New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
P1068 R11 Vector API for random number generation #409
Comments
Option to generate the same distribution sequence as sequential is important (for consistency) Option to generate a distribution sequence different from sequential is important (for performance) We must have an iterator based 'generate' interface We must have a simd-type-producing interface |
P1068R2 Vector API for random number generation (Ilya Burylov, Pavel Dyakov, Ruslan Arutyunyan, Andrey Nikolaev) |
We like the new algorithms for vectorized random number distributions, as a new Forward to LEWG for target at LEWG's choosing: |
SG6 does not favor change; the original API doesn't promise repeatability, so saying the vector API is consistent with the existing API doesn't really mean anything. |
P1068R3 Vector API for random number generation (Ilya Burylov, Pavel Dyakov, Ruslan Arutyunyan, Andrey Nikolaev) |
Putting this on the LEWGI agenda as we're already seeing some random number papers and LEWG likely won't have time at this meeting for this. |
P1068R3 Make Random Number Generation Vectorizable: Design Review Chair: Nevin Liber Champion: Ilya Burylov Minute Taker: Marco Foco Start Review: 2020-02-11 14:50 Prior art: Yes, in paper. This proposal changes the definition of the Regarding whether this provides enough implementation freedom, that's an SG1 or SG6 question. SG1 reviewed this and this is the amount of freedom they felt they needed. Do we need 3 APIs:
Clarify the design intent for Start Polling: 15:26 POLL: Ask SG1 to consider the capability of a vector URBG that is unconstrained in reproducing the scalar results (the goal being as much performance as possible).
Attendance: 22 # of Authors: 1 Author Position: SA That has no consensus. End: 15:34 CONSENSUS: LEWGI sends P1068R3 (Make Random Number Generation Vectorizable), with the guidance below, to LEWG.
|
P1068R4 Vector API for random number generation (Ilya Burylov, Pavel Dyakov, Ruslan Arutyunyan, Andrey Nikolaev) |
P1068R8 Vector API for random number generation (Ilya Burylov, Pavel Dyakov, Ruslan Arutyunyan, Andrey Nikolaev, Alina Elizarova) |
2023-11-28 Library Evolution TeleconP1068R9: Vector API for random number generation 2023-11-28 Library Evolution Telecon Minutes Champion: Ruslan Arutyunyan Chair: Ben Craig Minute Taker: Mark Hoemmen SummaryLWG wants changing the POLL: Change
Attendance: 14 # of Authors: 2 Author Position: N, WF Outcome: Consensus in favor WA: because it prevents a third party from optimizing a generator for vectorization Next StepsLEWG approved the recommended change. The paper will go back to LWG to continue the wording review. |
P1068R10 Vector API for random number generation (Ilya Burylov, Pavel Dyakov, Ruslan Arutyunyan, Andrey Nikolaev, Alina Elizarova) |
P1068R11 Vector API for random number generation (Ilya Burylov, Pavel Dyakov, Ruslan Arutyunyan, Andrey Nikolaev, Alina Elizarova) |
P1068R1 Vector API for random number generation (Ilya Burylov, Pavel Dyakov, Ruslan Arutyunyan, Andrey Nikolaev)
The text was updated successfully, but these errors were encountered: