This is the mail archive of the
gcc@gcc.gnu.org
mailing list for the GCC project.
Re: VLIW architectures
- To: gcc at gcc dot gnu dot org, badrian at ss dot pub dot ro
- Subject: Re: VLIW architectures
- From: Jason Eckhardt <jle at cygnus dot com>
- Date: Tue, 21 Mar 2000 07:53:56 -0800 (PST)
>
>
>>
>> There is little to no support in gcc explicitly for VLIW. In two
>> years time, there will be more.
>>
>
> Well, but do you have any suggestions for describing such an
>architecture by present means? For instance, do you think I could make use
>of define_function_unit and insn attributes to get parallelism from the
>instruction scheduler ?
>
One may attempt to use the current define_function_unit, etc. attributes to
describe a VLIW machine to the instruction scheduler. However, the current
mechanism is very imprecise with respect to modelling the machine. This
makes it difficult to schedule non-interlocked VLIWs since one may get
undefined results/wrong answers/processor hangs if resources (e.g. read/
write ports, functional units, etc.) are oversubscribed or latencies are
not honored.
One could write a target-specific function to pack operations, insert
NOPs, ensure resource and latency constraints are honored, etc. But
this isn't the best long-term solution.
In any case, support for precise scheduling (and hence VLIWs) is probably
not far off.
jason.