Possible API for vector operations: did you seen ForwardCompatible Arhitecture?

[Lev Serebryakov]

-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA512

On 16.09.2016 13:35, Vladimir Ivanov wrote:

> Overall, the idea you refer to is not new. Though x86 & ARM have 
> fixed-width vector extensions, variable-width vector extensions
> have been present for a long time in vector machines (e.g, long
> vectors in Cray).
 As far as I know, Cary needed code adaptation for different vector
lengths. Simple ones, yes.

> Nowadays, ARMv8-A SVE [2] features vector-length agnostic (VLA)
 I've seen press-release from ARM, but there are no specific details
yet :(

> programming model and RISC-V [3] chose variable width vectors as
> well.
 O! I didn't see RISC-V vector extensions!

> There's a nice comparison of different approaches [4] including
> SIMT & VT: "Exploring the Tradeoffs between Programmability and
> Efficiency in Data-Parallel Accelerators"
 Thank you, I'll look at it.

- -- 
// Black Lion AKA Lev Serebryakov
-----BEGIN PGP SIGNATURE-----
Version: GnuPG v2
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=4of4
-----END PGP SIGNATURE-----