qbit/go
1
0
Fork 0
mirror of https://github.com/golang/go synced 2024-11-23 00:20:12 -07:00
Code Releases Activity

cmd/compile: add cache of sizeable objects so they can be reused

Browse Source 
We kind of have this mechanism already, just normalizing it and
using it in a bunch of places. Previously a bunch of places cached
slices only for the duration of a single function compilation. Now
we can reuse slices across a whole compiler run.

Use a sync.Pool of powers-of-two sizes. This lets us use not
too much memory, and avoid holding onto memory we're no longer
using when a GC happens.

There's a few different types we need, so generate the code for it.
Generics would be useful here, but we can't use generics in the
compiler because of bootstrapping.

Change-Id: I6cf37e7b7b2e802882aaa723a0b29770511ccd82
Reviewed-on: https://go-review.googlesource.com/c/go/+/444820
Run-TryBot: Keith Randall <khr@golang.org>
Reviewed-by: Heschi Kreinick <heschi@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
This commit is contained in:
Keith Randall 2022-10-18 16:07:36 -07:00 committed by Keith Randall
parent 7ddc45263c
commit 68bd383368
21 changed files with 656 additions and 213 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

198
src/cmd/compile/internal/ssa/_gen/allocators.go Normal file
View File

@ -0,0 +1,198 @@
package main
// TODO: should we share backing storage for similarly-shaped types?
// e.g. []*Value and []*Block, or even []int32 and []bool.
import (
"bytes"
"fmt"
"go/format"
"io"
"log"
"os"
)
type allocator struct {
name string // name for alloc/free functions
typ string // the type they return/accept
mak string // code to make a new object (takes power-of-2 size as fmt arg)
capacity string // code to calculate the capacity of an object. Should always report a power of 2.
resize string // code to shrink to sub-power-of-two size (takes size as fmt arg)
clear string // code for clearing object before putting it on the free list
minLog int // log_2 of minimum allocation size
maxLog int // log_2 of maximum allocation size
}
func genAllocators() {
allocators := []allocator{
{
name: "ValueSlice",
typ: "[]*Value",
capacity: "cap(%s)",
mak: "make([]*Value, %s)",
resize: "%s[:%s]",
clear: "for i := range %[1]s {\n%[1]s[i] = nil\n}",
minLog: 5,
maxLog: 32,
},
{
name: "BlockSlice",
typ: "[]*Block",
capacity: "cap(%s)",
mak: "make([]*Block, %s)",
resize: "%s[:%s]",
clear: "for i := range %[1]s {\n%[1]s[i] = nil\n}",
minLog: 5,
maxLog: 32,
},
{
name: "BoolSlice",
typ: "[]bool",
capacity: "cap(%s)",
mak: "make([]bool, %s)",
resize: "%s[:%s]",
clear: "for i := range %[1]s {\n%[1]s[i] = false\n}",
minLog: 8,
maxLog: 32,
},
{
name: "IntSlice",
typ: "[]int",
capacity: "cap(%s)",
mak: "make([]int, %s)",
resize: "%s[:%s]",
clear: "for i := range %[1]s {\n%[1]s[i] = 0\n}",
minLog: 5,
maxLog: 32,
},
{
name: "Int32Slice",
typ: "[]int32",
capacity: "cap(%s)",
mak: "make([]int32, %s)",
resize: "%s[:%s]",
clear: "for i := range %[1]s {\n%[1]s[i] = 0\n}",
minLog: 6,
maxLog: 32,
},
{
name: "Int8Slice",
typ: "[]int8",
capacity: "cap(%s)",
mak: "make([]int8, %s)",
resize: "%s[:%s]",
clear: "for i := range %[1]s {\n%[1]s[i] = 0\n}",
minLog: 8,
maxLog: 32,
},
{
name: "IDSlice",
typ: "[]ID",
capacity: "cap(%s)",
mak: "make([]ID, %s)",
resize: "%s[:%s]",
clear: "for i := range %[1]s {\n%[1]s[i] = 0\n}",
minLog: 6,
maxLog: 32,
},
{
name: "SparseSet",
typ: "*sparseSet",
capacity: "%s.cap()",
mak: "newSparseSet(%s)",
resize: "", // larger-sized sparse sets are ok
clear: "%s.clear()",
minLog: 5,
maxLog: 32,
},
{
name: "SparseMap",
typ: "*sparseMap",
capacity: "%s.cap()",
mak: "newSparseMap(%s)",
resize: "", // larger-sized sparse maps are ok
clear: "%s.clear()",
minLog: 5,
maxLog: 32,
},
{
name: "SparseMapPos",
typ: "*sparseMapPos",
capacity: "%s.cap()",
mak: "newSparseMapPos(%s)",
resize: "", // larger-sized sparse maps are ok
clear: "%s.clear()",
minLog: 5,
maxLog: 32,
},
}
w := new(bytes.Buffer)
fmt.Fprintf(w, "// Code generated from _gen/allocators.go; DO NOT EDIT.\n")
fmt.Fprintln(w)
fmt.Fprintln(w, "package ssa")
fmt.Fprintln(w, "import (")
fmt.Fprintln(w, "\"math/bits\"")
fmt.Fprintln(w, "\"sync\"")
fmt.Fprintln(w, ")")
for _, a := range allocators {
genAllocator(w, a)
}
// gofmt result
b := w.Bytes()
var err error
b, err = format.Source(b)
if err != nil {
fmt.Printf("%s\n", w.Bytes())
panic(err)
}
if err := os.WriteFile("../allocators.go", b, 0666); err != nil {
log.Fatalf("can't write output: %v\n", err)
}
}
func genAllocator(w io.Writer, a allocator) {
fmt.Fprintf(w, "var poolFree%s [%d]sync.Pool\n", a.name, a.maxLog-a.minLog)
fmt.Fprintf(w, "func (c *Cache) alloc%s(n int) %s {\n", a.name, a.typ)
fmt.Fprintf(w, "var s %s\n", a.typ)
fmt.Fprintf(w, "n2 := n\n")
fmt.Fprintf(w, "if n2 < %d { n2 = %d }\n", 1<<a.minLog, 1<<a.minLog)
fmt.Fprintf(w, "b := bits.Len(uint(n2-1))\n")
fmt.Fprintf(w, "v := poolFree%s[b-%d].Get()\n", a.name, a.minLog)
fmt.Fprintf(w, "if v == nil {\n")
fmt.Fprintf(w, " s = %s\n", fmt.Sprintf(a.mak, "1<<b"))
fmt.Fprintf(w, "} else {\n")
if a.typ[0] == '*' {
fmt.Fprintf(w, "s = v.(%s)\n", a.typ)
} else {
fmt.Fprintf(w, "sp := v.(*%s)\n", a.typ)
fmt.Fprintf(w, "s = *sp\n")
fmt.Fprintf(w, "*sp = nil\n")
fmt.Fprintf(w, "c.hdr%s = append(c.hdr%s, sp)\n", a.name, a.name)
}
fmt.Fprintf(w, "}\n")
if a.resize != "" {
fmt.Fprintf(w, "s = %s\n", fmt.Sprintf(a.resize, "s", "n"))
}
fmt.Fprintf(w, "return s\n")
fmt.Fprintf(w, "}\n")
fmt.Fprintf(w, "func (c *Cache) free%s(s %s) {\n", a.name, a.typ)
fmt.Fprintf(w, "%s\n", fmt.Sprintf(a.clear, "s"))
fmt.Fprintf(w, "b := bits.Len(uint(%s) - 1)\n", fmt.Sprintf(a.capacity, "s"))
if a.typ[0] == '*' {
fmt.Fprintf(w, "poolFree%s[b-%d].Put(s)\n", a.name, a.minLog)
} else {
fmt.Fprintf(w, "var sp *%s\n", a.typ)
fmt.Fprintf(w, "if len(c.hdr%s) == 0 {\n", a.name)
fmt.Fprintf(w, " sp = new(%s)\n", a.typ)
fmt.Fprintf(w, "} else {\n")
fmt.Fprintf(w, " sp = c.hdr%s[len(c.hdr%s)-1]\n", a.name, a.name)
fmt.Fprintf(w, " c.hdr%s[len(c.hdr%s)-1] = nil\n", a.name, a.name)
fmt.Fprintf(w, " c.hdr%s = c.hdr%s[:len(c.hdr%s)-1]\n", a.name, a.name, a.name)
fmt.Fprintf(w, "}\n")
fmt.Fprintf(w, "*sp = s\n")
fmt.Fprintf(w, "poolFree%s[b-%d].Put(sp)\n", a.name, a.minLog)
}
fmt.Fprintf(w, "}\n")
}

1
src/cmd/compile/internal/ssa/_gen/main.go
View File

@ -153,6 +153,7 @@ func main() {
tasks := []func(){
genOp,
genAllocators,
}
for _, a := range archs {
a := a // the funcs are ran concurrently at a later time

343
src/cmd/compile/internal/ssa/allocators.go Normal file
View File

@ -0,0 +1,343 @@
// Code generated from _gen/allocators.go; DO NOT EDIT.
package ssa
import (
"math/bits"
"sync"
)
var poolFreeValueSlice [27]sync.Pool
func (c *Cache) allocValueSlice(n int) []*Value {
var s []*Value
n2 := n
if n2 < 32 {
n2 = 32
}
b := bits.Len(uint(n2 - 1))
v := poolFreeValueSlice[b-5].Get()
if v == nil {
s = make([]*Value, 1<<b)
} else {
sp := v.(*[]*Value)
s = *sp
*sp = nil
c.hdrValueSlice = append(c.hdrValueSlice, sp)
}
s = s[:n]
return s
}
func (c *Cache) freeValueSlice(s []*Value) {
for i := range s {
s[i] = nil
}
b := bits.Len(uint(cap(s)) - 1)
var sp *[]*Value
if len(c.hdrValueSlice) == 0 {
sp = new([]*Value)
} else {
sp = c.hdrValueSlice[len(c.hdrValueSlice)-1]
c.hdrValueSlice[len(c.hdrValueSlice)-1] = nil
c.hdrValueSlice = c.hdrValueSlice[:len(c.hdrValueSlice)-1]
}
*sp = s
poolFreeValueSlice[b-5].Put(sp)
}
var poolFreeBlockSlice [27]sync.Pool
func (c *Cache) allocBlockSlice(n int) []*Block {
var s []*Block
n2 := n
if n2 < 32 {
n2 = 32
}
b := bits.Len(uint(n2 - 1))
v := poolFreeBlockSlice[b-5].Get()
if v == nil {
s = make([]*Block, 1<<b)
} else {
sp := v.(*[]*Block)
s = *sp
*sp = nil
c.hdrBlockSlice = append(c.hdrBlockSlice, sp)
}
s = s[:n]
return s
}
func (c *Cache) freeBlockSlice(s []*Block) {
for i := range s {
s[i] = nil
}
b := bits.Len(uint(cap(s)) - 1)
var sp *[]*Block
if len(c.hdrBlockSlice) == 0 {
sp = new([]*Block)
} else {
sp = c.hdrBlockSlice[len(c.hdrBlockSlice)-1]
c.hdrBlockSlice[len(c.hdrBlockSlice)-1] = nil
c.hdrBlockSlice = c.hdrBlockSlice[:len(c.hdrBlockSlice)-1]
}
*sp = s
poolFreeBlockSlice[b-5].Put(sp)
}
var poolFreeBoolSlice [24]sync.Pool
func (c *Cache) allocBoolSlice(n int) []bool {
var s []bool
n2 := n
if n2 < 256 {
n2 = 256
}
b := bits.Len(uint(n2 - 1))
v := poolFreeBoolSlice[b-8].Get()
if v == nil {
s = make([]bool, 1<<b)
} else {
sp := v.(*[]bool)
s = *sp
*sp = nil
c.hdrBoolSlice = append(c.hdrBoolSlice, sp)
}
s = s[:n]
return s
}
func (c *Cache) freeBoolSlice(s []bool) {
for i := range s {
s[i] = false
}
b := bits.Len(uint(cap(s)) - 1)
var sp *[]bool
if len(c.hdrBoolSlice) == 0 {
sp = new([]bool)
} else {
sp = c.hdrBoolSlice[len(c.hdrBoolSlice)-1]
c.hdrBoolSlice[len(c.hdrBoolSlice)-1] = nil
c.hdrBoolSlice = c.hdrBoolSlice[:len(c.hdrBoolSlice)-1]
}
*sp = s
poolFreeBoolSlice[b-8].Put(sp)
}
var poolFreeIntSlice [27]sync.Pool
func (c *Cache) allocIntSlice(n int) []int {
var s []int
n2 := n
if n2 < 32 {
n2 = 32
}
b := bits.Len(uint(n2 - 1))
v := poolFreeIntSlice[b-5].Get()
if v == nil {
s = make([]int, 1<<b)
} else {
sp := v.(*[]int)
s = *sp
*sp = nil
c.hdrIntSlice = append(c.hdrIntSlice, sp)
}
s = s[:n]
return s
}
func (c *Cache) freeIntSlice(s []int) {
for i := range s {
s[i] = 0
}
b := bits.Len(uint(cap(s)) - 1)
var sp *[]int
if len(c.hdrIntSlice) == 0 {
sp = new([]int)
} else {
sp = c.hdrIntSlice[len(c.hdrIntSlice)-1]
c.hdrIntSlice[len(c.hdrIntSlice)-1] = nil
c.hdrIntSlice = c.hdrIntSlice[:len(c.hdrIntSlice)-1]
}
*sp = s
poolFreeIntSlice[b-5].Put(sp)
}
var poolFreeInt32Slice [26]sync.Pool
func (c *Cache) allocInt32Slice(n int) []int32 {
var s []int32
n2 := n
if n2 < 64 {
n2 = 64
}
b := bits.Len(uint(n2 - 1))
v := poolFreeInt32Slice[b-6].Get()
if v == nil {
s = make([]int32, 1<<b)
} else {
sp := v.(*[]int32)
s = *sp
*sp = nil
c.hdrInt32Slice = append(c.hdrInt32Slice, sp)
}
s = s[:n]
return s
}
func (c *Cache) freeInt32Slice(s []int32) {
for i := range s {
s[i] = 0
}
b := bits.Len(uint(cap(s)) - 1)
var sp *[]int32
if len(c.hdrInt32Slice) == 0 {
sp = new([]int32)
} else {
sp = c.hdrInt32Slice[len(c.hdrInt32Slice)-1]
c.hdrInt32Slice[len(c.hdrInt32Slice)-1] = nil
c.hdrInt32Slice = c.hdrInt32Slice[:len(c.hdrInt32Slice)-1]
}
*sp = s
poolFreeInt32Slice[b-6].Put(sp)
}
var poolFreeInt8Slice [24]sync.Pool
func (c *Cache) allocInt8Slice(n int) []int8 {
var s []int8
n2 := n
if n2 < 256 {
n2 = 256
}
b := bits.Len(uint(n2 - 1))
v := poolFreeInt8Slice[b-8].Get()
if v == nil {
s = make([]int8, 1<<b)
} else {
sp := v.(*[]int8)
s = *sp
*sp = nil
c.hdrInt8Slice = append(c.hdrInt8Slice, sp)
}
s = s[:n]
return s
}
func (c *Cache) freeInt8Slice(s []int8) {
for i := range s {
s[i] = 0
}
b := bits.Len(uint(cap(s)) - 1)
var sp *[]int8
if len(c.hdrInt8Slice) == 0 {
sp = new([]int8)
} else {
sp = c.hdrInt8Slice[len(c.hdrInt8Slice)-1]
c.hdrInt8Slice[len(c.hdrInt8Slice)-1] = nil
c.hdrInt8Slice = c.hdrInt8Slice[:len(c.hdrInt8Slice)-1]
}
*sp = s
poolFreeInt8Slice[b-8].Put(sp)
}
var poolFreeIDSlice [26]sync.Pool
func (c *Cache) allocIDSlice(n int) []ID {
var s []ID
n2 := n
if n2 < 64 {
n2 = 64
}
b := bits.Len(uint(n2 - 1))
v := poolFreeIDSlice[b-6].Get()
if v == nil {
s = make([]ID, 1<<b)
} else {
sp := v.(*[]ID)
s = *sp
*sp = nil
c.hdrIDSlice = append(c.hdrIDSlice, sp)
}
s = s[:n]
return s
}
func (c *Cache) freeIDSlice(s []ID) {
for i := range s {
s[i] = 0
}
b := bits.Len(uint(cap(s)) - 1)
var sp *[]ID
if len(c.hdrIDSlice) == 0 {
sp = new([]ID)
} else {
sp = c.hdrIDSlice[len(c.hdrIDSlice)-1]
c.hdrIDSlice[len(c.hdrIDSlice)-1] = nil
c.hdrIDSlice = c.hdrIDSlice[:len(c.hdrIDSlice)-1]
}
*sp = s
poolFreeIDSlice[b-6].Put(sp)
}
var poolFreeSparseSet [27]sync.Pool
func (c *Cache) allocSparseSet(n int) *sparseSet {
var s *sparseSet
n2 := n
if n2 < 32 {
n2 = 32
}
b := bits.Len(uint(n2 - 1))
v := poolFreeSparseSet[b-5].Get()
if v == nil {
s = newSparseSet(1 << b)
} else {
s = v.(*sparseSet)
}
return s
}
func (c *Cache) freeSparseSet(s *sparseSet) {
s.clear()
b := bits.Len(uint(s.cap()) - 1)
poolFreeSparseSet[b-5].Put(s)
}
var poolFreeSparseMap [27]sync.Pool
func (c *Cache) allocSparseMap(n int) *sparseMap {
var s *sparseMap
n2 := n
if n2 < 32 {
n2 = 32
}
b := bits.Len(uint(n2 - 1))
v := poolFreeSparseMap[b-5].Get()
if v == nil {
s = newSparseMap(1 << b)
} else {
s = v.(*sparseMap)
}
return s
}
func (c *Cache) freeSparseMap(s *sparseMap) {
s.clear()
b := bits.Len(uint(s.cap()) - 1)
poolFreeSparseMap[b-5].Put(s)
}
var poolFreeSparseMapPos [27]sync.Pool
func (c *Cache) allocSparseMapPos(n int) *sparseMapPos {
var s *sparseMapPos
n2 := n
if n2 < 32 {
n2 = 32
}
b := bits.Len(uint(n2 - 1))
v := poolFreeSparseMapPos[b-5].Get()
if v == nil {
s = newSparseMapPos(1 << b)
} else {
s = v.(*sparseMapPos)
}
return s
}
func (c *Cache) freeSparseMapPos(s *sparseMapPos) {
s.clear()
b := bits.Len(uint(s.cap()) - 1)
poolFreeSparseMapPos[b-5].Put(s)
}

39
src/cmd/compile/internal/ssa/cache.go
View File

@ -21,18 +21,8 @@ type Cache struct {
// See stackalloc.go's {new,put}StackAllocState.
stackAllocState *stackAllocState
domblockstore []ID // scratch space for computing dominators
scrSparseSet []*sparseSet // scratch sparse sets to be re-used.
scrSparseMap []*sparseMap // scratch sparse maps to be re-used.
scrSparseMapPos []*sparseMapPos // scratch sparse maps to be re-used.
scrPoset []*poset // scratch poset to be reused
// deadcode contains reusable slices specifically for the deadcode pass.
// It gets special treatment because of the frequency with which it is run.
deadcode struct {
liveOrderStmts []*Value
live []bool
q []*Value
}
scrPoset []*poset // scratch poset to be reused
// Reusable regalloc state.
regallocValues []valState
@ -40,6 +30,16 @@ type Cache struct {
debugState debugState
Liveness interface{} // *gc.livenessFuncCache
// Free "headers" for use by the allocators in allocators.go.
// Used to put slices in sync.Pools without allocation.
hdrValueSlice []*[]*Value
hdrBlockSlice []*[]*Block
hdrBoolSlice []*[]bool
hdrIntSlice []*[]int
hdrInt32Slice []*[]int32
hdrInt8Slice []*[]int8
hdrIDSlice []*[]ID
}
func (c *Cache) Reset() {
@ -64,19 +64,4 @@ func (c *Cache) Reset() {
for i := range c.regallocValues {
c.regallocValues[i] = valState{}
}
// liveOrderStmts gets used multiple times during compilation of a function.
// We don't know where the high water mark was, so reslice to cap and search.
c.deadcode.liveOrderStmts = c.deadcode.liveOrderStmts[:cap(c.deadcode.liveOrderStmts)]
no := sort.Search(len(c.deadcode.liveOrderStmts), func(i int) bool { return c.deadcode.liveOrderStmts[i] == nil })
xo := c.deadcode.liveOrderStmts[:no]
for i := range xo {
xo[i] = nil
}
c.deadcode.q = c.deadcode.q[:cap(c.deadcode.q)]
nq := sort.Search(len(c.deadcode.q), func(i int) bool { return c.deadcode.q[i] == nil })
xq := c.deadcode.q[:nq]
for i := range xq {
xq[i] = nil
}
}

3
src/cmd/compile/internal/ssa/critical.go
View File

@ -9,7 +9,8 @@ package ssa
// Regalloc wants a critical-edge-free CFG so it can implement phi values.
func critical(f *Func) {
// maps from phi arg ID to the new block created for that argument
blocks := make([]*Block, f.NumValues())
blocks := f.Cache.allocBlockSlice(f.NumValues())
defer f.Cache.freeBlockSlice(blocks)
// need to iterate over f.Blocks without range, as we might
// need to split critical edges on newly constructed blocks
for j := 0; j < len(f.Blocks); j++ {

10
src/cmd/compile/internal/ssa/cse.go
View File

@ -31,7 +31,9 @@ func cse(f *Func) {
// until it reaches a fixed point.
// Make initial coarse partitions by using a subset of the conditions above.
a := make([]*Value, 0, f.NumValues())
a := f.Cache.allocValueSlice(f.NumValues())
defer func() { f.Cache.freeValueSlice(a) }() // inside closure to use final value of a
a = a[:0]
if f.auxmap == nil {
f.auxmap = auxmap{}
}
@ -49,7 +51,8 @@ func cse(f *Func) {
partition := partitionValues(a, f.auxmap)
// map from value id back to eqclass id
valueEqClass := make([]ID, f.NumValues())
valueEqClass := f.Cache.allocIDSlice(f.NumValues())
defer f.Cache.freeIDSlice(valueEqClass)
for _, b := range f.Blocks {
for _, v := range b.Values {
// Use negative equivalence class #s for unique values.
@ -159,7 +162,8 @@ func cse(f *Func) {
// Compute substitutions we would like to do. We substitute v for w
// if v and w are in the same equivalence class and v dominates w.
rewrite := make([]*Value, f.NumValues())
rewrite := f.Cache.allocValueSlice(f.NumValues())
defer f.Cache.freeValueSlice(rewrite)
byDom := new(partitionByDom) // reusable partitionByDom to reduce allocs
for _, e := range partition {
byDom.a = e

29
src/cmd/compile/internal/ssa/deadcode.go
View File

@ -9,12 +9,12 @@ import (
)
// findlive returns the reachable blocks and live values in f.
// The caller should call f.retDeadcodeLive(live) when it is done with it.
// The caller should call f.Cache.freeBoolSlice(live) when it is done with it.
func findlive(f *Func) (reachable []bool, live []bool) {
reachable = ReachableBlocks(f)
var order []*Value
live, order = liveValues(f, reachable)
f.retDeadcodeLiveOrderStmts(order)
f.Cache.freeValueSlice(order)
return
}
@ -51,21 +51,11 @@ func ReachableBlocks(f *Func) []bool {
// to be statements in reversed data flow order.
// The second result is used to help conserve statement boundaries for debugging.
// reachable is a map from block ID to whether the block is reachable.
// The caller should call f.retDeadcodeLive(live) and f.retDeadcodeLiveOrderStmts(liveOrderStmts)
// The caller should call f.Cache.freeBoolSlice(live) and f.Cache.freeValueSlice(liveOrderStmts).
// when they are done with the return values.
func liveValues(f *Func, reachable []bool) (live []bool, liveOrderStmts []*Value) {
live = f.newDeadcodeLive()
if cap(live) < f.NumValues() {
live = make([]bool, f.NumValues())
} else {
live = live[:f.NumValues()]
for i := range live {
live[i] = false
}
}
liveOrderStmts = f.newDeadcodeLiveOrderStmts()
liveOrderStmts = liveOrderStmts[:0]
live = f.Cache.allocBoolSlice(f.NumValues())
liveOrderStmts = f.Cache.allocValueSlice(f.NumValues())[:0]
// After regalloc, consider all values to be live.
// See the comment at the top of regalloc.go and in deadcode for details.
@ -101,8 +91,8 @@ func liveValues(f *Func, reachable []bool) (live []bool, liveOrderStmts []*Value
}
// Find all live values
q := f.Cache.deadcode.q[:0]
defer func() { f.Cache.deadcode.q = q }()
q := f.Cache.allocValueSlice(f.NumValues())[:0]
defer f.Cache.freeValueSlice(q)
// Starting set: all control values of reachable blocks are live.
// Calls are live (because callee can observe the memory state).
@ -149,6 +139,7 @@ func liveValues(f *Func, reachable []bool) (live []bool, liveOrderStmts []*Value
for len(q) > 0 {
// pop a reachable value
v := q[len(q)-1]
q[len(q)-1] = nil
q = q[:len(q)-1]
for i, x := range v.Args {
if v.Op == OpPhi && !reachable[v.Block.Preds[i].b.ID] {
@ -213,8 +204,8 @@ func deadcode(f *Func) {
// Find live values.
live, order := liveValues(f, reachable)
defer f.retDeadcodeLive(live)
defer f.retDeadcodeLiveOrderStmts(order)
defer func() { f.Cache.freeBoolSlice(live) }()
defer func() { f.Cache.freeValueSlice(order) }()
// Remove dead & duplicate entries from namedValues map.
s := f.newSparseSet(f.NumValues())

57
src/cmd/compile/internal/ssa/dom.go
View File

@ -21,7 +21,8 @@ type blockAndIndex struct {
// postorderWithNumbering provides a DFS postordering.
// This seems to make loop-finding more robust.
func postorderWithNumbering(f *Func, ponums []int32) []*Block {
seen := make([]bool, f.NumBlocks())
seen := f.Cache.allocBoolSlice(f.NumBlocks())
defer f.Cache.freeBoolSlice(seen)
// result ordering
order := make([]*Block, 0, len(f.Blocks))
@ -56,44 +57,6 @@ func postorderWithNumbering(f *Func, ponums []int32) []*Block {
type linkedBlocks func(*Block) []Edge
const nscratchslices = 7
// experimentally, functions with 512 or fewer blocks account
// for 75% of memory (size) allocation for dominator computation
// in make.bash.
const minscratchblocks = 512
func (cache *Cache) scratchBlocksForDom(maxBlockID int) (a, b, c, d, e, f, g []ID) {
tot := maxBlockID * nscratchslices
scratch := cache.domblockstore
if len(scratch) < tot {
// req = min(1.5*tot, nscratchslices*minscratchblocks)
// 50% padding allows for graph growth in later phases.
req := (tot * 3) >> 1
if req < nscratchslices*minscratchblocks {
req = nscratchslices * minscratchblocks
}
scratch = make([]ID, req)
cache.domblockstore = scratch
} else {
// Clear as much of scratch as we will (re)use
scratch = scratch[0:tot]
for i := range scratch {
scratch[i] = 0
}
}
a = scratch[0*maxBlockID : 1*maxBlockID]
b = scratch[1*maxBlockID : 2*maxBlockID]
c = scratch[2*maxBlockID : 3*maxBlockID]
d = scratch[3*maxBlockID : 4*maxBlockID]
e = scratch[4*maxBlockID : 5*maxBlockID]
f = scratch[5*maxBlockID : 6*maxBlockID]
g = scratch[6*maxBlockID : 7*maxBlockID]
return
}
func dominators(f *Func) []*Block {
preds := func(b *Block) []Edge { return b.Preds }
succs := func(b *Block) []Edge { return b.Succs }
@ -110,12 +73,21 @@ func (f *Func) dominatorsLTOrig(entry *Block, predFn linkedBlocks, succFn linked
// Adapted directly from the original TOPLAS article's "simple" algorithm
maxBlockID := entry.Func.NumBlocks()
semi, vertex, label, parent, ancestor, bucketHead, bucketLink := f.Cache.scratchBlocksForDom(maxBlockID)
scratch := f.Cache.allocIDSlice(7 * maxBlockID)
defer f.Cache.freeIDSlice(scratch)
semi := scratch[0*maxBlockID : 1*maxBlockID]
vertex := scratch[1*maxBlockID : 2*maxBlockID]
label := scratch[2*maxBlockID : 3*maxBlockID]
parent := scratch[3*maxBlockID : 4*maxBlockID]
ancestor := scratch[4*maxBlockID : 5*maxBlockID]
bucketHead := scratch[5*maxBlockID : 6*maxBlockID]
bucketLink := scratch[6*maxBlockID : 7*maxBlockID]
// This version uses integers for most of the computation,
// to make the work arrays smaller and pointer-free.
// fromID translates from ID to *Block where that is needed.
fromID := make([]*Block, maxBlockID)
fromID := f.Cache.allocBlockSlice(maxBlockID)
defer f.Cache.freeBlockSlice(fromID)
for _, v := range f.Blocks {
fromID[v.ID] = v
}
@ -243,7 +215,8 @@ func dominatorsSimple(f *Func) []*Block {
post := f.postorder()
// Make map from block id to order index (for intersect call)
postnum := make([]int, f.NumBlocks())
postnum := f.Cache.allocIntSlice(f.NumBlocks())
defer f.Cache.freeIntSlice(postnum)
for i, b := range post {
postnum[b.ID] = i
}

3
src/cmd/compile/internal/ssa/flagalloc.go
View File

@ -11,7 +11,8 @@ func flagalloc(f *Func) {
// Compute the in-register flag value we want at the end of
// each block. This is basically a best-effort live variable
// analysis, so it can be much simpler than a full analysis.
end := make([]*Value, f.NumBlocks())
end := f.Cache.allocValueSlice(f.NumBlocks())
defer f.Cache.freeValueSlice(end)
po := f.postorder()
for n := 0; n < 2; n++ {
for _, b := range po {

78
src/cmd/compile/internal/ssa/func.go
View File

@ -105,74 +105,35 @@ func (f *Func) NumValues() int {
// newSparseSet returns a sparse set that can store at least up to n integers.
func (f *Func) newSparseSet(n int) *sparseSet {
for i, scr := range f.Cache.scrSparseSet {
if scr != nil && scr.cap() >= n {
f.Cache.scrSparseSet[i] = nil
scr.clear()
return scr
}
}
return newSparseSet(n)
return f.Cache.allocSparseSet(n)
}
// retSparseSet returns a sparse set to the config's cache of sparse
// sets to be reused by f.newSparseSet.
func (f *Func) retSparseSet(ss *sparseSet) {
for i, scr := range f.Cache.scrSparseSet {
if scr == nil {
f.Cache.scrSparseSet[i] = ss
return
}
}
f.Cache.scrSparseSet = append(f.Cache.scrSparseSet, ss)
f.Cache.freeSparseSet(ss)
}
// newSparseMap returns a sparse map that can store at least up to n integers.
func (f *Func) newSparseMap(n int) *sparseMap {
for i, scr := range f.Cache.scrSparseMap {
if scr != nil && scr.cap() >= n {
f.Cache.scrSparseMap[i] = nil
scr.clear()
return scr
}
}
return newSparseMap(n)
return f.Cache.allocSparseMap(n)
}
// retSparseMap returns a sparse map to the config's cache of sparse
// sets to be reused by f.newSparseMap.
func (f *Func) retSparseMap(ss *sparseMap) {
for i, scr := range f.Cache.scrSparseMap {
if scr == nil {
f.Cache.scrSparseMap[i] = ss
return
}
}
f.Cache.scrSparseMap = append(f.Cache.scrSparseMap, ss)
f.Cache.freeSparseMap(ss)
}
// newSparseMapPos returns a sparse map that can store at least up to n integers.
func (f *Func) newSparseMapPos(n int) *sparseMapPos {
for i, scr := range f.Cache.scrSparseMapPos {
if scr != nil && scr.cap() >= n {
f.Cache.scrSparseMapPos[i] = nil
scr.clear()
return scr
}
}
return newSparseMapPos(n)
return f.Cache.allocSparseMapPos(n)
}
// retSparseMapPos returns a sparse map to the config's cache of sparse
// sets to be reused by f.newSparseMapPos.
func (f *Func) retSparseMapPos(ss *sparseMapPos) {
for i, scr := range f.Cache.scrSparseMapPos {
if scr == nil {
f.Cache.scrSparseMapPos[i] = ss
return
}
}
f.Cache.scrSparseMapPos = append(f.Cache.scrSparseMapPos, ss)
f.Cache.freeSparseMapPos(ss)
}
// newPoset returns a new poset from the internal cache
@ -190,33 +151,6 @@ func (f *Func) retPoset(po *poset) {
f.Cache.scrPoset = append(f.Cache.scrPoset, po)
}
// newDeadcodeLive returns a slice for the
// deadcode pass to use to indicate which values are live.
func (f *Func) newDeadcodeLive() []bool {
r := f.Cache.deadcode.live
f.Cache.deadcode.live = nil
return r
}
// retDeadcodeLive returns a deadcode live value slice for re-use.
func (f *Func) retDeadcodeLive(live []bool) {
f.Cache.deadcode.live = live
}
// newDeadcodeLiveOrderStmts returns a slice for the
// deadcode pass to use to indicate which values
// need special treatment for statement boundaries.
func (f *Func) newDeadcodeLiveOrderStmts() []*Value {
r := f.Cache.deadcode.liveOrderStmts
f.Cache.deadcode.liveOrderStmts = nil
return r
}
// retDeadcodeLiveOrderStmts returns a deadcode liveOrderStmts slice for re-use.
func (f *Func) retDeadcodeLiveOrderStmts(liveOrderStmts []*Value) {
f.Cache.deadcode.liveOrderStmts = liveOrderStmts
}
func (f *Func) localSlotAddr(slot LocalSlot) *LocalSlot {
a, ok := f.CanonicalLocalSlots[slot]
if !ok {

9
src/cmd/compile/internal/ssa/layout.go
View File

@ -20,9 +20,12 @@ func layoutRegallocOrder(f *Func) []*Block {
func layoutOrder(f *Func) []*Block {
order := make([]*Block, 0, f.NumBlocks())
scheduled := make([]bool, f.NumBlocks())
idToBlock := make([]*Block, f.NumBlocks())
indegree := make([]int, f.NumBlocks())
scheduled := f.Cache.allocBoolSlice(f.NumBlocks())
defer f.Cache.freeBoolSlice(scheduled)
idToBlock := f.Cache.allocBlockSlice(f.NumBlocks())
defer f.Cache.freeBlockSlice(idToBlock)
indegree := f.Cache.allocIntSlice(f.NumBlocks())
defer f.Cache.freeIntSlice(indegree)
posdegree := f.newSparseSet(f.NumBlocks()) // blocks with positive remaining degree
defer f.retSparseSet(posdegree)
// blocks with zero remaining degree. Use slice to simulate a LIFO queue to implement

12
src/cmd/compile/internal/ssa/likelyadjust.go
View File

@ -117,8 +117,10 @@ func likelyadjust(f *Func) {
// in their rank order. 0 is default, more positive
// is less likely. It's possible to assign a negative
// unlikeliness (though not currently the case).
certain := make([]int8, f.NumBlocks()) // In the long run, all outcomes are at least this bad. Mainly for Exit
local := make([]int8, f.NumBlocks()) // for our immediate predecessors.
certain := f.Cache.allocInt8Slice(f.NumBlocks()) // In the long run, all outcomes are at least this bad. Mainly for Exit
defer f.Cache.freeInt8Slice(certain)
local := f.Cache.allocInt8Slice(f.NumBlocks()) // for our immediate predecessors.
defer f.Cache.freeInt8Slice(local)
po := f.postorder()
nest := f.loopnest()
@ -277,7 +279,8 @@ func loopnestfor(f *Func) *loopnest {
sdom := f.Sdom()
b2l := make([]*loop, f.NumBlocks())
loops := make([]*loop, 0)
visited := make([]bool, f.NumBlocks())
visited := f.Cache.allocBoolSlice(f.NumBlocks())
defer f.Cache.freeBoolSlice(visited)
sawIrred := false
if f.pass.debug > 2 {
@ -369,7 +372,8 @@ func loopnestfor(f *Func) *loopnest {
ln := &loopnest{f: f, b2l: b2l, po: po, sdom: sdom, loops: loops, hasIrreducible: sawIrred}
// Calculate containsUnavoidableCall for regalloc
dominatedByCall := make([]bool, f.NumBlocks())
dominatedByCall := f.Cache.allocBoolSlice(f.NumBlocks())
defer f.Cache.freeBoolSlice(dominatedByCall)
for _, b := range po {
if checkContainsCall(b) {
dominatedByCall[b.ID] = true

6
src/cmd/compile/internal/ssa/loopreschedchecks.go
View File

@ -94,7 +94,8 @@ func insertLoopReschedChecks(f *Func) {
lastMems[f.Entry.ID] = f.Entry.NewValue0(f.Entry.Pos, OpInitMem, types.TypeMem)
}
memDefsAtBlockEnds := make([]*Value, f.NumBlocks()) // For each block, the mem def seen at its bottom. Could be from earlier block.
memDefsAtBlockEnds := f.Cache.allocValueSlice(f.NumBlocks()) // For each block, the mem def seen at its bottom. Could be from earlier block.
defer f.Cache.freeValueSlice(memDefsAtBlockEnds)
// Propagate last mem definitions forward through successor blocks.
for i := len(po) - 1; i >= 0; i-- {
@ -404,7 +405,8 @@ outer:
func findLastMems(f *Func) []*Value {
var stores []*Value
lastMems := make([]*Value, f.NumBlocks())
lastMems := f.Cache.allocValueSlice(f.NumBlocks())
defer f.Cache.freeValueSlice(lastMems)
storeUse := f.newSparseSet(f.NumValues())
defer f.retSparseSet(storeUse)
for _, b := range f.Blocks {

16
src/cmd/compile/internal/ssa/looprotate.go
View File

@ -30,7 +30,8 @@ func loopRotate(f *Func) {
return
}
idToIdx := make([]int, f.NumBlocks())
idToIdx := f.Cache.allocIntSlice(f.NumBlocks())
defer f.Cache.freeIntSlice(idToIdx)
for i, b := range f.Blocks {
idToIdx[b.ID] = i
}
@ -92,20 +93,21 @@ func loopRotate(f *Func) {
// Some blocks that are not part of a loop may be placed
// between loop blocks. In order to avoid these blocks from
// being overwritten, use a temporary slice.
newOrder := make([]*Block, 0, f.NumBlocks())
for _, b := range f.Blocks {
oldOrder := f.Cache.allocBlockSlice(len(f.Blocks))
defer f.Cache.freeBlockSlice(oldOrder)
copy(oldOrder, f.Blocks)
for _, b := range oldOrder {
if _, ok := move[b.ID]; ok {
continue
}
newOrder = append(newOrder, b)
f.Blocks[j] = b
j++
for _, a := range after[b.ID] {
newOrder = append(newOrder, a)
f.Blocks[j] = a
j++
}
}
if j != len(f.Blocks) {
if j != len(oldOrder) {
f.Fatalf("bad reordering in looprotate")
}
f.Blocks = newOrder
}

6
src/cmd/compile/internal/ssa/nilcheck.go
View File

@ -41,7 +41,8 @@ func nilcheckelim(f *Func) {
// map from value ID to bool indicating if value is known to be non-nil
// in the current dominator path being walked. This slice is updated by
// walkStates to maintain the known non-nil values.
nonNilValues := make([]bool, f.NumValues())
nonNilValues := f.Cache.allocBoolSlice(f.NumValues())
defer f.Cache.freeBoolSlice(nonNilValues)
// make an initial pass identifying any non-nil values
for _, b := range f.Blocks {
@ -86,7 +87,8 @@ func nilcheckelim(f *Func) {
// allocate auxiliary date structures for computing store order
sset := f.newSparseSet(f.NumValues())
defer f.retSparseSet(sset)
storeNumber := make([]int32, f.NumValues())
storeNumber := f.Cache.allocInt32Slice(f.NumValues())
defer f.Cache.freeInt32Slice(storeNumber)
// perform a depth first walk of the dominee tree
for len(work) > 0 {

2
src/cmd/compile/internal/ssa/print.go
View File

@ -124,7 +124,7 @@ func (p stringFuncPrinter) named(n LocalSlot, vals []*Value) {
func fprintFunc(p funcPrinter, f *Func) {
reachable, live := findlive(f)
defer f.retDeadcodeLive(live)
defer f.Cache.freeBoolSlice(live)
p.header(f)
printed := make([]bool, f.NumValues())
for _, b := range f.Blocks {

13
src/cmd/compile/internal/ssa/regalloc.go
View File

@ -146,6 +146,7 @@ func regalloc(f *Func) {
var s regAllocState
s.init(f)
s.regalloc(f)
s.close()
}
type register uint8
@ -357,6 +358,12 @@ func (s *regAllocState) clobberRegs(m regMask) {
// setOrig records that c's original value is the same as
// v's original value.
func (s *regAllocState) setOrig(c *Value, v *Value) {
if int(c.ID) >= cap(s.orig) {
x := s.f.Cache.allocValueSlice(int(c.ID) + 1)
copy(x, s.orig)
s.f.Cache.freeValueSlice(s.orig)
s.orig = x
}
for int(c.ID) >= len(s.orig) {
s.orig = append(s.orig, nil)
}
@ -664,7 +671,7 @@ func (s *regAllocState) init(f *Func) {
s.f.Cache.regallocValues = make([]valState, nv)
}
s.values = s.f.Cache.regallocValues
s.orig = make([]*Value, nv)
s.orig = s.f.Cache.allocValueSlice(nv)
s.copies = make(map[*Value]bool)
for _, b := range s.visitOrder {
for _, v := range b.Values {
@ -728,6 +735,10 @@ func (s *regAllocState) init(f *Func) {
}
}
func (s *regAllocState) close() {
s.f.Cache.freeValueSlice(s.orig)
}
// Adds a use record for id at distance dist from the start of the block.
// All calls to addUse must happen with nonincreasing dist.
func (s *regAllocState) addUse(id ID, dist int32, pos src.XPos) {

9
src/cmd/compile/internal/ssa/schedule.go
View File

@ -94,13 +94,15 @@ func (op Op) isLoweredGetClosurePtr() bool {
func schedule(f *Func) {
// For each value, the number of times it is used in the block
// by values that have not been scheduled yet.
uses := make([]int32, f.NumValues())
uses := f.Cache.allocInt32Slice(f.NumValues())
defer f.Cache.freeInt32Slice(uses)
// reusable priority queue
priq := new(ValHeap)
// "priority" for a value
score := make([]int8, f.NumValues())
score := f.Cache.allocInt8Slice(f.NumValues())
defer f.Cache.freeInt8Slice(score)
// scheduling order. We queue values in this list in reverse order.
// A constant bound allows this to be stack-allocated. 64 is
@ -108,7 +110,8 @@ func schedule(f *Func) {
order := make([]*Value, 0, 64)
// maps mem values to the next live memory value
nextMem := make([]*Value, f.NumValues())
nextMem := f.Cache.allocValueSlice(f.NumValues())
defer f.Cache.freeValueSlice(nextMem)
// additional pretend arguments for each Value. Used to enforce load/store ordering.
additionalArgs := make([][]*Value, f.NumValues())

26
src/cmd/compile/internal/ssa/stackalloc.go
View File

@ -25,8 +25,6 @@ type stackAllocState struct {
values []stackValState
interfere [][]ID // interfere[v.id] = values that interfere with v.
names []LocalSlot
slots []int
used []bool
nArgSlot, // Number of Values sourced to arg slot
nNotNeed, // Number of Values not needing a stack slot
@ -57,12 +55,6 @@ func putStackAllocState(s *stackAllocState) {
for i := range s.names {
s.names[i] = LocalSlot{}
}
for i := range s.slots {
s.slots[i] = 0
}
for i := range s.used {
s.used[i] = false
}
s.f.Cache.stackAllocState = s
s.f = nil
s.live = nil
@ -218,25 +210,15 @@ func (s *stackAllocState) stackalloc() {
// Each time we assign a stack slot to a value v, we remember
// the slot we used via an index into locations[v.Type].
slots := s.slots
if n := f.NumValues(); cap(slots) >= n {
slots = slots[:n]
} else {
slots = make([]int, n)
s.slots = slots
}
slots := f.Cache.allocIntSlice(f.NumValues())
defer f.Cache.freeIntSlice(slots)
for i := range slots {
slots[i] = -1
}
// Pick a stack slot for each value needing one.
var used []bool
if n := f.NumValues(); cap(s.used) >= n {
used = s.used[:n]
} else {
used = make([]bool, n)
s.used = used
}
used := f.Cache.allocBoolSlice(f.NumValues())
defer f.Cache.freeBoolSlice(used)
for _, b := range f.Blocks {
for _, v := range b.Values {
if !s.values[v.ID].needSlot {

6
src/cmd/compile/internal/ssa/tighten.go
View File

@ -10,7 +10,8 @@ package ssa
// A Value can be moved to any block that
// dominates all blocks in which it is used.
func tighten(f *Func) {
canMove := make([]bool, f.NumValues())
canMove := f.Cache.allocBoolSlice(f.NumValues())
defer f.Cache.freeBoolSlice(canMove)
for _, b := range f.Blocks {
for _, v := range b.Values {
if v.Op.isLoweredGetClosurePtr() {
@ -52,7 +53,8 @@ func tighten(f *Func) {
lca := makeLCArange(f)
// For each moveable value, record the block that dominates all uses found so far.
target := make([]*Block, f.NumValues())
target := f.Cache.allocBlockSlice(f.NumValues())
defer f.Cache.freeBlockSlice(target)
// Grab loop information.
// We use this to make sure we don't tighten a value into a (deeper) loop.

3
src/cmd/compile/internal/ssa/writebarrier.go
View File

@ -139,7 +139,8 @@ func writebarrier(f *Func) {
// allocate auxiliary data structures for computing store order
sset = f.newSparseSet(f.NumValues())
defer f.retSparseSet(sset)
storeNumber = make([]int32, f.NumValues())
storeNumber = f.Cache.allocInt32Slice(f.NumValues())
defer f.Cache.freeInt32Slice(storeNumber)
}
// order values in store order