[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Public WebGL] Feature proposal for TypedArrays



On May 10, 2010, at 3:54 PM, Kenneth Russell wrote:

> On Mon, May 10, 2010 at 10:57 AM, Vladimir Vukicevic
> <vladimir@mozilla.com> wrote:
>> On 5/10/2010 9:49 AM, Chris Marrin wrote:
>>> 
>>> But fp32 and fp64 are supported in compilers and hardware FPUs. I don't
>>> know of any CPU's that support fp16 natively. So full support would require
>>> conversion to and from a support FP format. I don't think it's worth doing
>>> that at this point.
>> 
>> This is the part that worries me the most about the proposal, but even if we
>> implemented fp16 support in software, the typical usage could be to do math
>> as fp32 and then convert to fp16 before uploading to the video card for
>> memory savings.  That would argue for having a fp16 type, but cautioning
>> developers to not actually do math with it, which is a little scary --
>> someone invariably will.  However, if the use case is mainly for
>> pre-computed data, it can already be stored and transmitted as a raw buffer
>> (and if light manipulation is needed, an unsigned short array would provide
>> access to the individual fp16 elements).
> 
> I think it would be a good idea to eventually support the "half" data
> type, but currently there are no planned WebGL extensions that would
> use it. The additional cost is significantly higher than for the other
> existing data types because most CPUs do not support this format
> natively.
> 
> I think that we should add this as a to-do item, but not add it to the
> TypedArray or WebGL specifications at the current time. There is
> already too much work that needs to be done to stabilize the
> specification for version 1.0.


Agreed

-----
~Chris
cmarrin@apple.com





-----------------------------------------------------------
You are currently subscribe to public_webgl@khronos.org.
To unsubscribe, send an email to majordomo@khronos.org with
the following command in the body of your email: