SI
SI
discoversearch

We've detected that you're using an ad content blocking browser plug-in or feature. Ads provide a critical source of revenue to the continued operation of Silicon Investor.  We ask that you disable ad blocking while on Silicon Investor in the best interests of our community.  If you are not using an ad blocker but are still receiving this message, make sure your browser's tracking protection is set to the 'standard' level.
Technology Stocks : Advanced Micro Devices - Moderated (AMD) -- Ignore unavailable to you. Want to Upgrade?


To: ptanner who wrote (78723)4/29/2002 11:44:22 AM
From: pgerassiRead Replies (1) | Respond to of 275872
 
Dear PT:

With VLIW you must realize the design goals and the reasons for them. VLIW thinks that make the hardware simple, no out of order execution, direct pipeline control and many others, and make the compiler figure the best way to use the hardware. Thus making the hardware perform is easy since many of the high complexities that cause much of the control logic is missing. However it requires the compilers to be very smart and know the architecture of the executing CPU very well.

So far Intel may have succeeded with the hardware (IMHO they screwed this up as well as it underperforms given the simplicity required for VLIW to work), but they failed with the compiler. Only very restricted domains does the compiler makes any real optimizations. On most code, it fails miserably. Since most of the performance goals would be found only in the compiler, this makes the performance much less than that claimed during the on paper hype fests.

Only hand coded assembly does the underlying performance start to show. However this is a very high cost slow ponderous method of software development. It makes sense only for embedded and single purpose applications where the development costs are small relative to the compute time and costs. To the original GP computing market, it fails miserably and this shows by the practically non existent consumer demand.

Pete