summaryrefslogtreecommitdiffstats
path: root/drivers/gpu/nvgpu/gv11b
diff options
context:
space:
mode:
authorDavid Li <davli@nvidia.com>2018-04-26 05:00:01 -0400
committermobile promotions <svcmobile_promotions@nvidia.com>2018-05-18 02:34:20 -0400
commita807cf20419af737a79a3d0c7fcc1068ac6b724a (patch)
tree4efc94d09217bd5e7fdad973b8dacfdee9bab8dd /drivers/gpu/nvgpu/gv11b
parent8ac538e1b16c68ef4a5b9d85a82bbfc2b3fabd72 (diff)
gpu: nvgpu: add NVGPU_IOCTL_CHANNEL_RESCHEDULE_RUNLIST
Add NVGPU_IOCTL_CHANNEL_RESCHEDULE_RUNLIST ioctl to reschedule runlist, and optionally check host and FECS status to preempt pending load of context not belonging to the calling channel on GR engine during context switch. This should be called immediately after a submit to decrease worst case submit to start latency for high interleave channel. There is less than 0.002% chance that the ioctl blocks up to couple miliseconds due to race condition of FECS status changing while being read. For GV11B it will always preempt pending load of unwanted context since there is no chance that ioctl blocks due to race condition. Also fix bug with host reschedule for multiple runlists which needs to write both runlist registers. Bug 1987640 Bug 1924808 Change-Id: I0b7e2f91bd18b0b20928e5a3311b9426b1bf1848 Signed-off-by: David Li <davli@nvidia.com> Reviewed-on: https://git-master.nvidia.com/r/1549050 Reviewed-by: mobile promotions <svcmobile_promotions@nvidia.com> Tested-by: mobile promotions <svcmobile_promotions@nvidia.com>
Diffstat (limited to 'drivers/gpu/nvgpu/gv11b')
-rw-r--r--drivers/gpu/nvgpu/gv11b/fifo_gv11b.c8
-rw-r--r--drivers/gpu/nvgpu/gv11b/fifo_gv11b.h1
-rw-r--r--drivers/gpu/nvgpu/gv11b/gv11b.c1
-rw-r--r--drivers/gpu/nvgpu/gv11b/hal_gv11b.c1
4 files changed, 10 insertions, 1 deletions
diff --git a/drivers/gpu/nvgpu/gv11b/fifo_gv11b.c b/drivers/gpu/nvgpu/gv11b/fifo_gv11b.c
index 30e03092..f00e806f 100644
--- a/drivers/gpu/nvgpu/gv11b/fifo_gv11b.c
+++ b/drivers/gpu/nvgpu/gv11b/fifo_gv11b.c
@@ -664,6 +664,13 @@ static u32 gv11b_fifo_get_runlists_mask(struct gk20a *g, u32 act_eng_bitmask,
664 return runlists_mask; 664 return runlists_mask;
665} 665}
666 666
667int gv11b_fifo_reschedule_runlist(struct channel_gk20a *ch, bool preempt_next)
668{
669 /* gv11b allows multiple outstanding preempts,
670 so always preempt next for best reschedule effect */
671 return nvgpu_fifo_reschedule_runlist(ch, true, false);
672}
673
667static void gv11b_fifo_issue_runlist_preempt(struct gk20a *g, 674static void gv11b_fifo_issue_runlist_preempt(struct gk20a *g,
668 u32 runlists_mask) 675 u32 runlists_mask)
669{ 676{
@@ -842,7 +849,6 @@ int gv11b_fifo_preempt_tsg(struct gk20a *g, u32 tsgid)
842 return ret; 849 return ret;
843} 850}
844 851
845
846static int gv11b_fifo_preempt_runlists(struct gk20a *g, u32 runlists_mask) 852static int gv11b_fifo_preempt_runlists(struct gk20a *g, u32 runlists_mask)
847{ 853{
848 int ret = 0; 854 int ret = 0;
diff --git a/drivers/gpu/nvgpu/gv11b/fifo_gv11b.h b/drivers/gpu/nvgpu/gv11b/fifo_gv11b.h
index 3f58f927..1ae3c93e 100644
--- a/drivers/gpu/nvgpu/gv11b/fifo_gv11b.h
+++ b/drivers/gpu/nvgpu/gv11b/fifo_gv11b.h
@@ -76,6 +76,7 @@ void gv11b_dump_channel_status_ramfc(struct gk20a *g,
76void gv11b_dump_eng_status(struct gk20a *g, 76void gv11b_dump_eng_status(struct gk20a *g,
77 struct gk20a_debug_output *o); 77 struct gk20a_debug_output *o);
78u32 gv11b_fifo_intr_0_error_mask(struct gk20a *g); 78u32 gv11b_fifo_intr_0_error_mask(struct gk20a *g);
79int gv11b_fifo_reschedule_runlist(struct channel_gk20a *ch, bool preempt_next);
79int gv11b_fifo_is_preempt_pending(struct gk20a *g, u32 id, 80int gv11b_fifo_is_preempt_pending(struct gk20a *g, u32 id,
80 unsigned int id_type, unsigned int timeout_rc_type); 81 unsigned int id_type, unsigned int timeout_rc_type);
81int gv11b_fifo_preempt_channel(struct gk20a *g, u32 chid); 82int gv11b_fifo_preempt_channel(struct gk20a *g, u32 chid);
diff --git a/drivers/gpu/nvgpu/gv11b/gv11b.c b/drivers/gpu/nvgpu/gv11b/gv11b.c
index c1ad7944..09fe9a45 100644
--- a/drivers/gpu/nvgpu/gv11b/gv11b.c
+++ b/drivers/gpu/nvgpu/gv11b/gv11b.c
@@ -154,6 +154,7 @@ int gv11b_init_gpu_characteristics(struct gk20a *g)
154 __nvgpu_set_enabled(g, NVGPU_SUPPORT_TSG_SUBCONTEXTS, true); 154 __nvgpu_set_enabled(g, NVGPU_SUPPORT_TSG_SUBCONTEXTS, true);
155 __nvgpu_set_enabled(g, NVGPU_SUPPORT_IO_COHERENCE, true); 155 __nvgpu_set_enabled(g, NVGPU_SUPPORT_IO_COHERENCE, true);
156 __nvgpu_set_enabled(g, NVGPU_SUPPORT_SCG, true); 156 __nvgpu_set_enabled(g, NVGPU_SUPPORT_SCG, true);
157 __nvgpu_set_enabled(g, NVGPU_SUPPORT_RESCHEDULE_RUNLIST, true);
157 __nvgpu_set_enabled(g, NVGPU_SUPPORT_SYNCPOINT_ADDRESS, true); 158 __nvgpu_set_enabled(g, NVGPU_SUPPORT_SYNCPOINT_ADDRESS, true);
158 __nvgpu_set_enabled(g, NVGPU_SUPPORT_USER_SYNCPOINT, true); 159 __nvgpu_set_enabled(g, NVGPU_SUPPORT_USER_SYNCPOINT, true);
159 160
diff --git a/drivers/gpu/nvgpu/gv11b/hal_gv11b.c b/drivers/gpu/nvgpu/gv11b/hal_gv11b.c
index 290a9452..ff779075 100644
--- a/drivers/gpu/nvgpu/gv11b/hal_gv11b.c
+++ b/drivers/gpu/nvgpu/gv11b/hal_gv11b.c
@@ -503,6 +503,7 @@ static const struct gpu_ops gv11b_ops = {
503 .tsg_verify_channel_status = gk20a_fifo_tsg_unbind_channel_verify_status, 503 .tsg_verify_channel_status = gk20a_fifo_tsg_unbind_channel_verify_status,
504 .tsg_verify_status_ctx_reload = gm20b_fifo_tsg_verify_status_ctx_reload, 504 .tsg_verify_status_ctx_reload = gm20b_fifo_tsg_verify_status_ctx_reload,
505 .tsg_verify_status_faulted = gv11b_fifo_tsg_verify_status_faulted, 505 .tsg_verify_status_faulted = gv11b_fifo_tsg_verify_status_faulted,
506 .reschedule_runlist = gv11b_fifo_reschedule_runlist,
506 .update_runlist = gk20a_fifo_update_runlist, 507 .update_runlist = gk20a_fifo_update_runlist,
507 .trigger_mmu_fault = NULL, 508 .trigger_mmu_fault = NULL,
508 .get_mmu_fault_info = NULL, 509 .get_mmu_fault_info = NULL,