[libre-riscv-dev] [isa-dev] Re: FP transcendentals (trigonometry, root/exp/log) proposal
luke.leighton at gmail.com
Thu Aug 8 13:44:49 BST 2019
On Thursday, August 8, 2019 at 6:09:28 PM UTC+8, Jacob Lifshay wrote:
> maybe a solution would be to add an extra field to the fp control csr (or isamux?) to allow selecting one of several accurate or fast modes:
No definitely not ISAMUX/ISANS, its purpose is for switching (paging in) actual opcodes. Not quite true, LE/BE kinda flips in LE variants of LD/ST.
An FP CSR (dedicated or fields) makes more sense I think because it's quite a few bits, and I can see some potential value in the same bits being applied to F, G, H and Q as well.
More information about the libre-riscv-dev