Intel's 8087 chip hides a surprisingly complex secret in its simplest instruction
Microcode in Intel's 8087 floating-point chip: the scale instruction
The 8087 floating-point coprocessor, introduced in 1980, set the standard for modern floating-point arithmetic. Ken Shirriff, part of the Opcode Collective, reverse-engineers the microcode for the FSCALE instruction—expected to be trivial—and finds it uses over 140 micro-instructions and three levels of subroutine calls to handle special cases. The analysis reveals the chip's internal datapath and a hidden feature.
I figured that FSCALE was a simple, almost trivial instruction that would be straightforward to understand and explain. Spoiler: it is not simple.
- garganzol
100x speed improvement of math operations by 8087 is not an overestimation. The difference for apps relying on math was crazy back then. I experienced this first-hand on my 80286 machine, where it was 3-second vs 300-second calculation results.
One neat feature of 8087 instruction set is that it can be interspersed with x86 instructions in the code stream, giving you a simultaneous access to two processor chips working in parallel. This combo forms a real asymmetrical multi-processor system with certain opportunities for hardware-assisted code parallelization. If a thoughtful instruction scheduling is used, floating operations executed by 8087 work in parallel with the usual integer x86 code.
- cogman10
x87 is such a weird architecture. It was designed the same way you'd design a chip for a scientific calculator. Heck, It's almost a perfect fit for an HP RPN calculator.
But for a compiler to target, it's just so painful. It's so different from almost all other ways CPUs work. There's a reason both CPU and compilers prefer to avoid x87 when possible and use regular SIMD (SSE/AVX) instead.
Also, the arbitrary "Oh, and the registers are 80 bits wide" is also just one of those weird "Where did that number come from?".
- kens
Author here for your 8087 questions...
- finn888
The jump from a simple scale instruction to 140 micro-instructions makes the special-case machinery much more concrete.