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
|
|
|
// errorcheck -0 -m -l
|
|
|
|
|
|
|
|
// Copyright 2015 The Go Authors. All rights reserved.
|
|
|
|
// Use of this source code is governed by a BSD-style
|
|
|
|
// license that can be found in the LICENSE file.
|
|
|
|
|
2015-05-20 13:16:34 -06:00
|
|
|
// Test escape analysis for arrays and some large things
|
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
|
|
|
|
|
|
|
package foo
|
|
|
|
|
|
|
|
var Ssink *string
|
|
|
|
|
|
|
|
type U [2]*string
|
|
|
|
|
|
|
|
func bar(a, b *string) U { // ERROR "leaking param: a to result ~r2 level=0$" "leaking param: b to result ~r2 level=0$"
|
|
|
|
return U{a, b}
|
|
|
|
}
|
|
|
|
|
|
|
|
func foo(x U) U { // ERROR "leaking param: x to result ~r1 level=0$"
|
|
|
|
return U{x[1], x[0]}
|
|
|
|
}
|
|
|
|
|
|
|
|
func bff(a, b *string) U { // ERROR "leaking param: a to result ~r2 level=0$" "leaking param: b to result ~r2 level=0$"
|
|
|
|
return foo(foo(bar(a, b)))
|
|
|
|
}
|
|
|
|
|
|
|
|
func tbff1() *string {
|
|
|
|
a := "cat"
|
|
|
|
b := "dog" // ERROR "moved to heap: b$"
|
|
|
|
u := bff(&a, &b) // ERROR "tbff1 &a does not escape$" "tbff1 &b does not escape$"
|
|
|
|
_ = u[0]
|
|
|
|
return &b // ERROR "&b escapes to heap$"
|
|
|
|
}
|
|
|
|
|
|
|
|
// BAD: need fine-grained analysis to track u[0] and u[1] differently.
|
|
|
|
func tbff2() *string {
|
|
|
|
a := "cat" // ERROR "moved to heap: a$"
|
|
|
|
b := "dog" // ERROR "moved to heap: b$"
|
|
|
|
u := bff(&a, &b) // ERROR "&a escapes to heap$" "&b escapes to heap$"
|
|
|
|
_ = u[0]
|
|
|
|
return u[1]
|
|
|
|
}
|
|
|
|
|
|
|
|
func car(x U) *string { // ERROR "leaking param: x to result ~r1 level=0$"
|
|
|
|
return x[0]
|
|
|
|
}
|
|
|
|
|
|
|
|
// BAD: need fine-grained analysis to track x[0] and x[1] differently.
|
|
|
|
func fun(x U, y *string) *string { // ERROR "leaking param: x to result ~r2 level=0$" "leaking param: y to result ~r2 level=0$"
|
|
|
|
x[0] = y
|
|
|
|
return x[1]
|
|
|
|
}
|
|
|
|
|
|
|
|
func fup(x *U, y *string) *string { // ERROR "leaking param: x to result ~r2 level=1$" "leaking param: y$"
|
|
|
|
x[0] = y // leaking y to heap is intended
|
|
|
|
return x[1]
|
|
|
|
}
|
|
|
|
|
|
|
|
func fum(x *U, y **string) *string { // ERROR "leaking param: x to result ~r2 level=1$" "leaking param content: y$"
|
|
|
|
x[0] = *y
|
|
|
|
return x[1]
|
|
|
|
}
|
|
|
|
|
|
|
|
func fuo(x *U, y *U) *string { // ERROR "leaking param: x to result ~r2 level=1$" "leaking param content: y$"
|
|
|
|
x[0] = y[0]
|
|
|
|
return x[1]
|
|
|
|
}
|
2015-05-20 13:16:34 -06:00
|
|
|
|
|
|
|
// These two tests verify that:
|
|
|
|
// small array literals are stack allocated;
|
|
|
|
// pointers stored in small array literals do not escape;
|
|
|
|
// large array literals are heap allocated;
|
|
|
|
// pointers stored in large array literals escape.
|
|
|
|
func hugeLeaks1(x **string, y **string) { // ERROR "leaking param content: x" "hugeLeaks1 y does not escape" "mark escaped content: x"
|
|
|
|
a := [10]*string{*y}
|
|
|
|
_ = a
|
|
|
|
// 4 x 4,000,000 exceeds MaxStackVarSize, therefore it must be heap allocated if pointers are 4 bytes or larger.
|
|
|
|
b := [4000000]*string{*x} // ERROR "moved to heap: b"
|
|
|
|
_ = b
|
|
|
|
}
|
|
|
|
|
|
|
|
func hugeLeaks2(x *string, y *string) { // ERROR "leaking param: x" "hugeLeaks2 y does not escape"
|
|
|
|
a := [10]*string{y}
|
|
|
|
_ = a
|
|
|
|
// 4 x 4,000,000 exceeds MaxStackVarSize, therefore it must be heap allocated if pointers are 4 bytes or larger.
|
|
|
|
b := [4000000]*string{x} // ERROR "moved to heap: b"
|
|
|
|
_ = b
|
|
|
|
}
|
|
|
|
|
|
|
|
// BAD: x need not leak.
|
|
|
|
func doesNew1(x *string, y *string) { // ERROR "leaking param: x" "leaking param: y"
|
|
|
|
a := new([10]*string) // ERROR "new\(\[10\]\*string\) does not escape"
|
|
|
|
a[0] = x
|
|
|
|
b := new([65537]*string) // ERROR "new\(\[65537\]\*string\) escapes to heap"
|
|
|
|
b[0] = y
|
|
|
|
}
|
|
|
|
|
|
|
|
type a10 struct {
|
|
|
|
s *string
|
|
|
|
i [10]int32
|
|
|
|
}
|
|
|
|
|
|
|
|
type a65537 struct {
|
|
|
|
s *string
|
|
|
|
i [65537]int32
|
|
|
|
}
|
|
|
|
|
|
|
|
// BAD: x need not leak.
|
|
|
|
func doesNew2(x *string, y *string) { // ERROR "leaking param: x" "leaking param: y"
|
|
|
|
a := new(a10) // ERROR "new\(a10\) does not escape"
|
|
|
|
a.s = x
|
|
|
|
b := new(a65537) // ERROR "new\(a65537\) escapes to heap"
|
|
|
|
b.s = y
|
|
|
|
}
|
|
|
|
|
|
|
|
// BAD: x need not leak.
|
|
|
|
func doesMakeSlice(x *string, y *string) { // ERROR "leaking param: x" "leaking param: y"
|
|
|
|
a := make([]*string, 10) // ERROR "make\(\[\]\*string, 10\) does not escape"
|
|
|
|
a[0] = x
|
|
|
|
b := make([]*string, 65537) // ERROR "make\(\[\]\*string, 65537\) escapes to heap"
|
|
|
|
b[0] = y
|
|
|
|
}
|