Compute partial IDCT of single row.
shift = left-shift amount
a1 = source address
+ a3 = row[2,0]
+ a4 = row[3,1]
Output in registers v1--v8
*/
.macro idct_row shift
- ldr a3, [a1] /* a3 = row[2,0] */
ldr ip, [pc, #(w42-.-8)] /* ip = W4 | (W2 << 16) */
ldr lr, [pc, #(w46-.-8)] /* lr = W4 | (W6 << 16) */
- ldr a4, [a1, #8] /* a4 = row[3,1] */
mov a2, #(1<<(\shift-1))
smlad v1, a3, ip, a2
smlsd v4, a3, ip, a2
idct_col_armv6:
stmfd sp!, {a2, lr}
+ ldr a3, [a1] /* a3 = row[2,0] */
+ ldr a4, [a1, #8] /* a4 = row[3,1] */
idct_row COL_SHIFT
ldr a2, [sp], #4
idct_finish_shift COL_SHIFT
idct_col_put_armv6:
stmfd sp!, {a2, a3, lr}
+ ldr a3, [a1] /* a3 = row[2,0] */
+ ldr a4, [a1, #8] /* a4 = row[3,1] */
idct_row COL_SHIFT
ldmfd sp!, {a2, a3}
idct_finish_shift_sat COL_SHIFT
idct_col_add_armv6:
stmfd sp!, {a2, a3, lr}
+ ldr a3, [a1] /* a3 = row[2,0] */
+ ldr a4, [a1, #8] /* a4 = row[3,1] */
idct_row COL_SHIFT
ldmfd sp!, {a2, a3}
idct_finish