[Mesa-dev] [PATCH] main: Match DispatchCompute* API validation from main specification

Fri Oct 30 21:27:12 PDT 2015

On 2015-10-30 00:16:43, Iago Toral wrote:
> On Wed, 2015-10-14 at 13:46 -0700, Jordan Justen wrote:
> > There is a discrepancy between the ARB_compute_shader specification,
> > and the OpenGL 4.3 and OpenGLES 3.1 specifications. With regards to
> > the indirect dispatch parameter, unsupported value errors should
> > return INVALID_VALUE according to the main specifications, whereas the
> > extension specification indicated INVALID_OPERATION should be
> > returned.
> > 
> > Here we update the code to match the main specifications, and update
> > the citations use the main specification rather than the extension
> > specification.
> > 
> > Signed-off-by: Jordan Justen <jordan.l.justen at intel.com>
> > ---
> >  Fixes ES31-CTS.compute_shader.api-indirect
> > 
> >  src/mesa/main/api_validate.c | 37 ++++++++++++++++++++++++++-----------
> >  1 file changed, 26 insertions(+), 11 deletions(-)
> > 
> > diff --git a/src/mesa/main/api_validate.c b/src/mesa/main/api_validate.c
> > index a46c194..c286945 100644
> > --- a/src/mesa/main/api_validate.c
> > +++ b/src/mesa/main/api_validate.c
> > @@ -895,6 +895,11 @@ check_valid_to_compute(struct gl_context *ctx, const char *function)
> >        return false;
> >     }
> >  
> > +   /* From the OpenGL 4.3 Core Specification, Chapter 19, Compute Shaders:
> > +    *
> > +    * "An INVALID_OPERATION error is generated if there is no active program
> > +    *  for the compute shader stage."
> > +    */
> >     prog = ctx->Shader.CurrentProgram[MESA_SHADER_COMPUTE];
> >     if (prog == NULL || prog->_LinkedShaders[MESA_SHADER_COMPUTE] == NULL) {
> >        _mesa_error(ctx, GL_INVALID_OPERATION,
> > @@ -917,6 +922,12 @@ _mesa_validate_DispatchCompute(struct gl_context *ctx,
> >        return GL_FALSE;
> >  
> >     for (i = 0; i < 3; i++) {
> > +      /* From the OpenGL 4.3 Core Specification, Chapter 19, Compute Shaders:
> > +       *
> > +       * "An INVALID_VALUE error is generated if any of num_groups_x,
> > +       *  num_groups_y and num_groups_z are greater than or equal to the
> > +       *  maximum work group count for the corresponding dimension."
> > +       */
> >        if (num_groups[i] > ctx->Const.MaxComputeWorkGroupCount[i]) {
> 
> This should be >= according to that spec quotation.

Wow. Good catch, except, I am fairly certain that this must be a spec
error. 'greater than' instead of 'greater than or equal' is more
consistent with the usage and other parts of the spec.

Unfortunately, this wording exists in the OpenGL 4.3 - 4.5 specs. What
a mess. :)

Here is some contradicting text from the DispatchCompute section of
4.3 spec:

"num groups x, num groups y and num groups z specify the number of
 local work groups that will be dispatched in the X, Y and Z
 dimensions, respectively."

"The maximum number of work groups that may be dispatched at one time
 may be determined by calling GetIntegeri v with target set to
 MAX_COMPUTE_WORK_GROUP_COUNT and index set to zero, one, or two,
 representing the X, Y, and Z dimensions respectively."

So, it is saying that MAX_COMPUTE_WORK_GROUP_COUNT is the maximum
values. Thus being 'equal' to the max should be valid.

And, then in the DispatchComputeIndirect section of the 4.3 spec:

"If any of num groups x, num groups y or num groups z is greater than
 the value of MAX_COMPUTE_WORK_GROUP_COUNT for the corresponding
 dimension then the results are undefined."

Note 'greater than' rather than 'greater than or equal'.

I'll ask the spec author about it, but in the meantime, I think I
should just skip adding this spec citation, or instead use the one
from above that starts with "The maximum number of " ...

What do you think?

> 
> >           _mesa_error(ctx, GL_INVALID_VALUE,
> >                       "glDispatchCompute(num_groups_%c)", 'x' + i);
> > @@ -937,24 +948,33 @@ valid_dispatch_indirect(struct gl_context *ctx,
> >     if (!check_valid_to_compute(ctx, name))
> >        return GL_FALSE;
> >  
> > -   /* From the ARB_compute_shader specification:
> > +   /* From the OpenGL 4.3 Core Specification, Chapter 19, Compute Shaders:
> >      *
> > -    * "An INVALID_OPERATION error is generated [...] if <indirect> is less
> > -    *  than zero or not a multiple of the size, in basic machine units, of
> > -    *  uint."
> > +    * "An INVALID_VALUE error is generated if indirect is negative or is not a
> > +    *  multiple of four."
> > +    * Note that the OpenGLES 3.1 specification matches this, but this is
> > +    * different than the extension specification which has a return of
> > +    * INVALID_OPERATION instead.
> 
> However, I read the following:
> 
> "void DispatchComputeIndirect(intptr indirect);"
> (...)
> The error INVALID_VALUE is generated if <indirect> is less than zero or
> is not a multiple of four.(...)"
> 
> and then again:
> 
> "Errors:
> (...)
> INVALID_VALUE is generated by DispatchComputeIndirect if <indirect> is
> less than zero or not a multiple of four."
> 
> From:
> https://www.opengl.org/registry/specs/ARB/compute_shader.txt

Hmm, so the ARB_compute_shader spec is inconsistent with itself. :(

How about I just drop the note that the ARB_compute_shader
contradicting the OpenGL spec? In other words:

-   /* From the ARB_compute_shader specification:
+   /* From the OpenGL 4.3 Core Specification, Chapter 19, Compute Shaders:
     *
-    * "An INVALID_OPERATION error is generated [...] if <indirect> is less
-    *  than zero or not a multiple of the size, in basic machine units, of
-    *  uint."
+    * "An INVALID_VALUE error is generated if indirect is negative or is not a
+    *  multiple of four."

Thanks,

-Jordan

> 
> >      */
> >     if ((GLintptr)indirect & (sizeof(GLuint) - 1)) {
> > -      _mesa_error(ctx, GL_INVALID_OPERATION,
> > +      _mesa_error(ctx, GL_INVALID_VALUE,
> >                    "%s(indirect is not aligned)", name);
> >        return GL_FALSE;
> >     }
> >  
> >     if ((GLintptr)indirect < 0) {
> > -      _mesa_error(ctx, GL_INVALID_OPERATION,
> > +      _mesa_error(ctx, GL_INVALID_VALUE,
> >                    "%s(indirect is less than zero)", name);
> >        return GL_FALSE;
> >     }
> >  
> > +   /* From the OpenGL 4.3 Core Specification, Chapter 19, Compute Shaders:
> > +    *
> > +    * "An INVALID_OPERATION error is generated if no buffer is bound to the
> > +    *  DRAW_INDIRECT_BUFFER binding, or if the command would source data
> > +    *  beyond the end of the buffer object."
> > +    */
> >     if (!_mesa_is_bufferobj(ctx->DispatchIndirectBuffer)) {
> >        _mesa_error(ctx, GL_INVALID_OPERATION,
> >                    "%s: no buffer bound to DISPATCH_INDIRECT_BUFFER", name);
> > @@ -967,11 +987,6 @@ valid_dispatch_indirect(struct gl_context *ctx,
> >        return GL_FALSE;
> >     }
> >  
> > -   /* From the ARB_compute_shader specification:
> > -    *
> > -    * "An INVALID_OPERATION error is generated if this command sources data
> > -    *  beyond the end of the buffer object [...]"
> > -    */
> >     if (ctx->DispatchIndirectBuffer->Size < end) {
> >        _mesa_error(ctx, GL_INVALID_OPERATION,
> >                    "%s(DISPATCH_INDIRECT_BUFFER too small)", name);
> 
> 