The name was discussed in SG1 and LEWG. SG1 chose datapar over simd and other options.
LEWG overruled that choice and renamed it to simd.
We should use the TS to discover the following:
Is there interest in implementations that use hardware parallelism that does not map (exclusively) to SIMD instructions & registers?
Do users choose to not use simd, because it appears too narrowly focused (from its name) on (certain) CPUs? Would such users choose differently if the name would focus on the parallelism it allows instead of a specific hardware implementation of such parallelism?
The name was discussed in SG1 and LEWG. SG1 chose
dataparoversimdand other options.LEWG overruled that choice and renamed it to
simd.We should use the TS to discover the following:
simd, because it appears too narrowly focused (from its name) on (certain) CPUs? Would such users choose differently if the name would focus on the parallelism it allows instead of a specific hardware implementation of such parallelism?