Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Jump to bottom

cmd/compile: regalloc restoring to dead register #14761

Closed
randall77 opened this issue Mar 11, 2016 · 1 comment
Closed

cmd/compile: regalloc restoring to dead register #14761

randall77 opened this issue Mar 11, 2016 · 1 comment
Labels
FrozenDueToAge NeedsFix The path to resolution is known, but the work has not been done. Performance
Milestone
Go1.9

Comments

@randall77
Copy link
Contributor 

func f(p *int, v int, b bool) int {
    if b {
        *p = v
    }
    runtime.GC()
    return *p
}

This compiles to:

    0x000f 00015 (/home/khr/go/tmp2.go:5)   MOVBLZX "".b+24(FP), AX
    0x0014 00020 (/home/khr/go/tmp2.go:5)   TESTB   AL, AL
    0x0016 00022 (/home/khr/go/tmp2.go:6)   JEQ 56
    0x0018 00024 (/home/khr/go/tmp2.go:7)   MOVQ    "".p+8(FP), CX
    0x001d 00029 (/home/khr/go/tmp2.go:7)   MOVQ    "".v+16(FP), DX
    0x0022 00034 (/home/khr/go/tmp2.go:7)   MOVQ    DX, (CX)
    0x0025 00037 (/home/khr/go/tmp2.go:9)   PCDATA  $0, $0
    0x0025 00037 (/home/khr/go/tmp2.go:9)   CALL    runtime.GC(SB)
        ...
    0x0038 00056 (/home/khr/go/tmp2.go:5)   MOVQ    "".p+8(FP), CX
    0x003d 00061 (/home/khr/go/tmp2.go:9)   JMP 37

Notice the restore of p at 56 which is unnecessary, as CX is dead. This is just the register allocator trying to get the same data in the same registers as the other predecessor of instruction 37. We currently match only the values in registers which are live. But we really should match values in registers which are live and which are used at least once.
@bradfitz bradfitz added this to the Go1.7 milestone Apr 10, 2016
@randall77 randall77 modified the milestones: Go1.8, Go1.7 Apr 29, 2016
@randall77 randall77 mentioned this issue Jun 27, 2016
Open
@quentinmit quentinmit added the NeedsFix The path to resolution is known, but the work has not been done. label Oct 11, 2016
@rsc rsc modified the milestones: Go1.9, Go1.8 Oct 21, 2016
@randall77 randall77 mentioned this issue Mar 23, 2017
Closed
@gopherbot
Copy link

CL https://golang.org/cl/38457 mentions this issue.

lparth pushed a commit to lparth/go that referenced this issue Apr 13, 2017
@cherrymui @lparth
cmd/compile: improve startRegs calculation
eca0434 
In register allocation, we calculate what values are used in
and after the current block. If a value is used only after a
function call, since registers are clobbered in call, we don't
need to mark the value live at the entrance of the block.
Before this CL it is considered live, and unnecessary copy or
load may be generated when resolving merge edge.

Fixes golang#14761.

On AMD64:
name                      old time/op    new time/op    delta
BinaryTree17-12              2.84s ± 1%     2.81s ± 1%   -1.06%  (p=0.000 n=10+9)
Fannkuch11-12                3.61s ± 0%     3.55s ± 1%   -1.77%  (p=0.000 n=10+9)
FmtFprintfEmpty-12          50.4ns ± 4%    50.0ns ± 1%     ~     (p=0.785 n=9+8)
FmtFprintfString-12         80.0ns ± 3%    78.2ns ± 3%   -2.35%  (p=0.004 n=10+9)
FmtFprintfInt-12            81.3ns ± 4%    81.8ns ± 2%     ~     (p=0.159 n=10+10)
FmtFprintfIntInt-12          120ns ± 4%     118ns ± 2%     ~     (p=0.218 n=10+10)
FmtFprintfPrefixedInt-12     152ns ± 3%     155ns ± 2%   +2.11%  (p=0.026 n=10+10)
FmtFprintfFloat-12           240ns ± 1%     238ns ± 1%   -0.79%  (p=0.005 n=9+9)
FmtManyArgs-12               504ns ± 1%     510ns ± 1%   +1.14%  (p=0.000 n=8+9)
GobDecode-12                7.00ms ± 1%    6.99ms ± 0%     ~     (p=0.497 n=9+10)
GobEncode-12                5.47ms ± 1%    5.48ms ± 1%     ~     (p=0.218 n=10+10)
Gzip-12                      258ms ± 2%     256ms ± 1%   -0.96%  (p=0.043 n=10+9)
Gunzip-12                   38.6ms ± 0%    38.3ms ± 0%   -0.64%  (p=0.000 n=9+8)
HTTPClientServer-12         90.4µs ± 3%    87.2µs ±11%     ~     (p=0.053 n=9+10)
JSONEncode-12               15.6ms ± 0%    15.6ms ± 1%     ~     (p=0.077 n=9+9)
JSONDecode-12               55.1ms ± 1%    54.6ms ± 1%   -0.85%  (p=0.010 n=10+9)
Mandelbrot200-12            4.49ms ± 0%    4.47ms ± 0%   -0.25%  (p=0.000 n=10+8)
GoParse-12                  3.38ms ± 0%    3.37ms ± 1%     ~     (p=0.315 n=8+10)
RegexpMatchEasy0_32-12      82.5ns ± 4%    82.0ns ± 0%     ~     (p=0.164 n=10+8)
RegexpMatchEasy0_1K-12       203ns ± 1%     202ns ± 1%   -0.85%  (p=0.000 n=9+10)
RegexpMatchEasy1_32-12      82.3ns ± 1%    81.1ns ± 0%   -1.39%  (p=0.000 n=10+8)
RegexpMatchEasy1_1K-12       357ns ± 1%     357ns ± 1%     ~     (p=0.697 n=8+9)
RegexpMatchMedium_32-12      125ns ± 2%     126ns ± 2%     ~     (p=0.197 n=10+10)
RegexpMatchMedium_1K-12     39.6µs ± 3%    39.6µs ± 1%     ~     (p=0.971 n=10+10)
RegexpMatchHard_32-12       1.99µs ± 2%    1.99µs ± 4%     ~     (p=0.891 n=10+9)
RegexpMatchHard_1K-12       60.1µs ± 3%    60.4µs ± 3%     ~     (p=0.684 n=10+10)
Revcomp-12                   531ms ± 6%     441ms ± 0%  -16.94%  (p=0.000 n=10+9)
Template-12                 58.9ms ± 1%    58.7ms ± 1%     ~     (p=0.315 n=10+10)
TimeParse-12                 319ns ± 1%     320ns ± 4%     ~     (p=0.215 n=9+9)
TimeFormat-12                345ns ± 0%     333ns ± 1%   -3.36%  (p=0.000 n=9+10)
[Geo mean]                  52.2µs         51.6µs        -1.13%

On ARM64:
name                     old time/op    new time/op    delta
BinaryTree17-8              8.53s ± 0%     8.36s ± 0%   -1.89%  (p=0.000 n=10+10)
Fannkuch11-8                6.15s ± 0%     6.10s ± 0%   -0.67%  (p=0.000 n=10+10)
FmtFprintfEmpty-8           117ns ± 0%     117ns ± 0%     ~     (all equal)
FmtFprintfString-8          192ns ± 0%     192ns ± 0%     ~     (all equal)
FmtFprintfInt-8             198ns ± 0%     198ns ± 0%     ~     (p=0.211 n=10+10)
FmtFprintfIntInt-8          289ns ± 0%     291ns ± 0%   +0.59%  (p=0.000 n=7+10)
FmtFprintfPrefixedInt-8     320ns ± 2%     317ns ± 0%     ~     (p=0.431 n=10+8)
FmtFprintfFloat-8           538ns ± 0%     538ns ± 0%     ~     (all equal)
FmtManyArgs-8              1.17µs ± 1%    1.18µs ± 1%     ~     (p=0.063 n=10+10)
GobDecode-8                17.0ms ± 1%    17.2ms ± 1%   +0.83%  (p=0.000 n=10+10)
GobEncode-8                14.2ms ± 0%    14.1ms ± 1%   -0.78%  (p=0.001 n=9+10)
Gzip-8                      806ms ± 0%     797ms ± 0%   -1.12%  (p=0.000 n=6+9)
Gunzip-8                    131ms ± 0%     130ms ± 0%   -0.51%  (p=0.000 n=10+9)
HTTPClientServer-8          206µs ± 9%     212µs ± 2%     ~     (p=0.829 n=10+8)
JSONEncode-8               40.1ms ± 0%    40.1ms ± 0%     ~     (p=0.136 n=9+9)
JSONDecode-8                157ms ± 0%     151ms ± 0%   -3.32%  (p=0.000 n=9+9)
Mandelbrot200-8            10.1ms ± 0%    10.1ms ± 0%   -0.05%  (p=0.000 n=9+8)
GoParse-8                  8.43ms ± 0%    8.43ms ± 0%     ~     (p=0.912 n=10+10)
RegexpMatchEasy0_32-8       228ns ± 1%     227ns ± 0%   -0.26%  (p=0.026 n=10+9)
RegexpMatchEasy0_1K-8      1.92µs ± 0%    1.63µs ± 0%  -15.18%  (p=0.001 n=7+7)
RegexpMatchEasy1_32-8       258ns ± 1%     250ns ± 0%   -2.83%  (p=0.000 n=10+10)
RegexpMatchEasy1_1K-8      2.39µs ± 0%    2.13µs ± 0%  -10.94%  (p=0.000 n=9+9)
RegexpMatchMedium_32-8      352ns ± 0%     351ns ± 0%   -0.29%  (p=0.004 n=9+10)
RegexpMatchMedium_1K-8      104µs ± 0%     105µs ± 0%   +0.58%  (p=0.000 n=8+9)
RegexpMatchHard_32-8       5.84µs ± 0%    5.82µs ± 0%   -0.27%  (p=0.000 n=9+10)
RegexpMatchHard_1K-8        177µs ± 0%     177µs ± 0%   -0.07%  (p=0.000 n=9+9)
Revcomp-8                   1.57s ± 1%     1.50s ± 1%   -4.60%  (p=0.000 n=9+10)
Template-8                  157ms ± 1%     153ms ± 1%   -2.28%  (p=0.000 n=10+9)
TimeParse-8                 779ns ± 1%     770ns ± 1%   -1.18%  (p=0.013 n=10+10)
TimeFormat-8                823ns ± 2%     826ns ± 1%     ~     (p=0.324 n=10+9)
[Geo mean]                  144µs          142µs        -1.45%

Reduce cmd/go text size by 0.5%.

Change-Id: I9288ff983c4a7cf03fc0cb35b9b1750828013117
Reviewed-on: https://go-review.googlesource.com/38457
Run-TryBot: Cherry Zhang <cherryyz@google.com>
TryBot-Result: Gobot Gobot <gobot@golang.org>
Reviewed-by: David Chase <drchase@google.com>
Reviewed-by: Keith Randall <khr@golang.org>
@golang golang locked and limited conversation to collaborators Mar 29, 2018
Sign up for free to subscribe to this conversation on GitHub. Already have an account? Sign in.
Assignees
No one assigned
Labels
FrozenDueToAge NeedsFix The path to resolution is known, but the work has not been done. Performance
Projects
None yet
Milestone
Go1.9
Development

No branches or pull requests
5 participants
@bradfitz @rsc @quentinmit @randall77 @gopherbot