Vol. 1 D-19
GUIDELINES FOR WRITING X87 FPU EXCEPTION HANDLERS
virtual machine subsystem. The MS-DOS program is set up in a virtual machine that provides a virtualized inter-
rupt table. The MS-DOS application hooks interrupt 16 in the virtual machine in the normal way. A numeric excep-
tion will trap to the kernel via the real INT 16 residing in the kernel at ring 0.
The INT 16 handler in the kernel then locates the correct MS-DOS virtual machine, and reflects the interrupt to the
virtual machine monitor. The virtual machine monitor then emulates an interrupt by jumping through the address
in the virtualized interrupt table, eventually reaching the application’s numeric exception handler.
D.3.6.5
Special Considerations for Operating Systems that Support Streaming SIMD Extensions
Operating systems that support Streaming SIMD Extensions instructions introduced with the Pentium III processor
should use the FXSAVE and FXRSTOR instructions to save and restore the new SIMD floating-point instruction
register state as well as the floating-point state. Such operating systems must consider the following issues:
1. Enlarged state save area — FNSAVE/FRSTOR instructions operate on a 94-byte or 108-byte memory region,
depending on whether they are executed in 16-bit or 32-bit mode. The FXSAVE/FXRSTOR instructions operate
on a 512-byte memory region.
2. Alignment requirements — FXSAVE/FXRSTOR instructions require the memory region on which they
operate to be 16-byte aligned (refer to the individual instruction instructions descriptions in Chapter 3 of the
Intel® 64 and IA-32 Architectures Software Developer’s Manual, Volume 2A, for information about exceptions
generated if the memory region is not aligned).
3. Maintaining compatibility with legacy applications/libraries — The operating system changes to
support Streaming SIMD Extensions must be invisible to legacy applications or libraries that deal only with
floating-point instructions. The layout of the memory region operated on by the FXSAVE/FXRSTOR instructions
is different from the layout for the FNSAVE/FRSTOR instructions. Specifically, the format of the x87 FPU tag
word and the length of the various fields in the memory region is different. Care must be taken to return the
x87 FPU state to a legacy application (e.g., when reporting FP exceptions) in the format it expects.
4. Instruction semantic differences — There are some semantic differences between the way the FXSAVE and
FSAVE/FNSAVE instructions operate. The FSAVE/FNSAVE instructions clear the x87 FPU after they save the
state while the FXSAVE instruction saves the x87 FPU/Streaming SIMD Extensions state but does not clear it.
Operating systems that use FXSAVE to save the x87 FPU state before making it available for another thread
(e.g., during thread switch time) should take precautions not to pass a “dirty” x87 FPU to another application.
D.4
DIFFERENCES FOR HANDLERS USING NATIVE MODE
The 8087 has an INT pin which it asserts when an unmasked exception occurs. But there is no interrupt input pin
in the 8086 or 8088 dedicated to its attachment, nor an interrupt vector in the 8086 or 8088 specific for an x87 FPU
error assertion. Beginning with the Intel 286 and Intel 287 hardware, a connection was dedicated to support the
x87 FPU exception and interrupt vector 16 was assigned to it.
D.4.1
Origin with the Intel 286 and Intel 287, and Intel386 and Intel 387 Processors
The Intel 286 and Intel 287, and Intel386 and Intel 387 processor/coprocessor pairs are each provided with
ERROR# pins that are recommended to be connected between the processor and x87 FPU. If this is done, when an
unmasked x87 FPU exception occurs, the x87 FPU records the exception, and asserts its ERROR# pin. The
processor recognizes this active condition of the ERROR# status line immediately before execution of the next
WAIT or x87 FPU instruction (except for the no-wait type) in its instruction stream, and branches to the handler of
interrupt 16. Thus an x87 FPU exception will be handled before any other x87 FPU instruction (after the one
causing the error) is executed (except for no-wait instructions, which will be executed without triggering the x87
FPU exception interrupt, but it will remain pending).
Using the dedicated INT 16 for x87 FPU exception handling is referred to as the native mode. It is the simplest
approach, and the one recommended most highly by Intel.