]> Cypherpunks.ru repositories - gostls13.git/commitdiff
compress/flate: remove var newH
authorTobias Klauser <tklauser@distanz.ch>
Tue, 17 May 2022 06:13:47 +0000 (08:13 +0200)
committerGopher Robot <gobot@golang.org>
Tue, 17 May 2022 19:51:27 +0000 (19:51 +0000)
After CL 404696 hewH is no longer used outide the loop and val can be
used inside the loop instead. This leads to another slight improvement
in some benchmarks (only non-zero results reported):

name                             old time/op    new time/op     delta
Decode/Digits/Huffman/1e4-4         125µs ±40%       96µs ± 7%  -22.72%  (p=0.000 n=10+10)
Decode/Digits/Huffman/1e5-4        1.02ms ± 4%     0.97ms ± 3%   -4.29%  (p=0.000 n=10+10)
Decode/Digits/Huffman/1e6-4        10.5ms ± 3%     10.1ms ± 2%   -3.34%  (p=0.000 n=10+9)
Decode/Digits/Speed/1e4-4           130µs ± 5%      119µs ± 3%   -8.33%  (p=0.000 n=10+10)
Decode/Digits/Speed/1e5-4          1.32ms ± 3%     1.26ms ± 4%   -4.14%  (p=0.001 n=10+10)
Decode/Digits/Speed/1e6-4          13.4ms ± 3%     12.8ms ± 3%   -4.64%  (p=0.000 n=10+10)
Decode/Digits/Default/1e4-4         142µs ±19%      124µs ± 3%  -12.75%  (p=0.000 n=10+9)
Decode/Digits/Default/1e5-4        1.33ms ± 4%     1.27ms ± 2%   -4.45%  (p=0.000 n=10+10)
Decode/Digits/Compression/1e4-4     132µs ± 4%      126µs ± 3%   -4.59%  (p=0.000 n=10+10)
Decode/Digits/Compression/1e5-4    1.31ms ± 4%     1.28ms ± 3%   -2.12%  (p=0.015 n=10+10)
Decode/Digits/Compression/1e6-4    13.2ms ± 3%     12.8ms ± 4%   -3.36%  (p=0.000 n=10+10)
Decode/Newton/Huffman/1e4-4         138µs ± 5%      128µs ± 3%   -7.55%  (p=0.000 n=10+9)
Decode/Newton/Huffman/1e5-4        1.25ms ± 1%     1.23ms ± 3%   -2.21%  (p=0.027 n=8+10)
Decode/Newton/Huffman/1e6-4        13.0ms ± 5%     12.2ms ± 5%   -6.54%  (p=0.000 n=10+10)
Decode/Newton/Speed/1e4-4           128µs ± 3%      118µs ± 4%   -7.34%  (p=0.000 n=9+10)
Decode/Newton/Speed/1e5-4          1.06ms ± 3%     1.02ms ± 3%   -3.58%  (p=0.001 n=9+10)
Decode/Newton/Speed/1e6-4          10.4ms ± 4%     10.1ms ± 3%   -3.15%  (p=0.003 n=10+10)
Decode/Newton/Compression/1e4-4     105µs ± 2%      108µs ± 5%   +2.82%  (p=0.043 n=9+10)
Encode/Digits/Speed/1e5-4          1.65ms ± 2%     1.70ms ± 4%   +2.77%  (p=0.003 n=8+10)
Encode/Newton/Default/1e6-4        58.0ms ± 2%     57.0ms ± 1%   -1.59%  (p=0.001 n=9+9)

name                             old speed      new speed       delta
Decode/Digits/Huffman/1e4-4      82.2MB/s ±30%  103.9MB/s ± 8%  +26.38%  (p=0.000 n=10+10)
Decode/Digits/Huffman/1e5-4      98.5MB/s ± 4%  102.9MB/s ± 3%   +4.46%  (p=0.000 n=10+10)
Decode/Digits/Huffman/1e6-4      95.6MB/s ± 3%   98.9MB/s ± 2%   +3.44%  (p=0.000 n=10+9)
Decode/Digits/Speed/1e4-4        76.9MB/s ± 5%   83.8MB/s ± 3%   +9.06%  (p=0.000 n=10+10)
Decode/Digits/Speed/1e5-4        75.8MB/s ± 3%   79.1MB/s ± 4%   +4.34%  (p=0.001 n=10+10)
Decode/Digits/Speed/1e6-4        74.4MB/s ± 3%   78.0MB/s ± 3%   +4.86%  (p=0.000 n=10+10)
Decode/Digits/Default/1e4-4      70.7MB/s ±17%   80.6MB/s ± 2%  +13.93%  (p=0.000 n=10+9)
Decode/Digits/Default/1e5-4      75.4MB/s ± 4%   78.9MB/s ± 3%   +4.60%  (p=0.000 n=10+10)
Decode/Digits/Compression/1e4-4  75.8MB/s ± 3%   79.4MB/s ± 3%   +4.79%  (p=0.000 n=10+10)
Decode/Digits/Compression/1e5-4  76.5MB/s ± 4%   78.1MB/s ± 3%   +2.15%  (p=0.015 n=10+10)
Decode/Digits/Compression/1e6-4  75.7MB/s ± 3%   78.3MB/s ± 4%   +3.49%  (p=0.000 n=10+10)
Decode/Newton/Huffman/1e4-4      72.4MB/s ± 5%   78.3MB/s ± 3%   +8.13%  (p=0.000 n=10+9)
Decode/Newton/Huffman/1e5-4      79.8MB/s ± 1%   81.6MB/s ± 3%   +2.29%  (p=0.026 n=8+10)
Decode/Newton/Huffman/1e6-4      76.9MB/s ± 5%   82.2MB/s ± 5%   +6.96%  (p=0.000 n=10+10)
Decode/Newton/Speed/1e4-4        78.4MB/s ± 3%   84.6MB/s ± 4%   +7.92%  (p=0.000 n=9+10)
Decode/Newton/Speed/1e5-4        94.2MB/s ± 3%   97.7MB/s ± 3%   +3.72%  (p=0.001 n=9+10)
Decode/Newton/Speed/1e6-4        96.0MB/s ± 4%   99.1MB/s ± 3%   +3.24%  (p=0.003 n=10+10)
Decode/Newton/Compression/1e4-4  95.2MB/s ± 2%   92.6MB/s ± 5%   -2.67%  (p=0.043 n=9+10)
Encode/Digits/Speed/1e5-4        60.6MB/s ± 2%   59.0MB/s ± 4%   -2.66%  (p=0.002 n=8+10)
Encode/Newton/Default/1e6-4      17.3MB/s ± 2%   17.5MB/s ± 1%   +1.60%  (p=0.001 n=9+9)

Change-Id: I833b008fe5b67cd17ffdfbec22a3e4b10fcddece
Reviewed-on: https://go-review.googlesource.com/c/go/+/406754
Reviewed-by: Joseph Tsai <joetsai@digital-static.net>
Reviewed-by: Ian Lance Taylor <iant@google.com>
Auto-Submit: Tobias Klauser <tobias.klauser@gmail.com>
Run-TryBot: Tobias Klauser <tobias.klauser@gmail.com>
Reviewed-by: Michael Knyszek <mknyszek@google.com>
TryBot-Result: Gopher Robot <gobot@golang.org>

src/compress/flate/deflate.go

index 7c2dad22d7452ba1cb1d90c764785c47016373c5..93efd7cafb82d356a0b1b787fe0353d1ab93efbd 100644 (file)
@@ -210,11 +210,9 @@ func (d *compressor) fillWindow(b []byte) {
 
                dst := d.hashMatch[:dstSize]
                d.bulkHasher(toCheck, dst)
-               var newH uint32
                for i, val := range dst {
                        di := i + index
-                       newH = val
-                       hh := &d.hashHead[newH&hashMask]
+                       hh := &d.hashHead[val&hashMask]
                        // Get previous value with the same hash.
                        // Our chain should point to the previous value.
                        d.hashPrev[di&windowMask] = *hh