Re: RFC: Re: shader IR vs. application IR

On 18 Aug 2017, at 1:00 am, David Neto <dneto@google.com> wrote:
> Just to be clear: for the web the attitude should be that the shader is untrusted.

Kind of... if a single process model is presented within the specs, then the shader is free to do whatever it likes with pointers.

Is it possible to unmap everything and monitor context calls to ensure that WebGPU shaders only have access to their own resources?

On all three platforms? If the implementation detail is accessible enough, then it could result in a change to the specifications.

--
  steve@adam.com.au

Received on Friday, 18 August 2017 12:49:43 UTC