[CK] Refactor container array and function sequence_reverse_inclusive_scan #3612

CongMa13 · 2026-01-20T00:13:30Z

Refactor cd sequence_reverse_inclusive_scan from recursive template to constexpr for-loop.

This pull request improves the utility array functionality and its integration with sequence utilities, as well as adds comprehensive unit tests for the Array class. The main changes include refactoring the implementation of reverse inclusive scan for sequences, updating header dependencies, optimizing the Array assignment operator, and introducing a full test suite for Array.

Array and Sequence Utility Improvements:

Refactored the implementation of sequence_reverse_inclusive_scan in sequence.hpp to use a more efficient, array-based approach, encapsulated in a new impl namespace. This change simplifies the logic and should improve compile-time performance.
Updated header dependencies in both array.hpp and sequence.hpp to remove unused includes and ensure correct dependency order, notably including array.hpp in sequence.hpp to support the new scan implementation. [1] [2]
Optimized the assignment operator in the Array struct by replacing the static_for loop with a standard for-loop and adding a pragma for potential unrolling, improving both readability and performance.

Testing Enhancements:

Added a new unit test file unit_array.cpp with comprehensive tests covering construction, element access, assignment, iterators, the make_array helper, and different data types for the Array class.
Updated the CMakeLists.txt to build and link the new unit_array test executable.

Tracking issue: #3575

…_scan

CongMa13 · 2026-01-20T00:15:29Z

include/ck/utility/array.hpp

+#pragma unroll
+        for(int i = 0; i < NSize; i++)
+        {
+            mData[i] = a[i];
+        }


Removed the dependency on static_for. Hence, only need to include type.hpp.

And the code is easier to read.

Copilot

Pull request overview

This PR refactors the sequence_reverse_inclusive_scan implementation from a recursive template-based approach to a more efficient constexpr for-loop using the Array utility class. The changes improve compile-time performance and simplify the logic.

Changes:

Refactored sequence_reverse_inclusive_scan to use an array-based iterative approach instead of recursive templates
Updated header dependencies to remove circular includes and optimize dependencies
Optimized Array::operator= by replacing static_for with a standard for-loop with unroll pragma
Added comprehensive unit tests for the Array class

Reviewed changes

Copilot reviewed 4 out of 4 changed files in this pull request and generated 2 comments.

File	Description
include/ck/utility/sequence.hpp	Refactored `sequence_reverse_inclusive_scan` from recursive templates to constexpr array-based implementation
include/ck/utility/array.hpp	Updated header dependencies and optimized assignment operator with standard for-loop
test/util/unit_array.cpp	Added comprehensive unit tests covering Array construction, access, assignment, iterators, and edge cases
test/util/CMakeLists.txt	Added build configuration for new unit_array test executable

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

Copilot · 2026-01-20T22:09:55Z

test/util/unit_array.cpp

+
+using namespace ck;
+
+// Test basic Sequence construction and properties


Comment mentions 'Sequence' but should refer to 'Array' since this test is for the Array class.

Suggested change

// Test basic Sequence construction and properties

// Test basic Array construction and properties

This comment should be updated.

Copilot · 2026-01-20T22:09:56Z

include/ck/utility/array.hpp


-        static_for<0, Size(), 1>{}([&](auto i) { operator()(i) = a[i]; });
+#pragma unroll
+        for(int i = 0; i < NSize; i++)


The loop variable uses int type while NSize is of type index_t (typically size_t or similar). For consistency and to avoid potential type conversion issues, use index_t i = 0 instead of int i = 0.

Suggested change

for(int i = 0; i < NSize; i++)

for(index_t i = 0; i < NSize; i++)

Yes, let's use index_t for consistency.

shumway · 2026-01-21T21:09:19Z

include/ck/utility/array.hpp


-        static_for<0, Size(), 1>{}([&](auto i) { operator()(i) = a[i]; });
+#pragma unroll
+        for(int i = 0; i < NSize; i++)


We don't need pragma unroll here. The compiler should optimize correctly.

shumway · 2026-01-21T21:35:59Z

include/ck/utility/sequence.hpp


-    using type = typename sequence_merge<Sequence<new_reduce>, old_scan>::type;
-};
+        if constexpr(size == 0)


Let's hard code for a few of the shorter sequences:

if constexpr(size == 0)
{
return Sequence<>{};
}
else if constexpr(size == 1)
{
constexpr index_t values[1] = {Is...};
return Sequence<Reduce{}(values[0], Init)>{};
}
else if constexpr(size == 2)
{
constexpr index_t values[2] = {Is...};
constexpr index_t r1 = Reduce{}(values[1], Init);
constexpr index_t r0 = Reduce{}(values[0], r1);
return Sequence<r0, r1>{};
}

shumway · 2026-01-21T21:46:38Z

include/ck/utility/sequence.hpp

+        }
+        else
+        {
+            constexpr Array<index_t, size> arr = []() {


It might be slightly faster to replace this lambda with a named constexpr function helper.

shumway · 2026-01-21T21:47:16Z

include/ck/utility/sequence.hpp

-template <index_t I, index_t... Is, typename Reduce, index_t Init>
-struct sequence_reverse_inclusive_scan<Sequence<I, Is...>, Reduce, Init>
+template <index_t... Is, typename Reduce, index_t Init>
+struct sequence_reverse_inclusive_scan_impl<Sequence<Is...>, Reduce, Init>


Do we have unit tests for the reverse inclusive scan?

yes, we have.

test/util/unit_sequence.cpp

ReverseInclusiveScan, ReverseExclusiveScan

shumway

This looks great. I just have a few small comments.

I think we may need unit tests for the reverse inclusive scan.

illsilin · 2026-01-23T20:51:05Z

Hey @CongMa13, please resolve conflicts for the headers, and make sure ci tests pass.

[CK] Refactor container array and function sequence_reverse_inclusive…

6579ce4

…_scan

CongMa13 requested review from Snektron, ThomasNing, afagaj, andriy-ca, aosewski, asleepzzz, bartekxk, carlushuang, cgmillette, coderfeli, geyyer, illsilin, poyenc, qianfengz, shumway, tenpercent, vidyasagar-amd and vpietila-amd as code owners January 20, 2026 00:13

CongMa13 commented Jan 20, 2026

View reviewed changes

afagaj requested a review from Copilot January 20, 2026 22:09

Copilot AI reviewed Jan 20, 2026

View reviewed changes

shumway reviewed Jan 21, 2026

View reviewed changes

shumway previously approved these changes Jan 21, 2026

View reviewed changes

[CK] refactoring according to the review feedback

bf61492

CongMa13 dismissed shumway’s stale review via bf61492 January 24, 2026 00:37


		using namespace ck;

		// Test basic Sequence construction and properties

	// Test basic Sequence construction and properties
	// Test basic Array construction and properties

	for(int i = 0; i < NSize; i++)
	for(index_t i = 0; i < NSize; i++)

[CK] Refactor container array and function sequence_reverse_inclusive_scan #3612

Are you sure you want to change the base?

[CK] Refactor container array and function sequence_reverse_inclusive_scan #3612

Conversation

CongMa13 commented Jan 20, 2026

Uh oh!

CongMa13 Jan 20, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI left a comment

Choose a reason for hiding this comment

Pull request overview

Reviewed changes

Uh oh!

Copilot AI Jan 20, 2026

Choose a reason for hiding this comment

Uh oh!

shumway Jan 21, 2026

Choose a reason for hiding this comment

Uh oh!

Copilot AI Jan 20, 2026

Choose a reason for hiding this comment

Uh oh!

shumway Jan 21, 2026

Choose a reason for hiding this comment

Uh oh!

shumway Jan 21, 2026

Choose a reason for hiding this comment

Uh oh!

shumway Jan 21, 2026

Choose a reason for hiding this comment

Uh oh!

shumway Jan 21, 2026

Choose a reason for hiding this comment

Uh oh!

shumway Jan 21, 2026

Choose a reason for hiding this comment

Uh oh!

CongMa13 Jan 24, 2026 • 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!

shumway left a comment

Choose a reason for hiding this comment

Uh oh!

illsilin commented Jan 23, 2026

Uh oh!

Reviewers

Assignees

Labels

Projects

Milestone

Development

Uh oh!

4 participants

CongMa13 Jan 24, 2026 •

edited

Loading