x87 FPUs (8087) were ended after release of Intel 80386 computer - x87 is the name for separete FPU of older processors. Further generations of Intel processors had FPU built-in. x86 is a general name for ALL PC processors (including Pentium and AMD ones as well). Current generations of processors are marked 'x64' usually. But they are still based on 'x86' architecture.
The fact that newer processors have no stack-based floating point processing does not mean its precision faded away that much. We still have 64-bit floating point numbers and that is pretty enough for most tasks.
As for the 'handmade optimizations' it is very possible that with the new architecture (with its comparably huge number of floating point registers) modern optimizing complier will be able to generate code that is faster than handmade assembler code. Number of registers available for optimizer means much.
In 1989 the i486 was released and it had an x86 unit and and x87 co-processor on the die. The instruction set from 387 co-pro migrated to the x86. This is why we talk ever since about x87 instructions for non vector FPU.
This process of new CPUs absorbing functions which were hitherto off the CPU (eg Athlon 64/Opteron with MMU on die) will no doubt continue ;-)
Considering how hot they are getting it is only a matter of time till the Toaster co - pro is integrated on die :-)
This topic was last updated 180 days ago, and thus it was archived. Replying is disabled for this topic.