- 06 Mar, 2017 10 commits
-
-
Josh Bleecher Snyder authored
We generate a lot of pointless dead blocks during the AST to SSA conversion. There are a few commonly occurring kinds of statements that contain neither variables nor code and that switch to a new block themselves. Stop making dead blocks for them. For the code in #19379, this reduces compilation wall time by 36% and max rss by 28%. This also helps a little for regular code, particularly code heavy on switch statements. name old time/op new time/op delta Template 231ms ± 3% 230ms ± 5% ~ (p=0.402 n=17+16) Unicode 101ms ± 4% 103ms ± 5% ~ (p=0.221 n=19+18) GoTypes 635ms ± 5% 625ms ± 4% ~ (p=0.063 n=20+18) Compiler 2.93s ± 2% 2.89s ± 2% -1.22% (p=0.003 n=20+19) SSA 4.53s ± 3% 4.52s ± 3% ~ (p=0.380 n=20+19) Flate 132ms ± 4% 133ms ± 5% ~ (p=0.647 n=20+19) GoParser 161ms ± 3% 161ms ± 4% ~ (p=0.749 n=20+19) Reflect 403ms ± 4% 397ms ± 3% -1.53% (p=0.030 n=20+19) Tar 121ms ± 2% 121ms ± 8% ~ (p=0.544 n=19+19) XML 225ms ± 3% 224ms ± 4% ~ (p=0.396 n=20+19) name old user-ns/op new user-ns/op delta Template 302user-ms ± 1% 297user-ms ± 7% -1.49% (p=0.048 n=15+18) Unicode 142user-ms ± 3% 143user-ms ± 5% ~ (p=0.363 n=19+17) GoTypes 852user-ms ± 5% 851user-ms ± 3% ~ (p=0.851 n=20+18) Compiler 4.11user-s ± 6% 3.98user-s ± 3% -3.08% (p=0.000 n=20+19) SSA 6.91user-s ± 5% 6.82user-s ± 7% ~ (p=0.113 n=20+19) Flate 164user-ms ± 4% 168user-ms ± 4% +2.42% (p=0.001 n=18+19) GoParser 207user-ms ± 4% 206user-ms ± 4% ~ (p=0.176 n=20+18) Reflect 509user-ms ± 4% 505user-ms ± 4% ~ (p=0.113 n=20+19) Tar 153user-ms ± 7% 151user-ms ± 9% ~ (p=0.283 n=20+19) XML 284user-ms ± 4% 282user-ms ± 4% ~ (p=0.270 n=20+19) name old alloc/op new alloc/op delta Template 42.6MB ± 0% 41.9MB ± 0% -1.55% (p=0.000 n=19+19) Unicode 31.7MB ± 0% 31.7MB ± 0% ~ (p=0.828 n=20+18) GoTypes 124MB ± 0% 121MB ± 0% -2.11% (p=0.000 n=20+17) Compiler 534MB ± 0% 523MB ± 0% -2.06% (p=0.000 n=20+19) SSA 989MB ± 0% 977MB ± 0% -1.28% (p=0.000 n=20+19) Flate 27.8MB ± 0% 27.5MB ± 0% -0.98% (p=0.000 n=20+19) GoParser 34.3MB ± 0% 34.0MB ± 0% -0.81% (p=0.000 n=20+19) Reflect 84.6MB ± 0% 82.9MB ± 0% -2.00% (p=0.000 n=17+18) Tar 28.8MB ± 0% 28.3MB ± 0% -1.52% (p=0.000 n=16+18) XML 47.2MB ± 0% 45.8MB ± 0% -2.99% (p=0.000 n=20+19) name old allocs/op new allocs/op delta Template 421k ± 1% 419k ± 1% -0.41% (p=0.001 n=20+19) Unicode 338k ± 1% 338k ± 1% ~ (p=0.478 n=20+19) GoTypes 1.28M ± 0% 1.28M ± 0% -0.36% (p=0.000 n=20+18) Compiler 5.06M ± 0% 5.03M ± 0% -0.63% (p=0.000 n=20+19) SSA 9.14M ± 0% 9.11M ± 0% -0.34% (p=0.000 n=20+19) Flate 267k ± 1% 266k ± 1% ~ (p=0.149 n=20+19) GoParser 347k ± 0% 347k ± 1% ~ (p=0.103 n=19+19) Reflect 1.07M ± 0% 1.07M ± 0% -0.42% (p=0.000 n=16+18) Tar 274k ± 0% 273k ± 1% ~ (p=0.116 n=19+19) XML 449k ± 0% 446k ± 1% -0.60% (p=0.000 n=20+19) Updates #19379 Change-Id: Ie798c347a0c081f5e349e1529880bebaae290967 Reviewed-on: https://go-review.googlesource.com/37760Reviewed-by:
David Chase <drchase@google.com>
-
Aliaksandr Valialkin authored
This makes Sym flags consistent with the rest of the code after the CL 37445. No functional changes. Passes toolstash -cmp. Change-Id: Ica919f2ab98581371c717fff9a70aeb11058ca17 Reviewed-on: https://go-review.googlesource.com/37847Reviewed-by:
Matthew Dempsky <mdempsky@google.com> Run-TryBot: Matthew Dempsky <mdempsky@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org>
-
Daniel Martí authored
Found by github.com/mvdan/unparam. Change-Id: Ic97f05a2ecb5b17caa36aafe403e2266abea3e0e Reviewed-on: https://go-review.googlesource.com/37836 Run-TryBot: Rob Pike <r@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by:
Rob Pike <r@golang.org>
-
Russ Cox authored
On overloaded machines once we get to big N, the machine slowness dominates. But we only retry once we get to a big N. Instead, retry for small N too, and die on the first big N that fails. Change-Id: I3ab9cfb88832ad86e2ba1389a926045091268aeb Reviewed-on: https://go-review.googlesource.com/37543 Run-TryBot: Russ Cox <rsc@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by:
Brad Fitzpatrick <bradfitz@golang.org>
-
Daniel Martí authored
Found via github.com/mvdan/unparam. Change-Id: I12cb0c35b14c880425c347fb3eb146712a86f310 Reviewed-on: https://go-review.googlesource.com/37834Reviewed-by:
Robert Griesemer <gri@golang.org> Run-TryBot: Robert Griesemer <gri@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
-
Daniel Martí authored
Found by github.com/mvdan/unparam. Change-Id: I5b0c7cfdc1ab4fe0d79ef4c5a31612bbcf2ff3ad Reviewed-on: https://go-review.googlesource.com/37833Reviewed-by:
-
Daniel Martí authored
Found by github.com/mvdan/unparam. Change-Id: Ie26e963176eb7afb35d16fed5cbca6530f7731c3 Reviewed-on: https://go-review.googlesource.com/37832Reviewed-by:
-
Eitan Adler authored
Change-Id: I6343c162e27e2e492547c96f1fc504909b1c03c0 Reviewed-on: https://go-review.googlesource.com/37793Reviewed-by:
Daniel Martí <mvdan@mvdan.cc> Reviewed-by:
Ian Lance Taylor <iant@golang.org> Run-TryBot: Ian Lance Taylor <iant@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org>
-
Daniel Martí authored
Found by github.com/mvdan/unparam. Change-Id: I5a6664cceeba1cf1c2f3236ddf4db5ce7a64b02a Reviewed-on: https://go-review.googlesource.com/37835Reviewed-by:
-
Daniel Martí authored
Found by github.com/mvdan/unparam. Small performance win when the utf8.RuneLen call is removed. name old time/op new time/op delta AppendQuoteRune-4 21.7ns ± 0% 21.4ns ± 0% -1.38% (p=0.008 n=5+5) Change-Id: Ieb3b3e1148db7a3d854c81555a491edeff549f43 Reviewed-on: https://go-review.googlesource.com/37831Reviewed-by:
-
- 05 Mar, 2017 2 commits
-
-
Cherry Zhang authored
The assembler back end uses F15 as a temporary register in these instructions. Checked the assembler back end and made sure that this is the only case clobbering F15. Fixes #19403. Change-Id: I02b9e00fdd9229db899f501c8e9b306e02912d83 Reviewed-on: https://go-review.googlesource.com/37792 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by:
Keith Randall <khr@golang.org>
-
Josh Bleecher Snyder authored
Change-Id: I666d2c6ea8d0b54a71260809d1a2573b122865b2 Reviewed-on: https://go-review.googlesource.com/37790 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by:
-
- 04 Mar, 2017 12 commits
-
-
Josh Bleecher Snyder authored
CL 37234 introduced string concatenation into some hot code. This CL does that work earlier and caches the result. Updates #19386 Performance impact vs master: name old time/op new time/op delta Template 223ms ± 5% 216ms ± 5% -2.98% (p=0.001 n=20+20) Unicode 98.7ms ± 4% 99.0ms ± 4% ~ (p=0.749 n=20+19) GoTypes 631ms ± 4% 626ms ± 4% ~ (p=0.253 n=20+20) Compiler 2.91s ± 1% 2.87s ± 3% -1.11% (p=0.005 n=18+20) SSA 4.48s ± 2% 4.36s ± 2% -2.77% (p=0.000 n=20+20) Flate 130ms ± 2% 129ms ± 6% ~ (p=0.428 n=19+20) GoParser 160ms ± 4% 157ms ± 3% -1.62% (p=0.005 n=20+18) Reflect 395ms ± 2% 394ms ± 4% ~ (p=0.445 n=20+20) Tar 120ms ± 5% 118ms ± 6% ~ (p=0.101 n=19+20) XML 224ms ± 3% 223ms ± 3% ~ (p=0.544 n=19+19) name old user-ns/op new user-ns/op delta Template 291user-ms ± 5% 265user-ms ± 5% -9.02% (p=0.000 n=20+19) Unicode 140user-ms ± 3% 139user-ms ± 8% ~ (p=0.904 n=20+20) GoTypes 844user-ms ± 3% 849user-ms ± 3% ~ (p=0.251 n=20+18) Compiler 4.06user-s ± 5% 3.98user-s ± 2% ~ (p=0.056 n=20+20) SSA 6.89user-s ± 5% 6.50user-s ± 3% -5.61% (p=0.000 n=20+20) Flate 164user-ms ± 5% 163user-ms ± 4% ~ (p=0.365 n=20+19) GoParser 206user-ms ± 6% 204user-ms ± 4% ~ (p=0.534 n=20+18) Reflect 501user-ms ± 4% 505user-ms ± 5% ~ (p=0.383 n=20+20) Tar 151user-ms ± 3% 152user-ms ± 7% ~ (p=0.798 n=17+20) XML 283user-ms ± 7% 280user-ms ± 5% ~ (p=0.301 n=20+20) name old alloc/op new alloc/op delta Template 42.5MB ± 0% 40.2MB ± 0% -5.59% (p=0.000 n=20+20) Unicode 31.7MB ± 0% 31.0MB ± 0% -2.19% (p=0.000 n=20+18) GoTypes 124MB ± 0% 117MB ± 0% -5.90% (p=0.000 n=20+20) Compiler 533MB ± 0% 490MB ± 0% -8.07% (p=0.000 n=20+20) SSA 989MB ± 0% 893MB ± 0% -9.74% (p=0.000 n=20+20) Flate 27.8MB ± 0% 26.1MB ± 0% -5.92% (p=0.000 n=20+20) GoParser 34.3MB ± 0% 32.1MB ± 0% -6.43% (p=0.000 n=19+20) Reflect 84.6MB ± 0% 81.4MB ± 0% -3.84% (p=0.000 n=20+20) Tar 28.8MB ± 0% 27.7MB ± 0% -3.89% (p=0.000 n=20+20) XML 47.2MB ± 0% 44.2MB ± 0% -6.45% (p=0.000 n=20+19) name old allocs/op new allocs/op delta Template 420k ± 1% 381k ± 1% -9.35% (p=0.000 n=20+20) Unicode 338k ± 1% 324k ± 1% -4.29% (p=0.000 n=20+19) GoTypes 1.28M ± 0% 1.15M ± 0% -10.30% (p=0.000 n=20+20) Compiler 5.06M ± 0% 4.41M ± 0% -12.92% (p=0.000 n=20+20) SSA 9.14M ± 0% 7.91M ± 0% -13.46% (p=0.000 n=19+20) Flate 267k ± 0% 241k ± 1% -9.53% (p=0.000 n=20+20) GoParser 347k ± 1% 312k ± 0% -10.15% (p=0.000 n=19+20) Reflect 1.07M ± 0% 1.00M ± 0% -6.86% (p=0.000 n=20+20) Tar 274k ± 1% 256k ± 1% -6.73% (p=0.000 n=20+20) XML 448k ± 0% 398k ± 0% -11.17% (p=0.000 n=20+18) Performance impact when applied together with CL 37234 atop CL 37234's parent commit (i.e. as if it were a part of CL 37234), to show that this commit makes CL 37234 completely performance-neutral: name old time/op new time/op delta Template 222ms ±14% 222ms ±14% ~ (p=1.000 n=14+15) Unicode 104ms ±18% 106ms ±18% ~ (p=0.650 n=13+14) GoTypes 653ms ± 7% 638ms ± 5% ~ (p=0.145 n=14+12) Compiler 3.10s ± 1% 3.13s ±10% ~ (p=1.000 n=2+2) SSA 4.73s ±11% 4.68s ±11% ~ (p=0.567 n=15+15) Flate 136ms ± 4% 133ms ± 7% ~ (p=0.231 n=12+14) GoParser 163ms ±11% 169ms ±10% ~ (p=0.352 n=14+14) Reflect 415ms ±15% 423ms ±20% ~ (p=0.715 n=15+14) Tar 133ms ±17% 130ms ±23% ~ (p=0.252 n=14+15) XML 236ms ±16% 235ms ±14% ~ (p=0.874 n=14+14) name old user-ns/op new user-ns/op delta Template 271user-ms ±10% 271user-ms ±10% ~ (p=0.780 n=14+15) Unicode 143user-ms ± 5% 146user-ms ±11% ~ (p=0.432 n=12+14) GoTypes 864user-ms ± 5% 866user-ms ± 9% ~ (p=0.905 n=14+13) Compiler 4.17user-s ± 1% 4.26user-s ± 7% ~ (p=1.000 n=2+2) SSA 6.79user-s ± 8% 6.79user-s ± 6% ~ (p=0.902 n=15+15) Flate 169user-ms ± 8% 164user-ms ± 5% -3.13% (p=0.014 n=14+14) GoParser 212user-ms ± 7% 217user-ms ±22% ~ (p=1.000 n=13+15) Reflect 521user-ms ± 7% 533user-ms ±15% ~ (p=0.511 n=14+14) Tar 165user-ms ±17% 161user-ms ±15% ~ (p=0.345 n=15+15) XML 294user-ms ±11% 292user-ms ±10% ~ (p=0.839 n=14+14) name old alloc/op new alloc/op delta Template 39.9MB ± 0% 39.9MB ± 0% ~ (p=0.621 n=15+14) Unicode 31.0MB ± 0% 31.0MB ± 0% ~ (p=0.098 n=13+15) GoTypes 117MB ± 0% 117MB ± 0% ~ (p=0.775 n=15+15) Compiler 488MB ± 0% 488MB ± 0% ~ (p=0.333 n=2+2) SSA 892MB ± 0% 892MB ± 0% +0.03% (p=0.000 n=15+15) Flate 26.1MB ± 0% 26.1MB ± 0% ~ (p=0.098 n=15+15) GoParser 31.8MB ± 0% 31.8MB ± 0% ~ (p=0.525 n=15+13) Reflect 81.2MB ± 0% 81.2MB ± 0% +0.06% (p=0.001 n=12+14) Tar 27.5MB ± 0% 27.5MB ± 0% ~ (p=0.595 n=15+15) XML 44.1MB ± 0% 44.1MB ± 0% ~ (p=0.486 n=15+15) name old allocs/op new allocs/op delta Template 378k ± 1% 378k ± 0% ~ (p=0.949 n=15+14) Unicode 324k ± 0% 324k ± 1% ~ (p=0.057 n=14+15) GoTypes 1.15M ± 0% 1.15M ± 0% ~ (p=0.461 n=15+15) Compiler 4.39M ± 0% 4.39M ± 0% ~ (p=0.333 n=2+2) SSA 7.90M ± 0% 7.90M ± 0% +0.06% (p=0.008 n=15+15) Flate 240k ± 1% 241k ± 0% ~ (p=0.233 n=15+15) GoParser 309k ± 1% 309k ± 0% ~ (p=0.867 n=15+12) Reflect 1.00M ± 0% 1.00M ± 0% ~ (p=0.139 n=12+15) Tar 254k ± 1% 253k ± 1% ~ (p=0.345 n=15+15) XML 398k ± 0% 397k ± 1% ~ (p=0.267 n=15+15) Change-Id: Ic999a0f456a371c99eebba0f9747263a13836e33 Reviewed-on: https://go-review.googlesource.com/37766 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by:
-
Kevin Burke authored
If you cross compile for a Unix target and call user.Lookup("root") or user.LookupId("0"), we'll try to read the answer out of /etc/passwd instead of returning an "unimplemented" error. The equivalent cgo function calls getpwuid_r in glibc, which may reach out to the NSS database or allow callers to register extensions. The pure Go implementation only reads from /etc/passwd. Change-Id: I56a302d634b15ba5097f9f0d6a758c68e486ba6d Reviewed-on: https://go-review.googlesource.com/37664Reviewed-by: -
Giovanni Bajo authored
Match more patterns generated by the compiler where the index for a bound check is bounded through a AND operation, with different register sizes. These rules trigger a dozen of times in a bootstrap. Change-Id: Ic9fff16f21d08580f19a366c3ee1a372e58357d1 Reviewed-on: https://go-review.googlesource.com/37442 TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by:
-
Austin Clements authored
Currently ReadMemStats stops the world for ~1.7 ms/GB of heap because it collects statistics from every single span. For large heaps, this can be quite costly. This is particularly unfortunate because many production infrastructures call this function regularly to collect and report statistics. Fix this by tracking the necessary cumulative statistics in the mcaches. ReadMemStats still has to stop the world to stabilize these statistics, but there are only O(GOMAXPROCS) mcaches to collect statistics from, so this pause is only 25µs even at GOMAXPROCS=100. Fixes #13613. Change-Id: I3c0a4e14833f4760dab675efc1916e73b4c0032a Reviewed-on: https://go-review.googlesource.com/34937 Run-TryBot: Austin Clements <austin@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by:
Rick Hudson <rlh@golang.org>
-
Austin Clements authored
The gcstats structure is no longer consumed by anything and no longer tracks statistics that are particularly relevant to the concurrent garbage collector. Remove it. (Having statistics is probably a good idea, but these aren't the stats we need these days and we don't have a way to get them out of the runtime.) In preparation for #13613. Change-Id: Ib63e2f9067850668f9dcbfd4ed89aab4a6622c3f Reviewed-on: https://go-review.googlesource.com/34936 Run-TryBot: Austin Clements <austin@google.com> Reviewed-by:
-
Elias Naur authored
The iOS test harness was recently changed in response to lldb bugs to replace breakpoints with the SIGUSR2 signal (CL 34926), and to pass the current directory in the test binary arguments (CL 35152). Both the signal sending and working directory setup is done from the go test driver. However, the new method doesn't work with tests where a C program is the test driver instead of go test: the current working directory will not be changed and SIGUSR2 is not raised. Instead of copying that logic into any C test program, rework the test harness (again) to move the setup logic to the early runtime cgo setup code. That way, the harness will run even in the library build modes. Then, use the app Info.plist file to pass the working directory, removing the need to alter the arguments after running. Finally, use the SIGINT signal instead of SIGUSR2 to avoid manipulating the signal masks or handlers. Fixes the testcarchive tests on iOS. With this CL, both darwin/arm and darwin/arm64 passes all.bash. This CL replaces CL 34926, CL 35152 as well as the fixup CL 35123 and CL 35255. They are reverted in CLs earlier in the relation chain. Change-Id: I8485c7db1404fbd8daa261efd1ea89e905121a3e Reviewed-on: https://go-review.googlesource.com/36090 Run-TryBot: Elias Naur <elias.naur@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by:
David Crawshaw <crawshaw@golang.org>
-
Elias Naur authored
This reverts commit 333f764d. Replaced by a improved strategy later in the CL relation chain. Change-Id: I70a1d2f0aa5aa0d3d0ec85b5a956c6fb60d88908 Reviewed-on: https://go-review.googlesource.com/36069 Run-TryBot: Elias Naur <elias.naur@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by:
-
Josh Bleecher Snyder authored
It was used with Node.Ullman, which is now gone. Change-Id: I83b167645659ae7ef70043b7915d642e42ca524f Reviewed-on: https://go-review.googlesource.com/37761 Run-TryBot: Josh Bleecher Snyder <josharian@gmail.com> Reviewed-by:
-
Dmitri Shuralyov authored
When calling build.Import, normally, an error is returned if the directory doesn't exist. However, that didn't happen for local import paths when build.FindOnly ImportMode was used. This change fixes that, and adds tests. It also makes the error value more consistent in all scenarios where it occurs. When calling build.Import with a local import path, the package can only exist in a single deterministic directory. That makes it possible verify that directory exists earlier in the path, and return a "cannot find package" error if it doesn't. Previously, this occurred only when build.FindOnly ImportMode was not set. It occurred quite late, after getting past Found label, to line that calls ctxt.readDir. Doing so would return an error like "no such file or directory" when the directory does not exist. Fixes #17863. Updates #17888 (relevant issue I ran into while working on this CL). Change-Id: If6a6996ac6176ac203a88bd31419748f88d89d7c Reviewed-on: https://go-review.googlesource.com/33158Reviewed-by:
-
Elias Naur authored
This reverts commit 4f0aac52. Replaced by a improved strategy later in the CL relation chain. Change-Id: Iff0333f172443bb5b01a42ad06b01edeb6aa15bc Reviewed-on: https://go-review.googlesource.com/36068 Run-TryBot: Elias Naur <elias.naur@gmail.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by:
-
Elias Naur authored
This reverts commit 593ea3b3. Replaced by a improved strategy later in the CL relation chain. Change-Id: I6963e4d1bf38e7028cf545a953e28054d83548 Change-Id: I6963e4d1bf38e7028cf545a953e28054d8354870 Reviewed-on: https://go-review.googlesource.com/36067Reviewed-by:
-
Elias Naur authored
This reverts commit 467109bf. Replaced by a improved strategy later in the CL relation chain. Change-Id: Ib90813b5a6c4716b563c8496013d2d57f9c022b8 Reviewed-on: https://go-review.googlesource.com/36066Reviewed-by:
-
- 03 Mar, 2017 16 commits
-
-
Matthew Dempsky authored
Since switching to SSA, the only remaining use for the Ullman field was in tracking whether or not an expression contained a function call. Give it a new name and encode it in our fancy new bitset field. Passes toolstash-check. Change-Id: I95b7f9cb053856320c0d66efe14996667e6011c2 Reviewed-on: https://go-review.googlesource.com/37721 Run-TryBot: Matthew Dempsky <mdempsky@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by:
Josh Bleecher Snyder <josharian@gmail.com>
-
David Lazar authored
This gives accurate line numbers to inlined functions from another package. Previously AST nodes from another package would get the line number of the import statement for that package. The following benchmark results show how the size of package export data is impacted by this change. These benchmarks were created by compiling the go1 benchmark and running `go tool pack x` to extract the export data from the resulting .a files. name old bytes new bytes delta bufio 3.59k ± 0% 4.17k ± 0% +16.25% bytes 5.51k ± 0% 6.40k ± 0% +16.21% compress/bzip2 2.69k ± 0% 3.21k ± 0% +19.74% compress/flate 5.14k ± 0% 5.57k ± 0% +8.43% compress/gzip 8.91k ± 0% 10.46k ± 0% +17.32% container/list 1.76k ± 0% 2.13k ± 0% +21.51% context 4.51k ± 0% 5.47k ± 0% +21.43% crypto 1.11k ± 0% 1.13k ± 0% +1.90% crypto/aes 475 ± 0% 475 ± 0% +0.00% crypto/cipher 1.18k ± 0% 1.18k ± 0% +0.00% crypto/des 502 ± 0% 502 ± 0% +0.00% crypto/dsa 5.96k ± 0% 6.54k ± 0% +9.82% crypto/ecdsa 6.93k ± 0% 7.69k ± 0% +10.91% crypto/elliptic 6.53k ± 0% 7.17k ± 0% +9.72% crypto/hmac 464 ± 0% 464 ± 0% +0.00% crypto/internal/cipherhw 313 ± 0% 313 ± 0% +0.00% crypto/md5 695 ± 0% 711 ± 0% +2.30% crypto/rand 5.62k ± 0% 6.21k ± 0% +10.44% crypto/rc4 512 ± 0% 512 ± 0% +0.00% crypto/rsa 7.31k ± 0% 8.10k ± 0% +10.86% crypto/sha1 760 ± 0% 777 ± 0% +2.24% crypto/sha256 523 ± 0% 523 ± 0% +0.00% crypto/sha512 663 ± 0% 663 ± 0% +0.00% crypto/subtle 873 ± 0% 1007 ± 0% +15.35% crypto/tls 29.6k ± 0% 33.8k ± 0% +14.03% crypto/x509 18.7k ± 0% 21.0k ± 0% +12.56% crypto/x509/pkix 10.6k ± 0% 12.2k ± 0% +15.22% encoding 473 ± 0% 473 ± 0% +0.00% encoding/asn1 1.42k ± 0% 1.50k ± 0% +5.99% encoding/base64 1.69k ± 0% 1.80k ± 0% +6.88% encoding/binary 2.76k ± 0% 3.51k ± 0% +27.09% encoding/gob 13.5k ± 0% 15.2k ± 0% +12.98% encoding/hex 857 ± 0% 881 ± 0% +2.80% encoding/json 12.5k ± 0% 14.9k ± 0% +19.37% encoding/pem 484 ± 0% 484 ± 0% +0.00% errors 361 ± 0% 370 ± 0% +2.49% flag 10.5k ± 0% 12.1k ± 0% +14.92% fmt 1.42k ± 0% 1.42k ± 0% +0.00% go/ast 15.8k ± 0% 17.5k ± 0% +10.31% go/parser 8.13k ± 0% 9.86k ± 0% +21.28% go/scanner 3.94k ± 0% 4.53k ± 0% +14.73% go/token 3.53k ± 0% 3.75k ± 0% +6.11% hash 507 ± 0% 507 ± 0% +0.00% hash/crc32 685 ± 0% 685 ± 0% +0.00% internal/nettrace 474 ± 0% 474 ± 0% +0.00% internal/poll 7.23k ± 0% 8.38k ± 0% +15.90% internal/race 511 ± 0% 515 ± 0% +0.78% internal/singleflight 969 ± 0% 1075 ± 0% +10.94% internal/syscall/unix 427 ± 0% 427 ± 0% +0.00% io 3.52k ± 0% 3.69k ± 0% +4.82% io/ioutil 8.48k ± 0% 9.90k ± 0% +16.72% log 5.06k ± 0% 5.98k ± 0% +18.15% math 4.02k ± 0% 4.35k ± 0% +8.26% math/big 9.28k ± 0% 9.94k ± 0% +7.13% math/bits 3.47k ± 0% 4.33k ± 0% +24.83% math/rand 1.30k ± 0% 1.32k ± 0% +2.00% mime 2.98k ± 0% 3.70k ± 0% +24.21% mime/multipart 3.68k ± 0% 4.22k ± 0% +14.65% mime/quotedprintable 2.26k ± 0% 2.65k ± 0% +17.60% net 23.0k ± 0% 25.7k ± 0% +11.74% net/http 59.1k ± 0% 66.7k ± 0% +13.00% net/http/httptest 35.3k ± 0% 40.9k ± 0% +15.80% net/http/httptrace 15.3k ± 0% 17.6k ± 0% +15.26% net/http/internal 2.77k ± 0% 3.27k ± 0% +17.89% net/textproto 4.60k ± 0% 5.25k ± 0% +14.22% net/url 1.73k ± 0% 1.84k ± 0% +6.59% os 14.3k ± 0% 16.4k ± 0% +14.86% path 589 ± 0% 606 ± 0% +2.89% path/filepath 5.07k ± 0% 6.17k ± 0% +21.79% reflect 6.43k ± 0% 6.81k ± 0% +5.90% regexp 5.88k ± 0% 6.46k ± 0% +9.77% regexp/syntax 3.24k ± 0% 3.29k ± 0% +1.73% runtime 13.1k ± 0% 14.9k ± 0% +13.73% runtime/cgo 229 ± 0% 229 ± 0% +0.00% runtime/debug 4.23k ± 0% 5.15k ± 0% +21.79% runtime/internal/atomic 905 ± 0% 905 ± 0% +0.00% runtime/internal/sys 2.04k ± 0% 2.20k ± 0% +7.64% runtime/pprof 4.73k ± 0% 5.65k ± 0% +19.41% runtime/trace 354 ± 0% 354 ± 0% +0.00% sort 1.68k ± 0% 1.85k ± 0% +10.17% strconv 1.85k ± 0% 1.95k ± 0% +5.51% strings 3.98k ± 0% 4.53k ± 0% +13.91% sync 1.52k ± 0% 1.58k ± 0% +4.28% sync/atomic 1.60k ± 0% 1.74k ± 0% +8.50% syscall 53.3k ± 0% 54.3k ± 0% +1.84% testing 8.77k ± 0% 10.09k ± 0% +14.96% testing/internal/testdeps 598 ± 0% 600 ± 0% +0.33% text/tabwriter 3.63k ± 0% 4.41k ± 0% +21.64% text/template 15.7k ± 0% 18.1k ± 0% +15.67% text/template/parse 9.12k ± 0% 10.35k ± 0% +13.48% time 6.38k ± 0% 7.14k ± 0% +11.81% unicode 4.62k ± 0% 4.66k ± 0% +0.98% unicode/utf16 707 ± 0% 791 ± 0% +11.88% unicode/utf8 1.06k ± 0% 1.20k ± 0% +12.63% vendor/golang_org/x/crypto/chacha20poly1305 1.26k ± 0% 1.43k ± 0% +13.54% vendor/golang_org/x/crypto/curve25519 392 ± 0% 392 ± 0% +0.00% vendor/golang_org/x/crypto/poly1305 426 ± 0% 426 ± 0% +0.00% vendor/golang_org/x/net/http2/hpack 4.75k ± 0% 5.77k ± 0% +21.42% vendor/golang_org/x/net/idna 355 ± 0% 355 ± 0% +0.00% vendor/golang_org/x/net/lex/httplex 616 ± 0% 644 ± 0% +4.55% vendor/golang_org/x/net/proxy 7.76k ± 0% 9.58k ± 0% +23.37% vendor/golang_org/x/text/transform 1.31k ± 0% 1.32k ± 0% +0.46% vendor/golang_org/x/text/unicode/norm 5.89k ± 0% 6.84k ± 0% +16.06% vendor/golang_org/x/text/width 1.24k ± 0% 1.27k ± 0% +2.66% [Geo mean] 2.51k 2.74k +9.14% Change-Id: I9ded911bb0ff63c530795fc85253d76b56d8abbc Reviewed-on: https://go-review.googlesource.com/37239 Run-TryBot: David Lazar <lazard@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by:
Austin Clements <austin@google.com>
-
David Lazar authored
For example, `-d pctab=pctoinline` prints the PC-inline table and inlining tree for every function. Change-Id: Ia6b9ce4d83eed0b494318d40ffe06481ec5d58ab Reviewed-on: https://go-review.googlesource.com/37235 Run-TryBot: David Lazar <lazard@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by:
-
David Lazar authored
The meaning of Version=1 was overloaded: it was reserved for file name symbols (to avoid conflicts with non-file name symbols), but was also used to mean "give me a fresh version number for this symbol." With the new inlining tree, the same file name symbol can appear in multiple entries, but each one would become a distinct symbol with its own version number. Now, we avoid duplicating symbols by using Version=0 for file name symbols and we avoid conflicts with other symbols by prefixing the symbol name with "gofile..". Change-Id: I8d0374053b8cdb6a9ca7fb71871b69b4dd369a9c Reviewed-on: https://go-review.googlesource.com/37234 Run-TryBot: David Lazar <lazard@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by:
Russ Cox <rsc@golang.org>
-
David Lazar authored
The code in https://play.golang.org/p/aYQPrTtzoK now produces the following stack trace: goroutine 1 [running]: main.(*point).negate(...) /tmp/go/main.go:8 main.main() /tmp/go/main.go:14 +0x23 Previously the stack trace missed the inlined call: goroutine 1 [running]: main.main() /tmp/go/main.go:14 +0x23 Fixes #10152. Updates #19348. Change-Id: Ib43c67012f53da0ef1a1e69bcafb65b57d9cecb2 Reviewed-on: https://go-review.googlesource.com/37233 Run-TryBot: David Lazar <lazard@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by:
-
David Lazar authored
Without this, literals keep their original source positions through inlining, which results in strange jumps in line numbers of inlined function bodies. By copying literals, inlining can update their source position like other nodes. Fixes #15453. Change-Id: Iad5d9bbfe183883794213266dc30e31bab89ee69 Reviewed-on: https://go-review.googlesource.com/37232 Run-TryBot: David Lazar <lazard@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by:
-
David Lazar authored
In order to generate accurate tracebacks, the runtime needs to know the inlined call stack for a given PC. This creates two tables per function for this purpose. The first table is the inlining tree (stored in the function's funcdata), which has a node containing the file, line, and function name for every inlined call. The second table is a PC-value table that maps each PC to a node in the inlining tree (or -1 if the PC is not the result of inlining). To give the appearance that inlining hasn't happened, the runtime also needs the original source position information of inlined AST nodes. Previously the compiler plastered over the line numbers of inlined AST nodes with the line number of the call. This meant that the PC-line table mapped each PC to line number of the outermost call in its inlined call stack, with no way to access the innermost line number. Now the compiler retains line numbers of inlined AST nodes and writes the innermost source position information to the PC-line and PC-file tables. Some tools and tests expect to see outermost line numbers, so we provide the OutermostLine function for displaying line info. To keep track of the inlined call stack for an AST node, we extend the src.PosBase type with an index into a global inlining tree. Every time the compiler inlines a call, it creates a node in the global inlining tree for the call, and writes its index to the PosBase of every inlined AST node. The parent of this node is the inlining tree index of the call. -1 signifies no parent. For each function, the compiler creates a local inlining tree and a PC-value table mapping each PC to an index in the local tree. These are written to an object file, which is read by the linker. The linker re-encodes these tables compactly by deduplicating function names and file names. This change increases the size of binaries by 4-5%. For example, this is how the go1 benchmark binary is impacted by this change: section old bytes new bytes delta .text 3.49M ± 0% 3.49M ± 0% +0.06% .rodata 1.12M ± 0% 1.21M ± 0% +8.21% .gopclntab 1.50M ± 0% 1.68M ± 0% +11.89% .debug_line 338k ± 0% 435k ± 0% +28.78% Total 9.21M ± 0% 9.58M ± 0% +4.01% Updates #19348. Change-Id: Ic4f180c3b516018138236b0c35e0218270d957d3 Reviewed-on: https://go-review.googlesource.com/37231 Run-TryBot: David Lazar <lazard@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by:
-
Aliaksandr Valialkin authored
This reduces compiler memory usage by up to 4% - see compilebench results below. name old time/op new time/op delta Template 245ms ± 4% 241ms ± 2% -1.88% (p=0.029 n=10+10) Unicode 126ms ± 3% 124ms ± 3% ~ (p=0.105 n=10+10) GoTypes 805ms ± 2% 813ms ± 3% ~ (p=0.515 n=8+10) Compiler 3.95s ± 2% 3.83s ± 1% -2.96% (p=0.000 n=9+10) MakeBash 47.4s ± 4% 46.6s ± 1% -1.59% (p=0.028 n=9+10) name old user-ns/op new user-ns/op delta Template 324M ± 5% 326M ± 3% ~ (p=0.935 n=10+10) Unicode 186M ± 5% 178M ±10% ~ (p=0.067 n=9+10) GoTypes 1.08G ± 7% 1.09G ± 4% ~ (p=0.956 n=10+10) Compiler 5.34G ± 4% 5.31G ± 1% ~ (p=0.501 n=10+8) name old alloc/op new alloc/op delta Template 41.0MB ± 0% 39.8MB ± 0% -3.03% (p=0.000 n=10+10) Unicode 32.3MB ± 0% 31.0MB ± 0% -4.13% (p=0.000 n=10+10) GoTypes 119MB ± 0% 116MB ± 0% -2.39% (p=0.000 n=10+10) Compiler 499MB ± 0% 487MB ± 0% -2.48% (p=0.000 n=10+10) name old allocs/op new allocs/op delta Template 380k ± 1% 379k ± 1% ~ (p=0.436 n=10+10) Unicode 324k ± 1% 324k ± 0% ~ (p=0.853 n=10+10) GoTypes 1.15M ± 0% 1.15M ± 0% ~ (p=0.481 n=10+10) Compiler 4.41M ± 0% 4.41M ± 0% -0.12% (p=0.007 n=10+10) name old text-bytes new text-bytes delta HelloSize 623k ± 0% 623k ± 0% ~ (all equal) CmdGoSize 6.64M ± 0% 6.64M ± 0% ~ (all equal) name old data-bytes new data-bytes delta HelloSize 5.81k ± 0% 5.81k ± 0% ~ (all equal) CmdGoSize 238k ± 0% 238k ± 0% ~ (all equal) name old bss-bytes new bss-bytes delta HelloSize 134k ± 0% 134k ± 0% ~ (all equal) CmdGoSize 152k ± 0% 152k ± 0% ~ (all equal) name old exe-bytes new exe-bytes delta HelloSize 967k ± 0% 967k ± 0% ~ (all equal) CmdGoSize 10.2M ± 0% 10.2M ± 0% ~ (all equal) Change-Id: I1f40af738254892bd6c8ba2eb43390b175753d52 Reviewed-on: https://go-review.googlesource.com/37445Reviewed-by:
-
Johan Brandhorst authored
Adds a function for easily accessing the x509.Certificate of a Server, if there is one. Also adds a helper function for getting a http.Client suitable for use with the server. This makes the steps required to test a httptest TLS server simpler. Fixes #18411 Change-Id: I2e78fe1e54e31bed9c641be2d9a099f698c7bbde Reviewed-on: https://go-review.googlesource.com/34639Reviewed-by:
-
Matthew Dempsky authored
Change-Id: I90865921584ae4bdfb6c220d439b14593d72b6f9 Reviewed-on: https://go-review.googlesource.com/37752 Run-TryBot: Matthew Dempsky <mdempsky@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by:
Cherry Zhang <cherryyz@google.com>
-
Cherry Zhang authored
The Zero op right after newobject has been removed. But this rule does not cover Store of constant zero (for SSA-able types). Add rules to cover Store op as well. Updates #19027. Change-Id: I5d2b62eeca0aa9ce8dc7205b264b779de01c660b Reviewed-on: https://go-review.googlesource.com/36836 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by:
-
Cherry Zhang authored
On amd64p32, PtrSize and RegSize don't agree, and function return value is aligned with RegSize. Fix this rule. Other architectures are not affected, where PtrSize and RegSize are the same. Change-Id: If187d3dfde3dc3b931b8e97db5eeff49a781551b Reviewed-on: https://go-review.googlesource.com/37720 Run-TryBot: Cherry Zhang <cherryyz@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by:
-
Matthew Dempsky authored
Previously the compiler rewrote constant division into OHMUL operations, but that rewriting was moved to SSA in CL 37015. Now OHMUL is unused, so we can get rid of it. Change-Id: Ib6fc7c2b6435510bafb5735b3b4f42cfd8ed8cdb Reviewed-on: https://go-review.googlesource.com/37750 Run-TryBot: Matthew Dempsky <mdempsky@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by:
-
Austin Clements authored
There are two accesses to mheap_.busy that are guarded by checks against len(mheap_.free). This works because both lists are (and must be) the same length, but it makes the code less clear. Change these to use len(mheap_.busy) so the access more clearly parallels the check. Fixes #18944. Change-Id: I9bacbd3663988df351ed4396ae9018bc71018311 Reviewed-on: https://go-review.googlesource.com/36354 Run-TryBot: Austin Clements <austin@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by:
-
Austin Clements authored
Currently sweep counts the number of allocated objects, computes the number of free objects from that, then re-computes the number of allocated objects from that. Simplify and clean this up by skipping these intermediate steps. Change-Id: I3ed98e371eb54bbcab7c8530466c4ab5fde35f0a Reviewed-on: https://go-review.googlesource.com/34935 Run-TryBot: Austin Clements <austin@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by:
Marvin Stenger <marvin.stenger94@gmail.com> Reviewed-by:
-
Austin Clements authored
Currently we scan the finalizers queue both during concurrent mark and during mark termination. This costs roughly 20ns per queued finalizer and about 1ns per unused finalizer queue slot (allocated queue length never decreases), which can drive up STW time if there are many finalizers. However, we only add finalizers to this queue during sweeping, which means that the second scan will never find anything new. Hence, we can fix this by simply not scanning the finalizers queue during mark termination. This brings the STW time under the 100µs goal even with 1,000,000 queued finalizers. Fixes #18869. Change-Id: I4ce5620c66fb7f13ebeb39ca313ce57047d1d0fb Reviewed-on: https://go-review.googlesource.com/36013 Run-TryBot: Austin Clements <austin@google.com> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by:
-
