Reverse engineering the Intel 8087 microcode for the FSCALE instruction
- Sources: post, HN discussion
- Summary: The post traces the 8087 microcode that implements FSCALE, which scales a floating-point value by a power of two. The nominally trivial exponent addition runs to over 140 micro-instructions and three levels of subroutine calls once empty, denormal, infinity and NaN operands are handled.
- Why it matters: Operand classification, not the arithmetic, dominates the microcode path, which is what an instruction's nominal description hides.