llvm dev - Sep 2016 - [PPC, APFloat] Add full PPCDoubleDouble to APFloat

I have found some internal test failures due to the wrong constant folding
on ppc_fp128.

As documented in APFloat::PPCDoubleDouble, APFloat doesn't support PowerPC
double-double correctly <
https://github.com/llvm-mirror/llvm/blob/492acdd450bcdf9837494d6da029ed064f14fc33/lib/Support/APFloat.cpp#L74
>.
To support this, we need to add a second tuple of (sign, exponent,
significand) to APFloat. I wonder where should I start to change, so that
it's less hacky? I certainly expect refactoring to come. :)

Thanks!
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20160930/789e507d/attachment.html>

Hi Tim,

How, in general, are you thinking about doing this? I ask because, as you
clearly know, the double-double format is formed by the sum of two
double-precision numbers, and the various arithmetic operations are formed
mostly in terms of double-precision arithmetic on the elements of the pairs. As
a result, I suspect we really just want to store the two double-precision
numbers, and essentially delegate to APFloat IEEEdouble in the implementation of
the various operations.

As such, one option might be that, when PPCDoubleDouble fltSemantics are
selected, the exponent and sign are ignored, and we just store the data for the
two double-precision floating-point numbers in the significand. We might even do
this directly, by making the significand something like this:

  union Significand {
    integerPart part;
    integerPart *parts;
    APFloat *fltparts; // used for PPCDoubleDouble
  } significand;

Thanks again,
Hal

----- Original Message -----
> From: "Tim Shen" <timshen at google.com>
> To: llvm-dev at lists.llvm.org
> Cc: resistor at mac.com, "Hal Finkel" <hfinkel at anl.gov>,
"Fiona Glaser" <escha at apple.com>, "Stephen Canon"
> <scanon at apple.com>, "Eric Christopher" <echristo at
gmail.com>
> Sent: Friday, September 30, 2016 4:43:12 PM
> Subject: [PPC, APFloat] Add full PPCDoubleDouble to APFloat
> 
> 
> 
> 
> I have found some internal test failures due to the wrong constant
> folding on ppc_fp128.
> 
> 
> As documented in APFloat::PPCDoubleDouble, APFloat doesn't support
> PowerPC double-double correctly <
>
https://github.com/llvm-mirror/llvm/blob/492acdd450bcdf9837494d6da029ed064f14fc33/lib/Support/APFloat.cpp#L74
> >.
> 
> 
> To support this, we need to add a second tuple of (sign, exponent,
> significand) to APFloat. I wonder where should I start to change, so
> that it's less hacky? I certainly expect refactoring to come. :)
> 
> 
> Thanks!
-- 
Hal Finkel
Lead, Compiler Technology and Programming Languages
Leadership Computing Facility
Argonne National Laboratory

Hi Hal,

On Sun, Oct 2, 2016 at 7:43 PM Hal Finkel <hfinkel at anl.gov> wrote:
> Hi Tim,
>
> How, in general, are you thinking about doing this? I ask because, as you
> clearly know, the double-double format is formed by the sum of two
> double-precision numbers, and the various arithmetic operations are formed
> mostly in terms of double-precision arithmetic on the elements of the
> pairs. As a result, I suspect we really just want to store the two
> double-precision numbers, and essentially delegate to APFloat IEEEdouble in
> the implementation of the various operations.
>
> As such, one option might be that, when PPCDoubleDouble fltSemantics are
> selected, the exponent and sign are ignored, and we just store the data for
> the two double-precision floating-point numbers in the significand. We
> might even do this directly, by making the significand something like this:
>
>   union Significand {
>     integerPart part;
>     integerPart *parts;
>     APFloat *fltparts; // used for PPCDoubleDouble
>   } significand;
>
We can do this, but my concern is "what's next?". What do we do
to significandParts()? It doesn't make sense to return an array of
integerParts for PPCDoubleDouble. Do we examine every call site
of significandParts(), and assert/dispatch on fltSemantics? Do we want to
build an abstraction between APFloat and the underlying data
representation? For example:

class APFloatPayload {
  union Significand { integerPart part; integerPart *parts; } significand;
  int exponent;

public:
  // *Significand() operations.
};

class APFloat {
  fltSemantics *semantics;
  fltCategory category;
  APFloatPayload first;
  Optional<APFloatPayload> second;  // for PPCDoubleDouble
};

>
> Thanks again,
> Hal
>
> ----- Original Message -----
> > From: "Tim Shen" <timshen at google.com>
> > To: llvm-dev at lists.llvm.org
> > Cc: resistor at mac.com, "Hal Finkel" <hfinkel at
anl.gov>, "Fiona Glaser" <
> escha at apple.com>, "Stephen Canon"
> > <scanon at apple.com>, "Eric Christopher" <echristo
at gmail.com>
> > Sent: Friday, September 30, 2016 4:43:12 PM
> > Subject: [PPC, APFloat] Add full PPCDoubleDouble to APFloat
> >
> >
> >
> >
> > I have found some internal test failures due to the wrong constant
> > folding on ppc_fp128.
> >
> >
> > As documented in APFloat::PPCDoubleDouble, APFloat doesn't support
> > PowerPC double-double correctly <
> >
>
https://github.com/llvm-mirror/llvm/blob/492acdd450bcdf9837494d6da029ed064f14fc33/lib/Support/APFloat.cpp#L74
> > >.
> >
> >
> > To support this, we need to add a second tuple of (sign, exponent,
> > significand) to APFloat. I wonder where should I start to change, so
> > that it's less hacky? I certainly expect refactoring to come. :)
> >
> >
> > Thanks!
>
> --
> Hal Finkel
> Lead, Compiler Technology and Programming Languages
> Leadership Computing Facility
> Argonne National Laboratory
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL:
<http://lists.llvm.org/pipermail/llvm-dev/attachments/20161003/6b9925f8/attachment.html>