1
0
mirror of https://github.com/golang/go synced 2024-10-02 08:18:32 -06:00
go/src/runtime/mcentral.go

213 lines
5.5 KiB
Go
Raw Normal View History

// Copyright 2009 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.
// Central free lists.
//
// See malloc.go for an overview.
//
// The MCentral doesn't actually contain the list of free objects; the MSpan does.
// Each MCentral is two lists of MSpans: those with free objects (c->nonempty)
// and those that are completely allocated (c->empty).
package runtime
// Central list of free objects of a given size.
type mcentral struct {
lock mutex
sizeclass int32
nonempty mSpanList // list of spans with a free object
empty mSpanList // list of spans with no free objects (or cached in an mcache)
}
// Initialize a single central free list.
func mCentral_Init(c *mcentral, sizeclass int32) {
c.sizeclass = sizeclass
mSpanList_Init(&c.nonempty)
mSpanList_Init(&c.empty)
}
// Allocate a span to use in an MCache.
func mCentral_CacheSpan(c *mcentral) *mspan {
runtime: make sweep proportional to spans bytes allocated Proportional concurrent sweep is currently based on a ratio of spans to be swept per bytes of object allocation. However, proportional sweeping is performed during span allocation, not object allocation, in order to minimize contention and overhead. Since objects are allocated from spans after those spans are allocated, the system tends to operate in debt, which means when the next GC cycle starts, there is often sweep debt remaining, so GC has to finish the sweep, which delays the start of the cycle and delays enabling mutator assists. For example, it's quite likely that many Ps will simultaneously refill their span caches immediately after a GC cycle (because GC flushes the span caches), but at this point, there has been very little object allocation since the end of GC, so very little sweeping is done. The Ps then allocate objects from these cached spans, which drives up the bytes of object allocation, but since these allocations are coming from cached spans, nothing considers whether more sweeping has to happen. If the sweep ratio is high enough (which can happen if the next GC trigger is very close to the retained heap size), this can easily represent a sweep debt of thousands of pages. Fix this by making proportional sweep proportional to the number of bytes of spans allocated, rather than the number of bytes of objects allocated. Prior to allocating a span, both the small object path and the large object path ensure credit for allocating that span, so the system operates in the black, rather than in the red. Combined with the previous commit, this should eliminate all sweeping from GC start up. On the stress test in issue #11911, this reduces the time spent sweeping during GC (and delaying start up) by several orders of magnitude: mean 99%ile max pre fix 1 ms 11 ms 144 ms post fix 270 ns 735 ns 916 ns Updates #11911. Change-Id: I89223712883954c9d6ec2a7a51ecb97172097df3 Reviewed-on: https://go-review.googlesource.com/13044 Reviewed-by: Rick Hudson <rlh@golang.org> Reviewed-by: Russ Cox <rsc@golang.org>
2015-08-03 07:46:50 -06:00
// Deduct credit for this span allocation and sweep if necessary.
deductSweepCredit(uintptr(class_to_size[c.sizeclass]), 0)
runtime: finish sweeping before concurrent GC starts Currently, the concurrent sweep follows a 1:1 rule: when allocation needs a span, it sweeps a span (likewise, when a large allocation needs N pages, it sweeps until it frees N pages). This rule worked well for the STW collector (especially when GOGC==100) because it did no more sweeping than necessary to keep the heap from growing, would generally finish sweeping just before GC, and ensured good temporal locality between sweeping a page and allocating from it. It doesn't work well with concurrent GC. Since concurrent GC requires starting GC earlier (sometimes much earlier), the sweep often won't be done when GC starts. Unfortunately, the first thing GC has to do is finish the sweep. In the mean time, the mutator can continue allocating, pushing the heap size even closer to the goal size. This worked okay with the 7/8ths trigger, but it gets into a vicious cycle with the GC trigger controller: if the mutator is allocating quickly and driving the trigger lower, more and more sweep work will be left to GC; this both causes GC to take longer (allowing the mutator to allocate more during GC) and delays the start of the concurrent mark phase, which throws off the GC controller's statistics and generally causes it to push the trigger even lower. As an example of a particularly bad case, the garbage benchmark with GOMAXPROCS=4 and -benchmem 512 (MB) spends the first 0.4-0.8 seconds of each GC cycle sweeping, during which the heap grows by between 109MB and 252MB. To fix this, this change replaces the 1:1 sweep rule with a proportional sweep rule. At the end of GC, GC knows exactly how much heap allocation will occur before the next concurrent GC as well as how many span pages must be swept. This change computes this "sweep ratio" and when the mallocgc asks for a span, the mcentral sweeps enough spans to bring the swept span count into ratio with the allocated byte count. On the benchmark from above, this entirely eliminates sweeping at the beginning of GC, which reduces the time between startGC readying the GC goroutine and GC stopping the world for sweep termination to ~100µs during which the heap grows at most 134KB. Change-Id: I35422d6bba0c2310d48bb1f8f30a72d29e98c1af Reviewed-on: https://go-review.googlesource.com/8921 Reviewed-by: Rick Hudson <rlh@golang.org>
2015-04-13 21:34:57 -06:00
lock(&c.lock)
sg := mheap_.sweepgen
retry:
var s *mspan
for s = c.nonempty.first; s != nil; s = s.next {
if s.sweepgen == sg-2 && cas(&s.sweepgen, sg-2, sg-1) {
mSpanList_Remove(&c.nonempty, s)
mSpanList_InsertBack(&c.empty, s)
unlock(&c.lock)
mSpan_Sweep(s, true)
goto havespan
}
if s.sweepgen == sg-1 {
// the span is being swept by background sweeper, skip
continue
}
// we have a nonempty span that does not require sweeping, allocate from it
mSpanList_Remove(&c.nonempty, s)
mSpanList_InsertBack(&c.empty, s)
unlock(&c.lock)
goto havespan
}
for s = c.empty.first; s != nil; s = s.next {
if s.sweepgen == sg-2 && cas(&s.sweepgen, sg-2, sg-1) {
// we have an empty span that requires sweeping,
// sweep it and see if we can free some space in it
mSpanList_Remove(&c.empty, s)
// swept spans are at the end of the list
mSpanList_InsertBack(&c.empty, s)
unlock(&c.lock)
mSpan_Sweep(s, true)
if s.freelist.ptr() != nil {
goto havespan
}
lock(&c.lock)
// the span is still empty after sweep
// it is already in the empty list, so just retry
goto retry
}
if s.sweepgen == sg-1 {
// the span is being swept by background sweeper, skip
continue
}
// already swept empty span,
// all subsequent ones must also be either swept or in process of sweeping
break
}
unlock(&c.lock)
// Replenish central list if empty.
s = mCentral_Grow(c)
if s == nil {
return nil
}
lock(&c.lock)
mSpanList_InsertBack(&c.empty, s)
unlock(&c.lock)
// At this point s is a non-empty span, queued at the end of the empty list,
// c is unlocked.
havespan:
cap := int32((s.npages << _PageShift) / s.elemsize)
n := cap - int32(s.ref)
if n == 0 {
throw("empty span")
}
if s.freelist.ptr() == nil {
throw("freelist empty")
}
s.incache = true
return s
}
// Return span from an MCache.
func mCentral_UncacheSpan(c *mcentral, s *mspan) {
lock(&c.lock)
s.incache = false
if s.ref == 0 {
throw("uncaching full span")
}
cap := int32((s.npages << _PageShift) / s.elemsize)
n := cap - int32(s.ref)
if n > 0 {
mSpanList_Remove(&c.empty, s)
mSpanList_Insert(&c.nonempty, s)
}
unlock(&c.lock)
}
// Free n objects from a span s back into the central free list c.
// Called during sweep.
// Returns true if the span was returned to heap. Sets sweepgen to
// the latest generation.
// If preserve=true, don't return the span to heap nor relink in MCentral lists;
// caller takes care of it.
func mCentral_FreeSpan(c *mcentral, s *mspan, n int32, start gclinkptr, end gclinkptr, preserve bool) bool {
if s.incache {
throw("freespan into cached span")
}
// Add the objects back to s's free list.
wasempty := s.freelist.ptr() == nil
end.ptr().next = s.freelist
s.freelist = start
s.ref -= uint16(n)
if preserve {
// preserve is set only when called from MCentral_CacheSpan above,
// the span must be in the empty list.
if !mSpan_InList(s) {
throw("can't preserve unlinked span")
}
atomicstore(&s.sweepgen, mheap_.sweepgen)
return false
}
lock(&c.lock)
// Move to nonempty if necessary.
if wasempty {
mSpanList_Remove(&c.empty, s)
mSpanList_Insert(&c.nonempty, s)
}
// delay updating sweepgen until here. This is the signal that
// the span may be used in an MCache, so it must come after the
// linked list operations above (actually, just after the
// lock of c above.)
atomicstore(&s.sweepgen, mheap_.sweepgen)
if s.ref != 0 {
unlock(&c.lock)
return false
}
// s is completely freed, return it to the heap.
mSpanList_Remove(&c.nonempty, s)
s.needzero = 1
s.freelist = 0
unlock(&c.lock)
heapBitsForSpan(s.base()).initSpan(s.layout())
mHeap_Free(&mheap_, s, 0)
return true
}
// Fetch a new span from the heap and carve into objects for the free list.
func mCentral_Grow(c *mcentral) *mspan {
npages := uintptr(class_to_allocnpages[c.sizeclass])
size := uintptr(class_to_size[c.sizeclass])
n := (npages << _PageShift) / size
s := mHeap_Alloc(&mheap_, npages, c.sizeclass, false, true)
if s == nil {
return nil
}
p := uintptr(s.start << _PageShift)
s.limit = p + size*n
head := gclinkptr(p)
tail := gclinkptr(p)
// i==0 iteration already done
for i := uintptr(1); i < n; i++ {
p += size
tail.ptr().next = gclinkptr(p)
tail = gclinkptr(p)
}
if s.freelist.ptr() != nil {
throw("freelist not empty")
}
tail.ptr().next = 0
s.freelist = head
heapBitsForSpan(s.base()).initSpan(s.layout())
return s
}