public inbox for gcc-bugs@sourceware.org help / color / mirror / Atom feed
* [Bug libstdc++/111747] New: Problem with large float list initialization @ 2023-10-09 22:04 oplata.kes1 at mail dot ru 2023-10-10 1:50 ` [Bug libstdc++/111747] " pinskia at gcc dot gnu.org 2023-10-10 19:52 ` oplata.kes1 at mail dot ru 0 siblings, 2 replies; 3+ messages in thread From: oplata.kes1 at mail dot ru @ 2023-10-09 22:04 UTC (permalink / raw) To: gcc-bugs https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111747 Bug ID: 111747 Summary: Problem with large float list initialization Product: gcc Version: 11.4.0 Status: UNCONFIRMED Severity: normal Priority: P3 Component: libstdc++ Assignee: unassigned at gcc dot gnu.org Reporter: oplata.kes1 at mail dot ru Target Milestone: --- Created attachment 56085 --> https://gcc.gnu.org/bugzilla/attachment.cgi?id=56085&action=edit *ii file and output I create a vector on stack of 50 (or 500mln) floats equal to 1.0 and simply add them. The sum is not equal to 50 mln (or 5 billions). 5 mln of floats initizalize and sum fine. The command is g++ -v -save-temps gcc.cpp && ./a.out If it is a stack overflow, shouldn't the code fail with stack overflow error? If not, what is it? I use GCC 11.4.0 in Ubuntu 22.04 under WSL 2 (!) ^ permalink raw reply [flat|nested] 3+ messages in thread
* [Bug libstdc++/111747] Problem with large float list initialization 2023-10-09 22:04 [Bug libstdc++/111747] New: Problem with large float list initialization oplata.kes1 at mail dot ru @ 2023-10-10 1:50 ` pinskia at gcc dot gnu.org 2023-10-10 19:52 ` oplata.kes1 at mail dot ru 1 sibling, 0 replies; 3+ messages in thread From: pinskia at gcc dot gnu.org @ 2023-10-10 1:50 UTC (permalink / raw) To: gcc-bugs https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111747 Andrew Pinski <pinskia at gcc dot gnu.org> changed: What |Removed |Added ---------------------------------------------------------------------------- Status|UNCONFIRMED |RESOLVED Resolution|--- |INVALID --- Comment #1 from Andrew Pinski <pinskia at gcc dot gnu.org> --- 32bit floating point has the following characteristics: Sign bit: 1 bit Exponent width: 8 bits Significand precision: 24 bits (23 explicitly stored) 50000000 is 0x2faf080 which is more than 24bits in precision which means it cannot be represented exactly and when you start to add 1 to something which is greater than 0xfffff0 (which is what 1.67772e+07 is), the value stays the same and you start to lose precision. ^ permalink raw reply [flat|nested] 3+ messages in thread
* [Bug libstdc++/111747] Problem with large float list initialization 2023-10-09 22:04 [Bug libstdc++/111747] New: Problem with large float list initialization oplata.kes1 at mail dot ru 2023-10-10 1:50 ` [Bug libstdc++/111747] " pinskia at gcc dot gnu.org @ 2023-10-10 19:52 ` oplata.kes1 at mail dot ru 1 sibling, 0 replies; 3+ messages in thread From: oplata.kes1 at mail dot ru @ 2023-10-10 19:52 UTC (permalink / raw) To: gcc-bugs https://gcc.gnu.org/bugzilla/show_bug.cgi?id=111747 --- Comment #2 from oplata.kes1 at mail dot ru --- (In reply to Andrew Pinski from comment #1) > > 32bit floating point has the following characteristics: > Sign bit: 1 bit > Exponent width: 8 bits > Significand precision: 24 bits (23 explicitly stored) > > > 50000000 is 0x2faf080 which is more than 24bits in precision which means it > cannot be represented exactly and when you start to add 1 to something which > is greater than 0xfffff0 (which is what 1.67772e+07 is), the value stays the > same and you start to lose precision. Thank you very much! ^ permalink raw reply [flat|nested] 3+ messages in thread
end of thread, other threads:[~2023-10-10 19:52 UTC | newest] Thread overview: 3+ messages (download: mbox.gz / follow: Atom feed) -- links below jump to the message on this page -- 2023-10-09 22:04 [Bug libstdc++/111747] New: Problem with large float list initialization oplata.kes1 at mail dot ru 2023-10-10 1:50 ` [Bug libstdc++/111747] " pinskia at gcc dot gnu.org 2023-10-10 19:52 ` oplata.kes1 at mail dot ru
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox; as well as URLs for read-only IMAP folder(s) and NNTP newsgroup(s).