Connor Abbott <cwabbo...@gmail.com> writes: > A few comments here... > > 1. We should really split this up into a patch that adds > nir_shader_compiler_options (which would be empty at this point) and > the nir_algebraic infrastructure for conditionalizing transforms > (actually, maybe we should split these up into two patches) and a > patch which adds lower_pow with the same name as the current patch.
Split version sent. > 2. While I see the performance reasons for pre-computing all the > conditions up-front before running the pass, I'm not sure that's the > best approach long-term. Soon, we'll have to support the precise and > invariant keywords, which are essentially per-instruction and affect > which algebraic optimizations we can do. We may have to do something > more clever for those, but in the meantime I'm not sure we'll want > this optimization down the road... so it's probably better to wait > until we get to that point and do the simple thing for now. The expression has to appear in the code somewhere, and the places I consume the precomputed expression flag are each looping over an array of structures with different flags used. Where are you proposing I put the expression, instead?
signature.asc
Description: PGP signature
_______________________________________________ mesa-dev mailing list mesa-dev@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/mesa-dev
