X5: Applied tweaks from the H300 lcd_yuv_blit to the X5 version. Smaller code and ca. 1% speedup.

git-svn-id: svn://svn.rockbox.org/rockbox/trunk@10729 a1c6a512-1295-4272-9138-f99709370657

1 files changed, 207 insertions, 269 deletions

diff --git a/firmware/target/coldfire/iaudio/x5/lcd-as-x5.S b/firmware/target/coldfire/iaudio/x5/lcd-as-x5.S
index 1a527bb..7c89fb9 100644
--- a/firmware/target/coldfire/iaudio/x5/lcd-as-x5.S
+++ b/firmware/target/coldfire/iaudio/x5/lcd-as-x5.S
@@ -40,316 +40,254 @@
  *   |G| = |1.000000 -0.334136 -0.714136| |Pb|
  *   |B|   |1.000000  1.772000  0.000000| |Pr|
  *   Scaled, normalized, rounded and tweaked to yield RGB 666:
- *   |R|   |74    0  102| |Y' -  16| / 256
- *   |G| = |74  -25  -52| |Cb - 128| / 256
- *   |B|   |74  129    0| |Cr - 128| / 256
+ *   |R|   |74    0  101| |Y' -  16| / 256
+ *   |G| = |74  -24  -51| |Cb - 128| / 256
+ *   |B|   |74  128    0| |Cr - 128| / 256
  */
     .align      2
     .global     lcd_write_yuv420_lines
     .type       lcd_write_yuv420_lines,@function
 lcd_write_yuv420_lines:
-    lea.l   (-40,%sp),%sp   /* free up some registers */
-    movem.l %d2-%d7/%a2-%a5,(%sp)
+    lea.l   (-36,%sp),%sp       /* free up some registers */
+    movem.l %d2-%d6/%a2-%a5,(%sp)
 
-    lea.l   0xf0008002,%a0  /* LCD data port */
-    move.l  (40+4,%sp),%a1  /* Y data */
-    move.l  (40+8,%sp),%a2  /* Cb data */
-    move.l  (40+12,%sp),%a3 /* guv storage */
-    move.l  (40+16,%sp),%a4 /* Cr data */
-    move.l  (40+20,%sp),%d0 /* width */
-    lea.l   (%a1,%d0.l),%a5 /* end address */
+    lea.l   0xf0008002,%a0      /* LCD data port */
+    movem.l (36+4,%sp),%a1-%a5  /* Y data, Cb data, guv storage, Cr data, width */
+    lea.l   (%a1,%a5),%a5       /* end address */
 
 .yuv_line_loop1:
     /** Write first pixel **/
-    clr.l   %d1             /* get y component */
-    move.b  (%a1)+,%d1
-    subq.l  #8,%d1
-    subq.l  #8,%d1
-    moveq.l #74,%d6
-    muls.w  %d6,%d1
-    asr.l   #8,%d1
-
-    clr     %d2             /* get bu component */
-    move.b  (%a2),%d2
-    moveq.l #-128,%d6
-    add.l   %d6,%d2
-    move.l  %d2,%d3         /* %d3 = cb component for guv */
-    move.w  #129,%d6
-    muls.w  %d6,%d2
+    clr.l   %d1             /* get bu component */
+    move.b  (%a2),%d1
+    clr.l   %d3             /* get rv component */
+    move.b  (%a4),%d3
+    moveq.l #-128,%d0
+    add.l   %d0,%d1
+    add.l   %d0,%d3
+
+    move.l  %d1,%d2         /* %d2 = cb component for guv */
+    asr.l   #1,%d1          /* %d1 = 128 * (Cb - 128) / 256 */
+    move.b  %d1,(%a2)+      /* save bu for next line */
+    moveq.l #-24,%d0        /* multiply first term of guv */
+    muls.w  %d0,%d2
+    moveq.l #-51,%d0        /* multiply second term of guv */
+    muls.w  %d3,%d0
+    add.l   %d0,%d2
     asr.l   #8,%d2
-    move.b  %d2,(%a2)+      /* save bu for next line */
-
-    moveq.l #-25,%d6        /* multiply first term of guv */
-    muls.w  %d6,%d3
-
-    clr     %d4             /* get rv component */
-    move.b  (%a4),%d4
-    moveq.l #-128,%d6
-    add.l   %d6,%d4
-    move.l  %d4,%d7         /* %d7 = cr component for guv */
-    moveq.l  #102,%d6
-    muls.w  %d6,%d4
+    move.b  %d2,(%a3)+      /* save guv for next line */
+    moveq.l  #101,%d0
+    muls.w  %d0,%d3
+    asr.l   #8,%d3
+    move.b  %d3,(%a4)+      /* save rv for next line */
+
+    clr.l   %d4             /* get y component */
+    move.b  (%a1)+,%d4
+    moveq.l #74,%d0
+    muls.w  %d0,%d4
     asr.l   #8,%d4
-    move.b  %d4,(%a4)+      /* save rv for next line */
+    subq.l  #4,%d4
+    move.l  %d4,%d5
+    move.l  %d4,%d6
+    /* : %d4,%d5,%d6 = Y, %d1 = bu, %d2 = guv, %d3 = rv */
 
-    moveq.l #-52,%d6        /* multiply second term of guv */
-    muls.w  %d6,%d7
-    add.l   %d7,%d3
-    asr.l   #8,%d3
-    move.b  %d3,(%a3)+      /* save guv for next line */
-    /* : %d1 = Y, %d2 = bu, %d3 = guv, %d4 = rv */
-
-    move.l  %d1,%d5         /* get r */
-    add.l   %d4,%d5
-    move.l  %d1,%d6         /* get g */
-    add.l   %d3,%d6
-    move.l  %d1,%d7         /* get b */
-    add.l   %d2,%d7
-
-    move.l  %d7,%d1         /* is clamping needed? */
-    or.l    %d6,%d1
-    or.l    %d5,%d1
-    asr.l   #6,%d1
+    add.l   %d3,%d4         /* get r */
+    add.l   %d2,%d5         /* get g */
+    add.l   %d1,%d6         /* get b */
+
+    move.l  %d6,%d0         /* is clamping needed? */
+    or.l    %d5,%d0
+    or.l    %d4,%d0
+    asr.l   #6,%d0
     beq.b   .yuv_no_clamp1  /* values in range: skip clamping */
-    bpl.b   .yuv_r63_test1  /* no negative values: skip to high bounds checks */
-.yuv_r0_test1:
-    clr.l   %d1             /* check for any values < 0 */
-    cmp.l   %d1,%d5
-    bgt.b   .yuv_g0_test1
-    clr.l   %d5
-.yuv_g0_test1:
-    cmp.l   %d1,%d6
-    bgt.b   .yuv_b0_test1
-    clr.l   %d6
-.yuv_b0_test1:
-    cmp.l   %d1,%d7
-    bgt.b   .yuv_r63_test1
-    clr.l   %d7
-.yuv_r63_test1:             /* check for any values > 63 */
-    moveq.l #63,%d1
-    cmp.l   %d1,%d5
-    blt.b   .yuv_g63_test1
-    move.l  %d1,%d5
-.yuv_g63_test1:
-    cmp.l   %d1,%d6
-    blt.b   .yuv_b63_test1
-    move.l  %d1,%d6
-.yuv_b63_test1:
-    cmp.l   %d1,%d7
-    blt.b   .yuv_no_clamp1
-    move.l  %d1,%d7
+    moveq.l #63, %d0
+    cmp.l   %d0, %d4
+    bls.s   .yuv_red_ok1
+    spl.b   %d4
+    and.l   %d0, %d4
+.yuv_red_ok1:
+    cmp.l   %d0, %d5
+    bls.s   .yuv_green_ok1
+    spl.b   %d5
+    and.l   %d0, %d5
+.yuv_green_ok1:
+    cmp.l   %d0, %d6
+    bls.s   .yuv_blue_ok1
+    spl.b   %d6
+    and.l   %d0, %d6
+.yuv_blue_ok1:
 .yuv_no_clamp1:
-    /* : %d5 = R, %d6 = G, %d7 = B */
+    /* : %d4 = R, %d5 = G, %d6 = B */
 
-    move.l  %d6,%d1         /* save g for lower 9 bits */
-    lsl.l   #3,%d5          /* R << 3 */
-    lsr.l   #3,%d1          /* G >> 3 */
-    or.l    %d5,%d1
-    move.w  %d1,(%a0)       /* |00000000|000000000|0000000r|rrrrrggg| */
-    lsl.l   #6,%d6          /* B << 6 */
-    or.l    %d6,%d7         /* |00000000|000000000|0000gggg|ggbbbbbb| */
-    move.w  %d7,(%a0)
+    move.l  %d5,%d0         /* save g for lower 9 bits */
+    lsl.l   #3,%d4          /* R << 3 */
+    lsr.l   #3,%d0          /* G >> 3 */
+    or.l    %d4,%d0
+    move.w  %d0,(%a0)       /* |00000000|000000000|0000000r|rrrrrggg| */
+    lsl.l   #6,%d5          /* B << 6 */
+    or.l    %d5,%d6         /* |00000000|000000000|0000gggg|ggbbbbbb| */
+    move.w  %d6,(%a0)
 
     /** Write second pixel **/
-    clr     %d1
-    move.b  (%a1)+,%d1      /* get y component */
-    subq.l  #8,%d1
-    subq.l  #8,%d1
-    moveq.l #74,%d6
-    muls.w  %d6,%d1
-    asr.l   #8,%d1
-    /* : %d1 = Y, %d2 = bu, %d3 = guv, %d4 = rv */
-
-    /* Add Y + each chroma component (can clobber %d2-%d4 values now) */
-    add.l   %d1,%d4         /* get r */
-    add.l   %d1,%d3         /* get g */
-    add.l   %d1,%d2         /* get b */
-
-    move.l  %d2,%d1         /* is clamping needed? */
-    or.l    %d3,%d1
-    or.l    %d4,%d1
-    asr.l   #6,%d1
-    beq.b   .yuv_no_clamp2  /* values in range: skip clamping */
-    bpl.b   .yuv_r63_test2  /* no negative values: skip to high bounds checks */
-.yuv_r0_test2:
-    clr.l   %d1             /* check for any values < 0 */
-    cmp.l   %d1,%d4
-    bgt.b   .yuv_g0_test2
     clr.l   %d4
-.yuv_g0_test2:
-    cmp.l   %d1,%d3
-    bgt.b   .yuv_b0_test2
-    clr.l   %d3
-.yuv_b0_test2:
-    cmp.l   %d1,%d2
-    bgt.b   .yuv_r63_test2
-    clr.l   %d2
-.yuv_r63_test2:             /* check for any values > 63 */
-    moveq.l #63,%d1
-    cmp.l   %d1,%d4
-    blt.b   .yuv_g63_test2
-    move.l  %d1,%d4
-.yuv_g63_test2:
-    cmp.l   %d1,%d3
-    blt.b   .yuv_b63_test2
-    move.l  %d1,%d3
-.yuv_b63_test2:
-    cmp.l   %d1,%d2
-    blt.b   .yuv_no_clamp2
-    move.l  %d1,%d2
+    move.b  (%a1)+,%d4      /* get y component */
+    moveq.l #74,%d0
+    muls.w  %d0,%d4
+    asr.l   #8,%d4
+    subq.l  #4,%d4
+    /* : %d4 = Y, %d1 = bu, %d2 = guv, %d3 = rv */
+
+    /* Add Y + each chroma component (can clobber %d1-%d3 values now) */
+    add.l   %d4,%d3         /* get r */
+    add.l   %d4,%d2         /* get g */
+    add.l   %d4,%d1         /* get b */
+
+    move.l  %d1,%d0         /* is clamping needed? */
+    or.l    %d2,%d0
+    or.l    %d3,%d0
+    asr.l   #6,%d0
+    beq.b   .yuv_no_clamp2  /* values in range: skip clamping */
+    moveq.l #63, %d0
+    cmp.l   %d0, %d3
+    bls.s   .yuv_red_ok2
+    spl.b   %d3
+    and.l   %d0, %d3
+.yuv_red_ok2:
+    cmp.l   %d0, %d2
+    bls.s   .yuv_green_ok2
+    spl.b   %d2
+    and.l   %d0, %d2
+.yuv_green_ok2:
+    cmp.l   %d0, %d1
+    bls.s   .yuv_blue_ok2
+    spl.b   %d1
+    and.l   %d0, %d1
+.yuv_blue_ok2:
 .yuv_no_clamp2:
-    /* : %d4 = R, %d3 = G, %d2 = B */
-
-    move.l  %d3,%d1         /* save g for lower 9 bits */
-    lsl.l   #3,%d4          /* R << 3 */
-    lsr.l   #3,%d1          /* G >> 3 */
-    or.l    %d4,%d1         /* |00000000|000000000|0000000r|rrrrrggg| */
+    /* : %d3 = R, %d2 = G, %d1 = B */
+
+    move.l  %d2,%d0         /* save g for lower 9 bits */
+    lsl.l   #3,%d3          /* R << 3 */
+    lsr.l   #3,%d0          /* G >> 3 */
+    or.l    %d3,%d0         /* |00000000|000000000|0000000r|rrrrrggg| */
+    move.w  %d0,(%a0)
+    lsl.l   #6,%d2          /* G << 6 */
+    or.l    %d2,%d1         /* |00000000|000000000|0000gggg|ggbbbbbb| */
     move.w  %d1,(%a0)
-    lsl.l   #6,%d3          /* G << 6 */
-    or.l    %d3,%d2         /* |00000000|000000000|0000gggg|ggbbbbbb| */
-    move.w  %d2,(%a0)
 
     cmp.l   %a1,%a5         /* run %a1 up to end of line */
     bhi.w   .yuv_line_loop1
 
     /* Rewind chroma pointers */
-    move.l  (40+8,%sp),%a2  /* bu data */
-    move.l  (40+12,%sp),%a3 /* guv data */
-    move.l  (40+16,%sp),%a4 /* rv data */
-    lea.l   (%a5,%d0),%a5   /* next end address */
+    movem.l (36+8, %sp), %a2-%a5    /* bu data, guv data, rv data, width */
+    lea.l   (%a1, %a5), %a5         /* next end address */
 
 .yuv_line_loop2:
-    clr     %d1
-    move.b  (%a1)+,%d1      /* get y component */
-    subq.l  #8,%d1
-    subq.l  #8,%d1
-    moveq.l #74,%d6
-    muls.w  %d6,%d1
-    asr.l   #8,%d1
-
-    move.b  (%a2)+,%d2      /* read save chromas and sign extend */
+    move.b  (%a2)+,%d1      /* read save chromas and sign extend */
+    extb.l  %d1
+    move.b  (%a3)+,%d2
     extb.l  %d2
-    move.b  (%a3)+,%d3
+    move.b  (%a4)+,%d3
     extb.l  %d3
-    move.b  (%a4)+,%d4
-    extb.l  %d4
-    /* : %d1 = Y, %d2 = bu, %d3 = guv, %d4 = rv */
-
-    move.l  %d1,%d5         /* get r */
-    add.l   %d4,%d5
-    move.l  %d1,%d6         /* get g */
-    add.l   %d3,%d6
-    move.l  %d1,%d7         /* get b */
-    add.l   %d2,%d7
-
-    move.l  %d7,%d1         /* is clamping needed? */
-    or.l    %d6,%d1
-    or.l    %d5,%d1
-    asr.l   #6,%d1
+
+    clr.l   %d4
+    move.b  (%a1)+,%d4      /* get y component */
+    moveq.l #74,%d0
+    muls.w  %d0,%d4
+    asr.l   #8,%d4
+    subq.l  #4,%d4
+    move.l  %d4,%d5
+    move.l  %d4,%d6
+    /* : %d4,%d5,%d6 = Y, %d1 = bu, %d2 = guv, %d3 = rv */
+
+    add.l   %d3,%d4         /* get r */
+    add.l   %d2,%d5         /* get g */
+    add.l   %d1,%d6         /* get b */
+
+    move.l  %d6,%d0         /* is clamping needed? */
+    or.l    %d5,%d0
+    or.l    %d4,%d0
+    asr.l   #6,%d0
     beq.b   .yuv_no_clamp3  /* values in range: skip clamping */
-    bpl.b   .yuv_r63_test3  /* no negative values: skip to high bounds checks */
-.yuv_r0_test3:
-    clr.l   %d1             /* check for any values < 0 */
-    cmp.l   %d1,%d5
-    bgt.b   .yuv_g0_test3
-    clr.l   %d5
-.yuv_g0_test3:
-    cmp.l   %d1,%d6
-    bgt.b   .yuv_b0_test3
-    clr.l   %d6
-.yuv_b0_test3:
-    cmp.l   %d1,%d7
-    bgt.b   .yuv_r63_test3
-    clr.l   %d7
-.yuv_r63_test3:             /* check for any values > 63 */
-    moveq.l #63,%d1
-    cmp.l   %d1,%d5
-    blt.b   .yuv_g63_test3
-    move.l  %d1,%d5
-.yuv_g63_test3:
-    cmp.l   %d1,%d6
-    blt.b   .yuv_b63_test3
-    move.l  %d1,%d6
-.yuv_b63_test3:
-    cmp.l   %d1,%d7
-    blt.b   .yuv_no_clamp3
-    move.l  %d1,%d7
+    moveq.l #63, %d0
+    cmp.l   %d0, %d4
+    bls.s   .yuv_red_ok3
+    spl.b   %d4
+    and.l   %d0, %d4
+.yuv_red_ok3:
+    cmp.l   %d0, %d5
+    bls.s   .yuv_green_ok3
+    spl.b   %d5
+    and.l   %d0, %d5
+.yuv_green_ok3:
+    cmp.l   %d0, %d6
+    bls.s   .yuv_blue_ok3
+    spl.b   %d6
+    and.l   %d0, %d6
+.yuv_blue_ok3:
 .yuv_no_clamp3:
-    /* : %d5 = R, %d6 = G, %d7 = B */
+    /* : %d4 = R, %d5 = G, %d6 = B */
 
-    move.l  %d6,%d1         /* save g for lower 9 bits */
-    lsl.l   #3,%d5          /* R << 3 */
-    lsr.l   #3,%d1          /* G >> 3 */
-    or.l    %d5,%d1
-    move.w  %d1,(%a0)       /* |00000000|000000000|0000000r|rrrrrggg| */
-    lsl.l   #6,%d6          /* B << 6 */
-    or.l    %d6,%d7         /* |00000000|000000000|0000gggg|ggbbbbbb| */
-    move.w  %d7,(%a0)
+    move.l  %d5,%d0         /* save g for lower 9 bits */
+    lsl.l   #3,%d4          /* R << 3 */
+    lsr.l   #3,%d0          /* G >> 3 */
+    or.l    %d4,%d0
+    move.w  %d0,(%a0)       /* |00000000|000000000|0000000r|rrrrrggg| */
+    lsl.l   #6,%d5          /* B << 6 */
+    or.l    %d5,%d6         /* |00000000|000000000|0000gggg|ggbbbbbb| */
+    move.w  %d6,(%a0)
 
     /** Write second pixel **/
-    clr     %d1
-    move.b  (%a1)+,%d1      /* get y component */
-    subq.l  #8,%d1
-    subq.l  #8,%d1
-    moveq.l #74,%d6
-    muls.w  %d6,%d1
-    asr.l   #8,%d1
-    /* : %d1 = Y, %d2 = bu, %d3 = guv, %d4 = rv */
-
-    /* Add Y + each chroma component (can clobber %d2-%d4 values now) */
-    add.l   %d1,%d4         /* get r */
-    add.l   %d1,%d3         /* get g */
-    add.l   %d1,%d2         /* get b */
-
-    move.l  %d2,%d1         /* is clamping needed? */
-    or.l    %d3,%d1
-    or.l    %d4,%d1
-    asr.l   #6,%d1
-    beq.b   .yuv_no_clamp4  /* values in range: skip clamping */
-    bpl.b   .yuv_r63_test4  /* no negative values: skip to high bounds checks */
-.yuv_r0_test4:
-    clr.l   %d1             /* check for any values < 0 */
-    cmp.l   %d1,%d4
-    bgt.b   .yuv_g0_test4
     clr.l   %d4
-.yuv_g0_test4:
-    cmp.l   %d1,%d3
-    bgt.b   .yuv_b0_test4
-    clr.l   %d3
-.yuv_b0_test4:
-    cmp.l   %d1,%d2
-    bgt.b   .yuv_r63_test4
-    clr.l   %d2
-.yuv_r63_test4:             /* check for any values > 63 */
-    moveq.l #63,%d1
-    cmp.l   %d1,%d4
-    blt.b   .yuv_g63_test4
-    move.l  %d1,%d4
-.yuv_g63_test4:
-    cmp.l   %d1,%d3
-    blt.b   .yuv_b63_test4
-    move.l  %d1,%d3
-.yuv_b63_test4:
-    cmp.l   %d1,%d2
-    blt.b   .yuv_no_clamp4
-    move.l  %d1,%d2
+    move.b  (%a1)+,%d4      /* get y component */
+    moveq.l #74,%d0
+    muls.w  %d0,%d4
+    asr.l   #8,%d4
+    subq.l  #4,%d4
+    /* : %d4 = Y, %d1 = bu, %d2 = guv, %d3 = rv */
+
+    /* Add Y + each chroma component (can clobber %d1-%d3 values now) */
+    add.l   %d4,%d3         /* get r */
+    add.l   %d4,%d2         /* get g */
+    add.l   %d4,%d1         /* get b */
+
+    move.l  %d1,%d0         /* is clamping needed? */
+    or.l    %d2,%d0
+    or.l    %d3,%d0
+    asr.l   #6,%d0
+    beq.b   .yuv_no_clamp4  /* values in range: skip clamping */
+    moveq.l #63, %d0
+    cmp.l   %d0, %d3
+    bls.s   .yuv_red_ok4
+    spl.b   %d3
+    and.l   %d0, %d3
+.yuv_red_ok4:
+    cmp.l   %d0, %d2
+    bls.s   .yuv_green_ok4
+    spl.b   %d2
+    and.l   %d0, %d2
+.yuv_green_ok4:
+    cmp.l   %d0, %d1
+    bls.s   .yuv_blue_ok4
+    spl.b   %d1
+    and.l   %d0, %d1
+.yuv_blue_ok4:
 .yuv_no_clamp4:
-    /* : %d4 = R, %d3 = G, %d2 = B */
-
-    move.l  %d3,%d1         /* save g for lower 9 bits */
-    lsl.l   #3,%d4          /* R << 3 */
-    lsr.l   #3,%d1          /* G >> 3 */
-    or.l    %d4,%d1         /* |00000000|000000000|0000000r|rrrrrggg| */
+    /* : %d3 = R, %d2 = G, %d1 = B */
+
+    move.l  %d2,%d0         /* save g for lower 9 bits */
+    lsl.l   #3,%d3          /* R << 3 */
+    lsr.l   #3,%d0          /* G >> 3 */
+    or.l    %d3,%d0         /* |00000000|000000000|0000000r|rrrrrggg| */
+    move.w  %d0,(%a0)
+    lsl.l   #6,%d2          /* G << 6 */
+    or.l    %d2,%d1         /* |00000000|000000000|0000gggg|ggbbbbbb| */
     move.w  %d1,(%a0)
-    lsl.l   #6,%d3          /* G << 6 */
-    or.l    %d3,%d2         /* |00000000|000000000|0000gggg|ggbbbbbb| */
-    move.w  %d2,(%a0)
 
     cmp.l   %a1,%a5         /* run %a0 up to end of line */
     bhi.w   .yuv_line_loop2
 
-    movem.l (%sp),%d2-%d7/%a2-%a5
-    lea.l   (40,%sp),%sp    /* restore registers */
+    movem.l (%sp),%d2-%d6/%a2-%a5
+    lea.l   (36,%sp),%sp    /* restore registers */
 
     rts
 /* end lcd_write_yuv420_lines */