Add vp8_mse16x16_armv6 function

Change-Id: I77e9f2f521a71089228f96e2db72524189364ffb

3 files changed, 139 insertions, 2 deletions

diff --git a/vp8/encoder/arm/arm_csystemdependent.c b/vp8/encoder/arm/arm_csystemdependent.c
index 73007d414..afd43042d 100644
--- a/vp8/encoder/arm/arm_csystemdependent.c
+++ b/vp8/encoder/arm/arm_csystemdependent.c
@@ -50,8 +50,8 @@ void vp8_arch_arm_encoder_init(VP8_COMP *cpi)
         cpi->rtcd.variance.halfpixvar16x16_v     = vp8_variance_halfpixvar16x16_v_armv6;
         cpi->rtcd.variance.halfpixvar16x16_hv    = vp8_variance_halfpixvar16x16_hv_armv6;
 
-        /*cpi->rtcd.variance.mse16x16              = vp8_mse16x16_c;
-        cpi->rtcd.variance.getmbss               = vp8_get_mb_ss_c;*/
+        cpi->rtcd.variance.mse16x16              = vp8_mse16x16_armv6;
+        /*cpi->rtcd.variance.getmbss               = vp8_get_mb_ss_c;*/
 
         /*cpi->rtcd.variance.get16x16prederror     = vp8_get16x16pred_error_c;
         cpi->rtcd.variance.get8x8var             = vp8_get8x8var_c;
diff --git a/vp8/encoder/arm/armv6/vp8_mse16x16_armv6.asm b/vp8/encoder/arm/armv6/vp8_mse16x16_armv6.asm
new file mode 100644
index 000000000..a9060d76f
--- /dev/null
+++ b/vp8/encoder/arm/armv6/vp8_mse16x16_armv6.asm
@@ -0,0 +1,133 @@
+;
+;  Copyright (c) 2011 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_mse16x16_armv6|
+
+    ARM
+
+    AREA ||.text||, CODE, READONLY, ALIGN=2
+
+; r0    unsigned char *src_ptr
+; r1    int source_stride
+; r2    unsigned char *ref_ptr
+; r3    int  recon_stride
+; stack unsigned int *sse
+;
+;note: Based on vp8_variance16x16_armv6. In this function, sum is never used.
+;      So, we can remove this part of calculation.
+
+|vp8_mse16x16_armv6| PROC
+
+    push    {r4-r9, lr}
+    mov     r12, #16            ; set loop counter to 16 (=block height)
+
+    mov     r4, #0              ; initialize sse = 0
+
+loop
+    ; 1st 4 pixels
+    ldr     r5, [r0, #0x0]      ; load 4 src pixels
+    ldr     r6, [r2, #0x0]      ; load 4 ref pixels
+
+    mov     lr, #0              ; constant zero
+
+    usub8   r8, r5, r6          ; calculate difference
+    sel     r7, r8, lr          ; select bytes with positive difference
+    usub8   r9, r6, r5          ; calculate difference with reversed operands
+    sel     r8, r9, lr          ; select bytes with negative difference
+
+    ; calculate partial sums
+    usad8   r5, r7, lr          ; calculate sum of positive differences
+    usad8   r6, r8, lr          ; calculate sum of negative differences
+    orr     r8, r8, r7          ; differences of all 4 pixels
+
+    ldr     r5, [r0, #0x4]      ; load 4 src pixels
+
+    ; calculate sse
+    uxtb16  r6, r8              ; byte (two pixels) to halfwords
+    uxtb16  r7, r8, ror #8      ; another two pixels to halfwords
+    smlad   r4, r6, r6, r4      ; dual signed multiply, add and accumulate (1)
+
+    ; 2nd 4 pixels
+    ldr     r6, [r2, #0x4]      ; load 4 ref pixels
+    smlad   r4, r7, r7, r4      ; dual signed multiply, add and accumulate (2)
+
+    usub8   r8, r5, r6          ; calculate difference
+    sel     r7, r8, lr          ; select bytes with positive difference
+    usub8   r9, r6, r5          ; calculate difference with reversed operands
+    sel     r8, r9, lr          ; select bytes with negative difference
+
+    ; calculate partial sums
+    usad8   r5, r7, lr          ; calculate sum of positive differences
+    usad8   r6, r8, lr          ; calculate sum of negative differences
+    orr     r8, r8, r7          ; differences of all 4 pixels
+    ldr     r5, [r0, #0x8]      ; load 4 src pixels
+    ; calculate sse
+    uxtb16  r6, r8              ; byte (two pixels) to halfwords
+    uxtb16  r7, r8, ror #8      ; another two pixels to halfwords
+    smlad   r4, r6, r6, r4      ; dual signed multiply, add and accumulate (1)
+
+    ; 3rd 4 pixels
+    ldr     r6, [r2, #0x8]      ; load 4 ref pixels
+    smlad   r4, r7, r7, r4      ; dual signed multiply, add and accumulate (2)
+
+    usub8   r8, r5, r6          ; calculate difference
+    sel     r7, r8, lr          ; select bytes with positive difference
+    usub8   r9, r6, r5          ; calculate difference with reversed operands
+    sel     r8, r9, lr          ; select bytes with negative difference
+
+    ; calculate partial sums
+    usad8   r5, r7, lr          ; calculate sum of positive differences
+    usad8   r6, r8, lr          ; calculate sum of negative differences
+    orr     r8, r8, r7          ; differences of all 4 pixels
+
+    ldr     r5, [r0, #0xc]      ; load 4 src pixels
+
+    ; calculate sse
+    uxtb16  r6, r8              ; byte (two pixels) to halfwords
+    uxtb16  r7, r8, ror #8      ; another two pixels to halfwords
+    smlad   r4, r6, r6, r4      ; dual signed multiply, add and accumulate (1)
+
+    ; 4th 4 pixels
+    ldr     r6, [r2, #0xc]      ; load 4 ref pixels
+    smlad   r4, r7, r7, r4      ; dual signed multiply, add and accumulate (2)
+
+    usub8   r8, r5, r6          ; calculate difference
+    add     r0, r0, r1          ; set src_ptr to next row
+    sel     r7, r8, lr          ; select bytes with positive difference
+    usub8   r9, r6, r5          ; calculate difference with reversed operands
+    add     r2, r2, r3          ; set dst_ptr to next row
+    sel     r8, r9, lr          ; select bytes with negative difference
+
+    ; calculate partial sums
+    usad8   r5, r7, lr          ; calculate sum of positive differences
+    usad8   r6, r8, lr          ; calculate sum of negative differences
+    orr     r8, r8, r7          ; differences of all 4 pixels
+
+    subs    r12, r12, #1        ; next row
+
+    ; calculate sse
+    uxtb16  r6, r8              ; byte (two pixels) to halfwords
+    uxtb16  r7, r8, ror #8      ; another two pixels to halfwords
+    smlad   r4, r6, r6, r4      ; dual signed multiply, add and accumulate (1)
+    smlad   r4, r7, r7, r4      ; dual signed multiply, add and accumulate (2)
+
+    bne     loop
+
+    ; return stuff
+    ldr     r1, [sp, #28]       ; get address of sse
+    mov     r0, r4              ; return sse
+    str     r4, [r1]            ; store sse
+
+    pop     {r4-r9, pc}
+
+    ENDP
+
+    END
diff --git a/vp8/encoder/arm/variance_arm.h b/vp8/encoder/arm/variance_arm.h
index 7ac0ac08e..7ad7c76d3 100644
--- a/vp8/encoder/arm/variance_arm.h
+++ b/vp8/encoder/arm/variance_arm.h
@@ -20,6 +20,7 @@ extern prototype_subpixvariance(vp8_sub_pixel_variance16x16_armv6);
 extern prototype_variance(vp8_variance_halfpixvar16x16_h_armv6);
 extern prototype_variance(vp8_variance_halfpixvar16x16_v_armv6);
 extern prototype_variance(vp8_variance_halfpixvar16x16_hv_armv6);
+extern prototype_variance(vp8_mse16x16_armv6);
 
 #if !CONFIG_RUNTIME_CPU_DETECT
 
@@ -32,6 +33,9 @@ extern prototype_variance(vp8_variance_halfpixvar16x16_hv_armv6);
 #undef  vp8_variance_var16x16
 #define vp8_variance_var16x16 vp8_variance16x16_armv6
 
+#undef  vp8_variance_mse16x16
+#define vp8_variance_mse16x16 vp8_mse16x16_armv6
+
 #undef  vp8_variance_halfpixvar16x16_h
 #define vp8_variance_halfpixvar16x16_h vp8_variance_halfpixvar16x16_h_armv6