remove iwmmxt optimizations

The were broken since August of 2010 without anyone noticing until
three weeks ago. Nobody cares about it anymore and hopefully Marvell
will support NEON like in the PXA978 from now on.

13 files changed, 1 insertions, 1475 deletions

diff --git a/configure b/configure
index 4f956bb6b6..29fb432fb9 100755
--- a/configure
+++ b/configure
@@ -234,7 +234,6 @@ Advanced options (experts only):
   --disable-armv6          disable armv6 optimizations
   --disable-armv6t2        disable armv6t2 optimizations
   --disable-armvfp         disable ARM VFP optimizations
-  --disable-iwmmxt         disable iwmmxt optimizations
   --disable-mmi            disable MMI optimizations
   --disable-neon           disable NEON optimizations
   --disable-vis            disable VIS optimizations
@@ -1032,7 +1031,6 @@ ARCH_EXT_LIST='
     armv6t2
     armvfp
     avx
-    iwmmxt
     mmi
     mmx
     mmx2
@@ -1229,7 +1227,6 @@ armv5te_deps="arm"
 armv6_deps="arm"
 armv6t2_deps="arm"
 armvfp_deps="arm"
-iwmmxt_deps="arm"
 neon_deps="arm"
 vfpv3_deps="armvfp"
 
@@ -2690,7 +2687,6 @@ EOF
     enabled armv6   && check_asm armv6   '"sadd16 r0, r0, r0"'
     enabled armv6t2 && check_asm armv6t2 '"movt r0, #0"'
     enabled armvfp  && check_asm armvfp  '"fadds s0, s0, s0"'
-    enabled iwmmxt  && check_asm iwmmxt  '"wunpckelub wr6, wr4"'
     enabled neon    && check_asm neon    '"vadd.i16 q0, q0, q0"'
     enabled vfpv3   && check_asm vfpv3   '"vmov.f32 s0, #1.0"'
 
@@ -3199,7 +3195,6 @@ if enabled arm; then
     echo "ARMv6 enabled             ${armv6-no}"
     echo "ARMv6T2 enabled           ${armv6t2-no}"
     echo "ARM VFP enabled           ${armvfp-no}"
-    echo "IWMMXT enabled            ${iwmmxt-no}"
     echo "NEON enabled              ${neon-no}"
 fi
 if enabled mips; then
diff --git a/libavcodec/arm/Makefile b/libavcodec/arm/Makefile
index e7fa7e511a..2f5cea7df1 100644
--- a/libavcodec/arm/Makefile
+++ b/libavcodec/arm/Makefile
@@ -43,9 +43,6 @@ OBJS-$(HAVE_ARMVFP)                    += arm/dsputil_vfp.o             \
                                           arm/dsputil_init_vfp.o        \
                                           $(VFP-OBJS-yes)
 
-OBJS-$(HAVE_IWMMXT)                    += arm/dsputil_iwmmxt.o          \
-                                          arm/mpegvideo_iwmmxt.o        \
-
 NEON-OBJS-$(CONFIG_FFT)                += arm/fft_neon.o                \
                                           arm/fft_fixed_neon.o          \
 
diff --git a/libavcodec/arm/dsputil_arm.h b/libavcodec/arm/dsputil_arm.h
index 6d7e6a6d16..36d5d56899 100644
--- a/libavcodec/arm/dsputil_arm.h
+++ b/libavcodec/arm/dsputil_arm.h
@@ -28,6 +28,5 @@ void ff_dsputil_init_armv5te(DSPContext* c, AVCodecContext *avctx);
 void ff_dsputil_init_armv6(DSPContext* c, AVCodecContext *avctx);
 void ff_dsputil_init_vfp(DSPContext* c, AVCodecContext *avctx);
 void ff_dsputil_init_neon(DSPContext *c, AVCodecContext *avctx);
-void ff_dsputil_init_iwmmxt(DSPContext* c, AVCodecContext *avctx);
 
 #endif /* AVCODEC_ARM_DSPUTIL_H */
diff --git a/libavcodec/arm/dsputil_init_arm.c b/libavcodec/arm/dsputil_init_arm.c
index 4295e7288f..21e1351e16 100644
--- a/libavcodec/arm/dsputil_init_arm.c
+++ b/libavcodec/arm/dsputil_init_arm.c
@@ -119,7 +119,6 @@ void ff_dsputil_init_arm(DSPContext* c, AVCodecContext *avctx)
 
     if (HAVE_ARMV5TE) ff_dsputil_init_armv5te(c, avctx);
     if (HAVE_ARMV6)   ff_dsputil_init_armv6(c, avctx);
-    if (HAVE_IWMMXT)  ff_dsputil_init_iwmmxt(c, avctx);
     if (HAVE_ARMVFP)  ff_dsputil_init_vfp(c, avctx);
     if (HAVE_NEON)    ff_dsputil_init_neon(c, avctx);
 }
diff --git a/libavcodec/arm/dsputil_iwmmxt.c b/libavcodec/arm/dsputil_iwmmxt.c
deleted file mode 100644
index dec5d77472..0000000000
--- a/libavcodec/arm/dsputil_iwmmxt.c
+++ /dev/null
@@ -1,210 +0,0 @@
-/*
- * iWMMXt optimized DSP utils
- * Copyright (c) 2004 AGAWA Koji
- *
- * This file is part of Libav.
- *
- * Libav is free software; you can redistribute it and/or
- * modify it under the terms of the GNU Lesser General Public
- * License as published by the Free Software Foundation; either
- * version 2.1 of the License, or (at your option) any later version.
- *
- * Libav is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public
- * License along with Libav; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
- */
-
-#include "libavutil/cpu.h"
-#include "libavcodec/dsputil.h"
-
-#define DEF(x, y) x ## _no_rnd_ ## y ##_iwmmxt
-#define SET_RND(regd)  __asm__ volatile ("mov r12, #1 \n\t tbcsth " #regd ", r12":::"r12");
-#define WAVG2B "wavg2b"
-#include "dsputil_iwmmxt_rnd_template.c"
-#undef DEF
-#undef SET_RND
-#undef WAVG2B
-
-#define DEF(x, y) x ## _ ## y ##_iwmmxt
-#define SET_RND(regd)  __asm__ volatile ("mov r12, #2 \n\t tbcsth " #regd ", r12":::"r12");
-#define WAVG2B "wavg2br"
-#include "dsputil_iwmmxt_rnd_template.c"
-#undef DEF
-#undef SET_RND
-#undef WAVG2BR
-
-// need scheduling
-#define OP(AVG)                                         \
-    __asm__ volatile (                                      \
-        /* alignment */                                 \
-        "and r12, %[pixels], #7 \n\t"                   \
-        "bic %[pixels], %[pixels], #7 \n\t"             \
-        "tmcr wcgr1, r12 \n\t"                          \
-                                                        \
-        "wldrd wr0, [%[pixels]] \n\t"                   \
-        "wldrd wr1, [%[pixels], #8] \n\t"               \
-        "add %[pixels], %[pixels], %[line_size] \n\t"   \
-        "walignr1 wr4, wr0, wr1 \n\t"                   \
-                                                        \
-        "1: \n\t"                                       \
-                                                        \
-        "wldrd wr2, [%[pixels]] \n\t"                   \
-        "wldrd wr3, [%[pixels], #8] \n\t"               \
-        "add %[pixels], %[pixels], %[line_size] \n\t"   \
-        "pld [%[pixels]] \n\t"                          \
-        "walignr1 wr5, wr2, wr3 \n\t"                   \
-        AVG " wr6, wr4, wr5 \n\t"                       \
-        "wstrd wr6, [%[block]] \n\t"                    \
-        "add %[block], %[block], %[line_size] \n\t"     \
-                                                        \
-        "wldrd wr0, [%[pixels]] \n\t"                   \
-        "wldrd wr1, [%[pixels], #8] \n\t"               \
-        "add %[pixels], %[pixels], %[line_size] \n\t"   \
-        "walignr1 wr4, wr0, wr1 \n\t"                   \
-        "pld [%[pixels]] \n\t"                          \
-        AVG " wr6, wr4, wr5 \n\t"                       \
-        "wstrd wr6, [%[block]] \n\t"                    \
-        "add %[block], %[block], %[line_size] \n\t"     \
-                                                        \
-        "subs %[h], %[h], #2 \n\t"                      \
-        "bne 1b \n\t"                                   \
-        : [block]"+r"(block), [pixels]"+r"(pixels), [h]"+r"(h)  \
-        : [line_size]"r"(line_size) \
-        : "memory", "r12");
-void put_pixels8_y2_iwmmxt(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
-{
-    OP("wavg2br");
-}
-void put_no_rnd_pixels8_y2_iwmmxt(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
-{
-    OP("wavg2b");
-}
-#undef OP
-
-void add_pixels_clamped_iwmmxt(const DCTELEM *block, uint8_t *pixels, int line_size)
-{
-    uint8_t *pixels2 = pixels + line_size;
-
-    __asm__ volatile (
-        "mov            r12, #4                 \n\t"
-        "1:                                     \n\t"
-        "pld            [%[pixels], %[line_size2]]              \n\t"
-        "pld            [%[pixels2], %[line_size2]]             \n\t"
-        "wldrd          wr4, [%[pixels]]        \n\t"
-        "wldrd          wr5, [%[pixels2]]       \n\t"
-        "pld            [%[block], #32]         \n\t"
-        "wunpckelub     wr6, wr4                \n\t"
-        "wldrd          wr0, [%[block]]         \n\t"
-        "wunpckehub     wr7, wr4                \n\t"
-        "wldrd          wr1, [%[block], #8]     \n\t"
-        "wunpckelub     wr8, wr5                \n\t"
-        "wldrd          wr2, [%[block], #16]    \n\t"
-        "wunpckehub     wr9, wr5                \n\t"
-        "wldrd          wr3, [%[block], #24]    \n\t"
-        "add            %[block], %[block], #32 \n\t"
-        "waddhss        wr10, wr0, wr6          \n\t"
-        "waddhss        wr11, wr1, wr7          \n\t"
-        "waddhss        wr12, wr2, wr8          \n\t"
-        "waddhss        wr13, wr3, wr9          \n\t"
-        "wpackhus       wr14, wr10, wr11        \n\t"
-        "wpackhus       wr15, wr12, wr13        \n\t"
-        "wstrd          wr14, [%[pixels]]       \n\t"
-        "add            %[pixels], %[pixels], %[line_size2]     \n\t"
-        "subs           r12, r12, #1            \n\t"
-        "wstrd          wr15, [%[pixels2]]      \n\t"
-        "add            %[pixels2], %[pixels2], %[line_size2]   \n\t"
-        "bne            1b                      \n\t"
-        : [block]"+r"(block), [pixels]"+r"(pixels), [pixels2]"+r"(pixels2)
-        : [line_size2]"r"(line_size << 1)
-        : "cc", "memory", "r12");
-}
-
-static void clear_blocks_iwmmxt(DCTELEM *blocks)
-{
-    __asm__ volatile(
-                "wzero wr0                      \n\t"
-                "mov r1, #(128 * 6 / 32)        \n\t"
-                "1:                             \n\t"
-                "wstrd wr0, [%0]                \n\t"
-                "wstrd wr0, [%0, #8]            \n\t"
-                "wstrd wr0, [%0, #16]           \n\t"
-                "wstrd wr0, [%0, #24]           \n\t"
-                "subs r1, r1, #1                \n\t"
-                "add %0, %0, #32                \n\t"
-                "bne 1b                         \n\t"
-                : "+r"(blocks)
-                :
-                : "r1"
-        );
-}
-
-static void nop(uint8_t *block, const uint8_t *pixels, int line_size, int h)
-{
-    return;
-}
-
-/* A run time test is not simple. If this file is compiled in
- * then we should install the functions
- */
-
-void ff_dsputil_init_iwmmxt(DSPContext* c, AVCodecContext *avctx)
-{
-    int mm_flags = AV_CPU_FLAG_IWMMXT; /* multimedia extension flags */
-    const int high_bit_depth = avctx->bits_per_raw_sample > 8;
-
-    if (avctx->dsp_mask) {
-        if (avctx->dsp_mask & AV_CPU_FLAG_FORCE)
-            mm_flags |= (avctx->dsp_mask & 0xffff);
-        else
-            mm_flags &= ~(avctx->dsp_mask & 0xffff);
-    }
-
-    if (!(mm_flags & AV_CPU_FLAG_IWMMXT)) return;
-
-    c->add_pixels_clamped = add_pixels_clamped_iwmmxt;
-
-    if (!high_bit_depth) {
-    c->clear_blocks = clear_blocks_iwmmxt;
-
-    c->put_pixels_tab[0][0] = put_pixels16_iwmmxt;
-    c->put_pixels_tab[0][1] = put_pixels16_x2_iwmmxt;
-    c->put_pixels_tab[0][2] = put_pixels16_y2_iwmmxt;
-    c->put_pixels_tab[0][3] = put_pixels16_xy2_iwmmxt;
-    c->put_no_rnd_pixels_tab[0][0] = put_pixels16_iwmmxt;
-    c->put_no_rnd_pixels_tab[0][1] = put_no_rnd_pixels16_x2_iwmmxt;
-    c->put_no_rnd_pixels_tab[0][2] = put_no_rnd_pixels16_y2_iwmmxt;
-    c->put_no_rnd_pixels_tab[0][3] = put_no_rnd_pixels16_xy2_iwmmxt;
-
-    c->put_pixels_tab[1][0] = put_pixels8_iwmmxt;
-    c->put_pixels_tab[1][1] = put_pixels8_x2_iwmmxt;
-    c->put_pixels_tab[1][2] = put_pixels8_y2_iwmmxt;
-    c->put_pixels_tab[1][3] = put_pixels8_xy2_iwmmxt;
-    c->put_no_rnd_pixels_tab[1][0] = put_pixels8_iwmmxt;
-    c->put_no_rnd_pixels_tab[1][1] = put_no_rnd_pixels8_x2_iwmmxt;
-    c->put_no_rnd_pixels_tab[1][2] = put_no_rnd_pixels8_y2_iwmmxt;
-    c->put_no_rnd_pixels_tab[1][3] = put_no_rnd_pixels8_xy2_iwmmxt;
-
-    c->avg_pixels_tab[0][0] = avg_pixels16_iwmmxt;
-    c->avg_pixels_tab[0][1] = avg_pixels16_x2_iwmmxt;
-    c->avg_pixels_tab[0][2] = avg_pixels16_y2_iwmmxt;
-    c->avg_pixels_tab[0][3] = avg_pixels16_xy2_iwmmxt;
-    c->avg_no_rnd_pixels_tab[0][0] = avg_pixels16_iwmmxt;
-    c->avg_no_rnd_pixels_tab[0][1] = avg_no_rnd_pixels16_x2_iwmmxt;
-    c->avg_no_rnd_pixels_tab[0][2] = avg_no_rnd_pixels16_y2_iwmmxt;
-    c->avg_no_rnd_pixels_tab[0][3] = avg_no_rnd_pixels16_xy2_iwmmxt;
-
-    c->avg_pixels_tab[1][0] = avg_pixels8_iwmmxt;
-    c->avg_pixels_tab[1][1] = avg_pixels8_x2_iwmmxt;
-    c->avg_pixels_tab[1][2] = avg_pixels8_y2_iwmmxt;
-    c->avg_pixels_tab[1][3] = avg_pixels8_xy2_iwmmxt;
-    c->avg_no_rnd_pixels_tab[1][0] = avg_no_rnd_pixels8_iwmmxt;
-    c->avg_no_rnd_pixels_tab[1][1] = avg_no_rnd_pixels8_x2_iwmmxt;
-    c->avg_no_rnd_pixels_tab[1][2] = avg_no_rnd_pixels8_y2_iwmmxt;
-    c->avg_no_rnd_pixels_tab[1][3] = avg_no_rnd_pixels8_xy2_iwmmxt;
-    }
-}
diff --git a/libavcodec/arm/dsputil_iwmmxt_rnd_template.c b/libavcodec/arm/dsputil_iwmmxt_rnd_template.c
deleted file mode 100644
index df0ead6d80..0000000000
--- a/libavcodec/arm/dsputil_iwmmxt_rnd_template.c
+++ /dev/null
@@ -1,1114 +0,0 @@
-/*
- * iWMMXt optimized DSP utils
- * copyright (c) 2004 AGAWA Koji
- *
- * This file is part of Libav.
- *
- * Libav is free software; you can redistribute it and/or
- * modify it under the terms of the GNU Lesser General Public
- * License as published by the Free Software Foundation; either
- * version 2.1 of the License, or (at your option) any later version.
- *
- * Libav is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public
- * License along with Libav; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
- */
-
-void DEF(put, pixels8)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
-{
-    int stride = line_size;
-    __asm__ volatile (
-        "and r12, %[pixels], #7 \n\t"
-        "bic %[pixels], %[pixels], #7 \n\t"
-        "tmcr wcgr1, r12 \n\t"
-        "add r4, %[pixels], %[line_size] \n\t"
-        "add r5, %[block], %[line_size] \n\t"
-        "mov %[line_size], %[line_size], lsl #1 \n\t"
-        "1: \n\t"
-        "wldrd wr0, [%[pixels]] \n\t"
-        "subs %[h], %[h], #2 \n\t"
-        "wldrd wr1, [%[pixels], #8] \n\t"
-        "add %[pixels], %[pixels], %[line_size] \n\t"
-        "wldrd wr3, [r4] \n\t"
-        "pld [%[pixels]] \n\t"
-        "pld [%[pixels], #32] \n\t"
-        "wldrd wr4, [r4, #8] \n\t"
-        "add r4, r4, %[line_size] \n\t"
-        "walignr1 wr8, wr0, wr1 \n\t"
-        "pld [r4] \n\t"
-        "pld [r4, #32] \n\t"
-        "walignr1 wr10, wr3, wr4 \n\t"
-        "wstrd wr8, [%[block]] \n\t"
-        "add %[block], %[block], %[line_size] \n\t"
-        "wstrd wr10, [r5] \n\t"
-        "add r5, r5, %[line_size] \n\t"
-        "bne 1b \n\t"
-        : [block]"+r"(block), [pixels]"+r"(pixels), [line_size]"+r"(stride), [h]"+r"(h)
-        :
-        : "memory", "r4", "r5", "r12");
-}
-
-void DEF(avg, pixels8)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
-{
-    int stride = line_size;
-    __asm__ volatile (
-        "and r12, %[pixels], #7 \n\t"
-        "bic %[pixels], %[pixels], #7 \n\t"
-        "tmcr wcgr1, r12 \n\t"
-        "add r4, %[pixels], %[line_size] \n\t"
-        "add r5, %[block], %[line_size] \n\t"
-        "mov %[line_size], %[line_size], lsl #1 \n\t"
-        "1: \n\t"
-        "wldrd wr0, [%[pixels]] \n\t"
-        "subs %[h], %[h], #2 \n\t"
-        "wldrd wr1, [%[pixels], #8] \n\t"
-        "add %[pixels], %[pixels], %[line_size] \n\t"
-        "wldrd wr3, [r4] \n\t"
-        "pld [%[pixels]] \n\t"
-        "pld [%[pixels], #32] \n\t"
-        "wldrd wr4, [r4, #8] \n\t"
-        "add r4, r4, %[line_size] \n\t"
-        "walignr1 wr8, wr0, wr1 \n\t"
-        "wldrd wr0, [%[block]] \n\t"
-        "wldrd wr2, [r5] \n\t"
-        "pld [r4] \n\t"
-        "pld [r4, #32] \n\t"
-        "walignr1 wr10, wr3, wr4 \n\t"
-        WAVG2B" wr8, wr8, wr0 \n\t"
-        WAVG2B" wr10, wr10, wr2 \n\t"
-        "wstrd wr8, [%[block]] \n\t"
-        "add %[block], %[block], %[line_size] \n\t"
-        "wstrd wr10, [r5] \n\t"
-        "pld [%[block]] \n\t"
-        "pld [%[block], #32] \n\t"
-        "add r5, r5, %[line_size] \n\t"
-        "pld [r5] \n\t"
-        "pld [r5, #32] \n\t"
-        "bne 1b \n\t"
-        : [block]"+r"(block), [pixels]"+r"(pixels), [line_size]"+r"(stride), [h]"+r"(h)
-        :
-        : "memory", "r4", "r5", "r12");
-}
-
-void DEF(put, pixels16)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
-{
-    int stride = line_size;
-    __asm__ volatile (
-        "and r12, %[pixels], #7 \n\t"
-        "bic %[pixels], %[pixels], #7 \n\t"
-        "tmcr wcgr1, r12 \n\t"
-        "add r4, %[pixels], %[line_size] \n\t"
-        "add r5, %[block], %[line_size] \n\t"
-        "mov %[line_size], %[line_size], lsl #1 \n\t"
-        "1: \n\t"
-        "wldrd wr0, [%[pixels]] \n\t"
-        "wldrd wr1, [%[pixels], #8] \n\t"
-        "subs %[h], %[h], #2 \n\t"
-        "wldrd wr2, [%[pixels], #16] \n\t"
-        "add %[pixels], %[pixels], %[line_size] \n\t"
-        "wldrd wr3, [r4] \n\t"
-        "pld [%[pixels]] \n\t"
-        "pld [%[pixels], #32] \n\t"
-        "walignr1 wr8, wr0, wr1 \n\t"
-        "wldrd wr4, [r4, #8] \n\t"
-        "walignr1 wr9, wr1, wr2 \n\t"
-        "wldrd wr5, [r4, #16] \n\t"
-        "add r4, r4, %[line_size] \n\t"
-        "pld [r4] \n\t"
-        "pld [r4, #32] \n\t"
-        "walignr1 wr10, wr3, wr4 \n\t"
-        "wstrd wr8, [%[block]] \n\t"
-        "walignr1 wr11, wr4, wr5 \n\t"
-        "wstrd wr9, [%[block], #8] \n\t"
-        "add %[block], %[block], %[line_size] \n\t"
-        "wstrd wr10, [r5] \n\t"
-        "wstrd wr11, [r5, #8] \n\t"
-        "add r5, r5, %[line_size] \n\t"
-        "bne 1b \n\t"
-        : [block]"+r"(block), [pixels]"+r"(pixels), [line_size]"+r"(stride), [h]"+r"(h)
-        :
-        : "memory", "r4", "r5", "r12");
-}
-
-void DEF(avg, pixels16)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
-{
-    int stride = line_size;
-    __asm__ volatile (
-        "pld [%[pixels]]                \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "pld [%[block]]                 \n\t"
-        "pld [%[block], #32]            \n\t"
-        "and r12, %[pixels], #7         \n\t"
-        "bic %[pixels], %[pixels], #7   \n\t"
-        "tmcr wcgr1, r12                \n\t"
-        "add r4, %[pixels], %[line_size]\n\t"
-        "add r5, %[block], %[line_size] \n\t"
-        "mov %[line_size], %[line_size], lsl #1 \n\t"
-        "1:                             \n\t"
-        "wldrd wr0, [%[pixels]]         \n\t"
-        "wldrd wr1, [%[pixels], #8]     \n\t"
-        "subs %[h], %[h], #2            \n\t"
-        "wldrd wr2, [%[pixels], #16]    \n\t"
-        "add %[pixels], %[pixels], %[line_size] \n\t"
-        "wldrd wr3, [r4]                \n\t"
-        "pld [%[pixels]]                \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "walignr1 wr8, wr0, wr1         \n\t"
-        "wldrd wr4, [r4, #8]            \n\t"
-        "walignr1 wr9, wr1, wr2         \n\t"
-        "wldrd wr5, [r4, #16]           \n\t"
-        "add r4, r4, %[line_size]       \n\t"
-        "wldrd wr0, [%[block]]          \n\t"
-        "pld [r4]                       \n\t"
-        "wldrd wr1, [%[block], #8]      \n\t"
-        "pld [r4, #32]                  \n\t"
-        "wldrd wr2, [r5]                \n\t"
-        "walignr1 wr10, wr3, wr4        \n\t"
-        "wldrd wr3, [r5, #8]            \n\t"
-        WAVG2B" wr8, wr8, wr0           \n\t"
-        WAVG2B" wr9, wr9, wr1           \n\t"
-        WAVG2B" wr10, wr10, wr2         \n\t"
-        "wstrd wr8, [%[block]]          \n\t"
-        "walignr1 wr11, wr4, wr5        \n\t"
-        WAVG2B" wr11, wr11, wr3         \n\t"
-        "wstrd wr9, [%[block], #8]      \n\t"
-        "add %[block], %[block], %[line_size] \n\t"
-        "wstrd wr10, [r5]               \n\t"
-        "pld [%[block]]                 \n\t"
-        "pld [%[block], #32]            \n\t"
-        "wstrd wr11, [r5, #8]           \n\t"
-        "add r5, r5, %[line_size]       \n\t"
-        "pld [r5]                       \n\t"
-        "pld [r5, #32]                  \n\t"
-        "bne 1b \n\t"
-        : [block]"+r"(block), [pixels]"+r"(pixels), [line_size]"+r"(stride), [h]"+r"(h)
-        :
-        : "memory", "r4", "r5", "r12");
-}
-
-void DEF(put, pixels8_x2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
-{
-    int stride = line_size;
-    // [wr0 wr1 wr2 wr3] for previous line
-    // [wr4 wr5 wr6 wr7] for current line
-    SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
-    __asm__ volatile(
-        "pld [%[pixels]]                \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "and r12, %[pixels], #7         \n\t"
-        "bic %[pixels], %[pixels], #7   \n\t"
-        "tmcr wcgr1, r12                \n\t"
-        "add r12, r12, #1               \n\t"
-        "add r4, %[pixels], %[line_size]\n\t"
-        "tmcr wcgr2, r12                \n\t"
-        "add r5, %[block], %[line_size] \n\t"
-        "mov %[line_size], %[line_size], lsl #1 \n\t"
-
-        "1:                             \n\t"
-        "wldrd wr10, [%[pixels]]        \n\t"
-        "cmp r12, #8                    \n\t"
-        "wldrd wr11, [%[pixels], #8]    \n\t"
-        "add %[pixels], %[pixels], %[line_size] \n\t"
-        "wldrd wr13, [r4]               \n\t"
-        "pld [%[pixels]]                \n\t"
-        "wldrd wr14, [r4, #8]           \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "add r4, r4, %[line_size]       \n\t"
-        "walignr1 wr0, wr10, wr11       \n\t"
-        "pld [r4]                       \n\t"
-        "pld [r4, #32]                  \n\t"
-        "walignr1 wr2, wr13, wr14       \n\t"
-        "wmoveq wr4, wr11               \n\t"
-        "wmoveq wr6, wr14               \n\t"
-        "walignr2ne wr4, wr10, wr11     \n\t"
-        "walignr2ne wr6, wr13, wr14     \n\t"
-        WAVG2B" wr0, wr0, wr4           \n\t"
-        WAVG2B" wr2, wr2, wr6           \n\t"
-        "wstrd wr0, [%[block]]          \n\t"
-        "subs %[h], %[h], #2            \n\t"
-        "wstrd wr2, [r5]                \n\t"
-        "add %[block], %[block], %[line_size]   \n\t"
-        "add r5, r5, %[line_size]       \n\t"
-        "bne 1b                         \n\t"
-        : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
-        :
-        : "r4", "r5", "r12", "memory");
-}
-
-void DEF(put, pixels16_x2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
-{
-    int stride = line_size;
-    // [wr0 wr1 wr2 wr3] for previous line
-    // [wr4 wr5 wr6 wr7] for current line
-    SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
-    __asm__ volatile(
-        "pld [%[pixels]]                \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "and r12, %[pixels], #7         \n\t"
-        "bic %[pixels], %[pixels], #7   \n\t"
-        "tmcr wcgr1, r12                \n\t"
-        "add r12, r12, #1               \n\t"
-        "add r4, %[pixels], %[line_size]\n\t"
-        "tmcr wcgr2, r12                \n\t"
-        "add r5, %[block], %[line_size] \n\t"
-        "mov %[line_size], %[line_size], lsl #1 \n\t"
-
-        "1:                             \n\t"
-        "wldrd wr10, [%[pixels]]        \n\t"
-        "cmp r12, #8                    \n\t"
-        "wldrd wr11, [%[pixels], #8]    \n\t"
-        "wldrd wr12, [%[pixels], #16]   \n\t"
-        "add %[pixels], %[pixels], %[line_size] \n\t"
-        "wldrd wr13, [r4]               \n\t"
-        "pld [%[pixels]]                \n\t"
-        "wldrd wr14, [r4, #8]           \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "wldrd wr15, [r4, #16]          \n\t"
-        "add r4, r4, %[line_size]       \n\t"
-        "walignr1 wr0, wr10, wr11       \n\t"
-        "pld [r4]                       \n\t"
-        "pld [r4, #32]                  \n\t"
-        "walignr1 wr1, wr11, wr12       \n\t"
-        "walignr1 wr2, wr13, wr14       \n\t"
-        "walignr1 wr3, wr14, wr15       \n\t"
-        "wmoveq wr4, wr11               \n\t"
-        "wmoveq wr5, wr12               \n\t"
-        "wmoveq wr6, wr14               \n\t"
-        "wmoveq wr7, wr15               \n\t"
-        "walignr2ne wr4, wr10, wr11     \n\t"
-        "walignr2ne wr5, wr11, wr12     \n\t"
-        "walignr2ne wr6, wr13, wr14     \n\t"
-        "walignr2ne wr7, wr14, wr15     \n\t"
-        WAVG2B" wr0, wr0, wr4           \n\t"
-        WAVG2B" wr1, wr1, wr5           \n\t"
-        "wstrd wr0, [%[block]]          \n\t"
-        WAVG2B" wr2, wr2, wr6           \n\t"
-        "wstrd wr1, [%[block], #8]      \n\t"
-        WAVG2B" wr3, wr3, wr7           \n\t"
-        "add %[block], %[block], %[line_size]   \n\t"
-        "wstrd wr2, [r5]                \n\t"
-        "subs %[h], %[h], #2            \n\t"
-        "wstrd wr3, [r5, #8]            \n\t"
-        "add r5, r5, %[line_size]       \n\t"
-        "bne 1b                         \n\t"
-        : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
-        :
-        : "r4", "r5", "r12", "memory");
-}
-
-void DEF(avg, pixels8_x2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
-{
-    int stride = line_size;
-    // [wr0 wr1 wr2 wr3] for previous line
-    // [wr4 wr5 wr6 wr7] for current line
-    SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
-    __asm__ volatile(
-        "pld [%[pixels]]                \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "pld [%[block]]                 \n\t"
-        "pld [%[block], #32]            \n\t"
-        "and r12, %[pixels], #7         \n\t"
-        "bic %[pixels], %[pixels], #7   \n\t"
-        "tmcr wcgr1, r12                \n\t"
-        "add r12, r12, #1               \n\t"
-        "add r4, %[pixels], %[line_size]\n\t"
-        "tmcr wcgr2, r12                \n\t"
-        "add r5, %[block], %[line_size] \n\t"
-        "mov %[line_size], %[line_size], lsl #1 \n\t"
-        "pld [r5]                       \n\t"
-        "pld [r5, #32]                  \n\t"
-
-        "1:                             \n\t"
-        "wldrd wr10, [%[pixels]]        \n\t"
-        "cmp r12, #8                    \n\t"
-        "wldrd wr11, [%[pixels], #8]    \n\t"
-        "add %[pixels], %[pixels], %[line_size] \n\t"
-        "wldrd wr13, [r4]               \n\t"
-        "pld [%[pixels]]                \n\t"
-        "wldrd wr14, [r4, #8]           \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "add r4, r4, %[line_size]       \n\t"
-        "walignr1 wr0, wr10, wr11       \n\t"
-        "pld [r4]                       \n\t"
-        "pld [r4, #32]                  \n\t"
-        "walignr1 wr2, wr13, wr14       \n\t"
-        "wmoveq wr4, wr11               \n\t"
-        "wmoveq wr6, wr14               \n\t"
-        "walignr2ne wr4, wr10, wr11     \n\t"
-        "wldrd wr10, [%[block]]         \n\t"
-        "walignr2ne wr6, wr13, wr14     \n\t"
-        "wldrd wr12, [r5]               \n\t"
-        WAVG2B" wr0, wr0, wr4           \n\t"
-        WAVG2B" wr2, wr2, wr6           \n\t"
-        WAVG2B" wr0, wr0, wr10          \n\t"
-        WAVG2B" wr2, wr2, wr12          \n\t"
-        "wstrd wr0, [%[block]]          \n\t"
-        "subs %[h], %[h], #2            \n\t"
-        "wstrd wr2, [r5]                \n\t"
-        "add %[block], %[block], %[line_size]   \n\t"
-        "add r5, r5, %[line_size]       \n\t"
-        "pld [%[block]]                 \n\t"
-        "pld [%[block], #32]            \n\t"
-        "pld [r5]                       \n\t"
-        "pld [r5, #32]                  \n\t"
-        "bne 1b                         \n\t"
-        : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
-        :
-        : "r4", "r5", "r12", "memory");
-}
-
-void DEF(avg, pixels16_x2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
-{
-    int stride = line_size;
-    // [wr0 wr1 wr2 wr3] for previous line
-    // [wr4 wr5 wr6 wr7] for current line
-    SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
-    __asm__ volatile(
-        "pld [%[pixels]]                \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "pld [%[block]]                 \n\t"
-        "pld [%[block], #32]            \n\t"
-        "and r12, %[pixels], #7         \n\t"
-        "bic %[pixels], %[pixels], #7   \n\t"
-        "tmcr wcgr1, r12                \n\t"
-        "add r12, r12, #1               \n\t"
-        "add r4, %[pixels], %[line_size]\n\t"
-        "tmcr wcgr2, r12                \n\t"
-        "add r5, %[block], %[line_size] \n\t"
-        "mov %[line_size], %[line_size], lsl #1 \n\t"
-        "pld [r5]                       \n\t"
-        "pld [r5, #32]                  \n\t"
-
-        "1:                             \n\t"
-        "wldrd wr10, [%[pixels]]        \n\t"
-        "cmp r12, #8                    \n\t"
-        "wldrd wr11, [%[pixels], #8]    \n\t"
-        "wldrd wr12, [%[pixels], #16]   \n\t"
-        "add %[pixels], %[pixels], %[line_size] \n\t"
-        "wldrd wr13, [r4]               \n\t"
-        "pld [%[pixels]]                \n\t"
-        "wldrd wr14, [r4, #8]           \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "wldrd wr15, [r4, #16]          \n\t"
-        "add r4, r4, %[line_size]       \n\t"
-        "walignr1 wr0, wr10, wr11       \n\t"
-        "pld [r4]                       \n\t"
-        "pld [r4, #32]                  \n\t"
-        "walignr1 wr1, wr11, wr12       \n\t"
-        "walignr1 wr2, wr13, wr14       \n\t"
-        "walignr1 wr3, wr14, wr15       \n\t"
-        "wmoveq wr4, wr11               \n\t"
-        "wmoveq wr5, wr12               \n\t"
-        "wmoveq wr6, wr14               \n\t"
-        "wmoveq wr7, wr15               \n\t"
-        "walignr2ne wr4, wr10, wr11     \n\t"
-        "walignr2ne wr5, wr11, wr12     \n\t"
-        "walignr2ne wr6, wr13, wr14     \n\t"
-        "walignr2ne wr7, wr14, wr15     \n\t"
-        "wldrd wr10, [%[block]]         \n\t"
-        WAVG2B" wr0, wr0, wr4           \n\t"
-        "wldrd wr11, [%[block], #8]     \n\t"
-        WAVG2B" wr1, wr1, wr5           \n\t"
-        "wldrd wr12, [r5]               \n\t"
-        WAVG2B" wr2, wr2, wr6           \n\t"
-        "wldrd wr13, [r5, #8]           \n\t"
-        WAVG2B" wr3, wr3, wr7           \n\t"
-        WAVG2B" wr0, wr0, wr10          \n\t"
-        WAVG2B" wr1, wr1, wr11          \n\t"
-        WAVG2B" wr2, wr2, wr12          \n\t"
-        WAVG2B" wr3, wr3, wr13          \n\t"
-        "wstrd wr0, [%[block]]          \n\t"
-        "subs %[h], %[h], #2            \n\t"
-        "wstrd wr1, [%[block], #8]      \n\t"
-        "add %[block], %[block], %[line_size]   \n\t"
-        "wstrd wr2, [r5]                \n\t"
-        "pld [%[block]]                 \n\t"
-        "wstrd wr3, [r5, #8]            \n\t"
-        "add r5, r5, %[line_size]       \n\t"
-        "pld [%[block], #32]            \n\t"
-        "pld [r5]                       \n\t"
-        "pld [r5, #32]                  \n\t"
-        "bne 1b                         \n\t"
-        : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
-        :
-        :"r4", "r5", "r12", "memory");
-}
-
-void DEF(avg, pixels8_y2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
-{
-    int stride = line_size;
-    // [wr0 wr1 wr2 wr3] for previous line
-    // [wr4 wr5 wr6 wr7] for current line
-    __asm__ volatile(
-        "pld            [%[pixels]]                             \n\t"
-        "pld            [%[pixels], #32]                        \n\t"
-        "and            r12, %[pixels], #7                      \n\t"
-        "tmcr           wcgr1, r12                              \n\t"
-        "bic            %[pixels], %[pixels], #7                \n\t"
-
-        "wldrd          wr10, [%[pixels]]                       \n\t"
-        "wldrd          wr11, [%[pixels], #8]                   \n\t"
-        "pld            [%[block]]                              \n\t"
-        "add            %[pixels], %[pixels], %[line_size]      \n\t"
-        "walignr1       wr0, wr10, wr11                         \n\t"
-        "pld            [%[pixels]]                             \n\t"
-        "pld            [%[pixels], #32]                        \n\t"
-
-      "1:                                                       \n\t"
-        "wldrd          wr10, [%[pixels]]                       \n\t"
-        "wldrd          wr11, [%[pixels], #8]                   \n\t"
-        "add            %[pixels], %[pixels], %[line_size]      \n\t"
-        "pld            [%[pixels]]                             \n\t"
-        "pld            [%[pixels], #32]                        \n\t"
-        "walignr1       wr4, wr10, wr11                         \n\t"
-        "wldrd          wr10, [%[block]]                        \n\t"
-         WAVG2B"        wr8, wr0, wr4                           \n\t"
-         WAVG2B"        wr8, wr8, wr10                          \n\t"
-        "wstrd          wr8, [%[block]]                         \n\t"
-        "add            %[block], %[block], %[line_size]        \n\t"
-
-        "wldrd          wr10, [%[pixels]]                       \n\t"
-        "wldrd          wr11, [%[pixels], #8]                   \n\t"
-        "pld            [%[block]]                              \n\t"
-        "add            %[pixels], %[pixels], %[line_size]      \n\t"
-        "pld            [%[pixels]]                             \n\t"
-        "pld            [%[pixels], #32]                        \n\t"
-        "walignr1       wr0, wr10, wr11                         \n\t"
-        "wldrd          wr10, [%[block]]                        \n\t"
-         WAVG2B"        wr8, wr0, wr4                           \n\t"
-         WAVG2B"        wr8, wr8, wr10                          \n\t"
-        "wstrd          wr8, [%[block]]                         \n\t"
-        "add            %[block], %[block], %[line_size]        \n\t"
-
-        "subs           %[h], %[h], #2                          \n\t"
-        "pld            [%[block]]                              \n\t"
-        "bne            1b                                      \n\t"
-        : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
-        :
-        : "cc", "memory", "r12");
-}
-
-void DEF(put, pixels16_y2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
-{
-    int stride = line_size;
-    // [wr0 wr1 wr2 wr3] for previous line
-    // [wr4 wr5 wr6 wr7] for current line
-    __asm__ volatile(
-        "pld [%[pixels]]                \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "and r12, %[pixels], #7         \n\t"
-        "tmcr wcgr1, r12                \n\t"
-        "bic %[pixels], %[pixels], #7   \n\t"
-
-        "wldrd wr10, [%[pixels]]        \n\t"
-        "wldrd wr11, [%[pixels], #8]    \n\t"
-        "wldrd wr12, [%[pixels], #16]   \n\t"
-        "add %[pixels], %[pixels], %[line_size] \n\t"
-        "pld [%[pixels]]                \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "walignr1 wr0, wr10, wr11       \n\t"
-        "walignr1 wr1, wr11, wr12       \n\t"
-
-        "1:                             \n\t"
-        "wldrd wr10, [%[pixels]]        \n\t"
-        "wldrd wr11, [%[pixels], #8]    \n\t"
-        "wldrd wr12, [%[pixels], #16]   \n\t"
-        "add %[pixels], %[pixels], %[line_size] \n\t"
-        "pld [%[pixels]]                \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "walignr1 wr4, wr10, wr11       \n\t"
-        "walignr1 wr5, wr11, wr12       \n\t"
-        WAVG2B" wr8, wr0, wr4           \n\t"
-        WAVG2B" wr9, wr1, wr5           \n\t"
-        "wstrd wr8, [%[block]]          \n\t"
-        "wstrd wr9, [%[block], #8]      \n\t"
-        "add %[block], %[block], %[line_size]   \n\t"
-
-        "wldrd wr10, [%[pixels]]        \n\t"
-        "wldrd wr11, [%[pixels], #8]    \n\t"
-        "wldrd wr12, [%[pixels], #16]   \n\t"
-        "add %[pixels], %[pixels], %[line_size] \n\t"
-        "pld [%[pixels]]                \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "walignr1 wr0, wr10, wr11       \n\t"
-        "walignr1 wr1, wr11, wr12       \n\t"
-        WAVG2B" wr8, wr0, wr4           \n\t"
-        WAVG2B" wr9, wr1, wr5           \n\t"
-        "wstrd wr8, [%[block]]          \n\t"
-        "wstrd wr9, [%[block], #8]      \n\t"
-        "add %[block], %[block], %[line_size]   \n\t"
-
-        "subs %[h], %[h], #2            \n\t"
-        "bne 1b                         \n\t"
-        : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
-        :
-        : "r4", "r5", "r12", "memory");
-}
-
-void DEF(avg, pixels16_y2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
-{
-    int stride = line_size;
-    // [wr0 wr1 wr2 wr3] for previous line
-    // [wr4 wr5 wr6 wr7] for current line
-    __asm__ volatile(
-        "pld [%[pixels]]                \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "and r12, %[pixels], #7         \n\t"
-        "tmcr wcgr1, r12                \n\t"
-        "bic %[pixels], %[pixels], #7   \n\t"
-
-        "wldrd wr10, [%[pixels]]        \n\t"
-        "wldrd wr11, [%[pixels], #8]    \n\t"
-        "pld [%[block]]                 \n\t"
-        "wldrd wr12, [%[pixels], #16]   \n\t"
-        "add %[pixels], %[pixels], %[line_size] \n\t"
-        "pld [%[pixels]]                \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "walignr1 wr0, wr10, wr11       \n\t"
-        "walignr1 wr1, wr11, wr12       \n\t"
-
-        "1:                             \n\t"
-        "wldrd wr10, [%[pixels]]        \n\t"
-        "wldrd wr11, [%[pixels], #8]    \n\t"
-        "wldrd wr12, [%[pixels], #16]   \n\t"
-        "add %[pixels], %[pixels], %[line_size] \n\t"
-        "pld [%[pixels]]                \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "walignr1 wr4, wr10, wr11       \n\t"
-        "walignr1 wr5, wr11, wr12       \n\t"
-        "wldrd wr10, [%[block]]         \n\t"
-        "wldrd wr11, [%[block], #8]     \n\t"
-        WAVG2B" wr8, wr0, wr4           \n\t"
-        WAVG2B" wr9, wr1, wr5           \n\t"
-        WAVG2B" wr8, wr8, wr10          \n\t"
-        WAVG2B" wr9, wr9, wr11          \n\t"
-        "wstrd wr8, [%[block]]          \n\t"
-        "wstrd wr9, [%[block], #8]      \n\t"
-        "add %[block], %[block], %[line_size]   \n\t"
-
-        "wldrd wr10, [%[pixels]]        \n\t"
-        "wldrd wr11, [%[pixels], #8]    \n\t"
-        "pld [%[block]]                 \n\t"
-        "wldrd wr12, [%[pixels], #16]   \n\t"
-        "add %[pixels], %[pixels], %[line_size] \n\t"
-        "pld [%[pixels]]                \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "walignr1 wr0, wr10, wr11       \n\t"
-        "walignr1 wr1, wr11, wr12       \n\t"
-        "wldrd wr10, [%[block]]         \n\t"
-        "wldrd wr11, [%[block], #8]     \n\t"
-        WAVG2B" wr8, wr0, wr4           \n\t"
-        WAVG2B" wr9, wr1, wr5           \n\t"
-        WAVG2B" wr8, wr8, wr10          \n\t"
-        WAVG2B" wr9, wr9, wr11          \n\t"
-        "wstrd wr8, [%[block]]          \n\t"
-        "wstrd wr9, [%[block], #8]      \n\t"
-        "add %[block], %[block], %[line_size]   \n\t"
-
-        "subs %[h], %[h], #2            \n\t"
-        "pld [%[block]]                 \n\t"
-        "bne 1b                         \n\t"
-        : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block), [line_size]"+r"(stride)
-        :
-        : "r4", "r5", "r12", "memory");
-}
-
-void DEF(put, pixels8_xy2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
-{
-    // [wr0 wr1 wr2 wr3] for previous line
-    // [wr4 wr5 wr6 wr7] for current line
-    SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
-    __asm__ volatile(
-        "pld [%[pixels]]                \n\t"
-        "mov r12, #2                    \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "tmcr wcgr0, r12                \n\t" /* for shift value */
-        "and r12, %[pixels], #7         \n\t"
-        "bic %[pixels], %[pixels], #7   \n\t"
-        "tmcr wcgr1, r12                \n\t"
-
-        // [wr0 wr1 wr2 wr3] <= *
-        // [wr4 wr5 wr6 wr7]
-        "wldrd wr12, [%[pixels]]        \n\t"
-        "add r12, r12, #1               \n\t"
-        "wldrd wr13, [%[pixels], #8]    \n\t"
-        "tmcr wcgr2, r12                \n\t"
-        "add %[pixels], %[pixels], %[line_size] \n\t"
-        "cmp r12, #8                    \n\t"
-        "pld [%[pixels]]                \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "walignr1 wr2, wr12, wr13       \n\t"
-        "wmoveq wr10, wr13              \n\t"
-        "walignr2ne wr10, wr12, wr13    \n\t"
-        "wunpckelub wr0, wr2            \n\t"
-        "wunpckehub wr1, wr2            \n\t"
-        "wunpckelub wr8, wr10           \n\t"
-        "wunpckehub wr9, wr10           \n\t"
-        "waddhus wr0, wr0, wr8          \n\t"
-        "waddhus wr1, wr1, wr9          \n\t"
-
-        "1:                             \n\t"
-        // [wr0 wr1 wr2 wr3]
-        // [wr4 wr5 wr6 wr7] <= *
-        "wldrd wr12, [%[pixels]]        \n\t"
-        "cmp r12, #8                    \n\t"
-        "wldrd wr13, [%[pixels], #8]    \n\t"
-        "add %[pixels], %[pixels], %[line_size] \n\t"
-        "walignr1 wr6, wr12, wr13       \n\t"
-        "pld [%[pixels]]                \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "wmoveq wr10, wr13              \n\t"
-        "walignr2ne wr10, wr12, wr13    \n\t"
-        "wunpckelub wr4, wr6            \n\t"
-        "wunpckehub wr5, wr6            \n\t"
-        "wunpckelub wr8, wr10           \n\t"
-        "wunpckehub wr9, wr10           \n\t"
-        "waddhus wr4, wr4, wr8          \n\t"
-        "waddhus wr5, wr5, wr9          \n\t"
-        "waddhus wr8, wr0, wr4          \n\t"
-        "waddhus wr9, wr1, wr5          \n\t"
-        "waddhus wr8, wr8, wr15         \n\t"
-        "waddhus wr9, wr9, wr15         \n\t"
-        "wsrlhg wr8, wr8, wcgr0         \n\t"
-        "wsrlhg wr9, wr9, wcgr0         \n\t"
-        "wpackhus wr8, wr8, wr9         \n\t"
-        "wstrd wr8, [%[block]]          \n\t"
-        "add %[block], %[block], %[line_size]   \n\t"
-
-        // [wr0 wr1 wr2 wr3] <= *
-        // [wr4 wr5 wr6 wr7]
-        "wldrd wr12, [%[pixels]]        \n\t"
-        "wldrd wr13, [%[pixels], #8]    \n\t"
-        "add %[pixels], %[pixels], %[line_size] \n\t"
-        "walignr1 wr2, wr12, wr13       \n\t"
-        "pld [%[pixels]]                \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "wmoveq wr10, wr13              \n\t"
-        "walignr2ne wr10, wr12, wr13    \n\t"
-        "wunpckelub wr0, wr2            \n\t"
-        "wunpckehub wr1, wr2            \n\t"
-        "wunpckelub wr8, wr10           \n\t"
-        "wunpckehub wr9, wr10           \n\t"
-        "waddhus wr0, wr0, wr8          \n\t"
-        "waddhus wr1, wr1, wr9          \n\t"
-        "waddhus wr8, wr0, wr4          \n\t"
-        "waddhus wr9, wr1, wr5          \n\t"
-        "waddhus wr8, wr8, wr15         \n\t"
-        "waddhus wr9, wr9, wr15         \n\t"
-        "wsrlhg wr8, wr8, wcgr0         \n\t"
-        "wsrlhg wr9, wr9, wcgr0         \n\t"
-        "wpackhus wr8, wr8, wr9         \n\t"
-        "subs %[h], %[h], #2            \n\t"
-        "wstrd wr8, [%[block]]          \n\t"
-        "add %[block], %[block], %[line_size]   \n\t"
-        "bne 1b                         \n\t"
-        : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block)
-        : [line_size]"r"(line_size)
-        : "r12", "memory");
-}
-
-void DEF(put, pixels16_xy2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
-{
-    // [wr0 wr1 wr2 wr3] for previous line
-    // [wr4 wr5 wr6 wr7] for current line
-    SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
-    __asm__ volatile(
-        "pld [%[pixels]]                \n\t"
-        "mov r12, #2                    \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "tmcr wcgr0, r12                \n\t" /* for shift value */
-        /* alignment */
-        "and r12, %[pixels], #7         \n\t"
-        "bic %[pixels], %[pixels], #7   \n\t"
-        "tmcr wcgr1, r12                \n\t"
-        "add r12, r12, #1               \n\t"
-        "tmcr wcgr2, r12                \n\t"
-
-        // [wr0 wr1 wr2 wr3] <= *
-        // [wr4 wr5 wr6 wr7]
-        "wldrd wr12, [%[pixels]]        \n\t"
-        "cmp r12, #8                    \n\t"
-        "wldrd wr13, [%[pixels], #8]    \n\t"
-        "wldrd wr14, [%[pixels], #16]   \n\t"
-        "add %[pixels], %[pixels], %[line_size] \n\t"
-        "pld [%[pixels]]                \n\t"
-        "walignr1 wr2, wr12, wr13       \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "walignr1 wr3, wr13, wr14       \n\t"
-        "wmoveq wr10, wr13              \n\t"
-        "wmoveq wr11, wr14              \n\t"
-        "walignr2ne wr10, wr12, wr13    \n\t"
-        "walignr2ne wr11, wr13, wr14    \n\t"
-        "wunpckelub wr0, wr2            \n\t"
-        "wunpckehub wr1, wr2            \n\t"
-        "wunpckelub wr2, wr3            \n\t"
-        "wunpckehub wr3, wr3            \n\t"
-        "wunpckelub wr8, wr10           \n\t"
-        "wunpckehub wr9, wr10           \n\t"
-        "wunpckelub wr10, wr11          \n\t"
-        "wunpckehub wr11, wr11          \n\t"
-        "waddhus wr0, wr0, wr8          \n\t"
-        "waddhus wr1, wr1, wr9          \n\t"
-        "waddhus wr2, wr2, wr10         \n\t"
-        "waddhus wr3, wr3, wr11         \n\t"
-
-        "1:                             \n\t"
-        // [wr0 wr1 wr2 wr3]
-        // [wr4 wr5 wr6 wr7] <= *
-        "wldrd wr12, [%[pixels]]        \n\t"
-        "cmp r12, #8                    \n\t"
-        "wldrd wr13, [%[pixels], #8]    \n\t"
-        "wldrd wr14, [%[pixels], #16]   \n\t"
-        "add %[pixels], %[pixels], %[line_size] \n\t"
-        "walignr1 wr6, wr12, wr13       \n\t"
-        "pld [%[pixels]]                \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "walignr1 wr7, wr13, wr14       \n\t"
-        "wmoveq wr10, wr13              \n\t"
-        "wmoveq wr11, wr14              \n\t"
-        "walignr2ne wr10, wr12, wr13    \n\t"
-        "walignr2ne wr11, wr13, wr14    \n\t"
-        "wunpckelub wr4, wr6            \n\t"
-        "wunpckehub wr5, wr6            \n\t"
-        "wunpckelub wr6, wr7            \n\t"
-        "wunpckehub wr7, wr7            \n\t"
-        "wunpckelub wr8, wr10           \n\t"
-        "wunpckehub wr9, wr10           \n\t"
-        "wunpckelub wr10, wr11          \n\t"
-        "wunpckehub wr11, wr11          \n\t"
-        "waddhus wr4, wr4, wr8          \n\t"
-        "waddhus wr5, wr5, wr9          \n\t"
-        "waddhus wr6, wr6, wr10         \n\t"
-        "waddhus wr7, wr7, wr11         \n\t"
-        "waddhus wr8, wr0, wr4          \n\t"
-        "waddhus wr9, wr1, wr5          \n\t"
-        "waddhus wr10, wr2, wr6         \n\t"
-        "waddhus wr11, wr3, wr7         \n\t"
-        "waddhus wr8, wr8, wr15         \n\t"
-        "waddhus wr9, wr9, wr15         \n\t"
-        "waddhus wr10, wr10, wr15       \n\t"
-        "waddhus wr11, wr11, wr15       \n\t"
-        "wsrlhg wr8, wr8, wcgr0         \n\t"
-        "wsrlhg wr9, wr9, wcgr0         \n\t"
-        "wsrlhg wr10, wr10, wcgr0       \n\t"
-        "wsrlhg wr11, wr11, wcgr0       \n\t"
-        "wpackhus wr8, wr8, wr9         \n\t"
-        "wpackhus wr9, wr10, wr11       \n\t"
-        "wstrd wr8, [%[block]]          \n\t"
-        "wstrd wr9, [%[block], #8]      \n\t"
-        "add %[block], %[block], %[line_size]   \n\t"
-
-        // [wr0 wr1 wr2 wr3] <= *
-        // [wr4 wr5 wr6 wr7]
-        "wldrd wr12, [%[pixels]]        \n\t"
-        "wldrd wr13, [%[pixels], #8]    \n\t"
-        "wldrd wr14, [%[pixels], #16]   \n\t"
-        "add %[pixels], %[pixels], %[line_size] \n\t"
-        "walignr1 wr2, wr12, wr13       \n\t"
-        "pld [%[pixels]]                \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "walignr1 wr3, wr13, wr14       \n\t"
-        "wmoveq wr10, wr13              \n\t"
-        "wmoveq wr11, wr14              \n\t"
-        "walignr2ne wr10, wr12, wr13    \n\t"
-        "walignr2ne wr11, wr13, wr14    \n\t"
-        "wunpckelub wr0, wr2            \n\t"
-        "wunpckehub wr1, wr2            \n\t"
-        "wunpckelub wr2, wr3            \n\t"
-        "wunpckehub wr3, wr3            \n\t"
-        "wunpckelub wr8, wr10           \n\t"
-        "wunpckehub wr9, wr10           \n\t"
-        "wunpckelub wr10, wr11          \n\t"
-        "wunpckehub wr11, wr11          \n\t"
-        "waddhus wr0, wr0, wr8          \n\t"
-        "waddhus wr1, wr1, wr9          \n\t"
-        "waddhus wr2, wr2, wr10         \n\t"
-        "waddhus wr3, wr3, wr11         \n\t"
-        "waddhus wr8, wr0, wr4          \n\t"
-        "waddhus wr9, wr1, wr5          \n\t"
-        "waddhus wr10, wr2, wr6         \n\t"
-        "waddhus wr11, wr3, wr7         \n\t"
-        "waddhus wr8, wr8, wr15         \n\t"
-        "waddhus wr9, wr9, wr15         \n\t"
-        "waddhus wr10, wr10, wr15       \n\t"
-        "waddhus wr11, wr11, wr15       \n\t"
-        "wsrlhg wr8, wr8, wcgr0         \n\t"
-        "wsrlhg wr9, wr9, wcgr0         \n\t"
-        "wsrlhg wr10, wr10, wcgr0       \n\t"
-        "wsrlhg wr11, wr11, wcgr0       \n\t"
-        "wpackhus wr8, wr8, wr9         \n\t"
-        "wpackhus wr9, wr10, wr11       \n\t"
-        "wstrd wr8, [%[block]]          \n\t"
-        "wstrd wr9, [%[block], #8]      \n\t"
-        "add %[block], %[block], %[line_size]   \n\t"
-
-        "subs %[h], %[h], #2            \n\t"
-        "bne 1b                         \n\t"
-        : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block)
-        : [line_size]"r"(line_size)
-        : "r12", "memory");
-}
-
-void DEF(avg, pixels8_xy2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
-{
-    // [wr0 wr1 wr2 wr3] for previous line
-    // [wr4 wr5 wr6 wr7] for current line
-    SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
-    __asm__ volatile(
-        "pld [%[block]]                 \n\t"
-        "pld [%[block], #32]            \n\t"
-        "pld [%[pixels]]                \n\t"
-        "mov r12, #2                    \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "tmcr wcgr0, r12                \n\t" /* for shift value */
-        "and r12, %[pixels], #7         \n\t"
-        "bic %[pixels], %[pixels], #7   \n\t"
-        "tmcr wcgr1, r12                \n\t"
-
-        // [wr0 wr1 wr2 wr3] <= *
-        // [wr4 wr5 wr6 wr7]
-        "wldrd wr12, [%[pixels]]        \n\t"
-        "add r12, r12, #1               \n\t"
-        "wldrd wr13, [%[pixels], #8]    \n\t"
-        "tmcr wcgr2, r12                \n\t"
-        "add %[pixels], %[pixels], %[line_size] \n\t"
-        "cmp r12, #8                    \n\t"
-        "pld [%[pixels]]                \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "walignr1 wr2, wr12, wr13       \n\t"
-        "wmoveq wr10, wr13              \n\t"
-        "walignr2ne wr10, wr12, wr13    \n\t"
-        "wunpckelub wr0, wr2            \n\t"
-        "wunpckehub wr1, wr2            \n\t"
-        "wunpckelub wr8, wr10           \n\t"
-        "wunpckehub wr9, wr10           \n\t"
-        "waddhus wr0, wr0, wr8          \n\t"
-        "waddhus wr1, wr1, wr9          \n\t"
-
-        "1:                             \n\t"
-        // [wr0 wr1 wr2 wr3]
-        // [wr4 wr5 wr6 wr7] <= *
-        "wldrd wr12, [%[pixels]]        \n\t"
-        "cmp r12, #8                    \n\t"
-        "wldrd wr13, [%[pixels], #8]    \n\t"
-        "add %[pixels], %[pixels], %[line_size] \n\t"
-        "walignr1 wr6, wr12, wr13       \n\t"
-        "pld [%[pixels]]                \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "wmoveq wr10, wr13              \n\t"
-        "walignr2ne wr10, wr12, wr13    \n\t"
-        "wunpckelub wr4, wr6            \n\t"
-        "wunpckehub wr5, wr6            \n\t"
-        "wunpckelub wr8, wr10           \n\t"
-        "wunpckehub wr9, wr10           \n\t"
-        "waddhus wr4, wr4, wr8          \n\t"
-        "waddhus wr5, wr5, wr9          \n\t"
-        "waddhus wr8, wr0, wr4          \n\t"
-        "waddhus wr9, wr1, wr5          \n\t"
-        "waddhus wr8, wr8, wr15         \n\t"
-        "waddhus wr9, wr9, wr15         \n\t"
-        "wldrd wr12, [%[block]]         \n\t"
-        "wsrlhg wr8, wr8, wcgr0         \n\t"
-        "wsrlhg wr9, wr9, wcgr0         \n\t"
-        "wpackhus wr8, wr8, wr9         \n\t"
-        WAVG2B" wr8, wr8, wr12          \n\t"
-        "wstrd wr8, [%[block]]          \n\t"
-        "add %[block], %[block], %[line_size]   \n\t"
-        "wldrd wr12, [%[pixels]]        \n\t"
-        "pld [%[block]]                 \n\t"
-        "pld [%[block], #32]            \n\t"
-
-        // [wr0 wr1 wr2 wr3] <= *
-        // [wr4 wr5 wr6 wr7]
-        "wldrd wr13, [%[pixels], #8]    \n\t"
-        "add %[pixels], %[pixels], %[line_size] \n\t"
-        "walignr1 wr2, wr12, wr13       \n\t"
-        "pld [%[pixels]]                \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "wmoveq wr10, wr13              \n\t"
-        "walignr2ne wr10, wr12, wr13    \n\t"
-        "wunpckelub wr0, wr2            \n\t"
-        "wunpckehub wr1, wr2            \n\t"
-        "wunpckelub wr8, wr10           \n\t"
-        "wunpckehub wr9, wr10           \n\t"
-        "waddhus wr0, wr0, wr8          \n\t"
-        "waddhus wr1, wr1, wr9          \n\t"
-        "waddhus wr8, wr0, wr4          \n\t"
-        "waddhus wr9, wr1, wr5          \n\t"
-        "waddhus wr8, wr8, wr15         \n\t"
-        "waddhus wr9, wr9, wr15         \n\t"
-        "wldrd wr12, [%[block]]         \n\t"
-        "wsrlhg wr8, wr8, wcgr0         \n\t"
-        "wsrlhg wr9, wr9, wcgr0         \n\t"
-        "wpackhus wr8, wr8, wr9         \n\t"
-        "subs %[h], %[h], #2            \n\t"
-        WAVG2B" wr8, wr8, wr12          \n\t"
-        "wstrd wr8, [%[block]]          \n\t"
-        "add %[block], %[block], %[line_size]   \n\t"
-        "pld [%[block]]                 \n\t"
-        "pld [%[block], #32]            \n\t"
-        "bne 1b                         \n\t"
-        : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block)
-        : [line_size]"r"(line_size)
-        : "r12", "memory");
-}
-
-void DEF(avg, pixels16_xy2)(uint8_t *block, const uint8_t *pixels, const int line_size, int h)
-{
-    // [wr0 wr1 wr2 wr3] for previous line
-    // [wr4 wr5 wr6 wr7] for current line
-    SET_RND(wr15); // =2 for rnd  and  =1 for no_rnd version
-    __asm__ volatile(
-        "pld [%[block]]                 \n\t"
-        "pld [%[block], #32]            \n\t"
-        "pld [%[pixels]]                \n\t"
-        "mov r12, #2                    \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "tmcr wcgr0, r12                \n\t" /* for shift value */
-        /* alignment */
-        "and r12, %[pixels], #7         \n\t"
-        "bic %[pixels], %[pixels], #7           \n\t"
-        "tmcr wcgr1, r12                \n\t"
-        "add r12, r12, #1               \n\t"
-        "tmcr wcgr2, r12                \n\t"
-
-        // [wr0 wr1 wr2 wr3] <= *
-        // [wr4 wr5 wr6 wr7]
-        "wldrd wr12, [%[pixels]]        \n\t"
-        "cmp r12, #8                    \n\t"
-        "wldrd wr13, [%[pixels], #8]    \n\t"
-        "wldrd wr14, [%[pixels], #16]   \n\t"
-        "add %[pixels], %[pixels], %[line_size] \n\t"
-        "pld [%[pixels]]                \n\t"
-        "walignr1 wr2, wr12, wr13       \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "walignr1 wr3, wr13, wr14       \n\t"
-        "wmoveq wr10, wr13              \n\t"
-        "wmoveq wr11, wr14              \n\t"
-        "walignr2ne wr10, wr12, wr13    \n\t"
-        "walignr2ne wr11, wr13, wr14    \n\t"
-        "wunpckelub wr0, wr2            \n\t"
-        "wunpckehub wr1, wr2            \n\t"
-        "wunpckelub wr2, wr3            \n\t"
-        "wunpckehub wr3, wr3            \n\t"
-        "wunpckelub wr8, wr10           \n\t"
-        "wunpckehub wr9, wr10           \n\t"
-        "wunpckelub wr10, wr11          \n\t"
-        "wunpckehub wr11, wr11          \n\t"
-        "waddhus wr0, wr0, wr8          \n\t"
-        "waddhus wr1, wr1, wr9          \n\t"
-        "waddhus wr2, wr2, wr10         \n\t"
-        "waddhus wr3, wr3, wr11         \n\t"
-
-        "1:                             \n\t"
-        // [wr0 wr1 wr2 wr3]
-        // [wr4 wr5 wr6 wr7] <= *
-        "wldrd wr12, [%[pixels]]        \n\t"
-        "cmp r12, #8                    \n\t"
-        "wldrd wr13, [%[pixels], #8]    \n\t"
-        "wldrd wr14, [%[pixels], #16]   \n\t"
-        "add %[pixels], %[pixels], %[line_size] \n\t"
-        "walignr1 wr6, wr12, wr13       \n\t"
-        "pld [%[pixels]]                \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "walignr1 wr7, wr13, wr14       \n\t"
-        "wmoveq wr10, wr13              \n\t"
-        "wmoveq wr11, wr14              \n\t"
-        "walignr2ne wr10, wr12, wr13    \n\t"
-        "walignr2ne wr11, wr13, wr14    \n\t"
-        "wunpckelub wr4, wr6            \n\t"
-        "wunpckehub wr5, wr6            \n\t"
-        "wunpckelub wr6, wr7            \n\t"
-        "wunpckehub wr7, wr7            \n\t"
-        "wunpckelub wr8, wr10           \n\t"
-        "wunpckehub wr9, wr10           \n\t"
-        "wunpckelub wr10, wr11          \n\t"
-        "wunpckehub wr11, wr11          \n\t"
-        "waddhus wr4, wr4, wr8          \n\t"
-        "waddhus wr5, wr5, wr9          \n\t"
-        "waddhus wr6, wr6, wr10         \n\t"
-        "waddhus wr7, wr7, wr11         \n\t"
-        "waddhus wr8, wr0, wr4          \n\t"
-        "waddhus wr9, wr1, wr5          \n\t"
-        "waddhus wr10, wr2, wr6         \n\t"
-        "waddhus wr11, wr3, wr7         \n\t"
-        "waddhus wr8, wr8, wr15         \n\t"
-        "waddhus wr9, wr9, wr15         \n\t"
-        "waddhus wr10, wr10, wr15       \n\t"
-        "waddhus wr11, wr11, wr15       \n\t"
-        "wsrlhg wr8, wr8, wcgr0         \n\t"
-        "wsrlhg wr9, wr9, wcgr0         \n\t"
-        "wldrd wr12, [%[block]]         \n\t"
-        "wldrd wr13, [%[block], #8]     \n\t"
-        "wsrlhg wr10, wr10, wcgr0       \n\t"
-        "wsrlhg wr11, wr11, wcgr0       \n\t"
-        "wpackhus wr8, wr8, wr9         \n\t"
-        "wpackhus wr9, wr10, wr11       \n\t"
-        WAVG2B" wr8, wr8, wr12          \n\t"
-        WAVG2B" wr9, wr9, wr13          \n\t"
-        "wstrd wr8, [%[block]]          \n\t"
-        "wstrd wr9, [%[block], #8]      \n\t"
-        "add %[block], %[block], %[line_size]   \n\t"
-
-        // [wr0 wr1 wr2 wr3] <= *
-        // [wr4 wr5 wr6 wr7]
-        "wldrd wr12, [%[pixels]]        \n\t"
-        "pld [%[block]]                 \n\t"
-        "wldrd wr13, [%[pixels], #8]    \n\t"
-        "pld [%[block], #32]            \n\t"
-        "wldrd wr14, [%[pixels], #16]   \n\t"
-        "add %[pixels], %[pixels], %[line_size] \n\t"
-        "walignr1 wr2, wr12, wr13       \n\t"
-        "pld [%[pixels]]                \n\t"
-        "pld [%[pixels], #32]           \n\t"
-        "walignr1 wr3, wr13, wr14       \n\t"
-        "wmoveq wr10, wr13              \n\t"
-        "wmoveq wr11, wr14              \n\t"
-        "walignr2ne wr10, wr12, wr13    \n\t"
-        "walignr2ne wr11, wr13, wr14    \n\t"
-        "wunpckelub wr0, wr2            \n\t"
-        "wunpckehub wr1, wr2            \n\t"
-        "wunpckelub wr2, wr3            \n\t"
-        "wunpckehub wr3, wr3            \n\t"
-        "wunpckelub wr8, wr10           \n\t"
-        "wunpckehub wr9, wr10           \n\t"
-        "wunpckelub wr10, wr11          \n\t"
-        "wunpckehub wr11, wr11          \n\t"
-        "waddhus wr0, wr0, wr8          \n\t"
-        "waddhus wr1, wr1, wr9          \n\t"
-        "waddhus wr2, wr2, wr10         \n\t"
-        "waddhus wr3, wr3, wr11         \n\t"
-        "waddhus wr8, wr0, wr4          \n\t"
-        "waddhus wr9, wr1, wr5          \n\t"
-        "waddhus wr10, wr2, wr6         \n\t"
-        "waddhus wr11, wr3, wr7         \n\t"
-        "waddhus wr8, wr8, wr15         \n\t"
-        "waddhus wr9, wr9, wr15         \n\t"
-        "waddhus wr10, wr10, wr15       \n\t"
-        "waddhus wr11, wr11, wr15       \n\t"
-        "wsrlhg wr8, wr8, wcgr0         \n\t"
-        "wsrlhg wr9, wr9, wcgr0         \n\t"
-        "wldrd wr12, [%[block]]         \n\t"
-        "wldrd wr13, [%[block], #8]     \n\t"
-        "wsrlhg wr10, wr10, wcgr0       \n\t"
-        "wsrlhg wr11, wr11, wcgr0       \n\t"
-        "wpackhus wr8, wr8, wr9         \n\t"
-        "wpackhus wr9, wr10, wr11       \n\t"
-        WAVG2B" wr8, wr8, wr12          \n\t"
-        WAVG2B" wr9, wr9, wr13          \n\t"
-        "wstrd wr8, [%[block]]          \n\t"
-        "wstrd wr9, [%[block], #8]      \n\t"
-        "add %[block], %[block], %[line_size]   \n\t"
-        "subs %[h], %[h], #2            \n\t"
-        "pld [%[block]]                 \n\t"
-        "pld [%[block], #32]            \n\t"
-        "bne 1b                         \n\t"
-        : [h]"+r"(h), [pixels]"+r"(pixels), [block]"+r"(block)
-        : [line_size]"r"(line_size)
-        : "r12", "memory");
-}
diff --git a/libavcodec/arm/mpegvideo_arm.c b/libavcodec/arm/mpegvideo_arm.c
index 0a66f2a981..26eed89930 100644
--- a/libavcodec/arm/mpegvideo_arm.c
+++ b/libavcodec/arm/mpegvideo_arm.c
@@ -40,16 +40,9 @@ void ff_dct_unquantize_h263_intra_neon(MpegEncContext *s, DCTELEM *block,
 
 void ff_MPV_common_init_arm(MpegEncContext *s)
 {
-    /* IWMMXT support is a superset of armv5te, so
-     * allow optimized functions for armv5te unless
-     * a better iwmmxt function exists
-     */
 #if HAVE_ARMV5TE
     ff_MPV_common_init_armv5te(s);
 #endif
-#if HAVE_IWMMXT
-    ff_MPV_common_init_iwmmxt(s);
-#endif
 
     if (HAVE_NEON) {
         s->dct_unquantize_h263_intra = ff_dct_unquantize_h263_intra_neon;
diff --git a/libavcodec/arm/mpegvideo_arm.h b/libavcodec/arm/mpegvideo_arm.h
index 57a228881c..d69fa2f541 100644
--- a/libavcodec/arm/mpegvideo_arm.h
+++ b/libavcodec/arm/mpegvideo_arm.h
@@ -21,7 +21,6 @@
 
 #include "libavcodec/mpegvideo.h"
 
-void ff_MPV_common_init_iwmmxt(MpegEncContext *s);
 void ff_MPV_common_init_armv5te(MpegEncContext *s);
 
 #endif /* AVCODEC_ARM_MPEGVIDEO_H */
diff --git a/libavcodec/arm/mpegvideo_iwmmxt.c b/libavcodec/arm/mpegvideo_iwmmxt.c
deleted file mode 100644
index a10ca17114..0000000000
--- a/libavcodec/arm/mpegvideo_iwmmxt.c
+++ /dev/null
@@ -1,101 +0,0 @@
-/*
- * copyright (c) 2004 AGAWA Koji
- *
- * This file is part of Libav.
- *
- * Libav is free software; you can redistribute it and/or
- * modify it under the terms of the GNU Lesser General Public
- * License as published by the Free Software Foundation; either
- * version 2.1 of the License, or (at your option) any later version.
- *
- * Libav is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public
- * License along with Libav; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
- */
-
-#include "libavutil/cpu.h"
-#include "libavcodec/avcodec.h"
-#include "libavcodec/dsputil.h"
-#include "libavcodec/mpegvideo.h"
-#include "mpegvideo_arm.h"
-
-static void dct_unquantize_h263_intra_iwmmxt(MpegEncContext *s,
-                                             DCTELEM *block, int n, int qscale)
-{
-    int level, qmul, qadd;
-    int nCoeffs;
-    DCTELEM *block_orig = block;
-
-    assert(s->block_last_index[n]>=0);
-
-    qmul = qscale << 1;
-
-    if (!s->h263_aic) {
-        if (n < 4)
-            level = block[0] * s->y_dc_scale;
-        else
-            level = block[0] * s->c_dc_scale;
-        qadd = (qscale - 1) | 1;
-    }else{
-        qadd = 0;
-        level = block[0];
-    }
-    if(s->ac_pred)
-        nCoeffs=63;
-    else
-        nCoeffs= s->inter_scantable.raster_end[ s->block_last_index[n] ];
-
-    __asm__ volatile (
-/*      "movd %1, %%mm6                 \n\t" //qmul */
-/*      "packssdw %%mm6, %%mm6          \n\t" */
-/*      "packssdw %%mm6, %%mm6          \n\t" */
-        "tbcsth wr6, %[qmul]            \n\t"
-/*      "movd %2, %%mm5                 \n\t" //qadd */
-/*      "packssdw %%mm5, %%mm5          \n\t" */
-/*      "packssdw %%mm5, %%mm5          \n\t" */
-        "tbcsth wr5, %[qadd]            \n\t"
-        "wzero wr7                      \n\t" /* "pxor %%mm7, %%mm7             \n\t" */
-        "wzero wr4                      \n\t" /* "pxor %%mm4, %%mm4             \n\t" */
-        "wsubh wr7, wr5, wr7            \n\t" /* "psubw %%mm5, %%mm7            \n\t" */
-        "1:                             \n\t"
-        "wldrd wr2, [%[block]]          \n\t" /* "movq (%0, %3), %%mm0          \n\t" */
-        "wldrd wr3, [%[block], #8]      \n\t" /* "movq 8(%0, %3), %%mm1         \n\t" */
-        "wmulsl wr0, wr6, wr2           \n\t" /* "pmullw %%mm6, %%mm0           \n\t" */
-        "wmulsl wr1, wr6, wr3           \n\t" /* "pmullw %%mm6, %%mm1           \n\t" */
-/*      "movq (%0, %3), %%mm2           \n\t" */
-/*      "movq 8(%0, %3), %%mm3          \n\t" */
-        "wcmpgtsh wr2, wr4, wr2         \n\t" /* "pcmpgtw %%mm4, %%mm2          \n\t" // block[i] < 0 ? -1 : 0 */
-        "wcmpgtsh wr3, wr4, wr2         \n\t" /* "pcmpgtw %%mm4, %%mm3          \n\t" // block[i] < 0 ? -1 : 0 */
-        "wxor wr0, wr2, wr0             \n\t" /* "pxor %%mm2, %%mm0             \n\t" */
-        "wxor wr1, wr3, wr1             \n\t" /* "pxor %%mm3, %%mm1             \n\t" */
-        "waddh wr0, wr7, wr0            \n\t" /* "paddw %%mm7, %%mm0            \n\t" */
-        "waddh wr1, wr7, wr1            \n\t" /* "paddw %%mm7, %%mm1            \n\t" */
-        "wxor wr2, wr0, wr2             \n\t" /* "pxor %%mm0, %%mm2             \n\t" */
-        "wxor wr3, wr1, wr3             \n\t" /* "pxor %%mm1, %%mm3             \n\t" */
-        "wcmpeqh wr0, wr7, wr0          \n\t" /* "pcmpeqw %%mm7, %%mm0          \n\t" // block[i] == 0 ? -1 : 0 */
-        "wcmpeqh wr1, wr7, wr1          \n\t" /* "pcmpeqw %%mm7, %%mm1          \n\t" // block[i] == 0 ? -1 : 0 */
-        "wandn wr0, wr2, wr0            \n\t" /* "pandn %%mm2, %%mm0            \n\t" */
-        "wandn wr1, wr3, wr1            \n\t" /* "pandn %%mm3, %%mm1            \n\t" */
-        "wstrd wr0, [%[block]]          \n\t" /* "movq %%mm0, (%0, %3)          \n\t" */
-        "wstrd wr1, [%[block], #8]      \n\t" /* "movq %%mm1, 8(%0, %3)         \n\t" */
-        "add %[block], %[block], #16    \n\t" /* "addl $16, %3                  \n\t" */
-        "subs %[i], %[i], #1            \n\t"
-        "bne 1b                         \n\t" /* "jng 1b                                \n\t" */
-        :[block]"+r"(block)
-        :[i]"r"((nCoeffs + 8) / 8), [qmul]"r"(qmul), [qadd]"r"(qadd)
-        :"memory");
-
-    block_orig[0] = level;
-}
-
-void ff_MPV_common_init_iwmmxt(MpegEncContext *s)
-{
-    if (!(mm_flags & AV_CPU_FLAG_IWMMXT)) return;
-
-    s->dct_unquantize_h263_intra = dct_unquantize_h263_intra_iwmmxt;
-}
diff --git a/libavutil/Makefile b/libavutil/Makefile
index 221d9ce9bd..f7c8a61bf8 100644
--- a/libavutil/Makefile
+++ b/libavutil/Makefile
@@ -72,7 +72,6 @@ OBJS = adler32.o                                                        \
        tree.o                                                           \
        utils.o                                                          \
 
-OBJS-$(ARCH_ARM) += arm/cpu.o
 OBJS-$(ARCH_PPC) += ppc/cpu.o
 OBJS-$(ARCH_X86) += x86/cpu.o
 
diff --git a/libavutil/arm/cpu.c b/libavutil/arm/cpu.c
deleted file mode 100644
index 835513ed84..0000000000
--- a/libavutil/arm/cpu.c
+++ /dev/null
@@ -1,25 +0,0 @@
-/*
- * This file is part of Libav.
- *
- * Libav is free software; you can redistribute it and/or
- * modify it under the terms of the GNU Lesser General Public
- * License as published by the Free Software Foundation; either
- * version 2.1 of the License, or (at your option) any later version.
- *
- * Libav is distributed in the hope that it will be useful,
- * but WITHOUT ANY WARRANTY; without even the implied warranty of
- * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the GNU
- * Lesser General Public License for more details.
- *
- * You should have received a copy of the GNU Lesser General Public
- * License along with Libav; if not, write to the Free Software
- * Foundation, Inc., 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301 USA
- */
-
-#include "libavutil/cpu.h"
-#include "config.h"
-
-int ff_get_cpu_flags_arm(void)
-{
-    return HAVE_IWMMXT * AV_CPU_FLAG_IWMMXT;
-}
diff --git a/libavutil/cpu.c b/libavutil/cpu.c
index c44075be29..6fc13adbca 100644
--- a/libavutil/cpu.c
+++ b/libavutil/cpu.c
@@ -28,7 +28,6 @@ int av_get_cpu_flags(void)
     if (checked)
         return flags;
 
-    if (ARCH_ARM) flags = ff_get_cpu_flags_arm();
     if (ARCH_PPC) flags = ff_get_cpu_flags_ppc();
     if (ARCH_X86) flags = ff_get_cpu_flags_x86();
 
@@ -53,9 +52,7 @@ static const struct {
     int flag;
     const char *name;
 } cpu_flag_tab[] = {
-#if   ARCH_ARM
-    { AV_CPU_FLAG_IWMMXT,    "iwmmxt"     },
-#elif ARCH_PPC
+#if   ARCH_PPC
     { AV_CPU_FLAG_ALTIVEC,   "altivec"    },
 #elif ARCH_X86
     { AV_CPU_FLAG_MMX,       "mmx"        },
diff --git a/libavutil/cpu.h b/libavutil/cpu.h
index 361fe9866a..55ad7d17a9 100644
--- a/libavutil/cpu.h
+++ b/libavutil/cpu.h
@@ -40,7 +40,6 @@
 #define AV_CPU_FLAG_AVX          0x4000 ///< AVX functions: requires OS support even if YMM registers aren't used
 #define AV_CPU_FLAG_XOP          0x0400 ///< Bulldozer XOP functions
 #define AV_CPU_FLAG_FMA4         0x0800 ///< Bulldozer FMA4 functions
-#define AV_CPU_FLAG_IWMMXT       0x0100 ///< XScale IWMMXT
 #define AV_CPU_FLAG_ALTIVEC      0x0001 ///< standard
 
 /**
@@ -57,7 +56,6 @@ int av_get_cpu_flags(void);
 void av_set_cpu_flags_mask(int mask);
 
 /* The following CPU-specific functions shall not be called directly. */
-int ff_get_cpu_flags_arm(void);
 int ff_get_cpu_flags_ppc(void);
 int ff_get_cpu_flags_x86(void);