History

ARM: libswresample: NEON optimised generic fltp to s16 conversion

Adapted to swr by: Michael Niedermayer <michaelni@gmx.at> Signed-off-by: Michael Niedermayer <michaelni@gmx.at>

Mans Rullgard committed 14 years ago

Commit 7e3208a1 parent 97599f82 files · +242 −0Browse files

Unified · Split

2 files
  1. libswresample/arm/audio_convert_init.c +9 −0
  2. libswresample/arm/audio_convert_neon.S +233 −0
libswresample/arm/audio_convert_init.c+9−0
@@ -28,6 +28,7 @@ · 27 unmodified lines
2828
2929 void swri_oldapi_conv_flt_to_s16_neon(int16_t *dst, const float *src, int len);
3030 void swri_oldapi_conv_fltp_to_s16_2ch_neon(int16_t *dst, float *const *src, int len, int channels);
31+void swri_oldapi_conv_fltp_to_s16_nch_neon(int16_t *dst, float *const *src, int len, int channels);
3132
3233 static void conv_flt_to_s16_neon(uint8_t **dst, const uint8_t **src, int len){
3334 swri_oldapi_conv_flt_to_s16_neon((int16_t*)*dst, (const float*)*src, len);
@@ -37,6 +38,12 @@ static void conv_fltp_to_s16_2ch_neon(uint8_t **dst, const uint8_t **src, int le · 3 unmodified lines
3738 swri_oldapi_conv_fltp_to_s16_2ch_neon((int16_t*)*dst, (float *const*)src, len, 2);
3839 }
3940
41+static void conv_fltp_to_s16_nch_neon(uint8_t **dst, const uint8_t **src, int len){
42+ int channels;
43+ for(channels=3; channels<SWR_CH_MAX && src[channels]; channels++)
44+ ;
45+ swri_oldapi_conv_fltp_to_s16_nch_neon((int16_t*)*dst, (float *const*)src, len, channels);
46+}
4047
4148 av_cold void swri_audio_convert_init_arm(struct AudioConvert *ac,
4249 enum AVSampleFormat out_fmt,
@@ -52,5 +59,7 @@ av_cold void swri_audio_convert_init_arm(struct AudioConvert *ac, · 9 unmodified lines
5259 ac->simd_f = conv_flt_to_s16_neon;
5360 if(out_fmt == AV_SAMPLE_FMT_S16 && in_fmt == AV_SAMPLE_FMT_FLTP && channels == 2)
5461 ac->simd_f = conv_fltp_to_s16_2ch_neon;
62+ if(out_fmt == AV_SAMPLE_FMT_S16 && in_fmt == AV_SAMPLE_FMT_FLTP && channels > 2)
63+ ac->simd_f = conv_fltp_to_s16_nch_neon;
5564 }
5665 }
libswresample/arm/audio_convert_neon.S+233−0
@@ -128,3 +128,236 @@ function swri_oldapi_conv_fltp_to_s16_2ch_neon, export=1 · 127 unmodified lines
128128 vst1.16 {q10-q11},[r0,:128]!
129129 bx lr
130130 endfunc
131+
132+function swri_oldapi_conv_fltp_to_s16_nch_neon, export=1
133+ cmp r3, #2
134+ itt lt
135+ ldrlt r1, [r1]
136+ blt swri_oldapi_conv_flt_to_s16_neon
137+ beq swri_oldapi_conv_fltp_to_s16_2ch_neon
138+
139+ push {r4-r8, lr}
140+ cmp r3, #4
141+ lsl r12, r3, #1
142+ blt 4f
143+
144+ @ 4 channels
145+5: ldm r1!, {r4-r7}
146+ mov lr, r2
147+ mov r8, r0
148+ vld1.32 {q8}, [r4,:128]!
149+ vcvt.s32.f32 q8, q8, #31
150+ vld1.32 {q9}, [r5,:128]!
151+ vcvt.s32.f32 q9, q9, #31
152+ vld1.32 {q10}, [r6,:128]!
153+ vcvt.s32.f32 q10, q10, #31
154+ vld1.32 {q11}, [r7,:128]!
155+ vcvt.s32.f32 q11, q11, #31
156+6: subs lr, lr, #8
157+ vld1.32 {q0}, [r4,:128]!
158+ vcvt.s32.f32 q0, q0, #31
159+ vsri.32 q9, q8, #16
160+ vld1.32 {q1}, [r5,:128]!
161+ vcvt.s32.f32 q1, q1, #31
162+ vsri.32 q11, q10, #16
163+ vld1.32 {q2}, [r6,:128]!
164+ vcvt.s32.f32 q2, q2, #31
165+ vzip.32 d18, d22
166+ vld1.32 {q3}, [r7,:128]!
167+ vcvt.s32.f32 q3, q3, #31
168+ vzip.32 d19, d23
169+ vst1.16 {d18}, [r8], r12
170+ vsri.32 q1, q0, #16
171+ vst1.16 {d22}, [r8], r12
172+ vsri.32 q3, q2, #16
173+ vst1.16 {d19}, [r8], r12
174+ vzip.32 d2, d6
175+ vst1.16 {d23}, [r8], r12
176+ vzip.32 d3, d7
177+ beq 7f
178+ vld1.32 {q8}, [r4,:128]!
179+ vcvt.s32.f32 q8, q8, #31
180+ vst1.16 {d2}, [r8], r12
181+ vld1.32 {q9}, [r5,:128]!
182+ vcvt.s32.f32 q9, q9, #31
183+ vst1.16 {d6}, [r8], r12
184+ vld1.32 {q10}, [r6,:128]!
185+ vcvt.s32.f32 q10, q10, #31
186+ vst1.16 {d3}, [r8], r12
187+ vld1.32 {q11}, [r7,:128]!
188+ vcvt.s32.f32 q11, q11, #31
189+ vst1.16 {d7}, [r8], r12
190+ b 6b
191+7: vst1.16 {d2}, [r8], r12
192+ vst1.16 {d6}, [r8], r12
193+ vst1.16 {d3}, [r8], r12
194+ vst1.16 {d7}, [r8], r12
195+ subs r3, r3, #4
196+ it eq
197+ popeq {r4-r8, pc}
198+ cmp r3, #4
199+ add r0, r0, #8
200+ bge 5b
201+
202+ @ 2 channels
203+4: cmp r3, #2
204+ blt 4f
205+ ldm r1!, {r4-r5}
206+ mov lr, r2
207+ mov r8, r0
208+ tst lr, #8
209+ vld1.32 {q8}, [r4,:128]!
210+ vcvt.s32.f32 q8, q8, #31
211+ vld1.32 {q9}, [r5,:128]!
212+ vcvt.s32.f32 q9, q9, #31
213+ vld1.32 {q10}, [r4,:128]!
214+ vcvt.s32.f32 q10, q10, #31
215+ vld1.32 {q11}, [r5,:128]!
216+ vcvt.s32.f32 q11, q11, #31
217+ beq 6f
218+ subs lr, lr, #8
219+ beq 7f
220+ vsri.32 d18, d16, #16
221+ vsri.32 d19, d17, #16
222+ vld1.32 {q8}, [r4,:128]!
223+ vcvt.s32.f32 q8, q8, #31
224+ vst1.32 {d18[0]}, [r8], r12
225+ vsri.32 d22, d20, #16
226+ vst1.32 {d18[1]}, [r8], r12
227+ vsri.32 d23, d21, #16
228+ vst1.32 {d19[0]}, [r8], r12
229+ vst1.32 {d19[1]}, [r8], r12
230+ vld1.32 {q9}, [r5,:128]!
231+ vcvt.s32.f32 q9, q9, #31
232+ vst1.32 {d22[0]}, [r8], r12
233+ vst1.32 {d22[1]}, [r8], r12
234+ vld1.32 {q10}, [r4,:128]!
235+ vcvt.s32.f32 q10, q10, #31
236+ vst1.32 {d23[0]}, [r8], r12
237+ vst1.32 {d23[1]}, [r8], r12
238+ vld1.32 {q11}, [r5,:128]!
239+ vcvt.s32.f32 q11, q11, #31
240+6: subs lr, lr, #16
241+ vld1.32 {q0}, [r4,:128]!
242+ vcvt.s32.f32 q0, q0, #31
243+ vsri.32 d18, d16, #16
244+ vld1.32 {q1}, [r5,:128]!
245+ vcvt.s32.f32 q1, q1, #31
246+ vsri.32 d19, d17, #16
247+ vld1.32 {q2}, [r4,:128]!
248+ vcvt.s32.f32 q2, q2, #31
249+ vld1.32 {q3}, [r5,:128]!
250+ vcvt.s32.f32 q3, q3, #31
251+ vst1.32 {d18[0]}, [r8], r12
252+ vsri.32 d22, d20, #16
253+ vst1.32 {d18[1]}, [r8], r12
254+ vsri.32 d23, d21, #16
255+ vst1.32 {d19[0]}, [r8], r12
256+ vsri.32 d2, d0, #16
257+ vst1.32 {d19[1]}, [r8], r12
258+ vsri.32 d3, d1, #16
259+ vst1.32 {d22[0]}, [r8], r12
260+ vsri.32 d6, d4, #16
261+ vst1.32 {d22[1]}, [r8], r12
262+ vsri.32 d7, d5, #16
263+ vst1.32 {d23[0]}, [r8], r12
264+ vst1.32 {d23[1]}, [r8], r12
265+ beq 6f
266+ vld1.32 {q8}, [r4,:128]!
267+ vcvt.s32.f32 q8, q8, #31
268+ vst1.32 {d2[0]}, [r8], r12
269+ vst1.32 {d2[1]}, [r8], r12
270+ vld1.32 {q9}, [r5,:128]!
271+ vcvt.s32.f32 q9, q9, #31
272+ vst1.32 {d3[0]}, [r8], r12
273+ vst1.32 {d3[1]}, [r8], r12
274+ vld1.32 {q10}, [r4,:128]!
275+ vcvt.s32.f32 q10, q10, #31
276+ vst1.32 {d6[0]}, [r8], r12
277+ vst1.32 {d6[1]}, [r8], r12
278+ vld1.32 {q11}, [r5,:128]!
279+ vcvt.s32.f32 q11, q11, #31
280+ vst1.32 {d7[0]}, [r8], r12
281+ vst1.32 {d7[1]}, [r8], r12
282+ bgt 6b
283+6: vst1.32 {d2[0]}, [r8], r12
284+ vst1.32 {d2[1]}, [r8], r12
285+ vst1.32 {d3[0]}, [r8], r12
286+ vst1.32 {d3[1]}, [r8], r12
287+ vst1.32 {d6[0]}, [r8], r12
288+ vst1.32 {d6[1]}, [r8], r12
289+ vst1.32 {d7[0]}, [r8], r12
290+ vst1.32 {d7[1]}, [r8], r12
291+ b 8f
292+7: vsri.32 d18, d16, #16
293+ vsri.32 d19, d17, #16
294+ vst1.32 {d18[0]}, [r8], r12
295+ vsri.32 d22, d20, #16
296+ vst1.32 {d18[1]}, [r8], r12
297+ vsri.32 d23, d21, #16
298+ vst1.32 {d19[0]}, [r8], r12
299+ vst1.32 {d19[1]}, [r8], r12
300+ vst1.32 {d22[0]}, [r8], r12
301+ vst1.32 {d22[1]}, [r8], r12
302+ vst1.32 {d23[0]}, [r8], r12
303+ vst1.32 {d23[1]}, [r8], r12
304+8: subs r3, r3, #2
305+ add r0, r0, #4
306+ it eq
307+ popeq {r4-r8, pc}
308+
309+ @ 1 channel
310+4: ldr r4, [r1]
311+ tst r2, #8
312+ mov lr, r2
313+ mov r5, r0
314+ vld1.32 {q0}, [r4,:128]!
315+ vcvt.s32.f32 q0, q0, #31
316+ vld1.32 {q1}, [r4,:128]!
317+ vcvt.s32.f32 q1, q1, #31
318+ bne 8f
319+6: subs lr, lr, #16
320+ vld1.32 {q2}, [r4,:128]!
321+ vcvt.s32.f32 q2, q2, #31
322+ vld1.32 {q3}, [r4,:128]!
323+ vcvt.s32.f32 q3, q3, #31
324+ vst1.16 {d0[1]}, [r5,:16], r12
325+ vst1.16 {d0[3]}, [r5,:16], r12
326+ vst1.16 {d1[1]}, [r5,:16], r12
327+ vst1.16 {d1[3]}, [r5,:16], r12
328+ vst1.16 {d2[1]}, [r5,:16], r12
329+ vst1.16 {d2[3]}, [r5,:16], r12
330+ vst1.16 {d3[1]}, [r5,:16], r12
331+ vst1.16 {d3[3]}, [r5,:16], r12
332+ beq 7f
333+ vld1.32 {q0}, [r4,:128]!
334+ vcvt.s32.f32 q0, q0, #31
335+ vld1.32 {q1}, [r4,:128]!
336+ vcvt.s32.f32 q1, q1, #31
337+7: vst1.16 {d4[1]}, [r5,:16], r12
338+ vst1.16 {d4[3]}, [r5,:16], r12
339+ vst1.16 {d5[1]}, [r5,:16], r12
340+ vst1.16 {d5[3]}, [r5,:16], r12
341+ vst1.16 {d6[1]}, [r5,:16], r12
342+ vst1.16 {d6[3]}, [r5,:16], r12
343+ vst1.16 {d7[1]}, [r5,:16], r12
344+ vst1.16 {d7[3]}, [r5,:16], r12
345+ bgt 6b
346+ pop {r4-r8, pc}
347+8: subs lr, lr, #8
348+ vst1.16 {d0[1]}, [r5,:16], r12
349+ vst1.16 {d0[3]}, [r5,:16], r12
350+ vst1.16 {d1[1]}, [r5,:16], r12
351+ vst1.16 {d1[3]}, [r5,:16], r12
352+ vst1.16 {d2[1]}, [r5,:16], r12
353+ vst1.16 {d2[3]}, [r5,:16], r12
354+ vst1.16 {d3[1]}, [r5,:16], r12
355+ vst1.16 {d3[3]}, [r5,:16], r12
356+ it eq
357+ popeq {r4-r8, pc}
358+ vld1.32 {q0}, [r4,:128]!
359+ vcvt.s32.f32 q0, q0, #31
360+ vld1.32 {q1}, [r4,:128]!
361+ vcvt.s32.f32 q1, q1, #31
362+ b 6b
363+endfunc

Keyboard shortcuts

General

Show keyboard shortcuts
?
Focus the current search
/

Issues and pull requests

Open filters
f

Issues, pull requests and Inbox

Select next item
j↓
Select previous item
k↑
Open selected item
Enter
Dismiss filters or clear selection
Esc

Repository code

Go to file
t