>The Alpha 21264 is much faster than an Intel 400 MHz if you compare
>non-optimized C code (FORTRAN like code with big arrays of global data). But
>if you compare a well-written C code on the Alpha (who write in assembler on
>this chip ?)
Who should write in assembler on any other chip ? Todays informatics goes that
you 'describe' your wishes by a high level language and let complier optimize
the code. With that you separate your ideas from that how a machine should
execute it at best. And I must say that I realy like to program on a RISC
machine, does`t matter which one. For education and for the speed it`s more
advanced than any other assembler. I especialy mean the SPARC assembler. Such
instucion code (R1 x R2 -> R3) is very good for parallelization and out-of-order
execution. I don`t know how the x86 assembler is structured (know few), but I
see very few hope for a headroom of further speed up by brainiac methods. Just
an example:
c:=a+b
RISC: add a,b,c
CISC: mov a,c (1)
add b,c (2)
When (2) depends on (1), there is no possibility for parallelization. Good, you
can argue that instructions on 3 operators are better than the one on 2
instructions.
>with a code in assembler on the PII/400, the Alpha is faster
>(not much). If you take into account the number of PII that you can buy for
>the price of one Alpha 21264, a set of PII is faster if the computation can
>be distributed.
I heared that AMD took over the bus protocol (EV6) of the Alpha processor. So
coming AMD processors will have the same pin order like Alphas. With that you
can use an motherboard for both processors, exept you have to change the BIOS.
And unified motherboards are cheaper to produce than two special one. So there
is a hope for cheaper Alphas ...
Bojan
