NEONize a few more blit types

[sdl_omap.git] / src / video / SDL_blit_neon.S

diff --git a/src/video/SDL_blit_neon.S b/src/video/SDL_blit_neon.S
index 438d9fc..344ae05 100644 (file)
--- a/src/video/SDL_blit_neon.S
+++ b/src/video/SDL_blit_neon.S
@@ -1,5 +1,5 @@
 /*
- * (C) Gražvydas "notaz" Ignotas, 2011
+ * (C) Gražvydas "notaz" Ignotas, 2011,2012
  *
  * This work is licensed under the terms of any of these licenses
  * (at your option):
@@ -11,6 +11,10 @@
 .text
 .align 2
 
+#define func(name) \
+    .global name; \
+    name
+
 @ void *dst, const void *src, int count, uint abits
 .macro do_argb bgr2rgb
     vdup.i8    d0, r3
@@ -28,17 +32,24 @@
     b          0b
 .endm
 
-@ void *dst, const void *src, int count
-.macro do_argb_alpha bgr2rgb
-    mov        r3, #0xff
-    vdup.i16   q12, r3
+@ void *dst, const void *src, int count, uint global_alpha
+.macro do_argb_alpha bgr2rgb global_alpha
+    mov        r12, #0xff
+.if \global_alpha
+    vdup.16    q11, r3
+.endif
+    vdup.i16   q12, r12
 0:
+    pld        [r1, #64*2]
+    pld        [r0, #64*2]
     vld4.8     {d4-d7}, [r1]!
     vld4.8     {d0-d3}, [r0]
 .if \bgr2rgb
     vswp       d4, d6          @ BGR->RGB
 .endif
+.if !\global_alpha
     vmovl.u8   q11, d7
+.endif
     @ d = (((s-d)*a+255)>>8)+d
     vsubl.u8   q8, d4, d0
     vsubl.u8   q9, d5, d1
@@ -60,6 +71,7 @@
     b          0b
 .endm
 
+
 do_argb_finish:
     add        r2, r2, #8
     vzip.8     d4, d5          @ RRR..|GGG.. -> RGRG..
@@ -88,20 +100,103 @@ do_argb_finish:
     bx         lr
 
 
-.global neon_ARGBtoXRGB
-neon_ARGBtoXRGB:
+@ void *dst, const void *src, int count, uint global_alpha
+.macro do_argb_to_rgb565_alpha bgr2rgb global_alpha
+    mov        r12, #0xff
+.if \global_alpha
+    vdup.16    q11, r3
+.endif
+    vdup.i16   q12, r12
+0:
+    pld        [r1, #64*2]
+    pld        [r0, #64*2]
+    vld4.8     {d4-d7}, [r1]!
+    vld2.8     {d1-d2}, [r0]
+.if \bgr2rgb
+    vswp       d4, d6          @ BGR->RGB
+.endif
+.if !\global_alpha
+    vmovl.u8   q11, d7
+.endif
+    vshl.i8    d0, d1, #3
+    vshr.u8    d1, d1, #3
+    vsri.i8    d0, d0, #5       @ B
+    vsli.i8    d1, d2, #5
+    vsri.i8    d2, d2, #5       @ R
+    vsri.i8    d1, d1, #6       @ G
+    @ d = (((s-d)*a+255)>>8)+d
+    vsubl.u8   q8, d4, d0
+    vsubl.u8   q9, d5, d1
+    vsubl.u8   q10,d6, d2
+    vmul.s16   q8, q8, q11
+    vmul.s16   q9, q9, q11
+    vmul.s16   q10,q10,q11
+    vaddhn.i16 d4, q8, q12
+    vaddhn.i16 d5, q9, q12
+    vaddhn.i16 d6, q10,q12
+    vadd.i8    q2, q0
+    vadd.i8    d2, d6           @ rrrr rrrr
+    vshr.u8    d0, d5, #2
+    vshr.u8    d1, d4, #3       @ 000b bbbb
+    vsri.i8    d2, d5, #5       @ rrrr rggg
+    vsli.i8    d1, d0, #5       @ gggb bbbb
+    subs       r2, r2, #8
+    blt        do_rgb565_finish
+    vst2.8     {d1-d2}, [r0]!
+    bxeq       lr
+    nop
+    b          0b
+.endm
+
+
+do_rgb565_finish:
+    vzip.8     d1, d2
+    add        r2, r2, #8
+
+    vst1.16    d1[0], [r0]!
+    cmp        r2, #1
+    bxle       lr
+    vst1.16    d1[1], [r0]!
+    cmp        r2, #2
+    bxle       lr
+    vst1.16    d1[2], [r0]!
+    cmp        r2, #3
+    bxle       lr
+    vst1.16    d1[3], [r0]!
+    cmp        r2, #4
+    bxle       lr
+    vst1.16    d2[0], [r0]!
+    cmp        r2, #5
+    bxle       lr
+    vst1.16    d2[1], [r0]!
+    cmp        r2, #6
+    bxle       lr
+    vst1.16    d2[2], [r0]!
+    bx         lr
+
+
+func(neon_ARGBtoXRGB):
     do_argb 0
 
-.global neon_ABGRtoXRGB
-neon_ABGRtoXRGB:
+func(neon_ABGRtoXRGB):
     do_argb 1
 
-.global neon_ARGBtoXRGBalpha
-neon_ARGBtoXRGBalpha:
-    do_argb_alpha 0
+func(neon_ARGBtoXRGBalpha):
+    do_argb_alpha 0, 0
+
+func(neon_ABGRtoXRGBalpha):
+    do_argb_alpha 1, 0
+
+func(neon_ARGBtoXRGBalphaS):
+    do_argb_alpha 0, 1
+
+func(neon_ABGRtoXRGBalphaS):
+    do_argb_alpha 1, 1
+
+func(neon_ARGBtoRGB565alpha):
+    do_argb_to_rgb565_alpha 0, 0
 
-.global neon_ABGRtoXRGBalpha
-neon_ABGRtoXRGBalpha:
-    do_argb_alpha 1
+func(neon_ABGRtoRGB565alpha):
+    do_argb_to_rgb565_alpha 1, 0
 
 @ vim:filetype=armasm