[Mesa-dev] [PATCH 17/19] mesa: Add notes about remaining NV_vertex_program code.

[Roland Scheidegger]

Am 10.10.2012 17:48, schrieb Brian Paul:
> On 10/10/2012 08:34 AM, Roland Scheidegger wrote:
>> Am 10.10.2012 03:06, schrieb Eric Anholt:
>>> ---
>>>   src/mesa/main/nvprogram.c   |    4 ++++
>>>   src/mesa/main/vtxfmt.c      |    2 +-
>>>   src/mesa/vbo/vbo_exec_api.c |    4 ++++
>>>   3 files changed, 9 insertions(+), 1 deletion(-)
>>>
>>> diff --git a/src/mesa/main/nvprogram.c b/src/mesa/main/nvprogram.c
>>> index 6ad846e3..4b24dde 100644
>>> --- a/src/mesa/main/nvprogram.c
>>> +++ b/src/mesa/main/nvprogram.c
>>> @@ -205,6 +205,10 @@ _mesa_GetProgramStringNV(GLuint id, GLenum
>>> pname, GLubyte *program)
>>>
>>>   /**
>>>    * Get a vertex array attribute pointer.
>>> + *
>>> + * While the function first appeared in GL_NV_vertex_program, it's
>>> + * aliased for use by GL_ARB_vertex_program.
>>> + *
>>>    * \note Not compiled into display lists.
>>>    * \note Called from the GL API dispatcher.
>>>    */
>>> diff --git a/src/mesa/main/vtxfmt.c b/src/mesa/main/vtxfmt.c
>>> index aa6cf28..462cc88 100644
>>> --- a/src/mesa/main/vtxfmt.c
>>> +++ b/src/mesa/main/vtxfmt.c
>>> @@ -135,7 +135,7 @@ install_vtxfmt(struct gl_context *ctx, struct
>>> _glapi_table *tab,
>>>                                                 
>>> vfmt->DrawTransformFeedbackStreamInstanced);
>>>      }
>>>
>>> -   /* GL_NV_vertex_program */
>>> +   /* Originally for GL_NV_vertex_program, this is also used by
>>> dlist.c */
>>>      if (ctx->API == API_OPENGL) {
>>>         SET_VertexAttrib1fNV(tab, vfmt->VertexAttrib1fNV);
>>>         SET_VertexAttrib1fvNV(tab, vfmt->VertexAttrib1fvNV);
>>> diff --git a/src/mesa/vbo/vbo_exec_api.c b/src/mesa/vbo/vbo_exec_api.c
>>> index 52b363d..2ddb715 100644
>>> --- a/src/mesa/vbo/vbo_exec_api.c
>>> +++ b/src/mesa/vbo/vbo_exec_api.c
>>> @@ -958,6 +958,10 @@ static void vbo_exec_vtxfmt_init( struct
>>> vbo_exec_context *exec )
>>>         vfmt->VertexAttrib4fvARB = vbo_VertexAttrib4fvARB;
>>>      }
>>>
>>> +   /* Note that VertexAttrib4fNV is used from dlist.c and
>>> api_arrayelt.c so
>>> +    * they can have a single entrypoint for updating any of the legacy
>>> +    * attribs.
>>> +    */
>>>      vfmt->VertexAttrib1fNV = vbo_VertexAttrib1fNV;
>>>      vfmt->VertexAttrib1fvNV = vbo_VertexAttrib1fvNV;
>>>      vfmt->VertexAttrib2fNV = vbo_VertexAttrib2fNV;
>>>
>>
>> I think the idea of removing NV_vertex_program is quite good, though I
>> wonder if these functions shouldn't be renamed and moved instead of
>> keeping them as is?
>> Supporting both NV_vertex_program ARB_vertex_program always seems to
>> have caused headaches (not least because of the different aliasing of
>> attribs), though for a long time ARB_vertex_program in mesa actually was
>> code dependent quite a bit on NV_vertex_program.
>> IIRC celestia used to use NV_vertex_program but it's quite possible it
>> no longer does.
> 
> I seem to recall testing Celestia with our VMware gallium driver in the
> past, and we've never supported GL_NV_vp/fp with gallium, so I think
> we're OK there too.

Oh yes that was just an example of an app which could use it if
supported. But certainly not like viewperf it actually checked what's
available. I don't know if there was a version which only supported
NV_vp and not ARB_vp but even if there was that has to be decades ago...

Roland