mirror of
https://github.com/Cian-H/fastmath.git
synced 2025-12-22 22:22:02 +00:00
First dev commit
This commit is contained in:
Cian Hughes
4
.gitignore
vendored
Normal file
4
.gitignore
vendored
Normal file
@@ -0,0 +1,4 @@
|
||||
/target
|
||||
/.vscode
|
||||
/tmp
|
||||
*.ipynb
|
||||
709
Cargo.lock
generated
Normal file
709
Cargo.lock
generated
Normal file
@@ -0,0 +1,709 @@
|
||||
# This file is automatically @generated by Cargo.
|
||||
# It is not intended for manual editing.
|
||||
version = 3
|
||||
|
||||
[[package]]
|
||||
name = "aho-corasick"
|
||||
version = "1.0.2"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "43f6cb1bf222025340178f382c426f13757b2960e89779dfcb319c32542a5a41"
|
||||
dependencies = [
|
||||
"memchr",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "anes"
|
||||
version = "0.1.6"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "4b46cbb362ab8752921c97e041f5e366ee6297bd428a31275b9fcf1e380f7299"
|
||||
|
||||
[[package]]
|
||||
name = "anstyle"
|
||||
version = "1.0.1"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "3a30da5c5f2d5e72842e00bcb57657162cdabef0931f40e2deb9b4140440cecd"
|
||||
|
||||
[[package]]
|
||||
name = "autocfg"
|
||||
version = "1.1.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "d468802bab17cbc0cc575e9b053f41e72aa36bfa6b7f55e3529ffa43161b97fa"
|
||||
|
||||
[[package]]
|
||||
name = "bincode"
|
||||
version = "1.3.3"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "b1f45e9417d87227c7a56d22e471c6206462cba514c7590c09aff4cf6d1ddcad"
|
||||
dependencies = [
|
||||
"serde",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "bitflags"
|
||||
version = "2.3.3"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "630be753d4e58660abd17930c71b647fe46c27ea6b63cc59e1e3851406972e42"
|
||||
|
||||
[[package]]
|
||||
name = "bumpalo"
|
||||
version = "3.13.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "a3e2c3daef883ecc1b5d58c15adae93470a91d425f3532ba1695849656af3fc1"
|
||||
|
||||
[[package]]
|
||||
name = "cast"
|
||||
version = "0.3.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "37b2a672a2cb129a2e41c10b1224bb368f9f37a2b16b612598138befd7b37eb5"
|
||||
|
||||
[[package]]
|
||||
name = "cc"
|
||||
version = "1.0.79"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "50d30906286121d95be3d479533b458f87493b30a4b5f79a607db8f5d11aa91f"
|
||||
|
||||
[[package]]
|
||||
name = "cfg-if"
|
||||
version = "1.0.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "baf1de4339761588bc0619e3cbc0120ee582ebb74b53b4efbf79117bd2da40fd"
|
||||
|
||||
[[package]]
|
||||
name = "ciborium"
|
||||
version = "0.2.1"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "effd91f6c78e5a4ace8a5d3c0b6bfaec9e2baaef55f3efc00e45fb2e477ee926"
|
||||
dependencies = [
|
||||
"ciborium-io",
|
||||
"ciborium-ll",
|
||||
"serde",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "ciborium-io"
|
||||
version = "0.2.1"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "cdf919175532b369853f5d5e20b26b43112613fd6fe7aee757e35f7a44642656"
|
||||
|
||||
[[package]]
|
||||
name = "ciborium-ll"
|
||||
version = "0.2.1"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "defaa24ecc093c77630e6c15e17c51f5e187bf35ee514f4e2d67baaa96dae22b"
|
||||
dependencies = [
|
||||
"ciborium-io",
|
||||
"half",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "clap"
|
||||
version = "4.3.15"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "8f644d0dac522c8b05ddc39aaaccc5b136d5dc4ff216610c5641e3be5becf56c"
|
||||
dependencies = [
|
||||
"clap_builder",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "clap_builder"
|
||||
version = "4.3.15"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "af410122b9778e024f9e0fb35682cc09cc3f85cad5e8d3ba8f47a9702df6e73d"
|
||||
dependencies = [
|
||||
"anstyle",
|
||||
"clap_lex",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "clap_lex"
|
||||
version = "0.5.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "2da6da31387c7e4ef160ffab6d5e7f00c42626fe39aea70a7b0f1773f7dd6c1b"
|
||||
|
||||
[[package]]
|
||||
name = "criterion"
|
||||
version = "0.5.1"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "f2b12d017a929603d80db1831cd3a24082f8137ce19c69e6447f54f5fc8d692f"
|
||||
dependencies = [
|
||||
"anes",
|
||||
"cast",
|
||||
"ciborium",
|
||||
"clap",
|
||||
"criterion-plot",
|
||||
"is-terminal",
|
||||
"itertools",
|
||||
"num-traits",
|
||||
"once_cell",
|
||||
"oorandom",
|
||||
"plotters",
|
||||
"rayon",
|
||||
"regex",
|
||||
"serde",
|
||||
"serde_derive",
|
||||
"serde_json",
|
||||
"tinytemplate",
|
||||
"walkdir",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "criterion-plot"
|
||||
version = "0.5.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "6b50826342786a51a89e2da3a28f1c32b06e387201bc2d19791f622c673706b1"
|
||||
dependencies = [
|
||||
"cast",
|
||||
"itertools",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "crossbeam-channel"
|
||||
version = "0.5.8"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "a33c2bf77f2df06183c3aa30d1e96c0695a313d4f9c453cc3762a6db39f99200"
|
||||
dependencies = [
|
||||
"cfg-if",
|
||||
"crossbeam-utils",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "crossbeam-deque"
|
||||
version = "0.8.3"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "ce6fd6f855243022dcecf8702fef0c297d4338e226845fe067f6341ad9fa0cef"
|
||||
dependencies = [
|
||||
"cfg-if",
|
||||
"crossbeam-epoch",
|
||||
"crossbeam-utils",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "crossbeam-epoch"
|
||||
version = "0.9.15"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "ae211234986c545741a7dc064309f67ee1e5ad243d0e48335adc0484d960bcc7"
|
||||
dependencies = [
|
||||
"autocfg",
|
||||
"cfg-if",
|
||||
"crossbeam-utils",
|
||||
"memoffset",
|
||||
"scopeguard",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "crossbeam-utils"
|
||||
version = "0.8.16"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "5a22b2d63d4d1dc0b7f1b6b2747dd0088008a9be28b6ddf0b1e7d335e3037294"
|
||||
dependencies = [
|
||||
"cfg-if",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "either"
|
||||
version = "1.8.1"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "7fcaabb2fef8c910e7f4c7ce9f67a1283a1715879a7c230ca9d6d1ae31f16d91"
|
||||
|
||||
[[package]]
|
||||
name = "errno"
|
||||
version = "0.3.1"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "4bcfec3a70f97c962c307b2d2c56e358cf1d00b558d74262b5f929ee8cc7e73a"
|
||||
dependencies = [
|
||||
"errno-dragonfly",
|
||||
"libc",
|
||||
"windows-sys",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "errno-dragonfly"
|
||||
version = "0.1.2"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "aa68f1b12764fab894d2755d2518754e71b4fd80ecfb822714a1206c2aab39bf"
|
||||
dependencies = [
|
||||
"cc",
|
||||
"libc",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "fastmath"
|
||||
version = "0.1.0"
|
||||
dependencies = [
|
||||
"bincode",
|
||||
"criterion",
|
||||
"num-traits",
|
||||
"once_cell",
|
||||
"serde",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "half"
|
||||
version = "1.8.2"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "eabb4a44450da02c90444cf74558da904edde8fb4e9035a9a6a4e15445af0bd7"
|
||||
|
||||
[[package]]
|
||||
name = "hermit-abi"
|
||||
version = "0.3.2"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "443144c8cdadd93ebf52ddb4056d257f5b52c04d3c804e657d19eb73fc33668b"
|
||||
|
||||
[[package]]
|
||||
name = "is-terminal"
|
||||
version = "0.4.9"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "cb0889898416213fab133e1d33a0e5858a48177452750691bde3666d0fdbaf8b"
|
||||
dependencies = [
|
||||
"hermit-abi",
|
||||
"rustix",
|
||||
"windows-sys",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "itertools"
|
||||
version = "0.10.5"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "b0fd2260e829bddf4cb6ea802289de2f86d6a7a690192fbe91b3f46e0f2c8473"
|
||||
dependencies = [
|
||||
"either",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "itoa"
|
||||
version = "1.0.9"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "af150ab688ff2122fcef229be89cb50dd66af9e01a4ff320cc137eecc9bacc38"
|
||||
|
||||
[[package]]
|
||||
name = "js-sys"
|
||||
version = "0.3.64"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "c5f195fe497f702db0f318b07fdd68edb16955aed830df8363d837542f8f935a"
|
||||
dependencies = [
|
||||
"wasm-bindgen",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "libc"
|
||||
version = "0.2.147"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "b4668fb0ea861c1df094127ac5f1da3409a82116a4ba74fca2e58ef927159bb3"
|
||||
|
||||
[[package]]
|
||||
name = "linux-raw-sys"
|
||||
version = "0.4.3"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "09fc20d2ca12cb9f044c93e3bd6d32d523e6e2ec3db4f7b2939cd99026ecd3f0"
|
||||
|
||||
[[package]]
|
||||
name = "log"
|
||||
version = "0.4.19"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "b06a4cde4c0f271a446782e3eff8de789548ce57dbc8eca9292c27f4a42004b4"
|
||||
|
||||
[[package]]
|
||||
name = "memchr"
|
||||
version = "2.5.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "2dffe52ecf27772e601905b7522cb4ef790d2cc203488bbd0e2fe85fcb74566d"
|
||||
|
||||
[[package]]
|
||||
name = "memoffset"
|
||||
version = "0.9.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "5a634b1c61a95585bd15607c6ab0c4e5b226e695ff2800ba0cdccddf208c406c"
|
||||
dependencies = [
|
||||
"autocfg",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "num-traits"
|
||||
version = "0.2.15"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "578ede34cf02f8924ab9447f50c28075b4d3e5b269972345e7e0372b38c6cdcd"
|
||||
dependencies = [
|
||||
"autocfg",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "num_cpus"
|
||||
version = "1.16.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "4161fcb6d602d4d2081af7c3a45852d875a03dd337a6bfdd6e06407b61342a43"
|
||||
dependencies = [
|
||||
"hermit-abi",
|
||||
"libc",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "once_cell"
|
||||
version = "1.18.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "dd8b5dd2ae5ed71462c540258bedcb51965123ad7e7ccf4b9a8cafaa4a63576d"
|
||||
|
||||
[[package]]
|
||||
name = "oorandom"
|
||||
version = "11.1.3"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "0ab1bc2a289d34bd04a330323ac98a1b4bc82c9d9fcb1e66b63caa84da26b575"
|
||||
|
||||
[[package]]
|
||||
name = "plotters"
|
||||
version = "0.3.5"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "d2c224ba00d7cadd4d5c660deaf2098e5e80e07846537c51f9cfa4be50c1fd45"
|
||||
dependencies = [
|
||||
"num-traits",
|
||||
"plotters-backend",
|
||||
"plotters-svg",
|
||||
"wasm-bindgen",
|
||||
"web-sys",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "plotters-backend"
|
||||
version = "0.3.5"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "9e76628b4d3a7581389a35d5b6e2139607ad7c75b17aed325f210aa91f4a9609"
|
||||
|
||||
[[package]]
|
||||
name = "plotters-svg"
|
||||
version = "0.3.5"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "38f6d39893cca0701371e3c27294f09797214b86f1fb951b89ade8ec04e2abab"
|
||||
dependencies = [
|
||||
"plotters-backend",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "proc-macro2"
|
||||
version = "1.0.66"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "18fb31db3f9bddb2ea821cde30a9f70117e3f119938b5ee630b7403aa6e2ead9"
|
||||
dependencies = [
|
||||
"unicode-ident",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "quote"
|
||||
version = "1.0.31"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "5fe8a65d69dd0808184ebb5f836ab526bb259db23c657efa38711b1072ee47f0"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "rayon"
|
||||
version = "1.7.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "1d2df5196e37bcc87abebc0053e20787d73847bb33134a69841207dd0a47f03b"
|
||||
dependencies = [
|
||||
"either",
|
||||
"rayon-core",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "rayon-core"
|
||||
version = "1.11.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "4b8f95bd6966f5c87776639160a66bd8ab9895d9d4ab01ddba9fc60661aebe8d"
|
||||
dependencies = [
|
||||
"crossbeam-channel",
|
||||
"crossbeam-deque",
|
||||
"crossbeam-utils",
|
||||
"num_cpus",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "regex"
|
||||
version = "1.9.1"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "b2eae68fc220f7cf2532e4494aded17545fce192d59cd996e0fe7887f4ceb575"
|
||||
dependencies = [
|
||||
"aho-corasick",
|
||||
"memchr",
|
||||
"regex-automata",
|
||||
"regex-syntax",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "regex-automata"
|
||||
version = "0.3.3"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "39354c10dd07468c2e73926b23bb9c2caca74c5501e38a35da70406f1d923310"
|
||||
dependencies = [
|
||||
"aho-corasick",
|
||||
"memchr",
|
||||
"regex-syntax",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "regex-syntax"
|
||||
version = "0.7.4"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "e5ea92a5b6195c6ef2a0295ea818b312502c6fc94dde986c5553242e18fd4ce2"
|
||||
|
||||
[[package]]
|
||||
name = "rustix"
|
||||
version = "0.38.4"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "0a962918ea88d644592894bc6dc55acc6c0956488adcebbfb6e273506b7fd6e5"
|
||||
dependencies = [
|
||||
"bitflags",
|
||||
"errno",
|
||||
"libc",
|
||||
"linux-raw-sys",
|
||||
"windows-sys",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "ryu"
|
||||
version = "1.0.15"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "1ad4cc8da4ef723ed60bced201181d83791ad433213d8c24efffda1eec85d741"
|
||||
|
||||
[[package]]
|
||||
name = "same-file"
|
||||
version = "1.0.6"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "93fc1dc3aaa9bfed95e02e6eadabb4baf7e3078b0bd1b4d7b6b0b68378900502"
|
||||
dependencies = [
|
||||
"winapi-util",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "scopeguard"
|
||||
version = "1.2.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "94143f37725109f92c262ed2cf5e59bce7498c01bcc1502d7b9afe439a4e9f49"
|
||||
|
||||
[[package]]
|
||||
name = "serde"
|
||||
version = "1.0.171"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "30e27d1e4fd7659406c492fd6cfaf2066ba8773de45ca75e855590f856dc34a9"
|
||||
dependencies = [
|
||||
"serde_derive",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "serde_derive"
|
||||
version = "1.0.171"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "389894603bd18c46fa56231694f8d827779c0951a667087194cf9de94ed24682"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "serde_json"
|
||||
version = "1.0.103"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "d03b412469450d4404fe8499a268edd7f8b79fecb074b0d812ad64ca21f4031b"
|
||||
dependencies = [
|
||||
"itoa",
|
||||
"ryu",
|
||||
"serde",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "syn"
|
||||
version = "2.0.26"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "45c3457aacde3c65315de5031ec191ce46604304d2446e803d71ade03308d970"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"unicode-ident",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "tinytemplate"
|
||||
version = "1.2.1"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "be4d6b5f19ff7664e8c98d03e2139cb510db9b0a60b55f8e8709b689d939b6bc"
|
||||
dependencies = [
|
||||
"serde",
|
||||
"serde_json",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "unicode-ident"
|
||||
version = "1.0.11"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "301abaae475aa91687eb82514b328ab47a211a533026cb25fc3e519b86adfc3c"
|
||||
|
||||
[[package]]
|
||||
name = "walkdir"
|
||||
version = "2.3.3"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "36df944cda56c7d8d8b7496af378e6b16de9284591917d307c9b4d313c44e698"
|
||||
dependencies = [
|
||||
"same-file",
|
||||
"winapi-util",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "wasm-bindgen"
|
||||
version = "0.2.87"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "7706a72ab36d8cb1f80ffbf0e071533974a60d0a308d01a5d0375bf60499a342"
|
||||
dependencies = [
|
||||
"cfg-if",
|
||||
"wasm-bindgen-macro",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "wasm-bindgen-backend"
|
||||
version = "0.2.87"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "5ef2b6d3c510e9625e5fe6f509ab07d66a760f0885d858736483c32ed7809abd"
|
||||
dependencies = [
|
||||
"bumpalo",
|
||||
"log",
|
||||
"once_cell",
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn",
|
||||
"wasm-bindgen-shared",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "wasm-bindgen-macro"
|
||||
version = "0.2.87"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "dee495e55982a3bd48105a7b947fd2a9b4a8ae3010041b9e0faab3f9cd028f1d"
|
||||
dependencies = [
|
||||
"quote",
|
||||
"wasm-bindgen-macro-support",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "wasm-bindgen-macro-support"
|
||||
version = "0.2.87"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "54681b18a46765f095758388f2d0cf16eb8d4169b639ab575a8f5693af210c7b"
|
||||
dependencies = [
|
||||
"proc-macro2",
|
||||
"quote",
|
||||
"syn",
|
||||
"wasm-bindgen-backend",
|
||||
"wasm-bindgen-shared",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "wasm-bindgen-shared"
|
||||
version = "0.2.87"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "ca6ad05a4870b2bf5fe995117d3728437bd27d7cd5f06f13c17443ef369775a1"
|
||||
|
||||
[[package]]
|
||||
name = "web-sys"
|
||||
version = "0.3.64"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "9b85cbef8c220a6abc02aefd892dfc0fc23afb1c6a426316ec33253a3877249b"
|
||||
dependencies = [
|
||||
"js-sys",
|
||||
"wasm-bindgen",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "winapi"
|
||||
version = "0.3.9"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "5c839a674fcd7a98952e593242ea400abe93992746761e38641405d28b00f419"
|
||||
dependencies = [
|
||||
"winapi-i686-pc-windows-gnu",
|
||||
"winapi-x86_64-pc-windows-gnu",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "winapi-i686-pc-windows-gnu"
|
||||
version = "0.4.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "ac3b87c63620426dd9b991e5ce0329eff545bccbbb34f3be09ff6fb6ab51b7b6"
|
||||
|
||||
[[package]]
|
||||
name = "winapi-util"
|
||||
version = "0.1.5"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "70ec6ce85bb158151cae5e5c87f95a8e97d2c0c4b001223f33a334e3ce5de178"
|
||||
dependencies = [
|
||||
"winapi",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "winapi-x86_64-pc-windows-gnu"
|
||||
version = "0.4.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "712e227841d057c1ee1cd2fb22fa7e5a5461ae8e48fa2ca79ec42cfc1931183f"
|
||||
|
||||
[[package]]
|
||||
name = "windows-sys"
|
||||
version = "0.48.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "677d2418bec65e3338edb076e806bc1ec15693c5d0104683f2efe857f61056a9"
|
||||
dependencies = [
|
||||
"windows-targets",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "windows-targets"
|
||||
version = "0.48.1"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "05d4b17490f70499f20b9e791dcf6a299785ce8af4d709018206dc5b4953e95f"
|
||||
dependencies = [
|
||||
"windows_aarch64_gnullvm",
|
||||
"windows_aarch64_msvc",
|
||||
"windows_i686_gnu",
|
||||
"windows_i686_msvc",
|
||||
"windows_x86_64_gnu",
|
||||
"windows_x86_64_gnullvm",
|
||||
"windows_x86_64_msvc",
|
||||
]
|
||||
|
||||
[[package]]
|
||||
name = "windows_aarch64_gnullvm"
|
||||
version = "0.48.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "91ae572e1b79dba883e0d315474df7305d12f569b400fcf90581b06062f7e1bc"
|
||||
|
||||
[[package]]
|
||||
name = "windows_aarch64_msvc"
|
||||
version = "0.48.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "b2ef27e0d7bdfcfc7b868b317c1d32c641a6fe4629c171b8928c7b08d98d7cf3"
|
||||
|
||||
[[package]]
|
||||
name = "windows_i686_gnu"
|
||||
version = "0.48.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "622a1962a7db830d6fd0a69683c80a18fda201879f0f447f065a3b7467daa241"
|
||||
|
||||
[[package]]
|
||||
name = "windows_i686_msvc"
|
||||
version = "0.48.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "4542c6e364ce21bf45d69fdd2a8e455fa38d316158cfd43b3ac1c5b1b19f8e00"
|
||||
|
||||
[[package]]
|
||||
name = "windows_x86_64_gnu"
|
||||
version = "0.48.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "ca2b8a661f7628cbd23440e50b05d705db3686f894fc9580820623656af974b1"
|
||||
|
||||
[[package]]
|
||||
name = "windows_x86_64_gnullvm"
|
||||
version = "0.48.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "7896dbc1f41e08872e9d5e8f8baa8fdd2677f29468c4e156210174edc7f7b953"
|
||||
|
||||
[[package]]
|
||||
name = "windows_x86_64_msvc"
|
||||
version = "0.48.0"
|
||||
source = "registry+https://github.com/rust-lang/crates.io-index"
|
||||
checksum = "1a515f5799fe4961cb532f983ce2b23082366b898e52ffbce459c86f67c8378a"
|
||||
30
Cargo.toml
Normal file
30
Cargo.toml
Normal file
@@ -0,0 +1,30 @@
|
||||
[package]
|
||||
name = "fastmath"
|
||||
version = "0.1.0"
|
||||
edition = "2021"
|
||||
|
||||
# See more keys and their definitions at https://doc.rust-lang.org/cargo/reference/manifest.html
|
||||
|
||||
[dependencies]
|
||||
bincode = "1.3.3"
|
||||
num-traits = "0.2.15"
|
||||
once_cell = "1.18.0"
|
||||
serde = {version = "1.0.171", features = ["derive"] }
|
||||
|
||||
[build-dependencies]
|
||||
bincode = "1.3.3"
|
||||
num-traits = "0.2.15"
|
||||
serde = {version = "1.0.171", features = ["derive"] }
|
||||
|
||||
[lib]
|
||||
name = "fastmath"
|
||||
path = "src/lib.rs"
|
||||
test = true
|
||||
bench = true
|
||||
|
||||
[[bench]]
|
||||
name = "bench"
|
||||
harness = false
|
||||
|
||||
[dev-dependencies]
|
||||
criterion = "0.5.1"
|
||||
111
benches/bench.rs
Normal file
111
benches/bench.rs
Normal file
@@ -0,0 +1,111 @@
|
||||
#![allow(unused_imports)]
|
||||
extern crate fastmath;
|
||||
|
||||
use fastmath::*;
|
||||
use criterion::{Criterion, BenchmarkGroup, measurement::WallTime};
|
||||
use criterion::{black_box, criterion_group, criterion_main};
|
||||
|
||||
fn pow2_benchmarks(group: &mut BenchmarkGroup<WallTime>, x_f64: &[f64], x_f32: &[f32]) {
|
||||
group.bench_function("f64_fast", |b| {
|
||||
b.iter(|| x_f64.iter().map(|&x| black_box(x).fast_pow2()).collect::<Vec<f64>>())
|
||||
});
|
||||
group.bench_function("f64_builtin_fn", |b| {
|
||||
b.iter(|| x_f64.iter().map(|&x| black_box(x).powi(2)).collect::<Vec<f64>>())
|
||||
});
|
||||
group.bench_function("f64_builtin_mul", |b| {
|
||||
b.iter(|| x_f64.iter().map(|&x| black_box(x) * x).collect::<Vec<f64>>())
|
||||
});
|
||||
group.bench_function("f32_fast", |b| {
|
||||
b.iter(|| x_f32.iter().map(|&x| black_box(x).fast_pow2()).collect::<Vec<f32>>())
|
||||
});
|
||||
group.bench_function("f32_builtin_fn", |b| {
|
||||
b.iter(|| x_f32.iter().map(|&x| black_box(x).powi(2)).collect::<Vec<f32>>())
|
||||
});
|
||||
group.bench_function("f32_builtin_mul", |b| {
|
||||
b.iter(|| x_f32.iter().map(|&x| black_box(x) * x).collect::<Vec<f32>>())
|
||||
});
|
||||
}
|
||||
|
||||
fn exp_benchmarks(group: &mut BenchmarkGroup<WallTime>, x_f64: &[f64], x_f32: &[f32]) {
|
||||
group.bench_function("f64_fast", |b| {
|
||||
b.iter(|| x_f64.iter().map(|&x| black_box(x).fast_exp()).collect::<Vec<f64>>())
|
||||
});
|
||||
group.bench_function("f64_builtin", |b| {
|
||||
b.iter(|| x_f64.iter().map(|&x| black_box(x).exp()).collect::<Vec<f64>>())
|
||||
});
|
||||
group.bench_function("f32_fast", |b| {
|
||||
b.iter(|| x_f32.iter().map(|&x| black_box(x).fast_exp()).collect::<Vec<f32>>())
|
||||
});
|
||||
group.bench_function("f32_builtin", |b| {
|
||||
b.iter(|| x_f32.iter().map(|&x| black_box(x).exp()).collect::<Vec<f32>>())
|
||||
});
|
||||
}
|
||||
|
||||
fn cos_benchmarks(group: &mut BenchmarkGroup<WallTime>, x_f64: &[f64], x_f32: &[f32]) {
|
||||
group.bench_function("f64_fast", |b| {
|
||||
b.iter(|| x_f64.iter().map(|&x| black_box(x).fast_cos()).collect::<Vec<f64>>())
|
||||
});
|
||||
group.bench_function("f64_lookup", |b| {
|
||||
b.iter(|| x_f64.iter().map(|&x| black_box(x).lookup_cos()).collect::<Vec<f64>>())
|
||||
});
|
||||
group.bench_function("f64_builtin", |b| {
|
||||
b.iter(|| x_f64.iter().map(|&x| black_box(x).cos()).collect::<Vec<f64>>())
|
||||
});
|
||||
group.bench_function("f32_fast", |b| {
|
||||
b.iter(|| x_f32.iter().map(|&x| black_box(x).fast_cos()).collect::<Vec<f32>>())
|
||||
});
|
||||
group.bench_function("f32_lookup", |b| {
|
||||
b.iter(|| x_f32.iter().map(|&x| black_box(x).lookup_cos()).collect::<Vec<f32>>())
|
||||
});
|
||||
group.bench_function("f32_builtin", |b| {
|
||||
b.iter(|| x_f32.iter().map(|&x| black_box(x).cos()).collect::<Vec<f32>>())
|
||||
});
|
||||
}
|
||||
|
||||
fn sigmoid_benchmarks(group: &mut BenchmarkGroup<WallTime>, x_f64: &[f64], x_f32: &[f32]) {
|
||||
group.bench_function("f64_fast", |b| {
|
||||
b.iter(|| x_f64.iter().map(|&x| black_box(x).fast_sigmoid()).collect::<Vec<f64>>())
|
||||
});
|
||||
group.bench_function("f64_builtin", |b| {
|
||||
b.iter(|| x_f64.iter().map(|&x| sigmoid_builtin_f64(black_box(x))).collect::<Vec<f64>>())
|
||||
});
|
||||
group.bench_function("f32_fast", |b| {
|
||||
b.iter(|| x_f32.iter().map(|&x| black_box(x).fast_sigmoid()).collect::<Vec<f32>>())
|
||||
});
|
||||
group.bench_function("f32_builtin", |b| {
|
||||
b.iter(|| x_f32.iter().map(|&x| sigmoid_builtin_f32(black_box(x))).collect::<Vec<f32>>())
|
||||
});
|
||||
}
|
||||
|
||||
fn criterion_benchmark(c: &mut Criterion) {
|
||||
// Prepare x values for testing functions
|
||||
let x_f64 = (-10000..10000)
|
||||
.map(|a| (a as f64) / 1000.)
|
||||
.collect::<Vec<f64>>();
|
||||
let x_f32 = (-10000..10000)
|
||||
.map(|a| (a as f32) / 1000.)
|
||||
.collect::<Vec<f32>>();
|
||||
// to ensure tests are fair, we need to instantiate the lookup tables
|
||||
1.0f64.lookup_cos();
|
||||
1.0f32.lookup_cos();
|
||||
// Then, tests can begin
|
||||
let mut group = c.benchmark_group("pow2");
|
||||
pow2_benchmarks(&mut group, &x_f64, &x_f32);
|
||||
group.finish();
|
||||
|
||||
let mut group = c.benchmark_group("exp");
|
||||
exp_benchmarks(&mut group, &x_f64, &x_f32);
|
||||
group.finish();
|
||||
|
||||
let mut group = c.benchmark_group("cos");
|
||||
cos_benchmarks(&mut group, &x_f64, &x_f32);
|
||||
group.finish();
|
||||
|
||||
let mut group = c.benchmark_group("sigmoid");
|
||||
sigmoid_benchmarks(&mut group, &x_f64, &x_f32);
|
||||
group.finish();
|
||||
}
|
||||
|
||||
criterion_group!(benches, criterion_benchmark);
|
||||
criterion_main!(benches);
|
||||
|
||||
49
build.rs
Normal file
49
build.rs
Normal file
@@ -0,0 +1,49 @@
|
||||
// build.rs
|
||||
|
||||
mod precalculate_lookup_tables {
|
||||
use std::fs::{create_dir_all, File};
|
||||
use std::io::Write;
|
||||
include!("src/lookup/lookup_table.rs");
|
||||
use bincode::serialize;
|
||||
|
||||
const PRECISION: usize = 1000;
|
||||
|
||||
fn precalculate_sin_tables() -> Result<(), Box<dyn std::error::Error>> {
|
||||
let data = serialize(&EndoSinLookupTable::<f32>::new(PRECISION))?;
|
||||
let mut file = File::create("src/lookup/data/sin_f32.bin")?;
|
||||
file.write_all(&data)?;
|
||||
|
||||
let data = serialize(&EndoSinLookupTable::<f64>::new(PRECISION))?;
|
||||
let mut file = File::create("src/lookup/data/sin_f64.bin")?;
|
||||
file.write_all(&data)?;
|
||||
|
||||
Ok(())
|
||||
}
|
||||
|
||||
fn precalculate_cos_tables() -> Result<(), Box<dyn std::error::Error>> {
|
||||
let data = serialize(&EndoCosLookupTable::<f32>::new(PRECISION))?;
|
||||
let mut file = File::create("src/lookup/data/cos_f32.bin")?;
|
||||
file.write_all(&data)?;
|
||||
|
||||
let data = serialize(&EndoCosLookupTable::<f64>::new(PRECISION))?;
|
||||
let mut file = File::create("src/lookup/data/cos_f64.bin")?;
|
||||
file.write_all(&data)?;
|
||||
|
||||
Ok(())
|
||||
}
|
||||
|
||||
pub fn generate() -> Result<(), Box<dyn std::error::Error>> {
|
||||
create_dir_all("src/lookup/data")?;
|
||||
|
||||
precalculate_sin_tables()?;
|
||||
precalculate_cos_tables()?;
|
||||
|
||||
Ok(())
|
||||
}
|
||||
}
|
||||
|
||||
fn main() -> Result<(), Box<dyn std::error::Error>> {
|
||||
precalculate_lookup_tables::generate()?;
|
||||
|
||||
Ok(())
|
||||
}
|
||||
208
src/fastmath.rs
Normal file
208
src/fastmath.rs
Normal file
@@ -0,0 +1,208 @@
|
||||
//! A collection of fast (often approximate) mathematical functions for accelerating mathematical functions
|
||||
|
||||
// Optimisation note: lookup tables become faster when calculation takes > ~1ms
|
||||
|
||||
use std::f32::consts as f32_consts;
|
||||
use std::f64::consts as f64_consts;
|
||||
use crate::lookup::*;
|
||||
|
||||
pub trait FastMath: FastCos + FastPow2 + FastExp + FastSigmoid {}
|
||||
impl FastMath for f32 {}
|
||||
impl FastMath for f64 {}
|
||||
|
||||
|
||||
pub trait LookupCos {
|
||||
fn lookup_cos(self: Self) -> Self;
|
||||
}
|
||||
impl LookupCos for f64 {
|
||||
#[inline]
|
||||
fn lookup_cos(self: Self) -> f64 {
|
||||
// Look up the value in the table
|
||||
COS_LOOKUP_F64.lookup(self)
|
||||
}
|
||||
}
|
||||
impl LookupCos for f32 {
|
||||
#[inline]
|
||||
fn lookup_cos(self: Self) -> f32 {
|
||||
// Look up the value in the table
|
||||
COS_LOOKUP_F32.lookup(self)
|
||||
}
|
||||
}
|
||||
|
||||
pub trait FastCos {
|
||||
fn fast_cos(self: Self) -> Self;
|
||||
}
|
||||
impl FastCos for f32 {
|
||||
#[inline]
|
||||
fn fast_cos(self: Self) -> f32 {
|
||||
const BITAND: u32 = u32::MAX / 2;
|
||||
const ONE: f32 = 1.0;
|
||||
let mod_x = (((self + f32_consts::PI).abs()) % f32_consts::TAU) - f32_consts::PI;
|
||||
let v = mod_x.to_bits() & BITAND;
|
||||
let qpprox = ONE - f32_consts::FRAC_2_PI * f32::from_bits(v);
|
||||
qpprox + f32_consts::FRAC_PI_6 * qpprox * (ONE - qpprox * qpprox)
|
||||
}
|
||||
}
|
||||
impl FastCos for f64 {
|
||||
#[inline]
|
||||
fn fast_cos(self: Self) -> f64 {
|
||||
const BITAND: u64 = u64::MAX / 2;
|
||||
const ONE: f64 = 1.0;
|
||||
let mod_x = (((self + f64_consts::PI).abs()) % f64_consts::TAU) - f64_consts::PI;
|
||||
let v = mod_x.to_bits() & BITAND;
|
||||
let qpprox = ONE - f64_consts::FRAC_2_PI * f64::from_bits(v);
|
||||
qpprox + f64_consts::FRAC_PI_6 * qpprox * (ONE - qpprox * qpprox)
|
||||
}
|
||||
}
|
||||
|
||||
pub trait FastPow2 {
|
||||
fn fast_pow2(self: Self) -> Self;
|
||||
}
|
||||
impl FastPow2 for f32 {
|
||||
#[inline]
|
||||
fn fast_pow2(self: Self) -> f32 {
|
||||
// Khinchins constant over 3. IDK why it gives the best fit, but it does
|
||||
const KHINCHIN_3: f32 = 2.68545200106530644530971483548179569382038229399446295305115234555721885953715200280114117493184769799515 / 3.0;
|
||||
const CLIPP_THRESH: f32 = 0.12847338;
|
||||
const V_SCALE: f32 = 8388608.0; // (1_i32 << 23) as f32
|
||||
const CLIPP_SHIFT: f32 = 126.67740855;
|
||||
let abs_p = self.abs();
|
||||
let clipp = abs_p.max(CLIPP_THRESH); // if abs_p < CLIPP_THRESH { CLIPP_THRESH } else { abs_p };
|
||||
let v = (V_SCALE * (clipp + CLIPP_SHIFT)) as u32;
|
||||
f32::from_bits(v) - KHINCHIN_3
|
||||
}
|
||||
}
|
||||
impl FastPow2 for f64 {
|
||||
#[inline]
|
||||
fn fast_pow2(self: Self) -> f64 {
|
||||
const KHINCHIN_3: f64 = 2.68545200106530644530971483548179569382038229399446295305115234555721885953715200280114117493184769799515 / 3.0;
|
||||
const CLIPP_THRESH: f64 = -45774.9247660416;
|
||||
const V_SCALE: f64 = 4503599627370496.0; // (1i64 << 52) as f64
|
||||
const CLIPP_SHIFT: f64 = 1022.6769200000002;
|
||||
const ZERO: f64 = 0.;
|
||||
let abs_p = self.abs();
|
||||
let clipp = abs_p.max(CLIPP_THRESH); // if abs_p < CLIPP_THRESH { CLIPP_THRESH } else { abs_p };
|
||||
let v = (V_SCALE * (clipp + CLIPP_SHIFT)) as u64;
|
||||
let y = f64::from_bits(v) - KHINCHIN_3;
|
||||
if y.is_sign_positive() {
|
||||
y
|
||||
} else {
|
||||
ZERO
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
pub trait FastExp {
|
||||
fn fast_exp(self: Self) -> Self;
|
||||
}
|
||||
impl FastExp for f32 {
|
||||
#[inline]
|
||||
fn fast_exp(self: Self) -> f32 {
|
||||
const CLIPP_THRESH: f32 = -126.0; // 0.12847338;
|
||||
const V_SCALE: f32 = 8388608.0; // (1_i32 << 23) as f32
|
||||
const CLIPP_SHIFT: f32 = 126.94269504; // 126.67740855;
|
||||
|
||||
let scaled_p = f32_consts::LOG2_E * self;
|
||||
let clipp = scaled_p.max(CLIPP_THRESH); // if scaled_p < CLIPP_THRESH { CLIPP_THRESH } else { scaled_p };
|
||||
let v = (V_SCALE * (clipp + CLIPP_SHIFT)) as u32;
|
||||
f32::from_bits(v)
|
||||
}
|
||||
}
|
||||
impl FastExp for f64 {
|
||||
#[inline]
|
||||
fn fast_exp(self: Self) -> f64 {
|
||||
const CLIPP_THRESH: f64 = -180335.51911105003;
|
||||
const V_SCALE: f64 = 4524653012949098.0;
|
||||
const CLIPP_SHIFT: f64 = 1018.1563534409383;
|
||||
|
||||
let scaled_p = f64_consts::LOG2_E * self;
|
||||
let clipp = scaled_p.max(CLIPP_THRESH); // let clipp = if scaled_p < CLIPP_THRESH { CLIPP_THRESH } else { scaled_p };
|
||||
let v = (V_SCALE * (clipp + CLIPP_SHIFT)) as u64;
|
||||
f64::from_bits(v)
|
||||
}
|
||||
}
|
||||
|
||||
pub trait FastSigmoid {
|
||||
fn fast_sigmoid(self: Self) -> Self;
|
||||
}
|
||||
impl FastSigmoid for f32 {
|
||||
#[inline]
|
||||
fn fast_sigmoid(self: Self) -> f32 {
|
||||
const ONE: f32 = 1.0;
|
||||
(ONE + (-self).fast_exp()).recip()
|
||||
}
|
||||
}
|
||||
impl FastSigmoid for f64 {
|
||||
#[inline]
|
||||
fn fast_sigmoid(self: Self) -> f64 {
|
||||
const ONE: f64 = 1.0;
|
||||
(ONE + (-self).fast_exp()).recip()
|
||||
}
|
||||
}
|
||||
|
||||
// A trait for testing and improving implementations of fast functions
|
||||
pub trait Test {
|
||||
fn test(self: Self) -> Self;
|
||||
}
|
||||
impl Test for f32 {
|
||||
#[inline]
|
||||
fn test(self: Self) -> f32 {
|
||||
// Khinchins constant over 3. IDK why it gives the best fit, but it does
|
||||
// const KHINCHIN_3: f32 = 2.68545200106530644530971483548179569382038229399446295305115234555721885953715200280114117493184769799515 / 3.0;
|
||||
const CLIPP_THRESH: f32 = -126.0; // 0.12847338;
|
||||
const V_SCALE: f32 = 8388608.0; // (1_i32 << 23) as f32
|
||||
const CLIPP_SHIFT: f32 = 126.94269504; // 126.67740855;
|
||||
|
||||
let scaled_p = f32_consts::LOG2_E * self;
|
||||
let clipp = if scaled_p < CLIPP_THRESH {
|
||||
CLIPP_THRESH
|
||||
} else {
|
||||
scaled_p
|
||||
};
|
||||
let v = (V_SCALE * (clipp + CLIPP_SHIFT)) as u32;
|
||||
f32::from_bits(v) // - KHINCHIN_3
|
||||
}
|
||||
}
|
||||
impl Test for f64 {
|
||||
#[inline]
|
||||
fn test(self: Self) -> f64 {
|
||||
const CLIPP_THRESH: f64 = -180335.51911105003;
|
||||
const V_SCALE: f64 = 4524653012949098.0;
|
||||
const CLIPP_SHIFT: f64 = 1018.1563534409383;
|
||||
|
||||
let scaled_p = f64_consts::LOG2_E * self;
|
||||
let clipp = if scaled_p < CLIPP_THRESH {
|
||||
CLIPP_THRESH
|
||||
} else {
|
||||
scaled_p
|
||||
};
|
||||
let v = (V_SCALE * (clipp + CLIPP_SHIFT)) as u64;
|
||||
f64::from_bits(v)
|
||||
}
|
||||
}
|
||||
|
||||
#[allow(non_snake_case, dead_code)]
|
||||
pub fn optimizing(p: f64, CLIPP_THRESH: f64, V_SCALE: f64, CLIPP_SHIFT: f64) -> f64 {
|
||||
// const CLIPP_THRESH: f64 = -45774.9247660416;
|
||||
// const V_SCALE: f64 = 4503599627370496.0;
|
||||
// const CLIPP_SHIFT: f64 = 1022.6769200000002;
|
||||
|
||||
let scaled_p = f64_consts::LOG2_E * p;
|
||||
let clipp = if scaled_p < CLIPP_THRESH {
|
||||
CLIPP_THRESH
|
||||
} else {
|
||||
scaled_p
|
||||
};
|
||||
let v = (V_SCALE * (clipp + CLIPP_SHIFT)) as u64;
|
||||
f64::from_bits(v)
|
||||
}
|
||||
|
||||
#[inline]
|
||||
pub fn sigmoid_builtin_f32(p: f32) -> f32 {
|
||||
(1. + (-p).exp()).recip()
|
||||
}
|
||||
|
||||
#[inline]
|
||||
pub fn sigmoid_builtin_f64(p: f64) -> f64 {
|
||||
(1. + (-p).exp()).recip()
|
||||
}
|
||||
10
src/lib.rs
Normal file
10
src/lib.rs
Normal file
@@ -0,0 +1,10 @@
|
||||
//lib.rs
|
||||
#![allow(unused_imports)]
|
||||
|
||||
pub mod lookup;
|
||||
mod fastmath;
|
||||
|
||||
pub use fastmath::*;
|
||||
|
||||
#[cfg(test)]
|
||||
mod tests;
|
||||
30
src/lookup/const_tables.rs
Normal file
30
src/lookup/const_tables.rs
Normal file
@@ -0,0 +1,30 @@
|
||||
// lookup/const_tables.rs
|
||||
|
||||
use once_cell::sync::Lazy;
|
||||
use std::fs::read;
|
||||
use bincode::deserialize;
|
||||
use super::lookup_table::*;
|
||||
|
||||
pub const SIN_LOOKUP_F32: Lazy<EndoSinLookupTable<f32>> = Lazy::new(|| {
|
||||
deserialize(
|
||||
&read("src/lookup/data/sin_f32.bin").expect("Failed to read sin_f64.bin")
|
||||
).expect("Failed to load SIN_LOOKUP_F32")
|
||||
});
|
||||
|
||||
pub const SIN_LOOKUP_F64: Lazy<EndoSinLookupTable<f64>> = Lazy::new(|| {
|
||||
deserialize(
|
||||
&read("src/lookup/data/sin_f64.bin").expect("Failed to read sin_f32.bin")
|
||||
).expect("Failed to load SIN_LOOKUP_F64")
|
||||
});
|
||||
|
||||
pub const COS_LOOKUP_F32: Lazy<EndoCosLookupTable<f32>> = Lazy::new(|| {
|
||||
deserialize(
|
||||
&read("src/lookup/data/cos_f32.bin").expect("Failed to read cos_f64.bin")
|
||||
).expect("Failed to load COS_LOOKUP_F32")
|
||||
});
|
||||
|
||||
pub const COS_LOOKUP_F64: Lazy<EndoCosLookupTable<f64>> = Lazy::new(|| {
|
||||
deserialize(
|
||||
&read("src/lookup/data/cos_f64.bin").expect("Failed to read cos_f32.bin")
|
||||
).expect("Failed to load COS_LOOKUP_F64")
|
||||
});
|
||||
BIN
src/lookup/data/cos_f32.bin
Normal file
BIN
src/lookup/data/cos_f32.bin
Normal file
Binary file not shown.
BIN
src/lookup/data/cos_f64.bin
Normal file
BIN
src/lookup/data/cos_f64.bin
Normal file
Binary file not shown.
BIN
src/lookup/data/sin_f32.bin
Normal file
BIN
src/lookup/data/sin_f32.bin
Normal file
Binary file not shown.
BIN
src/lookup/data/sin_f64.bin
Normal file
BIN
src/lookup/data/sin_f64.bin
Normal file
Binary file not shown.
149
src/lookup/lookup_table.rs
Normal file
149
src/lookup/lookup_table.rs
Normal file
@@ -0,0 +1,149 @@
|
||||
use num_traits::sign::Signed;
|
||||
use num_traits::float::{Float, FloatConst};
|
||||
use num_traits::NumCast;
|
||||
use std::ops::{Sub, Rem};
|
||||
use serde::{Serialize, Deserialize};
|
||||
// use packed_simd::f64x4;
|
||||
|
||||
#[derive(Default, Debug, Serialize, Deserialize, Clone)]
|
||||
pub struct FloatLookupTable<T1, T2>
|
||||
where T1: Float,
|
||||
T2: Float,
|
||||
{
|
||||
keys: Vec<T1>,
|
||||
values: Vec<T2>,
|
||||
}
|
||||
impl<T1, T2> FloatLookupTable<T1, T2>
|
||||
where T1: Float,
|
||||
T2: Float,
|
||||
{
|
||||
pub fn new(mut keys: Vec<T1>, mut values: Vec<T2>) -> Self {
|
||||
let mut indices: Vec<_> = (0..keys.len()).collect();
|
||||
indices.sort_by(|&i, &j| keys[i].partial_cmp(&keys[j]).unwrap());
|
||||
for i in 0..keys.len() {
|
||||
while i != indices[i] {
|
||||
let swap_index = indices[i];
|
||||
keys.swap(i, swap_index);
|
||||
values.swap(i, swap_index);
|
||||
indices.swap(i, swap_index);
|
||||
}
|
||||
}
|
||||
FloatLookupTable { keys, values }
|
||||
}
|
||||
|
||||
#[allow(dead_code)]
|
||||
pub fn lookup(&self, key: T1) -> T2 {
|
||||
match self.keys.binary_search_by(|probe| probe.partial_cmp(&key).unwrap()) {
|
||||
Ok(index) => self.values[index],
|
||||
Err(index) => {
|
||||
let upper_key = &self.keys[index];
|
||||
let upper_val = &self.values[index];
|
||||
let low_index = index - 1;
|
||||
let lower_key = &self.keys[low_index];
|
||||
let lower_val = &self.values[low_index];
|
||||
// select nearest neighbour
|
||||
let diff_upper = (key - *upper_key).abs();
|
||||
let diff_lower = (key - *lower_key).abs();
|
||||
let mask = diff_lower <= diff_upper;
|
||||
(*lower_val * T2::from(mask as u8).expect("Failed to unwrap mask")) +
|
||||
(*upper_val * T2::from(!mask as u8).expect("Failed to unwrap !mask"))
|
||||
}
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
#[derive(Default, Debug, Serialize, Deserialize, Clone)]
|
||||
pub struct CyclingFloatLookupTable<T1, T2>
|
||||
where T1: Float,
|
||||
T2: Float,
|
||||
{
|
||||
lookup_table: FloatLookupTable<T1, T2>,
|
||||
lower_bound: T1,
|
||||
upper_bound: T1,
|
||||
bound_range: T1,
|
||||
}
|
||||
impl<T1, T2> CyclingFloatLookupTable<T1, T2>
|
||||
where T1: Float,
|
||||
T2: Float,
|
||||
{
|
||||
pub fn new(keys: Vec<T1>, values: Vec<T2>, lower_bound: T1, upper_bound: T1) -> Self {
|
||||
CyclingFloatLookupTable {
|
||||
lookup_table: FloatLookupTable::new(keys, values),
|
||||
lower_bound: lower_bound,
|
||||
upper_bound: upper_bound,
|
||||
bound_range: upper_bound - lower_bound,
|
||||
}
|
||||
}
|
||||
|
||||
pub fn lookup(&self, key: T1) -> T2 {
|
||||
let key = (key % self.bound_range) + self.lower_bound;
|
||||
self.lookup_table.lookup(key)
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
#[derive(Default, Debug, Serialize, Deserialize, Clone)]
|
||||
pub struct EndoSinLookupTable<T>
|
||||
where
|
||||
T: Float + FloatConst,
|
||||
{
|
||||
lookup_table: CyclingFloatLookupTable<T, T>,
|
||||
}
|
||||
impl<T> EndoSinLookupTable<T>
|
||||
where
|
||||
T: Float + FloatConst,
|
||||
{
|
||||
pub fn new(precision: usize) -> Self {
|
||||
let mut keys = Vec::with_capacity(precision);
|
||||
let mut values = Vec::with_capacity(precision);
|
||||
let upper_bound = T::PI();
|
||||
let step = T::FRAC_PI_2() / <T as NumCast>::from(precision).unwrap();
|
||||
for i in 0..precision+1 {
|
||||
let key = step * <T as NumCast>::from(i).unwrap();
|
||||
let value = key.sin();
|
||||
keys.push(key);
|
||||
values.push(value);
|
||||
}
|
||||
EndoSinLookupTable {
|
||||
lookup_table: CyclingFloatLookupTable::new(keys, values, T::zero(), upper_bound),
|
||||
}
|
||||
}
|
||||
|
||||
#[allow(dead_code)]
|
||||
pub fn lookup(&self, key: T) -> T {
|
||||
if key < T::zero() {
|
||||
-self.lookup(-key)
|
||||
} else if key < T::FRAC_PI_2() {
|
||||
self.lookup_table.lookup(key)
|
||||
} else if key < T::PI() {
|
||||
self.lookup_table.lookup(T::PI() - key)
|
||||
} else {
|
||||
-self.lookup(key - T::PI())
|
||||
}
|
||||
}
|
||||
}
|
||||
|
||||
|
||||
#[derive(Default, Debug, Serialize, Deserialize, Clone)]
|
||||
pub struct EndoCosLookupTable<T>
|
||||
where
|
||||
T: Float + FloatConst + Signed + Sub<Output = T> + Rem<Output = T> + NumCast + From<u8>,
|
||||
{
|
||||
lookup_table: EndoSinLookupTable<T>,
|
||||
}
|
||||
impl<T> EndoCosLookupTable<T>
|
||||
where
|
||||
T: Float + FloatConst + Signed + Sub<Output = T> + Rem<Output = T> + NumCast + From<u8>,
|
||||
{
|
||||
pub fn new(precision: usize) -> Self {
|
||||
EndoCosLookupTable {
|
||||
lookup_table: EndoSinLookupTable::new(precision),
|
||||
}
|
||||
}
|
||||
|
||||
#[allow(dead_code)]
|
||||
pub fn lookup(&self, key: T) -> T {
|
||||
self.lookup_table.lookup(key + T::FRAC_PI_2())
|
||||
}
|
||||
}
|
||||
4
src/lookup/mod.rs
Normal file
4
src/lookup/mod.rs
Normal file
@@ -0,0 +1,4 @@
|
||||
mod const_tables;
|
||||
pub mod lookup_table;
|
||||
|
||||
pub use const_tables::*;
|
||||
156
src/tests.rs
Normal file
156
src/tests.rs
Normal file
@@ -0,0 +1,156 @@
|
||||
//tests.rs
|
||||
|
||||
use num_traits::Float;
|
||||
|
||||
fn calculate_percentage_error<T>(vector1: &[T], vector2: &[T]) -> T
|
||||
where T: Float + std::ops::AddAssign,
|
||||
{
|
||||
let n = vector1.len();
|
||||
assert_eq!(n, vector2.len(), "Vectors must have equal lengths.");
|
||||
|
||||
let mut total_error = T::zero();
|
||||
for i in 0..n {
|
||||
let diff = (vector1[i] - vector2[i]).abs();
|
||||
let error = diff / if vector1[i] == T::zero() { T::min_positive_value() } else { vector1[i] };
|
||||
total_error += error;
|
||||
}
|
||||
|
||||
let average_error = total_error / T::from(n).unwrap();
|
||||
let percentage_error = average_error * T::from(100).expect("Cannot convert 100 to type T");
|
||||
percentage_error
|
||||
}
|
||||
|
||||
#[cfg(test)]
|
||||
mod tests {
|
||||
mod f64_error {
|
||||
use crate::*;
|
||||
use super::super::calculate_percentage_error;
|
||||
use once_cell::sync::Lazy;
|
||||
|
||||
const TOLERANCE: f64 = 2.5;
|
||||
|
||||
static X: Lazy<Vec<f64>> = Lazy::new(|| {
|
||||
(-10000..10000)
|
||||
.map(|a| (a as f64) / 1000.)
|
||||
.collect::<Vec<f64>>()
|
||||
});
|
||||
|
||||
#[test]
|
||||
fn pow2() -> Result<(), Box<dyn std::error::Error>> {
|
||||
let percentage_error = calculate_percentage_error(
|
||||
&X.iter().map(|&x| x.fast_pow2()).collect::<Vec<f64>>(),
|
||||
&X.iter().map(|&x| x.powi(2)).collect::<Vec<f64>>()
|
||||
);
|
||||
assert!(!percentage_error.is_nan(), "fast_pow2<f64> percentage error is NaN");
|
||||
assert!(
|
||||
percentage_error < TOLERANCE,
|
||||
"fast_pow2<f64> percentage error: {0}",
|
||||
percentage_error
|
||||
);
|
||||
Ok(())
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn exp() -> Result<(), Box<dyn std::error::Error>> {
|
||||
let percentage_error = calculate_percentage_error(
|
||||
&X.iter().map(|&x| x.fast_exp()).collect::<Vec<f64>>(),
|
||||
&X.iter().map(|&x| x.exp()).collect::<Vec<f64>>()
|
||||
);
|
||||
assert!(!percentage_error.is_nan(), "fast_exp<f64> percentage error is NaN");
|
||||
assert!(
|
||||
percentage_error < TOLERANCE,
|
||||
"fast_exp<f64> percentage error: {0}",
|
||||
percentage_error
|
||||
);
|
||||
Ok(())
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn cos() -> Result<(), Box<dyn std::error::Error>> {
|
||||
let percentage_error = calculate_percentage_error(
|
||||
&X.iter().map(|&x| x.fast_cos()).collect::<Vec<f64>>(),
|
||||
&X.iter().map(|&x| x.cos()).collect::<Vec<f64>>()
|
||||
);
|
||||
assert!(!percentage_error.is_nan(), "fast_cos<f64> percentage error is NaN");
|
||||
assert!(
|
||||
percentage_error < TOLERANCE,
|
||||
"fast_cos<f64> percentage error: {0}",
|
||||
percentage_error
|
||||
);
|
||||
Ok(())
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn sigmoid() -> Result<(), Box<dyn std::error::Error>> {
|
||||
let percentage_error = calculate_percentage_error(
|
||||
&X.iter().map(|&x| x.fast_sigmoid()).collect::<Vec<f64>>(),
|
||||
&X.iter().map(|&x| sigmoid_builtin_f64(x)).collect::<Vec<f64>>()
|
||||
);
|
||||
assert!(!percentage_error.is_nan(), "fast_sigmoid<f64> percentage error is NaN");
|
||||
assert!(
|
||||
percentage_error < TOLERANCE,
|
||||
"fast_sigmoid<f64> percentage error: {0}",
|
||||
percentage_error
|
||||
);
|
||||
Ok(())
|
||||
}
|
||||
}
|
||||
|
||||
mod f32_error {
|
||||
use crate::*;
|
||||
use super::super::calculate_percentage_error;
|
||||
use once_cell::sync::Lazy;
|
||||
|
||||
const TOLERANCE: f32 = 2.5;
|
||||
|
||||
static X: Lazy<Vec<f32>> = Lazy::new(|| {
|
||||
(-10000..10000)
|
||||
.map(|a| (a as f32) / 1000.)
|
||||
.collect::<Vec<f32>>()
|
||||
});
|
||||
|
||||
#[test]
|
||||
fn pow2() -> Result<(), Box<dyn std::error::Error>> {
|
||||
assert!(
|
||||
calculate_percentage_error(
|
||||
&X.iter().map(|&x| x.fast_pow2()).collect::<Vec<f32>>(),
|
||||
&X.iter().map(|&x| x.powi(2)).collect::<Vec<f32>>()
|
||||
) < TOLERANCE
|
||||
);
|
||||
Ok(())
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn exp() -> Result<(), Box<dyn std::error::Error>> {
|
||||
assert!(
|
||||
calculate_percentage_error(
|
||||
&X.iter().map(|&x| x.fast_exp()).collect::<Vec<f32>>(),
|
||||
&X.iter().map(|&x| x.exp()).collect::<Vec<f32>>()
|
||||
) < TOLERANCE
|
||||
);
|
||||
Ok(())
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn cos() -> Result<(), Box<dyn std::error::Error>> {
|
||||
assert!(
|
||||
calculate_percentage_error(
|
||||
&X.iter().map(|&x| x.fast_cos()).collect::<Vec<f32>>(),
|
||||
&X.iter().map(|&x| x.cos()).collect::<Vec<f32>>()
|
||||
) < TOLERANCE
|
||||
);
|
||||
Ok(())
|
||||
}
|
||||
|
||||
#[test]
|
||||
fn sigmoid() -> Result<(), Box<dyn std::error::Error>> {
|
||||
assert!(
|
||||
calculate_percentage_error(
|
||||
&X.iter().map(|&x| x.fast_sigmoid()).collect::<Vec<f32>>(),
|
||||
&X.iter().map(|&x| sigmoid_builtin_f32(x)).collect::<Vec<f32>>()
|
||||
) < TOLERANCE
|
||||
);
|
||||
Ok(())
|
||||
}
|
||||
}
|
||||
}
|
||||
Reference in New Issue
Block a user