From dc0db358abe19481e475e10c32149b53370f1a1c Mon Sep 17 00:00:00 2001 From: Daniel Baumann Date: Thu, 30 May 2024 05:57:31 +0200 Subject: Merging upstream version 1.72.1+dfsg1. Signed-off-by: Daniel Baumann --- vendor/compiler_builtins/.cargo-checksum.json | 2 +- vendor/compiler_builtins/Cargo.lock | 2 +- vendor/compiler_builtins/Cargo.toml | 2 +- vendor/compiler_builtins/README.md | 4 +- vendor/compiler_builtins/build.rs | 61 +- vendor/compiler_builtins/examples/intrinsics.rs | 1 + vendor/compiler_builtins/src/aarch64_linux.rs | 277 ++++++++ vendor/compiler_builtins/src/float/cmp.rs | 14 + vendor/compiler_builtins/src/float/conv.rs | 2 +- vendor/compiler_builtins/src/float/div.rs | 857 ++++++++++++++++++------ vendor/compiler_builtins/src/lib.rs | 4 + vendor/compiler_builtins/src/macros.rs | 20 +- vendor/compiler_builtins/src/math.rs | 5 +- 13 files changed, 1032 insertions(+), 219 deletions(-) create mode 100644 vendor/compiler_builtins/src/aarch64_linux.rs (limited to 'vendor/compiler_builtins') diff --git a/vendor/compiler_builtins/.cargo-checksum.json b/vendor/compiler_builtins/.cargo-checksum.json index 7f15b6576..82891782c 100644 --- a/vendor/compiler_builtins/.cargo-checksum.json +++ b/vendor/compiler_builtins/.cargo-checksum.json @@ -1 +1 @@ -{"files":{"Cargo.lock":"9ecf79f50992d034b2d83daf32ae512b7175916fe9a4b02d0e9b230543cd263c","Cargo.toml":"a5ba02ac8045d9d01708808f8a1d5a74ee82927d2f1c26ada47af280a05da88d","LICENSE.txt":"0e13fed90654e0bc677d624a2d770833a09541fe0c0bdb3d051b3d081207393a","README.md":"5eb36fbab30693dbbe9f0de54749c95bd06fd6e42013b5b9eff3c062b9fdd34f","build.rs":"45ab03c8f5369df7579abb8adc2554a892f415187a384c85832e75abe898d357","examples/intrinsics.rs":"a7aa69c17af3aa8f6edff32c214e80827d3cbe3aea386a2be42244444752d253","libm/src/math/acos.rs":"fb066ba84aba1372d706425ec14f35ff8d971756d15eeebd22ecf42a716493bb","libm/src/math/acosf.rs":"a112b82309bba1d35c4e3d6ad4d6c21ef305343d9ab601ddf4bc61d43bc9f1af","libm/src/math/acosh.rs":"99de01ded7922bb93a882ad5ad8b472b5cae0059dea0bdca2077f65e94483150","libm/src/math/acoshf.rs":"10750c4d39ef6717b20a15ef1ce43e15eb851682d2f820f7e94501adec98b9a5","libm/src/math/asin.rs":"095a1e98996daff45df0b154ca0ec35bbf31db964ee9fdda0207308cb20df441","libm/src/math/asinf.rs":"49cccb4db2881982643a4a7d5453f4f8daf527711bbb67313607a3c178856d61","libm/src/math/asinh.rs":"4dd51affa71cce34a192ad66154e248f8d1c4b40fb497f29052333e425bb740f","libm/src/math/asinhf.rs":"914bfecf449f5e2bce786aa12c056d419073c6011d41c1bab7c39ba765fa4c53","libm/src/math/atan.rs":"d4fe46e1c5739dd09997869dcfbc3c85f03c534af52e700d6c6bcf9c3fedda07","libm/src/math/atan2.rs":"2623bc8ca707d13a7092ce49adf68e9cbf4452ad1bf4a861dc40ca858606a747","libm/src/math/atan2f.rs":"dd01943e0e1f1955912e5c3ffc9467529cf64bd02ac0a6ad5ab31dbe6657f05d","libm/src/math/atanf.rs":"e41b41569474a59c970ede3538e00bda4072cf4d90040017101cc79d7dc28caa","libm/src/math/atanh.rs":"57a8fb3f0f116fa4a966ac6bc2abd5f80236ead8e79013f468bd3786921f7110","libm/src/math/atanhf.rs":"6f2e57aaec1b5fc7609cb3938b3d155f51b4237dbda530739c34a0448cd9beb9","libm/src/math/cbrt.rs":"f2c45612d2eecd93cfcdd9ebf824c754fc8f8dfd6d16862c0b9c4ccea78c2a0f","libm/src/math/cbrtf.rs":"ad0b483854aa9f17a44d36c049bf0e8ebab34c27e90b787c05f45cc230ec7d19","libm/src/math/ceil.rs":"57ba5b6e207a0ccbd34190d1aa544389ca12126be23821dfb5746497f620ce03","libm/src/math/ceilf.rs":"c922a0475a599b9ea5473e615f74700b99707cebd6927f24ea59cb2a3cb3bbc3","libm/src/math/copysign.rs":"8b6440a251f0f1509d87f18122f74d0d5c03d0b60517e89e441434a3c5d84591","libm/src/math/copysignf.rs":"87d35436d224852ada93a2e93f6730cf1a727b808dd10e7d49ab4585866e336b","libm/src/math/cos.rs":"74babdc13ede78e400c5ca1854c3e22d2e08cbdc5618aefa5bba6f9303ef65b6","libm/src/math/cosf.rs":"09c40f93c445b741e22477ceedf163ca33b6a47f973f7c9876cfba2692edb29c","libm/src/math/cosh.rs":"0d0a7cef18577f321996b8b87561963139f754ad7f2ea0a3b3883811f3f0693a","libm/src/math/coshf.rs":"be8ca8739e4cf1978425b349f941cb4838bba8c10cb559c7940b9fd4fdde21ad","libm/src/math/erf.rs":"de69e6669ce1014e5b5086a7a6d01c4755f2f0590e204d2a77bea455764114f7","libm/src/math/erff.rs":"6acdbb07f74296067bb0380b850918cfb5806a89f9ff04352a7a0b921d728944","libm/src/math/exp.rs":"ca7405ad0d1993fffcf9aae96f9256307bed3c4916545aaebd1cf1d2df1807fa","libm/src/math/exp10.rs":"2e136c6ecedd8e57a6c31796f57fae4546fcfd8bc6be66c836f553df9c74b907","libm/src/math/exp10f.rs":"9a3ce506ec587066a355ab74e0eb69a03a214ac405718087ae9772365050b20b","libm/src/math/exp2.rs":"94a9304a2ce3bc81f6d2aefd3cde6faa30f13260d46cb13692863cdea1c9a3a1","libm/src/math/exp2f.rs":"785f2630accd35118ec07bf60273e219ed91a215b956b1552eeea5bc2a708cc8","libm/src/math/expf.rs":"ec14c18f891a9e37735ec39e6fc2e9bf674a2c2e083f22e2533b481177359c98","libm/src/math/expm1.rs":"124069f456c8ad331f265c7509d9e223b2a300e461bbfd3d6adfdcdd2ee5b8ac","libm/src/math/expm1f.rs":"18e2116d31ea8410051cc709b9d04b754b0e3ba6758ee1bf0b48749f4999b840","libm/src/math/expo2.rs":"4f4f9fecfccb43f30c2784aa7c0bb656754a52b8ab431f7d1b551c673ab133f1","libm/src/math/fabs.rs":"e6c7db39f98508098cdf64ac0c2f53866c466149a7490afb9fe22b44c4dd81b3","libm/src/math/fabsf.rs":"83a1f5f4d9ca899ba2b701d7332e18b40258b83e111db4c5d8fab2cc1be58aa3","libm/src/math/fdim.rs":"8ec091996005207297c2389ae563e1b18dbc6a9eac951de29a976c5cd7bc32a7","libm/src/math/fdimf.rs":"c7f3f2269834d55be26b6580ddc07c42531577955fa4de35bad1e2a361085614","libm/src/math/fenv.rs":"916ae11e4763588518d64dee82afb41be9d1ee38ecc0679c821d4e7e22cd3dc5","libm/src/math/floor.rs":"5050804cae173af6775c0678d6c1aafb5ca2b744bc8a2f50d9d03b95dcee1fb0","libm/src/math/floorf.rs":"c903e0c57bc60a888c513eb7a873a87a4759ba68fc791b6b931652f8ee74cc03","libm/src/math/fma.rs":"d87963472cd5bfcb83eb4010c67f3653857cf28f11378e06d63abae14c723e5d","libm/src/math/fmaf.rs":"1db6ee0d47ddbdb441cfe167edf89b431239f5805708fd0376cf5c01349a4bd6","libm/src/math/fmax.rs":"f6c8e96a8b1a170648d2fa3513e7b6b459085d708c839869f82e305fe58fac37","libm/src/math/fmaxf.rs":"dff0025433232e8a5ec7bd54d847ccf596d762ea4e35f5c54fbaac9404d732fd","libm/src/math/fmin.rs":"95b6cb66ca0e0e22276f0bf88dbe8fb69796a69a196a7491bd4802efbcf2e298","libm/src/math/fminf.rs":"304bc839b15ea3d84e68d2af9f40524ec120d30a36a667b22fcb98a6c258f4c7","libm/src/math/fmod.rs":"a1c0550fc7df8164733d914e222ff0966a2ab886d6e75a1098f24fe0283ae227","libm/src/math/fmodf.rs":"ee51ed092c0eeb8195f35735ff725cfd46612e0d689a7c483538bd92fbe61828","libm/src/math/frexp.rs":"28af70026922a8ab979744c7ad4d8faba6079c4743b7eeb6d14c983a982fbbcc","libm/src/math/frexpf.rs":"2e2593ae8002ba420809ebfaf737ef001cdc912354be3d978a8c0cb930350d4d","libm/src/math/hypot.rs":"841131c4a0cea75bc8a86e29f3f6d0815a61fc99731c9984651ce83d3050d218","libm/src/math/hypotf.rs":"5f317323edc2eb699580fe54b074b7e570a7734d51a0a149c0b49b54470a836c","libm/src/math/ilogb.rs":"d178ad7ca3439f82d565962b143f20448e45b2e2c51357b127abaec683297e32","libm/src/math/ilogbf.rs":"00f2b1b0496e21c6a42d68aea74d7156fa2ff0a735741b9051f3ca1cf0f57586","libm/src/math/j0.rs":"9572b6396c489927d332d0e717920e61ec0618e5e9c31f7eeeec70f5e4abab06","libm/src/math/j0f.rs":"802c8254bded9b3afb6eea8b9af240038a5a4a5d811396729f69ca509e3e7d87","libm/src/math/j1.rs":"97b1af1611fa3d110c2b349ee8e4176100132ea1391b619086b47ac063b81803","libm/src/math/j1f.rs":"9c9b128752e8ea2e7d81b637ba84907ab54a545e7602c49167b313743927930b","libm/src/math/jn.rs":"847d122334e5707ad9627146cddccc082a1f2f5bcd3e5ef54399013a7007ce88","libm/src/math/jnf.rs":"4045076f7d1a1b89882ed60d4dd60a4cbbc66b85cfb90491378c8015effcc476","libm/src/math/k_cos.rs":"f34a69e44d6b8901b03b578a75972f438ab20a7b98a0903fc1903d6fde3899be","libm/src/math/k_cosf.rs":"8f7117ff21cebf8e890a5bcfd7ea858a94172f4172b79a66d53824c2cb0888b1","libm/src/math/k_expo2.rs":"eb4ca9e6a525b7ea6da868c3cb136896682cc46f8396ba2a2ebc3ae9e9ba54b0","libm/src/math/k_expo2f.rs":"d51ad5df61cb5d1258bdb90c52bfed4572bb446a9337de9c04411ed9454ae0cb","libm/src/math/k_sin.rs":"14b2aba6ca07150c92768b5a72acaf5cde6a11d6619e14896512a7ba242e289a","libm/src/math/k_sinf.rs":"2775fcc710807164e6f37a4f8da3c8143cd5f16e19ce7c31c5591522151d7a96","libm/src/math/k_tan.rs":"a72beae4ccd9631eeeb61d6365bbeecae81c8411f3120a999c515cca0d5ea5c5","libm/src/math/k_tanf.rs":"6a794be56fa4b2f60452b9bab19af01c388f174560acbf829a351378ea39495d","libm/src/math/ldexp.rs":"b647f0096e80e4d926d8dd18d294c892ee2cb1778effe2c5e1b2664ae5cb1a4e","libm/src/math/ldexpf.rs":"98743fad2cd97a7be496f40ba3157ac1438fce0d0c25d5ab90c3b8c71c3fd0ed","libm/src/math/lgamma.rs":"0edd18e4f96bfcbe8b1b5af3eeca5208cd6d2d479dfa5ad117c9dfeccecf614f","libm/src/math/lgamma_r.rs":"f44a37aeccd56559ef784ae8edf217d14ad5cc2d910f0a65e70ffc86d7dc23dd","libm/src/math/lgammaf.rs":"967845357758b868a571857ec001f9f9154001110b8e97c08b6d10586bed9c49","libm/src/math/lgammaf_r.rs":"7143016d60e11fa235d53968125e57231b1104ce52149b5e1eed39629e0d1ff0","libm/src/math/log.rs":"b5e0c5f30d9e94351488732801be3107c12b854c3f95ad37e256dd88eeca408f","libm/src/math/log10.rs":"3425ff8be001fd1646ba15e254eb6ef4bdc6ccaf0cbee27ddf1fa84e04178b90","libm/src/math/log10f.rs":"fee4f71879bc4c99259e68c0c641364901629fb29a8ebddfcc0d090102cceddd","libm/src/math/log1p.rs":"9cf400852f165e6be19b97036ae9521fb9ca857d0a9a91c117d9123221622185","libm/src/math/log1pf.rs":"2716e6d2afa271996b7c8f47fd9e4952c88f4c1fd8c07c3e8ce8c62794bf71d8","libm/src/math/log2.rs":"dbbbfbaaa8aa6a4dbefea554ea3983090a9691228b011910c751f6adca912c40","libm/src/math/log2f.rs":"92a90350d8edce21c31c285c3e620fca7c62a2366008921715945c2c73b5b79f","libm/src/math/logf.rs":"845342cffc34d3db1f5ec12d8e5b773cd5a79056e28662fcb9bcd80207596f50","libm/src/math/mod.rs":"d694260529d51d0bc17f88ad557d852b9bb0bc3f7466cf7f62b679dc95ebba42","libm/src/math/modf.rs":"d012ed5a708ef52b6d1313c22a46cadaf5764dde1220816e3df2f03a0fcc60ae","libm/src/math/modff.rs":"f8f1e4c27a85d2cdb3c8e74439d59ef64aa543b948f22c23227d02d8388d61c2","libm/src/math/nextafter.rs":"3282e7eef214a32736fb6928d490198ad394b26b402b45495115b104839eebfe","libm/src/math/nextafterf.rs":"0937dc8a8155c19842c12181e741cec1f7df1f7a00cee81fcb2475e2842761b7","libm/src/math/pow.rs":"17c38297c5bf99accd915f292b777f8716ecf328916297c8bb9dfde6fd8ce522","libm/src/math/powf.rs":"2c423a0ea57fdc4e20f3533f744c6e6288c998b4de8f2914fafaa0e78be81b04","libm/src/math/rem_pio2.rs":"3e53234977daf61c89c29c940791714aad2f676a6f38188c7d17543a2aa8806f","libm/src/math/rem_pio2_large.rs":"482f31ff4e4eacf885f6130ae26a1d59f76b382059d6c742f30e5036811d3ca8","libm/src/math/rem_pio2f.rs":"07fb48f6d5cbadfd32ce4124b2b74af98b8391a2a6f36ce2a7d32e4500cb65ac","libm/src/math/remainder.rs":"63865f4370853c476b45bb27a5c54a4072146aa4a626835ae5263871a4e7e5dc","libm/src/math/remainderf.rs":"dd3fa432dbda8f2135428198be7bd69c57f8d13df3f365b12f52bf6a82352ac4","libm/src/math/remquo.rs":"3cc0bf55069f165c4843f2c358b3a27279c01e8cdd99f9057a3f7f31f45408f2","libm/src/math/remquof.rs":"cc749e18ecb7e766b8b8eeabdbf89ac99087d3d587e71e30f690676a3d2c1f9b","libm/src/math/rint.rs":"2c17047bcfd0ccdca8669f7cf70c628154ae4abc142660f30e37f9c073928706","libm/src/math/rintf.rs":"3b54af9eaa1bb6808159ca435246acf6a4e7aebbc344e3f4a4c5636345155897","libm/src/math/round.rs":"f10797ef15dd34a74e912ba8621d60bc0200c87b94308c9de3cc88d7aec4feb4","libm/src/math/roundf.rs":"27e37cfcf82373709e7debf9c0c18f7ed00ae0f5d97a214c388041f7a6996d35","libm/src/math/scalbn.rs":"b5c9d6d4177fe393cbfe1c634d75ce14b754f6cbce87c5bf979a9661491748a2","libm/src/math/scalbnf.rs":"4f198d06db1896386256fb9a5ac5b805b16b836226c18780a475cf18d7c1449c","libm/src/math/sin.rs":"bb483a2138ca779e03a191222636f0c60fd75a77a2a12f263bda4b6aa9136317","libm/src/math/sincos.rs":"1cf62a16c215e367f51078a3ba23a3f257682032a8f3c657293029a886b18d82","libm/src/math/sincosf.rs":"b0f589e6ada8215944d7784f420c6721c90387d799e349ce7676674f3c475e75","libm/src/math/sinf.rs":"dcddac1d56b084cbb8d0e019433c9c5fe2201d9b257a7dcf2f85c9a8f14b79cf","libm/src/math/sinh.rs":"d8ee4c7af883a526f36c1a6da13bb81fba9181b477e2f2538161a2bee97edc35","libm/src/math/sinhf.rs":"d06eb030ba9dbf7094df127262bfe99f149b4db49fa8ab8c15499660f1e46b26","libm/src/math/sqrt.rs":"5f3a0a582b174fcfccb9c5274899cb05b664ccb92bf1d42caa58890947b68256","libm/src/math/sqrtf.rs":"da926ac27af6eecdf8b62d8baeefcfe1627110592e44298f6b7f05b7ac12fe7e","libm/src/math/tan.rs":"930ecedaadc60f704c2dfa4e15186f59713c1ba7d948529d215223b424827db5","libm/src/math/tanf.rs":"894156a3b107aee08461eb4e7e412fc049aa237d176ae705c6e3e2d7060d94e3","libm/src/math/tanh.rs":"f1f08eb98ed959a17370a7aaf0177be36e3764543424e78feb033ed3f5e8ec98","libm/src/math/tanhf.rs":"74027b0c672a4e64bdef6d7a3069b90caec50e1e7dbb2c12d2828f310502f41e","libm/src/math/tgamma.rs":"c889cfa49bbeb4dbb0941fe9fac3b4da7d5879dcf04a3b9bb6e56de529baf374","libm/src/math/tgammaf.rs":"0737b34777095d0e4d07fe533e8f105082dd4e8ece411bba6ae5993b45b9388c","libm/src/math/trunc.rs":"642264897cc1505e720c8cf313be81aa9fd53aae866644a2e988d01dbc77fd8a","libm/src/math/truncf.rs":"dee3607baf1af0f01deae46e429e097234c50b268eaefebbe716f19f38597900","src/arm.rs":"3a7a2e21de7f475fbc2765109a18a1301bfb5c7be361c8344dde5ec23e8d7ace","src/arm_linux.rs":"35a4cb7b75015543feb15b0c692da0faf0e6037d3b97a4a18067ba416eae1a70","src/float/add.rs":"3ec32ceaf470a89777b54f9cde61832fdadeade0f4894f268a949e968520bc57","src/float/cmp.rs":"79b1fdc8d5f943c4ad5ea4ad32623b18f63e17ac3852fbc64a4942228007e1fc","src/float/conv.rs":"d95b386e483d2bc77b2d5c41b62d01a8cc791fb3fb18ce97317947ecd5a3c02b","src/float/div.rs":"fe21115ecb1b3330569fd85cb51c650bf80683f152333db988d8e0d564a9ae11","src/float/extend.rs":"180b2e791c58e0526de0a798845c580ce3222c8a15c8665e6e6a4bf5cf1a34aa","src/float/mod.rs":"a91cf65abb6e715c5559e3e4bd87a69cd99a9552d54804d0b7137c02c513f158","src/float/mul.rs":"0d0c1f0c28c149ecadeafd459d3c4c9327e4cfcae2cba479957bb8010ef51a01","src/float/pow.rs":"2ada190738731eb6f24104f8fb8c4d6f03cfb16451536dbee32f2b33db0c4b19","src/float/sub.rs":"c2a87f4628f51d5d908d0f25b5d51ce0599dc559d5a72b20e131261f484d5848","src/float/trunc.rs":"d21d2a2f9a1918b4bbb594691e397972a7c04b74b2acf04016c55693abf6d24b","src/int/addsub.rs":"7ec45ce1ba15b56a5b7129d3e5722c4db764c6545306d3fa9090983bcabd6f17","src/int/leading_zeros.rs":"ccf5e9d098c80034dcf6e38437c9a2eb670fa8043558bbfb574f2293164729a6","src/int/mod.rs":"bab1b77535ceebdebb89fd4e59e7105f8c45347bb638351a626615b24544a0b1","src/int/mul.rs":"bb48d8fd42d8f9f5fe9271d8d0f7a92dbae320bf4346e19d1071eb2093cb8ed9","src/int/sdiv.rs":"ace4cb0ec388a38834e01cab2c5bc87182d31588dfc0b1ae117c11ed0c4781cf","src/int/shift.rs":"ff1e0dab608f2b9b3c68c7dfabd1e9bf6500518f5926a9ce4c1d84178bfe19f4","src/int/specialized_div_rem/asymmetric.rs":"27f5bf70a35109f9d4e4e1ad1e8003aa17da5a1e436bf3e63a493d7528a3a566","src/int/specialized_div_rem/binary_long.rs":"9f1ced81a394f000a21a329683144d68ee431a954136a3634eb55b1ee2cf6d51","src/int/specialized_div_rem/delegate.rs":"9df141af98e391361e25d71ae38d5e845a91d896edd2c041132fd46af8268e85","src/int/specialized_div_rem/mod.rs":"73c98b9f69cc9b101ae4c9081e82d66af1df4a58cf0c9bb2a8c8659265687f12","src/int/specialized_div_rem/norm_shift.rs":"3be7ee0dea545c1f702d9daf67dad2b624bf7b17b075c8b90d3d3f7b53df4c21","src/int/specialized_div_rem/trifecta.rs":"87eef69da255b809fd710b14f2eb3f9f59e3dac625f8564ebc8ba78f9763523b","src/int/udiv.rs":"3732b490a472505411577f008b92f489287745968ce6791665201201377d3475","src/lib.rs":"bbdb98c20ac43df24dcde957d8cfb066abdf09b7d62eb893f4d978fb8ab5332a","src/macros.rs":"85a9c368671803a7cd46365173cf4999b00f8f2b2da4a8834015e98cd3429eed","src/math.rs":"840f609cd634b8ffc05c0ad4d8f0b3ce35d7d4c7e7f009bbadeaa6e0d46c873d","src/mem/impls.rs":"8b389b9aeb43dd55351a86abd4b5fc311f7161e1a4023ca3c5a4c57b49650881","src/mem/mod.rs":"714763d045a20e0a68c04f929d14fb3d7b28662dda4a2622970416642af833dc","src/mem/x86_64.rs":"2f29fb392086b3f7e2e78fcfcbf0f0e205822eb4599f1bdf93e41833e1bd2766","src/probestack.rs":"ef5c07e9b95de7b2b77a937789fcfefd9846274317489ad6d623e377c9888601","src/riscv.rs":"50ddd6c732a9f810ab6e15a97b22fdc94cfc1dea09c45d87c833937f9206bee0","src/x86.rs":"117b50d6725ee0af0a7b3d197ea580655561f66a870ebc450d96af22bf7f39f6","src/x86_64.rs":"4f16bc9fad7757d48a6da3a078c715dd3a22154aadb4f1998d4c1b5d91396f9e"},"package":"64518f1ae689f74db058bbfb3238dfe6eb53f59f4ae712f1ff4348628522e190"} \ No newline at end of file +{"files":{"Cargo.lock":"a93bdfef36f65551ed4fd8a99679c6c9a7661dd3b9236f90a2a1a38f65d93e44","Cargo.toml":"14821c5a0ba22dfb7450dbcccd28d6a864120db28bad293845da3d8c4bab8899","LICENSE.txt":"0e13fed90654e0bc677d624a2d770833a09541fe0c0bdb3d051b3d081207393a","README.md":"693b529db2e5dd13069c05ef2e76e26c1fe1b2a590b07cd8e26dfb2df087ba62","build.rs":"3f2f9589f896c5dc7e7ce39f5809f4f63d1199e5f80ae23b343a7f1d889d0206","examples/intrinsics.rs":"cdf17d36ed38e703e954ccb0d6626037afcef9b036b4a12b1f61b729ca21079a","libm/src/math/acos.rs":"fb066ba84aba1372d706425ec14f35ff8d971756d15eeebd22ecf42a716493bb","libm/src/math/acosf.rs":"a112b82309bba1d35c4e3d6ad4d6c21ef305343d9ab601ddf4bc61d43bc9f1af","libm/src/math/acosh.rs":"99de01ded7922bb93a882ad5ad8b472b5cae0059dea0bdca2077f65e94483150","libm/src/math/acoshf.rs":"10750c4d39ef6717b20a15ef1ce43e15eb851682d2f820f7e94501adec98b9a5","libm/src/math/asin.rs":"095a1e98996daff45df0b154ca0ec35bbf31db964ee9fdda0207308cb20df441","libm/src/math/asinf.rs":"49cccb4db2881982643a4a7d5453f4f8daf527711bbb67313607a3c178856d61","libm/src/math/asinh.rs":"4dd51affa71cce34a192ad66154e248f8d1c4b40fb497f29052333e425bb740f","libm/src/math/asinhf.rs":"914bfecf449f5e2bce786aa12c056d419073c6011d41c1bab7c39ba765fa4c53","libm/src/math/atan.rs":"d4fe46e1c5739dd09997869dcfbc3c85f03c534af52e700d6c6bcf9c3fedda07","libm/src/math/atan2.rs":"2623bc8ca707d13a7092ce49adf68e9cbf4452ad1bf4a861dc40ca858606a747","libm/src/math/atan2f.rs":"dd01943e0e1f1955912e5c3ffc9467529cf64bd02ac0a6ad5ab31dbe6657f05d","libm/src/math/atanf.rs":"e41b41569474a59c970ede3538e00bda4072cf4d90040017101cc79d7dc28caa","libm/src/math/atanh.rs":"57a8fb3f0f116fa4a966ac6bc2abd5f80236ead8e79013f468bd3786921f7110","libm/src/math/atanhf.rs":"6f2e57aaec1b5fc7609cb3938b3d155f51b4237dbda530739c34a0448cd9beb9","libm/src/math/cbrt.rs":"f2c45612d2eecd93cfcdd9ebf824c754fc8f8dfd6d16862c0b9c4ccea78c2a0f","libm/src/math/cbrtf.rs":"ad0b483854aa9f17a44d36c049bf0e8ebab34c27e90b787c05f45cc230ec7d19","libm/src/math/ceil.rs":"57ba5b6e207a0ccbd34190d1aa544389ca12126be23821dfb5746497f620ce03","libm/src/math/ceilf.rs":"c922a0475a599b9ea5473e615f74700b99707cebd6927f24ea59cb2a3cb3bbc3","libm/src/math/copysign.rs":"8b6440a251f0f1509d87f18122f74d0d5c03d0b60517e89e441434a3c5d84591","libm/src/math/copysignf.rs":"87d35436d224852ada93a2e93f6730cf1a727b808dd10e7d49ab4585866e336b","libm/src/math/cos.rs":"74babdc13ede78e400c5ca1854c3e22d2e08cbdc5618aefa5bba6f9303ef65b6","libm/src/math/cosf.rs":"09c40f93c445b741e22477ceedf163ca33b6a47f973f7c9876cfba2692edb29c","libm/src/math/cosh.rs":"0d0a7cef18577f321996b8b87561963139f754ad7f2ea0a3b3883811f3f0693a","libm/src/math/coshf.rs":"be8ca8739e4cf1978425b349f941cb4838bba8c10cb559c7940b9fd4fdde21ad","libm/src/math/erf.rs":"de69e6669ce1014e5b5086a7a6d01c4755f2f0590e204d2a77bea455764114f7","libm/src/math/erff.rs":"6acdbb07f74296067bb0380b850918cfb5806a89f9ff04352a7a0b921d728944","libm/src/math/exp.rs":"ca7405ad0d1993fffcf9aae96f9256307bed3c4916545aaebd1cf1d2df1807fa","libm/src/math/exp10.rs":"2e136c6ecedd8e57a6c31796f57fae4546fcfd8bc6be66c836f553df9c74b907","libm/src/math/exp10f.rs":"9a3ce506ec587066a355ab74e0eb69a03a214ac405718087ae9772365050b20b","libm/src/math/exp2.rs":"94a9304a2ce3bc81f6d2aefd3cde6faa30f13260d46cb13692863cdea1c9a3a1","libm/src/math/exp2f.rs":"785f2630accd35118ec07bf60273e219ed91a215b956b1552eeea5bc2a708cc8","libm/src/math/expf.rs":"ec14c18f891a9e37735ec39e6fc2e9bf674a2c2e083f22e2533b481177359c98","libm/src/math/expm1.rs":"124069f456c8ad331f265c7509d9e223b2a300e461bbfd3d6adfdcdd2ee5b8ac","libm/src/math/expm1f.rs":"18e2116d31ea8410051cc709b9d04b754b0e3ba6758ee1bf0b48749f4999b840","libm/src/math/expo2.rs":"4f4f9fecfccb43f30c2784aa7c0bb656754a52b8ab431f7d1b551c673ab133f1","libm/src/math/fabs.rs":"e6c7db39f98508098cdf64ac0c2f53866c466149a7490afb9fe22b44c4dd81b3","libm/src/math/fabsf.rs":"83a1f5f4d9ca899ba2b701d7332e18b40258b83e111db4c5d8fab2cc1be58aa3","libm/src/math/fdim.rs":"8ec091996005207297c2389ae563e1b18dbc6a9eac951de29a976c5cd7bc32a7","libm/src/math/fdimf.rs":"c7f3f2269834d55be26b6580ddc07c42531577955fa4de35bad1e2a361085614","libm/src/math/fenv.rs":"916ae11e4763588518d64dee82afb41be9d1ee38ecc0679c821d4e7e22cd3dc5","libm/src/math/floor.rs":"5050804cae173af6775c0678d6c1aafb5ca2b744bc8a2f50d9d03b95dcee1fb0","libm/src/math/floorf.rs":"c903e0c57bc60a888c513eb7a873a87a4759ba68fc791b6b931652f8ee74cc03","libm/src/math/fma.rs":"d87963472cd5bfcb83eb4010c67f3653857cf28f11378e06d63abae14c723e5d","libm/src/math/fmaf.rs":"1db6ee0d47ddbdb441cfe167edf89b431239f5805708fd0376cf5c01349a4bd6","libm/src/math/fmax.rs":"f6c8e96a8b1a170648d2fa3513e7b6b459085d708c839869f82e305fe58fac37","libm/src/math/fmaxf.rs":"dff0025433232e8a5ec7bd54d847ccf596d762ea4e35f5c54fbaac9404d732fd","libm/src/math/fmin.rs":"95b6cb66ca0e0e22276f0bf88dbe8fb69796a69a196a7491bd4802efbcf2e298","libm/src/math/fminf.rs":"304bc839b15ea3d84e68d2af9f40524ec120d30a36a667b22fcb98a6c258f4c7","libm/src/math/fmod.rs":"a1c0550fc7df8164733d914e222ff0966a2ab886d6e75a1098f24fe0283ae227","libm/src/math/fmodf.rs":"ee51ed092c0eeb8195f35735ff725cfd46612e0d689a7c483538bd92fbe61828","libm/src/math/frexp.rs":"28af70026922a8ab979744c7ad4d8faba6079c4743b7eeb6d14c983a982fbbcc","libm/src/math/frexpf.rs":"2e2593ae8002ba420809ebfaf737ef001cdc912354be3d978a8c0cb930350d4d","libm/src/math/hypot.rs":"841131c4a0cea75bc8a86e29f3f6d0815a61fc99731c9984651ce83d3050d218","libm/src/math/hypotf.rs":"5f317323edc2eb699580fe54b074b7e570a7734d51a0a149c0b49b54470a836c","libm/src/math/ilogb.rs":"d178ad7ca3439f82d565962b143f20448e45b2e2c51357b127abaec683297e32","libm/src/math/ilogbf.rs":"00f2b1b0496e21c6a42d68aea74d7156fa2ff0a735741b9051f3ca1cf0f57586","libm/src/math/j0.rs":"9572b6396c489927d332d0e717920e61ec0618e5e9c31f7eeeec70f5e4abab06","libm/src/math/j0f.rs":"802c8254bded9b3afb6eea8b9af240038a5a4a5d811396729f69ca509e3e7d87","libm/src/math/j1.rs":"97b1af1611fa3d110c2b349ee8e4176100132ea1391b619086b47ac063b81803","libm/src/math/j1f.rs":"9c9b128752e8ea2e7d81b637ba84907ab54a545e7602c49167b313743927930b","libm/src/math/jn.rs":"847d122334e5707ad9627146cddccc082a1f2f5bcd3e5ef54399013a7007ce88","libm/src/math/jnf.rs":"4045076f7d1a1b89882ed60d4dd60a4cbbc66b85cfb90491378c8015effcc476","libm/src/math/k_cos.rs":"f34a69e44d6b8901b03b578a75972f438ab20a7b98a0903fc1903d6fde3899be","libm/src/math/k_cosf.rs":"8f7117ff21cebf8e890a5bcfd7ea858a94172f4172b79a66d53824c2cb0888b1","libm/src/math/k_expo2.rs":"eb4ca9e6a525b7ea6da868c3cb136896682cc46f8396ba2a2ebc3ae9e9ba54b0","libm/src/math/k_expo2f.rs":"d51ad5df61cb5d1258bdb90c52bfed4572bb446a9337de9c04411ed9454ae0cb","libm/src/math/k_sin.rs":"14b2aba6ca07150c92768b5a72acaf5cde6a11d6619e14896512a7ba242e289a","libm/src/math/k_sinf.rs":"2775fcc710807164e6f37a4f8da3c8143cd5f16e19ce7c31c5591522151d7a96","libm/src/math/k_tan.rs":"a72beae4ccd9631eeeb61d6365bbeecae81c8411f3120a999c515cca0d5ea5c5","libm/src/math/k_tanf.rs":"6a794be56fa4b2f60452b9bab19af01c388f174560acbf829a351378ea39495d","libm/src/math/ldexp.rs":"b647f0096e80e4d926d8dd18d294c892ee2cb1778effe2c5e1b2664ae5cb1a4e","libm/src/math/ldexpf.rs":"98743fad2cd97a7be496f40ba3157ac1438fce0d0c25d5ab90c3b8c71c3fd0ed","libm/src/math/lgamma.rs":"0edd18e4f96bfcbe8b1b5af3eeca5208cd6d2d479dfa5ad117c9dfeccecf614f","libm/src/math/lgamma_r.rs":"f44a37aeccd56559ef784ae8edf217d14ad5cc2d910f0a65e70ffc86d7dc23dd","libm/src/math/lgammaf.rs":"967845357758b868a571857ec001f9f9154001110b8e97c08b6d10586bed9c49","libm/src/math/lgammaf_r.rs":"7143016d60e11fa235d53968125e57231b1104ce52149b5e1eed39629e0d1ff0","libm/src/math/log.rs":"b5e0c5f30d9e94351488732801be3107c12b854c3f95ad37e256dd88eeca408f","libm/src/math/log10.rs":"3425ff8be001fd1646ba15e254eb6ef4bdc6ccaf0cbee27ddf1fa84e04178b90","libm/src/math/log10f.rs":"fee4f71879bc4c99259e68c0c641364901629fb29a8ebddfcc0d090102cceddd","libm/src/math/log1p.rs":"9cf400852f165e6be19b97036ae9521fb9ca857d0a9a91c117d9123221622185","libm/src/math/log1pf.rs":"2716e6d2afa271996b7c8f47fd9e4952c88f4c1fd8c07c3e8ce8c62794bf71d8","libm/src/math/log2.rs":"dbbbfbaaa8aa6a4dbefea554ea3983090a9691228b011910c751f6adca912c40","libm/src/math/log2f.rs":"92a90350d8edce21c31c285c3e620fca7c62a2366008921715945c2c73b5b79f","libm/src/math/logf.rs":"845342cffc34d3db1f5ec12d8e5b773cd5a79056e28662fcb9bcd80207596f50","libm/src/math/mod.rs":"d694260529d51d0bc17f88ad557d852b9bb0bc3f7466cf7f62b679dc95ebba42","libm/src/math/modf.rs":"d012ed5a708ef52b6d1313c22a46cadaf5764dde1220816e3df2f03a0fcc60ae","libm/src/math/modff.rs":"f8f1e4c27a85d2cdb3c8e74439d59ef64aa543b948f22c23227d02d8388d61c2","libm/src/math/nextafter.rs":"3282e7eef214a32736fb6928d490198ad394b26b402b45495115b104839eebfe","libm/src/math/nextafterf.rs":"0937dc8a8155c19842c12181e741cec1f7df1f7a00cee81fcb2475e2842761b7","libm/src/math/pow.rs":"17c38297c5bf99accd915f292b777f8716ecf328916297c8bb9dfde6fd8ce522","libm/src/math/powf.rs":"2c423a0ea57fdc4e20f3533f744c6e6288c998b4de8f2914fafaa0e78be81b04","libm/src/math/rem_pio2.rs":"3e53234977daf61c89c29c940791714aad2f676a6f38188c7d17543a2aa8806f","libm/src/math/rem_pio2_large.rs":"482f31ff4e4eacf885f6130ae26a1d59f76b382059d6c742f30e5036811d3ca8","libm/src/math/rem_pio2f.rs":"07fb48f6d5cbadfd32ce4124b2b74af98b8391a2a6f36ce2a7d32e4500cb65ac","libm/src/math/remainder.rs":"63865f4370853c476b45bb27a5c54a4072146aa4a626835ae5263871a4e7e5dc","libm/src/math/remainderf.rs":"dd3fa432dbda8f2135428198be7bd69c57f8d13df3f365b12f52bf6a82352ac4","libm/src/math/remquo.rs":"3cc0bf55069f165c4843f2c358b3a27279c01e8cdd99f9057a3f7f31f45408f2","libm/src/math/remquof.rs":"cc749e18ecb7e766b8b8eeabdbf89ac99087d3d587e71e30f690676a3d2c1f9b","libm/src/math/rint.rs":"2c17047bcfd0ccdca8669f7cf70c628154ae4abc142660f30e37f9c073928706","libm/src/math/rintf.rs":"3b54af9eaa1bb6808159ca435246acf6a4e7aebbc344e3f4a4c5636345155897","libm/src/math/round.rs":"f10797ef15dd34a74e912ba8621d60bc0200c87b94308c9de3cc88d7aec4feb4","libm/src/math/roundf.rs":"27e37cfcf82373709e7debf9c0c18f7ed00ae0f5d97a214c388041f7a6996d35","libm/src/math/scalbn.rs":"b5c9d6d4177fe393cbfe1c634d75ce14b754f6cbce87c5bf979a9661491748a2","libm/src/math/scalbnf.rs":"4f198d06db1896386256fb9a5ac5b805b16b836226c18780a475cf18d7c1449c","libm/src/math/sin.rs":"bb483a2138ca779e03a191222636f0c60fd75a77a2a12f263bda4b6aa9136317","libm/src/math/sincos.rs":"1cf62a16c215e367f51078a3ba23a3f257682032a8f3c657293029a886b18d82","libm/src/math/sincosf.rs":"b0f589e6ada8215944d7784f420c6721c90387d799e349ce7676674f3c475e75","libm/src/math/sinf.rs":"dcddac1d56b084cbb8d0e019433c9c5fe2201d9b257a7dcf2f85c9a8f14b79cf","libm/src/math/sinh.rs":"d8ee4c7af883a526f36c1a6da13bb81fba9181b477e2f2538161a2bee97edc35","libm/src/math/sinhf.rs":"d06eb030ba9dbf7094df127262bfe99f149b4db49fa8ab8c15499660f1e46b26","libm/src/math/sqrt.rs":"5f3a0a582b174fcfccb9c5274899cb05b664ccb92bf1d42caa58890947b68256","libm/src/math/sqrtf.rs":"da926ac27af6eecdf8b62d8baeefcfe1627110592e44298f6b7f05b7ac12fe7e","libm/src/math/tan.rs":"930ecedaadc60f704c2dfa4e15186f59713c1ba7d948529d215223b424827db5","libm/src/math/tanf.rs":"894156a3b107aee08461eb4e7e412fc049aa237d176ae705c6e3e2d7060d94e3","libm/src/math/tanh.rs":"f1f08eb98ed959a17370a7aaf0177be36e3764543424e78feb033ed3f5e8ec98","libm/src/math/tanhf.rs":"74027b0c672a4e64bdef6d7a3069b90caec50e1e7dbb2c12d2828f310502f41e","libm/src/math/tgamma.rs":"c889cfa49bbeb4dbb0941fe9fac3b4da7d5879dcf04a3b9bb6e56de529baf374","libm/src/math/tgammaf.rs":"0737b34777095d0e4d07fe533e8f105082dd4e8ece411bba6ae5993b45b9388c","libm/src/math/trunc.rs":"642264897cc1505e720c8cf313be81aa9fd53aae866644a2e988d01dbc77fd8a","libm/src/math/truncf.rs":"dee3607baf1af0f01deae46e429e097234c50b268eaefebbe716f19f38597900","src/aarch64_linux.rs":"a4bf136ba1624f253132a8588aac438ce23244a16655149320338ac980ad48cc","src/arm.rs":"3a7a2e21de7f475fbc2765109a18a1301bfb5c7be361c8344dde5ec23e8d7ace","src/arm_linux.rs":"35a4cb7b75015543feb15b0c692da0faf0e6037d3b97a4a18067ba416eae1a70","src/float/add.rs":"3ec32ceaf470a89777b54f9cde61832fdadeade0f4894f268a949e968520bc57","src/float/cmp.rs":"5a9d28640f76b3009f0829421f15898932d7d38da123b3e6215415d11221f91a","src/float/conv.rs":"8bf710288f88cfbf67e510f68abbb5a4f7173d2ea9ef32f98d594935fc051641","src/float/div.rs":"9b569e3f40135d4a3bfb59a53b1b65991c1b3cde38c7f9bb0f5597d2a6222151","src/float/extend.rs":"180b2e791c58e0526de0a798845c580ce3222c8a15c8665e6e6a4bf5cf1a34aa","src/float/mod.rs":"a91cf65abb6e715c5559e3e4bd87a69cd99a9552d54804d0b7137c02c513f158","src/float/mul.rs":"0d0c1f0c28c149ecadeafd459d3c4c9327e4cfcae2cba479957bb8010ef51a01","src/float/pow.rs":"2ada190738731eb6f24104f8fb8c4d6f03cfb16451536dbee32f2b33db0c4b19","src/float/sub.rs":"c2a87f4628f51d5d908d0f25b5d51ce0599dc559d5a72b20e131261f484d5848","src/float/trunc.rs":"d21d2a2f9a1918b4bbb594691e397972a7c04b74b2acf04016c55693abf6d24b","src/int/addsub.rs":"7ec45ce1ba15b56a5b7129d3e5722c4db764c6545306d3fa9090983bcabd6f17","src/int/leading_zeros.rs":"ccf5e9d098c80034dcf6e38437c9a2eb670fa8043558bbfb574f2293164729a6","src/int/mod.rs":"bab1b77535ceebdebb89fd4e59e7105f8c45347bb638351a626615b24544a0b1","src/int/mul.rs":"bb48d8fd42d8f9f5fe9271d8d0f7a92dbae320bf4346e19d1071eb2093cb8ed9","src/int/sdiv.rs":"ace4cb0ec388a38834e01cab2c5bc87182d31588dfc0b1ae117c11ed0c4781cf","src/int/shift.rs":"ff1e0dab608f2b9b3c68c7dfabd1e9bf6500518f5926a9ce4c1d84178bfe19f4","src/int/specialized_div_rem/asymmetric.rs":"27f5bf70a35109f9d4e4e1ad1e8003aa17da5a1e436bf3e63a493d7528a3a566","src/int/specialized_div_rem/binary_long.rs":"9f1ced81a394f000a21a329683144d68ee431a954136a3634eb55b1ee2cf6d51","src/int/specialized_div_rem/delegate.rs":"9df141af98e391361e25d71ae38d5e845a91d896edd2c041132fd46af8268e85","src/int/specialized_div_rem/mod.rs":"73c98b9f69cc9b101ae4c9081e82d66af1df4a58cf0c9bb2a8c8659265687f12","src/int/specialized_div_rem/norm_shift.rs":"3be7ee0dea545c1f702d9daf67dad2b624bf7b17b075c8b90d3d3f7b53df4c21","src/int/specialized_div_rem/trifecta.rs":"87eef69da255b809fd710b14f2eb3f9f59e3dac625f8564ebc8ba78f9763523b","src/int/udiv.rs":"3732b490a472505411577f008b92f489287745968ce6791665201201377d3475","src/lib.rs":"0a8a36f51f093f3063aebcbd7a5dc3d7d67efab598f246ba1fe5ac323c0c08ed","src/macros.rs":"a1872f50851bfcb822c1e90a98baa20f38cf1f470124b8213eebd37a92de7c37","src/math.rs":"84c616b3a8504594266a758719074e4b5406337b696b534495c123d56ec0326e","src/mem/impls.rs":"8b389b9aeb43dd55351a86abd4b5fc311f7161e1a4023ca3c5a4c57b49650881","src/mem/mod.rs":"714763d045a20e0a68c04f929d14fb3d7b28662dda4a2622970416642af833dc","src/mem/x86_64.rs":"2f29fb392086b3f7e2e78fcfcbf0f0e205822eb4599f1bdf93e41833e1bd2766","src/probestack.rs":"ef5c07e9b95de7b2b77a937789fcfefd9846274317489ad6d623e377c9888601","src/riscv.rs":"50ddd6c732a9f810ab6e15a97b22fdc94cfc1dea09c45d87c833937f9206bee0","src/x86.rs":"117b50d6725ee0af0a7b3d197ea580655561f66a870ebc450d96af22bf7f39f6","src/x86_64.rs":"4f16bc9fad7757d48a6da3a078c715dd3a22154aadb4f1998d4c1b5d91396f9e"},"package":"6866e0f3638013234db3c89ead7a14d278354338e7237257407500009012b23f"} \ No newline at end of file diff --git a/vendor/compiler_builtins/Cargo.lock b/vendor/compiler_builtins/Cargo.lock index 295e480ec..14025dbbf 100644 --- a/vendor/compiler_builtins/Cargo.lock +++ b/vendor/compiler_builtins/Cargo.lock @@ -10,7 +10,7 @@ checksum = "7db2f146208d7e0fbee761b09cd65a7f51ccc38705d4e7262dad4d73b12a76b1" [[package]] name = "compiler_builtins" -version = "0.1.92" +version = "0.1.95" dependencies = [ "cc", "rustc-std-workspace-core", diff --git a/vendor/compiler_builtins/Cargo.toml b/vendor/compiler_builtins/Cargo.toml index c0ea1fd6e..ff4c0328d 100644 --- a/vendor/compiler_builtins/Cargo.toml +++ b/vendor/compiler_builtins/Cargo.toml @@ -11,7 +11,7 @@ [package] name = "compiler_builtins" -version = "0.1.92" +version = "0.1.95" authors = ["Jorge Aparicio "] links = "compiler-rt" include = [ diff --git a/vendor/compiler_builtins/README.md b/vendor/compiler_builtins/README.md index 8b25558a8..da0adbce7 100644 --- a/vendor/compiler_builtins/README.md +++ b/vendor/compiler_builtins/README.md @@ -59,8 +59,8 @@ features = ["c"] 5. Once the PR passes our extensive [testing infrastructure][4], we'll merge it! 6. Celebrate :tada: -[1]: https://github.com/rust-lang/compiler-rt/tree/8598065bd965d9713bfafb6c1e766d63a7b17b89/test/builtins/Unit -[2]: https://github.com/rust-lang/compiler-rt/tree/8598065bd965d9713bfafb6c1e766d63a7b17b89/lib/builtins +[1]: https://github.com/rust-lang/llvm-project/tree/9e3de9490ff580cd484fbfa2908292b4838d56e7/compiler-rt/test/builtins/Unit +[2]: https://github.com/rust-lang/llvm-project/tree/9e3de9490ff580cd484fbfa2908292b4838d56e7/compiler-rt/lib/builtins [3]: https://github.com/rust-lang/compiler-builtins/blob/0ba07e49264a54cb5bbd4856fcea083bb3fbec15/build.rs#L180-L265 [4]: https://travis-ci.org/rust-lang/compiler-builtins diff --git a/vendor/compiler_builtins/build.rs b/vendor/compiler_builtins/build.rs index 766dec05d..4549d0b4f 100644 --- a/vendor/compiler_builtins/build.rs +++ b/vendor/compiler_builtins/build.rs @@ -1,4 +1,4 @@ -use std::env; +use std::{collections::HashMap, env, sync::atomic::Ordering}; fn main() { println!("cargo:rerun-if-changed=build.rs"); @@ -90,6 +90,65 @@ fn main() { { println!("cargo:rustc-cfg=kernel_user_helpers") } + + if llvm_target[0] == "aarch64" { + generate_aarch64_outlined_atomics(); + } +} + +fn aarch64_symbol(ordering: Ordering) -> &'static str { + match ordering { + Ordering::Relaxed => "relax", + Ordering::Acquire => "acq", + Ordering::Release => "rel", + Ordering::AcqRel => "acq_rel", + _ => panic!("unknown symbol for {:?}", ordering), + } +} + +/// The `concat_idents` macro is extremely annoying and doesn't allow us to define new items. +/// Define them from the build script instead. +/// Note that the majority of the code is still defined in `aarch64.rs` through inline macros. +fn generate_aarch64_outlined_atomics() { + use std::fmt::Write; + // #[macro_export] so that we can use this in tests + let gen_macro = + |name| format!("#[macro_export] macro_rules! foreach_{name} {{ ($macro:path) => {{\n"); + + // Generate different macros for add/clr/eor/set so that we can test them separately. + let sym_names = ["cas", "ldadd", "ldclr", "ldeor", "ldset", "swp"]; + let mut macros = HashMap::new(); + for sym in sym_names { + macros.insert(sym, gen_macro(sym)); + } + + // Only CAS supports 16 bytes, and it has a different implementation that uses a different macro. + let mut cas16 = gen_macro("cas16"); + + for ordering in [ + Ordering::Relaxed, + Ordering::Acquire, + Ordering::Release, + Ordering::AcqRel, + ] { + let sym_ordering = aarch64_symbol(ordering); + for size in [1, 2, 4, 8] { + for (sym, macro_) in &mut macros { + let name = format!("__aarch64_{sym}{size}_{sym_ordering}"); + writeln!(macro_, "$macro!( {ordering:?}, {size}, {name} );").unwrap(); + } + } + let name = format!("__aarch64_cas16_{sym_ordering}"); + writeln!(cas16, "$macro!( {ordering:?}, {name} );").unwrap(); + } + + let mut buf = String::new(); + for macro_def in macros.values().chain(std::iter::once(&cas16)) { + buf += macro_def; + buf += "}; }"; + } + let dst = std::env::var("OUT_DIR").unwrap() + "/outlined_atomics.rs"; + std::fs::write(dst, buf).unwrap(); } #[cfg(feature = "c")] diff --git a/vendor/compiler_builtins/examples/intrinsics.rs b/vendor/compiler_builtins/examples/intrinsics.rs index 0ca30c215..19bb569b5 100644 --- a/vendor/compiler_builtins/examples/intrinsics.rs +++ b/vendor/compiler_builtins/examples/intrinsics.rs @@ -4,6 +4,7 @@ // to link due to the missing intrinsic (symbol). #![allow(unused_features)] +#![allow(stable_features)] // bench_black_box feature is stable, leaving for backcompat #![cfg_attr(thumb, no_main)] #![deny(dead_code)] #![feature(bench_black_box)] diff --git a/vendor/compiler_builtins/src/aarch64_linux.rs b/vendor/compiler_builtins/src/aarch64_linux.rs new file mode 100644 index 000000000..62144e531 --- /dev/null +++ b/vendor/compiler_builtins/src/aarch64_linux.rs @@ -0,0 +1,277 @@ +//! Aarch64 targets have two possible implementations for atomics: +//! 1. Load-Locked, Store-Conditional (LL/SC), older and slower. +//! 2. Large System Extensions (LSE), newer and faster. +//! To avoid breaking backwards compat, C toolchains introduced a concept of "outlined atomics", +//! where atomic operations call into the compiler runtime to dispatch between two depending on +//! which is supported on the current CPU. +//! See https://community.arm.com/arm-community-blogs/b/tools-software-ides-blog/posts/making-the-most-of-the-arm-architecture-in-gcc-10#:~:text=out%20of%20line%20atomics for more discussion. +//! +//! Currently we only support LL/SC, because LSE requires `getauxval` from libc in order to do runtime detection. +//! Use the `compiler-rt` intrinsics if you want LSE support. +//! +//! Ported from `aarch64/lse.S` in LLVM's compiler-rt. +//! +//! Generate functions for each of the following symbols: +//! __aarch64_casM_ORDER +//! __aarch64_swpN_ORDER +//! __aarch64_ldaddN_ORDER +//! __aarch64_ldclrN_ORDER +//! __aarch64_ldeorN_ORDER +//! __aarch64_ldsetN_ORDER +//! for N = {1, 2, 4, 8}, M = {1, 2, 4, 8, 16}, ORDER = { relax, acq, rel, acq_rel } +//! +//! The original `lse.S` has some truly horrifying code that expects to be compiled multiple times with different constants. +//! We do something similar, but with macro arguments. +#![cfg_attr(feature = "c", allow(unused_macros))] // avoid putting the macros into a submodule + +// We don't do runtime dispatch so we don't have to worry about the `__aarch64_have_lse_atomics` global ctor. + +/// Translate a byte size to a Rust type. +#[rustfmt::skip] +macro_rules! int_ty { + (1) => { i8 }; + (2) => { i16 }; + (4) => { i32 }; + (8) => { i64 }; + (16) => { i128 }; +} + +/// Given a byte size and a register number, return a register of the appropriate size. +/// +/// See . +#[rustfmt::skip] +macro_rules! reg { + (1, $num:literal) => { concat!("w", $num) }; + (2, $num:literal) => { concat!("w", $num) }; + (4, $num:literal) => { concat!("w", $num) }; + (8, $num:literal) => { concat!("x", $num) }; +} + +/// Given an atomic ordering, translate it to the acquire suffix for the lxdr aarch64 ASM instruction. +#[rustfmt::skip] +macro_rules! acquire { + (Relaxed) => { "" }; + (Acquire) => { "a" }; + (Release) => { "" }; + (AcqRel) => { "a" }; +} + +/// Given an atomic ordering, translate it to the release suffix for the stxr aarch64 ASM instruction. +#[rustfmt::skip] +macro_rules! release { + (Relaxed) => { "" }; + (Acquire) => { "" }; + (Release) => { "l" }; + (AcqRel) => { "l" }; +} + +/// Given a size in bytes, translate it to the byte suffix for an aarch64 ASM instruction. +#[rustfmt::skip] +macro_rules! size { + (1) => { "b" }; + (2) => { "h" }; + (4) => { "" }; + (8) => { "" }; + (16) => { "" }; +} + +/// Given a byte size, translate it to an Unsigned eXTend instruction +/// with the correct semantics. +/// +/// See +#[rustfmt::skip] +macro_rules! uxt { + (1) => { "uxtb" }; + (2) => { "uxth" }; + ($_:tt) => { "mov" }; +} + +/// Given an atomic ordering and byte size, translate it to a LoaD eXclusive Register instruction +/// with the correct semantics. +/// +/// See . +macro_rules! ldxr { + ($ordering:ident, $bytes:tt) => { + concat!("ld", acquire!($ordering), "xr", size!($bytes)) + }; +} + +/// Given an atomic ordering and byte size, translate it to a STore eXclusive Register instruction +/// with the correct semantics. +/// +/// See . +macro_rules! stxr { + ($ordering:ident, $bytes:tt) => { + concat!("st", release!($ordering), "xr", size!($bytes)) + }; +} + +/// Given an atomic ordering and byte size, translate it to a LoaD eXclusive Pair of registers instruction +/// with the correct semantics. +/// +/// See +macro_rules! ldxp { + ($ordering:ident) => { + concat!("ld", acquire!($ordering), "xp") + }; +} + +/// Given an atomic ordering and byte size, translate it to a STore eXclusive Pair of registers instruction +/// with the correct semantics. +/// +/// See . +macro_rules! stxp { + ($ordering:ident) => { + concat!("st", release!($ordering), "xp") + }; +} + +/// See . +macro_rules! compare_and_swap { + ($ordering:ident, $bytes:tt, $name:ident) => { + intrinsics! { + #[maybe_use_optimized_c_shim] + #[naked] + pub unsafe extern "C" fn $name ( + expected: int_ty!($bytes), desired: int_ty!($bytes), ptr: *mut int_ty!($bytes) + ) -> int_ty!($bytes) { + // We can't use `AtomicI8::compare_and_swap`; we *are* compare_and_swap. + unsafe { core::arch::asm! { + // UXT s(tmp0), s(0) + concat!(uxt!($bytes), " ", reg!($bytes, 16), ", ", reg!($bytes, 0)), + "0:", + // LDXR s(0), [x2] + concat!(ldxr!($ordering, $bytes), " ", reg!($bytes, 0), ", [x2]"), + // cmp s(0), s(tmp0) + concat!("cmp ", reg!($bytes, 0), ", ", reg!($bytes, 16)), + "bne 1f", + // STXR w(tmp1), s(1), [x2] + concat!(stxr!($ordering, $bytes), " w17, ", reg!($bytes, 1), ", [x2]"), + "cbnz w17, 0b", + "1:", + "ret", + options(noreturn) + } } + } + } + }; +} + +// i128 uses a completely different impl, so it has its own macro. +macro_rules! compare_and_swap_i128 { + ($ordering:ident, $name:ident) => { + intrinsics! { + #[maybe_use_optimized_c_shim] + #[naked] + pub unsafe extern "C" fn $name ( + expected: i128, desired: i128, ptr: *mut i128 + ) -> i128 { + unsafe { core::arch::asm! { + "mov x16, x0", + "mov x17, x1", + "0:", + // LDXP x0, x1, [x4] + concat!(ldxp!($ordering), " x0, x1, [x4]"), + "cmp x0, x16", + "ccmp x1, x17, #0, eq", + "bne 1f", + // STXP w(tmp2), x2, x3, [x4] + concat!(stxp!($ordering), " w15, x2, x3, [x4]"), + "cbnz w15, 0b", + "1:", + "ret", + options(noreturn) + } } + } + } + }; +} + +/// See . +macro_rules! swap { + ($ordering:ident, $bytes:tt, $name:ident) => { + intrinsics! { + #[maybe_use_optimized_c_shim] + #[naked] + pub unsafe extern "C" fn $name ( + left: int_ty!($bytes), right_ptr: *mut int_ty!($bytes) + ) -> int_ty!($bytes) { + unsafe { core::arch::asm! { + // mov s(tmp0), s(0) + concat!("mov ", reg!($bytes, 16), ", ", reg!($bytes, 0)), + "0:", + // LDXR s(0), [x1] + concat!(ldxr!($ordering, $bytes), " ", reg!($bytes, 0), ", [x1]"), + // STXR w(tmp1), s(tmp0), [x1] + concat!(stxr!($ordering, $bytes), " w17, ", reg!($bytes, 16), ", [x1]"), + "cbnz w17, 0b", + "ret", + options(noreturn) + } } + } + } + }; +} + +/// See (e.g.) . +macro_rules! fetch_op { + ($ordering:ident, $bytes:tt, $name:ident, $op:literal) => { + intrinsics! { + #[maybe_use_optimized_c_shim] + #[naked] + pub unsafe extern "C" fn $name ( + val: int_ty!($bytes), ptr: *mut int_ty!($bytes) + ) -> int_ty!($bytes) { + unsafe { core::arch::asm! { + // mov s(tmp0), s(0) + concat!("mov ", reg!($bytes, 16), ", ", reg!($bytes, 0)), + "0:", + // LDXR s(0), [x1] + concat!(ldxr!($ordering, $bytes), " ", reg!($bytes, 0), ", [x1]"), + // OP s(tmp1), s(0), s(tmp0) + concat!($op, " ", reg!($bytes, 17), ", ", reg!($bytes, 0), ", ", reg!($bytes, 16)), + // STXR w(tmp2), s(tmp1), [x1] + concat!(stxr!($ordering, $bytes), " w15, ", reg!($bytes, 17), ", [x1]"), + "cbnz w15, 0b", + "ret", + options(noreturn) + } } + } + } + } +} + +// We need a single macro to pass to `foreach_ldadd`. +macro_rules! add { + ($ordering:ident, $bytes:tt, $name:ident) => { + fetch_op! { $ordering, $bytes, $name, "add" } + }; +} + +macro_rules! and { + ($ordering:ident, $bytes:tt, $name:ident) => { + fetch_op! { $ordering, $bytes, $name, "bic" } + }; +} + +macro_rules! xor { + ($ordering:ident, $bytes:tt, $name:ident) => { + fetch_op! { $ordering, $bytes, $name, "eor" } + }; +} + +macro_rules! or { + ($ordering:ident, $bytes:tt, $name:ident) => { + fetch_op! { $ordering, $bytes, $name, "orr" } + }; +} + +// See `generate_aarch64_outlined_atomics` in build.rs. +include!(concat!(env!("OUT_DIR"), "/outlined_atomics.rs")); +foreach_cas!(compare_and_swap); +foreach_cas16!(compare_and_swap_i128); +foreach_swp!(swap); +foreach_ldadd!(add); +foreach_ldclr!(and); +foreach_ldeor!(xor); +foreach_ldset!(or); diff --git a/vendor/compiler_builtins/src/float/cmp.rs b/vendor/compiler_builtins/src/float/cmp.rs index 1d4e38433..1bd7aa284 100644 --- a/vendor/compiler_builtins/src/float/cmp.rs +++ b/vendor/compiler_builtins/src/float/cmp.rs @@ -99,60 +99,74 @@ fn unord(a: F, b: F) -> bool { } intrinsics! { + #[avr_skip] pub extern "C" fn __lesf2(a: f32, b: f32) -> i32 { cmp(a, b).to_le_abi() } + #[avr_skip] pub extern "C" fn __gesf2(a: f32, b: f32) -> i32 { cmp(a, b).to_ge_abi() } + #[avr_skip] #[arm_aeabi_alias = __aeabi_fcmpun] pub extern "C" fn __unordsf2(a: f32, b: f32) -> i32 { unord(a, b) as i32 } + #[avr_skip] pub extern "C" fn __eqsf2(a: f32, b: f32) -> i32 { cmp(a, b).to_le_abi() } + #[avr_skip] pub extern "C" fn __ltsf2(a: f32, b: f32) -> i32 { cmp(a, b).to_le_abi() } + #[avr_skip] pub extern "C" fn __nesf2(a: f32, b: f32) -> i32 { cmp(a, b).to_le_abi() } + #[avr_skip] pub extern "C" fn __gtsf2(a: f32, b: f32) -> i32 { cmp(a, b).to_ge_abi() } + #[avr_skip] pub extern "C" fn __ledf2(a: f64, b: f64) -> i32 { cmp(a, b).to_le_abi() } + #[avr_skip] pub extern "C" fn __gedf2(a: f64, b: f64) -> i32 { cmp(a, b).to_ge_abi() } + #[avr_skip] #[arm_aeabi_alias = __aeabi_dcmpun] pub extern "C" fn __unorddf2(a: f64, b: f64) -> i32 { unord(a, b) as i32 } + #[avr_skip] pub extern "C" fn __eqdf2(a: f64, b: f64) -> i32 { cmp(a, b).to_le_abi() } + #[avr_skip] pub extern "C" fn __ltdf2(a: f64, b: f64) -> i32 { cmp(a, b).to_le_abi() } + #[avr_skip] pub extern "C" fn __nedf2(a: f64, b: f64) -> i32 { cmp(a, b).to_le_abi() } + #[avr_skip] pub extern "C" fn __gtdf2(a: f64, b: f64) -> i32 { cmp(a, b).to_ge_abi() } diff --git a/vendor/compiler_builtins/src/float/conv.rs b/vendor/compiler_builtins/src/float/conv.rs index a27d542fa..790c0ab9f 100644 --- a/vendor/compiler_builtins/src/float/conv.rs +++ b/vendor/compiler_builtins/src/float/conv.rs @@ -3,7 +3,7 @@ /// These are hand-optimized bit twiddling code, /// which unfortunately isn't the easiest kind of code to read. /// -/// The algorithm is explained here: https://blog.m-ou.se/floats/ +/// The algorithm is explained here: mod int_to_float { pub fn u32_to_f32_bits(i: u32) -> u32 { if i == 0 { diff --git a/vendor/compiler_builtins/src/float/div.rs b/vendor/compiler_builtins/src/float/div.rs index c2d6c07e7..c0aae34fb 100644 --- a/vendor/compiler_builtins/src/float/div.rs +++ b/vendor/compiler_builtins/src/float/div.rs @@ -12,11 +12,17 @@ where i32: CastInto, F::Int: CastInto, F::Int: HInt, + ::Int: core::ops::Mul, { + const NUMBER_OF_HALF_ITERATIONS: usize = 0; + const NUMBER_OF_FULL_ITERATIONS: usize = 3; + const USE_NATIVE_FULL_ITERATIONS: bool = true; + let one = F::Int::ONE; let zero = F::Int::ZERO; + let hw = F::BITS / 2; + let lo_mask = u32::MAX >> hw; - // let bits = F::BITS; let significand_bits = F::SIGNIFICAND_BITS; let max_exponent = F::EXPONENT_MAX; @@ -109,101 +115,341 @@ where } } - // Or in the implicit significand bit. (If we fell through from the + // Set the implicit significand bit. If we fell through from the // denormal path it was already set by normalize( ), but setting it twice - // won't hurt anything.) + // won't hurt anything. a_significand |= implicit_bit; b_significand |= implicit_bit; - let mut quotient_exponent: i32 = CastInto::::cast(a_exponent) - .wrapping_sub(CastInto::::cast(b_exponent)) - .wrapping_add(scale); - - // Align the significand of b as a Q31 fixed-point number in the range - // [1, 2.0) and get a Q32 approximate reciprocal using a small minimax - // polynomial approximation: reciprocal = 3/4 + 1/sqrt(2) - b/2. This - // is accurate to about 3.5 binary digits. - let q31b = CastInto::::cast(b_significand << 8.cast()); - let mut reciprocal = (0x7504f333u32).wrapping_sub(q31b); - - // Now refine the reciprocal estimate using a Newton-Raphson iteration: - // - // x1 = x0 * (2 - x0 * b) - // - // This doubles the number of correct binary digits in the approximation - // with each iteration, so after three iterations, we have about 28 binary - // digits of accuracy. - - let mut correction: u32 = - negate_u32(((reciprocal as u64).wrapping_mul(q31b as u64) >> 32) as u32); - reciprocal = ((reciprocal as u64).wrapping_mul(correction as u64) >> 31) as u32; - correction = negate_u32(((reciprocal as u64).wrapping_mul(q31b as u64) >> 32) as u32); - reciprocal = ((reciprocal as u64).wrapping_mul(correction as u64) >> 31) as u32; - correction = negate_u32(((reciprocal as u64).wrapping_mul(q31b as u64) >> 32) as u32); - reciprocal = ((reciprocal as u64).wrapping_mul(correction as u64) >> 31) as u32; - - // Exhaustive testing shows that the error in reciprocal after three steps - // is in the interval [-0x1.f58108p-31, 0x1.d0e48cp-29], in line with our - // expectations. We bump the reciprocal by a tiny value to force the error - // to be strictly positive (in the range [0x1.4fdfp-37,0x1.287246p-29], to - // be specific). This also causes 1/1 to give a sensible approximation - // instead of zero (due to overflow). - reciprocal = reciprocal.wrapping_sub(2); - - // The numerical reciprocal is accurate to within 2^-28, lies in the - // interval [0x1.000000eep-1, 0x1.fffffffcp-1], and is strictly smaller - // than the true reciprocal of b. Multiplying a by this reciprocal thus - // gives a numerical q = a/b in Q24 with the following properties: - // - // 1. q < a/b - // 2. q is in the interval [0x1.000000eep-1, 0x1.fffffffcp0) - // 3. the error in q is at most 2^-24 + 2^-27 -- the 2^24 term comes - // from the fact that we truncate the product, and the 2^27 term - // is the error in the reciprocal of b scaled by the maximum - // possible value of a. As a consequence of this error bound, - // either q or nextafter(q) is the correctly rounded - let mut quotient = (a_significand << 1).widen_mul(reciprocal.cast()).hi(); - - // Two cases: quotient is in [0.5, 1.0) or quotient is in [1.0, 2.0). - // In either case, we are going to compute a residual of the form - // - // r = a - q*b + + let written_exponent: i32 = CastInto::::cast( + a_exponent + .wrapping_sub(b_exponent) + .wrapping_add(scale.cast()), + ) + .wrapping_add(exponent_bias) as i32; + let b_uq1 = b_significand << (F::BITS - significand_bits - 1); + + // Align the significand of b as a UQ1.(n-1) fixed-point number in the range + // [1.0, 2.0) and get a UQ0.n approximate reciprocal using a small minimax + // polynomial approximation: x0 = 3/4 + 1/sqrt(2) - b/2. + // The max error for this approximation is achieved at endpoints, so + // abs(x0(b) - 1/b) <= abs(x0(1) - 1/1) = 3/4 - 1/sqrt(2) = 0.04289..., + // which is about 4.5 bits. + // The initial approximation is between x0(1.0) = 0.9571... and x0(2.0) = 0.4571... + + // Then, refine the reciprocal estimate using a quadratically converging + // Newton-Raphson iteration: + // x_{n+1} = x_n * (2 - x_n * b) // - // We know from the construction of q that r satisfies: + // Let b be the original divisor considered "in infinite precision" and + // obtained from IEEE754 representation of function argument (with the + // implicit bit set). Corresponds to rep_t-sized b_UQ1 represented in + // UQ1.(W-1). // - // 0 <= r < ulp(q)*b + // Let b_hw be an infinitely precise number obtained from the highest (HW-1) + // bits of divisor significand (with the implicit bit set). Corresponds to + // half_rep_t-sized b_UQ1_hw represented in UQ1.(HW-1) that is a **truncated** + // version of b_UQ1. // - // if r is greater than 1/2 ulp(q)*b, then q rounds up. Otherwise, we - // already have the correct result. The exact halfway case cannot occur. - // We also take this time to right shift quotient if it falls in the [1,2) - // range and adjust the exponent accordingly. - let residual = if quotient < (implicit_bit << 1) { - quotient_exponent = quotient_exponent.wrapping_sub(1); - (a_significand << (significand_bits + 1)).wrapping_sub(quotient.wrapping_mul(b_significand)) + // Let e_n := x_n - 1/b_hw + // E_n := x_n - 1/b + // abs(E_n) <= abs(e_n) + (1/b_hw - 1/b) + // = abs(e_n) + (b - b_hw) / (b*b_hw) + // <= abs(e_n) + 2 * 2^-HW + + // rep_t-sized iterations may be slower than the corresponding half-width + // variant depending on the handware and whether single/double/quad precision + // is selected. + // NB: Using half-width iterations increases computation errors due to + // rounding, so error estimations have to be computed taking the selected + // mode into account! + + #[allow(clippy::absurd_extreme_comparisons)] + let mut x_uq0 = if NUMBER_OF_HALF_ITERATIONS > 0 { + // Starting with (n-1) half-width iterations + let b_uq1_hw: u16 = + (CastInto::::cast(b_significand) >> (significand_bits + 1 - hw)) as u16; + + // C is (3/4 + 1/sqrt(2)) - 1 truncated to W0 fractional bits as UQ0.HW + // with W0 being either 16 or 32 and W0 <= HW. + // That is, C is the aforementioned 3/4 + 1/sqrt(2) constant (from which + // b/2 is subtracted to obtain x0) wrapped to [0, 1) range. + + // HW is at least 32. Shifting into the highest bits if needed. + let c_hw = (0x7504_u32 as u16).wrapping_shl(hw.wrapping_sub(32)); + + // b >= 1, thus an upper bound for 3/4 + 1/sqrt(2) - b/2 is about 0.9572, + // so x0 fits to UQ0.HW without wrapping. + let x_uq0_hw: u16 = { + let mut x_uq0_hw: u16 = c_hw.wrapping_sub(b_uq1_hw /* exact b_hw/2 as UQ0.HW */); + // An e_0 error is comprised of errors due to + // * x0 being an inherently imprecise first approximation of 1/b_hw + // * C_hw being some (irrational) number **truncated** to W0 bits + // Please note that e_0 is calculated against the infinitely precise + // reciprocal of b_hw (that is, **truncated** version of b). + // + // e_0 <= 3/4 - 1/sqrt(2) + 2^-W0 + + // By construction, 1 <= b < 2 + // f(x) = x * (2 - b*x) = 2*x - b*x^2 + // f'(x) = 2 * (1 - b*x) + // + // On the [0, 1] interval, f(0) = 0, + // then it increses until f(1/b) = 1 / b, maximum on (0, 1), + // then it decreses to f(1) = 2 - b + // + // Let g(x) = x - f(x) = b*x^2 - x. + // On (0, 1/b), g(x) < 0 <=> f(x) > x + // On (1/b, 1], g(x) > 0 <=> f(x) < x + // + // For half-width iterations, b_hw is used instead of b. + #[allow(clippy::reversed_empty_ranges)] + for _ in 0..NUMBER_OF_HALF_ITERATIONS { + // corr_UQ1_hw can be **larger** than 2 - b_hw*x by at most 1*Ulp + // of corr_UQ1_hw. + // "0.0 - (...)" is equivalent to "2.0 - (...)" in UQ1.(HW-1). + // On the other hand, corr_UQ1_hw should not overflow from 2.0 to 0.0 provided + // no overflow occurred earlier: ((rep_t)x_UQ0_hw * b_UQ1_hw >> HW) is + // expected to be strictly positive because b_UQ1_hw has its highest bit set + // and x_UQ0_hw should be rather large (it converges to 1/2 < 1/b_hw <= 1). + let corr_uq1_hw: u16 = + 0.wrapping_sub((x_uq0_hw as u32).wrapping_mul(b_uq1_hw.cast()) >> hw) as u16; + + // Now, we should multiply UQ0.HW and UQ1.(HW-1) numbers, naturally + // obtaining an UQ1.(HW-1) number and proving its highest bit could be + // considered to be 0 to be able to represent it in UQ0.HW. + // From the above analysis of f(x), if corr_UQ1_hw would be represented + // without any intermediate loss of precision (that is, in twice_rep_t) + // x_UQ0_hw could be at most [1.]000... if b_hw is exactly 1.0 and strictly + // less otherwise. On the other hand, to obtain [1.]000..., one have to pass + // 1/b_hw == 1.0 to f(x), so this cannot occur at all without overflow (due + // to 1.0 being not representable as UQ0.HW). + // The fact corr_UQ1_hw was virtually round up (due to result of + // multiplication being **first** truncated, then negated - to improve + // error estimations) can increase x_UQ0_hw by up to 2*Ulp of x_UQ0_hw. + x_uq0_hw = ((x_uq0_hw as u32).wrapping_mul(corr_uq1_hw as u32) >> (hw - 1)) as u16; + // Now, either no overflow occurred or x_UQ0_hw is 0 or 1 in its half_rep_t + // representation. In the latter case, x_UQ0_hw will be either 0 or 1 after + // any number of iterations, so just subtract 2 from the reciprocal + // approximation after last iteration. + + // In infinite precision, with 0 <= eps1, eps2 <= U = 2^-HW: + // corr_UQ1_hw = 2 - (1/b_hw + e_n) * b_hw + 2*eps1 + // = 1 - e_n * b_hw + 2*eps1 + // x_UQ0_hw = (1/b_hw + e_n) * (1 - e_n*b_hw + 2*eps1) - eps2 + // = 1/b_hw - e_n + 2*eps1/b_hw + e_n - e_n^2*b_hw + 2*e_n*eps1 - eps2 + // = 1/b_hw + 2*eps1/b_hw - e_n^2*b_hw + 2*e_n*eps1 - eps2 + // e_{n+1} = -e_n^2*b_hw + 2*eps1/b_hw + 2*e_n*eps1 - eps2 + // = 2*e_n*eps1 - (e_n^2*b_hw + eps2) + 2*eps1/b_hw + // \------ >0 -------/ \-- >0 ---/ + // abs(e_{n+1}) <= 2*abs(e_n)*U + max(2*e_n^2 + U, 2 * U) + } + // For initial half-width iterations, U = 2^-HW + // Let abs(e_n) <= u_n * U, + // then abs(e_{n+1}) <= 2 * u_n * U^2 + max(2 * u_n^2 * U^2 + U, 2 * U) + // u_{n+1} <= 2 * u_n * U + max(2 * u_n^2 * U + 1, 2) + + // Account for possible overflow (see above). For an overflow to occur for the + // first time, for "ideal" corr_UQ1_hw (that is, without intermediate + // truncation), the result of x_UQ0_hw * corr_UQ1_hw should be either maximum + // value representable in UQ0.HW or less by 1. This means that 1/b_hw have to + // be not below that value (see g(x) above), so it is safe to decrement just + // once after the final iteration. On the other hand, an effective value of + // divisor changes after this point (from b_hw to b), so adjust here. + x_uq0_hw.wrapping_sub(1_u16) + }; + + // Error estimations for full-precision iterations are calculated just + // as above, but with U := 2^-W and taking extra decrementing into account. + // We need at least one such iteration. + + // Simulating operations on a twice_rep_t to perform a single final full-width + // iteration. Using ad-hoc multiplication implementations to take advantage + // of particular structure of operands. + + let blo: u32 = (CastInto::::cast(b_uq1)) & lo_mask; + // x_UQ0 = x_UQ0_hw * 2^HW - 1 + // x_UQ0 * b_UQ1 = (x_UQ0_hw * 2^HW) * (b_UQ1_hw * 2^HW + blo) - b_UQ1 + // + // <--- higher half ---><--- lower half ---> + // [x_UQ0_hw * b_UQ1_hw] + // + [ x_UQ0_hw * blo ] + // - [ b_UQ1 ] + // = [ result ][.... discarded ...] + let corr_uq1 = negate_u32( + (x_uq0_hw as u32) * (b_uq1_hw as u32) + (((x_uq0_hw as u32) * (blo)) >> hw) - 1, + ); // account for *possible* carry + let lo_corr = corr_uq1 & lo_mask; + let hi_corr = corr_uq1 >> hw; + // x_UQ0 * corr_UQ1 = (x_UQ0_hw * 2^HW) * (hi_corr * 2^HW + lo_corr) - corr_UQ1 + let mut x_uq0: ::Int = ((((x_uq0_hw as u32) * hi_corr) << 1) + .wrapping_add(((x_uq0_hw as u32) * lo_corr) >> (hw - 1)) + .wrapping_sub(2)) + .cast(); // 1 to account for the highest bit of corr_UQ1 can be 1 + // 1 to account for possible carry + // Just like the case of half-width iterations but with possibility + // of overflowing by one extra Ulp of x_UQ0. + x_uq0 -= one; + // ... and then traditional fixup by 2 should work + + // On error estimation: + // abs(E_{N-1}) <= (u_{N-1} + 2 /* due to conversion e_n -> E_n */) * 2^-HW + // + (2^-HW + 2^-W)) + // abs(E_{N-1}) <= (u_{N-1} + 3.01) * 2^-HW + + // Then like for the half-width iterations: + // With 0 <= eps1, eps2 < 2^-W + // E_N = 4 * E_{N-1} * eps1 - (E_{N-1}^2 * b + 4 * eps2) + 4 * eps1 / b + // abs(E_N) <= 2^-W * [ 4 * abs(E_{N-1}) + max(2 * abs(E_{N-1})^2 * 2^W + 4, 8)) ] + // abs(E_N) <= 2^-W * [ 4 * (u_{N-1} + 3.01) * 2^-HW + max(4 + 2 * (u_{N-1} + 3.01)^2, 8) ] + x_uq0 } else { - quotient >>= 1; - (a_significand << significand_bits).wrapping_sub(quotient.wrapping_mul(b_significand)) + // C is (3/4 + 1/sqrt(2)) - 1 truncated to 32 fractional bits as UQ0.n + let c: ::Int = (0x7504F333 << (F::BITS - 32)).cast(); + let x_uq0: ::Int = c.wrapping_sub(b_uq1); + // E_0 <= 3/4 - 1/sqrt(2) + 2 * 2^-32 + x_uq0 + }; + + let mut x_uq0 = if USE_NATIVE_FULL_ITERATIONS { + for _ in 0..NUMBER_OF_FULL_ITERATIONS { + let corr_uq1: u32 = 0.wrapping_sub( + ((CastInto::::cast(x_uq0) as u64) * (CastInto::::cast(b_uq1) as u64)) + >> F::BITS, + ) as u32; + x_uq0 = ((((CastInto::::cast(x_uq0) as u64) * (corr_uq1 as u64)) >> (F::BITS - 1)) + as u32) + .cast(); + } + x_uq0 + } else { + // not using native full iterations + x_uq0 }; - let written_exponent = quotient_exponent.wrapping_add(exponent_bias as i32); + // Finally, account for possible overflow, as explained above. + x_uq0 = x_uq0.wrapping_sub(2.cast()); + + // u_n for different precisions (with N-1 half-width iterations): + // W0 is the precision of C + // u_0 = (3/4 - 1/sqrt(2) + 2^-W0) * 2^HW + + // Estimated with bc: + // define half1(un) { return 2.0 * (un + un^2) / 2.0^hw + 1.0; } + // define half2(un) { return 2.0 * un / 2.0^hw + 2.0; } + // define full1(un) { return 4.0 * (un + 3.01) / 2.0^hw + 2.0 * (un + 3.01)^2 + 4.0; } + // define full2(un) { return 4.0 * (un + 3.01) / 2.0^hw + 8.0; } + + // | f32 (0 + 3) | f32 (2 + 1) | f64 (3 + 1) | f128 (4 + 1) + // u_0 | < 184224974 | < 2812.1 | < 184224974 | < 791240234244348797 + // u_1 | < 15804007 | < 242.7 | < 15804007 | < 67877681371350440 + // u_2 | < 116308 | < 2.81 | < 116308 | < 499533100252317 + // u_3 | < 7.31 | | < 7.31 | < 27054456580 + // u_4 | | | | < 80.4 + // Final (U_N) | same as u_3 | < 72 | < 218 | < 13920 + + // Add 2 to U_N due to final decrement. + + let reciprocal_precision: ::Int = 10.cast(); + + // Suppose 1/b - P * 2^-W < x < 1/b + P * 2^-W + let x_uq0 = x_uq0 - reciprocal_precision; + // Now 1/b - (2*P) * 2^-W < x < 1/b + // FIXME Is x_UQ0 still >= 0.5? + + let mut quotient: ::Int = x_uq0.widen_mul(a_significand << 1).hi(); + // Now, a/b - 4*P * 2^-W < q < a/b for q= in UQ1.(SB+1+W). + + // quotient_UQ1 is in [0.5, 2.0) as UQ1.(SB+1), + // adjust it to be in [1.0, 2.0) as UQ1.SB. + let (mut residual, written_exponent) = if quotient < (implicit_bit << 1) { + // Highest bit is 0, so just reinterpret quotient_UQ1 as UQ1.SB, + // effectively doubling its value as well as its error estimation. + let residual_lo = (a_significand << (significand_bits + 1)).wrapping_sub( + (CastInto::::cast(quotient).wrapping_mul(CastInto::::cast(b_significand))) + .cast(), + ); + a_significand <<= 1; + (residual_lo, written_exponent.wrapping_sub(1)) + } else { + // Highest bit is 1 (the UQ1.(SB+1) value is in [1, 2)), convert it + // to UQ1.SB by right shifting by 1. Least significant bit is omitted. + quotient >>= 1; + let residual_lo = (a_significand << significand_bits).wrapping_sub( + (CastInto::::cast(quotient).wrapping_mul(CastInto::::cast(b_significand))) + .cast(), + ); + (residual_lo, written_exponent) + }; + //drop mutability + let quotient = quotient; + + // NB: residualLo is calculated above for the normal result case. + // It is re-computed on denormal path that is expected to be not so + // performance-sensitive. + + // Now, q cannot be greater than a/b and can differ by at most 8*P * 2^-W + 2^-SB + // Each NextAfter() increments the floating point value by at least 2^-SB + // (more, if exponent was incremented). + // Different cases (<---> is of 2^-SB length, * = a/b that is shown as a midpoint): + // q + // | | * | | | | | + // <---> 2^t + // | | | | | * | | + // q + // To require at most one NextAfter(), an error should be less than 1.5 * 2^-SB. + // (8*P) * 2^-W + 2^-SB < 1.5 * 2^-SB + // (8*P) * 2^-W < 0.5 * 2^-SB + // P < 2^(W-4-SB) + // Generally, for at most R NextAfter() to be enough, + // P < (2*R - 1) * 2^(W-4-SB) + // For f32 (0+3): 10 < 32 (OK) + // For f32 (2+1): 32 < 74 < 32 * 3, so two NextAfter() are required + // For f64: 220 < 256 (OK) + // For f128: 4096 * 3 < 13922 < 4096 * 5 (three NextAfter() are required) + + // If we have overflowed the exponent, return infinity if written_exponent >= max_exponent as i32 { - // If we have overflowed the exponent, return infinity. return F::from_repr(inf_rep | quotient_sign); - } else if written_exponent < 1 { - // Flush denormals to zero. In the future, it would be nice to add - // code to round them correctly. - return F::from_repr(quotient_sign); - } else { - let round = ((residual << 1) > b_significand) as u32; - // Clear the implicit bits - let mut abs_result = quotient & significand_mask; - // Insert the exponent - abs_result |= written_exponent.cast() << significand_bits; - // Round - abs_result = abs_result.wrapping_add(round.cast()); - // Insert the sign and return - return F::from_repr(abs_result | quotient_sign); } + + // Now, quotient <= the correctly-rounded result + // and may need taking NextAfter() up to 3 times (see error estimates above) + // r = a - b * q + let abs_result = if written_exponent > 0 { + let mut ret = quotient & significand_mask; + ret |= ((written_exponent as u32) << significand_bits).cast(); + residual <<= 1; + ret + } else { + if (significand_bits as i32 + written_exponent) < 0 { + return F::from_repr(quotient_sign); + } + let ret = quotient.wrapping_shr(negate_u32(CastInto::::cast(written_exponent)) + 1); + residual = (CastInto::::cast( + a_significand.wrapping_shl( + significand_bits.wrapping_add(CastInto::::cast(written_exponent)), + ), + ) + .wrapping_sub( + (CastInto::::cast(ret).wrapping_mul(CastInto::::cast(b_significand))) << 1, + )) + .cast(); + ret + }; + // Round + let abs_result = { + residual += abs_result & one; // tie to even + // The above line conditionally turns the below LT comparison into LTE + + if residual > b_significand { + abs_result + one + } else { + abs_result + } + }; + F::from_repr(abs_result | quotient_sign) } fn div64(a: F, b: F) -> F @@ -218,10 +464,15 @@ where F::Int: CastInto, F::Int: HInt, { + const NUMBER_OF_HALF_ITERATIONS: usize = 3; + const NUMBER_OF_FULL_ITERATIONS: usize = 1; + const USE_NATIVE_FULL_ITERATIONS: bool = false; + let one = F::Int::ONE; let zero = F::Int::ZERO; + let hw = F::BITS / 2; + let lo_mask = u64::MAX >> hw; - // let bits = F::BITS; let significand_bits = F::SIGNIFICAND_BITS; let max_exponent = F::EXPONENT_MAX; @@ -235,12 +486,6 @@ where let inf_rep = exponent_mask; let quiet_bit = implicit_bit >> 1; let qnan_rep = exponent_mask | quiet_bit; - // let exponent_bits = F::EXPONENT_BITS; - - #[inline(always)] - fn negate_u32(a: u32) -> u32 { - (::wrapping_neg(a as i32)) as u32 - } #[inline(always)] fn negate_u64(a: u64) -> u64 { @@ -320,128 +565,340 @@ where } } - // Or in the implicit significand bit. (If we fell through from the + // Set the implicit significand bit. If we fell through from the // denormal path it was already set by normalize( ), but setting it twice - // won't hurt anything.) + // won't hurt anything. a_significand |= implicit_bit; b_significand |= implicit_bit; - let mut quotient_exponent: i32 = CastInto::::cast(a_exponent) - .wrapping_sub(CastInto::::cast(b_exponent)) - .wrapping_add(scale); - - // Align the significand of b as a Q31 fixed-point number in the range - // [1, 2.0) and get a Q32 approximate reciprocal using a small minimax - // polynomial approximation: reciprocal = 3/4 + 1/sqrt(2) - b/2. This - // is accurate to about 3.5 binary digits. - let q31b = CastInto::::cast(b_significand >> 21.cast()); - let mut recip32 = (0x7504f333u32).wrapping_sub(q31b); - - // Now refine the reciprocal estimate using a Newton-Raphson iteration: - // - // x1 = x0 * (2 - x0 * b) - // - // This doubles the number of correct binary digits in the approximation - // with each iteration, so after three iterations, we have about 28 binary - // digits of accuracy. - - let mut correction32: u32 = - negate_u32(((recip32 as u64).wrapping_mul(q31b as u64) >> 32) as u32); - recip32 = ((recip32 as u64).wrapping_mul(correction32 as u64) >> 31) as u32; - correction32 = negate_u32(((recip32 as u64).wrapping_mul(q31b as u64) >> 32) as u32); - recip32 = ((recip32 as u64).wrapping_mul(correction32 as u64) >> 31) as u32; - correction32 = negate_u32(((recip32 as u64).wrapping_mul(q31b as u64) >> 32) as u32); - recip32 = ((recip32 as u64).wrapping_mul(correction32 as u64) >> 31) as u32; - - // recip32 might have overflowed to exactly zero in the preceeding - // computation if the high word of b is exactly 1.0. This would sabotage - // the full-width final stage of the computation that follows, so we adjust - // recip32 downward by one bit. - recip32 = recip32.wrapping_sub(1); - - // We need to perform one more iteration to get us to 56 binary digits; - // The last iteration needs to happen with extra precision. - let q63blo = CastInto::::cast(b_significand << 11.cast()); - - let correction: u64 = negate_u64( - (recip32 as u64) - .wrapping_mul(q31b as u64) - .wrapping_add((recip32 as u64).wrapping_mul(q63blo as u64) >> 32), - ); - let c_hi = (correction >> 32) as u32; - let c_lo = correction as u32; - let mut reciprocal: u64 = (recip32 as u64) - .wrapping_mul(c_hi as u64) - .wrapping_add((recip32 as u64).wrapping_mul(c_lo as u64) >> 32); - - // We already adjusted the 32-bit estimate, now we need to adjust the final - // 64-bit reciprocal estimate downward to ensure that it is strictly smaller - // than the infinitely precise exact reciprocal. Because the computation - // of the Newton-Raphson step is truncating at every step, this adjustment - // is small; most of the work is already done. - reciprocal = reciprocal.wrapping_sub(2); - - // The numerical reciprocal is accurate to within 2^-56, lies in the - // interval [0.5, 1.0), and is strictly smaller than the true reciprocal - // of b. Multiplying a by this reciprocal thus gives a numerical q = a/b - // in Q53 with the following properties: - // - // 1. q < a/b - // 2. q is in the interval [0.5, 2.0) - // 3. the error in q is bounded away from 2^-53 (actually, we have a - // couple of bits to spare, but this is all we need). - - // We need a 64 x 64 multiply high to compute q, which isn't a basic - // operation in C, so we need to be a little bit fussy. - // let mut quotient: F::Int = ((((reciprocal as u64) - // .wrapping_mul(CastInto::::cast(a_significand << 1) as u64)) - // >> 32) as u32) - // .cast(); - - // We need a 64 x 64 multiply high to compute q, which isn't a basic - // operation in C, so we need to be a little bit fussy. - let mut quotient = (a_significand << 2).widen_mul(reciprocal.cast()).hi(); - - // Two cases: quotient is in [0.5, 1.0) or quotient is in [1.0, 2.0). - // In either case, we are going to compute a residual of the form - // - // r = a - q*b + + let written_exponent: i64 = CastInto::::cast( + a_exponent + .wrapping_sub(b_exponent) + .wrapping_add(scale.cast()), + ) + .wrapping_add(exponent_bias as u64) as i64; + let b_uq1 = b_significand << (F::BITS - significand_bits - 1); + + // Align the significand of b as a UQ1.(n-1) fixed-point number in the range + // [1.0, 2.0) and get a UQ0.n approximate reciprocal using a small minimax + // polynomial approximation: x0 = 3/4 + 1/sqrt(2) - b/2. + // The max error for this approximation is achieved at endpoints, so + // abs(x0(b) - 1/b) <= abs(x0(1) - 1/1) = 3/4 - 1/sqrt(2) = 0.04289..., + // which is about 4.5 bits. + // The initial approximation is between x0(1.0) = 0.9571... and x0(2.0) = 0.4571... + + // Then, refine the reciprocal estimate using a quadratically converging + // Newton-Raphson iteration: + // x_{n+1} = x_n * (2 - x_n * b) // - // We know from the construction of q that r satisfies: + // Let b be the original divisor considered "in infinite precision" and + // obtained from IEEE754 representation of function argument (with the + // implicit bit set). Corresponds to rep_t-sized b_UQ1 represented in + // UQ1.(W-1). // - // 0 <= r < ulp(q)*b + // Let b_hw be an infinitely precise number obtained from the highest (HW-1) + // bits of divisor significand (with the implicit bit set). Corresponds to + // half_rep_t-sized b_UQ1_hw represented in UQ1.(HW-1) that is a **truncated** + // version of b_UQ1. // - // if r is greater than 1/2 ulp(q)*b, then q rounds up. Otherwise, we - // already have the correct result. The exact halfway case cannot occur. - // We also take this time to right shift quotient if it falls in the [1,2) - // range and adjust the exponent accordingly. - let residual = if quotient < (implicit_bit << 1) { - quotient_exponent = quotient_exponent.wrapping_sub(1); - (a_significand << (significand_bits + 1)).wrapping_sub(quotient.wrapping_mul(b_significand)) + // Let e_n := x_n - 1/b_hw + // E_n := x_n - 1/b + // abs(E_n) <= abs(e_n) + (1/b_hw - 1/b) + // = abs(e_n) + (b - b_hw) / (b*b_hw) + // <= abs(e_n) + 2 * 2^-HW + + // rep_t-sized iterations may be slower than the corresponding half-width + // variant depending on the handware and whether single/double/quad precision + // is selected. + // NB: Using half-width iterations increases computation errors due to + // rounding, so error estimations have to be computed taking the selected + // mode into account! + + let mut x_uq0 = if NUMBER_OF_HALF_ITERATIONS > 0 { + // Starting with (n-1) half-width iterations + let b_uq1_hw: u32 = + (CastInto::::cast(b_significand) >> (significand_bits + 1 - hw)) as u32; + + // C is (3/4 + 1/sqrt(2)) - 1 truncated to W0 fractional bits as UQ0.HW + // with W0 being either 16 or 32 and W0 <= HW. + // That is, C is the aforementioned 3/4 + 1/sqrt(2) constant (from which + // b/2 is subtracted to obtain x0) wrapped to [0, 1) range. + + // HW is at least 32. Shifting into the highest bits if needed. + let c_hw = (0x7504F333_u64 as u32).wrapping_shl(hw.wrapping_sub(32)); + + // b >= 1, thus an upper bound for 3/4 + 1/sqrt(2) - b/2 is about 0.9572, + // so x0 fits to UQ0.HW without wrapping. + let x_uq0_hw: u32 = { + let mut x_uq0_hw: u32 = c_hw.wrapping_sub(b_uq1_hw /* exact b_hw/2 as UQ0.HW */); + // dbg!(x_uq0_hw); + // An e_0 error is comprised of errors due to + // * x0 being an inherently imprecise first approximation of 1/b_hw + // * C_hw being some (irrational) number **truncated** to W0 bits + // Please note that e_0 is calculated against the infinitely precise + // reciprocal of b_hw (that is, **truncated** version of b). + // + // e_0 <= 3/4 - 1/sqrt(2) + 2^-W0 + + // By construction, 1 <= b < 2 + // f(x) = x * (2 - b*x) = 2*x - b*x^2 + // f'(x) = 2 * (1 - b*x) + // + // On the [0, 1] interval, f(0) = 0, + // then it increses until f(1/b) = 1 / b, maximum on (0, 1), + // then it decreses to f(1) = 2 - b + // + // Let g(x) = x - f(x) = b*x^2 - x. + // On (0, 1/b), g(x) < 0 <=> f(x) > x + // On (1/b, 1], g(x) > 0 <=> f(x) < x + // + // For half-width iterations, b_hw is used instead of b. + for _ in 0..NUMBER_OF_HALF_ITERATIONS { + // corr_UQ1_hw can be **larger** than 2 - b_hw*x by at most 1*Ulp + // of corr_UQ1_hw. + // "0.0 - (...)" is equivalent to "2.0 - (...)" in UQ1.(HW-1). + // On the other hand, corr_UQ1_hw should not overflow from 2.0 to 0.0 provided + // no overflow occurred earlier: ((rep_t)x_UQ0_hw * b_UQ1_hw >> HW) is + // expected to be strictly positive because b_UQ1_hw has its highest bit set + // and x_UQ0_hw should be rather large (it converges to 1/2 < 1/b_hw <= 1). + let corr_uq1_hw: u32 = + 0.wrapping_sub(((x_uq0_hw as u64).wrapping_mul(b_uq1_hw as u64)) >> hw) as u32; + // dbg!(corr_uq1_hw); + + // Now, we should multiply UQ0.HW and UQ1.(HW-1) numbers, naturally + // obtaining an UQ1.(HW-1) number and proving its highest bit could be + // considered to be 0 to be able to represent it in UQ0.HW. + // From the above analysis of f(x), if corr_UQ1_hw would be represented + // without any intermediate loss of precision (that is, in twice_rep_t) + // x_UQ0_hw could be at most [1.]000... if b_hw is exactly 1.0 and strictly + // less otherwise. On the other hand, to obtain [1.]000..., one have to pass + // 1/b_hw == 1.0 to f(x), so this cannot occur at all without overflow (due + // to 1.0 being not representable as UQ0.HW). + // The fact corr_UQ1_hw was virtually round up (due to result of + // multiplication being **first** truncated, then negated - to improve + // error estimations) can increase x_UQ0_hw by up to 2*Ulp of x_UQ0_hw. + x_uq0_hw = ((x_uq0_hw as u64).wrapping_mul(corr_uq1_hw as u64) >> (hw - 1)) as u32; + // dbg!(x_uq0_hw); + // Now, either no overflow occurred or x_UQ0_hw is 0 or 1 in its half_rep_t + // representation. In the latter case, x_UQ0_hw will be either 0 or 1 after + // any number of iterations, so just subtract 2 from the reciprocal + // approximation after last iteration. + + // In infinite precision, with 0 <= eps1, eps2 <= U = 2^-HW: + // corr_UQ1_hw = 2 - (1/b_hw + e_n) * b_hw + 2*eps1 + // = 1 - e_n * b_hw + 2*eps1 + // x_UQ0_hw = (1/b_hw + e_n) * (1 - e_n*b_hw + 2*eps1) - eps2 + // = 1/b_hw - e_n + 2*eps1/b_hw + e_n - e_n^2*b_hw + 2*e_n*eps1 - eps2 + // = 1/b_hw + 2*eps1/b_hw - e_n^2*b_hw + 2*e_n*eps1 - eps2 + // e_{n+1} = -e_n^2*b_hw + 2*eps1/b_hw + 2*e_n*eps1 - eps2 + // = 2*e_n*eps1 - (e_n^2*b_hw + eps2) + 2*eps1/b_hw + // \------ >0 -------/ \-- >0 ---/ + // abs(e_{n+1}) <= 2*abs(e_n)*U + max(2*e_n^2 + U, 2 * U) + } + // For initial half-width iterations, U = 2^-HW + // Let abs(e_n) <= u_n * U, + // then abs(e_{n+1}) <= 2 * u_n * U^2 + max(2 * u_n^2 * U^2 + U, 2 * U) + // u_{n+1} <= 2 * u_n * U + max(2 * u_n^2 * U + 1, 2) + + // Account for possible overflow (see above). For an overflow to occur for the + // first time, for "ideal" corr_UQ1_hw (that is, without intermediate + // truncation), the result of x_UQ0_hw * corr_UQ1_hw should be either maximum + // value representable in UQ0.HW or less by 1. This means that 1/b_hw have to + // be not below that value (see g(x) above), so it is safe to decrement just + // once after the final iteration. On the other hand, an effective value of + // divisor changes after this point (from b_hw to b), so adjust here. + x_uq0_hw.wrapping_sub(1_u32) + }; + + // Error estimations for full-precision iterations are calculated just + // as above, but with U := 2^-W and taking extra decrementing into account. + // We need at least one such iteration. + + // Simulating operations on a twice_rep_t to perform a single final full-width + // iteration. Using ad-hoc multiplication implementations to take advantage + // of particular structure of operands. + let blo: u64 = (CastInto::::cast(b_uq1)) & lo_mask; + // x_UQ0 = x_UQ0_hw * 2^HW - 1 + // x_UQ0 * b_UQ1 = (x_UQ0_hw * 2^HW) * (b_UQ1_hw * 2^HW + blo) - b_UQ1 + // + // <--- higher half ---><--- lower half ---> + // [x_UQ0_hw * b_UQ1_hw] + // + [ x_UQ0_hw * blo ] + // - [ b_UQ1 ] + // = [ result ][.... discarded ...] + let corr_uq1 = negate_u64( + (x_uq0_hw as u64) * (b_uq1_hw as u64) + (((x_uq0_hw as u64) * (blo)) >> hw) - 1, + ); // account for *possible* carry + let lo_corr = corr_uq1 & lo_mask; + let hi_corr = corr_uq1 >> hw; + // x_UQ0 * corr_UQ1 = (x_UQ0_hw * 2^HW) * (hi_corr * 2^HW + lo_corr) - corr_UQ1 + let mut x_uq0: ::Int = ((((x_uq0_hw as u64) * hi_corr) << 1) + .wrapping_add(((x_uq0_hw as u64) * lo_corr) >> (hw - 1)) + .wrapping_sub(2)) + .cast(); // 1 to account for the highest bit of corr_UQ1 can be 1 + // 1 to account for possible carry + // Just like the case of half-width iterations but with possibility + // of overflowing by one extra Ulp of x_UQ0. + x_uq0 -= one; + // ... and then traditional fixup by 2 should work + + // On error estimation: + // abs(E_{N-1}) <= (u_{N-1} + 2 /* due to conversion e_n -> E_n */) * 2^-HW + // + (2^-HW + 2^-W)) + // abs(E_{N-1}) <= (u_{N-1} + 3.01) * 2^-HW + + // Then like for the half-width iterations: + // With 0 <= eps1, eps2 < 2^-W + // E_N = 4 * E_{N-1} * eps1 - (E_{N-1}^2 * b + 4 * eps2) + 4 * eps1 / b + // abs(E_N) <= 2^-W * [ 4 * abs(E_{N-1}) + max(2 * abs(E_{N-1})^2 * 2^W + 4, 8)) ] + // abs(E_N) <= 2^-W * [ 4 * (u_{N-1} + 3.01) * 2^-HW + max(4 + 2 * (u_{N-1} + 3.01)^2, 8) ] + x_uq0 } else { - quotient >>= 1; - (a_significand << significand_bits).wrapping_sub(quotient.wrapping_mul(b_significand)) + // C is (3/4 + 1/sqrt(2)) - 1 truncated to 64 fractional bits as UQ0.n + let c: ::Int = (0x7504F333 << (F::BITS - 32)).cast(); + let x_uq0: ::Int = c.wrapping_sub(b_uq1); + // E_0 <= 3/4 - 1/sqrt(2) + 2 * 2^-64 + x_uq0 + }; + + let mut x_uq0 = if USE_NATIVE_FULL_ITERATIONS { + for _ in 0..NUMBER_OF_FULL_ITERATIONS { + let corr_uq1: u64 = 0.wrapping_sub( + (CastInto::::cast(x_uq0) * (CastInto::::cast(b_uq1))) >> F::BITS, + ); + x_uq0 = ((((CastInto::::cast(x_uq0) as u128) * (corr_uq1 as u128)) + >> (F::BITS - 1)) as u64) + .cast(); + } + x_uq0 + } else { + // not using native full iterations + x_uq0 }; - let written_exponent = quotient_exponent.wrapping_add(exponent_bias as i32); + // Finally, account for possible overflow, as explained above. + x_uq0 = x_uq0.wrapping_sub(2.cast()); + + // u_n for different precisions (with N-1 half-width iterations): + // W0 is the precision of C + // u_0 = (3/4 - 1/sqrt(2) + 2^-W0) * 2^HW + + // Estimated with bc: + // define half1(un) { return 2.0 * (un + un^2) / 2.0^hw + 1.0; } + // define half2(un) { return 2.0 * un / 2.0^hw + 2.0; } + // define full1(un) { return 4.0 * (un + 3.01) / 2.0^hw + 2.0 * (un + 3.01)^2 + 4.0; } + // define full2(un) { return 4.0 * (un + 3.01) / 2.0^hw + 8.0; } + + // | f32 (0 + 3) | f32 (2 + 1) | f64 (3 + 1) | f128 (4 + 1) + // u_0 | < 184224974 | < 2812.1 | < 184224974 | < 791240234244348797 + // u_1 | < 15804007 | < 242.7 | < 15804007 | < 67877681371350440 + // u_2 | < 116308 | < 2.81 | < 116308 | < 499533100252317 + // u_3 | < 7.31 | | < 7.31 | < 27054456580 + // u_4 | | | | < 80.4 + // Final (U_N) | same as u_3 | < 72 | < 218 | < 13920 + + // Add 2 to U_N due to final decrement. + + let reciprocal_precision: ::Int = 220.cast(); + + // Suppose 1/b - P * 2^-W < x < 1/b + P * 2^-W + let x_uq0 = x_uq0 - reciprocal_precision; + // Now 1/b - (2*P) * 2^-W < x < 1/b + // FIXME Is x_UQ0 still >= 0.5? + + let mut quotient: ::Int = x_uq0.widen_mul(a_significand << 1).hi(); + // Now, a/b - 4*P * 2^-W < q < a/b for q= in UQ1.(SB+1+W). + + // quotient_UQ1 is in [0.5, 2.0) as UQ1.(SB+1), + // adjust it to be in [1.0, 2.0) as UQ1.SB. + let (mut residual, written_exponent) = if quotient < (implicit_bit << 1) { + // Highest bit is 0, so just reinterpret quotient_UQ1 as UQ1.SB, + // effectively doubling its value as well as its error estimation. + let residual_lo = (a_significand << (significand_bits + 1)).wrapping_sub( + (CastInto::::cast(quotient).wrapping_mul(CastInto::::cast(b_significand))) + .cast(), + ); + a_significand <<= 1; + (residual_lo, written_exponent.wrapping_sub(1)) + } else { + // Highest bit is 1 (the UQ1.(SB+1) value is in [1, 2)), convert it + // to UQ1.SB by right shifting by 1. Least significant bit is omitted. + quotient >>= 1; + let residual_lo = (a_significand << significand_bits).wrapping_sub( + (CastInto::::cast(quotient).wrapping_mul(CastInto::::cast(b_significand))) + .cast(), + ); + (residual_lo, written_exponent) + }; - if written_exponent >= max_exponent as i32 { - // If we have overflowed the exponent, return infinity. + //drop mutability + let quotient = quotient; + + // NB: residualLo is calculated above for the normal result case. + // It is re-computed on denormal path that is expected to be not so + // performance-sensitive. + + // Now, q cannot be greater than a/b and can differ by at most 8*P * 2^-W + 2^-SB + // Each NextAfter() increments the floating point value by at least 2^-SB + // (more, if exponent was incremented). + // Different cases (<---> is of 2^-SB length, * = a/b that is shown as a midpoint): + // q + // | | * | | | | | + // <---> 2^t + // | | | | | * | | + // q + // To require at most one NextAfter(), an error should be less than 1.5 * 2^-SB. + // (8*P) * 2^-W + 2^-SB < 1.5 * 2^-SB + // (8*P) * 2^-W < 0.5 * 2^-SB + // P < 2^(W-4-SB) + // Generally, for at most R NextAfter() to be enough, + // P < (2*R - 1) * 2^(W-4-SB) + // For f32 (0+3): 10 < 32 (OK) + // For f32 (2+1): 32 < 74 < 32 * 3, so two NextAfter() are required + // For f64: 220 < 256 (OK) + // For f128: 4096 * 3 < 13922 < 4096 * 5 (three NextAfter() are required) + + // If we have overflowed the exponent, return infinity + if written_exponent >= max_exponent as i64 { return F::from_repr(inf_rep | quotient_sign); - } else if written_exponent < 1 { - // Flush denormals to zero. In the future, it would be nice to add - // code to round them correctly. - return F::from_repr(quotient_sign); - } else { - let round = ((residual << 1) > b_significand) as u32; - // Clear the implicit bits - let mut abs_result = quotient & significand_mask; - // Insert the exponent - abs_result |= written_exponent.cast() << significand_bits; - // Round - abs_result = abs_result.wrapping_add(round.cast()); - // Insert the sign and return - return F::from_repr(abs_result | quotient_sign); } + + // Now, quotient <= the correctly-rounded result + // and may need taking NextAfter() up to 3 times (see error estimates above) + // r = a - b * q + let abs_result = if written_exponent > 0 { + let mut ret = quotient & significand_mask; + ret |= ((written_exponent as u64) << significand_bits).cast(); + residual <<= 1; + ret + } else { + if (significand_bits as i64 + written_exponent) < 0 { + return F::from_repr(quotient_sign); + } + let ret = + quotient.wrapping_shr((negate_u64(CastInto::::cast(written_exponent)) + 1) as u32); + residual = (CastInto::::cast( + a_significand.wrapping_shl( + significand_bits.wrapping_add(CastInto::::cast(written_exponent)), + ), + ) + .wrapping_sub( + (CastInto::::cast(ret).wrapping_mul(CastInto::::cast(b_significand))) << 1, + )) + .cast(); + ret + }; + // Round + let abs_result = { + residual += abs_result & one; // tie to even + // conditionally turns the below LT comparison into LTE + if residual > b_significand { + abs_result + one + } else { + abs_result + } + }; + F::from_repr(abs_result | quotient_sign) } intrinsics! { diff --git a/vendor/compiler_builtins/src/lib.rs b/vendor/compiler_builtins/src/lib.rs index 71f249c8e..a6b61bdf5 100644 --- a/vendor/compiler_builtins/src/lib.rs +++ b/vendor/compiler_builtins/src/lib.rs @@ -48,6 +48,7 @@ pub mod int; all(target_arch = "x86_64", target_os = "uefi"), all(target_arch = "arm", target_os = "none"), all(target_arch = "xtensa", target_os = "none"), + all(target_arch = "mips", target_os = "none"), target_os = "xous", all(target_vendor = "fortanix", target_env = "sgx") ))] @@ -57,6 +58,9 @@ pub mod mem; #[cfg(target_arch = "arm")] pub mod arm; +#[cfg(all(target_arch = "aarch64", target_os = "linux", not(feature = "no-asm"),))] +pub mod aarch64_linux; + #[cfg(all( kernel_user_helpers, any(target_os = "linux", target_os = "android"), diff --git a/vendor/compiler_builtins/src/macros.rs b/vendor/compiler_builtins/src/macros.rs index 59f25317e..b11114f12 100644 --- a/vendor/compiler_builtins/src/macros.rs +++ b/vendor/compiler_builtins/src/macros.rs @@ -33,7 +33,7 @@ macro_rules! public_test_dep { /// /// This macro is structured to be invoked with a bunch of functions that looks /// like: -/// +/// ```ignore /// intrinsics! { /// pub extern "C" fn foo(a: i32) -> u32 { /// // ... @@ -44,6 +44,7 @@ macro_rules! public_test_dep { /// // ... /// } /// } +/// ``` /// /// Each function is defined in a manner that looks like a normal Rust function. /// The macro then accepts a few nonstandard attributes that can decorate @@ -203,7 +204,7 @@ macro_rules! intrinsics { ( #[maybe_use_optimized_c_shim] $(#[$($attr:tt)*])* - pub extern $abi:tt fn $name:ident( $($argname:ident: $ty:ty),* ) $(-> $ret:ty)? { + pub $(unsafe $(@ $empty:tt)? )? extern $abi:tt fn $name:ident( $($argname:ident: $ty:ty),* ) $(-> $ret:ty)? { $($body:tt)* } @@ -211,7 +212,7 @@ macro_rules! intrinsics { ) => ( #[cfg($name = "optimized-c")] #[cfg_attr(feature = "weak-intrinsics", linkage = "weak")] - pub extern $abi fn $name( $($argname: $ty),* ) $(-> $ret)? { + pub $(unsafe $($empty)? )? extern $abi fn $name( $($argname: $ty),* ) $(-> $ret)? { extern $abi { fn $name($($argname: $ty),*) $(-> $ret)?; } @@ -223,7 +224,7 @@ macro_rules! intrinsics { #[cfg(not($name = "optimized-c"))] intrinsics! { $(#[$($attr)*])* - pub extern $abi fn $name( $($argname: $ty),* ) $(-> $ret)? { + pub $(unsafe $($empty)? )? extern $abi fn $name( $($argname: $ty),* ) $(-> $ret)? { $($body)* } } @@ -437,12 +438,11 @@ macro_rules! intrinsics { intrinsics!($($rest)*); ); - // For division and modulo, AVR uses a custom calling convention¹ that does - // not match our definitions here. Ideally we would just use hand-written - // naked functions, but that's quite a lot of code to port² - so for the - // time being we are just ignoring the problematic functions, letting - // avr-gcc (which is required to compile to AVR anyway) link them from - // libgcc. + // For some intrinsics, AVR uses a custom calling convention¹ that does not + // match our definitions here. Ideally we would just use hand-written naked + // functions, but that's quite a lot of code to port² - so for the time + // being we are just ignoring the problematic functions, letting avr-gcc + // (which is required to compile to AVR anyway) link them from libgcc. // // ¹ https://gcc.gnu.org/wiki/avr-gcc (see "Exceptions to the Calling // Convention") diff --git a/vendor/compiler_builtins/src/math.rs b/vendor/compiler_builtins/src/math.rs index 498e4d85f..b4e5fc113 100644 --- a/vendor/compiler_builtins/src/math.rs +++ b/vendor/compiler_builtins/src/math.rs @@ -136,11 +136,12 @@ no_mangle! { fn truncf(x: f32) -> f32; } -// only for the thumb*-none-eabi*, riscv32*-none-elf and x86_64-unknown-none targets that lack the floating point instruction set +// only for the thumb*-none-eabi*, riscv32*-none-elf, x86_64-unknown-none and mips*-unknown-none targets that lack the floating point instruction set #[cfg(any( all(target_arch = "arm", target_os = "none"), all(target_arch = "riscv32", not(target_feature = "f"), target_os = "none"), - all(target_arch = "x86_64", target_os = "none") + all(target_arch = "x86_64", target_os = "none"), + all(target_arch = "mips", target_os = "none"), ))] no_mangle! { fn fmin(x: f64, y: f64) -> f64; -- cgit v1.2.3