Support of 'long double' floating point data type on Intel CPUs ( A collection of threads )

[SergeyKostrov]

Support of 'long double' floating point data type on Intel CPUs ( A collection of threads )

Forum topic: Mathimf and Windows
Web-link: software.intel.com/en-us/forums/topic/357759

Forum topic: Support of Extended or Quad IEEE FP formats
Web-link: software.intel.com/en-us/forums/topic/358472

Forum topic: Using 'long double' in Parallel Studio?
Web-link: software.intel.com/en-us/forums/topic/266290

Forum topic: Why function printf does not support long double?
Web-link: software.intel.com/en-us/forums/topic/372720

Forum topic: Mixing of Floating-Point Types ( MFPT ) when performing calculations. Does it improve accuracy?
Web-link: software.intel.com/en-us/forums/topic/361134

[Bernard]

Hi Sergey!

What do you think about the idea of creating threads which contain a collection of links to specific topics?

[SergeyKostrov]

>>...What do you think about the idea of creating threads which contain a collection of links to specific topics? Intel already does it and such threads are marked as Sticky. For all the rest it makes sense only if subject of a thread is really hot.

[Bernard]

Thanks for replaying.

[SergeyKostrov · Accepted solution]

Updated list of threads is as follows: Forum topic: Mathimf and Windows Web-link: http://software.intel.com/en-us/forums/topic/357759 Forum topic: Support of Extended or Quad IEEE FP formats Web-link: http://software.intel.com/en-us/forums/topic/358472 Forum topic: Using 'long double' in Parallel Studio? Web-link: http://software.intel.com/en-us/forums/topic/266290 Forum topic: Why function printf does not support long double? Web-link: http://software.intel.com/en-us/forums/topic/372720 Forum topic: Mixing of Floating-Point Types ( MFPT ) when performing calculations. Does it improve accuracy? Web-link: http://software.intel.com/en-us/forums/topic/361134 Forum topic: Test results for CRT-function 'sqrt' for different Floating Point Models Web-link: http://software.intel.com/en-us/forums/topic/368241

[SergeyKostrov]

A new very interesting issue was detected recently: Forum topic: Precision Issues Web-link: http://software.intel.com/en-us/forums/topic/487265

[QIAOMIN_Q_]

Good catch ,thanks Sergey~

Best, Qiao

[QIAOMIN_Q_]

Good catch ,thanks Sergey~

Best, Qiao

[SergeyKostrov]

It is hard to believe but Single-Precision floating point arithmetics posesses some very interesting surprises, like: ... 16777216.0 + 1.0 = 16777216.0 16777216.0 + 2.0 = 16777218.0 16777216.0 + 3.0 = 16777220.0 ...

[Patrick_F_Intel1]

Hello Sergey.

It is not really hard to believe right? This is a feature of the single precision IEEE 754 spec and the number of significant digits that can be represented.

One can get into the bits of how the numbers are represented in the FP standard but I haven't done those exercises since grad school.

Pat

[SergeyKostrov]

>>...It is not really hard to believe right?.. Yes, but when it happens a developer is ready to assume that something is wrong with Intel software, like Intel C++ compiler, or Intel hardware, like FPU. The developer actually can not assume that the core problem is in sources, because float data type is used instead of double or long double data types.