[v8-users] Re: Persistent::WrapperClassId size reduction

[Mike Moening]

255 isn't a very big number.  For debugging purposes its almost nicer a 
have a string (i.e.  Ben's BABE below)
Numbers are harder on the eyes and less meaningful.

On Monday, August 5, 2013 5:56:56 AM UTC-5, Dan Carney wrote:
>
> Hi all,
>
> is anyone using more than a few bits of the Persistent::WrapperClassId 
> field?  I'm not seeing a lot of uses in the wild, and I'd like to cut it 
> down from a uint16_t to a uint8_t.
>
> Thanks,
> Dan
>

-- 
-- 
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/groups/opt_out.

Re: [v8-users] Soft Floating Point support

[Rodolph Perfetta]

V8 used to have a soft-float option but this has been deprecated and the
code removed.

I believe 3.17 is the last branch with this option if you want to try. Two
quick notes:
 * it wasn't thoroughly tested
 * when VFP is off V8 doesn't use the optimising JIT

Regards,
Rodolph.



On 5 August 2013 12:25, Thulasidhar JK  wrote:

> Thanks for your feedback.
> I understand that there will be a performance drop if we go for soft-float.
>
> On this site (
> http://blogs.arm.com/software-enablement/456-googles-v8-on-arm-five-times-better/),
> there is mention of V8 running without VFP (in the Crankshaft section).
> Is there any option to build V8 without vfp support?
> Even if it is only for developers, we would like to try and see if it
> meets with our performance requirements.
>
> Warm Regards,
> JK
>
>
> On Monday, August 5, 2013 6:58:47 PM UTC+8, Rodolph Perfetta wrote:
>
>> Using library calls will force you to do a lot of push/pop/register
>> shuffling and the performance will be poor. That being said if you are
>> moving form a VFP soc to a non VFP soc you may not have an L2 cache either
>> in which case performance is going to be low anyway.
>>
>> HTH,
>> Rodolph.
>>
>>
>> On 5 August 2013 11:21, Ben Noordhuis  wrote:
>>
>>> On Mon, Aug 5, 2013 at 12:07 PM, Rodolph Perfetta
>>>  wrote:
>>> > The question is does v8 work when there is no hardware floating point
>>> > present and the answer is no.
>>> >
>>> > softfp is the name of the abi which assumes hardware floating point
>>> but uses
>>> > integer result for function call.
>>> >
>>> > Nowadays a lot of software on ARM (all the ubuntu packages for
>>> example) are
>>> > compile for hardware floating point.
>>>
>>> Oh, I guess you're right - I think I misread VFP as VFP3 in the OP's
>>> post.
>>>
>>> @OP: softfloat (as opposed to softfp) is not supported.  In theory,
>>> it'd be possible for V8 to emit library calls rather than VFP
>>> instructions but in practice it doesn't.  Making it softfloat
>>> compatible would be a fair amount of work, I think.
>>>
>>> --
>>> --
>>> 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/**groups/opt_out
>>> .
>>>
>>>
>>>
>>  --
> --
> 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/groups/opt_out.
>
>
>

-- 
-- 
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/groups/opt_out.

Re: [v8-users] Persistent::WrapperClassId size reduction

[Ben Noordhuis]

On Mon, Aug 5, 2013 at 12:56 PM, Dan Carney  wrote:
> Hi all,
>
> is anyone using more than a few bits of the Persistent::WrapperClassId
> field?  I'm not seeing a lot of uses in the wild, and I'd like to cut it
> down from a uint16_t to a uint8_t.
>
> Thanks,
> Dan

I'll have to change my 0xBABE ID to e.g. 0xBA but that's something I
can live with.

-- 
-- 
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/groups/opt_out.

Re: [v8-users] Soft Floating Point support

[Thulasidhar JK]

Thanks for your feedback.
I understand that there will be a performance drop if we go for soft-float.

On this site 
(http://blogs.arm.com/software-enablement/456-googles-v8-on-arm-five-times-better/),
 
there is mention of V8 running without VFP (in the Crankshaft section).
Is there any option to build V8 without vfp support?
Even if it is only for developers, we would like to try and see if it meets 
with our performance requirements.

Warm Regards,
JK

On Monday, August 5, 2013 6:58:47 PM UTC+8, Rodolph Perfetta wrote:
>
> Using library calls will force you to do a lot of push/pop/register 
> shuffling and the performance will be poor. That being said if you are 
> moving form a VFP soc to a non VFP soc you may not have an L2 cache either 
> in which case performance is going to be low anyway.
>
> HTH,
> Rodolph.
>
>
> On 5 August 2013 11:21, Ben Noordhuis  >wrote:
>
>> On Mon, Aug 5, 2013 at 12:07 PM, Rodolph Perfetta
>> > wrote:
>> > The question is does v8 work when there is no hardware floating point
>> > present and the answer is no.
>> >
>> > softfp is the name of the abi which assumes hardware floating point but 
>> uses
>> > integer result for function call.
>> >
>> > Nowadays a lot of software on ARM (all the ubuntu packages for example) 
>> are
>> > compile for hardware floating point.
>>
>> Oh, I guess you're right - I think I misread VFP as VFP3 in the OP's post.
>>
>> @OP: softfloat (as opposed to softfp) is not supported.  In theory,
>> it'd be possible for V8 to emit library calls rather than VFP
>> instructions but in practice it doesn't.  Making it softfloat
>> compatible would be a fair amount of work, I think.
>>
>> --
>> --
>> 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/groups/opt_out.
>>
>>
>>
>

-- 
-- 
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/groups/opt_out.

Re: [v8-users] Re: Holding v8:Persistent handles

[Dan Carney]

We like it too, which is why we are planning on having a second template 
parameter for persistent, a traits class, which will control whether a 
handle is copyable, etc.  We're undecided on what the default template 
parameter will be, but at first it will have to implement the current 
semantics to keep things sane.

On Monday, August 5, 2013 11:38:38 AM UTC+2, Ben Noordhuis wrote:
>
> On Mon, Aug 5, 2013 at 9:01 AM, Dan Carney > 
> wrote: 
> >> Good. 
> >> Will we be able to copy persistent?  In my case I have to be able to 
> pass 
> >> around a persistent and store it in several places at a time. 
> >> Or should I wrap it in another object and ref count it myself? 
> > 
> > 
> > It will be copyable, but copying will become more expensive, as the 
> backing 
> > store needs to be copied. 
>
> As a counterpoint, I like that Persistent is non-copyable now.  It 
> makes it much easier to reason about its life cycle and ownership.  I 
> would move to keep it like this. 
>

-- 
-- 
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/groups/opt_out.

Re: [v8-users] Soft Floating Point support

[Rodolph Perfetta]

Using library calls will force you to do a lot of push/pop/register
shuffling and the performance will be poor. That being said if you are
moving form a VFP soc to a non VFP soc you may not have an L2 cache either
in which case performance is going to be low anyway.

HTH,
Rodolph.


On 5 August 2013 11:21, Ben Noordhuis  wrote:

> On Mon, Aug 5, 2013 at 12:07 PM, Rodolph Perfetta
>  wrote:
> > The question is does v8 work when there is no hardware floating point
> > present and the answer is no.
> >
> > softfp is the name of the abi which assumes hardware floating point but
> uses
> > integer result for function call.
> >
> > Nowadays a lot of software on ARM (all the ubuntu packages for example)
> are
> > compile for hardware floating point.
>
> Oh, I guess you're right - I think I misread VFP as VFP3 in the OP's post.
>
> @OP: softfloat (as opposed to softfp) is not supported.  In theory,
> it'd be possible for V8 to emit library calls rather than VFP
> instructions but in practice it doesn't.  Making it softfloat
> compatible would be a fair amount of work, I think.
>
> --
> --
> 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/groups/opt_out.
>
>
>

-- 
-- 
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/groups/opt_out.

[v8-users] Persistent::WrapperClassId size reduction

[Dan Carney]

Hi all,

is anyone using more than a few bits of the Persistent::WrapperClassId 
field?  I'm not seeing a lot of uses in the wild, and I'd like to cut it 
down from a uint16_t to a uint8_t.

Thanks,
Dan

-- 
-- 
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/groups/opt_out.

Re: [v8-users] Soft Floating Point support

[Ben Noordhuis]

On Mon, Aug 5, 2013 at 12:07 PM, Rodolph Perfetta
 wrote:
> The question is does v8 work when there is no hardware floating point
> present and the answer is no.
>
> softfp is the name of the abi which assumes hardware floating point but uses
> integer result for function call.
>
> Nowadays a lot of software on ARM (all the ubuntu packages for example) are
> compile for hardware floating point.

Oh, I guess you're right - I think I misread VFP as VFP3 in the OP's post.

@OP: softfloat (as opposed to softfp) is not supported.  In theory,
it'd be possible for V8 to emit library calls rather than VFP
instructions but in practice it doesn't.  Making it softfloat
compatible would be a fair amount of work, I think.

-- 
-- 
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/groups/opt_out.

Re: [v8-users] Soft Floating Point support

[Rodolph Perfetta]

The question is does v8 work when there is no hardware floating point
present and the answer is no.

softfp is the name of the abi which assumes hardware floating point but
uses integer result for function call.

Nowadays a lot of software on ARM (all the ubuntu packages for example) are
compile for hardware floating point.


On 5 August 2013 10:45, Ben Noordhuis  wrote:

> On Mon, Aug 5, 2013 at 11:25 AM, Thulasidhar JK 
> wrote:
> > Dear All,
> >
> > We are planning to use Blink+V8 as the browser in our embedded devices
> > (blu-ray players)
> >
> > We are currently using Webkit+JSC on ARMv6z CPU and are moving into a new
> > variant of this chipset, which does not have VFP copro.
> > The JSC JIT is heavily dependent on hw vfp and hence we are blocked on
> our
> > development.
> >
> > Now we are thinking to shift to Blink+V8 and in this regard we would to
> > understand if V8 can be used in this scenario.
> > Please let me know your thoughts on this.
> >
> > Warm Regards,
> > JK
>
> If your question is "does V8 work with softfp?" then the answer is
> "yes, it does."  Compile it with `make arm.release
> armfloatabi=softfp`.
>
> There are a few more ARM-specific switches, you can find them in the
> Makefile.  You probably want to tweak armfpu and (maybe) armthumb.
>
> --
> --
> 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/groups/opt_out.
>
>
>

-- 
-- 
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/groups/opt_out.

Re: [v8-users] Soft Floating Point support

[Ben Noordhuis]

On Mon, Aug 5, 2013 at 11:25 AM, Thulasidhar JK  wrote:
> Dear All,
>
> We are planning to use Blink+V8 as the browser in our embedded devices
> (blu-ray players)
>
> We are currently using Webkit+JSC on ARMv6z CPU and are moving into a new
> variant of this chipset, which does not have VFP copro.
> The JSC JIT is heavily dependent on hw vfp and hence we are blocked on our
> development.
>
> Now we are thinking to shift to Blink+V8 and in this regard we would to
> understand if V8 can be used in this scenario.
> Please let me know your thoughts on this.
>
> Warm Regards,
> JK

If your question is "does V8 work with softfp?" then the answer is
"yes, it does."  Compile it with `make arm.release
armfloatabi=softfp`.

There are a few more ARM-specific switches, you can find them in the
Makefile.  You probably want to tweak armfpu and (maybe) armthumb.

-- 
-- 
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/groups/opt_out.

Re: [v8-users] Re: Holding v8:Persistent handles

[Ben Noordhuis]

On Mon, Aug 5, 2013 at 9:01 AM, Dan Carney  wrote:
>> Good.
>> Will we be able to copy persistent?  In my case I have to be able to pass
>> around a persistent and store it in several places at a time.
>> Or should I wrap it in another object and ref count it myself?
>
>
> It will be copyable, but copying will become more expensive, as the backing
> store needs to be copied.

As a counterpoint, I like that Persistent is non-copyable now.  It
makes it much easier to reason about its life cycle and ownership.  I
would move to keep it like this.

-- 
-- 
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/groups/opt_out.

[v8-users] Soft Floating Point support

[Thulasidhar JK]

Dear All,

We are planning to use Blink+V8 as the browser in our embedded devices 
(blu-ray players)

We are currently using Webkit+JSC on ARMv6z CPU and are moving into a new 
variant of this chipset, which does not have VFP copro.
The JSC JIT is heavily dependent on hw vfp and hence we are blocked on our 
development.

Now we are thinking to shift to Blink+V8 and in this regard we would to 
understand if V8 can be used in this scenario.
Please let me know your thoughts on this.

Warm Regards,
JK

-- 
-- 
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/groups/opt_out.

Re: [v8-users] Re: Holding v8:Persistent handles

[Dan Carney]

>
> Good.  
> Will we be able to copy persistent?  In my case I have to be able to pass 
> around a persistent and store it in several places at a time.
> Or should I wrap it in another object and ref count it myself?
>

It will be copyable, but copying will become more expensive, as the backing 
store needs to be copied.

-- 
-- 
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/groups/opt_out.