Comments (53)
chick
commented on September 13, 2024
1
@azidar all the wrap cases I did made sense except for
val base = Wire(Interval(range"[-4, 6]"))
val disjointRight = Wire(Interval(range"[7,11]"))
val w6 = base.wrap(disjointRight)
assert(w6 === 10.I()) // FAILS says it should be 6.U
// even though derived interval is disjointRight Closed(7) Closed(11)Any ideas
from dsptools.
shunshou
commented on September 13, 2024
Branches I'm using (based off the above):
Branches
| project | branch |
|---|---|
| firrtl | intervals-oct |
| chisel3 | interval-oct |
| firrtl-interpreter | ??? |
from dsptools.
shunshou
commented on September 13, 2024
@chick can you make sure ops like the ones in https://github.com/ucb-bar/dsptools/tree/master/src/main/scala/dsptools/numbers are also supported? i.e. <<, >> not implemented:
scala.NotImplementedError: an implementation is missing
[info] at scala.Predef$.$qmark$qmark$qmark(Predef.scala:225)
[info] at chisel3.internal.firrtl.IntervalRange.$less$less(IR.scala:359)
from dsptools.
chick
commented on September 13, 2024
Working on it. Is there general agreement on what it means. I would guess for shift that:
- Shifts the point of Closed and Open and BinaryPoint by the shift amount
- UnknownBounds and BinaryPoint remain unchanged.
And what about merge. Not mentioned in Wikipedia Intervals
My instinct would be:
- increase the bounds of the interval with the smaller binary point so it matches the binary point of the larger
- min of lows
- max of uppers
- All this when defined.
Or maybe more naively
- Low is min of bounds if defined, unknown otherwise
- Upper is max of bounds if defined, unknown otherwise
- max of binary point
@shunshou , @azidar @grebe Your thoughts?
from dsptools.
shunshou
commented on September 13, 2024
So two things:
-
We need to look at how Adam's firrtl code does this -- probably would be useful to have Adam document user-facing features; or that might already be in his tests.
-
For dsptools:
a >> n
Arithmetic shift right where n is either Int or UInt
Note: precision loss will result since the decimal point location is not adjusted
AND
a.div2(n)
a/2^n
UInt: Consistent with a >> n (i.e. rounds 1/2 to 0)
SInt: Round output to nearest SInt via DspContext.trimType
FixedPoint: Adjusts decimal point; up to you how much precision to keep via DspContext.trimType and DspContext.binaryPointGrowth (adds more fractional bits on top of the input amount)
For both FixedPoint and SInt, again, be aware of overflow behavior when RoundHalfUp is used for the most positive input
^ Both should be eventually supported (but not necessarily in Chisel; can be in dsptools).
What is merge?
It seems like, for example, with multiply, you were intending to have firrtl infer everything if uninferred i.e. maybe the naive approach?
Are we going at this expecting that Chisel basically should have no knowledge of intervals besides the ones that are user assigned and letting firrtl take care of pretty much everything?
from dsptools.
shunshou
commented on September 13, 2024
Also, fyi, I pushed more changes to aforementioned branches to fix some bugs.
from dsptools.
chick
commented on September 13, 2024
I added the unimplemented shift functions IntervalRange and Interval. I need to add more tests but I wanted to get the implementations out there first. I left merge() in as unimplemented. Not sure what it is intended for.
from dsptools.
chick
commented on September 13, 2024
And yes, I believe @azidar thought the naive approach of anything not simple with intervals should be deferred to Firrtl as best for now.
from dsptools.
shunshou
commented on September 13, 2024
Ok then beyond that, the ops that should take precedence are clip, wrap, etc. (overflow behavior and truncation/round, etc.). After that, I think we're good for being able to try some more complicated legitimate hardware...
I will also add some tests... (but in my FFT repo for now)
from dsptools.
shunshou
commented on September 13, 2024
Here are the ops Adam checked in IntervalMathSpec:
+
*
-
>, >=, <, <=, ===, =/=
>> int
<< int
<< uint
>> uint
[]()
arith assign?
cat
Is cat something built into the type in Chisel or an external helper? o_O Otherwise, assuming the dynamic/static shifts work, I think we've covered all of our bases there.
from dsptools.
shunshou
commented on September 13, 2024
And as far as the rest of Adam's non-math tests, really it's:
Cannot defined unknown BP for interval (Chisel restriction) -- just my note
wrap
clip
Is there a setBinaryPoint equivalent for Interval? That might be more necessary for stuff like rounding ops.
from dsptools.
grebe
commented on September 13, 2024
I can't find a reference to merge in our doc. @chick, your instinct matches mine, but I'm still trying to think of why it would be useful.
from dsptools.
chick
commented on September 13, 2024
@shunshou @azidar are there any explicit docs on clip, wrap, etc. I am working from what I can see in intervals-oct, but any other materials would be useful
from dsptools.
chick
commented on September 13, 2024
@shunshou
I've pushed naive wrap and clip to chisel3 intervals-oct, they don't really figure anything out right now. just make things Unknown so firrtl can figure it all out. I need to talk to Adam to proceed on this.
Also I did not see a problem with running the interpreter's instrumenting-sizes branch with the intervals-oct branches of firrtl and chisel3, if you have an example can you pass it to me
from dsptools.
shunshou
commented on September 13, 2024
Yeah, so Adam's super busy this week--he asked to reschedule our meeting (will send out an email shortly), but if he has a few minutes to meet with you tomorrow to discuss next steps, that'd be great. The one thing I'd like to see is being able to do clip based off of some other signal's range * # of loops, as an example, to handle: multiply by a constant and accumulate. I'll test firrtl-interpreter again with various branches and copy+paste the error I get in a few minutes.
from dsptools.
shunshou
commented on September 13, 2024
Also, are you on the same branch of chisel-testers/dsptools as I am?
Note: the last working verison of firrtl-interpreter is 5dd2f5b
When I switch over to firrtl-interpreter/instrumenting-sizes, and I try to re-publish-local chisel-testers with hash cb82340, I get:
[error] /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/src/main/scala/chisel3/iotesters/Driver.scala:130: type mismatch;
[error] found : optionsManager.type (with underlying type chisel3.iotesters.ReplOptionsManager)
[error] required: firrtl_interpreter.InterpreterOptionsManager with firrtl_interpreter.HasReplConfig
[error] FirrtlRepl.execute(optionsManager)
[error] ^
[error] /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/src/main/scala/chisel3/iotesters/Driver.scala:130: type mismatch;
[error] found : optionsManager.type (with underlying type chisel3.iotesters.ReplOptionsManager)
[error] required: firrtl_interpreter.InterpreterOptionsManager with firrtl_interpreter.HasReplConfig
[error] FirrtlRepl.execute(optionsManager)
[error] ^
[error] /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/src/main/scala/chisel3/iotesters/FirrtlTerpBackend.scala:14: type mismatch;
[error] found : chisel3.iotesters.TesterOptionsManager
[error] required: firrtl_interpreter.HasInterpreterSuite
[error] val interpretiveTester = new InterpretiveTester(firrtlIR, optionsManager)
[error] ^
[error] /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/src/main/scala/chisel3/iotesters/FirrtlTerpBackend.scala:14: type mismatch;
[error] found : chisel3.iotesters.TesterOptionsManager
[error] required: firrtl_interpreter.HasInterpreterSuite
[error] val interpretiveTester = new InterpretiveTester(firrtlIR, optionsManager)
If you are on a better version of chisel-testers (and corresponding dsptools), please let me know!
from dsptools.
shunshou
commented on September 13, 2024
Before you delve too deep into this, as a basic use case, I just want to see that
val base = Wire(Interval(range"[0, 9]"))
val out = base.wrap(range"[0,7]")
base := 9.I
--> out == 1.I without any mux logic. That's the most common use-case for wrap. Does that work?
from dsptools.
chick
commented on September 13, 2024
@shunshou I'll need to change chisel-testers to fix its publishLocal problems with the instrumenting-sizes branch of firrtl-interpreter. Can there be a intervals-oct branch of chisel-testers that I can make my changes in. I'm not sure what to do with cb82340
from dsptools.
shunshou
commented on September 13, 2024
yea go for it--make a branch and i'll point my stuff to it. thanks!! :D
from dsptools.
chick
commented on September 13, 2024
@shunshou pushed my latest wrap stuff to intervals-oct. This includes wrap(that: IntervalRange) which lets you say myInterval.wrap(range"[-4, 11)") if you are inclined to say such things
Will try to do clip next
from dsptools.
chick
commented on September 13, 2024
@shunshou just pushed intervals-oct branch with interpreter fixes for instrumenting-sizes branch.
It's based on the cb82340 (I hope that's ok) this builds for me, hope it works for you
from dsptools.
shunshou
commented on September 13, 2024
Hmm... I get
Can you report latest hashes of everything so I can double check?
] Executing in batch mode.
[warn] For better performance, hit [ENTER] to switch to interactive mode, or
[warn] consider launching sbt without any commands, or explicitly passing 'shell'
[info] Loading project definition from /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/project
[info] Set current project to Chisel.iotesters (in build file:/Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/)
[info] Packaging /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/target/scala-2.11/chisel-iotesters_2.11-1.2-SNAPSHOT-sources.jar ...
[info] Done packaging.
[info] Main Scala API documentation to /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/target/scala-2.11/api...
[info] Wrote /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/target/scala-2.11/chisel-iotesters_2.11-1.2-SNAPSHOT.pom
[info] :: delivering :: edu.berkeley.cs#chisel-iotesters_2.11;1.2-SNAPSHOT :: 1.2-SNAPSHOT :: integration :: Mon Oct 16 21:58:03 PDT 2017
[info] delivering ivy file to /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/target/scala-2.11/ivy-1.2-SNAPSHOT.xml
[info] Compiling 20 Scala sources to /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/target/scala-2.11/classes...
[error] /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/src/main/scala/chisel3/iotesters/ChiselPokeSpec.scala:15: type mismatch;
[error] found : options.type (with underlying type chisel3.iotesters.TesterOptionsManager)
[error] required: chisel3.iotesters.TesterOptionsManager with firrtl_interpreter.HasInterpreterSuite
[error] setupFirrtlTerpBackend(dutGen, options)
[error] ^
[error] /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/src/main/scala/chisel3/iotesters/ChiselPokeSpec.scala:15: type mismatch;
[error] found : options.type (with underlying type chisel3.iotesters.TesterOptionsManager)
[error] required: chisel3.iotesters.TesterOptionsManager with firrtl_interpreter.HasInterpreterSuite
[error] setupFirrtlTerpBackend(dutGen, options)
[error] ^
[error] /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/src/main/scala/chisel3/iotesters/Driver.scala:51: type mismatch;
[error] found : optionsManager.type (with underlying type chisel3.iotesters.TesterOptionsManager)
[error] required: chisel3.iotesters.TesterOptionsManager with firrtl_interpreter.HasInterpreterSuite
[error] setupFirrtlTerpBackend(dutGenerator, optionsManager)
[error] ^
[error] /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/src/main/scala/chisel3/iotesters/Driver.scala:60: type mismatch;
[error] found : Any
[error] required: chisel3.iotesters.Backend
[error] backendVar.withValue(Some(backend)) {
[error] ^
[error] /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/src/main/scala/chisel3/iotesters/Driver.scala:62: type mismatch;
[error] found : dut.type (with underlying type Any)
[error] required: T
[error] testerGen(dut).finish
[error] ^
[error] /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/src/main/scala/chisel3/iotesters/Driver.scala:130: type mismatch;
[error] found : optionsManager.type (with underlying type chisel3.iotesters.ReplOptionsManager)
[error] required: firrtl_interpreter.InterpreterOptionsManager with firrtl_interpreter.HasReplConfig
[error] FirrtlRepl.execute(optionsManager)
[error] ^
[error] /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/src/main/scala/chisel3/iotesters/Driver.scala:51: type mismatch;
[error] found : optionsManager.type (with underlying type chisel3.iotesters.TesterOptionsManager)
[error] required: chisel3.iotesters.TesterOptionsManager with firrtl_interpreter.HasInterpreterSuite
[error] setupFirrtlTerpBackend(dutGenerator, optionsManager)
[error] ^
[error] /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/src/main/scala/chisel3/iotesters/Driver.scala:60: type mismatch;
[error] found : Any
[error] required: chisel3.iotesters.Backend
[error] backendVar.withValue(Some(backend)) {
[error] ^
[error] /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/src/main/scala/chisel3/iotesters/Driver.scala:62: type mismatch;
[error] found : dut.type (with underlying type Any)
[error] required: T
[error] testerGen(dut).finish
[error] ^
[error] /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/src/main/scala/chisel3/iotesters/Driver.scala:130: type mismatch;
[error] found : optionsManager.type (with underlying type chisel3.iotesters.ReplOptionsManager)
[error] required: firrtl_interpreter.InterpreterOptionsManager with firrtl_interpreter.HasReplConfig
[error] FirrtlRepl.execute(optionsManager)
[error] ^
[error] /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/src/main/scala/chisel3/iotesters/FirrtlTerpBackend.scala:12: type mismatch;
[error] found : chisel3.iotesters.TesterOptionsManager
[error] required: chisel3.iotesters.TesterOptionsManager with firrtl_interpreter.HasInterpreterSuite
[error] optionsManager: TesterOptionsManager with HasInterpreterSuite = new TesterOptionsManager)
[error] ^
[error] /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/src/main/scala/chisel3/iotesters/FirrtlTerpBackend.scala:106: type mismatch;
[error] found : chisel3.iotesters.TesterOptionsManager
[error] required: chisel3.iotesters.TesterOptionsManager with firrtl_interpreter.HasInterpreterSuite
[error] optionsManager: TesterOptionsManager with HasInterpreterSuite = new TesterOptionsManager): (T, Backend) = {
[error] ^
[error] /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/src/main/scala/chisel3/iotesters/FirrtlTerpBackend.scala:12: type mismatch;
[error] found : chisel3.iotesters.TesterOptionsManager
[error] required: chisel3.iotesters.TesterOptionsManager with firrtl_interpreter.HasInterpreterSuite
[error] optionsManager: TesterOptionsManager with HasInterpreterSuite = new TesterOptionsManager)
[error] ^
[error] /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/src/main/scala/chisel3/iotesters/FirrtlTerpBackend.scala:106: type mismatch;
[error] found : chisel3.iotesters.TesterOptionsManager
[error] required: chisel3.iotesters.TesterOptionsManager with firrtl_interpreter.HasInterpreterSuite
from dsptools.
chick
commented on September 13, 2024
@shunshou just pushed clip (saturate) operator for intervals
BTW Aaaargh, on weird errors with testers, I'll try to reproduce
from dsptools.
shunshou
commented on September 13, 2024
Thx! Finally have a ~10 pg outline of the ~6 pg paper. :P Will look into this after I'm done drafing background + intro hopefully...
from dsptools.
chick
commented on September 13, 2024
Hmnm, I am building with
(cd firrtl ; git status | grep 'On branch')
On branch intervals-oct
(cd firrtl-interpreter ; git status | grep 'On branch')
On branch instrumenting-sizes
(cd chisel3 ; git status | grep 'On branch')
On branch intervals-oct
(cd chisel-testers ; git status | grep 'On branch')
On branch intervals-oct
(cd dsptools ; git status | grep 'On branch')
On branch master
from dsptools.
shunshou
commented on September 13, 2024
@chick, I'm still getting:
error] /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/src/main/scala/chisel3/iotesters/ChiselPokeSpec.scala:15: type mismatch;
[error] found : options.type (with underlying type chisel3.iotesters.TesterOptionsManager)
[error] required: chisel3.iotesters.TesterOptionsManager with firrtl_interpreter.HasInterpreterSuite
[error] setupFirrtlTerpBackend(dutGen, options)
[error] ^
[error] /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/src/main/scala/chisel3/iotesters/ChiselPokeSpec.scala:15: type mismatch;
[error] found : options.type (with underlying type chisel3.iotesters.TesterOptionsManager)
[error] required: chisel3.iotesters.TesterOptionsManager with firrtl_interpreter.HasInterpreterSuite
[error] setupFirrtlTerpBackend(dutGen, options)
[error] ^
[error] /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/src/main/scala/chisel3/iotesters/Driver.scala:51: type mismatch;
[error] found : optionsManager.type (with underlying type chisel3.iotesters.TesterOptionsManager)
[error] required: chisel3.iotesters.TesterOptionsManager with firrtl_interpreter.HasInterpreterSuite
[error] setupFirrtlTerpBackend(dutGenerator, optionsManager)
[error] ^
[error] /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/src/main/scala/chisel3/iotesters/Driver.scala:60: type mismatch;
[error] found : Any
[error] required: chisel3.iotesters.Backend
[error] backendVar.withValue(Some(backend)) {
[error] ^
[error] /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/src/main/scala/chisel3/iotesters/Driver.scala:62: type mismatch;
[error] found : dut.type (with underlying type Any)
[error] required: T
[error] testerGen(dut).finish
[error] ^
[error] /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/src/main/scala/chisel3/iotesters/Driver.scala:130: type mismatch;
[error] found : optionsManager.type (with underlying type chisel3.iotesters.ReplOptionsManager)
[error] required: firrtl_interpreter.InterpreterOptionsManager with firrtl_interpreter.HasReplConfig
[error] FirrtlRepl.execute(optionsManager)
[error] ^
[error] /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/src/main/scala/chisel3/iotesters/Driver.scala:51: type mismatch;
[error] found : optionsManager.type (with underlying type chisel3.iotesters.TesterOptionsManager)
[error] required: chisel3.iotesters.TesterOptionsManager with firrtl_interpreter.HasInterpreterSuite
[error] setupFirrtlTerpBackend(dutGenerator, optionsManager)
[error] ^
[error] /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/src/main/scala/chisel3/iotesters/Driver.scala:60: type mismatch;
[error] found : Any
[error] required: chisel3.iotesters.Backend
[error] backendVar.withValue(Some(backend)) {
[error] ^
[error] /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/src/main/scala/chisel3/iotesters/Driver.scala:62: type mismatch;
[error] found : dut.type (with underlying type Any)
[error] required: T
[error] testerGen(dut).finish
[error] ^
[error] /Users/angie.wang/Chisel3FFAST/Chisel3DSPDependencies/chisel-testers/src/main/scala/chisel3/iotesters/Driver.scala:130: type mismatch;
[error] found : optionsManager.type (with underlying type chisel3.iotesters.ReplOptionsManager)
[error] required: firrtl_interpreter.InterpreterOptionsManager with firrtl_interpreter.HasReplConfig
[error] FirrtlRepl.execute(optionsManager)
with the branches you've pointed me to :(
from dsptools.
chick
commented on September 13, 2024
Still trying to get this
Maybe now (had to push to interpreter and testers)
firrtl
3abda10e5e82b481f58c73adf26c48678e496574
firrtl-interpreter
4f0c623ad03b1b7f5c484aa99575ea324d8cdfbd
chisel3
14cc99a8880ba2ee735f888cec70c7cc99235a4b
chisel-testers
496bd86063b9a47a0bb13264068efdf94bd98a66
from dsptools.
shunshou
commented on September 13, 2024
Yay looks like i was able to sbt publish-local everything this time around, but dsptools still needs editing :( --
(When I run FFT-local tests):
java.lang.NoSuchMethodError: firrtl_interpreter.InterpreterOptions.copy(ZZZZZJLscala/collection/Seq;JZZ)Lfirrtl_interpreter/InterpreterOptions;
at dsptools.Driver$$anonfun$execute$1.apply$mcZ$sp(Driver.scala:27)
at dsptools.Driver$$anonfun$execute$1.apply(Driver.scala:26)
at dsptools.Driver$$anonfun$execute$1.apply(Driver.scala:26)
at scala.util.DynamicVariable.withValue(DynamicVariable.scala:58)
at dsptools.Driver$.execute(Driver.scala:26)
at dsptools.intervals.tests.BasicIASpec$$anonfun$2.apply(BasicIA.scala:134)
at dsptools.intervals.tests.BasicIASpec$$anonfun$2.apply(BasicIA.scala:135)
at org.scalatest.OutcomeOf$class.outcomeOf(OutcomeOf.scala:85)
at org.scalatest.OutcomeOf$.outcomeOf(OutcomeOf.scala:104)
at org.scalatest.Transformer.apply(Transformer.scala:22)
at org.scalatest.Transformer.apply(Transformer.scala:20)
at org.scalatest.FlatSpecLike$$anon$1.apply(FlatSpecLike.scala:1682)
at org.scalatest.TestSuite$class.withFixture(TestSuite.scala:196)
at org.scalatest.FlatSpec.withFixture(FlatSpec.scala:1685)
at org.scalatest.FlatSpecLike$class.invokeWithFixture$1(FlatSpecLike.scala:1679)
at org.scalatest.FlatSpecLike$$anonfun$runTest$1.apply(FlatSpecLike.scala:1692)
at org.scalatest.FlatSpecLike$$anonfun$runTest$1.apply(FlatSpecLike.scala:1692)
at org.scalatest.SuperEngine.runTestImpl(Engine.scala:289)
at org.scalatest.FlatSpecLike$class.runTest(FlatSpecLike.scala:1692)
at org.scalatest.FlatSpec.runTest(FlatSpec.scala:1685)
at org.scalatest.FlatSpecLike$$anonfun$runTests$1.apply(FlatSpecLike.scala:1750)
at org.scalatest.FlatSpecLike$$anonfun$runTests$1.apply(FlatSpecLike.scala:1750)
at org.scalatest.SuperEngine$$anonfun$traverseSubNodes$1$1.apply(Engine.scala:396)
at org.scalatest.SuperEngine$$anonfun$traverseSubNodes$1$1.apply(Engine.scala:384)
at scala.collection.immutable.List.foreach(List.scala:381)
at org.scalatest.SuperEngine.traverseSubNodes$1(Engine.scala:384)
from dsptools.
chick
commented on September 13, 2024
@shunshou What's the branch of dsptools
from dsptools.
shunshou
commented on September 13, 2024
I assumed you were on master, so I switched to that (but that's probably just not compatible w/ old stuff...).
from dsptools.
shunshou
commented on September 13, 2024
https://github.com/ucb-art/Chisel3DSPDependencies/tree/ranges (range branch!) points to stuff that should hopefully work now:
firrtl: intervals-oct
chisel3: intervals-oct
firrtl-interpreter: instrumenting-sizes
chisel-testers: intervals-oct
dsptools: intervals-oct
(Note: not sure if the default versions of chisel3/chisel-testers need updating in build.sbt? -- more generally; not for this particular issue).
from dsptools.
shunshou
commented on September 13, 2024
@chick bpsl, bpsr, bpset should also be implemented in chisel, if it hasn't already.
from dsptools.
chick
commented on September 13, 2024
I have just pushed to intervals-oct new methods on Interval
.setBinaryPoint(that: Int)
.shiftLeftBinaryPoint(that: Int)
.shiftRightBinaryPoint(that: Int)
I am not sure they are implemented exactly right, I need to check with Adam, but you can start to try them
from dsptools.
shunshou
commented on September 13, 2024
Cool. I'm still in the process of debugging some of the shifts... Adam and I will be having a working meeting today ~1 to maybe up to 3ish? We should just clean out these details and make basic tests for everything while people are here... :o
from dsptools.
chick
commented on September 13, 2024
@shunshou I had to push some small changes to instrumenting-sizes branch of the interpreter.
I wrote up a little bit here Instrumenting signals
from dsptools.
chick
commented on September 13, 2024
I have started a documentation page on chisel3 wiki for Intervals
from dsptools.
shunshou
commented on September 13, 2024
Ok, I'm still debugging some intermediate shr weirdness -- doesn't seem like I'm getting the right answer when I'm shifting by less than the width either :, but need to finish up slides for another conference tonight. Thanks for starting the intervals documentation :D
from dsptools.
shunshou
commented on September 13, 2024
@chick what implications are there if chisel range inference is wrong? (and firrtl is right?)
from dsptools.
chick
commented on September 13, 2024
It would imply to me that things are broken. In general, @azidar and I have tried to match the firrtl behavior, and in many cases actually use the firrtl code generate answers when we can. I think there a number of cases where chisel might defer to firrtl, by just leaving things unknown so firrtl will figure it out, but it shouldn't contradict it
from dsptools.
shunshou
commented on September 13, 2024
OK. I went ahead and double checked stuff and made things unknown wherever I felt there was a contradiction... Still not quite at the point of passing enough of my own tests, but getting closer.
from dsptools.
shunshou
commented on September 13, 2024
@chick -- looks like x.clip(uint/sint) and x.wrap(uint/sint) don't work. Namely, there's no firrtl PrimOp that operates on 1 Interval and 1 UInt/SInt type... I thought I could handle this at the Chisel level by first doing asInterval on the UInt -- when the width is known, everything's good... unfortunately, when the width is unknown, I don't think there's any mechanism to pass the unknown (?) through from the Chisel emitter... [at least for asInterval?].
from dsptools.
chick
commented on September 13, 2024
Hmm, it looks to me like x.wrap(sint) might be a problem, but the
rest of the family convert the argument to an interval before passing down to firrtl.
But clearly I need to expand the tests to include use of the SInt and UInt versions of
clip and wrap, also to add tests for all these with some arguments with unknown ranges. Sorry, but it's going to be later tonight before I can get to it.
from dsptools.
shunshou
commented on September 13, 2024
Hm ok I made some changes assuming known UInt and SInt. I'm making my own
tests for them locally (basically assuming my model is intended behavior).
If you can generalize to handle unknown ranges, that'd be great. Also,
looks like some of my changes broke some of your tests. I made some of them
based off of my understanding of how functions should operate (from various
discussions), so we should make sure our understandings of behaviors match
and better document.
…On Sunday, October 22, 2017, Chick Markley ***@***.***> wrote:
Hmm, it looks to me like x.wrap(sint) might be a problem, but the
rest of the family convert the argument to an interval before passing down
to firrtl.
But clearly I need to expand the tests to include use of the SInt and UInt
versions of
clip and wrap, also to add tests for all these with some arguments with
unknown ranges. Sorry, but it's going to be later tonight before I can get
to it.
—
You are receiving this because you were mentioned.
Reply to this email directly, view it on GitHub
<#112 (comment)>,
or mute the thread
<https://github.com/notifications/unsubscribe-auth/AGTTFl1oN83VZZ82hlcQqV4k6qHLIzk-ks5su6figaJpZM4PzGSp>
.
from dsptools.
shunshou
commented on September 13, 2024
FYI--when adding tests, it's important to test stuff that's not just integers. Running into annoying binary point alignment problems, etc.
from dsptools.
shunshou
commented on September 13, 2024
@chick to save me some time of trying to figure this out -- can you make a toy example for how firrtl interpreter stuff works with the driver in DspTools? (Likely stuff will need modification?)
from dsptools.
chick
commented on September 13, 2024
Almost got it working, but found some edge case problems :-(.
from dsptools.
shunshou
commented on September 13, 2024
no worries. i'm still trying to get accumulate working. :\ slow process...
from dsptools.
shunshou
commented on September 13, 2024
Re: Interpreter --
I think we agreed on trimming widths based on some n sigma criteria or min/max. The other thing that's useful is that, if we're trimming based off of sigma, it should be trim and optionally clip to maximum.
from dsptools.
chick
commented on September 13, 2024
The toy example works with my latest push/merges to
firrtl-interpreter and dsptools on intervals-oct branches
See InstrumentingSpec for how.
Time to start working on feedback system
from dsptools.
shunshou
commented on September 13, 2024
Yay! OK. I will try to run the interpreter on my interval arithmetic testbench and get back to you.
from dsptools.
shunshou
commented on September 13, 2024
@azidar What do you expect should happen in these cases?
-
Directly assign a <= b where a has known range that's smaller than b. Should that be an error without explicit b.wrap(a), b.clip(a), or b.assignInterval(a)? Also, re: back-propagation of ranges, if b has some smaller bound and b <= x - y, I assume you can't make any assumptions about the ranges of x and y. However, if you've constrained x and y to be only positive and use only multiply and add ops i.e. b <= x + y, you can further optimize x and y ranges -- should be min(x_range, b_range). That's not something the tool should be able to figure out, maybe? But it'd be nice to have some optional Firrtl pass go and trim widths that way, maybe (unless synthesis tools are smart about that). More generally, does a always keep its range if known or will it take on b's range if b's range is smaller?
-
Directly assign a <= b where a's bp = 5 and b's bp = 12 (both known). You're losing precision, which means that your interval also changes. In that case, before range inference, it might make sense to explicitly add b.bpset(a)? -- Otherwise, you're propagating the wrong ranges, which potentially accumulates and becomes probelamtic.
Looks like a always keeps a's range...
from dsptools.
shunshou
commented on September 13, 2024
@chick @azidar @grebe fyi firrtl is on intervals-oct2 now.
from dsptools.
shunshou
commented on September 13, 2024
@chick @grebe -- Intervals type classes should hopefully be in a workable state, so can actually really start using them. QOR for the firrtl output looks like what I calculated + the firrtl-interpreter sims got the right mins + maxes for the one test I paid closest attention to. I'm going to port over my testing infrastructure from my local repo to dsptools so hopefully it'll be easier for you to play with stuff. Should be up in the next hour or so? Then I'll go and try more complicated circuits and run them through some real synthesis tool.
from dsptools.
Related Issues (20)
- dsptools/blob/master/rocket/README.md links not working
- DspComplex[T].zero not initialized HOT 1
- Instantiating constants of type DspComplex[DspReal] HOT 7
- Compilation fails in scala 2.12.8 HOT 2
- Pokeable Typeclasses for Testers HOT 1
- Remove updateDependencies script
- Document Width Relationships for Operations
- Add -D flag for rocket source dependency
- Is there any plan to update the SAM module? HOT 1
- Issue with TileLink ready-valid logic in TLMasterModel HOT 2
- Run-time failures for DspComplex isLit()
- Read the docs
- rocket-dsptools DmaSpec test failure with current rocket-chip
- Working with large FixedPoint and Intervals should use BigDecimal not Double
- Set precision in div2 fixedpoint and interval may have provblems
- Update the link to numeric classes description HOT 2
- Failing Tests on 3.4.1 HOT 1
- example code not working
- Library conflict HOT 1
- Arithmetic between types
Recommend Projects
-
React
A declarative, efficient, and flexible JavaScript library for building user interfaces.
-
Vue.js
🖖 Vue.js is a progressive, incrementally-adoptable JavaScript framework for building UI on the web.
-
Typescript
TypeScript is a superset of JavaScript that compiles to clean JavaScript output.
-
TensorFlow
An Open Source Machine Learning Framework for Everyone
-
Django
The Web framework for perfectionists with deadlines.
-
Laravel
A PHP framework for web artisans
-
D3
Bring data to life with SVG, Canvas and HTML. 📊📈🎉
-
Recommend Topics
-
javascript
JavaScript (JS) is a lightweight interpreted programming language with first-class functions.
-
web
Some thing interesting about web. New door for the world.
-
server
A server is a program made to process requests and deliver data to clients.
-
Machine learning
Machine learning is a way of modeling and interpreting data that allows a piece of software to respond intelligently.
-
Visualization
Some thing interesting about visualization, use data art
-
Game
Some thing interesting about game, make everyone happy.
Recommend Org
-
Facebook
We are working to build community through open source technology. NB: members must have two-factor auth.
-
Microsoft
Open source projects and samples from Microsoft.
-
Google
Google ❤️ Open Source for everyone.
-
Alibaba
Alibaba Open Source for everyone
-
D3
Data-Driven Documents codes.
-
Tencent
China tencent open source team.
from dsptools.