diff --git a/libavcodec/arm/simple_idct_armv6.S b/libavcodec/arm/simple_idct_armv6.S index 4f3330d223..d276fdb570 100644 --- a/libavcodec/arm/simple_idct_armv6.S +++ b/libavcodec/arm/simple_idct_armv6.S @@ -52,158 +52,158 @@ w57: .long W57 /* Compute partial IDCT of single row. shift = left-shift amount - a1 = source address - a3 = row[2,0] <= 2 cycles - a4 = row[3,1] + r0 = source address + r2 = row[2,0] <= 2 cycles + r3 = row[3,1] ip = w42 <= 2 cycles - Output in registers v1--v8 + Output in registers r4--r11 */ .macro idct_row shift ldr lr, [pc, #(w46-.-8)] /* lr = W4 | (W6 << 16) */ - mov a2, #(1<<(\shift-1)) - smlad v1, a3, ip, a2 - smlsd v4, a3, ip, a2 + mov r1, #(1<<(\shift-1)) + smlad r4, r2, ip, r1 + smlsd r7, r2, ip, r1 ldr ip, [pc, #(w13-.-8)] /* ip = W1 | (W3 << 16) */ - ldr v7, [pc, #(w57-.-8)] /* v7 = W5 | (W7 << 16) */ - smlad v2, a3, lr, a2 - smlsd v3, a3, lr, a2 - - smuad v5, a4, ip /* v5 = B0 = W1*row[1] + W3*row[3] */ - smusdx fp, a4, v7 /* fp = B3 = W7*row[1] - W5*row[3] */ - ldr lr, [a1, #12] /* lr = row[7,5] */ - pkhtb a3, ip, v7, asr #16 /* a4 = W7 | (W3 << 16) */ - pkhbt a2, ip, v7, lsl #16 /* a2 = W1 | (W5 << 16) */ - smusdx v6, a3, a4 /* v6 = -B1 = W7*row[3] - W3*row[1] */ - smlad v5, lr, v7, v5 /* B0 += W5*row[5] + W7*row[7] */ - smusdx v7, a4, a2 /* v7 = B2 = W5*row[1] - W1*row[3] */ - - ldr a4, [pc, #(w42n-.-8)] /* a4 = -W4 | (-W2 << 16) */ - smlad v7, lr, a3, v7 /* B2 += W7*row[5] + W3*row[7] */ - ldr a3, [a1, #4] /* a3 = row[6,4] */ - smlsdx fp, lr, ip, fp /* B3 += W3*row[5] - W1*row[7] */ + ldr r10,[pc, #(w57-.-8)] /* r10 = W5 | (W7 << 16) */ + smlad r5, r2, lr, r1 + smlsd r6, r2, lr, r1 + + smuad r8, r3, ip /* r8 = B0 = W1*row[1] + W3*row[3] */ + smusdx r11,r3, r10 /* r11 = B3 = W7*row[1] - W5*row[3] */ + ldr lr, [r0, #12] /* lr = row[7,5] */ + pkhtb r2, ip, r10,asr #16 /* r3 = W7 | (W3 << 16) */ + pkhbt r1, ip, r10,lsl #16 /* r1 = W1 | (W5 << 16) */ + smusdx r9, r2, r3 /* r9 = -B1 = W7*row[3] - W3*row[1] */ + smlad r8, lr, r10,r8 /* B0 += W5*row[5] + W7*row[7] */ + smusdx r10,r3, r1 /* r10 = B2 = W5*row[1] - W1*row[3] */ + + ldr r3, [pc, #(w42n-.-8)] /* r3 = -W4 | (-W2 << 16) */ + smlad r10,lr, r2, r10 /* B2 += W7*row[5] + W3*row[7] */ + ldr r2, [r0, #4] /* r2 = row[6,4] */ + smlsdx r11,lr, ip, r11 /* B3 += W3*row[5] - W1*row[7] */ ldr ip, [pc, #(w46-.-8)] /* ip = W4 | (W6 << 16) */ - smlad v6, lr, a2, v6 /* B1 -= W1*row[5] + W5*row[7] */ + smlad r9, lr, r1, r9 /* B1 -= W1*row[5] + W5*row[7] */ - smlad v2, a3, a4, v2 /* A1 += -W4*row[4] - W2*row[6] */ - smlsd v3, a3, a4, v3 /* A2 += -W4*row[4] + W2*row[6] */ - smlad v1, a3, ip, v1 /* A0 += W4*row[4] + W6*row[6] */ - smlsd v4, a3, ip, v4 /* A3 += W4*row[4] - W6*row[6] */ + smlad r5, r2, r3, r5 /* A1 += -W4*row[4] - W2*row[6] */ + smlsd r6, r2, r3, r6 /* A2 += -W4*row[4] + W2*row[6] */ + smlad r4, r2, ip, r4 /* A0 += W4*row[4] + W6*row[6] */ + smlsd r7, r2, ip, r7 /* A3 += W4*row[4] - W6*row[6] */ .endm /* Compute partial IDCT of half row. shift = left-shift amount - a3 = row[2,0] - a4 = row[3,1] + r2 = row[2,0] + r3 = row[3,1] ip = w42 - Output in registers v1--v8 + Output in registers r4--r11 */ .macro idct_row4 shift ldr lr, [pc, #(w46-.-8)] /* lr = W4 | (W6 << 16) */ - ldr v7, [pc, #(w57-.-8)] /* v7 = W5 | (W7 << 16) */ - mov a2, #(1<<(\shift-1)) - smlad v1, a3, ip, a2 - smlsd v4, a3, ip, a2 + ldr r10,[pc, #(w57-.-8)] /* r10 = W5 | (W7 << 16) */ + mov r1, #(1<<(\shift-1)) + smlad r4, r2, ip, r1 + smlsd r7, r2, ip, r1 ldr ip, [pc, #(w13-.-8)] /* ip = W1 | (W3 << 16) */ - smlad v2, a3, lr, a2 - smlsd v3, a3, lr, a2 - smusdx fp, a4, v7 /* fp = B3 = W7*row[1] - W5*row[3] */ - smuad v5, a4, ip /* v5 = B0 = W1*row[1] + W3*row[3] */ - pkhtb a3, ip, v7, asr #16 /* a4 = W7 | (W3 << 16) */ - pkhbt a2, ip, v7, lsl #16 /* a2 = W1 | (W5 << 16) */ - smusdx v6, a3, a4 /* v6 = -B1 = W7*row[3] - W3*row[1] */ - smusdx v7, a4, a2 /* v7 = B2 = W5*row[1] - W1*row[3] */ + smlad r5, r2, lr, r1 + smlsd r6, r2, lr, r1 + smusdx r11,r3, r10 /* r11 = B3 = W7*row[1] - W5*row[3] */ + smuad r8, r3, ip /* r8 = B0 = W1*row[1] + W3*row[3] */ + pkhtb r2, ip, r10,asr #16 /* r3 = W7 | (W3 << 16) */ + pkhbt r1, ip, r10,lsl #16 /* r1 = W1 | (W5 << 16) */ + smusdx r9, r2, r3 /* r9 = -B1 = W7*row[3] - W3*row[1] */ + smusdx r10,r3, r1 /* r10 = B2 = W5*row[1] - W1*row[3] */ .endm /* Compute final part of IDCT single row without shift. - Input in registers v1--v8 - Output in registers ip, v1--v3, lr, v5--v7 + Input in registers r4--r11 + Output in registers ip, r4--r6, lr, r8--r10 */ .macro idct_finish - add ip, v1, v5 /* a2 = A0 + B0 */ - sub lr, v1, v5 /* a3 = A0 - B0 */ - sub v1, v2, v6 /* a3 = A1 + B1 */ - add v5, v2, v6 /* a3 = A1 - B1 */ - add v2, v3, v7 /* a2 = A2 + B2 */ - sub v6, v3, v7 /* a2 = A2 - B2 */ - add v3, v4, fp /* a3 = A3 + B3 */ - sub v7, v4, fp /* a3 = A3 - B3 */ + add ip, r4, r8 /* r1 = A0 + B0 */ + sub lr, r4, r8 /* r2 = A0 - B0 */ + sub r4, r5, r9 /* r2 = A1 + B1 */ + add r8, r5, r9 /* r2 = A1 - B1 */ + add r5, r6, r10 /* r1 = A2 + B2 */ + sub r9, r6, r10 /* r1 = A2 - B2 */ + add r6, r7, r11 /* r2 = A3 + B3 */ + sub r10,r7, r11 /* r2 = A3 - B3 */ .endm /* Compute final part of IDCT single row. shift = right-shift amount - Input/output in registers v1--v8 + Input/output in registers r4--r11 */ .macro idct_finish_shift shift - add a4, v1, v5 /* a4 = A0 + B0 */ - sub a3, v1, v5 /* a3 = A0 - B0 */ - mov v1, a4, asr #\shift - mov v5, a3, asr #\shift - - sub a4, v2, v6 /* a4 = A1 + B1 */ - add a3, v2, v6 /* a3 = A1 - B1 */ - mov v2, a4, asr #\shift - mov v6, a3, asr #\shift - - add a4, v3, v7 /* a4 = A2 + B2 */ - sub a3, v3, v7 /* a3 = A2 - B2 */ - mov v3, a4, asr #\shift - mov v7, a3, asr #\shift - - add a4, v4, fp /* a4 = A3 + B3 */ - sub a3, v4, fp /* a3 = A3 - B3 */ - mov v4, a4, asr #\shift - mov fp, a3, asr #\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 + + sub r3, r5, r9 /* r3 = A1 + B1 */ + add r2, r5, r9 /* r2 = A1 - B1 */ + mov r5, r3, asr #\shift + mov r9, r2, asr #\shift + + add r3, r6, r10 /* r3 = A2 + B2 */ + sub r2, r6, r10 /* r2 = A2 - B2 */ + mov r6, r3, asr #\shift + mov r10,r2, asr #\shift + + add r3, r7, r11 /* r3 = A3 + B3 */ + sub r2, r7, r11 /* r2 = A3 - B3 */ + mov r7, r3, asr #\shift + mov r11,r2, asr #\shift .endm /* Compute final part of IDCT single row, saturating results at 8 bits. shift = right-shift amount - Input/output in registers v1--v8 + Input/output in registers r4--r11 */ .macro idct_finish_shift_sat shift - add a4, v1, v5 /* a4 = A0 + B0 */ - sub ip, v1, v5 /* ip = A0 - B0 */ - usat v1, #8, a4, asr #\shift - usat v5, #8, ip, asr #\shift - - sub a4, v2, v6 /* a4 = A1 + B1 */ - add ip, v2, v6 /* ip = A1 - B1 */ - usat v2, #8, a4, asr #\shift - usat v6, #8, ip, asr #\shift - - add a4, v3, v7 /* a4 = A2 + B2 */ - sub ip, v3, v7 /* ip = A2 - B2 */ - usat v3, #8, a4, asr #\shift - usat v7, #8, ip, asr #\shift - - add a4, v4, fp /* a4 = A3 + B3 */ - sub ip, v4, fp /* ip = A3 - B3 */ - usat v4, #8, a4, asr #\shift - usat fp, #8, ip, asr #\shift + 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 + + 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 + + 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 + + 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 .endm /* Compute IDCT of single row, storing as column. - a1 = source - a2 = dest + r0 = source + r1 = dest */ function idct_row_armv6 str lr, [sp, #-4]! - ldr lr, [a1, #12] /* lr = row[7,5] */ - ldr ip, [a1, #4] /* ip = row[6,4] */ - ldr a4, [a1, #8] /* a4 = row[3,1] */ - ldr a3, [a1] /* a3 = row[2,0] */ + ldr lr, [r0, #12] /* lr = row[7,5] */ + ldr ip, [r0, #4] /* ip = row[6,4] */ + ldr r3, [r0, #8] /* r3 = row[3,1] */ + ldr r2, [r0] /* r2 = row[2,0] */ orrs lr, lr, ip - cmpeq lr, a4 - cmpeq lr, a3, lsr #16 + cmpeq lr, r3 + cmpeq lr, r2, lsr #16 beq 1f - str a2, [sp, #-4]! + str r1, [sp, #-4]! ldr ip, [pc, #(w42-.-8)] /* ip = W4 | (W2 << 16) */ cmp lr, #0 beq 2f @@ -213,140 +213,140 @@ function idct_row_armv6 2: idct_row4 ROW_SHIFT -3: ldr a2, [sp], #4 +3: ldr r1, [sp], #4 idct_finish_shift ROW_SHIFT - strh v1, [a2] - strh v2, [a2, #(16*2)] - strh v3, [a2, #(16*4)] - strh v4, [a2, #(16*6)] - strh fp, [a2, #(16*1)] - strh v7, [a2, #(16*3)] - strh v6, [a2, #(16*5)] - strh v5, [a2, #(16*7)] + strh r4, [r1] + strh r5, [r1, #(16*2)] + strh r6, [r1, #(16*4)] + strh r7, [r1, #(16*6)] + strh r11,[r1, #(16*1)] + strh r10,[r1, #(16*3)] + strh r9, [r1, #(16*5)] + strh r8, [r1, #(16*7)] ldr pc, [sp], #4 -1: mov a3, a3, lsl #3 - strh a3, [a2] - strh a3, [a2, #(16*2)] - strh a3, [a2, #(16*4)] - strh a3, [a2, #(16*6)] - strh a3, [a2, #(16*1)] - strh a3, [a2, #(16*3)] - strh a3, [a2, #(16*5)] - strh a3, [a2, #(16*7)] +1: mov r2, r2, lsl #3 + strh r2, [r1] + strh r2, [r1, #(16*2)] + strh r2, [r1, #(16*4)] + strh r2, [r1, #(16*6)] + strh r2, [r1, #(16*1)] + strh r2, [r1, #(16*3)] + strh r2, [r1, #(16*5)] + strh r2, [r1, #(16*7)] ldr pc, [sp], #4 .endfunc /* Compute IDCT of single column, read as row. - a1 = source - a2 = dest + r0 = source + r1 = dest */ function idct_col_armv6 - stmfd sp!, {a2, lr} + stmfd sp!, {r1, lr} - ldr a3, [a1] /* a3 = row[2,0] */ + ldr r2, [r0] /* r2 = row[2,0] */ ldr ip, [pc, #(w42-.-8)] /* ip = W4 | (W2 << 16) */ - ldr a4, [a1, #8] /* a4 = row[3,1] */ + ldr r3, [r0, #8] /* r3 = row[3,1] */ idct_row COL_SHIFT - ldr a2, [sp], #4 + ldr r1, [sp], #4 idct_finish_shift COL_SHIFT - strh v1, [a2] - strh v2, [a2, #(16*1)] - strh v3, [a2, #(16*2)] - strh v4, [a2, #(16*3)] - strh fp, [a2, #(16*4)] - strh v7, [a2, #(16*5)] - strh v6, [a2, #(16*6)] - strh v5, [a2, #(16*7)] + strh r4, [r1] + strh r5, [r1, #(16*1)] + strh r6, [r1, #(16*2)] + strh r7, [r1, #(16*3)] + strh r11,[r1, #(16*4)] + strh r10,[r1, #(16*5)] + strh r9, [r1, #(16*6)] + strh r8, [r1, #(16*7)] ldr pc, [sp], #4 .endfunc /* Compute IDCT of single column, read as row, store saturated 8-bit. - a1 = source - a2 = dest - a3 = line size + r0 = source + r1 = dest + r2 = line size */ function idct_col_put_armv6 - stmfd sp!, {a2, a3, lr} + stmfd sp!, {r1, r2, lr} - ldr a3, [a1] /* a3 = row[2,0] */ + ldr r2, [r0] /* r2 = row[2,0] */ ldr ip, [pc, #(w42-.-8)] /* ip = W4 | (W2 << 16) */ - ldr a4, [a1, #8] /* a4 = row[3,1] */ + ldr r3, [r0, #8] /* r3 = row[3,1] */ idct_row COL_SHIFT - ldmfd sp!, {a2, a3} + ldmfd sp!, {r1, r2} idct_finish_shift_sat COL_SHIFT - strb v1, [a2], a3 - strb v2, [a2], a3 - strb v3, [a2], a3 - strb v4, [a2], a3 - strb fp, [a2], a3 - strb v7, [a2], a3 - strb v6, [a2], a3 - strb v5, [a2], a3 + strb r4, [r1], r2 + strb r5, [r1], r2 + strb r6, [r1], r2 + strb r7, [r1], r2 + strb r11,[r1], r2 + strb r10,[r1], r2 + strb r9, [r1], r2 + strb r8, [r1], r2 - sub a2, a2, a3, lsl #3 + sub r1, r1, r2, lsl #3 ldr pc, [sp], #4 .endfunc /* Compute IDCT of single column, read as row, add/store saturated 8-bit. - a1 = source - a2 = dest - a3 = line size + r0 = source + r1 = dest + r2 = line size */ function idct_col_add_armv6 - stmfd sp!, {a2, a3, lr} + stmfd sp!, {r1, r2, lr} - ldr a3, [a1] /* a3 = row[2,0] */ + ldr r2, [r0] /* r2 = row[2,0] */ ldr ip, [pc, #(w42-.-8)] /* ip = W4 | (W2 << 16) */ - ldr a4, [a1, #8] /* a4 = row[3,1] */ + ldr r3, [r0, #8] /* r3 = row[3,1] */ idct_row COL_SHIFT - ldmfd sp!, {a2, a3} + ldmfd sp!, {r1, r2} idct_finish - ldrb a4, [a2] - ldrb v4, [a2, a3] - ldrb fp, [a2, a3, lsl #2] - add ip, a4, ip, asr #COL_SHIFT + ldrb r3, [r1] + ldrb r7, [r1, r2] + ldrb r11,[r1, r2, lsl #2] + add ip, r3, ip, asr #COL_SHIFT usat ip, #8, ip - add v1, v4, v1, asr #COL_SHIFT - strb ip, [a2], a3 - ldrb ip, [a2, a3] - usat v1, #8, v1 - ldrb fp, [a2, a3, lsl #2] - add v2, ip, v2, asr #COL_SHIFT - usat v2, #8, v2 - strb v1, [a2], a3 - ldrb a4, [a2, a3] - ldrb ip, [a2, a3, lsl #2] - strb v2, [a2], a3 - ldrb v4, [a2, a3] - ldrb v1, [a2, a3, lsl #2] - add v3, a4, v3, asr #COL_SHIFT - usat v3, #8, v3 - add v7, v4, v7, asr #COL_SHIFT - usat v7, #8, v7 - add v6, fp, v6, asr #COL_SHIFT - usat v6, #8, v6 - add v5, ip, v5, asr #COL_SHIFT - usat v5, #8, v5 - add lr, v1, lr, asr #COL_SHIFT + add r4, r7, r4, asr #COL_SHIFT + strb ip, [r1], r2 + ldrb ip, [r1, r2] + usat r4, #8, r4 + ldrb r11,[r1, r2, lsl #2] + add r5, ip, r5, asr #COL_SHIFT + usat r5, #8, r5 + strb r4, [r1], r2 + ldrb r3, [r1, r2] + ldrb ip, [r1, r2, lsl #2] + strb r5, [r1], r2 + ldrb r7, [r1, r2] + ldrb r4, [r1, r2, lsl #2] + add r6, r3, r6, asr #COL_SHIFT + usat r6, #8, r6 + add r10,r7, r10,asr #COL_SHIFT + usat r10,#8, r10 + add r9, r11,r9, asr #COL_SHIFT + usat r9, #8, r9 + add r8, ip, r8, asr #COL_SHIFT + usat r8, #8, r8 + add lr, r4, lr, asr #COL_SHIFT usat lr, #8, lr - strb v3, [a2], a3 - strb v7, [a2], a3 - strb v6, [a2], a3 - strb v5, [a2], a3 - strb lr, [a2], a3 + strb r6, [r1], r2 + strb r10,[r1], r2 + strb r9, [r1], r2 + strb r8, [r1], r2 + strb lr, [r1], r2 - sub a2, a2, a3, lsl #3 + sub r1, r1, r2, lsl #3 ldr pc, [sp], #4 .endfunc @@ -358,76 +358,76 @@ function idct_col_add_armv6 */ .macro idct_rows func width bl \func - add a1, a1, #(16*2) - add a2, a2, #\width + add r0, r0, #(16*2) + add r1, r1, #\width bl \func - add a1, a1, #(16*2) - add a2, a2, #\width + add r0, r0, #(16*2) + add r1, r1, #\width bl \func - add a1, a1, #(16*2) - add a2, a2, #\width + add r0, r0, #(16*2) + add r1, r1, #\width bl \func - sub a1, a1, #(16*5) - add a2, a2, #\width + sub r0, r0, #(16*5) + add r1, r1, #\width bl \func - add a1, a1, #(16*2) - add a2, a2, #\width + add r0, r0, #(16*2) + add r1, r1, #\width bl \func - add a1, a1, #(16*2) - add a2, a2, #\width + add r0, r0, #(16*2) + add r1, r1, #\width bl \func - add a1, a1, #(16*2) - add a2, a2, #\width + add r0, r0, #(16*2) + add r1, r1, #\width bl \func - sub a1, a1, #(16*7) + sub r0, r0, #(16*7) .endm /* void ff_simple_idct_armv6(DCTELEM *data); */ function ff_simple_idct_armv6, export=1 - stmfd sp!, {v1, v2, v3, v4, v5, v6, v7, fp, lr} + stmfd sp!, {r4-r11, lr} sub sp, sp, #128 - mov a2, sp + mov r1, sp idct_rows idct_row_armv6, 2 - mov a2, a1 - mov a1, sp + mov r1, r0 + mov r0, sp idct_rows idct_col_armv6, 2 add sp, sp, #128 - ldmfd sp!, {v1, v2, v3, v4, v5, v6, v7, fp, pc} + ldmfd sp!, {r4-r11, pc} .endfunc /* ff_simple_idct_add_armv6(uint8_t *dest, int line_size, DCTELEM *data); */ function ff_simple_idct_add_armv6, export=1 - stmfd sp!, {a1, a2, v1, v2, v3, v4, v5, v6, v7, fp, lr} + stmfd sp!, {r0, r1, r4-r11, lr} sub sp, sp, #128 - mov a1, a3 - mov a2, sp + mov r0, r2 + mov r1, sp idct_rows idct_row_armv6, 2 - mov a1, sp - ldr a2, [sp, #128] - ldr a3, [sp, #(128+4)] + mov r0, sp + ldr r1, [sp, #128] + ldr r2, [sp, #(128+4)] idct_rows idct_col_add_armv6, 1 add sp, sp, #(128+8) - ldmfd sp!, {v1, v2, v3, v4, v5, v6, v7, fp, pc} + ldmfd sp!, {r4-r11, pc} .endfunc /* ff_simple_idct_put_armv6(uint8_t *dest, int line_size, DCTELEM *data); */ function ff_simple_idct_put_armv6, export=1 - stmfd sp!, {a1, a2, v1, v2, v3, v4, v5, v6, v7, fp, lr} + stmfd sp!, {r0, r1, r4-r11, lr} sub sp, sp, #128 - mov a1, a3 - mov a2, sp + mov r0, r2 + mov r1, sp idct_rows idct_row_armv6, 2 - mov a1, sp - ldr a2, [sp, #128] - ldr a3, [sp, #(128+4)] + mov r0, sp + ldr r1, [sp, #128] + ldr r2, [sp, #(128+4)] idct_rows idct_col_put_armv6, 1 add sp, sp, #(128+8) - ldmfd sp!, {v1, v2, v3, v4, v5, v6, v7, fp, pc} + ldmfd sp!, {r4-r11, pc} .endfunc