[libre-riscv-dev] fp special functions

Hendrik Boom hendrik at topoi.pooq.com
Mon Aug 5 05:15:24 BST 2019

On Mon, Aug 05, 2019 at 04:49:13AM +0100, Luke Kenneth Casson Leighton wrote:
> am a fan of just saying "IEEE754" here.  exact results.  makes testing
> easier, yes.  my feeling is: if people *really want* inaccuracy, it
> should be as a sub-extension.

Amdahl's 360 clones did their operations lat-bit accurately.  IBM's 
original didn't do them all correctly.  Amdahl was forced by its 
customers to install a switch that would force the machine to produce 
the same wrong answer as IBM's.

-- hendrik

More information about the libre-riscv-dev mailing list