Adding the x86 part of behavioural testing for std::arch intrinsics #1814

madhav-madhusoodanan · 2025-05-30T11:46:36Z

No description provided.

madhav-madhusoodanan · 2025-06-11T08:56:31Z

crates/intrinsic-test/src/common/intrinsic_helpers.rs

            match self {
                Self::BFloat => "bfloat",
                Self::Float => "float",
-                Self::Int => "int",
-                Self::UInt => "uint",
+                Self::Double => "double",
+                Self::Int(true) => "int",
+                Self::Int(false) => "uint",
                Self::Poly => "poly",
                Self::Void => "void",
+                Self::Char(true) => "char",
+                Self::Char(false) => "unsigned char",
+                Self::Short(true) => "short",
+                Self::Short(false) => "unsigned short",


@Amanieu @adamgemmell Are we using this trait (fmt::Display for TypeKind) anywhere in the crate? I was searching for its usage and I was unable to find any.

If we are not using this trait, may I delete the definition?

I suspect this may be useful for debug output during development. However since it isn't used for functionality then it's fine to make any changes you want to it.

madhav-madhusoodanan · 2025-06-11T08:57:15Z

Pardon the git history, I will be sorting that out shortly

Amanieu · 2025-06-18T08:11:38Z

crates/intrinsic-test/src/common/intrinsic_helpers.rs

-    UInt,
+
+    // if signed, then the inner value is true
+    Int(bool),


Rather than using a bool, considering instead something like:

enum Signed { Signed, Unsigned, }

This ends up making the code much more readable.

Amanieu · 2025-06-18T08:12:22Z

crates/intrinsic-test/src/x86/config.rs

+    format!(
+        "\
+{line_prefix}This is a transient test file, not intended for distribution. Some aspects of the
+{line_prefix}test are derived from a JSON specification, published under the same license as the


x86 uses an XML specification, not JSON.

madhav-madhusoodanan · 2025-06-19T06:54:16Z

The CI is giving me an odd error, even after I rebased from master:

  Updating crates.io index
     Locking 5 packages to latest compatible versions
      Adding quick-xml v0.37.5
      Adding serde-xml-rs v0.8.1
      Adding thiserror v1.0.69
      Adding thiserror-impl v1.0.69
      Adding xml-rs v0.8.26
   Compiling core_arch v0.1.5 (/Users/runner/work/stdarch/stdarch/crates/core_arch)
error: the feature `generic_arg_infer` has been stable since 1.89.0-nightly and no longer requires an attribute to enable
  --> crates/core_arch/src/lib.rs:34:5
   |
34 |     generic_arg_infer,
   |     ^^^^^^^^^^^^^^^^^
   |
   = note: `-D stable-features` implied by `-D warnings`
   = help: to override `-D warnings` add `#[allow(stable_features)]`

Amanieu · 2025-06-19T14:38:25Z

The CI is giving me an odd error, even after I rebased from master:

  Updating crates.io index
     Locking 5 packages to latest compatible versions
      Adding quick-xml v0.37.5
      Adding serde-xml-rs v0.8.1
      Adding thiserror v1.0.69
      Adding thiserror-impl v1.0.69
      Adding xml-rs v0.8.26
   Compiling core_arch v0.1.5 (/Users/runner/work/stdarch/stdarch/crates/core_arch)
error: the feature `generic_arg_infer` has been stable since 1.89.0-nightly and no longer requires an attribute to enable
  --> crates/core_arch/src/lib.rs:34:5
   |
34 |     generic_arg_infer,
   |     ^^^^^^^^^^^^^^^^^
   |
   = note: `-D stable-features` implied by `-D warnings`
   = help: to override `-D warnings` add `#[allow(stable_features)]`

Yes, this regularly happens due to changes in nightly Rust. In this case you can just remove the feature from lib.rs.

trait to make the same dyn-compatible.

…g capabilities

intrinsic-test's X86IntrinsicType struct

IntrinsicType impl

1. filtered any intrinsic arguments where the only parameter is of type "void" 2. Better parsing for intrinsic return type 3. Filtering intrinsics that are difficult to test (void returns) 4. Reduced the variants in TypeKind (which differed only in bit_len)

necessary. Further notes: 1. IntrinsicType.target -> IntrinsicType.metadata 2. Maintaining a HashMap to support arbitrary data storage for other arch 3. Create a dedicated "from_param" associated function for X86IntrinsicType

…ectly instead of constructing it. Further details: 1. moved "type_and_name_from_c" from Argument<T> to Argument<ArmIntrinsicType> 2. removed "from_c" in Argument<T>

clarity

from_param of X86IntrinsicType

vector register as an array of fundamental types

Notes: 1. rust_type deinition may need future work for accuracy

folkertdev · 2025-07-10T10:50:50Z

Hi, cool project!

I'm currently investigating how we can speed up the arm/aarch64 version of intrinsic-test. The CI job with it currently takes 28 minutes, and that is just not acceptible; it slows everything/everyone down.

On the other hand, I don't want to create horrible merge conflicts for you. The two main changes that I want to make are

combine tests into larger files (one per core), this doesn't really change anything relevant to you
write into a &mut impl std::io::Write instead of concatenating strings

The latter change mostly just helps with making the former possible. In the experimental #1856 I've now gotten CI time down to 8 minutes.

So, to prevent merge conflicts (especially now that we have additional reviewing capacity) I'd encourage you to split out and merge changes early if they mostly touch the existing code, instead of accumulating them in one big PR. Some changes that look like they can be split out are:

removing some of the Box wrapping (e.g. some of the changes in crates/intrinsic-test/src/arm/types.rs)
the pub enum Sign and its use in TypeKind::Int
the cleanup in crates/intrinsic-test/src/common/gen_c.rs

There might be others that I missed.

Feel free to r? @folkertdev me on any such cleanup PRs.

madhav-madhusoodanan · 2025-07-10T17:35:51Z

Hi @folkertdev, thank you so much for the suggestion and the concern about merge conflicts (and about looking into how we can optimize compile times!)
I was checking out #1856 and I agree with you on the need for the same.

I shall create small PRs with the changes that needs to be made on the ARM and the common modules, once I reconcile the changes with this PR itself.

madhav-madhusoodanan force-pushed the x86_extension_intrinsic_test branch 3 times, most recently from b2777ac to 7047369 Compare June 8, 2025 17:19

madhav-madhusoodanan commented Jun 11, 2025

View reviewed changes

madhav-madhusoodanan force-pushed the x86_extension_intrinsic_test branch 5 times, most recently from ed8dfde to e6ff768 Compare June 16, 2025 08:28

Amanieu reviewed Jun 18, 2025

View reviewed changes

madhav-madhusoodanan force-pushed the x86_extension_intrinsic_test branch 2 times, most recently from e113a61 to bc5da7d Compare June 19, 2025 06:52

madhav-madhusoodanan force-pushed the x86_extension_intrinsic_test branch from bc5da7d to c5dec86 Compare June 19, 2025 11:24

madhav-madhusoodanan added 10 commits June 20, 2025 10:56

fix: updated function definition of "from_c" in IntrinsicTypeDefinition

4a3d05f

trait to make the same dyn-compatible.

Feat: started the skeleton for x86 module, added XML intrinsic parsin…

38fd826

…g capabilities

feat: added functionality to convert XML description of intrinsics to

a786643

intrinsic-test's X86IntrinsicType struct

feat: added the simple set of argument types for X86 intrinsics

7b8e08c

feat: added X86IntrinsicType parsing from string.

e95bd9b

fix: removing Box<> types from IntrinsicType in "from_c" definition for

241f309

IntrinsicType impl

feat: implemented c_type for X86IntrinsicType

c77800a

Add x86/config.rs to intrinsic-test

16bd12c

madhav-madhusoodanan force-pushed the x86_extension_intrinsic_test branch 2 times, most recently from 9c953a7 to c6491da Compare June 20, 2025 05:38

madhav-madhusoodanan added 2 commits June 20, 2025 11:56

Fix: unused variables.

e827a04

feat: fetching c_type representation from IntrinsicType's hashmap dir…

2784b45

…ectly instead of constructing it. Further details: 1. moved "type_and_name_from_c" from Argument<T> to Argument<ArmIntrinsicType> 2. removed "from_c" in Argument<T>

feat: changed from TypeKind::Int(bool) to TypeKind::Int(Sign) for more

d0f2d78

clarity

madhav-madhusoodanan force-pushed the x86_extension_intrinsic_test branch from c6491da to d0f2d78 Compare June 20, 2025 06:31

madhav-madhusoodanan added 4 commits June 24, 2025 11:29

Feat: setup load function for x86 intrinsics

758d7ec

feat: implemented c_single_vector_type and fixed logical errors in

6e190b2

from_param of X86IntrinsicType

feat: added vector types to support intrinsics that does not represent a

e947072

vector register as an array of fundamental types

feat: added memsize and rust_type implementation

e57c4bf

Notes: 1. rust_type deinition may need future work for accuracy

adamgemmell mentioned this pull request Jul 8, 2025

Speeding up intrinsic-test #1851

Open

feat: added initial functionality for building Rust files

5b89089

madhav-madhusoodanan force-pushed the x86_extension_intrinsic_test branch from 5817342 to 5b89089 Compare July 10, 2025 06:53

Added output comparison functionality

2624722

feat: moved to_range to to_vector for extended iteration capabilities

9cd1cb0

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Uh oh!

Adding the x86 part of behavioural testing for std::arch intrinsics #1814

Adding the x86 part of behavioural testing for std::arch intrinsics #1814

madhav-madhusoodanan commented May 30, 2025

Uh oh!

madhav-madhusoodanan Jun 11, 2025

Uh oh!

Amanieu Jun 12, 2025

Uh oh!

madhav-madhusoodanan Jun 12, 2025

Uh oh!

madhav-madhusoodanan commented Jun 11, 2025

Uh oh!

Amanieu Jun 18, 2025

Uh oh!

Amanieu Jun 18, 2025

Uh oh!

madhav-madhusoodanan commented Jun 19, 2025

Uh oh!

Amanieu commented Jun 19, 2025

Uh oh!

folkertdev commented Jul 10, 2025

Uh oh!

madhav-madhusoodanan commented Jul 10, 2025 •

edited

Loading

Uh oh!

Uh oh!

Adding the x86 part of behavioural testing for std::arch intrinsics #1814

Are you sure you want to change the base?

Adding the x86 part of behavioural testing for std::arch intrinsics #1814

Conversation

madhav-madhusoodanan commented May 30, 2025

Uh oh!

madhav-madhusoodanan Jun 11, 2025

Choose a reason for hiding this comment

Uh oh!

Amanieu Jun 12, 2025

Choose a reason for hiding this comment

Uh oh!

madhav-madhusoodanan Jun 12, 2025

Choose a reason for hiding this comment

Uh oh!

madhav-madhusoodanan commented Jun 11, 2025

Uh oh!

Amanieu Jun 18, 2025

Choose a reason for hiding this comment

Uh oh!

Amanieu Jun 18, 2025

Choose a reason for hiding this comment

Uh oh!

madhav-madhusoodanan commented Jun 19, 2025

Uh oh!

Amanieu commented Jun 19, 2025

Uh oh!

folkertdev commented Jul 10, 2025

Uh oh!

madhav-madhusoodanan commented Jul 10, 2025 • edited Loading Uh oh! There was an error while loading. Please reload this page.

Uh oh!

Uh oh!

Uh oh!

madhav-madhusoodanan commented Jul 10, 2025 •

edited

Loading