1@/*****************************************************************************
2@*
3@* Copyright (C) 2012 Ittiam Systems Pvt Ltd, Bangalore
4@*
5@* Licensed under the Apache License, Version 2.0 (the "License");
6@* you may not use this file except in compliance with the License.
7@* You may obtain a copy of the License at:
8@*
9@* http://www.apache.org/licenses/LICENSE-2.0
10@*
11@* Unless required by applicable law or agreed to in writing, software
12@* distributed under the License is distributed on an "AS IS" BASIS,
13@* WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
14@* See the License for the specific language governing permissions and
15@* limitations under the License.
16@*
17@*****************************************************************************/
18@/**
19@*******************************************************************************
20@* @file
21@*  ihevc_inter_pred_chroma_copy_w16out_neon.s
22@*
23@* @brief
24@*  contains function definitions for inter prediction  interpolation.
25@* functions are coded using neon  intrinsics and can be compiled using
26
27@* rvct
28@*
29@* @author
30@*  yogeswaran rs
31@*
32@* @par list of functions:
33@*
34@*
35@* @remarks
36@*  none
37@*
38@*******************************************************************************
39@*/
40@/**
41@*******************************************************************************
42@*
43@* @brief
44@*   chroma interprediction filter for copy
45@*
46@* @par description:
47@*    copies the array of width 'wd' and height 'ht' from the  location pointed
48@*    by 'src' to the location pointed by 'dst'
49@*
50@* @param[in] pu1_src
51@*  uword8 pointer to the source
52@*
53@* @param[out] pu1_dst
54@*  uword8 pointer to the destination
55@*
56@* @param[in] src_strd
57@*  integer source stride
58@*
59@* @param[in] dst_strd
60@*  integer destination stride
61@*
62@* @param[in] pi1_coeff
63@*  word8 pointer to the filter coefficients
64@*
65@* @param[in] ht
66@*  integer height of the array
67@*
68@* @param[in] wd
69@*  integer width of the array
70@*
71@* @returns
72@*
73@* @remarks
74@*  none
75@*
76@*******************************************************************************
77@*/
78
79@void ihevc_inter_pred_chroma_copy_w16out(uword8 *pu1_src,
80@                                           word16 *pi2_dst,
81@                                           word32 src_strd,
82@                                           word32 dst_strd,
83@                                           word8 *pi1_coeff,
84@                                           word32 ht,
85@                                           word32 wd)
86@**************variables vs registers*****************************************
87@r0 => *pu1_src
88@r1 => *pi2_dst
89@r2 =>  src_strd
90@r3 =>  dst_strd
91@r4 => *pi1_coeff
92@r5 =>  ht
93@r6 =>  wd
94
95.text
96.align 4
97
98
99
100
101.globl ihevc_inter_pred_chroma_copy_w16out_a9q
102
103.type ihevc_inter_pred_chroma_copy_w16out_a9q, %function
104
105ihevc_inter_pred_chroma_copy_w16out_a9q:
106
107    stmfd       sp!, {r4-r12, r14}          @stack stores the values of the arguments
108    ldr         r12,[sp,#48]                @loads wd
109    lsl         r12,r12,#1                  @2*wd
110    ldr         r7,[sp,#44]                 @loads ht
111    cmp         r7,#0                       @ht condition(ht == 0)
112    ble         end_loops                   @loop
113    and         r8,r7,#3                    @check ht for mul of 2
114    sub         r9,r7,r8                    @check the rounded height value
115    and         r11,r7,#6
116    cmp         r11,#6
117    beq         loop_ht_6
118    tst         r12,#7                      @conditional check for wd (multiples)
119    beq         core_loop_wd_8
120
121loop_ht_6:
122    sub         r11,r12,#4
123    lsls        r6,r3,#1
124    cmp         r9,#0
125    beq         outer_loop_wd_4_ht_2
126
127outer_loop_wd_4:
128    subs        r4,r12,#0                   @wd conditional subtract
129    ble         end_inner_loop_wd_4
130
131inner_loop_wd_4:
132    vld1.8      {d0},[r0]                   @vld1_u8(pu1_src_tmp)
133    add         r5,r0,r2                    @pu1_src +src_strd
134    vmovl.u8    q0,d0                       @vmovl_u8(vld1_u8(pu1_src_tmp)
135    add         r10,r1,r6
136    subs        r4,r4,#4                    @wd - 4
137    vshl.i64    q0,q0,#6                    @vshlq_n_s64(temp, 6)
138    vld1.8      {d22},[r5],r2               @vld1_u8(pu1_src_tmp)
139    add         r0,r0,#4                    @pu1_src += 4
140    vst1.64     {d0},[r1]                   @vst1q_lane_s64(pi2_dst_tmp, temp, 0)
141    add         r1,r1,#8
142    vmovl.u8    q11,d22                     @vmovl_u8(vld1_u8(pu1_src_tmp)
143    vld1.8      {d24},[r5],r2               @vld1_u8(pu1_src_tmp)
144    vshl.i64    q11,q11,#6                  @vshlq_n_s64(temp, 6)
145    vmovl.u8    q12,d24                     @vmovl_u8(vld1_u8(pu1_src_tmp)
146    vst1.64     {d22},[r10],r6              @vst1q_lane_s64(pi2_dst_tmp, temp, 0)
147    vshl.i64    q12,q12,#6                  @vshlq_n_s64(temp, 6)
148    vld1.8      {d26},[r5],r2               @vld1_u8(pu1_src_tmp)
149    vst1.64     {d24},[r10],r6              @vst1q_lane_s64(pi2_dst_tmp, temp, 0)
150    vmovl.u8    q13,d26                     @vmovl_u8(vld1_u8(pu1_src_tmp)
151    vshl.i64    q13,q13,#6                  @vshlq_n_s64(temp, 6)
152    vst1.64     {d26},[r10],r6              @vst1q_lane_s64(pi2_dst_tmp, temp, 0)
153    bgt         inner_loop_wd_4
154
155end_inner_loop_wd_4:
156    subs        r9,r9,#4                    @ht - 4
157    sub         r0,r5,r11
158    sub         r1,r10,r11,lsl #1
159    bgt         outer_loop_wd_4
160    cmp         r8,#0
161    bgt         outer_loop_wd_4_ht_2
162
163
164end_loops:
165    ldmfd       sp!,{r4-r12,r15}            @reload the registers from sp
166
167
168outer_loop_wd_4_ht_2:
169    subs        r4,r12,#0                   @wd conditional subtract
170    ble         end_inner_loop_wd_4
171
172inner_loop_wd_4_ht_2:
173    vld1.8      {d0},[r0]                   @vld1_u8(pu1_src_tmp)
174    add         r5,r0,r2                    @pu1_src +src_strd
175    vmovl.u8    q0,d0                       @vmovl_u8(vld1_u8(pu1_src_tmp)
176    add         r10,r1,r6
177    subs        r4,r4,#4                    @wd - 4
178    vshl.i64    q0,q0,#6                    @vshlq_n_s64(temp, 6)
179    vld1.8      {d22},[r5],r2               @vld1_u8(pu1_src_tmp)
180    add         r0,r0,#4                    @pu1_src += 4
181    vst1.64     {d0},[r1]                   @vst1q_lane_s64(pi2_dst_tmp, temp, 0)
182    add         r1,r1,#8
183    vmovl.u8    q11,d22                     @vmovl_u8(vld1_u8(pu1_src_tmp)
184    vld1.8      {d24},[r5],r2               @vld1_u8(pu1_src_tmp)
185    vshl.i64    q11,q11,#6                  @vshlq_n_s64(temp, 6)
186    vmovl.u8    q12,d24                     @vmovl_u8(vld1_u8(pu1_src_tmp)
187    vst1.64     {d22},[r10],r6              @vst1q_lane_s64(pi2_dst_tmp, temp, 0)
188    bgt         inner_loop_wd_4_ht_2
189    b           end_loops
190
191
192core_loop_wd_8:
193    @sub            r11,r12,#8
194    lsls        r5,r3,#1
195    rsb         r11,r12,r3, lsl #2          @ r11 = (dst_strd * 4) - width
196    rsb         r8,r12,r2,lsl #2            @r2->src_strd
197    mov         r4,r12, lsr #3              @ divide by 8
198    mov         r7,r9
199    mul         r7, r4
200    sub         r4,r12,#0                   @wd conditional check
201    sub         r7,r7,#4                    @subtract one for epilog
202    cmp         r9,#0
203    beq         core_loop_wd_8_ht_2
204
205prolog:
206    add         r6,r0,r2                    @pu1_src_tmp += src_strd
207    add         r10,r1,r5
208    vld1.8      {d8},[r0]!                  @vld1_u8(pu1_src_tmp)
209    vld1.8      {d10},[r6],r2               @vld1_u8(pu1_src_tmp)
210    vld1.8      {d12},[r6],r2               @vld1_u8(pu1_src_tmp)
211    vld1.8      {d14},[r6],r2               @vld1_u8(pu1_src_tmp)
212    vmovl.u8    q8,d8                       @vmovl_u8(vld1_u8(pu1_src_tmp))
213    vmovl.u8    q9,d10                      @vmovl_u8(vld1_u8(pu1_src_tmp)
214    vmovl.u8    q10,d12                     @vmovl_u8(vld1_u8(pu1_src_tmp)
215    vmovl.u8    q11,d14                     @vmovl_u8(vld1_u8(pu1_src_tmp)
216    subs        r4,r4,#8                    @wd decrements by 8
217    vshl.i16    q0,q8,#6                    @vshlq_n_s16(tmp, 6)
218    vshl.i16    q1,q9,#6                    @vshlq_n_s16(tmp, 6)
219    vshl.i16    q2,q10,#6                   @vshlq_n_s16(tmp, 6)
220    vshl.i16    q3,q11,#6                   @vshlq_n_s16(tmp, 6)
221    addle       r0,r0,r8
222    add         r6,r0,r2                    @pu1_src_tmp += src_strd
223    vld1.8      {d8},[r0]!                  @vld1_u8(pu1_src_tmp)
224    vld1.8      {d10},[r6],r2               @vld1_u8(pu1_src_tmp)
225    vld1.8      {d12},[r6],r2               @vld1_u8(pu1_src_tmp)
226    vld1.8      {d14},[r6],r2               @vld1_u8(pu1_src_tmp)
227
228    vst1.16     {d0,d1},[r1]!               @vst1q_s16(pi2_dst_tmp, tmp)
229    addle       r1,r1,r11,lsl #1
230    suble       r4,r12,#0                   @wd conditional check
231
232    subs        r7,r7,#4                    @ht - 4
233
234    blt         epilog_end                  @jumps to epilog_end
235    beq         epilog                      @jumps to epilog
236
237
238
239outer_loop_wd_8:
240
241    vst1.16     {d2,d3},[r10],r5            @vst1q_s16(pi2_dst_tmp, tmp)
242    vmovl.u8    q8,d8                       @vmovl_u8(vld1_u8(pu1_src_tmp))
243
244    vst1.16     {d4,d5},[r10],r5            @vst1q_s16(pi2_dst_tmp, tmp)
245    vmovl.u8    q9,d10                      @vmovl_u8(vld1_u8(pu1_src_tmp)
246
247    vst1.16     {d6,d7},[r10],r5            @vst1q_s16(pi2_dst_tmp, tmp)
248    vmovl.u8    q10,d12                     @vmovl_u8(vld1_u8(pu1_src_tmp)
249
250    vmovl.u8    q11,d14                     @vmovl_u8(vld1_u8(pu1_src_tmp)
251
252    subs        r4,r4,#8                    @wd decrements by 8
253    addle       r0,r0,r8
254
255    add         r6,r0,r2                    @pu1_src_tmp += src_strd
256
257    vld1.8      {d8},[r0]!                  @vld1_u8(pu1_src_tmp)
258    vshl.i16    q0,q8,#6                    @vshlq_n_s16(tmp, 6)
259
260    vld1.8      {d10},[r6],r2               @vld1_u8(pu1_src_tmp)
261    vshl.i16    q1,q9,#6                    @vshlq_n_s16(tmp, 6)
262
263    vld1.8      {d12},[r6],r2               @vld1_u8(pu1_src_tmp)
264    vshl.i16    q2,q10,#6                   @vshlq_n_s16(tmp, 6)
265
266    vld1.8      {d14},[r6],r2               @vld1_u8(pu1_src_tmp)
267    add         r10,r1,r5
268
269    vshl.i16    q3,q11,#6                   @vshlq_n_s16(tmp, 6)
270
271    vst1.16     {d0,d1},[r1]!               @vst1q_s16(pi2_dst_tmp, tmp)
272
273    addle       r1,r1,r11,lsl #1
274    suble       r4,r12,#0                   @wd conditional check
275
276    subs        r7,r7,#4                    @ht - 4
277    bgt         outer_loop_wd_8
278
279epilog:
280    vst1.16     {d2,d3},[r10],r5            @vst1q_s16(pi2_dst_tmp, tmp)
281    vmovl.u8    q8,d8                       @vmovl_u8(vld1_u8(pu1_src_tmp))
282
283    vst1.16     {d4,d5},[r10],r5            @vst1q_s16(pi2_dst_tmp, tmp)
284    vmovl.u8    q9,d10                      @vmovl_u8(vld1_u8(pu1_src_tmp)
285
286    vst1.16     {d6,d7},[r10],r5            @vst1q_s16(pi2_dst_tmp, tmp)
287    vmovl.u8    q10,d12                     @vmovl_u8(vld1_u8(pu1_src_tmp)
288
289    vmovl.u8    q11,d14                     @vmovl_u8(vld1_u8(pu1_src_tmp)
290    @add        r6,r0,r2                @pu1_src_tmp += src_strd
291
292    vshl.i16    q0,q8,#6                    @vshlq_n_s16(tmp, 6)
293    vshl.i16    q1,q9,#6                    @vshlq_n_s16(tmp, 6)
294    vshl.i16    q2,q10,#6                   @vshlq_n_s16(tmp, 6)
295    add         r10,r1,r5
296    vshl.i16    q3,q11,#6                   @vshlq_n_s16(tmp, 6)
297
298    vst1.16     {d0,d1},[r1]!               @vst1q_s16(pi2_dst_tmp, tmp)
299epilog_end:
300    vst1.16     {d2,d3},[r10],r5            @vst1q_s16(pi2_dst_tmp, tmp)
301    vst1.16     {d4,d5},[r10],r5            @vst1q_s16(pi2_dst_tmp, tmp)
302    vst1.16     {d6,d7},[r10],r5            @vst1q_s16(pi2_dst_tmp, tmp)
303    b           end_loops
304
305core_loop_wd_8_ht_2:
306    add         r6,r0,r2                    @pu1_src_tmp += src_strd
307    add         r10,r1,r5
308    vld1.8      {d8},[r0]!                  @vld1_u8(pu1_src_tmp)
309    vld1.8      {d10},[r6],r2               @vld1_u8(pu1_src_tmp)
310    vmovl.u8    q8,d8                       @vmovl_u8(vld1_u8(pu1_src_tmp))
311    vmovl.u8    q9,d10                      @vmovl_u8(vld1_u8(pu1_src_tmp)
312    subs        r12,r12,#8                  @wd decrements by 8
313    vshl.i16    q0,q8,#6                    @vshlq_n_s16(tmp, 6)
314    vshl.i16    q1,q9,#6                    @vshlq_n_s16(tmp, 6)
315    vst1.16     {d0,d1},[r1]!               @vst1q_s16(pi2_dst_tmp, tmp)
316    vst1.16     {d2,d3},[r10],r5            @vst1q_s16(pi2_dst_tmp, tmp)
317    bgt         core_loop_wd_8_ht_2
318
319    ldmfd       sp!,{r4-r12,r15}            @reload the registers from sp
320
321
322
323
324
325
326