update extend frame borders

the neon code made several assumptions which were broken by a recent
change: https://review.webmproject.org/2676

update the code with new assumptions and guard them with a compile time
assert

Change-Id: I32a8378030759966068f34618d7b4b1b02e101a0

1 files changed, 60 insertions, 343 deletions

diff --git a/vpx_scale/arm/neon/vp8_vpxyv12_extendframeborders_neon.asm b/vpx_scale/arm/neon/vp8_vpxyv12_extendframeborders_neon.asm
index b0a3b93a2..8444b8e03 100644
--- a/vpx_scale/arm/neon/vp8_vpxyv12_extendframeborders_neon.asm
+++ b/vpx_scale/arm/neon/vp8_vpxyv12_extendframeborders_neon.asm
@@ -18,42 +18,32 @@
 
     AREA ||.text||, CODE, READONLY, ALIGN=2
 ;void vp8_yv12_extend_frame_borders_neon (YV12_BUFFER_CONFIG *ybf);
-;Note: this is VP8 function, which has border=32 and 16. Internal y_width and y_height
-; are always multiples of 16.
+; we depend on VP8BORDERINPIXELS being 32
 
 |vp8_yv12_extend_frame_borders_neon| PROC
     push            {r4 - r10, lr}
     vpush           {d8 - d15}
 
-    ;Not need to load y_width, since: y_width = y_stride - 2*border
-    ldr             r3, [r0, #yv12_buffer_config_border]
-    ldr             r1, [r0, #yv12_buffer_config_y_buffer]       ;srcptr1
-    ldr             r4, [r0, #yv12_buffer_config_y_height]
-    ldr             lr, [r0, #yv12_buffer_config_y_stride]
+    ; Border = 32
+    ldr             r3, [r0, #yv12_buffer_config_y_width]  ; plane_width
+    ldr             r1, [r0, #yv12_buffer_config_y_buffer] ; src_ptr1
+    ldr             r4, [r0, #yv12_buffer_config_y_height] ; plane_height
+    ldr             lr, [r0, #yv12_buffer_config_y_stride] ; plane_stride
 
-    cmp             r3, #16
-    beq             b16_extend_frame_borders
+; Border copy for Y plane
+; copy the left and right most columns out
+    add             r6, r1, r3              ; dest_ptr2 = src_ptr2 + 1 (src_ptr1 + plane_width)
+    sub             r2, r6, #1              ; src_ptr2 = src_ptr1 + plane_width - 1
+    sub             r5, r1, #32             ; dest_ptr1 = src_ptr1 - Border
 
-;=======================
-b32_extend_frame_borders
-;border = 32
-;=======================
-;Border copy for Y plane
-;copy the left and right most columns out
-    sub             r5, r1, r3              ;destptr1
-    add             r6, r1, lr
-    sub             r6, r6, r3, lsl #1      ;destptr2
-    sub             r2, r6, #1              ;srcptr2
-
-    ;Do four rows at one time
-    mov             r12, r4, lsr #2
+    mov             r12, r4, lsr #2         ; plane_height / 4
 
 copy_left_right_y
     vld1.8          {d0[], d1[]}, [r1], lr
     vld1.8          {d4[], d5[]}, [r2], lr
     vld1.8          {d8[], d9[]}, [r1], lr
     vld1.8          {d12[], d13[]}, [r2], lr
-    vld1.8          {d16[], d17[]},  [r1], lr
+    vld1.8          {d16[], d17[]}, [r1], lr
     vld1.8          {d20[], d21[]}, [r2], lr
     vld1.8          {d24[], d25[]}, [r1], lr
     vld1.8          {d28[], d29[]}, [r2], lr
@@ -81,15 +71,16 @@ copy_left_right_y
     bne             copy_left_right_y
 
 ;Now copy the top and bottom source lines into each line of the respective borders
-    ldr             r7, [r0, #yv12_buffer_config_y_buffer]       ;srcptr1
-    mul             r8, r3, lr
+    ldr             r1, [r0, #yv12_buffer_config_y_buffer] ; y_buffer
+    mul             r8, r4, lr              ; plane_height * plane_stride
 
-    mov             r12, lr, lsr #7
+    ; copy width is plane_stride
+    mov             r12, lr, lsr #7         ; plane_stride / 128
 
-    sub             r6, r1, r3              ;destptr2
-    sub             r2, r6, lr              ;srcptr2
-    sub             r1, r7, r3              ;srcptr1
-    sub             r5, r1, r8              ;destptr1
+    sub             r1, r1, #32             ; src_ptr1 = y_buffer - Border
+    add             r6, r1, r8              ; dest_ptr2 = src_ptr2 - plane_stride (src_ptr1 + (plane_height * plane_stride))
+    sub             r2, r6, lr              ; src_ptr2 = src_ptr1 + (plane_height * plane_stride) - plane_stride
+    sub             r5, r1, lr, asl #5      ; dest_ptr1 = src_ptr1 - (Border * plane_stride)
 
 copy_top_bottom_y
     vld1.8          {q0, q1}, [r1]!
@@ -101,7 +92,7 @@ copy_top_bottom_y
     vld1.8          {q6, q7}, [r1]!
     vld1.8          {q14, q15}, [r2]!
 
-    mov             r7, r3
+    mov             r7, #32                 ; Border
 
 top_bottom_32
     subs            r7, r7, #1
@@ -115,44 +106,41 @@ top_bottom_32
     vst1.8          {q6, q7}, [r5]!
     vst1.8          {q14, q15}, [r6]!
 
-    add             r5, r5, lr
-    sub             r5, r5, #128
-    add             r6, r6, lr
-    sub             r6, r6, #128
+    add             r5, r5, lr              ; dest_ptr1 += plane_stride
+    sub             r5, r5, #128            ; dest_ptr1 -= 128
+    add             r6, r6, lr              ; dest_ptr2 += plane_stride
+    sub             r6, r6, #128            ; dest_ptr2 -= 128
 
     bne             top_bottom_32
 
-    sub             r5, r1, r8
-    add             r6, r2, lr
+    sub             r5, r1, lr, asl #5      ; src_ptr1 - (Border* plane_stride)
+    add             r6, r2, lr              ; src_ptr2 + plane_stride
 
     subs            r12, r12, #1
     bne             copy_top_bottom_y
 
-    mov             r7, lr, lsr #4              ;check to see if extra copy is needed
+    mov             r7, lr, lsr #4          ; check to see if extra copy is needed
     ands            r7, r7, #0x7
     bne             extra_top_bottom_y
 end_of_border_copy_y
 
 ;Border copy for U, V planes
-    ldr             r1, [r0, #yv12_buffer_config_u_buffer]       ;srcptr1
-    mov             lr, lr, lsr #1              ;uv_stride
-    mov             r3, r3, lsr #1              ;border
-    mov             r4, r4, lsr #1              ;uv_height
-    mov             r8, r8, lsr #2
+; Border = 16
+    ldr             r7, [r0, #yv12_buffer_config_u_buffer]  ; src_ptr1
+    ldr             lr, [r0, #yv12_buffer_config_uv_stride] ; plane_stride
+    ldr             r3, [r0, #yv12_buffer_config_uv_width]  ; plane_width
+    ldr             r4, [r0, #yv12_buffer_config_uv_height] ; plane_height
 
     mov             r10, #2
 
 ;copy the left and right most columns out
 border_copy_uv
-    sub             r5, r1, r3              ;destptr1
-    add             r6, r1, lr
-    sub             r6, r6, r3, lsl #1      ;destptr2
-    sub             r2, r6, #1              ;srcptr2
+    mov             r1, r7                  ; src_ptr1 needs to be saved for second half of loop
+    sub             r5, r1, #16             ; dest_ptr1 = src_ptr1 - Border
+    add             r6, r1, r3              ; dest_ptr2 = src_ptr2 + 1 (src_ptr1 + plane_width)
+    sub             r2, r6, #1              ; src_ptr2 = src_ptr1 + plane_width - 1
 
-    mov             r7, r1
-
-    ;Do eight rows at one time
-    mov             r12, r4, lsr #3
+    mov             r12, r4, lsr #3         ; plane_height / 8
 
 copy_left_right_uv
     vld1.8          {d0[], d1[]}, [r1], lr
@@ -167,7 +155,7 @@ copy_left_right_uv
     vld1.8          {d18[], d19[]}, [r2], lr
     vld1.8          {d20[], d21[]}, [r1], lr
     vld1.8          {d22[], d23[]}, [r2], lr
-    vld1.8          {d24[], d25[]},  [r1], lr
+    vld1.8          {d24[], d25[]}, [r1], lr
     vld1.8          {d26[], d27[]}, [r2], lr
     vld1.8          {d28[], d29[]}, [r1], lr
     vld1.8          {d30[], d31[]}, [r2], lr
@@ -194,12 +182,14 @@ copy_left_right_uv
     bne             copy_left_right_uv
 
 ;Now copy the top and bottom source lines into each line of the respective borders
-    mov             r12, lr, lsr #6
+    mov             r1, r7
+    mul             r8, r4, lr              ; plane_height * plane_stride
+    mov             r12, lr, lsr #6         ; plane_stride / 64
 
-    sub             r6, r1, r3              ;destptr2
-    sub             r2, r6, lr              ;srcptr2
-    sub             r1, r7, r3              ;srcptr1
-    sub             r5, r1, r8              ;destptr1
+    sub             r1, r1, #16             ; src_ptr1 = u_buffer - Border
+    add             r6, r1, r8              ; dest_ptr2 = src_ptr2 + plane_stride (src_ptr1 + (plane_height * plane_stride)
+    sub             r2, r6, lr              ; src_ptr2 = src_ptr1 + (plane_height * plane_stride) - plane_stride
+    sub             r5, r1, lr, asl #4      ; dest_ptr1 = src_ptr1 - (Border * plane_stride)
 
 copy_top_bottom_uv
     vld1.8          {q0, q1}, [r1]!
@@ -207,7 +197,7 @@ copy_top_bottom_uv
     vld1.8          {q2, q3}, [r1]!
     vld1.8          {q10, q11}, [r2]!
 
-    mov             r7, r3
+    mov             r7, #16                 ; Border
 
 top_bottom_16
     subs            r7, r7, #1
@@ -217,38 +207,37 @@ top_bottom_16
     vst1.8          {q2, q3}, [r5]!
     vst1.8          {q10, q11}, [r6]!
 
-    add             r5, r5, lr
+    add             r5, r5, lr              ; dest_ptr1 += plane_stride
     sub             r5, r5, #64
-    add             r6, r6, lr
+    add             r6, r6, lr              ; dest_ptr2 += plane_stride
     sub             r6, r6, #64
 
     bne             top_bottom_16
 
-    sub             r5, r1, r8
-    add             r6, r2, lr
+    sub             r5, r1, lr, asl #4      ; dest_ptr1 = src_ptr1 - (Border * plane_stride)
+    add             r6, r2, lr              ; dest_ptr2 = src_ptr2 + plane_stride
 
     subs            r12, r12, #1
     bne             copy_top_bottom_uv
 
-    mov             r7, lr, lsr #3              ;check to see if extra copy is needed
+    mov             r7, lr, lsr #3          ; check to see if extra copy is needed
     ands            r7, r7, #0x7
     bne             extra_top_bottom_uv
 
 end_of_border_copy_uv
     subs            r10, r10, #1
-    ldrne           r1, [r0, #yv12_buffer_config_v_buffer]       ;srcptr1
+    ldrne           r7, [r0, #yv12_buffer_config_v_buffer] ; src_ptr1
     bne             border_copy_uv
 
     vpop            {d8 - d15}
     pop             {r4 - r10, pc}
 
 ;;;;;;;;;;;;;;;;;;;;;;
-;extra copy part for Y
 extra_top_bottom_y
     vld1.8          {q0}, [r1]!
     vld1.8          {q2}, [r2]!
 
-    mov             r9, r3, lsr #3
+    mov             r9, #4                  ; 32 >> 3
 
 extra_top_bottom_32
     subs            r9, r9, #1
@@ -271,19 +260,18 @@ extra_top_bottom_32
     vst1.8          {q2}, [r6], lr
     bne             extra_top_bottom_32
 
-    sub             r5, r1, r8
-    add             r6, r2, lr
+    sub             r5, r1, lr, asl #5      ; src_ptr1 - (Border * plane_stride)
+    add             r6, r2, lr              ; src_ptr2 + plane_stride
     subs            r7, r7, #1
     bne             extra_top_bottom_y
 
     b               end_of_border_copy_y
 
-;extra copy part for UV
 extra_top_bottom_uv
     vld1.8          {d0}, [r1]!
     vld1.8          {d8}, [r2]!
 
-    mov             r9, r3, lsr #3
+    mov             r9, #2                  ; 16 >> 3
 
 extra_top_bottom_16
     subs            r9, r9, #1
@@ -306,283 +294,12 @@ extra_top_bottom_16
     vst1.8          {d8}, [r6], lr
     bne             extra_top_bottom_16
 
-    sub             r5, r1, r8
-    add             r6, r2, lr
+    sub             r5, r1, lr, asl #4      ; src_ptr1 - (Border * plane_stride)
+    add             r6, r2, lr              ; src_ptr2 + plane_stride
     subs            r7, r7, #1
     bne             extra_top_bottom_uv
 
     b               end_of_border_copy_uv
 
-
-;=======================
-b16_extend_frame_borders
-;border = 16
-;=======================
-;Border copy for Y plane
-;copy the left and right most columns out
-    sub             r5, r1, r3              ;destptr1
-    add             r6, r1, lr
-    sub             r6, r6, r3, lsl #1      ;destptr2
-    sub             r2, r6, #1              ;srcptr2
-
-    ;Do four rows at one time
-    mov             r12, r4, lsr #2
-
-copy_left_right_y_b16
-    vld1.8          {d0[], d1[]}, [r1], lr
-    vld1.8          {d4[], d5[]}, [r2], lr
-    vld1.8          {d8[], d9[]}, [r1], lr
-    vld1.8          {d12[], d13[]}, [r2], lr
-    vld1.8          {d16[], d17[]},  [r1], lr
-    vld1.8          {d20[], d21[]}, [r2], lr
-    vld1.8          {d24[], d25[]}, [r1], lr
-    vld1.8          {d28[], d29[]}, [r2], lr
-
-    subs            r12, r12, #1
-
-    vst1.8          {q0}, [r5], lr
-    vst1.8          {q2}, [r6], lr
-    vst1.8          {q4}, [r5], lr
-    vst1.8          {q6}, [r6], lr
-    vst1.8          {q8}, [r5], lr
-    vst1.8          {q10}, [r6], lr
-    vst1.8          {q12}, [r5], lr
-    vst1.8          {q14}, [r6], lr
-
-    bne             copy_left_right_y_b16
-
-;Now copy the top and bottom source lines into each line of the respective borders
-    ldr             r7, [r0, #yv12_buffer_config_y_buffer]       ;srcptr1
-    mul             r8, r3, lr
-
-    mov             r12, lr, lsr #7
-
-    sub             r6, r1, r3              ;destptr2
-    sub             r2, r6, lr              ;srcptr2
-    sub             r1, r7, r3              ;srcptr1
-    sub             r5, r1, r8              ;destptr1
-
-copy_top_bottom_y_b16
-    vld1.8          {q0, q1}, [r1]!
-    vld1.8          {q8, q9}, [r2]!
-    vld1.8          {q2, q3}, [r1]!
-    vld1.8          {q10, q11}, [r2]!
-    vld1.8          {q4, q5}, [r1]!
-    vld1.8          {q12, q13}, [r2]!
-    vld1.8          {q6, q7}, [r1]!
-    vld1.8          {q14, q15}, [r2]!
-
-    mov             r7, r3
-
-top_bottom_16_b16
-    subs            r7, r7, #1
-
-    vst1.8          {q0, q1}, [r5]!
-    vst1.8          {q8, q9}, [r6]!
-    vst1.8          {q2, q3}, [r5]!
-    vst1.8          {q10, q11}, [r6]!
-    vst1.8          {q4, q5}, [r5]!
-    vst1.8          {q12, q13}, [r6]!
-    vst1.8          {q6, q7}, [r5]!
-    vst1.8          {q14, q15}, [r6]!
-
-    add             r5, r5, lr
-    sub             r5, r5, #128
-    add             r6, r6, lr
-    sub             r6, r6, #128
-
-    bne             top_bottom_16_b16
-
-    sub             r5, r1, r8
-    add             r6, r2, lr
-
-    subs            r12, r12, #1
-    bne             copy_top_bottom_y_b16
-
-    mov             r7, lr, lsr #4              ;check to see if extra copy is needed
-    ands            r7, r7, #0x7
-    bne             extra_top_bottom_y_b16
-end_of_border_copy_y_b16
-
-;Border copy for U, V planes
-    ldr             r1, [r0, #yv12_buffer_config_u_buffer]       ;srcptr1
-    mov             lr, lr, lsr #1              ;uv_stride
-    mov             r3, r3, lsr #1              ;border
-    mov             r4, r4, lsr #1              ;uv_height
-    mov             r8, r8, lsr #2
-
-    mov             r10, #2
-
-;copy the left and right most columns out
-border_copy_uv_b16
-    sub             r5, r1, r3              ;destptr1
-    add             r6, r1, lr
-    sub             r6, r6, r3, lsl #1      ;destptr2
-    sub             r2, r6, #1              ;srcptr2
-
-    mov             r7, r1
-
-    ;Do eight rows at one time
-    mov             r12, r4, lsr #3
-
-copy_left_right_uv_b16
-    vld1.8          {d0[]}, [r1], lr
-    vld1.8          {d2[]}, [r2], lr
-    vld1.8          {d4[]}, [r1], lr
-    vld1.8          {d6[]}, [r2], lr
-    vld1.8          {d8[]},  [r1], lr
-    vld1.8          {d10[]}, [r2], lr
-    vld1.8          {d12[]}, [r1], lr
-    vld1.8          {d14[]}, [r2], lr
-    vld1.8          {d16[]}, [r1], lr
-    vld1.8          {d18[]}, [r2], lr
-    vld1.8          {d20[]}, [r1], lr
-    vld1.8          {d22[]}, [r2], lr
-    vld1.8          {d24[]},  [r1], lr
-    vld1.8          {d26[]}, [r2], lr
-    vld1.8          {d28[]}, [r1], lr
-    vld1.8          {d30[]}, [r2], lr
-
-    subs            r12, r12, #1
-
-    vst1.8          {d0}, [r5], lr
-    vst1.8          {d2}, [r6], lr
-    vst1.8          {d4}, [r5], lr
-    vst1.8          {d6}, [r6], lr
-    vst1.8          {d8}, [r5], lr
-    vst1.8          {d10}, [r6], lr
-    vst1.8          {d12}, [r5], lr
-    vst1.8          {d14}, [r6], lr
-    vst1.8          {d16}, [r5], lr
-    vst1.8          {d18}, [r6], lr
-    vst1.8          {d20}, [r5], lr
-    vst1.8          {d22}, [r6], lr
-    vst1.8          {d24}, [r5], lr
-    vst1.8          {d26}, [r6], lr
-    vst1.8          {d28}, [r5], lr
-    vst1.8          {d30}, [r6], lr
-
-    bne             copy_left_right_uv_b16
-
-;Now copy the top and bottom source lines into each line of the respective borders
-    mov             r12, lr, lsr #6
-
-    sub             r6, r1, r3              ;destptr2
-    sub             r2, r6, lr              ;srcptr2
-    sub             r1, r7, r3              ;srcptr1
-    sub             r5, r1, r8              ;destptr1
-
-copy_top_bottom_uv_b16
-    vld1.8          {q0, q1}, [r1]!
-    vld1.8          {q8, q9}, [r2]!
-    vld1.8          {q2, q3}, [r1]!
-    vld1.8          {q10, q11}, [r2]!
-
-    mov             r7, r3
-
-top_bottom_8_b16
-    subs            r7, r7, #1
-
-    vst1.8          {q0, q1}, [r5]!
-    vst1.8          {q8, q9}, [r6]!
-    vst1.8          {q2, q3}, [r5]!
-    vst1.8          {q10, q11}, [r6]!
-
-    add             r5, r5, lr
-    sub             r5, r5, #64
-    add             r6, r6, lr
-    sub             r6, r6, #64
-
-    bne             top_bottom_8_b16
-
-    sub             r5, r1, r8
-    add             r6, r2, lr
-
-    subs            r12, r12, #1
-    bne             copy_top_bottom_uv_b16
-
-    mov             r7, lr, lsr #3              ;check to see if extra copy is needed
-    ands            r7, r7, #0x7
-    bne             extra_top_bottom_uv_b16
-
-end_of_border_copy_uv_b16
-    subs            r10, r10, #1
-    ldrne           r1, [r0, #yv12_buffer_config_v_buffer]       ;srcptr1
-    bne             border_copy_uv_b16
-
-    vpop            {d8-d15}
-    pop             {r4 - r10, pc}
-
-;;;;;;;;;;;;;;;;;;;;;;
-;extra copy part for Y
-extra_top_bottom_y_b16
-    vld1.8          {q0}, [r1]!
-    vld1.8          {q2}, [r2]!
-
-    mov             r9, r3, lsr #3
-
-extra_top_bottom_16_b16
-    subs            r9, r9, #1
-
-    vst1.8          {q0}, [r5], lr
-    vst1.8          {q2}, [r6], lr
-    vst1.8          {q0}, [r5], lr
-    vst1.8          {q2}, [r6], lr
-    vst1.8          {q0}, [r5], lr
-    vst1.8          {q2}, [r6], lr
-    vst1.8          {q0}, [r5], lr
-    vst1.8          {q2}, [r6], lr
-    vst1.8          {q0}, [r5], lr
-    vst1.8          {q2}, [r6], lr
-    vst1.8          {q0}, [r5], lr
-    vst1.8          {q2}, [r6], lr
-    vst1.8          {q0}, [r5], lr
-    vst1.8          {q2}, [r6], lr
-    vst1.8          {q0}, [r5], lr
-    vst1.8          {q2}, [r6], lr
-    bne             extra_top_bottom_16_b16
-
-    sub             r5, r1, r8
-    add             r6, r2, lr
-    subs            r7, r7, #1
-    bne             extra_top_bottom_y_b16
-
-    b               end_of_border_copy_y_b16
-
-;extra copy part for UV
-extra_top_bottom_uv_b16
-    vld1.8          {d0}, [r1]!
-    vld1.8          {d8}, [r2]!
-
-    mov             r9, r3, lsr #3
-
-extra_top_bottom_8_b16
-    subs            r9, r9, #1
-
-    vst1.8          {d0}, [r5], lr
-    vst1.8          {d8}, [r6], lr
-    vst1.8          {d0}, [r5], lr
-    vst1.8          {d8}, [r6], lr
-    vst1.8          {d0}, [r5], lr
-    vst1.8          {d8}, [r6], lr
-    vst1.8          {d0}, [r5], lr
-    vst1.8          {d8}, [r6], lr
-    vst1.8          {d0}, [r5], lr
-    vst1.8          {d8}, [r6], lr
-    vst1.8          {d0}, [r5], lr
-    vst1.8          {d8}, [r6], lr
-    vst1.8          {d0}, [r5], lr
-    vst1.8          {d8}, [r6], lr
-    vst1.8          {d0}, [r5], lr
-    vst1.8          {d8}, [r6], lr
-    bne             extra_top_bottom_8_b16
-
-    sub             r5, r1, r8
-    add             r6, r2, lr
-    subs            r7, r7, #1
-    bne             extra_top_bottom_uv_b16
-
-    b               end_of_border_copy_uv_b16
-
     ENDP
     END