--- /Users/yonas/ffmpeg-r29748-orig/libavcodec/arm/dsputil_arm_s.S 2009-10-03 20:58:30.000000000 -0400 +++ /Users/yonas/ffmpeg-r29748-orig/libavcodec/arm/dsputil_arm_s.S 2009-10-03 19:13:34.000000000 -0400 @@ -22,7 +22,7 @@ #include "config.h" #include "asm.S" - preserve8 + //preserve8 #if !HAVE_PLD .macro pld reg @@ -36,67 +36,64 @@ add r0, r0, r1 bne ff_prefetch_arm bx lr - .endfunc + @.endfunc #endif -.macro ALIGN_QWORD_D shift, Rd0, Rd1, Rd2, Rd3, Rn0, Rn1, Rn2, Rn3, Rn4 - mov \Rd0, \Rn0, lsr #(\shift * 8) - mov \Rd1, \Rn1, lsr #(\shift * 8) - mov \Rd2, \Rn2, lsr #(\shift * 8) - mov \Rd3, \Rn3, lsr #(\shift * 8) - orr \Rd0, \Rd0, \Rn1, lsl #(32 - \shift * 8) - orr \Rd1, \Rd1, \Rn2, lsl #(32 - \shift * 8) - orr \Rd2, \Rd2, \Rn3, lsl #(32 - \shift * 8) - orr \Rd3, \Rd3, \Rn4, lsl #(32 - \shift * 8) +.macro ALIGN_QWORD_D + mov $1, $5, lsr #($0 * 8) + mov $2, $6, lsr #($0 * 8) + mov $3, $7, lsr #($0 * 8) + mov $4, $8, lsr #($0 * 8) + orr $1, $1, $6, lsl #(32 - $0 * 8) + orr $2, $2, $7, lsl #(32 - $0 * 8) + orr $3, $3, $8, lsl #(32 - $0 * 8) + orr $4, $4, $9, lsl #(32 - $0 * 8) .endm -.macro ALIGN_DWORD shift, R0, R1, R2 - mov \R0, \R0, lsr #(\shift * 8) - orr \R0, \R0, \R1, lsl #(32 - \shift * 8) - mov \R1, \R1, lsr #(\shift * 8) - orr \R1, \R1, \R2, lsl #(32 - \shift * 8) +.macro ALIGN_DWORD + mov $1, $1, lsr #($0 * 8) + orr $1, $1, $2, lsl #(32 - $0 * 8) + mov $2, $2, lsr #($0 * 8) + orr $2, $2, $3, lsl #(32 - $0 * 8) .endm -.macro ALIGN_DWORD_D shift, Rdst0, Rdst1, Rsrc0, Rsrc1, Rsrc2 - mov \Rdst0, \Rsrc0, lsr #(\shift * 8) - mov \Rdst1, \Rsrc1, lsr #(\shift * 8) - orr \Rdst0, \Rdst0, \Rsrc1, lsl #(32 - (\shift * 8)) - orr \Rdst1, \Rdst1, \Rsrc2, lsl #(32 - (\shift * 8)) +.macro ALIGN_DWORD_D + mov $1, $3, lsr #($0 * 8) + mov $2, $4, lsr #($0 * 8) + orr $1, $1, $4, lsl #(32 - ($0 * 8)) + orr $2, $2, $5, lsl #(32 - ($0 * 8)) .endm - -.macro RND_AVG32 Rd0, Rd1, Rn0, Rn1, Rm0, Rm1, Rmask +.macro RND_AVG32 @ Rd = (Rn | Rm) - (((Rn ^ Rm) & ~0x01010101) >> 1) @ Rmask = 0xFEFEFEFE @ Rn = destroy - eor \Rd0, \Rn0, \Rm0 - eor \Rd1, \Rn1, \Rm1 - orr \Rn0, \Rn0, \Rm0 - orr \Rn1, \Rn1, \Rm1 - and \Rd0, \Rd0, \Rmask - and \Rd1, \Rd1, \Rmask - sub \Rd0, \Rn0, \Rd0, lsr #1 - sub \Rd1, \Rn1, \Rd1, lsr #1 + eor $0, $2, $4 + eor $1, $3, $5 + orr $2, $2, $4 + orr $3, $3, $5 + and $0, $0, $6 + and $1, $1, $6 + sub $0, $2, $0, lsr #1 + sub $1, $3, $1, lsr #1 .endm - -.macro NO_RND_AVG32 Rd0, Rd1, Rn0, Rn1, Rm0, Rm1, Rmask +.macro NO_RND_AVG32 @ Rd = (Rn & Rm) - (((Rn ^ Rm) & ~0x01010101) >> 1) @ Rmask = 0xFEFEFEFE @ Rn = destroy - eor \Rd0, \Rn0, \Rm0 - eor \Rd1, \Rn1, \Rm1 - and \Rn0, \Rn0, \Rm0 - and \Rn1, \Rn1, \Rm1 - and \Rd0, \Rd0, \Rmask - and \Rd1, \Rd1, \Rmask - add \Rd0, \Rn0, \Rd0, lsr #1 - add \Rd1, \Rn1, \Rd1, lsr #1 + eor $0, $2, $4 + eor $1, $3, $5 + and $2, $2, $4 + and $3, $3, $5 + and $0, $0, $6 + and $1, $1, $6 + add $0, $2, $0, lsr #1 + add $1, $3, $1, lsr #1 .endm - .macro JMP_ALIGN tmp, reg - ands \tmp, \reg, #3 - bic \reg, \reg, #3 + ands $0, $1, #3 + bic $1, $1, #3 beq 1f - subs \tmp, \tmp, #1 + subs $0, $0, #1 beq 2f - subs \tmp, \tmp, #1 + subs $0, $0, #1 beq 3f b 4f .endm @@ -107,51 +104,51 @@ @ void func(uint8_t *block, const uint8_t *pixels, int line_size, int h) @ block = word aligned, pixles = unaligned pld [r1] - push {r4-r11, lr} + stmfd sp!, {r4-r11, lr} @ R14 is also called LR JMP_ALIGN r5, r1 1: - ldm r1, {r4-r7} + ldmia r1, {r4-r7} add r1, r1, r2 - stm r0, {r4-r7} + stmia r0, {r4-r7} pld [r1] subs r3, r3, #1 add r0, r0, r2 bne 1b - pop {r4-r11, pc} + ldmfd sp!, {r4-r11, pc} .align 5 2: - ldm r1, {r4-r8} + ldmia r1, {r4-r8} add r1, r1, r2 - ALIGN_QWORD_D 1, r9, r10, r11, r12, r4, r5, r6, r7, r8 + ALIGN_QWORD_D 1, r9, r10, r11, r12, r4, r5, r6, r7, r8 pld [r1] subs r3, r3, #1 - stm r0, {r9-r12} + stmia r0, {r9-r12} add r0, r0, r2 bne 2b - pop {r4-r11, pc} + ldmfd sp!, {r4-r11, pc} .align 5 3: - ldm r1, {r4-r8} + ldmia r1, {r4-r8} add r1, r1, r2 - ALIGN_QWORD_D 2, r9, r10, r11, r12, r4, r5, r6, r7, r8 + ALIGN_QWORD_D 2, r9, r10, r11, r12, r4, r5, r6, r7, r8 pld [r1] subs r3, r3, #1 - stm r0, {r9-r12} + stmia r0, {r9-r12} add r0, r0, r2 bne 3b - pop {r4-r11, pc} + ldmfd sp!, {r4-r11, pc} .align 5 4: - ldm r1, {r4-r8} + ldmia r1, {r4-r8} add r1, r1, r2 - ALIGN_QWORD_D 3, r9, r10, r11, r12, r4, r5, r6, r7, r8 + ALIGN_QWORD_D 3, r9, r10, r11, r12, r4, r5, r6, r7, r8 pld [r1] subs r3, r3, #1 - stm r0, {r9-r12} + stmia r0, {r9-r12} add r0, r0, r2 bne 4b - pop {r4-r11,pc} - .endfunc + ldmfd sp!, {r4-r11,pc} + @.endfunc @ ---------------------------------------------------------------- .align 5 @@ -159,51 +156,51 @@ @ void func(uint8_t *block, const uint8_t *pixels, int line_size, int h) @ block = word aligned, pixles = unaligned pld [r1] - push {r4-r5,lr} + stmfd sp!, {r4-r5,lr} @ R14 is also called LR JMP_ALIGN r5, r1 1: - ldm r1, {r4-r5} + ldmia r1, {r4-r5} add r1, r1, r2 subs r3, r3, #1 pld [r1] - stm r0, {r4-r5} + stmia r0, {r4-r5} add r0, r0, r2 bne 1b - pop {r4-r5,pc} + ldmfd sp!, {r4-r5,pc} .align 5 2: - ldm r1, {r4-r5, r12} + ldmia r1, {r4-r5, r12} add r1, r1, r2 - ALIGN_DWORD 1, r4, r5, r12 + ALIGN_DWORD 1, r4, r5, r12 pld [r1] subs r3, r3, #1 - stm r0, {r4-r5} + stmia r0, {r4-r5} add r0, r0, r2 bne 2b - pop {r4-r5,pc} + ldmfd sp!, {r4-r5,pc} .align 5 3: - ldm r1, {r4-r5, r12} + ldmia r1, {r4-r5, r12} add r1, r1, r2 - ALIGN_DWORD 2, r4, r5, r12 + ALIGN_DWORD 2, r4, r5, r12 pld [r1] subs r3, r3, #1 - stm r0, {r4-r5} + stmia r0, {r4-r5} add r0, r0, r2 bne 3b - pop {r4-r5,pc} + ldmfd sp!, {r4-r5,pc} .align 5 4: - ldm r1, {r4-r5, r12} + ldmia r1, {r4-r5, r12} add r1, r1, r2 - ALIGN_DWORD 3, r4, r5, r12 + ALIGN_DWORD 3, r4, r5, r12 pld [r1] subs r3, r3, #1 - stm r0, {r4-r5} + stmia r0, {r4-r5} add r0, r0, r2 bne 4b - pop {r4-r5,pc} - .endfunc + ldmfd sp!, {r4-r5,pc} + @.endfunc @ ---------------------------------------------------------------- .align 5 @@ -211,118 +208,128 @@ @ void func(uint8_t *block, const uint8_t *pixels, int line_size, int h) @ block = word aligned, pixles = unaligned pld [r1] - push {r4-r10,lr} - ldr r12, =0xfefefefe + stmfd sp!, {r4-r10,lr} @ R14 is also called LR + @ldr r12, =0xfefefefe + adr r12, 5f + ldr r12, [r12, #0] JMP_ALIGN r5, r1 1: - ldm r1, {r4-r5, r10} + ldmia r1, {r4-r5, r10} add r1, r1, r2 - ALIGN_DWORD_D 1, r6, r7, r4, r5, r10 + ALIGN_DWORD_D 1, r6, r7, r4, r5, r10 pld [r1] RND_AVG32 r8, r9, r4, r5, r6, r7, r12 subs r3, r3, #1 - stm r0, {r8-r9} + stmia r0, {r8-r9} add r0, r0, r2 bne 1b - pop {r4-r10,pc} + ldmfd sp!, {r4-r10,pc} .align 5 2: - ldm r1, {r4-r5, r10} + ldmia r1, {r4-r5, r10} add r1, r1, r2 - ALIGN_DWORD_D 1, r6, r7, r4, r5, r10 - ALIGN_DWORD_D 2, r8, r9, r4, r5, r10 + ALIGN_DWORD_D 1, r6, r7, r4, r5, r10 + ALIGN_DWORD_D 2, r8, r9, r4, r5, r10 pld [r1] RND_AVG32 r4, r5, r6, r7, r8, r9, r12 subs r3, r3, #1 - stm r0, {r4-r5} + stmia r0, {r4-r5} add r0, r0, r2 bne 2b - pop {r4-r10,pc} + ldmfd sp!, {r4-r10,pc} .align 5 3: - ldm r1, {r4-r5, r10} + ldmia r1, {r4-r5, r10} add r1, r1, r2 - ALIGN_DWORD_D 2, r6, r7, r4, r5, r10 - ALIGN_DWORD_D 3, r8, r9, r4, r5, r10 + ALIGN_DWORD_D 2, r6, r7, r4, r5, r10 + ALIGN_DWORD_D 3, r8, r9, r4, r5, r10 pld [r1] RND_AVG32 r4, r5, r6, r7, r8, r9, r12 subs r3, r3, #1 - stm r0, {r4-r5} + stmia r0, {r4-r5} add r0, r0, r2 bne 3b - pop {r4-r10,pc} + ldmfd sp!, {r4-r10,pc} .align 5 4: - ldm r1, {r4-r5, r10} + ldmia r1, {r4-r5, r10} add r1, r1, r2 - ALIGN_DWORD_D 3, r6, r7, r4, r5, r10 + ALIGN_DWORD_D 3, r6, r7, r4, r5, r10 pld [r1] RND_AVG32 r8, r9, r6, r7, r5, r10, r12 subs r3, r3, #1 - stm r0, {r8-r9} + stmia r0, {r8-r9} add r0, r0, r2 bne 4b - pop {r4-r10,pc} - .endfunc + ldmfd sp!, {r4-r10,pc} @@ update PC with LR content. + .align 5 +5: + .long 0xFEFEFEFE + @.endfunc .align 5 function put_no_rnd_pixels8_x2_arm, export=1 @ void func(uint8_t *block, const uint8_t *pixels, int line_size, int h) @ block = word aligned, pixles = unaligned pld [r1] - push {r4-r10,lr} - ldr r12, =0xfefefefe + stmfd sp!, {r4-r10,lr} @ R14 is also called LR + @ldr r12, =0xfefefefe + adr r12, 5f + ldr r12, [r12, #0] JMP_ALIGN r5, r1 1: - ldm r1, {r4-r5, r10} + ldmia r1, {r4-r5, r10} add r1, r1, r2 - ALIGN_DWORD_D 1, r6, r7, r4, r5, r10 + ALIGN_DWORD_D 1, r6, r7, r4, r5, r10 pld [r1] NO_RND_AVG32 r8, r9, r4, r5, r6, r7, r12 subs r3, r3, #1 - stm r0, {r8-r9} + stmia r0, {r8-r9} add r0, r0, r2 bne 1b - pop {r4-r10,pc} + ldmfd sp!, {r4-r10,pc} .align 5 2: - ldm r1, {r4-r5, r10} + ldmia r1, {r4-r5, r10} add r1, r1, r2 - ALIGN_DWORD_D 1, r6, r7, r4, r5, r10 - ALIGN_DWORD_D 2, r8, r9, r4, r5, r10 + ALIGN_DWORD_D 1, r6, r7, r4, r5, r10 + ALIGN_DWORD_D 2, r8, r9, r4, r5, r10 pld [r1] NO_RND_AVG32 r4, r5, r6, r7, r8, r9, r12 subs r3, r3, #1 - stm r0, {r4-r5} + stmia r0, {r4-r5} add r0, r0, r2 bne 2b - pop {r4-r10,pc} + ldmfd sp!, {r4-r10,pc} .align 5 3: - ldm r1, {r4-r5, r10} + ldmia r1, {r4-r5, r10} add r1, r1, r2 - ALIGN_DWORD_D 2, r6, r7, r4, r5, r10 - ALIGN_DWORD_D 3, r8, r9, r4, r5, r10 + ALIGN_DWORD_D 2, r6, r7, r4, r5, r10 + ALIGN_DWORD_D 3, r8, r9, r4, r5, r10 pld [r1] NO_RND_AVG32 r4, r5, r6, r7, r8, r9, r12 subs r3, r3, #1 - stm r0, {r4-r5} + stmia r0, {r4-r5} add r0, r0, r2 bne 3b - pop {r4-r10,pc} + ldmfd sp!, {r4-r10,pc} .align 5 4: - ldm r1, {r4-r5, r10} + ldmia r1, {r4-r5, r10} add r1, r1, r2 - ALIGN_DWORD_D 3, r6, r7, r4, r5, r10 + ALIGN_DWORD_D 3, r6, r7, r4, r5, r10 pld [r1] NO_RND_AVG32 r8, r9, r6, r7, r5, r10, r12 subs r3, r3, #1 - stm r0, {r8-r9} + stmia r0, {r8-r9} add r0, r0, r2 bne 4b - pop {r4-r10,pc} - .endfunc + ldmfd sp!, {r4-r10,pc} @@ update PC with LR content. + .align 5 +5: + .long 0xFEFEFEFE + @.endfunc @ ---------------------------------------------------------------- @@ -331,200 +338,212 @@ @ void func(uint8_t *block, const uint8_t *pixels, int line_size, int h) @ block = word aligned, pixles = unaligned pld [r1] - push {r4-r11,lr} + stmfd sp!, {r4-r11,lr} @ R14 is also called LR mov r3, r3, lsr #1 - ldr r12, =0xfefefefe + @ldr r12, =0xfefefefe + adr r12, 5f + ldr r12, [r12, #0] JMP_ALIGN r5, r1 1: - ldm r1, {r4-r5} + ldmia r1, {r4-r5} add r1, r1, r2 -6: ldm r1, {r6-r7} +6: ldmia r1, {r6-r7} add r1, r1, r2 pld [r1] RND_AVG32 r8, r9, r4, r5, r6, r7, r12 - ldm r1, {r4-r5} + ldmia r1, {r4-r5} add r1, r1, r2 - stm r0, {r8-r9} + stmia r0, {r8-r9} add r0, r0, r2 pld [r1] RND_AVG32 r8, r9, r6, r7, r4, r5, r12 subs r3, r3, #1 - stm r0, {r8-r9} + stmia r0, {r8-r9} add r0, r0, r2 bne 6b - pop {r4-r11,pc} + ldmfd sp!, {r4-r11,pc} .align 5 2: - ldm r1, {r4-r6} + ldmia r1, {r4-r6} add r1, r1, r2 pld [r1] - ALIGN_DWORD 1, r4, r5, r6 -6: ldm r1, {r7-r9} + ALIGN_DWORD 1, r4, r5, r6 +6: ldmia r1, {r7-r9} add r1, r1, r2 pld [r1] - ALIGN_DWORD 1, r7, r8, r9 + ALIGN_DWORD 1, r7, r8, r9 RND_AVG32 r10, r11, r4, r5, r7, r8, r12 - stm r0, {r10-r11} + stmia r0, {r10-r11} add r0, r0, r2 - ldm r1, {r4-r6} + ldmia r1, {r4-r6} add r1, r1, r2 pld [r1] - ALIGN_DWORD 1, r4, r5, r6 + ALIGN_DWORD 1, r4, r5, r6 subs r3, r3, #1 RND_AVG32 r10, r11, r7, r8, r4, r5, r12 - stm r0, {r10-r11} + stmia r0, {r10-r11} add r0, r0, r2 bne 6b - pop {r4-r11,pc} + ldmfd sp!, {r4-r11,pc} .align 5 3: - ldm r1, {r4-r6} + ldmia r1, {r4-r6} add r1, r1, r2 pld [r1] - ALIGN_DWORD 2, r4, r5, r6 -6: ldm r1, {r7-r9} + ALIGN_DWORD 2, r4, r5, r6 +6: ldmia r1, {r7-r9} add r1, r1, r2 pld [r1] - ALIGN_DWORD 2, r7, r8, r9 + ALIGN_DWORD 2, r7, r8, r9 RND_AVG32 r10, r11, r4, r5, r7, r8, r12 - stm r0, {r10-r11} + stmia r0, {r10-r11} add r0, r0, r2 - ldm r1, {r4-r6} + ldmia r1, {r4-r6} add r1, r1, r2 pld [r1] - ALIGN_DWORD 2, r4, r5, r6 + ALIGN_DWORD 2, r4, r5, r6 subs r3, r3, #1 RND_AVG32 r10, r11, r7, r8, r4, r5, r12 - stm r0, {r10-r11} + stmia r0, {r10-r11} add r0, r0, r2 bne 6b - pop {r4-r11,pc} + ldmfd sp!, {r4-r11,pc} .align 5 4: - ldm r1, {r4-r6} + ldmia r1, {r4-r6} add r1, r1, r2 pld [r1] - ALIGN_DWORD 3, r4, r5, r6 -6: ldm r1, {r7-r9} + ALIGN_DWORD 3, r4, r5, r6 +6: ldmia r1, {r7-r9} add r1, r1, r2 pld [r1] - ALIGN_DWORD 3, r7, r8, r9 + ALIGN_DWORD 3, r7, r8, r9 RND_AVG32 r10, r11, r4, r5, r7, r8, r12 - stm r0, {r10-r11} + stmia r0, {r10-r11} add r0, r0, r2 - ldm r1, {r4-r6} + ldmia r1, {r4-r6} add r1, r1, r2 pld [r1] - ALIGN_DWORD 3, r4, r5, r6 + ALIGN_DWORD 3, r4, r5, r6 subs r3, r3, #1 RND_AVG32 r10, r11, r7, r8, r4, r5, r12 - stm r0, {r10-r11} + stmia r0, {r10-r11} add r0, r0, r2 bne 6b - pop {r4-r11,pc} - .endfunc + ldmfd sp!, {r4-r11,pc} + .align 5 +5: + .long 0xfefefefe + + @.endfunc .align 5 function put_no_rnd_pixels8_y2_arm, export=1 @ void func(uint8_t *block, const uint8_t *pixels, int line_size, int h) @ block = word aligned, pixles = unaligned pld [r1] - push {r4-r11,lr} + stmfd sp!, {r4-r11,lr} @ R14 is also called LR mov r3, r3, lsr #1 - ldr r12, =0xfefefefe + @ldr r12, =0xfefefefe + adr r12, 5f + ldr r12, [r12, #0] JMP_ALIGN r5, r1 1: - ldm r1, {r4-r5} + ldmia r1, {r4-r5} add r1, r1, r2 -6: ldm r1, {r6-r7} +6: ldmia r1, {r6-r7} add r1, r1, r2 pld [r1] NO_RND_AVG32 r8, r9, r4, r5, r6, r7, r12 - ldm r1, {r4-r5} + ldmia r1, {r4-r5} add r1, r1, r2 - stm r0, {r8-r9} + stmia r0, {r8-r9} add r0, r0, r2 pld [r1] NO_RND_AVG32 r8, r9, r6, r7, r4, r5, r12 subs r3, r3, #1 - stm r0, {r8-r9} + stmia r0, {r8-r9} add r0, r0, r2 bne 6b - pop {r4-r11,pc} + ldmfd sp!, {r4-r11,pc} .align 5 2: - ldm r1, {r4-r6} + ldmia r1, {r4-r6} add r1, r1, r2 pld [r1] - ALIGN_DWORD 1, r4, r5, r6 -6: ldm r1, {r7-r9} + ALIGN_DWORD 1, r4, r5, r6 +6: ldmia r1, {r7-r9} add r1, r1, r2 pld [r1] - ALIGN_DWORD 1, r7, r8, r9 + ALIGN_DWORD 1, r7, r8, r9 NO_RND_AVG32 r10, r11, r4, r5, r7, r8, r12 - stm r0, {r10-r11} + stmia r0, {r10-r11} add r0, r0, r2 - ldm r1, {r4-r6} + ldmia r1, {r4-r6} add r1, r1, r2 pld [r1] - ALIGN_DWORD 1, r4, r5, r6 + ALIGN_DWORD 1, r4, r5, r6 subs r3, r3, #1 NO_RND_AVG32 r10, r11, r7, r8, r4, r5, r12 - stm r0, {r10-r11} + stmia r0, {r10-r11} add r0, r0, r2 bne 6b - pop {r4-r11,pc} + ldmfd sp!, {r4-r11,pc} .align 5 3: - ldm r1, {r4-r6} + ldmia r1, {r4-r6} add r1, r1, r2 pld [r1] - ALIGN_DWORD 2, r4, r5, r6 -6: ldm r1, {r7-r9} + ALIGN_DWORD 2, r4, r5, r6 +6: ldmia r1, {r7-r9} add r1, r1, r2 pld [r1] - ALIGN_DWORD 2, r7, r8, r9 + ALIGN_DWORD 2, r7, r8, r9 NO_RND_AVG32 r10, r11, r4, r5, r7, r8, r12 - stm r0, {r10-r11} + stmia r0, {r10-r11} add r0, r0, r2 - ldm r1, {r4-r6} + ldmia r1, {r4-r6} add r1, r1, r2 pld [r1] - ALIGN_DWORD 2, r4, r5, r6 + ALIGN_DWORD 2, r4, r5, r6 subs r3, r3, #1 NO_RND_AVG32 r10, r11, r7, r8, r4, r5, r12 - stm r0, {r10-r11} + stmia r0, {r10-r11} add r0, r0, r2 bne 6b - pop {r4-r11,pc} + ldmfd sp!, {r4-r11,pc} .align 5 4: - ldm r1, {r4-r6} + ldmia r1, {r4-r6} add r1, r1, r2 pld [r1] - ALIGN_DWORD 3, r4, r5, r6 -6: ldm r1, {r7-r9} + ALIGN_DWORD 3, r4, r5, r6 +6: ldmia r1, {r7-r9} add r1, r1, r2 pld [r1] - ALIGN_DWORD 3, r7, r8, r9 + ALIGN_DWORD 3, r7, r8, r9 NO_RND_AVG32 r10, r11, r4, r5, r7, r8, r12 - stm r0, {r10-r11} + stmia r0, {r10-r11} add r0, r0, r2 - ldm r1, {r4-r6} + ldmia r1, {r4-r6} add r1, r1, r2 pld [r1] - ALIGN_DWORD 3, r4, r5, r6 + ALIGN_DWORD 3, r4, r5, r6 subs r3, r3, #1 NO_RND_AVG32 r10, r11, r7, r8, r4, r5, r12 - stm r0, {r10-r11} + stmia r0, {r10-r11} add r0, r0, r2 bne 6b - pop {r4-r11,pc} - .endfunc + ldmfd sp!, {r4-r11,pc} + .align 5 +5: + .long 0xfefefefe + @.endfunc - .ltorg + @.ltorg @ ---------------------------------------------------------------- + .macro RND_XY2_IT align, rnd @ l1= (a & 0x03030303) + (b & 0x03030303) ?(+ 0x02020202) @ h1= ((a & 0xFCFCFCFCUL) >> 2) + ((b & 0xFCFCFCFCUL) >> 2) @@ -528,65 +547,68 @@ .macro RND_XY2_IT align, rnd @ l1= (a & 0x03030303) + (b & 0x03030303) ?(+ 0x02020202) @ h1= ((a & 0xFCFCFCFCUL) >> 2) + ((b & 0xFCFCFCFCUL) >> 2) -.if \align == 0 - ldm r1, {r6-r8} -.elseif \align == 3 - ldm r1, {r5-r7} +.if $0 == 0 + ldmia r1, {r6-r8} +.elseif $0 == 3 + ldmia r1, {r5-r7} .else - ldm r1, {r8-r10} + ldmia r1, {r8-r10} .endif add r1, r1, r2 pld [r1] -.if \align == 0 - ALIGN_DWORD_D 1, r4, r5, r6, r7, r8 -.elseif \align == 1 - ALIGN_DWORD_D 1, r4, r5, r8, r9, r10 - ALIGN_DWORD_D 2, r6, r7, r8, r9, r10 -.elseif \align == 2 - ALIGN_DWORD_D 2, r4, r5, r8, r9, r10 - ALIGN_DWORD_D 3, r6, r7, r8, r9, r10 -.elseif \align == 3 - ALIGN_DWORD_D 3, r4, r5, r5, r6, r7 +.if $0 == 0 + ALIGN_DWORD_D 1, r4, r5, r6, r7, r8 +.elseif $0 == 1 + ALIGN_DWORD_D 1, r4, r5, r8, r9, r10 + ALIGN_DWORD_D 2, r6, r7, r8, r9, r10 +.elseif $0 == 2 + ALIGN_DWORD_D 2, r4, r5, r8, r9, r10 + ALIGN_DWORD_D 3, r6, r7, r8, r9, r10 +.elseif $0 == 3 + ALIGN_DWORD_D 3, r4, r5, r5, r6, r7 .endif - ldr r14, =0x03030303 + @ldr r14, =0x03030303 + ldr r14, [r12, #0] tst r3, #1 and r8, r4, r14 and r9, r5, r14 and r10, r6, r14 and r11, r7, r14 - andeq r14, r14, r14, \rnd #1 + andeq r14, r14, r14, $1 #1 add r8, r8, r10 add r9, r9, r11 - ldr r12, =0xfcfcfcfc >> 2 addeq r8, r8, r14 addeq r9, r9, r14 - and r4, r12, r4, lsr #2 - and r5, r12, r5, lsr #2 - and r6, r12, r6, lsr #2 - and r7, r12, r7, lsr #2 + @ldr r12, =0xfcfcfcfc >> 2 + ldr r14, [r12, #4] @ 0xFCFCFCFC >> 2 + and r4, r14, r4, lsr #2 + and r5, r14, r5, lsr #2 + and r6, r14, r6, lsr #2 + and r7, r14, r7, lsr #2 add r10, r4, r6 add r11, r5, r7 subs r3, r3, #1 .endm .macro RND_XY2_EXPAND align, rnd - RND_XY2_IT \align, \rnd -6: push {r8-r11} - RND_XY2_IT \align, \rnd - pop {r4-r7} + RND_XY2_IT $0, $1 +6: stmfd sp!, {r8-r11} + RND_XY2_IT $0, $1 + ldmfd sp!, {r4-r7} add r4, r4, r8 add r5, r5, r9 - ldr r14, =0x0f0f0f0f + @ldr r14, =0x0f0f0f0f + ldr r14, [r12, #8] add r6, r6, r10 add r7, r7, r11 and r4, r14, r4, lsr #2 and r5, r14, r5, lsr #2 add r4, r4, r6 add r5, r5, r7 - stm r0, {r4-r5} + stmia r0, {r4-r5} add r0, r0, r2 bge 6b - pop {r4-r11,pc} + ldmfd sp!, {r4-r11,pc} .endm .align 5 @@ -594,7 +616,8 @@ @ void func(uint8_t *block, const uint8_t *pixels, int line_size, int h) @ block = word aligned, pixles = unaligned pld [r1] - push {r4-r11,lr} @ R14 is also called LR + stmfd sp!, {r4-r11,lr} @ R14 is also called LR + adr r12, 5f JMP_ALIGN r5, r1 1: RND_XY2_EXPAND 0, lsl @@ -610,14 +633,21 @@ .align 5 4: RND_XY2_EXPAND 3, lsl - .endfunc + + .align 5 +5: + .long 0x03030303 + .long 0xFCFCFCFC >> 2 + .long 0x0F0F0F0F .align 5 function put_no_rnd_pixels8_xy2_arm, export=1 @ void func(uint8_t *block, const uint8_t *pixels, int line_size, int h) @ block = word aligned, pixles = unaligned pld [r1] - push {r4-r11,lr} + stmfd sp!, {r4-r11,lr} @ R14 is also called LR + adr r12, 5f + JMP_ALIGN r5, r1 1: RND_XY2_EXPAND 0, lsr @@ -633,12 +663,19 @@ .align 5 4: RND_XY2_EXPAND 3, lsr - .endfunc + + .align 5 +5: + .long 0x03030303 + .long 0xFCFCFCFC >> 2 + .long 0x0F0F0F0F + @.endfunc .align 5 @ void ff_add_pixels_clamped_ARM(int16_t *block, uint8_t *dest, int stride) function ff_add_pixels_clamped_ARM, export=1 - push {r4-r10} + @push {r4-r10} + stmfd sp!, {r4-r10} mov r10, #8 1: ldr r4, [r1] /* load dest */ @@ -721,6 +758,8 @@ add r1, r1, r2 bne 1b - pop {r4-r10} + @pop {r4-r10} + ldmfd sp!, {r4-r10} bx lr - .endfunc + @.endfunc + diff -wBNru /Users/yonas/ffmpeg-r29748-orig/libavcodec/arm/simple_idct_armv6.S /Users/yonas/ffmpeg-r29748-iphone/libavcodec/arm/simple_idct_armv6.S --- /Users/yonas/ffmpeg-r29748-orig/libavcodec/arm/simple_idct_armv6.S 2009-01-30 19:37:45.000000000 -0500 +++ /Users/yonas/ffmpeg-r29748-iphone/libavcodec/arm/simple_idct_armv6.S 2009-01-30 20:30:18.000000000 -0500 @@ -59,9 +59,9 @@ Output in registers r4--r11 */ - .macro idct_row shift + .macro idct_row ldr lr, w46 /* lr = W4 | (W6 << 16) */ - mov r1, #(1<<(\shift-1)) + mov r1, #(1<<($0-1)) smlad r4, r2, ip, r1 smlsd r7, r2, ip, r1 ldr ip, w13 /* ip = W1 | (W3 << 16) */ @@ -100,10 +100,10 @@ Output in registers r4--r11 */ - .macro idct_row4 shift + .macro idct_row4 ldr lr, w46 /* lr = W4 | (W6 << 16) */ ldr r10,w57 /* r10 = W5 | (W7 << 16) */ - mov r1, #(1<<(\shift-1)) + mov r1, #(1<<($0-1)) smlad r4, r2, ip, r1 smlsd r7, r2, ip, r1 ldr ip, w13 /* ip = W1 | (W3 << 16) */ @@ -138,28 +138,29 @@ shift = right-shift amount Input/output in registers r4--r11 */ - .macro idct_finish_shift shift + .macro idct_finish_shift add r3, r4, r8 /* r3 = A0 + B0 */ sub r2, r4, r8 /* r2 = A0 - B0 */ - mov r4, r3, asr #\shift - mov r8, r2, asr #\shift + mov r4, r3, asr $0 + mov r8, r2, asr $0 sub r3, r5, r9 /* r3 = A1 + B1 */ add r2, r5, r9 /* r2 = A1 - B1 */ - mov r5, r3, asr #\shift - mov r9, r2, asr #\shift + mov r5, r3, asr $0 + mov r9, r2, asr $0 add r3, r6, r10 /* r3 = A2 + B2 */ sub r2, r6, r10 /* r2 = A2 - B2 */ - mov r6, r3, asr #\shift - mov r10,r2, asr #\shift + mov r6, r3, asr $0 + mov r10,r2, asr $0 add r3, r7, r11 /* r3 = A3 + B3 */ sub r2, r7, r11 /* r2 = A3 - B3 */ - mov r7, r3, asr #\shift - mov r11,r2, asr #\shift + mov r7, r3, asr $0 + mov r11,r2, asr $0 .endm + /* Compute final part of IDCT single row, saturating results at 8 bits. shift = right-shift amount @@ -165,26 +166,26 @@ shift = right-shift amount Input/output in registers r4--r11 */ - .macro idct_finish_shift_sat shift + .macro idct_finish_shift_sat add r3, r4, r8 /* r3 = A0 + B0 */ sub ip, r4, r8 /* ip = A0 - B0 */ - usat r4, #8, r3, asr #\shift - usat r8, #8, ip, asr #\shift + usat r4, #8, r3, asr $0 + usat r8, #8, ip, asr $0 sub r3, r5, r9 /* r3 = A1 + B1 */ add ip, r5, r9 /* ip = A1 - B1 */ - usat r5, #8, r3, asr #\shift - usat r9, #8, ip, asr #\shift + usat r5, #8, r3, asr $0 + usat r9, #8, ip, asr $0 add r3, r6, r10 /* r3 = A2 + B2 */ sub ip, r6, r10 /* ip = A2 - B2 */ - usat r6, #8, r3, asr #\shift - usat r10,#8, ip, asr #\shift + usat r6, #8, r3, asr $0 + usat r10,#8, ip, asr $0 add r3, r7, r11 /* r3 = A3 + B3 */ sub ip, r7, r11 /* ip = A3 - B3 */ - usat r7, #8, r3, asr #\shift - usat r11,#8, ip, asr #\shift + usat r7, #8, r3, asr $0 + usat r11,#8, ip, asr $0 .endm /* @@ -193,7 +194,7 @@ r1 = dest */ function idct_row_armv6 - push {lr} + str lr, [sp, #-4]! ldr lr, [r0, #12] /* lr = row[7,5] */ ldr ip, [r0, #4] /* ip = row[6,4] */ @@ -203,7 +204,7 @@ cmpeq lr, r3 cmpeq lr, r2, lsr #16 beq 1f - push {r1} + str r1, [sp, #-4]! ldr ip, w42 /* ip = W4 | (W2 << 16) */ cmp lr, #0 beq 2f @@ -213,8 +214,8 @@ 2: idct_row4 ROW_SHIFT -3: pop {r1} - idct_finish_shift ROW_SHIFT +3: ldr r1, [sp], #4 + idct_finish_shift #ROW_SHIFT strh r4, [r1] strh r5, [r1, #(16*2)] @@ -225,7 +226,7 @@ strh r9, [r1, #(16*5)] strh r8, [r1, #(16*7)] - pop {pc} + ldr pc, [sp], #4 1: mov r2, r2, lsl #3 strh r2, [r1] @@ -236,8 +237,7 @@ strh r2, [r1, #(16*3)] strh r2, [r1, #(16*5)] strh r2, [r1, #(16*7)] - pop {pc} - .endfunc + ldr pc, [sp], #4 /* Compute IDCT of single column, read as row. @@ -245,14 +245,14 @@ r1 = dest */ function idct_col_armv6 - push {r1, lr} + stmfd sp!, {r1, lr} ldr r2, [r0] /* r2 = row[2,0] */ ldr ip, w42 /* ip = W4 | (W2 << 16) */ ldr r3, [r0, #8] /* r3 = row[3,1] */ idct_row COL_SHIFT - pop {r1} - idct_finish_shift COL_SHIFT + ldr r1, [sp], #4 + idct_finish_shift #COL_SHIFT strh r4, [r1] strh r5, [r1, #(16*1)] @@ -263,8 +263,7 @@ strh r9, [r1, #(16*6)] strh r8, [r1, #(16*7)] - pop {pc} - .endfunc + ldr pc, [sp], #4 /* Compute IDCT of single column, read as row, store saturated 8-bit. @@ -273,14 +272,14 @@ r2 = line size */ function idct_col_put_armv6 - push {r1, r2, lr} + stmfd sp!, {r1, r2, lr} ldr r2, [r0] /* r2 = row[2,0] */ ldr ip, w42 /* ip = W4 | (W2 << 16) */ ldr r3, [r0, #8] /* r3 = row[3,1] */ idct_row COL_SHIFT - pop {r1, r2} - idct_finish_shift_sat COL_SHIFT + ldmfd sp!, {r1, r2} + idct_finish_shift_sat #COL_SHIFT strb r4, [r1], r2 strb r5, [r1], r2 @@ -293,8 +292,7 @@ sub r1, r1, r2, lsl #3 - pop {pc} - .endfunc + ldr pc, [sp], #4 /* Compute IDCT of single column, read as row, add/store saturated 8-bit. @@ -303,13 +301,13 @@ r2 = line size */ function idct_col_add_armv6 - push {r1, r2, lr} + stmfd sp!, {r1, r2, lr} ldr r2, [r0] /* r2 = row[2,0] */ ldr ip, w42 /* ip = W4 | (W2 << 16) */ ldr r3, [r0, #8] /* r3 = row[3,1] */ idct_row COL_SHIFT - pop {r1, r2} + ldmfd sp!, {r1, r2} idct_finish ldrb r3, [r1] @@ -348,86 +346,83 @@ sub r1, r1, r2, lsl #3 - pop {pc} - .endfunc + ldr pc, [sp], #4 /* Compute 8 IDCT row transforms. func = IDCT row->col function width = width of columns in bytes */ - .macro idct_rows func width - bl \func + .macro idct_rows + bl $0 add r0, r0, #(16*2) - add r1, r1, #\width - bl \func + add r1, r1, $1 + bl $0 add r0, r0, #(16*2) - add r1, r1, #\width - bl \func + add r1, r1, $1 + bl $0 add r0, r0, #(16*2) - add r1, r1, #\width - bl \func + add r1, r1, $1 + bl $0 sub r0, r0, #(16*5) - add r1, r1, #\width - bl \func + add r1, r1, $1 + bl $0 add r0, r0, #(16*2) - add r1, r1, #\width - bl \func + add r1, r1, $1 + bl $0 add r0, r0, #(16*2) - add r1, r1, #\width - bl \func + add r1, r1, $1 + bl $0 add r0, r0, #(16*2) - add r1, r1, #\width - bl \func + add r1, r1, $1 + bl $0 sub r0, r0, #(16*7) .endm /* void ff_simple_idct_armv6(DCTELEM *data); */ function ff_simple_idct_armv6, export=1 - push {r4-r11, lr} + stmfd sp!, {r4-r11, lr} sub sp, sp, #128 mov r1, sp - idct_rows idct_row_armv6, 2 + idct_rows _idct_row_armv6, #2 mov r1, r0 mov r0, sp - idct_rows idct_col_armv6, 2 + idct_rows _idct_col_armv6, #2 add sp, sp, #128 - pop {r4-r11, pc} - .endfunc + ldmfd sp!, {r4-r11, pc} /* ff_simple_idct_add_armv6(uint8_t *dest, int line_size, DCTELEM *data); */ function ff_simple_idct_add_armv6, export=1 - push {r0, r1, r4-r11, lr} + stmfd sp!, {r0, r1, r4-r11, lr} sub sp, sp, #128 mov r0, r2 mov r1, sp - idct_rows idct_row_armv6, 2 + idct_rows _idct_row_armv6, #2 mov r0, sp ldr r1, [sp, #128] ldr r2, [sp, #(128+4)] - idct_rows idct_col_add_armv6, 1 + idct_rows _idct_col_add_armv6, #1 add sp, sp, #(128+8) - pop {r4-r11, pc} - .endfunc + ldmfd sp!, {r4-r11, pc} /* ff_simple_idct_put_armv6(uint8_t *dest, int line_size, DCTELEM *data); */ function ff_simple_idct_put_armv6, export=1 - push {r0, r1, r4-r11, lr} + stmfd sp!, {r0, r1, r4-r11, lr} sub sp, sp, #128 mov r0, r2 mov r1, sp - idct_rows idct_row_armv6, 2 + idct_rows _idct_row_armv6, #2 mov r0, sp ldr r1, [sp, #128] ldr r2, [sp, #(128+4)] - idct_rows idct_col_put_armv6, 1 + idct_rows _idct_col_put_armv6, #1 add sp, sp, #(128+8) - pop {r4-r11, pc} - .endfunc + ldmfd sp!, {r4-r11, pc} + diff -wBNru /Users/yonas/ffmpeg-r29748-orig/libavutil/internal.h /Users/yonas/ffmpeg-r29748-iphone/libavutil/internal.h --- /Users/yonas/ffmpeg-r29748-orig/libavutil/internal.h 2009-08-07 19:25:56.000000000 -0400 +++ /Users/yonas/ffmpeg-r29748-iphone/libavutil/internal.h 2009-08-07 19:39:45.000000000 -0400 @@ -140,14 +140,14 @@ #elif HAVE_ARMV6 && HAVE_INLINE_ASM static inline av_const int FASTDIV(int a, int b) { - int r, t; - __asm__ volatile("cmp %3, #2 \n\t" - "ldr %1, [%4, %3, lsl #2] \n\t" - "lsrle %0, %2, #1 \n\t" - "smmulgt %0, %1, %2 \n\t" - : "=&r"(r), "=&r"(t) : "r"(a), "r"(b), "r"(ff_inverse)); + int r; + __asm__ volatile("cmp %2, #0 \n\t" + "smmul %0, %1, %2 \n\t" + "rsblt %0, %0, #0 \n\t" + : "=r"(r) : "r"(a), "r"(ff_inverse[b])); return r; } + #elif ARCH_ARM && HAVE_INLINE_ASM static inline av_const int FASTDIV(int a, int b) { diff -wBNru /Users/yonas/ffmpeg-r29748-orig/libavcodec/arm.S /Users/yonas/ffmpeg-r29748-iphone/libavcodec/arm.S --- /Users/yonas/ffmpeg-r29748-iphone/libavcodec/arm/asm.S 2009-10-03 21:28:49.000000000 -0400 +++ /Users/yonas/ffmpeg-r29748-iphone/libavcodec/arm/asm.S 2009-10-03 21:38:54.000000000 -0400 @@ -35,13 +35,8 @@ .endm .macro function name, export=0 -.if \export - .global EXTERN_ASM\name -EXTERN_ASM\name: -.endif -ELF .type \name, %function - .func \name -\name: + .globl _$0 + _$0: .endm .macro movrel rd, val