diff options
Diffstat (limited to 'vendor/github.com/klauspost/compress/s2')
10 files changed, 784 insertions, 645 deletions
diff --git a/vendor/github.com/klauspost/compress/s2/README.md b/vendor/github.com/klauspost/compress/s2/README.md index 73c0c462d..1d80c42a5 100644 --- a/vendor/github.com/klauspost/compress/s2/README.md +++ b/vendor/github.com/klauspost/compress/s2/README.md @@ -325,35 +325,35 @@ The content compressed in this mode is fully compatible with the standard decode Snappy vs S2 **compression** speed on 16 core (32 thread) computer, using all threads and a single thread (1 CPU): -| File | S2 speed | S2 Throughput | S2 % smaller | S2 "better" | "better" throughput | "better" % smaller | -|-----------------------------------------------------------------------------------------------------|----------|---------------|--------------|-------------|---------------------|--------------------| -| [rawstudio-mint14.tar](https://files.klauspost.com/compress/rawstudio-mint14.7z) | 12.70x | 10556 MB/s | 7.35% | 4.15x | 3455 MB/s | 12.79% | -| (1 CPU) | 1.14x | 948 MB/s | - | 0.42x | 349 MB/s | - | -| [github-june-2days-2019.json](https://files.klauspost.com/compress/github-june-2days-2019.json.zst) | 17.13x | 14484 MB/s | 31.60% | 10.09x | 8533 MB/s | 37.71% | -| (1 CPU) | 1.33x | 1127 MB/s | - | 0.70x | 589 MB/s | - | -| [github-ranks-backup.bin](https://files.klauspost.com/compress/github-ranks-backup.bin.zst) | 15.14x | 12000 MB/s | -5.79% | 6.59x | 5223 MB/s | 5.80% | -| (1 CPU) | 1.11x | 877 MB/s | - | 0.47x | 370 MB/s | - | -| [consensus.db.10gb](https://files.klauspost.com/compress/consensus.db.10gb.zst) | 14.62x | 12116 MB/s | 15.90% | 5.35x | 4430 MB/s | 16.08% | -| (1 CPU) | 1.38x | 1146 MB/s | - | 0.38x | 312 MB/s | - | -| [adresser.json](https://files.klauspost.com/compress/adresser.json.zst) | 8.83x | 17579 MB/s | 43.86% | 6.54x | 13011 MB/s | 47.23% | -| (1 CPU) | 1.14x | 2259 MB/s | - | 0.74x | 1475 MB/s | - | -| [gob-stream](https://files.klauspost.com/compress/gob-stream.7z) | 16.72x | 14019 MB/s | 24.02% | 10.11x | 8477 MB/s | 30.48% | -| (1 CPU) | 1.24x | 1043 MB/s | - | 0.70x | 586 MB/s | - | -| [10gb.tar](http://mattmahoney.net/dc/10gb.html) | 13.33x | 9254 MB/s | 1.84% | 6.75x | 4686 MB/s | 6.72% | -| (1 CPU) | 0.97x | 672 MB/s | - | 0.53x | 366 MB/s | - | -| sharnd.out.2gb | 2.11x | 12639 MB/s | 0.01% | 1.98x | 11833 MB/s | 0.01% | -| (1 CPU) | 0.93x | 5594 MB/s | - | 1.34x | 8030 MB/s | - | -| [enwik9](http://mattmahoney.net/dc/textdata.html) | 19.34x | 8220 MB/s | 3.98% | 7.87x | 3345 MB/s | 15.82% | -| (1 CPU) | 1.06x | 452 MB/s | - | 0.50x | 213 MB/s | - | -| [silesia.tar](http://sun.aei.polsl.pl/~sdeor/corpus/silesia.zip) | 10.48x | 6124 MB/s | 5.67% | 3.76x | 2197 MB/s | 12.60% | -| (1 CPU) | 0.97x | 568 MB/s | - | 0.46x | 271 MB/s | - | -| [enwik10](https://encode.su/threads/3315-enwik10-benchmark-results) | 21.07x | 9020 MB/s | 6.36% | 6.91x | 2959 MB/s | 16.95% | -| (1 CPU) | 1.07x | 460 MB/s | - | 0.51x | 220 MB/s | - | +| File | S2 Speed | S2 Throughput | S2 % smaller | S2 "better" | "better" throughput | "better" % smaller | +|---------------------------------------------------------------------------------------------------------|----------|---------------|--------------|-------------|---------------------|--------------------| +| [rawstudio-mint14.tar](https://files.klauspost.com/compress/rawstudio-mint14.7z) | 16.33x | 10556 MB/s | 8.0% | 6.04x | 5252 MB/s | 14.7% | +| (1 CPU) | 1.08x | 940 MB/s | - | 0.46x | 400 MB/s | - | +| [github-june-2days-2019.json](https://files.klauspost.com/compress/github-june-2days-2019.json.zst) | 16.51x | 15224 MB/s | 31.70% | 9.47x | 8734 MB/s | 37.71% | +| (1 CPU) | 1.26x | 1157 MB/s | - | 0.60x | 556 MB/s | - | +| [github-ranks-backup.bin](https://files.klauspost.com/compress/github-ranks-backup.bin.zst) | 15.14x | 12598 MB/s | -5.76% | 6.23x | 5675 MB/s | 3.62% | +| (1 CPU) | 1.02x | 932 MB/s | - | 0.47x | 432 MB/s | - | +| [consensus.db.10gb](https://files.klauspost.com/compress/consensus.db.10gb.zst) | 11.21x | 12116 MB/s | 15.95% | 3.24x | 3500 MB/s | 18.00% | +| (1 CPU) | 1.05x | 1135 MB/s | - | 0.27x | 292 MB/s | - | +| [apache.log](https://files.klauspost.com/compress/apache.log.zst) | 8.55x | 16673 MB/s | 20.54% | 5.85x | 11420 MB/s | 24.97% | +| (1 CPU) | 1.91x | 1771 MB/s | - | 0.53x | 1041 MB/s | - | +| [gob-stream](https://files.klauspost.com/compress/gob-stream.7z) | 15.76x | 14357 MB/s | 24.01% | 8.67x | 7891 MB/s | 33.68% | +| (1 CPU) | 1.17x | 1064 MB/s | - | 0.65x | 595 MB/s | - | +| [10gb.tar](http://mattmahoney.net/dc/10gb.html) | 13.33x | 9835 MB/s | 2.34% | 6.85x | 4863 MB/s | 9.96% | +| (1 CPU) | 0.97x | 689 MB/s | - | 0.55x | 387 MB/s | - | +| sharnd.out.2gb | 9.11x | 13213 MB/s | 0.01% | 1.49x | 9184 MB/s | 0.01% | +| (1 CPU) | 0.88x | 5418 MB/s | - | 0.77x | 5417 MB/s | - | +| [sofia-air-quality-dataset csv](https://files.klauspost.com/compress/sofia-air-quality-dataset.tar.zst) | 22.00x | 11477 MB/s | 18.73% | 11.15x | 5817 MB/s | 27.88% | +| (1 CPU) | 1.23x | 642 MB/s | - | 0.71x | 642 MB/s | - | +| [silesia.tar](http://sun.aei.polsl.pl/~sdeor/corpus/silesia.zip) | 11.23x | 6520 MB/s | 5.9% | 5.35x | 3109 MB/s | 15.88% | +| (1 CPU) | 1.05x | 607 MB/s | - | 0.52x | 304 MB/s | - | +| [enwik9](https://files.klauspost.com/compress/enwik9.zst) | 19.28x | 8440 MB/s | 4.04% | 9.31x | 4076 MB/s | 18.04% | +| (1 CPU) | 1.12x | 488 MB/s | - | 0.57x | 250 MB/s | - | ### Legend -* `S2 speed`: Speed of S2 compared to Snappy, using 16 cores and 1 core. -* `S2 throughput`: Throughput of S2 in MB/s. +* `S2 Speed`: Speed of S2 compared to Snappy, using 16 cores and 1 core. +* `S2 Throughput`: Throughput of S2 in MB/s. * `S2 % smaller`: How many percent of the Snappy output size is S2 better. * `S2 "better"`: Speed when enabling "better" compression mode in S2 compared to Snappy. * `"better" throughput`: Speed when enabling "better" compression mode in S2 compared to Snappy. @@ -361,7 +361,7 @@ Snappy vs S2 **compression** speed on 16 core (32 thread) computer, using all th There is a good speedup across the board when using a single thread and a significant speedup when using multiple threads. -Machine generated data gets by far the biggest compression boost, with size being being reduced by up to 45% of Snappy size. +Machine generated data gets by far the biggest compression boost, with size being reduced by up to 35% of Snappy size. The "better" compression mode sees a good improvement in all cases, but usually at a performance cost. @@ -404,15 +404,15 @@ The "better" compression mode will actively look for shorter matches, which is w Without assembly decompression is also very fast; single goroutine decompression speed. No assembly: | File | S2 Throughput | S2 throughput | -|--------------------------------|--------------|---------------| -| consensus.db.10gb.s2 | 1.84x | 2289.8 MB/s | -| 10gb.tar.s2 | 1.30x | 867.07 MB/s | -| rawstudio-mint14.tar.s2 | 1.66x | 1329.65 MB/s | -| github-june-2days-2019.json.s2 | 2.36x | 1831.59 MB/s | -| github-ranks-backup.bin.s2 | 1.73x | 1390.7 MB/s | -| enwik9.s2 | 1.67x | 681.53 MB/s | -| adresser.json.s2 | 3.41x | 4230.53 MB/s | -| silesia.tar.s2 | 1.52x | 811.58 | +|--------------------------------|---------------|---------------| +| consensus.db.10gb.s2 | 1.84x | 2289.8 MB/s | +| 10gb.tar.s2 | 1.30x | 867.07 MB/s | +| rawstudio-mint14.tar.s2 | 1.66x | 1329.65 MB/s | +| github-june-2days-2019.json.s2 | 2.36x | 1831.59 MB/s | +| github-ranks-backup.bin.s2 | 1.73x | 1390.7 MB/s | +| enwik9.s2 | 1.67x | 681.53 MB/s | +| adresser.json.s2 | 3.41x | 4230.53 MB/s | +| silesia.tar.s2 | 1.52x | 811.58 | Even though S2 typically compresses better than Snappy, decompression speed is always better. @@ -450,14 +450,14 @@ The most reliable is a wide dataset. For this we use [`webdevdata.org-2015-01-07-subset`](https://files.klauspost.com/compress/webdevdata.org-2015-01-07-4GB-subset.7z), 53927 files, total input size: 4,014,735,833 bytes. Single goroutine used. -| * | Input | Output | Reduction | MB/s | -|-------------------|------------|------------|-----------|--------| -| S2 | 4014735833 | 1059723369 | 73.60% | **934.34** | -| S2 Better | 4014735833 | 969670507 | 75.85% | 532.70 | -| S2 Best | 4014735833 | 906625668 | **77.85%** | 46.84 | -| Snappy | 4014735833 | 1128706759 | 71.89% | 762.59 | -| S2, Snappy Output | 4014735833 | 1093821420 | 72.75% | 908.60 | -| LZ4 | 4014735833 | 1079259294 | 73.12% | 526.94 | +| * | Input | Output | Reduction | MB/s | +|-------------------|------------|------------|------------|------------| +| S2 | 4014735833 | 1059723369 | 73.60% | **936.73** | +| S2 Better | 4014735833 | 961580539 | 76.05% | 451.10 | +| S2 Best | 4014735833 | 899182886 | **77.60%** | 46.84 | +| Snappy | 4014735833 | 1128706759 | 71.89% | 790.15 | +| S2, Snappy Output | 4014735833 | 1093823291 | 72.75% | 936.60 | +| LZ4 | 4014735833 | 1063768713 | 73.50% | 452.02 | S2 delivers both the best single threaded throughput with regular mode and the best compression rate with "best". "Better" mode provides the same compression speed as LZ4 with better compression ratio. @@ -489,42 +489,23 @@ AMD64 assembly is use for both S2 and Snappy. | Absolute Perf | Snappy size | S2 Size | Snappy Speed | S2 Speed | Snappy dec | S2 dec | |-----------------------|-------------|---------|--------------|-------------|-------------|-------------| -| html | 22843 | 21111 | 16246 MB/s | 17438 MB/s | 40972 MB/s | 49263 MB/s | -| urls.10K | 335492 | 287326 | 7943 MB/s | 9693 MB/s | 22523 MB/s | 26484 MB/s | -| fireworks.jpeg | 123034 | 123100 | 349544 MB/s | 273889 MB/s | 718321 MB/s | 827552 MB/s | -| fireworks.jpeg (200B) | 146 | 155 | 8869 MB/s | 17773 MB/s | 33691 MB/s | 52421 MB/s | -| paper-100k.pdf | 85304 | 84459 | 167546 MB/s | 101263 MB/s | 326905 MB/s | 291944 MB/s | -| html_x_4 | 92234 | 21113 | 15194 MB/s | 50670 MB/s | 30843 MB/s | 32217 MB/s | -| alice29.txt | 88034 | 85975 | 5936 MB/s | 6139 MB/s | 12882 MB/s | 20044 MB/s | -| asyoulik.txt | 77503 | 79650 | 5517 MB/s | 6366 MB/s | 12735 MB/s | 22806 MB/s | -| lcet10.txt | 234661 | 220670 | 6235 MB/s | 6067 MB/s | 14519 MB/s | 18697 MB/s | -| plrabn12.txt | 319267 | 317985 | 5159 MB/s | 5726 MB/s | 11923 MB/s | 19901 MB/s | -| geo.protodata | 23335 | 18690 | 21220 MB/s | 26529 MB/s | 56271 MB/s | 62540 MB/s | -| kppkn.gtb | 69526 | 65312 | 9732 MB/s | 8559 MB/s | 18491 MB/s | 18969 MB/s | -| alice29.txt (128B) | 80 | 82 | 6691 MB/s | 15489 MB/s | 31883 MB/s | 38874 MB/s | -| alice29.txt (1000B) | 774 | 774 | 12204 MB/s | 13000 MB/s | 48056 MB/s | 52341 MB/s | -| alice29.txt (10000B) | 6648 | 6933 | 10044 MB/s | 12806 MB/s | 32378 MB/s | 46322 MB/s | -| alice29.txt (20000B) | 12686 | 13574 | 7733 MB/s | 11210 MB/s | 30566 MB/s | 58969 MB/s | - - -| Relative Perf | Snappy size | S2 size improved | S2 Speed | S2 Dec Speed | -|-----------------------|-------------|------------------|----------|--------------| -| html | 22.31% | 7.58% | 1.07x | 1.20x | -| urls.10K | 47.78% | 14.36% | 1.22x | 1.18x | -| fireworks.jpeg | 99.95% | -0.05% | 0.78x | 1.15x | -| fireworks.jpeg (200B) | 73.00% | -6.16% | 2.00x | 1.56x | -| paper-100k.pdf | 83.30% | 0.99% | 0.60x | 0.89x | -| html_x_4 | 22.52% | 77.11% | 3.33x | 1.04x | -| alice29.txt | 57.88% | 2.34% | 1.03x | 1.56x | -| asyoulik.txt | 61.91% | -2.77% | 1.15x | 1.79x | -| lcet10.txt | 54.99% | 5.96% | 0.97x | 1.29x | -| plrabn12.txt | 66.26% | 0.40% | 1.11x | 1.67x | -| geo.protodata | 19.68% | 19.91% | 1.25x | 1.11x | -| kppkn.gtb | 37.72% | 6.06% | 0.88x | 1.03x | -| alice29.txt (128B) | 62.50% | -2.50% | 2.31x | 1.22x | -| alice29.txt (1000B) | 77.40% | 0.00% | 1.07x | 1.09x | -| alice29.txt (10000B) | 66.48% | -4.29% | 1.27x | 1.43x | -| alice29.txt (20000B) | 63.43% | -7.00% | 1.45x | 1.93x | +| html | 22843 | 20868 | 16246 MB/s | 18617 MB/s | 40972 MB/s | 49263 MB/s | +| urls.10K | 335492 | 286541 | 7943 MB/s | 10201 MB/s | 22523 MB/s | 26484 MB/s | +| fireworks.jpeg | 123034 | 123100 | 349544 MB/s | 303228 MB/s | 718321 MB/s | 827552 MB/s | +| fireworks.jpeg (200B) | 146 | 155 | 8869 MB/s | 20180 MB/s | 33691 MB/s | 52421 MB/s | +| paper-100k.pdf | 85304 | 84202 | 167546 MB/s | 112988 MB/s | 326905 MB/s | 291944 MB/s | +| html_x_4 | 92234 | 20870 | 15194 MB/s | 54457 MB/s | 30843 MB/s | 32217 MB/s | +| alice29.txt | 88034 | 85934 | 5936 MB/s | 6540 MB/s | 12882 MB/s | 20044 MB/s | +| asyoulik.txt | 77503 | 79575 | 5517 MB/s | 6657 MB/s | 12735 MB/s | 22806 MB/s | +| lcet10.txt | 234661 | 220383 | 6235 MB/s | 6303 MB/s | 14519 MB/s | 18697 MB/s | +| plrabn12.txt | 319267 | 318196 | 5159 MB/s | 6074 MB/s | 11923 MB/s | 19901 MB/s | +| geo.protodata | 23335 | 18606 | 21220 MB/s | 25432 MB/s | 56271 MB/s | 62540 MB/s | +| kppkn.gtb | 69526 | 65019 | 9732 MB/s | 8905 MB/s | 18491 MB/s | 18969 MB/s | +| alice29.txt (128B) | 80 | 82 | 6691 MB/s | 17179 MB/s | 31883 MB/s | 38874 MB/s | +| alice29.txt (1000B) | 774 | 774 | 12204 MB/s | 13273 MB/s | 48056 MB/s | 52341 MB/s | +| alice29.txt (10000B) | 6648 | 6933 | 10044 MB/s | 12824 MB/s | 32378 MB/s | 46322 MB/s | +| alice29.txt (20000B) | 12686 | 13516 | 7733 MB/s | 12160 MB/s | 30566 MB/s | 58969 MB/s | + Speed is generally at or above Snappy. Small blocks gets a significant speedup, although at the expense of size. @@ -543,42 +524,23 @@ So individual benchmarks should only be seen as a guideline and the overall pict | Absolute Perf | Snappy size | Better Size | Snappy Speed | Better Speed | Snappy dec | Better dec | |-----------------------|-------------|-------------|--------------|--------------|-------------|-------------| -| html | 22843 | 19833 | 16246 MB/s | 7731 MB/s | 40972 MB/s | 40292 MB/s | -| urls.10K | 335492 | 253529 | 7943 MB/s | 3980 MB/s | 22523 MB/s | 20981 MB/s | -| fireworks.jpeg | 123034 | 123100 | 349544 MB/s | 9760 MB/s | 718321 MB/s | 823698 MB/s | -| fireworks.jpeg (200B) | 146 | 142 | 8869 MB/s | 594 MB/s | 33691 MB/s | 30101 MB/s | -| paper-100k.pdf | 85304 | 82915 | 167546 MB/s | 7470 MB/s | 326905 MB/s | 198869 MB/s | -| html_x_4 | 92234 | 19841 | 15194 MB/s | 23403 MB/s | 30843 MB/s | 30937 MB/s | -| alice29.txt | 88034 | 73218 | 5936 MB/s | 2945 MB/s | 12882 MB/s | 16611 MB/s | -| asyoulik.txt | 77503 | 66844 | 5517 MB/s | 2739 MB/s | 12735 MB/s | 14975 MB/s | -| lcet10.txt | 234661 | 190589 | 6235 MB/s | 3099 MB/s | 14519 MB/s | 16634 MB/s | -| plrabn12.txt | 319267 | 270828 | 5159 MB/s | 2600 MB/s | 11923 MB/s | 13382 MB/s | -| geo.protodata | 23335 | 18278 | 21220 MB/s | 11208 MB/s | 56271 MB/s | 57961 MB/s | -| kppkn.gtb | 69526 | 61851 | 9732 MB/s | 4556 MB/s | 18491 MB/s | 16524 MB/s | -| alice29.txt (128B) | 80 | 81 | 6691 MB/s | 529 MB/s | 31883 MB/s | 34225 MB/s | -| alice29.txt (1000B) | 774 | 748 | 12204 MB/s | 1943 MB/s | 48056 MB/s | 42068 MB/s | -| alice29.txt (10000B) | 6648 | 6234 | 10044 MB/s | 2949 MB/s | 32378 MB/s | 28813 MB/s | -| alice29.txt (20000B) | 12686 | 11584 | 7733 MB/s | 2822 MB/s | 30566 MB/s | 27315 MB/s | - - -| Relative Perf | Snappy size | Better size | Better Speed | Better dec | -|-----------------------|-------------|-------------|--------------|------------| -| html | 22.31% | 13.18% | 0.48x | 0.98x | -| urls.10K | 47.78% | 24.43% | 0.50x | 0.93x | -| fireworks.jpeg | 99.95% | -0.05% | 0.03x | 1.15x | -| fireworks.jpeg (200B) | 73.00% | 2.74% | 0.07x | 0.89x | -| paper-100k.pdf | 83.30% | 2.80% | 0.07x | 0.61x | -| html_x_4 | 22.52% | 78.49% | 0.04x | 1.00x | -| alice29.txt | 57.88% | 16.83% | 1.54x | 1.29x | -| asyoulik.txt | 61.91% | 13.75% | 0.50x | 1.18x | -| lcet10.txt | 54.99% | 18.78% | 0.50x | 1.15x | -| plrabn12.txt | 66.26% | 15.17% | 0.50x | 1.12x | -| geo.protodata | 19.68% | 21.67% | 0.50x | 1.03x | -| kppkn.gtb | 37.72% | 11.04% | 0.53x | 0.89x | -| alice29.txt (128B) | 62.50% | -1.25% | 0.47x | 1.07x | -| alice29.txt (1000B) | 77.40% | 3.36% | 0.08x | 0.88x | -| alice29.txt (10000B) | 66.48% | 6.23% | 0.16x | 0.89x | -| alice29.txt (20000B) | 63.43% | 8.69% | 0.29x | 0.89x | +| html | 22843 | 18972 | 16246 MB/s | 8621 MB/s | 40972 MB/s | 40292 MB/s | +| urls.10K | 335492 | 248079 | 7943 MB/s | 5104 MB/s | 22523 MB/s | 20981 MB/s | +| fireworks.jpeg | 123034 | 123100 | 349544 MB/s | 84429 MB/s | 718321 MB/s | 823698 MB/s | +| fireworks.jpeg (200B) | 146 | 149 | 8869 MB/s | 7125 MB/s | 33691 MB/s | 30101 MB/s | +| paper-100k.pdf | 85304 | 82887 | 167546 MB/s | 11087 MB/s | 326905 MB/s | 198869 MB/s | +| html_x_4 | 92234 | 18982 | 15194 MB/s | 29316 MB/s | 30843 MB/s | 30937 MB/s | +| alice29.txt | 88034 | 71611 | 5936 MB/s | 3709 MB/s | 12882 MB/s | 16611 MB/s | +| asyoulik.txt | 77503 | 65941 | 5517 MB/s | 3380 MB/s | 12735 MB/s | 14975 MB/s | +| lcet10.txt | 234661 | 184939 | 6235 MB/s | 3537 MB/s | 14519 MB/s | 16634 MB/s | +| plrabn12.txt | 319267 | 264990 | 5159 MB/s | 2960 MB/s | 11923 MB/s | 13382 MB/s | +| geo.protodata | 23335 | 17689 | 21220 MB/s | 10859 MB/s | 56271 MB/s | 57961 MB/s | +| kppkn.gtb | 69526 | 55398 | 9732 MB/s | 5206 MB/s | 18491 MB/s | 16524 MB/s | +| alice29.txt (128B) | 80 | 78 | 6691 MB/s | 7422 MB/s | 31883 MB/s | 34225 MB/s | +| alice29.txt (1000B) | 774 | 746 | 12204 MB/s | 5734 MB/s | 48056 MB/s | 42068 MB/s | +| alice29.txt (10000B) | 6648 | 6218 | 10044 MB/s | 6055 MB/s | 32378 MB/s | 28813 MB/s | +| alice29.txt (20000B) | 12686 | 11492 | 7733 MB/s | 3143 MB/s | 30566 MB/s | 27315 MB/s | + Except for the mostly incompressible JPEG image compression is better and usually in the double digits in terms of percentage reduction over Snappy. @@ -605,29 +567,29 @@ Some examples compared on 16 core CPU, amd64 assembly used: ``` * enwik10 -Default... 10000000000 -> 4761467548 [47.61%]; 1.098s, 8685.6MB/s -Better... 10000000000 -> 4219438251 [42.19%]; 1.925s, 4954.2MB/s -Best... 10000000000 -> 3627364337 [36.27%]; 43.051s, 221.5MB/s +Default... 10000000000 -> 4759950115 [47.60%]; 1.03s, 9263.0MB/s +Better... 10000000000 -> 4084706676 [40.85%]; 2.16s, 4415.4MB/s +Best... 10000000000 -> 3615520079 [36.16%]; 42.259s, 225.7MB/s * github-june-2days-2019.json -Default... 6273951764 -> 1043196283 [16.63%]; 431ms, 13882.3MB/s -Better... 6273951764 -> 949146808 [15.13%]; 547ms, 10938.4MB/s -Best... 6273951764 -> 832855506 [13.27%]; 9.455s, 632.8MB/s +Default... 6273951764 -> 1041700255 [16.60%]; 431ms, 13882.3MB/s +Better... 6273951764 -> 945841238 [15.08%]; 547ms, 10938.4MB/s +Best... 6273951764 -> 826392576 [13.17%]; 9.455s, 632.8MB/s * nyc-taxi-data-10M.csv -Default... 3325605752 -> 1095998837 [32.96%]; 324ms, 9788.7MB/s -Better... 3325605752 -> 954776589 [28.71%]; 491ms, 6459.4MB/s -Best... 3325605752 -> 779098746 [23.43%]; 8.29s, 382.6MB/s +Default... 3325605752 -> 1093516949 [32.88%]; 324ms, 9788.7MB/s +Better... 3325605752 -> 885394158 [26.62%]; 491ms, 6459.4MB/s +Best... 3325605752 -> 773681257 [23.26%]; 8.29s, 412.0MB/s * 10gb.tar -Default... 10065157632 -> 5916578242 [58.78%]; 1.028s, 9337.4MB/s -Better... 10065157632 -> 5649207485 [56.13%]; 1.597s, 6010.6MB/s -Best... 10065157632 -> 5208719802 [51.75%]; 32.78s, 292.8MB/ +Default... 10065157632 -> 5915541066 [58.77%]; 1.028s, 9337.4MB/s +Better... 10065157632 -> 5453844650 [54.19%]; 1.597s, 4862.7MB/s +Best... 10065157632 -> 5192495021 [51.59%]; 32.78s, 308.2MB/ * consensus.db.10gb -Default... 10737418240 -> 4562648848 [42.49%]; 882ms, 11610.0MB/s -Better... 10737418240 -> 4542428129 [42.30%]; 1.533s, 6679.7MB/s -Best... 10737418240 -> 4244773384 [39.53%]; 42.96s, 238.4MB/s +Default... 10737418240 -> 4549762344 [42.37%]; 882ms, 12118.4MB/s +Better... 10737418240 -> 4438535064 [41.34%]; 1.533s, 3500.9MB/s +Best... 10737418240 -> 4210602774 [39.21%]; 42.96s, 254.4MB/s ``` Decompression speed should be around the same as using the 'better' compression mode. @@ -648,10 +610,10 @@ If you would like more control, you can use the s2 package as described below: Snappy compatible blocks can be generated with the S2 encoder. Compression and speed is typically a bit better `MaxEncodedLen` is also smaller for smaller memory usage. Replace -| Snappy | S2 replacement | -|----------------------------|-------------------------| -| snappy.Encode(...) | s2.EncodeSnappy(...) | -| snappy.MaxEncodedLen(...) | s2.MaxEncodedLen(...) | +| Snappy | S2 replacement | +|---------------------------|-----------------------| +| snappy.Encode(...) | s2.EncodeSnappy(...) | +| snappy.MaxEncodedLen(...) | s2.MaxEncodedLen(...) | `s2.EncodeSnappy` can be replaced with `s2.EncodeSnappyBetter` or `s2.EncodeSnappyBest` to get more efficiently compressed snappy compatible output. @@ -660,12 +622,12 @@ Compression and speed is typically a bit better `MaxEncodedLen` is also smaller Comparison of [`webdevdata.org-2015-01-07-subset`](https://files.klauspost.com/compress/webdevdata.org-2015-01-07-4GB-subset.7z), 53927 files, total input size: 4,014,735,833 bytes. amd64, single goroutine used: -| Encoder | Size | MB/s | Reduction | -|-----------------------|------------|------------|------------ -| snappy.Encode | 1128706759 | 725.59 | 71.89% | -| s2.EncodeSnappy | 1093823291 | **899.16** | 72.75% | -| s2.EncodeSnappyBetter | 1001158548 | 578.49 | 75.06% | -| s2.EncodeSnappyBest | 944507998 | 66.00 | **76.47%**| +| Encoder | Size | MB/s | Reduction | +|-----------------------|------------|------------|------------| +| snappy.Encode | 1128706759 | 725.59 | 71.89% | +| s2.EncodeSnappy | 1093823291 | **899.16** | 72.75% | +| s2.EncodeSnappyBetter | 1001158548 | 578.49 | 75.06% | +| s2.EncodeSnappyBest | 944507998 | 66.00 | **76.47%** | ## Streams @@ -835,6 +797,13 @@ This is done using the regular "Skip" function: This will ensure that we are at exactly the offset we want, and reading from `dec` will start at the requested offset. +# Compact storage + +For compact storage [RemoveIndexHeaders](https://pkg.go.dev/github.com/klauspost/compress/s2#RemoveIndexHeaders) can be used to remove any redundant info from +a serialized index. If you remove the header it must be restored before [Loading](https://pkg.go.dev/github.com/klauspost/compress/s2#Index.Load). + +This is expected to save 20 bytes. These can be restored using [RestoreIndexHeaders](https://pkg.go.dev/github.com/klauspost/compress/s2#RestoreIndexHeaders). This removes a layer of security, but is the most compact representation. Returns nil if headers contains errors. + ## Index Format: Each block is structured as a snappy skippable block, with the chunk ID 0x99. @@ -844,20 +813,20 @@ The block can be read from the front, but contains information so it can be read Numbers are stored as fixed size little endian values or [zigzag encoded](https://developers.google.com/protocol-buffers/docs/encoding#signed_integers) [base 128 varints](https://developers.google.com/protocol-buffers/docs/encoding), with un-encoded value length of 64 bits, unless other limits are specified. -| Content | Format | -|---------------------------------------------------------------------------|-----------------------------------------------------------------------------------------------------------------------------| -| ID, `[1]byte` | Always 0x99. | -| Data Length, `[3]byte` | 3 byte little-endian length of the chunk in bytes, following this. | -| Header `[6]byte` | Header, must be `[115, 50, 105, 100, 120, 0]` or in text: "s2idx\x00". | -| UncompressedSize, Varint | Total Uncompressed size. | -| CompressedSize, Varint | Total Compressed size if known. Should be -1 if unknown. | -| EstBlockSize, Varint | Block Size, used for guessing uncompressed offsets. Must be >= 0. | -| Entries, Varint | Number of Entries in index, must be < 65536 and >=0. | -| HasUncompressedOffsets `byte` | 0 if no uncompressed offsets are present, 1 if present. Other values are invalid. | -| UncompressedOffsets, [Entries]VarInt | Uncompressed offsets. See below how to decode. | -| CompressedOffsets, [Entries]VarInt | Compressed offsets. See below how to decode. | -| Block Size, `[4]byte` | Little Endian total encoded size (including header and trailer). Can be used for searching backwards to start of block. | -| Trailer `[6]byte` | Trailer, must be `[0, 120, 100, 105, 50, 115]` or in text: "\x00xdi2s". Can be used for identifying block from end of stream. | +| Content | Format | +|--------------------------------------|-------------------------------------------------------------------------------------------------------------------------------| +| ID, `[1]byte` | Always 0x99. | +| Data Length, `[3]byte` | 3 byte little-endian length of the chunk in bytes, following this. | +| Header `[6]byte` | Header, must be `[115, 50, 105, 100, 120, 0]` or in text: "s2idx\x00". | +| UncompressedSize, Varint | Total Uncompressed size. | +| CompressedSize, Varint | Total Compressed size if known. Should be -1 if unknown. | +| EstBlockSize, Varint | Block Size, used for guessing uncompressed offsets. Must be >= 0. | +| Entries, Varint | Number of Entries in index, must be < 65536 and >=0. | +| HasUncompressedOffsets `byte` | 0 if no uncompressed offsets are present, 1 if present. Other values are invalid. | +| UncompressedOffsets, [Entries]VarInt | Uncompressed offsets. See below how to decode. | +| CompressedOffsets, [Entries]VarInt | Compressed offsets. See below how to decode. | +| Block Size, `[4]byte` | Little Endian total encoded size (including header and trailer). Can be used for searching backwards to start of block. | +| Trailer `[6]byte` | Trailer, must be `[0, 120, 100, 105, 50, 115]` or in text: "\x00xdi2s". Can be used for identifying block from end of stream. | For regular streams the uncompressed offsets are fully predictable, so `HasUncompressedOffsets` allows to specify that compressed blocks all have @@ -929,6 +898,7 @@ To decode from any given uncompressed offset `(wantOffset)`: See [using indexes](https://github.com/klauspost/compress/tree/master/s2#using-indexes) for functions that perform the operations with a simpler interface. + # Format Extensions * Frame [Stream identifier](https://github.com/google/snappy/blob/master/framing_format.txt#L68) changed from `sNaPpY` to `S2sTwO`. @@ -951,10 +921,11 @@ The length is specified by reading the 3-bit length specified in the tag and dec | 7 | 65540 + read 3 bytes | This allows any repeat offset + length to be represented by 2 to 5 bytes. +It also allows to emit matches longer than 64 bytes with one copy + one repeat instead of several 64 byte copies. Lengths are stored as little endian values. -The first copy of a block cannot be a repeat offset and the offset is not carried across blocks in streams. +The first copy of a block cannot be a repeat offset and the offset is reset on every block in streams. Default streaming block size is 1MB. diff --git a/vendor/github.com/klauspost/compress/s2/decode.go b/vendor/github.com/klauspost/compress/s2/decode.go index 27c0f3c2c..00c5cc72c 100644 --- a/vendor/github.com/klauspost/compress/s2/decode.go +++ b/vendor/github.com/klauspost/compress/s2/decode.go @@ -952,7 +952,11 @@ func (r *Reader) ReadSeeker(random bool, index []byte) (*ReadSeeker, error) { // Seek allows seeking in compressed data. func (r *ReadSeeker) Seek(offset int64, whence int) (int64, error) { if r.err != nil { - return 0, r.err + if !errors.Is(r.err, io.EOF) { + return 0, r.err + } + // Reset on EOF + r.err = nil } if offset == 0 && whence == io.SeekCurrent { return r.blockStart + int64(r.i), nil diff --git a/vendor/github.com/klauspost/compress/s2/decode_other.go b/vendor/github.com/klauspost/compress/s2/decode_other.go index 1074ebd21..11300c3a8 100644 --- a/vendor/github.com/klauspost/compress/s2/decode_other.go +++ b/vendor/github.com/klauspost/compress/s2/decode_other.go @@ -28,6 +28,9 @@ func s2Decode(dst, src []byte) int { // As long as we can read at least 5 bytes... for s < len(src)-5 { + // Removing bounds checks is SLOWER, when if doing + // in := src[s:s+5] + // Checked on Go 1.18 switch src[s] & 0x03 { case tagLiteral: x := uint32(src[s] >> 2) @@ -38,14 +41,19 @@ func s2Decode(dst, src []byte) int { s += 2 x = uint32(src[s-1]) case x == 61: + in := src[s : s+3] + x = uint32(in[1]) | uint32(in[2])<<8 s += 3 - x = uint32(src[s-2]) | uint32(src[s-1])<<8 case x == 62: + in := src[s : s+4] + // Load as 32 bit and shift down. + x = uint32(in[0]) | uint32(in[1])<<8 | uint32(in[2])<<16 | uint32(in[3])<<24 + x >>= 8 s += 4 - x = uint32(src[s-3]) | uint32(src[s-2])<<8 | uint32(src[s-1])<<16 case x == 63: + in := src[s : s+5] + x = uint32(in[1]) | uint32(in[2])<<8 | uint32(in[3])<<16 | uint32(in[4])<<24 s += 5 - x = uint32(src[s-4]) | uint32(src[s-3])<<8 | uint32(src[s-2])<<16 | uint32(src[s-1])<<24 } length = int(x) + 1 if length > len(dst)-d || length > len(src)-s || (strconv.IntSize == 32 && length <= 0) { @@ -62,8 +70,8 @@ func s2Decode(dst, src []byte) int { case tagCopy1: s += 2 - length = int(src[s-2]) >> 2 & 0x7 toffset := int(uint32(src[s-2])&0xe0<<3 | uint32(src[s-1])) + length = int(src[s-2]) >> 2 & 0x7 if toffset == 0 { if debug { fmt.Print("(repeat) ") @@ -71,14 +79,16 @@ func s2Decode(dst, src []byte) int { // keep last offset switch length { case 5: + length = int(src[s]) + 4 s += 1 - length = int(uint32(src[s-1])) + 4 case 6: + in := src[s : s+2] + length = int(uint32(in[0])|(uint32(in[1])<<8)) + (1 << 8) s += 2 - length = int(uint32(src[s-2])|(uint32(src[s-1])<<8)) + (1 << 8) case 7: + in := src[s : s+3] + length = int((uint32(in[2])<<16)|(uint32(in[1])<<8)|uint32(in[0])) + (1 << 16) s += 3 - length = int(uint32(src[s-3])|(uint32(src[s-2])<<8)|(uint32(src[s-1])<<16)) + (1 << 16) default: // 0-> 4 } } else { @@ -86,14 +96,16 @@ func s2Decode(dst, src []byte) int { } length += 4 case tagCopy2: + in := src[s : s+3] + offset = int(uint32(in[1]) | uint32(in[2])<<8) + length = 1 + int(in[0])>>2 s += 3 - length = 1 + int(src[s-3])>>2 - offset = int(uint32(src[s-2]) | uint32(src[s-1])<<8) case tagCopy4: + in := src[s : s+5] + offset = int(uint32(in[1]) | uint32(in[2])<<8 | uint32(in[3])<<16 | uint32(in[4])<<24) + length = 1 + int(in[0])>>2 s += 5 - length = 1 + int(src[s-5])>>2 - offset = int(uint32(src[s-4]) | uint32(src[s-3])<<8 | uint32(src[s-2])<<16 | uint32(src[s-1])<<24) } if offset <= 0 || d < offset || length > len(dst)-d { diff --git a/vendor/github.com/klauspost/compress/s2/encode_all.go b/vendor/github.com/klauspost/compress/s2/encode_all.go index 8b16c38a6..54c71d3b5 100644 --- a/vendor/github.com/klauspost/compress/s2/encode_all.go +++ b/vendor/github.com/klauspost/compress/s2/encode_all.go @@ -58,8 +58,9 @@ func encodeGo(dst, src []byte) []byte { // been written. // // It also assumes that: +// // len(dst) >= MaxEncodedLen(len(src)) && -// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize +// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize func encodeBlockGo(dst, src []byte) (d int) { // Initialize the hash table. const ( diff --git a/vendor/github.com/klauspost/compress/s2/encode_amd64.go b/vendor/github.com/klauspost/compress/s2/encode_amd64.go index e612225f4..6b93daa5a 100644 --- a/vendor/github.com/klauspost/compress/s2/encode_amd64.go +++ b/vendor/github.com/klauspost/compress/s2/encode_amd64.go @@ -8,8 +8,9 @@ package s2 // been written. // // It also assumes that: +// // len(dst) >= MaxEncodedLen(len(src)) && -// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize +// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize func encodeBlock(dst, src []byte) (d int) { const ( // Use 12 bit table when less than... @@ -43,8 +44,9 @@ func encodeBlock(dst, src []byte) (d int) { // been written. // // It also assumes that: +// // len(dst) >= MaxEncodedLen(len(src)) && -// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize +// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize func encodeBlockBetter(dst, src []byte) (d int) { const ( // Use 12 bit table when less than... @@ -78,8 +80,9 @@ func encodeBlockBetter(dst, src []byte) (d int) { // been written. // // It also assumes that: +// // len(dst) >= MaxEncodedLen(len(src)) && -// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize +// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize func encodeBlockSnappy(dst, src []byte) (d int) { const ( // Use 12 bit table when less than... @@ -112,8 +115,9 @@ func encodeBlockSnappy(dst, src []byte) (d int) { // been written. // // It also assumes that: +// // len(dst) >= MaxEncodedLen(len(src)) && -// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize +// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize func encodeBlockBetterSnappy(dst, src []byte) (d int) { const ( // Use 12 bit table when less than... diff --git a/vendor/github.com/klauspost/compress/s2/encode_best.go b/vendor/github.com/klauspost/compress/s2/encode_best.go index 4bc80bc6a..1b7ea394f 100644 --- a/vendor/github.com/klauspost/compress/s2/encode_best.go +++ b/vendor/github.com/klauspost/compress/s2/encode_best.go @@ -15,8 +15,9 @@ import ( // been written. // // It also assumes that: +// // len(dst) >= MaxEncodedLen(len(src)) && -// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize +// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize func encodeBlockBest(dst, src []byte) (d int) { // Initialize the hash tables. const ( @@ -176,14 +177,21 @@ func encodeBlockBest(dst, src []byte) (d int) { best = bestOf(best, matchAt(getPrev(nextLong), s, uint32(cv), false)) } // Search for a match at best match end, see if that is better. - if sAt := best.s + best.length; sAt < sLimit { - sBack := best.s - backL := best.length + // Allow some bytes at the beginning to mismatch. + // Sweet spot is around 1-2 bytes, but depends on input. + // The skipped bytes are tested in Extend backwards, + // and still picked up as part of the match if they do. + const skipBeginning = 2 + const skipEnd = 1 + if sAt := best.s + best.length - skipEnd; sAt < sLimit { + + sBack := best.s + skipBeginning - skipEnd + backL := best.length - skipBeginning // Load initial values cv = load64(src, sBack) - // Search for mismatch + + // Grab candidates... next := lTable[hash8(load64(src, sAt), lTableBits)] - //next := sTable[hash4(load64(src, sAt), sTableBits)] if checkAt := getCur(next) - backL; checkAt > 0 { best = bestOf(best, matchAt(checkAt, sBack, uint32(cv), false)) @@ -191,6 +199,16 @@ func encodeBlockBest(dst, src []byte) (d int) { if checkAt := getPrev(next) - backL; checkAt > 0 { best = bestOf(best, matchAt(checkAt, sBack, uint32(cv), false)) } + // Disabled: Extremely small gain + if false { + next = sTable[hash4(load64(src, sAt), sTableBits)] + if checkAt := getCur(next) - backL; checkAt > 0 { + best = bestOf(best, matchAt(checkAt, sBack, uint32(cv), false)) + } + if checkAt := getPrev(next) - backL; checkAt > 0 { + best = bestOf(best, matchAt(checkAt, sBack, uint32(cv), false)) + } + } } } } @@ -288,8 +306,9 @@ emitRemainder: // been written. // // It also assumes that: +// // len(dst) >= MaxEncodedLen(len(src)) && -// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize +// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize func encodeBlockBestSnappy(dst, src []byte) (d int) { // Initialize the hash tables. const ( @@ -546,6 +565,7 @@ emitRemainder: // emitCopySize returns the size to encode the offset+length // // It assumes that: +// // 1 <= offset && offset <= math.MaxUint32 // 4 <= length && length <= 1 << 24 func emitCopySize(offset, length int) int { @@ -584,6 +604,7 @@ func emitCopySize(offset, length int) int { // emitCopyNoRepeatSize returns the size to encode the offset+length // // It assumes that: +// // 1 <= offset && offset <= math.MaxUint32 // 4 <= length && length <= 1 << 24 func emitCopyNoRepeatSize(offset, length int) int { diff --git a/vendor/github.com/klauspost/compress/s2/encode_better.go b/vendor/github.com/klauspost/compress/s2/encode_better.go index 943215b8a..3b66ba42b 100644 --- a/vendor/github.com/klauspost/compress/s2/encode_better.go +++ b/vendor/github.com/klauspost/compress/s2/encode_better.go @@ -42,8 +42,9 @@ func hash8(u uint64, h uint8) uint32 { // been written. // // It also assumes that: +// // len(dst) >= MaxEncodedLen(len(src)) && -// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize +// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize func encodeBlockBetterGo(dst, src []byte) (d int) { // sLimit is when to stop looking for offset/length copies. The inputMargin // lets us use a fast path for emitLiteral in the main loop, while we are @@ -56,7 +57,7 @@ func encodeBlockBetterGo(dst, src []byte) (d int) { // Initialize the hash tables. const ( // Long hash matches. - lTableBits = 16 + lTableBits = 17 maxLTableSize = 1 << lTableBits // Short hash matches. @@ -97,9 +98,26 @@ func encodeBlockBetterGo(dst, src []byte) (d int) { lTable[hashL] = uint32(s) sTable[hashS] = uint32(s) + valLong := load64(src, candidateL) + valShort := load64(src, candidateS) + + // If long matches at least 8 bytes, use that. + if cv == valLong { + break + } + if cv == valShort { + candidateL = candidateS + break + } + // Check repeat at offset checkRep. const checkRep = 1 - if false && uint32(cv>>(checkRep*8)) == load32(src, s-repeat+checkRep) { + // Minimum length of a repeat. Tested with various values. + // While 4-5 offers improvements in some, 6 reduces + // regressions significantly. + const wantRepeatBytes = 6 + const repeatMask = ((1 << (wantRepeatBytes * 8)) - 1) << (8 * checkRep) + if false && repeat > 0 && cv&repeatMask == load64(src, s-repeat)&repeatMask { base := s + checkRep // Extend back for i := base - repeat; base > nextEmit && i > 0 && src[i-1] == src[base-1]; { @@ -109,8 +127,8 @@ func encodeBlockBetterGo(dst, src []byte) (d int) { d += emitLiteral(dst[d:], src[nextEmit:base]) // Extend forward - candidate := s - repeat + 4 + checkRep - s += 4 + checkRep + candidate := s - repeat + wantRepeatBytes + checkRep + s += wantRepeatBytes + checkRep for s < len(src) { if len(src)-s < 8 { if src[s] == src[candidate] { @@ -127,28 +145,40 @@ func encodeBlockBetterGo(dst, src []byte) (d int) { s += 8 candidate += 8 } - if nextEmit > 0 { - // same as `add := emitCopy(dst[d:], repeat, s-base)` but skips storing offset. - d += emitRepeat(dst[d:], repeat, s-base) - } else { - // First match, cannot be repeat. - d += emitCopy(dst[d:], repeat, s-base) - } + // same as `add := emitCopy(dst[d:], repeat, s-base)` but skips storing offset. + d += emitRepeat(dst[d:], repeat, s-base) nextEmit = s if s >= sLimit { goto emitRemainder } + // Index in-between + index0 := base + 1 + index1 := s - 2 + + cv = load64(src, s) + for index0 < index1 { + cv0 := load64(src, index0) + cv1 := load64(src, index1) + lTable[hash7(cv0, lTableBits)] = uint32(index0) + sTable[hash4(cv0>>8, sTableBits)] = uint32(index0 + 1) + + lTable[hash7(cv1, lTableBits)] = uint32(index1) + sTable[hash4(cv1>>8, sTableBits)] = uint32(index1 + 1) + index0 += 2 + index1 -= 2 + } cv = load64(src, s) continue } - if uint32(cv) == load32(src, candidateL) { + // Long likely matches 7, so take that. + if uint32(cv) == uint32(valLong) { break } // Check our short candidate - if uint32(cv) == load32(src, candidateS) { + if uint32(cv) == uint32(valShort) { // Try a long candidate at s+1 hashL = hash7(cv>>8, lTableBits) candidateL = int(lTable[hashL]) @@ -227,21 +257,29 @@ func encodeBlockBetterGo(dst, src []byte) (d int) { // Do we have space for more, if not bail. return 0 } - // Index match start+1 (long) and start+2 (short) + + // Index short & long index0 := base + 1 - // Index match end-2 (long) and end-1 (short) index1 := s - 2 cv0 := load64(src, index0) cv1 := load64(src, index1) - cv = load64(src, s) lTable[hash7(cv0, lTableBits)] = uint32(index0) - lTable[hash7(cv0>>8, lTableBits)] = uint32(index0 + 1) - lTable[hash7(cv1, lTableBits)] = uint32(index1) - lTable[hash7(cv1>>8, lTableBits)] = uint32(index1 + 1) sTable[hash4(cv0>>8, sTableBits)] = uint32(index0 + 1) - sTable[hash4(cv0>>16, sTableBits)] = uint32(index0 + 2) + + lTable[hash7(cv1, lTableBits)] = uint32(index1) sTable[hash4(cv1>>8, sTableBits)] = uint32(index1 + 1) + index0 += 1 + index1 -= 1 + cv = load64(src, s) + + // index every second long in between. + for index0 < index1 { + lTable[hash7(load64(src, index0), lTableBits)] = uint32(index0) + lTable[hash7(load64(src, index1), lTableBits)] = uint32(index1) + index0 += 2 + index1 -= 2 + } } emitRemainder: @@ -260,8 +298,9 @@ emitRemainder: // been written. // // It also assumes that: +// // len(dst) >= MaxEncodedLen(len(src)) && -// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize +// minNonLiteralBlockSize <= len(src) && len(src) <= maxBlockSize func encodeBlockBetterSnappyGo(dst, src []byte) (d int) { // sLimit is when to stop looking for offset/length copies. The inputMargin // lets us use a fast path for emitLiteral in the main loop, while we are @@ -402,21 +441,29 @@ func encodeBlockBetterSnappyGo(dst, src []byte) (d int) { // Do we have space for more, if not bail. return 0 } - // Index match start+1 (long) and start+2 (short) + + // Index short & long index0 := base + 1 - // Index match end-2 (long) and end-1 (short) index1 := s - 2 cv0 := load64(src, index0) cv1 := load64(src, index1) - cv = load64(src, s) lTable[hash7(cv0, lTableBits)] = uint32(index0) - lTable[hash7(cv0>>8, lTableBits)] = uint32(index0 + 1) - lTable[hash7(cv1, lTableBits)] = uint32(index1) - lTable[hash7(cv1>>8, lTableBits)] = uint32(index1 + 1) sTable[hash4(cv0>>8, sTableBits)] = uint32(index0 + 1) - sTable[hash4(cv0>>16, sTableBits)] = uint32(index0 + 2) + + lTable[hash7(cv1, lTableBits)] = uint32(index1) sTable[hash4(cv1>>8, sTableBits)] = uint32(index1 + 1) + index0 += 1 + index1 -= 1 + cv = load64(src, s) + + // index every second long in between. + for index0 < index1 { + lTable[hash7(load64(src, index0), lTableBits)] = uint32(index0) + lTable[hash7(load64(src, index1), lTableBits)] = uint32(index1) + index0 += 2 + index1 -= 2 + } } emitRemainder: diff --git a/vendor/github.com/klauspost/compress/s2/encode_go.go b/vendor/github.com/klauspost/compress/s2/encode_go.go index 94784b82a..db08fc355 100644 --- a/vendor/github.com/klauspost/compress/s2/encode_go.go +++ b/vendor/github.com/klauspost/compress/s2/encode_go.go @@ -12,6 +12,7 @@ import ( // been written. // // It also assumes that: +// // len(dst) >= MaxEncodedLen(len(src)) func encodeBlock(dst, src []byte) (d int) { if len(src) < minNonLiteralBlockSize { @@ -25,6 +26,7 @@ func encodeBlock(dst, src []byte) (d int) { // been written. // // It also assumes that: +// // len(dst) >= MaxEncodedLen(len(src)) func encodeBlockBetter(dst, src []byte) (d int) { return encodeBlockBetterGo(dst, src) @@ -35,6 +37,7 @@ func encodeBlockBetter(dst, src []byte) (d int) { // been written. // // It also assumes that: +// // len(dst) >= MaxEncodedLen(len(src)) func encodeBlockBetterSnappy(dst, src []byte) (d int) { return encodeBlockBetterSnappyGo(dst, src) @@ -45,6 +48,7 @@ func encodeBlockBetterSnappy(dst, src []byte) (d int) { // been written. // // It also assumes that: +// // len(dst) >= MaxEncodedLen(len(src)) func encodeBlockSnappy(dst, src []byte) (d int) { if len(src) < minNonLiteralBlockSize { @@ -56,6 +60,7 @@ func encodeBlockSnappy(dst, src []byte) (d int) { // emitLiteral writes a literal chunk and returns the number of bytes written. // // It assumes that: +// // dst is long enough to hold the encoded bytes // 0 <= len(lit) && len(lit) <= math.MaxUint32 func emitLiteral(dst, lit []byte) int { @@ -146,6 +151,7 @@ func emitRepeat(dst []byte, offset, length int) int { // emitCopy writes a copy chunk and returns the number of bytes written. // // It assumes that: +// // dst is long enough to hold the encoded bytes // 1 <= offset && offset <= math.MaxUint32 // 4 <= length && length <= 1 << 24 @@ -214,6 +220,7 @@ func emitCopy(dst []byte, offset, length int) int { // emitCopyNoRepeat writes a copy chunk and returns the number of bytes written. // // It assumes that: +// // dst is long enough to hold the encoded bytes // 1 <= offset && offset <= math.MaxUint32 // 4 <= length && length <= 1 << 24 @@ -273,8 +280,8 @@ func emitCopyNoRepeat(dst []byte, offset, length int) int { // matchLen returns how many bytes match in a and b // // It assumes that: -// len(a) <= len(b) // +// len(a) <= len(b) func matchLen(a []byte, b []byte) int { b = b[:len(a)] var checked int diff --git a/vendor/github.com/klauspost/compress/s2/encodeblock_amd64.go b/vendor/github.com/klauspost/compress/s2/encodeblock_amd64.go index 88f27c099..7e00bac3e 100644 --- a/vendor/github.com/klauspost/compress/s2/encodeblock_amd64.go +++ b/vendor/github.com/klauspost/compress/s2/encodeblock_amd64.go @@ -1,7 +1,6 @@ // Code generated by command: go run gen.go -out ../encodeblock_amd64.s -stubs ../encodeblock_amd64.go -pkg=s2. DO NOT EDIT. //go:build !appengine && !noasm && gc && !noasm -// +build !appengine,!noasm,gc,!noasm package s2 @@ -150,8 +149,9 @@ func encodeSnappyBetterBlockAsm8B(dst []byte, src []byte) int // emitLiteral writes a literal chunk and returns the number of bytes written. // // It assumes that: -// dst is long enough to hold the encoded bytes with margin of 0 bytes -// 0 <= len(lit) && len(lit) <= math.MaxUint32 +// +// dst is long enough to hold the encoded bytes with margin of 0 bytes +// 0 <= len(lit) && len(lit) <= math.MaxUint32 // //go:noescape func emitLiteral(dst []byte, lit []byte) int @@ -165,9 +165,10 @@ func emitRepeat(dst []byte, offset int, length int) int // emitCopy writes a copy chunk and returns the number of bytes written. // // It assumes that: -// dst is long enough to hold the encoded bytes -// 1 <= offset && offset <= math.MaxUint32 -// 4 <= length && length <= 1 << 24 +// +// dst is long enough to hold the encoded bytes +// 1 <= offset && offset <= math.MaxUint32 +// 4 <= length && length <= 1 << 24 // //go:noescape func emitCopy(dst []byte, offset int, length int) int @@ -175,9 +176,10 @@ func emitCopy(dst []byte, offset int, length int) int // emitCopyNoRepeat writes a copy chunk and returns the number of bytes written. // // It assumes that: -// dst is long enough to hold the encoded bytes -// 1 <= offset && offset <= math.MaxUint32 -// 4 <= length && length <= 1 << 24 +// +// dst is long enough to hold the encoded bytes +// 1 <= offset && offset <= math.MaxUint32 +// 4 <= length && length <= 1 << 24 // //go:noescape func emitCopyNoRepeat(dst []byte, offset int, length int) int @@ -185,7 +187,8 @@ func emitCopyNoRepeat(dst []byte, offset int, length int) int // matchLen returns how many bytes match in a and b // // It assumes that: -// len(a) <= len(b) +// +// len(a) <= len(b) // //go:noescape func matchLen(a []byte, b []byte) int diff --git a/vendor/github.com/klauspost/compress/s2/encodeblock_amd64.s b/vendor/github.com/klauspost/compress/s2/encodeblock_amd64.s index 36915d949..81a487d6d 100644 --- a/vendor/github.com/klauspost/compress/s2/encodeblock_amd64.s +++ b/vendor/github.com/klauspost/compress/s2/encodeblock_amd64.s @@ -1,7 +1,6 @@ // Code generated by command: go run gen.go -out ../encodeblock_amd64.s -stubs ../encodeblock_amd64.go -pkg=s2. DO NOT EDIT. //go:build !appengine && !noasm && gc && !noasm -// +build !appengine,!noasm,gc,!noasm #include "textflag.h" @@ -5743,9 +5742,9 @@ emit_literal_done_emit_remainder_encodeBlockAsm8B: // func encodeBetterBlockAsm(dst []byte, src []byte) int // Requires: BMI, SSE2 -TEXT ·encodeBetterBlockAsm(SB), $327704-56 +TEXT ·encodeBetterBlockAsm(SB), $589848-56 MOVQ dst_base+0(FP), AX - MOVQ $0x00000a00, CX + MOVQ $0x00001200, CX LEAQ 24(SP), DX PXOR X0, X0 @@ -5797,27 +5796,37 @@ check_maxskip_cont_encodeBetterBlockAsm: MOVQ DI, R11 SHLQ $0x08, R10 IMULQ R9, R10 - SHRQ $0x30, R10 + SHRQ $0x2f, R10 SHLQ $0x20, R11 IMULQ SI, R11 SHRQ $0x32, R11 MOVL 24(SP)(R10*4), SI - MOVL 262168(SP)(R11*4), R8 + MOVL 524312(SP)(R11*4), R8 MOVL CX, 24(SP)(R10*4) - MOVL CX, 262168(SP)(R11*4) - CMPL (DX)(SI*1), DI + MOVL CX, 524312(SP)(R11*4) + MOVQ (DX)(SI*1), R10 + MOVQ (DX)(R8*1), R11 + CMPQ R10, DI JEQ candidate_match_encodeBetterBlockAsm - CMPL (DX)(R8*1), DI - JEQ candidateS_match_encodeBetterBlockAsm - MOVL 20(SP), CX - JMP search_loop_encodeBetterBlockAsm + CMPQ R11, DI + JNE no_short_found_encodeBetterBlockAsm + MOVL R8, SI + JMP candidate_match_encodeBetterBlockAsm + +no_short_found_encodeBetterBlockAsm: + CMPL R10, DI + JEQ candidate_match_encodeBetterBlockAsm + CMPL R11, DI + JEQ candidateS_match_encodeBetterBlockAsm + MOVL 20(SP), CX + JMP search_loop_encodeBetterBlockAsm candidateS_match_encodeBetterBlockAsm: SHRQ $0x08, DI MOVQ DI, R10 SHLQ $0x08, R10 IMULQ R9, R10 - SHRQ $0x30, R10 + SHRQ $0x2f, R10 MOVL 24(SP)(R10*4), SI INCL CX MOVL CX, 24(SP)(R10*4) @@ -6590,52 +6599,49 @@ match_nolit_emitcopy_end_encodeBetterBlockAsm: match_nolit_dst_ok_encodeBetterBlockAsm: MOVQ $0x00cf1bbcdcbfa563, SI MOVQ $0x9e3779b1, R8 - INCL DI - MOVQ (DX)(DI*1), R9 - MOVQ R9, R10 - MOVQ R9, R11 - MOVQ R9, R12 - SHRQ $0x08, R11 - MOVQ R11, R13 - SHRQ $0x10, R12 - LEAL 1(DI), R14 - LEAL 2(DI), R15 - MOVQ -2(DX)(CX*1), R9 + LEAQ 1(DI), DI + LEAQ -2(CX), R9 + MOVQ (DX)(DI*1), R10 + MOVQ 1(DX)(DI*1), R11 + MOVQ (DX)(R9*1), R12 + MOVQ 1(DX)(R9*1), R13 SHLQ $0x08, R10 IMULQ SI, R10 - SHRQ $0x30, R10 - SHLQ $0x08, R13 - IMULQ SI, R13 - SHRQ $0x30, R13 + SHRQ $0x2f, R10 SHLQ $0x20, R11 IMULQ R8, R11 SHRQ $0x32, R11 - SHLQ $0x20, R12 - IMULQ R8, R12 - SHRQ $0x32, R12 + SHLQ $0x08, R12 + IMULQ SI, R12 + SHRQ $0x2f, R12 + SHLQ $0x20, R13 + IMULQ R8, R13 + SHRQ $0x32, R13 + LEAQ 1(DI), R8 + LEAQ 1(R9), R14 MOVL DI, 24(SP)(R10*4) - MOVL R14, 24(SP)(R13*4) - MOVL R14, 262168(SP)(R11*4) - MOVL R15, 262168(SP)(R12*4) - MOVQ R9, R10 - MOVQ R9, R11 - SHRQ $0x08, R11 - MOVQ R11, R13 - LEAL -2(CX), R9 - LEAL -1(CX), DI + MOVL R9, 24(SP)(R12*4) + MOVL R8, 524312(SP)(R11*4) + MOVL R14, 524312(SP)(R13*4) + ADDQ $0x01, DI + SUBQ $0x01, R9 + +index_loop_encodeBetterBlockAsm: + CMPQ DI, R9 + JAE search_loop_encodeBetterBlockAsm + MOVQ (DX)(DI*1), R8 + MOVQ (DX)(R9*1), R10 + SHLQ $0x08, R8 + IMULQ SI, R8 + SHRQ $0x2f, R8 SHLQ $0x08, R10 IMULQ SI, R10 - SHRQ $0x30, R10 - SHLQ $0x20, R11 - IMULQ R8, R11 - SHRQ $0x32, R11 - SHLQ $0x08, R13 - IMULQ SI, R13 - SHRQ $0x30, R13 + SHRQ $0x2f, R10 + MOVL DI, 24(SP)(R8*4) MOVL R9, 24(SP)(R10*4) - MOVL DI, 262168(SP)(R11*4) - MOVL DI, 24(SP)(R13*4) - JMP search_loop_encodeBetterBlockAsm + ADDQ $0x02, DI + SUBQ $0x02, R9 + JMP index_loop_encodeBetterBlockAsm emit_remainder_encodeBetterBlockAsm: MOVQ src_len+32(FP), CX @@ -6815,9 +6821,9 @@ emit_literal_done_emit_remainder_encodeBetterBlockAsm: // func encodeBetterBlockAsm4MB(dst []byte, src []byte) int // Requires: BMI, SSE2 -TEXT ·encodeBetterBlockAsm4MB(SB), $327704-56 +TEXT ·encodeBetterBlockAsm4MB(SB), $589848-56 MOVQ dst_base+0(FP), AX - MOVQ $0x00000a00, CX + MOVQ $0x00001200, CX LEAQ 24(SP), DX PXOR X0, X0 @@ -6869,27 +6875,37 @@ check_maxskip_cont_encodeBetterBlockAsm4MB: MOVQ DI, R11 SHLQ $0x08, R10 IMULQ R9, R10 - SHRQ $0x30, R10 + SHRQ $0x2f, R10 SHLQ $0x20, R11 IMULQ SI, R11 SHRQ $0x32, R11 MOVL 24(SP)(R10*4), SI - MOVL 262168(SP)(R11*4), R8 + MOVL 524312(SP)(R11*4), R8 MOVL CX, 24(SP)(R10*4) - MOVL CX, 262168(SP)(R11*4) - CMPL (DX)(SI*1), DI + MOVL CX, 524312(SP)(R11*4) + MOVQ (DX)(SI*1), R10 + MOVQ (DX)(R8*1), R11 + CMPQ R10, DI JEQ candidate_match_encodeBetterBlockAsm4MB - CMPL (DX)(R8*1), DI - JEQ candidateS_match_encodeBetterBlockAsm4MB - MOVL 20(SP), CX - JMP search_loop_encodeBetterBlockAsm4MB + CMPQ R11, DI + JNE no_short_found_encodeBetterBlockAsm4MB + MOVL R8, SI + JMP candidate_match_encodeBetterBlockAsm4MB + +no_short_found_encodeBetterBlockAsm4MB: + CMPL R10, DI + JEQ candidate_match_encodeBetterBlockAsm4MB + CMPL R11, DI + JEQ candidateS_match_encodeBetterBlockAsm4MB + MOVL 20(SP), CX + JMP search_loop_encodeBetterBlockAsm4MB candidateS_match_encodeBetterBlockAsm4MB: SHRQ $0x08, DI MOVQ DI, R10 SHLQ $0x08, R10 IMULQ R9, R10 - SHRQ $0x30, R10 + SHRQ $0x2f, R10 MOVL 24(SP)(R10*4), SI INCL CX MOVL CX, 24(SP)(R10*4) @@ -7600,52 +7616,49 @@ match_nolit_emitcopy_end_encodeBetterBlockAsm4MB: match_nolit_dst_ok_encodeBetterBlockAsm4MB: MOVQ $0x00cf1bbcdcbfa563, SI MOVQ $0x9e3779b1, R8 - INCL DI - MOVQ (DX)(DI*1), R9 - MOVQ R9, R10 - MOVQ R9, R11 - MOVQ R9, R12 - SHRQ $0x08, R11 - MOVQ R11, R13 - SHRQ $0x10, R12 - LEAL 1(DI), R14 - LEAL 2(DI), R15 - MOVQ -2(DX)(CX*1), R9 + LEAQ 1(DI), DI + LEAQ -2(CX), R9 + MOVQ (DX)(DI*1), R10 + MOVQ 1(DX)(DI*1), R11 + MOVQ (DX)(R9*1), R12 + MOVQ 1(DX)(R9*1), R13 SHLQ $0x08, R10 IMULQ SI, R10 - SHRQ $0x30, R10 - SHLQ $0x08, R13 - IMULQ SI, R13 - SHRQ $0x30, R13 + SHRQ $0x2f, R10 SHLQ $0x20, R11 IMULQ R8, R11 SHRQ $0x32, R11 - SHLQ $0x20, R12 - IMULQ R8, R12 - SHRQ $0x32, R12 + SHLQ $0x08, R12 + IMULQ SI, R12 + SHRQ $0x2f, R12 + SHLQ $0x20, R13 + IMULQ R8, R13 + SHRQ $0x32, R13 + LEAQ 1(DI), R8 + LEAQ 1(R9), R14 MOVL DI, 24(SP)(R10*4) - MOVL R14, 24(SP)(R13*4) - MOVL R14, 262168(SP)(R11*4) - MOVL R15, 262168(SP)(R12*4) - MOVQ R9, R10 - MOVQ R9, R11 - SHRQ $0x08, R11 - MOVQ R11, R13 - LEAL -2(CX), R9 - LEAL -1(CX), DI + MOVL R9, 24(SP)(R12*4) + MOVL R8, 524312(SP)(R11*4) + MOVL R14, 524312(SP)(R13*4) + ADDQ $0x01, DI + SUBQ $0x01, R9 + +index_loop_encodeBetterBlockAsm4MB: + CMPQ DI, R9 + JAE search_loop_encodeBetterBlockAsm4MB + MOVQ (DX)(DI*1), R8 + MOVQ (DX)(R9*1), R10 + SHLQ $0x08, R8 + IMULQ SI, R8 + SHRQ $0x2f, R8 SHLQ $0x08, R10 IMULQ SI, R10 - SHRQ $0x30, R10 - SHLQ $0x20, R11 - IMULQ R8, R11 - SHRQ $0x32, R11 - SHLQ $0x08, R13 - IMULQ SI, R13 - SHRQ $0x30, R13 + SHRQ $0x2f, R10 + MOVL DI, 24(SP)(R8*4) MOVL R9, 24(SP)(R10*4) - MOVL DI, 262168(SP)(R11*4) - MOVL DI, 24(SP)(R13*4) - JMP search_loop_encodeBetterBlockAsm4MB + ADDQ $0x02, DI + SUBQ $0x02, R9 + JMP index_loop_encodeBetterBlockAsm4MB emit_remainder_encodeBetterBlockAsm4MB: MOVQ src_len+32(FP), CX @@ -7871,12 +7884,22 @@ search_loop_encodeBetterBlockAsm12B: MOVL 65560(SP)(R11*4), R8 MOVL CX, 24(SP)(R10*4) MOVL CX, 65560(SP)(R11*4) - CMPL (DX)(SI*1), DI + MOVQ (DX)(SI*1), R10 + MOVQ (DX)(R8*1), R11 + CMPQ R10, DI JEQ candidate_match_encodeBetterBlockAsm12B - CMPL (DX)(R8*1), DI - JEQ candidateS_match_encodeBetterBlockAsm12B - MOVL 20(SP), CX - JMP search_loop_encodeBetterBlockAsm12B + CMPQ R11, DI + JNE no_short_found_encodeBetterBlockAsm12B + MOVL R8, SI + JMP candidate_match_encodeBetterBlockAsm12B + +no_short_found_encodeBetterBlockAsm12B: + CMPL R10, DI + JEQ candidate_match_encodeBetterBlockAsm12B + CMPL R11, DI + JEQ candidateS_match_encodeBetterBlockAsm12B + MOVL 20(SP), CX + JMP search_loop_encodeBetterBlockAsm12B candidateS_match_encodeBetterBlockAsm12B: SHRQ $0x08, DI @@ -8447,52 +8470,49 @@ match_nolit_emitcopy_end_encodeBetterBlockAsm12B: match_nolit_dst_ok_encodeBetterBlockAsm12B: MOVQ $0x0000cf1bbcdcbf9b, SI MOVQ $0x9e3779b1, R8 - INCL DI - MOVQ (DX)(DI*1), R9 - MOVQ R9, R10 - MOVQ R9, R11 - MOVQ R9, R12 - SHRQ $0x08, R11 - MOVQ R11, R13 - SHRQ $0x10, R12 - LEAL 1(DI), R14 - LEAL 2(DI), R15 - MOVQ -2(DX)(CX*1), R9 + LEAQ 1(DI), DI + LEAQ -2(CX), R9 + MOVQ (DX)(DI*1), R10 + MOVQ 1(DX)(DI*1), R11 + MOVQ (DX)(R9*1), R12 + MOVQ 1(DX)(R9*1), R13 SHLQ $0x10, R10 IMULQ SI, R10 SHRQ $0x32, R10 - SHLQ $0x10, R13 - IMULQ SI, R13 - SHRQ $0x32, R13 SHLQ $0x20, R11 IMULQ R8, R11 SHRQ $0x34, R11 - SHLQ $0x20, R12 - IMULQ R8, R12 - SHRQ $0x34, R12 + SHLQ $0x10, R12 + IMULQ SI, R12 + SHRQ $0x32, R12 + SHLQ $0x20, R13 + IMULQ R8, R13 + SHRQ $0x34, R13 + LEAQ 1(DI), R8 + LEAQ 1(R9), R14 MOVL DI, 24(SP)(R10*4) - MOVL R14, 24(SP)(R13*4) - MOVL R14, 65560(SP)(R11*4) - MOVL R15, 65560(SP)(R12*4) - MOVQ R9, R10 - MOVQ R9, R11 - SHRQ $0x08, R11 - MOVQ R11, R13 - LEAL -2(CX), R9 - LEAL -1(CX), DI + MOVL R9, 24(SP)(R12*4) + MOVL R8, 65560(SP)(R11*4) + MOVL R14, 65560(SP)(R13*4) + ADDQ $0x01, DI + SUBQ $0x01, R9 + +index_loop_encodeBetterBlockAsm12B: + CMPQ DI, R9 + JAE search_loop_encodeBetterBlockAsm12B + MOVQ (DX)(DI*1), R8 + MOVQ (DX)(R9*1), R10 + SHLQ $0x10, R8 + IMULQ SI, R8 + SHRQ $0x32, R8 SHLQ $0x10, R10 IMULQ SI, R10 SHRQ $0x32, R10 - SHLQ $0x20, R11 - IMULQ R8, R11 - SHRQ $0x34, R11 - SHLQ $0x10, R13 - IMULQ SI, R13 - SHRQ $0x32, R13 + MOVL DI, 24(SP)(R8*4) MOVL R9, 24(SP)(R10*4) - MOVL DI, 65560(SP)(R11*4) - MOVL DI, 24(SP)(R13*4) - JMP search_loop_encodeBetterBlockAsm12B + ADDQ $0x02, DI + SUBQ $0x02, R9 + JMP index_loop_encodeBetterBlockAsm12B emit_remainder_encodeBetterBlockAsm12B: MOVQ src_len+32(FP), CX @@ -8707,12 +8727,22 @@ search_loop_encodeBetterBlockAsm10B: MOVL 16408(SP)(R11*4), R8 MOVL CX, 24(SP)(R10*4) MOVL CX, 16408(SP)(R11*4) - CMPL (DX)(SI*1), DI + MOVQ (DX)(SI*1), R10 + MOVQ (DX)(R8*1), R11 + CMPQ R10, DI JEQ candidate_match_encodeBetterBlockAsm10B - CMPL (DX)(R8*1), DI - JEQ candidateS_match_encodeBetterBlockAsm10B - MOVL 20(SP), CX - JMP search_loop_encodeBetterBlockAsm10B + CMPQ R11, DI + JNE no_short_found_encodeBetterBlockAsm10B + MOVL R8, SI + JMP candidate_match_encodeBetterBlockAsm10B + +no_short_found_encodeBetterBlockAsm10B: + CMPL R10, DI + JEQ candidate_match_encodeBetterBlockAsm10B + CMPL R11, DI + JEQ candidateS_match_encodeBetterBlockAsm10B + MOVL 20(SP), CX + JMP search_loop_encodeBetterBlockAsm10B candidateS_match_encodeBetterBlockAsm10B: SHRQ $0x08, DI @@ -9283,52 +9313,49 @@ match_nolit_emitcopy_end_encodeBetterBlockAsm10B: match_nolit_dst_ok_encodeBetterBlockAsm10B: MOVQ $0x0000cf1bbcdcbf9b, SI MOVQ $0x9e3779b1, R8 - INCL DI - MOVQ (DX)(DI*1), R9 - MOVQ R9, R10 - MOVQ R9, R11 - MOVQ R9, R12 - SHRQ $0x08, R11 - MOVQ R11, R13 - SHRQ $0x10, R12 - LEAL 1(DI), R14 - LEAL 2(DI), R15 - MOVQ -2(DX)(CX*1), R9 + LEAQ 1(DI), DI + LEAQ -2(CX), R9 + MOVQ (DX)(DI*1), R10 + MOVQ 1(DX)(DI*1), R11 + MOVQ (DX)(R9*1), R12 + MOVQ 1(DX)(R9*1), R13 SHLQ $0x10, R10 IMULQ SI, R10 SHRQ $0x34, R10 - SHLQ $0x10, R13 - IMULQ SI, R13 - SHRQ $0x34, R13 SHLQ $0x20, R11 IMULQ R8, R11 SHRQ $0x36, R11 - SHLQ $0x20, R12 - IMULQ R8, R12 - SHRQ $0x36, R12 + SHLQ $0x10, R12 + IMULQ SI, R12 + SHRQ $0x34, R12 + SHLQ $0x20, R13 + IMULQ R8, R13 + SHRQ $0x36, R13 + LEAQ 1(DI), R8 + LEAQ 1(R9), R14 MOVL DI, 24(SP)(R10*4) - MOVL R14, 24(SP)(R13*4) - MOVL R14, 16408(SP)(R11*4) - MOVL R15, 16408(SP)(R12*4) - MOVQ R9, R10 - MOVQ R9, R11 - SHRQ $0x08, R11 - MOVQ R11, R13 - LEAL -2(CX), R9 - LEAL -1(CX), DI + MOVL R9, 24(SP)(R12*4) + MOVL R8, 16408(SP)(R11*4) + MOVL R14, 16408(SP)(R13*4) + ADDQ $0x01, DI + SUBQ $0x01, R9 + +index_loop_encodeBetterBlockAsm10B: + CMPQ DI, R9 + JAE search_loop_encodeBetterBlockAsm10B + MOVQ (DX)(DI*1), R8 + MOVQ (DX)(R9*1), R10 + SHLQ $0x10, R8 + IMULQ SI, R8 + SHRQ $0x34, R8 SHLQ $0x10, R10 IMULQ SI, R10 SHRQ $0x34, R10 - SHLQ $0x20, R11 - IMULQ R8, R11 - SHRQ $0x36, R11 - SHLQ $0x10, R13 - IMULQ SI, R13 - SHRQ $0x34, R13 + MOVL DI, 24(SP)(R8*4) MOVL R9, 24(SP)(R10*4) - MOVL DI, 16408(SP)(R11*4) - MOVL DI, 24(SP)(R13*4) - JMP search_loop_encodeBetterBlockAsm10B + ADDQ $0x02, DI + SUBQ $0x02, R9 + JMP index_loop_encodeBetterBlockAsm10B emit_remainder_encodeBetterBlockAsm10B: MOVQ src_len+32(FP), CX @@ -9543,12 +9570,22 @@ search_loop_encodeBetterBlockAsm8B: MOVL 4120(SP)(R11*4), R8 MOVL CX, 24(SP)(R10*4) MOVL CX, 4120(SP)(R11*4) - CMPL (DX)(SI*1), DI + MOVQ (DX)(SI*1), R10 + MOVQ (DX)(R8*1), R11 + CMPQ R10, DI JEQ candidate_match_encodeBetterBlockAsm8B - CMPL (DX)(R8*1), DI - JEQ candidateS_match_encodeBetterBlockAsm8B - MOVL 20(SP), CX - JMP search_loop_encodeBetterBlockAsm8B + CMPQ R11, DI + JNE no_short_found_encodeBetterBlockAsm8B + MOVL R8, SI + JMP candidate_match_encodeBetterBlockAsm8B + +no_short_found_encodeBetterBlockAsm8B: + CMPL R10, DI + JEQ candidate_match_encodeBetterBlockAsm8B + CMPL R11, DI + JEQ candidateS_match_encodeBetterBlockAsm8B + MOVL 20(SP), CX + JMP search_loop_encodeBetterBlockAsm8B candidateS_match_encodeBetterBlockAsm8B: SHRQ $0x08, DI @@ -10105,52 +10142,49 @@ match_nolit_emitcopy_end_encodeBetterBlockAsm8B: match_nolit_dst_ok_encodeBetterBlockAsm8B: MOVQ $0x0000cf1bbcdcbf9b, SI MOVQ $0x9e3779b1, R8 - INCL DI - MOVQ (DX)(DI*1), R9 - MOVQ R9, R10 - MOVQ R9, R11 - MOVQ R9, R12 - SHRQ $0x08, R11 - MOVQ R11, R13 - SHRQ $0x10, R12 - LEAL 1(DI), R14 - LEAL 2(DI), R15 - MOVQ -2(DX)(CX*1), R9 + LEAQ 1(DI), DI + LEAQ -2(CX), R9 + MOVQ (DX)(DI*1), R10 + MOVQ 1(DX)(DI*1), R11 + MOVQ (DX)(R9*1), R12 + MOVQ 1(DX)(R9*1), R13 SHLQ $0x10, R10 IMULQ SI, R10 SHRQ $0x36, R10 - SHLQ $0x10, R13 - IMULQ SI, R13 - SHRQ $0x36, R13 SHLQ $0x20, R11 IMULQ R8, R11 SHRQ $0x38, R11 - SHLQ $0x20, R12 - IMULQ R8, R12 - SHRQ $0x38, R12 + SHLQ $0x10, R12 + IMULQ SI, R12 + SHRQ $0x36, R12 + SHLQ $0x20, R13 + IMULQ R8, R13 + SHRQ $0x38, R13 + LEAQ 1(DI), R8 + LEAQ 1(R9), R14 MOVL DI, 24(SP)(R10*4) - MOVL R14, 24(SP)(R13*4) - MOVL R14, 4120(SP)(R11*4) - MOVL R15, 4120(SP)(R12*4) - MOVQ R9, R10 - MOVQ R9, R11 - SHRQ $0x08, R11 - MOVQ R11, R13 - LEAL -2(CX), R9 - LEAL -1(CX), DI + MOVL R9, 24(SP)(R12*4) + MOVL R8, 4120(SP)(R11*4) + MOVL R14, 4120(SP)(R13*4) + ADDQ $0x01, DI + SUBQ $0x01, R9 + +index_loop_encodeBetterBlockAsm8B: + CMPQ DI, R9 + JAE search_loop_encodeBetterBlockAsm8B + MOVQ (DX)(DI*1), R8 + MOVQ (DX)(R9*1), R10 + SHLQ $0x10, R8 + IMULQ SI, R8 + SHRQ $0x36, R8 SHLQ $0x10, R10 IMULQ SI, R10 SHRQ $0x36, R10 - SHLQ $0x20, R11 - IMULQ R8, R11 - SHRQ $0x38, R11 - SHLQ $0x10, R13 - IMULQ SI, R13 - SHRQ $0x36, R13 + MOVL DI, 24(SP)(R8*4) MOVL R9, 24(SP)(R10*4) - MOVL DI, 4120(SP)(R11*4) - MOVL DI, 24(SP)(R13*4) - JMP search_loop_encodeBetterBlockAsm8B + ADDQ $0x02, DI + SUBQ $0x02, R9 + JMP index_loop_encodeBetterBlockAsm8B emit_remainder_encodeBetterBlockAsm8B: MOVQ src_len+32(FP), CX @@ -14287,9 +14321,9 @@ emit_literal_done_emit_remainder_encodeSnappyBlockAsm8B: // func encodeSnappyBetterBlockAsm(dst []byte, src []byte) int // Requires: BMI, SSE2 -TEXT ·encodeSnappyBetterBlockAsm(SB), $327704-56 +TEXT ·encodeSnappyBetterBlockAsm(SB), $589848-56 MOVQ dst_base+0(FP), AX - MOVQ $0x00000a00, CX + MOVQ $0x00001200, CX LEAQ 24(SP), DX PXOR X0, X0 @@ -14341,27 +14375,37 @@ check_maxskip_cont_encodeSnappyBetterBlockAsm: MOVQ DI, R11 SHLQ $0x08, R10 IMULQ R9, R10 - SHRQ $0x30, R10 + SHRQ $0x2f, R10 SHLQ $0x20, R11 IMULQ SI, R11 SHRQ $0x32, R11 MOVL 24(SP)(R10*4), SI - MOVL 262168(SP)(R11*4), R8 + MOVL 524312(SP)(R11*4), R8 MOVL CX, 24(SP)(R10*4) - MOVL CX, 262168(SP)(R11*4) - CMPL (DX)(SI*1), DI + MOVL CX, 524312(SP)(R11*4) + MOVQ (DX)(SI*1), R10 + MOVQ (DX)(R8*1), R11 + CMPQ R10, DI JEQ candidate_match_encodeSnappyBetterBlockAsm - CMPL (DX)(R8*1), DI - JEQ candidateS_match_encodeSnappyBetterBlockAsm - MOVL 20(SP), CX - JMP search_loop_encodeSnappyBetterBlockAsm + CMPQ R11, DI + JNE no_short_found_encodeSnappyBetterBlockAsm + MOVL R8, SI + JMP candidate_match_encodeSnappyBetterBlockAsm + +no_short_found_encodeSnappyBetterBlockAsm: + CMPL R10, DI + JEQ candidate_match_encodeSnappyBetterBlockAsm + CMPL R11, DI + JEQ candidateS_match_encodeSnappyBetterBlockAsm + MOVL 20(SP), CX + JMP search_loop_encodeSnappyBetterBlockAsm candidateS_match_encodeSnappyBetterBlockAsm: SHRQ $0x08, DI MOVQ DI, R10 SHLQ $0x08, R10 IMULQ R9, R10 - SHRQ $0x30, R10 + SHRQ $0x2f, R10 MOVL 24(SP)(R10*4), SI INCL CX MOVL CX, 24(SP)(R10*4) @@ -14685,52 +14729,49 @@ match_nolit_emitcopy_end_encodeSnappyBetterBlockAsm: match_nolit_dst_ok_encodeSnappyBetterBlockAsm: MOVQ $0x00cf1bbcdcbfa563, SI MOVQ $0x9e3779b1, R8 - INCL DI - MOVQ (DX)(DI*1), R9 - MOVQ R9, R10 - MOVQ R9, R11 - MOVQ R9, R12 - SHRQ $0x08, R11 - MOVQ R11, R13 - SHRQ $0x10, R12 - LEAL 1(DI), R14 - LEAL 2(DI), R15 - MOVQ -2(DX)(CX*1), R9 + LEAQ 1(DI), DI + LEAQ -2(CX), R9 + MOVQ (DX)(DI*1), R10 + MOVQ 1(DX)(DI*1), R11 + MOVQ (DX)(R9*1), R12 + MOVQ 1(DX)(R9*1), R13 SHLQ $0x08, R10 IMULQ SI, R10 - SHRQ $0x30, R10 - SHLQ $0x08, R13 - IMULQ SI, R13 - SHRQ $0x30, R13 + SHRQ $0x2f, R10 SHLQ $0x20, R11 IMULQ R8, R11 SHRQ $0x32, R11 - SHLQ $0x20, R12 - IMULQ R8, R12 - SHRQ $0x32, R12 + SHLQ $0x08, R12 + IMULQ SI, R12 + SHRQ $0x2f, R12 + SHLQ $0x20, R13 + IMULQ R8, R13 + SHRQ $0x32, R13 + LEAQ 1(DI), R8 + LEAQ 1(R9), R14 MOVL DI, 24(SP)(R10*4) - MOVL R14, 24(SP)(R13*4) - MOVL R14, 262168(SP)(R11*4) - MOVL R15, 262168(SP)(R12*4) - MOVQ R9, R10 - MOVQ R9, R11 - SHRQ $0x08, R11 - MOVQ R11, R13 - LEAL -2(CX), R9 - LEAL -1(CX), DI + MOVL R9, 24(SP)(R12*4) + MOVL R8, 524312(SP)(R11*4) + MOVL R14, 524312(SP)(R13*4) + ADDQ $0x01, DI + SUBQ $0x01, R9 + +index_loop_encodeSnappyBetterBlockAsm: + CMPQ DI, R9 + JAE search_loop_encodeSnappyBetterBlockAsm + MOVQ (DX)(DI*1), R8 + MOVQ (DX)(R9*1), R10 + SHLQ $0x08, R8 + IMULQ SI, R8 + SHRQ $0x2f, R8 SHLQ $0x08, R10 IMULQ SI, R10 - SHRQ $0x30, R10 - SHLQ $0x20, R11 - IMULQ R8, R11 - SHRQ $0x32, R11 - SHLQ $0x08, R13 - IMULQ SI, R13 - SHRQ $0x30, R13 + SHRQ $0x2f, R10 + MOVL DI, 24(SP)(R8*4) MOVL R9, 24(SP)(R10*4) - MOVL DI, 262168(SP)(R11*4) - MOVL DI, 24(SP)(R13*4) - JMP search_loop_encodeSnappyBetterBlockAsm + ADDQ $0x02, DI + SUBQ $0x02, R9 + JMP index_loop_encodeSnappyBetterBlockAsm emit_remainder_encodeSnappyBetterBlockAsm: MOVQ src_len+32(FP), CX @@ -14964,12 +15005,22 @@ search_loop_encodeSnappyBetterBlockAsm64K: MOVL 262168(SP)(R11*4), R8 MOVL CX, 24(SP)(R10*4) MOVL CX, 262168(SP)(R11*4) - CMPL (DX)(SI*1), DI + MOVQ (DX)(SI*1), R10 + MOVQ (DX)(R8*1), R11 + CMPQ R10, DI JEQ candidate_match_encodeSnappyBetterBlockAsm64K - CMPL (DX)(R8*1), DI - JEQ candidateS_match_encodeSnappyBetterBlockAsm64K - MOVL 20(SP), CX - JMP search_loop_encodeSnappyBetterBlockAsm64K + CMPQ R11, DI + JNE no_short_found_encodeSnappyBetterBlockAsm64K + MOVL R8, SI + JMP candidate_match_encodeSnappyBetterBlockAsm64K + +no_short_found_encodeSnappyBetterBlockAsm64K: + CMPL R10, DI + JEQ candidate_match_encodeSnappyBetterBlockAsm64K + CMPL R11, DI + JEQ candidateS_match_encodeSnappyBetterBlockAsm64K + MOVL 20(SP), CX + JMP search_loop_encodeSnappyBetterBlockAsm64K candidateS_match_encodeSnappyBetterBlockAsm64K: SHRQ $0x08, DI @@ -15248,52 +15299,49 @@ match_nolit_emitcopy_end_encodeSnappyBetterBlockAsm64K: match_nolit_dst_ok_encodeSnappyBetterBlockAsm64K: MOVQ $0x00cf1bbcdcbfa563, SI MOVQ $0x9e3779b1, R8 - INCL DI - MOVQ (DX)(DI*1), R9 - MOVQ R9, R10 - MOVQ R9, R11 - MOVQ R9, R12 - SHRQ $0x08, R11 - MOVQ R11, R13 - SHRQ $0x10, R12 - LEAL 1(DI), R14 - LEAL 2(DI), R15 - MOVQ -2(DX)(CX*1), R9 + LEAQ 1(DI), DI + LEAQ -2(CX), R9 + MOVQ (DX)(DI*1), R10 + MOVQ 1(DX)(DI*1), R11 + MOVQ (DX)(R9*1), R12 + MOVQ 1(DX)(R9*1), R13 SHLQ $0x08, R10 IMULQ SI, R10 SHRQ $0x30, R10 - SHLQ $0x08, R13 - IMULQ SI, R13 - SHRQ $0x30, R13 SHLQ $0x20, R11 IMULQ R8, R11 SHRQ $0x32, R11 - SHLQ $0x20, R12 - IMULQ R8, R12 - SHRQ $0x32, R12 + SHLQ $0x08, R12 + IMULQ SI, R12 + SHRQ $0x30, R12 + SHLQ $0x20, R13 + IMULQ R8, R13 + SHRQ $0x32, R13 + LEAQ 1(DI), R8 + LEAQ 1(R9), R14 MOVL DI, 24(SP)(R10*4) - MOVL R14, 24(SP)(R13*4) - MOVL R14, 262168(SP)(R11*4) - MOVL R15, 262168(SP)(R12*4) - MOVQ R9, R10 - MOVQ R9, R11 - SHRQ $0x08, R11 - MOVQ R11, R13 - LEAL -2(CX), R9 - LEAL -1(CX), DI + MOVL R9, 24(SP)(R12*4) + MOVL R8, 262168(SP)(R11*4) + MOVL R14, 262168(SP)(R13*4) + ADDQ $0x01, DI + SUBQ $0x01, R9 + +index_loop_encodeSnappyBetterBlockAsm64K: + CMPQ DI, R9 + JAE search_loop_encodeSnappyBetterBlockAsm64K + MOVQ (DX)(DI*1), R8 + MOVQ (DX)(R9*1), R10 + SHLQ $0x08, R8 + IMULQ SI, R8 + SHRQ $0x30, R8 SHLQ $0x08, R10 IMULQ SI, R10 SHRQ $0x30, R10 - SHLQ $0x20, R11 - IMULQ R8, R11 - SHRQ $0x32, R11 - SHLQ $0x08, R13 - IMULQ SI, R13 - SHRQ $0x30, R13 + MOVL DI, 24(SP)(R8*4) MOVL R9, 24(SP)(R10*4) - MOVL DI, 262168(SP)(R11*4) - MOVL DI, 24(SP)(R13*4) - JMP search_loop_encodeSnappyBetterBlockAsm64K + ADDQ $0x02, DI + SUBQ $0x02, R9 + JMP index_loop_encodeSnappyBetterBlockAsm64K emit_remainder_encodeSnappyBetterBlockAsm64K: MOVQ src_len+32(FP), CX @@ -15508,12 +15556,22 @@ search_loop_encodeSnappyBetterBlockAsm12B: MOVL 65560(SP)(R11*4), R8 MOVL CX, 24(SP)(R10*4) MOVL CX, 65560(SP)(R11*4) - CMPL (DX)(SI*1), DI + MOVQ (DX)(SI*1), R10 + MOVQ (DX)(R8*1), R11 + CMPQ R10, DI JEQ candidate_match_encodeSnappyBetterBlockAsm12B - CMPL (DX)(R8*1), DI - JEQ candidateS_match_encodeSnappyBetterBlockAsm12B - MOVL 20(SP), CX - JMP search_loop_encodeSnappyBetterBlockAsm12B + CMPQ R11, DI + JNE no_short_found_encodeSnappyBetterBlockAsm12B + MOVL R8, SI + JMP candidate_match_encodeSnappyBetterBlockAsm12B + +no_short_found_encodeSnappyBetterBlockAsm12B: + CMPL R10, DI + JEQ candidate_match_encodeSnappyBetterBlockAsm12B + CMPL R11, DI + JEQ candidateS_match_encodeSnappyBetterBlockAsm12B + MOVL 20(SP), CX + JMP search_loop_encodeSnappyBetterBlockAsm12B candidateS_match_encodeSnappyBetterBlockAsm12B: SHRQ $0x08, DI @@ -15792,52 +15850,49 @@ match_nolit_emitcopy_end_encodeSnappyBetterBlockAsm12B: match_nolit_dst_ok_encodeSnappyBetterBlockAsm12B: MOVQ $0x0000cf1bbcdcbf9b, SI MOVQ $0x9e3779b1, R8 - INCL DI - MOVQ (DX)(DI*1), R9 - MOVQ R9, R10 - MOVQ R9, R11 - MOVQ R9, R12 - SHRQ $0x08, R11 - MOVQ R11, R13 - SHRQ $0x10, R12 - LEAL 1(DI), R14 - LEAL 2(DI), R15 - MOVQ -2(DX)(CX*1), R9 + LEAQ 1(DI), DI + LEAQ -2(CX), R9 + MOVQ (DX)(DI*1), R10 + MOVQ 1(DX)(DI*1), R11 + MOVQ (DX)(R9*1), R12 + MOVQ 1(DX)(R9*1), R13 SHLQ $0x10, R10 IMULQ SI, R10 SHRQ $0x32, R10 - SHLQ $0x10, R13 - IMULQ SI, R13 - SHRQ $0x32, R13 SHLQ $0x20, R11 IMULQ R8, R11 SHRQ $0x34, R11 - SHLQ $0x20, R12 - IMULQ R8, R12 - SHRQ $0x34, R12 + SHLQ $0x10, R12 + IMULQ SI, R12 + SHRQ $0x32, R12 + SHLQ $0x20, R13 + IMULQ R8, R13 + SHRQ $0x34, R13 + LEAQ 1(DI), R8 + LEAQ 1(R9), R14 MOVL DI, 24(SP)(R10*4) - MOVL R14, 24(SP)(R13*4) - MOVL R14, 65560(SP)(R11*4) - MOVL R15, 65560(SP)(R12*4) - MOVQ R9, R10 - MOVQ R9, R11 - SHRQ $0x08, R11 - MOVQ R11, R13 - LEAL -2(CX), R9 - LEAL -1(CX), DI + MOVL R9, 24(SP)(R12*4) + MOVL R8, 65560(SP)(R11*4) + MOVL R14, 65560(SP)(R13*4) + ADDQ $0x01, DI + SUBQ $0x01, R9 + +index_loop_encodeSnappyBetterBlockAsm12B: + CMPQ DI, R9 + JAE search_loop_encodeSnappyBetterBlockAsm12B + MOVQ (DX)(DI*1), R8 + MOVQ (DX)(R9*1), R10 + SHLQ $0x10, R8 + IMULQ SI, R8 + SHRQ $0x32, R8 SHLQ $0x10, R10 IMULQ SI, R10 SHRQ $0x32, R10 - SHLQ $0x20, R11 - IMULQ R8, R11 - SHRQ $0x34, R11 - SHLQ $0x10, R13 - IMULQ SI, R13 - SHRQ $0x32, R13 + MOVL DI, 24(SP)(R8*4) MOVL R9, 24(SP)(R10*4) - MOVL DI, 65560(SP)(R11*4) - MOVL DI, 24(SP)(R13*4) - JMP search_loop_encodeSnappyBetterBlockAsm12B + ADDQ $0x02, DI + SUBQ $0x02, R9 + JMP index_loop_encodeSnappyBetterBlockAsm12B emit_remainder_encodeSnappyBetterBlockAsm12B: MOVQ src_len+32(FP), CX @@ -16052,12 +16107,22 @@ search_loop_encodeSnappyBetterBlockAsm10B: MOVL 16408(SP)(R11*4), R8 MOVL CX, 24(SP)(R10*4) MOVL CX, 16408(SP)(R11*4) - CMPL (DX)(SI*1), DI + MOVQ (DX)(SI*1), R10 + MOVQ (DX)(R8*1), R11 + CMPQ R10, DI JEQ candidate_match_encodeSnappyBetterBlockAsm10B - CMPL (DX)(R8*1), DI - JEQ candidateS_match_encodeSnappyBetterBlockAsm10B - MOVL 20(SP), CX - JMP search_loop_encodeSnappyBetterBlockAsm10B + CMPQ R11, DI + JNE no_short_found_encodeSnappyBetterBlockAsm10B + MOVL R8, SI + JMP candidate_match_encodeSnappyBetterBlockAsm10B + +no_short_found_encodeSnappyBetterBlockAsm10B: + CMPL R10, DI + JEQ candidate_match_encodeSnappyBetterBlockAsm10B + CMPL R11, DI + JEQ candidateS_match_encodeSnappyBetterBlockAsm10B + MOVL 20(SP), CX + JMP search_loop_encodeSnappyBetterBlockAsm10B candidateS_match_encodeSnappyBetterBlockAsm10B: SHRQ $0x08, DI @@ -16336,52 +16401,49 @@ match_nolit_emitcopy_end_encodeSnappyBetterBlockAsm10B: match_nolit_dst_ok_encodeSnappyBetterBlockAsm10B: MOVQ $0x0000cf1bbcdcbf9b, SI MOVQ $0x9e3779b1, R8 - INCL DI - MOVQ (DX)(DI*1), R9 - MOVQ R9, R10 - MOVQ R9, R11 - MOVQ R9, R12 - SHRQ $0x08, R11 - MOVQ R11, R13 - SHRQ $0x10, R12 - LEAL 1(DI), R14 - LEAL 2(DI), R15 - MOVQ -2(DX)(CX*1), R9 + LEAQ 1(DI), DI + LEAQ -2(CX), R9 + MOVQ (DX)(DI*1), R10 + MOVQ 1(DX)(DI*1), R11 + MOVQ (DX)(R9*1), R12 + MOVQ 1(DX)(R9*1), R13 SHLQ $0x10, R10 IMULQ SI, R10 SHRQ $0x34, R10 - SHLQ $0x10, R13 - IMULQ SI, R13 - SHRQ $0x34, R13 SHLQ $0x20, R11 IMULQ R8, R11 SHRQ $0x36, R11 - SHLQ $0x20, R12 - IMULQ R8, R12 - SHRQ $0x36, R12 + SHLQ $0x10, R12 + IMULQ SI, R12 + SHRQ $0x34, R12 + SHLQ $0x20, R13 + IMULQ R8, R13 + SHRQ $0x36, R13 + LEAQ 1(DI), R8 + LEAQ 1(R9), R14 MOVL DI, 24(SP)(R10*4) - MOVL R14, 24(SP)(R13*4) - MOVL R14, 16408(SP)(R11*4) - MOVL R15, 16408(SP)(R12*4) - MOVQ R9, R10 - MOVQ R9, R11 - SHRQ $0x08, R11 - MOVQ R11, R13 - LEAL -2(CX), R9 - LEAL -1(CX), DI + MOVL R9, 24(SP)(R12*4) + MOVL R8, 16408(SP)(R11*4) + MOVL R14, 16408(SP)(R13*4) + ADDQ $0x01, DI + SUBQ $0x01, R9 + +index_loop_encodeSnappyBetterBlockAsm10B: + CMPQ DI, R9 + JAE search_loop_encodeSnappyBetterBlockAsm10B + MOVQ (DX)(DI*1), R8 + MOVQ (DX)(R9*1), R10 + SHLQ $0x10, R8 + IMULQ SI, R8 + SHRQ $0x34, R8 SHLQ $0x10, R10 IMULQ SI, R10 SHRQ $0x34, R10 - SHLQ $0x20, R11 - IMULQ R8, R11 - SHRQ $0x36, R11 - SHLQ $0x10, R13 - IMULQ SI, R13 - SHRQ $0x34, R13 + MOVL DI, 24(SP)(R8*4) MOVL R9, 24(SP)(R10*4) - MOVL DI, 16408(SP)(R11*4) - MOVL DI, 24(SP)(R13*4) - JMP search_loop_encodeSnappyBetterBlockAsm10B + ADDQ $0x02, DI + SUBQ $0x02, R9 + JMP index_loop_encodeSnappyBetterBlockAsm10B emit_remainder_encodeSnappyBetterBlockAsm10B: MOVQ src_len+32(FP), CX @@ -16596,12 +16658,22 @@ search_loop_encodeSnappyBetterBlockAsm8B: MOVL 4120(SP)(R11*4), R8 MOVL CX, 24(SP)(R10*4) MOVL CX, 4120(SP)(R11*4) - CMPL (DX)(SI*1), DI + MOVQ (DX)(SI*1), R10 + MOVQ (DX)(R8*1), R11 + CMPQ R10, DI JEQ candidate_match_encodeSnappyBetterBlockAsm8B - CMPL (DX)(R8*1), DI - JEQ candidateS_match_encodeSnappyBetterBlockAsm8B - MOVL 20(SP), CX - JMP search_loop_encodeSnappyBetterBlockAsm8B + CMPQ R11, DI + JNE no_short_found_encodeSnappyBetterBlockAsm8B + MOVL R8, SI + JMP candidate_match_encodeSnappyBetterBlockAsm8B + +no_short_found_encodeSnappyBetterBlockAsm8B: + CMPL R10, DI + JEQ candidate_match_encodeSnappyBetterBlockAsm8B + CMPL R11, DI + JEQ candidateS_match_encodeSnappyBetterBlockAsm8B + MOVL 20(SP), CX + JMP search_loop_encodeSnappyBetterBlockAsm8B candidateS_match_encodeSnappyBetterBlockAsm8B: SHRQ $0x08, DI @@ -16878,52 +16950,49 @@ match_nolit_emitcopy_end_encodeSnappyBetterBlockAsm8B: match_nolit_dst_ok_encodeSnappyBetterBlockAsm8B: MOVQ $0x0000cf1bbcdcbf9b, SI MOVQ $0x9e3779b1, R8 - INCL DI - MOVQ (DX)(DI*1), R9 - MOVQ R9, R10 - MOVQ R9, R11 - MOVQ R9, R12 - SHRQ $0x08, R11 - MOVQ R11, R13 - SHRQ $0x10, R12 - LEAL 1(DI), R14 - LEAL 2(DI), R15 - MOVQ -2(DX)(CX*1), R9 + LEAQ 1(DI), DI + LEAQ -2(CX), R9 + MOVQ (DX)(DI*1), R10 + MOVQ 1(DX)(DI*1), R11 + MOVQ (DX)(R9*1), R12 + MOVQ 1(DX)(R9*1), R13 SHLQ $0x10, R10 IMULQ SI, R10 SHRQ $0x36, R10 - SHLQ $0x10, R13 - IMULQ SI, R13 - SHRQ $0x36, R13 SHLQ $0x20, R11 IMULQ R8, R11 SHRQ $0x38, R11 - SHLQ $0x20, R12 - IMULQ R8, R12 - SHRQ $0x38, R12 + SHLQ $0x10, R12 + IMULQ SI, R12 + SHRQ $0x36, R12 + SHLQ $0x20, R13 + IMULQ R8, R13 + SHRQ $0x38, R13 + LEAQ 1(DI), R8 + LEAQ 1(R9), R14 MOVL DI, 24(SP)(R10*4) - MOVL R14, 24(SP)(R13*4) - MOVL R14, 4120(SP)(R11*4) - MOVL R15, 4120(SP)(R12*4) - MOVQ R9, R10 - MOVQ R9, R11 - SHRQ $0x08, R11 - MOVQ R11, R13 - LEAL -2(CX), R9 - LEAL -1(CX), DI + MOVL R9, 24(SP)(R12*4) + MOVL R8, 4120(SP)(R11*4) + MOVL R14, 4120(SP)(R13*4) + ADDQ $0x01, DI + SUBQ $0x01, R9 + +index_loop_encodeSnappyBetterBlockAsm8B: + CMPQ DI, R9 + JAE search_loop_encodeSnappyBetterBlockAsm8B + MOVQ (DX)(DI*1), R8 + MOVQ (DX)(R9*1), R10 + SHLQ $0x10, R8 + IMULQ SI, R8 + SHRQ $0x36, R8 SHLQ $0x10, R10 IMULQ SI, R10 SHRQ $0x36, R10 - SHLQ $0x20, R11 - IMULQ R8, R11 - SHRQ $0x38, R11 - SHLQ $0x10, R13 - IMULQ SI, R13 - SHRQ $0x36, R13 + MOVL DI, 24(SP)(R8*4) MOVL R9, 24(SP)(R10*4) - MOVL DI, 4120(SP)(R11*4) - MOVL DI, 24(SP)(R13*4) - JMP search_loop_encodeSnappyBetterBlockAsm8B + ADDQ $0x02, DI + SUBQ $0x02, R9 + JMP index_loop_encodeSnappyBetterBlockAsm8B emit_remainder_encodeSnappyBetterBlockAsm8B: MOVQ src_len+32(FP), CX |