Switch pending event clearing to an atomic swap operation

Rather than performing an atomic bit clearing for every encountered
event bit set we can adjust the code to perform an atomic swap of a
single row of the events array and decrease the amount of expensive
atomic operations.

From FreeBSD.

diff --git a/sys/dev/pv/hyperv.c b/sys/dev/pv/hyperv.c
index 1aba897..7a8e22c 100644 (file)
--- a/sys/dev/pv/hyperv.c
+++ b/sys/dev/pv/hyperv.c
@@ -534,7 +534,7 @@ hv_event_intr(struct hv_softc *sc)
        int cpu = CPU_INFO_UNIT(ci);
        int bit, dword, maxdword, relid;
        struct hv_channel *ch;
-       uint32_t *revents;
+       uint32_t *revents, pending;
 
        evt = (struct hv_synic_event_flags *)sc->sc_siep[cpu] + HV_MESSAGE_SINT;
        if ((sc->sc_proto == HV_VMBUS_VERSION_WS2008) ||
@@ -559,8 +559,9 @@ hv_event_intr(struct hv_softc *sc)
        for (dword = 0; dword < maxdword; dword++) {
                if (revents[dword] == 0)
                        continue;
-               for (bit = 0; bit < 32; bit++) {
-                       if (!atomic_clearbit_ptr(&revents[dword], bit))
+               pending = atomic_swap_uint(&revents[dword], 0);
+               for (bit = 0; pending > 0; pending >>= 1, bit++) {
+                       if ((pending & 1) == 0)
                                continue;
                        relid = (dword << 5) + bit;
                        /* vmbus channel protocol message */