go/src/cmd/internal/obj/link.go

888 lines
27 KiB
Go
Raw Normal View History

// Derived from Inferno utils/6l/l.h and related files.
// https://bitbucket.org/inferno-os/inferno-os/src/default/utils/6l/l.h
//
// Copyright © 1994-1999 Lucent Technologies Inc. All rights reserved.
// Portions Copyright © 1995-1997 C H Forsyth (forsyth@terzarima.net)
// Portions Copyright © 1997-1999 Vita Nuova Limited
// Portions Copyright © 2000-2007 Vita Nuova Holdings Limited (www.vitanuova.com)
// Portions Copyright © 2004,2006 Bruce Ellis
// Portions Copyright © 2005-2007 C H Forsyth (forsyth@terzarima.net)
// Revisions Copyright © 2000-2007 Lucent Technologies Inc. and others
// Portions Copyright © 2009 The Go Authors. All rights reserved.
//
// Permission is hereby granted, free of charge, to any person obtaining a copy
// of this software and associated documentation files (the "Software"), to deal
// in the Software without restriction, including without limitation the rights
// to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
// copies of the Software, and to permit persons to whom the Software is
// furnished to do so, subject to the following conditions:
//
// The above copyright notice and this permission notice shall be included in
// all copies or substantial portions of the Software.
//
// THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
// IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
// FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
// AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
// LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
// OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN
// THE SOFTWARE.
package obj
import (
"bufio"
"cmd/internal/dwarf"
"cmd/internal/src"
"cmd/internal/sys"
"fmt"
)
// An Addr is an argument to an instruction.
// The general forms and their encodings are:
//
// sym±offset(symkind)(reg)(index*scale)
// Memory reference at address &sym(symkind) + offset + reg + index*scale.
// Any of sym(symkind), ±offset, (reg), (index*scale), and *scale can be omitted.
// If (reg) and *scale are both omitted, the resulting expression (index) is parsed as (reg).
// To force a parsing as index*scale, write (index*1).
// Encoding:
// type = TYPE_MEM
// name = symkind (NAME_AUTO, ...) or 0 (NAME_NONE)
// sym = sym
// offset = ±offset
// reg = reg (REG_*)
// index = index (REG_*)
// scale = scale (1, 2, 4, 8)
//
// $<mem>
// Effective address of memory reference <mem>, defined above.
// Encoding: same as memory reference, but type = TYPE_ADDR.
//
// $<±integer value>
// This is a special case of $<mem>, in which only ±offset is present.
// It has a separate type for easy recognition.
// Encoding:
// type = TYPE_CONST
// offset = ±integer value
//
// *<mem>
// Indirect reference through memory reference <mem>, defined above.
// Only used on x86 for CALL/JMP *sym(SB), which calls/jumps to a function
// pointer stored in the data word sym(SB), not a function named sym(SB).
// Encoding: same as above, but type = TYPE_INDIR.
//
// $*$<mem>
// No longer used.
// On machines with actual SB registers, $*$<mem> forced the
// instruction encoding to use a full 32-bit constant, never a
// reference relative to SB.
//
// $<floating point literal>
// Floating point constant value.
// Encoding:
// type = TYPE_FCONST
// val = floating point value
//
// $<string literal, up to 8 chars>
// String literal value (raw bytes used for DATA instruction).
// Encoding:
// type = TYPE_SCONST
// val = string
//
// <register name>
// Any register: integer, floating point, control, segment, and so on.
// If looking for specific register kind, must check type and reg value range.
// Encoding:
// type = TYPE_REG
// reg = reg (REG_*)
//
// x(PC)
// Encoding:
// type = TYPE_BRANCH
// val = Prog* reference OR ELSE offset = target pc (branch takes priority)
//
// $±x-±y
// Final argument to TEXT, specifying local frame size x and argument size y.
// In this form, x and y are integer literals only, not arbitrary expressions.
// This avoids parsing ambiguities due to the use of - as a separator.
// The ± are optional.
// If the final argument to TEXT omits the -±y, the encoding should still
// use TYPE_TEXTSIZE (not TYPE_CONST), with u.argsize = ArgsSizeUnknown.
// Encoding:
// type = TYPE_TEXTSIZE
// offset = x
// val = int32(y)
//
// reg<<shift, reg>>shift, reg->shift, reg@>shift
// Shifted register value, for ARM and ARM64.
// In this form, reg must be a register and shift can be a register or an integer constant.
// Encoding:
// type = TYPE_SHIFT
// On ARM:
// offset = (reg&15) | shifttype<<5 | count
// shifttype = 0, 1, 2, 3 for <<, >>, ->, @>
// count = (reg&15)<<8 | 1<<4 for a register shift count, (n&31)<<7 for an integer constant.
// On ARM64:
// offset = (reg&31)<<16 | shifttype<<22 | (count&63)<<10
// shifttype = 0, 1, 2 for <<, >>, ->
//
// (reg, reg)
// A destination register pair. When used as the last argument of an instruction,
// this form makes clear that both registers are destinations.
// Encoding:
// type = TYPE_REGREG
// reg = first register
// offset = second register
//
// [reg, reg, reg-reg]
// Register list for ARM.
// Encoding:
// type = TYPE_REGLIST
// offset = bit mask of registers in list; R0 is low bit.
//
// reg, reg
// Register pair for ARM.
// TYPE_REGREG2
//
// (reg+reg)
// Register pair for PPC64.
// Encoding:
// type = TYPE_MEM
// reg = first register
// index = second register
// scale = 1
//
type Addr struct {
Reg int16
Index int16
Scale int16 // Sometimes holds a register.
Type AddrType
Name AddrName
Class int8
Offset int64
Sym *LSym
// argument value:
// for TYPE_SCONST, a string
// for TYPE_FCONST, a float64
// for TYPE_BRANCH, a *Prog (optional)
// for TYPE_TEXTSIZE, an int32 (optional)
Val interface{}
}
type AddrName int8
const (
NAME_NONE AddrName = iota
NAME_EXTERN
NAME_STATIC
NAME_AUTO
NAME_PARAM
// A reference to name@GOT(SB) is a reference to the entry in the global offset
// table for 'name'.
NAME_GOTREF
)
type AddrType uint8
const (
TYPE_NONE AddrType = iota
TYPE_BRANCH
TYPE_TEXTSIZE
TYPE_MEM
TYPE_CONST
TYPE_FCONST
TYPE_SCONST
TYPE_REG
TYPE_ADDR
TYPE_SHIFT
TYPE_REGREG
TYPE_REGREG2
TYPE_INDIR
TYPE_REGLIST
)
// Prog describes a single machine instruction.
//
// The general instruction form is:
//
// As.Scond From, Reg, From3, To, RegTo2
//
// where As is an opcode and the others are arguments:
// From, Reg, From3 are sources, and To, RegTo2 are destinations.
// Usually, not all arguments are present.
// For example, MOVL R1, R2 encodes using only As=MOVL, From=R1, To=R2.
// The Scond field holds additional condition bits for systems (like arm)
// that have generalized conditional execution.
//
// Jump instructions use the Pcond field to point to the target instruction,
// which must be in the same linked list as the jump instruction.
//
// The Progs for a given function are arranged in a list linked through the Link field.
//
// Each Prog is charged to a specific source line in the debug information,
// specified by Pos.Line().
// Every Prog has a Ctxt field that defines its context.
cmd/compile: teach assemblers to accept a Prog allocator The existing bulk Prog allocator is not concurrency-safe. To allow for concurrency-safe bulk allocation of Progs, I want to move Prog allocation and caching upstream, to the clients of cmd/internal/obj. This is a preliminary enabling refactoring. After this CL, instead of calling Ctxt.NewProg throughout the assemblers, we thread through a newprog function that returns a new Prog. That function is set up to be Ctxt.NewProg, so there are no real changes in this CL; this CL only establishes the plumbing. Passes toolstash-check -all. Negligible compiler performance impact. Updates #15756 name old time/op new time/op delta Template 213ms ± 3% 214ms ± 4% ~ (p=0.574 n=49+47) Unicode 90.1ms ± 5% 89.9ms ± 4% ~ (p=0.417 n=50+49) GoTypes 585ms ± 4% 584ms ± 3% ~ (p=0.466 n=49+49) SSA 6.50s ± 3% 6.52s ± 2% ~ (p=0.251 n=49+49) Flate 128ms ± 4% 128ms ± 4% ~ (p=0.673 n=49+50) GoParser 152ms ± 3% 152ms ± 3% ~ (p=0.810 n=48+49) Reflect 372ms ± 4% 372ms ± 5% ~ (p=0.778 n=49+50) Tar 113ms ± 5% 111ms ± 4% -0.98% (p=0.016 n=50+49) XML 208ms ± 3% 208ms ± 2% ~ (p=0.483 n=47+49) [Geo mean] 285ms 285ms -0.17% name old user-ns/op new user-ns/op delta Template 253M ± 8% 254M ± 9% ~ (p=0.899 n=50+50) Unicode 106M ± 9% 106M ±11% ~ (p=0.642 n=50+50) GoTypes 736M ± 4% 740M ± 4% ~ (p=0.121 n=50+49) SSA 8.82G ± 3% 8.88G ± 2% +0.65% (p=0.006 n=49+48) Flate 147M ± 4% 147M ± 5% ~ (p=0.844 n=47+48) GoParser 179M ± 4% 178M ± 6% ~ (p=0.785 n=50+50) Reflect 443M ± 6% 441M ± 5% ~ (p=0.850 n=48+47) Tar 126M ± 5% 126M ± 5% ~ (p=0.734 n=50+50) XML 244M ± 5% 244M ± 5% ~ (p=0.594 n=49+50) [Geo mean] 341M 341M +0.11% Change-Id: Ice962f61eb3a524c2db00a166cb582c22caa7d68 Reviewed-on: https://go-review.googlesource.com/39633 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-04 14:31:55 -07:00
// For performance reasons, Progs usually are usually bulk allocated, cached, and reused;
// those bulk allocators should always be used, rather than new(Prog).
//
// The other fields not yet mentioned are for use by the back ends and should
// be left zeroed by creators of Prog lists.
type Prog struct {
Ctxt *Link // linker context
Link *Prog // next Prog in linked list
From Addr // first source operand
From3 *Addr // third source operand (second is Reg below)
To Addr // destination operand (second is RegTo2 below)
Pcond *Prog // target of conditional jump
Forwd *Prog // for x86 back end
Rel *Prog // for x86, arm back ends
Pc int64 // for back ends or assembler: virtual or actual program counter, depending on phase
Pos src.XPos // source position of this instruction
Spadj int32 // effect of instruction on stack pointer (increment or decrement amount)
As As // assembler opcode
Reg int16 // 2nd source operand
RegTo2 int16 // 2nd destination operand
Mark uint16 // bitmask of arch-specific items
Optab uint16 // arch-specific opcode index
Scond uint8 // condition bits for conditional instruction (e.g., on ARM)
Back uint8 // for x86 back end: backwards branch state
Ft uint8 // for x86 back end: type index of Prog.From
Tt uint8 // for x86 back end: type index of Prog.To
Isize uint8 // for x86 back end: size of the instruction in bytes
}
// From3Type returns From3.Type, or TYPE_NONE when From3 is nil.
func (p *Prog) From3Type() AddrType {
if p.From3 == nil {
return TYPE_NONE
}
return p.From3.Type
}
// An As denotes an assembler opcode.
// There are some portable opcodes, declared here in package obj,
// that are common to all architectures.
// However, the majority of opcodes are arch-specific
// and are declared in their respective architecture's subpackage.
type As int16
// These are the portable opcodes.
const (
AXXX As = iota
ACALL
ADUFFCOPY
ADUFFZERO
AEND
AFUNCDATA
AJMP
ANOP
APCDATA
ARET
ATEXT
AUNDEF
A_ARCHSPECIFIC
)
// Each architecture is allotted a distinct subspace of opcode values
// for declaring its arch-specific opcodes.
// Within this subspace, the first arch-specific opcode should be
// at offset A_ARCHSPECIFIC.
//
// Subspaces are aligned to a power of two so opcodes can be masked
// with AMask and used as compact array indices.
const (
ABase386 = (1 + iota) << 10
ABaseARM
ABaseAMD64
ABasePPC64
ABaseARM64
ABaseMIPS
ABaseS390X
AllowedOpCodes = 1 << 10 // The number of opcodes available for any given architecture.
AMask = AllowedOpCodes - 1 // AND with this to use the opcode as an array index.
)
// An LSym is the sort of symbol that is written to an object file.
type LSym struct {
Name string
Type SymKind
Version int16
Attribute
RefIdx int // Index of this symbol in the symbol reference list.
Size int64
Gotype *LSym
P []byte
R []Reloc
// TODO(mdempsky): De-anonymize field.
*FuncInfo
}
// A FuncInfo contains extra fields for STEXT symbols.
type FuncInfo struct {
Args int32
Locals int32
Text *Prog
Autom []*Auto
Pcln Pcln
}
// Attribute is a set of symbol attributes.
type Attribute int16
const (
AttrDuplicateOK Attribute = 1 << iota
AttrCFunc
AttrNoSplit
AttrLeaf
cmd/internal/obj: stop storing Text flags in From3 Prior to this CL, flags such as NOSPLIT on ATEXT Progs were stored in From3.Offset. Some but not all of those flags were also duplicated into From.Sym.Attribute. This CL migrates all of those flags into From.Sym.Attribute and stops creating a From3. A side-effect of this is that printing an ATEXT Prog can no longer simply dump From3.Offset. That's kind of good, since the raw flag value wasn't very informative anyway, but it did necessitate a bunch of updates to the cmd/asm tests. The reason I'm doing this work now is that avoiding storing flags in both From.Sym and From3.Offset simplifies some other changes to fix the data race first described in CL 40254. This CL almost passes toolstash-check -all. The only changes are in cases where the assembler has decided that a function's flags may be altered, e.g. to make a function with no calls in it NOSPLIT. Prior to this CL, that information was not printed. Sample before: "".Ctz64 t=1 size=63 args=0x10 locals=0x0 0x0000 00000 (/Users/josh/go/tip/src/runtime/internal/sys/intrinsics.go:35) TEXT "".Ctz64(SB), $0-16 0x0000 00000 (/Users/josh/go/tip/src/runtime/internal/sys/intrinsics.go:35) FUNCDATA $0, gclocals·f207267fbf96a0178e8758c6e3e0ce28(SB) Sample after: "".Ctz64 t=1 nosplit size=63 args=0x10 locals=0x0 0x0000 00000 (/Users/josh/go/tip/src/runtime/internal/sys/intrinsics.go:35) TEXT "".Ctz64(SB), NOSPLIT, $0-16 0x0000 00000 (/Users/josh/go/tip/src/runtime/internal/sys/intrinsics.go:35) FUNCDATA $0, gclocals·f207267fbf96a0178e8758c6e3e0ce28(SB) Observe the additional "nosplit" in the first line and the additional "NOSPLIT" in the second line. Updates #15756 Change-Id: I5c59bd8f3bdc7c780361f801d94a261f0aef3d13 Reviewed-on: https://go-review.googlesource.com/40495 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2017-04-11 15:15:04 -07:00
AttrWrapper
AttrNeedCtxt
AttrNoFrame
AttrSeenGlobl
AttrOnList
// MakeTypelink means that the type should have an entry in the typelink table.
AttrMakeTypelink
// ReflectMethod means the function may call reflect.Type.Method or
// reflect.Type.MethodByName. Matching is imprecise (as reflect.Type
// can be used through a custom interface), so ReflectMethod may be
// set in some cases when the reflect package is not called.
//
// Used by the linker to determine what methods can be pruned.
AttrReflectMethod
// Local means make the symbol local even when compiling Go code to reference Go
// symbols in other shared libraries, as in this mode symbols are global by
// default. "local" here means in the sense of the dynamic linker, i.e. not
// visible outside of the module (shared library or executable) that contains its
// definition. (When not compiling to support Go shared libraries, all symbols are
// local in this sense unless there is a cgo_export_* directive).
AttrLocal
)
func (a Attribute) DuplicateOK() bool { return a&AttrDuplicateOK != 0 }
func (a Attribute) MakeTypelink() bool { return a&AttrMakeTypelink != 0 }
func (a Attribute) CFunc() bool { return a&AttrCFunc != 0 }
func (a Attribute) NoSplit() bool { return a&AttrNoSplit != 0 }
func (a Attribute) Leaf() bool { return a&AttrLeaf != 0 }
func (a Attribute) SeenGlobl() bool { return a&AttrSeenGlobl != 0 }
func (a Attribute) OnList() bool { return a&AttrOnList != 0 }
func (a Attribute) ReflectMethod() bool { return a&AttrReflectMethod != 0 }
func (a Attribute) Local() bool { return a&AttrLocal != 0 }
cmd/internal/obj: stop storing Text flags in From3 Prior to this CL, flags such as NOSPLIT on ATEXT Progs were stored in From3.Offset. Some but not all of those flags were also duplicated into From.Sym.Attribute. This CL migrates all of those flags into From.Sym.Attribute and stops creating a From3. A side-effect of this is that printing an ATEXT Prog can no longer simply dump From3.Offset. That's kind of good, since the raw flag value wasn't very informative anyway, but it did necessitate a bunch of updates to the cmd/asm tests. The reason I'm doing this work now is that avoiding storing flags in both From.Sym and From3.Offset simplifies some other changes to fix the data race first described in CL 40254. This CL almost passes toolstash-check -all. The only changes are in cases where the assembler has decided that a function's flags may be altered, e.g. to make a function with no calls in it NOSPLIT. Prior to this CL, that information was not printed. Sample before: "".Ctz64 t=1 size=63 args=0x10 locals=0x0 0x0000 00000 (/Users/josh/go/tip/src/runtime/internal/sys/intrinsics.go:35) TEXT "".Ctz64(SB), $0-16 0x0000 00000 (/Users/josh/go/tip/src/runtime/internal/sys/intrinsics.go:35) FUNCDATA $0, gclocals·f207267fbf96a0178e8758c6e3e0ce28(SB) Sample after: "".Ctz64 t=1 nosplit size=63 args=0x10 locals=0x0 0x0000 00000 (/Users/josh/go/tip/src/runtime/internal/sys/intrinsics.go:35) TEXT "".Ctz64(SB), NOSPLIT, $0-16 0x0000 00000 (/Users/josh/go/tip/src/runtime/internal/sys/intrinsics.go:35) FUNCDATA $0, gclocals·f207267fbf96a0178e8758c6e3e0ce28(SB) Observe the additional "nosplit" in the first line and the additional "NOSPLIT" in the second line. Updates #15756 Change-Id: I5c59bd8f3bdc7c780361f801d94a261f0aef3d13 Reviewed-on: https://go-review.googlesource.com/40495 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2017-04-11 15:15:04 -07:00
func (a Attribute) Wrapper() bool { return a&AttrWrapper != 0 }
func (a Attribute) NeedCtxt() bool { return a&AttrNeedCtxt != 0 }
func (a Attribute) NoFrame() bool { return a&AttrNoFrame != 0 }
func (a *Attribute) Set(flag Attribute, value bool) {
if value {
*a |= flag
} else {
*a &^= flag
}
}
cmd/internal/obj: stop storing Text flags in From3 Prior to this CL, flags such as NOSPLIT on ATEXT Progs were stored in From3.Offset. Some but not all of those flags were also duplicated into From.Sym.Attribute. This CL migrates all of those flags into From.Sym.Attribute and stops creating a From3. A side-effect of this is that printing an ATEXT Prog can no longer simply dump From3.Offset. That's kind of good, since the raw flag value wasn't very informative anyway, but it did necessitate a bunch of updates to the cmd/asm tests. The reason I'm doing this work now is that avoiding storing flags in both From.Sym and From3.Offset simplifies some other changes to fix the data race first described in CL 40254. This CL almost passes toolstash-check -all. The only changes are in cases where the assembler has decided that a function's flags may be altered, e.g. to make a function with no calls in it NOSPLIT. Prior to this CL, that information was not printed. Sample before: "".Ctz64 t=1 size=63 args=0x10 locals=0x0 0x0000 00000 (/Users/josh/go/tip/src/runtime/internal/sys/intrinsics.go:35) TEXT "".Ctz64(SB), $0-16 0x0000 00000 (/Users/josh/go/tip/src/runtime/internal/sys/intrinsics.go:35) FUNCDATA $0, gclocals·f207267fbf96a0178e8758c6e3e0ce28(SB) Sample after: "".Ctz64 t=1 nosplit size=63 args=0x10 locals=0x0 0x0000 00000 (/Users/josh/go/tip/src/runtime/internal/sys/intrinsics.go:35) TEXT "".Ctz64(SB), NOSPLIT, $0-16 0x0000 00000 (/Users/josh/go/tip/src/runtime/internal/sys/intrinsics.go:35) FUNCDATA $0, gclocals·f207267fbf96a0178e8758c6e3e0ce28(SB) Observe the additional "nosplit" in the first line and the additional "NOSPLIT" in the second line. Updates #15756 Change-Id: I5c59bd8f3bdc7c780361f801d94a261f0aef3d13 Reviewed-on: https://go-review.googlesource.com/40495 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> Reviewed-by: Brad Fitzpatrick <bradfitz@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2017-04-11 15:15:04 -07:00
var textAttrStrings = [...]struct {
bit Attribute
s string
}{
{bit: AttrDuplicateOK, s: "DUPOK"},
{bit: AttrMakeTypelink, s: ""},
{bit: AttrCFunc, s: "CFUNC"},
{bit: AttrNoSplit, s: "NOSPLIT"},
{bit: AttrLeaf, s: "LEAF"},
{bit: AttrSeenGlobl, s: ""},
{bit: AttrOnList, s: ""},
{bit: AttrReflectMethod, s: "REFLECTMETHOD"},
{bit: AttrLocal, s: "LOCAL"},
{bit: AttrWrapper, s: "WRAPPER"},
{bit: AttrNeedCtxt, s: "NEEDCTXT"},
{bit: AttrNoFrame, s: "NOFRAME"},
}
// TextAttrString formats a for printing in as part of a TEXT prog.
func (a Attribute) TextAttrString() string {
var s string
for _, x := range textAttrStrings {
if a&x.bit != 0 {
if x.s != "" {
s += x.s + "|"
}
a &^= x.bit
}
}
if a != 0 {
s += fmt.Sprintf("UnknownAttribute(%d)|", a)
}
// Chop off trailing |, if present.
if len(s) > 0 {
s = s[:len(s)-1]
}
return s
}
// The compiler needs LSym to satisfy fmt.Stringer, because it stores
// an LSym in ssa.ExternSymbol.
func (s *LSym) String() string {
return s.Name
}
type Pcln struct {
Pcsp Pcdata
Pcfile Pcdata
Pcline Pcdata
cmd/compile,link: generate PC-value tables with inlining information In order to generate accurate tracebacks, the runtime needs to know the inlined call stack for a given PC. This creates two tables per function for this purpose. The first table is the inlining tree (stored in the function's funcdata), which has a node containing the file, line, and function name for every inlined call. The second table is a PC-value table that maps each PC to a node in the inlining tree (or -1 if the PC is not the result of inlining). To give the appearance that inlining hasn't happened, the runtime also needs the original source position information of inlined AST nodes. Previously the compiler plastered over the line numbers of inlined AST nodes with the line number of the call. This meant that the PC-line table mapped each PC to line number of the outermost call in its inlined call stack, with no way to access the innermost line number. Now the compiler retains line numbers of inlined AST nodes and writes the innermost source position information to the PC-line and PC-file tables. Some tools and tests expect to see outermost line numbers, so we provide the OutermostLine function for displaying line info. To keep track of the inlined call stack for an AST node, we extend the src.PosBase type with an index into a global inlining tree. Every time the compiler inlines a call, it creates a node in the global inlining tree for the call, and writes its index to the PosBase of every inlined AST node. The parent of this node is the inlining tree index of the call. -1 signifies no parent. For each function, the compiler creates a local inlining tree and a PC-value table mapping each PC to an index in the local tree. These are written to an object file, which is read by the linker. The linker re-encodes these tables compactly by deduplicating function names and file names. This change increases the size of binaries by 4-5%. For example, this is how the go1 benchmark binary is impacted by this change: section old bytes new bytes delta .text 3.49M ± 0% 3.49M ± 0% +0.06% .rodata 1.12M ± 0% 1.21M ± 0% +8.21% .gopclntab 1.50M ± 0% 1.68M ± 0% +11.89% .debug_line 338k ± 0% 435k ± 0% +28.78% Total 9.21M ± 0% 9.58M ± 0% +4.01% Updates #19348. Change-Id: Ic4f180c3b516018138236b0c35e0218270d957d3 Reviewed-on: https://go-review.googlesource.com/37231 Run-TryBot: David Lazar <lazard@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Austin Clements <austin@google.com>
2017-02-17 12:28:05 -05:00
Pcinline Pcdata
Pcdata []Pcdata
Funcdata []*LSym
Funcdataoff []int64
File []string
Lastfile string
Lastindex int
cmd/compile,link: generate PC-value tables with inlining information In order to generate accurate tracebacks, the runtime needs to know the inlined call stack for a given PC. This creates two tables per function for this purpose. The first table is the inlining tree (stored in the function's funcdata), which has a node containing the file, line, and function name for every inlined call. The second table is a PC-value table that maps each PC to a node in the inlining tree (or -1 if the PC is not the result of inlining). To give the appearance that inlining hasn't happened, the runtime also needs the original source position information of inlined AST nodes. Previously the compiler plastered over the line numbers of inlined AST nodes with the line number of the call. This meant that the PC-line table mapped each PC to line number of the outermost call in its inlined call stack, with no way to access the innermost line number. Now the compiler retains line numbers of inlined AST nodes and writes the innermost source position information to the PC-line and PC-file tables. Some tools and tests expect to see outermost line numbers, so we provide the OutermostLine function for displaying line info. To keep track of the inlined call stack for an AST node, we extend the src.PosBase type with an index into a global inlining tree. Every time the compiler inlines a call, it creates a node in the global inlining tree for the call, and writes its index to the PosBase of every inlined AST node. The parent of this node is the inlining tree index of the call. -1 signifies no parent. For each function, the compiler creates a local inlining tree and a PC-value table mapping each PC to an index in the local tree. These are written to an object file, which is read by the linker. The linker re-encodes these tables compactly by deduplicating function names and file names. This change increases the size of binaries by 4-5%. For example, this is how the go1 benchmark binary is impacted by this change: section old bytes new bytes delta .text 3.49M ± 0% 3.49M ± 0% +0.06% .rodata 1.12M ± 0% 1.21M ± 0% +8.21% .gopclntab 1.50M ± 0% 1.68M ± 0% +11.89% .debug_line 338k ± 0% 435k ± 0% +28.78% Total 9.21M ± 0% 9.58M ± 0% +4.01% Updates #19348. Change-Id: Ic4f180c3b516018138236b0c35e0218270d957d3 Reviewed-on: https://go-review.googlesource.com/37231 Run-TryBot: David Lazar <lazard@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Austin Clements <austin@google.com>
2017-02-17 12:28:05 -05:00
InlTree InlTree // per-function inlining tree extracted from the global tree
}
// A SymKind describes the kind of memory represented by a symbol.
type SymKind int16
// Defined SymKind values.
//
// TODO(rsc): Give idiomatic Go names.
// TODO(rsc): Reduce the number of symbol types in the object files.
//go:generate stringer -type=SymKind
const (
Sxxx SymKind = iota
STEXT
SELFRXSECT
// Read-only sections.
STYPE
SSTRING
SGOSTRING
SGOFUNC
SGCBITS
SRODATA
SFUNCTAB
SELFROSECT
SMACHOPLT
// Read-only sections with relocations.
//
// Types STYPE-SFUNCTAB above are written to the .rodata section by default.
// When linking a shared object, some conceptually "read only" types need to
// be written to by relocations and putting them in a section called
// ".rodata" interacts poorly with the system linkers. The GNU linkers
// support this situation by arranging for sections of the name
// ".data.rel.ro.XXX" to be mprotected read only by the dynamic linker after
// relocations have applied, so when the Go linker is creating a shared
// object it checks all objects of the above types and bumps any object that
// has a relocation to it to the corresponding type below, which are then
// written to sections with appropriate magic names.
STYPERELRO
SSTRINGRELRO
SGOSTRINGRELRO
SGOFUNCRELRO
SGCBITSRELRO
SRODATARELRO
SFUNCTABRELRO
// Part of .data.rel.ro if it exists, otherwise part of .rodata.
STYPELINK
SITABLINK
SSYMTAB
SPCLNTAB
// Writable sections.
SELFSECT
SMACHO
SMACHOGOT
SWINDOWS
SELFGOT
SNOPTRDATA
SINITARR
SDATA
SBSS
SNOPTRBSS
STLSBSS
SXREF
SMACHOSYMSTR
SMACHOSYMTAB
SMACHOINDIRECTPLT
SMACHOINDIRECTGOT
SFILE
SFILEPATH
SCONST
SDYNIMPORT
SHOSTOBJ
SDWARFSECT
SDWARFINFO
SSUB = SymKind(1 << 8)
SMASK = SymKind(SSUB - 1)
SHIDDEN = SymKind(1 << 9)
SCONTAINER = SymKind(1 << 10) // has a sub-symbol
)
// ReadOnly are the symbol kinds that form read-only sections. In some
// cases, if they will require relocations, they are transformed into
// rel-ro sections using RelROMap.
var ReadOnly = []SymKind{
STYPE,
SSTRING,
SGOSTRING,
SGOFUNC,
SGCBITS,
SRODATA,
SFUNCTAB,
}
// RelROMap describes the transformation of read-only symbols to rel-ro
// symbols.
var RelROMap = map[SymKind]SymKind{
STYPE: STYPERELRO,
SSTRING: SSTRINGRELRO,
SGOSTRING: SGOSTRINGRELRO,
SGOFUNC: SGOFUNCRELRO,
SGCBITS: SGCBITSRELRO,
SRODATA: SRODATARELRO,
SFUNCTAB: SFUNCTABRELRO,
}
type Reloc struct {
Off int32
Siz uint8
Type RelocType
Add int64
Sym *LSym
}
type RelocType int32
//go:generate stringer -type=RelocType
const (
R_ADDR RelocType = 1 + iota
// R_ADDRPOWER relocates a pair of "D-form" instructions (instructions with 16-bit
// immediates in the low half of the instruction word), usually addis followed by
// another add or a load, inserting the "high adjusted" 16 bits of the address of
// the referenced symbol into the immediate field of the first instruction and the
// low 16 bits into that of the second instruction.
R_ADDRPOWER
// R_ADDRARM64 relocates an adrp, add pair to compute the address of the
// referenced symbol.
R_ADDRARM64
// R_ADDRMIPS (only used on mips/mips64) resolves to the low 16 bits of an external
// address, by encoding it into the instruction.
R_ADDRMIPS
cmd/compile, etc: store method tables as offsets This CL introduces the typeOff type and a lookup method of the same name that can turn a typeOff offset into an *rtype. In a typical Go binary (built with buildmode=exe, pie, c-archive, or c-shared), there is one moduledata and all typeOff values are offsets relative to firstmoduledata.types. This makes computing the pointer cheap in typical programs. With buildmode=shared (and one day, buildmode=plugin) there are multiple modules whose relative offset is determined at runtime. We identify a type in the general case by the pair of the original *rtype that references it and its typeOff value. We determine the module from the original pointer, and then use the typeOff from there to compute the final *rtype. To ensure there is only one *rtype representing each type, the runtime initializes a typemap for each module, using any identical type from an earlier module when resolving that offset. This means that types computed from an offset match the type mapped by the pointer dynamic relocations. A series of followup CLs will replace other *rtype values with typeOff (and name/*string with nameOff). For types created at runtime by reflect, type offsets are treated as global IDs and reference into a reflect offset map kept by the runtime. darwin/amd64: cmd/go: -57KB (0.6%) jujud: -557KB (0.8%) linux/amd64 PIE: cmd/go: -361KB (3.0%) jujud: -3.5MB (4.2%) For #6853. Change-Id: Icf096fd884a0a0cb9f280f46f7a26c70a9006c96 Reviewed-on: https://go-review.googlesource.com/21285 Reviewed-by: Ian Lance Taylor <iant@golang.org> Run-TryBot: David Crawshaw <crawshaw@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-03-28 10:32:27 -04:00
// R_ADDROFF resolves to a 32-bit offset from the beginning of the section
// holding the data being relocated to the referenced symbol.
R_ADDROFF
// R_WEAKADDROFF resolves just like R_ADDROFF but is a weak relocation.
// A weak relocation does not make the symbol it refers to reachable,
// and is only honored by the linker if the symbol is in some other way
// reachable.
R_WEAKADDROFF
R_SIZE
R_CALL
R_CALLARM
R_CALLARM64
R_CALLIND
R_CALLPOWER
// R_CALLMIPS (only used on mips64) resolves to non-PC-relative target address
// of a CALL (JAL) instruction, by encoding the address into the instruction.
R_CALLMIPS
R_CONST
R_PCREL
// R_TLS_LE, used on 386, amd64, and ARM, resolves to the offset of the
// thread-local symbol from the thread local base and is used to implement the
// "local exec" model for tls access (r.Sym is not set on intel platforms but is
// set to a TLS symbol -- runtime.tlsg -- in the linker when externally linking).
R_TLS_LE
// R_TLS_IE, used 386, amd64, and ARM resolves to the PC-relative offset to a GOT
// slot containing the offset from the thread-local symbol from the thread local
// base and is used to implemented the "initial exec" model for tls access (r.Sym
// is not set on intel platforms but is set to a TLS symbol -- runtime.tlsg -- in
// the linker when externally linking).
R_TLS_IE
R_GOTOFF
R_PLT0
R_PLT1
R_PLT2
R_USEFIELD
// R_USETYPE resolves to an *rtype, but no relocation is created. The
// linker uses this as a signal that the pointed-to type information
// should be linked into the final binary, even if there are no other
// direct references. (This is used for types reachable by reflection.)
R_USETYPE
cmd/compile, etc: store method tables as offsets This CL introduces the typeOff type and a lookup method of the same name that can turn a typeOff offset into an *rtype. In a typical Go binary (built with buildmode=exe, pie, c-archive, or c-shared), there is one moduledata and all typeOff values are offsets relative to firstmoduledata.types. This makes computing the pointer cheap in typical programs. With buildmode=shared (and one day, buildmode=plugin) there are multiple modules whose relative offset is determined at runtime. We identify a type in the general case by the pair of the original *rtype that references it and its typeOff value. We determine the module from the original pointer, and then use the typeOff from there to compute the final *rtype. To ensure there is only one *rtype representing each type, the runtime initializes a typemap for each module, using any identical type from an earlier module when resolving that offset. This means that types computed from an offset match the type mapped by the pointer dynamic relocations. A series of followup CLs will replace other *rtype values with typeOff (and name/*string with nameOff). For types created at runtime by reflect, type offsets are treated as global IDs and reference into a reflect offset map kept by the runtime. darwin/amd64: cmd/go: -57KB (0.6%) jujud: -557KB (0.8%) linux/amd64 PIE: cmd/go: -361KB (3.0%) jujud: -3.5MB (4.2%) For #6853. Change-Id: Icf096fd884a0a0cb9f280f46f7a26c70a9006c96 Reviewed-on: https://go-review.googlesource.com/21285 Reviewed-by: Ian Lance Taylor <iant@golang.org> Run-TryBot: David Crawshaw <crawshaw@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-03-28 10:32:27 -04:00
// R_METHODOFF resolves to a 32-bit offset from the beginning of the section
// holding the data being relocated to the referenced symbol.
// It is a variant of R_ADDROFF used when linking from the uncommonType of a
// *rtype, and may be set to zero by the linker if it determines the method
// text is unreachable by the linked program.
R_METHODOFF
R_POWER_TOC
R_GOTPCREL
// R_JMPMIPS (only used on mips64) resolves to non-PC-relative target address
// of a JMP instruction, by encoding the address into the instruction.
// The stack nosplit check ignores this since it is not a function call.
R_JMPMIPS
// R_DWARFREF resolves to the offset of the symbol from its section.
R_DWARFREF
// Platform dependent relocations. Architectures with fixed width instructions
// have the inherent issue that a 32-bit (or 64-bit!) displacement cannot be
// stuffed into a 32-bit instruction, so an address needs to be spread across
// several instructions, and in turn this requires a sequence of relocations, each
// updating a part of an instruction. This leads to relocation codes that are
// inherently processor specific.
// Arm64.
// Set a MOV[NZ] immediate field to bits [15:0] of the offset from the thread
// local base to the thread local variable defined by the referenced (thread
// local) symbol. Error if the offset does not fit into 16 bits.
R_ARM64_TLS_LE
// Relocates an ADRP; LD64 instruction sequence to load the offset between
// the thread local base and the thread local variable defined by the
// referenced (thread local) symbol from the GOT.
R_ARM64_TLS_IE
// R_ARM64_GOTPCREL relocates an adrp, ld64 pair to compute the address of the GOT
// slot of the referenced symbol.
R_ARM64_GOTPCREL
// PPC64.
// R_POWER_TLS_LE is used to implement the "local exec" model for tls
// access. It resolves to the offset of the thread-local symbol from the
// thread pointer (R13) and inserts this value into the low 16 bits of an
// instruction word.
R_POWER_TLS_LE
// R_POWER_TLS_IE is used to implement the "initial exec" model for tls access. It
// relocates a D-form, DS-form instruction sequence like R_ADDRPOWER_DS. It
// inserts to the offset of GOT slot for the thread-local symbol from the TOC (the
// GOT slot is filled by the dynamic linker with the offset of the thread-local
// symbol from the thread pointer (R13)).
R_POWER_TLS_IE
// R_POWER_TLS marks an X-form instruction such as "MOVD 0(R13)(R31*1), g" as
// accessing a particular thread-local symbol. It does not affect code generation
// but is used by the system linker when relaxing "initial exec" model code to
// "local exec" model code.
R_POWER_TLS
// R_ADDRPOWER_DS is similar to R_ADDRPOWER above, but assumes the second
// instruction is a "DS-form" instruction, which has an immediate field occupying
// bits [15:2] of the instruction word. Bits [15:2] of the address of the
// relocated symbol are inserted into this field; it is an error if the last two
// bits of the address are not 0.
R_ADDRPOWER_DS
cmd/compile, cmd/link, runtime: on ppc64x, maintain the TOC pointer in R2 when compiling PIC The PowerPC ISA does not have a PC-relative load instruction, which poses obvious challenges when generating position-independent code. The way the ELFv2 ABI addresses this is to specify that r2 points to a per "module" (shared library or executable) TOC pointer. Maintaining this pointer requires cooperation between codegen and the system linker: * Non-leaf functions leave space on the stack at r1+24 to save the TOC pointer. * A call to a function that *might* have to go via a PLT stub must be followed by a nop instruction that the system linker can replace with "ld r1, 24(r1)" to restore the TOC pointer (only when dynamically linking Go code). * When calling a function via a function pointer, the address of the function must be in r12, and the first couple of instructions (the "global entry point") of the called function use this to derive the address of the TOC for the module it is in. * When calling a function that is implemented in the same module, the system linker adjusts the call to skip over the instructions mentioned above (the "local entry point"), assuming that r2 is already correctly set. So this changeset adds the global entry point instructions, sets the metadata so the system linker knows where the local entry point is, inserts code to save the TOC pointer at 24(r1), adds a nop after any call not known to be local and copes with the odd non-local code transfer in the runtime (e.g. the stuff around jmpdefer). It does not actually compile PIC yet. Change-Id: I7522e22bdfd2f891745a900c60254fe9e372c854 Reviewed-on: https://go-review.googlesource.com/15967 Reviewed-by: Russ Cox <rsc@golang.org>
2015-10-16 15:42:09 +13:00
// R_ADDRPOWER_PCREL relocates a D-form, DS-form instruction sequence like
// R_ADDRPOWER_DS but inserts the offset of the GOT slot for the referenced symbol
// from the TOC rather than the symbol's address.
R_ADDRPOWER_GOT
cmd/compile, cmd/link, runtime: on ppc64x, maintain the TOC pointer in R2 when compiling PIC The PowerPC ISA does not have a PC-relative load instruction, which poses obvious challenges when generating position-independent code. The way the ELFv2 ABI addresses this is to specify that r2 points to a per "module" (shared library or executable) TOC pointer. Maintaining this pointer requires cooperation between codegen and the system linker: * Non-leaf functions leave space on the stack at r1+24 to save the TOC pointer. * A call to a function that *might* have to go via a PLT stub must be followed by a nop instruction that the system linker can replace with "ld r1, 24(r1)" to restore the TOC pointer (only when dynamically linking Go code). * When calling a function via a function pointer, the address of the function must be in r12, and the first couple of instructions (the "global entry point") of the called function use this to derive the address of the TOC for the module it is in. * When calling a function that is implemented in the same module, the system linker adjusts the call to skip over the instructions mentioned above (the "local entry point"), assuming that r2 is already correctly set. So this changeset adds the global entry point instructions, sets the metadata so the system linker knows where the local entry point is, inserts code to save the TOC pointer at 24(r1), adds a nop after any call not known to be local and copes with the odd non-local code transfer in the runtime (e.g. the stuff around jmpdefer). It does not actually compile PIC yet. Change-Id: I7522e22bdfd2f891745a900c60254fe9e372c854 Reviewed-on: https://go-review.googlesource.com/15967 Reviewed-by: Russ Cox <rsc@golang.org>
2015-10-16 15:42:09 +13:00
// R_ADDRPOWER_PCREL relocates two D-form instructions like R_ADDRPOWER, but
// inserts the displacement from the place being relocated to the address of the
// the relocated symbol instead of just its address.
R_ADDRPOWER_PCREL
// R_ADDRPOWER_TOCREL relocates two D-form instructions like R_ADDRPOWER, but
// inserts the offset from the TOC to the address of the relocated symbol
// rather than the symbol's address.
R_ADDRPOWER_TOCREL
// R_ADDRPOWER_TOCREL relocates a D-form, DS-form instruction sequence like
// R_ADDRPOWER_DS but inserts the offset from the TOC to the address of the the
// relocated symbol rather than the symbol's address.
R_ADDRPOWER_TOCREL_DS
// R_PCRELDBL relocates s390x 2-byte aligned PC-relative addresses.
// TODO(mundaym): remove once variants can be serialized - see issue 14218.
R_PCRELDBL
// R_ADDRMIPSU (only used on mips/mips64) resolves to the sign-adjusted "upper" 16
// bits (bit 16-31) of an external address, by encoding it into the instruction.
R_ADDRMIPSU
// R_ADDRMIPSTLS (only used on mips64) resolves to the low 16 bits of a TLS
// address (offset from thread pointer), by encoding it into the instruction.
R_ADDRMIPSTLS
)
cmd/link: insert trampolines for too-far jumps on ARM ARM direct CALL/JMP instruction has 24 bit offset, which can only encodes jumps within +/-32M. When the target is too far, the top bits get truncated and the program jumps wild. This CL detects too-far jumps and automatically insert trampolines, currently only internal linking on ARM. It is necessary to make the following changes to the linker: - Resolve direct jump relocs when assigning addresses to functions. this allows trampoline insertion without moving all code that already laid down. - Lay down packages in dependency order, so that when resolving a inter-package direct jump reloc, the target address is already known. Intra-package jumps are assumed never too far. - a linker flag -debugtramp is added for debugging trampolines: "-debugtramp=1 -v" prints trampoline debug message "-debugtramp=2" forces all inter-package jump to use trampolines (currently ARM only) "-debugtramp=2 -v" does both - Some data structures are changed for bookkeeping. On ARM, pseudo DIV/DIVU/MOD/MODU instructions now clobber R8 (unfortunate). In the standard library there is no ARM assembly code that uses these instructions, and the compiler no longer emits them (CL 29390). all.bash passes with -debugtramp=2, except a disassembly test (this is unavoidable as we changed the instruction). TBD: debug info of trampolines? Fixes #17028. Change-Id: Idcce347ea7e0af77c4079041a160b2f6e114b474 Reviewed-on: https://go-review.googlesource.com/29397 Reviewed-by: David Crawshaw <crawshaw@golang.org> Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org>
2016-09-14 14:47:12 -04:00
// IsDirectJump returns whether r is a relocation for a direct jump.
// A direct jump is a CALL or JMP instruction that takes the target address
// as immediate. The address is embedded into the instruction, possibly
// with limited width.
// An indirect jump is a CALL or JMP instruction that takes the target address
// in register or memory.
func (r RelocType) IsDirectJump() bool {
switch r {
case R_CALL, R_CALLARM, R_CALLARM64, R_CALLPOWER, R_CALLMIPS, R_JMPMIPS:
return true
}
return false
}
type Auto struct {
Asym *LSym
Aoffset int32
Name AddrName
Gotype *LSym
}
// Auto.name
const (
A_AUTO = 1 + iota
A_PARAM
)
type Pcdata struct {
P []byte
}
// Link holds the context for writing object code from a compiler
// to be linker input or for reading that input into the linker.
type Link struct {
Headtype HeadType
Arch *LinkArch
Debugasm bool
Debugvlog bool
Debugpcln string
Flag_shared bool
Flag_dynlink bool
Flag_optimize bool
Bso *bufio.Writer
Pathname string
Hash map[SymVer]*LSym
2016-12-15 17:17:01 -08:00
PosTable src.PosTable
cmd/compile,link: generate PC-value tables with inlining information In order to generate accurate tracebacks, the runtime needs to know the inlined call stack for a given PC. This creates two tables per function for this purpose. The first table is the inlining tree (stored in the function's funcdata), which has a node containing the file, line, and function name for every inlined call. The second table is a PC-value table that maps each PC to a node in the inlining tree (or -1 if the PC is not the result of inlining). To give the appearance that inlining hasn't happened, the runtime also needs the original source position information of inlined AST nodes. Previously the compiler plastered over the line numbers of inlined AST nodes with the line number of the call. This meant that the PC-line table mapped each PC to line number of the outermost call in its inlined call stack, with no way to access the innermost line number. Now the compiler retains line numbers of inlined AST nodes and writes the innermost source position information to the PC-line and PC-file tables. Some tools and tests expect to see outermost line numbers, so we provide the OutermostLine function for displaying line info. To keep track of the inlined call stack for an AST node, we extend the src.PosBase type with an index into a global inlining tree. Every time the compiler inlines a call, it creates a node in the global inlining tree for the call, and writes its index to the PosBase of every inlined AST node. The parent of this node is the inlining tree index of the call. -1 signifies no parent. For each function, the compiler creates a local inlining tree and a PC-value table mapping each PC to an index in the local tree. These are written to an object file, which is read by the linker. The linker re-encodes these tables compactly by deduplicating function names and file names. This change increases the size of binaries by 4-5%. For example, this is how the go1 benchmark binary is impacted by this change: section old bytes new bytes delta .text 3.49M ± 0% 3.49M ± 0% +0.06% .rodata 1.12M ± 0% 1.21M ± 0% +8.21% .gopclntab 1.50M ± 0% 1.68M ± 0% +11.89% .debug_line 338k ± 0% 435k ± 0% +28.78% Total 9.21M ± 0% 9.58M ± 0% +4.01% Updates #19348. Change-Id: Ic4f180c3b516018138236b0c35e0218270d957d3 Reviewed-on: https://go-review.googlesource.com/37231 Run-TryBot: David Lazar <lazard@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Austin Clements <austin@google.com>
2017-02-17 12:28:05 -05:00
InlTree InlTree // global inlining tree used by gc/inl.go
Imports []string
DiagFunc func(string, ...interface{})
DebugInfo func(fn *LSym, curfn interface{}) []*dwarf.Var // if non-nil, curfn is a *gc.Node
Errors int
Framepointer_enabled bool
// state for writing objects
Text []*LSym
Data []*LSym
}
func (ctxt *Link) Diag(format string, args ...interface{}) {
ctxt.Errors++
ctxt.DiagFunc(format, args...)
}
func (ctxt *Link) Logf(format string, args ...interface{}) {
fmt.Fprintf(ctxt.Bso, format, args...)
ctxt.Bso.Flush()
}
// The smallest possible offset from the hardware stack pointer to a local
// variable on the stack. Architectures that use a link register save its value
// on the stack in the function prologue and so always have a pointer between
// the hardware stack pointer and the local variable area.
func (ctxt *Link) FixedFrameSize() int64 {
switch ctxt.Arch.Family {
case sys.AMD64, sys.I386:
return 0
case sys.PPC64:
// PIC code on ppc64le requires 32 bytes of stack, and it's easier to
// just use that much stack always on ppc64x.
return int64(4 * ctxt.Arch.PtrSize)
default:
return int64(ctxt.Arch.PtrSize)
}
}
type SymVer struct {
Name string
Version int // TODO: make int16 to match LSym.Version?
}
// LinkArch is the definition of a single architecture.
type LinkArch struct {
*sys.Arch
Init func(*Link)
cmd/compile: teach assemblers to accept a Prog allocator The existing bulk Prog allocator is not concurrency-safe. To allow for concurrency-safe bulk allocation of Progs, I want to move Prog allocation and caching upstream, to the clients of cmd/internal/obj. This is a preliminary enabling refactoring. After this CL, instead of calling Ctxt.NewProg throughout the assemblers, we thread through a newprog function that returns a new Prog. That function is set up to be Ctxt.NewProg, so there are no real changes in this CL; this CL only establishes the plumbing. Passes toolstash-check -all. Negligible compiler performance impact. Updates #15756 name old time/op new time/op delta Template 213ms ± 3% 214ms ± 4% ~ (p=0.574 n=49+47) Unicode 90.1ms ± 5% 89.9ms ± 4% ~ (p=0.417 n=50+49) GoTypes 585ms ± 4% 584ms ± 3% ~ (p=0.466 n=49+49) SSA 6.50s ± 3% 6.52s ± 2% ~ (p=0.251 n=49+49) Flate 128ms ± 4% 128ms ± 4% ~ (p=0.673 n=49+50) GoParser 152ms ± 3% 152ms ± 3% ~ (p=0.810 n=48+49) Reflect 372ms ± 4% 372ms ± 5% ~ (p=0.778 n=49+50) Tar 113ms ± 5% 111ms ± 4% -0.98% (p=0.016 n=50+49) XML 208ms ± 3% 208ms ± 2% ~ (p=0.483 n=47+49) [Geo mean] 285ms 285ms -0.17% name old user-ns/op new user-ns/op delta Template 253M ± 8% 254M ± 9% ~ (p=0.899 n=50+50) Unicode 106M ± 9% 106M ±11% ~ (p=0.642 n=50+50) GoTypes 736M ± 4% 740M ± 4% ~ (p=0.121 n=50+49) SSA 8.82G ± 3% 8.88G ± 2% +0.65% (p=0.006 n=49+48) Flate 147M ± 4% 147M ± 5% ~ (p=0.844 n=47+48) GoParser 179M ± 4% 178M ± 6% ~ (p=0.785 n=50+50) Reflect 443M ± 6% 441M ± 5% ~ (p=0.850 n=48+47) Tar 126M ± 5% 126M ± 5% ~ (p=0.734 n=50+50) XML 244M ± 5% 244M ± 5% ~ (p=0.594 n=49+50) [Geo mean] 341M 341M +0.11% Change-Id: Ice962f61eb3a524c2db00a166cb582c22caa7d68 Reviewed-on: https://go-review.googlesource.com/39633 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2017-04-04 14:31:55 -07:00
Preprocess func(*Link, *LSym, ProgAlloc)
Assemble func(*Link, *LSym, ProgAlloc)
Progedit func(*Link, *Prog, ProgAlloc)
UnaryDst map[As]bool // Instruction takes one operand, a destination.
}
// HeadType is the executable header type.
type HeadType uint8
const (
Hunknown HeadType = iota
Hdarwin
Hdragonfly
Hfreebsd
Hlinux
Hnacl
Hnetbsd
Hopenbsd
Hplan9
Hsolaris
Hwindows
)
func (h *HeadType) Set(s string) error {
switch s {
case "darwin":
*h = Hdarwin
case "dragonfly":
*h = Hdragonfly
case "freebsd":
*h = Hfreebsd
case "linux", "android":
*h = Hlinux
case "nacl":
*h = Hnacl
case "netbsd":
*h = Hnetbsd
case "openbsd":
*h = Hopenbsd
case "plan9":
*h = Hplan9
case "solaris":
*h = Hsolaris
case "windows":
*h = Hwindows
default:
return fmt.Errorf("invalid headtype: %q", s)
}
return nil
}
func (h *HeadType) String() string {
switch *h {
case Hdarwin:
return "darwin"
case Hdragonfly:
return "dragonfly"
case Hfreebsd:
return "freebsd"
case Hlinux:
return "linux"
case Hnacl:
return "nacl"
case Hnetbsd:
return "netbsd"
case Hopenbsd:
return "openbsd"
case Hplan9:
return "plan9"
case Hsolaris:
return "solaris"
case Hwindows:
return "windows"
}
return fmt.Sprintf("HeadType(%d)", *h)
}