Use the correct position in the SSE resamplers for left-over processing

This commit is contained in:
Chris Robinson 2015-10-25 14:07:58 -07:00
parent fc03a955e9
commit 90349b384f
4 changed files with 23 additions and 10 deletions

View File

@ -182,6 +182,10 @@ void Mix_SSE(const ALfloat *data, ALuint OutChans, ALfloat (*restrict OutBuffer)
_mm_store_ps(&OutBuffer[c][OutPos+pos], dry4);
pos += 4;
} while(minsize-pos > 3);
/* NOTE: gain4 now represents the next four gains after the
* last four mixed samples, so the lowest element represents
* the next gain to apply.
*/
gain = _mm_cvtss_f32(gain4);
}
/* Mix with applying left over gain steps that aren't aligned multiples of 4. */

View File

@ -63,7 +63,10 @@ const ALfloat *Resample_lerp32_SSE2(const ALfloat *src, ALuint frac, ALuint incr
_mm_store_ps(pos_.f, _mm_castsi128_ps(pos4));
}
pos = pos_.i[3];
/* NOTE: These four elements represent the position *after* the last four
* samples, so the lowest element is the next position to resample.
*/
pos = pos_.i[0];
frac = _mm_cvtsi128_si32(frac4);
for(;i < numsamples;i++)

View File

@ -78,8 +78,11 @@ const ALfloat *Resample_fir4_32_SSE3(const ALfloat *src, ALuint frac, ALuint inc
_mm_store_ps(frac_.f, _mm_castsi128_ps(frac4));
}
pos = pos_.i[3];
frac = frac_.i[3];
/* NOTE: These four elements represent the position *after* the last four
* samples, so the lowest element is the next position to resample.
*/
pos = pos_.i[0];
frac = frac_.i[0];
for(;i < numsamples;i++)
{
@ -143,8 +146,8 @@ const ALfloat *Resample_fir8_32_SSE3(const ALfloat *src, ALuint frac, ALuint inc
_mm_store_ps(frac_.f, _mm_castsi128_ps(frac4));
}
pos = pos_.i[3];
frac = frac_.i[3];
pos = pos_.i[0];
frac = frac_.i[0];
for(;i < numsamples;i++)
{

View File

@ -67,7 +67,10 @@ const ALfloat *Resample_lerp32_SSE41(const ALfloat *src, ALuint frac, ALuint inc
pos_.i[3] = _mm_extract_epi32(pos4, 3);
}
pos = pos_.i[3];
/* NOTE: These four elements represent the position *after* the last four
* samples, so the lowest element is the next position to resample.
*/
pos = pos_.i[0];
frac = _mm_cvtsi128_si32(frac4);
for(;i < numsamples;i++)
@ -134,8 +137,8 @@ const ALfloat *Resample_fir4_32_SSE41(const ALfloat *src, ALuint frac, ALuint in
frac_.i[3] = _mm_extract_epi32(frac4, 3);
}
pos = pos_.i[3];
frac = frac_.i[3];
pos = pos_.i[0];
frac = frac_.i[0];
for(;i < numsamples;i++)
{
@ -205,8 +208,8 @@ const ALfloat *Resample_fir8_32_SSE41(const ALfloat *src, ALuint frac, ALuint in
frac_.i[3] = _mm_extract_epi32(frac4, 3);
}
pos = pos_.i[3];
frac = frac_.i[3];
pos = pos_.i[0];
frac = frac_.i[0];
for(;i < numsamples;i++)
{