Other machine-level facilities

In this section:

• Introduction
• The execution environment
• The CODE/EDOC facility
• Mixing of Intel 32-bit Assembler and Fortran
• Labels
• Referencing Fortran variables
• Literals
• Halfword and byte forms of instructions
• Using the coprocessor
• Instruction prefixes
• Other assembler facilities
• Other machine-level facilities
• Error messages
• 64-bit assembler

It is always inconvenient to have to descend to assembler, even in the form of a CODE/EDOC sequence, and a number of special Fortran constructions have been introduced for convenience.

• The intrinsic functions CCORE1, CORE1, CORE2, CORE4, FCORE4, DCORE8, XCORE10, ZCORE8, ZCORE16, and ZCORE20 are available to examine the contents of a given location.

Each function takes an INTEGER(KIND=7) argument which is an address and returns the following value at that address.

These functions may also be used on the left hand side of an assignment. For example:

CORE2(L) = CORE2(L)+1
CCORE1(P)=' '
CORE1(PTR)=123

If an argument to a routine is one of these functions the actual address is passed, for example:

INTEGER(7) L
INTRINSIC LOC,CORE2
K = 4
L = LOC(K)
CALL FRED(CORE2(L))
PRINT *,K
END
SUBROUTINE FRED(M)
M = M + 2
END

would print 6.

• A special form of the SUBROUTINE statement is available for x86 (Win32) thus:

      SPECIAL SUBROUTINE JACK

Special routines must have no arguments, and contain no preamble to set EBX%,EBP% etc. They can only really be followed by CODE/EDOC sequences, and no reference to dynamic variables must be made in such a routine. Static variables may be referenced and will use the full address form of the instruction (rather than EBX% relative). Special subroutines may contain additional entry points coded as special entries:

      SPECIAL ENTRY BILL

Special routines may not contain ordinary entry points and vice-versa. The return from a special subroutine must be via a RET instruction and not as a result of executing a RETURN or END statement. The main purpose of the special subroutine is as a routine which can be called from assembler without altering the contents of the registers.

An additional use of this facility is in conjunction with the SET_TRAP@ routine. A control break or floating point fault can take place at an arbitrary point in a program, and it is important to be able to save the registers etc. before they are overwritten. Although this can be done with an interrupt subroutine without the use of CODE/EDOC, the latter offers the ability to inspect and alter the contents of the registers if desired.

• Circular shifts are available as intrinsic functions and thus do not require the use of assembler.

• The LOC intrinsic function returns the address of its argument as a 32 bit number (64 bit for x64).