From: drh <> Date: Wed, 13 Sep 2023 20:35:04 +0000 (+0000) Subject: Improved comments on the hasHighPrecisionDouble() routine. No changes to X-Git-Tag: version-3.44.0~203^2 X-Git-Url: http://git.ipfire.org/?a=commitdiff_plain;h=refs%2Fheads%2Fruntime-longdouble-test;p=thirdparty%2Fsqlite.git Improved comments on the hasHighPrecisionDouble() routine. No changes to the underlying code. FossilOrigin-Name: 810c635ce063d873e969bf83339c654f6008e84ce8a61f0ffc61806e98d13dde --- diff --git a/manifest b/manifest index 05f0ad474e..806405336e 100644 --- a/manifest +++ b/manifest @@ -1,5 +1,5 @@ -C Determine\sat\sstart\stime\swhether\sor\snot\sthe\sunderlying\shardware\ssupports\nhigh-precision\slong\sdouble\scomputations. -D 2023-09-13T20:06:46.382 +C Improved\scomments\son\sthe\shasHighPrecisionDouble()\sroutine.\s\sNo\schanges\sto\nthe\sunderlying\scode. +D 2023-09-13T20:35:04.920 F .fossil-settings/empty-dirs dbb81e8fc0401ac46a1491ab34a7f2c7c0452f2f06b54ebb845d024ca8283ef1 F .fossil-settings/ignore-glob 35175cdfcf539b2318cb04a9901442804be81cd677d8b889fcc9149c21f239ea F LICENSE.md df5091916dbb40e6e9686186587125e1b2ff51f022cc334e886c19a0e9982724 @@ -672,7 +672,7 @@ F src/insert.c 3f0a94082d978bbdd33c38fefea15346c6c6bffb70bc645a71dc0f1f87dd3276 F src/json.c 51141f1c09ccb177057e5813e6302a5e32e5ba88cc4a756318a35081010fc6df F src/legacy.c d7874bc885906868cd51e6c2156698f2754f02d9eee1bae2d687323c3ca8e5aa F src/loadext.c 98cfba10989b3da6f1807ad42444017742db7f100a54f1032af7a8b1295912c0 -F src/main.c d3489e9c6bb2953c341a1d21905e938cc637cf3eda3d45763c8783ad3e8a5df8 +F src/main.c 618aeb399e993cf561864f4b0cf6a331ee4f355cf663635f8d9da3193a46aa40 F src/malloc.c 47b82c5daad557d9b963e3873e99c22570fb470719082c6658bf64e3012f7d23 F src/mem0.c 6a55ebe57c46ca1a7d98da93aaa07f99f1059645 F src/mem1.c 3bb59158c38e05f6270e761a9f435bf19827a264c13d1631c58b84bdc96d73b2 @@ -2120,11 +2120,8 @@ F vsixtest/vsixtest.tcl 6a9a6ab600c25a91a7acc6293828957a386a8a93 F vsixtest/vsixtest.vcxproj.data 2ed517e100c66dc455b492e1a33350c1b20fbcdc F vsixtest/vsixtest.vcxproj.filters 37e51ffedcdb064aad6ff33b6148725226cd608e F vsixtest/vsixtest_TemporaryKey.pfx e5b1b036facdb453873e7084e1cae9102ccc67a0 -P 9ea0a9f39d03707d251af5af08d1ba94958704ba74019d7e8823ccd21936023c -R 9d032fc0954884df48b056f822f415bb -T *branch * runtime-longdouble-test -T *sym-runtime-longdouble-test * -T -sym-trunk * +P 9a854b919667e0e679a259542b2ee444ee416dbd73ecd9458f6ced35d9d3f264 +R 2cf5225da68e808b8d6be96ec87494aa U drh -Z 4050589208bc39d1cdb1d56c5679577d +Z 7ae23f2f1935a56e3223edbdedc613fe # Remove this line to create a well-formed Fossil manifest. diff --git a/manifest.uuid b/manifest.uuid index 8466fcf0c1..5b62515d2a 100644 --- a/manifest.uuid +++ b/manifest.uuid @@ -1 +1 @@ -9a854b919667e0e679a259542b2ee444ee416dbd73ecd9458f6ced35d9d3f264 \ No newline at end of file +810c635ce063d873e969bf83339c654f6008e84ce8a61f0ffc61806e98d13dde \ No newline at end of file diff --git a/src/main.c b/src/main.c index 59f90197fa..a43afab168 100644 --- a/src/main.c +++ b/src/main.c @@ -160,15 +160,21 @@ char *sqlite3_temp_directory = 0; char *sqlite3_data_directory = 0; /* -** Determine what the default bUseLongDouble value should be and set it. +** Determine whether or not high-precision (long double) floating point +** math works correctly on CPU currently running. */ static SQLITE_NOINLINE int hasHighPrecisionDouble(int rc){ if( sizeof(LONGDOUBLE_TYPE)<=8 ){ + /* If the size of "long double" is not more than 8, then + ** high-precision math is not possible. */ return 0; }else{ /* Just because sizeof(long double)>8 does not mean that the underlying - ** hardware actually supports high-precision floating point. Do a test - ** to verify that it really does */ + ** hardware actually supports high-precision floating point. For example, + ** clearing the 0x100 bit in the floating-point control word on Intel + ** processors will make long double work like double, even though long + ** double takes up more space. The only way to determine if long double + ** actually works is to run an experiment. */ LONGDOUBLE_TYPE a, b, c; rc++; a = 1.0+rc*0.1;