ARM asm LCD and ATA driver functions: Don't save r12 as it is a scratch reg. Saves a bit of stack and execution time.

git-svn-id: svn://svn.rockbox.org/rockbox/trunk@21795 a1c6a512-1295-4272-9138-f99709370657

12 files changed, 256 insertions, 261 deletions

diff --git a/firmware/target/arm/as3525/lcd-as-e200v2-fuze.S b/firmware/target/arm/as3525/lcd-as-e200v2-fuze.S
index 254816cd0e..71d997d044 100644
--- a/firmware/target/arm/as3525/lcd-as-e200v2-fuze.S
+++ b/firmware/target/arm/as3525/lcd-as-e200v2-fuze.S
@@ -51,7 +51,7 @@ lcd_write_yuv420_lines:
                                         @ r0 = yuv_src
                                         @ r1 = width
                                         @ r2 = stride
-    stmfd       sp!, { r4-r12 }         @ save non-scratch
+    stmfd       sp!, { r4-r10, lr }     @ save non-scratch
     ldmia       r0, { r4, r5, r6 }      @ r4 = yuv_src[0] = Y'_p
                                         @ r5 = yuv_src[1] = Cb_p
                                         @ r6 = yuv_src[2] = Cr_p
@@ -76,9 +76,9 @@ lcd_write_yuv420_lines:
     add         r10, r10, r8, asl #3    @
     add         r10, r10, r8, asl #4    @
                                         @
-    add         r11, r9, r9, asl #2     @ r9 = Cr*101
-    add         r11, r11, r9, asl #5    @
-    add         r9, r11, r9, asl #6     @
+    add         lr, r9, r9, asl #2      @ r9 = Cr*101
+    add         lr, lr, r9, asl #5      @
+    add         r9, lr, r9, asl #6      @
                                         @
     add         r8, r8, #2              @ r8 = bu = (Cb*128 + 128) >> 8
     mov         r8, r8, asr #2          @
@@ -88,19 +88,19 @@ lcd_write_yuv420_lines:
     mov         r10, r10, asr #8        @
                                         @ compute R, G, and B
     add         r0, r8, r7, asr #8      @ r0  = b = (Y >> 9) + bu
-    add         r11, r9, r7, asr #8     @ r11 = r = (Y >> 9) + rv
+    add         lr, r9, r7, asr #8      @ lr = r = (Y >> 9) + rv
     add         r7, r10, r7, asr #7     @ r7  = g = (Y >> 8) + guv
                                         @
-    orr         r12, r0, r11            @ check if clamping is needed...
+    orr         r12, r0, lr             @ check if clamping is needed...
     orr         r12, r12, r7, asr #1    @ ...at all
     cmp         r12, #31                @
     bls         15f @ no clamp          @
     cmp         r0, #31                 @ clamp b
     mvnhi       r0, r0, asr #31         @
     andhi       r0, r0, #31             @
-    cmp         r11, #31                @ clamp r
-    mvnhi       r11, r11, asr #31       @
-    andhi       r11, r11, #31           @
+    cmp         lr, #31                 @ clamp r
+    mvnhi       lr, lr, asr #31         @
+    andhi       lr, lr, #31             @
     cmp         r7, #63                 @ clamp g
     mvnhi       r7, r7, asr #31         @
     andhi       r7, r7, #63             @
@@ -108,7 +108,7 @@ lcd_write_yuv420_lines:
                                         @
     ldrb        r12, [r4, r2]           @ r12 = Y' = *(Y'_p + stride)
                                         @
-    orr         r0, r0, r11, lsl #11    @ r0 = (r << 11) | b
+    orr         r0, r0, lr, lsl #11     @ r0 = (r << 11) | b
     orr         r0, r0, r7, lsl #5      @ r0 = (r << 11) | (g << 5) | b
     strh        r0, [r3, #0x10]         @ write pixel
 1: @ busy                               @
@@ -121,19 +121,19 @@ lcd_write_yuv420_lines:
     add         r7, r12, r7, asl #5     @
                                         @ compute R, G, and B
     add         r0, r8, r7, asr #8      @ r0  = b = (Y >> 9) + bu
-    add         r11, r9, r7, asr #8     @ r11 = r = (Y >> 9) + rv
+    add         lr, r9, r7, asr #8      @ lr = r = (Y >> 9) + rv
     add         r7, r10, r7, asr #7     @ r7  = g = (Y >> 8) + guv
                                         @
-    orr         r12, r0, r11            @ check if clamping is needed...
+    orr         r12, r0, lr             @ check if clamping is needed...
     orr         r12, r12, r7, asr #1    @ ...at all
     cmp         r12, #31                @
     bls         15f @ no clamp          @
     cmp         r0, #31                 @ clamp b
     mvnhi       r0, r0, asr #31         @
     andhi       r0, r0, #31             @
-    cmp         r11, #31                @ clamp r
-    mvnhi       r11, r11, asr #31       @
-    andhi       r11, r11, #31           @
+    cmp         lr, #31                 @ clamp r
+    mvnhi       lr, lr, asr #31         @
+    andhi       lr, lr, #31             @
     cmp         r7, #63                 @ clamp g
     mvnhi       r7, r7, asr #31         @
     andhi       r7, r7, #63             @
@@ -141,7 +141,7 @@ lcd_write_yuv420_lines:
                                         @
     ldrb        r12, [r4], #1           @ r12 = Y' = *(Y'_p++)
                                         @
-    orr         r0, r0, r11, lsl #11    @ r0 = (r << 11) | b
+    orr         r0, r0, lr, lsl #11     @ r0 = (r << 11) | b
     orr         r0, r0, r7, lsl #5      @ r0 = (r << 11) | (g << 5) | b
     strh        r0, [r3, #0x10]         @ write pixel
 1: @ busy                               @
@@ -154,19 +154,19 @@ lcd_write_yuv420_lines:
     add         r7, r12, r7, asl #5     @
                                         @ compute R, G, and B
     add         r0, r8, r7, asr #8      @ r0  = b = (Y >> 9) + bu
-    add         r11, r9, r7, asr #8     @ r11 = r = (Y >> 9) + rv
+    add         lr, r9, r7, asr #8      @ lr = r = (Y >> 9) + rv
     add         r7, r10, r7, asr #7     @ r7  = g = (Y >> 8) + guv
                                         @
-    orr         r12, r0, r11            @ check if clamping is needed...
+    orr         r12, r0, lr             @ check if clamping is needed...
     orr         r12, r12, r7, asr #1    @ ...at all
     cmp         r12, #31                @
     bls         15f @ no clamp          @
     cmp         r0, #31                 @ clamp b
     mvnhi       r0, r0, asr #31         @
     andhi       r0, r0, #31             @
-    cmp         r11, #31                @ clamp r
-    mvnhi       r11, r11, asr #31       @
-    andhi       r11, r11, #31           @
+    cmp         lr, #31                 @ clamp r
+    mvnhi       lr, lr, asr #31         @
+    andhi       lr, lr, #31             @
     cmp         r7, #63                 @ clamp g
     mvnhi       r7, r7, asr #31         @
     andhi       r7, r7, #63             @
@@ -175,7 +175,7 @@ lcd_write_yuv420_lines:
     ldrb        r12, [r4, r2]           @ r12 = Y' = *(Y'_p + stride)
                                         @
                                         @
-    orr         r0, r0, r11, lsl #11    @ r0 = (r << 11) | b
+    orr         r0, r0, lr, lsl #11     @ r0 = (r << 11) | b
     orr         r0, r0, r7, lsl #5      @ r0 = (r << 11) | (g << 5) | b
     strh        r0, [r3, #0x10]         @ write pixel
 1: @ busy                               @
@@ -188,25 +188,25 @@ lcd_write_yuv420_lines:
     add         r7, r12, r7, asl #5     @
                                         @ compute R, G, and B
     add         r0, r8, r7, asr #8      @ r0  = b = (Y >> 9) + bu
-    add         r11, r9, r7, asr #8     @ r11 = r = (Y >> 9) + rv
+    add         lr, r9, r7, asr #8      @ lr = r = (Y >> 9) + rv
     add         r7, r10, r7, asr #7     @ r7  = g = (Y >> 8) + guv
                                         @
-    orr         r12, r0, r11            @ check if clamping is needed...
+    orr         r12, r0, lr             @ check if clamping is needed...
     orr         r12, r12, r7, asr #1    @ ...at all
     cmp         r12, #31                @
     bls         15f @ no clamp          @
     cmp         r0, #31                 @ clamp b
     mvnhi       r0, r0, asr #31         @
     andhi       r0, r0, #31             @
-    cmp         r11, #31                @ clamp r
-    mvnhi       r11, r11, asr #31       @
-    andhi       r11, r11, #31           @
+    cmp         lr, #31                 @ clamp r
+    mvnhi       lr, lr, asr #31         @
+    andhi       lr, lr, #31             @
     cmp         r7, #63                 @ clamp g
     mvnhi       r7, r7, asr #31         @
     andhi       r7, r7, #63             @
 15: @ no clamp                          @
                                         @
-    orr         r0, r0, r11, lsl #11    @ r0 = (r << 11) | b
+    orr         r0, r0, lr, lsl #11     @ r0 = (r << 11) | b
     orr         r0, r0, r7, lsl #5      @ r0 = (r << 11) | (g << 5) | b
     strh        r0, [r3, #0x10]         @ write pixel
 1: @ busy                               @
@@ -217,7 +217,7 @@ lcd_write_yuv420_lines:
     subs        r1, r1, #2              @ subtract block from width
     bgt         10b @ loop line         @
                                         @
-    ldmfd       sp!, { r4-r12 }         @ restore registers and return
+    ldmfd       sp!, { r4-r10, pc }     @ restore registers and return
     bx          lr                      @
     .ltorg                              @ dump constant pool
     .size   lcd_write_yuv420_lines, .-lcd_write_yuv420_lines
@@ -258,7 +258,7 @@ lcd_write_yuv420_lines_odither:
                                         @ r2 = stride
                                         @ r3 = x_screen
                                         @ [sp] = y_screen
-    stmfd       sp!, { r4-r12, lr }     @ save non-scratch
+    stmfd       sp!, { r4-r11, lr }     @ save non-scratch
     ldmia       r0, { r4, r5, r6 }      @ r4 = yuv_src[0] = Y'_p
                                         @ r5 = yuv_src[1] = Cb_p
                                         @ r6 = yuv_src[2] = Cr_p
@@ -496,6 +496,6 @@ lcd_write_yuv420_lines_odither:
     subs        r1, r1, #2              @ subtract block from width
     bgt         10b @ loop line         @
                                         @
-    ldmfd       sp!, { r4-r12, pc }     @ restore registers and return
+    ldmfd       sp!, { r4-r11, pc }     @ restore registers and return
     .ltorg                              @ dump constant pool
     .size   lcd_write_yuv420_lines_odither, .-lcd_write_yuv420_lines_odither
diff --git a/firmware/target/arm/as3525/sansa-clip/lcd-as-clip.S b/firmware/target/arm/as3525/sansa-clip/lcd-as-clip.S
index ed430178b9..5a52eb032c 100644
--- a/firmware/target/arm/as3525/sansa-clip/lcd-as-clip.S
+++ b/firmware/target/arm/as3525/sansa-clip/lcd-as-clip.S
@@ -39,15 +39,15 @@
  *   r3/r4 - current block of phases
  *   r5/r6 - current block of values
  *   r7 - lcd data accumulator
- *   r8 - phase signs mask
+ *   r12 - phase signs mask
  *   lr - lcd bridge address
  */
 
 lcd_grey_data:
-    stmfd   sp!, {r4-r8, lr}
-    mov     r8, #0x80
-    orr     r8, r8, r8, lsl #8
-    orr     r8, r8, r8, lsl #16
+    stmfd   sp!, {r4-r7, lr}
+    mov     r12, #0x80
+    orr     r12, r12, r12, lsl #8
+    orr     r12, r12, r12, lsl #16
 
     ldr     lr, =GPIOA_BASE
     mov     r3, #(1<<5)
@@ -70,7 +70,7 @@ lcd_grey_data:
     orrne   r7, r7, #0x2000
     tst     r3, #0x80000000
     orrne   r7, r7, #0x1000
-    bic     r3, r3, r8
+    bic     r3, r3, r12
     add     r3, r3, r5
 
     /* set bits 3..0 */
@@ -82,7 +82,7 @@ lcd_grey_data:
     orrne   r7, r7, #0x02
     tst     r4, #0x80000000
     orrne   r7, r7, #0x01
-    bic     r4, r4, r8
+    bic     r4, r4, r12
     add     r4, r4, r6
 
     stmia   r1!, {r3-r4}
@@ -102,5 +102,5 @@ lcd_grey_data:
     ands    r5, r5, #(1<<10)    @ wait until push fifo empties
     beq     1b
 
-    ldmfd   sp!, {r4-r8, pc}
+    ldmfd   sp!, {r4-r7, pc}
     .size   lcd_grey_data,.-lcd_grey_data
diff --git a/firmware/target/arm/ata-as-arm.S b/firmware/target/arm/ata-as-arm.S
index 31b753c421..c1475c0a5a 100644
--- a/firmware/target/arm/ata-as-arm.S
+++ b/firmware/target/arm/ata-as-arm.S
@@ -56,11 +56,11 @@
  *   r0 - current address
  *   r1 - word count
  *   r2 - ata port
- *   r3..r6, lr - read buffers
+ *   r3..r5, r12, lr - read buffers
  */
 
 copy_read_sectors:
-    stmfd   sp!, {r4, r5, r6, lr}
+    stmfd   sp!, {r4, r5, lr}
     ldr     r2, =.ata_port
     tst     r0, #1              /* 16 bit aligned? */
     beq     .r_aligned
@@ -92,16 +92,16 @@ copy_read_sectors:
     ldrh    r5, [r2]
     orr     r4, r4, r5, lsl #24
     mov     r5, r5, lsr #8
-    ldrh    r6, [r2]
-    orr     r5, r5, r6, lsl #8
-    ldrh    r6, [r2]
-    orr     r5, r5, r6, lsl #24
-    mov     r6, r6, lsr #8
+    ldrh    r12, [r2]
+    orr     r5, r5, r12, lsl #8
+    ldrh    r12, [r2]
+    orr     r5, r5, r12, lsl #24
+    mov     r12, r12, lsr #8
     ldrh    lr, [r2]
-    orr     r6, r6, lr, lsl #8
+    orr     r12, r12, lr, lsl #8
     ldrh    lr, [r2]
-    orr     r6, r6, lr, lsl #24
-    stmia   r0!, {r3, r4, r5, r6}
+    orr     r12, r12, lr, lsl #24
+    stmia   r0!, {r3, r4, r5, r12}
     mov     r3, lr, lsr #8
     subs    r1, r1, #8          /* 8 or more halfwords left? */
     bge     .r_loop_u
@@ -140,7 +140,7 @@ copy_read_sectors:
 
     strb    r3, [r0], #1        /* store final byte */
 
-    ldmfd   sp!, {r4, r5, r6, pc}
+    ldmfd   sp!, {r4, r5, pc}
 
     /* 16-bit aligned */
 .r_aligned:
@@ -158,12 +158,12 @@ copy_read_sectors:
     ldrh    r5, [r2]
     orr     r4, r4, r5, lsl #16
     ldrh    r5, [r2]
-    ldrh    r6, [r2]
-    orr     r5, r5, r6, lsl #16
-    ldrh    r6, [r2]
+    ldrh    r12, [r2]
+    orr     r5, r5, r12, lsl #16
+    ldrh    r12, [r2]
     ldrh    lr, [r2]
-    orr     r6, r6, lr, lsl #16
-    stmia   r0!, {r3, r4, r5, r6}
+    orr     r12, r12, lr, lsl #16
+    stmia   r0!, {r3, r4, r5, r12}
     subs    r1, r1, #8          /* 8 or more halfwords left? */
     bge     .r_loop_a
     
@@ -189,7 +189,7 @@ copy_read_sectors:
     ldrneh  r3, [r2]
     strneh  r3, [r0], #2
 
-    ldmfd   sp!, {r4, r5, r6, pc}
+    ldmfd   sp!, {r4, r5, pc}
 
 .r_end:
     .size   copy_read_sectors,.r_end-copy_read_sectors
@@ -210,11 +210,11 @@ copy_read_sectors:
  *   r0 - current address
  *   r1 - word count
  *   r2 - ata port
- *   r3..r6, lr - read buffers
+ *   r3..r5, r12, lr - read buffers
  */
 
 copy_write_sectors:
-    stmfd   sp!, {r4, r5, r6, lr}
+    stmfd   sp!, {r4, r5, lr}
     ldr     r2, =.ata_port
     tst     r0, #1              /* 16 bit aligned? */
     beq     .w_aligned
@@ -234,7 +234,7 @@ copy_write_sectors:
 
     sub     r1, r1, #8          /* adjust for zero-check and doing 8 halfwords/loop */
 .w_loop_u:
-    ldmia   r0!, {r4, r5, r6, lr}
+    ldmia   r0!, {r4, r5, r12, lr}
     orr     r3, r3, r4, lsl #8  /* Load 4 words at once and decompose them */
     strh    r3, [r2]            /*  into 8 halfwords in a way that the words */
     mov     r3, r3, lsr #16     /*  are shifted by 8 bits, putting the high */
@@ -245,15 +245,15 @@ copy_write_sectors:
     mov     r4, r4, lsr #16
     strh    r4, [r2]
     mov     r5, r5, lsr #24
-    orr     r5, r5, r6, lsl #8
+    orr     r5, r5, r12, lsl #8
     strh    r5, [r2]
     mov     r5, r5, lsr #16
     strh    r5, [r2]
-    mov     r6, r6, lsr #24
-    orr     r6, r6, lr, lsl #8
-    strh    r6, [r2]
-    mov     r6, r6, lsr #16
-    strh    r6, [r2]
+    mov     r12, r12, lsr #24
+    orr     r12, r12, lr, lsl #8
+    strh    r12, [r2]
+    mov     r12, r12, lsr #16
+    strh    r12, [r2]
     mov     r3, lr, lsr #24
     subs    r1, r1, #8          /* 8 or more halfwords left? */
     bge     .w_loop_u
@@ -294,7 +294,7 @@ copy_write_sectors:
     orr     r3, r3, r4, lsl #8
     strh    r3, [r2]            /* write final halfword */
 
-    ldmfd   sp!, {r4, r5, r6, pc}
+    ldmfd   sp!, {r4, r5, pc}
 
     /* 16-bit aligned */
 .w_aligned:
@@ -305,7 +305,7 @@ copy_write_sectors:
 
     sub     r1, r1, #8          /* adjust for zero-check and doing 8 halfwords/loop */
 .w_loop_a:
-    ldmia   r0!, {r3, r4, r5, r6}
+    ldmia   r0!, {r3, r4, r5, r12}
     strh    r3, [r2]            /* Load 4 words and decompose them into */
     mov     r3, r3, lsr #16     /*  2 halfwords each, and write those. */
     strh    r3, [r2]
@@ -315,9 +315,9 @@ copy_write_sectors:
     strh    r5, [r2]
     mov     r5, r5, lsr #16
     strh    r5, [r2]
-    strh    r6, [r2]
-    mov     r6, r6, lsr #16
-    strh    r6, [r2]
+    strh    r12, [r2]
+    mov     r12, r12, lsr #16
+    strh    r12, [r2]
     subs    r1, r1, #8          /* 8 or more halfwords left? */
     bge     .w_loop_a
 
@@ -343,7 +343,7 @@ copy_write_sectors:
     ldrneh  r3, [r0], #2
     strneh  r3, [r2]
 
-    ldmfd   sp!, {r4, r5, r6, pc}
+    ldmfd   sp!, {r4, r5, pc}
 
 .w_end:
     .size   copy_write_sectors,.w_end-copy_write_sectors
diff --git a/firmware/target/arm/ipod/lcd-as-gray.S b/firmware/target/arm/ipod/lcd-as-gray.S
index a066e55a05..b6878a1a78 100644
--- a/firmware/target/arm/ipod/lcd-as-gray.S
+++ b/firmware/target/arm/ipod/lcd-as-gray.S
@@ -34,8 +34,7 @@
     .type       lcd_write_data,%function
 
 lcd_write_data:
-    str     lr, [sp, #-4]!
-    ldr     lr, =LCD1_BASE
+    ldr     r12, =LCD1_BASE
 
 .loop:
     ldrb    r2, [r0], #1
@@ -45,29 +44,29 @@ lcd_write_data:
     orr     r2, r3, r2, lsl #8
     orr     r2, r2, #0x760000
 1:
-    ldr     r3, [lr]
+    ldr     r3, [r12]
     tst     r3, #LCD1_BUSY_MASK
     bne     1b
-    str     r2, [lr, #0x08]
+    str     r2, [r12, #0x08]
 #else
 1:
-    ldr     r3, [lr]
+    ldr     r3, [r12]
     tst     r3, #LCD1_BUSY_MASK
     bne     1b
-    str     r2, [lr, #0x10]
+    str     r2, [r12, #0x10]
 
     ldrb    r2, [r0], #1
 1:
-    ldr     r3, [lr]
+    ldr     r3, [r12]
     tst     r3, #LCD1_BUSY_MASK
     bne     1b
-    str     r2, [lr, #0x10]
+    str     r2, [r12, #0x10]
 #endif
 
     subs    r1, r1, #1
     bne     .loop
 
-    ldr     pc, [sp], #4
+    bx      lr
     .size   lcd_write_data,.-lcd_write_data
     
 
@@ -77,9 +76,9 @@ lcd_write_data:
     .type       lcd_write_data_shifted,%function
     
 lcd_write_data_shifted:
-    stmfd   sp!, {r4, r5, lr}
+    stmfd   sp!, {r4, lr}
     ldr     lr, =LCD1_BASE
-    mov     r5, #0x760000
+    mov     r12, #0x760000
     ldrb    r2, [r0], #1
 
 .sloop:
@@ -88,7 +87,7 @@ lcd_write_data_shifted:
     ldrb    r3, [r0], #1
     orr     r2, r3, r2, lsl #8
     mov     r4, r2, lsl #12
-    orr     r4, r5, r4, lsr #16
+    orr     r4, r12, r4, lsr #16
 1:
     ldr     r3, [lr]
     tst     r3, #LCD1_BUSY_MASK
@@ -98,7 +97,7 @@ lcd_write_data_shifted:
     subs    r1, r1, #1
     bne     .sloop
 
-    ldmfd   sp!, {r4, r5, pc}
+    ldmfd   sp!, {r4, pc}
     .size   lcd_write_data_shifted,.-lcd_write_data_shifted
     
 #elif defined IPOD_MINI
@@ -107,33 +106,33 @@ lcd_write_data_shifted:
     .type       lcd_write_data_shifted,%function
     
 lcd_write_data_shifted:
-    stmfd   sp!, {r4, lr}
+    str     lr, [sp, #-4]!
     ldr     lr, =LCD1_BASE
     ldrb    r2, [r0], #1
 
 .sloop:
     ldrb    r3, [r0], #1
     orr     r2, r3, r2, lsl #8
-    mov     r4, r2, lsr #4
+    mov     r12, r2, lsr #4
 1:
     ldr     r3, [lr]
     tst     r3, #LCD1_BUSY_MASK
     bne     1b
-    str     r4, [lr, #0x10]
+    str     r12, [lr, #0x10]
 
     ldrb    r3, [r0], #1
     orr     r2, r3, r2, lsl #8
-    mov     r4, r2, lsr #4
+    mov     r12, r2, lsr #4
 1:
     ldr     r3, [lr]
     tst     r3, #LCD1_BUSY_MASK
     bne     1b
-    str     r4, [lr, #0x10]
+    str     r12, [lr, #0x10]
 
     subs    r1, r1, #1
     bne     .sloop
 
-    ldmfd   sp!, {r4, pc}
+    ldr     pc, [sp], #4
     .size   lcd_write_data_shifted,.-lcd_write_data_shifted
 
 #endif
@@ -142,18 +141,18 @@ lcd_write_data_shifted:
     .type       lcd_mono_data,%function
     
 lcd_mono_data:
-    stmfd   sp!, {r4, r5, lr}
+    stmfd   sp!, {r4, lr}
     ldr     lr, =LCD1_BASE
-    adr     r5, .dibits
+    adr     r12, .dibits
 
 .mloop:
     ldrb    r2, [r0], #1
     mov     r3, r2, lsr #4
-    ldrb    r4, [r5, r3]
+    ldrb    r4, [r12, r3]
 
 #ifdef IPOD_MINI2G
     and     r3, r2, #0x0f
-    ldrb    r3, [r5, r3]
+    ldrb    r3, [r12, r3]
     orr     r4, r3, r4, lsl #8
     orr     r4, r4, #0x760000
 1:
@@ -169,7 +168,7 @@ lcd_mono_data:
     str     r4, [lr, #0x10]
 
     and     r3, r2, #0x0f
-    ldrb    r4, [r5, r3]
+    ldrb    r4, [r12, r3]
 1:
     ldr     r3, [lr]
     tst     r3, #LCD1_BUSY_MASK
@@ -180,7 +179,7 @@ lcd_mono_data:
     subs    r1, r1, #1
     bne     .mloop
 
-    ldmfd   sp!, {r4, r5, pc}
+    ldmfd   sp!, {r4, pc}
 
 .dibits:
     .byte   0x00, 0x03, 0x0C, 0x0F, 0x30, 0x33, 0x3C, 0x3F
@@ -203,16 +202,16 @@ lcd_mono_data:
  * Register usage:
  *   r3/r4 - current block of phases
  *   r5/r6 - current block of values
- *   r7 - lcd data accumulator
- *   r8 - phase signs mask
- *   lr - lcd bridge address
+ *   r7  - lcd data accumulator
+ *   r12 - phase signs mask
+ *   lr  - lcd bridge address
  */
  
 lcd_grey_data:
-    stmfd   sp!, {r4-r8, lr}
-    mov     r8, #0x80
-    orr     r8, r8, r8, lsl #8
-    orr     r8, r8, r8, lsl #16
+    stmfd   sp!, {r4-r7, lr}
+    mov     r12, #0x80
+    orr     r12, r12, r12, lsl #8
+    orr     r12, r12, r12, lsl #16
     ldr     lr, =LCD1_BASE
 
 .greyloop:
@@ -229,7 +228,7 @@ lcd_grey_data:
     orreq   r7, r7, #0x0c00
     tst     r3, #0x80000000
     orreq   r7, r7, #0x0300
-    bic     r3, r3, r8
+    bic     r3, r3, r12
     add     r3, r3, r5
 #else                           /* Parallel bridge mode */
     mov     r7, #0
@@ -241,7 +240,7 @@ lcd_grey_data:
     orreq   r7, r7, #0x0c
     tst     r3, #0x80000000
     orreq   r7, r7, #0x03
-    bic     r3, r3, r8
+    bic     r3, r3, r12
     add     r3, r3, r5
 
 1:
@@ -261,7 +260,7 @@ lcd_grey_data:
     orreq   r7, r7, #0x0c
     tst     r4, #0x80000000
     orreq   r7, r7, #0x03
-    bic     r4, r4, r8
+    bic     r4, r4, r12
     add     r4, r4, r6
 
     stmia   r1!, {r3-r4}
@@ -279,6 +278,6 @@ lcd_grey_data:
     subs    r2, r2, #1
     bne     .greyloop
 
-    ldmfd   sp!, {r4-r8, pc}
+    ldmfd   sp!, {r4-r7, pc}
     .size   lcd_grey_data,.-lcd_grey_data
 
diff --git a/firmware/target/arm/ipod/video/lcd-as-video.S b/firmware/target/arm/ipod/video/lcd-as-video.S
index e14b9dafb8..4b0cabcec2 100644
--- a/firmware/target/arm/ipod/video/lcd-as-video.S
+++ b/firmware/target/arm/ipod/video/lcd-as-video.S
@@ -32,22 +32,22 @@
     .type   lcd_write_data, %function
                                       /* r0 = addr, must be aligned */
 lcd_write_data:                       /* r1 = pixel count, must be even */
-    stmfd   sp!, {r4-r5, lr}
+    stmfd   sp!, {r4, lr}
     mov     lr, #0x30000000           /* LCD data port */
 
     subs    r1, r1, #16
 .loop16:
-    ldmgeia r0!, {r2-r5}
-    stmgeia lr, {r2-r5}
-    ldmgeia r0!, {r2-r5}
-    stmgeia lr, {r2-r5}
+    ldmgeia r0!, {r2-r4, r12}
+    stmgeia lr, {r2-r4, r12}
+    ldmgeia r0!, {r2-r4, r12}
+    stmgeia lr, {r2-r4, r12}
     subges  r1, r1, #16
     bge     .loop16
 
     /* no need to correct the count, we're just checking bits from now */
     tst     r1, #8
-    ldmneia r0!, {r2-r5}
-    stmneia lr, {r2-r5}
+    ldmneia r0!, {r2-r4, r12}
+    stmneia lr, {r2-r4, r12}
     tst     r1, #4
     ldmneia r0!, {r2-r3}
     stmneia lr, {r2-r3}
@@ -55,7 +55,7 @@ lcd_write_data:                       /* r1 = pixel count, must be even */
     ldrne   r3, [r0], #4
     strne   r3, [lr]
 
-    ldmfd   sp!, {r4-r5, pc}
+    ldmfd   sp!, {r4, pc}
 
 /****************************************************************************
  * extern void lcd_write_yuv420_lines(unsigned char const * const src[3],
@@ -87,10 +87,10 @@ lcd_write_yuv420_lines:
                                       /* r1 = dst = bcmaddr */
                                       /* r2 = width */
                                       /* r3 = stride */                
-    stmfd       sp!, { r4-r11, lr }   /* save non-scratch */
-    ldmia       r0, { r9-r11 }        /* r9 = yuv_src[0] = Y'_p */
+    stmfd       sp!, { r4-r10, lr }   /* save non-scratch */
+    ldmia       r0, { r9, r10, r12 }  /* r9 = yuv_src[0] = Y'_p */
                                       /* r10 = yuv_src[1] = Cb_p */
-                                      /* r11 = yuv_src[2] = Cr_p */
+                                      /* r12 = yuv_src[2] = Cr_p */
     add         r3, r9, r3            /* r3 = &ysrc[stride] */
     add         r4, r2, r2, asr #1    /* chroma buffer lenght = width/2 *3 */
     mov         r4, r4, asl #2        /*   use words for str/ldm possibility */
@@ -116,7 +116,7 @@ lcd_write_yuv420_lines:
 10:                                   /* loop start */
 
     ldrb        r0, [r10], #1         /* r0 = *usrc++ = *Cb_p++ */
-    ldrb        r1, [r11], #1         /* r1 = *vsrc++ = *Cr_p++ */
+    ldrb        r1, [r12], #1         /* r1 = *vsrc++ = *Cr_p++ */
 
     sub         r0, r0, #128          /* r0 = Cb-128 */
     sub         r1, r1, #128          /* r1 = Cr-128 */
@@ -289,7 +289,7 @@ lcd_write_yuv420_lines:
 
     ldr         r3, [sp, #12]
     add         sp, sp, r3            /* deallocate buffer */
-    ldmfd       sp!, { r4-r11, pc }   /* restore registers */
+    ldmfd       sp!, { r4-r10, pc }   /* restore registers */
 
     .ltorg
     .size   lcd_write_yuv420_lines, .-lcd_write_yuv420_lines
diff --git a/firmware/target/arm/iriver/h10/lcd-as-h10.S b/firmware/target/arm/iriver/h10/lcd-as-h10.S
index ba6c876d44..e99eee50c9 100644
--- a/firmware/target/arm/iriver/h10/lcd-as-h10.S
+++ b/firmware/target/arm/iriver/h10/lcd-as-h10.S
@@ -49,7 +49,7 @@ lcd_write_yuv420_lines:
                                         @ r0 = yuv_src
                                         @ r1 = width
                                         @ r2 = stride
-    stmfd       sp!, { r4-r12, lr }     @ save non-scratch
+    stmfd       sp!, { r4-r11, lr }     @ save non-scratch
     ldmia       r0, { r4, r5, r6 }      @ r4 = yuv_src[0] = Y'_p
                                         @ r5 = yuv_src[1] = Cb_p
                                         @ r6 = yuv_src[2] = Cr_p
@@ -232,7 +232,7 @@ lcd_write_yuv420_lines:
     subs        r1, r1, #2              @ subtract block from width
     bgt         10b @ loop line         @
                                         @
-    ldmfd       sp!, { r4-r12, pc }     @ restore registers and return
+    ldmfd       sp!, { r4-r11, pc }     @ restore registers and return
     .ltorg                              @ dump constant pool
     .size   lcd_write_yuv420_lines, .-lcd_write_yuv420_lines
 
@@ -273,7 +273,7 @@ lcd_write_yuv420_lines_odither:
                                         @ r2   = stride
                                         @ r3   = x_screen
                                         @ [sp] = y_screen
-    stmfd       sp!, { r4-r12, lr }     @ save non-scratch
+    stmfd       sp!, { r4-r11, lr }     @ save non-scratch
     ldmia       r0, { r4, r5, r6 }      @ r4 = yuv_src[0] = Y'_p
                                         @ r5 = yuv_src[1] = Cb_p
                                         @ r6 = yuv_src[2] = Cr_p
@@ -533,6 +533,6 @@ lcd_write_yuv420_lines_odither:
     subs        r1, r1, #2              @ subtract block from width
     bgt         10b @ loop line         @
                                         @
-    ldmfd       sp!, { r4-r12, pc }     @ restore registers and return
+    ldmfd       sp!, { r4-r11, pc }     @ restore registers and return
     .ltorg                              @ dump constant pool
     .size   lcd_write_yuv420_lines_odither, .-lcd_write_yuv420_lines_odither
diff --git a/firmware/target/arm/lcd-as-memframe.S b/firmware/target/arm/lcd-as-memframe.S
index 84f549036f..2d0aff937f 100644
--- a/firmware/target/arm/lcd-as-memframe.S
+++ b/firmware/target/arm/lcd-as-memframe.S
@@ -37,7 +37,7 @@
                                         @ r2 = width
                                         @ r3 = height
 lcd_copy_buffer_rect:                   @
-    stmfd   sp!, { r4-r12, lr }         @ save non-scratch regs
+    stmfd   sp!, { r4-r11, lr }         @ save non-scratch regs
     mov     r5, r2                      @ r5 = cached width
     rsb     r4, r2, #LCD_WIDTH          @ r4 = LCD_WIDTH - width
 10: @ copy line                         @
@@ -101,7 +101,7 @@ lcd_copy_buffer_rect:                   @
     add     r0, r0, r4, lsl #1          @
     subs    r3, r3, #1                  @ next line
     bgt     10b @ copy line             @
-    ldmfd   sp!, { r4-r12, pc }         @ restore regs and return
+    ldmfd   sp!, { r4-r11, pc }         @ restore regs and return
     .ltorg                              @ dump constant pool
     .size   lcd_copy_buffer_rect, .-lcd_copy_buffer_rect
 
@@ -132,7 +132,7 @@ lcd_write_yuv420_lines:
                                         @ r1 = yuv_src
                                         @ r2 = width
                                         @ r3 = stride
-    stmfd       sp!, { r4-r12 }         @ save non-scratch
+    stmfd       sp!, { r4-r10, lr }     @ save non-scratch
     ldmia       r1, { r4, r5, r6 }      @ r4 = yuv_src[0] = Y'_p
                                         @ r5 = yuv_src[1] = Cb_p
                                         @ r6 = yuv_src[2] = Cr_p
@@ -155,9 +155,9 @@ lcd_write_yuv420_lines:
     add         r10, r10, r8, asl #3    @
     add         r10, r10, r8, asl #4    @
                                         @
-    add         r11, r9, r9, asl #2     @ r9 = Cr*101
-    add         r11, r11, r9, asl #5    @
-    add         r9, r11, r9, asl #6     @
+    add         lr, r9, r9, asl #2      @ r9 = Cr*101
+    add         lr, lr, r9, asl #5      @
+    add         r9, lr, r9, asl #6      @
                                         @
     add         r8, r8, #2              @ r8 = bu = (Cb*128 + 128) >> 8
     mov         r8, r8, asr #2          @
@@ -167,19 +167,19 @@ lcd_write_yuv420_lines:
     mov         r10, r10, asr #8        @
                                         @ compute R, G, and B
     add         r1, r8, r7, asr #8      @ r1  = b = (Y >> 9) + bu
-    add         r11, r9, r7, asr #8     @ r11 = r = (Y >> 9) + rv
+    add         lr, r9, r7, asr #8      @ lr = r = (Y >> 9) + rv
     add         r7, r10, r7, asr #7     @ r7  = g = (Y >> 8) + guv
                                         @
-    orr         r12, r1, r11            @ check if clamping is needed...
+    orr         r12, r1, lr             @ check if clamping is needed...
     orr         r12, r12, r7, asr #1    @ ...at all
     cmp         r12, #31                @
     bls         15f @ no clamp          @
     cmp         r1, #31                 @ clamp b
     mvnhi       r1, r1, asr #31         @
     andhi       r1, r1, #31             @
-    cmp         r11, #31                @ clamp r
-    mvnhi       r11, r11, asr #31       @
-    andhi       r11, r11, #31           @
+    cmp         lr, #31                 @ clamp r
+    mvnhi       lr, lr, asr #31         @
+    andhi       lr, lr, #31             @
     cmp         r7, #63                 @ clamp g
     mvnhi       r7, r7, asr #31         @
     andhi       r7, r7, #63             @
@@ -188,7 +188,7 @@ lcd_write_yuv420_lines:
     ldrb        r12, [r4, r3]           @ r12 = Y' = *(Y'_p + stride)
                                         @
     orr         r1, r1, r7, lsl #5      @ r4 |= (g << 5)
-    orr         r1, r1, r11, lsl #11    @ r4 = b | (r << 11)
+    orr         r1, r1, lr, lsl #11     @ r4 = b | (r << 11)
 
 #if LCD_WIDTH >= LCD_HEIGHT
     strh        r1, [r0]                @
@@ -203,19 +203,19 @@ lcd_write_yuv420_lines:
     add         r7, r12, r7, asl #5     @
                                         @ compute R, G, and B
     add         r1, r8, r7, asr #8      @ r1  = b = (Y >> 9) + bu
-    add         r11, r9, r7, asr #8     @ r11 = r = (Y >> 9) + rv
+    add         lr, r9, r7, asr #8      @ lr = r = (Y >> 9) + rv
     add         r7, r10, r7, asr #7     @ r7  = g = (Y >> 8) + guv
                                         @
-    orr         r12, r1, r11            @ check if clamping is needed...
+    orr         r12, r1, lr             @ check if clamping is needed...
     orr         r12, r12, r7, asr #1    @ ...at all
     cmp         r12, #31                @
     bls         15f @ no clamp          @
     cmp         r1, #31                 @ clamp b
     mvnhi       r1, r1, asr #31         @
     andhi       r1, r1, #31             @
-    cmp         r11, #31                @ clamp r
-    mvnhi       r11, r11, asr #31       @
-    andhi       r11, r11, #31           @
+    cmp         lr, #31                 @ clamp r
+    mvnhi       lr, lr, asr #31         @
+    andhi       lr, lr, #31             @
     cmp         r7, #63                 @ clamp g
     mvnhi       r7, r7, asr #31         @
     andhi       r7, r7, #63             @
@@ -223,7 +223,7 @@ lcd_write_yuv420_lines:
                                         @
     ldrb        r12, [r4], #1           @ r12 = Y' = *(Y'_p++)
                                         @
-    orr         r1, r1, r11, lsl #11    @ r1 = b | (r << 11)
+    orr         r1, r1, lr, lsl #11     @ r1 = b | (r << 11)
     orr         r1, r1, r7, lsl #5      @ r1 |= (g << 5)
 
 #if LCD_WIDTH >= LCD_HEIGHT
@@ -242,19 +242,19 @@ lcd_write_yuv420_lines:
     add         r7, r12, r7, asl #5     @
                                         @ compute R, G, and B
     add         r1, r8, r7, asr #8      @ r1  = b = (Y >> 9) + bu
-    add         r11, r9, r7, asr #8     @ r11 = r = (Y >> 9) + rv
+    add         lr, r9, r7, asr #8      @ lr = r = (Y >> 9) + rv
     add         r7, r10, r7, asr #7     @ r7  = g = (Y >> 8) + guv
                                         @
-    orr         r12, r1, r11            @ check if clamping is needed...
+    orr         r12, r1, lr             @ check if clamping is needed...
     orr         r12, r12, r7, asr #1    @ ...at all
     cmp         r12, #31                @
     bls         15f @ no clamp          @
     cmp         r1, #31                 @ clamp b
     mvnhi       r1, r1, asr #31         @
     andhi       r1, r1, #31             @
-    cmp         r11, #31                @ clamp r
-    mvnhi       r11, r11, asr #31       @
-    andhi       r11, r11, #31           @
+    cmp         lr, #31                 @ clamp r
+    mvnhi       lr, lr, asr #31         @
+    andhi       lr, lr, #31             @
     cmp         r7, #63                 @ clamp g
     mvnhi       r7, r7, asr #31         @
     andhi       r7, r7, #63             @
@@ -263,7 +263,7 @@ lcd_write_yuv420_lines:
     ldrb        r12, [r4, r3]           @ r12 = Y' = *(Y'_p + stride)
                                         @
     orr         r1, r1, r7, lsl #5      @ r1 = b | (g << 5)
-    orr         r1, r1, r11, lsl #11    @ r1 |= (r << 11)
+    orr         r1, r1, lr, lsl #11     @ r1 |= (r << 11)
 
 #if LCD_WIDTH >= LCD_HEIGHT
     strh        r1, [r0, #2]
@@ -278,25 +278,25 @@ lcd_write_yuv420_lines:
     add         r7, r12, r7, asl #5     @
                                         @ compute R, G, and B
     add         r1, r8, r7, asr #8      @ r1  = b = (Y >> 9) + bu
-    add         r11, r9, r7, asr #8     @ r11 = r = (Y >> 9) + rv
+    add         lr, r9, r7, asr #8      @ lr = r = (Y >> 9) + rv
     add         r7, r10, r7, asr #7     @ r7  = g = (Y >> 8) + guv
                                         @
-    orr         r12, r1, r11            @ check if clamping is needed...
+    orr         r12, r1, lr             @ check if clamping is needed...
     orr         r12, r12, r7, asr #1    @ ...at all
     cmp         r12, #31                @
     bls         15f @ no clamp          @
     cmp         r1, #31                 @ clamp b
     mvnhi       r1, r1, asr #31         @
     andhi       r1, r1, #31             @
-    cmp         r11, #31                @ clamp r
-    mvnhi       r11, r11, asr #31       @
-    andhi       r11, r11, #31           @
+    cmp         lr, #31                 @ clamp r
+    mvnhi       lr, lr, asr #31         @
+    andhi       lr, lr, #31             @
     cmp         r7, #63                 @ clamp g
     mvnhi       r7, r7, asr #31         @
     andhi       r7, r7, #63             @
 15: @ no clamp                          @
                                         @
-    orr         r12, r1, r11, lsl #11   @ r12 = b | (r << 11)
+    orr         r12, r1, lr, lsl #11    @ r12 = b | (r << 11)
     orr         r12, r12, r7, lsl #5    @ r12 |= (g << 5)
 
 #if LCD_WIDTH >= LCD_HEIGHT
@@ -320,8 +320,7 @@ lcd_write_yuv420_lines:
     subs        r2, r2, #2              @ subtract block from width
     bgt         10b @ loop line         @
                                         @
-    ldmfd       sp!, { r4-r12 }         @ restore registers and return
-    bx          lr                      @
+    ldmfd       sp!, { r4-r10, pc }     @ restore registers and return
     .ltorg                              @ dump constant pool
     .size   lcd_write_yuv420_lines, .-lcd_write_yuv420_lines
 
@@ -364,7 +363,7 @@ lcd_write_yuv420_lines_odither:
                                         @ r3 = stride
                                         @ [sp]   = x_screen
                                         @ [sp+4] = y_screen
-    stmfd       sp!, { r4-r12, lr }     @ save non-scratch
+    stmfd       sp!, { r4-r11, lr }     @ save non-scratch
     ldmia       r1, { r4, r5, r6 }      @ r4 = yuv_src[0] = Y'_p
                                         @ r5 = yuv_src[1] = Cb_p
                                         @ r6 = yuv_src[2] = Cr_p
@@ -626,6 +625,6 @@ lcd_write_yuv420_lines_odither:
     subs        r2, r2, #2              @ subtract block from width
     bgt         10b @ loop line         @
                                         @
-    ldmfd       sp!, { r4-r12, pc }     @ restore registers and return
+    ldmfd       sp!, { r4-r11, pc }     @ restore registers and return
     .ltorg                              @ dump constant pool
     .size   lcd_write_yuv420_lines_odither, .-lcd_write_yuv420_lines_odither
diff --git a/firmware/target/arm/olympus/mrobe-100/lcd-as-mr100.S b/firmware/target/arm/olympus/mrobe-100/lcd-as-mr100.S
index 84f47addd8..2aede6d5e7 100644
--- a/firmware/target/arm/olympus/mrobe-100/lcd-as-mr100.S
+++ b/firmware/target/arm/olympus/mrobe-100/lcd-as-mr100.S
@@ -30,21 +30,20 @@
     .type       lcd_write_data,%function
 
 lcd_write_data:
-    str     lr, [sp, #-4]!
-    ldr     lr, =LCD1_BASE
+    ldr     r12, =LCD1_BASE
 
 .loop:
     ldrb    r2, [r0], #1
 1:
-    ldr     r3, [lr]
+    ldr     r3, [r12]
     tst     r3, #LCD1_BUSY_MASK
     bne     1b
-    str     r2, [lr, #0x10]
+    str     r2, [r12, #0x10]
 
     subs    r1, r1, #1
     bne     .loop
 
-    ldr     pc, [sp], #4
+    bx      lr
     .size   lcd_write_data,.-lcd_write_data
 
 
@@ -62,16 +61,16 @@ lcd_write_data:
  * Register usage:
  *   r3/r4 - current block of phases
  *   r5/r6 - current block of values
- *   r7 - lcd data accumulator
- *   r8 - phase signs mask
- *   lr - lcd bridge address
+ *   r7  - lcd data accumulator
+ *   r12 - phase signs mask
+ *   lr  - lcd bridge address
  */
  
 lcd_grey_data:
-    stmfd   sp!, {r4-r8, lr}
-    mov     r8, #0x80
-    orr     r8, r8, r8, lsl #8
-    orr     r8, r8, r8, lsl #16
+    stmfd   sp!, {r4-r7, lr}
+    mov     r12, #0x80
+    orr     r12, r12, r12, lsl #8
+    orr     r12, r12, r12, lsl #16
     ldr     lr, =LCD1_BASE
 
 .greyloop:
@@ -87,7 +86,7 @@ lcd_grey_data:
     biceq   r7, r7, #0x20
     tst     r3, #0x80000000
     biceq   r7, r7, #0x10
-    bic     r3, r3, r8
+    bic     r3, r3, r12
     add     r3, r3, r5
 
     tst     r4, #0x80
@@ -98,7 +97,7 @@ lcd_grey_data:
     biceq   r7, r7, #0x02
     tst     r4, #0x80000000
     biceq   r7, r7, #0x01
-    bic     r4, r4, r8
+    bic     r4, r4, r12
     add     r4, r4, r6
 
     stmia   r1!, {r3-r4}
@@ -112,6 +111,6 @@ lcd_grey_data:
     subs    r2, r2, #1
     bne     .greyloop
 
-    ldmfd   sp!, {r4-r8, pc}
+    ldmfd   sp!, {r4-r7, pc}
     .size   lcd_grey_data,.-lcd_grey_data
 
diff --git a/firmware/target/arm/philips/hdd1630/lcd-as-hdd1630.S b/firmware/target/arm/philips/hdd1630/lcd-as-hdd1630.S
index 8f3d12e727..74327dddaf 100644
--- a/firmware/target/arm/philips/hdd1630/lcd-as-hdd1630.S
+++ b/firmware/target/arm/philips/hdd1630/lcd-as-hdd1630.S
@@ -49,7 +49,7 @@ lcd_write_yuv420_lines:
                                         @ r0 = yuv_src
                                         @ r1 = width
                                         @ r2 = stride
-    stmfd       sp!, { r4-r12, lr }     @ save non-scratch
+    stmfd       sp!, { r4-r11, lr }     @ save non-scratch
     ldmia       r0, { r4, r5, r6 }      @ r4 = yuv_src[0] = Y'_p
                                         @ r5 = yuv_src[1] = Cb_p
                                         @ r6 = yuv_src[2] = Cr_p
@@ -248,7 +248,7 @@ lcd_write_yuv420_lines:
     subs        r1, r1, #2              @ subtract block from width
     bgt         10b @ loop line         @
                                         @
-    ldmfd       sp!, { r4-r12, pc }     @ restore registers and return
+    ldmfd       sp!, { r4-r11, pc }     @ restore registers and return
     .ltorg                              @ dump constant pool
     .size   lcd_write_yuv420_lines, .-lcd_write_yuv420_lines
 
@@ -289,7 +289,7 @@ lcd_write_yuv420_lines_odither:
                                         @ r2   = stride
                                         @ r3   = x_screen
                                         @ [sp] = y_screen
-    stmfd       sp!, { r4-r12, lr }     @ save non-scratch
+    stmfd       sp!, { r4-r11, lr }     @ save non-scratch
     ldmia       r0, { r4, r5, r6 }      @ r4 = yuv_src[0] = Y'_p
                                         @ r5 = yuv_src[1] = Cb_p
                                         @ r6 = yuv_src[2] = Cr_p
@@ -565,6 +565,6 @@ lcd_write_yuv420_lines_odither:
     subs        r1, r1, #2              @ subtract block from width
     bgt         10b @ loop line         @
                                         @
-    ldmfd       sp!, { r4-r12, pc }     @ restore registers and return
+    ldmfd       sp!, { r4-r11, pc }     @ restore registers and return
     .ltorg                              @ dump constant pool
     .size   lcd_write_yuv420_lines_odither, .-lcd_write_yuv420_lines_odither
diff --git a/firmware/target/arm/samsung/yh820/lcd-as-yh820.S b/firmware/target/arm/samsung/yh820/lcd-as-yh820.S
index e466624d04..7187b83661 100644
--- a/firmware/target/arm/samsung/yh820/lcd-as-yh820.S
+++ b/firmware/target/arm/samsung/yh820/lcd-as-yh820.S
@@ -48,7 +48,7 @@ lcd_write_yuv420_lines:
                                         @ r0 = yuv_src
                                         @ r1 = width
                                         @ r2 = stride
-    stmfd       sp!, { r4-r12 }         @ save non-scratch
+    stmfd       sp!, { r4-r10, lr }     @ save non-scratch
     ldmia       r0, { r4, r5, r6 }      @ r4 = yuv_src[0] = Y'_p
                                         @ r5 = yuv_src[1] = Cb_p
                                         @ r6 = yuv_src[2] = Cr_p
@@ -73,9 +73,9 @@ lcd_write_yuv420_lines:
     add         r10, r10, r8, asl #3    @
     add         r10, r10, r8, asl #4    @
                                         @
-    add         r11, r9, r9, asl #2     @ r9 = Cr*101
-    add         r11, r11, r9, asl #5    @
-    add         r9, r11, r9, asl #6     @
+    add         lr, r9, r9, asl #2      @ r9 = Cr*101
+    add         lr, lr, r9, asl #5      @
+    add         r9, lr, r9, asl #6      @
                                         @
     add         r8, r8, #2              @ r8 = bu = (Cb*128 + 128) >> 8
     mov         r8, r8, asr #2          @
@@ -85,19 +85,19 @@ lcd_write_yuv420_lines:
     mov         r10, r10, asr #8        @
                                         @ compute R, G, and B
     add         r0, r8, r7, asr #8      @ r0  = b = (Y >> 9) + bu
-    add         r11, r9, r7, asr #8     @ r11 = r = (Y >> 9) + rv
+    add         lr, r9, r7, asr #8      @ lr = r = (Y >> 9) + rv
     add         r7, r10, r7, asr #7     @ r7  = g = (Y >> 8) + guv
                                         @
-    orr         r12, r0, r11            @ check if clamping is needed...
+    orr         r12, r0, lr             @ check if clamping is needed...
     orr         r12, r12, r7, asr #1    @ ...at all
     cmp         r12, #31                @
     bls         15f @ no clamp          @
     cmp         r0, #31                 @ clamp b
     mvnhi       r0, r0, asr #31         @
     andhi       r0, r0, #31             @
-    cmp         r11, #31                @ clamp r
-    mvnhi       r11, r11, asr #31       @
-    andhi       r11, r11, #31           @
+    cmp         lr, #31                 @ clamp r
+    mvnhi       lr, lr, asr #31         @
+    andhi       lr, lr, #31             @
     cmp         r7, #63                 @ clamp g
     mvnhi       r7, r7, asr #31         @
     andhi       r7, r7, #63             @
@@ -105,14 +105,14 @@ lcd_write_yuv420_lines:
                                         @
     ldrb        r12, [r4, r2]           @ r12 = Y' = *(Y'_p + stride)
                                         @
-    mov         r11, r11, lsl #3        @
-    orr         r11, r11, r7, lsr #3    @ r11 = (r << 3) | (g >> 3)
+    mov         lr, lr, lsl #3          @
+    orr         lr, lr, r7, lsr #3      @ lr = (r << 3) | (g >> 3)
     orr         r0, r0, r7, lsl #5      @ r0 = (g << 5) | b
 1: @ busy                               @
     ldr         r7, [r3]                @ r7 = LCD1_BASE
     tst         r7, #LCD1_BUSY_MASK     @ bridge busy?
     bne         1b                      @
-    str         r11, [r3, #0x10]        @ send MSB
+    str         lr, [r3, #0x10]         @ send MSB
 1: @busy                                @
     ldr         r7, [r3]                @ r7 = LCD1_BASE
     tst         r7, #LCD1_BUSY_MASK     @ bridge busy?
@@ -124,19 +124,19 @@ lcd_write_yuv420_lines:
     add         r7, r12, r7, asl #5     @
                                         @ compute R, G, and B
     add         r0, r8, r7, asr #8      @ r0  = b = (Y >> 9) + bu
-    add         r11, r9, r7, asr #8     @ r11 = r = (Y >> 9) + rv
+    add         lr, r9, r7, asr #8      @ lr = r = (Y >> 9) + rv
     add         r7, r10, r7, asr #7     @ r7  = g = (Y >> 8) + guv
                                         @
-    orr         r12, r0, r11            @ check if clamping is needed...
+    orr         r12, r0, lr             @ check if clamping is needed...
     orr         r12, r12, r7, asr #1    @ ...at all
     cmp         r12, #31                @
     bls         15f @ no clamp          @
     cmp         r0, #31                 @ clamp b
     mvnhi       r0, r0, asr #31         @
     andhi       r0, r0, #31             @
-    cmp         r11, #31                @ clamp r
-    mvnhi       r11, r11, asr #31       @
-    andhi       r11, r11, #31           @
+    cmp         lr, #31                 @ clamp r
+    mvnhi       lr, lr, asr #31         @
+    andhi       lr, lr, #31             @
     cmp         r7, #63                 @ clamp g
     mvnhi       r7, r7, asr #31         @
     andhi       r7, r7, #63             @
@@ -144,14 +144,14 @@ lcd_write_yuv420_lines:
                                         @
     ldrb        r12, [r4], #1           @ r12 = Y' = *(Y'_p++)
                                         @
-    mov         r11, r11, lsl #3        @
-    orr         r11, r11, r7, lsr #3    @ r11 = (r << 3) | (g >> 3)
+    mov         lr, lr, lsl #3          @
+    orr         lr, lr, r7, lsr #3      @ lr = (r << 3) | (g >> 3)
     orr         r0, r0, r7, lsl #5      @ r0 = (g << 5) | b
 1: @ busy                               @
     ldr         r7, [r3]                @ r7 = LCD1_BASE
     tst         r7, #LCD1_BUSY_MASK     @ bridge busy?
     bne         1b                      @
-    str         r11, [r3, #0x10]        @ send MSB
+    str         lr, [r3, #0x10]         @ send MSB
 1: @ busy                               @
     ldr         r7, [r3]                @ r7 = LCD1_BASE
     tst         r7, #LCD1_BUSY_MASK     @ bridge busy?
@@ -163,19 +163,19 @@ lcd_write_yuv420_lines:
     add         r7, r12, r7, asl #5     @
                                         @ compute R, G, and B
     add         r0, r8, r7, asr #8      @ r0  = b = (Y >> 9) + bu
-    add         r11, r9, r7, asr #8     @ r11 = r = (Y >> 9) + rv
+    add         lr, r9, r7, asr #8      @ lr = r = (Y >> 9) + rv
     add         r7, r10, r7, asr #7     @ r7  = g = (Y >> 8) + guv
                                         @
-    orr         r12, r0, r11            @ check if clamping is needed...
+    orr         r12, r0, lr             @ check if clamping is needed...
     orr         r12, r12, r7, asr #1    @ ...at all
     cmp         r12, #31                @
     bls         15f @ no clamp          @
     cmp         r0, #31                 @ clamp b
     mvnhi       r0, r0, asr #31         @
     andhi       r0, r0, #31             @
-    cmp         r11, #31                @ clamp r
-    mvnhi       r11, r11, asr #31       @
-    andhi       r11, r11, #31           @
+    cmp         lr, #31                 @ clamp r
+    mvnhi       lr, lr, asr #31         @
+    andhi       lr, lr, #31             @
     cmp         r7, #63                 @ clamp g
     mvnhi       r7, r7, asr #31         @
     andhi       r7, r7, #63             @
@@ -184,14 +184,14 @@ lcd_write_yuv420_lines:
     ldrb        r12, [r4, r2]           @ r12 = Y' = *(Y'_p + stride)
                                         @
                                         @
-    mov         r11, r11, lsl #3        @
-    orr         r11, r11, r7, lsr #3    @ r11 = (r << 3) | (g >> 3)
+    mov         lr, lr, lsl #3          @
+    orr         lr, lr, r7, lsr #3      @ lr = (r << 3) | (g >> 3)
     orr         r0, r0, r7, lsl #5      @ r0 = (g << 5) | b
 1: @ busy                               @
     ldr         r7, [r3]                @ r7 = LCD1_BASE
     tst         r7, #LCD1_BUSY_MASK     @ bridge busy?
     bne         1b                      @
-    str         r11, [r3, #0x10]        @ send MSB
+    str         lr, [r3, #0x10]         @ send MSB
 1: @ busy                               @
     ldr         r7, [r3]                @ r7 = LCD1_BASE
     tst         r7, #LCD1_BUSY_MASK     @ bridge busy?
@@ -203,32 +203,32 @@ lcd_write_yuv420_lines:
     add         r7, r12, r7, asl #5     @
                                         @ compute R, G, and B
     add         r0, r8, r7, asr #8      @ r0  = b = (Y >> 9) + bu
-    add         r11, r9, r7, asr #8     @ r11 = r = (Y >> 9) + rv
+    add         lr, r9, r7, asr #8      @ lr = r = (Y >> 9) + rv
     add         r7, r10, r7, asr #7     @ r7  = g = (Y >> 8) + guv
                                         @
-    orr         r12, r0, r11            @ check if clamping is needed...
+    orr         r12, r0, lr             @ check if clamping is needed...
     orr         r12, r12, r7, asr #1    @ ...at all
     cmp         r12, #31                @
     bls         15f @ no clamp          @
     cmp         r0, #31                 @ clamp b
     mvnhi       r0, r0, asr #31         @
     andhi       r0, r0, #31             @
-    cmp         r11, #31                @ clamp r
-    mvnhi       r11, r11, asr #31       @
-    andhi       r11, r11, #31           @
+    cmp         lr, #31                 @ clamp r
+    mvnhi       lr, lr, asr #31         @
+    andhi       lr, lr, #31             @
     cmp         r7, #63                 @ clamp g
     mvnhi       r7, r7, asr #31         @
     andhi       r7, r7, #63             @
 15: @ no clamp                          @
                                         @
-    mov         r11, r11, lsl #3        @
-    orr         r11, r11, r7, lsr #3    @ r11 = (r << 3) | (g >> 3)
+    mov         lr, lr, lsl #3          @
+    orr         lr, lr, r7, lsr #3      @ lr = (r << 3) | (g >> 3)
     orr         r0, r0, r7, lsl #5      @ r0 = (g << 5) | b
 1: @ busy                               @
     ldr         r7, [r3]                @ r7 = LCD1_BASE
     tst         r7, #LCD1_BUSY_MASK     @ bridge busy?
     bne         1b                      @
-    str         r11, [r3, #0x10]        @ send MSB
+    str         lr, [r3, #0x10]         @ send MSB
 1: @ busy                               @
     ldr         r7, [r3]                @ r7 = LCD1_BASE
     tst         r7, #LCD1_BUSY_MASK     @ bridge busy?
@@ -238,8 +238,7 @@ lcd_write_yuv420_lines:
     subs        r1, r1, #2              @ subtract block from width
     bgt         10b @ loop line         @
                                         @
-    ldmfd       sp!, { r4-r12 }         @ restore registers and return
-    bx          lr                      @
+    ldmfd       sp!, { r4-r10, pc }     @ restore registers and return
     .ltorg                              @ dump constant pool
     .size   lcd_write_yuv420_lines, .-lcd_write_yuv420_lines
 
@@ -279,7 +278,7 @@ lcd_write_yuv420_lines_odither:
                                         @ r2 = stride
                                         @ r3 = x_screen
                                         @ [sp] = y_screen
-    stmfd       sp!, { r4-r12, lr }     @ save non-scratch
+    stmfd       sp!, { r4-r11, lr }     @ save non-scratch
     ldmia       r0, { r4, r5, r6 }      @ r4 = yuv_src[0] = Y'_p
                                         @ r5 = yuv_src[1] = Cb_p
                                         @ r6 = yuv_src[2] = Cr_p
@@ -546,6 +545,6 @@ lcd_write_yuv420_lines_odither:
     subs        r1, r1, #2              @ subtract block from width
     bgt         10b @ loop line         @
                                         @
-    ldmfd       sp!, { r4-r12, pc }     @ restore registers and return
+    ldmfd       sp!, { r4-r11, pc }     @ restore registers and return
     .ltorg                              @ dump constant pool
     .size   lcd_write_yuv420_lines_odither, .-lcd_write_yuv420_lines_odither
diff --git a/firmware/target/arm/samsung/yh925/lcd-as-yh925.S b/firmware/target/arm/samsung/yh925/lcd-as-yh925.S
index ba6c876d44..e99eee50c9 100644
--- a/firmware/target/arm/samsung/yh925/lcd-as-yh925.S
+++ b/firmware/target/arm/samsung/yh925/lcd-as-yh925.S
@@ -49,7 +49,7 @@ lcd_write_yuv420_lines:
                                         @ r0 = yuv_src
                                         @ r1 = width
                                         @ r2 = stride
-    stmfd       sp!, { r4-r12, lr }     @ save non-scratch
+    stmfd       sp!, { r4-r11, lr }     @ save non-scratch
     ldmia       r0, { r4, r5, r6 }      @ r4 = yuv_src[0] = Y'_p
                                         @ r5 = yuv_src[1] = Cb_p
                                         @ r6 = yuv_src[2] = Cr_p
@@ -232,7 +232,7 @@ lcd_write_yuv420_lines:
     subs        r1, r1, #2              @ subtract block from width
     bgt         10b @ loop line         @
                                         @
-    ldmfd       sp!, { r4-r12, pc }     @ restore registers and return
+    ldmfd       sp!, { r4-r11, pc }     @ restore registers and return
     .ltorg                              @ dump constant pool
     .size   lcd_write_yuv420_lines, .-lcd_write_yuv420_lines
 
@@ -273,7 +273,7 @@ lcd_write_yuv420_lines_odither:
                                         @ r2   = stride
                                         @ r3   = x_screen
                                         @ [sp] = y_screen
-    stmfd       sp!, { r4-r12, lr }     @ save non-scratch
+    stmfd       sp!, { r4-r11, lr }     @ save non-scratch
     ldmia       r0, { r4, r5, r6 }      @ r4 = yuv_src[0] = Y'_p
                                         @ r5 = yuv_src[1] = Cb_p
                                         @ r6 = yuv_src[2] = Cr_p
@@ -533,6 +533,6 @@ lcd_write_yuv420_lines_odither:
     subs        r1, r1, #2              @ subtract block from width
     bgt         10b @ loop line         @
                                         @
-    ldmfd       sp!, { r4-r12, pc }     @ restore registers and return
+    ldmfd       sp!, { r4-r11, pc }     @ restore registers and return
     .ltorg                              @ dump constant pool
     .size   lcd_write_yuv420_lines_odither, .-lcd_write_yuv420_lines_odither
diff --git a/firmware/target/arm/sandisk/sansa-c200/lcd-as-c200.S b/firmware/target/arm/sandisk/sansa-c200/lcd-as-c200.S
index e466624d04..7187b83661 100644
--- a/firmware/target/arm/sandisk/sansa-c200/lcd-as-c200.S
+++ b/firmware/target/arm/sandisk/sansa-c200/lcd-as-c200.S
@@ -48,7 +48,7 @@ lcd_write_yuv420_lines:
                                         @ r0 = yuv_src
                                         @ r1 = width
                                         @ r2 = stride
-    stmfd       sp!, { r4-r12 }         @ save non-scratch
+    stmfd       sp!, { r4-r10, lr }     @ save non-scratch
     ldmia       r0, { r4, r5, r6 }      @ r4 = yuv_src[0] = Y'_p
                                         @ r5 = yuv_src[1] = Cb_p
                                         @ r6 = yuv_src[2] = Cr_p
@@ -73,9 +73,9 @@ lcd_write_yuv420_lines:
     add         r10, r10, r8, asl #3    @
     add         r10, r10, r8, asl #4    @
                                         @
-    add         r11, r9, r9, asl #2     @ r9 = Cr*101
-    add         r11, r11, r9, asl #5    @
-    add         r9, r11, r9, asl #6     @
+    add         lr, r9, r9, asl #2      @ r9 = Cr*101
+    add         lr, lr, r9, asl #5      @
+    add         r9, lr, r9, asl #6      @
                                         @
     add         r8, r8, #2              @ r8 = bu = (Cb*128 + 128) >> 8
     mov         r8, r8, asr #2          @
@@ -85,19 +85,19 @@ lcd_write_yuv420_lines:
     mov         r10, r10, asr #8        @
                                         @ compute R, G, and B
     add         r0, r8, r7, asr #8      @ r0  = b = (Y >> 9) + bu
-    add         r11, r9, r7, asr #8     @ r11 = r = (Y >> 9) + rv
+    add         lr, r9, r7, asr #8      @ lr = r = (Y >> 9) + rv
     add         r7, r10, r7, asr #7     @ r7  = g = (Y >> 8) + guv
                                         @
-    orr         r12, r0, r11            @ check if clamping is needed...
+    orr         r12, r0, lr             @ check if clamping is needed...
     orr         r12, r12, r7, asr #1    @ ...at all
     cmp         r12, #31                @
     bls         15f @ no clamp          @
     cmp         r0, #31                 @ clamp b
     mvnhi       r0, r0, asr #31         @
     andhi       r0, r0, #31             @
-    cmp         r11, #31                @ clamp r
-    mvnhi       r11, r11, asr #31       @
-    andhi       r11, r11, #31           @
+    cmp         lr, #31                 @ clamp r
+    mvnhi       lr, lr, asr #31         @
+    andhi       lr, lr, #31             @
     cmp         r7, #63                 @ clamp g
     mvnhi       r7, r7, asr #31         @
     andhi       r7, r7, #63             @
@@ -105,14 +105,14 @@ lcd_write_yuv420_lines:
                                         @
     ldrb        r12, [r4, r2]           @ r12 = Y' = *(Y'_p + stride)
                                         @
-    mov         r11, r11, lsl #3        @
-    orr         r11, r11, r7, lsr #3    @ r11 = (r << 3) | (g >> 3)
+    mov         lr, lr, lsl #3          @
+    orr         lr, lr, r7, lsr #3      @ lr = (r << 3) | (g >> 3)
     orr         r0, r0, r7, lsl #5      @ r0 = (g << 5) | b
 1: @ busy                               @
     ldr         r7, [r3]                @ r7 = LCD1_BASE
     tst         r7, #LCD1_BUSY_MASK     @ bridge busy?
     bne         1b                      @
-    str         r11, [r3, #0x10]        @ send MSB
+    str         lr, [r3, #0x10]         @ send MSB
 1: @busy                                @
     ldr         r7, [r3]                @ r7 = LCD1_BASE
     tst         r7, #LCD1_BUSY_MASK     @ bridge busy?
@@ -124,19 +124,19 @@ lcd_write_yuv420_lines:
     add         r7, r12, r7, asl #5     @
                                         @ compute R, G, and B
     add         r0, r8, r7, asr #8      @ r0  = b = (Y >> 9) + bu
-    add         r11, r9, r7, asr #8     @ r11 = r = (Y >> 9) + rv
+    add         lr, r9, r7, asr #8      @ lr = r = (Y >> 9) + rv
     add         r7, r10, r7, asr #7     @ r7  = g = (Y >> 8) + guv
                                         @
-    orr         r12, r0, r11            @ check if clamping is needed...
+    orr         r12, r0, lr             @ check if clamping is needed...
     orr         r12, r12, r7, asr #1    @ ...at all
     cmp         r12, #31                @
     bls         15f @ no clamp          @
     cmp         r0, #31                 @ clamp b
     mvnhi       r0, r0, asr #31         @
     andhi       r0, r0, #31             @
-    cmp         r11, #31                @ clamp r
-    mvnhi       r11, r11, asr #31       @
-    andhi       r11, r11, #31           @
+    cmp         lr, #31                 @ clamp r
+    mvnhi       lr, lr, asr #31         @
+    andhi       lr, lr, #31             @
     cmp         r7, #63                 @ clamp g
     mvnhi       r7, r7, asr #31         @
     andhi       r7, r7, #63             @
@@ -144,14 +144,14 @@ lcd_write_yuv420_lines:
                                         @
     ldrb        r12, [r4], #1           @ r12 = Y' = *(Y'_p++)
                                         @
-    mov         r11, r11, lsl #3        @
-    orr         r11, r11, r7, lsr #3    @ r11 = (r << 3) | (g >> 3)
+    mov         lr, lr, lsl #3          @
+    orr         lr, lr, r7, lsr #3      @ lr = (r << 3) | (g >> 3)
     orr         r0, r0, r7, lsl #5      @ r0 = (g << 5) | b
 1: @ busy                               @
     ldr         r7, [r3]                @ r7 = LCD1_BASE
     tst         r7, #LCD1_BUSY_MASK     @ bridge busy?
     bne         1b                      @
-    str         r11, [r3, #0x10]        @ send MSB
+    str         lr, [r3, #0x10]         @ send MSB
 1: @ busy                               @
     ldr         r7, [r3]                @ r7 = LCD1_BASE
     tst         r7, #LCD1_BUSY_MASK     @ bridge busy?
@@ -163,19 +163,19 @@ lcd_write_yuv420_lines:
     add         r7, r12, r7, asl #5     @
                                         @ compute R, G, and B
     add         r0, r8, r7, asr #8      @ r0  = b = (Y >> 9) + bu
-    add         r11, r9, r7, asr #8     @ r11 = r = (Y >> 9) + rv
+    add         lr, r9, r7, asr #8      @ lr = r = (Y >> 9) + rv
     add         r7, r10, r7, asr #7     @ r7  = g = (Y >> 8) + guv
                                         @
-    orr         r12, r0, r11            @ check if clamping is needed...
+    orr         r12, r0, lr             @ check if clamping is needed...
     orr         r12, r12, r7, asr #1    @ ...at all
     cmp         r12, #31                @
     bls         15f @ no clamp          @
     cmp         r0, #31                 @ clamp b
     mvnhi       r0, r0, asr #31         @
     andhi       r0, r0, #31             @
-    cmp         r11, #31                @ clamp r
-    mvnhi       r11, r11, asr #31       @
-    andhi       r11, r11, #31           @
+    cmp         lr, #31                 @ clamp r
+    mvnhi       lr, lr, asr #31         @
+    andhi       lr, lr, #31             @
     cmp         r7, #63                 @ clamp g
     mvnhi       r7, r7, asr #31         @
     andhi       r7, r7, #63             @
@@ -184,14 +184,14 @@ lcd_write_yuv420_lines:
     ldrb        r12, [r4, r2]           @ r12 = Y' = *(Y'_p + stride)
                                         @
                                         @
-    mov         r11, r11, lsl #3        @
-    orr         r11, r11, r7, lsr #3    @ r11 = (r << 3) | (g >> 3)
+    mov         lr, lr, lsl #3          @
+    orr         lr, lr, r7, lsr #3      @ lr = (r << 3) | (g >> 3)
     orr         r0, r0, r7, lsl #5      @ r0 = (g << 5) | b
 1: @ busy                               @
     ldr         r7, [r3]                @ r7 = LCD1_BASE
     tst         r7, #LCD1_BUSY_MASK     @ bridge busy?
     bne         1b                      @
-    str         r11, [r3, #0x10]        @ send MSB
+    str         lr, [r3, #0x10]         @ send MSB
 1: @ busy                               @
     ldr         r7, [r3]                @ r7 = LCD1_BASE
     tst         r7, #LCD1_BUSY_MASK     @ bridge busy?
@@ -203,32 +203,32 @@ lcd_write_yuv420_lines:
     add         r7, r12, r7, asl #5     @
                                         @ compute R, G, and B
     add         r0, r8, r7, asr #8      @ r0  = b = (Y >> 9) + bu
-    add         r11, r9, r7, asr #8     @ r11 = r = (Y >> 9) + rv
+    add         lr, r9, r7, asr #8      @ lr = r = (Y >> 9) + rv
     add         r7, r10, r7, asr #7     @ r7  = g = (Y >> 8) + guv
                                         @
-    orr         r12, r0, r11            @ check if clamping is needed...
+    orr         r12, r0, lr             @ check if clamping is needed...
     orr         r12, r12, r7, asr #1    @ ...at all
     cmp         r12, #31                @
     bls         15f @ no clamp          @
     cmp         r0, #31                 @ clamp b
     mvnhi       r0, r0, asr #31         @
     andhi       r0, r0, #31             @
-    cmp         r11, #31                @ clamp r
-    mvnhi       r11, r11, asr #31       @
-    andhi       r11, r11, #31           @
+    cmp         lr, #31                 @ clamp r
+    mvnhi       lr, lr, asr #31         @
+    andhi       lr, lr, #31             @
     cmp         r7, #63                 @ clamp g
     mvnhi       r7, r7, asr #31         @
     andhi       r7, r7, #63             @
 15: @ no clamp                          @
                                         @
-    mov         r11, r11, lsl #3        @
-    orr         r11, r11, r7, lsr #3    @ r11 = (r << 3) | (g >> 3)
+    mov         lr, lr, lsl #3          @
+    orr         lr, lr, r7, lsr #3      @ lr = (r << 3) | (g >> 3)
     orr         r0, r0, r7, lsl #5      @ r0 = (g << 5) | b
 1: @ busy                               @
     ldr         r7, [r3]                @ r7 = LCD1_BASE
     tst         r7, #LCD1_BUSY_MASK     @ bridge busy?
     bne         1b                      @
-    str         r11, [r3, #0x10]        @ send MSB
+    str         lr, [r3, #0x10]         @ send MSB
 1: @ busy                               @
     ldr         r7, [r3]                @ r7 = LCD1_BASE
     tst         r7, #LCD1_BUSY_MASK     @ bridge busy?
@@ -238,8 +238,7 @@ lcd_write_yuv420_lines:
     subs        r1, r1, #2              @ subtract block from width
     bgt         10b @ loop line         @
                                         @
-    ldmfd       sp!, { r4-r12 }         @ restore registers and return
-    bx          lr                      @
+    ldmfd       sp!, { r4-r10, pc }     @ restore registers and return
     .ltorg                              @ dump constant pool
     .size   lcd_write_yuv420_lines, .-lcd_write_yuv420_lines
 
@@ -279,7 +278,7 @@ lcd_write_yuv420_lines_odither:
                                         @ r2 = stride
                                         @ r3 = x_screen
                                         @ [sp] = y_screen
-    stmfd       sp!, { r4-r12, lr }     @ save non-scratch
+    stmfd       sp!, { r4-r11, lr }     @ save non-scratch
     ldmia       r0, { r4, r5, r6 }      @ r4 = yuv_src[0] = Y'_p
                                         @ r5 = yuv_src[1] = Cb_p
                                         @ r6 = yuv_src[2] = Cr_p
@@ -546,6 +545,6 @@ lcd_write_yuv420_lines_odither:
     subs        r1, r1, #2              @ subtract block from width
     bgt         10b @ loop line         @
                                         @
-    ldmfd       sp!, { r4-r12, pc }     @ restore registers and return
+    ldmfd       sp!, { r4-r11, pc }     @ restore registers and return
     .ltorg                              @ dump constant pool
     .size   lcd_write_yuv420_lines_odither, .-lcd_write_yuv420_lines_odither