May 2019 - MPlayer-cvslog - lists.mplayerhq.hu

r38147 - trunk/libao2/ao_kai.c
by komh 29 May '19

29 May '19

Author: komh Date: Wed May 29 12:57:32 2019 New Revision: 38147 Log: kai: check boundaries more strictly Suggested by reimar. Modified: trunk/libao2/ao_kai.c Modified: trunk/libao2/ao_kai.c ============================================================================== --- trunk/libao2/ao_kai.c Wed May 29 12:56:50 2019 (r38146) +++ trunk/libao2/ao_kai.c Wed May 29 12:57:32 2019 (r38147) @@ -166,7 +166,8 @@ static int init(int rate, int channels, {NULL} }; - const char *audioDriver[] = {"DART", "UNIAUD",}; + const char *audioDriver[] = {"UNKNOWN", "DART", "UNIAUD",}; + int nAudioDriver = sizeof(audioDriver) / sizeof(*audioDriver); if (subopt_parse(ao_subdevice, subopts) != 0) { print_help(); @@ -190,7 +191,7 @@ static int init(int rate, int channels, kaiCaps(&kc); mp_msg(MSGT_AO, MSGL_V, "KAI: selected audio driver = %s\n", - audioDriver[kc.ulMode - 1]); + audioDriver[kc.ulMode >= nAudioDriver ? 0 : kc.ulMode]); mp_msg(MSGT_AO, MSGL_V, "KAI: PDD name = %s, maximum channels = %lu\n", kc.szPDDName, kc.ulMaxChannels);

1 0

r38146 - trunk/libvo/vo_kva.c
by komh 29 May '19

29 May '19

Author: komh Date: Wed May 29 12:56:50 2019 New Revision: 38146 Log: kva: check boundaries more strictly Suggested by reimar. Modified: trunk/libvo/vo_kva.c Modified: trunk/libvo/vo_kva.c ============================================================================== --- trunk/libvo/vo_kva.c Mon May 27 23:02:14 2019 (r38145) +++ trunk/libvo/vo_kva.c Wed May 29 12:56:50 2019 (r38146) @@ -608,7 +608,9 @@ static int preinit(const char *arg) {NULL, 0, NULL, NULL} }; - PCSZ pcszVideoModeStr[] = {"DIVE", "WarpOverlay!", "SNAP", "VMAN"}; + PCSZ pcszVideoModeStr[] = {"UNKNOWN", "DIVE", "WarpOverlay!", "SNAP", + "VMAN"}; + int nVideoModeStr = sizeof(pcszVideoModeStr) / sizeof(*pcszVideoModeStr); if (subopt_parse(arg, subopts) != 0) return -1; @@ -688,7 +690,8 @@ static int preinit(const char *arg) kvaCaps(&m_int.kvac); mp_msg(MSGT_VO, MSGL_V, "KVA: Selected video mode = %s\n", - pcszVideoModeStr[m_int.kvac.ulMode - 1]); + pcszVideoModeStr[m_int.kvac.ulMode >= nVideoModeStr ? + 0 : m_int.kvac.ulMode]); kvaDisableScreenSaver();

1 0

r38145 - in trunk/libmpcodecs: vf_eq.c vf_eq2.c
by reimar 27 May '19

27 May '19

Author: reimar Date: Mon May 27 23:02:14 2019 New Revision: 38145 Log: vf_eq.c, vf_eq2.c: Add SSE2 support. The code is actually identical, it would be even nicer if it was actually shared. Modified: trunk/libmpcodecs/vf_eq.c trunk/libmpcodecs/vf_eq2.c Modified: trunk/libmpcodecs/vf_eq.c ============================================================================== --- trunk/libmpcodecs/vf_eq.c Mon May 27 23:02:13 2019 (r38144) +++ trunk/libmpcodecs/vf_eq.c Mon May 27 23:02:14 2019 (r38145) @@ -104,6 +104,45 @@ static void process_MMX(unsigned char *d } #endif +#if HAVE_EMMINTRIN_H +#include <emmintrin.h> + +ATTR_TARGET_SSE2 +static void process_SSE2(unsigned char *dest, int dstride, unsigned char *src, int sstride, + int w, int h, int brightness, int contrast) +{ + int scaled_contrast = ((contrast+100)*256*16)/100; + int scaled_brightness = ((brightness+100)*511)/200-128 - scaled_contrast/32; + __m128i mmcontrast = _mm_set1_epi16(scaled_contrast); + __m128i mmbrightness = _mm_set1_epi16(scaled_brightness); + __m128i zero = _mm_setzero_si128(); + while (h--) { + int i; + for (i = 0; i < w - 15; i += 16) + { + __m128i mmsrc = _mm_loadu_si128((const __m128i *)(src + i)); + __m128i srclo = _mm_unpacklo_epi8(mmsrc, zero); + __m128i srchi = _mm_unpackhi_epi8(mmsrc, zero); + srclo = _mm_slli_epi16(srclo, 4); + srchi = _mm_slli_epi16(srchi, 4); + srclo = _mm_mulhi_epu16(srclo, mmcontrast); + srchi = _mm_mulhi_epu16(srchi, mmcontrast); + srclo = _mm_add_epi16(srclo, mmbrightness); + srchi = _mm_add_epi16(srchi, mmbrightness); + _mm_storeu_si128((__m128i *)(dest + i), _mm_packus_epi16(srclo, srchi)); + } + for (; i < w; i++) + { + int pel = ((src[i] * scaled_contrast)>>12) + scaled_brightness; + if(pel&768) pel = (-pel)>>31; + dest[i] = pel; + } + src += sstride; + dest += dstride; + } +} +#endif + static void process_C(unsigned char *dest, int dstride, unsigned char *src, int sstride, int w, int h, int brightness, int contrast) { @@ -230,6 +269,9 @@ static int vf_open(vf_instance_t *vf, ch #if HAVE_MMX_INLINE if(gCpuCaps.hasMMX) process = process_MMX; #endif +#if HAVE_EMMINTRIN_H + if(gCpuCaps.hasSSE2) process = process_SSE2; +#endif return 1; } Modified: trunk/libmpcodecs/vf_eq2.c ============================================================================== --- trunk/libmpcodecs/vf_eq2.c Mon May 27 23:02:13 2019 (r38144) +++ trunk/libmpcodecs/vf_eq2.c Mon May 27 23:02:14 2019 (r38145) @@ -188,6 +188,45 @@ void affine_1d_MMX (eq2_param_t *par, un } #endif +#if HAVE_EMMINTRIN_H +#include <emmintrin.h> + +ATTR_TARGET_SSE2 +static void affine_1d_SSE2(eq2_param_t *par, unsigned char *dst, unsigned char *src, + unsigned w, unsigned h, unsigned dstride, unsigned sstride) +{ + int scaled_contrast = par->c * 256 * 16; + int scaled_brightness = ((par->b+1.0)*511)/2-128 - scaled_contrast/32; + __m128i mmcontrast = _mm_set1_epi16(scaled_contrast); + __m128i mmbrightness = _mm_set1_epi16(scaled_brightness); + __m128i zero = _mm_setzero_si128(); + while (h--) { + int i; + for (i = 0; i < w - 15; i += 16) + { + __m128i mmsrc = _mm_loadu_si128((const __m128i *)(src + i)); + __m128i srclo = _mm_unpacklo_epi8(mmsrc, zero); + __m128i srchi = _mm_unpackhi_epi8(mmsrc, zero); + srclo = _mm_slli_epi16(srclo, 4); + srchi = _mm_slli_epi16(srchi, 4); + srclo = _mm_mulhi_epu16(srclo, mmcontrast); + srchi = _mm_mulhi_epu16(srchi, mmcontrast); + srclo = _mm_add_epi16(srclo, mmbrightness); + srchi = _mm_add_epi16(srchi, mmbrightness); + _mm_storeu_si128((__m128i *)(dst + i), _mm_packus_epi16(srclo, srchi)); + } + for (; i < w; i++) + { + int pel = ((src[i] * scaled_contrast)>>12) + scaled_brightness; + if(pel&768) pel = (-pel)>>31; + dst[i] = pel; + } + src += sstride; + dst += dstride; + } +} +#endif + static void apply_lut (eq2_param_t *par, unsigned char *dst, unsigned char *src, unsigned w, unsigned h, unsigned dstride, unsigned sstride) @@ -289,6 +328,11 @@ void check_values (eq2_param_t *par) if ((par->c == 1.0) && (par->b == 0.0) && (par->g == 1.0)) { par->adjust = NULL; } +#if HAVE_EMMINTRIN_H + else if (par->g == 1.0 && gCpuCaps.hasSSE2) { + par->adjust = &affine_1d_SSE2; + } +#endif #if HAVE_MMX_INLINE else if (par->g == 1.0 && gCpuCaps.hasMMX) { par->adjust = &affine_1d_MMX;

1 0

r38144 - trunk/libmpcodecs/pullup.c
by reimar 27 May '19

27 May '19

Author: reimar Date: Mon May 27 23:02:13 2019 New Revision: 38144 Log: pullup.c: Add SSE2 support to pullup filtering. About 2x faster vf_pullup. Modified: trunk/libmpcodecs/pullup.c Modified: trunk/libmpcodecs/pullup.c ============================================================================== --- trunk/libmpcodecs/pullup.c Mon May 27 23:02:12 2019 (r38143) +++ trunk/libmpcodecs/pullup.c Mon May 27 23:02:13 2019 (r38144) @@ -24,8 +24,6 @@ #include "config.h" #include "pullup.h" - - #if ARCH_X86 #if HAVE_MMX_INLINE static int diff_y_mmx(unsigned char *a, unsigned char *b, int s) @@ -212,6 +210,106 @@ static int var_y_mmx(unsigned char *a, u #endif #endif +#if HAVE_EMMINTRIN_H +#include <emmintrin.h> + +ATTR_TARGET_SSE2 +static int diff_y_sse2(unsigned char *a, unsigned char *b, int s) +{ + __m128i res; + __m128i mma = _mm_loadl_epi64((const __m128i *)a); + __m128i mmb = _mm_loadl_epi64((const __m128i *)b); + a+=s; b+=s; + mma = _mm_castpd_si128(_mm_loadh_pd(_mm_castsi128_pd(mma), (const double *)a)); + mmb = _mm_castpd_si128(_mm_loadh_pd(_mm_castsi128_pd(mmb), (const double *)b)); + a+=s; b+=s; + res = _mm_sad_epu8(mma, mmb); + + mma = _mm_loadl_epi64((const __m128i *)a); + mmb = _mm_loadl_epi64((const __m128i *)b); + a+=s; b+=s; + mma = _mm_castpd_si128(_mm_loadh_pd(_mm_castsi128_pd(mma), (const double *)a)); + mmb = _mm_castpd_si128(_mm_loadh_pd(_mm_castsi128_pd(mmb), (const double *)b)); + a+=s; b+=s; + res = _mm_add_epi16(res, _mm_sad_epu8(mma, mmb)); + + // sum high and low result + res = _mm_add_epi16(res, _mm_srli_si128(res, 8)); + + return _mm_extract_epi16(res, 0); +} + +ATTR_TARGET_SSE2 +static inline __m128i loadu8u16(const unsigned char *p) +{ + __m128i res = _mm_loadl_epi64((const __m128i *)p); + res = _mm_unpacklo_epi8(res, _mm_setzero_si128()); + return res; +} + +ATTR_TARGET_SSE2 +static inline __m128i mmabs16(__m128i v) +{ + // For the cases this is used, the subtraction + // cannot overflow, so skip saturation for slight + // better performance on some architectures + __m128i neg = _mm_sub_epi16(_mm_setzero_si128(), v); + return _mm_max_epi16(neg, v); +} + +ATTR_TARGET_SSE2 +static int licomb_y_sse2(unsigned char *a, unsigned char *b, int s) +{ + int i; + __m128i res = _mm_setzero_si128(); + __m128i mmbprev = loadu8u16(b-s); + __m128i mma = loadu8u16(a); + + for (i=4; i; i--) { + __m128i tmpa, tmpb; + __m128i mmb = loadu8u16(b); + __m128i mmanext = loadu8u16(a+s); + + tmpa = _mm_add_epi16(mma, mma); + tmpa = _mm_sub_epi16(tmpa, mmbprev); + tmpa = _mm_sub_epi16(tmpa, mmb); + + res = _mm_add_epi16(res, mmabs16(tmpa)); + + tmpb = _mm_add_epi16(mmb, mmb); + tmpb = _mm_sub_epi16(tmpb, mma); + tmpb = _mm_sub_epi16(tmpb, mmanext); + + res = _mm_add_epi16(res, mmabs16(tmpb)); + + mma = mmanext; + mmbprev = mmb; + a += s; b += s; + } + + // Sum reduce 8 to 4 to 2 to 1 elements. + // A bit wasteful to do 8 adds for a single result we + // care about, but _mm_extract_epi16 has high latency. + res = _mm_add_epi16(res, _mm_srli_si128(res, 8)); + res = _mm_add_epi16(res, _mm_srli_si128(res, 4)); + res = _mm_add_epi16(res, _mm_srli_si128(res, 2)); + return _mm_extract_epi16(res, 0); +} + +ATTR_TARGET_SSE2 +static int var_y_sse2(unsigned char *a, unsigned char *b, int s) +{ + __m128i a0 = _mm_loadl_epi64((const __m128i *)a); + __m128i a1 = _mm_loadl_epi64((const __m128i *)(a + s)); + __m128i a2 = _mm_loadl_epi64((const __m128i *)(a + 2*s)); + __m128i a3 = _mm_loadl_epi64((const __m128i *)(a + 3*s)); + __m128i res = _mm_sad_epu8(a0, a1); + res = _mm_add_epi16(res, _mm_sad_epu8(a1, a2)); + res = _mm_add_epi16(res, _mm_sad_epu8(a2, a3)); + return 4*_mm_extract_epi16(res, 0); +} +#endif + #define ABS(a) (((a)^((a)>>31))-((a)>>31)) static int diff_y(unsigned char *a, unsigned char *b, int s) @@ -252,8 +350,10 @@ static int licomb_y_test(unsigned char * { int c = licomb_y(a,b,s); int m = licomb_y_mmx(a,b,s); + int sse2 = licomb_y_sse2(a,b,s); if (c != m) printf("%d != %d\n", c, m); - return m; + if (c != sse2) printf("%d != %d (SSE2)\n", c, sse2); + return sse2; } #endif @@ -792,6 +892,13 @@ void pullup_init_context(struct pullup_c c->var = var_y_mmx; } #endif +#if HAVE_EMMINTRIN_H + if (c->cpu & PULLUP_CPU_SSE) { + c->diff = diff_y_sse2; + c->comb = licomb_y_sse2; + c->var = var_y_sse2; + } +#endif #endif /* c->comb = qpcomb_y; */ break;

1 0

r38143 - in trunk: configure sub/osd.c sub/osd_template.c sub/sub.c
by reimar 27 May '19

27 May '19

Author: reimar Date: Mon May 27 23:02:12 2019 New Revision: 38143 Log: osd_template.c: Add intrinsic based SSE2 support. Intrinsics are not well liked for good reason, but it is much faster to implement and compilers are no longer completely horrible for x86. Speed is about 2x-3x and it matches the C code instead of the less accurate MMX code. Primary motivation is though that MMX is starting to have worse and worse support in CPUs. To benchmark, use this command and remove the reading code from rawvideo and the memcpy_pic from vf_expand. ./mplayer /dev/zero -demuxer rawvideo -rawvideo w=1920:h=1080:format=yuy2 -vo null -osdlevel 3 -benchmark -quiet -frames 50000 -vf expand=osd=1 -subfont-osd-scale 20 Modified: trunk/configure trunk/sub/osd.c trunk/sub/osd_template.c trunk/sub/sub.c Modified: trunk/configure ============================================================================== --- trunk/configure Sun May 26 21:56:32 2019 (r38142) +++ trunk/configure Mon May 27 23:02:12 2019 (r38143) @@ -3883,6 +3883,26 @@ header_check sys/poll.h && poll_h=yes && def_poll_h='#define HAVE_POLL_H 1' echores "$poll_h" +echocheck "emmintrin.h (SSE intrinsics)" +emmintrin_h=no +def_emmintrin_h='#define HAVE_EMMINTRIN_H 0' + cat > $TMPC << EOF +#include <emmintrin.h> + +__attribute__((target("sse2"))) +static int sse2test(int i) { + __m128i mmi = _mm_set1_epi16(i); + mmi = _mm_add_epi16(mmi, mmi); + return _mm_extract_epi16(mmi, 2); +} + +int main(int argc, char **argv) { + return sse2test(argc); +} +EOF +cc_check && emmintrin_h=yes && + def_emmintrin_h='#define HAVE_EMMINTRIN_H 1' +echores "$emmintrin_h" echocheck "inttypes.h (required)" _inttypes=no @@ -9243,6 +9263,12 @@ $def_direct_h $def_io_h $def_poll_h $def_windows_h +$def_emmintrin_h +#if ARCH_X86_32 +#define ATTR_TARGET_SSE2 __attribute__((target("sse2"))) +#else +#define ATTR_TARGET_SSE2 +#endif /* external libraries */ $def_bzlib Modified: trunk/sub/osd.c ============================================================================== --- trunk/sub/osd.c Sun May 26 21:56:32 2019 (r38142) +++ trunk/sub/osd.c Mon May 27 23:02:12 2019 (r38143) @@ -31,7 +31,12 @@ #include "libmpcodecs/img_format.h" #include "cpudetect.h" -#if ARCH_X86 +#if !HAVE_EMMINTRIN_H +#undef HAVE_SSE2 +#define HAVE_SSE2 0 +#endif + +#if ARCH_X86 && (!HAVE_SSE2 || CONFIG_RUNTIME_CPUDETECT) static const uint64_t bFF __attribute__((aligned(8))) = 0xFFFFFFFFFFFFFFFFULL; static const unsigned long long mask24lh __attribute__((aligned(8))) = 0xFFFF000000000000ULL; static const unsigned long long mask24hl __attribute__((aligned(8))) = 0x0000FFFFFFFFFFFFULL; @@ -39,32 +44,38 @@ static const unsigned long long mask24hl //Note: we have C, X86-nommx, MMX, MMX2, 3DNOW version therse no 3DNOW+MMX2 one //Plain C versions -#if !HAVE_MMX || CONFIG_RUNTIME_CPUDETECT +#if (!HAVE_MMX && !HAVE_SSE2) || CONFIG_RUNTIME_CPUDETECT #define COMPILE_C #endif #if ARCH_X86 -#if (HAVE_MMX && !HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT +#if (HAVE_MMX && !HAVE_AMD3DNOW && !HAVE_MMX2 && !HAVE_SSE2) || CONFIG_RUNTIME_CPUDETECT #define COMPILE_MMX #endif -#if HAVE_MMX2 || CONFIG_RUNTIME_CPUDETECT +#if (HAVE_MMX2 && !HAVE_SSE2) || CONFIG_RUNTIME_CPUDETECT #define COMPILE_MMX2 #endif -#if (HAVE_AMD3DNOW && !HAVE_MMX2) || CONFIG_RUNTIME_CPUDETECT +#if (HAVE_AMD3DNOW && !HAVE_MMX2 && !HAVE_SSE2) || CONFIG_RUNTIME_CPUDETECT #define COMPILE_3DNOW #endif +#if HAVE_SSE2 || CONFIG_RUNTIME_CPUDETECT +#define COMPILE_SSE2 +#endif + #endif /* ARCH_X86 */ #undef HAVE_MMX #undef HAVE_MMX2 #undef HAVE_AMD3DNOW +#undef HAVE_SSE2 #define HAVE_MMX 0 #define HAVE_MMX2 0 #define HAVE_AMD3DNOW 0 +#define HAVE_SSE2 0 #if ! ARCH_X86 @@ -72,9 +83,11 @@ static const unsigned long long mask24hl #undef HAVE_MMX #undef HAVE_MMX2 #undef HAVE_AMD3DNOW +#undef HAVE_SSE2 #define HAVE_MMX 0 #define HAVE_MMX2 0 #define HAVE_AMD3DNOW 0 +#define HAVE_SSE2 0 #define RENAME(a) a ## _C #include "osd_template.c" #endif @@ -87,9 +100,11 @@ static const unsigned long long mask24hl #undef HAVE_MMX #undef HAVE_MMX2 #undef HAVE_AMD3DNOW +#undef HAVE_SSE2 #define HAVE_MMX 0 #define HAVE_MMX2 0 #define HAVE_AMD3DNOW 0 +#define HAVE_SSE2 0 #define RENAME(a) a ## _X86 #include "osd_template.c" #endif @@ -100,9 +115,11 @@ static const unsigned long long mask24hl #undef HAVE_MMX #undef HAVE_MMX2 #undef HAVE_AMD3DNOW +#undef HAVE_SSE2 #define HAVE_MMX 1 #define HAVE_MMX2 0 #define HAVE_AMD3DNOW 0 +#define HAVE_SSE2 0 #define RENAME(a) a ## _MMX #include "osd_template.c" #endif @@ -113,9 +130,11 @@ static const unsigned long long mask24hl #undef HAVE_MMX #undef HAVE_MMX2 #undef HAVE_AMD3DNOW +#undef HAVE_SSE2 #define HAVE_MMX 1 #define HAVE_MMX2 1 #define HAVE_AMD3DNOW 0 +#define HAVE_SSE2 0 #define RENAME(a) a ## _MMX2 #include "osd_template.c" #endif @@ -126,20 +145,39 @@ static const unsigned long long mask24hl #undef HAVE_MMX #undef HAVE_MMX2 #undef HAVE_AMD3DNOW +#undef HAVE_SSE2 #define HAVE_MMX 1 #define HAVE_MMX2 0 #define HAVE_AMD3DNOW 1 +#define HAVE_SSE2 0 #define RENAME(a) a ## _3DNow #include "osd_template.c" #endif +//SSE2 versions +#ifdef COMPILE_SSE2 +#undef RENAME +#undef HAVE_MMX +#undef HAVE_MMX2 +#undef HAVE_AMD3DNOW +#undef HAVE_SSE2 +#define HAVE_MMX 0 +#define HAVE_MMX2 0 +#define HAVE_AMD3DNOW 0 +#define HAVE_SSE2 1 +#define RENAME(a) a ## _SSE2 +#include <emmintrin.h> +#include "osd_template.c" +#endif #endif /* ARCH_X86 */ void vo_draw_alpha_yv12(int w,int h, unsigned char* src, unsigned char *srca, int srcstride, unsigned char* dstbase,int dststride){ #if CONFIG_RUNTIME_CPUDETECT #if ARCH_X86 // ordered by speed / fastest first - if(gCpuCaps.hasMMX2) + if(HAVE_EMMINTRIN_H && gCpuCaps.hasSSE2) + vo_draw_alpha_yv12_SSE2(w, h, src, srca, srcstride, dstbase, dststride); + else if(gCpuCaps.hasMMX2) vo_draw_alpha_yv12_MMX2(w, h, src, srca, srcstride, dstbase, dststride); else if(gCpuCaps.has3DNow) vo_draw_alpha_yv12_3DNow(w, h, src, srca, srcstride, dstbase, dststride); @@ -151,7 +189,9 @@ void vo_draw_alpha_yv12(int w,int h, uns vo_draw_alpha_yv12_C(w, h, src, srca, srcstride, dstbase, dststride); #endif #else //CONFIG_RUNTIME_CPUDETECT -#if HAVE_MMX2 +#if HAVE_SSE2 + vo_draw_alpha_yv12_SSE2(w, h, src, srca, srcstride, dstbase, dststride); +#elif HAVE_MMX2 vo_draw_alpha_yv12_MMX2(w, h, src, srca, srcstride, dstbase, dststride); #elif HAVE_AMD3DNOW vo_draw_alpha_yv12_3DNow(w, h, src, srca, srcstride, dstbase, dststride); @@ -169,7 +209,9 @@ void vo_draw_alpha_yuy2(int w,int h, uns #if CONFIG_RUNTIME_CPUDETECT #if ARCH_X86 // ordered by speed / fastest first - if(gCpuCaps.hasMMX2) + if(HAVE_EMMINTRIN_H && gCpuCaps.hasSSE2) + vo_draw_alpha_yuy2_SSE2(w, h, src, srca, srcstride, dstbase, dststride); + else if(gCpuCaps.hasMMX2) vo_draw_alpha_yuy2_MMX2(w, h, src, srca, srcstride, dstbase, dststride); else if(gCpuCaps.has3DNow) vo_draw_alpha_yuy2_3DNow(w, h, src, srca, srcstride, dstbase, dststride); @@ -181,7 +223,9 @@ void vo_draw_alpha_yuy2(int w,int h, uns vo_draw_alpha_yuy2_C(w, h, src, srca, srcstride, dstbase, dststride); #endif #else //CONFIG_RUNTIME_CPUDETECT -#if HAVE_MMX2 +#if HAVE_SSE2 + vo_draw_alpha_yuy2_SSE2(w, h, src, srca, srcstride, dstbase, dststride); +#elif HAVE_MMX2 vo_draw_alpha_yuy2_MMX2(w, h, src, srca, srcstride, dstbase, dststride); #elif HAVE_AMD3DNOW vo_draw_alpha_yuy2_3DNow(w, h, src, srca, srcstride, dstbase, dststride); @@ -199,7 +243,9 @@ void vo_draw_alpha_uyvy(int w,int h, uns #if CONFIG_RUNTIME_CPUDETECT #if ARCH_X86 // ordered by speed / fastest first - if(gCpuCaps.hasMMX2) + if(HAVE_EMMINTRIN_H && gCpuCaps.hasSSE2) + vo_draw_alpha_uyvy_SSE2(w, h, src, srca, srcstride, dstbase, dststride); + else if(gCpuCaps.hasMMX2) vo_draw_alpha_uyvy_MMX2(w, h, src, srca, srcstride, dstbase, dststride); else if(gCpuCaps.has3DNow) vo_draw_alpha_uyvy_3DNow(w, h, src, srca, srcstride, dstbase, dststride); @@ -211,7 +257,9 @@ void vo_draw_alpha_uyvy(int w,int h, uns vo_draw_alpha_uyvy_C(w, h, src, srca, srcstride, dstbase, dststride); #endif #else //CONFIG_RUNTIME_CPUDETECT -#if HAVE_MMX2 +#if HAVE_SSE2 + vo_draw_alpha_uyvy_SSE2(w, h, src, srca, srcstride, dstbase, dststride); +#elif HAVE_MMX2 vo_draw_alpha_uyvy_MMX2(w, h, src, srca, srcstride, dstbase, dststride); #elif HAVE_AMD3DNOW vo_draw_alpha_uyvy_3DNow(w, h, src, srca, srcstride, dstbase, dststride); @@ -229,7 +277,9 @@ void vo_draw_alpha_rgb24(int w,int h, un #if CONFIG_RUNTIME_CPUDETECT #if ARCH_X86 // ordered by speed / fastest first - if(gCpuCaps.hasMMX2) + if(HAVE_EMMINTRIN_H && gCpuCaps.hasSSE2) + vo_draw_alpha_rgb24_SSE2(w, h, src, srca, srcstride, dstbase, dststride); + else if(gCpuCaps.hasMMX2) vo_draw_alpha_rgb24_MMX2(w, h, src, srca, srcstride, dstbase, dststride); else if(gCpuCaps.has3DNow) vo_draw_alpha_rgb24_3DNow(w, h, src, srca, srcstride, dstbase, dststride); @@ -241,7 +291,9 @@ void vo_draw_alpha_rgb24(int w,int h, un vo_draw_alpha_rgb24_C(w, h, src, srca, srcstride, dstbase, dststride); #endif #else //CONFIG_RUNTIME_CPUDETECT -#if HAVE_MMX2 +#if HAVE_SSE2 + vo_draw_alpha_rgb24_SSE2(w, h, src, srca, srcstride, dstbase, dststride); +#elif HAVE_MMX2 vo_draw_alpha_rgb24_MMX2(w, h, src, srca, srcstride, dstbase, dststride); #elif HAVE_AMD3DNOW vo_draw_alpha_rgb24_3DNow(w, h, src, srca, srcstride, dstbase, dststride); @@ -259,7 +311,9 @@ void vo_draw_alpha_rgb32(int w,int h, un #if CONFIG_RUNTIME_CPUDETECT #if ARCH_X86 // ordered by speed / fastest first - if(gCpuCaps.hasMMX2) + if(HAVE_EMMINTRIN_H && gCpuCaps.hasSSE2) + vo_draw_alpha_rgb32_SSE2(w, h, src, srca, srcstride, dstbase, dststride); + else if(gCpuCaps.hasMMX2) vo_draw_alpha_rgb32_MMX2(w, h, src, srca, srcstride, dstbase, dststride); else if(gCpuCaps.has3DNow) vo_draw_alpha_rgb32_3DNow(w, h, src, srca, srcstride, dstbase, dststride); @@ -271,7 +325,9 @@ void vo_draw_alpha_rgb32(int w,int h, un vo_draw_alpha_rgb32_C(w, h, src, srca, srcstride, dstbase, dststride); #endif #else //CONFIG_RUNTIME_CPUDETECT -#if HAVE_MMX2 +#if HAVE_SSE2 + vo_draw_alpha_rgb32_SSE2(w, h, src, srca, srcstride, dstbase, dststride); +#elif HAVE_MMX2 vo_draw_alpha_rgb32_MMX2(w, h, src, srca, srcstride, dstbase, dststride); #elif HAVE_AMD3DNOW vo_draw_alpha_rgb32_3DNow(w, h, src, srca, srcstride, dstbase, dststride); @@ -306,7 +362,9 @@ void vo_draw_alpha_init(void){ #if CONFIG_RUNTIME_CPUDETECT #if ARCH_X86 // ordered per speed fasterst first - if(gCpuCaps.hasMMX2) + if(HAVE_EMMINTRIN_H && gCpuCaps.hasSSE2) + mp_msg(MSGT_OSD,MSGL_INFO,"Using SSE2 Optimized OnScreenDisplay\n"); + else if(gCpuCaps.hasMMX2) mp_msg(MSGT_OSD,MSGL_INFO,"Using MMX (with tiny bit MMX2) Optimized OnScreenDisplay\n"); else if(gCpuCaps.has3DNow) mp_msg(MSGT_OSD,MSGL_INFO,"Using MMX (with tiny bit 3DNow) Optimized OnScreenDisplay\n"); @@ -318,7 +376,9 @@ void vo_draw_alpha_init(void){ mp_msg(MSGT_OSD,MSGL_INFO,"Using Unoptimized OnScreenDisplay\n"); #endif #else //CONFIG_RUNTIME_CPUDETECT -#if HAVE_MMX2 +#if HAVE_SSE2 + mp_msg(MSGT_OSD,MSGL_INFO,"Using SSE2 Optimized OnScreenDisplay\n"); +#elif HAVE_MMX2 mp_msg(MSGT_OSD,MSGL_INFO,"Using MMX (with tiny bit MMX2) Optimized OnScreenDisplay\n"); #elif HAVE_AMD3DNOW mp_msg(MSGT_OSD,MSGL_INFO,"Using MMX (with tiny bit 3DNow) Optimized OnScreenDisplay\n"); Modified: trunk/sub/osd_template.c ============================================================================== --- trunk/sub/osd_template.c Sun May 26 21:56:32 2019 (r38142) +++ trunk/sub/osd_template.c Mon May 27 23:02:12 2019 (r38143) @@ -44,9 +44,60 @@ #define EMMS "emms" #endif +#if HAVE_SSE2 +ATTR_TARGET_SSE2 +static inline __m128i muladd_src_unpacked(__m128i dstlo, __m128i dsthi, __m128i src, __m128i srcalo, __m128i srcahi) +{ + // (mmhigh,mmlow) = (dst * (srca * 256)) / 65536 (= (dst * srca) >> 8) + __m128i mmlow = _mm_mulhi_epu16(dstlo, srcalo); + __m128i mmhigh = _mm_mulhi_epu16(dsthi, srcahi); + + __m128i res = _mm_packus_epi16(mmlow, mmhigh); + + return _mm_add_epi8(res, src); +} + +ATTR_TARGET_SSE2 +static inline __m128i muladd_src(__m128i dst, __m128i src, __m128i srca) +{ + __m128i zero = _mm_setzero_si128(); + __m128i dstlo = _mm_unpacklo_epi8(dst, zero); + __m128i dsthi = _mm_unpackhi_epi8(dst, zero); + __m128i srcalo = _mm_unpacklo_epi8(zero, srca); + __m128i srcahi = _mm_unpackhi_epi8(zero, srca); + return muladd_src_unpacked(dstlo, dsthi, src, srcalo, srcahi); +} + +ATTR_TARGET_SSE2 +static inline __m128i alphamask(__m128i orig, __m128i blended, __m128i srca) +{ + __m128i zero = _mm_setzero_si128(); + // if (!srca) res |= dst --- assumes srca == 0 implies src == 0, + // thus no need to mask res + __m128i mask = _mm_cmpeq_epi8(srca, zero); + orig = _mm_and_si128(orig, mask); + return _mm_or_si128(blended, orig); +} + +// Special version that compares alpha in 16 bit chunks instead of bytewise +ATTR_TARGET_SSE2 +static inline __m128i alphamask16(__m128i orig, __m128i blended, __m128i srca) +{ + __m128i zero = _mm_setzero_si128(); + // if (!srca) res |= dst --- assumes srca == 0 implies src == 0, + // thus no need to mask res + __m128i mask = _mm_cmpeq_epi16(srca, zero); + orig = _mm_and_si128(orig, mask); + return _mm_or_si128(blended, orig); +} +#endif + +#if HAVE_SSE2 +ATTR_TARGET_SSE2 +#endif static inline void RENAME(vo_draw_alpha_yv12)(int w,int h, unsigned char* src, unsigned char *srca, int srcstride, unsigned char* dstbase,int dststride){ int y; -#if defined(FAST_OSD) && !HAVE_MMX +#if defined(FAST_OSD) && !HAVE_MMX && !HAVE_SSE2 w=w>>1; #endif #if HAVE_MMX @@ -94,7 +145,28 @@ static inline void RENAME(vo_draw_alpha_ :: "m" (dstbase[x]), "m" (srca[x]), "m" (src[x]) : "%eax"); } -#else +#elif HAVE_SSE2 + __m128i zero = _mm_setzero_si128(); + for(x=0;x+8<w;x+=16){ + __m128i mmsrc, mmdst, res; + __m128i mmsrca = _mm_load_si128((const __m128i *)(srca + x)); + + int alpha = _mm_movemask_epi8(_mm_cmpeq_epi8(mmsrca, zero)); + if (alpha == 0xffff) continue; + + mmdst = _mm_loadu_si128((const __m128i *)(dstbase + x)); + mmsrc = _mm_load_si128((const __m128i *)(src + x)); + + res = muladd_src(mmdst, mmsrc, mmsrca); + + // _mm_maskmoveu_s128 would be an alternative but slower + res = alphamask(mmdst, res, mmsrca); + _mm_storeu_si128((__m128i *)(dstbase + x), res); + } + for(;x<w;x++){ + if(srca[x]) dstbase[x]=((dstbase[x]*srca[x])>>8)+src[x]; + } +#else /* HAVE_SSE2 */ for(x=0;x<w;x++){ #ifdef FAST_OSD if(srca[2*x+0]) dstbase[2*x+0]=src[2*x+0]; @@ -114,6 +186,9 @@ static inline void RENAME(vo_draw_alpha_ return; } +#if HAVE_SSE2 +ATTR_TARGET_SSE2 +#endif static inline void RENAME(vo_draw_alpha_yuy2)(int w,int h, unsigned char* src, unsigned char *srca, int srcstride, unsigned char* dstbase,int dststride){ int y; #if defined(FAST_OSD) && !HAVE_MMX @@ -155,7 +230,7 @@ static inline void RENAME(vo_draw_alpha_ "psrlw $8, %%mm0\n\t" "pand %%mm5, %%mm1\n\t" //U0V0U0V0 "movd %2, %%mm2\n\t" //src 0000DCBA - "punpcklbw %%mm7, %%mm2\n\t" //srca 0D0C0B0A + "punpcklbw %%mm7, %%mm2\n\t" //src 0D0C0B0A "por %%mm1, %%mm0\n\t" "paddb %%mm2, %%mm0\n\t" "movq %%mm0, %0\n\t" @@ -163,7 +238,56 @@ static inline void RENAME(vo_draw_alpha_ :: "m" (dstbase[x*2]), "m" (srca[x]), "m" (src[x]) : "%eax"); } -#else +#elif HAVE_SSE2 + __m128i zero = _mm_setzero_si128(); + __m128i ymask = _mm_set1_epi16(0xff); + __m128i uvofs = _mm_set1_epi16(0x8000); + for(x=0;x+12<w;x+=16){ + __m128i mmsrc, mmsrcalo, mmsrcahi, mmdst, mmdst2, mmlow, mmhigh, mmy, mmuv, res; + __m128i mmsrca = _mm_load_si128((const __m128i *)(srca + x)); + int alpha = _mm_movemask_epi8(_mm_cmpeq_epi8(mmsrca, zero)); + if (alpha == 0xffff) continue; + + mmdst = _mm_loadu_si128((const __m128i *)(dstbase + 2*x)); + mmdst2 = _mm_loadu_si128((const __m128i *)(dstbase + 2*x + 16)); + + // convert UV to signed + mmdst = _mm_xor_si128(mmdst, uvofs); + mmdst2 = _mm_xor_si128(mmdst2, uvofs); + + mmsrc = _mm_load_si128((const __m128i *)(src + x)); + mmsrcalo = _mm_unpacklo_epi8(zero, mmsrca); + mmsrcahi = _mm_unpackhi_epi8(zero, mmsrca); + + mmy = muladd_src_unpacked(_mm_and_si128(mmdst, ymask), _mm_and_si128(mmdst2, ymask), mmsrc, mmsrcalo, mmsrcahi); + + // mmuv = ((dst(uv) ^ 128) * (srca * 256)) / 65536 ^ 128 (= (((dst - 128) * srca) >> 8)) + 128 + mmlow = _mm_srai_epi16(mmdst, 8); + mmlow = _mm_mulhi_epi16(mmlow, mmsrcalo); + mmhigh = _mm_srai_epi16(mmdst2, 8); + mmhigh = _mm_mulhi_epi16(mmhigh, mmsrcahi); + + mmuv = _mm_packs_epi16(mmlow, mmhigh); + + res = _mm_unpacklo_epi8(mmy, mmuv); + res = alphamask16(mmdst, res, mmsrcalo); + // convert UV to unsigned + res = _mm_xor_si128(res, uvofs); + _mm_storeu_si128((__m128i *)(dstbase + 2 * x), res); + + res = _mm_unpackhi_epi8(mmy, mmuv); + res = alphamask16(mmdst2, res, mmsrcahi); + // convert UV to unsigned + res = _mm_xor_si128(res, uvofs); + _mm_storeu_si128((__m128i *)(dstbase + 2 * x + 16), res); + } + for(;x<w;x++){ + if(srca[x]) { + dstbase[2*x]=((dstbase[2*x]*srca[x])>>8)+src[x]; + dstbase[2*x+1]=((((signed)dstbase[2*x+1]-128)*srca[x])>>8)+128; + } + } +#else /* HAVE_SSE2 */ for(x=0;x<w;x++){ #ifdef FAST_OSD if(srca[2*x+0]) dstbase[4*x+0]=src[2*x+0]; @@ -186,6 +310,9 @@ static inline void RENAME(vo_draw_alpha_ return; } +#if HAVE_SSE2 +ATTR_TARGET_SSE2 +#endif static inline void RENAME(vo_draw_alpha_uyvy)(int w,int h, unsigned char* src, unsigned char *srca, int srcstride, unsigned char* dstbase,int dststride){ int y; #if defined(FAST_OSD) @@ -193,6 +320,58 @@ static inline void RENAME(vo_draw_alpha_ #endif for(y=0;y<h;y++){ register int x; +#if HAVE_SSE2 + __m128i zero = _mm_setzero_si128(); + __m128i uvofs = _mm_set1_epi16(0x80); + for(x=0;x+12<w;x+=16){ + __m128i mmsrc, mmsrcalo, mmsrcahi, mmdst, mmdst2, mmlow, mmhigh, mmy, mmuv, res; + __m128i mmsrca = _mm_load_si128((const __m128i *)(srca + x)); + int alpha = _mm_movemask_epi8(_mm_cmpeq_epi8(mmsrca, zero)); + if (alpha == 0xffff) continue; + + mmdst = _mm_loadu_si128((const __m128i *)(dstbase + 2*x)); + mmdst2 = _mm_loadu_si128((const __m128i *)(dstbase + 2*x + 16)); + + // convert UV to signed + mmdst = _mm_xor_si128(mmdst, uvofs); + mmdst2 = _mm_xor_si128(mmdst2, uvofs); + + mmsrc = _mm_load_si128((const __m128i *)(src + x)); + mmsrcalo = _mm_unpacklo_epi8(zero, mmsrca); + mmsrcahi = _mm_unpackhi_epi8(zero, mmsrca); + + mmy = muladd_src_unpacked(_mm_srli_epi16(mmdst, 8), _mm_srli_epi16(mmdst2, 8), mmsrc, mmsrcalo, mmsrcahi); + + // mmuv = ((dst(uv) ^ 128) * (srca * 256)) / 65536 ^ 128 (= (((dst - 128) * srca) >> 8)) + 128 + // sign-extend and multiply + mmlow = _mm_slli_epi16(mmdst, 8); + mmlow = _mm_srai_epi16(mmlow, 8); + mmlow = _mm_mulhi_epi16(mmlow, mmsrcalo); + mmhigh = _mm_slli_epi16(mmdst2, 8); + mmhigh = _mm_srai_epi16(mmhigh, 8); + mmhigh = _mm_mulhi_epi16(mmhigh, mmsrcahi); + + mmuv = _mm_packs_epi16(mmlow, mmhigh); + + res = _mm_unpacklo_epi8(mmuv, mmy); + res = alphamask16(mmdst, res, mmsrcalo); + // convert UV to unsigned + res = _mm_xor_si128(res, uvofs); + _mm_storeu_si128((__m128i *)(dstbase + 2 * x), res); + + res = _mm_unpackhi_epi8(mmuv, mmy); + res = alphamask16(mmdst2, res, mmsrcahi); + // convert UV to unsigned + res = _mm_xor_si128(res, uvofs); + _mm_storeu_si128((__m128i *)(dstbase + 2 * x + 16), res); + } + for(;x<w;x++){ + if(srca[x]) { + dstbase[2*x+1]=((dstbase[2*x+1]*srca[x])>>8)+src[x]; + dstbase[2*x]=((((signed)dstbase[2*x]-128)*srca[x])>>8)+128; + } + } +#else /* HAVE_SSE2 */ for(x=0;x<w;x++){ #ifdef FAST_OSD if(srca[2*x+0]) dstbase[4*x+2]=src[2*x+0]; @@ -204,12 +383,25 @@ static inline void RENAME(vo_draw_alpha_ } #endif } +#endif src+=srcstride; srca+=srcstride; dstbase+=dststride; } } +#define REPL3X(out, sd1, sa1, sd2, sa2, in) \ +do { \ + __m128i shuf012 = _mm_shufflelo_epi16(in, 0x40); \ + __m128i shuf345 = _mm_shufflelo_epi16(in, 0xA5); \ + __m128i repl3x_mmtmp = _mm_unpacklo_epi64(shuf012, shuf345); \ + repl3x_mmtmp = _mm_and_si128(repl3x_mmtmp, one_in_three_mask); \ + out = _mm_or_si128(_mm_or_si128(repl3x_mmtmp, _mm_s##sd1##li_si128(repl3x_mmtmp, sa1)), _mm_s##sd2##li_si128(repl3x_mmtmp, sa2)); \ +} while (0) + +#if HAVE_SSE2 +ATTR_TARGET_SSE2 +#endif static inline void RENAME(vo_draw_alpha_rgb24)(int w,int h, unsigned char* src, unsigned char *srca, int srcstride, unsigned char* dstbase,int dststride){ int y; #if HAVE_MMX @@ -221,7 +413,7 @@ static inline void RENAME(vo_draw_alpha_ for(y=0;y<h;y++){ register unsigned char *dst = dstbase; register int x; -#if ARCH_X86 && (!ARCH_X86_64 || HAVE_MMX) +#if ARCH_X86 && (!ARCH_X86_64 || HAVE_MMX || HAVE_SSE2) #if HAVE_MMX __asm__ volatile( PREFETCHW" %0\n\t" @@ -264,7 +456,62 @@ static inline void RENAME(vo_draw_alpha_ :: "m" (dst[0]), "m" (srca[x]), "m" (src[x]), "m"(mask24hl), "m"(mask24lh)); dst += 6; } -#else /* HAVE_MMX */ +#elif HAVE_SSE2 + __m128i one_in_three_mask = _mm_set_epi32(0xff0000ffu, 0x0000ff00u, 0x00ff0000u, 0xff0000ffu); + __m128i zero = _mm_setzero_si128(); + for(x=0;x+14<w;x+=16){ + __m128i mmsrc, mmtmp, mmtmpa, mmdst, res; + __m128i mmsrca = _mm_load_si128((const __m128i *)(srca + x)); + int alpha = _mm_movemask_epi8(_mm_cmpeq_epi8(mmsrca, zero)); + if (alpha == 0xffff) { dst += 48; continue; } + + mmsrc = _mm_load_si128((const __m128i *)(src + x)); + + if ((alpha & 0x3f) != 0x3f) { + mmdst = _mm_loadu_si128((const __m128i *)dst); + REPL3X(mmtmpa, l, 1, l, 2, mmsrca); + REPL3X(mmtmp, l, 1, l, 2, mmsrc); + res = muladd_src(mmdst, mmtmp, mmtmpa); + res = alphamask(mmdst, res, mmtmpa); + _mm_storeu_si128((__m128i *)dst, res); + } + dst += 16; + + mmsrca = _mm_srli_si128(mmsrca, 5); + mmsrc = _mm_srli_si128(mmsrc, 5); + + if ((alpha & 0x7e0) != 0x7e0) { + mmdst = _mm_loadu_si128((const __m128i *)dst); + REPL3X(mmtmpa, l, 1, r, 1, mmsrca); + REPL3X(mmtmp, l, 1, r, 1, mmsrc); + res = muladd_src(mmdst, mmtmp, mmtmpa); + res = alphamask(mmdst, res, mmtmpa); + _mm_storeu_si128((__m128i *)dst, res); + } + dst += 16; + + mmsrc = _mm_srli_si128(mmsrc, 5); + mmsrca = _mm_srli_si128(mmsrca, 5); + + if ((alpha & 0xfc00) != 0xfc00) { + mmdst = _mm_loadu_si128((const __m128i *)dst); + REPL3X(mmtmpa, r, 1, r, 2, mmsrca); + REPL3X(mmtmp, r, 1, r, 2, mmsrc); + res = muladd_src(mmdst, mmtmp, mmtmpa); + res = alphamask(mmdst, res, mmtmpa); + _mm_storeu_si128((__m128i *)dst, res); + } + dst += 16; + } + for(;x<w;x++){ + if(srca[x]){ + dst[0]=((dst[0]*srca[x])>>8)+src[x]; + dst[1]=((dst[1]*srca[x])>>8)+src[x]; + dst[2]=((dst[2]*srca[x])>>8)+src[x]; + } + dst+=3; // 24bpp + } +#else /* HAVE_SSE2 */ for(x=0;x<w;x++){ if(srca[x]){ __asm__ volatile( @@ -294,7 +541,7 @@ static inline void RENAME(vo_draw_alpha_ dst += 3; } #endif /* !HAVE_MMX */ -#else /*non x86 arch or x86_64 with MMX disabled */ +#else /*non x86 arch or x86_64 with MMX and SSE2 disabled */ for(x=0;x<w;x++){ if(srca[x]){ #ifdef FAST_OSD @@ -318,6 +565,9 @@ static inline void RENAME(vo_draw_alpha_ return; } +#if HAVE_SSE2 +ATTR_TARGET_SSE2 +#endif static inline void RENAME(vo_draw_alpha_rgb32)(int w,int h, unsigned char* src, unsigned char *srca, int srcstride, unsigned char* dstbase,int dststride){ int y; #if HAVE_BIGENDIAN @@ -341,7 +591,7 @@ static inline void RENAME(vo_draw_alpha_ #endif /* HAVE_MMX */ for(y=0;y<h;y++){ register int x; -#if ARCH_X86 && (!ARCH_X86_64 || HAVE_MMX) +#if ARCH_X86 && (!ARCH_X86_64 || HAVE_MMX || HAVE_SSE2) #if HAVE_MMX #if HAVE_AMD3DNOW __asm__ volatile( @@ -431,7 +681,42 @@ static inline void RENAME(vo_draw_alpha_ : "%eax"); } #endif -#else /* HAVE_MMX */ +#elif HAVE_SSE2 + __m128i zero = _mm_setzero_si128(); + __m128i mmsrca = _mm_setzero_si128(); + for(x=0;x<w;x+=4){ + __m128i mmsrc, mmdst, mmsrcexp, mmsrcaexp, res; + if ((x & 15) == 0) { + int alpha; + mmsrca = _mm_load_si128((const __m128i *)(srca + x)); + alpha = _mm_movemask_epi8(_mm_cmpeq_epi8(mmsrca, zero)); + if (alpha == 0xffff) { x += 12; continue; } + mmsrc = _mm_load_si128((const __m128i *)(src + x)); + } + + mmdst = _mm_loadu_si128((const __m128i *)(dstbase + 4*x)); + + mmsrcaexp = _mm_unpacklo_epi8(mmsrca, mmsrca); + mmsrcaexp = _mm_unpacklo_epi8(mmsrcaexp, mmsrcaexp); + mmsrcexp = _mm_unpacklo_epi8(mmsrc, mmsrc); + mmsrcexp = _mm_unpacklo_epi8(mmsrcexp, mmsrcexp); + + res = muladd_src(mmdst, mmsrcexp, mmsrcaexp); + + res = alphamask(mmdst, res, mmsrcaexp); + _mm_storeu_si128((__m128i *)(dstbase + 4*x), res); + + mmsrca = _mm_srli_si128(mmsrca, 4); + mmsrc = _mm_srli_si128(mmsrc, 4); + } + for(;x<w;x++){ + if(srca[x]){ + dstbase[4*x+0]=((dstbase[4*x+0]*srca[x])>>8)+src[x]; + dstbase[4*x+1]=((dstbase[4*x+1]*srca[x])>>8)+src[x]; + dstbase[4*x+2]=((dstbase[4*x+2]*srca[x])>>8)+src[x]; + } + } +#else /* HAVE_SSE2 */ for(x=0;x<w;x++){ if(srca[x]){ __asm__ volatile( Modified: trunk/sub/sub.c ============================================================================== --- trunk/sub/sub.c Sun May 26 21:56:32 2019 (r38142) +++ trunk/sub/sub.c Mon May 27 23:02:12 2019 (r38143) @@ -152,7 +152,7 @@ static void alloc_buf(mp_osd_obj_t* obj) int len; if (obj->bbox.x2 < obj->bbox.x1) obj->bbox.x2 = obj->bbox.x1; if (obj->bbox.y2 < obj->bbox.y1) obj->bbox.y2 = obj->bbox.y1; - obj->stride = ((obj->bbox.x2-obj->bbox.x1)+7)&(~7); + obj->stride = ((obj->bbox.x2-obj->bbox.x1)+15)&(~15); len = obj->stride*(obj->bbox.y2-obj->bbox.y1); if (obj->allocated<len) { obj->allocated = len;

1 0

r38140 - trunk/libvo/vo_kva.c
by komh 27 May '19

27 May '19

Author: komh Date: Fri May 24 13:42:53 2019 New Revision: 38140 Log: kva: fix buffer overflow Modified: trunk/libvo/vo_kva.c Modified: trunk/libvo/vo_kva.c ============================================================================== --- trunk/libvo/vo_kva.c Tue May 21 21:21:41 2019 (r38139) +++ trunk/libvo/vo_kva.c Fri May 24 13:42:53 2019 (r38140) @@ -608,7 +608,7 @@ static int preinit(const char *arg) {NULL, 0, NULL, NULL} }; - PCSZ pcszVideoModeStr[3] = {"DIVE", "WarpOverlay!", "SNAP", "VMAN"}; + PCSZ pcszVideoModeStr[] = {"DIVE", "WarpOverlay!", "SNAP", "VMAN"}; if (subopt_parse(arg, subopts) != 0) return -1;

3 2

r38142 - trunk/configure
by reimar 26 May '19

26 May '19

Author: reimar Date: Sun May 26 21:56:32 2019 New Revision: 38142 Log: configure: update zlib-dependent codec list. Fixes compilation without zlib. Modified: trunk/configure Modified: trunk/configure ============================================================================== --- trunk/configure Sun May 26 21:56:31 2019 (r38141) +++ trunk/configure Sun May 26 21:56:32 2019 (r38142) @@ -6605,7 +6605,7 @@ if test "$_zlib" = yes ; then mplayer_encoders="$mplayer_encoders PNG_ENCODER" else def_zlib='#define CONFIG_ZLIB 0' - libavdecoders=$(filter_out_component decoder 'APNG FLASHSV FLASHSV2 PNG ZMBV ZLIB DXA EXR G2M MSCC MWSC RASC RSCC SCREENPRESSO SRGC TDSC TSCC ZEROCODEC WCMV') + libavdecoders=$(filter_out_component decoder 'APNG FLASHSV FLASHSV2 PNG ZMBV ZLIB DXA EXR G2M LSCR MSCC MWSC RASC RSCC SCREENPRESSO SRGC TDSC TSCC ZEROCODEC WCMV') libavencoders=$(filter_out_component encoder 'APNG FLASHSV FLASHSV2 PNG ZMBV ZLIB') fi echores "$_zlib"

1 0

r38141 - in trunk: configure libaf/af_format.c libaf/af_volume.c libmpcodecs/ad_ffmpeg.c
by reimar 26 May '19

26 May '19

Author: reimar Date: Sun May 26 21:56:31 2019 New Revision: 38141 Log: configure: improve aarch64 handling. Automatically detect 64- vs. 32-bit build like for other architectures. Also detect NEON and VFP support correctly. Lastly, disable inline assembly NEON code on AARCH64 that only works on 32-bit. Modified: trunk/configure trunk/libaf/af_format.c trunk/libaf/af_volume.c trunk/libmpcodecs/ad_ffmpeg.c Modified: trunk/configure ============================================================================== --- trunk/configure Fri May 24 13:42:53 2019 (r38140) +++ trunk/configure Sun May 26 21:56:31 2019 (r38141) @@ -270,9 +270,16 @@ alpha() { esac } +aarch64() { + case "$host_arch" in + aarch64*) return 0;; + *) return 1;; + esac +} + arm() { case "$host_arch" in - arm*) return 0;; + arm*|aarch64*) return 0;; *) return 1;; esac } @@ -1895,6 +1902,9 @@ fi if sparc ; then cc_check && host_arch=sparc64 || host_arch=sparc fi +if arm ; then + cc_check && host_arch=aarch64 || host_arch=arm +fi fi echo "Detected operating system: $system_name" @@ -3223,6 +3233,7 @@ if arm ; then if test $_armvfp = "auto" ; then _armvfp=no inline_asm_check '"fadds s0, s0, s0"' && _armvfp=yes + inline_asm_check '"fmadd d0, d0, d1, d2"' && _armvfp=yes fi echores "$_armvfp" @@ -3254,9 +3265,15 @@ if arm ; then if test $neon = "auto" ; then neon=no inline_asm_check '"vadd.i16 q0, q0, q0"' && neon=yes + inline_asm_check '"ext v0.8B, v0.8B, v1.8B, #1"' && neon=yes fi echores "$neon" + echocheck "ARM NEON intrinsics" + neon_intrinsics=no + statement_check "arm_neon.h" "int16x8_t test = vdupq_n_s16(0)" && neon_intrinsics=yes + echores "$neon_intrinsics" + echocheck "ARM THUMB" if test $armthumb = "auto" ; then armthumb=no @@ -3265,7 +3282,7 @@ if arm ; then extra_cflags="$extra_cflags -mthumb" def_armthumb='#define CONFIG_THUMB 1' else - extra_cflags="$extra_cflags -marm" + aarch64 || extra_cflags="$extra_cflags -marm" def_armthumb='#define CONFIG_THUMB 0' fi echores "$armthumb" @@ -3278,7 +3295,8 @@ if arm ; then echores "$_iwmmxt" fi -cpuexts_all='ALTIVEC XOP AVX AVX2 AVX512 FMA3 FMA4 MMX MMX2 MMXEXT AMD3DNOW AMD3DNOWEXT SSE SSE2 SSE3 SSSE3 SSE4 SSE42 FAST_CMOV I686 FAST_CLZ ARMV5TE ARMV6 ARMV6T2 VFP VFPV3 SETEND NEON IWMMXT MMI VIS MVI' +intrinsics="none" +cpuexts_all='ALTIVEC XOP AVX AVX2 AVX512 FMA3 FMA4 MMX MMX2 MMXEXT AMD3DNOW AMD3DNOWEXT SSE SSE2 SSE3 SSSE3 SSE4 SSE42 FAST_CMOV I686 FAST_CLZ ARMV5TE ARMV6 ARMV6T2 ARMV8 VFP VFPV3 SETEND NEON INTRINSICS_NEON IWMMXT MMI VIS MVI' test "$_altivec" = yes && cpuexts="ALTIVEC $cpuexts" test "$_mmx" = yes && cpuexts="MMX $cpuexts" test "$_mmxext" = yes && cpuexts="MMX2 $cpuexts" @@ -3303,10 +3321,12 @@ test "$_fast_clz" = yes && cpuexts="FAS test "$_armv5te" = yes && cpuexts="ARMV5TE $cpuexts" test "$_armv6" = yes && cpuexts="ARMV6 $cpuexts" test "$_armv6t2" = yes && cpuexts="ARMV6T2 $cpuexts" +aarch64 && cpuexts="ARMV8 $cpuexts" test "$_armvfp" = yes && cpuexts="VFP $cpuexts" test "$vfpv3" = yes && cpuexts="VFPV3 $cpuexts" test "$setend" = yes && cpuexts="SETEND $cpuexts" test "$neon" = yes && cpuexts="NEON $cpuexts" +test "$neon_intrinsics" = yes && cpuexts="INTRINSICS_NEON $cpuexts" && intrinsics="neon" test "$_iwmmxt" = yes && cpuexts="IWMMXT $cpuexts" test "$_vis" = yes && cpuexts="VIS $cpuexts" test "$_mvi" = yes && cpuexts="MVI $cpuexts" @@ -8790,7 +8810,7 @@ HAVE_W32THREADS = $_w32threads HAVE_THREADS = $_threads HAVE_YASM = $have_yasm HAVE_X86ASM = $have_yasm -INTRINSICS = none +INTRINSICS = $intrinsics CONFIG_LIBXVID = $_xvid_lavc $(mak_enable "$libavdecoders_all" "$libavdecoders" CONFIG) @@ -9340,7 +9360,6 @@ $def_openssl #define HAVE_INLINE_ASM 1 #define HAVE_INLINE_ASM_NONLOCAL_LABELS 1 #define HAVE_INLINE_ASM_DIRECT_SYMBOL_REFS 1 -#define HAVE_INTRINSICS_NEON 0 #define HAVE_ISATTY 0 #define HAVE_LDBRX 0 #define HAVE_LIBC_MSVCRT 0 Modified: trunk/libaf/af_format.c ============================================================================== --- trunk/libaf/af_format.c Fri May 24 13:42:53 2019 (r38140) +++ trunk/libaf/af_format.c Sun May 26 21:56:31 2019 (r38141) @@ -494,7 +494,7 @@ static void float2int(const float* in, v ((int8_t *)out)[i] = av_clip_int8(lrintf(128.0f * in[i])); break; case(2): -#if HAVE_NEON +#if HAVE_NEON && !ARCH_AARCH64 { const float *in_end = in + len; while (in < in_end - 7) { Modified: trunk/libaf/af_volume.c ============================================================================== --- trunk/libaf/af_volume.c Fri May 24 13:42:53 2019 (r38140) +++ trunk/libaf/af_volume.c Sun May 26 21:56:31 2019 (r38141) @@ -125,7 +125,7 @@ static av_always_inline void s16_inner_l static av_always_inline void float_inner_loop(float *data, int len, int offset, int step, float level, int softclip) { int i; -#if HAVE_NEON +#if HAVE_NEON && !ARCH_AARCH64 if (offset == 0 && step == 1 && !softclip && len >= 8) { __asm__( Modified: trunk/libmpcodecs/ad_ffmpeg.c ============================================================================== --- trunk/libmpcodecs/ad_ffmpeg.c Fri May 24 13:42:53 2019 (r38140) +++ trunk/libmpcodecs/ad_ffmpeg.c Sun May 26 21:56:31 2019 (r38141) @@ -223,7 +223,7 @@ static av_always_inline void copy_sample { size_t s, c, o = 0; -#if HAVE_NEON_INLINE +#if HAVE_NEON_INLINE && !ARCH_AARCH64 if (nb_channels == 2 && bps == 4) { const unsigned char *src0 = src[0]; const unsigned char *src1 = src[1];

1 0

r38139 - trunk/libmpcodecs/vf_noise.c
by reimar 21 May '19

21 May '19

Author: reimar Date: Tue May 21 21:21:41 2019 New Revision: 38139 Log: vf_noise: make static array const. Modified: trunk/libmpcodecs/vf_noise.c Modified: trunk/libmpcodecs/vf_noise.c ============================================================================== --- trunk/libmpcodecs/vf_noise.c Mon May 20 19:45:17 2019 (r38138) +++ trunk/libmpcodecs/vf_noise.c Tue May 21 21:21:41 2019 (r38139) @@ -72,7 +72,7 @@ struct vf_priv_s { static int nonTempRandShift_init; static int nonTempRandShift[MAX_RES]; -static int patt[4] = { +static const int patt[4] = { -1,0,1,0 };

1 0

r38138 - trunk/configure
by reimar 20 May '19

20 May '19

Author: reimar Date: Mon May 20 19:45:17 2019 New Revision: 38138 Log: Add check for stdatomic.h If not found, try adding the gcc compat header. Modified: trunk/configure Modified: trunk/configure ============================================================================== --- trunk/configure Mon May 20 19:45:16 2019 (r38137) +++ trunk/configure Mon May 20 19:45:17 2019 (r38138) @@ -3897,6 +3897,14 @@ fi echores "$_fast_inttypes" +echocheck "stdatomic.h" +stdatomic_h=no +header_check stdatomic.h && stdatomic_h=yes +echores "$stdatomic_h" +if test "$stdatomic_h" = no ; then + CFLAGS="-Icompat/atomics/gcc $CFLAGS" +fi + echocheck "malloc.h" _malloc=no header_check malloc.h && _malloc=yes

1 0