Difference between revisions of "Code Generation/Duplication of Floating Point Numbers on the Stack"

Revision as of 08:59, 2 May 2018

The following pure-Postfix program illustrates duplication of double-precision floating point numbers on the stack.

The C equivalent would be (roughly):

<c> int main() {

 extern void printd(double);
 double d, e;
 d = e = 3.3e-2;
 printd(d + e);
 return 0;

} </c>

Note that this program's code does not necessarily correspond to any particular code generator output.

Postfix code
<asm> -- generation of the "_main" function TEXT ALIGN GLOBAL _main, FUNC LABEL _main ENTER 16 ; d@-8 and e@-16 -- put double literal in RODATA RODATA ALIGN LABEL _L123 SDOUBLE 3.3e-2 -- load literal onto stack TEXT ADDR _L123 LDDOUBLE -- to perform the first assignment (to "e"), duplicate the value on the stack DUP64 LOCAL -16 ; write value to "e" STDOUBLE -- to perform the second assignment (to "d"), duplicate the value on the stack DUP64 LOCAL -8 ; write value to "d" STDOUBLE -- the assignments are an instruction trash the value left on the stack TRASH 8 -- now to perform the add operation LOCAL -8 LDDOUBLE ; load "d" LOCAL -16 LDDOUBLE ; load "e" DADD ; leaves result on the stack EXTERN printd CALL printd TRASH 8 ; argument value no longer needed INT 0 STFVAL32 LEAVE RET </asm>

Assuming that the code above is in file dbl.pf, the following commands would compile and run the program:

pf2asm dbl.pf
yasm -felf dbl.asm
ld -o dbl dbl.o -lrts
./dbl

@@ Line 27: / Line 27: @@
 ALIGN
 LABEL _L123
-DOUBLE 3.3e-2
+SDOUBLE 3.3e-2
 ;;-- load literal onto stack
 TEXT
 ADDR _L123
-DLOAD
+LDDOUBLE
 ;;-- to perform the first assignment (to "e"), duplicate the value on the stack
-DDUP
+DUP64
 LOCAL -16  ; write value to "e"
-DSTORE
+STDOUBLE
 ;;-- to perform the second assignment (to "d"), duplicate the value on the stack
-DDUP
+DUP64
 LOCAL -8  ; write value to "d"
-DSTORE
+STDOUBLE
 ;;-- the assignments are an instruction: trash the value left on the stack
@@ Line 51: / Line 51: @@
 ;;-- now to perform the add operation
 LOCAL -8
-DLOAD   ; load "d"
+LDDOUBLE   ; load "d"
 LOCAL -16
-DLOAD ; load "e"
+LDDOUBLE ; load "e"
 DADD   ; leaves result on the stack
@@ Line 63: / Line 63: @@
 INT 0
-POP
+STFVAL32
 LEAVE
 RET