The concern is that a massive int sample, like 0x7FFFFFFF, won't fit in a
float32, which doesn't have enough bits to hold a whole number this large,
just to divide it to get a value between 0 and 1.
Previously we would convert to double, to get more bits, do the division, and
cast back to a float, but this is expensive.
Casting to double is more accurate, but it's 2x to 3x slower. Shifting out
the least significant byte of an int32, so it'll definitely fit in a float,
and dividing by 0x7FFFFF is still accurate to about 5 decimal places, and the
difference doesn't appear to be perceptable.
SDL now builds with gcc 7.2 with the following command line options:
-Wall -pedantic-errors -Wno-deprecated-declarations -Wno-overlength-strings --std=c99
XAudio2 doesn't have capture support, so WASAPI was to replace it; the holdout
was WinRT, which still needed it as its primary audio target until the WASAPI
code code be made to work.
The support matrix now looks like:
WinXP: directsound by default, winmm as a fallback for buggy drivers.
Vista+: WASAPI (directsound and winmm as fallbacks for debugging).
WinRT: WASAPI
Simon Hug
Patch that adds [-1, 1] clamping to the scalar audio type conversions.
This may come from the SDL_Convert_F32_to_X_Scalar functions. They don't clamp the float value to [-1, 1] and when they cast it to the target integer it may be too large or too small for the type and get truncated, causing horrible noise.
The attached patch throws clamping in, but I don't know if that's the preferred way to fix this. For x86 (without SSE) the compiler (I tested MSVC) seems to throw a horrible amount of x87 code in it. It's a bit better with SSE, but probably still quite the performance hit. And SSE2 uses a branchless approach with maxss and minss.
The audioqueue thread needs to keep running, and processing the CFRunLoop
until the AudioQueue is disposed of, otherwise CoreAudio will hang waiting for
final data to feed the device.
At least, I think this is how it all works. It definitely fixes the bug here!
Since AudioQueueDispose() calls AudioQueueStop() internally, there's no need
for our thread to handle this, either, which is good because the AudioQueue
would be disposed by this point. So now the AudioQueue is disposed first, and
then our thread is joined, and everything works out okay.
Just in case, we mark the device "paused" before setting everything in motion,
so any further callbacks from CoreAudio will write silence and not fire the
app's audio callback again.
Fixes Bugzilla #3868.
(I thought padding size ranged from 5 frames to ~30 frames (based around
RESAMPLER_ZERO_CROSSINGS, which is 5), but it's actually between 512 and
several thousands (based on RESAMPLER_SAMPLES_PER_ZERO_CROSSING)). It gets
big fast when downsampling.
Previously, the padding was silence, which was a problem when streaming since
you would sample a little bit of this silence between each buffer.
We still need a means to get padding data for the right hand side, but this
patch makes the resampler output more correct.
This time it's using real math from a real whitepaper instead of my previous
amateur, fast-but-low-quality attempt. The new resampler does "bandlimited
interpolation," as described here: https://ccrma.stanford.edu/~jos/resample/
The output appears to sound cleaner, especially at high frequencies, and of
course works with non-power-of-two rate conversions.
There are some obvious optimizations to be done to this still, and there is
other fallout: this doesn't resample a buffer in-place, the 2-channels-Sint16
fast path is gone because this resampler does a _lot_ of floating point math.
There is a nasty hack to make it work with SDL_AudioCVT.
It's possible these issues are solvable, but they aren't solved as of yet.
Still, I hope this effort is slouching in the right direction.
This would cause playback problems in certain situations, such as on the
Raspberry Pi. The device that the wait was added for seems to not benefit from
it in modern times, and standard desktop Linux seems to do the right thing
when a USB device is unplugged now, without this patch.
Fixes Bugzilla #3599.
Simon Hug
This issue actually raises the question if this API change (requirement of initialized audio subsystem) is breaking backwards compatibility. I don't see the documentation saying it is needed in 2.0.5.
"Major changes, roughly in order of appearance:
- Use float math everywhere, instead of promoting to double and casting back
all the time.
- Conserve sound energy when downmixing any channel into two other channels.
- Add a QuadToStereo filter. (The previous technique of reusing StereoToMono
never worked, since it assumed an incorrect channel layout for 4.0.)
- Add a 71to51 filter. This removes just under half of the cases the previous
code would silently break in.
- Add a QuadTo51 filter. More silent breakage fixed.
- Add a 51to71 filter, removing another almost-half of the silently broken
cases.
- Add 8 to the list of values SDL_SupportedChannelCount will accept.
- Change SDL_BuildAudioCVT's channel-related logic to handle every case, and
to actually fail if it fails instead of silently corrupting sound data and/or
crashing down the road."
(Note that SDL doesn't otherwise support 7.1 audio yet, but hopefully it will
soon and the 7.1 converters are an important piece of that. --ryan.)
Fixes Bugzilla #3727.
