)]}' { "commit": "ae7b6a3b779c4d6de96f59efbfed0b899c3ff6df", "tree": "6c1195a807d207f75bcf5b201fde66e785af127e", "parents": [ "55cf84b0778c382967b46f3f2ffd402fba005a26" ], "author": { "name": "Xiangdong Ji", "email": "xiangdong.ji@arm.com", "time": "Sat Feb 22 08:12:54 2020 +0000" }, "committer": { "name": "Cherry Zhang", "email": "cherryyz@google.com", "time": "Fri Aug 28 16:40:41 2020 +0000" }, "message": "math/big: tune addVW/subVW performance on arm64\n\nAdd an optimization for addVW and subVW over large-sized vectors, it switches\nfrom add/sub with carry to copy the rest of the vector when we are done with\ncarries. Consistent performance improvement are observed on various arm64\nmachines.\n\nAdd additional tests and benchmarks to increase the test coverage.\nTestFunVWExt:\n Testing with various types of input vector, using the result from go-version\n addVW/subVW as golden reference.\nBenchmarkAddVWext and BenchmarkSubVWext:\n Benchmarking using input vector having all 1s or all 0s, for evaluating the\n overhead of worst case.\n\n1. Perf. comparison over randomly generated input vectors:\n\nServer 1:\nname old time/op new time/op delta\nAddVW/1 12.3ns ± 3% 12.0ns ± 0% -2.60% (p\u003d0.001 n\u003d10+8)\nAddVW/2 12.5ns ± 2% 12.3ns ± 0% -1.84% (p\u003d0.001 n\u003d10+8)\nAddVW/3 12.6ns ± 2% 12.3ns ± 0% -1.91% (p\u003d0.009 n\u003d10+10)\nAddVW/4 13.1ns ± 3% 12.7ns ± 0% -2.98% (p\u003d0.006 n\u003d10+8)\nAddVW/5 14.4ns ± 1% 13.9ns ± 0% -3.81% (p\u003d0.000 n\u003d10+10)\nAddVW/10 11.7ns ± 0% 11.7ns ± 0% ~ (all equal)\nAddVW/100 47.8ns ± 0% 29.9ns ± 2% -37.38% (p\u003d0.000 n\u003d10+9)\nAddVW/1000 446ns ± 0% 207ns ± 0% -53.59% (p\u003d0.000 n\u003d10+10)\nAddVW/10000 4.35µs ± 1% 2.92µs ± 0% -32.85% (p\u003d0.000 n\u003d10+10)\nAddVW/100000 43.6µs ± 0% 29.7µs ± 0% -31.92% (p\u003d0.000 n\u003d8+10)\nSubVW/1 12.6ns ± 0% 12.3ns ± 2% -2.22% (p\u003d0.000 n\u003d7+10)\nSubVW/2 12.7ns ± 0% 12.6ns ± 1% -0.39% (p\u003d0.046 n\u003d8+10)\nSubVW/3 12.7ns ± 1% 12.6ns ± 1% ~ (p\u003d0.410 n\u003d10+10)\nSubVW/4 13.3ns ± 3% 13.1ns ± 3% ~ (p\u003d0.072 n\u003d10+10)\nSubVW/5 14.2ns ± 0% 14.1ns ± 1% -0.63% (p\u003d0.046 n\u003d8+10)\nSubVW/10 11.7ns ± 0% 11.7ns ± 0% ~ (all equal)\nSubVW/100 47.8ns ± 0% 33.1ns ±19% -30.71% (p\u003d0.000 n\u003d10+10)\nSubVW/1000 446ns ± 0% 207ns ± 0% -53.59% (p\u003d0.000 n\u003d10+10)\nSubVW/10000 4.33µs ± 1% 2.92µs ± 0% -32.66% (p\u003d0.000 n\u003d10+6)\nSubVW/100000 43.4µs ± 0% 29.6µs ± 0% -31.90% (p\u003d0.000 n\u003d10+9)\n\nServer 2:\nname old time/op new time/op delta\nAddVW/1 5.49ns ± 0% 5.53ns ± 2% ~ (p\u003d1.000 n\u003d9+10)\nAddVW/2 5.96ns ± 2% 5.92ns ± 1% -0.69% (p\u003d0.039 n\u003d10+10)\nAddVW/3 6.72ns ± 0% 6.73ns ± 0% ~ (p\u003d0.078 n\u003d10+10)\nAddVW/4 7.07ns ± 0% 6.75ns ± 2% -4.55% (p\u003d0.000 n\u003d10+10)\nAddVW/5 8.14ns ± 0% 8.17ns ± 0% +0.46% (p\u003d0.003 n\u003d8+8)\nAddVW/10 10.0ns ± 0% 10.1ns ± 1% +0.70% (p\u003d0.003 n\u003d10+10)\nAddVW/100 43.0ns ± 0% 33.5ns ± 0% -22.09% (p\u003d0.000 n\u003d9+9)\nAddVW/1000 394ns ± 0% 278ns ± 0% -29.44% (p\u003d0.000 n\u003d10+10)\nAddVW/10000 4.18µs ± 0% 3.14µs ± 0% -24.81% (p\u003d0.000 n\u003d8+8)\nAddVW/100000 68.3µs ± 3% 62.1µs ± 5% -9.13% (p\u003d0.000 n\u003d10+10)\nSubVW/1 5.37ns ± 2% 5.42ns ± 1% ~ (p\u003d0.990 n\u003d10+10)\nSubVW/2 5.89ns ± 0% 5.92ns ± 1% +0.58% (p\u003d0.000 n\u003d8+10)\nSubVW/3 6.64ns ± 1% 6.82ns ± 3% +2.63% (p\u003d0.000 n\u003d9+10)\nSubVW/4 7.17ns ± 0% 6.69ns ± 2% -6.74% (p\u003d0.000 n\u003d10+9)\nSubVW/5 8.22ns ± 0% 8.18ns ± 0% -0.46% (p\u003d0.001 n\u003d8+9)\nSubVW/10 10.0ns ± 1% 10.1ns ± 1% ~ (p\u003d0.341 n\u003d10+10)\nSubVW/100 43.0ns ± 0% 33.5ns ± 0% -22.09% (p\u003d0.000 n\u003d7+10)\nSubVW/1000 394ns ± 0% 278ns ± 0% -29.44% (p\u003d0.000 n\u003d10+10)\nSubVW/10000 4.18µs ± 0% 3.15µs ± 0% -24.62% (p\u003d0.000 n\u003d9+9)\nSubVW/100000 67.7µs ± 4% 62.4µs ± 2% -7.92% (p\u003d0.000 n\u003d10+10)\n\n2. Perf. comparison over input vectors of all 1s or all 0s\n\nServer 1:\nname old time/op new time/op delta\nAddVWext/1 12.6ns ± 0% 12.0ns ± 0% -4.76% (p\u003d0.000 n\u003d6+10)\nAddVWext/2 12.7ns ± 0% 12.4ns ± 1% -2.52% (p\u003d0.000 n\u003d10+10)\nAddVWext/3 12.7ns ± 0% 12.4ns ± 0% -2.36% (p\u003d0.000 n\u003d9+7)\nAddVWext/4 13.2ns ± 4% 12.7ns ± 0% -3.71% (p\u003d0.001 n\u003d10+9)\nAddVWext/5 14.6ns ± 0% 13.9ns ± 0% -4.79% (p\u003d0.000 n\u003d10+8)\nAddVWext/10 11.7ns ± 0% 11.7ns ± 0% ~ (all equal)\nAddVWext/100 47.8ns ± 0% 47.4ns ± 0% -0.84% (p\u003d0.000 n\u003d10+10)\nAddVWext/1000 446ns ± 0% 399ns ± 0% -10.54% (p\u003d0.000 n\u003d10+10)\nAddVWext/10000 4.34µs ± 1% 3.90µs ± 0% -10.12% (p\u003d0.000 n\u003d10+10)\nAddVWext/100000 43.9µs ± 1% 39.4µs ± 0% -10.18% (p\u003d0.000 n\u003d10+10)\nSubVWext/1 12.6ns ± 0% 12.3ns ± 2% -2.70% (p\u003d0.000 n\u003d7+10)\nSubVWext/2 12.6ns ± 1% 12.6ns ± 2% ~ (p\u003d0.234 n\u003d10+10)\nSubVWext/3 12.7ns ± 0% 12.6ns ± 2% -0.71% (p\u003d0.033 n\u003d10+10)\nSubVWext/4 13.4ns ± 0% 13.1ns ± 3% -2.01% (p\u003d0.006 n\u003d8+10)\nSubVWext/5 14.2ns ± 0% 14.1ns ± 1% -0.85% (p\u003d0.003 n\u003d10+10)\nSubVWext/10 11.7ns ± 0% 11.7ns ± 0% ~ (all equal)\nSubVWext/100 47.8ns ± 0% 47.4ns ± 0% -0.84% (p\u003d0.000 n\u003d10+10)\nSubVWext/1000 446ns ± 0% 399ns ± 0% -10.54% (p\u003d0.000 n\u003d10+10)\nSubVWext/10000 4.33µs ± 1% 3.90µs ± 0% -10.02% (p\u003d0.000 n\u003d10+10)\nSubVWext/100000 43.5µs ± 0% 39.5µs ± 1% -9.16% (p\u003d0.000 n\u003d7+10)\n\nServer 2:\nname old time/op new time/op delta\nAddVWext/1 5.48ns ± 0% 5.43ns ± 1% -0.97% (p\u003d0.000 n\u003d9+9)\nAddVWext/2 5.99ns ± 2% 5.93ns ± 1% ~ (p\u003d0.054 n\u003d10+10)\nAddVWext/3 6.74ns ± 0% 6.79ns ± 1% +0.80% (p\u003d0.000 n\u003d9+10)\nAddVWext/4 7.18ns ± 0% 7.21ns ± 1% +0.36% (p\u003d0.034 n\u003d9+10)\nAddVWext/5 7.93ns ± 3% 8.18ns ± 0% +3.18% (p\u003d0.000 n\u003d10+8)\nAddVWext/10 10.0ns ± 0% 10.1ns ± 1% +0.60% (p\u003d0.011 n\u003d10+10)\nAddVWext/100 43.0ns ± 0% 47.7ns ± 0% +10.93% (p\u003d0.000 n\u003d9+10)\nAddVWext/1000 394ns ± 0% 399ns ± 0% +1.27% (p\u003d0.000 n\u003d10+10)\nAddVWext/10000 4.18µs ± 0% 4.50µs ± 0% +7.73% (p\u003d0.000 n\u003d9+10)\nAddVWext/100000 67.6µs ± 2% 68.4µs ± 3% ~ (p\u003d0.139 n\u003d9+8)\nSubVWext/1 5.46ns ± 1% 5.43ns ± 0% -0.55% (p\u003d0.002 n\u003d9+9)\nSubVWext/2 5.89ns ± 0% 5.93ns ± 1% +0.68% (p\u003d0.000 n\u003d8+10)\nSubVWext/3 6.72ns ± 1% 6.79ns ± 1% +1.07% (p\u003d0.000 n\u003d10+10)\nSubVWext/4 6.98ns ± 1% 7.21ns ± 0% +3.25% (p\u003d0.000 n\u003d10+10)\nSubVWext/5 8.22ns ± 0% 7.99ns ± 3% -2.83% (p\u003d0.000 n\u003d8+10)\nSubVWext/10 10.0ns ± 1% 10.1ns ± 1% ~ (p\u003d0.239 n\u003d10+10)\nSubVWext/100 43.0ns ± 0% 47.7ns ± 0% +10.93% (p\u003d0.000 n\u003d8+10)\nSubVWext/1000 394ns ± 0% 399ns ± 0% +1.27% (p\u003d0.000 n\u003d10+10)\nSubVWext/10000 4.18µs ± 0% 4.51µs ± 0% +7.86% (p\u003d0.000 n\u003d8+8)\nSubVWext/100000 68.3µs ± 2% 68.0µs ± 3% ~ (p\u003d0.515 n\u003d10+8)\n\nChange-Id: I134a5194b8a2deaaebbaa2b771baf72846971d58\nReviewed-on: https://go-review.googlesource.com/c/go/+/229739\nReviewed-by: Cherry Zhang \u003ccherryyz@google.com\u003e\nReviewed-by: Robert Griesemer \u003cgri@golang.org\u003e\nRun-TryBot: Cherry Zhang \u003ccherryyz@google.com\u003e\nTryBot-Result: Gobot Gobot \u003cgobot@golang.org\u003e\n", "tree_diff": [ { "type": "modify", "old_id": "18e513e2c38d5c7fa542be5355a35b777699a84e", "old_mode": 33188, "old_path": "src/math/big/arith_arm64.s", "new_id": "da6e408e19f726c76d3a27eb804849fcb2034947", "new_mode": 33188, "new_path": "src/math/big/arith_arm64.s" }, { "type": "modify", "old_id": "e2b982c89c256be3c57c9acb6350dae938868de0", "old_mode": 33188, "old_path": "src/math/big/arith_test.go", "new_id": "fc205934c5ca92f568c60a1f2a347e3746efaa33", "new_mode": 33188, "new_path": "src/math/big/arith_test.go" } ] }