64-bit computing and x86-64

even though x86 ISA had lots of drawbacks but software based on this ISA were already huge and transition to RISC ISA costs more.

 

Intel's IA-64 and AMD's x86-64

IA-64 (Itanic) == Explicitly Parallel Instruction Computing (EPIC) + very long instruction word(VLIW) ISA + predication , speculative loading , other performance enhnacing technique

but it had compatibilty issue with x86 and left big hole on 64-bit sized server markets and commodity workstation

AMD worked hard at set of 64-bit extension tothe x86 ISA == x86-64 , midrange x86 processors and baackward-compatible with previous x86 codes. (Hammer)

 

What is 64 bit computing

nubmer of bits that each of the processor's general purpose register(GPR) can hold. same of 64 bit instruction, operates on 64 bit nubmers that are stored in 64-bit GPRs

nothing special jsut register sized and PC got doubled. notice IR are still same.

 

Current 64-bit Applications 

1. dynamic range == range of integers that a processor can perform increased by a factor of 4.3billion.

2. overflow,underflow == use 64bit integer consturct rpovided by a high level language like C. compiler uses 2 registers per integer to do 64 bit calcuations in 32bit hardware.

3. virtual address == x86-64 can supprot 1 terabye of physical RAM 

 

The 64-bit alternative: x86 -64

more register is always better. modern x86 cpu did register renmaing to overcome the number of reigster limitation. 

notice they have left x87 floating point stack alone

because SSE is much prefer. because x87 uses 80 bits internally it can cause optimizing compiler very hard to predict the result. so it is ahrd to verify that you are getting the correct answer when doing regression tests.

notice PC got larger too. 

it is essential to AMD's plans for x86-64 that there is no perfomance penalty for running legacy