* ARM asm DSP and codec/plugin functions: Use r12 scratch register properly

* Fix saving another unused reg in dsp code
* Use less regs in the generic ARM mpegplayer adding idct pure DC case
* Fix ARMv6 mpegplayer adding idct using an unsaved register in pure DC case


git-svn-id: svn://svn.rockbox.org/rockbox/trunk@21803 a1c6a512-1295-4272-9138-f99709370657

6 files changed, 125 insertions, 127 deletions

diff --git a/apps/codecs/lib/mdct_arm.S b/apps/codecs/lib/mdct_arm.S
index f262951..d082b4e 100644
--- a/apps/codecs/lib/mdct_arm.S
+++ b/apps/codecs/lib/mdct_arm.S
@@ -119,7 +119,7 @@ mdct_butterfly_16:
     ldr     pc, [sp], #4
 
 mdct_butterfly_32:
-    stmdb   sp!, {r4, r5, r6, r7, r8, r9, r10, r11, r12, lr}
+    stmdb   sp!, {r4-r11, lr}
 
     add     r1, r0, #16*4
 
@@ -247,13 +247,13 @@ mdct_butterfly_32:
     add     r0, r0, #16*4
     bl      mdct_butterfly_16
 
-    ldmia   sp!, {r4, r5, r6, r7, r8, r9, r10, r11, r12, pc}
+    ldmia   sp!, {r4-r11, pc}
 
     @ mdct_butterfly_generic_loop(x1, x2, T0, step, Ttop)
 mdct_butterfly_generic_loop:
-    stmdb   sp!, {r4, r5, r6, r7, r8, r9, r10, r11, r12, lr}
+    stmdb   sp!, {r4-r11, lr}
     str     r2, [sp, #-4]
-    ldr     r4, [sp, #40]
+    ldr     r4, [sp, #36]
 1:
     ldmdb   r0, {r6, r7, r8, r9}
     ldmdb   r1, {r10, r11, r12, r14}
@@ -339,7 +339,7 @@ mdct_butterfly_generic_loop:
     cmp     r2, r4
     bhi     1b
 
-    ldr     r4, [sp, #40]
+    ldr     r4, [sp, #36]
 1:
     ldmdb   r0, {r6, r7, r8, r9}
     ldmdb   r1, {r10, r11, r12, r14}
@@ -425,5 +425,5 @@ mdct_butterfly_generic_loop:
     cmp     r2, r4
     bhi     1b
 
-    ldmia   sp!, {r4, r5, r6, r7, r8, r9, r10, r11, r12, pc}
+    ldmia   sp!, {r4-r11, pc}
 
diff --git a/apps/codecs/libmad/synth_full_arm.S b/apps/codecs/libmad/synth_full_arm.S
index 99a223e..419bf2b 100644
--- a/apps/codecs/libmad/synth_full_arm.S
+++ b/apps/codecs/libmad/synth_full_arm.S
@@ -32,8 +32,8 @@
     ;; r3 = D0ptr
     ;; r4 = D1ptr
 synth_full1:
-    stmdb   sp!, {r4, r5, r6, r7, r8, r9, r10, r11, r12, lr}
-    ldr     r4, [sp, #40]
+    stmdb   sp!, {r4-r11, lr}
+    ldr     r4, [sp, #36]
     ldr     r5, =synth_full_sp
     str     sp, [r5]
     mov     r5, #15
@@ -135,11 +135,11 @@ synth_full1:
 
     ldr     r5, =synth_full_sp
     ldr     sp, [r5]
-    ldmia   sp!, {r4, r5, r6, r7, r8, r9, r10, r11, r12, pc}
+    ldmia   sp!, {r4-r11, pc}
 
 synth_full2:
-    stmdb   sp!, {r4, r5, r6, r7, r8, r9, r10, r11, r12, lr}
-    ldr     r4, [sp, #40]
+    stmdb   sp!, {r4-r11, lr}
+    ldr     r4, [sp, #36]
     ldr     r5, =synth_full_sp
     str     sp, [r5]
     mov     r5, #15
@@ -241,12 +241,12 @@ synth_full2:
 
     ldr     r5, =synth_full_sp
     ldr     sp, [r5]
-    ldmia   sp!, {r4, r5, r6, r7, r8, r9, r10, r11, r12, pc}
+    ldmia   sp!, {r4-r11, pc}
 
     .global III_aliasreduce
 
 III_aliasreduce:
-    stmdb   sp!, {r4, r5, r6, r7, r8, r9, r10, r11, r12, lr}
+    stmdb   sp!, {r4-r11, lr}
     add     r1, r0, r1, lsl #2
     add     r0, r0, #72
 .arl1:
@@ -289,7 +289,7 @@ III_aliasreduce:
     add     r0, r0, #72
     cmp     r0, r1
     blo     .arl1
-    ldmia   sp!, {r4, r5, r6, r7, r8, r9, r10, r11, r12, pc}
+    ldmia   sp!, {r4-r11, pc}
 
 csa:
     .word +0x0db84a81
@@ -311,14 +311,14 @@ csa:
 
     .global III_overlap
 III_overlap:
-    stmdb   sp!, {r4, r5, r6, r7, r8, lr}
+    stmdb   sp!, {r4-r7, lr}
     add     r2, r2, r3, lsl #2
     mov     r3, #6
 .ol:
     ldmia r0!, {r4, r5, r6}
-    ldmia r1!, {r7, r8, lr}
+    ldmia r1!, {r7, r12, lr}
     add   r4, r4, r7
-    add   r5, r5, r8
+    add   r5, r5, r12
     add   r6, r6, lr
     str   r4, [r2], #128
     str   r5, [r2], #128
@@ -326,13 +326,13 @@ III_overlap:
     subs  r3, r3, #1
     bne   .ol
     sub   r1, r1, #72
-    ldmia r0!, {r4, r5, r6, r7, r8, lr}
-    stmia r1!, {r4, r5, r6, r7, r8, lr}
-    ldmia r0!, {r4, r5, r6, r7, r8, lr}
-    stmia r1!, {r4, r5, r6, r7, r8, lr}
-    ldmia r0!, {r4, r5, r6, r7, r8, lr}
-    stmia r1!, {r4, r5, r6, r7, r8, lr}
-    ldmia   sp!, {r4, r5, r6, r7, r8, pc}
+    ldmia r0!, {r4, r5, r6, r7, r12, lr}
+    stmia r1!, {r4, r5, r6, r7, r12, lr}
+    ldmia r0!, {r4, r5, r6, r7, r12, lr}
+    stmia r1!, {r4, r5, r6, r7, r12, lr}
+    ldmia r0!, {r4, r5, r6, r7, r12, lr}
+    stmia r1!, {r4, r5, r6, r7, r12, lr}
+    ldmia   sp!, {r4-r7, pc}
 
     .section    IBSS_SECTION_MPA_ARM,"aw",%nobits
 synth_full_sp:
diff --git a/apps/codecs/libmusepack/synth_filter_arm.S b/apps/codecs/libmusepack/synth_filter_arm.S
index c2b28bd..8c87b61 100755
--- a/apps/codecs/libmusepack/synth_filter_arm.S
+++ b/apps/codecs/libmusepack/synth_filter_arm.S
@@ -9,10 +9,10 @@
  *

  * Copyright (C) 2008 by Andree Buschmann

  *

- * This program is free software; you can redistribute it and/or
- * modify it under the terms of the GNU General Public License
- * as published by the Free Software Foundation; either version 2
- * of the License, or (at your option) any later version.
+ * This program is free software; you can redistribute it and/or

+ * modify it under the terms of the GNU General Public License

+ * as published by the Free Software Foundation; either version 2

+ * of the License, or (at your option) any later version.

  *

  * This software is distributed on an "AS IS" basis, WITHOUT WARRANTY OF ANY

  * KIND, either express or implied.

@@ -41,7 +41,7 @@ mpc_decoder_windowing_D:
     /* r2 = D[] */

     /* lr = counter */

     

-    stmfd   sp!, {r4-r12, lr}

+    stmfd   sp!, {r4-r11, lr}

     

     mov lr, #32

 .loop32:

@@ -86,7 +86,7 @@ mpc_decoder_windowing_D:
     subs lr, lr, #1

     bgt .loop32

     

-    ldmfd   sp!, {r4-r12, pc}

+    ldmfd   sp!, {r4-r11, pc}

 .mpc_dewindowing_end:

     .size   mpc_decoder_windowing_D,.mpc_dewindowing_end-mpc_decoder_windowing_D

 #else

@@ -110,55 +110,55 @@ mpc_decoder_windowing_D:
     /************************************************************************

      * Reference implementation.

      ***********************************************************************/

-    stmfd   sp!, {r4-r9, lr}

+    stmfd   sp!, {r4-r8, lr}

     

     mov lr, #32

 .loop32:

     ldmia r2!, { r3-r6 }    /* load D[00..03] */

     ldr r7, [r1]            /*  0 */

-    smull r8, r9, r7, r3

+    smull r8, r12, r7, r3

     ldr r7, [r1, #96*4]     /*  1 */

-    smlal r8, r9, r7, r4

+    smlal r8, r12, r7, r4

     ldr r7, [r1, #128*4]    /*  2 */

-    smlal r8, r9, r7, r5

+    smlal r8, r12, r7, r5

     ldr r7, [r1, #224*4]    /*  3 */

-    smlal r8, r9, r7, r6

+    smlal r8, r12, r7, r6

     ldmia r2!, { r3-r6 }    /* load D[04..07] */

     ldr r7, [r1, #256*4]    /*  4 */

-    smlal r8, r9, r7, r3

+    smlal r8, r12, r7, r3

     ldr r7, [r1, #352*4]    /*  5 */

-    smlal r8, r9, r7, r4

+    smlal r8, r12, r7, r4

     ldr r7, [r1, #384*4]    /*  6 */

-    smlal r8, r9, r7, r5

+    smlal r8, r12, r7, r5

     ldr r7, [r1, #480*4]    /*  7 */

-    smlal r8, r9, r7, r6

+    smlal r8, r12, r7, r6

     ldmia r2!, { r3-r6 }    /* load D[08..11] */

     ldr r7, [r1, #512*4]    /*  8 */

-    smlal r8, r9, r7, r3

+    smlal r8, r12, r7, r3

     ldr r7, [r1, #608*4]    /*  9 */

-    smlal r8, r9, r7, r4

+    smlal r8, r12, r7, r4

     ldr r7, [r1, #640*4]    /* 10 */

-    smlal r8, r9, r7, r5

+    smlal r8, r12, r7, r5

     ldr r7, [r1, #736*4]    /* 11 */

-    smlal r8, r9, r7, r6

+    smlal r8, r12, r7, r6

     ldmia r2!, { r3-r6 }    /* load D[12..15] */

     ldr r7, [r1, #768*4]    /* 12 */

-    smlal r8, r9, r7, r3

+    smlal r8, r12, r7, r3

     ldr r7, [r1, #864*4]    /* 13 */

-    smlal r8, r9, r7, r4

+    smlal r8, r12, r7, r4

     ldr r7, [r1, #896*4]    /* 14 */

-    smlal r8, r9, r7, r5

+    smlal r8, r12, r7, r5

     ldr r7, [r1, #992*4]    /* 15 */

-    smlal r8, r9, r7, r6

+    smlal r8, r12, r7, r6

     mov r8, r8, lsr #16

-    orr r8, r8, r9, lsl #16 /* (lo>>16) || (hi<<16) */

+    orr r8, r8, r12, lsl #16 /* (lo>>16) || (hi<<16) */

     str r8, [r0], #4        /* store Data */

     add r1, r1, #4          /* V++ */

     

     subs lr, lr, #1

     bgt .loop32

     

-    ldmfd   sp!, {r4-r9, pc}

+    ldmfd   sp!, {r4-r8, pc}

 #else

 mpc_decoder_windowing_D:

     /* r0 = Data[] */

@@ -174,7 +174,7 @@ mpc_decoder_windowing_D:
      * The row V[16] can be extracted as it has symmetries within this single

      * row. 8 smull/mlal and 8 ldr's can be saved.

      ***********************************************************************/

-    stmfd   sp!, {r4-r12, lr}

+    stmfd   sp!, {r4-r11, lr}

     

     /******************************************

      * row 0 with internal symmetry

@@ -356,7 +356,7 @@ mpc_decoder_windowing_D:
     str r8, [r0], #4            /* store Data */

     add r1, r1, #4              /* V++ */

     

-    ldmfd   sp!, {r4-r12, pc}

+    ldmfd   sp!, {r4-r11, pc}

 #endif

 .mpc_dewindowing_end:

     .size   mpc_decoder_windowing_D,.mpc_dewindowing_end-mpc_decoder_windowing_D

diff --git a/apps/dsp_arm.S b/apps/dsp_arm.S
index 11b7ba7..df05520 100644
--- a/apps/dsp_arm.S
+++ b/apps/dsp_arm.S
@@ -32,14 +32,14 @@
     .type   channels_process_sound_chan_mono, %function
 channels_process_sound_chan_mono:
     @ input: r0 = count, r1 = buf
-    stmfd   sp!, {r4-r6, lr}
+    stmfd   sp!, {r4-r5, lr}
     ldmia   r1, {r2-r3}                @ r4 = buf[0], r5 = buf[1]
 
 .monoloop:
     ldmia   r2, {r4-r5}
-    ldmia   r3, {r6,lr}
+    ldmia   r3, {r12,lr}
     mov     r4, r4, asr #1             @ r4 = r4/2
-    add     r4, r4, r6, asr #1         @ r4 = r4 + r6/2 = (buf[0]+buf[1])/2
+    add     r4, r4, r12, asr #1        @ r4 = r4 + r12/2 = (buf[0]+buf[1])/2
     mov     r5, r5, asr #1             @ r5 = r5/2
     add     r5, r5, lr, asr #1         @ r5 = r5 + lr/2 = (buf[0]+buf[1])/2
     stmia   r2!, {r4-r5}
@@ -47,7 +47,7 @@ channels_process_sound_chan_mono:
     subs    r0, r0, #2
     bgt     .monoloop
     
-    ldmfd   sp!, {r4-r6, pc}
+    ldmfd   sp!, {r4-r5, pc}
 .monoend:
     .size   channels_process_sound_chan_mono,.monoend-channels_process_sound_chan_mono 
  
@@ -63,24 +63,24 @@ channels_process_sound_chan_mono:
     .type   channels_process_sound_chan_karaoke, %function
 channels_process_sound_chan_karaoke:
     @ input: r0 = count, r1 = buf
-    stmfd   sp!, {r4-r6, lr}
+    stmfd   sp!, {r4-r5, lr}
     ldmia   r1, {r2-r3}                @ r4 = buf[0], r5 = buf[1]
 
 .karaokeloop:
     ldmia   r2, {r4-r5}
-    ldmia   r3, {r6,lr} 
-    mov     r6, r6, asr #1             @ r6 = r6/2
-    rsb     r4, r6, r4, asr #1         @ r4 = -r6 + r4/2 = (buf[0]-buf[1])/2
-    rsb     r6, r4, #0                 @ r6 = -r4
+    ldmia   r3, {r12,lr}
+    mov     r12, r12, asr #1           @ r12 = r12/2
+    rsb     r4, r12, r4, asr #1        @ r4 = -r12 + r4/2 = (buf[0]-buf[1])/2
+    rsb     r12, r4, #0                @ r12 = -r4
     mov     lr, lr, asr #1             @ lr = lr/2
     rsb     r5, lr, r5, asr #1         @ r5 = -lr + r5/2 = (buf[0]-buf[1])/2
     rsb     lr, r5, #0                 @ lr = -r5  
     stmia   r2!, {r4-r5}
-    stmia   r3!, {r6,lr}
+    stmia   r3!, {r12,lr}
     subs    r0, r0, #2
     bgt     .karaokeloop
 
-    ldmfd   sp!, {r4-r6, pc}
+    ldmfd   sp!, {r4-r5, pc}
 .karaokeend:
     .size   channels_process_sound_chan_karaoke,.karaokeend-channels_process_sound_chan_karaoke
     
@@ -97,7 +97,7 @@ channels_process_sound_chan_karaoke:
     .type   sample_output_mono, %function
 sample_output_mono:
     @ input: r0 = count, r1 = data, r2 = src, r3 = dst
-    stmfd   sp!, {r4-r9, lr}
+    stmfd   sp!, {r4-r7, lr}
 
     ldr     r4, [r2]                   @ r4 = src[0]
     ldr     r5, [r1]                   @ lr = data->output_scale
@@ -105,8 +105,8 @@ sample_output_mono:
     mov     r2, #1
     mov     r2, r2, asl r1             @ r2 = 1<<r1 = 1 << (scale-1)
     mvn     r1, #0x8000                @ r1 needed for clipping
-    mov     r8, #0xff00
-    orr     r8, r8, #0xff              @ r8 needed for masking
+    mov     r12, #0xff00
+    orr     r12, r12, #0xff            @ r12 needed for masking
 
 .somloop:
     ldmia   r4!, {r6-r7}
@@ -121,16 +121,16 @@ sample_output_mono:
     teq     lr, lr, asr #31
     eorne   r7, r1, lr, asr #31        @ Clip (-32768...+32767)
     
-    and     r6, r6, r8
+    and     r6, r6, r12
     orr     r6, r6, r6, asl #16        @ pack first 2 halfwords into 1 word
-    and     r7, r7, r8
+    and     r7, r7, r12
     orr     r7, r7, r7, asl #16        @ pack last 2 halfwords into 1 word
     stmia   r3!, {r6-r7}
     
     subs    r0, r0, #2
     bgt     .somloop     
        
-    ldmfd   sp!, {r4-r9, pc}
+    ldmfd   sp!, {r4-r7, pc}
 .somend:
     .size   sample_output_mono,.somend-sample_output_mono
     
@@ -147,7 +147,7 @@ sample_output_mono:
     .type   sample_output_stereo, %function
 sample_output_stereo:
     @ input: r0 = count, r1 = data, r2 = src, r3 = dst
-    stmfd   sp!, {r4-r11, lr}
+    stmfd   sp!, {r4-r10, lr}
 
     ldmia   r2, {r4-r5}                @ r4 = src[0], r5 = src[1]
     ldr     r6, [r1]                   @ r6 = data->output_scale
@@ -155,8 +155,8 @@ sample_output_stereo:
     mov     r2, #1
     mov     r2, r2, asl r1             @ r2 = 1<<r1 = 1 << (scale-1)
     mvn     r1, #0x8000                @ r1 needed for clipping
-    mov     r11, #0xff00
-    orr     r11, r11, #0xff            @ r11 needed for masking
+    mov     r12, #0xff00
+    orr     r12, r12, #0xff            @ r12 needed for masking
 
 .sosloop:
     ldmia   r4!, {r7-r8}
@@ -183,16 +183,16 @@ sample_output_stereo:
     teq     lr, lr, asr #31
     eorne   r10, r1, lr, asr #31       @ Clip (-32768...+32767)
     
-    and     r7, r7, r11
+    and     r7, r7, r12
     orr     r9, r7, r9, asl #16        @ pack first 2 halfwords into 1 word
-    and     r8, r8, r11
+    and     r8, r8, r12
     orr     r10, r8, r10, asl #16      @ pack last 2 halfwords into 1 word
     stmia   r3!, {r9-r10}
 
     subs    r0, r0, #2
     bgt     .sosloop
 
-    ldmfd   sp!, {r4-r11, pc}
+    ldmfd   sp!, {r4-r10, pc}
 .sosend:
     .size   sample_output_stereo,.sosend-sample_output_stereo
 
diff --git a/apps/plugins/mpegplayer/idct_arm.S b/apps/plugins/mpegplayer/idct_arm.S
index 7de2d7f..3cb1554 100644
--- a/apps/plugins/mpegplayer/idct_arm.S
+++ b/apps/plugins/mpegplayer/idct_arm.S
@@ -268,7 +268,7 @@
     bx     lr
 
 mpeg2_idct_copy:
-    stmfd  sp!, { r1-r2, r4-r12, lr }
+    stmfd  sp!, { r1-r2, r4-r11, lr }
     bl     .idct
     ldmfd  sp!, { r1-r2 }
     mov    r11, #0
@@ -313,7 +313,7 @@ mpeg2_idct_copy:
     add    r1, r1, r2
     cmp    r0, r12
     blo    1b
-    ldmfd  sp!, { r4-r12, pc }
+    ldmfd  sp!, { r4-r11, pc }
 
 mpeg2_idct_add:
     cmp    r0, #129
@@ -324,7 +324,7 @@ mpeg2_idct_add:
     cmp    r1, #0x40
     bne    3f
 1:
-    stmfd  sp!, { r2-r12, lr }
+    stmfd  sp!, { r2-r11, lr }
     bl     .idct
     ldmfd  sp!, { r1-r2 }
     mov    r11, #0
@@ -385,24 +385,20 @@ mpeg2_idct_add:
     add    r1, r1, r2
     cmp    r0, r12
     blo    2b
-    ldmfd  sp!, { r4-r12, pc }
+    ldmfd  sp!, { r4-r11, pc }
 3:
-    stmfd  sp!, { r4-r11 }
+    stmfd  sp!, { r4-r6, lr }
     ldrsh  r1, [r0, #0]           /* r1 = block[0] */
-    mov    r11, #0
-    strh   r11, [r0, #0]          /* block[0] = 0 */
-    strh   r11, [r0, #126]        /* block[63] = 0 */
+    mov    r4, #0
+    strh   r4, [r0, #0]           /* block[0] = 0 */
+    strh   r4, [r0, #126]         /* block[63] = 0 */
     add    r1, r1, #64            /* r1 = DC << 7 */
     add    r0, r2, r3, asl #3
 4:
     ldrb   r4, [r2, #0]
     ldrb   r5, [r2, #1]
     ldrb   r6, [r2, #2]
-    ldrb   r7, [r2, #3]
-    ldrb   r8, [r2, #4]
-    ldrb   r9, [r2, #5]
-    ldrb   r10, [r2, #6]
-    ldrb   r11, [r2, #7]
+    ldrb   lr, [r2, #3]
     add    r4, r4, r1, asr #7
     cmp    r4, #255
     mvnhi  r4, r4, asr #31
@@ -415,28 +411,31 @@ mpeg2_idct_add:
     cmp    r6, #255
     mvnhi  r6, r6, asr #31
     strb   r6, [r2, #2]
-    add    r7, r7, r1, asr #7
-    cmp    r7, #255
-    mvnhi  r7, r7, asr #31
-    strb   r7, [r2, #3]
-    add    r8, r8, r1, asr #7
-    cmp    r8, #255
-    mvnhi  r8, r8, asr #31
-    strb   r8, [r2, #4]
-    add    r9, r9, r1, asr #7
-    cmp    r9, #255
-    mvnhi  r9, r9, asr #31
-    strb   r9, [r2, #5]
-    add    r10, r10, r1, asr #7
-    cmp    r10, #255
-    mvnhi  r10, r10, asr #31
-    strb   r10, [r2, #6]
-    add    r11, r11, r1, asr #7
-    cmp    r11, #255
-    mvnhi  r11, r11, asr #31
-    strb   r11, [r2, #7]
+    add    lr, lr, r1, asr #7
+    cmp    lr, #255
+    mvnhi  lr, lr, asr #31
+    strb   lr, [r2, #3]
+    ldrb   r4, [r2, #4]
+    ldrb   r5, [r2, #5]
+    ldrb   r6, [r2, #6]
+    ldrb   lr, [r2, #7]
+    add    r4, r4, r1, asr #7
+    cmp    r4, #255
+    mvnhi  r4, r4, asr #31
+    strb   r4, [r2, #4]
+    add    r5, r5, r1, asr #7
+    cmp    r5, #255
+    mvnhi  r5, r5, asr #31
+    strb   r5, [r2, #5]
+    add    r6, r6, r1, asr #7
+    cmp    r6, #255
+    mvnhi  r6, r6, asr #31
+    strb   r6, [r2, #6]
+    add    lr, lr, r1, asr #7
+    cmp    lr, #255
+    mvnhi  lr, lr, asr #31
+    strb   lr, [r2, #7]
     add    r2, r2, r3
     cmp    r2, r0
     blo    4b
-    ldmfd  sp!, { r4-r11 }
-    bx     lr
+    ldmfd  sp!, { r4-r6, pc }
diff --git a/apps/plugins/mpegplayer/idct_armv6.S b/apps/plugins/mpegplayer/idct_armv6.S
index 73feed4..ca97d12 100644
--- a/apps/plugins/mpegplayer/idct_armv6.S
+++ b/apps/plugins/mpegplayer/idct_armv6.S
@@ -196,7 +196,7 @@ L_W0246:
 
 
 mpeg2_idct_copy:
-    stmfd  sp!, {r1-r2, r4-r12, lr}
+    stmfd  sp!, {r1-r2, r4-r11, lr}
     bl     .idct
     ldmfd  sp!, {r1-r2}
 
@@ -230,7 +230,7 @@ mpeg2_idct_copy:
     cmp    r0, r12
     blo    1b
 
-    ldmfd  sp!, {r4-r12, pc}
+    ldmfd  sp!, {r4-r11, pc}
 
 mpeg2_idct_add:
     cmp    r0, #129
@@ -241,7 +241,7 @@ mpeg2_idct_add:
     cmp    r1, #0x40
     bne    3f
 1:
-    stmfd  sp!, {r2-r12, lr}
+    stmfd  sp!, {r2-r11, lr}
     bl     .idct
     ldmfd  sp!, {r1-r2}
     mov    r11, #0
@@ -287,21 +287,21 @@ mpeg2_idct_add:
     add    r1, r1, r2
     cmp    r0, r12
     blo    2b
-    ldmfd  sp!, {r4-r12, pc}
+    ldmfd  sp!, {r4-r11, pc}
 
 3:
-    stmfd  sp!, {r4-r7}
+    stmfd  sp!, {r4-r6, lr}
     ldrsh  r1, [r0, #0]           /* r1 = block[0] */
-    mov    r11, #0
-    strh   r11, [r0, #0]          /* block[0] = 0 */
-    strh   r11, [r0, #126]        /* block[63] = 0 */
+    mov    r4, #0
+    strh   r4, [r0, #0]           /* block[0] = 0 */
+    strh   r4, [r0, #126]         /* block[63] = 0 */
     add    r1, r1, #64            /* r1 = DC << 7 */
     add    r0, r2, r3, asl #3
 4:
     ldrb   r4, [r2, #0]
     ldrb   r5, [r2, #1]
     ldrb   r6, [r2, #2]
-    ldrb   r7, [r2, #3]
+    ldrb   lr, [r2, #3]
     add    r4, r4, r1, asr #7
     usat   r4, #8, r4
     strb   r4, [r2, #0]
@@ -311,13 +311,13 @@ mpeg2_idct_add:
     add    r6, r6, r1, asr #7
     usat   r6, #8, r6
     strb   r6, [r2, #2]
-    add    r7, r7, r1, asr #7
-    usat   r7, #8, r7
-    strb   r7, [r2, #3]
+    add    lr, lr, r1, asr #7
+    usat   lr, #8, lr
+    strb   lr, [r2, #3]
     ldrb   r4, [r2, #4]
     ldrb   r5, [r2, #5]
     ldrb   r6, [r2, #6]
-    ldrb   r7, [r2, #7]
+    ldrb   lr, [r2, #7]
     add    r4, r4, r1, asr #7
     usat   r4, #8, r4
     strb   r4, [r2, #4]
@@ -327,11 +327,10 @@ mpeg2_idct_add:
     add    r6, r6, r1, asr #7
     usat   r6, #8, r6
     strb   r6, [r2, #6]
-    add    r7, r7, r1, asr #7
-    usat   r7, #8, r7
-    strb   r7, [r2, #7]
+    add    lr, lr, r1, asr #7
+    usat   lr, #8, lr
+    strb   lr, [r2, #7]
     add    r2, r2, r3
     cmp    r2, r0
     blo    4b
-    ldmfd  sp!, {r4-r7}
-    bx     lr
+    ldmfd  sp!, {r4-r6, pc}