Chapter 3, Problem 3.73HW

Explanation of Solution

Given assembly code:

x in %xmm0

find_range:

vxorps %xmm1, %xmm1, %xmm1

vucomiss %xmm1, %xmm0

ja .L5

vucomiss %xmm1, %xmm0

jp .L8

movl $1, %eax

je .L3

.L8:

vucomiss %xmm1, %xmm0

setbe %al

movzbl %al, %eax

addl $2, %eax

ret

.L5:

movl $0, %eax

.L3:

Rep;ret

Explanation:

• The assembly function “find_range” illustrates conditional branching in floating point.
• It compares a value “x” with 0.
• If value is less than 0, then result is negative.
• If value is greater than 0, then result is positive.
• If value equals 0, then result is zero.
• The instruction “vxorps %xmm1, %xmm1, %xmm1” sets value of “%xmm1” to 0.
• The instruction “vucomiss %xmm1, %xmm0” compares values.
• The instruction “jp .L8” jumps to label “.L8”.
• The instruction “ja .L5” jumps to label “.L5” if first value is greater than second.
• The instruction “movl $0, %eax” moves immediate value to register “%eax”.
• The instruction “movl $1, %eax” moves immediate value to register “%eax”.

  Function find_range:

  #Define function

range_t find_range(float x) {

  __asm__(

      "vxorps %xmm1, %xmm1, %xmm1\n\t"

      "vucomiss %xmm1, %xmm0\n\t"

      "jp .P\n\t"

      "ja .A\n\t"

      "jb .B\n\t"

      "je .E\n\t"

      "...