Create provisional and compat features around the vect type #328

mandolaerik · 2024-10-21T09:42:55Z

Include docs for compat and provisionals in 1.2 refman too
Allow provisional declarations in DML 1.2
Repair pickling of provisionals
Create provisional and compat features around the vect type
Dogfood the c_vect provisional

py/dml/messages.py

py/dml/provisional.py

test/1.4/legacy/T_c_vect_without_provisional_enabled.dml

lib/1.4/dml-builtins.dml

lib/1.2/dml-builtins.dml

syssimics · 2024-10-21T12:31:48Z

Verification #14005157: ❌ fail

syssimics · 2024-10-22T11:25:01Z

Verification #14009292: ❌ fail

syssimics · 2024-10-22T13:24:15Z

Verification #14009657: ❌ fail

syssimics · 2024-10-22T16:54:40Z

Verification #14010442: ❌ fail

syssimics · 2024-10-22T19:20:38Z

Verification #14010884: ✅ pass

The duplicated headers are unfortunate, and is there because one of them mentions "1.4" specifically. It's not a problem if we let the 1.2 versions rot.

RELEASENOTES.md

doc/1.2/deprecations-header.md

lwaern-intel · 2025-06-25T21:36:46Z

lib/1.2/utility.dml

            log_level = 2;
-        } else {
+        } else if ($dev._compat_warning_statement) {
            _warning "The write_only template only makes sense on registers."


I don't get why this works. Why don't you get ESYNTAX on this when the compat feature is disabled? #if shouldn't stop the AST from being built. Or... does it??!

Haha, it's because we suppress the warning while building dml-builtins.dmlast. After that the parser will not look at dml-builtins.dml again.

Oh good lord. So you're saying this code is only meaningful because we can leverage dmlast. Let me see if I get this right:

The .dmlast is prebuilt, and it is so with the compat feature enabled. Because of that, it's able to use the _warning statement without provoking ESYNTAX, no matter if the compat feature is enabled/disabled for the individual device.

But the value of $dev._compat_warning_statement is resolved only when it's time to actually build the device. Because of that, if the individual device does not have that compat feature enabled, then compilation will end up leveraging error instead of _warning .

Is that correct? That's...
I...
I don't like that. I don't like that we have code in dml-builtins/utility that only works because of dmlast caching; because we're able to parse and build the AST using a different set of flags compared to when building the device. That feels extremely bad.
Not to mention, people (at least I) look towards dml-builtins/utility for reference and inspiration -- this code snippet could leave people thinking that the pattern

#if ($dev._compat_warning_statement) { _warning "..."; }

would work in general, when the fact is it only works due to extremely niche circumstances. Worse yet, they may extrapolate to expect that pattern would work for other syntactical features controlled by a compatibility feature.

You must at the very least add a comment about this. This should be treated as a HACK. A pretty egregious one at that -- you would never expect the correctness of a piece of model code to be dependent on whether or not it gets cached.

Sure, I can add a comment.

Not to mention, people (at least I) look towards dml-builtins/utility for reference and inspiration

If you are in search of inspiration, then visiting to DML 1.2 code is strongly discouraged. Not good for your health.

(I would also have been concerned about the hackiness of this code if this wasn't dying code)

lib/1.2/utility.dml

lwaern-intel · 2025-06-25T21:39:46Z

py/dml/provisional.py

+      }
+      ```
+
+    * DML natively supports `vect` types in `foreach` statements.


Only true in DML 1.2!

Added a test to prove you wrong

Added a comment to prove you proving me wrong, wrong!

py/dml/provisional.py

syssimics · 2025-06-25T22:18:16Z

PR Verification: ✅ success

syssimics · 2025-06-25T22:43:38Z

PR Verification: ❌ failure

syssimics · 2025-06-26T06:33:02Z

PR Verification: ✅ success

test/1.4/misc/T_vect.dml

lwaern-intel

See my existential crisis re. abusing .dmlast caching

syssimics · 2025-06-26T09:56:28Z

PR Verification: ✅ success

On popular demand. Also, rename ECVECT -> EOLDVECT and c_vect_without_provisional -> experimental_vect

lwaern-intel

Approved modulo verification. Though I do feel inclined to bikeshed a bit.

lwaern-intel · 2025-06-26T10:19:21Z

py/dml/provisional.py

+    error unless the [`experimental_vect` compatibility
+    feature](deprecations-auto.html#experimental_vect) is enabled.
+    '''
+    short = "Allow vect syntax based on the VECT macro"


I'm inclined to bikeshed this short desc a bit, for multiple reasons:

To make it more obvious as to what the syntax actually is and does (it's possibly to read this as the DML syntax also being VECT)

To explicitly state the connection to the Simics API

And to somewhat future-proof the formulation against the introduction of RAII vectors

My suggestion would be:

Enable 'vect' type syntax that compile to usages of the Simics 'VECT' macro

However, I don't insist on any on this, and this is not blocking.

I prefer to leave disambiguation to the full docs. I guess most of the time when you look at the list it's one of the other features you are looking for, which means the main use case for the short desc is to quickly convince you that this is not the thing you are looking for. So better to be terse and simple. If you don't know what the VECT macro is about, then you don't use the feature and the short desc is not for you.

syssimics · 2025-06-26T10:53:28Z

PR Verification: ✅ success

mandolaerik requested a review from lwaern-intel October 21, 2024 09:43

mandolaerik changed the title ~~Dogfood the c_vect provisional~~ Create provisional and compat features around the vect type Oct 21, 2024

lwaern-intel requested changes Oct 21, 2024

View reviewed changes

mandolaerik force-pushed the pr/dogfood-the-c_vect-provisional branch from 6e31f31 to 1d81bd5 Compare October 21, 2024 12:02

mandolaerik requested a review from lwaern-intel October 21, 2024 12:02

mandolaerik force-pushed the pr/dogfood-the-c_vect-provisional branch from 1d81bd5 to b2a4cfd Compare October 22, 2024 10:57

mandolaerik force-pushed the pr/dogfood-the-c_vect-provisional branch from b2a4cfd to c587b73 Compare October 22, 2024 12:37

mandolaerik force-pushed the pr/dogfood-the-c_vect-provisional branch from c587b73 to 015229f Compare October 22, 2024 15:45

mandolaerik added 3 commits June 25, 2025 23:34

Include docs for compat and provisionals in 1.2 refman too

3d817de

The duplicated headers are unfortunate, and is there because one of them mentions "1.4" specifically. It's not a problem if we let the 1.2 versions rot.

Allow provisional declarations in DML 1.2

dbc0d50

Repair pickling of provisionals

66e27a7

mandolaerik force-pushed the pr/dogfood-the-c_vect-provisional branch from 72156e4 to a4065c4 Compare June 25, 2025 21:37

lwaern-intel requested changes Jun 25, 2025

View reviewed changes

mandolaerik force-pushed the pr/dogfood-the-c_vect-provisional branch from a4065c4 to a4864cf Compare June 25, 2025 22:12

mandolaerik requested a review from lwaern-intel June 25, 2025 22:13

mandolaerik force-pushed the pr/dogfood-the-c_vect-provisional branch from a4864cf to acea79a Compare June 26, 2025 05:45

lwaern-intel requested changes Jun 26, 2025

View reviewed changes

test/1.4/misc/T_vect.dml Outdated Show resolved Hide resolved

mandolaerik added 2 commits June 26, 2025 11:18

Create provisional and compat features around the vect type

3835a44

Dogfood the c_vect provisional

fbfceab

mandolaerik force-pushed the pr/dogfood-the-c_vect-provisional branch from acea79a to 70d4d0e Compare June 26, 2025 09:19

syssimics requested a review from lwaern-intel June 26, 2025 09:19

lwaern-intel requested changes Jun 26, 2025

View reviewed changes

mandolaerik added 5 commits June 26, 2025 12:17

Deprecate the _warning statement

0ee40d8

Rename c_vect -> simics_util_vect

97cff78

On popular demand. Also, rename ECVECT -> EOLDVECT and c_vect_without_provisional -> experimental_vect

Avoid warning

6ca7575

Adjust script to cope with warning statement

4d17335

Add rudimentary test for vect in 1.4

8e97202

mandolaerik force-pushed the pr/dogfood-the-c_vect-provisional branch from 70d4d0e to 8e97202 Compare June 26, 2025 10:17

syssimics requested a review from lwaern-intel June 26, 2025 10:18

lwaern-intel approved these changes Jun 26, 2025

View reviewed changes

mandolaerik merged commit 2f0f7b1 into intel:main Jun 26, 2025

mandolaerik deleted the pr/dogfood-the-c_vect-provisional branch June 26, 2025 11:26

Create provisional and compat features around the vect type #328

Create provisional and compat features around the vect type #328

Uh oh!

Conversation

mandolaerik commented Oct 21, 2024 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

Uh oh!

syssimics commented Oct 21, 2024

Uh oh!

syssimics commented Oct 22, 2024

Uh oh!

syssimics commented Oct 22, 2024

Uh oh!

syssimics commented Oct 22, 2024

Uh oh!

syssimics commented Oct 22, 2024

Uh oh!

Uh oh!

Uh oh!

lwaern-intel Jun 25, 2025

Choose a reason for hiding this comment

Uh oh!

mandolaerik Jun 25, 2025

Choose a reason for hiding this comment

Uh oh!

lwaern-intel Jun 26, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

lwaern-intel Jun 26, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

mandolaerik Jun 26, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Choose a reason for hiding this comment

Uh oh!

Uh oh!

lwaern-intel Jun 25, 2025

Choose a reason for hiding this comment

Uh oh!

mandolaerik Jun 25, 2025

Choose a reason for hiding this comment

Uh oh!

lwaern-intel Jun 26, 2025

Choose a reason for hiding this comment

Uh oh!

Uh oh!

syssimics commented Jun 25, 2025

Uh oh!

syssimics commented Jun 25, 2025

Uh oh!

syssimics commented Jun 26, 2025

Uh oh!

Uh oh!

lwaern-intel left a comment

Choose a reason for hiding this comment

Uh oh!

syssimics commented Jun 26, 2025

Uh oh!

lwaern-intel left a comment

Choose a reason for hiding this comment

Uh oh!

lwaern-intel Jun 26, 2025

Choose a reason for hiding this comment

Uh oh!

mandolaerik Jun 26, 2025

Choose a reason for hiding this comment

Uh oh!

syssimics commented Jun 26, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

mandolaerik commented Oct 21, 2024 •

edited

Loading

lwaern-intel Jun 26, 2025 •

edited

Loading

lwaern-intel Jun 26, 2025 •

edited

Loading

mandolaerik Jun 26, 2025 •

edited

Loading