Re: [v8-users] internal fields

[J Decker]

On Mon, Nov 6, 2017 at 12:47 PM, Jakob Kummerow 
wrote:

> 3. What difference does it make to v8 if the internal field is an aligned
>>> pointer or not? Is the ability to set/get aligned pointers a consistency
>>> check so assumptions can be made? Does the interface check the alignment?
>>> (Not critical for me, I don't think, but I'd like to understand.)
>>>
>>
>> I doubt it matters... basically internal fields seem to be user-data
>> fields that store the value so your user code can later retrieve it.
>> Internally I wouldn't expect V8 to ever actually do anything with those
>> fields. Since they are usually pointers that are stored, aligned buffers
>> will be more optimal.
>>
>
> It actually does matter, and the interface does check for alignment. The
> reason is that V8 relies on those pointers being aligned: it uses this fact
> as part of the mechanism for deciding to ignore them.
>

Well 'aligned' is pretty generous... it only requires being 16 bit
aligned...  (2 byte)


from https://github.com/v8/v8/blob/master/src/api.cc#L1186

const int kSmiTag = 0;
const int kSmiTagSize = 1;
const intptr_t kSmiTagMask = (1 << kSmiTagSize) - 1;

#define HAS_SMI_TAG(value) \
  ((reinterpret_cast(value) & ::i::kSmiTagMask) == ::i::kSmiTag)

bool Object::IsSmi() const { return HAS_SMI_TAG(this); }

Utils::ApiCheck(smi->IsSmi(), location, "Pointer is not aligned");

Which since this is ObjectWrap feature, it's passed a pointer to a class,
which I would imagine 'new' will return a 4 or 8 byte aligned pointer; but
on further research there is no guarantee.

-- 
> --
> v8-users mailing list
> v8-users@googlegroups.com
> http://groups.google.com/group/v8-users
> ---
> You received this message because you are subscribed to the Google Groups
> "v8-users" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to v8-users+unsubscr...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.
>

-- 
-- 
v8-users mailing list
v8-users@googlegroups.com
http://groups.google.com/group/v8-users
--- 
You received this message because you are subscribed to the Google Groups 
"v8-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to v8-users+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Re: [v8-users] internal fields

[Jakob Kummerow]

>
> 3. What difference does it make to v8 if the internal field is an aligned
>> pointer or not? Is the ability to set/get aligned pointers a consistency
>> check so assumptions can be made? Does the interface check the alignment?
>> (Not critical for me, I don't think, but I'd like to understand.)
>>
>
> I doubt it matters... basically internal fields seem to be user-data
> fields that store the value so your user code can later retrieve it.
> Internally I wouldn't expect V8 to ever actually do anything with those
> fields. Since they are usually pointers that are stored, aligned buffers
> will be more optimal.
>

It actually does matter, and the interface does check for alignment. The
reason is that V8 relies on those pointers being aligned: it uses this fact
as part of the mechanism for deciding to ignore them.

-- 
-- 
v8-users mailing list
v8-users@googlegroups.com
http://groups.google.com/group/v8-users
--- 
You received this message because you are subscribed to the Google Groups 
"v8-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to v8-users+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Re: [v8-users] internal fields

[Bruce MacNaughton]

>
> Does it use Wrap and/or as classes subclassed with ObjectWrap?  Wrap uses 
> internal field 0 to store the class so it can be later unwrapped from the 
> V8 object.

https://github.com/nodejs/node/blob/master/src/node_object_wrap.h#L75 (near 
> that is also SetWeak reference)


Yes, I think you just answered my question. So
 object->Wrap(info.This());

in the Object::New() method stores the instance in internal field 0, right?


>

-- 
-- 
v8-users mailing list
v8-users@googlegroups.com
http://groups.google.com/group/v8-users
--- 
You received this message because you are subscribed to the Google Groups 
"v8-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to v8-users+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Re: [v8-users] internal fields

[Bruce MacNaughton]

Thanks for pointing me in the right direction.

On Saturday, November 4, 2017 at 9:52:23 AM UTC-7, J Decker wrote:
>
>
>
> On Sat, Nov 4, 2017 at 9:29 AM, Bruce MacNaughton  > wrote:
>
>> I am new to Nan, V8, and C++ (so if I haven't put a big enough target on 
>> my back I don't know what else I can add). I've written a lot of JavaScript 
>> and, in the past, C, assembler, and kernel mode code, so hopefully the 
>> bulls-eye is a little smaller now.
>>
>> I'm working with an existing code base and am trying to understand why 
>> things were done the way they were. It uses Nan to create an addon for 
>> nodejs. I'm hoping someone here can help me understand some pieces that 
>> escape me.
>>
>  
> Nan is really a nodejs thing, and not V8... so this is sort of the wrong 
> place for these questions...
>  
>
>>
>> 1. The code sets internal field count for each class - sometimes to 1 and 
>> sometimes to 2 - but never invokes "setInternalField()" or 
>> "getInternalField()". Is there some reason, possibly historical, that 
>> "setInternalFieldCount()" needed to be called to set a value? The way I 
>> have interpreted what I've read is that my code needs to set and get the 
>> value explicitly, so setting a value but never storing anything there makes 
>> no sense to me.
>>
>>   // Prepare constructor template
>>  v8::Local ctor = 
>> Nan::New(New);
>>  ctor->InstanceTemplate()->SetInternalFieldCount(2);
>>  ctor->SetClassName(Nan::New("MyClass").ToLocalChecked());
>>
>>
> Does it use Wrap and/or as classes subclassed with ObjectWrap?  Wrap uses 
> internal field 0 to store the class so it can be later unwrapped from the 
> V8 object.
> https://github.com/nodejs/node/blob/master/src/node_object_wrap.h#L75 
> (near that is also SetWeak reference)
>  
>
>> 2. Given that I'm storing something in internal fields, my understanding 
>> is that I need to free any resources (memory, etc.) that are used by the 
>> internal field if the object is GC'd. Doing that in the destructor seems to 
>> be the right way to handle that. Is that all there is to it?
>>
>> the destructor is really too late, at the point the destructor is called, 
> the Object holding it would have also disappeared If the destructor is 
> getting called, it's probably because of an ObjectWrapped thing 
> disappearing, which internally stores the object in the class as a 
> Persistent<> that is SetWeak()'d.  SetWeak takes a callback which is called 
> when the object is GC'd.
>  
>
>> 3. What difference does it make to v8 if the internal field is an aligned 
>> pointer or not? Is the ability to set/get aligned pointers a consistency 
>> check so assumptions can be made? Does the interface check the alignment? 
>> (Not critical for me, I don't think, but I'd like to understand.)
>>
>>
> I dooubt it matters... basically internal fields seem to be user-data 
> fields that store the value so your user code can later retrieve it.  
> Internally I wouldn't expect V8 to ever actually do anything with those 
> fields. Since they are usually pointers that are stored, aligned buffers 
> will be more optimal.
>  
>
>>
>>
>>
>>
>> -- 
>> -- 
>> v8-users mailing list
>> v8-u...@googlegroups.com 
>> http://groups.google.com/group/v8-users
>> --- 
>> You received this message because you are subscribed to the Google Groups 
>> "v8-users" group.
>> To unsubscribe from this group and stop receiving emails from it, send an 
>> email to v8-users+u...@googlegroups.com .
>> For more options, visit https://groups.google.com/d/optout.
>>
>
>

-- 
-- 
v8-users mailing list
v8-users@googlegroups.com
http://groups.google.com/group/v8-users
--- 
You received this message because you are subscribed to the Google Groups 
"v8-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to v8-users+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

Re: [v8-users] internal fields

[J Decker]

On Sat, Nov 4, 2017 at 9:29 AM, Bruce MacNaughton 
wrote:

> I am new to Nan, V8, and C++ (so if I haven't put a big enough target on
> my back I don't know what else I can add). I've written a lot of JavaScript
> and, in the past, C, assembler, and kernel mode code, so hopefully the
> bulls-eye is a little smaller now.
>
> I'm working with an existing code base and am trying to understand why
> things were done the way they were. It uses Nan to create an addon for
> nodejs. I'm hoping someone here can help me understand some pieces that
> escape me.
>

Nan is really a nodejs thing, and not V8... so this is sort of the wrong
place for these questions...


>
> 1. The code sets internal field count for each class - sometimes to 1 and
> sometimes to 2 - but never invokes "setInternalField()" or
> "getInternalField()". Is there some reason, possibly historical, that
> "setInternalFieldCount()" needed to be called to set a value? The way I
> have interpreted what I've read is that my code needs to set and get the
> value explicitly, so setting a value but never storing anything there makes
> no sense to me.
>
>   // Prepare constructor template
>  v8::Local ctor = Nan::New
> (New);
>  ctor->InstanceTemplate()->SetInternalFieldCount(2);
>  ctor->SetClassName(Nan::New("MyClass").ToLocalChecked());
>
>
Does it use Wrap and/or as classes subclassed with ObjectWrap?  Wrap uses
internal field 0 to store the class so it can be later unwrapped from the
V8 object.
https://github.com/nodejs/node/blob/master/src/node_object_wrap.h#L75 (near
that is also SetWeak reference)


> 2. Given that I'm storing something in internal fields, my understanding
> is that I need to free any resources (memory, etc.) that are used by the
> internal field if the object is GC'd. Doing that in the destructor seems to
> be the right way to handle that. Is that all there is to it?
>
> the destructor is really too late, at the point the destructor is called,
the Object holding it would have also disappeared If the destructor is
getting called, it's probably because of an ObjectWrapped thing
disappearing, which internally stores the object in the class as a
Persistent<> that is SetWeak()'d.  SetWeak takes a callback which is called
when the object is GC'd.


> 3. What difference does it make to v8 if the internal field is an aligned
> pointer or not? Is the ability to set/get aligned pointers a consistency
> check so assumptions can be made? Does the interface check the alignment?
> (Not critical for me, I don't think, but I'd like to understand.)
>
>
I dooubt it matters... basically internal fields seem to be user-data
fields that store the value so your user code can later retrieve it.
Internally I wouldn't expect V8 to ever actually do anything with those
fields. Since they are usually pointers that are stored, aligned buffers
will be more optimal.


>
>
>
>
> --
> --
> v8-users mailing list
> v8-users@googlegroups.com
> http://groups.google.com/group/v8-users
> ---
> You received this message because you are subscribed to the Google Groups
> "v8-users" group.
> To unsubscribe from this group and stop receiving emails from it, send an
> email to v8-users+unsubscr...@googlegroups.com.
> For more options, visit https://groups.google.com/d/optout.
>

-- 
-- 
v8-users mailing list
v8-users@googlegroups.com
http://groups.google.com/group/v8-users
--- 
You received this message because you are subscribed to the Google Groups 
"v8-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to v8-users+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.

[v8-users] internal fields

[Bruce MacNaughton]

I am new to Nan, V8, and C++ (so if I haven't put a big enough target on my 
back I don't know what else I can add). I've written a lot of JavaScript 
and, in the past, C, assembler, and kernel mode code, so hopefully the 
bulls-eye is a little smaller now.

I'm working with an existing code base and am trying to understand why 
things were done the way they were. It uses Nan to create an addon for 
nodejs. I'm hoping someone here can help me understand some pieces that 
escape me.

1. The code sets internal field count for each class - sometimes to 1 and 
sometimes to 2 - but never invokes "setInternalField()" or 
"getInternalField()". Is there some reason, possibly historical, that 
"setInternalFieldCount()" needed to be called to set a value? The way I 
have interpreted what I've read is that my code needs to set and get the 
value explicitly, so setting a value but never storing anything there makes 
no sense to me.

  // Prepare constructor template
 v8::Local ctor = Nan::New(New);
 ctor->InstanceTemplate()->SetInternalFieldCount(2);
 ctor->SetClassName(Nan::New("MyClass").ToLocalChecked());

2. Given that I'm storing something in internal fields, my understanding is 
that I need to free any resources (memory, etc.) that are used by the 
internal field if the object is GC'd. Doing that in the destructor seems to 
be the right way to handle that. Is that all there is to it?

3. What difference does it make to v8 if the internal field is an aligned 
pointer or not? Is the ability to set/get aligned pointers a consistency 
check so assumptions can be made? Does the interface check the alignment? 
(Not critical for me, I don't think, but I'd like to understand.)





-- 
-- 
v8-users mailing list
v8-users@googlegroups.com
http://groups.google.com/group/v8-users
--- 
You received this message because you are subscribed to the Google Groups 
"v8-users" group.
To unsubscribe from this group and stop receiving emails from it, send an email 
to v8-users+unsubscr...@googlegroups.com.
For more options, visit https://groups.google.com/d/optout.