ieee754fpu.git
2 years agoworking on implementing fma, f16 rtz formal proof seems likely to work
Jacob Lifshay [Mon, 4 Jul 2022 10:49:45 +0000 (03:49 -0700)]
working on implementing fma, f16 rtz formal proof seems likely to work

bitwuzla has been running the formal proof for the last 23min, seems
like it'll probably succeed after a bunch of time.

2 years agoadd FPFormat.get_exponent_value to get an unbiased exponent corrected for subnormals
Jacob Lifshay [Mon, 4 Jul 2022 10:42:27 +0000 (03:42 -0700)]
add FPFormat.get_exponent_value to get an unbiased exponent corrected for subnormals

2 years agomisc bug fixes and cleanup in fpbase.py
Jacob Lifshay [Mon, 4 Jul 2022 10:41:24 +0000 (03:41 -0700)]
misc bug fixes and cleanup in fpbase.py

2 years agoadd tests for non-power-of-2 shifts for MultiShift* classes
Jacob Lifshay [Mon, 4 Jul 2022 06:18:20 +0000 (23:18 -0700)]
add tests for non-power-of-2 shifts for MultiShift* classes

2 years agofix MultiShift* classes to correctly calculate shift-amount bit width
Jacob Lifshay [Mon, 4 Jul 2022 06:15:14 +0000 (23:15 -0700)]
fix MultiShift* classes to correctly calculate shift-amount bit width

2 years agostart implementing fp fused-mul-add pipeline
Jacob Lifshay [Sat, 2 Jul 2022 02:59:39 +0000 (19:59 -0700)]
start implementing fp fused-mul-add pipeline

https://bugs.libre-soc.org/show_bug.cgi?id=877

2 years agodisable fadd f32 formal proofs by default -- they're too slow
Jacob Lifshay [Sat, 2 Jul 2022 02:54:20 +0000 (19:54 -0700)]
disable fadd f32 formal proofs by default -- they're too slow

2 years agofix copyright years
Jacob Lifshay [Fri, 1 Jul 2022 22:01:37 +0000 (15:01 -0700)]
fix copyright years

[skip ci]

2 years agoupdate README, include links to source code
Luke Kenneth Casson Leighton [Fri, 1 Jul 2022 08:03:36 +0000 (09:03 +0100)]
update README, include links to source code
mention grant numbers, and that nmigen is Trademarked.
+nmigen is a Registered Trademark of M-Labs
+https://uspto.report/TM/88980893

2 years agochange gitlab-ci to show test names
Jacob Lifshay [Fri, 1 Jul 2022 08:00:10 +0000 (01:00 -0700)]
change gitlab-ci to show test names

2 years agofix my name
Jacob Lifshay [Fri, 1 Jul 2022 07:59:05 +0000 (00:59 -0700)]
fix my name

2 years agoupdate Copyright year
Luke Kenneth Casson Leighton [Fri, 1 Jul 2022 07:38:51 +0000 (08:38 +0100)]
update Copyright year

2 years agoadd fsub support to fadd pipeline
Jacob Lifshay [Fri, 1 Jul 2022 05:07:27 +0000 (22:07 -0700)]
add fsub support to fadd pipeline

2 years agofinish adding all rounding modes to fadd -- formal proof passes
Jacob Lifshay [Fri, 1 Jul 2022 03:34:24 +0000 (20:34 -0700)]
finish adding all rounding modes to fadd -- formal proof passes

2 years agoadd rounding modes to fpbase.Overflow
Jacob Lifshay [Wed, 29 Jun 2022 04:39:22 +0000 (21:39 -0700)]
add rounding modes to fpbase.Overflow

2 years agof32 fadd formal proof is fast enough -- don't skip it anymore
Jacob Lifshay [Tue, 28 Jun 2022 05:53:06 +0000 (22:53 -0700)]
f32 fadd formal proof is fast enough -- don't skip it anymore

2 years agoadd correct NaN propagation to the fadd pipeline and formal proof
Jacob Lifshay [Tue, 28 Jun 2022 05:27:39 +0000 (22:27 -0700)]
add correct NaN propagation to the fadd pipeline and formal proof

2 years agoadd bitwuzla to gitlab-ci
Jacob Lifshay [Tue, 28 Jun 2022 05:27:28 +0000 (22:27 -0700)]
add bitwuzla to gitlab-ci

2 years agoadd formal proofs for other fadd widths, but with unittest.skip
Jacob Lifshay [Tue, 28 Jun 2022 03:44:28 +0000 (20:44 -0700)]
add formal proofs for other fadd widths, but with unittest.skip

2 years agoadd initial f16 fadd formal proof
Jacob Lifshay [Sat, 25 Jun 2022 01:04:09 +0000 (18:04 -0700)]
add initial f16 fadd formal proof

2 years agochange .gitlab-ci.yml to use nmigen with smtlib2 support
Jacob Lifshay [Fri, 24 Jun 2022 05:23:41 +0000 (22:23 -0700)]
change .gitlab-ci.yml to use nmigen with smtlib2 support

2 years agoadd set-logic ALL clause to stop cvc5 warning
Luke Kenneth Casson Leighton [Tue, 17 May 2022 08:59:29 +0000 (09:59 +0100)]
add set-logic ALL clause to stop cvc5 warning

2 years agochange run command
Jacob Lifshay [Tue, 17 May 2022 03:42:26 +0000 (20:42 -0700)]
change run command

2 years agoadd fp16mul_test.smt2
Jacob Lifshay [Tue, 17 May 2022 03:38:16 +0000 (20:38 -0700)]
add fp16mul_test.smt2

2 years agoremove redundant bvadd with 1 input
Jacob Lifshay [Fri, 13 May 2022 22:36:01 +0000 (15:36 -0700)]
remove redundant bvadd with 1 input

2 years agoadd missing inf * 0 -> NaN cases
Jacob Lifshay [Fri, 13 May 2022 22:10:17 +0000 (15:10 -0700)]
add missing inf * 0 -> NaN cases

2 years agoadd fpmul_test.smt2 as a test to see if we should bother trying to wire-up smtlib2...
Jacob Lifshay [Fri, 13 May 2022 08:23:32 +0000 (01:23 -0700)]
add fpmul_test.smt2 as a test to see if we should bother trying to wire-up smtlib2 real and fp support in yosys and nmigen

It's probably correct, z3 ran longer than I bothered to wait, it usually stops relatively quickly if there's an error in the logic.

2 years agofix ci not finding sfpy
Jacob Lifshay [Thu, 12 May 2022 03:58:48 +0000 (20:58 -0700)]
fix ci not finding sfpy

2 years agofix tests/mark as expected failure
Jacob Lifshay [Thu, 12 May 2022 03:06:31 +0000 (20:06 -0700)]
fix tests/mark as expected failure

2 years agoformat code
Jacob Lifshay [Thu, 12 May 2022 03:05:15 +0000 (20:05 -0700)]
format code

2 years agopin some dependency versions
Jacob Lifshay [Thu, 12 May 2022 01:34:17 +0000 (18:34 -0700)]
pin some dependency versions

2 years agoremove redundant arguments that are now specified in pyproject.toml
Jacob Lifshay [Wed, 11 May 2022 08:26:41 +0000 (01:26 -0700)]
remove redundant arguments that are now specified in pyproject.toml

2 years agofix some borked imports
Jacob Lifshay [Wed, 11 May 2022 08:24:04 +0000 (01:24 -0700)]
fix some borked imports

2 years agoadd pytest config -- it ignores some borked files
Jacob Lifshay [Wed, 11 May 2022 08:21:54 +0000 (01:21 -0700)]
add pytest config -- it ignores some borked files

2 years agomake sfpy build
Jacob Lifshay [Tue, 10 May 2022 04:45:43 +0000 (21:45 -0700)]
make sfpy build

2 years agofix .gitlab-ci.yml
Jacob Lifshay [Tue, 10 May 2022 04:13:17 +0000 (21:13 -0700)]
fix .gitlab-ci.yml

2 years agoswitch to using nmutil's FHDLTestCase
Jacob Lifshay [Thu, 2 Dec 2021 01:56:09 +0000 (17:56 -0800)]
switch to using nmutil's FHDLTestCase

3 years agoallow alias SimdSignal<-PartitionedSignal to allow tracking back through
Luke Kenneth Casson Leighton [Mon, 8 Nov 2021 11:25:28 +0000 (11:25 +0000)]
allow alias SimdSignal<-PartitionedSignal to allow tracking back through
git revision history

3 years agocomments
Luke Kenneth Casson Leighton [Sat, 30 Oct 2021 13:20:13 +0000 (14:20 +0100)]
comments

3 years agoallow scope to be None in SimdShape, it becomes an empty
Luke Kenneth Casson Leighton [Sat, 30 Oct 2021 13:13:44 +0000 (14:13 +0100)]
allow scope to be None in SimdShape, it becomes an empty
storage shell for parameters until a scope is given,
at which point a new SindShape is created and the PartitionPoints
can all be set up.

3 years agoadd SimdScope.__call__ function needed to copy and adapt scope
Luke Kenneth Casson Leighton [Sat, 30 Oct 2021 10:35:07 +0000 (11:35 +0100)]
add SimdScope.__call__ function needed to copy and adapt scope

3 years agoadd setmodule function to SimdScope
Luke Kenneth Casson Leighton [Sat, 30 Oct 2021 10:26:31 +0000 (11:26 +0100)]
add setmodule function to SimdScope

3 years agoput in (unused) code for SimdShape.__mul__ when both LHS and RHS
Luke Kenneth Casson Leighton [Thu, 28 Oct 2021 11:30:55 +0000 (12:30 +0100)]
put in (unused) code for SimdShape.__mul__ when both LHS and RHS
are SimdShapes.  this is unused at present, needs review

3 years agostart filling in integer SimdShape.__mul__ case with list of uses, based
Luke Kenneth Casson Leighton [Thu, 28 Oct 2021 11:15:40 +0000 (12:15 +0100)]
start filling in integer SimdShape.__mul__ case with list of uses, based
on whether the SimdShape was initially created as FIXED priority or
ELWID priority.

3 years agoadd SimdShape "priority" mode flag
Luke Kenneth Casson Leighton [Thu, 28 Oct 2021 10:41:49 +0000 (11:41 +0100)]
add SimdShape "priority" mode flag
(not used, yet)

3 years agoadd initial SimdShape.__add__
Jacob Lifshay [Thu, 28 Oct 2021 04:06:15 +0000 (21:06 -0700)]
add initial SimdShape.__add__

3 years agoadd initial SimdShape.__mul__
Jacob Lifshay [Thu, 28 Oct 2021 03:58:46 +0000 (20:58 -0700)]
add initial SimdShape.__mul__

3 years agoformat code
Jacob Lifshay [Thu, 28 Oct 2021 03:57:29 +0000 (20:57 -0700)]
format code

3 years agoremove SimdScope.get() and friends
Jacob Lifshay [Thu, 28 Oct 2021 03:41:34 +0000 (20:41 -0700)]
remove SimdScope.get() and friends

3 years agoformat code
Jacob Lifshay [Thu, 28 Oct 2021 03:36:51 +0000 (20:36 -0700)]
format code

3 years agoadd tests for SimdMap and friends
Jacob Lifshay [Wed, 27 Oct 2021 09:11:14 +0000 (02:11 -0700)]
add tests for SimdMap and friends

3 years agoadd SimdWHintMap to support tracking width_hint for XLEN
Jacob Lifshay [Wed, 27 Oct 2021 09:10:24 +0000 (02:10 -0700)]
add SimdWHintMap to support tracking width_hint for XLEN

3 years agohooray, convert PartitionedCat over to new PartType API, using
Luke Kenneth Casson Leighton [Mon, 25 Oct 2021 17:52:48 +0000 (18:52 +0100)]
hooray, convert PartitionedCat over to new PartType API, using
get_num_elements and get_el_range

3 years agoadd two new functions to PartType: get_num_elements and get_el_range
Luke Kenneth Casson Leighton [Mon, 25 Oct 2021 17:30:53 +0000 (18:30 +0100)]
add two new functions to PartType: get_num_elements and get_el_range

3 years agoremove fixed_width parameter from SimdScope.Signal
Luke Kenneth Casson Leighton [Mon, 25 Oct 2021 13:36:55 +0000 (14:36 +0100)]
remove fixed_width parameter from SimdScope.Signal
after working out how to explicitly use a SimdShape

3 years agouse explicit SimdShape for minitest example rather than
Luke Kenneth Casson Leighton [Mon, 25 Oct 2021 13:35:57 +0000 (14:35 +0100)]
use explicit SimdShape for minitest example rather than
fixed_width parameter

3 years agohad to add fixed_width parameter temporarily to confirm that the
Luke Kenneth Casson Leighton [Mon, 25 Oct 2021 13:28:55 +0000 (14:28 +0100)]
had to add fixed_width parameter temporarily to confirm that the
minitest, test_partsig_scope.py, worked (which it did).
now can work out how to remove it

3 years agodebugging of initial test_partsig_scope.py mini-test
Luke Kenneth Casson Leighton [Mon, 25 Oct 2021 13:19:02 +0000 (14:19 +0100)]
debugging of initial test_partsig_scope.py mini-test
scope and shape need to be stored in SimdSignal in order to get at them
later

3 years agoadapt/debug SimdSignal when using ElwidPartType
Luke Kenneth Casson Leighton [Mon, 25 Oct 2021 12:29:51 +0000 (13:29 +0100)]
adapt/debug SimdSignal when using ElwidPartType
example mini-test with test_partsig_scope.py working through issues

3 years agoremove unnecessary imports
Luke Kenneth Casson Leighton [Mon, 25 Oct 2021 12:29:06 +0000 (13:29 +0100)]
remove unnecessary imports

3 years agoremove unnecessary code which creates complications from SimdScope
Luke Kenneth Casson Leighton [Mon, 25 Oct 2021 12:14:06 +0000 (13:14 +0100)]
remove unnecessary code which creates complications from SimdScope
constructor.

elwid is mandatory, vec_el_counts is mandatory.  no need for complicated
types, complicated adaptation, complicated interlocking conditional
behaviour.

pass in elwid (Signal)
pass in vec_el_counts.

done.

3 years agorename the arguments to SimdShape() so as to match up with Shape() params
Luke Kenneth Casson Leighton [Mon, 25 Oct 2021 11:19:04 +0000 (12:19 +0100)]
rename the arguments to SimdShape() so as to match up with Shape() params

3 years agoadd SimdScope.Shape redirector which switches from scalar to simd behaviour
Luke Kenneth Casson Leighton [Mon, 25 Oct 2021 11:10:46 +0000 (12:10 +0100)]
add SimdScope.Shape redirector which switches from scalar to simd behaviour
depending on context.

for compatibility with nmigen Shape() the width parameter is passed
through to widths_at_elwid in SimdShape, allowing the layout() function
the opportunity to turn a scalar width into fixed element widths at
all elwids.

3 years agostart filling in ElwidPartType switch/case and other info
Luke Kenneth Casson Leighton [Mon, 25 Oct 2021 10:33:51 +0000 (11:33 +0100)]
start filling in ElwidPartType switch/case and other info
needed to get it functional and a drop-in replacement for the old
PartitionedSignal PartType behaviour

3 years agocreate a SimdShape class and begin to investigate how to tie it in
Luke Kenneth Casson Leighton [Mon, 25 Oct 2021 10:25:03 +0000 (11:25 +0100)]
create a SimdShape class and begin to investigate how to tie it in
to ElwidPartType.

3 years agoadd more code-comments
Luke Kenneth Casson Leighton [Mon, 25 Oct 2021 09:47:08 +0000 (10:47 +0100)]
add more code-comments

3 years agoreturn layout points from layout()
Luke Kenneth Casson Leighton [Sun, 24 Oct 2021 18:32:01 +0000 (19:32 +0100)]
return layout points from layout()

3 years agospelling corrections after typing on a phone
Luke Kenneth Casson Leighton [Sun, 24 Oct 2021 18:26:47 +0000 (19:26 +0100)]
spelling corrections after typing on a phone

3 years agoadd dictionary of list of layout points, start-end
Luke Kenneth Casson Leighton [Sun, 24 Oct 2021 16:14:43 +0000 (17:14 +0100)]
add dictionary of list of layout points, start-end
ElwidPartType it looks like it needs to know where each element
starts and ends, therefore provide that info at the point it was
created

3 years agoadd docstrings for additional necessary context-aware versions of
Luke Kenneth Casson Leighton [Sun, 24 Oct 2021 10:13:56 +0000 (11:13 +0100)]
add docstrings for additional necessary context-aware versions of
Signal.like and Shape in SimdScope.

3 years agofrom mobile reslly bad spelling comment why layout should remain a fn
Luke Kenneth Casson Leighton [Fri, 22 Oct 2021 16:41:20 +0000 (17:41 +0100)]
from mobile reslly bad spelling comment why layout should remain a fn

3 years agoadd back-link to why SimdScope has to get the hell out the way when
Luke Kenneth Casson Leighton [Fri, 22 Oct 2021 14:29:18 +0000 (15:29 +0100)]
add back-link to why SimdScope has to get the hell out the way when
scalar=True

3 years agoadd in code-comments
Luke Kenneth Casson Leighton [Fri, 22 Oct 2021 14:21:51 +0000 (15:21 +0100)]
add in code-comments
noticed a lot of weird code (including the hint that there might be
an intention to treat SimdSignal as a scalar ("one elwid, full width")
which is very dangerous (explained already)

added in particular some notes about the parameters that go through
to layout()

3 years agoadd comments / docstrings for layout function to illustrate where
Luke Kenneth Casson Leighton [Fri, 22 Oct 2021 14:14:13 +0000 (15:14 +0100)]
add comments / docstrings for layout function to illustrate where
the parameters actually come from

3 years agoonly set standard PartType in non-Simd-mode
Luke Kenneth Casson Leighton [Fri, 22 Oct 2021 14:03:22 +0000 (15:03 +0100)]
only set standard PartType in non-Simd-mode

3 years agoin SimdScope only update module AST Typecast function in SIMD mode
Luke Kenneth Casson Leighton [Fri, 22 Oct 2021 14:01:16 +0000 (15:01 +0100)]
in SimdScope only update module AST Typecast function in SIMD mode
do not whatever you do change the mode over to Simd casting in Scalar mode

3 years agoadd the beginnings of a usage docstring for the module and add a Signal
Luke Kenneth Casson Leighton [Fri, 22 Oct 2021 13:41:41 +0000 (14:41 +0100)]
add the beginnings of a usage docstring for the module and add a Signal
function which shows how to connect things together.

with SimdScope being passed in as the "mask" parameter, the SimdSignal
knows to switch over to "ElwidPartType mode"

3 years agoremove simd_full_width_hint, it is down to individual Signals within the
Luke Kenneth Casson Leighton [Fri, 22 Oct 2021 13:39:19 +0000 (14:39 +0100)]
remove simd_full_width_hint, it is down to individual Signals within the
context to explicitly declare their width, if in fact they need one.
some cases (as shown in the layout() function) no fixed (full) width
is required to explicitly be specified, it is determined instead from
the element widths

3 years agoadd in TODO notes tying in SimdScope/SimdMode
Luke Kenneth Casson Leighton [Fri, 22 Oct 2021 13:15:18 +0000 (14:15 +0100)]
add in TODO notes tying in SimdScope/SimdMode

3 years agoadd type annotations .pyi file for SimdScope
Jacob Lifshay [Fri, 22 Oct 2021 07:40:09 +0000 (00:40 -0700)]
add type annotations .pyi file for SimdScope

3 years agoupdate SimdScope to use vec_el_counts
Jacob Lifshay [Fri, 22 Oct 2021 07:39:42 +0000 (00:39 -0700)]
update SimdScope to use vec_el_counts

3 years agomove SimdScope to separate file
Jacob Lifshay [Fri, 22 Oct 2021 05:58:33 +0000 (22:58 -0700)]
move SimdScope to separate file

3 years agoadd LHS support into PartitionedCat. amazingly - stunningly - it works
Luke Kenneth Casson Leighton [Thu, 21 Oct 2021 15:56:51 +0000 (16:56 +0100)]
add LHS support into PartitionedCat. amazingly - stunningly - it works
https://bugs.libre-soc.org/show_bug.cgi?id=731

3 years agoconfirmed (in prototype form that LHS Cat will cause conflict
Luke Kenneth Casson Leighton [Thu, 21 Oct 2021 15:50:01 +0000 (16:50 +0100)]
confirmed (in prototype form that LHS Cat will cause conflict

3 years agocontinue truly awful hack which, in SimdSignal.__Assign__, detects the
Luke Kenneth Casson Leighton [Thu, 21 Oct 2021 14:04:33 +0000 (15:04 +0100)]
continue truly awful hack which, in SimdSignal.__Assign__, detects the
back-link to the submodule (PartitionedCat) in its return result,
and calls set_lhs_mode(True) or (False) on LHS and RHS as appropriate.

the default value is *NOT* set in the PartitionedCat constructor very very
deliberately so as to show up any bugs.  it is particularly fortunate that
this was chosen to be done because there was, in fact, a bug in the
TestCatMod unit test, which assumed that it was ok to splat a Cat() result
of a pair of SimdSignals directly onto a Signal().

it *is* in fact "technically allowed" by nmigen due to automatic casting
of UserValue, but should not strictly have been done.

3 years agoadd back-link in the return result of PartitionedCat to allow access
Luke Kenneth Casson Leighton [Thu, 21 Oct 2021 13:31:36 +0000 (14:31 +0100)]
add back-link in the return result of PartitionedCat to allow access
to the submodule.
PartitionedAssign can then detect this and alter the conditions to
LHS *before* PartitionedCat.elaborate() is called

it was already established in commit 494757caa1f that the elaborates
are all called later

3 years agoadd quick print statements to show that elaborate() gets called as a
Luke Kenneth Casson Leighton [Thu, 21 Oct 2021 12:40:29 +0000 (13:40 +0100)]
add quick print statements to show that elaborate() gets called as a
second phase after the creation of the AST tree
this gives a window of opportunity to tree-walk and set whether SimdSignals
are LHS or RHS as determined by encountering SimdSignal.__Assign__

3 years agofound an error in PartitionedAssign and PartitionedRepl
Luke Kenneth Casson Leighton [Thu, 21 Oct 2021 12:06:35 +0000 (13:06 +0100)]
found an error in PartitionedAssign and PartitionedRepl
where Slice was accidentally being done on SimdSignal rather than
SimdSignals internal sig.  whilst this was a legitimate oversight
the bug should have been found when a NotImplemented SimdSignal.__Slice__
was added.

Project Development Practices were violated here by unit tests not
having been run, which would have easily detected the bug

3 years agoremove duplicate function definition
Luke Kenneth Casson Leighton [Thu, 21 Oct 2021 11:50:58 +0000 (12:50 +0100)]
remove duplicate function definition

3 years agoremove reference to use of Swizzled class, due to it violating
Luke Kenneth Casson Leighton [Thu, 21 Oct 2021 11:49:53 +0000 (12:49 +0100)]
remove reference to use of Swizzled class, due to it violating
Project Development Practices
documented here:
https://bugs.libre-soc.org/show_bug.cgi?id=731#c10

3 years agoadd forgotten files from last commit
Jacob Lifshay [Sat, 16 Oct 2021 01:19:11 +0000 (18:19 -0700)]
add forgotten files from last commit

3 years agoadd WIP code for handling Slice and Cat in a unified way, supporting assignment
Jacob Lifshay [Sat, 16 Oct 2021 01:12:49 +0000 (18:12 -0700)]
add WIP code for handling Slice and Cat in a unified way, supporting assignment

3 years agoformat code
Jacob Lifshay [Fri, 15 Oct 2021 23:58:24 +0000 (16:58 -0700)]
format code

3 years agofix bmask calculation
Jacob Lifshay [Fri, 15 Oct 2021 05:05:23 +0000 (22:05 -0700)]
fix bmask calculation

3 years agosplit out end_bit
Jacob Lifshay [Fri, 15 Oct 2021 03:53:09 +0000 (20:53 -0700)]
split out end_bit

3 years agosplit out start_bit
Jacob Lifshay [Fri, 15 Oct 2021 03:51:09 +0000 (20:51 -0700)]
split out start_bit

3 years agoremove redundant plist variable
Jacob Lifshay [Fri, 15 Oct 2021 03:45:49 +0000 (20:45 -0700)]
remove redundant plist variable

3 years agosort dpoints keys
Jacob Lifshay [Fri, 15 Oct 2021 03:27:42 +0000 (20:27 -0700)]
sort dpoints keys

3 years agosimplify dpoints computation
Jacob Lifshay [Fri, 15 Oct 2021 03:21:12 +0000 (20:21 -0700)]
simplify dpoints computation

width is already set to fixed_width, we don't need to have a separate case

3 years agodedup dpoints
Jacob Lifshay [Fri, 15 Oct 2021 03:20:43 +0000 (20:20 -0700)]
dedup dpoints