Gentle ping to Jason, as he probably missed that I was trying (but failed) to make a question. See below (also skipped most of the original email)
On 18/01/18 13:31, Alejandro Piñeiro wrote: > > + > +#include "nir_spirv.h" > + > +#include "vtn_private.h" > +#include "spirv_info.h" > + > +static bool > +vtn_validate_preamble_instruction(struct vtn_builder *b, SpvOp > opcode, > + const uint32_t *w, unsigned count) > > > I think you could probably re-use all of > vtn_handle_preamble_instruction. It would do a bit more than strictly > needed (like handle capabilities) but I don't see any harm in it. >> Ok, will try to re-use it. > Well, I tried, and here the situation: as the validation is doing the > barely minimum to check for the errors defined at the method > glSpecializeShader, we are also passing it the barely minimum parameters > needed. So we are not passing spirv_to_nir_options. So if we try to > reuse vtn_handle_preamble_instruction during the validation, we start to > get several "Unsupported SPIR-V capabilities" vtn_warnings. So the > option is passing the spirv_to_nir_options here too, or just keep the > simplified version that this patch already includes. What option would you prefer? Pass the spirv_to_nir_options to the validation method in order to be able to reuse vtn_handle_preamble_instruction, or as it is not really needed for this validation, not pass the spirv_to_nir_options and keep a simplified version of such method in order to avoid those vtn_warnings? BR _______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org https://lists.freedesktop.org/mailman/listinfo/mesa-dev
