[OpenSmalltalk/opensmalltalk-vm] Aarch64 integer compare bug (#<) (#571)

 

16r80000000 < 16r8000000000000000. "==> false "

[Intel reports true]

—
You are receiving this because you are subscribed to this thread.
Reply to this email directly, view it on GitHub, or unsubscribe.

<script type="application/ld+json">[ { "@context": "http://schema.org", "@type": "EmailMessage", "potentialAction": { "@type": "ViewAction", "target": "https://github.com/OpenSmalltalk/opensmalltalk-vm/issues/571", "url": "https://github.com/OpenSmalltalk/opensmalltalk-vm/issues/571", "name": "View Issue" }, "description": "View this Issue on GitHub", "publisher": { "@type": "Organization", "name": "GitHub", "url": "https://github.com" } } ]</script>

 
I think "good grief" is the only polite response here...


tim
--
tim Rowledge; [hidden email]; http://www.rowledge.org/tim
Strange OpCodes: BBL: Branch on Burned out Light

 

 

That is indeed a great test case that might help narrow it down:

If you "debug into it" you should see that primitive 3 failed for SmallInteger, and it falls back to Integer's version of <. If it doesn't, then it's either the JIT incorrectly short-circuiting the comparison, or primitive 3 incorrectly succeeding.

If instead you do see the super send (as is the case on Intel), then it must be primitive primDigitCompare in LargeIntegers plugin.

Very much sounds like a signed/unsigned problem.

—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub, or unsubscribe.

<script type="application/ld+json">[ { "@context": "http://schema.org", "@type": "EmailMessage", "potentialAction": { "@type": "ViewAction", "target": "https://github.com/OpenSmalltalk/opensmalltalk-vm/issues/571#issuecomment-863488032", "url": "https://github.com/OpenSmalltalk/opensmalltalk-vm/issues/571#issuecomment-863488032", "name": "View Issue" }, "description": "View this Issue on GitHub", "publisher": { "@type": "Organization", "name": "GitHub", "url": "https://github.com" } } ]</script>

 

On 2021-06-17 11:57, Vanessa Freudenberg wrote:

> That is indeed a great test case that might help narrow it down:
>
> If you "debug into it" you should see that primitive 3 failed for
> SmallInteger, and it falls back to Integer's version of <. If it
> doesn't, then it's either the JIT incorrectly short-circuiting the
> comparison, or primitive 3 incorrectly succeeding.

Tried "Debug It" + "Setp Into" in Squeak and Cuis. Both return 'false'
without interior detail.

Also same with current
~/OpenSmalltalk/oscogvm/build.linux64ARMv8/squeak.cog.spur/build
as well as the "ARMv8 BitBLT speedups provided by RPi/Ben Avison" VM.

vvv==========================vvv
Chromebook:Linux:Arm64:~/Squeak5.3 >>> squeak --version
5.0-202103220146 Sat May 22 19:42:47 PDT 2021 gcc 8 [Production Spur
64-bit VM]
CoInterpreter VMMaker.oscog-eem.2948 uuid:
935d1ee7-0e61-47a5-92e2-953224c7ffcf May 22 2021
StackToRegisterMappingCogit VMMaker.oscog-eem.2946 uuid:
461911e1-d450-4f05-8a40-7b0470487041 May 22 2021
VM: 202103220146 ***@***.***:OpenSmalltalk/Bavison
Date: Sun Mar 21 21:46:24 2021 CommitHash: d494240
Plugins: 202103220146 ***@***.***:OpenSmalltalk/Bavison
Linux penguin 5.4.99-12983-g8b7876ab9f5e #1 SMP PREEMPT Sun Feb 21
11:49:16 PST 2021 aarch64 GNU/Linux
plugin path: /usr/local/bin/../lib/squeak/5.0-202103220146 [default:
/usr/local/lib/squeak/5.0-202103220146/]
Chromebook:Linux:Arm64:~/Squeak5.3 >>>
...
Chromebook:Linux:Arm64:~/OpenSmalltalk/oscogvm/products/sqcogspur64ARMv8linuxht/lib/squeak/5.0-202106162331
>>> ./squeak --version
5.0-202106162331 Wed Jun 16 16:39:14 PDT 2021 gcc 8 [Production Spur
64-bit VM]
CoInterpreter VMMaker.oscog-eem.2967 uuid:
57b9e5f9-0212-436d-acaf-4e501e470621 Jun 16 2021
StackToRegisterMappingCogit VMMaker.oscog-eem.2953 uuid:
9f3d924e-9226-4242-9b6f-3dad93c7a837 Jun 16 2021
VM: 202106162331 ***@***.***:OpenSmalltalk/oscogvm
Date: Wed Jun 16 16:31:37 2021 CommitHash: 58e8b33db
Plugins: 202106162331 ***@***.***:OpenSmalltalk/oscogvm
Linux penguin 5.4.109-26077-g12746d86825a #1 SMP PREEMPT Mon May 17
21:46:50 PDT 2021 aarch64 GNU/Linux
plugin path:
/home/kendi3he/OpenSmalltalk/oscogvm/products/sqcogspur64ARMv8linuxht/lib/squeak/5.0-202106162331/
[default:
/home/kendi3he/OpenSmalltalk/oscogvm/products/sqcogspur64ARMv8linuxht/lib/squeak/5.0-202106162331/]

^^^==========================^^^

—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub, or unsubscribe.

<script type="application/ld+json">[ { "@context": "http://schema.org", "@type": "EmailMessage", "potentialAction": { "@type": "ViewAction", "target": "https://github.com/OpenSmalltalk/opensmalltalk-vm/issues/571#issuecomment-863531725", "url": "https://github.com/OpenSmalltalk/opensmalltalk-vm/issues/571#issuecomment-863531725", "name": "View Issue" }, "description": "View this Issue on GitHub", "publisher": { "@type": "Organization", "name": "GitHub", "url": "https://github.com" } } ]</script>

 

Thanks for testing that. My guess then would be that the jitted code for < fails to recognize 16r8000000000000000 as being a LargeInteger (which is unlikely), or knowing it's a LargeInt but thinking it can handle the comparison quickly and failing to do so correctly.

—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub, or unsubscribe.

<script type="application/ld+json">[ { "@context": "http://schema.org", "@type": "EmailMessage", "potentialAction": { "@type": "ViewAction", "target": "https://github.com/OpenSmalltalk/opensmalltalk-vm/issues/571#issuecomment-863568544", "url": "https://github.com/OpenSmalltalk/opensmalltalk-vm/issues/571#issuecomment-863568544", "name": "View Issue" }, "description": "View this Issue on GitHub", "publisher": { "@type": "Organization", "name": "GitHub", "url": "https://github.com" } } ]</script>

 

Has anyone built the GdbARMv8Plugin on a Raspberry Pi?

—
You are receiving this because you commented.
Reply to this email directly, view it on GitHub, or unsubscribe.

<script type="application/ld+json">[ { "@context": "http://schema.org", "@type": "EmailMessage", "potentialAction": { "@type": "ViewAction", "target": "https://github.com/OpenSmalltalk/opensmalltalk-vm/issues/571#issuecomment-864541571", "url": "https://github.com/OpenSmalltalk/opensmalltalk-vm/issues/571#issuecomment-864541571", "name": "View Issue" }, "description": "View this Issue on GitHub", "publisher": { "@type": "Organization", "name": "GitHub", "url": "https://github.com" } } ]</script>

 


> On 2021-06-20, at 4:47 AM, Craig Latta <[hidden email]> wrote:
>
>
> Has anyone built the GdbARMv8Plugin on a Raspberry Pi?

Not to the best of my knowledge. I'd imagine it should work but the performance might be too slow to make it practical for anything beyond basic generate-method, disassemble etc.

tim
--
tim Rowledge; [hidden email]; http://www.rowledge.org/tim
Oxymorons: Alone together