habanalabs/gaudi: do not set EB in collective slave queues

We don't need to set EB on signal packets from collective slave
queues as it degrades performance. Because the slaves are the network
queues, the engine barrier doesn't actually guarantee that the
packet has been sent.

Signed-off-by: Alon Mizrahi <amizrahi@habana.ai>
Reviewed-by: Oded Gabbay <ogabbay@kernel.org>
Signed-off-by: Oded Gabbay <ogabbay@kernel.org>

diff --git a/drivers/misc/habanalabs/common/habanalabs.h b/drivers/misc/habanalabs/common/habanalabs.h
index 571eda6ef5ab09b77e2ce2a7a43c609efd82397d..70b778a0d60e23f0ff59e0d23667e5210ebd642c 100644 (file)
--- a/drivers/misc/habanalabs/common/habanalabs.h
+++ b/drivers/misc/habanalabs/common/habanalabs.h
@@ -944,7 +944,7 @@ struct hl_asic_funcs {
        u32 (*get_signal_cb_size)(struct hl_device *hdev);
        u32 (*get_wait_cb_size)(struct hl_device *hdev);
        u32 (*gen_signal_cb)(struct hl_device *hdev, void *data, u16 sob_id,
-                       u32 size);
+                       u32 size, bool eb);
        u32 (*gen_wait_cb)(struct hl_device *hdev,
                        struct hl_gen_wait_properties *prop);
        void (*reset_sob)(struct hl_device *hdev, void *data);

diff --git a/drivers/misc/habanalabs/common/hw_queue.c b/drivers/misc/habanalabs/common/hw_queue.c
index 7caf868d1585c2fbd1d5cb033cedde1ddbfb035e..76217258780a4b2649154e041a08e111e121f389 100644 (file)
--- a/drivers/misc/habanalabs/common/hw_queue.c
+++ b/drivers/misc/habanalabs/common/hw_queue.c
@@ -418,8 +418,11 @@ static void init_signal_cs(struct hl_device *hdev,
                "generate signal CB, sob_id: %d, sob val: 0x%x, q_idx: %d\n",
                cs_cmpl->hw_sob->sob_id, cs_cmpl->sob_val, q_idx);
 
+       /* we set an EB since we must make sure all oeprations are done
+        * when sending the signal
+        */
        hdev->asic_funcs->gen_signal_cb(hdev, job->patched_cb,
-                               cs_cmpl->hw_sob->sob_id, 0);
+                               cs_cmpl->hw_sob->sob_id, 0, true);
 
        kref_get(&hw_sob->kref);
 

diff --git a/drivers/misc/habanalabs/gaudi/gaudi.c b/drivers/misc/habanalabs/gaudi/gaudi.c
index e465c158eaeb8af4c127b01c89015b8ff084be3a..65895ba075fe2fefdaead3417810442958d2ea2e 100644 (file)
--- a/drivers/misc/habanalabs/gaudi/gaudi.c
+++ b/drivers/misc/habanalabs/gaudi/gaudi.c
@@ -361,7 +361,7 @@ static int gaudi_cpucp_info_get(struct hl_device *hdev);
 static void gaudi_disable_clock_gating(struct hl_device *hdev);
 static void gaudi_mmu_prepare(struct hl_device *hdev, u32 asid);
 static u32 gaudi_gen_signal_cb(struct hl_device *hdev, void *data, u16 sob_id,
-                               u32 size);
+                               u32 size, bool eb);
 static u32 gaudi_gen_wait_cb(struct hl_device *hdev,
                                struct hl_gen_wait_properties *prop);
 
@@ -1064,7 +1064,7 @@ static void gaudi_collective_slave_init_job(struct hl_device *hdev,
                prop->collective_sob_id, queue_id);
 
        cb_size += gaudi_gen_signal_cb(hdev, job->user_cb,
-                       prop->collective_sob_id, cb_size);
+                       prop->collective_sob_id, cb_size, false);
 }
 
 static void gaudi_collective_wait_init_cs(struct hl_cs *cs)
@@ -7893,7 +7893,7 @@ static u32 gaudi_get_wait_cb_size(struct hl_device *hdev)
 }
 
 static u32 gaudi_gen_signal_cb(struct hl_device *hdev, void *data, u16 sob_id,
-                               u32 size)
+                               u32 size, bool eb)
 {
        struct hl_cb *cb = (struct hl_cb *) data;
        struct packet_msg_short *pkt;
@@ -7910,7 +7910,7 @@ static u32 gaudi_gen_signal_cb(struct hl_device *hdev, void *data, u16 sob_id,
        ctl |= FIELD_PREP(GAUDI_PKT_SHORT_CTL_OP_MASK, 0); /* write the value */
        ctl |= FIELD_PREP(GAUDI_PKT_SHORT_CTL_BASE_MASK, 3); /* W_S SOB base */
        ctl |= FIELD_PREP(GAUDI_PKT_SHORT_CTL_OPCODE_MASK, PACKET_MSG_SHORT);
-       ctl |= FIELD_PREP(GAUDI_PKT_SHORT_CTL_EB_MASK, 1);
+       ctl |= FIELD_PREP(GAUDI_PKT_SHORT_CTL_EB_MASK, eb);
        ctl |= FIELD_PREP(GAUDI_PKT_SHORT_CTL_RB_MASK, 1);
        ctl |= FIELD_PREP(GAUDI_PKT_SHORT_CTL_MB_MASK, 1);
 

diff --git a/drivers/misc/habanalabs/goya/goya.c b/drivers/misc/habanalabs/goya/goya.c
index d61177bf36a5a5b511c5cb52af18de8a18ffb8e7..b8b4aa636b7cb652a4e7e5790ef26cfb6792fee6 100644 (file)
--- a/drivers/misc/habanalabs/goya/goya.c
+++ b/drivers/misc/habanalabs/goya/goya.c
@@ -5339,7 +5339,7 @@ static u32 goya_get_wait_cb_size(struct hl_device *hdev)
 }
 
 static u32 goya_gen_signal_cb(struct hl_device *hdev, void *data, u16 sob_id,
-               u32 size)
+                               u32 size, bool eb)
 {
        return 0;
 }