diff options
Diffstat (limited to 'liba52/liba52_changes.diff')
| -rw-r--r-- | liba52/liba52_changes.diff | 34 |
1 files changed, 17 insertions, 17 deletions
diff --git a/liba52/liba52_changes.diff b/liba52/liba52_changes.diff index 09eefbd617..10cbc94b08 100644 --- a/liba52/liba52_changes.diff +++ b/liba52/liba52_changes.diff @@ -35,13 +35,13 @@ expbap_t * expbap); int a52_downmix_init (int input, int flags, sample_t * level, - sample_t clev, sample_t slev); + sample_t clev, sample_t slev); +void downmix_accel_init(uint32_t mm_accel); int a52_downmix_coeff (sample_t * coeff, int acmod, int output, sample_t level, - sample_t clev, sample_t slev); + sample_t clev, sample_t slev); -void a52_downmix (sample_t * samples, int acmod, int output, sample_t bias, +extern void (*a52_downmix) (sample_t * samples, int acmod, int output, sample_t bias, - sample_t clev, sample_t slev); + sample_t clev, sample_t slev); -void a52_upmix (sample_t * samples, int acmod, int output); +extern void (*a52_upmix) (sample_t * samples, int acmod, int output); @@ -120,7 +120,7 @@ +#endif //!ARCH_X86 + +#endif -+ ++ /* (stolen from the kernel) */ #ifdef WORDS_BIGENDIAN @@ -254,9 +254,9 @@ + if(mm_accel & MM_ACCEL_X86_3DNOW) a52_downmix= downmix_3dnow; +#endif +} -+ ++ int a52_downmix_init (int input, int flags, sample_t * level, - sample_t clev, sample_t slev) + sample_t clev, sample_t slev) { @@ -451,7 +480,7 @@ samples[i] = 0; @@ -264,7 +264,7 @@ -void a52_downmix (sample_t * samples, int acmod, int output, sample_t bias, +void downmix_C (sample_t * samples, int acmod, int output, sample_t bias, - sample_t clev, sample_t slev) + sample_t clev, sample_t slev) { switch (CONVERT (acmod, output & A52_CHANNEL_MASK)) { @@ -563,7 +592,7 @@ @@ -1533,14 +1533,14 @@ +static float __attribute__((aligned(16))) *sseW[7]= + {NULL /*sseW0*/,sseW1,sseW2,sseW3,sseW4,sseW5,sseW6}; +static float __attribute__((aligned(16))) sseWindow[512]; -+#endif -+ ++#endif ++ /* Root values for IFFT */ static sample_t roots16[3]; static sample_t roots32[7]; @@ -245,7 +322,7 @@ ifft_pass (buf, roots128 - 32, 32); - } + } -void a52_imdct_512 (sample_t * data, sample_t * delay, sample_t bias) +void imdct_do_512 (sample_t * data, sample_t * delay, sample_t bias) @@ -1973,7 +1973,7 @@ + "add $16, %%"REG_S" \n\t" + "add $2, %%"REG_a" \n\t" // avoid complex addressing for P4 crap + "sub $16, %%"REG_D" \n\t" -+ " jnc 1b \n\t" ++ "jnc 1b \n\t" + "pop %%"REG_BP" \n\t"//no we didnt touch ebp *g* + :: "b" (data), "c" (buf) + : "%"REG_S, "%"REG_D, "%"REG_a, "%"REG_d @@ -2125,7 +2125,7 @@ + "movaps %%xmm0, (%%"REG_d", %%"REG_D") \n\t" + "add $16, %%"REG_D" \n\t" + "cmp %3, %%"REG_D" \n\t" //FIXME (opt) count against 0 -+ " jb 2b \n\t" ++ "jb 2b \n\t" + "add %2, %%"REG_S" \n\t" + "cmp %1, %%"REG_S" \n\t" + " jb 1b \n\t" @@ -2265,7 +2265,7 @@ @@ -368,7 +1153,7 @@ void a52_imdct_init (uint32_t mm_accel) - { + { - int i, k; + int i, j, k; double sum; @@ -2274,7 +2274,7 @@ @@ -420,6 +1205,99 @@ post2[i].real = cos ((M_PI / 128) * (i + 0.5)); post2[i].imag = sin ((M_PI / 128) * (i + 0.5)); - } + } + for (i = 0; i < 128; i++) { + xcos1[i] = -cos ((M_PI / 2048) * (8 * i + 1)); + xsin1[i] = -sin ((M_PI / 2048) * (8 * i + 1)); @@ -2345,7 +2345,7 @@ + { + fprintf (stderr, "Using SSE optimized IMDCT transform\n"); + a52_imdct_512 = imdct_do_512_sse; -+ } ++ } + else + if(mm_accel & MM_ACCEL_X86_3DNOWEXT) + { @@ -2366,7 +2366,7 @@ + fprintf(stderr, "Using AltiVec optimized IMDCT transform\n"); + a52_imdct_512 = imdct_do_512_altivec; + } -+ else ++ else +#endif #ifdef LIBA52_DJBFFT @@ -2437,7 +2437,7 @@ + mm_accel &=~MM_ACCEL_X86_SSE; + fprintf(stderr, "liba52: unable to get 16 byte aligned memory disabling usage of SSE instructions\n"); + } -+ ++ if (state->samples == NULL) { free (state); return NULL; |