Merge from branch bqresample, with some resampler fixes
Hm, those settings weren't as well-behaved as I thought. I wonder if there is some unpleasant randomness here
337343780d3c — Chris Cannam 2 days ago
Settle on a single set of params that seems to work ok in tests
30b4a4910637 — Chris Cannam 3 days ago
Remove support for resampler from IPP versions older than v7
845a708dfb49 — Chris Cannam 3 days ago
Fix buffer underrun in case where rounded sample count accumulator exceeds length
dd9a2edad80e — Chris Cannam 3 days ago
Also avoid passing a genuinely long block when the ratio has just changed
9258a040f625 — Chris Cannam 3 days ago
Avoid passing a very oversize outcount for the reasons given
f82e54c2c4f8 — Chris Cannam 3 days ago
Resampler code pulled back in from bqresample, with safer api
ef795a7d6172 — Chris Cannam 4 days ago
Merge from default branch
8f1295fa2ed4 — Chris Cannam default 17 days ago
Avoid compiler complaints about dynamic array syntax. This is no better code, bu
t it's less sensitive to compile options.
74a8a12b8d3c — Chris Cannam 26 days ago
Prefer libsamplerate at any performance setting, now we know it is better-behaved when changing rates
b7a7fba73c31 — Chris Cannam 30 days ago
Add -fembed-bitcode
64e90588b59d — Chris Cannam a month ago
Note about preference for libsamplerate
b5f64f41c499 — Chris Cannam a month ago
In this emergency, increase the buffer size by a factor rather than only by a constant amount - because if this arises because of an overlong increment being broken down into littler bits, we would end up doing it again and again, which is much more expensive. This makes a huge performance difference in cases where a time map ends up mapping very little input into a great expanse of output
6433ca1736b8 — Chris Cannam a month ago
In this emergency, increase the buffer size by a factor rather than only by a constant amount - because if this arises because of an overlong increment being broken down into littler bits, we would end up doing it again and again, which is much more expensive. This makes a huge performance difference in cases where a time map ends up mapping very little input into a great expanse of output
23cd34a14117 — Chris Cannam a month ago
Clarify text about overall stretch factor
2a1a9c6a0596 — Chris Cannam a month ago
Same fix in resampleInterleaved as earlier committed in resample
6c0f84bf30b3 — Chris Cannam a month ago
Fix potential discontinuity when window size is reduced due to change in ratio
dbfc749bce16 — Chris Cannam a month ago
It appears this may be all we need - apparently the occasional odd extra sample (due to incremental rounding) added up and eventually overran the filter. I believe +1 is perfectly good here, but we can afford to be more accommodating just in case
0022acaba4a9 — Chris Cannam a month ago
Fix potential discontinuity when window size is reduced due to change in ratio
Next