[FFmpeg-devel] [PATCH/RFC]Allow compilation without HAVE_6REGS
Carl Eugen Hoyos
cehoyos at ag.or.at
Sat Apr 19 00:18:34 CEST 2014
On Friday 18 April 2014 11:18:00 pm Michael Niedermayer wrote:
> On Fri, Apr 18, 2014 at 08:00:32PM +0000, Carl Eugen Hoyos wrote:
> > Michael Niedermayer <michaelni <at> gmx.at> writes:
> > > On Fri, Apr 18, 2014 at 02:48:00PM +0000, Carl Eugen Hoyos wrote:
> > > > Carl Eugen Hoyos <cehoyos <at> ag.or.at> writes:
> > > > > Attached patch is needed here for compilation without HAVE_6REGS
> > >
> > > should be ok if these all use 6 or more registers
> >
> > I am not sure which parameters must be counted:
>
> [...]
>
> > libavcodec/x86/cavsdsp.c QPEL_CAVS():
> > : "+a"(src), "+c"(dst), "+m"(h)
> > : "d"((x86_reg)srcStride), "S"((x86_reg)dstStride),
> >
> > "m"(ff_pw_5), "m"(ff_pw_4)
> >
> > : "memory"
>
> thats not 6
New patch attached.
> you could try to access the ff_pw* through
> MANGLE & NAMED_CONSTRAINTS_ADD
Patch sent.
Thank you, Carl Eugen
-------------- next part --------------
diff --git a/libavcodec/x86/mathops.h b/libavcodec/x86/mathops.h
index 1f049c0..1cca05d 100644
--- a/libavcodec/x86/mathops.h
+++ b/libavcodec/x86/mathops.h
@@ -24,6 +24,7 @@
#include "config.h"
#include "libavutil/common.h"
+#include "libavutil/x86/asm.h"
#if HAVE_INLINE_ASM
@@ -88,6 +89,7 @@ static inline av_const int mid_pred(int a, int b, int c)
return i;
}
+#if HAVE_6REGS
#define COPY3_IF_LT(x, y, a, b, c, d)\
__asm__ volatile(\
"cmpl %0, %3 \n\t"\
@@ -97,6 +99,7 @@ __asm__ volatile(\
: "+&r" (x), "+&r" (a), "+r" (c)\
: "r" (y), "r" (b), "r" (d)\
);
+#endif /* HAVE_6REGS */
#endif /* HAVE_I686 */
#define MASK_ABS(mask, level) \
diff --git a/libavcodec/x86/mpegaudiodsp.c b/libavcodec/x86/mpegaudiodsp.c
index 5dc7bc6..2edf157 100644
--- a/libavcodec/x86/mpegaudiodsp.c
+++ b/libavcodec/x86/mpegaudiodsp.c
@@ -45,7 +45,7 @@ void ff_four_imdct36_float_avx(float *out, float *buf, float *in, float *win,
DECLARE_ALIGNED(16, static float, mdct_win_sse)[2][4][4*40];
-#if HAVE_SSE2_INLINE
+#if HAVE_6REGS && HAVE_SSE2_INLINE
#define MACS(rt, ra, rb) rt+=(ra)*(rb)
#define MLSS(rt, ra, rb) rt-=(ra)*(rb)
@@ -189,7 +189,7 @@ static void apply_window_mp3(float *in, float *win, int *unused, float *out,
*out = sum;
}
-#endif /* HAVE_SSE2_INLINE */
+#endif /* HAVE_6REGS && HAVE_SSE2_INLINE */
#if HAVE_YASM
#define DECL_IMDCT_BLOCKS(CPU1, CPU2) \
@@ -255,7 +255,7 @@ av_cold void ff_mpadsp_init_x86(MPADSPContext *s)
}
}
-#if HAVE_SSE2_INLINE
+#if HAVE_6REGS && HAVE_SSE2_INLINE
if (INLINE_SSE2(cpu_flags)) {
s->apply_window_float = apply_window_mp3;
}
diff --git a/libavcodec/x86/mpegvideoenc.c b/libavcodec/x86/mpegvideoenc.c
index e24e5a6..f14a268 100644
--- a/libavcodec/x86/mpegvideoenc.c
+++ b/libavcodec/x86/mpegvideoenc.c
@@ -30,6 +30,8 @@
/* not permutated inverse zigzag_direct + 1 for MMX quantizer */
DECLARE_ALIGNED(16, static uint16_t, inv_zigzag_direct16)[64];
+#if HAVE_6REGS
+
#if HAVE_MMX_INLINE
#define COMPILE_TEMPLATE_MMXEXT 0
#define COMPILE_TEMPLATE_SSE2 0
@@ -81,6 +83,8 @@ DECLARE_ALIGNED(16, static uint16_t, inv_zigzag_direct16)[64];
#include "mpegvideoenc_template.c"
#endif /* HAVE_SSSE3_INLINE */
+#endif /* HAVE_6REGS */
+
#if HAVE_INLINE_ASM
static void denoise_dct_mmx(MpegEncContext *s, int16_t *block){
const int intra= s->mb_intra;
@@ -205,21 +209,25 @@ av_cold void ff_dct_encode_init_x86(MpegEncContext *s)
#if HAVE_MMX_INLINE
int cpu_flags = av_get_cpu_flags();
if (INLINE_MMX(cpu_flags)) {
+#if HAVE_6REGS
s->dct_quantize = dct_quantize_MMX;
+#endif
s->denoise_dct = denoise_dct_mmx;
}
#endif
-#if HAVE_MMXEXT_INLINE
+#if HAVE_6REGS && HAVE_MMXEXT_INLINE
if (INLINE_MMXEXT(cpu_flags))
s->dct_quantize = dct_quantize_MMXEXT;
#endif
#if HAVE_SSE2_INLINE
if (INLINE_SSE2(cpu_flags)) {
+#if HAVE_6REGS
s->dct_quantize = dct_quantize_SSE2;
+#endif
s->denoise_dct = denoise_dct_sse2;
}
#endif
-#if HAVE_SSSE3_INLINE
+#if HAVE_6REGS && HAVE_SSSE3_INLINE
if (INLINE_SSSE3(cpu_flags))
s->dct_quantize = dct_quantize_SSSE3;
#endif
diff --git a/libavcodec/x86/snowdsp.c b/libavcodec/x86/snowdsp.c
index 711af5f..67cb868 100644
--- a/libavcodec/x86/snowdsp.c
+++ b/libavcodec/x86/snowdsp.c
@@ -606,6 +606,7 @@ static void ff_snow_vertical_compose97i_mmx(IDWTELEM *b0, IDWTELEM *b1, IDWTELEM
}
#endif //HAVE_7REGS
+#if HAVE_6REGS
#define snow_inner_add_yblock_sse2_header \
IDWTELEM * * dst_array = sb->line + src_y;\
x86_reg tmp;\
@@ -873,6 +874,7 @@ static void ff_snow_inner_add_yblock_mmx(const uint8_t *obmc, const int obmc_str
else
ff_snow_inner_add_yblock(obmc, obmc_stride, block, b_w, b_h, src_x,src_y, src_stride, sb, add, dst8);
}
+#endif /* HAVE_6REGS */
#endif /* HAVE_INLINE_ASM */
@@ -887,7 +889,9 @@ void ff_dwt_init_x86(SnowDWTContext *c)
#if HAVE_7REGS
c->vertical_compose97i = ff_snow_vertical_compose97i_sse2;
#endif
+#if HAVE_6REGS
c->inner_add_yblock = ff_snow_inner_add_yblock_sse2;
+#endif
}
else{
if (mm_flags & AV_CPU_FLAG_MMXEXT) {
@@ -896,7 +900,9 @@ void ff_dwt_init_x86(SnowDWTContext *c)
c->vertical_compose97i = ff_snow_vertical_compose97i_mmx;
#endif
}
+#if HAVE_6REGS
c->inner_add_yblock = ff_snow_inner_add_yblock_mmx;
+#endif
}
}
#endif /* HAVE_INLINE_ASM */
diff --git a/libavcodec/x86/vc1dsp_init.c b/libavcodec/x86/vc1dsp_init.c
index aff4b26..9256be3 100644
--- a/libavcodec/x86/vc1dsp_init.c
+++ b/libavcodec/x86/vc1dsp_init.c
@@ -27,6 +27,7 @@
#include "libavutil/attributes.h"
#include "libavutil/cpu.h"
#include "libavutil/x86/cpu.h"
+#include "libavutil/x86/asm.h"
#include "libavcodec/vc1dsp.h"
#include "fpel.h"
#include "vc1dsp.h"
@@ -86,10 +87,10 @@ av_cold void ff_vc1dsp_init_x86(VC1DSPContext *dsp)
{
int cpu_flags = av_get_cpu_flags();
- if (INLINE_MMX(cpu_flags))
+ if (HAVE_6REGS && INLINE_MMX(cpu_flags))
ff_vc1dsp_init_mmx(dsp);
- if (INLINE_MMXEXT(cpu_flags))
+ if (HAVE_6REGS && INLINE_MMXEXT(cpu_flags))
ff_vc1dsp_init_mmxext(dsp);
#define ASSIGN_LF(EXT) \
diff --git a/libavcodec/x86/vc1dsp_mmx.c b/libavcodec/x86/vc1dsp_mmx.c
index 5b71de3..4867795 100644
--- a/libavcodec/x86/vc1dsp_mmx.c
+++ b/libavcodec/x86/vc1dsp_mmx.c
@@ -33,7 +33,7 @@
#include "fpel.h"
#include "vc1dsp.h"
-#if HAVE_INLINE_ASM
+#if HAVE_6REGS && HAVE_INLINE_ASM
#define OP_PUT(S,D)
#define OP_AVG(S,D) "pavgb " #S ", " #D " \n\t"
@@ -760,4 +762,4 @@ av_cold void ff_vc1dsp_init_mmxext(VC1DSPContext *dsp)
dsp->vc1_inv_trans_8x4_dc = vc1_inv_trans_8x4_dc_mmxext;
dsp->vc1_inv_trans_4x4_dc = vc1_inv_trans_4x4_dc_mmxext;
}
-#endif /* HAVE_INLINE_ASM */
+#endif /* HAVE_6REGS && HAVE_INLINE_ASM */
diff --git a/libavcodec/x86/vp3dsp_init.c b/libavcodec/x86/vp3dsp_init.c
index 709e49a..cc3eba4 100644
--- a/libavcodec/x86/vp3dsp_init.c
+++ b/libavcodec/x86/vp3dsp_init.c
@@ -63,6 +63,7 @@ void ff_vp3_h_loop_filter_mmxext(uint8_t *src, int stride,
"paddb "#regb", "#regr" \n\t" \
"paddb "#regd", "#regp" \n\t"
+#if HAVE_6REGS
static void put_vp_no_rnd_pixels8_l2_mmx(uint8_t *dst, const uint8_t *a, const uint8_t *b, ptrdiff_t stride, int h)
{
// START_TIMER
@@ -94,15 +95,16 @@ static void put_vp_no_rnd_pixels8_l2_mmx(uint8_t *dst, const uint8_t *a, const u
:"memory");
// STOP_TIMER("put_vp_no_rnd_pixels8_l2_mmx")
}
+#endif /*HAVE_6REGS */
#endif /* HAVE_MMX_INLINE */
av_cold void ff_vp3dsp_init_x86(VP3DSPContext *c, int flags)
{
int cpu_flags = av_get_cpu_flags();
-#if HAVE_MMX_INLINE
+#if HAVE_6REGS && HAVE_MMX_INLINE
c->put_no_rnd_pixels_l2 = put_vp_no_rnd_pixels8_l2_mmx;
-#endif /* HAVE_MMX_INLINE */
+#endif /* HAVE_6REGS && HAVE_MMX_INLINE */
#if ARCH_X86_32
if (EXTERNAL_MMX(cpu_flags)) {
diff --git a/libavcodec/x86/vp56_arith.h b/libavcodec/x86/vp56_arith.h
index e560cd4..810cc8d 100644
--- a/libavcodec/x86/vp56_arith.h
+++ b/libavcodec/x86/vp56_arith.h
@@ -24,7 +24,7 @@
#ifndef AVCODEC_X86_VP56_ARITH_H
#define AVCODEC_X86_VP56_ARITH_H
-#if HAVE_INLINE_ASM && HAVE_FAST_CMOV
+#if HAVE_INLINE_ASM && HAVE_FAST_CMOV && HAVE_6REGS
#define vp56_rac_get_prob vp56_rac_get_prob
static av_always_inline int vp56_rac_get_prob(VP56RangeCoder *c, uint8_t prob)
{
diff --git a/libavfilter/vf_noise.c b/libavfilter/vf_noise.c
index c29afa2..07d1c27 100644
--- a/libavfilter/vf_noise.c
+++ b/libavfilter/vf_noise.c
@@ -289,7 +289,7 @@ static inline void line_noise_avg_c(uint8_t *dst, const uint8_t *src,
static inline void line_noise_avg_mmx(uint8_t *dst, const uint8_t *src,
int len, int8_t **shift)
{
-#if HAVE_MMX_INLINE
+#if HAVE_MMX_INLINE && HAVE_6REGS
x86_reg mmx_len= len&(~7);
__asm__ volatile(
@@ -438,7 +438,9 @@ static av_cold int init(AVFilterContext *ctx)
if (HAVE_MMX_INLINE &&
cpu_flags & AV_CPU_FLAG_MMX) {
n->line_noise = line_noise_mmx;
+#if HAVE_6REGS
n->line_noise_avg = line_noise_avg_mmx;
+#endif
}
if (HAVE_MMXEXT_INLINE &&
cpu_flags & AV_CPU_FLAG_MMXEXT)
diff --git a/libswscale/x86/swscale_template.c b/libswscale/x86/swscale_template.c
index a2f0d6e..1213197 100644
--- a/libswscale/x86/swscale_template.c
+++ b/libswscale/x86/swscale_template.c
@@ -666,6 +666,7 @@ static void RENAME(yuv2rgb555_X)(SwsContext *c, const int16_t *lumFilter,
#define WRITEBGR24(dst, dstw, index) WRITEBGR24MMX(dst, dstw, index)
#endif
+#if HAVE_6REGS
static void RENAME(yuv2bgr24_X_ar)(SwsContext *c, const int16_t *lumFilter,
const int16_t **lumSrc, int lumFilterSize,
const int16_t *chrFilter, const int16_t **chrUSrc,
@@ -715,6 +716,7 @@ static void RENAME(yuv2bgr24_X)(SwsContext *c, const int16_t *lumFilter,
: "%"REG_a, "%"REG_c, "%"REG_d, "%"REG_S
);
}
+#endif /* HAVE_6REGS */
#define REAL_WRITEYUY2(dst, dstw, index) \
"packuswb %%mm3, %%mm3 \n\t"\
@@ -1664,7 +1666,9 @@ static av_cold void RENAME(sws_init_swscale)(SwsContext *c)
if (!(c->flags & SWS_FULL_CHR_H_INT)) {
switch (c->dstFormat) {
case AV_PIX_FMT_RGB32: c->yuv2packedX = RENAME(yuv2rgb32_X_ar); break;
+#if HAVE_6REGS
case AV_PIX_FMT_BGR24: c->yuv2packedX = RENAME(yuv2bgr24_X_ar); break;
+#endif
case AV_PIX_FMT_RGB555: c->yuv2packedX = RENAME(yuv2rgb555_X_ar); break;
case AV_PIX_FMT_RGB565: c->yuv2packedX = RENAME(yuv2rgb565_X_ar); break;
case AV_PIX_FMT_YUYV422: c->yuv2packedX = RENAME(yuv2yuyv422_X_ar); break;
@@ -1677,7 +1681,9 @@ static av_cold void RENAME(sws_init_swscale)(SwsContext *c)
if (!(c->flags & SWS_FULL_CHR_H_INT)) {
switch (c->dstFormat) {
case AV_PIX_FMT_RGB32: c->yuv2packedX = RENAME(yuv2rgb32_X); break;
+#if HAVE_6REGS
case AV_PIX_FMT_BGR24: c->yuv2packedX = RENAME(yuv2bgr24_X); break;
+#endif
case AV_PIX_FMT_RGB555: c->yuv2packedX = RENAME(yuv2rgb555_X); break;
case AV_PIX_FMT_RGB565: c->yuv2packedX = RENAME(yuv2rgb565_X); break;
case AV_PIX_FMT_YUYV422: c->yuv2packedX = RENAME(yuv2yuyv422_X); break;
diff --git a/libswscale/x86/yuv2rgb.c b/libswscale/x86/yuv2rgb.c
index a3370ee..5e2f77c 100644
--- a/libswscale/x86/yuv2rgb.c
+++ b/libswscale/x86/yuv2rgb.c
@@ -50,28 +50,28 @@ DECLARE_ASM_CONST(8, uint64_t, pb_03) = 0x0303030303030303ULL;
DECLARE_ASM_CONST(8, uint64_t, pb_07) = 0x0707070707070707ULL;
//MMX versions
-#if HAVE_MMX_INLINE
+#if HAVE_MMX_INLINE && HAVE_6REGS
#undef RENAME
#undef COMPILE_TEMPLATE_MMXEXT
#define COMPILE_TEMPLATE_MMXEXT 0
#define RENAME(a) a ## _mmx
#include "yuv2rgb_template.c"
-#endif /* HAVE_MMX_INLINE */
+#endif /* HAVE_MMX_INLINE && HAVE_6REGS */
// MMXEXT versions
-#if HAVE_MMXEXT_INLINE
+#if HAVE_MMXEXT_INLINE && HAVE_6REGS
#undef RENAME
#undef COMPILE_TEMPLATE_MMXEXT
#define COMPILE_TEMPLATE_MMXEXT 1
#define RENAME(a) a ## _mmxext
#include "yuv2rgb_template.c"
-#endif /* HAVE_MMXEXT_INLINE */
+#endif /* HAVE_MMXEXT_INLINE && HAVE_6REGS */
#endif /* HAVE_INLINE_ASM */
av_cold SwsFunc ff_yuv2rgb_init_x86(SwsContext *c)
{
-#if HAVE_MMX_INLINE
+#if HAVE_MMX_INLINE && HAVE_6REGS
int cpu_flags = av_get_cpu_flags();
#if HAVE_MMXEXT_INLINE
@@ -113,7 +113,7 @@ av_cold SwsFunc ff_yuv2rgb_init_x86(SwsContext *c)
return yuv420_rgb15_mmx;
}
}
-#endif /* HAVE_MMX_INLINE */
+#endif /* HAVE_MMX_INLINE && HAVE_6REGS */
return NULL;
}
More information about the ffmpeg-devel
mailing list