Currently there is a wild mix of 3dn2/3dnow2/3dnowext. Switching to
"3dnowext", which is the actual name of the CPU flag unifies this.
---
libavcodec/x86/Makefile | 4 +-
libavcodec/x86/dsputil_mmx.c | 14 ++++++------
libavcodec/x86/fft.c | 6 ++--
libavcodec/x86/fft.h | 6 ++--
libavcodec/x86/{fft_3dn.c => fft_3dnow.c} | 2 +-
libavcodec/x86/{fft_3dn2.c => fft_3dnowext.c} | 26 ++++++++++++------------
libavcodec/x86/fft_mmx.asm | 8 +++---
libavcodec/x86/fmtconvert.asm | 6 ++--
libavcodec/x86/fmtconvert_mmx.c | 10 +++++---
libavutil/x86/x86inc.asm | 2 +-
10 files changed, 43 insertions(+), 41 deletions(-)
rename libavcodec/x86/{fft_3dn.c => fft_3dnow.c} (97%)
rename libavcodec/x86/{fft_3dn2.c => fft_3dnowext.c} (85%)
diff --git a/libavcodec/x86/Makefile b/libavcodec/x86/Makefile
index 1d26357..8cd7bb0 100644
--- a/libavcodec/x86/Makefile
+++ b/libavcodec/x86/Makefile
@@ -38,8 +38,8 @@ YASM-OBJS-$(CONFIG_AAC_DECODER) += x86/sbrdsp.o
YASM-OBJS-$(CONFIG_AC3DSP) += x86/ac3dsp.o
YASM-OBJS-$(CONFIG_DCT) += x86/dct32_sse.o
YASM-OBJS-$(CONFIG_ENCODERS) += x86/dsputilenc_yasm.o
-YASM-OBJS-FFT-$(HAVE_AMD3DNOW) += x86/fft_3dn.o
-YASM-OBJS-FFT-$(HAVE_AMD3DNOWEXT) += x86/fft_3dn2.o
+YASM-OBJS-FFT-$(HAVE_AMD3DNOW) += x86/fft_3dnow.o
+YASM-OBJS-FFT-$(HAVE_AMD3DNOWEXT) += x86/fft_3dnowext.o
YASM-OBJS-$(CONFIG_FFT) += x86/fft_mmx.o \
$(YASM-OBJS-FFT-yes)
YASM-OBJS-$(CONFIG_H264CHROMA) += x86/h264_chromamc.o \
diff --git a/libavcodec/x86/dsputil_mmx.c b/libavcodec/x86/dsputil_mmx.c
index afbb531..2663c39 100644
--- a/libavcodec/x86/dsputil_mmx.c
+++ b/libavcodec/x86/dsputil_mmx.c
@@ -2387,9 +2387,9 @@ static void ac3_downmix_sse(float (*samples)[256], float
(*matrix)[2],
}
#if HAVE_6REGS
-static void vector_fmul_window_3dnow2(float *dst, const float *src0,
- const float *src1, const float *win,
- int len)
+static void vector_fmul_window_3dnowext(float *dst, const float *src0,
+ const float *src1, const float *win,
+ int len)
{
x86_reg i = -len * 4;
x86_reg j = len * 4 - 8;
@@ -2838,11 +2838,11 @@ static void dsputil_init_3dnow(DSPContext *c,
AVCodecContext *avctx,
#endif
}
-static void dsputil_init_3dnow2(DSPContext *c, AVCodecContext *avctx,
- int mm_flags)
+static void dsputil_init_3dnowext(DSPContext *c, AVCodecContext *avctx,
+ int mm_flags)
{
#if HAVE_6REGS && HAVE_INLINE_ASM
- c->vector_fmul_window = vector_fmul_window_3dnow2;
+ c->vector_fmul_window = vector_fmul_window_3dnowext;
#endif
}
@@ -3093,7 +3093,7 @@ void ff_dsputil_init_mmx(DSPContext *c, AVCodecContext
*avctx)
dsputil_init_3dnow(c, avctx, mm_flags);
if (mm_flags & AV_CPU_FLAG_3DNOWEXT && HAVE_AMD3DNOWEXT)
- dsputil_init_3dnow2(c, avctx, mm_flags);
+ dsputil_init_3dnowext(c, avctx, mm_flags);
if (mm_flags & AV_CPU_FLAG_SSE && HAVE_SSE)
dsputil_init_sse(c, avctx, mm_flags);
diff --git a/libavcodec/x86/fft.c b/libavcodec/x86/fft.c
index 6349c23..ae7a08e 100644
--- a/libavcodec/x86/fft.c
+++ b/libavcodec/x86/fft.c
@@ -33,9 +33,9 @@ av_cold void ff_fft_init_mmx(FFTContext *s)
}
if (has_vectors & AV_CPU_FLAG_3DNOWEXT && HAVE_AMD3DNOWEXT) {
/* 3DNowEx for K7 */
- s->imdct_calc = ff_imdct_calc_3dnow2;
- s->imdct_half = ff_imdct_half_3dnow2;
- s->fft_calc = ff_fft_calc_3dnow2;
+ s->imdct_calc = ff_imdct_calc_3dnowext;
+ s->imdct_half = ff_imdct_half_3dnowext;
+ s->fft_calc = ff_fft_calc_3dnowext;
}
if (has_vectors & AV_CPU_FLAG_SSE && HAVE_SSE) {
/* SSE for P3/P4/K8 */
diff --git a/libavcodec/x86/fft.h b/libavcodec/x86/fft.h
index 1cefe7a..6e80b95 100644
--- a/libavcodec/x86/fft.h
+++ b/libavcodec/x86/fft.h
@@ -25,12 +25,12 @@ void ff_fft_permute_sse(FFTContext *s, FFTComplex *z);
void ff_fft_calc_avx(FFTContext *s, FFTComplex *z);
void ff_fft_calc_sse(FFTContext *s, FFTComplex *z);
void ff_fft_calc_3dnow(FFTContext *s, FFTComplex *z);
-void ff_fft_calc_3dnow2(FFTContext *s, FFTComplex *z);
+void ff_fft_calc_3dnowext(FFTContext *s, FFTComplex *z);
void ff_imdct_calc_3dnow(FFTContext *s, FFTSample *output, const FFTSample
*input);
void ff_imdct_half_3dnow(FFTContext *s, FFTSample *output, const FFTSample
*input);
-void ff_imdct_calc_3dnow2(FFTContext *s, FFTSample *output, const FFTSample
*input);
-void ff_imdct_half_3dnow2(FFTContext *s, FFTSample *output, const FFTSample
*input);
+void ff_imdct_calc_3dnowext(FFTContext *s, FFTSample *output, const FFTSample
*input);
+void ff_imdct_half_3dnowext(FFTContext *s, FFTSample *output, const FFTSample
*input);
void ff_imdct_calc_sse(FFTContext *s, FFTSample *output, const FFTSample
*input);
void ff_imdct_half_sse(FFTContext *s, FFTSample *output, const FFTSample
*input);
void ff_imdct_half_avx(FFTContext *s, FFTSample *output, const FFTSample
*input);
diff --git a/libavcodec/x86/fft_3dn.c b/libavcodec/x86/fft_3dnow.c
similarity index 97%
rename from libavcodec/x86/fft_3dn.c
rename to libavcodec/x86/fft_3dnow.c
index 5a4d3ad..8778260 100644
--- a/libavcodec/x86/fft_3dn.c
+++ b/libavcodec/x86/fft_3dnow.c
@@ -20,4 +20,4 @@
*/
#define EMULATE_3DNOWEXT
-#include "fft_3dn2.c"
+#include "fft_3dnowext.c"
diff --git a/libavcodec/x86/fft_3dn2.c b/libavcodec/x86/fft_3dnowext.c
similarity index 85%
rename from libavcodec/x86/fft_3dn2.c
rename to libavcodec/x86/fft_3dnowext.c
index e684cc7..47d912d 100644
--- a/libavcodec/x86/fft_3dn2.c
+++ b/libavcodec/x86/fft_3dnowext.c
@@ -30,30 +30,30 @@ DECLARE_ALIGNED(8, static const unsigned int, m1m1)[2] = {
1U<<31, 1U<<31 };
"movq "#s","#d"\n"\
"psrlq $32,"#d"\n"\
"punpckldq "#s","#d"\n"
-#define ff_fft_calc_3dnow2 ff_fft_calc_3dnow
-#define ff_fft_dispatch_3dnow2 ff_fft_dispatch_3dnow
-#define ff_fft_dispatch_interleave_3dnow2 ff_fft_dispatch_interleave_3dnow
-#define ff_imdct_calc_3dnow2 ff_imdct_calc_3dnow
-#define ff_imdct_half_3dnow2 ff_imdct_half_3dnow
+#define ff_fft_calc_3dnowext ff_fft_calc_3dnow
+#define ff_fft_dispatch_3dnowext ff_fft_dispatch_3dnow
+#define ff_fft_dispatch_interleave_3dnowext ff_fft_dispatch_interleave_3dnow
+#define ff_imdct_calc_3dnowext ff_imdct_calc_3dnow
+#define ff_imdct_half_3dnowext ff_imdct_half_3dnow
#else
#define PSWAPD(s,d) "pswapd "#s","#d"\n"
#endif
-void ff_fft_dispatch_3dnow2(FFTComplex *z, int nbits);
-void ff_fft_dispatch_interleave_3dnow2(FFTComplex *z, int nbits);
+void ff_fft_dispatch_3dnowext(FFTComplex *z, int nbits);
+void ff_fft_dispatch_interleave_3dnowext(FFTComplex *z, int nbits);
-void ff_fft_calc_3dnow2(FFTContext *s, FFTComplex *z)
+void ff_fft_calc_3dnowext(FFTContext *s, FFTComplex *z)
{
int n = 1<<s->nbits;
int i;
- ff_fft_dispatch_interleave_3dnow2(z, s->nbits);
+ ff_fft_dispatch_interleave_3dnowext(z, s->nbits);
__asm__ volatile("femms");
if(n <= 8)
for(i=0; i<n; i+=2)
FFSWAP(FFTSample, z[i].im, z[i+1].re);
}
-void ff_imdct_half_3dnow2(FFTContext *s, FFTSample *output, const FFTSample
*input)
+void ff_imdct_half_3dnowext(FFTContext *s, FFTSample *output, const FFTSample
*input)
{
x86_reg j, k;
long n = s->mdct_size;
@@ -101,7 +101,7 @@ void ff_imdct_half_3dnow2(FFTContext *s, FFTSample *output,
const FFTSample *inp
);
}
- ff_fft_dispatch_3dnow2(z, s->nbits);
+ ff_fft_dispatch_3dnowext(z, s->nbits);
#define CMUL(j,mm0,mm1)\
"movq (%2,"#j",2), %%mm6 \n"\
@@ -144,13 +144,13 @@ void ff_imdct_half_3dnow2(FFTContext *s, FFTSample
*output, const FFTSample *inp
__asm__ volatile("femms");
}
-void ff_imdct_calc_3dnow2(FFTContext *s, FFTSample *output, const FFTSample
*input)
+void ff_imdct_calc_3dnowext(FFTContext *s, FFTSample *output, const FFTSample
*input)
{
x86_reg j, k;
long n = s->mdct_size;
long n4 = n >> 2;
- ff_imdct_half_3dnow2(s, output+n4, input);
+ ff_imdct_half_3dnowext(s, output+n4, input);
j = -n;
k = n-8;
diff --git a/libavcodec/x86/fft_mmx.asm b/libavcodec/x86/fft_mmx.asm
index 5c6583b..d39806d 100644
--- a/libavcodec/x86/fft_mmx.asm
+++ b/libavcodec/x86/fft_mmx.asm
@@ -498,7 +498,7 @@ fft8 %+ SUFFIX:
ret
%endmacro
-INIT_MMX 3dnow2
+INIT_MMX 3dnowext
FFT48_3DN
%macro pswapd 2
@@ -695,8 +695,8 @@ INIT_MMX 3dnow
%define unpckhps punpckhdq
DECL_PASS pass_3dnow, PASS_SMALL 1, [wq], [wq+o1q]
DECL_PASS pass_interleave_3dnow, PASS_BIG 0
-%define pass_3dnow2 pass_3dnow
-%define pass_interleave_3dnow2 pass_interleave_3dnow
+%define pass_3dnowext pass_3dnow
+%define pass_interleave_3dnowext pass_interleave_3dnow
%ifdef PIC
%define SECTION_REL - $$
@@ -763,7 +763,7 @@ DECL_FFT 5, _interleave
INIT_MMX 3dnow
DECL_FFT 4
DECL_FFT 4, _interleave
-INIT_MMX 3dnow2
+INIT_MMX 3dnowext
DECL_FFT 4
DECL_FFT 4, _interleave
diff --git a/libavcodec/x86/fmtconvert.asm b/libavcodec/x86/fmtconvert.asm
index 4916e7a..0fd14fe 100644
--- a/libavcodec/x86/fmtconvert.asm
+++ b/libavcodec/x86/fmtconvert.asm
@@ -249,7 +249,7 @@ FLOAT_TO_INT16_INTERLEAVE2 sse2
%macro PSWAPD_SSE 2
pshufw %1, %2, 0x4e
%endmacro
-%macro PSWAPD_3DN1 2
+%macro PSWAPD_3DNOW 2
movq %1, %2
psrlq %1, 32
punpckldq %1, %2
@@ -306,10 +306,10 @@ cglobal float_to_int16_interleave6_%1, 2,8,0, dst, src,
src1, src2, src3, src4,
%define pswapd PSWAPD_SSE
FLOAT_TO_INT16_INTERLEAVE6 sse
%define cvtps2pi pf2id
-%define pswapd PSWAPD_3DN1
+%define pswapd PSWAPD_3DNOW
FLOAT_TO_INT16_INTERLEAVE6 3dnow
%undef pswapd
-FLOAT_TO_INT16_INTERLEAVE6 3dn2
+FLOAT_TO_INT16_INTERLEAVE6 3dnowext
%undef cvtps2pi
;-----------------------------------------------------------------------------
diff --git a/libavcodec/x86/fmtconvert_mmx.c b/libavcodec/x86/fmtconvert_mmx.c
index aaf634d..fbdc526 100644
--- a/libavcodec/x86/fmtconvert_mmx.c
+++ b/libavcodec/x86/fmtconvert_mmx.c
@@ -46,7 +46,7 @@ void ff_float_to_int16_interleave2_sse2 (int16_t *dst, const
float **src, long l
void ff_float_to_int16_interleave6_sse(int16_t *dst, const float **src, int
len);
void ff_float_to_int16_interleave6_3dnow(int16_t *dst, const float **src, int
len);
-void ff_float_to_int16_interleave6_3dn2(int16_t *dst, const float **src, int
len);
+void ff_float_to_int16_interleave6_3dnowext(int16_t *dst, const float **src,
int len);
#define ff_float_to_int16_interleave6_sse2 ff_float_to_int16_interleave6_sse
@@ -74,9 +74,11 @@ FLOAT_TO_INT16_INTERLEAVE(3dnow)
FLOAT_TO_INT16_INTERLEAVE(sse)
FLOAT_TO_INT16_INTERLEAVE(sse2)
-static void float_to_int16_interleave_3dn2(int16_t *dst, const float **src,
long len, int channels){
+static void float_to_int16_interleave_3dnowext(int16_t *dst, const float **src,
+ long len, int channels)
+{
if(channels==6)
- ff_float_to_int16_interleave6_3dn2(dst, src, len);
+ ff_float_to_int16_interleave6_3dnowext(dst, src, len);
else
float_to_int16_interleave_3dnow(dst, src, len, channels);
}
@@ -126,7 +128,7 @@ void ff_fmt_convert_init_x86(FmtConvertContext *c,
AVCodecContext *avctx)
}
if (HAVE_AMD3DNOWEXT && mm_flags & AV_CPU_FLAG_3DNOWEXT) {
if(!(avctx->flags & CODEC_FLAG_BITEXACT)){
- c->float_to_int16_interleave = float_to_int16_interleave_3dn2;
+ c->float_to_int16_interleave =
float_to_int16_interleave_3dnowext;
}
}
if (HAVE_SSE && mm_flags & AV_CPU_FLAG_SSE) {
diff --git a/libavutil/x86/x86inc.asm b/libavutil/x86/x86inc.asm
index 7a75951..03e6c07 100644
--- a/libavutil/x86/x86inc.asm
+++ b/libavutil/x86/x86inc.asm
@@ -557,7 +557,7 @@ SECTION .note.GNU-stack noalloc noexec nowrite progbits
%assign cpuflags_mmx (1<<0)
%assign cpuflags_mmx2 (1<<1) | cpuflags_mmx
%assign cpuflags_3dnow (1<<2) | cpuflags_mmx
-%assign cpuflags_3dnow2 (1<<3) | cpuflags_3dnow
+%assign cpuflags_3dnowext (1<<3) | cpuflags_3dnow
%assign cpuflags_sse (1<<4) | cpuflags_mmx2
%assign cpuflags_sse2 (1<<5) | cpuflags_sse
%assign cpuflags_sse2slow (1<<6) | cpuflags_sse2
--
1.7.2.5
_______________________________________________
libav-devel mailing list
[email protected]
https://lists.libav.org/mailman/listinfo/libav-devel
