National Instruments 320685D-01 00ProRef.book : 06chap04.fm Page 6 Monday, March 9, 1998 3:23 PM

Chapter 4 Windows3.1 Compiler/Linker Issues
LabWindows/CVI Programmer Reference Manual 4-6
© National Instruments Corporation
regs.x.si = HIWORD(size);
int86(0x31, &regs, &regs);
return regs.x.cflag;
}
After the DMA is complete, you must unlock the buffer. You can unlock the buffer using
the DPMILock function, if you set regs.x.ax to 0x601, instead of 0x600.
• If you compile the DLL with the /FPi or /FPc switches or with no /FP switches
(/FPiis the default), the DLL uses the WIN87EM.DLL floating point emulator.
LabWindows/CVI does not use WIN87EM.DLL. If the DLL uses WIN87EM.DLL, use the
following strategy in the DLL to prevent conflicts:
1. Structure the code so that all functions that perform any floating-point math have
known entry and exit points. Ideally, specify a particular set of exported entry
points as the only ways into the floating-point code.
2. Call the Windows SDK function FPInit in each of these entry points. Store the
previous signal handler in a function pointer.
3. If the DLL has its own exception handler, call signal to register the DLL’s own
signal handler.
4. Perform the floating-point math.
5. Upon exiting through one of the well-defined DLL exit points, call the Windows
SDK function FPTerm to restore the previous exception handler and terminate
the DLL’s use of WIN87EM.DLL.
typedef void (*LPFNSIGNALPROC) (int, int);
/* prototypes for functions in WIN87EM.d11 */
LPFNSIGNALPROC PASCAL_FPInit (void);
VOID PASCAL_FPTerm (LPFNSIGNALPROC);
void DllFunction (void)
{LPFNSIGNALPROC OldFPHandler;
/* save the floating point state, and setup the */
/* floating point exception handler for this DLL.*/
OldFPHandler = _FPInit ();
signal ( SIGFPE, DLLsFPEHandler); /* optional */
.
.
.
00ProRef.book : 06chap04.fm Page 6 Monday, March 9, 1998 3:23 PM