm7ErrorOrISt10unique_ptrINS_12MemoryBufferESt14default_deleteIS2_EEE10getStorageEv, i64 0, i64 0)) #7
|
2014-08-15 21:51:57 +08:00
|
|
|
unreachable
|
|
|
|
|
|
|
|
_ZN4llvm7ErrorOrISt10unique_ptrINS_12MemoryBufferESt14default_deleteIS2_EEE3getEv.exit: ; preds = %entry
|
|
|
|
%_M_head_impl.i.i.i.i.i = bitcast %"class.llvm::ErrorOr"* %FileOrErr to %"class.llvm::MemoryBuffer"**
|
2015-02-28 05:17:42 +08:00
|
|
|
%37 = load %"class.llvm::MemoryBuffer"*, %"class.llvm::MemoryBuffer"** %_M_head_impl.i.i.i.i.i, align 8, !tbaa !27
|
2014-08-15 21:51:57 +08:00
|
|
|
%call9 = call %"class.llvm::Module"* @_ZN4llvm7ParseIREPNS_12MemoryBufferERNS_12SMDiagnosticERNS_11LLVMContextE(%"class.llvm::MemoryBuffer"* %37, %"class.llvm::SMDiagnostic"* dereferenceable(200) %Err, %"class.llvm::LLVMContext"* dereferenceable(8) %Context)
|
|
|
|
br label %cleanup
|
|
|
|
|
|
|
|
cleanup: ; preds = %_ZN4llvm7ErrorOrISt10unique_ptrINS_12MemoryBufferESt14default_deleteIS2_EEE3getEv.exit, %_ZNSsD1Ev.exit62
|
|
|
|
%retval.0 = phi %"class.llvm::Module"* [ null, %_ZNSsD1Ev.exit62 ], [ %call9, %_ZN4llvm7ErrorOrISt10unique_ptrINS_12MemoryBufferESt14default_deleteIS2_EEE3getEv.exit ]
|
2015-02-28 05:17:42 +08:00
|
|
|
%bf.load.i = load i8, i8* %HasError.i24, align 8
|
2014-08-15 21:51:57 +08:00
|
|
|
%38 = and i8 %bf.load.i, 1
|
|
|
|
%bf.cast.i = icmp eq i8 %38, 0
|
|
|
|
br i1 %bf.cast.i, label %_ZN4llvm7ErrorOrISt10unique_ptrINS_12MemoryBufferESt14default_deleteIS2_EEE10getStorageEv.exit.i, label %_ZN4llvm7ErrorOrISt10unique_ptrINS_12MemoryBufferESt14default_deleteIS2_EEED2Ev.exit
|
|
|
|
|
|
|
|
_ZN4llvm7ErrorOrISt10unique_ptrINS_12MemoryBufferESt14default_deleteIS2_EEE10getStorageEv.exit.i: ; preds = %cleanup
|
|
|
|
%_M_head_impl.i.i.i.i.i.i = bitcast %"class.llvm::ErrorOr"* %FileOrErr to %"class.llvm::MemoryBuffer"**
|
2015-02-28 05:17:42 +08:00
|
|
|
%39 = load %"class.llvm::MemoryBuffer"*, %"class.llvm::MemoryBuffer"** %_M_head_impl.i.i.i.i.i.i, align 8, !tbaa !27
|
2014-08-15 21:51:57 +08:00
|
|
|
%cmp.i.i = icmp eq %"class.llvm::MemoryBuffer"* %39, null
|
|
|
|
br i1 %cmp.i.i, label %_ZNSt10unique_ptrIN4llvm12MemoryBufferESt14default_deleteIS1_EED2Ev.exit.i, label %_ZNKSt14default_deleteIN4llvm12MemoryBufferEEclEPS1_.exit.i.i
|
|
|
|
|
|
|
|
_ZNKSt14default_deleteIN4llvm12MemoryBufferEEclEPS1_.exit.i.i: ; preds = %_ZN4llvm7ErrorOrISt10unique_ptrINS_12MemoryBufferESt14default_deleteIS2_EEE10getStorageEv.exit.i
|
|
|
|
%40 = bitcast %"class.llvm::MemoryBuffer"* %39 to void (%"class.llvm::MemoryBuffer"*)***
|
2015-02-28 05:17:42 +08:00
|
|
|
%vtable.i.i.i = load void (%"class.llvm::MemoryBuffer"*)**, void (%"class.llvm::MemoryBuffer"*)*** %40, align 8, !tbaa !11
|
[opaque pointer type] Add textual IR support for explicit type parameter to getelementptr instruction
One of several parallel first steps to remove the target type of pointers,
replacing them with a single opaque pointer type.
This adds an explicit type parameter to the gep instruction so that when the
first parameter becomes an opaque pointer type, the type to gep through is
still available to the instructions.
* This doesn't modify gep operators, only instructions (operators will be
handled separately)
* Textual IR changes only. Bitcode (including upgrade) and changing the
in-memory representation will be in separate changes.
* geps of vectors are transformed as:
getelementptr <4 x float*> %x, ...
->getelementptr float, <4 x float*> %x, ...
Then, once the opaque pointer type is introduced, this will ultimately look
like:
getelementptr float, <4 x ptr> %x
with the unambiguous interpretation that it is a vector of pointers to float.
* address spaces remain on the pointer, not the type:
getelementptr float addrspace(1)* %x
->getelementptr float, float addrspace(1)* %x
Then, eventually:
getelementptr float, ptr addrspace(1) %x
Importantly, the massive amount of test case churn has been automated by
same crappy python code. I had to manually update a few test cases that
wouldn't fit the script's model (r228970,r229196,r229197,r229198). The
python script just massages stdin and writes the result to stdout, I
then wrapped that in a shell script to handle replacing files, then
using the usual find+xargs to migrate all the files.
update.py:
import fileinput
import sys
import re
ibrep = re.compile(r"(^.*?[^%\w]getelementptr inbounds )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
normrep = re.compile( r"(^.*?[^%\w]getelementptr )(((?:<\d* x )?)(.*?)(| addrspace\(\d\)) *\*(|>)(?:$| *(?:%|@|null|undef|blockaddress|getelementptr|addrspacecast|bitcast|inttoptr|\[\[[a-zA-Z]|\{\{).*$))")
def conv(match, line):
if not match:
return line
line = match.groups()[0]
if len(match.groups()[5]) == 0:
line += match.groups()[2]
line += match.groups()[3]
line += ", "
line += match.groups()[1]
line += "\n"
return line
for line in sys.stdin:
if line.find("getelementptr ") == line.find("getelementptr inbounds"):
if line.find("getelementptr inbounds") != line.find("getelementptr inbounds ("):
line = conv(re.match(ibrep, line), line)
elif line.find("getelementptr ") != line.find("getelementptr ("):
line = conv(re.match(normrep, line), line)
sys.stdout.write(line)
apply.sh:
for name in "$@"
do
python3 `dirname "$0"`/update.py < "$name" > "$name.tmp" && mv "$name.tmp" "$name"
rm -f "$name.tmp"
done
The actual commands:
From llvm/src:
find test/ -name *.ll | xargs ./apply.sh
From llvm/src/tools/clang:
find test/ -name *.mm -o -name *.m -o -name *.cpp -o -name *.c | xargs -I '{}' ../../apply.sh "{}"
From llvm/src/tools/polly:
find test/ -name *.ll | xargs ./apply.sh
After that, check-all (with llvm, clang, clang-tools-extra, lld,
compiler-rt, and polly all checked out).
The extra 'rm' in the apply.sh script is due to a few files in clang's test
suite using interesting unicode stuff that my python script was throwing
exceptions on. None of those files needed to be migrated, so it seemed
sufficient to ignore those cases.
Reviewers: rafael, dexonsmith, grosser
Differential Revision: http://reviews.llvm.org/D7636
llvm-svn: 230786
2015-02-28 03:29:02 +08:00
|
|
|
%vfn.i.i.i = getelementptr inbounds void (%"class.llvm::MemoryBuffer"*)*, void (%"class.llvm::MemoryBuffer"*)** %vtable.i.i.i, i64 1
|
2015-02-28 05:17:42 +08:00
|
|
|
%41 = load void (%"class.llvm::MemoryBuffer"*)*, void (%"class.llvm::MemoryBuffer"*)** %vfn.i.i.i, align 8
|
2014-08-15 21:51:57 +08:00
|
|
|
call void %41(%"class.llvm::MemoryBuffer"* %39) #3
|
|
|
|
br label %_ZNSt10unique_ptrIN4llvm12MemoryBufferESt14default_deleteIS1_EED2Ev.exit.i
|
|
|
|
|
|
|
|
_ZNSt10unique_ptrIN4llvm12MemoryBufferESt14default_deleteIS1_EED2Ev.exit.i: ; preds = %_ZNKSt14default_deleteIN4llvm12MemoryBufferEEclEPS1_.exit.i.i, %_ZN4llvm7ErrorOrISt10unique_ptrINS_12MemoryBufferESt14default_deleteIS2_EEE10getStorageEv.exit.i
|
|
|
|
store %"class.llvm::MemoryBuffer"* null, %"class.llvm::MemoryBuffer"** %_M_head_impl.i.i.i.i.i.i, align 8, !tbaa !27
|
|
|
|
br label %_ZN4llvm7ErrorOrISt10unique_ptrINS_12MemoryBufferESt14default_deleteIS2_EEED2Ev.exit
|
|
|
|
|
|
|
|
_ZN4llvm7ErrorOrISt10unique_ptrINS_12MemoryBufferESt14default_deleteIS2_EEED2Ev.exit: ; preds = %cleanup, %_ZNSt10unique_ptrIN4llvm12MemoryBufferESt14default_deleteIS1_EED2Ev.exit.i
|
|
|
|
ret %"class.llvm::Module"* %retval.0
|
|
|
|
}
|
|
|
|
|
|
|
|
; Function Attrs: nounwind
|
2017-04-11 04:18:21 +08:00
|
|
|
declare void @llvm.lifetime.start.p0i8(i64, i8* nocapture) #3
|
2014-08-15 21:51:57 +08:00
|
|
|
|
|
|
|
; Function Attrs: nounwind
|
2017-04-11 04:18:21 +08:00
|
|
|
declare void @llvm.lifetime.end.p0i8(i64, i8* nocapture) #3
|
2014-08-15 21:51:57 +08:00
|
|
|
|
|
|
|
; Function Attrs: noreturn nounwind
|
|
|
|
declare void @__assert_fail(i8*, i8*, i32 zeroext, i8*) #4
|
|
|
|
|
|
|
|
declare dereferenceable(8) %"class.std::basic_string"* @_ZNSs6insertEmPKcm(%"class.std::basic_string"*, i64, i8*, i64) #1
|
|
|
|
|
|
|
|
; Function Attrs: nounwind
|
Remove alignment argument from memcpy/memmove/memset in favour of alignment attributes (Step 1)
Summary:
This is a resurrection of work first proposed and discussed in Aug 2015:
http://lists.llvm.org/pipermail/llvm-dev/2015-August/089384.html
and initially landed (but then backed out) in Nov 2015:
http://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20151109/312083.html
The @llvm.memcpy/memmove/memset intrinsics currently have an explicit argument
which is required to be a constant integer. It represents the alignment of the
dest (and source), and so must be the minimum of the actual alignment of the
two.
This change is the first in a series that allows source and dest to each
have their own alignments by using the alignment attribute on their arguments.
In this change we:
1) Remove the alignment argument.
2) Add alignment attributes to the source & dest arguments. We, temporarily,
require that the alignments for source & dest be equal.
For example, code which used to read:
call void @llvm.memcpy.p0i8.p0i8.i32(i8* %dest, i8* %src, i32 100, i32 4, i1 false)
will now read
call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %dest, i8* align 4 %src, i32 100, i1 false)
Downstream users may have to update their lit tests that check for
@llvm.memcpy/memmove/memset call/declaration patterns. The following extended sed script
may help with updating the majority of your tests, but it does not catch all possible
patterns so some manual checking and updating will be required.
s~declare void @llvm\.mem(set|cpy|move)\.p([^(]*)\((.*), i32, i1\)~declare void @llvm.mem\1.p\2(\3, i1)~g
s~call void @llvm\.memset\.p([^(]*)i8\(i8([^*]*)\* (.*), i8 (.*), i8 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i8(i8\2* \3, i8 \4, i8 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i16\(i8([^*]*)\* (.*), i8 (.*), i16 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i16(i8\2* \3, i8 \4, i16 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i32\(i8([^*]*)\* (.*), i8 (.*), i32 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i32(i8\2* \3, i8 \4, i32 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i64\(i8([^*]*)\* (.*), i8 (.*), i64 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i64(i8\2* \3, i8 \4, i64 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i128\(i8([^*]*)\* (.*), i8 (.*), i128 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i128(i8\2* \3, i8 \4, i128 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i8\(i8([^*]*)\* (.*), i8 (.*), i8 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i8(i8\2* align \6 \3, i8 \4, i8 \5, i1 \7)~g
s~call void @llvm\.memset\.p([^(]*)i16\(i8([^*]*)\* (.*), i8 (.*), i16 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i16(i8\2* align \6 \3, i8 \4, i16 \5, i1 \7)~g
s~call void @llvm\.memset\.p([^(]*)i32\(i8([^*]*)\* (.*), i8 (.*), i32 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i32(i8\2* align \6 \3, i8 \4, i32 \5, i1 \7)~g
s~call void @llvm\.memset\.p([^(]*)i64\(i8([^*]*)\* (.*), i8 (.*), i64 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i64(i8\2* align \6 \3, i8 \4, i64 \5, i1 \7)~g
s~call void @llvm\.memset\.p([^(]*)i128\(i8([^*]*)\* (.*), i8 (.*), i128 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i128(i8\2* align \6 \3, i8 \4, i128 \5, i1 \7)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i8\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i8 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i8(i8\3* \4, i8\5* \6, i8 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i16\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i16 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i16(i8\3* \4, i8\5* \6, i16 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i32\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i32 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i32(i8\3* \4, i8\5* \6, i32 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i64\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i64 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i64(i8\3* \4, i8\5* \6, i64 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i128\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i128 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i128(i8\3* \4, i8\5* \6, i128 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i8\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i8 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i8(i8\3* align \8 \4, i8\5* align \8 \6, i8 \7, i1 \9)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i16\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i16 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i16(i8\3* align \8 \4, i8\5* align \8 \6, i16 \7, i1 \9)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i32\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i32 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i32(i8\3* align \8 \4, i8\5* align \8 \6, i32 \7, i1 \9)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i64\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i64 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i64(i8\3* align \8 \4, i8\5* align \8 \6, i64 \7, i1 \9)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i128\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i128 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i128(i8\3* align \8 \4, i8\5* align \8 \6, i128 \7, i1 \9)~g
The remaining changes in the series will:
Step 2) Expand the IRBuilder API to allow creation of memcpy/memmove with differing
source and dest alignments.
Step 3) Update Clang to use the new IRBuilder API.
Step 4) Update Polly to use the new IRBuilder API.
Step 5) Update LLVM passes that create memcpy/memmove calls to use the new IRBuilder API,
and those that use use MemIntrinsicInst::[get|set]Alignment() to use
getDestAlignment() and getSourceAlignment() instead.
Step 6) Remove the single-alignment IRBuilder API for memcpy/memmove, and the
MemIntrinsicInst::[get|set]Alignment() methods.
Reviewers: pete, hfinkel, lhames, reames, bollu
Reviewed By: reames
Subscribers: niosHD, reames, jholewinski, qcolombet, jfb, sanjoy, arsenm, dschuff, dylanmckay, mehdi_amini, sdardis, nemanjai, david2050, nhaehnle, javed.absar, sbc100, jgravelle-google, eraman, aheejin, kbarton, JDevlieghere, asb, rbar, johnrusso, simoncook, jordy.potman.lists, apazos, sabuasal, llvm-commits
Differential Revision: https://reviews.llvm.org/D41675
llvm-svn: 322965
2018-01-20 01:13:12 +08:00
|
|
|
declare void @llvm.memcpy.p0i8.p0i8.i64(i8* nocapture, i8* nocapture readonly, i64, i1) #3
|
2014-08-15 21:51:57 +08:00
|
|
|
|
|
|
|
; Function Attrs: nounwind
|
|
|
|
declare void @_ZNSs4_Rep10_M_destroyERKSaIcE(%"struct.std::basic_string<char, std::char_traits<char>, std::allocator<char> >::_Rep"*, %"class.std::allocator"* dereferenceable(1)) #0
|
|
|
|
|
|
|
|
; Function Attrs: nounwind
|
|
|
|
declare extern_weak signext i32 @__pthread_key_create(i32*, void (i8*)*) #0
|
|
|
|
|
|
|
|
; Function Attrs: nobuiltin nounwind
|
|
|
|
declare void @_ZdlPv(i8*) #6
|
|
|
|
|
|
|
|
declare void @_ZNSsC1EPKcmRKSaIcE(%"class.std::basic_string"*, i8*, i64, %"class.std::allocator"* dereferenceable(1)) #1
|
|
|
|
|
|
|
|
declare hidden void @_ZN4llvm12SMDiagnosticD2Ev(%"class.llvm::SMDiagnostic"* readonly %this) unnamed_addr #2 align 2
|
|
|
|
|
|
|
|
declare dereferenceable(48) %"class.llvm::SmallVectorImpl.85"* @_ZN4llvm15SmallVectorImplINS_7SMFixItEEaSEOS2_(%"class.llvm::SmallVectorImpl.85"* %this, %"class.llvm::SmallVectorImpl.85"* dereferenceable(48) %RHS) #0 align 2
|
|
|
|
|
|
|
|
declare %"class.llvm::Module"* @_ZN4llvm7ParseIREPNS_12MemoryBufferERNS_12SMDiagnosticERNS_11LLVMContextE(%"class.llvm::MemoryBuffer"* %Buffer, %"class.llvm::SMDiagnostic"* dereferenceable(200) %Err, %"class.llvm::LLVMContext"* dereferenceable(8) %Context) #0
|
|
|
|
|
|
|
|
declare void @_ZNSs4swapERSs(%"class.std::basic_string"*, %"class.std::basic_string"* dereferenceable(8)) #1
|
|
|
|
|
|
|
|
; Function Attrs: nounwind
|
Remove alignment argument from memcpy/memmove/memset in favour of alignment attributes (Step 1)
Summary:
This is a resurrection of work first proposed and discussed in Aug 2015:
http://lists.llvm.org/pipermail/llvm-dev/2015-August/089384.html
and initially landed (but then backed out) in Nov 2015:
http://lists.llvm.org/pipermail/llvm-commits/Week-of-Mon-20151109/312083.html
The @llvm.memcpy/memmove/memset intrinsics currently have an explicit argument
which is required to be a constant integer. It represents the alignment of the
dest (and source), and so must be the minimum of the actual alignment of the
two.
This change is the first in a series that allows source and dest to each
have their own alignments by using the alignment attribute on their arguments.
In this change we:
1) Remove the alignment argument.
2) Add alignment attributes to the source & dest arguments. We, temporarily,
require that the alignments for source & dest be equal.
For example, code which used to read:
call void @llvm.memcpy.p0i8.p0i8.i32(i8* %dest, i8* %src, i32 100, i32 4, i1 false)
will now read
call void @llvm.memcpy.p0i8.p0i8.i32(i8* align 4 %dest, i8* align 4 %src, i32 100, i1 false)
Downstream users may have to update their lit tests that check for
@llvm.memcpy/memmove/memset call/declaration patterns. The following extended sed script
may help with updating the majority of your tests, but it does not catch all possible
patterns so some manual checking and updating will be required.
s~declare void @llvm\.mem(set|cpy|move)\.p([^(]*)\((.*), i32, i1\)~declare void @llvm.mem\1.p\2(\3, i1)~g
s~call void @llvm\.memset\.p([^(]*)i8\(i8([^*]*)\* (.*), i8 (.*), i8 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i8(i8\2* \3, i8 \4, i8 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i16\(i8([^*]*)\* (.*), i8 (.*), i16 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i16(i8\2* \3, i8 \4, i16 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i32\(i8([^*]*)\* (.*), i8 (.*), i32 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i32(i8\2* \3, i8 \4, i32 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i64\(i8([^*]*)\* (.*), i8 (.*), i64 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i64(i8\2* \3, i8 \4, i64 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i128\(i8([^*]*)\* (.*), i8 (.*), i128 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.memset.p\1i128(i8\2* \3, i8 \4, i128 \5, i1 \6)~g
s~call void @llvm\.memset\.p([^(]*)i8\(i8([^*]*)\* (.*), i8 (.*), i8 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i8(i8\2* align \6 \3, i8 \4, i8 \5, i1 \7)~g
s~call void @llvm\.memset\.p([^(]*)i16\(i8([^*]*)\* (.*), i8 (.*), i16 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i16(i8\2* align \6 \3, i8 \4, i16 \5, i1 \7)~g
s~call void @llvm\.memset\.p([^(]*)i32\(i8([^*]*)\* (.*), i8 (.*), i32 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i32(i8\2* align \6 \3, i8 \4, i32 \5, i1 \7)~g
s~call void @llvm\.memset\.p([^(]*)i64\(i8([^*]*)\* (.*), i8 (.*), i64 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i64(i8\2* align \6 \3, i8 \4, i64 \5, i1 \7)~g
s~call void @llvm\.memset\.p([^(]*)i128\(i8([^*]*)\* (.*), i8 (.*), i128 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.memset.p\1i128(i8\2* align \6 \3, i8 \4, i128 \5, i1 \7)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i8\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i8 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i8(i8\3* \4, i8\5* \6, i8 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i16\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i16 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i16(i8\3* \4, i8\5* \6, i16 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i32\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i32 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i32(i8\3* \4, i8\5* \6, i32 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i64\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i64 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i64(i8\3* \4, i8\5* \6, i64 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i128\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i128 (.*), i32 [01], i1 ([^)]*)\)~call void @llvm.mem\1.p\2i128(i8\3* \4, i8\5* \6, i128 \7, i1 \8)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i8\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i8 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i8(i8\3* align \8 \4, i8\5* align \8 \6, i8 \7, i1 \9)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i16\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i16 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i16(i8\3* align \8 \4, i8\5* align \8 \6, i16 \7, i1 \9)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i32\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i32 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i32(i8\3* align \8 \4, i8\5* align \8 \6, i32 \7, i1 \9)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i64\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i64 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i64(i8\3* align \8 \4, i8\5* align \8 \6, i64 \7, i1 \9)~g
s~call void @llvm\.mem(cpy|move)\.p([^(]*)i128\(i8([^*]*)\* (.*), i8([^*]*)\* (.*), i128 (.*), i32 ([0-9]*), i1 ([^)]*)\)~call void @llvm.mem\1.p\2i128(i8\3* align \8 \4, i8\5* align \8 \6, i128 \7, i1 \9)~g
The remaining changes in the series will:
Step 2) Expand the IRBuilder API to allow creation of memcpy/memmove with differing
source and dest alignments.
Step 3) Update Clang to use the new IRBuilder API.
Step 4) Update Polly to use the new IRBuilder API.
Step 5) Update LLVM passes that create memcpy/memmove calls to use the new IRBuilder API,
and those that use use MemIntrinsicInst::[get|set]Alignment() to use
getDestAlignment() and getSourceAlignment() instead.
Step 6) Remove the single-alignment IRBuilder API for memcpy/memmove, and the
MemIntrinsicInst::[get|set]Alignment() methods.
Reviewers: pete, hfinkel, lhames, reames, bollu
Reviewed By: reames
Subscribers: niosHD, reames, jholewinski, qcolombet, jfb, sanjoy, arsenm, dschuff, dylanmckay, mehdi_amini, sdardis, nemanjai, david2050, nhaehnle, javed.absar, sbc100, jgravelle-google, eraman, aheejin, kbarton, JDevlieghere, asb, rbar, johnrusso, simoncook, jordy.potman.lists, apazos, sabuasal, llvm-commits
Differential Revision: https://reviews.llvm.org/D41675
llvm-svn: 322965
2018-01-20 01:13:12 +08:00
|
|
|
declare void @llvm.memset.p0i8.i64(i8* nocapture, i8, i64, i1) #3
|
2014-08-15 21:51:57 +08:00
|
|
|
|
2019-12-25 08:11:33 +08:00
|
|
|
attributes #0 = { nounwind "less-precise-fpmad"="false" "frame-pointer"="none" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
|
|
|
|
attributes #1 = { "less-precise-fpmad"="false" "frame-pointer"="none" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
|
|
|
|
attributes #2 = { inlinehint nounwind "less-precise-fpmad"="false" "frame-pointer"="none" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
|
2014-08-15 21:51:57 +08:00
|
|
|
attributes #3 = { nounwind }
|
2019-12-25 08:11:33 +08:00
|
|
|
attributes #4 = { noreturn nounwind "less-precise-fpmad"="false" "frame-pointer"="none" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
|
|
|
|
attributes #5 = { nounwind readonly "less-precise-fpmad"="false" "frame-pointer"="none" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
|
|
|
|
attributes #6 = { nobuiltin nounwind "less-precise-fpmad"="false" "frame-pointer"="none" "no-infs-fp-math"="false" "no-nans-fp-math"="false" "stack-protector-buffer-size"="8" "unsafe-fp-math"="false" "use-soft-float"="false" }
|
2014-08-15 21:51:57 +08:00
|
|
|
attributes #7 = { noreturn nounwind }
|
|
|
|
|
|
|
|
!llvm.ident = !{!0}
|
|
|
|
|
IR: Make metadata typeless in assembly
Now that `Metadata` is typeless, reflect that in the assembly. These
are the matching assembly changes for the metadata/value split in
r223802.
- Only use the `metadata` type when referencing metadata from a call
intrinsic -- i.e., only when it's used as a `Value`.
- Stop pretending that `ValueAsMetadata` is wrapped in an `MDNode`
when referencing it from call intrinsics.
So, assembly like this:
define @foo(i32 %v) {
call void @llvm.foo(metadata !{i32 %v}, metadata !0)
call void @llvm.foo(metadata !{i32 7}, metadata !0)
call void @llvm.foo(metadata !1, metadata !0)
call void @llvm.foo(metadata !3, metadata !0)
call void @llvm.foo(metadata !{metadata !3}, metadata !0)
ret void, !bar !2
}
!0 = metadata !{metadata !2}
!1 = metadata !{i32* @global}
!2 = metadata !{metadata !3}
!3 = metadata !{}
turns into this:
define @foo(i32 %v) {
call void @llvm.foo(metadata i32 %v, metadata !0)
call void @llvm.foo(metadata i32 7, metadata !0)
call void @llvm.foo(metadata i32* @global, metadata !0)
call void @llvm.foo(metadata !3, metadata !0)
call void @llvm.foo(metadata !{!3}, metadata !0)
ret void, !bar !2
}
!0 = !{!2}
!1 = !{i32* @global}
!2 = !{!3}
!3 = !{}
I wrote an upgrade script that handled almost all of the tests in llvm
and many of the tests in cfe (even handling many `CHECK` lines). I've
attached it (or will attach it in a moment if you're speedy) to PR21532
to help everyone update their out-of-tree testcases.
This is part of PR21532.
llvm-svn: 224257
2014-12-16 03:07:53 +08:00
|
|
|
!0 = !{!"clang version 3.6.0 (trunk 215115) (llvm/trunk 215117)"}
|
|
|
|
!1 = !{!2, !4, i64 0}
|
|
|
|
!2 = !{!"_ZTSSs", !3, i64 0}
|
|
|
|
!3 = !{!"_ZTSNSs12_Alloc_hiderE", !4, i64 0}
|
|
|
|
!4 = !{!"any pointer", !5, i64 0}
|
|
|
|
!5 = !{!"omnipotent char", !6, i64 0}
|
|
|
|
!6 = !{!"Simple C/C++ TBAA"}
|
|
|
|
!7 = !{!8, !9, i64 0}
|
|
|
|
!8 = !{!"_ZTSNSs9_Rep_baseE", !9, i64 0, !9, i64 8, !10, i64 16}
|
|
|
|
!9 = !{!"long", !5, i64 0}
|
|
|
|
!10 = !{!"int", !5, i64 0}
|
|
|
|
!11 = !{!12, !12, i64 0}
|
|
|
|
!12 = !{!"vtable pointer", !6, i64 0}
|
|
|
|
!13 = !{!3, !4, i64 0}
|
|
|
|
!14 = !{!15, !10, i64 24}
|
|
|
|
!15 = !{!"_ZTSN4llvm12SMDiagnosticE", !4, i64 0, !16, i64 8, !2, i64 16, !10, i64 24, !10, i64 28, !17, i64 32, !2, i64 40, !2, i64 48, !18, i64 56, !19, i64 80}
|
|
|
|
!16 = !{!"_ZTSN4llvm5SMLocE", !4, i64 0}
|
|
|
|
!17 = !{!"_ZTSN4llvm9SourceMgr8DiagKindE", !5, i64 0}
|
|
|
|
!18 = !{!"_ZTSSt6vectorISt4pairIjjESaIS1_EE"}
|
|
|
|
!19 = !{!"_ZTSN4llvm11SmallVectorINS_7SMFixItELj4EEE", !20, i64 48}
|
|
|
|
!20 = !{!"_ZTSN4llvm18SmallVectorStorageINS_7SMFixItELj4EEE", !5, i64 0}
|
|
|
|
!21 = !{!15, !10, i64 28}
|
|
|
|
!22 = !{!15, !17, i64 32}
|
|
|
|
!23 = !{!24, !4, i64 0}
|
|
|
|
!24 = !{!"_ZTSN4llvm15SmallVectorBaseE", !4, i64 0, !4, i64 8, !4, i64 16}
|
|
|
|
!25 = !{!24, !4, i64 8}
|
|
|
|
!26 = !{!24, !4, i64 16}
|
|
|
|
!27 = !{!4, !4, i64 0}
|
|
|
|
!28 = !{!"branch_weights", i32 64, i32 4}
|
|
|
|
!29 = !{!10, !10, i64 0}
|
|
|
|
!30 = !{!31, !4, i64 8}
|
|
|
|
!31 = !{!"_ZTSN4llvm12MemoryBufferE", !4, i64 8, !4, i64 16}
|
|
|
|
!32 = !{!31, !4, i64 16}
|
|
|
|
!33 = !{!5, !5, i64 0}
|
|
|
|
!34 = !{!35, !4, i64 0}
|
|
|
|
!35 = !{!"_ZTSSt12_Vector_baseISt4pairIjjESaIS1_EE", !36, i64 0}
|
|
|
|
!36 = !{!"_ZTSNSt12_Vector_baseISt4pairIjjESaIS1_EE12_Vector_implE", !4, i64 0, !4, i64 8, !4, i64 16}
|
|
|
|
!37 = !{!38, !38, i64 0}
|
|
|
|
!38 = !{!"bool", !5, i64 0}
|
|
|
|
!39 = !{i8 0, i8 2}
|
|
|
|
!40 = !{!41, !4, i64 0}
|
|
|
|
!41 = !{!"_ZTSN4llvm10TimeRegionE", !4, i64 0}
|
|
|
|
!42 = !{!43, !44, i64 32}
|
|
|
|
!43 = !{!"_ZTSN4llvm11raw_ostreamE", !4, i64 8, !4, i64 16, !4, i64 24, !44, i64 32}
|
|
|
|
!44 = !{!"_ZTSN4llvm11raw_ostream10BufferKindE", !5, i64 0}
|
|
|
|
!45 = !{!43, !4, i64 24}
|
|
|
|
!46 = !{!43, !4, i64 8}
|
|
|
|
!47 = !{i64 0, i64 8, !27, i64 8, i64 8, !27}
|