Voice Allocation

Overview

• Call AXAcquireVoice() to acquire a voice. The caller must specify a priority for the voice.
• Call AXFreeVoice() to free a voice once the voice is no longer needed.
• Call AXSetVoicePriority() to adjust, as needed, the priority of an acquired voice.
• Call AXGetDSPCycles() to retrieve the number of DSP cycles used by AX during the last audio frame.
• Call AXSetMaxDSPCycles() to adjust the maximum number of DSP cycles per frame that AX may use.
• Call AXGetMaxDSPCycles() to retrieve the maximum number of DSP cycles per frame that AX may use.

Voice Priority

• AX will first expend DSP cycles on the highest priority voices that are newest, during each audio frame.
• Thus, older voices of lower priority will be dropped if AX exhausts the allowed DSP cycles before processing all voices.

Voice Reacquisition

• If the number of active voices reaches AX_MAX_VOICES, the lowest priority voice of greatest age will be forcibly reacquired for subsequent acquisition requests.
• Upon reacquiring a voice, AX will invoke the voice's associated callback, if any.
  Note: Your application specifies the callback when calling AXAcquireVoice().
• Upon receiving a 'dropped voice' callback, the user application must remove any references to that voice, to ensure that it is no longer serviced by the user's abstraction layer/voice management scheme.
• The user application must not free the voice upon receiving the dropped-voice callback. This is because the voice has been forcibly reacquired by AX. In other words, your application must no longer process this voice.

DSP Load Management

• If the DSP takes longer than expected or is otherwise late when processing a frame, AX will begin dropping voices based on priority, and then age. The oldest of the lowest priority voices will be dropped first.
• AX will invoke the associated callback (if any) for each voice, as it is being dropped.
• Upon receiving a 'dropped voice' callback, you must remove any references to that voice, to ensure that it is no longer serviced by your abstraction layer/voice management scheme.
• Your application must not free the voice during the 'dropped-voice' callback.

Revision History

03/01/2006 Initial version.