* commit '4efab89332ea39a77145e8b15562b981d9dbde68': x86: Use *_FAST/*_SLOW CPU feature detection macros where appropriate
Unified · Split
| @@ -228,16 +228,19 @@ av_cold void ff_ac3dsp_init_x86(AC3DSPContext *c, int bit_exact) · 227 unmodified lines | |||
| 228 | c->float_to_fixed24 = ff_float_to_fixed24_sse2; | 228 | c->float_to_fixed24 = ff_float_to_fixed24_sse2; |
| 229 | c->compute_mantissa_size = ff_ac3_compute_mantissa_size_sse2; | 229 | c->compute_mantissa_size = ff_ac3_compute_mantissa_size_sse2; |
| 230 | c->extract_exponents = ff_ac3_extract_exponents_sse2; | 230 | c->extract_exponents = ff_ac3_extract_exponents_sse2; |
| 231 | if (!(cpu_flags & AV_CPU_FLAG_SSE2SLOW)) { | ||
| 232 | c->ac3_lshift_int16 = ff_ac3_lshift_int16_sse2; | ||
| 233 | c->ac3_rshift_int32 = ff_ac3_rshift_int32_sse2; | ||
| 234 | } | ||
| 235 | if (bit_exact) { | 231 | if (bit_exact) { |
| 236 | c->apply_window_int16 = ff_apply_window_int16_sse2; | 232 | c->apply_window_int16 = ff_apply_window_int16_sse2; |
| 237 | } else if (!(cpu_flags & AV_CPU_FLAG_SSE2SLOW)) { | 233 | } |
| 234 | } | ||
| 235 | |||
| 236 | if (EXTERNAL_SSE2_FAST(cpu_flags)) { | ||
| 237 | c->ac3_lshift_int16 = ff_ac3_lshift_int16_sse2; | ||
| 238 | c->ac3_rshift_int32 = ff_ac3_rshift_int32_sse2; | ||
| 239 | if (!bit_exact) { | ||
| 238 | c->apply_window_int16 = ff_apply_window_int16_round_sse2; | 240 | c->apply_window_int16 = ff_apply_window_int16_round_sse2; |
| 239 | } | 241 | } |
| 240 | } | 242 | } |
| 243 | |||
| 241 | if (EXTERNAL_SSSE3(cpu_flags)) { | 244 | if (EXTERNAL_SSSE3(cpu_flags)) { |
| 242 | c->ac3_max_msb_abs_int16 = ff_ac3_max_msb_abs_int16_ssse3; | 245 | c->ac3_max_msb_abs_int16 = ff_ac3_max_msb_abs_int16_ssse3; |
| 243 | if (cpu_flags & AV_CPU_FLAG_ATOM) { | 246 | if (cpu_flags & AV_CPU_FLAG_ATOM) { |
| @@ -562,11 +562,6 @@ av_cold void ff_h264qpel_init_x86(H264QpelContext *c, int bit_depth) · 561 unmodified lines | |||
| 562 | } | 562 | } |
| 563 | 563 | ||
| 564 | if (EXTERNAL_SSE2(cpu_flags)) { | 564 | if (EXTERNAL_SSE2(cpu_flags)) { |
| 565 | if (!(cpu_flags & AV_CPU_FLAG_SSE2SLOW) && !high_bit_depth) { | ||
| 566 | // these functions are slower than mmx on AMD, but faster on Intel | ||
| 567 | H264_QPEL_FUNCS(0, 0, sse2); | ||
| 568 | } | ||
| 569 | |||
| 570 | if (!high_bit_depth) { | 565 | if (!high_bit_depth) { |
| 571 | H264_QPEL_FUNCS(0, 1, sse2); | 566 | H264_QPEL_FUNCS(0, 1, sse2); |
| 572 | H264_QPEL_FUNCS(0, 2, sse2); | 567 | H264_QPEL_FUNCS(0, 2, sse2); |
| @@ -593,6 +588,12 @@ av_cold void ff_h264qpel_init_x86(H264QpelContext *c, int bit_depth) · 20 unmodified lines | |||
| 593 | } | 588 | } |
| 594 | } | 589 | } |
| 595 | 590 | ||
| 591 | if (EXTERNAL_SSE2_FAST(cpu_flags)) { | ||
| 592 | if (!high_bit_depth) { | ||
| 593 | H264_QPEL_FUNCS(0, 0, sse2); | ||
| 594 | } | ||
| 595 | } | ||
| 596 | |||
| 596 | if (EXTERNAL_SSSE3(cpu_flags)) { | 597 | if (EXTERNAL_SSSE3(cpu_flags)) { |
| 597 | if (!high_bit_depth) { | 598 | if (!high_bit_depth) { |
| 598 | H264_QPEL_FUNCS(1, 0, ssse3); | 599 | H264_QPEL_FUNCS(1, 0, ssse3); |
| @@ -154,7 +154,7 @@ av_cold void ff_lpc_init_x86(LPCContext *c) · 153 unmodified lines | |||
| 154 | #if HAVE_SSE2_INLINE | 154 | #if HAVE_SSE2_INLINE |
| 155 | int cpu_flags = av_get_cpu_flags(); | 155 | int cpu_flags = av_get_cpu_flags(); |
| 156 | 156 | ||
| 157 | if (HAVE_SSE2_INLINE && cpu_flags & (AV_CPU_FLAG_SSE2 | AV_CPU_FLAG_SSE2SLOW)) { | 157 | if (INLINE_SSE2(cpu_flags) || INLINE_SSE2_SLOW(cpu_flags)) { |
| 158 | c->lpc_apply_welch_window = lpc_apply_welch_window_sse2; | 158 | c->lpc_apply_welch_window = lpc_apply_welch_window_sse2; |
| 159 | c->lpc_compute_autocorr = lpc_compute_autocorr_sse2; | 159 | c->lpc_compute_autocorr = lpc_compute_autocorr_sse2; |
| 160 | } | 160 | } |
| @@ -346,7 +346,7 @@ av_cold void ff_vp78dsp_init_x86(VP8DSPContext *c) · 345 unmodified lines | |||
| 346 | c->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_sse; | 346 | c->put_vp8_bilinear_pixels_tab[0][0][0] = ff_put_vp8_pixels16_sse; |
| 347 | } | 347 | } |
| 348 | 348 | ||
| 349 | if (HAVE_SSE2_EXTERNAL && cpu_flags & (AV_CPU_FLAG_SSE2 | AV_CPU_FLAG_SSE2SLOW)) { | 349 | if (EXTERNAL_SSE2(cpu_flags) || EXTERNAL_SSE2_SLOW(cpu_flags)) { |
| 350 | VP8_LUMA_MC_FUNC(0, 16, sse2); | 350 | VP8_LUMA_MC_FUNC(0, 16, sse2); |
| 351 | VP8_MC_FUNC(1, 8, sse2); | 351 | VP8_MC_FUNC(1, 8, sse2); |
| 352 | VP8_BILINEAR_MC_FUNC(0, 16, sse2); | 352 | VP8_BILINEAR_MC_FUNC(0, 16, sse2); |
| @@ -416,7 +416,7 @@ av_cold void ff_vp8dsp_init_x86(VP8DSPContext *c) · 63 unmodified lines | |||
| 416 | c->vp8_luma_dc_wht = ff_vp8_luma_dc_wht_sse; | 416 | c->vp8_luma_dc_wht = ff_vp8_luma_dc_wht_sse; |
| 417 | } | 417 | } |
| 418 | 418 | ||
| 419 | if (HAVE_SSE2_EXTERNAL && cpu_flags & (AV_CPU_FLAG_SSE2 | AV_CPU_FLAG_SSE2SLOW)) { | 419 | if (EXTERNAL_SSE2(cpu_flags) || EXTERNAL_SSE2_SLOW(cpu_flags)) { |
| 420 | c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_sse2; | 420 | c->vp8_v_loop_filter_simple = ff_vp8_v_loop_filter_simple_sse2; |
| 421 | 421 | ||
| 422 | c->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16y_inner_sse2; | 422 | c->vp8_v_loop_filter16y_inner = ff_vp8_v_loop_filter16y_inner_sse2; |