1
0
mirror of https://github.com/golang/go synced 2024-11-24 06:50:17 -07:00
go/test
Austin Clements b49b71ae19 runtime: don't rescan globals
Currently the runtime rescans globals during mark 2 and mark
termination. This costs as much as 500µs/MB in STW time, which is
enough to surpass the 10ms STW limit with only 20MB of globals.

It's also basically unnecessary. The compiler already generates write
barriers for global -> heap pointer updates and the regular write
barrier doesn't check whether the slot is a global or in the heap.
Some less common write barriers do cause problems.
heapBitsBulkBarrier, which is used by typedmemmove and related
functions, currently depends on having access to the pointer bitmap
and as a result ignores writes to globals. Likewise, the
reflect-related write barriers reflect_typedmemmovepartial and
callwritebarrier ignore non-heap destinations; though it appears they
can never be called with global pointers anyway.

This commit makes heapBitsBulkBarrier issue write barriers for writes
to global pointers using the data and BSS pointer bitmaps, removes the
inheap checks from the reflection write barriers, and eliminates the
rescans during mark 2 and mark termination. It also adds a test that
writes to globals have write barriers.

Programs with large data+BSS segments (with pointers) aren't common,
but for programs that do have large data+BSS segments, this
significantly reduces pause time:

name \ 95%ile-time/markTerm              old         new  delta
LargeBSS/bss:1GB/gomaxprocs:4  148200µs ± 6%  302µs ±52%  -99.80% (p=0.008 n=5+5)

This very slightly improves the go1 benchmarks:

name                      old time/op    new time/op    delta
BinaryTree17-12              2.62s ± 3%     2.62s ± 4%    ~     (p=0.904 n=20+20)
Fannkuch11-12                2.15s ± 1%     2.13s ± 0%  -1.29%  (p=0.000 n=18+20)
FmtFprintfEmpty-12          48.3ns ± 2%    47.6ns ± 1%  -1.52%  (p=0.000 n=20+16)
FmtFprintfString-12          152ns ± 0%     152ns ± 1%    ~     (p=0.725 n=18+18)
FmtFprintfInt-12             150ns ± 1%     149ns ± 1%  -1.14%  (p=0.000 n=19+20)
FmtFprintfIntInt-12          250ns ± 0%     244ns ± 1%  -2.12%  (p=0.000 n=20+18)
FmtFprintfPrefixedInt-12     219ns ± 1%     217ns ± 1%  -1.20%  (p=0.000 n=19+20)
FmtFprintfFloat-12           280ns ± 0%     281ns ± 1%  +0.47%  (p=0.000 n=19+19)
FmtManyArgs-12               928ns ± 0%     923ns ± 1%  -0.53%  (p=0.000 n=19+18)
GobDecode-12                7.21ms ± 1%    7.24ms ± 2%    ~     (p=0.091 n=19+19)
GobEncode-12                6.07ms ± 1%    6.05ms ± 1%  -0.36%  (p=0.002 n=20+17)
Gzip-12                      265ms ± 1%     265ms ± 1%    ~     (p=0.496 n=20+19)
Gunzip-12                   39.6ms ± 1%    39.3ms ± 1%  -0.85%  (p=0.000 n=19+19)
HTTPClientServer-12         74.0µs ± 2%    73.8µs ± 1%    ~     (p=0.569 n=20+19)
JSONEncode-12               15.4ms ± 1%    15.3ms ± 1%  -0.25%  (p=0.049 n=17+17)
JSONDecode-12               53.7ms ± 2%    53.0ms ± 1%  -1.29%  (p=0.000 n=18+17)
Mandelbrot200-12            3.97ms ± 1%    3.97ms ± 0%    ~     (p=0.072 n=17+18)
GoParse-12                  3.35ms ± 2%    3.36ms ± 1%  +0.51%  (p=0.005 n=18+20)
RegexpMatchEasy0_32-12      72.7ns ± 2%    72.2ns ± 1%  -0.70%  (p=0.005 n=19+19)
RegexpMatchEasy0_1K-12       246ns ± 1%     245ns ± 0%  -0.60%  (p=0.000 n=18+16)
RegexpMatchEasy1_32-12      72.8ns ± 1%    72.5ns ± 1%  -0.37%  (p=0.011 n=18+18)
RegexpMatchEasy1_1K-12       380ns ± 1%     385ns ± 1%  +1.34%  (p=0.000 n=20+19)
RegexpMatchMedium_32-12      115ns ± 2%     115ns ± 1%  +0.44%  (p=0.047 n=20+20)
RegexpMatchMedium_1K-12     35.4µs ± 1%    35.5µs ± 1%    ~     (p=0.079 n=18+19)
RegexpMatchHard_32-12       1.83µs ± 0%    1.80µs ± 1%  -1.76%  (p=0.000 n=18+18)
RegexpMatchHard_1K-12       55.1µs ± 0%    54.3µs ± 1%  -1.42%  (p=0.000 n=18+19)
Revcomp-12                   386ms ± 1%     381ms ± 1%  -1.14%  (p=0.000 n=18+18)
Template-12                 61.5ms ± 2%    61.5ms ± 2%    ~     (p=0.647 n=19+20)
TimeParse-12                 338ns ± 0%     336ns ± 1%  -0.72%  (p=0.000 n=14+19)
TimeFormat-12                350ns ± 0%     357ns ± 0%  +2.05%  (p=0.000 n=19+18)
[Geo mean]                  55.3µs         55.0µs       -0.41%

Change-Id: I57e8720385a1b991aeebd111b6874354308e2a6b
Reviewed-on: https://go-review.googlesource.com/20829
Run-TryBot: Austin Clements <austin@google.com>
Reviewed-by: Rick Hudson <rlh@golang.org>
2016-04-27 18:48:16 +00:00
..
bench cmd/compile: enable const division for arm64 2016-04-27 17:47:49 +00:00
bugs test: migrate remaining tests to run.go 2014-12-22 22:41:34 +00:00
chan Fix several out of date references to 4g/5g/6g/8g/9g. 2015-06-26 03:38:21 +00:00
ddd2.dir
dwarf
fixedbugs cmd/compile: treat empty and absent struct field tags as identical 2016-04-25 21:28:45 +00:00
import2.dir
import4.dir test: adjust gccgo expected import errors 2015-10-07 14:37:44 +00:00
interface cmd/internal/gc: improve "type *X has no field or method M" message 2015-05-07 16:21:57 +00:00
intrinsic.dir cmd/compile: added some intrinsics to SSA back end 2016-03-28 16:29:59 +00:00
ken all: fix spelling mistakes 2016-04-03 17:03:15 +00:00
method4.dir cmd/gc: do not omit wrapper for expression (interface{...}).F 2014-01-07 13:26:48 +01:00
safe
stress build: adjustments for move from src/pkg to src 2014-09-08 00:06:45 -04:00
syntax cmd/compile: use bufio.Reader directly in lexer 2016-03-11 21:27:23 +00:00
64bit.go
235.go
alg.go test: add test that required algs get generated 2016-04-24 21:36:52 +00:00
alias1.go
alias.go
append.go
args.go
assign1.go
assign.go cmd/gc: prohibit short variable declarations containing duplicate symbols 2014-10-06 17:16:39 -04:00
atomicload.go cmd/compile: keep value use counts in SSA 2016-03-17 04:20:02 +00:00
bigalg.go
bigmap.go
blank1.go cmd/compile: don't allow blank method declarations on builtins 2015-12-02 18:26:38 +00:00
blank.go
bom.go
bombad.go
bounds.go
chancap.go
chanlinear.go runtime: test for linear enqueue/dequeue behavior 2014-12-08 22:18:17 +00:00
char_lit1.go
char_lit.go
checkbce.go cmd/compile: bce when max and limit are consts 2016-04-11 16:01:22 +00:00
clearfat.go [dev.power64] 9g: fix under-zeroing in clearfat 2014-10-31 11:08:27 -04:00
closedchan.go
closure1.go cmd/gc: capture variables by value 2015-01-29 13:07:30 +00:00
closure2.go cmd/gc: fix capturing by value for range statements 2015-02-03 15:48:48 +00:00
closure.go
cmp6.go cmd/gc: do not consider length zero arrays as comparable. 2014-01-31 00:30:56 +01:00
cmp.go Fix several out of date references to 4g/5g/6g/8g/9g. 2015-06-26 03:38:21 +00:00
cmplx.go
cmplxdivide1.go
cmplxdivide.c test: comment the behavior and use of cmplxdivide* 2015-01-15 00:00:06 +00:00
cmplxdivide.go test: comment the behavior and use of cmplxdivide* 2015-01-15 00:00:06 +00:00
complit1.go cmd/internal/gc: accept map literals with omitted key type 2015-06-04 02:31:38 +00:00
complit.go
compos.go
const1.go
const2.go
const3.go
const4.go all: link to https instead of http 2015-07-11 14:36:33 +00:00
const5.go all: link to https instead of http 2015-07-11 14:36:33 +00:00
const6.go
const.go cmd/compile: allow huge rsh in constants arithmetic 2015-08-21 20:27:22 +00:00
convert1.go
convert3.go
convert.go
convlit1.go
convlit.go cmd/internal/gc: unsafe.Pointer constants may only be converted to uintptr 2015-05-07 23:54:28 +00:00
convT2X.go
copy.go
crlf.go
ddd1.go
ddd2.go
ddd.go
decl.go
declbad.go
defer.go
deferfin.go runtime: combine small NoScan allocations 2014-01-24 22:35:11 +04:00
deferprint.go
deferprint.out
divide.go
divmod.go
empty.go
env.go
eof1.go
eof.go
errchk Fix several out of date references to 4g/5g/6g/8g/9g. 2015-06-26 03:38:21 +00:00
escape2.go cmd/compile: better modeling of escape across loop levels 2016-01-13 04:01:00 +00:00
escape2n.go cmd/compile: better modeling of escape across loop levels 2016-01-13 04:01:00 +00:00
escape3.go
escape4.go cmd/internal/gc: transitive inlining 2015-02-26 17:36:00 +00:00
escape5.go cmd/internal/gc: move check for large-hence-heap-allocated types into escape analysis 2015-05-22 02:13:54 +00:00
escape_array.go cmd/internal/gc: move check for large-hence-heap-allocated types into escape analysis 2015-05-22 02:13:54 +00:00
escape_because.go cmd/compile: escape analysis explanations added to -m -m output 2016-03-17 13:29:48 +00:00
escape_calls.go cmd/internal/gc: make indirect calls properly escape-y 2015-05-22 01:36:11 +00:00
escape_closure.go cmd/compile: note escape of parts of closured-capture vars 2016-04-05 18:10:09 +00:00
escape_field.go cmd/internal/gc: improve flow of input params to output params 2015-05-01 13:47:20 +00:00
escape_iface.go cmd/compile: avoid leak of dottype expression if type does not contain pointers. 2016-02-19 16:10:14 +00:00
escape_indir.go cmd/internal/gc: improve flow of input params to output params 2015-05-01 13:47:20 +00:00
escape_level.go cmd/internal/gc: improve flow of input params to output params 2015-05-01 13:47:20 +00:00
escape_map.go test: add tests for escape analysis of interface conversions 2015-03-28 16:15:27 +00:00
escape_param.go cmd/internal/gc: improve flow of input params to output params 2015-05-01 13:47:20 +00:00
escape_slice.go cmd/internal/gc: extend escape analysis to pointers in slices 2015-05-18 15:34:39 +00:00
escape_struct_param1.go cmd/internal/gc: improve flow of input params to output params 2015-05-01 13:47:20 +00:00
escape_struct_param2.go cmd/internal/gc: improve flow of input params to output params 2015-05-01 13:47:20 +00:00
escape_struct_return.go cmd/internal/gc: improve flow of input params to output params 2015-05-01 13:47:20 +00:00
escape.go
fibo.go Fix whitespace (use blanks consistently). 2014-12-08 22:22:58 +00:00
finprofiled.go runtime: fix finalization and profiling of tiny allocations 2015-11-03 18:57:18 +00:00
float_lit2.go test: fix two typos in float_lit2.go 2014-05-21 17:19:12 -04:00
float_lit3.go test/float_lit2.go: rewrite to test values near boundaries 2014-05-21 17:12:06 -04:00
float_lit.go
floatcmp.go
for.go
func1.go
func2.go
func3.go
func4.go
func5.go
func6.go Fix several out of date references to 4g/5g/6g/8g/9g. 2015-06-26 03:38:21 +00:00
func7.go Fix several out of date references to 4g/5g/6g/8g/9g. 2015-06-26 03:38:21 +00:00
func8.go test: use go:noinline consistently 2015-11-03 02:01:34 +00:00
func.go
funcdup2.go
funcdup.go
gc1.go
gc2.go test: check for build constraints only upto the first blank line 2015-02-06 05:36:26 +00:00
gc.go
gcstring.go runtime: fix empty string handling in garbage collector 2014-03-11 23:58:39 -04:00
goprint.go test: make goprint.go wait for goroutine termination 2016-04-06 18:45:00 +00:00
goprint.out
goto.go [dev.ssa] test: gofmt {goto,label,label1}.go 2015-07-20 20:09:34 +00:00
heapsampling.go test: really force heap profile update in heapsampling.go 2015-11-12 18:32:13 +00:00
helloworld.go
helloworld.out
if.go
import1.go
import2.go
import4.go
import5.go
import.go
index0.go
index1.go
index2.go
index.go Fix several out of date references to 4g/5g/6g/8g/9g. 2015-06-26 03:38:21 +00:00
indirect1.go
indirect.go
init1.go test: use correct value in error message in init1.go 2016-04-14 18:35:09 +00:00
init.go
initcomma.go
initialize.go
initializerr.go
initloop.go cmd/compile: use []*Node instead of NodeList in sinit 2015-09-05 02:25:01 +00:00
inline.go cmd/compile: allow inlining of functions with switch statements 2016-03-21 23:05:10 +00:00
int_lit.go
intcvt.go
intrinsic.go cmd/compile: added some intrinsics to SSA back end 2016-03-28 16:29:59 +00:00
iota.go
label1.go [dev.ssa] cmd/compile: implement control flow handling 2015-07-23 00:45:26 +00:00
label.go [dev.ssa] cmd/compile: implement control flow handling 2015-07-23 00:45:26 +00:00
linkmain_run.go test: fix linkmain test 2015-12-17 23:36:13 +00:00
linkmain.go cmd/link: reject non-package main toplevel.a file, remove dead code 2015-12-17 20:59:51 +00:00
linkx_run.go test: print output on link -X failure 2016-04-20 17:37:35 +00:00
linkx.go cmd/link: detect -X setting non-string variable 2015-06-29 20:28:36 +00:00
literal.go
live1.go cmd/gc: correct liveness for fat variables 2014-02-15 10:58:55 -05:00
live2.go [dev.ssa] cmd/compile: default compile+test with SSA 2015-10-30 20:35:20 +00:00
live_ssa.go cmd/compile: fix x=x assignments 2016-04-06 15:04:32 +00:00
live_syscall.go cmd/compile: recognize Syscall-like functions for liveness analysis 2016-01-14 01:16:45 +00:00
live.go [dev.ssa] cmd/compile: default compile+test with SSA 2015-10-30 20:35:20 +00:00
loopbce.go cmd/compile: bce when max and limit are consts 2016-04-11 16:01:22 +00:00
mallocfin.go
map1.go
map.go test/mapnan.go: add regression test for non-empty interfaces. 2014-08-06 17:02:55 -04:00
maplinear.go test: make maplinear more robust 2014-10-27 18:59:02 -04:00
method1.go cmd/compile: ignore receiver parameters in Eqtype 2016-03-17 00:38:15 +00:00
method2.go
method3.go
method4.go
method5.go
method.go
named1.go cmd/gc: comma-ok assignments produce untyped bool as 2nd result 2014-08-11 16:11:55 -07:00
named.go
nil.go
nilcheck.go [dev.ssa] cmd/compile: make cse faster 2016-01-28 20:59:20 +00:00
nilptr2.go
nilptr3_ssa.go cmd/compile: don't nilcheck newobject and return values from mapaccess{1,2} 2016-04-22 16:18:42 +00:00
nilptr3.go cmd/compile: don't nilcheck newobject and return values from mapaccess{1,2} 2016-04-22 16:18:42 +00:00
nilptr4.go cmd/gc: fix bad checknil with ints on 32 bit compilers 2014-02-26 12:25:13 -08:00
nilptr.go
nosplit.go test: add s390x case to nosplit test 2016-03-21 08:14:51 +00:00
nul1.go
opt_branchlikely.go [dev.ssa] cmd/compile: adjust branch likeliness for calls/loops 2016-03-01 20:09:41 +00:00
parentype.go
peano.go
phiopt.go cmd/compile: introduce bool operations. 2016-04-25 20:43:04 +00:00
print.go runtime: move built-in print routines to go. 2014-07-31 13:48:48 -07:00
print.out runtime: move built-in print routines to go. 2014-07-31 13:48:48 -07:00
printbig.go
printbig.out
prove.go cmd/compile: transform some Phis into Or8. 2016-04-19 22:04:30 +00:00
range.go cmd/gc: implement 'for range x {' 2014-07-16 19:27:10 -04:00
recover1.go
recover2.go
recover3.go
recover4.go test: move allocation before munmap in recover4 2015-09-16 08:51:48 +00:00
recover.go test: add more cases to recover.go 2014-10-22 08:06:15 -07:00
reflectmethod1.go cmd/compile: track reflect.Type.Method in deadcode 2016-03-11 21:19:20 +00:00
reflectmethod2.go cmd/compile: track reflect.Type.Method in deadcode 2016-03-11 21:19:20 +00:00
reflectmethod3.go cmd/compile: track reflect.Type.Method in deadcode 2016-03-11 21:19:20 +00:00
reflectmethod4.go cmd/link: treat reflect.Value.Method like Call 2016-03-11 22:07:02 +00:00
rename1.go
rename.go
reorder2.go test: use go:noinline consistently 2015-11-03 02:01:34 +00:00
reorder.go
return.go
rotate0.go test: re-enable rotate tests 2015-05-28 19:40:58 +00:00
rotate1.go test: re-enable rotate tests 2015-05-28 19:40:58 +00:00
rotate2.go test: re-enable rotate tests 2015-05-28 19:40:58 +00:00
rotate3.go test: re-enable rotate tests 2015-05-28 19:40:58 +00:00
rotate.go
run.go cmd/compile: fix build 2016-03-31 18:00:36 +00:00
rune.go
runtime.go
shift1.go test: avoid "declared but not used" errors in shift1.go 2014-07-20 12:25:24 -07:00
shift2.go
sieve.go
sigchld.go test: skip SIGCHLD test on Plan 9 2014-01-29 09:28:23 +01:00
sigchld.out
simassign.go
sinit_run.go all: retire architecture letter in file names, public API 2015-05-21 17:32:17 +00:00
sinit.go cmd/compile: allow static init for unsafe.Pointer(&x) where x is global 2015-07-07 21:51:57 +00:00
sizeof.go
slice3.go cmd/gc, runtime: treat slices and strings like pointers in garbage collection 2014-08-25 14:38:19 -04:00
slice3err.go
slicecap.go cmd/gc, runtime: treat slices and strings like pointers in garbage collection 2014-08-25 14:38:19 -04:00
sliceopt.go [dev.ssa] cmd/compile: default compile+test with SSA 2015-10-30 20:35:20 +00:00
solitaire.go
stack.go
strength.go cmd/compile: generalize strength reduction of mulq 2016-03-30 22:27:13 +00:00
string_lit.go
stringrange.go
struct0.go
switch2.go cmd/compile: address several more 1.6 TODOs in parser 2015-11-21 07:21:23 +00:00
switch3.go
switch4.go
switch.go
tinyfin.go runtime: fix finalization and profiling of tiny allocations 2015-11-03 18:57:18 +00:00
torture.go cmd/8g: don't allocate a register early for cap(CHAN). 2014-07-01 09:20:51 +02:00
turing.go
typecheck.go
typecheckloop.go cmd/compile: convert typecheck_stack to []*Node 2015-09-06 23:50:51 +00:00
typeswitch1.go
typeswitch2.go
typeswitch3.go
typeswitch.go
undef.go
utf.go
varerr.go
varinit.go
writebarrier.go runtime: don't rescan globals 2016-04-27 18:48:16 +00:00
zerodivide.go