A flowgraph that shows how to issue FLT/VOL commands from top to bottom, instead of from left to right, to allow an EQ/CMP to be processed in parallel and improve performance.

As a best practice for ordering flowgraphs among commands of a particular type, favor depth through a voice's processing path, up to a submix stage or up to reused SHAPE components. Then, favor breadth for voices that share common processing orders. Figure 1 shows that the optimal (and typical) approach is to create FLT/VOL commands for the leftmost column of FLT/VOLs first, and then to pivot the remaining FLT/VOL commands for the two channels of the top stereo voice. ### Allocation of mix buffer instances Mix buffers are locked while being written to. No output can be read from them until all inputs have contributed to the mix. From the perspective of implementation, mix buffers use the `numIn` and `numOut` fields to manage this. Therefore, it's possible to construct scenarios that require more than 128 physical mix buffers at the same time. SHAPE virtualizes mix buffers so that, even for high polyphony, this scenario usually won't cause an issue. However, you can construct flowgraphs that require high numbers of mix buffers to be concurrently accessed. In particular, mix buffers that are submixing or mastering among multiple voices are locked until the last voice is mixed into them and until the last output of the mix buffers is consumed by a subsequent SHAPE block. In a typical scenario - a single 7.1 mastering voice into which all SHAPE voices are mixed - 8 of the 128 voices are locked throughout the frame. Additional mix buffers could lock for a significant portion of the frame (reducing the capacity for the virtual mix buffers) if there's extensive, multichannel submixing with flowgraph-dependency chains. Alternatively, additional mastering voices are created if, for instance, you were to create both a main speaker and a 7.1 headset-mix that's per-player for a four-player title, which could consume 40 concurrent mix buffers. In general, a case that requires more than 128 concurrent physical mix buffers is pathologic. However, mix buffers that are locked unnecessarily, for the entire frame, reduce your title's ability to achieve the potential throughput. Figure 2 represents unnecessary submixing that could reduce the number of mix buffers that are available at a particular time. Depending on command ordering (figure 2), the 5 separate 7.1 mastering mix buffers could be locked throughout almost the entire frame, which represents 40 of the 128 physical mix buffers. Although this path requires only 2 other mix buffers to be in use at a time - that is, the paired input and output for an FLT/VOL or EQ/CMP - consider first processing all the mix buffers going into speakers. You can then free those mix buffers and process headsets 1, 2, and more. **Figure 2. A flowgraph that shows unnecessary submixing and locking of allocated mix buffer instances due to poor command ordering.**

A flowgraph that shows unnecessary submixing and locking of allocated mix buffer instances due to poor command ordering.

### Selection and balancing of SHAPE components (FLT/VOL vs. EQ/CMP) If you use the same SHAPE components repeatedly in sequence, you can reduce the ability to reach the maximum throughput. By comparison, by alternating different components, you ensure that each SHAPE block is doing meaningful parallel processing. Figure 3 shows a common example: FLT/VOLs can't all be processed in parallel. Performance might suffer. **Figure 3. A flowgraph that shows an attempt to process all FLT/VOLs in parallel.**

A flowgraph that shows an attempt to process all FLT/VOLs in parallel.

The second set of FLT/VOLs must wait for the first to be completed, potentially while other SHAPE blocks remain unused. If either set were used exclusively for filtering, the EQ/CMP might have represented a better choice for one set, especially if the flowgraph (figure 4) with the same configuration had more voices. This could then begin their FLT/VOL while the first set was processing its EQ/CMP blocks. Figure 4 shows performance that more closely approaches the ideal SHAPE capabilities. This flowgraph replaces one set of FLT/VOLs, presumed to be used for filtering only (see figure 3), with EQ/CMP, which allows for parallel processing of the two SHAPE components. **Figure 4. A flowgraph that shows an attempt to process FLT/VOLs and EQ/CMPs in parallel.**

A flowgraph that shows an attempt to process FLT/VOLs and EQ/CMPs in parallel.

Balancing of SHAPE components is also relevant when you consider the maximum instancing of the various components. The 2560 FLT/VOL components that can occur within a frame allow the FLT/VOL hardware to run approximately five times more calculations than is possible with the 512 EQ/CMP blocks. If you run an EQ/CMP block in series with a single FLT/VOL, you'll realize less performance than you would by running multiple FLT/VOL components that are coupled with an EQ/CMP because the former's processing will be gated by the latter's consumption. ### Too many or frequent low-value DMA roundtrips DMA is bound by the read/write bandwidth available to the SHAPE's bus. If you consume all available bandwidth - though this scenario is usually pathological - DMA blocks will stall and reduce throughput. As discussed earlier, all processing that occurs after a DMA input block will have to wait until the DMA is completed. Therefore, you can create flowgraphs with much of the flowgraph blocked and waiting for a DMA to be completed before the flowgraph can begin processing. You should also evaluate DMA roundtrips in relation to the value they present to the audio stream. The minimal addition of latency (which you can control in terms of the number of frames accessed via DMA at multiples of the 2.667 ms size of the audio frame) might not be a concern for most audio playback scenarios. This is because blocks to the right of a DMA will be processing audio data from previous frames. In particular, sending a DMA back into SHAPE just to perform a final mix might be unnecessary when you also plan to send that final mix as DMA back out to memory for presentation to an audio endpoint. Figure 5 shows an example of a potentially low value, where a voice sent via DMA to memory for CPU/GPU processing is brought back into SHAPE, only to be 7.1-panned into the mastering mix. The voice is then sent via DMA back out to memory. Figure 5 assumes that no filtering is being applied to voices in the FLT/VOL blocks and that the 7.1 mastering mix buffers have additional voices routed to them. **Figure 5. A flowgraph that shows a potentially low-value routing, involving unnecessary FLT/VOL components and unnecessary DMA roundtrips.**

A flowgraph that shows a potentially low-value routing, involving unnecessary FLT/VOL components and unnecessary DMA roundtrips.

Figure 6 shows potentially more-optimal routing of figure 5's intent. The voice is sent via DMA to memory for CPU/GPU processing. The voice can then be 7.1-panned and combined with the rest of the 7.1 mix on the CPU itself without the overhead of additional DMAs, additional audio frame latency, or the use of seven FLT/VOL components. **Figure 6. A flowgraph that shows a more optimal routing without additional DMA roundtrips or unnecessary FLT/VOL components.**

A flowgraph that shows a more optimal routing without additional DMA roundtrips or unnecessary FLT/VOL components.