Add Preference to disable runtime invalidation #22

topolarity · 2025-11-19T22:55:49Z

More conservative version of #21 (thanks to @el-oso for the inspiration).

Adds three preferences:

cpu_target: if provided, use this for feature detection instead of JULIA_CPU_TARGET
freeze_cpu_target: if true, "freeze" the features detected based on your precompile-time CPU target
allow_runtime_invalidation: if false, warn instead of invalidating when CPU features don't match precompile-time

┌ Warning: Runtime invalidation was disabled, but the CPU info is out-of-date.
│ Will continue with incorrect CPU name (from build time).
└ @ HostCPUFeatures ~/repos/HostCPUFeatures.jl/src/HostCPUFeatures.jl:62

This makes the package --trim compatible and also provides a nice preference for users who do not wish for HostCPUFeatures.jl to ever cause an "invalidation storm", even if it means running with the wrong CPU info.

This makes the package (nearly) `--trim` compatible and also provides a nice preference for users who do not wish for HostCPUFeatures.jl to ever cause an "invalidation storm", even if it means running with the wrong CPU info.

oscardssmith · 2025-11-20T14:51:08Z

src/cpu_info_x86.jl

 fast_half() = False()

-@noinline function setfeaturefalse(s)
+@inline function setfeaturefalse(s)


why the inlining switch?

This was to improve inferrability - has_feature(Val(s)) is only inferrable if the literal value of s is available in the function, which is true in all of the callers (the argument is always a literal Symbol)

codecov · 2025-11-20T14:51:26Z

Codecov Report

❌ Patch coverage is 15.38462% with 33 lines in your changes missing coverage. Please review.
✅ Project coverage is 31.45%. Comparing base (8ebc3d2) to head (bddfae2).
⚠️ Report is 12 commits behind head on main.

Files with missing lines	Patch %	Lines
src/cpu_info.jl	0.00%	9 Missing ⚠️
src/cpu_info_x86.jl	0.00%	9 Missing ⚠️
src/HostCPUFeatures.jl	42.85%	8 Missing ⚠️
src/cpu_info_aarch64.jl	0.00%	7 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main      #22      +/-   ##
==========================================
- Coverage   36.12%   31.45%   -4.68%     
==========================================
  Files           6        6              
  Lines         191      213      +22     
==========================================
- Hits           69       67       -2     
- Misses        122      146      +24

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

🚀 New features to boost your workflow:

❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

src/cpu_info.jl

tz-lom · 2025-11-20T15:11:52Z

src/HostCPUFeatures.jl

  reset_extra_features!()
 end
 const BASELINE_CPU_NAME = get_cpu_name()
+const allow_eval = @load_preference("allow_runtime_invalidation", false)


Why not to check that --trim is enabled in JLOptions ?

everyone who will use it with trim will have to find out this culprit themself

The main thinking is that this preference is usable by more than just JuliaC.jl, since some users may wish to be opt-in to error / warn on invalidation storms from this package. JuliaC.jl can set this preference automatically, so the end-user experience is the same.

Also technically checking JLOptions at pre-compilation time won't detect --trim properly, but something like JuliaLang/JuliaC.jl#31 would work

tz-lom · 2025-11-20T15:15:20Z

As a source of the problem I understand the approach, but it actually downgrades package HostCPUFeatures to MaybeSomeOtherHostCPUFeatures, while what we really need is to use some CompileTargetCPUFeatures in LoopVectorization etc.

Pragmatically it is "easier" upgrade standard interface of HostCPUFeatures to return features for the compile target and provide API to retrieve real Host features if that what user really wants (like printing statistics on which CPU we are running now)

topolarity · 2025-11-21T15:17:40Z

As a source of the problem I understand the approach, but it actually downgrades package HostCPUFeatures to MaybeSomeOtherHostCPUFeatures, while what we really need is to use some CompileTargetCPUFeatures in LoopVectorization etc.

That's already true if "native" is missing from your JULIA_CPU_TARGET string though - I agree that it might be a good idea to require that we are specifically operating in the non-native case though (where we already under-approximate feature flags)

…time We already support this dynamically, but for some reason it wasn't possible to get this configuration at precompilation time. Also adds a "freeze_cpu_target" preference to enable this behavior by default.

gbaraldi · 2025-12-04T13:25:32Z

src/cpu_info.jl

        if _has_feature(feature) ≠ has
-            @debug "Defining $(has ? "presence" : "absense") of feature $feature."
-            set_feature(feature, has)
+            if allow_eval


This code doesn't execute if there's a preference right?

Depends on which preference you mean

Are you saying you don't think this guard is necessary?

oscardssmith · 2025-12-04T15:14:13Z

Looks like the Julia 1.5 CI is failing. Time to bump minimum version to LTS?

oscardssmith · 2025-12-04T15:14:22Z

Other than that, LGTM.

oscardssmith · 2025-12-04T15:15:11Z

@tz-lom are you good with this in current state? If so, I'll merge and release.

tz-lom · 2025-12-04T15:31:25Z

@tz-lom are you good with this in current state? If so, I'll merge and release.

@oscardssmith , thanks, it looks ok.
I think you need to fix test matrix to solve failing tests, and it would be nice if you put description of preferences in documentation, because now you have to dig through the code to find out this feature.

topolarity · 2025-12-05T18:53:47Z

@tz-lom Anything else you need from me on this?

tz-lom · 2025-12-06T07:41:56Z

@topolarity , no that is enough, thank you very much for your work

tz-lom · 2025-12-08T16:07:52Z

@topolarity hi Cody, I think we did whoopsie here
I think version was not bumped, and I see that link to documentation from the badge in README is not working (wrong link?)

also check open issue #15 , maybe it can be closed?

oscardssmith · 2025-12-08T17:14:47Z

I'll bump and release.

oscardssmith · 2025-12-08T18:23:58Z

@topolarity you missed a compat entry on Preferences.jl

Add Preference to disable runtime invalidation

723446a

This makes the package (nearly) `--trim` compatible and also provides a nice preference for users who do not wish for HostCPUFeatures.jl to ever cause an "invalidation storm", even if it means running with the wrong CPU info.

topolarity mentioned this pull request Nov 19, 2025

Add HostCPUFeatures.jl compatibility by providing freeze_cpu_target = true Preference JuliaLang/JuliaC.jl#80

Open

oscardssmith reviewed Nov 20, 2025

View reviewed changes

tz-lom reviewed Nov 20, 2025

View reviewed changes

src/cpu_info.jl Show resolved Hide resolved

tz-lom reviewed Nov 20, 2025

View reviewed changes

topolarity force-pushed the ct/no-eval-pref branch 7 times, most recently from 116a624 to 09a8d77 Compare December 4, 2025 00:58

Add build_cpu_target global to opt-in to generic CPU at precompile-…

bbd20e1

…time We already support this dynamically, but for some reason it wasn't possible to get this configuration at precompilation time. Also adds a "freeze_cpu_target" preference to enable this behavior by default.

topolarity force-pushed the ct/no-eval-pref branch from 09a8d77 to bbd20e1 Compare December 4, 2025 01:18

gbaraldi reviewed Dec 4, 2025

View reviewed changes

topolarity added 2 commits December 4, 2025 10:42

Bump minimum Julia version to 1.10

c118075

Add preferences documentation

bddfae2

oscardssmith approved these changes Dec 4, 2025

View reviewed changes

oscardssmith merged commit ae4fdac into JuliaSIMD:main Dec 7, 2025
16 of 18 checks passed

oscardssmith mentioned this pull request Dec 8, 2025

bump version #23

Merged

Add Preference to disable runtime invalidation #22

Add Preference to disable runtime invalidation #22

Uh oh!

Conversation

topolarity commented Nov 19, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

oscardssmith Nov 20, 2025

Choose a reason for hiding this comment

Uh oh!

topolarity Nov 21, 2025

Choose a reason for hiding this comment

Uh oh!

codecov bot commented Nov 20, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Codecov Report

Uh oh!

Uh oh!

tz-lom Nov 20, 2025

Choose a reason for hiding this comment

Uh oh!

topolarity Nov 21, 2025

Choose a reason for hiding this comment

Uh oh!

tz-lom commented Nov 20, 2025

Uh oh!

topolarity commented Nov 21, 2025

Uh oh!

gbaraldi Dec 4, 2025

Choose a reason for hiding this comment

Uh oh!

topolarity Dec 4, 2025

Choose a reason for hiding this comment

Uh oh!

oscardssmith commented Dec 4, 2025

Uh oh!

oscardssmith commented Dec 4, 2025

Uh oh!

oscardssmith commented Dec 4, 2025

Uh oh!

tz-lom commented Dec 4, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

topolarity commented Dec 5, 2025

Uh oh!

tz-lom commented Dec 6, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

tz-lom commented Dec 8, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

oscardssmith commented Dec 8, 2025

Uh oh!

oscardssmith commented Dec 8, 2025

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

topolarity commented Nov 19, 2025 •

edited

Loading

codecov bot commented Nov 20, 2025 •

edited

Loading

tz-lom commented Dec 4, 2025 •

edited

Loading

tz-lom commented Dec 6, 2025 •

edited

Loading

tz-lom commented Dec 8, 2025 •

edited

Loading