why moving float to xmm0 lead to output 0

Asked Sep 16 '16 at 15:37

Active Sep 16 '16 at 15:37

Viewed 86 times

Want to print floating-point value via printf in assembly

    segment .data
float_fmt   db  "%f", 0xa, 0
fp  dd  1.1

    segment .text
    global  main
    extern  printf
main:
    push    rbp
    mov rbp, rsp
    lea rdi, [float_fmt]
    movss   xmm0, [fp]
    mov eax, 1
    call    printf
    leave
    ret

I want to print single-precision floating-point number(dd, 4 bytes), however, printf prints 0.000000 Changing

fp dd 1.1

fp dq 1.1 ; store double-precision floating-point(dq, 8 bytes)

and

movss xmm0, [fp]

movsd xmm0, [fp] ; move double-precision floating-point (8 bytes)

solves the problem. Could you explain why?

asked Sep 16 '16 at 15:37

Bulat M.

2

floats varargs are promoted to double. – Margaret Bloom Sep 16 '16 at 15:47
In this case how to print exactly floats(4 bytes) with format string "%f" not doubles? Should one always cast float to double before passing in xmm* registers? – Bulat M. Sep 16 '16 at 15:48
@BulatM.: yes, you can't tell printf to interpret an arg as a float, only double or long double, so you should use cvtss2sd instead of movss. – Peter Cordes Sep 16 '16 at 15:55
Is double 4 bytes in size and long double 8 bytes in size? – Bulat M. Sep 16 '16 at 16:02
2

double is 8 bytes, long double 10 bytes (if IIRC) – Margaret Bloom Sep 16 '16 at 16:59

why moving float to xmm0 lead to output 0

0 Answers0