• Cherry Zhang's avatar
    cmd/internal/obj: remove Follow pass · bed8129e
    Cherry Zhang authored
    The Follow pass in the assembler backend reorders and copies
    instructions. This even applies to hand-written assembly code,
    which in many cases don't want to be reordered. Now that the
    SSA compiler does a good job for laying out instructions, the
    benefit of this pass is very little:
    
    AMD64: (old = with Follow, new = without Follow)
    name                      old time/op    new time/op    delta
    BinaryTree17-12              2.78s ± 1%     2.79s ± 1%  +0.44%  (p=0.000 n=20+19)
    Fannkuch11-12                3.11s ± 0%     3.31s ± 1%  +6.16%  (p=0.000 n=19+19)
    FmtFprintfEmpty-12          50.9ns ± 1%    51.6ns ± 3%  +1.40%  (p=0.000 n=17+20)
    FmtFprintfString-12          127ns ± 0%     128ns ± 1%  +0.88%  (p=0.000 n=17+17)
    FmtFprintfInt-12             122ns ± 0%     123ns ± 1%  +0.76%  (p=0.000 n=20+19)
    FmtFprintfIntInt-12          185ns ± 1%     186ns ± 1%  +0.65%  (p=0.000 n=20+19)
    FmtFprintfPrefixedInt-12     192ns ± 1%     202ns ± 1%  +4.99%  (p=0.000 n=20+19)
    FmtFprintfFloat-12           284ns ± 0%     288ns ± 0%  +1.33%  (p=0.000 n=15+19)
    FmtManyArgs-12               807ns ± 0%     804ns ± 0%  -0.44%  (p=0.000 n=16+18)
    GobDecode-12                7.23ms ± 1%    7.21ms ± 1%    ~     (p=0.052 n=20+20)
    GobEncode-12                6.09ms ± 1%    6.12ms ± 1%  +0.41%  (p=0.002 n=19+19)
    Gzip-12                      253ms ± 1%     255ms ± 1%  +0.95%  (p=0.000 n=18+20)
    Gunzip-12                   38.4ms ± 0%    38.5ms ± 0%  +0.34%  (p=0.000 n=17+17)
    HTTPClientServer-12         95.4µs ± 2%    96.1µs ± 1%  +0.78%  (p=0.002 n=19+19)
    JSONEncode-12               16.5ms ± 1%    16.6ms ± 1%  +1.17%  (p=0.000 n=19+19)
    JSONDecode-12               54.6ms ± 1%    55.3ms ± 1%  +1.23%  (p=0.000 n=18+18)
    Mandelbrot200-12            4.47ms ± 0%    4.47ms ± 0%  +0.06%  (p=0.000 n=18+18)
    GoParse-12                  3.47ms ± 1%    3.47ms ± 1%    ~     (p=0.583 n=20+20)
    RegexpMatchEasy0_32-12      84.8ns ± 1%    85.2ns ± 2%  +0.51%  (p=0.022 n=20+20)
    RegexpMatchEasy0_1K-12       206ns ± 1%     206ns ± 1%    ~     (p=0.770 n=20+20)
    RegexpMatchEasy1_32-12      82.8ns ± 1%    83.4ns ± 1%  +0.64%  (p=0.000 n=20+19)
    RegexpMatchEasy1_1K-12       363ns ± 1%     361ns ± 1%  -0.48%  (p=0.007 n=20+20)
    RegexpMatchMedium_32-12      126ns ± 1%     126ns ± 0%  +0.72%  (p=0.000 n=20+20)
    RegexpMatchMedium_1K-12     39.1µs ± 1%    39.8µs ± 0%  +1.73%  (p=0.000 n=19+19)
    RegexpMatchHard_32-12       1.97µs ± 0%    1.98µs ± 1%  +0.29%  (p=0.005 n=18+20)
    RegexpMatchHard_1K-12       59.5µs ± 1%    59.8µs ± 1%  +0.36%  (p=0.000 n=18+20)
    Revcomp-12                   442ms ± 1%     445ms ± 2%  +0.67%  (p=0.000 n=19+20)
    Template-12                 58.0ms ± 1%    57.5ms ± 1%  -0.85%  (p=0.000 n=19+19)
    TimeParse-12                 311ns ± 0%     314ns ± 0%  +0.94%  (p=0.000 n=20+18)
    TimeFormat-12                350ns ± 3%     346ns ± 0%    ~     (p=0.076 n=20+19)
    [Geo mean]                  55.9µs         56.4µs       +0.80%
    
    ARM32:
    name                     old time/op    new time/op    delta
    BinaryTree17-4              30.4s ± 0%     30.1s ± 0%  -1.14%  (p=0.000 n=10+8)
    Fannkuch11-4                13.7s ± 0%     13.6s ± 0%  -0.75%  (p=0.000 n=10+10)
    FmtFprintfEmpty-4           664ns ± 1%     651ns ± 1%  -1.96%  (p=0.000 n=7+8)
    FmtFprintfString-4         1.83µs ± 2%    1.77µs ± 2%  -3.21%  (p=0.000 n=10+10)
    FmtFprintfInt-4            1.57µs ± 2%    1.54µs ± 2%  -2.25%  (p=0.007 n=10+10)
    FmtFprintfIntInt-4         2.37µs ± 2%    2.31µs ± 1%  -2.68%  (p=0.000 n=10+10)
    FmtFprintfPrefixedInt-4    2.14µs ± 2%    2.10µs ± 1%  -1.83%  (p=0.006 n=10+10)
    FmtFprintfFloat-4          3.69µs ± 2%    3.74µs ± 1%  +1.60%  (p=0.000 n=10+10)
    FmtManyArgs-4              9.43µs ± 1%    9.17µs ± 1%  -2.70%  (p=0.000 n=10+10)
    GobDecode-4                76.3ms ± 1%    75.5ms ± 1%  -1.14%  (p=0.003 n=10+10)
    GobEncode-4                70.7ms ± 2%    69.0ms ± 1%  -2.36%  (p=0.000 n=10+10)
    Gzip-4                      2.64s ± 1%     2.65s ± 0%  +0.59%  (p=0.002 n=10+10)
    Gunzip-4                    402ms ± 0%     398ms ± 0%  -1.11%  (p=0.000 n=10+9)
    HTTPClientServer-4          458µs ± 0%     457µs ± 0%    ~     (p=0.247 n=10+10)
    JSONEncode-4                171ms ± 0%     172ms ± 0%  +0.56%  (p=0.000 n=10+10)
    JSONDecode-4                672ms ± 1%     668ms ± 1%    ~     (p=0.105 n=10+10)
    Mandelbrot200-4            33.5ms ± 0%    33.5ms ± 0%    ~     (p=0.156 n=9+10)
    GoParse-4                  33.9ms ± 0%    34.0ms ± 0%  +0.36%  (p=0.031 n=9+9)
    RegexpMatchEasy0_32-4       823ns ± 1%     835ns ± 1%  +1.49%  (p=0.000 n=8+8)
    RegexpMatchEasy0_1K-4      3.99µs ± 0%    4.02µs ± 1%  +0.92%  (p=0.000 n=8+10)
    RegexpMatchEasy1_32-4       877ns ± 3%     904ns ± 2%  +3.07%  (p=0.012 n=10+10)
    RegexpMatchEasy1_1K-4      5.99µs ± 0%    5.97µs ± 1%  -0.38%  (p=0.023 n=8+8)
    RegexpMatchMedium_32-4     1.40µs ± 2%    1.40µs ± 2%    ~     (p=0.590 n=10+9)
    RegexpMatchMedium_1K-4      357µs ± 0%     355µs ± 1%  -0.72%  (p=0.000 n=7+8)
    RegexpMatchHard_32-4       22.3µs ± 0%    22.1µs ± 0%  -0.49%  (p=0.000 n=8+7)
    RegexpMatchHard_1K-4        661µs ± 0%     658µs ± 0%  -0.42%  (p=0.000 n=8+7)
    Revcomp-4                  46.3ms ± 0%    46.3ms ± 0%    ~     (p=0.393 n=10+10)
    Template-4                  753ms ± 1%     750ms ± 0%    ~     (p=0.211 n=10+9)
    TimeParse-4                4.28µs ± 1%    4.22µs ± 1%  -1.34%  (p=0.000 n=8+10)
    TimeFormat-4               9.00µs ± 0%    9.05µs ± 0%  +0.59%  (p=0.000 n=10+10)
    [Geo mean]                  538µs          535µs       -0.55%
    
    ARM64:
    name                     old time/op    new time/op    delta
    BinaryTree17-8              8.39s ± 0%     8.39s ± 0%    ~     (p=0.684 n=10+10)
    Fannkuch11-8                5.95s ± 0%     5.99s ± 0%  +0.63%  (p=0.000 n=10+10)
    FmtFprintfEmpty-8           116ns ± 0%     116ns ± 0%    ~     (all equal)
    FmtFprintfString-8          361ns ± 0%     360ns ± 0%  -0.31%  (p=0.003 n=8+6)
    FmtFprintfInt-8             290ns ± 0%     290ns ± 0%    ~     (p=0.620 n=9+9)
    FmtFprintfIntInt-8          476ns ± 1%     469ns ± 0%  -1.47%  (p=0.000 n=10+6)
    FmtFprintfPrefixedInt-8     412ns ± 2%     417ns ± 2%  +1.39%  (p=0.006 n=9+10)
    FmtFprintfFloat-8           652ns ± 1%     652ns ± 0%    ~     (p=0.161 n=10+8)
    FmtManyArgs-8              1.94µs ± 0%    1.94µs ± 2%    ~     (p=0.781 n=10+10)
    GobDecode-8                17.7ms ± 1%    17.7ms ± 0%    ~     (p=0.962 n=10+7)
    GobEncode-8                15.6ms ± 0%    15.6ms ± 1%    ~     (p=0.063 n=10+10)
    Gzip-8                      786ms ± 0%     787ms ± 0%    ~     (p=0.356 n=10+9)
    Gunzip-8                    127ms ± 0%     127ms ± 0%  +0.08%  (p=0.028 n=10+9)
    HTTPClientServer-8          198µs ± 6%     198µs ± 7%    ~     (p=0.796 n=10+10)
    JSONEncode-8               42.5ms ± 0%    42.2ms ± 0%  -0.73%  (p=0.000 n=9+8)
    JSONDecode-8                158ms ± 1%     162ms ± 0%  +2.28%  (p=0.000 n=10+9)
    Mandelbrot200-8            10.1ms ± 0%    10.1ms ± 0%  -0.01%  (p=0.000 n=10+9)
    GoParse-8                  8.54ms ± 1%    8.63ms ± 1%  +1.06%  (p=0.000 n=10+9)
    RegexpMatchEasy0_32-8       231ns ± 1%     225ns ± 0%  -2.52%  (p=0.000 n=9+10)
    RegexpMatchEasy0_1K-8      1.63µs ± 0%    1.63µs ± 0%    ~     (p=0.170 n=10+10)
    RegexpMatchEasy1_32-8       253ns ± 0%     249ns ± 0%  -1.41%  (p=0.000 n=9+10)
    RegexpMatchEasy1_1K-8      2.08µs ± 0%    2.08µs ± 0%  -0.32%  (p=0.000 n=9+10)
    RegexpMatchMedium_32-8      355ns ± 1%     351ns ± 0%  -1.04%  (p=0.007 n=10+7)
    RegexpMatchMedium_1K-8      104µs ± 0%     104µs ± 0%    ~     (p=0.148 n=10+10)
    RegexpMatchHard_32-8       5.79µs ± 0%    5.79µs ± 0%    ~     (p=0.578 n=10+10)
    RegexpMatchHard_1K-8        176µs ± 0%     176µs ± 0%    ~     (p=0.137 n=10+10)
    Revcomp-8                   1.37s ± 1%     1.36s ± 1%  -0.26%  (p=0.023 n=10+10)
    Template-8                  151ms ± 1%     154ms ± 1%  +2.14%  (p=0.000 n=9+10)
    TimeParse-8                 723ns ± 2%     721ns ± 1%    ~     (p=0.592 n=10+10)
    TimeFormat-8                804ns ± 2%     798ns ± 3%    ~     (p=0.344 n=10+10)
    [Geo mean]                  154µs          154µs       -0.02%
    
    Therefore remove this pass. Also reduce text size by 0.5~2%.
    
    Comment out some dead code in runtime/sys_nacl_amd64p32.s
    which contains undefined symbols.
    
    Change-Id: I1473986fe5b18b3d2554ce96cdc6f0999b8d955d
    Reviewed-on: https://go-review.googlesource.com/36205
    Run-TryBot: Cherry Zhang <cherryyz@google.com>
    TryBot-Result: Gobot Gobot <gobot@golang.org>
    Reviewed-by: 's avatarRuss Cox <rsc@golang.org>
    bed8129e
Name
Last commit
Last update
..
a.out.go Loading commit data...
anames.go Loading commit data...
anames0.go Loading commit data...
asm0.go Loading commit data...
list0.go Loading commit data...
obj0.go Loading commit data...