-
Carlo Alberto Ferraris authored
During benchmark of an internal tool we found out that (*Buffer).Reset() was surprisingly showing up in CPU profiles. This CL contains two related changes aimed at speeding up Reset(): 1. Create a fast path for Truncate(0) by moving the logic to Reset() (this makes Reset() a simple leaf func that gets inlined since it gets compiled to 3 MOVx instructions). Accordingly change calls in the rest of the Buffer methods to call Reset() instead of Truncate(0). 2. Reorder the fields in the Buffer struct so that frequently accessed fields are packed together (buf, off, lastRead). This also make them likely to be in the same cacheline. Ideally it would be advisable to have Buffer{} cacheline-aligned, but I couldn't find a way to do this without changing the size of the bootstrap array (but this will cause some regressions, because it will make duffcopy show up in CPU profiles where it wasn't showing up before). go1 benchmarks are not really affected, but some other benchmarks that exercise Buffer more show improvements: name old time/op new time/op delta BinaryTree17-4 2.46s ± 9% 2.43s ± 3% ~ (p=0.982 n=14+14) Fannkuch11-4 2.98s ± 1% 2.90s ± 1% -2.58% (p=0.000 n=15+14) FmtFprintfEmpty-4 45.2ns ± 1% 45.2ns ± 1% ~ (p=0.494 n=14+15) FmtFprintfString-4 76.8ns ± 1% 83.1ns ± 2% +8.23% (p=0.000 n=10+15) FmtFprintfInt-4 78.0ns ± 2% 74.6ns ± 1% -4.46% (p=0.000 n=15+15) FmtFprintfIntInt-4 113ns ± 1% 109ns ± 2% -2.91% (p=0.000 n=13+15) FmtFprintfPrefixedInt-4 152ns ± 2% 143ns ± 2% -6.04% (p=0.000 n=15+14) FmtFprintfFloat-4 224ns ± 1% 222ns ± 2% -1.08% (p=0.001 n=15+14) FmtManyArgs-4 464ns ± 2% 463ns ± 2% ~ (p=0.303 n=14+15) GobDecode-4 6.25ms ± 2% 6.32ms ± 3% +1.20% (p=0.002 n=14+14) GobEncode-4 5.41ms ± 2% 5.41ms ± 2% ~ (p=0.967 n=15+15) Gzip-4 215ms ± 2% 218ms ± 2% +1.35% (p=0.002 n=15+15) Gunzip-4 34.3ms ± 2% 34.2ms ± 2% ~ (p=0.539 n=15+15) HTTPClientServer-4 76.4µs ± 2% 75.4µs ± 1% -1.31% (p=0.000 n=15+15) JSONEncode-4 14.7ms ± 2% 14.6ms ± 3% ~ (p=0.094 n=14+14) JSONDecode-4 48.0ms ± 1% 48.5ms ± 1% +0.92% (p=0.001 n=14+12) Mandelbrot200-4 4.04ms ± 2% 4.06ms ± 1% ~ (p=0.108 n=15+13) GoParse-4 2.99ms ± 2% 3.00ms ± 1% ~ (p=0.130 n=15+13) RegexpMatchEasy0_32-4 78.3ns ± 1% 79.5ns ± 1% +1.51% (p=0.000 n=15+14) RegexpMatchEasy0_1K-4 185ns ± 1% 186ns ± 1% +0.76% (p=0.005 n=15+15) RegexpMatchEasy1_32-4 79.0ns ± 2% 76.7ns ± 1% -2.87% (p=0.000 n=14+15) name old speed new speed delta GobDecode-4 123MB/s ± 2% 121MB/s ± 3% -1.18% (p=0.002 n=14+14) GobEncode-4 142MB/s ± 2% 142MB/s ± 1% ~ (p=0.959 n=15+15) Gzip-4 90.3MB/s ± 2% 89.1MB/s ± 2% -1.34% (p=0.002 n=15+15) Gunzip-4 565MB/s ± 2% 567MB/s ± 2% ~ (p=0.539 n=15+15) JSONEncode-4 132MB/s ± 2% 133MB/s ± 3% ~ (p=0.091 n=14+14) JSONDecode-4 40.4MB/s ± 1% 40.0MB/s ± 1% -0.92% (p=0.001 n=14+12) GoParse-4 19.4MB/s ± 2% 19.3MB/s ± 1% ~ (p=0.121 n=15+13) RegexpMatchEasy0_32-4 409MB/s ± 1% 403MB/s ± 1% -1.47% (p=0.000 n=15+14) RegexpMatchEasy0_1K-4 5.53GB/s ± 1% 5.49GB/s ± 1% -0.86% (p=0.002 n=15+15) RegexpMatchEasy1_32-4 405MB/s ± 2% 417MB/s ± 1% +2.94% (p=0.000 n=14+15) name old time/op new time/op delta PoolsSingle1K-4 34.9ns ± 2% 30.4ns ± 4% -12.80% (p=0.000 n=15+15) PoolsSingle64K-4 36.9ns ± 1% 34.4ns ± 4% -6.72% (p=0.000 n=14+15) PoolsRandomSmall-4 34.8ns ± 3% 29.5ns ± 1% -15.19% (p=0.000 n=15+14) PoolsRandomLarge-4 38.6ns ± 1% 34.3ns ± 3% -11.17% (p=0.000 n=14+15) PoolSingle1K-4 26.1ns ± 1% 21.2ns ± 2% -18.59% (p=0.000 n=15+14) PoolSingle64K-4 26.7ns ± 2% 21.5ns ± 2% -19.72% (p=0.000 n=15+15) MakeSingle1K-4 24.2ns ± 2% 24.3ns ± 3% ~ (p=0.132 n=13+15) MakeSingle64K-4 6.76µs ± 1% 6.96µs ± 5% +2.94% (p=0.002 n=13+13) MakeRandomSmall-4 531ns ± 4% 538ns ± 5% ~ (p=0.066 n=14+15) MakeRandomLarge-4 152µs ± 0% 152µs ± 1% -0.31% (p=0.001 n=14+13) Change-Id: I86d7d9d2cac65335baf62214fbb35ba0fd8f9528 Reviewed-on: https://go-review.googlesource.com/37416 Run-TryBot: Ian Lance Taylor <iant@golang.org> TryBot-Result: Gobot Gobot <gobot@golang.org> Reviewed-by:Ian Lance Taylor <iant@golang.org>
55310403
