anv: Enable KHR_shader_integer_dot_product
Open questions:
Should we mark the integerDotProduct8Bit*Accelerated
? It's only the 4 element vector cases that get automatically turned into the DP4A
instruction.
Should we mark all of the (non-64-bit) AccumulatingSaturating
instructions as accelerated? On all platforms? Applications are unlikely to use the "add with saturate" opcodes from VK_INTEL_shader_integer_functions2
, so, technically, this would "[provide] a performance advantage over user-provided code composed from elementary instructions..."