2019-07-25 13:54:03 -06:00
|
|
|
// errorcheck -0 -m -l
|
2015-02-19 09:10:31 -07:00
|
|
|
|
2016-04-10 15:32:26 -06:00
|
|
|
// Copyright 2015 The Go Authors. All rights reserved.
|
2015-02-19 09:10:31 -07:00
|
|
|
// Use of this source code is governed by a BSD-style
|
|
|
|
// license that can be found in the LICENSE file.
|
|
|
|
|
|
|
|
// Test escape analysis for function parameters.
|
|
|
|
|
|
|
|
// In this test almost everything is BAD except the simplest cases
|
|
|
|
// where input directly flows to output.
|
|
|
|
|
|
|
|
package escape
|
|
|
|
|
2018-09-04 08:17:32 -06:00
|
|
|
func zero() int { return 0 }
|
|
|
|
|
2015-02-19 09:10:31 -07:00
|
|
|
var sink interface{}
|
|
|
|
|
|
|
|
// in -> out
|
cmd/internal/gc: improve flow of input params to output params
This includes the following information in the per-function summary:
outK = paramJ encoded in outK bits for paramJ
outK = *paramJ encoded in outK bits for paramJ
heap = paramJ EscHeap
heap = *paramJ EscContentEscapes
Note that (currently) if the address of a parameter is taken and
returned, necessarily a heap allocation occurred to contain that
reference, and the heap can never refer to stack, therefore the
parameter and everything downstream from it escapes to the heap.
The per-function summary information now has a tuneable number of bits
(2 is probably noticeably better than 1, 3 is likely overkill, but it
is now easy to check and the -m debugging output includes information
that allows you to figure out if more would be better.)
A new test was added to check pointer flow through struct-typed and
*struct-typed parameters and returns; some of these are sensitive to
the number of summary bits, and ought to yield better results with a
more competent escape analysis algorithm. Another new test checks
(some) correctness with array parameters, results, and operations.
The old analysis inferred a piece of plan9 runtime was non-escaping by
counteracting overconservative analysis with buggy analysis; with the
bug fixed, the result was too conservative (and it's not easy to fix
in this framework) so the source code was tweaked to get the desired
result. A test was added against the discovered bug.
The escape analysis was further improved splitting the "level" into
3 parts, one tracking the conventional "level" and the other two
computing the highest-level-suffix-from-copy, which is used to
generally model the cancelling effect of indirection applied to
address-of.
With the improved escape analysis enabled, it was necessary to
modify one of the runtime tests because it now attempts to allocate
too much on the (small, fixed-size) G0 (system) stack and this
failed the test.
Compiling src/std after touching src/runtime/*.go with -m logging
turned on shows 420 fewer heap allocation sites (10538 vs 10968).
Profiling allocations in src/html/template with
for i in {1..5} ;
do go tool 6g -memprofile=mastx.${i}.prof -memprofilerate=1 *.go;
go tool pprof -alloc_objects -text mastx.${i}.prof ;
done
showed a 15% reduction in allocations performed by the compiler.
Update #3753
Update #4720
Fixes #10466
Change-Id: I0fd97d5f5ac527b45f49e2218d158a6e89951432
Reviewed-on: https://go-review.googlesource.com/8202
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
2015-03-26 14:36:15 -06:00
|
|
|
func param0(p *int) *int { // ERROR "leaking param: p to result ~r1"
|
2015-02-19 09:10:31 -07:00
|
|
|
return p
|
|
|
|
}
|
|
|
|
|
|
|
|
func caller0a() {
|
|
|
|
i := 0
|
2019-04-01 12:58:33 -06:00
|
|
|
_ = param0(&i)
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
func caller0b() {
|
2020-09-08 22:09:01 -06:00
|
|
|
i := 0 // ERROR "moved to heap: i$"
|
cmd/compile: silence esc diagnostics about directiface OCONVIFACEs
In general, a conversion to interface type may require values to be
boxed, which in turn necessitates escape analysis to determine whether
the boxed representation can be stack allocated.
However, esc.go used to unconditionally print escape analysis
decisions about OCONVIFACE, even for conversions that don't require
boxing (e.g., pointers, channels, maps, functions).
For test compatibility with esc.go, escape.go similarly printed these
useless diagnostics. This CL removes the diagnostics, and updates test
expectations accordingly.
Change-Id: I97c57a4a08e44d265bba516c78426ff4f2bf1e12
Reviewed-on: https://go-review.googlesource.com/c/go/+/192697
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
2019-08-30 11:56:30 -06:00
|
|
|
sink = param0(&i)
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
// in, in -> out, out
|
cmd/internal/gc: improve flow of input params to output params
This includes the following information in the per-function summary:
outK = paramJ encoded in outK bits for paramJ
outK = *paramJ encoded in outK bits for paramJ
heap = paramJ EscHeap
heap = *paramJ EscContentEscapes
Note that (currently) if the address of a parameter is taken and
returned, necessarily a heap allocation occurred to contain that
reference, and the heap can never refer to stack, therefore the
parameter and everything downstream from it escapes to the heap.
The per-function summary information now has a tuneable number of bits
(2 is probably noticeably better than 1, 3 is likely overkill, but it
is now easy to check and the -m debugging output includes information
that allows you to figure out if more would be better.)
A new test was added to check pointer flow through struct-typed and
*struct-typed parameters and returns; some of these are sensitive to
the number of summary bits, and ought to yield better results with a
more competent escape analysis algorithm. Another new test checks
(some) correctness with array parameters, results, and operations.
The old analysis inferred a piece of plan9 runtime was non-escaping by
counteracting overconservative analysis with buggy analysis; with the
bug fixed, the result was too conservative (and it's not easy to fix
in this framework) so the source code was tweaked to get the desired
result. A test was added against the discovered bug.
The escape analysis was further improved splitting the "level" into
3 parts, one tracking the conventional "level" and the other two
computing the highest-level-suffix-from-copy, which is used to
generally model the cancelling effect of indirection applied to
address-of.
With the improved escape analysis enabled, it was necessary to
modify one of the runtime tests because it now attempts to allocate
too much on the (small, fixed-size) G0 (system) stack and this
failed the test.
Compiling src/std after touching src/runtime/*.go with -m logging
turned on shows 420 fewer heap allocation sites (10538 vs 10968).
Profiling allocations in src/html/template with
for i in {1..5} ;
do go tool 6g -memprofile=mastx.${i}.prof -memprofilerate=1 *.go;
go tool pprof -alloc_objects -text mastx.${i}.prof ;
done
showed a 15% reduction in allocations performed by the compiler.
Update #3753
Update #4720
Fixes #10466
Change-Id: I0fd97d5f5ac527b45f49e2218d158a6e89951432
Reviewed-on: https://go-review.googlesource.com/8202
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
2015-03-26 14:36:15 -06:00
|
|
|
func param1(p1, p2 *int) (*int, *int) { // ERROR "leaking param: p1 to result ~r2" "leaking param: p2 to result ~r3"
|
2015-02-19 09:10:31 -07:00
|
|
|
return p1, p2
|
|
|
|
}
|
|
|
|
|
|
|
|
func caller1() {
|
|
|
|
i := 0 // ERROR "moved to heap: i$"
|
|
|
|
j := 0
|
2019-04-01 12:58:33 -06:00
|
|
|
sink, _ = param1(&i, &j)
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
// in -> other in
|
2019-09-12 11:18:03 -06:00
|
|
|
func param2(p1 *int, p2 **int) { // ERROR "leaking param: p1$" "p2 does not escape$"
|
2015-02-19 09:10:31 -07:00
|
|
|
*p2 = p1
|
|
|
|
}
|
|
|
|
|
|
|
|
func caller2a() {
|
|
|
|
i := 0 // ERROR "moved to heap: i$"
|
|
|
|
var p *int
|
2019-04-01 12:58:33 -06:00
|
|
|
param2(&i, &p)
|
2015-02-19 09:10:31 -07:00
|
|
|
_ = p
|
|
|
|
}
|
|
|
|
|
|
|
|
func caller2b() {
|
|
|
|
i := 0 // ERROR "moved to heap: i$"
|
|
|
|
var p *int
|
2019-04-01 12:58:33 -06:00
|
|
|
param2(&i, &p)
|
cmd/compile: silence esc diagnostics about directiface OCONVIFACEs
In general, a conversion to interface type may require values to be
boxed, which in turn necessitates escape analysis to determine whether
the boxed representation can be stack allocated.
However, esc.go used to unconditionally print escape analysis
decisions about OCONVIFACE, even for conversions that don't require
boxing (e.g., pointers, channels, maps, functions).
For test compatibility with esc.go, escape.go similarly printed these
useless diagnostics. This CL removes the diagnostics, and updates test
expectations accordingly.
Change-Id: I97c57a4a08e44d265bba516c78426ff4f2bf1e12
Reviewed-on: https://go-review.googlesource.com/c/go/+/192697
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
2019-08-30 11:56:30 -06:00
|
|
|
sink = p
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
|
|
|
|
2018-07-27 10:32:17 -06:00
|
|
|
func paramArraySelfAssign(p *PairOfPairs) { // ERROR "p does not escape"
|
|
|
|
p.pairs[0] = p.pairs[1] // ERROR "ignoring self-assignment in p.pairs\[0\] = p.pairs\[1\]"
|
|
|
|
}
|
|
|
|
|
2018-09-04 08:17:32 -06:00
|
|
|
func paramArraySelfAssignUnsafeIndex(p *PairOfPairs) { // ERROR "leaking param content: p"
|
|
|
|
// Function call inside index disables self-assignment case to trigger.
|
|
|
|
p.pairs[zero()] = p.pairs[1]
|
|
|
|
p.pairs[zero()+1] = p.pairs[1]
|
|
|
|
}
|
|
|
|
|
2018-07-27 10:32:17 -06:00
|
|
|
type PairOfPairs struct {
|
|
|
|
pairs [2]*Pair
|
|
|
|
}
|
|
|
|
|
|
|
|
type BoxedPair struct {
|
|
|
|
pair *Pair
|
|
|
|
}
|
|
|
|
|
|
|
|
type WrappedPair struct {
|
|
|
|
pair Pair
|
|
|
|
}
|
|
|
|
|
|
|
|
func leakParam(x interface{}) { // ERROR "leaking param: x"
|
|
|
|
sink = x
|
|
|
|
}
|
|
|
|
|
|
|
|
func sinkAfterSelfAssignment1(box *BoxedPair) { // ERROR "leaking param content: box"
|
|
|
|
box.pair.p1 = box.pair.p2 // ERROR "ignoring self-assignment in box.pair.p1 = box.pair.p2"
|
cmd/compile: silence esc diagnostics about directiface OCONVIFACEs
In general, a conversion to interface type may require values to be
boxed, which in turn necessitates escape analysis to determine whether
the boxed representation can be stack allocated.
However, esc.go used to unconditionally print escape analysis
decisions about OCONVIFACE, even for conversions that don't require
boxing (e.g., pointers, channels, maps, functions).
For test compatibility with esc.go, escape.go similarly printed these
useless diagnostics. This CL removes the diagnostics, and updates test
expectations accordingly.
Change-Id: I97c57a4a08e44d265bba516c78426ff4f2bf1e12
Reviewed-on: https://go-review.googlesource.com/c/go/+/192697
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
2019-08-30 11:56:30 -06:00
|
|
|
sink = box.pair.p2
|
2018-07-27 10:32:17 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
func sinkAfterSelfAssignment2(box *BoxedPair) { // ERROR "leaking param content: box"
|
|
|
|
box.pair.p1 = box.pair.p2 // ERROR "ignoring self-assignment in box.pair.p1 = box.pair.p2"
|
cmd/compile: silence esc diagnostics about directiface OCONVIFACEs
In general, a conversion to interface type may require values to be
boxed, which in turn necessitates escape analysis to determine whether
the boxed representation can be stack allocated.
However, esc.go used to unconditionally print escape analysis
decisions about OCONVIFACE, even for conversions that don't require
boxing (e.g., pointers, channels, maps, functions).
For test compatibility with esc.go, escape.go similarly printed these
useless diagnostics. This CL removes the diagnostics, and updates test
expectations accordingly.
Change-Id: I97c57a4a08e44d265bba516c78426ff4f2bf1e12
Reviewed-on: https://go-review.googlesource.com/c/go/+/192697
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
2019-08-30 11:56:30 -06:00
|
|
|
sink = box.pair
|
2018-07-27 10:32:17 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
func sinkAfterSelfAssignment3(box *BoxedPair) { // ERROR "leaking param content: box"
|
|
|
|
box.pair.p1 = box.pair.p2 // ERROR "ignoring self-assignment in box.pair.p1 = box.pair.p2"
|
cmd/compile: silence esc diagnostics about directiface OCONVIFACEs
In general, a conversion to interface type may require values to be
boxed, which in turn necessitates escape analysis to determine whether
the boxed representation can be stack allocated.
However, esc.go used to unconditionally print escape analysis
decisions about OCONVIFACE, even for conversions that don't require
boxing (e.g., pointers, channels, maps, functions).
For test compatibility with esc.go, escape.go similarly printed these
useless diagnostics. This CL removes the diagnostics, and updates test
expectations accordingly.
Change-Id: I97c57a4a08e44d265bba516c78426ff4f2bf1e12
Reviewed-on: https://go-review.googlesource.com/c/go/+/192697
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
2019-08-30 11:56:30 -06:00
|
|
|
leakParam(box.pair.p2)
|
2018-07-27 10:32:17 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
func sinkAfterSelfAssignment4(box *BoxedPair) { // ERROR "leaking param content: box"
|
|
|
|
box.pair.p1 = box.pair.p2 // ERROR "ignoring self-assignment in box.pair.p1 = box.pair.p2"
|
cmd/compile: silence esc diagnostics about directiface OCONVIFACEs
In general, a conversion to interface type may require values to be
boxed, which in turn necessitates escape analysis to determine whether
the boxed representation can be stack allocated.
However, esc.go used to unconditionally print escape analysis
decisions about OCONVIFACE, even for conversions that don't require
boxing (e.g., pointers, channels, maps, functions).
For test compatibility with esc.go, escape.go similarly printed these
useless diagnostics. This CL removes the diagnostics, and updates test
expectations accordingly.
Change-Id: I97c57a4a08e44d265bba516c78426ff4f2bf1e12
Reviewed-on: https://go-review.googlesource.com/c/go/+/192697
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
2019-08-30 11:56:30 -06:00
|
|
|
leakParam(box.pair)
|
2018-07-27 10:32:17 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
func selfAssignmentAndUnrelated(box1, box2 *BoxedPair) { // ERROR "leaking param content: box2" "box1 does not escape"
|
|
|
|
box1.pair.p1 = box1.pair.p2 // ERROR "ignoring self-assignment in box1.pair.p1 = box1.pair.p2"
|
cmd/compile: silence esc diagnostics about directiface OCONVIFACEs
In general, a conversion to interface type may require values to be
boxed, which in turn necessitates escape analysis to determine whether
the boxed representation can be stack allocated.
However, esc.go used to unconditionally print escape analysis
decisions about OCONVIFACE, even for conversions that don't require
boxing (e.g., pointers, channels, maps, functions).
For test compatibility with esc.go, escape.go similarly printed these
useless diagnostics. This CL removes the diagnostics, and updates test
expectations accordingly.
Change-Id: I97c57a4a08e44d265bba516c78426ff4f2bf1e12
Reviewed-on: https://go-review.googlesource.com/c/go/+/192697
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
2019-08-30 11:56:30 -06:00
|
|
|
leakParam(box2.pair.p2)
|
2018-07-27 10:32:17 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
func notSelfAssignment1(box1, box2 *BoxedPair) { // ERROR "leaking param content: box2" "box1 does not escape"
|
|
|
|
box1.pair.p1 = box2.pair.p1
|
|
|
|
}
|
|
|
|
|
|
|
|
func notSelfAssignment2(p1, p2 *PairOfPairs) { // ERROR "leaking param content: p2" "p1 does not escape"
|
|
|
|
p1.pairs[0] = p2.pairs[1]
|
|
|
|
}
|
|
|
|
|
|
|
|
func notSelfAssignment3(p1, p2 *PairOfPairs) { // ERROR "leaking param content: p2" "p1 does not escape"
|
|
|
|
p1.pairs[0].p1 = p2.pairs[1].p1
|
|
|
|
}
|
|
|
|
|
|
|
|
func boxedPairSelfAssign(box *BoxedPair) { // ERROR "box does not escape"
|
|
|
|
box.pair.p1 = box.pair.p2 // ERROR "ignoring self-assignment in box.pair.p1 = box.pair.p2"
|
|
|
|
}
|
|
|
|
|
|
|
|
func wrappedPairSelfAssign(w *WrappedPair) { // ERROR "w does not escape"
|
|
|
|
w.pair.p1 = w.pair.p2 // ERROR "ignoring self-assignment in w.pair.p1 = w.pair.p2"
|
|
|
|
}
|
|
|
|
|
2015-02-19 09:10:31 -07:00
|
|
|
// in -> in
|
|
|
|
type Pair struct {
|
|
|
|
p1 *int
|
|
|
|
p2 *int
|
|
|
|
}
|
|
|
|
|
2019-09-12 11:18:03 -06:00
|
|
|
func param3(p *Pair) { // ERROR "p does not escape"
|
2018-07-27 10:32:17 -06:00
|
|
|
p.p1 = p.p2 // ERROR "param3 ignoring self-assignment in p.p1 = p.p2"
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
func caller3a() {
|
2018-07-27 10:32:17 -06:00
|
|
|
i := 0
|
|
|
|
j := 0
|
2019-04-01 12:58:33 -06:00
|
|
|
p := Pair{&i, &j}
|
|
|
|
param3(&p)
|
2015-02-19 09:10:31 -07:00
|
|
|
_ = p
|
|
|
|
}
|
|
|
|
|
|
|
|
func caller3b() {
|
2020-09-08 22:09:01 -06:00
|
|
|
i := 0 // ERROR "moved to heap: i$"
|
|
|
|
j := 0 // ERROR "moved to heap: j$"
|
2019-04-01 12:58:33 -06:00
|
|
|
p := Pair{&i, &j}
|
|
|
|
param3(&p)
|
2020-09-08 22:09:01 -06:00
|
|
|
sink = p // ERROR "p escapes to heap$"
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
// in -> rcvr
|
2019-09-12 11:18:03 -06:00
|
|
|
func (p *Pair) param4(i *int) { // ERROR "p does not escape$" "leaking param: i$"
|
2015-02-19 09:10:31 -07:00
|
|
|
p.p1 = i
|
|
|
|
}
|
|
|
|
|
|
|
|
func caller4a() {
|
|
|
|
i := 0 // ERROR "moved to heap: i$"
|
|
|
|
p := Pair{}
|
2019-04-01 12:58:33 -06:00
|
|
|
p.param4(&i)
|
2015-02-19 09:10:31 -07:00
|
|
|
_ = p
|
|
|
|
}
|
|
|
|
|
|
|
|
func caller4b() {
|
|
|
|
i := 0 // ERROR "moved to heap: i$"
|
|
|
|
p := Pair{}
|
2019-04-01 12:58:33 -06:00
|
|
|
p.param4(&i)
|
2020-09-08 22:09:01 -06:00
|
|
|
sink = p // ERROR "p escapes to heap$"
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
// in -> heap
|
|
|
|
func param5(i *int) { // ERROR "leaking param: i$"
|
cmd/compile: silence esc diagnostics about directiface OCONVIFACEs
In general, a conversion to interface type may require values to be
boxed, which in turn necessitates escape analysis to determine whether
the boxed representation can be stack allocated.
However, esc.go used to unconditionally print escape analysis
decisions about OCONVIFACE, even for conversions that don't require
boxing (e.g., pointers, channels, maps, functions).
For test compatibility with esc.go, escape.go similarly printed these
useless diagnostics. This CL removes the diagnostics, and updates test
expectations accordingly.
Change-Id: I97c57a4a08e44d265bba516c78426ff4f2bf1e12
Reviewed-on: https://go-review.googlesource.com/c/go/+/192697
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
2019-08-30 11:56:30 -06:00
|
|
|
sink = i
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
func caller5() {
|
2020-09-08 22:09:01 -06:00
|
|
|
i := 0 // ERROR "moved to heap: i$"
|
2019-04-01 12:58:33 -06:00
|
|
|
param5(&i)
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
// *in -> heap
|
cmd/internal/gc: improve flow of input params to output params
This includes the following information in the per-function summary:
outK = paramJ encoded in outK bits for paramJ
outK = *paramJ encoded in outK bits for paramJ
heap = paramJ EscHeap
heap = *paramJ EscContentEscapes
Note that (currently) if the address of a parameter is taken and
returned, necessarily a heap allocation occurred to contain that
reference, and the heap can never refer to stack, therefore the
parameter and everything downstream from it escapes to the heap.
The per-function summary information now has a tuneable number of bits
(2 is probably noticeably better than 1, 3 is likely overkill, but it
is now easy to check and the -m debugging output includes information
that allows you to figure out if more would be better.)
A new test was added to check pointer flow through struct-typed and
*struct-typed parameters and returns; some of these are sensitive to
the number of summary bits, and ought to yield better results with a
more competent escape analysis algorithm. Another new test checks
(some) correctness with array parameters, results, and operations.
The old analysis inferred a piece of plan9 runtime was non-escaping by
counteracting overconservative analysis with buggy analysis; with the
bug fixed, the result was too conservative (and it's not easy to fix
in this framework) so the source code was tweaked to get the desired
result. A test was added against the discovered bug.
The escape analysis was further improved splitting the "level" into
3 parts, one tracking the conventional "level" and the other two
computing the highest-level-suffix-from-copy, which is used to
generally model the cancelling effect of indirection applied to
address-of.
With the improved escape analysis enabled, it was necessary to
modify one of the runtime tests because it now attempts to allocate
too much on the (small, fixed-size) G0 (system) stack and this
failed the test.
Compiling src/std after touching src/runtime/*.go with -m logging
turned on shows 420 fewer heap allocation sites (10538 vs 10968).
Profiling allocations in src/html/template with
for i in {1..5} ;
do go tool 6g -memprofile=mastx.${i}.prof -memprofilerate=1 *.go;
go tool pprof -alloc_objects -text mastx.${i}.prof ;
done
showed a 15% reduction in allocations performed by the compiler.
Update #3753
Update #4720
Fixes #10466
Change-Id: I0fd97d5f5ac527b45f49e2218d158a6e89951432
Reviewed-on: https://go-review.googlesource.com/8202
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
2015-03-26 14:36:15 -06:00
|
|
|
func param6(i ***int) { // ERROR "leaking param content: i$"
|
cmd/compile: silence esc diagnostics about directiface OCONVIFACEs
In general, a conversion to interface type may require values to be
boxed, which in turn necessitates escape analysis to determine whether
the boxed representation can be stack allocated.
However, esc.go used to unconditionally print escape analysis
decisions about OCONVIFACE, even for conversions that don't require
boxing (e.g., pointers, channels, maps, functions).
For test compatibility with esc.go, escape.go similarly printed these
useless diagnostics. This CL removes the diagnostics, and updates test
expectations accordingly.
Change-Id: I97c57a4a08e44d265bba516c78426ff4f2bf1e12
Reviewed-on: https://go-review.googlesource.com/c/go/+/192697
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
2019-08-30 11:56:30 -06:00
|
|
|
sink = *i
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
func caller6a() {
|
2020-09-08 22:09:01 -06:00
|
|
|
i := 0 // ERROR "moved to heap: i$"
|
|
|
|
p := &i // ERROR "moved to heap: p$"
|
2019-04-01 12:58:33 -06:00
|
|
|
p2 := &p
|
|
|
|
param6(&p2)
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
// **in -> heap
|
cmd/internal/gc: improve flow of input params to output params
This includes the following information in the per-function summary:
outK = paramJ encoded in outK bits for paramJ
outK = *paramJ encoded in outK bits for paramJ
heap = paramJ EscHeap
heap = *paramJ EscContentEscapes
Note that (currently) if the address of a parameter is taken and
returned, necessarily a heap allocation occurred to contain that
reference, and the heap can never refer to stack, therefore the
parameter and everything downstream from it escapes to the heap.
The per-function summary information now has a tuneable number of bits
(2 is probably noticeably better than 1, 3 is likely overkill, but it
is now easy to check and the -m debugging output includes information
that allows you to figure out if more would be better.)
A new test was added to check pointer flow through struct-typed and
*struct-typed parameters and returns; some of these are sensitive to
the number of summary bits, and ought to yield better results with a
more competent escape analysis algorithm. Another new test checks
(some) correctness with array parameters, results, and operations.
The old analysis inferred a piece of plan9 runtime was non-escaping by
counteracting overconservative analysis with buggy analysis; with the
bug fixed, the result was too conservative (and it's not easy to fix
in this framework) so the source code was tweaked to get the desired
result. A test was added against the discovered bug.
The escape analysis was further improved splitting the "level" into
3 parts, one tracking the conventional "level" and the other two
computing the highest-level-suffix-from-copy, which is used to
generally model the cancelling effect of indirection applied to
address-of.
With the improved escape analysis enabled, it was necessary to
modify one of the runtime tests because it now attempts to allocate
too much on the (small, fixed-size) G0 (system) stack and this
failed the test.
Compiling src/std after touching src/runtime/*.go with -m logging
turned on shows 420 fewer heap allocation sites (10538 vs 10968).
Profiling allocations in src/html/template with
for i in {1..5} ;
do go tool 6g -memprofile=mastx.${i}.prof -memprofilerate=1 *.go;
go tool pprof -alloc_objects -text mastx.${i}.prof ;
done
showed a 15% reduction in allocations performed by the compiler.
Update #3753
Update #4720
Fixes #10466
Change-Id: I0fd97d5f5ac527b45f49e2218d158a6e89951432
Reviewed-on: https://go-review.googlesource.com/8202
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
2015-03-26 14:36:15 -06:00
|
|
|
func param7(i ***int) { // ERROR "leaking param content: i$"
|
cmd/compile: silence esc diagnostics about directiface OCONVIFACEs
In general, a conversion to interface type may require values to be
boxed, which in turn necessitates escape analysis to determine whether
the boxed representation can be stack allocated.
However, esc.go used to unconditionally print escape analysis
decisions about OCONVIFACE, even for conversions that don't require
boxing (e.g., pointers, channels, maps, functions).
For test compatibility with esc.go, escape.go similarly printed these
useless diagnostics. This CL removes the diagnostics, and updates test
expectations accordingly.
Change-Id: I97c57a4a08e44d265bba516c78426ff4f2bf1e12
Reviewed-on: https://go-review.googlesource.com/c/go/+/192697
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
2019-08-30 11:56:30 -06:00
|
|
|
sink = **i
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
func caller7() {
|
2020-09-08 22:09:01 -06:00
|
|
|
i := 0 // ERROR "moved to heap: i$"
|
2019-09-26 18:21:50 -06:00
|
|
|
p := &i
|
2019-04-01 12:58:33 -06:00
|
|
|
p2 := &p
|
|
|
|
param7(&p2)
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
// **in -> heap
|
2019-09-12 11:18:03 -06:00
|
|
|
func param8(i **int) { // ERROR "i does not escape$"
|
2020-12-10 18:55:10 -07:00
|
|
|
sink = **i // ERROR "\*\(\*i\) escapes to heap"
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
func caller8() {
|
|
|
|
i := 0
|
2019-04-01 12:58:33 -06:00
|
|
|
p := &i
|
|
|
|
param8(&p)
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
// *in -> out
|
cmd/internal/gc: improve flow of input params to output params
This includes the following information in the per-function summary:
outK = paramJ encoded in outK bits for paramJ
outK = *paramJ encoded in outK bits for paramJ
heap = paramJ EscHeap
heap = *paramJ EscContentEscapes
Note that (currently) if the address of a parameter is taken and
returned, necessarily a heap allocation occurred to contain that
reference, and the heap can never refer to stack, therefore the
parameter and everything downstream from it escapes to the heap.
The per-function summary information now has a tuneable number of bits
(2 is probably noticeably better than 1, 3 is likely overkill, but it
is now easy to check and the -m debugging output includes information
that allows you to figure out if more would be better.)
A new test was added to check pointer flow through struct-typed and
*struct-typed parameters and returns; some of these are sensitive to
the number of summary bits, and ought to yield better results with a
more competent escape analysis algorithm. Another new test checks
(some) correctness with array parameters, results, and operations.
The old analysis inferred a piece of plan9 runtime was non-escaping by
counteracting overconservative analysis with buggy analysis; with the
bug fixed, the result was too conservative (and it's not easy to fix
in this framework) so the source code was tweaked to get the desired
result. A test was added against the discovered bug.
The escape analysis was further improved splitting the "level" into
3 parts, one tracking the conventional "level" and the other two
computing the highest-level-suffix-from-copy, which is used to
generally model the cancelling effect of indirection applied to
address-of.
With the improved escape analysis enabled, it was necessary to
modify one of the runtime tests because it now attempts to allocate
too much on the (small, fixed-size) G0 (system) stack and this
failed the test.
Compiling src/std after touching src/runtime/*.go with -m logging
turned on shows 420 fewer heap allocation sites (10538 vs 10968).
Profiling allocations in src/html/template with
for i in {1..5} ;
do go tool 6g -memprofile=mastx.${i}.prof -memprofilerate=1 *.go;
go tool pprof -alloc_objects -text mastx.${i}.prof ;
done
showed a 15% reduction in allocations performed by the compiler.
Update #3753
Update #4720
Fixes #10466
Change-Id: I0fd97d5f5ac527b45f49e2218d158a6e89951432
Reviewed-on: https://go-review.googlesource.com/8202
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
2015-03-26 14:36:15 -06:00
|
|
|
func param9(p ***int) **int { // ERROR "leaking param: p to result ~r1 level=1"
|
2015-02-19 09:10:31 -07:00
|
|
|
return *p
|
|
|
|
}
|
|
|
|
|
|
|
|
func caller9a() {
|
cmd/internal/gc: improve flow of input params to output params
This includes the following information in the per-function summary:
outK = paramJ encoded in outK bits for paramJ
outK = *paramJ encoded in outK bits for paramJ
heap = paramJ EscHeap
heap = *paramJ EscContentEscapes
Note that (currently) if the address of a parameter is taken and
returned, necessarily a heap allocation occurred to contain that
reference, and the heap can never refer to stack, therefore the
parameter and everything downstream from it escapes to the heap.
The per-function summary information now has a tuneable number of bits
(2 is probably noticeably better than 1, 3 is likely overkill, but it
is now easy to check and the -m debugging output includes information
that allows you to figure out if more would be better.)
A new test was added to check pointer flow through struct-typed and
*struct-typed parameters and returns; some of these are sensitive to
the number of summary bits, and ought to yield better results with a
more competent escape analysis algorithm. Another new test checks
(some) correctness with array parameters, results, and operations.
The old analysis inferred a piece of plan9 runtime was non-escaping by
counteracting overconservative analysis with buggy analysis; with the
bug fixed, the result was too conservative (and it's not easy to fix
in this framework) so the source code was tweaked to get the desired
result. A test was added against the discovered bug.
The escape analysis was further improved splitting the "level" into
3 parts, one tracking the conventional "level" and the other two
computing the highest-level-suffix-from-copy, which is used to
generally model the cancelling effect of indirection applied to
address-of.
With the improved escape analysis enabled, it was necessary to
modify one of the runtime tests because it now attempts to allocate
too much on the (small, fixed-size) G0 (system) stack and this
failed the test.
Compiling src/std after touching src/runtime/*.go with -m logging
turned on shows 420 fewer heap allocation sites (10538 vs 10968).
Profiling allocations in src/html/template with
for i in {1..5} ;
do go tool 6g -memprofile=mastx.${i}.prof -memprofilerate=1 *.go;
go tool pprof -alloc_objects -text mastx.${i}.prof ;
done
showed a 15% reduction in allocations performed by the compiler.
Update #3753
Update #4720
Fixes #10466
Change-Id: I0fd97d5f5ac527b45f49e2218d158a6e89951432
Reviewed-on: https://go-review.googlesource.com/8202
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
2015-03-26 14:36:15 -06:00
|
|
|
i := 0
|
2019-04-01 12:58:33 -06:00
|
|
|
p := &i
|
|
|
|
p2 := &p
|
|
|
|
_ = param9(&p2)
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
func caller9b() {
|
2020-09-08 22:09:01 -06:00
|
|
|
i := 0 // ERROR "moved to heap: i$"
|
|
|
|
p := &i // ERROR "moved to heap: p$"
|
2019-04-01 12:58:33 -06:00
|
|
|
p2 := &p
|
cmd/compile: silence esc diagnostics about directiface OCONVIFACEs
In general, a conversion to interface type may require values to be
boxed, which in turn necessitates escape analysis to determine whether
the boxed representation can be stack allocated.
However, esc.go used to unconditionally print escape analysis
decisions about OCONVIFACE, even for conversions that don't require
boxing (e.g., pointers, channels, maps, functions).
For test compatibility with esc.go, escape.go similarly printed these
useless diagnostics. This CL removes the diagnostics, and updates test
expectations accordingly.
Change-Id: I97c57a4a08e44d265bba516c78426ff4f2bf1e12
Reviewed-on: https://go-review.googlesource.com/c/go/+/192697
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
2019-08-30 11:56:30 -06:00
|
|
|
sink = param9(&p2)
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
// **in -> out
|
cmd/internal/gc: improve flow of input params to output params
This includes the following information in the per-function summary:
outK = paramJ encoded in outK bits for paramJ
outK = *paramJ encoded in outK bits for paramJ
heap = paramJ EscHeap
heap = *paramJ EscContentEscapes
Note that (currently) if the address of a parameter is taken and
returned, necessarily a heap allocation occurred to contain that
reference, and the heap can never refer to stack, therefore the
parameter and everything downstream from it escapes to the heap.
The per-function summary information now has a tuneable number of bits
(2 is probably noticeably better than 1, 3 is likely overkill, but it
is now easy to check and the -m debugging output includes information
that allows you to figure out if more would be better.)
A new test was added to check pointer flow through struct-typed and
*struct-typed parameters and returns; some of these are sensitive to
the number of summary bits, and ought to yield better results with a
more competent escape analysis algorithm. Another new test checks
(some) correctness with array parameters, results, and operations.
The old analysis inferred a piece of plan9 runtime was non-escaping by
counteracting overconservative analysis with buggy analysis; with the
bug fixed, the result was too conservative (and it's not easy to fix
in this framework) so the source code was tweaked to get the desired
result. A test was added against the discovered bug.
The escape analysis was further improved splitting the "level" into
3 parts, one tracking the conventional "level" and the other two
computing the highest-level-suffix-from-copy, which is used to
generally model the cancelling effect of indirection applied to
address-of.
With the improved escape analysis enabled, it was necessary to
modify one of the runtime tests because it now attempts to allocate
too much on the (small, fixed-size) G0 (system) stack and this
failed the test.
Compiling src/std after touching src/runtime/*.go with -m logging
turned on shows 420 fewer heap allocation sites (10538 vs 10968).
Profiling allocations in src/html/template with
for i in {1..5} ;
do go tool 6g -memprofile=mastx.${i}.prof -memprofilerate=1 *.go;
go tool pprof -alloc_objects -text mastx.${i}.prof ;
done
showed a 15% reduction in allocations performed by the compiler.
Update #3753
Update #4720
Fixes #10466
Change-Id: I0fd97d5f5ac527b45f49e2218d158a6e89951432
Reviewed-on: https://go-review.googlesource.com/8202
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
2015-03-26 14:36:15 -06:00
|
|
|
func param10(p ***int) *int { // ERROR "leaking param: p to result ~r1 level=2"
|
2015-02-19 09:10:31 -07:00
|
|
|
return **p
|
|
|
|
}
|
|
|
|
|
|
|
|
func caller10a() {
|
cmd/internal/gc: improve flow of input params to output params
This includes the following information in the per-function summary:
outK = paramJ encoded in outK bits for paramJ
outK = *paramJ encoded in outK bits for paramJ
heap = paramJ EscHeap
heap = *paramJ EscContentEscapes
Note that (currently) if the address of a parameter is taken and
returned, necessarily a heap allocation occurred to contain that
reference, and the heap can never refer to stack, therefore the
parameter and everything downstream from it escapes to the heap.
The per-function summary information now has a tuneable number of bits
(2 is probably noticeably better than 1, 3 is likely overkill, but it
is now easy to check and the -m debugging output includes information
that allows you to figure out if more would be better.)
A new test was added to check pointer flow through struct-typed and
*struct-typed parameters and returns; some of these are sensitive to
the number of summary bits, and ought to yield better results with a
more competent escape analysis algorithm. Another new test checks
(some) correctness with array parameters, results, and operations.
The old analysis inferred a piece of plan9 runtime was non-escaping by
counteracting overconservative analysis with buggy analysis; with the
bug fixed, the result was too conservative (and it's not easy to fix
in this framework) so the source code was tweaked to get the desired
result. A test was added against the discovered bug.
The escape analysis was further improved splitting the "level" into
3 parts, one tracking the conventional "level" and the other two
computing the highest-level-suffix-from-copy, which is used to
generally model the cancelling effect of indirection applied to
address-of.
With the improved escape analysis enabled, it was necessary to
modify one of the runtime tests because it now attempts to allocate
too much on the (small, fixed-size) G0 (system) stack and this
failed the test.
Compiling src/std after touching src/runtime/*.go with -m logging
turned on shows 420 fewer heap allocation sites (10538 vs 10968).
Profiling allocations in src/html/template with
for i in {1..5} ;
do go tool 6g -memprofile=mastx.${i}.prof -memprofilerate=1 *.go;
go tool pprof -alloc_objects -text mastx.${i}.prof ;
done
showed a 15% reduction in allocations performed by the compiler.
Update #3753
Update #4720
Fixes #10466
Change-Id: I0fd97d5f5ac527b45f49e2218d158a6e89951432
Reviewed-on: https://go-review.googlesource.com/8202
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
2015-03-26 14:36:15 -06:00
|
|
|
i := 0
|
2019-04-01 12:58:33 -06:00
|
|
|
p := &i
|
|
|
|
p2 := &p
|
|
|
|
_ = param10(&p2)
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
func caller10b() {
|
2020-09-08 22:09:01 -06:00
|
|
|
i := 0 // ERROR "moved to heap: i$"
|
2019-04-01 12:58:33 -06:00
|
|
|
p := &i
|
|
|
|
p2 := &p
|
cmd/compile: silence esc diagnostics about directiface OCONVIFACEs
In general, a conversion to interface type may require values to be
boxed, which in turn necessitates escape analysis to determine whether
the boxed representation can be stack allocated.
However, esc.go used to unconditionally print escape analysis
decisions about OCONVIFACE, even for conversions that don't require
boxing (e.g., pointers, channels, maps, functions).
For test compatibility with esc.go, escape.go similarly printed these
useless diagnostics. This CL removes the diagnostics, and updates test
expectations accordingly.
Change-Id: I97c57a4a08e44d265bba516c78426ff4f2bf1e12
Reviewed-on: https://go-review.googlesource.com/c/go/+/192697
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
2019-08-30 11:56:30 -06:00
|
|
|
sink = param10(&p2)
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
|
|
|
|
cmd/internal/gc: improve flow of input params to output params
This includes the following information in the per-function summary:
outK = paramJ encoded in outK bits for paramJ
outK = *paramJ encoded in outK bits for paramJ
heap = paramJ EscHeap
heap = *paramJ EscContentEscapes
Note that (currently) if the address of a parameter is taken and
returned, necessarily a heap allocation occurred to contain that
reference, and the heap can never refer to stack, therefore the
parameter and everything downstream from it escapes to the heap.
The per-function summary information now has a tuneable number of bits
(2 is probably noticeably better than 1, 3 is likely overkill, but it
is now easy to check and the -m debugging output includes information
that allows you to figure out if more would be better.)
A new test was added to check pointer flow through struct-typed and
*struct-typed parameters and returns; some of these are sensitive to
the number of summary bits, and ought to yield better results with a
more competent escape analysis algorithm. Another new test checks
(some) correctness with array parameters, results, and operations.
The old analysis inferred a piece of plan9 runtime was non-escaping by
counteracting overconservative analysis with buggy analysis; with the
bug fixed, the result was too conservative (and it's not easy to fix
in this framework) so the source code was tweaked to get the desired
result. A test was added against the discovered bug.
The escape analysis was further improved splitting the "level" into
3 parts, one tracking the conventional "level" and the other two
computing the highest-level-suffix-from-copy, which is used to
generally model the cancelling effect of indirection applied to
address-of.
With the improved escape analysis enabled, it was necessary to
modify one of the runtime tests because it now attempts to allocate
too much on the (small, fixed-size) G0 (system) stack and this
failed the test.
Compiling src/std after touching src/runtime/*.go with -m logging
turned on shows 420 fewer heap allocation sites (10538 vs 10968).
Profiling allocations in src/html/template with
for i in {1..5} ;
do go tool 6g -memprofile=mastx.${i}.prof -memprofilerate=1 *.go;
go tool pprof -alloc_objects -text mastx.${i}.prof ;
done
showed a 15% reduction in allocations performed by the compiler.
Update #3753
Update #4720
Fixes #10466
Change-Id: I0fd97d5f5ac527b45f49e2218d158a6e89951432
Reviewed-on: https://go-review.googlesource.com/8202
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
2015-03-26 14:36:15 -06:00
|
|
|
// in escapes to heap (address of param taken and returned)
|
2015-02-19 09:10:31 -07:00
|
|
|
func param11(i **int) ***int { // ERROR "moved to heap: i$"
|
2019-04-01 12:58:33 -06:00
|
|
|
return &i
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
func caller11a() {
|
2020-09-08 22:09:01 -06:00
|
|
|
i := 0 // ERROR "moved to heap: i"
|
|
|
|
p := &i // ERROR "moved to heap: p"
|
2019-04-01 12:58:33 -06:00
|
|
|
_ = param11(&p)
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
func caller11b() {
|
2020-09-08 22:09:01 -06:00
|
|
|
i := 0 // ERROR "moved to heap: i$"
|
|
|
|
p := &i // ERROR "moved to heap: p$"
|
cmd/compile: silence esc diagnostics about directiface OCONVIFACEs
In general, a conversion to interface type may require values to be
boxed, which in turn necessitates escape analysis to determine whether
the boxed representation can be stack allocated.
However, esc.go used to unconditionally print escape analysis
decisions about OCONVIFACE, even for conversions that don't require
boxing (e.g., pointers, channels, maps, functions).
For test compatibility with esc.go, escape.go similarly printed these
useless diagnostics. This CL removes the diagnostics, and updates test
expectations accordingly.
Change-Id: I97c57a4a08e44d265bba516c78426ff4f2bf1e12
Reviewed-on: https://go-review.googlesource.com/c/go/+/192697
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
2019-08-30 11:56:30 -06:00
|
|
|
sink = param11(&p)
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
|
|
|
|
cmd/internal/gc: improve flow of input params to output params
This includes the following information in the per-function summary:
outK = paramJ encoded in outK bits for paramJ
outK = *paramJ encoded in outK bits for paramJ
heap = paramJ EscHeap
heap = *paramJ EscContentEscapes
Note that (currently) if the address of a parameter is taken and
returned, necessarily a heap allocation occurred to contain that
reference, and the heap can never refer to stack, therefore the
parameter and everything downstream from it escapes to the heap.
The per-function summary information now has a tuneable number of bits
(2 is probably noticeably better than 1, 3 is likely overkill, but it
is now easy to check and the -m debugging output includes information
that allows you to figure out if more would be better.)
A new test was added to check pointer flow through struct-typed and
*struct-typed parameters and returns; some of these are sensitive to
the number of summary bits, and ought to yield better results with a
more competent escape analysis algorithm. Another new test checks
(some) correctness with array parameters, results, and operations.
The old analysis inferred a piece of plan9 runtime was non-escaping by
counteracting overconservative analysis with buggy analysis; with the
bug fixed, the result was too conservative (and it's not easy to fix
in this framework) so the source code was tweaked to get the desired
result. A test was added against the discovered bug.
The escape analysis was further improved splitting the "level" into
3 parts, one tracking the conventional "level" and the other two
computing the highest-level-suffix-from-copy, which is used to
generally model the cancelling effect of indirection applied to
address-of.
With the improved escape analysis enabled, it was necessary to
modify one of the runtime tests because it now attempts to allocate
too much on the (small, fixed-size) G0 (system) stack and this
failed the test.
Compiling src/std after touching src/runtime/*.go with -m logging
turned on shows 420 fewer heap allocation sites (10538 vs 10968).
Profiling allocations in src/html/template with
for i in {1..5} ;
do go tool 6g -memprofile=mastx.${i}.prof -memprofilerate=1 *.go;
go tool pprof -alloc_objects -text mastx.${i}.prof ;
done
showed a 15% reduction in allocations performed by the compiler.
Update #3753
Update #4720
Fixes #10466
Change-Id: I0fd97d5f5ac527b45f49e2218d158a6e89951432
Reviewed-on: https://go-review.googlesource.com/8202
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
2015-03-26 14:36:15 -06:00
|
|
|
func caller11c() { // GOOD
|
2020-09-08 22:09:01 -06:00
|
|
|
i := 0 // ERROR "moved to heap: i$"
|
|
|
|
p := &i // ERROR "moved to heap: p"
|
cmd/compile: silence esc diagnostics about directiface OCONVIFACEs
In general, a conversion to interface type may require values to be
boxed, which in turn necessitates escape analysis to determine whether
the boxed representation can be stack allocated.
However, esc.go used to unconditionally print escape analysis
decisions about OCONVIFACE, even for conversions that don't require
boxing (e.g., pointers, channels, maps, functions).
For test compatibility with esc.go, escape.go similarly printed these
useless diagnostics. This CL removes the diagnostics, and updates test
expectations accordingly.
Change-Id: I97c57a4a08e44d265bba516c78426ff4f2bf1e12
Reviewed-on: https://go-review.googlesource.com/c/go/+/192697
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
2019-08-30 11:56:30 -06:00
|
|
|
sink = *param11(&p)
|
cmd/internal/gc: improve flow of input params to output params
This includes the following information in the per-function summary:
outK = paramJ encoded in outK bits for paramJ
outK = *paramJ encoded in outK bits for paramJ
heap = paramJ EscHeap
heap = *paramJ EscContentEscapes
Note that (currently) if the address of a parameter is taken and
returned, necessarily a heap allocation occurred to contain that
reference, and the heap can never refer to stack, therefore the
parameter and everything downstream from it escapes to the heap.
The per-function summary information now has a tuneable number of bits
(2 is probably noticeably better than 1, 3 is likely overkill, but it
is now easy to check and the -m debugging output includes information
that allows you to figure out if more would be better.)
A new test was added to check pointer flow through struct-typed and
*struct-typed parameters and returns; some of these are sensitive to
the number of summary bits, and ought to yield better results with a
more competent escape analysis algorithm. Another new test checks
(some) correctness with array parameters, results, and operations.
The old analysis inferred a piece of plan9 runtime was non-escaping by
counteracting overconservative analysis with buggy analysis; with the
bug fixed, the result was too conservative (and it's not easy to fix
in this framework) so the source code was tweaked to get the desired
result. A test was added against the discovered bug.
The escape analysis was further improved splitting the "level" into
3 parts, one tracking the conventional "level" and the other two
computing the highest-level-suffix-from-copy, which is used to
generally model the cancelling effect of indirection applied to
address-of.
With the improved escape analysis enabled, it was necessary to
modify one of the runtime tests because it now attempts to allocate
too much on the (small, fixed-size) G0 (system) stack and this
failed the test.
Compiling src/std after touching src/runtime/*.go with -m logging
turned on shows 420 fewer heap allocation sites (10538 vs 10968).
Profiling allocations in src/html/template with
for i in {1..5} ;
do go tool 6g -memprofile=mastx.${i}.prof -memprofilerate=1 *.go;
go tool pprof -alloc_objects -text mastx.${i}.prof ;
done
showed a 15% reduction in allocations performed by the compiler.
Update #3753
Update #4720
Fixes #10466
Change-Id: I0fd97d5f5ac527b45f49e2218d158a6e89951432
Reviewed-on: https://go-review.googlesource.com/8202
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
2015-03-26 14:36:15 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
func caller11d() {
|
2020-09-08 22:09:01 -06:00
|
|
|
i := 0 // ERROR "moved to heap: i$"
|
|
|
|
p := &i // ERROR "moved to heap: p"
|
2019-04-01 12:58:33 -06:00
|
|
|
p2 := &p
|
cmd/compile: silence esc diagnostics about directiface OCONVIFACEs
In general, a conversion to interface type may require values to be
boxed, which in turn necessitates escape analysis to determine whether
the boxed representation can be stack allocated.
However, esc.go used to unconditionally print escape analysis
decisions about OCONVIFACE, even for conversions that don't require
boxing (e.g., pointers, channels, maps, functions).
For test compatibility with esc.go, escape.go similarly printed these
useless diagnostics. This CL removes the diagnostics, and updates test
expectations accordingly.
Change-Id: I97c57a4a08e44d265bba516c78426ff4f2bf1e12
Reviewed-on: https://go-review.googlesource.com/c/go/+/192697
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
2019-08-30 11:56:30 -06:00
|
|
|
sink = param11(p2)
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
// &in -> rcvr
|
|
|
|
type Indir struct {
|
|
|
|
p ***int
|
|
|
|
}
|
|
|
|
|
2019-09-12 11:18:03 -06:00
|
|
|
func (r *Indir) param12(i **int) { // ERROR "r does not escape$" "moved to heap: i$"
|
2019-04-01 12:58:33 -06:00
|
|
|
r.p = &i
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
func caller12a() {
|
|
|
|
i := 0 // ERROR "moved to heap: i$"
|
2019-04-01 12:58:33 -06:00
|
|
|
p := &i // ERROR "moved to heap: p$"
|
2015-02-19 09:10:31 -07:00
|
|
|
var r Indir
|
2019-04-01 12:58:33 -06:00
|
|
|
r.param12(&p)
|
2015-02-19 09:10:31 -07:00
|
|
|
_ = r
|
|
|
|
}
|
|
|
|
|
|
|
|
func caller12b() {
|
|
|
|
i := 0 // ERROR "moved to heap: i$"
|
2019-04-01 12:58:33 -06:00
|
|
|
p := &i // ERROR "moved to heap: p$"
|
2020-09-08 22:09:01 -06:00
|
|
|
r := &Indir{} // ERROR "&Indir{} does not escape$"
|
2019-04-01 12:58:33 -06:00
|
|
|
r.param12(&p)
|
2015-02-19 09:10:31 -07:00
|
|
|
_ = r
|
|
|
|
}
|
|
|
|
|
|
|
|
func caller12c() {
|
|
|
|
i := 0 // ERROR "moved to heap: i$"
|
2019-04-01 12:58:33 -06:00
|
|
|
p := &i // ERROR "moved to heap: p$"
|
2015-02-19 09:10:31 -07:00
|
|
|
r := Indir{}
|
2019-04-01 12:58:33 -06:00
|
|
|
r.param12(&p)
|
cmd/compile: silence esc diagnostics about directiface OCONVIFACEs
In general, a conversion to interface type may require values to be
boxed, which in turn necessitates escape analysis to determine whether
the boxed representation can be stack allocated.
However, esc.go used to unconditionally print escape analysis
decisions about OCONVIFACE, even for conversions that don't require
boxing (e.g., pointers, channels, maps, functions).
For test compatibility with esc.go, escape.go similarly printed these
useless diagnostics. This CL removes the diagnostics, and updates test
expectations accordingly.
Change-Id: I97c57a4a08e44d265bba516c78426ff4f2bf1e12
Reviewed-on: https://go-review.googlesource.com/c/go/+/192697
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
2019-08-30 11:56:30 -06:00
|
|
|
sink = r
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
func caller12d() {
|
|
|
|
i := 0 // ERROR "moved to heap: i$"
|
2019-04-01 12:58:33 -06:00
|
|
|
p := &i // ERROR "moved to heap: p$"
|
2015-02-19 09:10:31 -07:00
|
|
|
r := Indir{}
|
2019-04-01 12:58:33 -06:00
|
|
|
r.param12(&p)
|
cmd/compile: silence esc diagnostics about directiface OCONVIFACEs
In general, a conversion to interface type may require values to be
boxed, which in turn necessitates escape analysis to determine whether
the boxed representation can be stack allocated.
However, esc.go used to unconditionally print escape analysis
decisions about OCONVIFACE, even for conversions that don't require
boxing (e.g., pointers, channels, maps, functions).
For test compatibility with esc.go, escape.go similarly printed these
useless diagnostics. This CL removes the diagnostics, and updates test
expectations accordingly.
Change-Id: I97c57a4a08e44d265bba516c78426ff4f2bf1e12
Reviewed-on: https://go-review.googlesource.com/c/go/+/192697
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
2019-08-30 11:56:30 -06:00
|
|
|
sink = **r.p
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
// in -> value rcvr
|
|
|
|
type Val struct {
|
|
|
|
p **int
|
|
|
|
}
|
|
|
|
|
2019-09-12 11:18:03 -06:00
|
|
|
func (v Val) param13(i *int) { // ERROR "v does not escape$" "leaking param: i$"
|
2015-02-19 09:10:31 -07:00
|
|
|
*v.p = i
|
|
|
|
}
|
|
|
|
|
|
|
|
func caller13a() {
|
|
|
|
i := 0 // ERROR "moved to heap: i$"
|
|
|
|
var p *int
|
|
|
|
var v Val
|
2019-04-01 12:58:33 -06:00
|
|
|
v.p = &p
|
|
|
|
v.param13(&i)
|
2015-02-19 09:10:31 -07:00
|
|
|
_ = v
|
|
|
|
}
|
|
|
|
|
|
|
|
func caller13b() {
|
|
|
|
i := 0 // ERROR "moved to heap: i$"
|
|
|
|
var p *int
|
2019-04-01 12:58:33 -06:00
|
|
|
v := Val{&p}
|
|
|
|
v.param13(&i)
|
2015-02-19 09:10:31 -07:00
|
|
|
_ = v
|
|
|
|
}
|
|
|
|
|
|
|
|
func caller13c() {
|
|
|
|
i := 0 // ERROR "moved to heap: i$"
|
|
|
|
var p *int
|
2020-09-08 22:09:01 -06:00
|
|
|
v := &Val{&p} // ERROR "&Val{...} does not escape$"
|
2019-04-01 12:58:33 -06:00
|
|
|
v.param13(&i)
|
2015-02-19 09:10:31 -07:00
|
|
|
_ = v
|
|
|
|
}
|
|
|
|
|
|
|
|
func caller13d() {
|
|
|
|
i := 0 // ERROR "moved to heap: i$"
|
|
|
|
var p *int // ERROR "moved to heap: p$"
|
|
|
|
var v Val
|
2019-04-01 12:58:33 -06:00
|
|
|
v.p = &p
|
|
|
|
v.param13(&i)
|
cmd/compile: silence esc diagnostics about directiface OCONVIFACEs
In general, a conversion to interface type may require values to be
boxed, which in turn necessitates escape analysis to determine whether
the boxed representation can be stack allocated.
However, esc.go used to unconditionally print escape analysis
decisions about OCONVIFACE, even for conversions that don't require
boxing (e.g., pointers, channels, maps, functions).
For test compatibility with esc.go, escape.go similarly printed these
useless diagnostics. This CL removes the diagnostics, and updates test
expectations accordingly.
Change-Id: I97c57a4a08e44d265bba516c78426ff4f2bf1e12
Reviewed-on: https://go-review.googlesource.com/c/go/+/192697
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
2019-08-30 11:56:30 -06:00
|
|
|
sink = v
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
func caller13e() {
|
2020-09-08 22:09:01 -06:00
|
|
|
i := 0 // ERROR "moved to heap: i$"
|
|
|
|
var p *int // ERROR "moved to heap: p$"
|
2019-04-01 12:58:33 -06:00
|
|
|
v := Val{&p}
|
|
|
|
v.param13(&i)
|
cmd/compile: silence esc diagnostics about directiface OCONVIFACEs
In general, a conversion to interface type may require values to be
boxed, which in turn necessitates escape analysis to determine whether
the boxed representation can be stack allocated.
However, esc.go used to unconditionally print escape analysis
decisions about OCONVIFACE, even for conversions that don't require
boxing (e.g., pointers, channels, maps, functions).
For test compatibility with esc.go, escape.go similarly printed these
useless diagnostics. This CL removes the diagnostics, and updates test
expectations accordingly.
Change-Id: I97c57a4a08e44d265bba516c78426ff4f2bf1e12
Reviewed-on: https://go-review.googlesource.com/c/go/+/192697
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
2019-08-30 11:56:30 -06:00
|
|
|
sink = v
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
func caller13f() {
|
|
|
|
i := 0 // ERROR "moved to heap: i$"
|
|
|
|
var p *int // ERROR "moved to heap: p$"
|
2020-09-08 22:09:01 -06:00
|
|
|
v := &Val{&p} // ERROR "&Val{...} escapes to heap$"
|
2019-04-01 12:58:33 -06:00
|
|
|
v.param13(&i)
|
cmd/compile: silence esc diagnostics about directiface OCONVIFACEs
In general, a conversion to interface type may require values to be
boxed, which in turn necessitates escape analysis to determine whether
the boxed representation can be stack allocated.
However, esc.go used to unconditionally print escape analysis
decisions about OCONVIFACE, even for conversions that don't require
boxing (e.g., pointers, channels, maps, functions).
For test compatibility with esc.go, escape.go similarly printed these
useless diagnostics. This CL removes the diagnostics, and updates test
expectations accordingly.
Change-Id: I97c57a4a08e44d265bba516c78426ff4f2bf1e12
Reviewed-on: https://go-review.googlesource.com/c/go/+/192697
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
2019-08-30 11:56:30 -06:00
|
|
|
sink = v
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
func caller13g() {
|
|
|
|
i := 0 // ERROR "moved to heap: i$"
|
|
|
|
var p *int
|
2019-04-01 12:58:33 -06:00
|
|
|
v := Val{&p}
|
|
|
|
v.param13(&i)
|
cmd/compile: silence esc diagnostics about directiface OCONVIFACEs
In general, a conversion to interface type may require values to be
boxed, which in turn necessitates escape analysis to determine whether
the boxed representation can be stack allocated.
However, esc.go used to unconditionally print escape analysis
decisions about OCONVIFACE, even for conversions that don't require
boxing (e.g., pointers, channels, maps, functions).
For test compatibility with esc.go, escape.go similarly printed these
useless diagnostics. This CL removes the diagnostics, and updates test
expectations accordingly.
Change-Id: I97c57a4a08e44d265bba516c78426ff4f2bf1e12
Reviewed-on: https://go-review.googlesource.com/c/go/+/192697
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
2019-08-30 11:56:30 -06:00
|
|
|
sink = *v.p
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
func caller13h() {
|
|
|
|
i := 0 // ERROR "moved to heap: i$"
|
|
|
|
var p *int
|
2020-09-08 22:09:01 -06:00
|
|
|
v := &Val{&p} // ERROR "&Val{...} does not escape$"
|
2019-04-01 12:58:33 -06:00
|
|
|
v.param13(&i)
|
2020-12-10 18:55:10 -07:00
|
|
|
sink = **v.p // ERROR "\*\(\*v\.p\) escapes to heap"
|
2015-02-19 09:10:31 -07:00
|
|
|
}
|
cmd/internal/gc: improve flow of input params to output params
This includes the following information in the per-function summary:
outK = paramJ encoded in outK bits for paramJ
outK = *paramJ encoded in outK bits for paramJ
heap = paramJ EscHeap
heap = *paramJ EscContentEscapes
Note that (currently) if the address of a parameter is taken and
returned, necessarily a heap allocation occurred to contain that
reference, and the heap can never refer to stack, therefore the
parameter and everything downstream from it escapes to the heap.
The per-function summary information now has a tuneable number of bits
(2 is probably noticeably better than 1, 3 is likely overkill, but it
is now easy to check and the -m debugging output includes information
that allows you to figure out if more would be better.)
A new test was added to check pointer flow through struct-typed and
*struct-typed parameters and returns; some of these are sensitive to
the number of summary bits, and ought to yield better results with a
more competent escape analysis algorithm. Another new test checks
(some) correctness with array parameters, results, and operations.
The old analysis inferred a piece of plan9 runtime was non-escaping by
counteracting overconservative analysis with buggy analysis; with the
bug fixed, the result was too conservative (and it's not easy to fix
in this framework) so the source code was tweaked to get the desired
result. A test was added against the discovered bug.
The escape analysis was further improved splitting the "level" into
3 parts, one tracking the conventional "level" and the other two
computing the highest-level-suffix-from-copy, which is used to
generally model the cancelling effect of indirection applied to
address-of.
With the improved escape analysis enabled, it was necessary to
modify one of the runtime tests because it now attempts to allocate
too much on the (small, fixed-size) G0 (system) stack and this
failed the test.
Compiling src/std after touching src/runtime/*.go with -m logging
turned on shows 420 fewer heap allocation sites (10538 vs 10968).
Profiling allocations in src/html/template with
for i in {1..5} ;
do go tool 6g -memprofile=mastx.${i}.prof -memprofilerate=1 *.go;
go tool pprof -alloc_objects -text mastx.${i}.prof ;
done
showed a 15% reduction in allocations performed by the compiler.
Update #3753
Update #4720
Fixes #10466
Change-Id: I0fd97d5f5ac527b45f49e2218d158a6e89951432
Reviewed-on: https://go-review.googlesource.com/8202
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
2015-03-26 14:36:15 -06:00
|
|
|
|
|
|
|
type Node struct {
|
|
|
|
p *Node
|
|
|
|
}
|
|
|
|
|
|
|
|
var Sink *Node
|
|
|
|
|
|
|
|
func f(x *Node) { // ERROR "leaking param content: x"
|
2020-09-08 22:09:01 -06:00
|
|
|
Sink = &Node{x.p} // ERROR "&Node{...} escapes to heap"
|
cmd/internal/gc: improve flow of input params to output params
This includes the following information in the per-function summary:
outK = paramJ encoded in outK bits for paramJ
outK = *paramJ encoded in outK bits for paramJ
heap = paramJ EscHeap
heap = *paramJ EscContentEscapes
Note that (currently) if the address of a parameter is taken and
returned, necessarily a heap allocation occurred to contain that
reference, and the heap can never refer to stack, therefore the
parameter and everything downstream from it escapes to the heap.
The per-function summary information now has a tuneable number of bits
(2 is probably noticeably better than 1, 3 is likely overkill, but it
is now easy to check and the -m debugging output includes information
that allows you to figure out if more would be better.)
A new test was added to check pointer flow through struct-typed and
*struct-typed parameters and returns; some of these are sensitive to
the number of summary bits, and ought to yield better results with a
more competent escape analysis algorithm. Another new test checks
(some) correctness with array parameters, results, and operations.
The old analysis inferred a piece of plan9 runtime was non-escaping by
counteracting overconservative analysis with buggy analysis; with the
bug fixed, the result was too conservative (and it's not easy to fix
in this framework) so the source code was tweaked to get the desired
result. A test was added against the discovered bug.
The escape analysis was further improved splitting the "level" into
3 parts, one tracking the conventional "level" and the other two
computing the highest-level-suffix-from-copy, which is used to
generally model the cancelling effect of indirection applied to
address-of.
With the improved escape analysis enabled, it was necessary to
modify one of the runtime tests because it now attempts to allocate
too much on the (small, fixed-size) G0 (system) stack and this
failed the test.
Compiling src/std after touching src/runtime/*.go with -m logging
turned on shows 420 fewer heap allocation sites (10538 vs 10968).
Profiling allocations in src/html/template with
for i in {1..5} ;
do go tool 6g -memprofile=mastx.${i}.prof -memprofilerate=1 *.go;
go tool pprof -alloc_objects -text mastx.${i}.prof ;
done
showed a 15% reduction in allocations performed by the compiler.
Update #3753
Update #4720
Fixes #10466
Change-Id: I0fd97d5f5ac527b45f49e2218d158a6e89951432
Reviewed-on: https://go-review.googlesource.com/8202
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
2015-03-26 14:36:15 -06:00
|
|
|
}
|
|
|
|
|
cmd/compile: update escape analysis tests for newescape
The new escape analysis implementation tries to emit debugging
diagnostics that are compatible with the existing implementation, but
there's a handful of cases that are easier to handle by updating the
test expectations instead.
For regress tests that need updating, the original file is copied to
oldescapeXXX.go.go with -newescape=false added to the //errorcheck
line, while the file is updated in place with -newescape=true and new
test requirements.
Notable test changes:
1) escape_because.go looks for a lot of detailed internal debugging
messages that are fairly particular to how esc.go works and that I
haven't attempted to port over to escape.go yet.
2) There are a lot of "leaking param: x to result ~r1 level=-1"
messages for code like
func(p *int) *T { return &T{p} }
that were simply wrong. Here &T must be heap allocated unconditionally
(because it's being returned); and since p is stored into it, p
escapes unconditionally too. esc.go incorrectly reports that p escapes
conditionally only if the returned pointer escaped.
3) esc.go used to print each "leaking param" analysis result as it
discovered them, which could lead to redundant messages (e.g., that a
param leaks at level=0 and level=1). escape.go instead prints
everything at the end, once it knows the shortest path to each sink.
4) esc.go didn't precisely model direct-interface types, resulting in
some values unnecessarily escaping to the heap when stored into
non-escaping interface values.
5) For functions written in assembly, esc.go only printed "does not
escape" messages, whereas escape.go prints "does not escape" or
"leaking param" as appropriate, consistent with the behavior for
functions written in Go.
6) 12 tests included "BAD" annotations identifying cases where esc.go
was unnecessarily heap allocating something. These are all fixed by
escape.go.
Updates #23109.
Change-Id: Iabc9eb14c94c9cadde3b183478d1fd54f013502f
Reviewed-on: https://go-review.googlesource.com/c/go/+/170447
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
2019-04-02 15:44:13 -06:00
|
|
|
func g(x *Node) *Node { // ERROR "leaking param content: x"
|
2020-09-08 22:09:01 -06:00
|
|
|
return &Node{x.p} // ERROR "&Node{...} escapes to heap"
|
cmd/internal/gc: improve flow of input params to output params
This includes the following information in the per-function summary:
outK = paramJ encoded in outK bits for paramJ
outK = *paramJ encoded in outK bits for paramJ
heap = paramJ EscHeap
heap = *paramJ EscContentEscapes
Note that (currently) if the address of a parameter is taken and
returned, necessarily a heap allocation occurred to contain that
reference, and the heap can never refer to stack, therefore the
parameter and everything downstream from it escapes to the heap.
The per-function summary information now has a tuneable number of bits
(2 is probably noticeably better than 1, 3 is likely overkill, but it
is now easy to check and the -m debugging output includes information
that allows you to figure out if more would be better.)
A new test was added to check pointer flow through struct-typed and
*struct-typed parameters and returns; some of these are sensitive to
the number of summary bits, and ought to yield better results with a
more competent escape analysis algorithm. Another new test checks
(some) correctness with array parameters, results, and operations.
The old analysis inferred a piece of plan9 runtime was non-escaping by
counteracting overconservative analysis with buggy analysis; with the
bug fixed, the result was too conservative (and it's not easy to fix
in this framework) so the source code was tweaked to get the desired
result. A test was added against the discovered bug.
The escape analysis was further improved splitting the "level" into
3 parts, one tracking the conventional "level" and the other two
computing the highest-level-suffix-from-copy, which is used to
generally model the cancelling effect of indirection applied to
address-of.
With the improved escape analysis enabled, it was necessary to
modify one of the runtime tests because it now attempts to allocate
too much on the (small, fixed-size) G0 (system) stack and this
failed the test.
Compiling src/std after touching src/runtime/*.go with -m logging
turned on shows 420 fewer heap allocation sites (10538 vs 10968).
Profiling allocations in src/html/template with
for i in {1..5} ;
do go tool 6g -memprofile=mastx.${i}.prof -memprofilerate=1 *.go;
go tool pprof -alloc_objects -text mastx.${i}.prof ;
done
showed a 15% reduction in allocations performed by the compiler.
Update #3753
Update #4720
Fixes #10466
Change-Id: I0fd97d5f5ac527b45f49e2218d158a6e89951432
Reviewed-on: https://go-review.googlesource.com/8202
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
2015-03-26 14:36:15 -06:00
|
|
|
}
|
|
|
|
|
|
|
|
func h(x *Node) { // ERROR "leaking param: x"
|
2020-09-08 22:09:01 -06:00
|
|
|
y := &Node{x} // ERROR "&Node{...} does not escape"
|
cmd/internal/gc: improve flow of input params to output params
This includes the following information in the per-function summary:
outK = paramJ encoded in outK bits for paramJ
outK = *paramJ encoded in outK bits for paramJ
heap = paramJ EscHeap
heap = *paramJ EscContentEscapes
Note that (currently) if the address of a parameter is taken and
returned, necessarily a heap allocation occurred to contain that
reference, and the heap can never refer to stack, therefore the
parameter and everything downstream from it escapes to the heap.
The per-function summary information now has a tuneable number of bits
(2 is probably noticeably better than 1, 3 is likely overkill, but it
is now easy to check and the -m debugging output includes information
that allows you to figure out if more would be better.)
A new test was added to check pointer flow through struct-typed and
*struct-typed parameters and returns; some of these are sensitive to
the number of summary bits, and ought to yield better results with a
more competent escape analysis algorithm. Another new test checks
(some) correctness with array parameters, results, and operations.
The old analysis inferred a piece of plan9 runtime was non-escaping by
counteracting overconservative analysis with buggy analysis; with the
bug fixed, the result was too conservative (and it's not easy to fix
in this framework) so the source code was tweaked to get the desired
result. A test was added against the discovered bug.
The escape analysis was further improved splitting the "level" into
3 parts, one tracking the conventional "level" and the other two
computing the highest-level-suffix-from-copy, which is used to
generally model the cancelling effect of indirection applied to
address-of.
With the improved escape analysis enabled, it was necessary to
modify one of the runtime tests because it now attempts to allocate
too much on the (small, fixed-size) G0 (system) stack and this
failed the test.
Compiling src/std after touching src/runtime/*.go with -m logging
turned on shows 420 fewer heap allocation sites (10538 vs 10968).
Profiling allocations in src/html/template with
for i in {1..5} ;
do go tool 6g -memprofile=mastx.${i}.prof -memprofilerate=1 *.go;
go tool pprof -alloc_objects -text mastx.${i}.prof ;
done
showed a 15% reduction in allocations performed by the compiler.
Update #3753
Update #4720
Fixes #10466
Change-Id: I0fd97d5f5ac527b45f49e2218d158a6e89951432
Reviewed-on: https://go-review.googlesource.com/8202
Run-TryBot: David Chase <drchase@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Russ Cox <rsc@golang.org>
2015-03-26 14:36:15 -06:00
|
|
|
Sink = g(y)
|
|
|
|
f(y)
|
|
|
|
}
|
cmd/compile: flow interface data to heap if CONVIFACE of a non-direct interface escapes
Consider the following code:
func f(x []*T) interface{} {
return x
}
It returns an interface that holds a heap copy of x (by calling
convT2I or friend), therefore x escape to heap. The current
escape analysis only recognizes that x flows to the result. This
is not sufficient, since if the result does not escape, x's
content may be stack allocated and this will result a
heap-to-stack pointer, which is bad.
Fix this by realizing that if a CONVIFACE escapes and we're
converting from a non-direct interface type, the data needs to
escape to heap.
Running "toolstash -cmp" on std & cmd, the generated machine code
are identical for all packages. However, the export data (escape
tags) differ in the following packages. It looks to me that all
are similar to the "f" above, where the parameter should escape
to heap.
io/ioutil/ioutil.go:118
old: leaking param: r to result ~r1 level=0
new: leaking param: r
image/image.go:943
old: leaking param: p to result ~r0 level=1
new: leaking param content: p
net/url/url.go:200
old: leaking param: s to result ~r2 level=0
new: leaking param: s
(as a consequence)
net/url/url.go:183
old: leaking param: s to result ~r1 level=0
new: leaking param: s
net/url/url.go:194
old: leaking param: s to result ~r1 level=0
new: leaking param: s
net/url/url.go:699
old: leaking param: u to result ~r0 level=1
new: leaking param: u
net/url/url.go:775
old: (*URL).String u does not escape
new: leaking param content: u
net/url/url.go:1038
old: leaking param: u to result ~r0 level=1
new: leaking param: u
net/url/url.go:1099
old: (*URL).MarshalBinary u does not escape
new: leaking param content: u
flag/flag.go:235
old: leaking param: s to result ~r0 level=1
new: leaking param content: s
go/scanner/errors.go:105
old: leaking param: p to result ~r0 level=0
new: leaking param: p
database/sql/sql.go:204
old: leaking param: ns to result ~r0 level=0
new: leaking param: ns
go/constant/value.go:303
old: leaking param: re to result ~r2 level=0, leaking param: im to result ~r2 level=0
new: leaking param: re, leaking param: im
go/constant/value.go:846
old: leaking param: x to result ~r1 level=0
new: leaking param: x
encoding/xml/xml.go:518
old: leaking param: d to result ~r1 level=2
new: leaking param content: d
encoding/xml/xml.go:122
old: leaking param: leaking param: t to result ~r1 level=0
new: leaking param: t
crypto/x509/verify.go:506
old: leaking param: c to result ~r8 level=0
new: leaking param: c
crypto/x509/verify.go:563
old: leaking param: c to result ~r3 level=0, leaking param content: c
new: leaking param: c
crypto/x509/verify.go:615
old: (nothing)
new: leaking closure reference c
crypto/x509/verify.go:996
old: leaking param: c to result ~r1 level=0, leaking param content: c
new: leaking param: c
net/http/filetransport.go:30
old: leaking param: fs to result ~r1 level=0
new: leaking param: fs
net/http/h2_bundle.go:2684
old: leaking param: mh to result ~r0 level=2
new: leaking param content: mh
net/http/h2_bundle.go:7352
old: http2checkConnHeaders req does not escape
new: leaking param content: req
net/http/pprof/pprof.go:221
old: leaking param: name to result ~r1 level=0
new: leaking param: name
cmd/internal/bio/must.go:21
old: leaking param: w to result ~r1 level=0
new: leaking param: w
Fixes #29353.
Change-Id: I7e7798ae773728028b0dcae5bccb3ada51189c68
Reviewed-on: https://go-review.googlesource.com/c/162829
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: David Chase <drchase@google.com>
2019-02-17 21:12:55 -07:00
|
|
|
|
|
|
|
// interface(in) -> out
|
|
|
|
// See also issue 29353.
|
|
|
|
|
|
|
|
// Convert to a non-direct interface, require an allocation and
|
|
|
|
// copy x to heap (not to result).
|
|
|
|
func param14a(x [4]*int) interface{} { // ERROR "leaking param: x$"
|
|
|
|
return x // ERROR "x escapes to heap"
|
|
|
|
}
|
|
|
|
|
|
|
|
// Convert to a direct interface, does not need an allocation.
|
|
|
|
// So x only leaks to result.
|
|
|
|
func param14b(x *int) interface{} { // ERROR "leaking param: x to result ~r1 level=0"
|
cmd/compile: silence esc diagnostics about directiface OCONVIFACEs
In general, a conversion to interface type may require values to be
boxed, which in turn necessitates escape analysis to determine whether
the boxed representation can be stack allocated.
However, esc.go used to unconditionally print escape analysis
decisions about OCONVIFACE, even for conversions that don't require
boxing (e.g., pointers, channels, maps, functions).
For test compatibility with esc.go, escape.go similarly printed these
useless diagnostics. This CL removes the diagnostics, and updates test
expectations accordingly.
Change-Id: I97c57a4a08e44d265bba516c78426ff4f2bf1e12
Reviewed-on: https://go-review.googlesource.com/c/go/+/192697
Run-TryBot: Matthew Dempsky <mdempsky@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Cherry Zhang <cherryyz@google.com>
2019-08-30 11:56:30 -06:00
|
|
|
return x
|
cmd/compile: flow interface data to heap if CONVIFACE of a non-direct interface escapes
Consider the following code:
func f(x []*T) interface{} {
return x
}
It returns an interface that holds a heap copy of x (by calling
convT2I or friend), therefore x escape to heap. The current
escape analysis only recognizes that x flows to the result. This
is not sufficient, since if the result does not escape, x's
content may be stack allocated and this will result a
heap-to-stack pointer, which is bad.
Fix this by realizing that if a CONVIFACE escapes and we're
converting from a non-direct interface type, the data needs to
escape to heap.
Running "toolstash -cmp" on std & cmd, the generated machine code
are identical for all packages. However, the export data (escape
tags) differ in the following packages. It looks to me that all
are similar to the "f" above, where the parameter should escape
to heap.
io/ioutil/ioutil.go:118
old: leaking param: r to result ~r1 level=0
new: leaking param: r
image/image.go:943
old: leaking param: p to result ~r0 level=1
new: leaking param content: p
net/url/url.go:200
old: leaking param: s to result ~r2 level=0
new: leaking param: s
(as a consequence)
net/url/url.go:183
old: leaking param: s to result ~r1 level=0
new: leaking param: s
net/url/url.go:194
old: leaking param: s to result ~r1 level=0
new: leaking param: s
net/url/url.go:699
old: leaking param: u to result ~r0 level=1
new: leaking param: u
net/url/url.go:775
old: (*URL).String u does not escape
new: leaking param content: u
net/url/url.go:1038
old: leaking param: u to result ~r0 level=1
new: leaking param: u
net/url/url.go:1099
old: (*URL).MarshalBinary u does not escape
new: leaking param content: u
flag/flag.go:235
old: leaking param: s to result ~r0 level=1
new: leaking param content: s
go/scanner/errors.go:105
old: leaking param: p to result ~r0 level=0
new: leaking param: p
database/sql/sql.go:204
old: leaking param: ns to result ~r0 level=0
new: leaking param: ns
go/constant/value.go:303
old: leaking param: re to result ~r2 level=0, leaking param: im to result ~r2 level=0
new: leaking param: re, leaking param: im
go/constant/value.go:846
old: leaking param: x to result ~r1 level=0
new: leaking param: x
encoding/xml/xml.go:518
old: leaking param: d to result ~r1 level=2
new: leaking param content: d
encoding/xml/xml.go:122
old: leaking param: leaking param: t to result ~r1 level=0
new: leaking param: t
crypto/x509/verify.go:506
old: leaking param: c to result ~r8 level=0
new: leaking param: c
crypto/x509/verify.go:563
old: leaking param: c to result ~r3 level=0, leaking param content: c
new: leaking param: c
crypto/x509/verify.go:615
old: (nothing)
new: leaking closure reference c
crypto/x509/verify.go:996
old: leaking param: c to result ~r1 level=0, leaking param content: c
new: leaking param: c
net/http/filetransport.go:30
old: leaking param: fs to result ~r1 level=0
new: leaking param: fs
net/http/h2_bundle.go:2684
old: leaking param: mh to result ~r0 level=2
new: leaking param content: mh
net/http/h2_bundle.go:7352
old: http2checkConnHeaders req does not escape
new: leaking param content: req
net/http/pprof/pprof.go:221
old: leaking param: name to result ~r1 level=0
new: leaking param: name
cmd/internal/bio/must.go:21
old: leaking param: w to result ~r1 level=0
new: leaking param: w
Fixes #29353.
Change-Id: I7e7798ae773728028b0dcae5bccb3ada51189c68
Reviewed-on: https://go-review.googlesource.com/c/162829
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: Matthew Dempsky <mdempsky@google.com>
Reviewed-by: David Chase <drchase@google.com>
2019-02-17 21:12:55 -07:00
|
|
|
}
|