diff options
| author | Nils Wallménius <nils@rockbox.org> | 2010-06-18 13:47:37 +0000 |
|---|---|---|
| committer | Nils Wallménius <nils@rockbox.org> | 2010-06-18 13:47:37 +0000 |
| commit | 273d05906de397785d97f71c221b8dde9e77e93e (patch) | |
| tree | 744fc563736500ca96b59f6391bce0e733399b89 | |
| parent | 68e2023184ab0138a0c475e2100c1d76891ef2af (diff) | |
| download | rockbox-273d05906de397785d97f71c221b8dde9e77e93e.tar.gz rockbox-273d05906de397785d97f71c221b8dde9e77e93e.zip | |
libmad: Optimize away 2 instructions from coldfire III_imdct, no measurable speed difference.
git-svn-id: svn://svn.rockbox.org/rockbox/trunk@26918 a1c6a512-1295-4272-9138-f99709370657
| -rw-r--r-- | apps/codecs/libmad/imdct_mcf5249.S | 12 |
1 files changed, 5 insertions, 7 deletions
diff --git a/apps/codecs/libmad/imdct_mcf5249.S b/apps/codecs/libmad/imdct_mcf5249.S index d6d4460986..b5e2a12b0f 100644 --- a/apps/codecs/libmad/imdct_mcf5249.S +++ b/apps/codecs/libmad/imdct_mcf5249.S | |||
| @@ -34,10 +34,10 @@ III_imdct_s: | |||
| 34 | * get more low bits out of the accext01 register _before_ doing the | 34 | * get more low bits out of the accext01 register _before_ doing the |
| 35 | * movclrs. | 35 | * movclrs. |
| 36 | */ | 36 | */ |
| 37 | sub.l %a0, %a0 /* clear loop variable */ | 37 | moveq.l #3, %d5 /* we do three outer loop iterations */ |
| 38 | .imdctloop: /* outer loop label */ | 38 | .imdctloop: /* outer loop label */ |
| 39 | lea.l imdct_s, %a1 /* load pointer to imdct coefs in a1 */ | 39 | lea.l imdct_s, %a1 /* load pointer to imdct coefs in a1 */ |
| 40 | movem.l (%a2), %d0-%d5 /* load some input data in d0-d5 */ | 40 | movem.l (%a2), %d0-%d4/%a0 /* load some input data in d0-d4/a0 */ |
| 41 | lea.l (6*4, %a2), %a2 | 41 | lea.l (6*4, %a2), %a2 |
| 42 | 42 | ||
| 43 | clr.l %d7 /* clear loop variable */ | 43 | clr.l %d7 /* clear loop variable */ |
| @@ -48,7 +48,7 @@ III_imdct_s: | |||
| 48 | mac.l %d2, %a4, (%a1)+, %a4, %acc0 | 48 | mac.l %d2, %a4, (%a1)+, %a4, %acc0 |
| 49 | mac.l %d3, %a4, (%a1)+, %a4, %acc0 | 49 | mac.l %d3, %a4, (%a1)+, %a4, %acc0 |
| 50 | mac.l %d4, %a4, (%a1)+, %a4, %acc0 | 50 | mac.l %d4, %a4, (%a1)+, %a4, %acc0 |
| 51 | mac.l %d5, %a4, (%a1)+, %a4, %acc0 | 51 | mac.l %a0, %a4, (%a1)+, %a4, %acc0 |
| 52 | movclr.l %acc0, %d6 /* get result, left shifted once */ | 52 | movclr.l %acc0, %d6 /* get result, left shifted once */ |
| 53 | asl.l #3, %d6 /* one shift free, shift three more */ | 53 | asl.l #3, %d6 /* one shift free, shift three more */ |
| 54 | move.l %d6, (%a3, %d7.l*4) /* yptr[i] = result */ | 54 | move.l %d6, (%a3, %d7.l*4) /* yptr[i] = result */ |
| @@ -60,7 +60,7 @@ III_imdct_s: | |||
| 60 | mac.l %d2, %a4, (%a1)+, %a4, %acc0 | 60 | mac.l %d2, %a4, (%a1)+, %a4, %acc0 |
| 61 | mac.l %d3, %a4, (%a1)+, %a4, %acc0 | 61 | mac.l %d3, %a4, (%a1)+, %a4, %acc0 |
| 62 | mac.l %d4, %a4, (%a1)+, %a4, %acc0 | 62 | mac.l %d4, %a4, (%a1)+, %a4, %acc0 |
| 63 | mac.l %d5, %a4, (%a1)+, %a4, %acc0 | 63 | mac.l %a0, %a4, (%a1)+, %a4, %acc0 |
| 64 | movclr.l %acc0, %d6 /* get result */ | 64 | movclr.l %acc0, %d6 /* get result */ |
| 65 | asl.l #3, %d6 | 65 | asl.l #3, %d6 |
| 66 | move.l %d6, (11*4, %a3, %d7.l*4) /* yptr[11 - i] = result */ | 66 | move.l %d6, (11*4, %a3, %d7.l*4) /* yptr[11 - i] = result */ |
| @@ -72,9 +72,7 @@ III_imdct_s: | |||
| 72 | jne .macloop | 72 | jne .macloop |
| 73 | 73 | ||
| 74 | lea.l (12*4, %a3), %a3 /* add pointer increment */ | 74 | lea.l (12*4, %a3), %a3 /* add pointer increment */ |
| 75 | addq.l #1, %a0 /* increment outer loop variable */ | 75 | subq.l #1, %d5 /* decrement outer loop variable */ |
| 76 | moveq.l #3, %d0 | ||
| 77 | cmp.l %d0, %a0 /* we do three outer loop iterations */ | ||
| 78 | jne .imdctloop | 76 | jne .imdctloop |
| 79 | 77 | ||
| 80 | /* windowing, overlapping and concatenation */ | 78 | /* windowing, overlapping and concatenation */ |