Libre RISC-V Mailing list Archives (accessible via public-inbox)
 help / color / Atom feed
* Re: [libre-riscv-dev] POWER virtual coffee 5 mins (now)
From: Roberto Innocenti @ 2020-04-16 22:04 UTC (permalink / raw)
  To: lkcl .; +Cc: Libre-RISCV General Development
In-Reply-To: <CAPweEDweHUYVnECNyN+9mmrrYTRg5G+0r737MTQLkz5euoZ8JQ@mail.gmail.com>

We are online now
Roberto

On apr 9 2020, at 11:56 pm, lkcl . <luke.leighton@gmail.com> wrote:

> https://openpowerfoundation.org/openpower-virtual-coffee-calls/

_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* [libre-riscv-dev] [Bug 137] NLNet 2019 Video Acceleration Proposal 2019-10-031
From: bugzilla-daemon @ 2020-04-18  7:03 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-137-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=137

--- Comment #21 from Michael Pham <pham.michael.98@gmail.com> ---
(In reply to cand from comment #20)
> I believe it's too late to add AAC, however due to how ffmpeg is structured,
> some of the work on the other codecs will speed up AAC too. Not as much as
> focusing on it, but more than plain sw.
> 

Ah ok, at least it will get some treatment.

> MPEG2, MP3, AC3 are all free, patents expired. Vorbis and Opus were
> explicitly designed for that.
> 
Ok, good to hear the other audio formats are safe!

> For the newer video codecs, it's possible any implementation would infringe
> patents. So does 90% of ffmpeg code. IANAL, but only the sw patents I
> believe, as the hw blocks we will have will not be specialized to any codec.
> Unlike a modern GPU that has "H.264 frame in, RGB out" blocks, we will have
> sub-operations such as "calculate transform XYZ for this data". Those are
> not patentable in general, some specific algorithm in hw may be.
> 
> So my conclusion is that the hw is safe, but if a commercial entity wishes
> to ship our software in the US, they will need to disable the newer video
> codecs or to license patents. I.e. the exact same situation as ffmpeg code,
> if they want to ship that in a product with new stuff enabled.

I'm not a lawyer either, though your conclusion sounds reasonable to me. :)

Patents are a minefield to navigate (sigh)

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* [libre-riscv-dev] [Bug 137] NLNet 2019 Video Acceleration Proposal 2019-10-031
From: bugzilla-daemon @ 2020-04-18  6:43 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-137-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=137

--- Comment #20 from cand@gmx.com ---
I believe it's too late to add AAC, however due to how ffmpeg is structured,
some of the work on the other codecs will speed up AAC too. Not as much as
focusing on it, but more than plain sw.

MPEG2, MP3, AC3 are all free, patents expired. Vorbis and Opus were explicitly
designed for that.

For the newer video codecs, it's possible any implementation would infringe
patents. So does 90% of ffmpeg code. IANAL, but only the sw patents I believe,
as the hw blocks we will have will not be specialized to any codec. Unlike a
modern GPU that has "H.264 frame in, RGB out" blocks, we will have
sub-operations such as "calculate transform XYZ for this data". Those are not
patentable in general, some specific algorithm in hw may be.

So my conclusion is that the hw is safe, but if a commercial entity wishes to
ship our software in the US, they will need to disable the newer video codecs
or to license patents. I.e. the exact same situation as ffmpeg code, if they
want to ship that in a product with new stuff enabled.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* [libre-riscv-dev] [Bug 137] NLNet 2019 Video Acceleration Proposal 2019-10-031
From: bugzilla-daemon @ 2020-04-18  0:34 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-137-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=137

Michael Pham <pham.michael.98@gmail.com> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
                 CC|                            |pham.michael.98@gmail.com

--- Comment #19 from Michael Pham <pham.michael.98@gmail.com> ---
I see that you already sent this to NLNet, but I hope it's not too late to add
AAC to the list of Audio Formats to accelerate. The reason being is that it is
one of the primary audio formats used on YouTube and also has common use as a
Bluetooth audio format. I would even argue that AAC is even more widely used
than AC3 and Vorbis and Opus at least in my experience as an average user.

In regards to the video formats, what is the legality of putting into the
hardware to help decoding for MPEG2, MPEG4, H.264, and H.265? I know these are
at least covered by the patent pool of MPEG LA (
https://en.wikipedia.org/wiki/MPEG_LA ). Some of the audio codecs may or may
not also be tangled in legalities (not sure, I haven't really looked into it
for the audio formats).

Would it be prudent to get legal advice first? Or do you feel it is safe doing
this?

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* [libre-riscv-dev] [Bug 208] implement CORDIC in a general way sufficient to do transcendentals
From: bugzilla-daemon @ 2020-04-17 19:59 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-208-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=208

--- Comment #32 from Jacob Lifshay <programmerjake@gmail.com> ---
(In reply to Michael Nolan from comment #29)
> (In reply to Jacob Lifshay from comment #28)
> > Please also add the libgmp-dev package (or whatever the correct name is) to
> > .gitlab-ci.yml
> 
> Done.

You used tabs instead of spaces, which broke the yaml parser, please fix.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* [libre-riscv-dev] [Bug 208] implement CORDIC in a general way sufficient to do transcendentals
From: bugzilla-daemon @ 2020-04-17 19:56 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-208-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=208

--- Comment #31 from Jacob Lifshay <programmerjake@gmail.com> ---
(In reply to Luke Kenneth Casson Leighton from comment #30)
> it's sorted now however i just found this:  python3-mpmath
> it's a pure-python-only arbitrary-precision FP math library.  it doesn't
> say whether it's IEEE754 compliant however.

ieee754 compliant basically means the results are rounded properly to an
ieee754 binary/decimal format, mpmath is instead designed to give as many
digits of accuracy as you want, it's designed to emulate mathematical
real/complex numbers rather than ieee 754 arithmetic.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* [libre-riscv-dev] [Bug 208] implement CORDIC in a general way sufficient to do transcendentals
From: bugzilla-daemon @ 2020-04-17 19:36 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-208-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=208

--- Comment #30 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
it's sorted now however i just found this:  python3-mpmath
it's a pure-python-only arbitrary-precision FP math library.  it doesn't
say whether it's IEEE754 compliant however.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* [libre-riscv-dev] [Bug 208] implement CORDIC in a general way sufficient to do transcendentals
From: bugzilla-daemon @ 2020-04-17 18:42 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-208-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=208

--- Comment #29 from Michael Nolan <mtnolan2640@gmail.com> ---
(In reply to Luke Kenneth Casson Leighton from comment #27)
> with bf.precision(300): # use 300 bits of precision
>     x = bf.atan(BigFloat(2) ....)
> 
> which is why i recommended bigfloat: you can specify extremely large
> FP precision.

Yup. I went with quad precision for now.


(In reply to Jacob Lifshay from comment #28)
> Please also add the libgmp-dev package (or whatever the correct name is) to
> .gitlab-ci.yml

Done.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* [libre-riscv-dev] [Bug 208] implement CORDIC in a general way sufficient to do transcendentals
From: bugzilla-daemon @ 2020-04-17 18:38 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-208-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=208

--- Comment #28 from Jacob Lifshay <programmerjake@gmail.com> ---
(In reply to Michael Nolan from comment #26)
> Ok, I reverted the commit containing the gmpy stuff and added a similar one
> calculating the atan tables with bigfloat

Please also add the libgmp-dev package (or whatever the correct name is) to
.gitlab-ci.yml

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* [libre-riscv-dev] [Bug 208] implement CORDIC in a general way sufficient to do transcendentals
From: bugzilla-daemon @ 2020-04-17 18:37 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-208-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=208

--- Comment #27 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
(In reply to Michael Nolan from comment #26)
> (In reply to Luke Kenneth Casson Leighton from comment #24)
> > https://pypi.org/project/bigfloat/
> > 
> > i used python bigfloat about... 8 years ago, and just updated it:
> > it works well.  it also happens to work in a similar way to
> > python-gmpy2 by cython-wrapping mpfr... except that the debian
> > package maintainer for python-gmpy2 is not experienced enough
> > to do multi-python-version compiles.
> 
> Ok, I reverted the commit containing the gmpy stuff and added a similar one
> calculating the atan tables with bigfloat

with the debian package being a mess that's probably the sensible option
for now.

there's a trick btw, i think it's:

with bf.precision(300): # use 300 bits of precision
    x = bf.atan(BigFloat(2) ....)

which is why i recommended bigfloat: you can specify extremely large
FP precision.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* [libre-riscv-dev] [Bug 208] implement CORDIC in a general way sufficient to do transcendentals
From: bugzilla-daemon @ 2020-04-17 18:02 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-208-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=208

--- Comment #26 from Michael Nolan <mtnolan2640@gmail.com> ---
(In reply to Luke Kenneth Casson Leighton from comment #24)
> https://pypi.org/project/bigfloat/
> 
> i used python bigfloat about... 8 years ago, and just updated it:
> it works well.  it also happens to work in a similar way to
> python-gmpy2 by cython-wrapping mpfr... except that the debian
> package maintainer for python-gmpy2 is not experienced enough
> to do multi-python-version compiles.

Ok, I reverted the commit containing the gmpy stuff and added a similar one
calculating the atan tables with bigfloat

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* [libre-riscv-dev] [Bug 208] implement CORDIC in a general way sufficient to do transcendentals
From: bugzilla-daemon @ 2020-04-17 17:31 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-208-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=208

--- Comment #25 from Jacob Lifshay <programmerjake@gmail.com> ---
I can add sin, cos, tan, etc. support to simple-soft-float.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* [libre-riscv-dev] [Bug 208] implement CORDIC in a general way sufficient to do transcendentals
From: bugzilla-daemon @ 2020-04-17 17:28 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-208-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=208

--- Comment #24 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
https://pypi.org/project/bigfloat/

i used python bigfloat about... 8 years ago, and just updated it:
it works well.  it also happens to work in a similar way to
python-gmpy2 by cython-wrapping mpfr... except that the debian
package maintainer for python-gmpy2 is not experienced enough
to do multi-python-version compiles.

python-bigfloat i specifically recommended because it has been
"wrapped" with additional functionality that makes it near-identical
to both the standard python "math" module *and* to the standard
python "float".

    from bigfloat import cos, sin, tan

therefore works

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* [libre-riscv-dev] [Bug 208] implement CORDIC in a general way sufficient to do transcendentals
From: bugzilla-daemon @ 2020-04-17 17:19 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-208-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=208

--- Comment #23 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
argh, michael, python3-gmpy2 is not properly packaged on debian.

i can't even *install* python3-gmpy2 for python3.7!

 apt-get install python3-gmpy2=2.1.0~a4-1
Reading package lists... Done
Building dependency tree       
Reading state information... Done
Some packages could not be installed. This may mean that you have
requested an impossible situation or if you are using the unstable
distribution that some required packages have not yet been created
or been moved out of Incoming.
The following information may help to resolve the situation:

The following packages have unmet dependencies:
 python3-gmpy2 : Depends: python3 (< 3.8) but 3.8.2-3 is to be installed
E: Unable to correct problems, you have held broken packages.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* [libre-riscv-dev] [Bug 287] corrections to commit f17100d6c435aec47b17acb188272a7e9b3cf67d
From: bugzilla-daemon @ 2020-04-17 16:41 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-287-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=287

Luke Kenneth Casson Leighton <lkcl@lkcl.net> changed:

           What    |Removed                     |Added
----------------------------------------------------------------------------
           Assignee|lkcl@lkcl.net               |libre-soc@platen-software.d
                   |                            |e

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* [libre-riscv-dev] [Bug 287] New: corrections to commit f17100d6c435aec47b17acb188272a7e9b3cf67d
From: bugzilla-daemon @ 2020-04-17 16:41 UTC (permalink / raw)
  To: libre-riscv-dev

https://bugs.libre-soc.org/show_bug.cgi?id=287

            Bug ID: 287
           Summary: corrections to commit
                    f17100d6c435aec47b17acb188272a7e9b3cf67d
           Product: Libre-SOC's first SoC
           Version: unspecified
          Hardware: PC
                OS: Linux
            Status: CONFIRMED
          Severity: enhancement
          Priority: ---
         Component: Source Code
          Assignee: lkcl@lkcl.net
          Reporter: lkcl@lkcl.net
                CC: libre-riscv-dev@lists.libre-riscv.org
   NLnet milestone: ---

* add suitable separate unit tests for both mul and carry.
  add them as *separate commits* with a message "unit test for sim multiply"
  and "unit tests for sim carry"

* remove the whitespace added around lines 140.  ensure that there is
  only one newline (not two)

* remove the redundant brackets in these two statements:

+        if(cry_out==1):
+            if(result > 0xFFFFFFFF):

* ensure that one space is between "if" and its "test"

   if cry_out == 1:

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* [libre-riscv-dev] style / commit messages
From: Luke Kenneth Casson Leighton @ 2020-04-17 16:38 UTC (permalink / raw)
  To: Tobias Platen, Tobias Platen, Libre-RISCV General Development

hi tobias, regarding this commit:

commit f17100d6c435aec47b17acb188272a7e9b3cf67d
Author: Tobias Platen <tplaten@posteo.de>
Date:   Fri Apr 17 16:06:25 2020 +0000

    add with carry instructions

* the addition of the MUL feature needed to be under a separate commit.
 (multiplication doesn't have anything to do with "add with carry instructions")

 you'll have to watch out for this in future.

* whitespace was added at the end: this you particularly have to watch out for
  with vi.  it's why it's specifically in the HDL_workflow.

 please therefore can you correct that one.

* spaces and brackets: please try to keep to PEP8 as much as you can

  the brackets are redundant in this case (unnecessary) and it's definitely a
  style irritation not to have a space between the if and the test.

+        if(cry_out==1):
+            if(result > 0xFFFFFFFF):

 please can you correct those.

* if adding some functionality it's critically important to also add a
corresponding
  unit test, of at least some basic functionality.

  this because if you don't, someone else will have to.  the worst outcome is if
  you add the functionality and someone *assumes* that a unit test has been
  added for it.

i'll assign a bugreport to you to make the necessary corrections.

l.

_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* [libre-riscv-dev] [Bug 208] implement CORDIC in a general way sufficient to do transcendentals
From: bugzilla-daemon @ 2020-04-17 12:29 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-208-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=208

--- Comment #22 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
michael you may find this extremely informative to experiment with:

http://weitz.de/ieee/

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* [libre-riscv-dev] [Bug 208] implement CORDIC in a general way sufficient to do transcendentals
From: bugzilla-daemon @ 2020-04-16 17:16 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-208-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=208

--- Comment #21 from Michael Nolan <mtnolan2640@gmail.com> ---
(In reply to Jacob Lifshay from comment #18)

> 
> I think that's done in ieee754.fpcommon.denorm:
> https://git.libre-soc.org/?p=ieee754fpu.git;a=blob;f=src/ieee754/fpcommon/
> denorm.py;h=350f413ede9858ae490059f6709b3454ea8a7f1e;hb=HEAD

Ah, thank you!

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* [libre-riscv-dev] [Bug 208] implement CORDIC in a general way sufficient to do transcendentals
From: bugzilla-daemon @ 2020-04-16 16:59 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-208-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=208

--- Comment #20 from Jacob Lifshay <programmerjake@gmail.com> ---
(In reply to Luke Kenneth Casson Leighton from comment #19)
> btw i don't recommend trying multiplying or dividing either the input
> or the output by pi inside the FP CORDIC code. we can use micro-coding for
> that (and in the unit tests do the divide and/or multiply using import math
> or import bigfloat).
> 
> so let us assume that we're implementing cospi and sinpi, here, *NOT*
> actual cos or actual sin, ok?

That will allow us to have a much more accurate cospi and sinpi because we
don't need multiple conversions. Also, since cospi and sinpi have a period of 2
which is a power of 2, it is super easy to do range reduction -- just some bit
masking.

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* [libre-riscv-dev] [Bug 208] implement CORDIC in a general way sufficient to do transcendentals
From: bugzilla-daemon @ 2020-04-16 16:46 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-208-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=208

--- Comment #19 from Luke Kenneth Casson Leighton <lkcl@lkcl.net> ---
(In reply to Jacob Lifshay from comment #18)
> (In reply to Michael Nolan from comment #17)
> > I'm working on converting the floating point input to the fixed point value
> > needed by the cordic. I've got my module decoding the float into sign,
> > exponent, and mantissa with FPNumDecode, and the mantissa is missing the
> > (implied) leading 1 bit. I'm also looking at the modules of fpdiv (div0.py
> > and nmigen_div_experiment.py), and they don't seem to add that bit back in.
> > Is there a way to have FPNumDecode generate a mantissa with the leading 1
> > bit?
> 
> I think that's done in ieee754.fpcommon.denorm:
> https://git.libre-soc.org/?p=ieee754fpu.git;a=blob;f=src/ieee754/fpcommon/
> denorm.py;h=350f413ede9858ae490059f6709b3454ea8a7f1e;hb=HEAD

like 41, yes.

ok.

i only sort-of "understood" IEEE754 FP retrospectively, through long exposure,
having started initially from jon dawson's "working" verilog code and making
absolutely damn sure, throughout the entire six months of morphing it, that
it passed the unit tests *at all times*.  this made for a bit of a mess at
times.

i therefore substituted "unit tests pass 100%" for "understanding what the
hell is really going on".

line 41.

https://git.libre-soc.org/?p=ieee754fpu.git;a=blob;f=src/ieee754/fpcommon/denorm.py;h=350f413ede9858ae490059f6709b3454ea8a7f1e;hb=HEAD#l41

here you can see that the top bit of a (and later b) are, yes, set to 1
if and only if the exponent is not equal to -126 [for FP32]

back-track a little: the data structure FPSCData has had one *extra* bit
already allocated at the top (m[-1]) which is *specifically* allocated
for the purpose of allowing that extra implicit mantissa "1" to be
inserted *if it is needed*.

the reason why it is sometimes not needed is this:

when the exponent is "-126" (on FP32), this is "ridiculously small number"
but it is not the limit of the range of IEEE754.

what they allowed was: when exp=-126, you can express smaller and smaller
numbers by simply having less bits in the *mantissa*.  obviously, those
numbers become less and less accurate however this is an acceptable tradeoff.

however when you do that, you have to go from "implicit 1 in the top digit"
to "explicit 1 in the mantissa itself".

thus, we have that Mux detecting the transition condition between these
"implicit 1" and "really small numbers with explicit 1 *ALREADY* in
the mantissa".

the whole process is called "de-normalisation".

so basically, you can just "trust" that after the number has gone through
FPAddDeNorm, it *will* have the mantissa and exponent correctly set, such
that you can treat that "expanded mantissa" as just "A Fixed Integer".

btw i don't recommend trying multiplying or dividing either the input
or the output by pi inside the FP CORDIC code. we can use micro-coding for
that (and in the unit tests do the divide and/or multiply using import math
or import bigfloat).

so let us assume that we're implementing cospi and sinpi, here, *NOT*
actual cos or actual sin, ok?

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* [libre-riscv-dev] [Bug 208] implement CORDIC in a general way sufficient to do transcendentals
From: bugzilla-daemon @ 2020-04-16 16:22 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-208-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=208

--- Comment #18 from Jacob Lifshay <programmerjake@gmail.com> ---
(In reply to Michael Nolan from comment #17)
> I'm working on converting the floating point input to the fixed point value
> needed by the cordic. I've got my module decoding the float into sign,
> exponent, and mantissa with FPNumDecode, and the mantissa is missing the
> (implied) leading 1 bit. I'm also looking at the modules of fpdiv (div0.py
> and nmigen_div_experiment.py), and they don't seem to add that bit back in.
> Is there a way to have FPNumDecode generate a mantissa with the leading 1
> bit?

I think that's done in ieee754.fpcommon.denorm:
https://git.libre-soc.org/?p=ieee754fpu.git;a=blob;f=src/ieee754/fpcommon/denorm.py;h=350f413ede9858ae490059f6709b3454ea8a7f1e;hb=HEAD

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* Re: [libre-riscv-dev] morphing 6600 code to use power decoder
From: Luke Kenneth Casson Leighton @ 2020-04-16 15:41 UTC (permalink / raw)
  To: Libre-RISCV General Development
In-Reply-To: <CAPweEDy19rdSn-ihjv7Dye8aay9t4NU4YfTgCKmnaEwVZYK-cQ@mail.gmail.com>

adding multi-read and multi-write register port selection is a massive
ripple-effect change, going through virtually absolutely every module
involved in the 6600 layout.

* PriorityPicker can no longer be used, it has to be
MultiPriorityPicker because that's the class which can arbitrate
exclusive access to multiple register file ports
* Group Picker uses PriorityPicker, which is told not just which
read/write ports are being requested, it has to be told which ones are
safe to read/write
* Function-Unit to Function-Unit Matrices need to be multi-port aware
* Function-Unit to Register Matrices need to be multi-port aware
* Computation Units need to be multi-port aware

all this because we need to have a separate register file for POWER
branch/condition/link registers, which needs its own separate port
"management" signals from the INT/FP regfile.

gaah.

l.

_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* [libre-riscv-dev] [Bug 208] implement CORDIC in a general way sufficient to do transcendentals
From: bugzilla-daemon @ 2020-04-16 15:36 UTC (permalink / raw)
  To: libre-riscv-dev
In-Reply-To: <bug-208-13@https.bugs.libre-soc.org/>

https://bugs.libre-soc.org/show_bug.cgi?id=208

--- Comment #17 from Michael Nolan <mtnolan2640@gmail.com> ---
I'm working on converting the floating point input to the fixed point value
needed by the cordic. I've got my module decoding the float into sign,
exponent, and mantissa with FPNumDecode, and the mantissa is missing the
(implied) leading 1 bit. I'm also looking at the modules of fpdiv (div0.py and
nmigen_div_experiment.py), and they don't seem to add that bit back in. Is
there a way to have FPNumDecode generate a mantissa with the leading 1 bit?

-- 
You are receiving this mail because:
You are on the CC list for the bug.
_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

* Re: [libre-riscv-dev] morphing 6600 code to use power decoder
From: Luke Kenneth Casson Leighton @ 2020-04-16 12:19 UTC (permalink / raw)
  To: Libre-RISCV General Development
In-Reply-To: <CAPweEDxn1wqweqoz+-g3OoKkKDrxLUZ=zmcF-vLTEWow1ccohg@mail.gmail.com>

ok i've got the computation unit "operational" with multiple
go-read/acknowledge signals.  i say operational, it's tricky code that
i'm not completely happy with, however it "works".  finally i can move
on to getting the comp unit and dependency matrices multi-rd/write
signals integrated.

l.

_______________________________________________
libre-riscv-dev mailing list
libre-riscv-dev@lists.libre-riscv.org
http://lists.libre-riscv.org/mailman/listinfo/libre-riscv-dev

^ permalink raw reply

page:   latest

Libre RISC-V Mailing list Archives (accessible via public-inbox)

Archives are clonable: git clone --mirror https://inbox.libre-soc.org/libre-riscv-dev

Example config snippet for mirrors


AGPL code for this site: git clone https://public-inbox.org/public-inbox.git