Merge pull request #465 from PoroCYon/eeeeeeeeeeeeeeeeeeeeeeee

Add/improve some assembly versions

e.asm

@@ -1,17 +1,56 @@
+; compile with:
+; $ [ny]asm -felf(32|64) -oe.o e.asm
+; $ (gcc|clang) -m(32|64) -oe e.o -nostdlib -nostartfiles
+
 section     .text
-global      _start                              
+global      _start
 
-_start:                                         
+%if __BITS__ == 32
+%define r(n) e%+n
+%define SYS_write 4
+%define rarg0 ebx
+%define rarg1 ecx
+%define rarg2 edx
+%define syscall int 0x80
+%else
+%define r(n) r%+n
+%define SYS_write 1
+%define rarg0 rdi
+%define rarg1 rsi
+%define rarg2 rdx
+default rel
+%endif
 
-	mov rax, 1
+; size of a Linux pipe buffer
-	mov rdi, 1
+%define PIPE_SIZE 0x10000
-	mov rsi, msg
+%define STDOUT_FILENO 1
-	mov rdx, len
-	loop:
-	syscall
-	jmp loop                             
 
-section     .data
+; Instead of simply storing a char in .rodata and write(2)-ing it
+; over and over again, we first fill a buffer full of e's, and *then*
+; write the entire buffer. This is much faster than the first option,
+; because we only need to issue a syscall once every 65536 bytes. (Remember
+; that doing a syscall requires the kernel to handle an interrupt etc etc etc.)
 
-msg: db "e"                 
+_start:
-len: equ $ - msg                             
+        ; allocate space for the message
+        mov r(cx), PIPE_SIZE
+        mov r(bx), r(cx) ; we'll need it later
+        sub r(sp), r(cx)
+
+        ; quick memset(3)
+        mov al, 'e'
+        mov r(di), r(sp)
+        rep stosb
+
+        ; push+pop is actually a smaller encoding than mov for ints that fit within 8 bit
+        push STDOUT_FILENO
+        pop rarg0
+        mov rarg1, r(sp)
+        mov rarg2, r(bx)
+
+.loop:
+        ; set this within the loop because the syscall's exit code is placed in r(ax)
+        push SYS_write
+        pop r(ax)
+        syscall
+        jmp short .loop

e.linux.arm.S

@@ -0,0 +1,36 @@
+#define SYS_write 4
+#define STDOUT 1
+#define PIPE_SIZE 0x10000
+
+.arm
+.align 4
+
+.section .text, "ax", %progbits
+
+type _start, %function
+globl _start
+_start:
+    mov r1, #PIPE_SIZE
+    
+    ldr r3, =('e'|('e'<<8)|('e'<<16)|('e'<<24))
+    mov r4, r3
+    mov r5, r3
+    mov r6, r3
+    
+.Lloop:
+    push {r3-r6}
+    sub r1, #(4*4)
+    cmp r1, #0
+    bgt .Lloop
+    
+    mov r7, #SYS_write
+    mov r1, sp
+    mov r2, #PIPE_SIZE
+
+.Lcall:
+    mov r0, #STDIN
+    swi #0
+    b .Lcall
+    
+.align 4
+.pool

e.mbr.asm

@@ -0,0 +1,31 @@
+bits 16
+org 0x7C00
+
+_start:
+    ; enable cursor
+    xor cx, cx
+    mov ch, 00100000b
+    mov ah, 1
+    int 0x10
+    
+    ; move cursor to top
+    xor dx, dx
+    xor bx, bx
+    inc ah
+    int 0x10
+    
+    ; print 1 char
+    mov ax, 'e'|(0x0E<<8)
+ .loop:
+    int 0x10
+    jmp short .loop
+
+END:
+    times 0x200-2-(END-_start) db 'e'
+    db 0x55,0xAA
+    
+stack:
+
+%if END-_org > 0x200-2
+%error "Not enough space!"
+%endif