swr: replace planar->planar/packed->packed FLT<->S16/S32 SIMD by new macros

this simplifies the code

Michael Niedermayer committed 14 years ago

Commit aae31191 parent 47055b81 file · +47 −145Browse files

Unified · Split

libswresample/x86/audio_convert.asm+47−145

@@ -62,142 +62,7 @@ int16_to_int32_u_int %+ SUFFIX · 61 unmodified lines
62	62	REP_RET
63	63	%endmacro
64	64
65		−%macro INT32_TO_FLOAT 1
66		−cglobal int32_to_float_%1, 3, 3, 3, dst, src, len
67		− mov srcq, [srcq]
68		− mov dstq, [dstq]
69		− shl lenq, 2
70		−%ifidn %1, a
71		− test dstq, mmsize-1
72		− jne int32_to_float_u_int %+ SUFFIX
73		− test srcq, mmsize-1
74		− jne int32_to_float_u_int %+ SUFFIX
75		−%else
76		−int32_to_float_u_int %+ SUFFIX
77		−%endif
78		− add srcq, lenq
79		− add dstq, lenq
80		− neg lenq
81		− mova m2, [flt2pm31]
82		−.next:
83		−%ifidn %1, a
84		− cvtdq2ps m0, [ srcq+lenq]
85		− cvtdq2ps m1, [mmsize + srcq+lenq]
86		−%else
87		− movu m0, [ srcq+lenq]
88		− movu m1, [mmsize + srcq+lenq]
89		− cvtdq2ps m0, m0
90		− cvtdq2ps m1, m1
91		−%endif
92		− mulps m0, m0, m2
93		− mulps m1, m1, m2
94		− mov%1 [ dstq+lenq], m0
95		− mov%1 [mmsize + dstq+lenq], m1
96		− add lenq, 2*mmsize
97		− jl .next
98		− REP_RET
99		−%endmacro
100	65
101		−%macro INT16_TO_FLOAT 1
102		−cglobal int16_to_float_%1, 3, 3, 4, dst, src, len
103		− mov srcq, [srcq]
104		− mov dstq, [dstq]
105		− shl lenq, 2
106		−%ifidn %1, a
107		− test dstq, mmsize-1
108		− jne int16_to_float_u_int %+ SUFFIX
109		− test srcq, mmsize-1
110		− jne int16_to_float_u_int %+ SUFFIX
111		−%else
112		−int16_to_float_u_int %+ SUFFIX
113		−%endif
114		− add dstq, lenq
115		− shr lenq, 1
116		− add srcq, lenq
117		− neg lenq
118		− mova m3, [flt2pm31]
119		−.next:
120		− mov%1 m2, [srcq+lenq]
121		− pxor m0, m0
122		− pxor m1, m1
123		− punpcklwd m0, m2
124		− punpckhwd m1, m2
125		− cvtdq2ps m0, m0
126		− cvtdq2ps m1, m1
127		− mulps m0, m3
128		− mulps m1, m3
129		− mov%1 [ dstq+2*lenq], m0
130		− mov%1 [mmsize + dstq+2*lenq], m1
131		− add lenq, mmsize
132		− jl .next
133		− REP_RET
134		−%endmacro
135		−
136		−%macro FLOAT_TO_INT32 1
137		−cglobal float_to_int32_%1, 3, 3, 5, dst, src, len
138		− mov srcq, [srcq]
139		− mov dstq, [dstq]
140		− shl lenq, 2
141		−%ifidn %1, a
142		− test dstq, mmsize-1
143		− jne float_to_int32_u_int %+ SUFFIX
144		− test srcq, mmsize-1
145		− jne float_to_int32_u_int %+ SUFFIX
146		−%else
147		−float_to_int32_u_int %+ SUFFIX
148		−%endif
149		− add srcq, lenq
150		− add dstq, lenq
151		− neg lenq
152		− mova m2, [flt2p31]
153		−.next:
154		− mov%1 m0, [ srcq+lenq]
155		− mov%1 m1, [mmsize + srcq+lenq]
156		− mulps m0, m2
157		− mulps m1, m2
158		− cvtps2dq m3, m0
159		− cvtps2dq m4, m1
160		− cmpnltps m0, m2
161		− cmpnltps m1, m2
162		− paddd m0, m3
163		− paddd m1, m4
164		− mov%1 [ dstq+lenq], m0
165		− mov%1 [mmsize + dstq+lenq], m1
166		− add lenq, 2*mmsize
167		− jl .next
168		− REP_RET
169		−%endmacro
170		−
171		−%macro FLOAT_TO_INT16 1
172		−cglobal float_to_int16_%1, 3, 3, 3, dst, src, len
173		− mov srcq, [srcq]
174		− mov dstq, [dstq]
175		− add lenq , lenq
176		−%ifidn %1, a
177		− test dstq, mmsize-1
178		− jne float_to_int16_u_int %+ SUFFIX
179		− test srcq, mmsize-1
180		− jne float_to_int16_u_int %+ SUFFIX
181		−%else
182		−float_to_int16_u_int %+ SUFFIX
183		−%endif
184		− lea srcq, [srcq + 2*lenq]
185		− add dstq, lenq
186		− neg lenq
187		− mova m2, [flt2p15]
188		−.next:
189		− mov%1 m0, [ srcq+2*lenq]
190		− mov%1 m1, [mmsize + srcq+2*lenq]
191		− mulps m0, m2
192		− mulps m1, m2
193		− cvtps2dq m0, m0
194		− cvtps2dq m1, m1
195		− packssdw m0, m1
196		− mov%1 [ dstq+lenq], m0
197		− add lenq, mmsize
198		− jl .next
199		− REP_RET
200		−%endmacro
201	66
202	67	%macro INT32_TO_INT16 1
203	68	cglobal int32_to_int16_%1, 3, 3, 2, dst, src, len
@@ -280,6 +145,43 @@ pack_2ch_%1_to_%2_u_int %+ SUFFIX · 76 unmodified lines
280	145	REP_RET
281	146	%endmacro
282	147
	148	+%macro CONV 5-7
	149	+cglobal %2_to_%1_%3, 3, 3, 6, dst, src, len
	150	+ mov srcq , [srcq]
	151	+ mov dstq , [dstq]
	152	+%ifidn %3, a
	153	+ test dstq, mmsize-1
	154	+ jne %2_to_%1_u_int %+ SUFFIX
	155	+ test srcq, mmsize-1
	156	+ jne %2_to_%1_u_int %+ SUFFIX
	157	+%else
	158	+%2_to_%1_u_int %+ SUFFIX
	159	+%endif
	160	+ lea srcq , [srcq + (1<<%5)*lenq]
	161	+ lea dstq , [dstq + (1<<%4)*lenq]
	162	+ neg lenq
	163	+ %7
	164	+.next:
	165	+ mov%3 m0, [ srcq +(1<<%5)*lenq]
	166	+ mov%3 m1, [ mmsize + srcq +(1<<%5)*lenq]
	167	+%if %4 < %5
	168	+ mov%3 m2, [2mmsize + srcq +(1<<%5)lenq]
	169	+ mov%3 m3, [3mmsize + srcq +(1<<%5)lenq]
	170	+%endif
	171	+ %6
	172	+ mov%3 [ dstq+(1<<%4)*lenq], m0
	173	+ mov%3 [ mmsize + dstq+(1<<%4)*lenq], m1
	174	+%if %4 > %5
	175	+ mov%3 [2mmsize + dstq+(1<<%4)lenq], m2
	176	+ mov%3 [3mmsize + dstq+(1<<%4)lenq], m3
	177	+ add lenq, 4*mmsize/(1<<%4)
	178	+%else
	179	+ add lenq, 2*mmsize/(1<<%4)
	180	+%endif
	181	+ jl .next
	182	+ REP_RET
	183	+%endmacro
	184	+
283	185	%macro INT16_TO_INT32_N 0
284	186	pxor m2, m2
285	187	pxor m3, m3
@@ -379,14 +281,14 @@ PACK_2CH int16, int32, u, 1, 2, INT32_TO_INT16_N · 93 unmodified lines
379	281	PACK_2CH int16, int32, a, 1, 2, INT32_TO_INT16_N
380	282
381	283	INIT_XMM sse2
382		−INT32_TO_FLOAT u
383		−INT32_TO_FLOAT a
384		−INT16_TO_FLOAT u
385		−INT16_TO_FLOAT a
386		−FLOAT_TO_INT32 u
387		−FLOAT_TO_INT32 a
388		−FLOAT_TO_INT16 u
389		−FLOAT_TO_INT16 a
	284	+CONV float, int32, u, 2, 2, INT32_TO_FLOAT_N, INT32_TO_FLOAT_INIT
	285	+CONV float, int32, a, 2, 2, INT32_TO_FLOAT_N, INT32_TO_FLOAT_INIT
	286	+CONV int32, float, u, 2, 2, FLOAT_TO_INT32_N, FLOAT_TO_INT32_INIT
	287	+CONV int32, float, a, 2, 2, FLOAT_TO_INT32_N, FLOAT_TO_INT32_INIT
	288	+CONV float, int16, u, 2, 1, INT16_TO_FLOAT_N, INT16_TO_FLOAT_INIT
	289	+CONV float, int16, a, 2, 1, INT16_TO_FLOAT_N, INT16_TO_FLOAT_INIT
	290	+CONV int16, float, u, 1, 2, FLOAT_TO_INT16_N, FLOAT_TO_INT16_INIT
	291	+CONV int16, float, a, 1, 2, FLOAT_TO_INT16_N, FLOAT_TO_INT16_INIT
390	292
391	293	PACK_2CH float, int32, u, 2, 2, INT32_TO_FLOAT_N, INT32_TO_FLOAT_INIT
392	294	PACK_2CH float, int32, a, 2, 2, INT32_TO_FLOAT_N, INT32_TO_FLOAT_INIT
@@ -400,6 +302,6 @@ PACK_2CH int16, float, a, 1, 2, FLOAT_TO_INT16_N, FLOAT_TO_INT16_INIT · 7 unmodified lines
400	302
401	303	%if HAVE_AVX
402	304	INIT_YMM avx
403		−INT32_TO_FLOAT u
404		−INT32_TO_FLOAT a
	305	+CONV float, int32, u, 2, 2, INT32_TO_FLOAT_N, INT32_TO_FLOAT_INIT
	306	+CONV float, int32, a, 2, 2, INT32_TO_FLOAT_N, INT32_TO_FLOAT_INIT
405	307	%endif

swr: replace planar->planar/packed->packed FLT<->S16/S32 SIMD by new macros

Keyboard shortcuts

General

Issues and pull requests

Issues, pull requests and Inbox

Repository code