[libre-riscv-dev] [isa-dev] Re: FP transcendentals (trigonometry, root/exp/log) proposal
luke.leighton at gmail.com
Mon Aug 12 04:20:28 BST 2019
(thank you to michael for the assistance, above).
as we had people say that it was "not obvious" what or why equivalence with
Khronos OpenCL opcodes is being proposed, we have created a special section
with an "equivalence" table between the proposed opcodes and the Khronos
this table makes it abundantly clear how obvious the need for these opcodes
is, how much of a no-brainer decision it is, and why "quantitative
analysis" - *for these opcodes and these opcodes alone* - is completely
other opcodes (SLERP, LERP, CORDIC, Vector Normalisation) are *NOT*
obvious, and *WILL* require further research and careful analysis.
More information about the libre-riscv-dev