From: Serge Petrenko <sergepetrenko@tarantool.org> To: vdavydov.dev@gmail.com Cc: tarantool-patches@freelists.org, Serge Petrenko <sergepetrenko@tarantool.org> Subject: [PATCH v3 1/5] decimal: fix ln hang on values between ~ 0.9 and 1.1 Date: Tue, 2 Jul 2019 20:27:48 +0300 [thread overview] Message-ID: <2536cd7f0c6aec76d62c2c0696b3449500e2c726.1562087728.git.sergepetrenko@tarantool.org> (raw) In-Reply-To: <cover.1562087728.git.sergepetrenko@tarantool.org> Turns out decNumberLn hangs when result is subnormal, according to the current context settings. To fix this, reset minimal allowed exponent to a smaller value during the ln operation and round the result afterwards. Follow-up 6d62c6c19ed418932ead1bba44fcd7cd84c78a19 --- src/lib/core/decimal.c | 19 +++++++++++++ test/unit/decimal.c | 5 +++- test/unit/decimal.result | 58 +++++++++++++++++++++++----------------- 3 files changed, 56 insertions(+), 26 deletions(-) diff --git a/src/lib/core/decimal.c b/src/lib/core/decimal.c index 1f9b5838d..c76387d76 100644 --- a/src/lib/core/decimal.c +++ b/src/lib/core/decimal.c @@ -264,11 +264,30 @@ decimal_log10(decimal_t *res, const decimal_t *lhs) decimal_t * decimal_ln(decimal_t *res, const decimal_t *lhs) { + /* + * ln hangs in an infinite loop when result is + * between -10 ^ emin and 10 ^ emin. + * For small x, ln(1 + x) = x. Say, we take the + * smallest allowed value for + * (1 + x) = 1 + 10 ^ -(DECIMAL_MAX_DIGITS - 1). + * For ln to work for this value we need to set emin to + * -DECIMAL_MAX_DIGITS. + */ + int32_t emin = decimal_context.emin; + decimal_context.emin = -DECIMAL_MAX_DIGITS; + decNumberLn(res, lhs, &decimal_context); + decimal_context.emin = emin; if (decimal_check_status(&decimal_context) != 0) { return NULL; } else { + /* + * The increased EMIN allows for scale up to + * 2 * (DECIMAL_MAX_DIGITS - 1). + * Round back to DECIMAL_MAX_DIGITS - 1. + */ + decimal_round(res, DECIMAL_MAX_DIGITS - 1); return res; } } diff --git a/test/unit/decimal.c b/test/unit/decimal.c index 7453d13ca..327a06d72 100644 --- a/test/unit/decimal.c +++ b/test/unit/decimal.c @@ -121,7 +121,7 @@ test_pack_unpack(void) int main(void) { - plan(258); + plan(266); dectest(314, 271, uint64, uint64_t); dectest(65535, 23456, uint64, uint64_t); @@ -152,6 +152,9 @@ main(void) dectest_op1(log10, 100, 2, 0); dectest_op1(ln, 10, 2.3, 2); + dectest_op1(ln, 1.1, 0.1, 1); + dectest_op1(ln, 1.0000000000000000000000000000000000001, + 0.0000000000000000000000000000000000001, 0); dectest_op1(exp, 2, 7.39, 2); dectest_op1(sqrt, 100, 10, 0); diff --git a/test/unit/decimal.result b/test/unit/decimal.result index 051dc7960..7e767dcd2 100644 --- a/test/unit/decimal.result +++ b/test/unit/decimal.result @@ -1,4 +1,4 @@ -1..258 +1..266 ok 1 - decimal(314) ok 2 - decimal(271) ok 3 - decimal(314) + decimal(271) @@ -233,29 +233,37 @@ ok 231 - decimal_from_string(10) ok 232 - decimal_from_string(2.3) ok 233 - decimal_ln(10) ok 234 - decimal_compare(2.3) -ok 235 - decimal_from_string(2) -ok 236 - decimal_from_string(7.39) -ok 237 - decimal_exp(2) -ok 238 - decimal_compare(7.39) -ok 239 - decimal_from_string(100) -ok 240 - decimal_from_string(10) -ok 241 - decimal_sqrt(100) -ok 242 - decimal_compare(10) -ok 243 - decimal construction from 2e38 failure -ok 244 - decimal construction from "1e38" failure -ok 245 - decimal construction from "100000000000000000000000000000000000000" failure -ok 246 - decimal construction from LONG_MIN success -ok 247 - decimal construction from LONG_MAX success -ok 248 - decimal construction from ULONG_MAX success -ok 249 - decimal_from_string(9e37) -ok 250 - decimal_from_string(1e37) -ok 251 - decimal_add(9e37, 1e37) - overflow -ok 252 - decimal_from_string(1e19) -ok 253 - decimal_from_string(1e19) -ok 254 - decimal_mul(1e19, 1e19) - overflow -ok 255 - decimal_from_string(1e19) -ok 256 - decimal_from_string(1e-19) -ok 257 - decimal_div(1e19, 1e-19) - overflow +ok 235 - decimal_from_string(1.1) +ok 236 - decimal_from_string(0.1) +ok 237 - decimal_ln(1.1) +ok 238 - decimal_compare(0.1) +ok 239 - decimal_from_string(1.0000000000000000000000000000000000001) +ok 240 - decimal_from_string(0.0000000000000000000000000000000000001) +ok 241 - decimal_ln(1.0000000000000000000000000000000000001) +ok 242 - decimal_compare(0.0000000000000000000000000000000000001) +ok 243 - decimal_from_string(2) +ok 244 - decimal_from_string(7.39) +ok 245 - decimal_exp(2) +ok 246 - decimal_compare(7.39) +ok 247 - decimal_from_string(100) +ok 248 - decimal_from_string(10) +ok 249 - decimal_sqrt(100) +ok 250 - decimal_compare(10) +ok 251 - decimal construction from 2e38 failure +ok 252 - decimal construction from "1e38" failure +ok 253 - decimal construction from "100000000000000000000000000000000000000" failure +ok 254 - decimal construction from LONG_MIN success +ok 255 - decimal construction from LONG_MAX success +ok 256 - decimal construction from ULONG_MAX success +ok 257 - decimal_from_string(9e37) +ok 258 - decimal_from_string(1e37) +ok 259 - decimal_add(9e37, 1e37) - overflow +ok 260 - decimal_from_string(1e19) +ok 261 - decimal_from_string(1e19) +ok 262 - decimal_mul(1e19, 1e19) - overflow +ok 263 - decimal_from_string(1e19) +ok 264 - decimal_from_string(1e-19) +ok 265 - decimal_div(1e19, 1e-19) - overflow 1..146 ok 1 - decimal_len(0) ok 2 - decimal_len(0) == len(decimal_pack(0) @@ -403,4 +411,4 @@ ok 257 - decimal_div(1e19, 1e-19) - overflow ok 144 - str(decimal_unpack(decimal_pack(-99999999999999999999999999999999999999)) == -99999999999999999999999999999999999999 ok 145 - unpack malformed decimal fails ok 146 - decode malformed decimal preserves buffer position -ok 258 - subtests +ok 266 - subtests -- 2.20.1 (Apple Git-117)
next prev parent reply other threads:[~2019-07-02 17:27 UTC|newest] Thread overview: 8+ messages / expand[flat|nested] mbox.gz Atom feed top 2019-07-02 17:27 [PATCH v3 0/5] decimal: expose decimal module to Lua Serge Petrenko 2019-07-02 17:27 ` Serge Petrenko [this message] 2019-07-02 17:27 ` [PATCH v3 2/5] decimal: diallow infinity and NaN entirely Serge Petrenko 2019-07-02 17:27 ` [PATCH v3 3/5] decimal: fix string formatting on construction from double Serge Petrenko 2019-07-02 17:27 ` [PATCH v3 4/5] lua/utils: add a function to register FFI metatypes Serge Petrenko 2019-07-02 17:27 ` [PATCH v3 5/5] decimal: expose decimal type to lua Serge Petrenko 2019-07-03 6:50 ` [tarantool-patches] " Serge Petrenko 2019-07-08 14:25 ` [PATCH v3 0/5] decimal: expose decimal module to Lua Vladimir Davydov
Reply instructions: You may reply publicly to this message via plain-text email using any one of the following methods: * Save the following mbox file, import it into your mail client, and reply-to-all from there: mbox Avoid top-posting and favor interleaved quoting: https://en.wikipedia.org/wiki/Posting_style#Interleaved_style * Reply using the --to, --cc, and --in-reply-to switches of git-send-email(1): git send-email \ --in-reply-to=2536cd7f0c6aec76d62c2c0696b3449500e2c726.1562087728.git.sergepetrenko@tarantool.org \ --to=sergepetrenko@tarantool.org \ --cc=tarantool-patches@freelists.org \ --cc=vdavydov.dev@gmail.com \ --subject='Re: [PATCH v3 1/5] decimal: fix ln hang on values between ~ 0.9 and 1.1' \ /path/to/YOUR_REPLY https://kernel.org/pub/software/scm/git/docs/git-send-email.html * If your mail client supports setting the In-Reply-To header via mailto: links, try the mailto: link
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox