@#########################################################
@
@ Scott R. Armstrong
@ UIUC CS433 Spring 2007
@ ASM Performance Improvement Project
@ Due by May 8, 2007
@
@#########################################################
        .file   "primes.c"
        .section        .rodata
        .text
        .align  2
        .global main
        .type   main, %function
main:
        @ args = 0, pretend = 0, frame = 28
        @ frame_needed = 1, uses_anonymous_args = 0
        mov     ip, sp
        stmfd   sp!, {fp, ip, lr, pc}
        sub     fp, ip, #4
        sub     sp, sp, #28
        
        @@ set up z (limit of numbers to check for prime-ness)
        mov     r3, #99328      @ compiler split z into 99328+672
        add     r3, r3, #672    @ uses shifted values to fit IMM field
                                @ or load a different IMM into r6
				@ assigned here to replace I/O for
				@ the purposes of this project
        mov     r6, r3          @ move value into z
        mov     r5, #10         @ y, initialized to 10
				@ to demonstrate loop unrolling

.L1:                            @@ y-loop-start
        cmp     r5, r6
        bgt     .L1_END

@@ Start of unrolled-loops section
@@ n+1 loop

	add	r5, r5, #1	@ n + 1

        mov     r7, #1          @@ prime, initialized to true
        mov     r4, #2          @@ x, initialized to 2

.L2_n1:                            @@ x-loop-start
	mov	r2, r5, asr #1	@@ y/2 (y DIV 2, no need for fractions)
	cmp     r4, r2          @@ Compare r2 (y/2) with x
        bgt     .L2_END_n1

@ DIV/MOD function - based on ARM Developer Suite 1.1 Divide Function
@ used here instead of linking a MOD function

        mov     r9, r4
        mov     r10, r5         @ r10 will contain any remainder
.L_DIV_n1:
        mov     r2, r9
        cmp     r2, r10, lsr #1
.L_DIV_1_n1:
        movls   r2, r2, lsl #1@ 
        cmp     r2, r10, lsr #1
        bls     .L_DIV_1_n1
        mov     r0, #0          @ Initialize quotient
.L_DIV_2_n1:
        cmp     r10, r2
        subcs   r10, r10, r2
        adc     r0, r0, r0
        mov     r2, r2, lsr #1
        cmp     r2, r9
        bhs     .L_DIV_2_n1

@ End DIV/MOD function

        cmp     r10, #0		@ if no remainder
        moveq   r7, #0		@ set prime to false
        cmp     r7, #1
        addeq   r4, r4, #1	@ if prime, increment x
        beq     .L2_n1		@ ...and branch to L2 (repeat)

.L2_END_n1:
        cmp	r7, #1
        moveq   r8, r5		@ store last prime number in r8

@@ n+3 loop

	add	r5, r5, #2	@ n + 3

        mov     r7, #1          @@ prime, initialized to true
        mov     r4, #2          @@ x, initialized to 2

.L2_n3:                            @@ x-loop-start
	mov	r2, r5, asr #1	@@ y/2 (y DIV 2, no need for fractions)
	cmp     r4, r2          @@ Compare r2 (y/2) with x
        bgt     .L2_END_n3

@ DIV/MOD function - based on ARM Developer Suite 1.1 Divide Function
@ used here instead of linking a MOD function

        mov     r9, r4
        mov     r10, r5         @ r10 will contain any remainder
.L_DIV_n3:
        mov     r2, r9
        cmp     r2, r10, lsr #1
.L_DIV_1_n3:
        movls   r2, r2, lsl #1@ 
        cmp     r2, r10, lsr #1
        bls     .L_DIV_1_n3
        mov     r0, #0          @ Initialize quotient
.L_DIV_2_n3:
        cmp     r10, r2
        subcs   r10, r10, r2
        adc     r0, r0, r0
        mov     r2, r2, lsr #1
        cmp     r2, r9
        bhs     .L_DIV_2_n3

@ End DIV/MOD function

        cmp     r10, #0		@ if no remainder
        moveq   r7, #0		@ set prime to false
        cmp     r7, #1
        addeq   r4, r4, #1	@ if prime, increment x
        beq     .L2_n3		@ ...and branch to L2 (repeat)

.L2_END_n3:
        cmp	r7, #1
        moveq   r8, r5		@ store last prime number in r8

@@ n+7 loop

	add	r5, r5, #4	@ n + 7

        mov     r7, #1          @@ prime, initialized to true
        mov     r4, #2          @@ x, initialized to 2

.L2_n7:                            @@ x-loop-start
	mov	r2, r5, asr #1	@@ y/2 (y DIV 2, no need for fractions)
	cmp     r4, r2          @@ Compare r2 (y/2) with x
        bgt     .L2_END_n7

@ DIV/MOD function - based on ARM Developer Suite 1.1 Divide Function
@ used here instead of linking a MOD function

        mov     r9, r4
        mov     r10, r5         @ r10 will contain any remainder
.L_DIV_n7:
        mov     r2, r9
        cmp     r2, r10, lsr #1
.L_DIV_1_n7:
        movls   r2, r2, lsl #1@ 
        cmp     r2, r10, lsr #1
        bls     .L_DIV_1_n7
        mov     r0, #0          @ Initialize quotient
.L_DIV_2_n7:
        cmp     r10, r2
        subcs   r10, r10, r2
        adc     r0, r0, r0
        mov     r2, r2, lsr #1
        cmp     r2, r9
        bhs     .L_DIV_2_n7

@ End DIV/MOD function

        cmp     r10, #0		@ if no remainder
        moveq   r7, #0		@ set prime to false
        cmp     r7, #1
        addeq   r4, r4, #1	@ if prime, increment x
        beq     .L2_n7		@ ...and branch to L2 (repeat)

.L2_END_n7:
        cmp	r7, #1
        moveq   r8, r5		@ store last prime number in r8

@@ n+9 loop
	add	r5, r5, #2	@ n + 9

        mov     r7, #1          @@ prime, initialized to true
        mov     r4, #2          @@ x, initialized to 2

.L2_n9:                            @@ x-loop-start
	mov	r2, r5, asr #1	@@ y/2 (y DIV 2, no need for fractions)
	cmp     r4, r2          @@ Compare r2 (y/2) with x
        bgt     .L2_END_n9

@ DIV/MOD function - based on ARM Developer Suite 1.1 Divide Function
@ used here instead of linking a MOD function

        mov     r9, r4
        mov     r10, r5         @ r10 will contain any remainder
.L_DIV_n9:
        mov     r2, r9
        cmp     r2, r10, lsr #1
.L_DIV_1_n9:
        movls   r2, r2, lsl #1@ 
        cmp     r2, r10, lsr #1
        bls     .L_DIV_1_n9
        mov     r0, #0          @ Initialize quotient
.L_DIV_2_n9:
        cmp     r10, r2
        subcs   r10, r10, r2
        adc     r0, r0, r0
        mov     r2, r2, lsr #1
        cmp     r2, r9
        bhs     .L_DIV_2_n9

@ End DIV/MOD function

        cmp     r10, #0		@ if no remainder
        moveq   r7, #0		@ set prime to false
        cmp     r7, #1
        addeq   r4, r4, #1	@ if prime, increment x
        beq     .L2_n9		@ ...and branch to L2 (repeat)

.L2_END_n9:
        cmp	r7, #1
        moveq   r8, r5		@ store last prime number in r8

@@ End of unrolled-loops section

        add     r5, r5, #1	@ add 1 to y, r5 adds total 10
	b       .L1

.L1_END:
        @mov     r3, #0
        @mov     r0, r3
        sub     sp, fp, #12
        ldmfd   sp, {fp, sp, pc}
        .align  2
        .size   main, .-main
        .ident  "GCC: (GNU) 4.1.0"

.end