From 7e26a2d9a80b825d019c2cdaf6437d89001506a9 Mon Sep 17 00:00:00 2001
From: Russ Cox <rsc@golang.org>
Date: Thu, 14 May 2015 17:27:04 -0400
Subject: [PATCH] runtime: allocate map element zero values for reflect-created
 types on demand

Preallocating them in reflect means that
(1) if you say _ = PtrTo(ArrayOf(1000000000, reflect.TypeOf(byte(0)))), you just allocated 1GB of data
(2) if you say it again, that's *another* GB of data.

The only use of t.zero in the runtime is for map elements.
Delay the allocation until the creation of a map with that element type,
and share the zeros.

The one downside of the shared zero is that it's not garbage collected,
but it's also never written, so the OS should be able to handle it fairly
efficiently.

Change-Id: I56b098a091abf3ac0945de28ebef9a6c08e76614
Reviewed-on: https://go-review.googlesource.com/10111
Reviewed-by: Keith Randall <khr@golang.org>
---
 src/reflect/type.go    |  9 --------
 src/runtime/hashmap.go | 48 ++++++++++++++++++++++++++++++++++++++++++
 2 files changed, 48 insertions(+), 9 deletions(-)

diff --git a/src/reflect/type.go b/src/reflect/type.go
index 5a43805626..bffe2595dd 100644
--- a/src/reflect/type.go
+++ b/src/reflect/type.go
@@ -1087,7 +1087,6 @@ func (t *rtype) ptrTo() *rtype {
 
 	p.uncommonType = nil
 	p.ptrToThis = nil
-	p.zero = unsafe.Pointer(&make([]byte, p.size)[0])
 	p.elem = t
 
 	ptrMap.m[t] = p
@@ -1467,7 +1466,6 @@ func ChanOf(dir ChanDir, t Type) Type {
 	ch.elem = typ
 	ch.uncommonType = nil
 	ch.ptrToThis = nil
-	ch.zero = unsafe.Pointer(&make([]byte, ch.size)[0])
 
 	return cachePut(ckey, &ch.rtype)
 }
@@ -1530,7 +1528,6 @@ func MapOf(key, elem Type) Type {
 	mt.reflexivekey = isReflexive(ktyp)
 	mt.uncommonType = nil
 	mt.ptrToThis = nil
-	mt.zero = unsafe.Pointer(&make([]byte, mt.size)[0])
 
 	return cachePut(ckey, &mt.rtype)
 }
@@ -1610,7 +1607,6 @@ func FuncOf(in, out []Type, variadic bool) Type {
 	ft.string = &str
 	ft.uncommonType = nil
 	ft.ptrToThis = nil
-	ft.zero = unsafe.Pointer(&make([]byte, ft.size)[0])
 	funcLookupCache.m[hash] = append(funcLookupCache.m[hash], &ft.rtype)
 
 	return ft
@@ -1857,7 +1853,6 @@ func SliceOf(t Type) Type {
 	slice.elem = typ
 	slice.uncommonType = nil
 	slice.ptrToThis = nil
-	slice.zero = unsafe.Pointer(&make([]byte, slice.size)[0])
 
 	return cachePut(ckey, &slice.rtype)
 }
@@ -1913,10 +1908,6 @@ func ArrayOf(count int, elem Type) Type {
 	array.fieldAlign = typ.fieldAlign
 	array.uncommonType = nil
 	array.ptrToThis = nil
-	if array.size > 0 {
-		zero := make([]byte, array.size)
-		array.zero = unsafe.Pointer(&zero[0])
-	}
 	array.len = uintptr(count)
 	array.slice = slice.(*rtype)
 
diff --git a/src/runtime/hashmap.go b/src/runtime/hashmap.go
index 9ca33992bb..2b3af301b3 100644
--- a/src/runtime/hashmap.go
+++ b/src/runtime/hashmap.go
@@ -233,6 +233,9 @@ func makemap(t *maptype, hint int64, h *hmap, bucket unsafe.Pointer) *hmap {
 		throw("need padding in bucket (value)")
 	}
 
+	// make sure zero of element type is available.
+	mapzero(t.elem)
+
 	// find size parameter which will hold the requested # of elements
 	B := uint8(0)
 	for ; hint > bucketCnt && float32(hint) > loadFactor*float32(uintptr(1)<<B); B++ {
@@ -990,3 +993,48 @@ func reflect_maplen(h *hmap) int {
 func reflect_ismapkey(t *_type) bool {
 	return ismapkey(t)
 }
+
+var zerobuf struct {
+	lock mutex
+	p    *byte
+	size uintptr
+}
+
+var zerotiny [1024]byte
+
+// mapzero ensures that t.zero points at a zero value for type t.
+// Types known to the compiler are in read-only memory and all point
+// to a single zero in the bss of a large enough size.
+// Types allocated by package reflect are in writable memory and
+// start out with zero set to nil; we initialize those on demand.
+func mapzero(t *_type) {
+	// Already done?
+	// Check without lock, so must use atomicload to sync with atomicstore in allocation case below.
+	if atomicloadp(unsafe.Pointer(&t.zero)) != nil {
+		return
+	}
+
+	// Small enough for static buffer?
+	if t.size <= uintptr(len(zerotiny)) {
+		atomicstorep(unsafe.Pointer(&t.zero), unsafe.Pointer(&zerotiny[0]))
+		return
+	}
+
+	// Use allocated buffer.
+	lock(&zerobuf.lock)
+	if zerobuf.size < t.size {
+		if zerobuf.size == 0 {
+			zerobuf.size = 4 * 1024
+		}
+		for zerobuf.size < t.size {
+			zerobuf.size *= 2
+			if zerobuf.size == 0 {
+				// need >2GB zero on 32-bit machine
+				throw("map element too large")
+			}
+		}
+		zerobuf.p = (*byte)(persistentalloc(zerobuf.size, 64, &memstats.other_sys))
+	}
+	atomicstorep(unsafe.Pointer(&t.zero), unsafe.Pointer(zerobuf.p))
+	unlock(&zerobuf.lock)
+}