usrp2: make the booty smaller than the number of recv frames

This may fix some of our woes when the host cannot keep up.
That is, with a smaller booty, the managed buffers will get
freed up and the call to get buffer will never block waiting
for a buffer to become free. This has several side effects:

1)
Overflows are more likely to occur in the pirate thread.
Pirate-based overflows will overwrite old packets, whereas
socket-based overflows will discard newer incoming packets.

2)
The pirate thread will continue to pull in async packets
rather than loosing them in a socket-based overflow.

1 files changed, 1 insertions, 1 deletions

diff --git a/host/lib/usrp/usrp2/io_impl.cpp b/host/lib/usrp/usrp2/io_impl.cpp
index eba704059..bbe9c273f 100644
--- a/host/lib/usrp/usrp2/io_impl.cpp
+++ b/host/lib/usrp/usrp2/io_impl.cpp
@@ -46,7 +46,7 @@ struct usrp2_impl::io_impl{
 
     io_impl(size_t num_frames, size_t width):
         packet_handler_recv_state(width),
-        recv_pirate_booty(alignment_buffer_type::make(num_frames, width)),
+        recv_pirate_booty(alignment_buffer_type::make(num_frames-3, width)),
         async_msg_fifo(bounded_buffer<async_metadata_t>::make(100/*messages deep*/))
     {
         /* NOP */