)]}' { "commit": "2dbfe9001fa8e8b979ff80a1c88b888d48e1b575", "tree": "de45872e0734c93f0e6f2668799b274cca52272e", "parents": [ "bd318be0434a57050ed475e0f45c3dbb16c09c2e" ], "author": { "name": "Filippo Valsorda", "email": "filippo@golang.org", "time": "Sat Mar 23 01:45:20 2019 -0400" }, "committer": { "name": "Filippo Valsorda", "email": "filippo@golang.org", "time": "Mon Nov 11 21:33:42 2019 +0000" }, "message": "poly1305: rewrite the Go implementation with 64-bit limbs\n\nThe new code is meant to be readable without external references for\nPoly1305, and explains the field logic. The generic code is now 30-50%\nfaster on a Intel(R) Xeon(R) CPU E5-2690 v3 @ 2.60GHz, and even better\non a 3.1 GHz i7 MacBook.\n\nname old time/op new time/op delta\n64-48 126ns ± 0% 80ns ± 1% -36.24% (p\u003d0.000 n\u003d16+20)\n1K-48 1.07µs ± 0% 0.81µs ± 2% -23.63% (p\u003d0.000 n\u003d19+20)\n2M-48 2.07ms ± 0% 1.61ms ± 1% -22.31% (p\u003d0.000 n\u003d20+20)\nWrite64-48 79.3ns ± 0% 58.0ns ± 1% -26.89% (p\u003d0.000 n\u003d20+19)\nWrite1K-48 1.02µs ± 0% 0.79µs ± 1% -22.91% (p\u003d0.000 n\u003d19+19)\nWrite2M-48 2.07ms ± 0% 1.61ms ± 2% -22.33% (p\u003d0.000 n\u003d17+20)\n\nname old speed new speed delta\n64-48 508MB/s ± 0% 797MB/s ± 1% +56.95% (p\u003d0.000 n\u003d16+20)\n1K-48 960MB/s ± 0% 1257MB/s ± 2% +30.94% (p\u003d0.000 n\u003d18+20)\n2M-48 1.01GB/s ± 0% 1.30GB/s ± 1% +28.73% (p\u003d0.000 n\u003d20+20)\nWrite64-48 807MB/s ± 0% 1104MB/s ± 1% +36.78% (p\u003d0.000 n\u003d18+19)\nWrite1K-48 1.00GB/s ± 0% 1.30GB/s ± 1% +29.71% (p\u003d0.000 n\u003d18+19)\nWrite2M-48 1.01GB/s ± 0% 1.31GB/s ± 2% +28.77% (p\u003d0.000 n\u003d17+20)\n\nThe assembly is still 50-90% faster on the Xeon, 30-60% on the MacBook.\nThe Go code does not use all the arithmetic tricks the assembly does,\nand it does not have access to the three operand wide shift instruction.\n\nname old time/op new time/op delta\n64-48 80.3ns ± 1% 54.2ns ± 0% -32.50% (p\u003d0.000 n\u003d20+17)\n1K-48 815ns ± 2% 446ns ± 1% -45.27% (p\u003d0.000 n\u003d20+20)\n2M-48 1.61ms ± 1% 0.86ms ± 0% -46.54% (p\u003d0.000 n\u003d20+17)\nWrite64-48 58.0ns ± 1% 34.0ns ± 0% -41.34% (p\u003d0.000 n\u003d19+20)\nWrite1K-48 790ns ± 1% 427ns ± 0% -45.92% (p\u003d0.000 n\u003d19+17)\nWrite2M-48 1.61ms ± 2% 0.86ms ± 0% -46.51% (p\u003d0.000 n\u003d20+20)\n\nname old speed new speed delta\n64-48 797MB/s ± 1% 1180MB/s ± 0% +48.09% (p\u003d0.000 n\u003d20+19)\n1K-48 1.26GB/s ± 2% 2.30GB/s ± 1% +82.71% (p\u003d0.000 n\u003d20+20)\n2M-48 1.30GB/s ± 1% 2.44GB/s ± 0% +87.04% (p\u003d0.000 n\u003d20+17)\nWrite64-48 1.10GB/s ± 1% 1.88GB/s ± 0% +70.52% (p\u003d0.000 n\u003d19+18)\nWrite1K-48 1.30GB/s ± 1% 2.40GB/s ± 0% +84.84% (p\u003d0.000 n\u003d19+18)\nWrite2M-48 1.31GB/s ± 2% 2.44GB/s ± 0% +86.93% (p\u003d0.000 n\u003d20+20)\n\nHopefully this will also avoid the need for an arm64 implementation.\n\nSince now the Go and the amd64/ppc64le assembly use the same limb\nschedule, drop the assembly initialize and finalize implementations,\nand make the wrapper code match. It comes with a minor slowdown.\n\nname old time/op new time/op delta\n64-48 50.3ns ± 0% 54.2ns ± 0% +7.73% (p\u003d0.000 n\u003d20+17)\n1K-48 441ns ± 0% 446ns ± 1% +1.10% (p\u003d0.000 n\u003d19+20)\n2M-48 860µs ± 0% 859µs ± 0% ~ (p\u003d0.178 n\u003d19+17)\nWrite64-48 34.0ns ± 0% 34.0ns ± 0% ~ (all equal)\nWrite1K-48 424ns ± 0% 427ns ± 0% +0.71% (p\u003d0.000 n\u003d17+17)\nWrite2M-48 860µs ± 0% 859µs ± 0% -0.04% (p\u003d0.000 n\u003d19+20)\n\nname old speed new speed delta\n64-48 1.27GB/s ± 0% 1.18GB/s ± 0% -7.20% (p\u003d0.000 n\u003d20+19)\n1K-48 2.32GB/s ± 0% 2.30GB/s ± 1% -1.07% (p\u003d0.000 n\u003d18+20)\n2M-48 2.44GB/s ± 0% 2.44GB/s ± 0% ~ (p\u003d0.173 n\u003d19+17)\nWrite64-48 1.88GB/s ± 0% 1.88GB/s ± 0% +0.04% (p\u003d0.000 n\u003d19+18)\nWrite1K-48 2.41GB/s ± 0% 2.40GB/s ± 0% -0.67% (p\u003d0.000 n\u003d19+18)\nWrite2M-48 2.44GB/s ± 0% 2.44GB/s ± 0% +0.04% (p\u003d0.000 n\u003d19+20)\n\nSince poly1305/sum_generic.go was almost entirely rewritten, it\u0027s\nprobably best reviewed on gitiles.\n\nThis is the implementation published at\nhttps://blog.filippo.io/a-literate-go-implementation-of-poly1305/\n\nUpdates #31470\n\nChange-Id: I74f9011d3ee317a43b05ae7f05d96081d08bffd3\nReviewed-on: https://go-review.googlesource.com/c/crypto/+/169037\nReviewed-by: Katie Hockman \u003ckatie@golang.org\u003e\n", "tree_diff": [ { "type": "add", "old_id": "0000000000000000000000000000000000000000", "old_mode": 0, "old_path": "/dev/null", "new_id": "157a69f61bd9a8c4344a944ab4dbbd1857452718", "new_mode": 33188, "new_path": "poly1305/bits_compat.go" }, { "type": "add", "old_id": "0000000000000000000000000000000000000000", "old_mode": 0, "old_path": "/dev/null", "new_id": "a0a185f0fc77187616eda36f23d69a94fa0af0f2", "new_mode": 33188, "new_path": "poly1305/bits_go1.13.go" }, { "type": "modify", "old_id": "d076a562351f4f01093c42749ca48c2442a9c133", "old_mode": 33188, "old_path": "poly1305/poly1305.go", "new_id": "066159b797dd8cdfd30ee9a1ec8bed8a41a6e704", "new_mode": 33188, "new_path": "poly1305/poly1305.go" }, { "type": "modify", "old_id": "e35928a0b6b6fd0e133de9c3dcaa4710d5032285", "old_mode": 33188, "old_path": "poly1305/poly1305_test.go", "new_id": "b258eed67948e37fb0dbe7b5f76a7f3ee4554bac", "new_mode": 33188, "new_path": "poly1305/poly1305_test.go" }, { "type": "modify", "old_id": "2dbf42aa537a5910eb40d0b6e767d7dc3121d5d8", "old_mode": 33188, "old_path": "poly1305/sum_amd64.go", "new_id": "df56a652ff0854876ad7671abb4f45d0f8b325fa", "new_mode": 33188, "new_path": "poly1305/sum_amd64.go" }, { "type": "modify", "old_id": "7d600f13cc87a51f57d75364fd1dae5d05aa73c6", "old_mode": 33188, "old_path": "poly1305/sum_amd64.s", "new_id": "8c0cefbb3cb7a959573cc0352667a4a31d02ec06", "new_mode": 33188, "new_path": "poly1305/sum_amd64.s" }, { "type": "modify", "old_id": "5dc321c2f39e67863917e0cbe6b7b76aa5361669", "old_mode": 33188, "old_path": "poly1305/sum_arm.go", "new_id": "6e695e4272e45c0d459fc64ddfb2ea897cab2a74", "new_mode": 33188, "new_path": "poly1305/sum_arm.go" }, { "type": "modify", "old_id": "bab76ef0d83b4d187bf2f22860aa71470898a6b3", "old_mode": 33188, "old_path": "poly1305/sum_generic.go", "new_id": "1187eab78fd449cb50da7d378bc9de6e098fef6b", "new_mode": 33188, "new_path": "poly1305/sum_generic.go" }, { "type": "modify", "old_id": "8a9c2070b9f2f63f4444b298ab5779f8962323dd", "old_mode": 33188, "old_path": "poly1305/sum_noasm.go", "new_id": "1682eda45f174c81dd91815f53fb0121a9154e49", "new_mode": 33188, "new_path": "poly1305/sum_noasm.go" }, { "type": "modify", "old_id": "2402b6371bfcc52851b7ae2dbd0399b03944f586", "old_mode": 33188, "old_path": "poly1305/sum_ppc64le.go", "new_id": "3233616935bd8f35369771f97cd61d17eb825e86", "new_mode": 33188, "new_path": "poly1305/sum_ppc64le.go" }, { "type": "modify", "old_id": "55c7167ec9813ee569f0568e2261316f2917a29b", "old_mode": 33188, "old_path": "poly1305/sum_ppc64le.s", "new_id": "4e20bf299a5ea11ac89f520ae2d896268d2b686e", "new_mode": 33188, "new_path": "poly1305/sum_ppc64le.s" }, { "type": "modify", "old_id": "ec99e07e9fb4963224bbb6f985c82357923b4ab9", "old_mode": 33188, "old_path": "poly1305/sum_s390x.go", "new_id": "a8920ee9d21d9cee5f3f5e26ff9b32f979d514f3", "new_mode": 33188, "new_path": "poly1305/sum_s390x.go" } ] }