CHAPTER 24. 64-BIT VALUES IN 32-BIT ENVIRONMENT CHAPTER 24. 64-BIT VALUES IN 32-BIT ENVIRONMENT
ret_f_div:
sub esp, 28
mov eax, DWORD PTR [esp+40]
mov edx, DWORD PTR [esp+44]
mov DWORD PTR [esp+8], eax
mov eax, DWORD PTR [esp+32]
mov DWORD PTR [esp+12], edx
mov edx, DWORD PTR [esp+36]
mov DWORD PTR [esp], eax
mov DWORD PTR [esp+4], edx
call ___udivdi3 ; unsigned division
add esp, 28
ret
_f_rem:
sub esp, 28
mov eax, DWORD PTR [esp+40]
mov edx, DWORD PTR [esp+44]
mov DWORD PTR [esp+8], eax
mov eax, DWORD PTR [esp+32]
mov DWORD PTR [esp+12], edx
mov edx, DWORD PTR [esp+36]
mov DWORD PTR [esp], eax
mov DWORD PTR [esp+4], edx
call ___umoddi3 ; unsigned modulo
add esp, 28
ret
GCC does the expected, but the multiplication code is inlined right in the function, thinking it could be more efficient. GCC
has different library function names:D on page 902.
24.3.2 ARM.
Keil for Thumb mode inserts library subroutine calls:
Listing 24.11: Optimizing Keil 6/2013 (Thumb mode)||f_mul|| PROC
PUSH {r4,lr}
BL __aeabi_lmul
POP {r4,pc}
ENDP
||f_div|| PROC
PUSH {r4,lr}
BL __aeabi_uldivmod
POP {r4,pc}
ENDP
||f_rem|| PROC
PUSH {r4,lr}
BL __aeabi_uldivmod
MOVS r0,r2
MOVS r1,r3
POP {r4,pc}
ENDP
Keil for ARM mode, on the other hand, is able to produce 64-bit multiplication code:
Listing 24.12: Optimizing Keil 6/2013 (ARM mode)||f_mul|| PROC
PUSH {r4,lr}
UMULL r12,r4,r0,r2
MLA r1,r2,r1,r4
MLA r1,r0,r3,r1
MOV r0,r12