diff options
| author |   Sebastian Sanchez <sebastian.sanchez@intel.com> | 2016-10-25 13:12:34 -0700 |
|---|---|---|
| committer |   Doug Ledford <dledford@redhat.com> | 2016-11-15 16:37:27 -0500 |
| commit | 8af8d2970ed98493a2db88dfcad88b0065e55e79 (patch) | |
| tree | 57841ee0bdf580dc5c23fe6de9a52dfe20cf005d /drivers/infiniband/hw/hfi1/pio.c | |
| parent | IB/hfi1: Get rid of divide in pio buffer allocator (diff) | |
| download | linux-dev-8af8d2970ed98493a2db88dfcad88b0065e55e79.tar.xz linux-dev-8af8d2970ed98493a2db88dfcad88b0065e55e79.zip | |
IB/hfi1: Optimize pio_buf and send_context structs
Both pio_buf and send_context structs have oversized
fields and have cachelines that can be optimized.
Reduce oversized fields for both structs.
Make sure pio_buf struct fits within a cacheline.
Move read-only fields to their own cacheline in
send_context struct.
All of this will avoid cacheline trading as the ring
progresses and pio buffers/send contexts are used.
Reviewed-by: Mike Marciniszyn <mike.marciniszyn@intel.com>
Signed-off-by: Sebastian Sanchez <sebastian.sanchez@intel.com>
Signed-off-by: Dennis Dalessandro <dennis.dalessandro@intel.com>
Signed-off-by: Doug Ledford <dledford@redhat.com>
Diffstat (limited to 'drivers/infiniband/hw/hfi1/pio.c')
| -rw-r--r-- | drivers/infiniband/hw/hfi1/pio.c | 5 |
1 files changed, 2 insertions, 3 deletions
diff --git a/drivers/infiniband/hw/hfi1/pio.c b/drivers/infiniband/hw/hfi1/pio.c index 516fac38d31e..86a7f365b624 100644 --- a/drivers/infiniband/hw/hfi1/pio.c +++ b/drivers/infiniband/hw/hfi1/pio.c @@ -765,6 +765,7 @@ struct send_context *sc_alloc(struct hfi1_devdata *dd, int type, sc->hw_context = hw_context; cr_group_addresses(sc, &dma); sc->credits = sci->credits; + sc->size = sc->credits * PIO_BLOCK_SIZE; /* PIO Send Memory Address details */ #define PIO_ADDR_CONTEXT_MASK 0xfful @@ -1470,9 +1471,7 @@ retry: /* finish filling in the buffer outside the lock */ pbuf->start = sc->base_addr + fill_wrap * PIO_BLOCK_SIZE; - pbuf->size = sc->credits * PIO_BLOCK_SIZE; - pbuf->end = sc->base_addr + pbuf->size; - pbuf->block_count = blocks; + pbuf->end = sc->base_addr + sc->size; pbuf->qw_written = 0; pbuf->carry_bytes = 0; pbuf->carry.val64 = 0; |