Aras Pranckevičius
Aras Pranckevičius
do_mat_op_to_vec could be called after constant folding. The kind of optimizations this does: Original code: vec3 N = normalize((sys_WorldMatrix \* vec4( vec3( 1.0, 0.0, 0.0 ), 0 )).xyz); Current: tmpvar_1...
Have to use "inout" instead of "in out".
So far I've found isinf() and isnan() will cause it to fail, even though there are GLSL equivalents.
HLSL2GLSL cannot figure out what sampler type to cast to if the parameter and argument sampler type differ: // You can't pass a generic sampler type ("sampler") to this float4...
GLSL does not have a generic sampler type ("sampler") and that's not usually a problem as HLSL2GLSL can usually figure out what a sampler's type needs to be from it's...
Requires explicit type constructors. Note that this only applies to arrays; non-array types now work fine either way. // This fails: static const float3 my_array[2] = { { 1, 1,...
The `-Gis` command line argument should effectively mark all variables behave as if they had `precise` HLSL modifier in them (that's what #2861 seems to be doing in DXIL output)....
At Unity we're finding that just updating the underlying ISPC compiler to 1.14 version gives a small compression speed increase (3-5% for BC7 & BC6H). That's cool! However the source...
- popcount() implementations - workaround the lack of std::aligned_alloc - fix cpu_coded.inl wrongly using #ifdef instead of #if for OpenMP flags (the flag is defined, just to zero). With OpenMP...