I read in wiki that ARM has a unique half precision standard and i want to know whether it's also true for mobile GPU like mali. I asked this because i do encounter a wired issue about half precision where same shade program with same data end with different result in different mobile gpu. The number i calculate in shader will be out of half precision range so it was supposed to be a NaN but instead its final output is NOT in some mali GPU while(i said some not all, cause in Mali-T830 the result is the same as other mobile gpu like in adreno or powervr while in Mali-G76, the result is different) . And i just curious whether it is related the ARM half standard I mentioned about. In wiki, it said"ARM alternative half-precision
ARM processors support (via a floating point control register bit) an "alternative half-precision" format, which does away with the special case for an exponent value of 31 (111112). It is almost identical to the IEEE format, but there is no encoding for infinity or NaNs; instead, an exponent of 31 encodes normalized numbers in the range 65536 to 131008.
"(arm community warn me that i shouldn't put link in my post so i just paste the content here, just google "Half-precision floating-point format wiki" and you will find the original post)