VP8 for ARMv8 by using NEON intrinsics 02

Add copymem_neon.c
- vp8_copy_mem16x16_neon
- vp8_copy_mem8x8_neon
- vp8_copy_mem8x4_neon

vpxdec  --summary --noblit ../videos/tears_of_steel_1080p.webm
Before => After, 13.25 => 13.25 (fps)

Change-Id: Ib956b5a20522ff57dc8a580bf0aef7b252bddba6
Signed-off-by: James Yu <james.yu@linaro.org>

diff --git a/vp8/common/arm/neon/copymem16x16_neon.asm b/vp8/common/arm/neon/copymem16x16_neon.asm
deleted file mode 100644 (file)
index bda4b96..0000000
--- a/vp8/common/arm/neon/copymem16x16_neon.asm
+++ /dev/null
@@ -1,59 +0,0 @@
-;
-;  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
-;
-;  Use of this source code is governed by a BSD-style license
-;  that can be found in the LICENSE file in the root of the source
-;  tree. An additional intellectual property rights grant can be found
-;  in the file PATENTS.  All contributing project authors may
-;  be found in the AUTHORS file in the root of the source tree.
-;
-
-
-    EXPORT  |vp8_copy_mem16x16_neon|
-    ; ARM
-    ; REQUIRE8
-    ; PRESERVE8
-
-    AREA    Block, CODE, READONLY ; name this block of code
-;void copy_mem16x16_neon( unsigned char *src, int src_stride, unsigned char *dst, int dst_stride)
-;-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
-|vp8_copy_mem16x16_neon| PROC
-
-    vld1.u8     {q0}, [r0], r1
-    vld1.u8     {q1}, [r0], r1
-    vld1.u8     {q2}, [r0], r1
-    vst1.u8     {q0}, [r2], r3
-    vld1.u8     {q3}, [r0], r1
-    vst1.u8     {q1}, [r2], r3
-    vld1.u8     {q4}, [r0], r1
-    vst1.u8     {q2}, [r2], r3
-    vld1.u8     {q5}, [r0], r1
-    vst1.u8     {q3}, [r2], r3
-    vld1.u8     {q6}, [r0], r1
-    vst1.u8     {q4}, [r2], r3
-    vld1.u8     {q7}, [r0], r1
-    vst1.u8     {q5}, [r2], r3
-    vld1.u8     {q8}, [r0], r1
-    vst1.u8     {q6}, [r2], r3
-    vld1.u8     {q9}, [r0], r1
-    vst1.u8     {q7}, [r2], r3
-    vld1.u8     {q10}, [r0], r1
-    vst1.u8     {q8}, [r2], r3
-    vld1.u8     {q11}, [r0], r1
-    vst1.u8     {q9}, [r2], r3
-    vld1.u8     {q12}, [r0], r1
-    vst1.u8     {q10}, [r2], r3
-    vld1.u8     {q13}, [r0], r1
-    vst1.u8     {q11}, [r2], r3
-    vld1.u8     {q14}, [r0], r1
-    vst1.u8     {q12}, [r2], r3
-    vld1.u8     {q15}, [r0], r1
-    vst1.u8     {q13}, [r2], r3
-    vst1.u8     {q14}, [r2], r3
-    vst1.u8     {q15}, [r2], r3
-
-    mov     pc, lr
-
-    ENDP  ; |vp8_copy_mem16x16_neon|
-
-    END

diff --git a/vp8/common/arm/neon/copymem8x4_neon.asm b/vp8/common/arm/neon/copymem8x4_neon.asm
deleted file mode 100644 (file)
index 35c0f67..0000000
--- a/vp8/common/arm/neon/copymem8x4_neon.asm
+++ /dev/null
@@ -1,34 +0,0 @@
-;
-;  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
-;
-;  Use of this source code is governed by a BSD-style license
-;  that can be found in the LICENSE file in the root of the source
-;  tree. An additional intellectual property rights grant can be found
-;  in the file PATENTS.  All contributing project authors may
-;  be found in the AUTHORS file in the root of the source tree.
-;
-
-
-    EXPORT  |vp8_copy_mem8x4_neon|
-    ; ARM
-    ; REQUIRE8
-    ; PRESERVE8
-
-    AREA    Block, CODE, READONLY ; name this block of code
-;void copy_mem8x4_neon( unsigned char *src, int src_stride, unsigned char *dst, int dst_stride)
-;-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
-|vp8_copy_mem8x4_neon| PROC
-    vld1.u8     {d0}, [r0], r1
-    vld1.u8     {d1}, [r0], r1
-    vst1.u8     {d0}, [r2], r3
-    vld1.u8     {d2}, [r0], r1
-    vst1.u8     {d1}, [r2], r3
-    vld1.u8     {d3}, [r0], r1
-    vst1.u8     {d2}, [r2], r3
-    vst1.u8     {d3}, [r2], r3
-
-    mov     pc, lr
-
-    ENDP  ; |vp8_copy_mem8x4_neon|
-
-    END

diff --git a/vp8/common/arm/neon/copymem8x8_neon.asm b/vp8/common/arm/neon/copymem8x8_neon.asm
deleted file mode 100644 (file)
index 1f5b941..0000000
--- a/vp8/common/arm/neon/copymem8x8_neon.asm
+++ /dev/null
@@ -1,43 +0,0 @@
-;
-;  Copyright (c) 2010 The WebM project authors. All Rights Reserved.
-;
-;  Use of this source code is governed by a BSD-style license
-;  that can be found in the LICENSE file in the root of the source
-;  tree. An additional intellectual property rights grant can be found
-;  in the file PATENTS.  All contributing project authors may
-;  be found in the AUTHORS file in the root of the source tree.
-;
-
-
-    EXPORT  |vp8_copy_mem8x8_neon|
-    ; ARM
-    ; REQUIRE8
-    ; PRESERVE8
-
-    AREA    Block, CODE, READONLY ; name this block of code
-;void copy_mem8x8_neon( unsigned char *src, int src_stride, unsigned char *dst, int dst_stride)
-;-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=-=
-|vp8_copy_mem8x8_neon| PROC
-
-    vld1.u8     {d0}, [r0], r1
-    vld1.u8     {d1}, [r0], r1
-    vst1.u8     {d0}, [r2], r3
-    vld1.u8     {d2}, [r0], r1
-    vst1.u8     {d1}, [r2], r3
-    vld1.u8     {d3}, [r0], r1
-    vst1.u8     {d2}, [r2], r3
-    vld1.u8     {d4}, [r0], r1
-    vst1.u8     {d3}, [r2], r3
-    vld1.u8     {d5}, [r0], r1
-    vst1.u8     {d4}, [r2], r3
-    vld1.u8     {d6}, [r0], r1
-    vst1.u8     {d5}, [r2], r3
-    vld1.u8     {d7}, [r0], r1
-    vst1.u8     {d6}, [r2], r3
-    vst1.u8     {d7}, [r2], r3
-
-    mov     pc, lr
-
-    ENDP  ; |vp8_copy_mem8x8_neon|
-
-    END

diff --git a/vp8/common/arm/neon/copymem_neon.c b/vp8/common/arm/neon/copymem_neon.c
new file mode 100644 (file)
index 0000000..deced11
--- /dev/null
+++ b/vp8/common/arm/neon/copymem_neon.c
@@ -0,0 +1,59 @@
+/*
+ *  Copyright (c) 2014 The WebM project authors. All Rights Reserved.
+ *
+ *  Use of this source code is governed by a BSD-style license
+ *  that can be found in the LICENSE file in the root of the source
+ *  tree. An additional intellectual property rights grant can be found
+ *  in the file PATENTS.  All contributing project authors may
+ *  be found in the AUTHORS file in the root of the source tree.
+ */
+
+#include <arm_neon.h>
+
+void vp8_copy_mem8x4_neon(
+        unsigned char *src,
+        int src_stride,
+        unsigned char *dst,
+        int dst_stride) {
+    uint8x8_t vtmp;
+    int r;
+
+    for (r = 0; r < 4; r++) {
+        vtmp = vld1_u8(src);
+        vst1_u8(dst, vtmp);
+        src += src_stride;
+        dst += dst_stride;
+    }
+}
+
+void vp8_copy_mem8x8_neon(
+        unsigned char *src,
+        int src_stride,
+        unsigned char *dst,
+        int dst_stride) {
+    uint8x8_t vtmp;
+    int r;
+
+    for (r = 0; r < 8; r++) {
+        vtmp = vld1_u8(src);
+        vst1_u8(dst, vtmp);
+        src += src_stride;
+        dst += dst_stride;
+    }
+}
+
+void vp8_copy_mem16x16_neon(
+        unsigned char *src,
+        int src_stride,
+        unsigned char *dst,
+        int dst_stride) {
+    int r;
+    uint8x16_t qtmp;
+
+    for (r = 0; r < 16; r++) {
+        qtmp = vld1q_u8(src);
+        vst1q_u8(dst, qtmp);
+        src += src_stride;
+        dst += dst_stride;
+    }
+}

diff --git a/vp8/vp8_common.mk b/vp8/vp8_common.mk
index a2127c9b7399f9e9334eab728421a073520caec1..d1eb44540460101674685379374d0ab8a1ac8277 100644 (file)
--- a/vp8/vp8_common.mk
+++ b/vp8/vp8_common.mk
@@ -159,9 +159,6 @@ VP8_COMMON_SRCS-$(HAVE_MEDIA)  += common/arm/armv6/vp8_variance_halfpixvar16x16_
 VP8_COMMON_SRCS-$(HAVE_MEDIA)  += common/arm/armv6/vp8_variance_halfpixvar16x16_hv_armv6$(ASM)
 
 # common (neon)
-VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/copymem8x4_neon$(ASM)
-VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/copymem8x8_neon$(ASM)
-VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/copymem16x16_neon$(ASM)
 VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/dc_only_idct_add_neon$(ASM)
 VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/iwalsh_neon$(ASM)
 VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/loopfilter_neon$(ASM)
@@ -189,6 +186,7 @@ VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/vp8_subpixelvariance16x16s_neon
 
 # common (neon intrinsics)
 VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/bilinearpredict_neon.c
+VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/copymem_neon.c
 
 
 $(eval $(call rtcd_h_template,vp8_rtcd,vp8/common/rtcd_defs.sh))