nuclear@22: ; vim:set ft=nasm:
nuclear@22: 	segment code use32
nuclear@22: 
nuclear@22: 	; void swap_buffers_asm(void *dest, void *src, int xsz, int ysz, int bpp)
nuclear@22: 	; dest	-> eax
nuclear@22: 	; src	-> edx
nuclear@22: 	; xsz	-> ebx
nuclear@22: 	; ysz	-> ecx
nuclear@22: 	; bpp	-> [ebp + 8] (after pushing ebp)
nuclear@22: 	global swap_buffers_asm_
nuclear@22: swap_buffers_asm_:
nuclear@22: 	push ebp
nuclear@22: 	mov ebp, esp
nuclear@22: 
nuclear@22: 	mov edi, eax	; let's hold dest ptr in edi, frees up eax
nuclear@22: 	mov esi, edx	; let's hold src ptr in esi, frees up edx
nuclear@22: 	; calculate pixel count -> ecx, frees up ebx
nuclear@22: 	mov eax, ebx
nuclear@22: 	mul ecx
nuclear@22: 	mov ecx, eax	; now ecx = xsz * ysz	
nuclear@22: 	
nuclear@22: 	mov eax, [ebp + 8]	; eax <- bpp
nuclear@22: 	cmp eax, 32
nuclear@22: 	je .bpp32
nuclear@22: 	cmp eax, 24
nuclear@22: 	je .bpp24
nuclear@22: 	cmp eax, 16
nuclear@22: 	je .bpp16
nuclear@22: 	; invalid bpp, ignore
nuclear@22: 	jmp .done
nuclear@22: 
nuclear@22: .bpp32:				; 32bit block transfer, no conversion
nuclear@22: 	rep movsd		; esi, edi, and ecx already loaded, just go...
nuclear@22: 	jmp .done
nuclear@22: 	
nuclear@22: .bpp24:				; 32bpp -> 24bpp conversion (LSB-first), 1 byte overrun!
nuclear@22: 	movsd			; transfer a full 32bit chunk and inc esi,edi by 4
nuclear@22: 	dec edi			; backtrack dest one byte after last transfer
nuclear@22: 	dec ecx
nuclear@22: 	jnz .bpp24
nuclear@22: 	jmp .done
nuclear@22: 
nuclear@22: .bpp16:				; fuck 16bpp for now (TODO)
nuclear@22: .done:
nuclear@22: 	pop ebp
nuclear@22: 	ret