On Fri Mar 17 16:07:31 2023 +0000, Matteo Bruni wrote:
I guess the first patch could be split further by introducing another test that directly uses a few of the same opcodes. Not a big deal though, it's fine as a single patch. More importantly, I think we should actually test in test_effect_value_expression() the new test passes you're introducing. I'm curious if d3d10 does any conversion when storing an int into a float state (i.e. the blend factor), but otherwise it's probably nicer to set the sample mask from the effect pass (it should be the third argument to SetBlendState()) and check it in the test, as that is supposed to be an integer value. Unless that won't work yet for some reason?
I don't understand the question about converting int to float. Such expression generates explicit itof's.