• Martin Möhrmann's avatar
    fmt: optimize struct layout and intbuf size · 7345fa54
    Martin Möhrmann authored
    Move boolean fields to the end of the pp struct for better packing.
    
    Increase the fmt.intbuf size to leave no padding space unused
    at the end of the fmt struct on 32 bit architectures.
    
    The size of the pp struct on amd64 is decreased from 192 byte
    to 184 byte and on 386 from 132 byte to 128 byte.
    
    Simplify buffer size calculation in fmt_integer.
    
    Consolidate test cases for intbuf overflow checks.
    
    amd64 Haswell:
    
    name                     old time/op  new time/op  delta
    SprintfPadding-2          227ns ± 2%   227ns ± 1%    ~     (p=0.941 n=25+25)
    SprintfEmpty-2           38.4ns ± 6%  35.8ns ±23%  -6.71%  (p=0.030 n=24+25)
    SprintfString-2           100ns ± 0%   101ns ± 0%  +1.00%  (p=0.000 n=19+18)
    SprintfTruncateString-2   142ns ± 1%   142ns ± 1%  +0.37%  (p=0.028 n=25+25)
    SprintfQuoteString-2      397ns ± 0%   393ns ± 0%  -0.82%  (p=0.000 n=21+24)
    SprintfInt-2              101ns ± 2%   102ns ± 4%    ~     (p=0.159 n=21+24)
    SprintfIntInt-2           155ns ± 3%   155ns ± 3%    ~     (p=0.934 n=25+25)
    SprintfPrefixedInt-2      252ns ± 2%   251ns ± 2%    ~     (p=0.198 n=25+25)
    SprintfFloat-2            184ns ± 3%   179ns ± 2%  -3.07%  (p=0.000 n=24+25)
    SprintfComplex-2          532ns ± 2%   535ns ± 2%  +0.64%  (p=0.046 n=25+24)
    SprintfBoolean-2         90.5ns ± 3%  91.6ns ± 2%  +1.17%  (p=0.000 n=22+22)
    SprintfHexString-2        164ns ± 2%   165ns ± 3%    ~     (p=0.066 n=25+25)
    SprintfHexBytes-2         171ns ± 3%   170ns ± 3%    ~     (p=0.941 n=25+24)
    SprintfBytes-2            320ns ± 1%   313ns ± 1%  -2.31%  (p=0.000 n=24+21)
    SprintfStringer-2         347ns ± 2%   348ns ± 2%    ~     (p=0.426 n=24+24)
    SprintfStructure-2        753ns ± 1%   742ns ± 1%  -1.49%  (p=0.000 n=24+25)
    FprintInt-2               145ns ± 0%   144ns ± 0%  -0.69%  (p=0.000 n=24+23)
    FprintfBytes-2            163ns ± 0%   163ns ± 0%  -0.27%  (p=0.000 n=25+25)
    FprintIntNoAlloc-2        108ns ± 0%   105ns ± 0%  -2.78%  (p=0.000 n=25+2
    
    386 Haswell:
    
    name                     old time/op  new time/op  delta
    SprintfPadding-2          426ns ± 2%   422ns ± 1%  -0.89%  (p=0.000 n=25+24)
    SprintfEmpty-2           24.6ns ± 1%  24.5ns ± 0%  -0.39%  (p=0.000 n=22+15)
    SprintfString-2          99.1ns ± 3%  95.6ns ± 0%  -3.52%  (p=0.000 n=25+24)
    SprintfTruncateString-2   156ns ± 4%   153ns ± 1%  -1.65%  (p=0.000 n=24+23)
    SprintfQuoteString-2      500ns ± 2%   493ns ± 1%  -1.49%  (p=0.000 n=25+24)
    SprintfInt-2             92.6ns ± 9%  88.3ns ± 1%  -4.72%  (p=0.000 n=25+25)
    SprintfIntInt-2           143ns ± 7%   137ns ± 2%  -4.01%  (p=0.000 n=23+25)
    SprintfPrefixedInt-2      268ns ±19%   264ns ±16%    ~     (p=0.826 n=24+24)
    SprintfFloat-2            242ns ± 4%   246ns ± 2%  +1.60%  (p=0.000 n=25+25)
    SprintfComplex-2         1.04µs ± 3%  1.03µs ± 1%  -0.89%  (p=0.026 n=25+25)
    SprintfBoolean-2         82.2ns ± 9%  80.7ns ± 2%    ~     (p=0.163 n=24+24)
    SprintfHexString-2        240ns ± 5%   224ns ± 2%  -6.94%  (p=0.000 n=25+25)
    SprintfHexBytes-2         245ns ± 3%   234ns ± 2%  -4.55%  (p=0.000 n=25+25)
    SprintfBytes-2            432ns ±13%   419ns ± 2%    ~     (p=0.081 n=23+25)
    SprintfStringer-2         356ns ± 4%   356ns ± 4%    ~     (p=0.988 n=25+25)
    SprintfStructure-2        968ns ± 5%   948ns ± 2%  -2.11%  (p=0.000 n=25+25)
    FprintInt-2               206ns ± 0%   201ns ± 0%  -2.43%  (p=0.000 n=25+21)
    FprintfBytes-2            187ns ± 1%   187ns ± 1%    ~     (p=0.420 n=25+25)
    FprintIntNoAlloc-2        173ns ± 0%   168ns ± 0%  -2.89%  (p=0.000 n=25+2
    
    amd64 Ivy Bridge:
    
    name                     old time/op  new time/op  delta
    SprintfPadding-4          203ns ± 4%   210ns ± 8%  +3.27%  (p=0.000 n=23+25)
    SprintfEmpty-4           24.4ns ± 2%  24.4ns ± 3%    ~     (p=0.487 n=24+25)
    SprintfString-4          92.4ns ± 2%  93.1ns ± 3%    ~     (p=0.087 n=22+25)
    SprintfTruncateString-4   137ns ± 3%   136ns ± 2%  -1.02%  (p=0.002 n=25+25)
    SprintfQuoteString-4      378ns ± 1%   373ns ± 1%  -1.32%  (p=0.000 n=24+22)
    SprintfInt-4             89.9ns ± 3%  90.3ns ± 4%    ~     (p=0.444 n=25+25)
    SprintfIntInt-4           137ns ± 4%   138ns ± 3%    ~     (p=0.112 n=25+23)
    SprintfPrefixedInt-4      155ns ±14%   154ns ±14%    ~     (p=0.791 n=25+25)
    SprintfFloat-4            154ns ± 2%   154ns ± 3%    ~     (p=0.789 n=25+25)
    SprintfComplex-4          396ns ± 2%   402ns ± 3%  +1.53%  (p=0.001 n=23+25)
    SprintfBoolean-4         71.0ns ± 3%  71.2ns ± 2%    ~     (p=0.515 n=25+24)
    SprintfHexString-4        156ns ± 3%   150ns ± 5%  -3.69%  (p=0.000 n=24+25)
    SprintfHexBytes-4         154ns ± 3%   157ns ± 5%  +1.72%  (p=0.003 n=24+25)
    SprintfBytes-4            297ns ± 4%   291ns ± 3%  -1.86%  (p=0.000 n=25+25)
    SprintfStringer-4         275ns ± 3%   265ns ± 3%  -3.51%  (p=0.000 n=25+25)
    SprintfStructure-4        878ns ± 2%   823ns ± 2%  -6.21%  (p=0.000 n=25+22)
    FprintInt-4               145ns ± 1%   147ns ± 2%  +0.94%  (p=0.001 n=23+25)
    FprintfBytes-4            166ns ± 1%   168ns ± 2%  +0.81%  (p=0.000 n=24+25)
    FprintIntNoAlloc-4        113ns ± 2%   109ns ± 2%  -3.79%  (p=0.000 n=24+25)
    
    386 Ivy Bridge:
    
    name                     old time/op  new time/op  delta
    SprintfPadding-4          353ns ± 4%   354ns ± 4%    ~     (p=0.769 n=25+24)
    SprintfEmpty-4           21.9ns ± 6%  21.1ns ± 3%  -3.45%  (p=0.000 n=24+25)
    SprintfString-4          94.7ns ± 1%  93.0ns ± 3%  -1.77%  (p=0.000 n=24+23)
    SprintfTruncateString-4   150ns ± 2%   147ns ± 0%  -1.71%  (p=0.000 n=25+21)
    SprintfQuoteString-4      472ns ± 1%   479ns ± 1%  +1.48%  (p=0.000 n=25+23)
    SprintfInt-4             87.0ns ± 2%  85.3ns ± 2%  -1.95%  (p=0.000 n=25+25)
    SprintfIntInt-4           137ns ± 2%   134ns ± 2%  -1.97%  (p=0.000 n=24+23)
    SprintfPrefixedInt-4      166ns ± 8%   161ns ± 8%  -3.07%  (p=0.023 n=24+24)
    SprintfFloat-4            226ns ± 1%   219ns ± 1%  -2.97%  (p=0.000 n=24+25)
    SprintfComplex-4          867ns ± 1%   784ns ± 1%  -9.47%  (p=0.000 n=24+23)
    SprintfBoolean-4         77.2ns ± 2%  76.0ns ± 2%  -1.63%  (p=0.000 n=25+25)
    SprintfHexString-4        212ns ± 2%   214ns ± 2%  +0.96%  (p=0.000 n=25+25)
    SprintfHexBytes-4         221ns ± 2%   218ns ± 1%  -1.42%  (p=0.000 n=25+24)
    SprintfBytes-4            423ns ± 3%   417ns ± 1%  -1.49%  (p=0.000 n=25+24)
    SprintfStringer-4         306ns ± 3%   298ns ± 3%  -2.57%  (p=0.000 n=24+25)
    SprintfStructure-4       1.00µs ± 2%  0.98µs ± 2%  -1.34%  (p=0.000 n=24+24)
    FprintInt-4               202ns ± 3%   197ns ± 2%  -2.04%  (p=0.000 n=25+25)
    FprintfBytes-4            186ns ± 2%   184ns ± 2%  -0.88%  (p=0.000 n=24+25)
    FprintIntNoAlloc-4        170ns ± 2%   166ns ± 2%  -2.26%  (p=0.000 n=24+25)
    
    Change-Id: I46e62bf8b6afa90a24f75b40f1d354b2084b910b
    Reviewed-on: https://go-review.googlesource.com/20984
    Run-TryBot: Rob Pike <r@golang.org>
    TryBot-Result: Gobot Gobot <gobot@golang.org>
    Reviewed-by: 's avatarRob Pike <r@golang.org>
    7345fa54
fmt_test.go 51.6 KB