runtime: avoid extra tophash check in mapassign when key comparison is cheap

mapaccess and mapdelete functions are already optimized to prefer direct
key comparison instead of tophash checks when key comparison is cheap.

Extended version of golang.org/cl/55235.

AMD64:
name                old time/op    new time/op    delta
MapPopulate/1         42.5ns ± 2%    40.3ns ± 2%  -5.37%  (p=0.000 n=9+10)
MapPopulate/10         558ns ± 1%     556ns ± 1%    ~     (p=0.157 n=10+10)
MapPopulate/100       7.75µs ± 1%    7.66µs ± 2%  -1.19%  (p=0.001 n=10+10)
MapPopulate/1000      92.6µs ± 1%    92.0µs ± 1%  -0.61%  (p=0.016 n=10+8)
MapPopulate/10000      817µs ± 1%     814µs ± 1%    ~     (p=0.247 n=10+10)
MapPopulate/100000    8.02ms ± 1%    7.90ms ± 2%  -1.47%  (p=0.007 n=10+10)

Change-Id: If0eca9931379cbbd37eb753e9bcd2888d8272153
Reviewed-on: https://go-review.googlesource.com/62050
Run-TryBot: Martin Möhrmann <moehrmann@google.com>
Reviewed-by: Daniel Martí <mvdan@mvdan.cc>
Reviewed-by: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>

diff --git a/src/runtime/hashmap_fast.go b/src/runtime/hashmap_fast.go
index 21e1f68bf778df85247a4a540243deef33e913ee..e7b84482032032bb628d74c42ce72356bce4ac06 100644 (file)
--- a/src/runtime/hashmap_fast.go
+++ b/src/runtime/hashmap_fast.go
@@ -373,15 +373,14 @@ again:
                growWork_fast32(t, h, bucket)
        }
        b := (*bmap)(unsafe.Pointer(uintptr(h.buckets) + bucket*uintptr(t.bucketsize)))
-       top := tophash(hash)
 
        var inserti *uint8
        var insertk unsafe.Pointer
        var val unsafe.Pointer
        for {
                for i := uintptr(0); i < bucketCnt; i++ {
-                       if b.tophash[i] != top {
-                               if b.tophash[i] == empty && inserti == nil {
+                       if b.tophash[i] == empty {
+                               if inserti == nil {
                                        inserti = &b.tophash[i]
                                        insertk = add(unsafe.Pointer(b), dataOffset+i*4)
                                        val = add(unsafe.Pointer(b), dataOffset+bucketCnt*4+i*uintptr(t.valuesize))
@@ -425,7 +424,8 @@ again:
        } else {
                *(*uint32)(insertk) = key
        }
-       *inserti = top
+
+       *inserti = tophash(hash)
        h.count++
 
 done:
@@ -462,15 +462,14 @@ again:
                growWork_fast64(t, h, bucket)
        }
        b := (*bmap)(unsafe.Pointer(uintptr(h.buckets) + bucket*uintptr(t.bucketsize)))
-       top := tophash(hash)
 
        var inserti *uint8
        var insertk unsafe.Pointer
        var val unsafe.Pointer
        for {
                for i := uintptr(0); i < bucketCnt; i++ {
-                       if b.tophash[i] != top {
-                               if b.tophash[i] == empty && inserti == nil {
+                       if b.tophash[i] == empty {
+                               if inserti == nil {
                                        inserti = &b.tophash[i]
                                        insertk = add(unsafe.Pointer(b), dataOffset+i*8)
                                        val = add(unsafe.Pointer(b), dataOffset+bucketCnt*8+i*uintptr(t.valuesize))
@@ -521,7 +520,7 @@ again:
                *(*uint64)(insertk) = key
        }
 
-       *inserti = top
+       *inserti = tophash(hash)
        h.count++
 
 done: