1 |
On Sat, 18 May 2019 20:47:28 +0200 |
2 |
Michał Górny <mgorny@g.o> wrote: |
3 |
|
4 |
> On Fri, 2019-05-03 at 23:34 +0200, Andreas K. Huettel wrote: |
5 |
> > * We will initially add two profiles to profile.desc: |
6 |
> > default/linux/riscv/17.0/rv64gc/lp64d (non-multilib, 64bit |
7 |
> > hardfloat) default/linux/riscv/17.0/rv64gc (multilib lp64d/lp64, |
8 |
> > i.e. hard/softfloat) |
9 |
> |
10 |
> I still don't understand the purpose of this multilib. If you have |
11 |
> a hardfloat CPU, why would you ever build some of the software |
12 |
> softfloat? |
13 |
|
14 |
|
15 |
One reason I could imagine is that the hardfloat isn't IEEE 754 |
16 |
compliant. Searching through the RISC-V spec, it does not seem to be |
17 |
the case here (ie: it is required to be compliant) so I'm also |
18 |
wondering what is the point here. |
19 |
|
20 |
|
21 |
Note that hard vs soft float will be the least of our concerns |
22 |
here: Based on wikipedia, I count 4 base ISA (2 stable, 2 dev) and 13 |
23 |
optional extensions (6 stable, 7 dev) so we potentially have an |
24 |
exponential number of ABIs here... |