P3287 R3 Exploration of namespaces for std::simd

[wg21bot]

P3287R0 Exploration of namespaces for std::simd (Matthias Kretz)

[inbal2l]

2024-06-26 Library Evolution St. Louis Meeting

P3287R0: Exploration of namespaces for std::simd

2024-06-26 Library Evolution St. Louis Meeting Minutes

Champion: Matthias Kretz
Chair: Robert/Ben
Minute Taker: Eddie

Summary

POLL: We should introduce a std::simd namespace in P1928 (realizing that this would mean we potentially have a type whose fully-qualified name is std::simd::simd and would involve removing “simd” from basic_simd_mask) and remove the “simd_” prefix from utilities moved into that namespace which currently have that prefix.

SF	F	N	A	SA
9	6	3	1	1

Attendance: 20 IP + 9 R
# of Authors: 2
Author's Position: 2xSF
Outcome: Consensus in favour

POLL: Operations on std::simd::simd should be available in the namespace std::simd.

SF	F	N	A	SA
12	7	1	1	0

Attendance: 20 IP + 9 R
# of Authors: 2
Author's Position: 2xSF
Outcome: Consensus in favour

POLL: Element-wise overloads of existing functions in std which operate on std::simd::simd should also be available in the namespace std.

SF	F	N	A	SA
4	7	5	1	1

Attendance: 20 IP + 9 R
# of Authors: 2
Author's Position: 1xSF, 1xF
Outcome: Consensus in favour

Next Steps

The authors got feedback for relevant design questions and will apply it to the new revision, LEWG will see the next revision of the paper.

[wg21bot]

P3287R1 Exploration of namespaces for std::simd (Matthias Kretz)

[mattkretz]

P3287R2 Exploration of namespaces for std::simd (Matthias Kretz)

[cor3ntin]

Should we close this?

[jensmaurer]

Let's leave that to the LEWG chair.

[cor3ntin]

@inbal2l @mattkretz

[mattkretz]

Should we close this?

Why?

[inbal2l]

This needs to stay open for now.
Only part of the design changes proposed here were discussed during St. Louis, we will see the rest in Hagenberg (tentatively - Monday). Thanks, all.

(Comment updated to preserve order)

2025-02-10 Library Evolution Hagenberg Meeting

P3287R2: Exploration of namespaces for std::simd

2025-02-10 Library Evolution Hagenberg Minutes

Champion: Matthias Kretz
Chair: Fabio Fracassi / Inbal Levi
Minute Taker: Eddie

Summary

POLL: We prefer the following namespace name (multiple votes are allowed)

Name	Votes
std::dp	6
std::datapar	20
std::data_parallel	20

Attendance: 30 (IP) + 4(R)
Outcome: No consensus between std::datapar and std::data_parallel, taking the next poll between two remaining options

POLL: std::datapar vs std::data_parallel

Name	Votes
std::datapar	13
Nutral	4
std::data_parallel	13

Attendance: 30 (IP) + 4(R)
Outcome: No consensus

POLL: In P3287R2 rename the std::simd namespace to std::data_parallel or std::datapar (author’s choice).

SF	F	N	A	SA
11	8	3	8	1

Attendance: 30 (IP) + 4 (R)
Author's Position: SF
Outcome: Weak consensus in favour

SA: The alternative names are too broad, not self-explanatory, and confusing.

POLL: In P3287R2 rename the datapar::[basic_]simd_mask to datapar::[basic_]mask.

SF	F	N	A	SA
2	7	7	8	1

Attendance: 30 (IP) + 4 (R)
Author's Position: SF
Outcome: No consensus for change

POLL: With the changed namespace, forward P3287R2 to LWG for C++26.

SF	F	N	A	SA
11	12	0	0	0

Attendance: 30 (IP) + 4 (R)
Author's Position: SF
Outcome: Unanimous consensus in favour

Next Steps

The paper (with changed namespace) was forwarded to LWG for C++26.

[bstamour]

LWG polled this in Hagenberg Thursday morning:

Poll: put P3287r3 (as reviewed by TK and JG) into C++26?
|F|A|N|
|7|0|0|

[wg21bot]

P3287R3 Exploration of namespaces for std::simd (Matthias Kretz)