VP8 for ARMv8 by using NEON intrinsics 05

Add dequantizeb_neon.c
- vp8_dequantize_b_loop_neon

vpxdec  --summary --noblit ../videos/tears_of_steel_1080p.webm
Before => After, 13.25 => 13.23 (fps)

Change-Id: Iebe3b0c6ed2359c778b0570763c5681ae25fef0c
Signed-off-by: James Yu <james.yu@linaro.org>

diff --git a/vp8/common/arm/neon/dequantizeb_neon.asm b/vp8/common/arm/neon/dequantizeb_neon.asm
deleted file mode 100644 (file)
index c8e0c31..0000000
--- a/vp8/common/arm/neon/dequantizeb_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_dequantize_b_loop_neon|
-    ARM
-    REQUIRE8
-    PRESERVE8
-
-    AREA ||.text||, CODE, READONLY, ALIGN=2
-; r0    short *Q,
-; r1    short *DQC
-; r2    short *DQ
-|vp8_dequantize_b_loop_neon| PROC
-    vld1.16         {q0, q1}, [r0]
-    vld1.16         {q2, q3}, [r1]
-
-    vmul.i16        q4, q0, q2
-    vmul.i16        q5, q1, q3
-
-    vst1.16         {q4, q5}, [r2]
-
-    bx             lr
-
-    ENDP
-
-    END

diff --git a/vp8/common/arm/neon/dequantizeb_neon.c b/vp8/common/arm/neon/dequantizeb_neon.c
new file mode 100644 (file)
index 0000000..60f69c8
--- /dev/null
+++ b/vp8/common/arm/neon/dequantizeb_neon.c
@@ -0,0 +1,27 @@
+/*
+ *  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_dequantize_b_loop_neon(
+        int16_t *Q,
+        int16_t *DQC,
+        int16_t *DQ) {
+    int16x8x2_t qQ, qDQC, qDQ;
+
+    qQ   = vld2q_s16(Q);
+    qDQC = vld2q_s16(DQC);
+
+    qDQ.val[0] = vmulq_s16(qQ.val[0], qDQC.val[0]);
+    qDQ.val[1] = vmulq_s16(qQ.val[1], qDQC.val[1]);
+
+    vst2q_s16(DQ, qDQ);
+    return;
+}

diff --git a/vp8/vp8_common.mk b/vp8/vp8_common.mk
index e9806439d4670eeb03083e7a8d3b91b2746ae7db..ac91d7af5ba63409b9e38b8f1208b7c6d63c7a16 100644 (file)
--- a/vp8/vp8_common.mk
+++ b/vp8/vp8_common.mk
@@ -175,7 +175,6 @@ VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/buildintrapredictorsmby_neon$(A
 VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/save_reg_neon$(ASM)
 VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/idct_dequant_full_2x_neon$(ASM)
 VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/idct_dequant_0_2x_neon$(ASM)
-VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/dequantizeb_neon$(ASM)
 VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/idct_blk_neon.c
 VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/variance_neon$(ASM)
 VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/vp8_subpixelvariance8x8_neon$(ASM)
@@ -187,6 +186,7 @@ VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/bilinearpredict_neon.c
 VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/copymem_neon.c
 VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/dc_only_idct_add_neon.c
 VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/dequant_idct_neon.c
+VP8_COMMON_SRCS-$(HAVE_NEON)  += common/arm/neon/dequantizeb_neon.c
 
 
 $(eval $(call rtcd_h_template,vp8_rtcd,vp8/common/rtcd_defs.sh))