1
0
mirror of https://github.com/golang/go synced 2024-11-19 21:14:43 -07:00
go/src/runtime/hashmap_fast.go

423 lines
12 KiB
Go
Raw Normal View History

// Copyright 2014 The Go Authors. All rights reserved.
// Use of this source code is governed by a BSD-style
// license that can be found in the LICENSE file.
package runtime
import (
"runtime/internal/sys"
"unsafe"
)
func mapaccess1_fast32(t *maptype, h *hmap, key uint32) unsafe.Pointer {
if raceenabled && h != nil {
callerpc := getcallerpc(unsafe.Pointer(&t))
racereadpc(unsafe.Pointer(h), callerpc, funcPC(mapaccess1_fast32))
}
if h == nil || h.count == 0 {
return unsafe.Pointer(&zeroVal[0])
}
if h.flags&hashWriting != 0 {
throw("concurrent map read and map write")
}
var b *bmap
if h.B == 0 {
// One-bucket table. No need to hash.
b = (*bmap)(h.buckets)
} else {
hash := t.key.alg.hash(noescape(unsafe.Pointer(&key)), uintptr(h.hash0))
m := uintptr(1)<<h.B - 1
b = (*bmap)(add(h.buckets, (hash&m)*uintptr(t.bucketsize)))
if c := h.oldbuckets; c != nil {
runtime: limit the number of map overflow buckets Consider repeatedly adding many items to a map and then deleting them all, as in #16070. The map itself doesn't need to grow above the high water mark of number of items. However, due to random collisions, the map can accumulate overflow buckets. Prior to this CL, those overflow buckets were never removed, which led to a slow memory leak. The problem with removing overflow buckets is iterators. The obvious approach is to repack keys and values and eliminate unused overflow buckets. However, keys, values, and overflow buckets cannot be manipulated without disrupting iterators. This CL takes a different approach, which is to reuse the existing map growth mechanism, which is well established, well tested, and safe in the presence of iterators. When a map has accumulated enough overflow buckets we trigger map growth, but grow into a map of the same size as before. The old overflow buckets will be left behind for garbage collection. For the code in #16070, instead of climbing (very slowly) forever, memory usage now cycles between 264mb and 483mb every 15 minutes or so. To avoid increasing the size of maps, the overflow bucket counter is only 16 bits. For large maps, the counter is incremented stochastically. Fixes #16070 Change-Id: If551d77613ec6836907efca58bda3deee304297e Reviewed-on: https://go-review.googlesource.com/25049 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2016-07-19 12:47:53 -06:00
if !h.sameSizeGrow() {
// There used to be half as many buckets; mask down one more power of two.
m >>= 1
}
oldb := (*bmap)(add(c, (hash&m)*uintptr(t.bucketsize)))
if !evacuated(oldb) {
b = oldb
}
}
}
for {
for i := uintptr(0); i < bucketCnt; i++ {
k := *((*uint32)(add(unsafe.Pointer(b), dataOffset+i*4)))
if k != key {
continue
}
x := *((*uint8)(add(unsafe.Pointer(b), i))) // b.topbits[i] without the bounds check
if x == empty {
continue
}
return add(unsafe.Pointer(b), dataOffset+bucketCnt*4+i*uintptr(t.valuesize))
}
b = b.overflow(t)
if b == nil {
return unsafe.Pointer(&zeroVal[0])
}
}
}
func mapaccess2_fast32(t *maptype, h *hmap, key uint32) (unsafe.Pointer, bool) {
if raceenabled && h != nil {
callerpc := getcallerpc(unsafe.Pointer(&t))
racereadpc(unsafe.Pointer(h), callerpc, funcPC(mapaccess2_fast32))
}
if h == nil || h.count == 0 {
return unsafe.Pointer(&zeroVal[0]), false
}
if h.flags&hashWriting != 0 {
throw("concurrent map read and map write")
}
var b *bmap
if h.B == 0 {
// One-bucket table. No need to hash.
b = (*bmap)(h.buckets)
} else {
hash := t.key.alg.hash(noescape(unsafe.Pointer(&key)), uintptr(h.hash0))
m := uintptr(1)<<h.B - 1
b = (*bmap)(add(h.buckets, (hash&m)*uintptr(t.bucketsize)))
if c := h.oldbuckets; c != nil {
runtime: limit the number of map overflow buckets Consider repeatedly adding many items to a map and then deleting them all, as in #16070. The map itself doesn't need to grow above the high water mark of number of items. However, due to random collisions, the map can accumulate overflow buckets. Prior to this CL, those overflow buckets were never removed, which led to a slow memory leak. The problem with removing overflow buckets is iterators. The obvious approach is to repack keys and values and eliminate unused overflow buckets. However, keys, values, and overflow buckets cannot be manipulated without disrupting iterators. This CL takes a different approach, which is to reuse the existing map growth mechanism, which is well established, well tested, and safe in the presence of iterators. When a map has accumulated enough overflow buckets we trigger map growth, but grow into a map of the same size as before. The old overflow buckets will be left behind for garbage collection. For the code in #16070, instead of climbing (very slowly) forever, memory usage now cycles between 264mb and 483mb every 15 minutes or so. To avoid increasing the size of maps, the overflow bucket counter is only 16 bits. For large maps, the counter is incremented stochastically. Fixes #16070 Change-Id: If551d77613ec6836907efca58bda3deee304297e Reviewed-on: https://go-review.googlesource.com/25049 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2016-07-19 12:47:53 -06:00
if !h.sameSizeGrow() {
// There used to be half as many buckets; mask down one more power of two.
m >>= 1
}
oldb := (*bmap)(add(c, (hash&m)*uintptr(t.bucketsize)))
if !evacuated(oldb) {
b = oldb
}
}
}
for {
for i := uintptr(0); i < bucketCnt; i++ {
k := *((*uint32)(add(unsafe.Pointer(b), dataOffset+i*4)))
if k != key {
continue
}
x := *((*uint8)(add(unsafe.Pointer(b), i))) // b.topbits[i] without the bounds check
if x == empty {
continue
}
return add(unsafe.Pointer(b), dataOffset+bucketCnt*4+i*uintptr(t.valuesize)), true
}
b = b.overflow(t)
if b == nil {
return unsafe.Pointer(&zeroVal[0]), false
}
}
}
func mapaccess1_fast64(t *maptype, h *hmap, key uint64) unsafe.Pointer {
if raceenabled && h != nil {
callerpc := getcallerpc(unsafe.Pointer(&t))
racereadpc(unsafe.Pointer(h), callerpc, funcPC(mapaccess1_fast64))
}
if h == nil || h.count == 0 {
return unsafe.Pointer(&zeroVal[0])
}
if h.flags&hashWriting != 0 {
throw("concurrent map read and map write")
}
var b *bmap
if h.B == 0 {
// One-bucket table. No need to hash.
b = (*bmap)(h.buckets)
} else {
hash := t.key.alg.hash(noescape(unsafe.Pointer(&key)), uintptr(h.hash0))
m := uintptr(1)<<h.B - 1
b = (*bmap)(add(h.buckets, (hash&m)*uintptr(t.bucketsize)))
if c := h.oldbuckets; c != nil {
runtime: limit the number of map overflow buckets Consider repeatedly adding many items to a map and then deleting them all, as in #16070. The map itself doesn't need to grow above the high water mark of number of items. However, due to random collisions, the map can accumulate overflow buckets. Prior to this CL, those overflow buckets were never removed, which led to a slow memory leak. The problem with removing overflow buckets is iterators. The obvious approach is to repack keys and values and eliminate unused overflow buckets. However, keys, values, and overflow buckets cannot be manipulated without disrupting iterators. This CL takes a different approach, which is to reuse the existing map growth mechanism, which is well established, well tested, and safe in the presence of iterators. When a map has accumulated enough overflow buckets we trigger map growth, but grow into a map of the same size as before. The old overflow buckets will be left behind for garbage collection. For the code in #16070, instead of climbing (very slowly) forever, memory usage now cycles between 264mb and 483mb every 15 minutes or so. To avoid increasing the size of maps, the overflow bucket counter is only 16 bits. For large maps, the counter is incremented stochastically. Fixes #16070 Change-Id: If551d77613ec6836907efca58bda3deee304297e Reviewed-on: https://go-review.googlesource.com/25049 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2016-07-19 12:47:53 -06:00
if !h.sameSizeGrow() {
// There used to be half as many buckets; mask down one more power of two.
m >>= 1
}
oldb := (*bmap)(add(c, (hash&m)*uintptr(t.bucketsize)))
if !evacuated(oldb) {
b = oldb
}
}
}
for {
for i := uintptr(0); i < bucketCnt; i++ {
k := *((*uint64)(add(unsafe.Pointer(b), dataOffset+i*8)))
if k != key {
continue
}
x := *((*uint8)(add(unsafe.Pointer(b), i))) // b.topbits[i] without the bounds check
if x == empty {
continue
}
return add(unsafe.Pointer(b), dataOffset+bucketCnt*8+i*uintptr(t.valuesize))
}
b = b.overflow(t)
if b == nil {
return unsafe.Pointer(&zeroVal[0])
}
}
}
func mapaccess2_fast64(t *maptype, h *hmap, key uint64) (unsafe.Pointer, bool) {
if raceenabled && h != nil {
callerpc := getcallerpc(unsafe.Pointer(&t))
racereadpc(unsafe.Pointer(h), callerpc, funcPC(mapaccess2_fast64))
}
if h == nil || h.count == 0 {
return unsafe.Pointer(&zeroVal[0]), false
}
if h.flags&hashWriting != 0 {
throw("concurrent map read and map write")
}
var b *bmap
if h.B == 0 {
// One-bucket table. No need to hash.
b = (*bmap)(h.buckets)
} else {
hash := t.key.alg.hash(noescape(unsafe.Pointer(&key)), uintptr(h.hash0))
m := uintptr(1)<<h.B - 1
b = (*bmap)(add(h.buckets, (hash&m)*uintptr(t.bucketsize)))
if c := h.oldbuckets; c != nil {
runtime: limit the number of map overflow buckets Consider repeatedly adding many items to a map and then deleting them all, as in #16070. The map itself doesn't need to grow above the high water mark of number of items. However, due to random collisions, the map can accumulate overflow buckets. Prior to this CL, those overflow buckets were never removed, which led to a slow memory leak. The problem with removing overflow buckets is iterators. The obvious approach is to repack keys and values and eliminate unused overflow buckets. However, keys, values, and overflow buckets cannot be manipulated without disrupting iterators. This CL takes a different approach, which is to reuse the existing map growth mechanism, which is well established, well tested, and safe in the presence of iterators. When a map has accumulated enough overflow buckets we trigger map growth, but grow into a map of the same size as before. The old overflow buckets will be left behind for garbage collection. For the code in #16070, instead of climbing (very slowly) forever, memory usage now cycles between 264mb and 483mb every 15 minutes or so. To avoid increasing the size of maps, the overflow bucket counter is only 16 bits. For large maps, the counter is incremented stochastically. Fixes #16070 Change-Id: If551d77613ec6836907efca58bda3deee304297e Reviewed-on: https://go-review.googlesource.com/25049 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2016-07-19 12:47:53 -06:00
if !h.sameSizeGrow() {
// There used to be half as many buckets; mask down one more power of two.
m >>= 1
}
oldb := (*bmap)(add(c, (hash&m)*uintptr(t.bucketsize)))
if !evacuated(oldb) {
b = oldb
}
}
}
for {
for i := uintptr(0); i < bucketCnt; i++ {
k := *((*uint64)(add(unsafe.Pointer(b), dataOffset+i*8)))
if k != key {
continue
}
x := *((*uint8)(add(unsafe.Pointer(b), i))) // b.topbits[i] without the bounds check
if x == empty {
continue
}
return add(unsafe.Pointer(b), dataOffset+bucketCnt*8+i*uintptr(t.valuesize)), true
}
b = b.overflow(t)
if b == nil {
return unsafe.Pointer(&zeroVal[0]), false
}
}
}
func mapaccess1_faststr(t *maptype, h *hmap, ky string) unsafe.Pointer {
if raceenabled && h != nil {
callerpc := getcallerpc(unsafe.Pointer(&t))
racereadpc(unsafe.Pointer(h), callerpc, funcPC(mapaccess1_faststr))
}
if h == nil || h.count == 0 {
return unsafe.Pointer(&zeroVal[0])
}
if h.flags&hashWriting != 0 {
throw("concurrent map read and map write")
}
key := stringStructOf(&ky)
if h.B == 0 {
// One-bucket table.
b := (*bmap)(h.buckets)
if key.len < 32 {
// short key, doing lots of comparisons is ok
for i := uintptr(0); i < bucketCnt; i++ {
x := *((*uint8)(add(unsafe.Pointer(b), i))) // b.topbits[i] without the bounds check
if x == empty {
continue
}
k := (*stringStruct)(add(unsafe.Pointer(b), dataOffset+i*2*sys.PtrSize))
if k.len != key.len {
continue
}
if k.str == key.str || memequal(k.str, key.str, uintptr(key.len)) {
return add(unsafe.Pointer(b), dataOffset+bucketCnt*2*sys.PtrSize+i*uintptr(t.valuesize))
}
}
return unsafe.Pointer(&zeroVal[0])
}
// long key, try not to do more comparisons than necessary
keymaybe := uintptr(bucketCnt)
for i := uintptr(0); i < bucketCnt; i++ {
x := *((*uint8)(add(unsafe.Pointer(b), i))) // b.topbits[i] without the bounds check
if x == empty {
continue
}
k := (*stringStruct)(add(unsafe.Pointer(b), dataOffset+i*2*sys.PtrSize))
if k.len != key.len {
continue
}
if k.str == key.str {
return add(unsafe.Pointer(b), dataOffset+bucketCnt*2*sys.PtrSize+i*uintptr(t.valuesize))
}
// check first 4 bytes
// TODO: on amd64/386 at least, make this compile to one 4-byte comparison instead of
// four 1-byte comparisons.
if *((*[4]byte)(key.str)) != *((*[4]byte)(k.str)) {
continue
}
// check last 4 bytes
if *((*[4]byte)(add(key.str, uintptr(key.len)-4))) != *((*[4]byte)(add(k.str, uintptr(key.len)-4))) {
continue
}
if keymaybe != bucketCnt {
// Two keys are potential matches. Use hash to distinguish them.
goto dohash
}
keymaybe = i
}
if keymaybe != bucketCnt {
k := (*stringStruct)(add(unsafe.Pointer(b), dataOffset+keymaybe*2*sys.PtrSize))
if memequal(k.str, key.str, uintptr(key.len)) {
return add(unsafe.Pointer(b), dataOffset+bucketCnt*2*sys.PtrSize+keymaybe*uintptr(t.valuesize))
}
}
return unsafe.Pointer(&zeroVal[0])
}
dohash:
hash := t.key.alg.hash(noescape(unsafe.Pointer(&ky)), uintptr(h.hash0))
m := uintptr(1)<<h.B - 1
b := (*bmap)(add(h.buckets, (hash&m)*uintptr(t.bucketsize)))
if c := h.oldbuckets; c != nil {
runtime: limit the number of map overflow buckets Consider repeatedly adding many items to a map and then deleting them all, as in #16070. The map itself doesn't need to grow above the high water mark of number of items. However, due to random collisions, the map can accumulate overflow buckets. Prior to this CL, those overflow buckets were never removed, which led to a slow memory leak. The problem with removing overflow buckets is iterators. The obvious approach is to repack keys and values and eliminate unused overflow buckets. However, keys, values, and overflow buckets cannot be manipulated without disrupting iterators. This CL takes a different approach, which is to reuse the existing map growth mechanism, which is well established, well tested, and safe in the presence of iterators. When a map has accumulated enough overflow buckets we trigger map growth, but grow into a map of the same size as before. The old overflow buckets will be left behind for garbage collection. For the code in #16070, instead of climbing (very slowly) forever, memory usage now cycles between 264mb and 483mb every 15 minutes or so. To avoid increasing the size of maps, the overflow bucket counter is only 16 bits. For large maps, the counter is incremented stochastically. Fixes #16070 Change-Id: If551d77613ec6836907efca58bda3deee304297e Reviewed-on: https://go-review.googlesource.com/25049 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2016-07-19 12:47:53 -06:00
if !h.sameSizeGrow() {
// There used to be half as many buckets; mask down one more power of two.
m >>= 1
}
oldb := (*bmap)(add(c, (hash&m)*uintptr(t.bucketsize)))
if !evacuated(oldb) {
b = oldb
}
}
top := uint8(hash >> (sys.PtrSize*8 - 8))
if top < minTopHash {
top += minTopHash
}
for {
for i := uintptr(0); i < bucketCnt; i++ {
x := *((*uint8)(add(unsafe.Pointer(b), i))) // b.topbits[i] without the bounds check
if x != top {
continue
}
k := (*stringStruct)(add(unsafe.Pointer(b), dataOffset+i*2*sys.PtrSize))
if k.len != key.len {
continue
}
if k.str == key.str || memequal(k.str, key.str, uintptr(key.len)) {
return add(unsafe.Pointer(b), dataOffset+bucketCnt*2*sys.PtrSize+i*uintptr(t.valuesize))
}
}
b = b.overflow(t)
if b == nil {
return unsafe.Pointer(&zeroVal[0])
}
}
}
func mapaccess2_faststr(t *maptype, h *hmap, ky string) (unsafe.Pointer, bool) {
if raceenabled && h != nil {
callerpc := getcallerpc(unsafe.Pointer(&t))
racereadpc(unsafe.Pointer(h), callerpc, funcPC(mapaccess2_faststr))
}
if h == nil || h.count == 0 {
return unsafe.Pointer(&zeroVal[0]), false
}
if h.flags&hashWriting != 0 {
throw("concurrent map read and map write")
}
key := stringStructOf(&ky)
if h.B == 0 {
// One-bucket table.
b := (*bmap)(h.buckets)
if key.len < 32 {
// short key, doing lots of comparisons is ok
for i := uintptr(0); i < bucketCnt; i++ {
x := *((*uint8)(add(unsafe.Pointer(b), i))) // b.topbits[i] without the bounds check
if x == empty {
continue
}
k := (*stringStruct)(add(unsafe.Pointer(b), dataOffset+i*2*sys.PtrSize))
if k.len != key.len {
continue
}
if k.str == key.str || memequal(k.str, key.str, uintptr(key.len)) {
return add(unsafe.Pointer(b), dataOffset+bucketCnt*2*sys.PtrSize+i*uintptr(t.valuesize)), true
}
}
return unsafe.Pointer(&zeroVal[0]), false
}
// long key, try not to do more comparisons than necessary
keymaybe := uintptr(bucketCnt)
for i := uintptr(0); i < bucketCnt; i++ {
x := *((*uint8)(add(unsafe.Pointer(b), i))) // b.topbits[i] without the bounds check
if x == empty {
continue
}
k := (*stringStruct)(add(unsafe.Pointer(b), dataOffset+i*2*sys.PtrSize))
if k.len != key.len {
continue
}
if k.str == key.str {
return add(unsafe.Pointer(b), dataOffset+bucketCnt*2*sys.PtrSize+i*uintptr(t.valuesize)), true
}
// check first 4 bytes
if *((*[4]byte)(key.str)) != *((*[4]byte)(k.str)) {
continue
}
// check last 4 bytes
if *((*[4]byte)(add(key.str, uintptr(key.len)-4))) != *((*[4]byte)(add(k.str, uintptr(key.len)-4))) {
continue
}
if keymaybe != bucketCnt {
// Two keys are potential matches. Use hash to distinguish them.
goto dohash
}
keymaybe = i
}
if keymaybe != bucketCnt {
k := (*stringStruct)(add(unsafe.Pointer(b), dataOffset+keymaybe*2*sys.PtrSize))
if memequal(k.str, key.str, uintptr(key.len)) {
return add(unsafe.Pointer(b), dataOffset+bucketCnt*2*sys.PtrSize+keymaybe*uintptr(t.valuesize)), true
}
}
return unsafe.Pointer(&zeroVal[0]), false
}
dohash:
hash := t.key.alg.hash(noescape(unsafe.Pointer(&ky)), uintptr(h.hash0))
m := uintptr(1)<<h.B - 1
b := (*bmap)(add(h.buckets, (hash&m)*uintptr(t.bucketsize)))
if c := h.oldbuckets; c != nil {
runtime: limit the number of map overflow buckets Consider repeatedly adding many items to a map and then deleting them all, as in #16070. The map itself doesn't need to grow above the high water mark of number of items. However, due to random collisions, the map can accumulate overflow buckets. Prior to this CL, those overflow buckets were never removed, which led to a slow memory leak. The problem with removing overflow buckets is iterators. The obvious approach is to repack keys and values and eliminate unused overflow buckets. However, keys, values, and overflow buckets cannot be manipulated without disrupting iterators. This CL takes a different approach, which is to reuse the existing map growth mechanism, which is well established, well tested, and safe in the presence of iterators. When a map has accumulated enough overflow buckets we trigger map growth, but grow into a map of the same size as before. The old overflow buckets will be left behind for garbage collection. For the code in #16070, instead of climbing (very slowly) forever, memory usage now cycles between 264mb and 483mb every 15 minutes or so. To avoid increasing the size of maps, the overflow bucket counter is only 16 bits. For large maps, the counter is incremented stochastically. Fixes #16070 Change-Id: If551d77613ec6836907efca58bda3deee304297e Reviewed-on: https://go-review.googlesource.com/25049 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by: Keith Randall <khr@golang.org>
2016-07-19 12:47:53 -06:00
if !h.sameSizeGrow() {
// There used to be half as many buckets; mask down one more power of two.
m >>= 1
}
oldb := (*bmap)(add(c, (hash&m)*uintptr(t.bucketsize)))
if !evacuated(oldb) {
b = oldb
}
}
top := uint8(hash >> (sys.PtrSize*8 - 8))
if top < minTopHash {
top += minTopHash
}
for {
for i := uintptr(0); i < bucketCnt; i++ {
x := *((*uint8)(add(unsafe.Pointer(b), i))) // b.topbits[i] without the bounds check
if x != top {
continue
}
k := (*stringStruct)(add(unsafe.Pointer(b), dataOffset+i*2*sys.PtrSize))
if k.len != key.len {
continue
}
if k.str == key.str || memequal(k.str, key.str, uintptr(key.len)) {
return add(unsafe.Pointer(b), dataOffset+bucketCnt*2*sys.PtrSize+i*uintptr(t.valuesize)), true
}
}
b = b.overflow(t)
if b == nil {
return unsafe.Pointer(&zeroVal[0]), false
}
}
}