Groups | Search | Server Info | Keyboard shortcuts | Login | Register [http] [https] [nntp] [nntps]
Groups > comp.lang.forth > #52039
| From | Gerry Jackson <gerry@jackson9000.fsnet.co.uk> |
|---|---|
| Newsgroups | comp.lang.forth |
| Subject | Re: Gforth snapshot for Windows |
| Date | 2016-10-19 10:50 +0100 |
| Organization | A noiseless patient Spider |
| Message-ID | <nu7fjp$4at$1@dont-email.me> (permalink) |
| References | (3 earlier) <20161012010234.3c552916@HP-6550b> <ntmkb2$qa9$2@dont-email.me> <ntr389$gv7$1@dont-email.me> <ntrppe$v0c$1@dont-email.me> <nu17cp$fc4$1@dont-email.me> |
On 17/10/2016 01:52, Bernd Paysan wrote:
> Am Fri, 14 Oct 2016 23:29:50 +0000 schrieb Bernd Paysan:
>
>> Am Fri, 14 Oct 2016 18:05:30 +0100 schrieb Gerry Jackson:
>>
>>> On 13/10/2016 01:26, Bernd Paysan wrote:
>>> gforth-0.7.9_20161006.exe fails 8 tests in SUBSTITUTE tests in file
>>> stringtests.fth under Windows 10. IIRC they used to run on a previous
>>> version Gforth 0.7.9_20160113
>>>
>>> Use the latest version of the tests:
>>> https://github.com/gerryjackson/forth2012-test-suite
>>
>> Yes, I really need to get around embedding all your new tests into the
>> test suite Gforth runs automatically after build.
>
> I started with your FP test, and I ran into problems pretty soon. First
> of all, older glibcs fail on a few of the precise number conversions.
> Second, the x87 coprocessor fails on the rounding tests, it is slightly
> off.
>
Thank you for the feedback on the FP tests I appreciate it, that is only
the second (IIRC) feedback since I packaged the tests from various
sources in May 2012 with the following comment on c.l.f:
<quote>I propose to package the programs, more or less, as is, into a
download on the FLAG site and invite implementers to run them on their
Forth system. If they encounter errors, non-compliances or other
problems in the test programs to raise it with me or on comp.lang.forth,
get it resolved and update the download. In this way we could converge
on a set of satisfactory tests.</quote>
Clearly that was an over-optimistic hope. As I remember at the time
GForth did much better than other major systems which either failed
abysmally or simply crashed - perhaps when I get time I'll try again and
'name and shame' the other systems :)
> I've identified these problems:
>
> old glibc:
>
> INCORRECT RESULT hex_t{ r8 2L@ -> 42c00000 00000002 }t
> Actual result[0]=1
> INCORRECT RESULT hex_t{ r8 2L@ -> 3fe00000 00000002 }t
> Actual result[0]=1
> INCORRECT RESULT hex_t{ r8 2L@ -> 404f44ab d5aa7ca4 }t
> Actual result[0]=D5AA7CA3
> INCORRECT RESULT hex_t{ r8 2L@ -> 3ff00000 00000001 }t
> Actual result[0]=0
>
If I was correct in my post on 4 May 2012 (sorry you'll have to scan
down the messages):
https://groups.google.com/forum/?hl=en#!searchin/comp.lang.forth/floating$20point$20tests%7Csort:relevance/comp.lang.forth/lJ1hdgZCpuo/Zp9H5xhKV0MJ
GForth 0.7.0 used to pass these tests in fpio-test.4th in 2012. Are you
suggesting that the tests are wrong or that old glibc has problems?
> The "Actual result[i]=value" is a modification of ttester.fs I made to
> make sure I can see what's going on right in the output, without having
> to check interactively; sometimes that's impossible, e.g. on the Travis
> setup.
That seems useful, I'll think about adding it to the ttester.fs in the
test suite.
>
> And this is x86 (the "spoil financial calculation" defect only with old
> glibc, so that is a number conversion problem):
>
> Page: 6
>
> Checking rounding on multiply, divide and add/subtract.
> ! F* is neither chopped nor correctly rounded.
> ! F/ is neither chopped nor correctly rounded.
> ! Addition/Subtraction neither rounds nor chops.
> ! Sticky bit used incorrectly or not at all.
> ! FLAW lack(s) of guard digits or failure(s) to correctly round or chop
> (noted above) count as one flaw in the final tally below.
> ...
> Test for sqrt monotonicity.
> sqrt has passed a test for Monotonicity.
> Testing whether sqrt is rounded or chopped.
> ! Square root is neither chopped nor correctly rounded.
> ! Observed errors run from -5.0000E-1 to 5.0000E-1 ulps.
> Diagnosis resumes after milestone Number 90
> ...
> This computed value is O.K.
>
> Testing X^((X + 1) / (X - 1)) vs. exp(2) = 7.3891E0 as X -> 1.
> ! DEFECT Calculated 7.3891E0 for
> ! (1 + (-1.1102E-16 ) ^ (-1.8014E16 );
> ! differs from correct value by -3.4413E-9
> ! This much error may spoil financial
> ! calculations involving tiny interest rates.
> Testing powers Z^Q at four nearly extreme values.
> ... no discrepancies found.
> ...
> ! DEFECTs discovered = 1
> ! FLAWs discovered = 1
>
> ! The arithmetic diagnosed may be Acceptable
> ! despite inconvenient Defects.
>
>
>
> I modified the code
The test code in paranoia.4th presumably
so that flaws merely emit warnings, and won't cause a
> fail, but the defects still flag it as failed, and I don't feel like
> turning defects into a warning ;-).
>
Of course you're free to modify your copy of tests for your own
convenience but I'd rather see what other systems do with those tests
before changing the master files.
> On the other hand, running those tests already revealed two bugs in Gforth
> (one in the dynamic code generator, and one in our ecvt_r replacement,
> which couldn't handle denormals with sufficient accuracy).
>
That's good to know. Once again thanks for the feedback.
--
Gerry
Back to comp.lang.forth | Previous | Next — Previous in thread | Next in thread | Find similar | Unroll thread
Gforth snapshot for Windows Bernd Paysan <bernd@net2o.de> - 2016-09-30 01:33 +0000
Re: Gforth snapshot for Windows Gerry Jackson <gerry@jackson9000.fsnet.co.uk> - 2016-09-30 15:16 +0100
Re: Gforth snapshot for Windows anton@mips.complang.tuwien.ac.at (Anton Ertl) - 2016-09-30 16:26 +0000
Re: Gforth snapshot for Windows Gerry Jackson <gerry@jackson9000.fsnet.co.uk> - 2016-09-30 19:50 +0100
Re: Gforth snapshot for Windows Bernd Paysan <bernd@net2o.de> - 2016-09-30 20:34 +0000
Re: Gforth snapshot for Windows Gerry Jackson <gerry@jackson9000.fsnet.co.uk> - 2016-10-02 11:14 +0100
Re: Gforth snapshot for Windows anton@mips.complang.tuwien.ac.at (Anton Ertl) - 2016-10-02 16:29 +0000
Re: Gforth snapshot for Windows Bernd Paysan <bernd@net2o.de> - 2016-10-02 22:47 +0000
Re: Gforth snapshot for Windows jjmail9@gmail.com - 2016-10-19 04:45 -0700
Re: Gforth snapshot for Windows Rod Pemberton <NoHaveNotOne@zxdehrnyya.cam> - 2016-09-30 17:28 -0400
Re: Gforth snapshot for Windows Bernd Paysan <bernd@net2o.de> - 2016-09-30 22:14 +0000
Re: Gforth snapshot for Windows Gerry Jackson <gerry@jackson9000.fsnet.co.uk> - 2016-10-01 19:42 +0100
Re: Gforth snapshot for Windows anton@mips.complang.tuwien.ac.at (Anton Ertl) - 2016-10-01 07:08 +0000
Re: Gforth snapshot for Windows Gerry Jackson <gerry@jackson9000.fsnet.co.uk> - 2016-10-02 11:23 +0100
Re: Gforth snapshot for Windows anton@mips.complang.tuwien.ac.at (Anton Ertl) - 2016-10-02 16:14 +0000
Re: Gforth snapshot for Windows Cecil Bayona <cbayona@cbayona.com> - 2016-10-02 12:14 -0500
Re: Gforth snapshot for Windows albert@cherry.spenarnc.xs4all.nl (Albert van der Horst) - 2016-10-02 19:45 +0200
Re: Gforth snapshot for Windows anton@mips.complang.tuwien.ac.at (Anton Ertl) - 2016-10-03 07:43 +0000
Re: Gforth snapshot for Windows alex <alex@rivadpm.com> - 2016-10-03 12:51 +0100
Re: Gforth snapshot for Windows Gerry Jackson <gerry@jackson9000.fsnet.co.uk> - 2016-10-02 12:07 +0100
Re: Gforth snapshot for Windows Bernd Paysan <bernd@net2o.de> - 2016-10-02 22:39 +0000
Re: Gforth snapshot for Windows Paul Rubin <no.email@nospam.invalid> - 2016-09-30 14:36 -0700
Re: Gforth snapshot for Windows anton@mips.complang.tuwien.ac.at (Anton Ertl) - 2016-10-01 07:46 +0000
Re: Gforth snapshot for Windows Paul Rubin <no.email@nospam.invalid> - 2016-10-01 12:12 -0700
Re: Gforth snapshot for Windows lawrencedo99@gmail.com - 2016-10-02 19:51 -0700
Re: Gforth snapshot for Windows Bernd Paysan <bernd@net2o.de> - 2016-09-30 22:09 +0000
Re: Gforth snapshot for Windows "Elizabeth D. Rather" <erather@forth.com> - 2016-09-30 13:25 -1000
Re: Gforth snapshot for Windows Bernd Paysan <bernd@net2o.de> - 2016-10-02 00:10 +0000
Re: Gforth snapshot for Windows "Elizabeth D. Rather" <erather@forth.com> - 2016-10-01 15:31 -1000
Re: Gforth snapshot for Windows Bernd Paysan <bernd@net2o.de> - 2016-10-02 23:00 +0000
Re: Gforth snapshot for Windows "Elizabeth D. Rather" <erather@forth.com> - 2016-10-02 21:32 -1000
Re: Gforth snapshot for Windows Cecil Bayona <cbayona@cbayona.com> - 2016-10-03 03:46 -0500
Re: Gforth snapshot for Windows "Elizabeth D. Rather" <erather@forth.com> - 2016-10-02 23:25 -1000
Re: Gforth snapshot for Windows albert@cherry.spenarnc.xs4all.nl (Albert van der Horst) - 2016-10-03 12:57 +0200
Re: Gforth snapshot for Windows lawrencedo99@gmail.com - 2016-10-03 04:31 -0700
Re: Gforth snapshot for Windows anton@mips.complang.tuwien.ac.at (Anton Ertl) - 2016-10-04 09:18 +0000
Re: Gforth snapshot for Windows Bernd Paysan <bernd@net2o.de> - 2016-10-04 21:57 +0000
Re: Gforth snapshot for Windows jjmail9@gmail.com - 2016-10-19 06:47 -0700
Re: Gforth snapshot for Windows Bernd Paysan <bernd@net2o.de> - 2016-10-21 20:22 +0000
Re: Gforth snapshot for Windows "Elizabeth D. Rather" <erather@forth.com> - 2016-10-04 12:40 -1000
Re: Gforth snapshot for Windows jjmail9@gmail.com - 2016-10-19 08:37 -0700
Re: Gforth snapshot for Windows Cecil Bayona <cbayona@cbayona.com> - 2016-10-19 11:29 -0500
Re: Gforth snapshot for Windows jjmail9@gmail.com - 2016-10-19 12:03 -0700
Re: Gforth snapshot for Windows anton@mips.complang.tuwien.ac.at (Anton Ertl) - 2016-10-20 12:15 +0000
Re: Gforth snapshot for Windows "Elizabeth D. Rather" <erather@forth.com> - 2016-10-20 07:36 -1000
Re: Gforth snapshot for Windows anton@mips.complang.tuwien.ac.at (Anton Ertl) - 2016-10-21 07:15 +0000
Re: Gforth snapshot for Windows "Elizabeth D. Rather" <erather@forth.com> - 2016-10-21 22:12 -1000
Re: Gforth snapshot for Windows albert@cherry.spenarnc.xs4all.nl (Albert van der Horst) - 2016-10-22 12:12 +0200
Re: Gforth snapshot for Windows anton@mips.complang.tuwien.ac.at (Anton Ertl) - 2016-10-22 13:52 +0000
Re: Gforth snapshot for Windows jjmail9@gmail.com - 2016-10-19 05:57 -0700
Re: Gforth snapshot for Windows Jason Damisch <jasondamisch@yahoo.com> - 2016-09-30 10:17 -0700
Re: Gforth snapshot for Windows Philip Smith <philipkingsmith@gmail.com> - 2016-09-30 11:45 -0700
Re: Gforth snapshot for Windows beeflo <beeflobill@gmail.com> - 2016-10-11 10:51 -0700
Re: Gforth snapshot for Windows Bernd Paysan <bernd@net2o.de> - 2016-10-11 20:44 +0000
Re: Gforth snapshot for Windows Jan Coombs <jenfhaomndgfwutc@murmic.plus.com> - 2016-10-12 01:02 +0100
Re: Gforth snapshot for Windows Bernd Paysan <bernd@net2o.de> - 2016-10-13 00:26 +0000
Re: Gforth snapshot for Windows Jan Coombs <jenfhaomndgfwutc@murmic.plus.com> - 2016-10-13 08:48 +0100
Re: Gforth snapshot for Windows lawrencedo99@gmail.com - 2016-10-13 01:09 -0700
Re: Gforth snapshot for Windows Bernd Paysan <bernd@net2o.de> - 2016-10-14 23:39 +0000
Re: Gforth snapshot for Windows Gerry Jackson <gerry@jackson9000.fsnet.co.uk> - 2016-10-14 18:05 +0100
Re: Gforth snapshot for Windows Bernd Paysan <bernd@net2o.de> - 2016-10-14 23:29 +0000
Re: Gforth snapshot for Windows anton@mips.complang.tuwien.ac.at (Anton Ertl) - 2016-10-15 16:26 +0000
Re: Gforth snapshot for Windows Bernd Paysan <bernd@net2o.de> - 2016-10-17 00:52 +0000
Re: Gforth snapshot for Windows Gerry Jackson <gerry@jackson9000.fsnet.co.uk> - 2016-10-19 10:50 +0100
Re: Gforth snapshot for Windows Spiros Lakkos <slakkos@hotmail.com> - 2016-10-12 06:50 +0300
Re: Gforth snapshot for Windows lawrencedo99@gmail.com - 2016-10-11 22:42 -0700
Re: Gforth snapshot for Windows beeflo <beeflobill@gmail.com> - 2016-10-12 09:41 -0700
Re: Gforth snapshot for Windows Bernd Paysan <bernd@net2o.de> - 2016-10-12 23:46 +0000
csiph-web