qbit/go
1
0
Fork 0
mirror of https://github.com/golang/go synced 2024-11-07 15:46:23 -07:00
Code Releases Activity
b294fe9b0a
go/test
History
Keith Randall 5fc70b6fac cmd/compile: set stricter inlining threshold in large functions 
If we're compiling a large function, be more picky about how big
the function we're inlining is.  If the function is >5000 nodes,
we lower the inlining threshold from a cost of 80 to 20.

Turns out reflect.Value's cost is exactly 80.  That's the function
at issue in #26546.

20 was chosen as a proxy for "inlined body is smaller than the call would be".
Simple functions still get inlined, like this one at cost 7:

func ifaceIndir(t *rtype) bool {
	return t.kind&kindDirectIface == 0
}

5000 nodes was chosen as the big function size.  Here are all the
5000+ node (~~1000+ lines) functions in the stdlib:

5187 cmd/internal/obj/arm (*ctxt5).asmout
6879 cmd/internal/obj/s390x (*ctxtz).asmout
6567 cmd/internal/obj/ppc64 (*ctxt9).asmout
9643 cmd/internal/obj/arm64 (*ctxt7).asmout
5042 cmd/internal/obj/x86 (*AsmBuf).doasm
8768 cmd/compile/internal/ssa rewriteBlockAMD64
8878 cmd/compile/internal/ssa rewriteBlockARM
8344 cmd/compile/internal/ssa rewriteValueARM64_OpARM64OR_20
7916 cmd/compile/internal/ssa rewriteValueARM64_OpARM64OR_30
5427 cmd/compile/internal/ssa rewriteBlockARM64
5126 cmd/compile/internal/ssa rewriteValuePPC64_OpPPC64OR_50
6152 cmd/compile/internal/ssa rewriteValuePPC64_OpPPC64OR_60
6412 cmd/compile/internal/ssa rewriteValuePPC64_OpPPC64OR_70
6486 cmd/compile/internal/ssa rewriteValuePPC64_OpPPC64OR_80
6534 cmd/compile/internal/ssa rewriteValuePPC64_OpPPC64OR_90
6534 cmd/compile/internal/ssa rewriteValuePPC64_OpPPC64OR_100
6534 cmd/compile/internal/ssa rewriteValuePPC64_OpPPC64OR_110
6675 cmd/compile/internal/gc typecheck1
5433 cmd/compile/internal/gc walkexpr
14070 cmd/vendor/golang.org/x/arch/arm64/arm64asm decodeArg

There are a lot more smaller (~1000 node) functions in the stdlib.
The function in #26546 has 12477 nodes.

At some point it might be nice to have a better heuristic for "inlined
body is smaller than the call", a non-cliff way to scale down the cost
as the function gets bigger, doing cheaper inlined calls first, etc.
All that can wait for another release. I'd like to do this CL for
1.11.

Fixes #26546
Update #17566

Change-Id: Idda13020e46ec2b28d79a17217f44b189f8139ac
Reviewed-on: https://go-review.googlesource.com/125516
Run-TryBot: Keith Randall <khr@golang.org>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
2018-07-24 16:11:08 +00:00
..
alias3.dir
bench
chan
closure3.dir cmd/compile: fix panic-okay-to-inline change; adjust tests 2018-06-06 20:35:23 +00:00
codegen test/codegen: add more test cases for arm64 2018-07-10 15:41:15 +00:00
ddd2.dir
dwarf
fixedbugs test: add test for gccgo bug #26495 2018-07-20 20:08:15 +00:00
import2.dir
import4.dir
interface
intrinsic.dir
ken test: remove rundircmpout and cmpout actions 2018-05-31 17:36:45 +00:00
linkname.dir
method4.dir
retjmp.dir
safe
stress
syntax
uintptrescapes.dir
64bit.go
235.go
alg.go
alias1.go
alias2.go
alias3.go
alias.go
align.go
append1.go
append.go
args.go
armimm.go
assign1.go
assign.go
atomicload.go
bigalg.go
bigmap.go
blank1.go
blank.go
bom.go
bombad.go
bounds.go test: gofmt bounds.go 2018-05-29 02:39:16 +00:00
chancap.go
chanlinear.go
char_lit1.go
char_lit.go
checkbce.go
clearfat.go
closedchan.go
closure1.go
closure2.go
closure3.go
closure.go
cmp6.go
cmp.go
cmplx.go
cmplxdivide1.go
cmplxdivide.c
cmplxdivide.go
complit1.go
complit.go
compos.go
const1.go
const2.go
const3.go
const4.go
const5.go
const6.go
const.go
convert1.go
convert2.go
convert3.go
convert.go
convlit1.go
convlit.go
convT2X.go
copy1.go
copy.go
crlf.go
ddd1.go
ddd2.go
ddd.go
decl.go
declbad.go
defer.go
deferfin.go
deferprint.go test: remove rundircmpout and cmpout actions 2018-05-31 17:36:45 +00:00
deferprint.out
devirt.go
divide.go
divmod.go
empty.go
env.go
eof1.go
eof.go
errchk
escape2.go
escape2n.go
escape3.go
escape4.go cmd/compile: fix panic-okay-to-inline change; adjust tests 2018-06-06 20:35:23 +00:00
escape5.go
escape_array.go
escape_because.go
escape_calls.go
escape_closure.go
escape_field.go
escape_iface.go
escape_indir.go
escape_level.go
escape_map.go
escape_param.go
escape_slice.go
escape_struct_param1.go
escape_struct_param2.go
escape_struct_return.go
escape.go
fibo.go
finprofiled.go
float_lit2.go
float_lit3.go
float_lit.go
floatcmp.go
for.go
func1.go
func2.go
func3.go
func4.go
func5.go
func6.go
func7.go
func8.go
func.go
funcdup2.go
funcdup.go
gc1.go
gc2.go
gc.go
gcgort.go
gcstring.go
goprint.go test: remove rundircmpout and cmpout actions 2018-05-31 17:36:45 +00:00
goprint.out
goto.go
heapsampling.go
helloworld.go test: remove rundircmpout and cmpout actions 2018-05-31 17:36:45 +00:00
helloworld.out
if.go
import1.go
import2.go
import4.go
import5.go
import6.go
import.go
index0.go
index1.go
index2.go
index.go
indirect1.go
indirect.go
init1.go
init.go
initcomma.go
initialize.go
initializerr.go
initloop.go
inline_big.go cmd/compile: set stricter inlining threshold in large functions 2018-07-24 16:11:08 +00:00
inline_caller.go
inline_callers.go
inline_literal.go
inline_variadic.go
inline.go
int_lit.go
intcvt.go
intrinsic_atomic.go
intrinsic.go
iota.go
label1.go
label.go
linkmain_run.go
linkmain.go
linkname.go
linkobj.go
linkx_run.go
linkx.go
literal.go
live1.go
live2.go
live_syscall.go
live.go cmd/compile: enable stack maps everywhere except unsafe points 2018-05-22 14:43:37 +00:00
locklinear.go
loopbce.go cmd/compile: ensure that loop condition is detected correctly 2018-07-09 18:23:39 +00:00
makechan.go
makemap.go
makenew.go
mallocfin.go
map1.go
map.go
mapclear.go
maplinear.go
mergemul.go
method1.go
method2.go
method3.go
method4.go
method5.go
method6.go
method7.go
method.go
named1.go
named.go
nil.go
nilcheck.go
nilptr2.go
nilptr3_wasm.go
nilptr3.go
nilptr4.go
nilptr.go
nosplit.go
notinheap2.go
notinheap3.go
notinheap.go
nowritebarrier.go
nul1.go
opt_branchlikely.go
parentype.go
peano.go
phiopt.go
print.go test: remove rundircmpout and cmpout actions 2018-05-31 17:36:45 +00:00
print.out
printbig.go test: remove rundircmpout and cmpout actions 2018-05-31 17:36:45 +00:00
printbig.out
prove.go cmd/compile: ensure that loop condition is detected correctly 2018-07-09 18:23:39 +00:00
range.go
README.md
recover1.go
recover2.go
recover3.go
recover4.go
recover5.go
recover.go
reflectmethod1.go
reflectmethod2.go
reflectmethod3.go
reflectmethod4.go
rename1.go
rename.go
reorder2.go
reorder.go
retjmp.go
return.go
rotate0.go
rotate1.go
rotate2.go
rotate3.go
rotate.go
run.go test: add comments for all test actions 2018-06-06 15:12:32 +00:00
rune.go
runtime.go
shift1.go
shift2.go
sieve.go
sigchld.go test: remove rundircmpout and cmpout actions 2018-05-31 17:36:45 +00:00
sigchld.out
simassign.go
sinit_run.go
sinit.go
sizeof.go
slice3.go
slice3err.go
slicecap.go
sliceopt.go
solitaire.go
stack.go
strcopy.go test: add test to verify that string copies don't get optimized away 2018-06-12 19:10:34 +00:00
strength.go
string_lit.go
stringrange.go
struct0.go
switch2.go
switch3.go
switch4.go
switch5.go
switch6.go
switch7.go
switch.go
tinyfin.go
torture.go
turing.go
typecheck.go
typecheckloop.go
typeswitch1.go
typeswitch2.go
typeswitch3.go
typeswitch.go
uintptrescapes2.go
uintptrescapes.go
undef.go
utf.go
varerr.go
varinit.go
writebarrier.go
zerodivide.go

README.md

The test directory contains tests of the Go tool chain and runtime. It includes black box tests, regression tests, and error output tests. They are run as part of all.bash.

To run just these tests, execute:

../bin/go run run.go

Standard library tests should be written as regular Go tests in the appropriate package.

The tool chain and runtime also have regular Go tests in their packages. The main reasons to add a new test to this directory are:

  • it is most naturally expressed using the test runner; or
  • it is also applicable to gccgo and other Go tool chains.