test_shader_instructions unconditionally tests a few double precision ops

vkd3dclosed

HansKristian-Work/vkd3d-proton#871 · opened 2021-10-25 by werman · updated 2021-11-09 · 1 comments · github

Wwerman 2021-10-25 github

    uint_tests[] =
    {
        {&ps_dadd,      {.d = {{2.5, 0.0}}},  {{0x20a80000, 0x400c0000, 0x20500000, 0x40120000}}},
        {&ps_dmax,      {.d = {{2.5, 0.0}}},  {{0x40200000, 0x40200000, 0x40200000, 0x40200000}}},
        ...

Supporting float64 isn't a hard requirement, right? There is a DoublePrecisionFloatShaderOps feature flag.

In Turnip we don't even have HW support for int64. I don't think there is a point in enabling float64 emulation for us.

HHansKristian-Work maintainer 2021-10-25 github

It is a feature flag, yes, so we should check that flag before running any 64-bit tests.

Nothing extracted yet.