gistfile1.txt

Byte shenanigans, little endian
uint32_t p(uint32_t pix) { uint8_t argb[4]; argb[3-1]=pix>>24; argb[3-2]=pix>>16; argb[3-3]=pix>>8; argb[3-0]=pix>>0; return *(uint32_t*)argb; }
_Z1pj:                # edi=1234
 mov eax, edi         # edi=1234 eax=1234
 mov edx, edi         # edi=1234 eax=1234 edx=1234
 shr eax, 24          # edi=1234 eax=0001 edx=1234
 shr edx, 16          # edi=1234 eax=0001 edx=0012
 sal eax, 16          # edi=1234 eax=0100 edx=0012
 mov ah, dl           # edi=1234 eax=0120 edx=0012
 mov edx, edi         # edi=1234 eax=0120 edx=1234
 sal edi, 24          # edi=4000 eax=0120 edx=1234
 shr edx, 8           # edi=4000 eax=0120 edx=0123
 mov al, dl           # edi=4000 eax=0123 edx=0123
 and eax, 16777215    # edi=4000 eax=0123 edx=0123 (useless opcode?)
 or eax, edi          # edi=4000 eax=4123 edx=0123
 ret                  # return = 4123

Byte shenanigans, big endian (I just removed the ^3s; still actually LE, but the registers have no endianness until you store them)
uint32_t p(uint32_t pix) { uint8_t argb[4]; argb[1]=pix>>24; argb[2]=pix>>16; argb[3]=pix>>8; argb[0]=pix>>0; return *(uint32_t*)argb; }
_Z1pj:             # edi=1234
 mov eax, edi      # edi=1234 eax=1234
 mov ecx, edi      # edi=1234 eax=1234 ecx=1234
 mov rdx, rdi      # edi=1234 eax=1234 ecx=1234 edx=1234
 shr eax, 24       # edi=1234 eax=0001 ecx=1234 edx=1234
 and ecx, 16711680 # edi=1234 eax=0001 ecx=0200 edx=1234
 movzx edx, dh     # edi=1234 eax=0001 ecx=0200 edx=0003 (nice catch, gcc)
 sal eax, 8        # edi=1234 eax=0010 ecx=0200 edx=0003
 sal edx, 24       # edi=1234 eax=0010 ecx=0200 edx=3000
 or eax, ecx       # edi=1234 eax=0210 ecx=0200 edx=3000
 or eax, edx       # edi=1234 eax=3210 ecx=0200 edx=3000
 mov al, dil       # edi=1234 eax=3214 ecx=0200 edx=3000
 ret               # return = 3214

Optimized
uint32_t p(uint32_t pix) { return pix<<24 | pix>>8; }
_Z1pj:        # edi=1234 eax=FFFF
 mov eax, edi # edi=1234 eax=1234
 ror eax, 8   # edi=1234 eax=4123
 ret          # return = 4123