|
---|
| | Christian Z
Posts 14 13 Feb 2021 11:01
| Excuse me for asking a probably dumb question. What is the precision of the FPU currently (i.e., with Build 7905 RELEASE5) available on the V4SA? I know from https://wiki.apollo-accelerators.com/doku.php/apollo_core:fpu that you omitted support for Motorola's proprietary 80-bit extended precision, which is completely understandable. But does the FPU on the V4SA make use of all *64* bits of double precision floating point numbers?
| |
| | Gerardo G.
Posts 54 14 Feb 2021 03:23
| I think it was 64 bits, at least months ago. That information may already be out of date with latest updates.
| |
| | Gunnar von Boehn (Apollo Team Member) Posts 6254 14 Feb 2021 06:15
| Christian Z wrote:
| Excuse me for asking a probably dumb question. What is the precision of the FPU currently (i.e., with Build 7905 RELEASE5) available on the V4SA?
|
R5 release supports 56 Bit FPU precision. The upcoming R6 release support 64 bit.
| |
| | Rollef 2000
Posts 29 14 Feb 2021 11:20
| Christian Z wrote:
| Motorola's proprietary 80-bit extended precision
|
no, it is part of IEEE754
| |
| | Christian Z
Posts 14 14 Feb 2021 12:46
| Gunnar von Boehn wrote:
| R5 release supports 56 Bit FPU precision. The upcoming R6 release support 64 bit.
|
Thank you for clarifying this! I was trying some numerical benchmarks that check themselves (to see if they're really being run in double precision) and I was already suspecting a bug in the compiler. Looking forward to R6, then.
| |
|