With the fix to the apic timer interrupts, the thread scheduling latency has

apparently vastly improved, so that audio hardware buffers can be configured
as small as on BeOS in the OSS node. This is especially nice for multi-media
apps since it reduces interactive response times.


git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@27795 a95241bf-73f2-0310-859d-f6bbb57e9c96

src/add-ons/media/media-add-ons/opensound/OpenSoundDeviceEngine.cpp

@@ -98,20 +98,8 @@ status_t OpenSoundDeviceEngine::Open(int mode)
 
 	// set driver buffer size by using the "fragments" API
 	// TODO: export this setting as a BParameter?
-
-	// NOTE stippi: 2048 bytes driver buffer is long enough for playback on
-	// BeOS. On Haiku, testing on HD Audio hardware, it is too short. However,
-	// I seem to remember the HD Audio supports 32 bit sample width (while
-	// C-Media supports "only" 16). If OSS uses the same 2048 bytes even for
-	// 32 bit/sample, then I could see how that would be asking for too much,
-	// since that would effectively half the latency.
-#ifdef HAIKU_TARGET_PLATFORM_HAIKU
-	uint32 bufferCount = 6;
-	uint32 bufferSize = 0x000b; // 1024 bytes
-#else
 	uint32 bufferCount = 4;
 	uint32 bufferSize = 0x000a; // 512 bytes
-#endif
 	v = (bufferCount << 16) | bufferSize;
 	if (ioctl(fFD, SNDCTL_DSP_SETFRAGMENT, &v, sizeof(int)) < 0) {
 		fInitCheckStatus = errno;