View Single Post
Old 07 June 2017, 02:52   #105
matthey
Banned
 
Join Date: Jan 2010
Location: Kansas
Posts: 1,284
Quote:
Originally Posted by Akira View Post
And let's not get this one closed too, i saw the shitshow that one was and would not like to see it again.
It was mainly 2 users in that thread who were the problems and acted like they wanted the thread closed so there would be no further discussions. I want discussions and consensus. I wanted A-Eon to work with Gunnar as I thought it would force him to give up some control and listen to reason. I hope there is a tech savvy group or board from the newly formed (A-Eon/Hyperion/iComp/Apollo Team) alliance making the decisions. At least Jens Schönfeld should have some tech and marketing knowledge even if he supported MIPS over the 68k for the Clone-A at one point.

Gunnar's ISA is all wrong. He is obsessed with adding registers even when consensus has said no. The way to overcome consensus is to take control (become the authoritarian) which is what he did. Adding registers is good for performance if they can be used but the integer registers he added and are shared with the SIMD are not orthogonal making them difficult to use especially for compilers, use more energy and use more memory (larger stacks are needed). His focus on high performance is like he is trying to compete with high end desktop processors in some ways but then he only has an old MMX SIMD unit and no virtual MMU. It looks like he is opposed to a compatible 68k FPU because he needs a new vector FPU since his SIMD unit can't have floating point or it would have floating point in the integer register file slowing down a critical path. He is opposed to modes even though a separate 68k_32 and 68k_64 mode may be the best option if/when 64 bit integer unit support is needed (I'm not a big fan of modes either but one for compatibility and code density and one for 64 bit support isn't too bad for a high end CPU). His ISA limits upgrade paths and compatibility. Gunnar may be talented at CPU design but he is terrible at ISA design. Am I the only one who sees this? What are the chances of getting a real development and standardization team?
matthey is offline  
 
Page generated in 0.06539 seconds with 9 queries