[Cfp-interest 2397] Re: WG14 IEEE 754-C binding meeting minutes 2022/03/16
Steve (Swift) Canon
scanon at apple.com
Thu Mar 17 08:48:18 PDT 2022
> On Mar 17, 2022, at 9:57 AM, Vincent Lefevre <vincent at vinc17.net> wrote:
>
> On 2022-03-16 16:55:49 +0000, Rajan Bhakta wrote:
>> Carry-over action items:
>> Fred: Ask C++ what their issues with *HAS_SUBFORM are and if they are OK with obsoleting it. - Still open
>
> I suppose that is should be *HAS_SUBNORM (in case this can be
> corrected for the future version on open-std.org).
>
>> fma subclause title [Cfp-interest 2388]
>> David H: We called it fused since there were implementations that had fused and unfused multiply-add's.
>> Jim: In C it is fused in the sense of only one rounding. The title says "Floating" though and Pavel wants it changed to "Fused".
>> Fred: The ARM hardware does double rounding.
>> Rajan: So they are already non-conforming?
>> Fred: Yes.
>
> This does not seem to be true. In the Assembler User Guide:
>
> https://www.keil.com/support/man/docs/armasm/armasm_pge1415706532461.htm
>
> 8.3 Fused Multiply-Add extension for VFP
>
> The Fused Multiply-Add extension optionally extends the VFPv3
> architecture.
>
> It provides VFP instructions that perform multiply and accumulate
> operations with a single rounding step, so suffers from less loss
> of accuracy than performing a multiplication followed by an add.
>
> The fused multiply-add instructions are only available on VFP
> systems that implement the fused multiply-add extension. The VFP
> system that implements the fused multiply-add extension is VFPv4.
Correct. Also note that:
- VFPv4 was released in 2011
- 64b ARM ISAs _only_ have the fused operation, so the unfused operation will disappear when the 32b ISA is end-of-lifed (2023, IIRC?)
– Steve
More information about the Cfp-interest
mailing list