From 87f5359d604a3d51526965a97c74896f392ed444 Mon Sep 17 00:00:00 2001 From: Jens Arnold Date: Fri, 19 Oct 2007 22:57:19 +0000 Subject: Shuffle some instructions around for that extra percent of performance. Fix a bunch of comments. git-svn-id: svn://svn.rockbox.org/rockbox/trunk@15216 a1c6a512-1295-4272-9138-f99709370657 --- apps/codecs/demac/libdemac/predictor-cf.S | 243 +++++++++++++++--------------- 1 file changed, 123 insertions(+), 120 deletions(-) diff --git a/apps/codecs/demac/libdemac/predictor-cf.S b/apps/codecs/demac/libdemac/predictor-cf.S index 19873420c3..3b9489e791 100644 --- a/apps/codecs/demac/libdemac/predictor-cf.S +++ b/apps/codecs/demac/libdemac/predictor-cf.S @@ -97,6 +97,8 @@ predictor_decode_stereo: sub.l %d3, %d2 neg.l %d2 | %d2 = %d3 - %d2 + move.l %d2, (YDELAYA-4,%a5) | p->buf[YDELAYA-1] = %d2 + movem.l (YcoeffsA,%a6), %a0-%a3 | %a0 = p->YcoeffsA[0] | %a1 = p->YcoeffsA[1] | %a2 = p->YcoeffsA[2] @@ -107,10 +109,7 @@ predictor_decode_stereo: mac.l %d1, %a2, %acc0 | %acc0 += p->buf[YDELAYA-2] * p->YcoeffsA[2] mac.l %d0, %a3, %acc0 | %acc0 += p->buf[YDELAYA-3] * p->YcoeffsA[3] - move.l %d2, (YDELAYA-4,%a5) | p->buf[YDELAYA-1] = %d2 move.l %d3, (YDELAYA,%a5) | p->buf[YDELAYA] = %d3 - - movclr.l %acc0, %d0 tst.l %d2 beq.s 1f @@ -118,16 +117,18 @@ predictor_decode_stereo: extb.l %d2 | pos: 0xffffffff, neg: 0x00000000 or.l #1, %d2 | pos: 0xffffffff, neg: 0x00000001 1: | %d2 = SIGN(%d2) + move.l %d2, (YADAPTCOEFFSA-4,%a5) | p->buf[YADAPTCOEFFSA-1] = %d2 + tst.l %d3 beq.s 1f spl.b %d3 extb.l %d3 or.l #1, %d3 1: | %d3 = SIGN(%d3) - - move.l %d2, (YADAPTCOEFFSA-4,%a5) | p->buf[YADAPTCOEFFSA-1] = %d2 move.l %d3, (YADAPTCOEFFSA,%a5) | p->buf[YADAPTCOEFFSA] = %d3 - + + movclr.l %acc0, %d0 + | NOTE: %d0 now contains predictionA - don't overwrite. | Predictor Y, Filter B @@ -149,11 +150,13 @@ predictor_decode_stereo: sub.l %d3, %d7 neg.l %d7 | %d7 = %d3 - %d7 - movem.l (YcoeffsB,%a6), %d2/%a0-%a3 | %d2 = p->YcoeffsB[0] - | %a0 = p->YcoeffsB[1] - | %a1 = p->YcoeffsB[2] - | %a2 = p->YcoeffsB[3] - | %a3 = p->YcoeffsB[4] + move.l %d7, (YDELAYB-4,%a5) | p->buf[YDELAYB-1] = %d7 + + movem.l (YcoeffsB,%a6), %d2/%a0-%a3 | %d2 = p->YcoeffsB[0] + | %a0 = p->YcoeffsB[1] + | %a1 = p->YcoeffsB[2] + | %a2 = p->YcoeffsB[3] + | %a3 = p->YcoeffsB[4] mac.l %d3, %d2, %acc0 | %acc0 = p->buf[YDELAYB] * p->YcoeffsB[0] mac.l %d7, %a0, %acc0 | %acc0 += p->buf[YDELAYB-1] * p->YcoeffsB[1] @@ -161,26 +164,24 @@ predictor_decode_stereo: mac.l %d5, %a2, %acc0 | %acc0 += p->buf[YDELAYB-3] * p->YcoeffsB[3] mac.l %d4, %a3, %acc0 | %acc0 += p->buf[YDELAYB-4] * p->YcoeffsB[4] - move.l %d7, (YDELAYB-4,%a5) | p->buf[YDELAYB-1] = %d7 move.l %d3, (YDELAYB, %a5) | p->buf[YDELAYB] = %d3 - movclr.l %acc0, %d1 - tst.l %d7 beq.s 1f spl.b %d7 extb.l %d7 or.l #1, %d7 1: | %d7 = SIGN(%d7) + move.l %d7, (YADAPTCOEFFSB-4,%a5) | p->buf[YADAPTCOEFFSB-1] = %d7 tst.l %d3 beq.s 1f spl.b %d3 extb.l %d3 or.l #1, %d3 1: | %d3 = SIGN(%d3) - - move.l %d7, (YADAPTCOEFFSB-4,%a5) | p->buf[YADAPTCOEFFSB-1] = %d7 move.l %d3, (YADAPTCOEFFSB, %a5) | p->buf[YADAPTCOEFFSB] = %d3 + + movclr.l %acc0, %d1 | %d0 still contains predictionA | %d1 contains predictionB @@ -215,70 +216,70 @@ predictor_decode_stereo: move.l %d1, (%a4)+ | *(decoded0++) = %d1 (p->YfilterA) move.l %a4, (%sp) | save decoded0 tst.l %d5 - beq.s 2f + beq.s 3f - movem.l (YADAPTCOEFFSB-16,%a5), %d4-%d6 | d4 = p->buf[YADAPTCOEFFSB-4] - | d5 = p->buf[YADAPTCOEFFSB-3] - | d6 = p->buf[YADAPTCOEFFSB-2] + movem.l (YADAPTCOEFFSB-16,%a5), %d4-%d6 | %d4 = p->buf[YADAPTCOEFFSB-4] + | %d5 = p->buf[YADAPTCOEFFSB-3] + | %d6 = p->buf[YADAPTCOEFFSB-2] bmi.s 1f | flags still valid here | *decoded0 > 0 - sub.l %d3, %d2 | d2 = p->YcoeffsB[0] - p->buf[YADAPTCOEFFSB] - sub.l %d7, %a0 | a0 = p->YcoeffsB[1] - p->buf[YADAPTCOEFFSB-1] - sub.l %d6, %a1 | a1 = p->YcoeffsB[2] - p->buf[YADAPTCOEFFSB-2] - sub.l %d5, %a2 | a2 = p->YcoeffsB[3] - p->buf[YADAPTCOEFFSB-3] - sub.l %d4, %a3 | a3 = p->YcoeffsB[4] - p->buf[YADAPTCOEFFSB-4] + sub.l %d3, %d2 | %d2 = p->YcoeffsB[0] - p->buf[YADAPTCOEFFSB] + sub.l %d7, %a0 | %a0 = p->YcoeffsB[1] - p->buf[YADAPTCOEFFSB-1] + sub.l %d6, %a1 | %a1 = p->YcoeffsB[2] - p->buf[YADAPTCOEFFSB-2] + sub.l %d5, %a2 | %a2 = p->YcoeffsB[3] - p->buf[YADAPTCOEFFSB-3] + sub.l %d4, %a3 | %a3 = p->YcoeffsB[4] - p->buf[YADAPTCOEFFSB-4] movem.l %d2/%a0-%a3, (YcoeffsB,%a6) | Save p->YcoeffsB[] - movem.l (YcoeffsA,%a6), %d4-%d7 | d4 = p->YcoeffsA[0] - | d5 = p->YcoeffsA[1] - | d6 = p->YcoeffsA[2] - | d7 = p->YcoeffsA[3] + movem.l (YcoeffsA,%a6), %d4-%d7 | %d4 = p->YcoeffsA[0] + | %d5 = p->YcoeffsA[1] + | %d6 = p->YcoeffsA[2] + | %d7 = p->YcoeffsA[3] - movem.l (YADAPTCOEFFSA-12,%a5), %d2/%a0-%a2 | d2 = p->buf[YADAPTCOEFFSA-3] - | a0 = p->buf[YADAPTCOEFFSA-2] - | a1 = p->buf[YADAPTCOEFFSA-1] - | a2 = p->buf[YADAPTCOEFFSA] + movem.l (YADAPTCOEFFSA-12,%a5), %d2/%a0-%a2 | %d2 = p->buf[YADAPTCOEFFSA-3] + | %a0 = p->buf[YADAPTCOEFFSA-2] + | %a1 = p->buf[YADAPTCOEFFSA-1] + | %a2 = p->buf[YADAPTCOEFFSA] - sub.l %a2, %d4 | d4 = p->YcoeffsA[0] - p->buf[YADAPTCOEFFSA] - sub.l %a1, %d5 | d5 = p->YcoeffsA[1] - p->buf[YADAPTCOEFFSA-1] - sub.l %a0, %d6 | d6 = p->YcoeffsA[2] - p->buf[YADAPTCOEFFSA-2] - sub.l %d2, %d7 | d7 = p->YcoeffsA[3] - p->buf[YADAPTCOEFFSA-3] + sub.l %a2, %d4 | %d4 = p->YcoeffsA[0] - p->buf[YADAPTCOEFFSA] + sub.l %a1, %d5 | %d5 = p->YcoeffsA[1] - p->buf[YADAPTCOEFFSA-1] + sub.l %a0, %d6 | %d6 = p->YcoeffsA[2] - p->buf[YADAPTCOEFFSA-2] + sub.l %d2, %d7 | %d7 = p->YcoeffsA[3] - p->buf[YADAPTCOEFFSA-3] - movem.l %d4-%d7, (YcoeffsA,%a6) | Save p->YcoeffsA[] bra.s 2f 1: | *decoded0 < 0 - add.l %d3, %d2 | d2 = p->YcoeffsB[0] + p->buf[YADAPTCOEFFSB] - add.l %d7, %a0 | a0 = p->YcoeffsB[1] + p->buf[YADAPTCOEFFSB-1] - add.l %d6, %a1 | a1 = p->YcoeffsB[2] + p->buf[YADAPTCOEFFSB-2] - add.l %d5, %a2 | a2 = p->YcoeffsB[3] + p->buf[YADAPTCOEFFSB-3] - add.l %d4, %a3 | a3 = p->YcoeffsB[4] + p->buf[YADAPTCOEFFSB-4] + add.l %d3, %d2 | %d2 = p->YcoeffsB[0] + p->buf[YADAPTCOEFFSB] + add.l %d7, %a0 | %a0 = p->YcoeffsB[1] + p->buf[YADAPTCOEFFSB-1] + add.l %d6, %a1 | %a1 = p->YcoeffsB[2] + p->buf[YADAPTCOEFFSB-2] + add.l %d5, %a2 | %a2 = p->YcoeffsB[3] + p->buf[YADAPTCOEFFSB-3] + add.l %d4, %a3 | %a3 = p->YcoeffsB[4] + p->buf[YADAPTCOEFFSB-4] movem.l %d2/%a0-%a3, (YcoeffsB,%a6) | Save p->YcoeffsB[] - movem.l (YcoeffsA,%a6), %d4-%d7 | d4 = p->YcoeffsA[0] - | d5 = p->YcoeffsA[1] - | d6 = p->YcoeffsA[2] - | d7 = p->YcoeffsA[3] + movem.l (YcoeffsA,%a6), %d4-%d7 | %d4 = p->YcoeffsA[0] + | %d5 = p->YcoeffsA[1] + | %d6 = p->YcoeffsA[2] + | %d7 = p->YcoeffsA[3] - movem.l (YADAPTCOEFFSA-12,%a5), %d2/%a0-%a2 | d2 = p->buf[YADAPTCOEFFSA-3] - | a0 = p->buf[YADAPTCOEFFSA-2] - | a1 = p->buf[YADAPTCOEFFSA-1] - | a2 = p->buf[YADAPTCOEFFSA] + movem.l (YADAPTCOEFFSA-12,%a5), %d2/%a0-%a2 | %d2 = p->buf[YADAPTCOEFFSA-3] + | %a0 = p->buf[YADAPTCOEFFSA-2] + | %a1 = p->buf[YADAPTCOEFFSA-1] + | %a2 = p->buf[YADAPTCOEFFSA] - add.l %a2, %d4 | d4 = p->YcoeffsA[0] + p->buf[YADAPTCOEFFSA] - add.l %a1, %d5 | d5 = p->YcoeffsA[1] + p->buf[YADAPTCOEFFSA-1] - add.l %a0, %d6 | d6 = p->YcoeffsA[2] + p->buf[YADAPTCOEFFSA-2] - add.l %d2, %d7 | d7 = p->YcoeffsA[3] + p->buf[YADAPTCOEFFSA-3] - - movem.l %d4-%d7, (YcoeffsA,%a6) | Save p->YcoeffsA[] + add.l %a2, %d4 | %d4 = p->YcoeffsA[0] + p->buf[YADAPTCOEFFSA] + add.l %a1, %d5 | %d5 = p->YcoeffsA[1] + p->buf[YADAPTCOEFFSA-1] + add.l %a0, %d6 | %d6 = p->YcoeffsA[2] + p->buf[YADAPTCOEFFSA-2] + add.l %d2, %d7 | %d7 = p->YcoeffsA[3] + p->buf[YADAPTCOEFFSA-3] 2: + movem.l %d4-%d7, (YcoeffsA,%a6) | Save p->YcoeffsA[] + +3: | ***** PREDICTOR X ***** @@ -293,6 +294,8 @@ predictor_decode_stereo: sub.l %d3, %d2 neg.l %d2 | %d2 = %d3 -%d2 + move.l %d2, (XDELAYA-4,%a5) | p->buf[XDELAYA-1] = %d2 + movem.l (XcoeffsA,%a6), %a0-%a3 | %a0 = p->XcoeffsA[0] | %a1 = p->XcoeffsA[1] | %a2 = p->XcoeffsA[2] @@ -303,27 +306,26 @@ predictor_decode_stereo: mac.l %d1, %a2, %acc0 | %acc0 += p->buf[XDELAYA-2] * p->XcoeffsA[2] mac.l %d0, %a3, %acc0 | %acc0 += p->buf[XDELAYA-3] * p->XcoeffsA[3] - move.l %d2, (XDELAYA-4,%a5) | p->buf[XDELAYA-1] = %d2 move.l %d3, (XDELAYA,%a5) | p->buf[XDELAYA] = %d3 - movclr.l %acc0, %d0 - tst.l %d2 beq.s 1f spl.b %d2 | pos: 0x??????ff, neg: 0x??????00 extb.l %d2 | pos: 0xffffffff, neg: 0x00000000 or.l #1, %d2 | pos: 0xffffffff, neg: 0x00000001 1: | %d2 = SIGN(%d2) + move.l %d2, (XADAPTCOEFFSA-4,%a5) | p->buf[XADAPTCOEFFSA-1] = %d2 + tst.l %d3 beq.s 1f spl.b %d3 extb.l %d3 or.l #1, %d3 1: | %d3 = SIGN(%d3) + move.l %d3, (XADAPTCOEFFSA,%a5) | p->buf[XADAPTCOEFFSA] = %d3 + + movclr.l %acc0, %d0 - move.l %d2, (XADAPTCOEFFSA-4,%a5) | p->buf[XADAPTCOEFFSA-1] = r2 - move.l %d3, (XADAPTCOEFFSA,%a5) | p->buf[XADAPTCOEFFSA] = r3 - | NOTE: %d0 now contains predictionA - don't overwrite. | Predictor X, Filter B @@ -345,11 +347,13 @@ predictor_decode_stereo: sub.l %d3, %d7 neg.l %d7 | %d7 = %d3 - %d7 - movem.l (XcoeffsB,%a6), %d2/%a0-%a3 | %d2 = p->XcoeffsB[0] - | %a0 = p->XcoeffsB[1] - | %a1 = p->XcoeffsB[2] - | %a2 = p->XcoeffsB[3] - | %a3 = p->XcoeffsB[4] + move.l %d7, (XDELAYB-4,%a5) | p->buf[XDELAYB-1] = %d7 + + movem.l (XcoeffsB,%a6), %d2/%a0-%a3 | %d2 = p->XcoeffsB[0] + | %a0 = p->XcoeffsB[1] + | %a1 = p->XcoeffsB[2] + | %a2 = p->XcoeffsB[3] + | %a3 = p->XcoeffsB[4] mac.l %d3, %d2, %acc0 | %acc0 = p->buf[XDELAYB] * p->XcoeffsB[0] mac.l %d7, %a0, %acc0 | %acc0 += p->buf[XDELAYB-1] * p->XcoeffsB[1] @@ -357,27 +361,26 @@ predictor_decode_stereo: mac.l %d5, %a2, %acc0 | %acc0 += p->buf[XDELAYB-3] * p->XcoeffsB[3] mac.l %d4, %a3, %acc0 | %acc0 += p->buf[XDELAYB-4] * p->XcoeffsB[4] - move.l %d7, (XDELAYB-4,%a5) | p->buf[XDELAYB-1] = %d7 move.l %d3, (XDELAYB, %a5) | p->buf[XDELAYB] = %d3 - movclr.l %acc0, %d1 - tst.l %d7 beq.s 1f spl.b %d7 extb.l %d7 or.l #1, %d7 1: | %d7 = SIGN(%d7) + move.l %d7, (XADAPTCOEFFSB-4,%a5) | p->buf[XADAPTCOEFFSB-1] = %d7 + tst.l %d3 beq.s 1f spl.b %d3 extb.l %d3 or.l #1, %d3 1: | %d3 = SIGN(%d3) - - move.l %d7, (XADAPTCOEFFSB-4,%a5) | p->buf[XADAPTCOEFFSB-1] = %d7 move.l %d3, (XADAPTCOEFFSB, %a5) | p->buf[XADAPTCOEFFSB] = %d3 - + + movclr.l %acc0, %d1 + | %d0 still contains predictionA | %d1 contains predictionB @@ -405,76 +408,76 @@ predictor_decode_stereo: | %a4 contains decoded1 | %d5 contains *decoded1 - | %d2, %a0, %a1, %a2, %a31 contain p->XcoeffsB[0..4] + | %d2, %a0, %a1, %a2, %a3 contain p->XcoeffsB[0..4] | %d7, %d3 contain p->buf[XADAPTCOEFFSB-1] and p->buf[XADAPTCOEFFSB] move.l %d1, (%a4)+ | *(decoded1++) = %d1 (p->XfilterA) move.l %a4, (4,%sp) | save decoded1 tst.l %d5 - beq.s 2f + beq.s 3f - movem.l (XADAPTCOEFFSB-16,%a5), %d4-%d6 | d4 = p->buf[XADAPTCOEFFSB-4] - | d5 = p->buf[XADAPTCOEFFSB-3] - | d6 = p->buf[XADAPTCOEFFSB-2] + movem.l (XADAPTCOEFFSB-16,%a5), %d4-%d6 | %d4 = p->buf[XADAPTCOEFFSB-4] + | %d5 = p->buf[XADAPTCOEFFSB-3] + | %d6 = p->buf[XADAPTCOEFFSB-2] bmi.s 1f | flags still valid here | *decoded1 > 0 - sub.l %d3, %d2 | d2 = p->XcoeffsB[0] - p->buf[XADAPTCOEFFSB] - sub.l %d7, %a0 | a0 = p->XcoeffsB[1] - p->buf[XADAPTCOEFFSB-1] - sub.l %d6, %a1 | a1 = p->XcoeffsB[2] - p->buf[XADAPTCOEFFSB-2] - sub.l %d5, %a2 | a2 = p->XcoeffsB[3] - p->buf[XADAPTCOEFFSB-3] - sub.l %d4, %a3 | a3 = p->XcoeffsB[4] - p->buf[XADAPTCOEFFSB-4] + sub.l %d3, %d2 | %d2 = p->XcoeffsB[0] - p->buf[XADAPTCOEFFSB] + sub.l %d7, %a0 | %a0 = p->XcoeffsB[1] - p->buf[XADAPTCOEFFSB-1] + sub.l %d6, %a1 | %a1 = p->XcoeffsB[2] - p->buf[XADAPTCOEFFSB-2] + sub.l %d5, %a2 | %a2 = p->XcoeffsB[3] - p->buf[XADAPTCOEFFSB-3] + sub.l %d4, %a3 | %a3 = p->XcoeffsB[4] - p->buf[XADAPTCOEFFSB-4] movem.l %d2/%a0-%a3, (XcoeffsB,%a6) | Save p->XcoeffsB[] - movem.l (XcoeffsA,%a6), %d4-%d7 | d4 = p->XcoeffsA[0] - | d5 = p->XcoeffsA[1] - | d6 = p->XcoeffsA[2] - | d7 = p->XcoeffsA[3] + movem.l (XcoeffsA,%a6), %d4-%d7 | %d4 = p->XcoeffsA[0] + | %d5 = p->XcoeffsA[1] + | %d6 = p->XcoeffsA[2] + | %d7 = p->XcoeffsA[3] - movem.l (XADAPTCOEFFSA-12,%a5), %d2/%a0-%a2 | d2 = p->buf[XADAPTCOEFFSA-3] - | a0 = p->buf[XADAPTCOEFFSA-2] - | a1 = p->buf[XADAPTCOEFFSA-1] - | a2 = p->buf[XADAPTCOEFFSA] + movem.l (XADAPTCOEFFSA-12,%a5), %d2/%a0-%a2 | %d2 = p->buf[XADAPTCOEFFSA-3] + | %a0 = p->buf[XADAPTCOEFFSA-2] + | %a1 = p->buf[XADAPTCOEFFSA-1] + | %a2 = p->buf[XADAPTCOEFFSA] - sub.l %a2, %d4 | d4 = p->XcoeffsA[0] - p->buf[XADAPTCOEFFSA] - sub.l %a1, %d5 | d5 = p->XcoeffsA[1] - p->buf[XADAPTCOEFFSA-1] - sub.l %a0, %d6 | d6 = p->XcoeffsA[2] - p->buf[XADAPTCOEFFSA-2] - sub.l %d2, %d7 | d7 = p->XcoeffsA[3] - p->buf[XADAPTCOEFFSA-3] + sub.l %a2, %d4 | %d4 = p->XcoeffsA[0] - p->buf[XADAPTCOEFFSA] + sub.l %a1, %d5 | %d5 = p->XcoeffsA[1] - p->buf[XADAPTCOEFFSA-1] + sub.l %a0, %d6 | %d6 = p->XcoeffsA[2] - p->buf[XADAPTCOEFFSA-2] + sub.l %d2, %d7 | %d7 = p->XcoeffsA[3] - p->buf[XADAPTCOEFFSA-3] - movem.l %d4-%d7, (XcoeffsA,%a6) | Save p->XcoeffsA[] bra.s 2f 1: | *decoded1 < 0 - add.l %d3, %d2 | d2 = p->XcoeffsB[0] + p->buf[XADAPTCOEFFSB] - add.l %d7, %a0 | a0 = p->XcoeffsB[1] + p->buf[XADAPTCOEFFSB-1] - add.l %d6, %a1 | a1 = p->XcoeffsB[2] + p->buf[XADAPTCOEFFSB-2] - add.l %d5, %a2 | a2 = p->XcoeffsB[3] + p->buf[XADAPTCOEFFSB-3] - add.l %d4, %a3 | a3 = p->XcoeffsB[4] + p->buf[XADAPTCOEFFSB-4] + add.l %d3, %d2 | %d2 = p->XcoeffsB[0] + p->buf[XADAPTCOEFFSB] + add.l %d7, %a0 | %a0 = p->XcoeffsB[1] + p->buf[XADAPTCOEFFSB-1] + add.l %d6, %a1 | %a1 = p->XcoeffsB[2] + p->buf[XADAPTCOEFFSB-2] + add.l %d5, %a2 | %a2 = p->XcoeffsB[3] + p->buf[XADAPTCOEFFSB-3] + add.l %d4, %a3 | %a3 = p->XcoeffsB[4] + p->buf[XADAPTCOEFFSB-4] movem.l %d2/%a0-%a3, (XcoeffsB,%a6) | Save p->XcoeffsB[] - movem.l (XcoeffsA,%a6), %d4-%d7 | d4 = p->XcoeffsA[0] - | d5 = p->XcoeffsA[1] - | d6 = p->XcoeffsA[2] - | d7 = p->XcoeffsA[3] + movem.l (XcoeffsA,%a6), %d4-%d7 | %d4 = p->XcoeffsA[0] + | %d5 = p->XcoeffsA[1] + | %d6 = p->XcoeffsA[2] + | %d7 = p->XcoeffsA[3] - movem.l (XADAPTCOEFFSA-12,%a5), %d2/%a0-%a2 | d2 = p->buf[XADAPTCOEFFSA-3] - | a0 = p->buf[XADAPTCOEFFSA-2] - | a1 = p->buf[XADAPTCOEFFSA-1] - | a2 = p->buf[XADAPTCOEFFSA] + movem.l (XADAPTCOEFFSA-12,%a5), %d2/%a0-%a2 | %d2 = p->buf[XADAPTCOEFFSA-3] + | %a0 = p->buf[XADAPTCOEFFSA-2] + | %a1 = p->buf[XADAPTCOEFFSA-1] + | %a2 = p->buf[XADAPTCOEFFSA] - add.l %a2, %d4 | d4 = p->XcoeffsA[0] + p->buf[XADAPTCOEFFSA] - add.l %a1, %d5 | d5 = p->XcoeffsA[1] + p->buf[XADAPTCOEFFSA-1] - add.l %a0, %d6 | d6 = p->XcoeffsA[2] + p->buf[XADAPTCOEFFSA-2] - add.l %d2, %d7 | d7 = p->XcoeffsA[3] + p->buf[XADAPTCOEFFSA-3] - - movem.l %d4-%d7, (XcoeffsA,%a6) | Save p->XcoeffsA[] + add.l %a2, %d4 | %d4 = p->XcoeffsA[0] + p->buf[XADAPTCOEFFSA] + add.l %a1, %d5 | %d5 = p->XcoeffsA[1] + p->buf[XADAPTCOEFFSA-1] + add.l %a0, %d6 | %d6 = p->XcoeffsA[2] + p->buf[XADAPTCOEFFSA-2] + add.l %d2, %d7 | %d7 = p->XcoeffsA[3] + p->buf[XADAPTCOEFFSA-3] 2: + movem.l %d4-%d7, (XcoeffsA,%a6) | Save p->XcoeffsA[] + +3: | ***** COMMON ***** @@ -484,10 +487,10 @@ predictor_decode_stereo: cmp.l %a3, %a5 bne.s .endofloop - + | The history buffer is full, we need to do a memmove: - - lea.l (historybuffer,%a6), %a3 + + lea.l (historybuffer,%a6), %a3 | dest = %a3 (p->historybuffer) | src = %a5 (p->buf) @@ -515,7 +518,7 @@ predictor_decode_stereo: lea.l (40,%a3), %a3 lea.l (historybuffer,%a6), %a5 | p->buf = &p->historybuffer[0] - + .endofloop: subq.l #1, (8,%sp) | decrease loop count bne.w .loop -- cgit v1.2.3