Re: iOS LDC 0.16.1 (2.067.1) binaries available

[Joakim via Digitalmars-d-announce]

On Saturday, 7 November 2015 at 20:34:06 UTC, Dan Olson wrote:

Joakim  writes:
Hmm, that's strange, this commit didn't fix the 64-bit issues 
for you? I believe it fixed them for me on Android/ARM:


https://github.com/ldc-developers/phobos/commit/9d1b49578ffa4b2e848159cfe5afe80b5e4c26eb


Yes, but there is still one case not working for both iOS armv7 
and arm64 in 0.16.1.  It is only off by one ulp so I'll make a 
PR for that.


https://github.com/ldc-developers/phobos/blob/ldc/std/internal/math/gammafunction.d#L540


OK, somehow doesn't assert for me on Android/ARMv7.


And then this new stuff in 2.068 is failing in various places.

https://github.com/ldc-developers/phobos/commit/9b86eebed53c800a58dfa7e065dcb9e11cdae5c5


Yeah, that's the new function I mentioned earlier.  Initializing 
the constant maxY ends up calling log2 through CTFE, the 
intrinsic for which doesn't exist in Kai's longdouble2 branch.  I 
was going to look at writing one, feel free to beat me to it. ;)


Or if you added that log2 to your branch already, could be other 
issues too, haven't gotten that far.

Re: iOS LDC 0.16.1 (2.067.1) binaries available

[Joakim via Digitalmars-d-announce]

On Saturday, 7 November 2015 at 19:20:02 UTC, Dan Olson wrote:

Dan Olson  writes:


Joakim  writes:
btw, std.internal.math.gammafunction hasn't given me a 
problem since 2.067.1, the Win64 guys fixed it.  2.068 added 
a function that needs a CTFE-able 64-bit log2, but other than 
that, it just works now.  You may want to revert your patch 
for that module and try it.


Thanks for the tip, I'll check it out.  It is the one 
remaining bad boy.


Still bad :-(


Hmm, that's strange, this commit didn't fix the 64-bit issues for 
you?  I believe it fixed them for me on Android/ARM:


https://github.com/ldc-developers/phobos/commit/9d1b49578ffa4b2e848159cfe5afe80b5e4c26eb

Re: iOS LDC 0.16.1 (2.067.1) binaries available

[Dan Olson via Digitalmars-d-announce]

Joakim  writes:

> On Saturday, 7 November 2015 at 19:20:02 UTC, Dan Olson wrote:
>> Dan Olson  writes:
>>
>>> Joakim  writes:
 btw, std.internal.math.gammafunction hasn't given me a problem
 since 2.067.1, the Win64 guys fixed it.  2.068 added a function
 that needs a CTFE-able 64-bit log2, but other than that, it just
 works now.  You may want to revert your patch for that module and
 try it.
>>>
>>> Thanks for the tip, I'll check it out.  It is the one remaining bad
>>> boy.
>>
>> Still bad :-(
>
> Hmm, that's strange, this commit didn't fix the 64-bit issues for you?
> I believe it fixed them for me on Android/ARM:
>
> https://github.com/ldc-developers/phobos/commit/9d1b49578ffa4b2e848159cfe5afe80b5e4c26eb

Yes, but there is still one case not working for both iOS armv7 and
arm64 in 0.16.1.  It is only off by one ulp so I'll make a PR for that.

https://github.com/ldc-developers/phobos/blob/ldc/std/internal/math/gammafunction.d#L540

And then this new stuff in 2.068 is failing in various places.

https://github.com/ldc-developers/phobos/commit/9b86eebed53c800a58dfa7e065dcb9e11cdae5c5

Re: iOS LDC 0.16.1 (2.067.1) binaries available

[Dan Olson via Digitalmars-d-announce]

Dan Olson  writes:

> Joakim  writes:
>> btw, std.internal.math.gammafunction hasn't given me a problem since
>> 2.067.1, the Win64 guys fixed it.  2.068 added a function that needs a
>> CTFE-able 64-bit log2, but other than that, it just works now.  You
>> may want to revert your patch for that module and try it.
>
> Thanks for the tip, I'll check it out.  It is the one remaining bad boy.

Still bad :-(

Re: iOS LDC 0.16.1 (2.067.1) binaries available

[Dan Olson via Digitalmars-d-announce]

Joakim  writes:

> On Saturday, 7 November 2015 at 20:34:06 UTC, Dan Olson wrote:
>> Joakim  writes:
>>> Hmm, that's strange, this commit didn't fix the 64-bit issues for
>>> you? I believe it fixed them for me on Android/ARM:
>>>
>>> https://github.com/ldc-developers/phobos/commit/9d1b49578ffa4b2e848159cfe5afe80b5e4c26eb
>>
>> Yes, but there is still one case not working for both iOS armv7 and
>> arm64 in 0.16.1.  It is only off by one ulp so I'll make a PR for
>> that.
>>
>> https://github.com/ldc-developers/phobos/blob/ldc/std/internal/math/gammafunction.d#L540
>
> OK, somehow doesn't assert for me on Android/ARMv7.

Different math libs I suspect is the cause.

>> And then this new stuff in 2.068 is failing in various places.
>>
>> https://github.com/ldc-developers/phobos/commit/9b86eebed53c800a58dfa7e065dcb9e11cdae5c5
>
> Yeah, that's the new function I mentioned earlier.  Initializing the
> constant maxY ends up calling log2 through CTFE, the intrinsic for
> which doesn't exist in Kai's longdouble2 branch.  I was going to look
> at writing one, feel free to beat me to it. ;)
>
> Or if you added that log2 to your branch already, could be other
> issues too, haven't gotten that far.

I don't use longdouble2 branch currently, but only because I had another
simpler 64-bit specific solution.  Yes, I had to add log() to it for
CTFE.

longdouble2 does need to get into LDC to make general support of
cross-compilers easier.  Kai's main concern was about CTFE accuracy for
functions like sin().  Someone or us should start on that.

I traced through the new gammafunction errors and several are due to
subnormals getting fed to log(). Many iOS math functions treat
subnormals as zero, producing different results.  I just have to find
them and revise the unitttests.

Re: iOS LDC 0.16.1 (2.067.1) binaries available

[Joakim via Digitalmars-d-announce]

On Thursday, 5 November 2015 at 07:44:48 UTC, Dan Olson wrote:
This is another set of binaries and universal libs for the 
experimental LDC iOS cross-compiler.  It is now based on LDC 
0.16.1 (2.067.1) and LLVM 3.6.2.


https://github.com/smolt/ldc-iphone-dev/releases/tag/ios-0.16.1-151104


btw, std.internal.math.gammafunction hasn't given me a problem 
since 2.067.1, the Win64 guys fixed it.  2.068 added a function 
that needs a CTFE-able 64-bit log2, but other than that, it just 
works now.  You may want to revert your patch for that module and 
try it.

Re: iOS LDC 0.16.1 (2.067.1) binaries available

[Dan Olson via Digitalmars-d-announce]

Joakim  writes:
> btw, std.internal.math.gammafunction hasn't given me a problem since
> 2.067.1, the Win64 guys fixed it.  2.068 added a function that needs a
> CTFE-able 64-bit log2, but other than that, it just works now.  You
> may want to revert your patch for that module and try it.

Thanks for the tip, I'll check it out.  It is the one remaining bad boy.

Re: iOS LDC 0.16.1 (2.067.1) binaries available

[Dan Olson via Digitalmars-d-announce]

Just a few more notes:

Dan Olson  writes:
> Feedback appreciated, especially since iOS support may get into official
> LDC sometime soon.  This is a good time to voice an opinion on the
> approach.

Just noticed that tvOS and watchOS are now present in LLVM, so I think
support for these could be added to LDC soon too.

> Xcode is not D aware and a working plugin is needed.  In the meantime,
> xc-iphoneos-dc in the bin dir can be used as a custom *.d build script.
> Or you can compile D source externally and add your libraries/object
> files to an Xcode project.

An Xcode plugin for D is badly needed.  I think someone could get this
one functioning: https://github.com/michelf/d-for-xcode

Re: iOS LDC 0.16.1 (2.067.1) binaries available

[Joakim via Digitalmars-d-announce]

Great, your last announcement was linked in reddit comments about 
the 2.069 release, when asked about iOS support.


On Thursday, 5 November 2015 at 08:05:39 UTC, Dan Olson wrote:
Just noticed that tvOS and watchOS are now present in LLVM, so 
I think support for these could be added to LDC soon too.


WatchOS and tvOS are bitcode-only, right?  Have they specified 
their bitcode format yet or is it just whatever clang spits out?  
I thought there were problems with that because of bitcode format 
changes over time and other platform issues, that the PNaCl guys 
had to work on solving.  I wonder if you'll have similar issues 
with the bitcode ldc spits out.

Re: iOS LDC 0.16.1 (2.067.1) binaries available

[Joakim via Digitalmars-d-announce]

On Thursday, 5 November 2015 at 15:45:35 UTC, Dan Olson wrote:
tvOS is essentially iOS and doesn't require bitcode (yet) like 
watchOS. I am looking at adding it soon because Xcode 7 enables 
it by default.


I just looked it up, their official docs say bitcode is required 
for both tvOS and watchOS:


https://developer.apple.com/library/watchos/documentation/IDEs/Conceptual/AppDistributionGuide/AppThinning/AppThinning.html

I don't totally appreciate all the possible bitcode problems, 
but one of the suggestions that apps can be updated for new 
CPUs without rebuilding doesn't make sense.  The IR/bitcode 
from clang for arm64 and armv7 is different.  They would have 
to make all the ABIs identical first or have LLVM backend do 
more of the ABI work.


Yeah, that's what google did with PNaCl, stabilize on one format 
and make it as architecture-agnostic as possible:


http://www.phoronix.com/scan.php?page=news_item=MTQyNTE

Maybe Apple doesn't really care about running on different 
architectures, but I wonder what they're doing to handle changes 
in the llvm bitcode format over time.

Re: iOS LDC 0.16.1 (2.067.1) binaries available

[Dan Olson via Digitalmars-d-announce]

Joakim  writes:

> Great, your last announcement was linked in reddit comments about the
> 2.069 release, when asked about iOS support.
>
> On Thursday, 5 November 2015 at 08:05:39 UTC, Dan Olson wrote:
>> Just noticed that tvOS and watchOS are now present in LLVM, so I
>> think support for these could be added to LDC soon too.
>
> WatchOS and tvOS are bitcode-only, right?  Have they specified their
> bitcode format yet or is it just whatever clang spits out?  I thought
> there were problems with that because of bitcode format changes over
> time and other platform issues, that the PNaCl guys had to work on
> solving.  I wonder if you'll have similar issues with the bitcode ldc
> spits out.

tvOS is essentially iOS and doesn't require bitcode (yet) like watchOS.
I am looking at adding it soon because Xcode 7 enables it by default.

I don't totally appreciate all the possible bitcode problems, but one of
the suggestions that apps can be updated for new CPUs without rebuilding
doesn't make sense.  The IR/bitcode from clang for arm64 and armv7 is
different.  They would have to make all the ABIs identical first or have
LLVM backend do more of the ABI work.
-- 
Dan