[ruby-core:68506] Seven stacks (and two questions)

[ruby-core:68506] Seven stacks (and two questions)

[Jakub Trzebiatowski]

[-- Attachment #1: Type: text/plain, Size: 772 bytes --]

The Ruby Hacking Guide says that Ruby has… seven stacks. Is it an implementation choice (and it could be implemented with one stack), or is there really a need for seven logical stacks? For example, Lua has one stack, and still closures with upvalues are totally possible (it’s like Ruby’s blocks that can reference local variables of their enclosing method, but it works for any function with any upvalues). 

1. Is there any Ruby feature that I’m not aware of, and it justifies the need of seven stacks?

2. When does Ruby generate `getdynamic` YARV opcode? Could you provide me with an example source code? (I really tried to make Ruby generate it. No results.)

---
Jakub Trzebiatowski
Computer science student
Gdańsk University of Technology




[-- Attachment #2: Type: text/html, Size: 1662 bytes --]

[ruby-core:68508] Re: Seven stacks (and two questions)

[Nobuyoshi Nakada]

On 2015/03/12 19:15, Jakub Trzebiatowski wrote:
> The Ruby Hacking Guide says that Ruby has… seven stacks. Is it an
> implementation choice (and it could be implemented with one stack),
> or is there really a need for seven logical stacks? For example, Lua
> has one stack, and still closures with upvalues are totally possible
> (it’s like Ruby’s blocks that can reference local variables of
> their enclosing method, but it works for any function with any
> upvalues).

RHG is a good book, but quite old for the core implementation.

> 2. When does Ruby generate `getdynamic` YARV opcode? Could you
> provide me with an example source code? (I really tried to make Ruby
> generate it. No results.)

No such instruction in YARV.
Do you mean `getlocal`?

-- 
Nobu Nakada

[ruby-core:68512] Re: Seven stacks (and two questions)

[SASADA Koichi]

On 2015/03/12 22:00, Nobuyoshi Nakada wrote:
>> > 2. When does Ruby generate `getdynamic` YARV opcode? Could you
>> > provide me with an example source code? (I really tried to make Ruby
>> > generate it. No results.)
> No such instruction in YARV.
> Do you mean `getlocal`?

Ruby 1.9 has. But it was removed.

-- 
// SASADA Koichi at atdot dot net

[ruby-core:68513] Re: Seven stacks (and two questions)

["Martin J. Dürst"]

Hello Jakub,

On 2015/03/12 22:00, Nobuyoshi Nakada wrote:
> On 2015/03/12 19:15, Jakub Trzebiatowski wrote:
>> The Ruby Hacking Guide says that Ruby has… seven stacks. Is it an
>> implementation choice (and it could be implemented with one stack),
>> or is there really a need for seven logical stacks? For example, Lua
>> has one stack, and still closures with upvalues are totally possible
>> (it’s like Ruby’s blocks that can reference local variables of
>> their enclosing method, but it works for any function with any
>> upvalues).
>
> RHG is a good book, but quite old for the core implementation.

I suggest you look at "Ruby Under a Microscope: An Illustrated Guide to 
Ruby Internals". It's much less detailed than the Ruby Hacking Guide, 
but much more up to date. And as far as I can remember, it talks about 2 
stacks, not seven.

Regards,   Martin.

[ruby-core:68514] Re: Seven stacks (and two questions)

[Nobuyoshi Nakada]

On 2015/03/12 22:00, Nobuyoshi Nakada wrote:
> On 2015/03/12 19:15, Jakub Trzebiatowski wrote:
>> The Ruby Hacking Guide says that Ruby has… seven stacks. Is it an
>> implementation choice (and it could be implemented with one stack),
>> or is there really a need for seven logical stacks? For example, Lua
>> has one stack, and still closures with upvalues are totally possible
>> (it’s like Ruby’s blocks that can reference local variables of
>> their enclosing method, but it works for any function with any
>> upvalues).
> 
> RHG is a good book, but quite old for the core implementation.

Perhaps, "stacks" might confuse you, I guess.

They are actually single-linked lists, stored on the machine stack but not separate areas, and called "stacks" as they are used as LIFO.
The reason counting separately is that all infos are not always saved/restored together.

-- 
Nobu Nakada