ARMv6 optimised put_pixels functions except xy2 variants

Originally committed as revision 21696 to svn://svn.ffmpeg.org/ffmpeg/trunk

2 files changed, 280 insertions, 0 deletions

diff --git a/libavcodec/arm/dsputil_armv6.S b/libavcodec/arm/dsputil_armv6.S
index c76d408809..6180378841 100644
--- a/libavcodec/arm/dsputil_armv6.S
+++ b/libavcodec/arm/dsputil_armv6.S
@@ -22,6 +22,244 @@
 
         .text
 
+.macro  call_2x_pixels  type, subp
+function ff_\type\()_pixels16\subp\()_armv6, export=1
+        push            {r0-r3, lr}
+        bl              ff_\type\()_pixels8\subp\()_armv6
+        pop             {r0-r3, lr}
+        add             r0,  r0,  #8
+        add             r1,  r1,  #8
+        b               ff_\type\()_pixels8\subp\()_armv6
+.endfunc
+.endm
+
+call_2x_pixels          avg
+call_2x_pixels          put, _x2
+call_2x_pixels          put, _y2
+call_2x_pixels          put, _x2_no_rnd
+call_2x_pixels          put, _y2_no_rnd
+
+function ff_put_pixels16_armv6, export=1
+        push            {r4-r11}
+1:
+        ldr             r5,  [r1, #4]
+        ldr             r6,  [r1, #8]
+        ldr             r7,  [r1, #12]
+        ldr             r4,  [r1], r2
+        strd            r6,  r7,  [r0, #8]
+        ldr             r9,  [r1, #4]
+        strd            r4,  r5,  [r0],  r2
+        ldr             r10, [r1, #8]
+        ldr             r11, [r1, #12]
+        ldr             r8,  [r1], r2
+        strd            r10, r11, [r0, #8]
+        subs            r3,  r3,  #2
+        strd            r8,  r9,  [r0],  r2
+        bne             1b
+
+        pop             {r4-r11}
+        bx              lr
+.endfunc
+
+function ff_put_pixels8_armv6, export=1
+        push            {r4-r7}
+1:
+        ldr             r5,  [r1, #4]
+        ldr             r4,  [r1], r2
+        ldr             r7,  [r1, #4]
+        strd            r4,  r5,  [r0],  r2
+        ldr             r6,  [r1], r2
+        subs            r3,  r3,  #2
+        strd            r6,  r7,  [r0],  r2
+        bne             1b
+
+        pop             {r4-r7}
+        bx              lr
+.endfunc
+
+function ff_put_pixels8_x2_armv6, export=1
+        push            {r4-r11, lr}
+        mov             r12, #1
+        orr             r12, r12, r12, lsl #8
+        orr             r12, r12, r12, lsl #16
+1:
+        ldr             r4,  [r1]
+        subs            r3,  r3,  #2
+        ldr             r5,  [r1, #4]
+        ldr             r7,  [r1, #5]
+        lsr             r6,  r4,  #8
+        ldr             r8,  [r1, r2]!
+        orr             r6,  r6,  r5,  lsl #24
+        ldr             r9,  [r1, #4]
+        ldr             r11, [r1, #5]
+        lsr             r10, r8,  #8
+        add             r1,  r1,  r2
+        orr             r10, r10, r9,  lsl #24
+        eor             r14, r4,  r6
+        uhadd8          r4,  r4,  r6
+        eor             r6,  r5,  r7
+        uhadd8          r5,  r5,  r7
+        and             r14, r14, r12
+        and             r6,  r6,  r12
+        uadd8           r4,  r4,  r14
+        eor             r14, r8,  r10
+        uadd8           r5,  r5,  r6
+        eor             r6,  r9,  r11
+        uhadd8          r8,  r8,  r10
+        and             r14, r14, r12
+        uhadd8          r9,  r9,  r11
+        and             r6,  r6,  r12
+        uadd8           r8,  r8,  r14
+        strd            r4,  r5,  [r0],  r2
+        uadd8           r9,  r9,  r6
+        strd            r8,  r9,  [r0],  r2
+        bne             1b
+
+        pop             {r4-r11, pc}
+.endfunc
+
+function ff_put_pixels8_y2_armv6, export=1
+        push            {r4-r11}
+        mov             r12, #1
+        orr             r12, r12, r12, lsl #8
+        orr             r12, r12, r12, lsl #16
+        ldr             r4,  [r1]
+        ldr             r5,  [r1, #4]
+        ldr             r6,  [r1, r2]!
+        ldr             r7,  [r1, #4]
+1:
+        subs            r3,  r3,  #2
+        uhadd8          r8,  r4,  r6
+        eor             r10, r4,  r6
+        uhadd8          r9,  r5,  r7
+        eor             r11, r5,  r7
+        and             r10, r10, r12
+        ldr             r4,  [r1, r2]!
+        uadd8           r8,  r8,  r10
+        and             r11, r11, r12
+        uadd8           r9,  r9,  r11
+        ldr             r5,  [r1, #4]
+        uhadd8          r10, r4,  r6
+        eor             r6,  r4,  r6
+        uhadd8          r11, r5,  r7
+        and             r6,  r6,  r12
+        eor             r7,  r5,  r7
+        uadd8           r10, r10, r6
+        and             r7,  r7,  r12
+        ldr             r6,  [r1, r2]!
+        uadd8           r11, r11, r7
+        strd            r8,  r9,  [r0],  r2
+        ldr             r7,  [r1, #4]
+        strd            r10, r11, [r0],  r2
+        bne             1b
+
+        pop             {r4-r11}
+        bx              lr
+.endfunc
+
+function ff_put_pixels8_x2_no_rnd_armv6, export=1
+        push            {r4-r9, lr}
+1:
+        subs            r3,  r3,  #2
+        ldr             r4,  [r1]
+        ldr             r5,  [r1, #4]
+        ldr             r7,  [r1, #5]
+        ldr             r8,  [r1, r2]!
+        ldr             r9,  [r1, #4]
+        ldr             r14, [r1, #5]
+        add             r1,  r1,  r2
+        lsr             r6,  r4,  #8
+        orr             r6,  r6,  r5,  lsl #24
+        lsr             r12, r8,  #8
+        orr             r12, r12, r9,  lsl #24
+        uhadd8          r4,  r4,  r6
+        uhadd8          r5,  r5,  r7
+        uhadd8          r8,  r8,  r12
+        uhadd8          r9,  r9,  r14
+        stm             r0,  {r4,r5}
+        add             r0,  r0,  r2
+        stm             r0,  {r8,r9}
+        add             r0,  r0,  r2
+        bne             1b
+
+        pop             {r4-r9, pc}
+.endfunc
+
+function ff_put_pixels8_y2_no_rnd_armv6, export=1
+        push            {r4-r9, lr}
+        ldr             r4,  [r1]
+        ldr             r5,  [r1, #4]
+        ldr             r6,  [r1, r2]!
+        ldr             r7,  [r1, #4]
+1:
+        subs            r3,  r3,  #2
+        uhadd8          r8,  r4,  r6
+        ldr             r4,  [r1, r2]!
+        uhadd8          r9,  r5,  r7
+        ldr             r5,  [r1, #4]
+        uhadd8          r12, r4,  r6
+        ldr             r6,  [r1, r2]!
+        uhadd8          r14, r5,  r7
+        ldr             r7,  [r1, #4]
+        stm             r0,  {r8,r9}
+        add             r0,  r0,  r2
+        stm             r0,  {r12,r14}
+        add             r0,  r0,  r2
+        bne             1b
+
+        pop             {r4-r9, pc}
+.endfunc
+
+function ff_avg_pixels8_armv6, export=1
+        pld             [r1, r2]
+        push            {r4-r10, lr}
+        mov             lr,  #1
+        orr             lr,  lr,  lr,  lsl #8
+        orr             lr,  lr,  lr,  lsl #16
+        ldrd            r4,  r5,  [r0]
+        ldr             r10, [r1, #4]
+        ldr             r9,  [r1], r2
+        subs            r3,  r3,  #2
+1:
+        pld             [r1, r2]
+        eor             r8,  r4,  r9
+        uhadd8          r4,  r4,  r9
+        eor             r12, r5,  r10
+        ldrd            r6,  r7,  [r0, r2]
+        uhadd8          r5,  r5,  r10
+        and             r8,  r8,  lr
+        ldr             r10, [r1, #4]
+        and             r12, r12, lr
+        uadd8           r4,  r4,  r8
+        ldr             r9,  [r1], r2
+        eor             r8,  r6,  r9
+        uadd8           r5,  r5,  r12
+        pld             [r1, r2,  lsl #1]
+        eor             r12, r7,  r10
+        uhadd8          r6,  r6,  r9
+        strd            r4,  r5,  [r0], r2
+        uhadd8          r7,  r7,  r10
+        beq             2f
+        and             r8,  r8,  lr
+        ldrd            r4,  r5,  [r0, r2]
+        uadd8           r6,  r6,  r8
+        ldr             r10, [r1, #4]
+        and             r12, r12, lr
+        subs            r3,  r3,  #2
+        uadd8           r7,  r7,  r12
+        ldr             r9,  [r1], r2
+        strd            r6,  r7,  [r0], r2
+        b               1b
+2:
+        and             r8,  r8,  lr
+        and             r12, r12, lr
+        uadd8           r6,  r6,  r8
+        uadd8           r7,  r7,  r12
+        strd            r6,  r7,  [r0], r2
+
+        pop             {r4-r10, pc}
+.endfunc
+
 function ff_add_pixels_clamped_armv6, export=1
         push            {r4-r8,lr}
         mov             r3,  #8
diff --git a/libavcodec/arm/dsputil_init_armv6.c b/libavcodec/arm/dsputil_init_armv6.c
index 76bde43f33..8dacc9df12 100644
--- a/libavcodec/arm/dsputil_init_armv6.c
+++ b/libavcodec/arm/dsputil_init_armv6.c
@@ -18,6 +18,9 @@
  * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
  */
 
+#include <stdint.h>
+
+#include "libavcodec/avcodec.h"
 #include "libavcodec/dsputil.h"
 #include "dsputil_arm.h"
 
@@ -25,6 +28,24 @@ void ff_simple_idct_armv6(DCTELEM *data);
 void ff_simple_idct_put_armv6(uint8_t *dest, int line_size, DCTELEM *data);
 void ff_simple_idct_add_armv6(uint8_t *dest, int line_size, DCTELEM *data);
 
+void ff_put_pixels16_armv6(uint8_t *, const uint8_t *, int, int);
+void ff_put_pixels16_x2_armv6(uint8_t *, const uint8_t *, int, int);
+void ff_put_pixels16_y2_armv6(uint8_t *, const uint8_t *, int, int);
+
+void ff_put_pixels16_x2_no_rnd_armv6(uint8_t *, const uint8_t *, int, int);
+void ff_put_pixels16_y2_no_rnd_armv6(uint8_t *, const uint8_t *, int, int);
+
+void ff_avg_pixels16_armv6(uint8_t *, const uint8_t *, int, int);
+
+void ff_put_pixels8_armv6(uint8_t *, const uint8_t *, int, int);
+void ff_put_pixels8_x2_armv6(uint8_t *, const uint8_t *, int, int);
+void ff_put_pixels8_y2_armv6(uint8_t *, const uint8_t *, int, int);
+
+void ff_put_pixels8_x2_no_rnd_armv6(uint8_t *, const uint8_t *, int, int);
+void ff_put_pixels8_y2_no_rnd_armv6(uint8_t *, const uint8_t *, int, int);
+
+void ff_avg_pixels8_armv6(uint8_t *, const uint8_t *, int, int);
+
 void ff_add_pixels_clamped_armv6(const DCTELEM *block,
                                  uint8_t *restrict pixels,
                                  int line_size);
@@ -39,5 +60,26 @@ void av_cold ff_dsputil_init_armv6(DSPContext* c, AVCodecContext *avctx)
         c->idct_permutation_type = FF_LIBMPEG2_IDCT_PERM;
     }
 
+    c->put_pixels_tab[0][0] = ff_put_pixels16_armv6;
+    c->put_pixels_tab[0][1] = ff_put_pixels16_x2_armv6;
+    c->put_pixels_tab[0][2] = ff_put_pixels16_y2_armv6;
+/*     c->put_pixels_tab[0][3] = ff_put_pixels16_xy2_armv6; */
+    c->put_pixels_tab[1][0] = ff_put_pixels8_armv6;
+    c->put_pixels_tab[1][1] = ff_put_pixels8_x2_armv6;
+    c->put_pixels_tab[1][2] = ff_put_pixels8_y2_armv6;
+/*     c->put_pixels_tab[1][3] = ff_put_pixels8_xy2_armv6; */
+
+    c->put_no_rnd_pixels_tab[0][0] = ff_put_pixels16_armv6;
+    c->put_no_rnd_pixels_tab[0][1] = ff_put_pixels16_x2_no_rnd_armv6;
+    c->put_no_rnd_pixels_tab[0][2] = ff_put_pixels16_y2_no_rnd_armv6;
+/*     c->put_no_rnd_pixels_tab[0][3] = ff_put_pixels16_xy2_no_rnd_armv6; */
+    c->put_no_rnd_pixels_tab[1][0] = ff_put_pixels8_armv6;
+    c->put_no_rnd_pixels_tab[1][1] = ff_put_pixels8_x2_no_rnd_armv6;
+    c->put_no_rnd_pixels_tab[1][2] = ff_put_pixels8_y2_no_rnd_armv6;
+/*     c->put_no_rnd_pixels_tab[1][3] = ff_put_pixels8_xy2_no_rnd_armv6; */
+
+    c->avg_pixels_tab[0][0] = ff_avg_pixels16_armv6;
+    c->avg_pixels_tab[1][0] = ff_avg_pixels8_armv6;
+
     c->add_pixels_clamped = ff_add_pixels_clamped_armv6;
 }