[erlang-questions] : Design of the Erlang VM

Serge Aleynikov <>
Thu Sep 6 12:39:06 CEST 2007


Could you shed some light on how non-tail recursive calls are being 
handled in the absence of use of the processor stack?

Serge

Raimo Niskanen wrote:
> No, they are not C function calls. C function calls push
> the arguments (that may come not only from the calling
> code, but also from other places, nevertheless there
> is explicit code to place them ---) on the processor stack,
> pushes the return addresss on the processor stack and then
> jumps to the function address. When the function returns
> the return instruction pops the return address from the
> stack and jumps back to where it came from. Then the
> calling code restores the stack pointer to remove the 
> call arguments from the processor stack.
> 
> BEAM instructions are the instruction word followed by
> zero or more operand words. The instruction word is
> an address to the processor instruction sequence that
> implements the BEAM instruction, as Bjorn said below.
> 
> The BEAM instructions are just straight processor code
> that do not expect arguments on the processor stack
> and do not end in a processor return instruction.
> 
> The BEAM instructions themselves increment the BEAM
> instruction pointer and pick up operand words into
> processor registers. They end by reading the next
> instruction and since it is a jump address - jump
> to that address and then BEAM is performing the next
> BEAM instruction.
> 
> If the BEAM instruction is a (conditional) jump or call
> et.al, the instruction just sets the BEAM instruction
> pointer to point to the next BEAM instruction to
> actually perform, reads it and jumps.
> 
> So there is no call setup overhead per instruction.
> 
> 
> On Wed, Aug 08, 2007 at 08:03:46PM +0100, Joel Reymont wrote:
>> On Aug 8, 2007, at 6:26 AM, Bjorn Gustavsson wrote:
>>
>>> The first word in each instruction points to the executable C code  
>>> for the instruction. That first instruction word is followed by  
>>> zero or more operand words.
>> Would it be right to say that Erlang VM instructions are C function  
>> calls? This is my understanding of the above.
>>
>> Is there significant overhead of function call setup per instruction?
>>
>> 	Thanks, Joel
>>
>> --
>> http://wagerlabs.com
>>
>>
>>
>>
>>
>> _______________________________________________
>> erlang-questions mailing list
>> 
>> http://www.erlang.org/mailman/listinfo/erlang-questions
> 




More information about the erlang-questions mailing list