This issue tracker has been migrated to GitHub, and is currently read-only.
For more information, see the GitHub FAQs in the Python's Developer Guide.

classification
Title: ICC compiler: ICC treats denormal floating point numbers as 0.0
Type: crash Stage: resolved
Components: Build, Windows Versions: Python 3.6, Python 3.5, Python 2.7
process
Status: closed Resolution: out of date
Dependencies: Superseder:
Assigned To: zach.ware Nosy List: mark.dickinson, paul.moore, pitrou, python-dev, r.david.murray, serhiy.storchaka, skrah, steve.dower, tim.golden, tim.peters, vstinner, zach.ware
Priority: normal Keywords: patch

Created on 2015-12-23 23:02 by skrah, last changed 2022-04-11 14:58 by admin. This issue is now closed.

Files
File name Uploaded Description Edit
issue25934.diff zach.ware, 2016-01-23 05:44 review
Messages (13)
msg256943 - (view) Author: Stefan Krah (skrah) * (Python committer) Date: 2015-12-23 23:02
Zachary in #24999:

"If that's the case, would anyone (in particular, Steve, Tim or Tim) mind if we just made the default (for MSVC as well as ICC) /fp:strict?  It would be much easier to just change the global default than to try to either make it settable or to change it just when ICC is used."
msg258631 - (view) Author: Zachary Ware (zach.ware) * (Python committer) Date: 2016-01-19 22:50
I've looked into this a bit, and tests won't pass with MSVC /fp:strict, so it would need to be ICC specific.  UNIX ICC builds do unconditionally use -fp-model strict now, though, so I think Windows ICC builds should as well.
msg258652 - (view) Author: Mark Dickinson (mark.dickinson) * (Python committer) Date: 2016-01-20 08:41
Which tests fail with MSVC /fp:strict? It's surprising that making behaviour stricter causes tests to fail that wouldn't have done otherwise. :-)
msg258685 - (view) Author: Zachary Ware (zach.ware) * (Python committer) Date: 2016-01-20 15:39
Ah, it's been a while since I tested that, so my reporting was inaccurate.  The problem is actually that the compile fails:

..\Modules\mathmodule.c(1924): error C2099: initializer is not a constant [C:\cpython\PCbuild\pythoncore.vcxproj]
..\Modules\mathmodule.c(1925): error C2099: initializer is not a constant [C:\cpython\PCbuild\pythoncore.vcxproj]

..\Python\dtoa.c(1215): error C2099: initializer is not a constant [C:\cpython\PCbuild\pythoncore.vcxproj]
msg258686 - (view) Author: Mark Dickinson (mark.dickinson) * (Python committer) Date: 2016-01-20 15:52
Thanks. The offending lines in the math module are:

static const double degToRad = Py_MATH_PI / 180.0;
static const double radToDeg = 180.0 / Py_MATH_PI;

It would be trivial to replace these with suitable constants.
msg258687 - (view) Author: Steve Dower (steve.dower) * (Python committer) Date: 2016-01-20 15:54
I'm -0 on changing the universal default (and want to call out that it's a relevant porting note for 3.6, not appropriate for 3.5), but we can presumably fix those in case someone wants to use strict for their own build.
msg258689 - (view) Author: Mark Dickinson (mark.dickinson) * (Python committer) Date: 2016-01-20 16:07
The dtoa.c occurrence is also straightforward to fix. It's coming from this declaration:

    static const double tinytens[] = { 1e-16, 1e-32, 1e-64, 1e-128,
                      9007199254740992.*9007199254740992.e-256};

We need to be a tiny bit careful here, since dtoa.c is fragile code and relies on exact representation of some floats. But this isn't one of them: the only thing that tinytens is used for is getting a first approximation to the correct strtod conversion before the main iteration kicks in.

So replacing that last tinytens value with a suitably precise constant should be okay. The *exact* value of the constant we need is 0x1.8062864ac6f43p-745, or in decimal: 

8.11296384146066798564414657510725945755617084762542409283393077939218873123696010224953185884350111344494845510586594487062499326481633636403965073720792027659736980514301906428513324403012698167973428740345918815412673646248939187125560378562250718330485204443480974365425682653185460026731686712157238961539250106472701178830382576220700577401565213825995547382545061286464557591871221948351734365233811730381171156150018612038234137396963100802781483162471536048255738998621259419238986828005847002315384502722411971989039148624688058131226853220141492784023284912109375E-225

But 8.112963841460668e-225 should be good enough (along with a comment explaining why the expression was changed). If MSVC supported C99's hex constants, we could just use 0x1.8062864ac6f43p-745. But it doesn't. :-(
msg258853 - (view) Author: Zachary Ware (zach.ware) * (Python committer) Date: 2016-01-23 05:44
Here's a patch that specifies /fp:strict if it looks like ICC is being used for the build.  It also adds a convenient property for checking whether it's an ICC build ($(ICCBuild)).  This doesn't change anything for MSVC builds but allows ICC builds to pass test_math, test_cmath, test_audioop, etc.; as such, it should be applied to all three branches.

As far as replacing the non-constants that MSVC doesn't like, I'm +1 but not too bothered about it.  I'm happy to test if anyone supplies a patch for it.
msg259227 - (view) Author: Roundup Robot (python-dev) (Python triager) Date: 2016-01-30 01:12
New changeset 296fb7c10a7d by Zachary Ware in branch '2.7':
Issue #25934: Default to /fp:strict for ICC builds
https://hg.python.org/cpython/rev/296fb7c10a7d

New changeset 747b415e96c4 by Zachary Ware in branch '3.5':
Issue #25934: Default to /fp:strict for ICC builds
https://hg.python.org/cpython/rev/747b415e96c4

New changeset c080ef5989f7 by Zachary Ware in branch 'default':
Issue #25934: Merge with 3.5
https://hg.python.org/cpython/rev/c080ef5989f7
msg259337 - (view) Author: Zachary Ware (zach.ware) * (Python committer) Date: 2016-02-01 17:52
With the buildbots not angry over this, I'm closing the issue.  I opened #26262 to follow the possibility of changing the initializers to allow /fp:strict with MSVC.
msg261976 - (view) Author: Zachary Ware (zach.ware) * (Python committer) Date: 2016-03-18 16:58
I closed this prematurely: I successfully added /fp:strict to ICC builds, but it didn't fix the underlying issue.
msg404221 - (view) Author: Zachary Ware (zach.ware) * (Python committer) Date: 2021-10-18 20:53
I no longer have access to ICC, and the ICC buildbots have been mothballed some years ago.  Closing this as out of date; the issue might not be, but the information about it here probably is :)
msg404315 - (view) Author: Mark Dickinson (mark.dickinson) * (Python committer) Date: 2021-10-19 16:45
> Closing this as out of date.

SGTM. Thanks.
History
Date User Action Args
2022-04-11 14:58:25adminsetgithub: 70122
2021-10-19 16:45:35mark.dickinsonsetmessages: + msg404315
2021-10-18 20:53:49zach.waresetstatus: open -> closed
resolution: out of date
messages: + msg404221

stage: resolved
2016-03-18 16:58:02zach.waresetstatus: closed -> open
resolution: fixed -> (no value)
messages: + msg261976

stage: resolved -> (no value)
2016-02-01 17:52:08zach.waresetstatus: open -> closed
messages: + msg259337

assignee: zach.ware
resolution: fixed
stage: resolved
2016-01-30 01:12:02python-devsetmessages: + msg259227
2016-01-23 05:44:35zach.waresetfiles: + issue25934.diff
keywords: + patch
messages: + msg258853

versions: + Python 2.7
2016-01-20 16:07:03mark.dickinsonsetmessages: + msg258689
2016-01-20 15:54:49steve.dowersetmessages: + msg258687
2016-01-20 15:52:03mark.dickinsonsetmessages: + msg258686
2016-01-20 15:39:45zach.waresetmessages: + msg258685
2016-01-20 08:41:38mark.dickinsonsetmessages: + msg258652
2016-01-19 22:50:06zach.waresetmessages: + msg258631
2015-12-24 06:32:54serhiy.storchakasettype: crash
components: + Build, Windows
versions: + Python 3.5, Python 3.6
2015-12-23 23:02:23skrahcreate