Do you want ":=" operator?

ruby-talk
1

It seems that I need more words to explain my CPU model. I'll show why
it is *impossible* to impletement *register assignment* by "=".
A HW register updates its output to its input value only when an event
occurs (typically a clock rising/falling edge). My CPU can process
serveral instrucions at the same time. For example, "R0=R0&R1; R2=R0"
means that, at next clock period, R0 will get the AND result and R2 the
current R0 value (not same as AND result). To support this feature,
registers must simulate the true HW properties like this:

class Reg
attr_accessor :value, :new_value, :data_enb

def initialize()
@value=@new_value=0
@data_enb=false
end

def assign(data)
@new_value = data
@data_enb=true
end

def update
@value, @data_enb=@new_value, false if @data_enb
end

end

"R0=R0&R1; R2=R0" will actually does "R0.assign(R0&R1);
R2.assign(R0.value)" (suppose Reg.& defined), and at later time all the
regs' update() will be called to simulate the clock event.
In short, Rx does not just represent a value, but a real register
*object*, so the variable Rx should always reference to the same
object. That's why "=" cannot be used despite of any forms.
What I want is an *object modification operator* ":=" provided by
syntax to save the translation effort, which is not so easy to do in
complex cases. What a nice thing if my CPU program "R0:=R0&R1; R2:=R0"
is just ruby program!

2

In article <cd7qpj$7cc@odbk17.prod.google.com>,

It seems that I need more words to explain my CPU model. I'll show why
it is *impossible* to impletement *register assignment* by "=".

Quite true, you would lose all your 'future' (as in events scheduled
to happen in the future). I mistakenly used '=' in my example whereas I
should have used the hypothetical ':='

A HW register updates its output to its input value only when an event
occurs (typically a clock rising/falling edge). My CPU can process
serveral instrucions at the same time. For example, "R0=R0&R1; R2=R0"
means that, at next clock period, R0 will get the AND result and R2 the
current R0 value (not same as AND result). To support this feature,
registers must simulate the true HW properties like this:

class Reg
attr_accessor :value, :new_value, :data_enb

def initialize()
@value=@new_value=0
@data_enb=false
end

def assign(data)
@new_value = data
@data_enb=true
end

def update
@value, @data_enb=@new_value, false if @data_enb
end

end

"R0=R0&R1; R2=R0" will actually does "R0.assign(R0&R1);
R2.assign(R0.value)" (suppose Reg.& defined), and at later time all the
regs' update() will be called to simulate the clock event.
In short, Rx does not just represent a value, but a real register
*object*, so the variable Rx should always reference to the same
object. That's why "=" cannot be used despite of any forms.
What I want is an *object modification operator* ":=" provided by
syntax to save the translation effort, which is not so easy to do in
complex cases. What a nice thing if my CPU program "R0:=R0&R1; R2:=R0"
is just ruby program!

Indeed. I understand exactly, but I suspect there are about 3 of us on
this list who have done hardware design (or used HDLs). I think perhaps
we're having a hard time conveying what we're trying to do and how a ':='
operator could really help. Maybe we need to come up with a more general use
for a ':=' operator that everyone can relate to. I'm sure some examples exist.

BTW: It's been mentioned, but you might find that RHDL would save you a
lot of work since it already takes care of event scheduling, process
launching based on sensitivity lists, concurrency (using continuations),
etc. The lack of a ':=' operator aside, Ruby's mix of features is really
great for hardware modelling. It would have been very much more difficult
to implement something like RHDL in Perl or Python, for example (no
continuations or code blocks).

Phil

···

Xiangyu Yang <xiangyu.yang@gmail.com> wrote:

3

Hi,

At Fri, 16 Jul 2004 15:02:19 +0900,
Xiangyu Yang wrote in [ruby-talk:106622]:

It seems that I need more words to explain my CPU model. I'll show why
it is *impossible* to impletement *register assignment* by "=".

Isn't this acceptable?

class CPU

class Reg
attr_accessor :value, :new_value, :data_enb

def initialize()
@value=@new_value=0
@data_enb=false
end

def assign(data)
@new_value = data
@data_enb=true
end

def update
@value, @data_enb=@new_value, false if @data_enb
end

end

  def initialize(regnum = 32)
    @regset = Array.new(regnum) {Reg.new}
  end
  def (idx)
    (0...@regset.size) === idx or raise RangeError
    @regset[idx]
  end
  def =(idx, data)
    (0...@regset.size) === idx or raise RangeError
    @regset[idx].assign(data)
  end
end

R = CPU.new

···

--
Nobu Nakada