On 2 March 2015 at 18:14, Keean Schupke <[email protected]> wrote:

> I am not sure how effective it will be. VLIW doesn't work well (including
> IA64 for general purpose computation) in my opinion because it
> underestimated the benefit of small instruction size and runtime
> optimisation in the OoOE unit. X86 code allows the first 8 registers
> (sufficient for most function calls) to be accessed from compact opcodes.
>
​If anyone ever wrote a book on common problems in VLIW implementations, it
could be as long as all three volumes of the Itanium Programmers Manual put
together.  I'm not sure if the impact of instruction size was measurable
(in time) in practice, and the typical L2 cache sizes on the Itanium made
up for that.  The impact of dynamic feedback is important however, which is
why common dynamic language runtimes beat out many straightforward
compilers today, so runtime instruction scheduling has huge benefits.

Oh - and really, the sort of architectural decisions you could make to
better support functional and dynamic languages and fast IPC are mostly
orthogonal to support for ILP, and I care about those more.

-- 
William Leslie

Notice:
Likely much of this email is, by the nature of copyright, covered under
copyright law.  You absolutely MAY reproduce any part of it in accordance
with the copyright law of the nation you are reading this in.  Any attempt
to DENY YOU THOSE RIGHTS would be illegal without prior contractual
agreement.
_______________________________________________
bitc-dev mailing list
[email protected]
http://www.coyotos.org/mailman/listinfo/bitc-dev

Reply via email to