*** Dedicated vmid per process v2 ***
Subsystems (3)
cover *** Dedicated vmid per process v2 ***
The current kernel implementation, which grabs the idle VMID from pool when emitting the job may:
The back-to-back submission from one process could use different VMID. The submission to different queues from single process could use different VMID
It works well in most case but cannot work for the SQ thread trace capture.
The VMID for the submission that set the {SQTT}_BASE, which refers to the address of the trace buffer, is stored in shader engine.
If the profiling application have to use different VMIDs to submit IBs in its life cycle:
Some trace is not captured since it actually uses different VMID to submit jobs.
Some part of captured trace may come from different application since they are accidentally uses the owner’s VMID to submit jobs.
V2:
1. address Christian's comments:
a. drop context flags for tag process, instead, add vm ioctl. b. change order of patches.
c. sync waiting only when vm flush needs.
2. address Alex's comments;
bump module version
Chunming Zhou (6):
drm/amdgpu: add vm ioctl
drm/amdgpu: add dedicated vmid field in vm struct drm/amdgpu: reserve vmid by vm ioctl drm/amdgpu: add limitation for dedicated vm number v2 drm/amdgpu: implement grab dedicated vmid V2 drm/amdgpu: bump module verion for reserved vmid
drivers/gpu/drm/amd/amdgpu/amdgpu.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 3 +- drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 159 ++++++++++++++++++++++++++++- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 5 + include/uapi/drm/amdgpu_drm.h | 20 ++++ 7 files changed, 188 insertions(+), 2 deletions(-)
Signature · 1 line
1.9.1
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
A message that is not in the archive (sent off-list or not yet received): 1493013460-13344-1-git-send-email-David1.Zhou-5C7GfCeVMHo@public.gmane.org
PATCH 1/6 drm/amdgpu: add vm ioctl
It will be used for reserving vmid.
--- drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 16 ++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 1 + include/uapi/drm/amdgpu_drm.h | 20 ++++++++++++++++++++ 4 files changed, 38 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +1 -0
@@ -1051,6 +1051,7 @@ int amdgpu_get_vblank_timestamp_kms(struct drm_device *dev, unsigned int pipe,const struct drm_ioctl_desc amdgpu_ioctls_kms[] = {DRM_IOCTL_DEF_DRV (AMDGPU_GEM_CREATE , amdgpu_gem_create_ioctl,DRM_AUTH |DRM_RENDER_ALLOW ),DRM_IOCTL_DEF_DRV (AMDGPU_CTX , amdgpu_ctx_ioctl,DRM_AUTH |DRM_RENDER_ALLOW ), +DRM_IOCTL_DEF_DRV (AMDGPU_VM , amdgpu_vm_ioctl,DRM_AUTH |DRM_RENDER_ALLOW ),DRM_IOCTL_DEF_DRV (AMDGPU_BO_LIST , amdgpu_bo_list_ioctl,DRM_AUTH |DRM_RENDER_ALLOW ),/* KMS */ DRM_IOCTL_DEF_DRV (AMDGPU_GEM_MMAP , amdgpu_gem_mmap_ioctl,DRM_AUTH |DRM_RENDER_ALLOW ),
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +16 -0
@@ -2358,3 +2358,19 @@ void amdgpu_vm_manager_fini(struct amdgpu_device *adev) } } } + +int amdgpu_vm_ioctl (struct drm_device *dev,void *data,struct drm_file *filp) +{ +union drm_amdgpu_vm *args = data; +struct amdgpu_device *adev = dev->dev_private; +struct amdgpu_fpriv *fpriv = filp->driver_priv; + +switch (args->in.op) { +case AMDGPU_VM_OP_RESERVE_VMID : +break ; +default : +return -EINVAL ; + } + +return 0 ; +}
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +1 -0
@@ -247,5 +247,6 @@ int amdgpu_vm_bo_clear_mappings(struct amdgpu_device *adev,void amdgpu_vm_bo_rmv (struct amdgpu_device *adev,struct amdgpu_bo_va *bo_va);void amdgpu_vm_adjust_size (struct amdgpu_device *adev,uint64_t vm_size); +int amdgpu_vm_ioctl (struct drm_device *dev,void *data,struct drm_file *filp);#endif
include/uapi/drm/amdgpu_drm.h +20 -0
@@ -51,6 +51,7 @@#define DRM_AMDGPU_GEM_OP 0x10#define DRM_AMDGPU_GEM_USERPTR 0x11#define DRM_AMDGPU_WAIT_FENCES 0x12 +#define DRM_AMDGPU_VM 0x13/* hybrid specific ioctls */ #define DRM_AMDGPU_SEM 0x5b
@@ -71,6 +72,7 @@#define DRM_IOCTL_AMDGPU_GEM_OP DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_GEM_OP, struct drm_amdgpu_gem_op)#define DRM_IOCTL_AMDGPU_GEM_USERPTR DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_GEM_USERPTR, struct drm_amdgpu_gem_userptr)#define DRM_IOCTL_AMDGPU_WAIT_FENCES DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_WAIT_FENCES, union drm_amdgpu_wait_fences) +#define DRM_IOCTL_AMDGPU_VM DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_VM, union drm_amdgpu_vm)/* hybrid specific ioctls */ #define DRM_IOCTL_AMDGPU_GEM_DGMA DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_GEM_DGMA, struct drm_amdgpu_gem_dgma)
@@ -212,6 +214,24 @@ struct drm_amdgpu_ctx_in {union drm_amdgpu_ctx_out out; }; +/* vm ioctl */ +#define AMDGPU_VM_OP_RESERVE_VMID 1 +struct drm_amdgpu_vm_in { +/** AMDGPU_VM_OP_* */ +__u32 op; +__u32 flags; +}; + +struct drm_amdgpu_vm_out { +/** For future use, no flags defined so far */ +__u64 flags; +}; + +union drm_amdgpu_vm { +struct drm_amdgpu_vm_in in; +struct drm_amdgpu_vm_out out; +}; +/* sem related */ #define AMDGPU_SEM_OP_CREATE_SEM 1#define AMDGPU_SEM_OP_WAIT_SEM 2
Signature · 6 lines
1.9.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
A message that is not in the archive (sent off-list or not yet received): 1493013460-13344-2-git-send-email-David1.Zhou-5C7GfCeVMHo@public.gmane.org
Re: [PATCH 1/6] drm/amdgpu: add vm ioctl
Quoted 30 lines · Chunming Zhou
It will be used for reserving vmid.
Change-Id Ib7169ea999690c8e82d0dcbccdd2d97760c0270aSigned-off-by Chunming Zhou <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 16 ++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 1 + include/uapi/drm/amdgpu_drm.h | 20 ++++++++++++++++++++ 4 files changed, 38 insertions(+)drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +1
-0@@ -1051,6 +1051,7 @@ int amdgpu_get_vblank_timestamp_kms(struct drm_device *dev, unsigned int pipe,const struct drm_ioctl_desc amdgpu_ioctls_kms[] = {DRM_IOCTL_DEF_DRV (AMDGPU_GEM_CREATE , amdgpu_gem_create_ioctl,DRM_AUTH |DRM_RENDER_ALLOW ),DRM_IOCTL_DEF_DRV (AMDGPU_CTX , amdgpu_ctx_ioctl,DRM_AUTH |DRM_RENDER_ALLOW ), +DRM_IOCTL_DEF_DRV (AMDGPU_VM , amdgpu_vm_ioctl,DRM_AUTH |DRM_RENDER_ALLOW ),DRM_IOCTL_DEF_DRV (AMDGPU_BO_LIST , amdgpu_bo_list_ioctl,DRM_AUTH |DRM_RENDER_ALLOW ),/* KMS */ DRM_IOCTL_DEF_DRV (AMDGPU_GEM_MMAP , amdgpu_gem_mmap_ioctl,DRM_AUTH |DRM_RENDER_ALLOW ),drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +0
-0@@ -2358,3 +2358,19 @@ void amdgpu_vm_manager_fini(struct amdgpu_device *adev) } }
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c void amdgpu_vm_manager_fini(struct amdgpu_device *adev) · line 2369
Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +9
-0} + +int amdgpu_vm_ioctl (struct drm_device *dev,void *data,struct drm_file *filp) +{ +union drm_amdgpu_vm *args = data; +struct amdgpu_device *adev = dev->dev_private; +struct amdgpu_fpriv *fpriv = filp->driver_priv; + +switch (args->in.op) { +case AMDGPU_VM_OP_RESERVE_VMID :
How do you think to add an op to release the reserved vmid as well?
Jerry
Quoted 54 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +7
-0+break ; +default : +return -EINVAL ; + } + +return 0 ; +}drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +1
-0@@ -247,5 +247,6 @@ int amdgpu_vm_bo_clear_mappings(struct amdgpu_device *adev,void amdgpu_vm_bo_rmv (struct amdgpu_device *adev,struct amdgpu_bo_va *bo_va);void amdgpu_vm_adjust_size (struct amdgpu_device *adev,uint64_t vm_size); +int amdgpu_vm_ioctl (struct drm_device *dev,void *data,struct drm_file *filp);#endif include/uapi/drm/amdgpu_drm.h +13
-0@@ -51,6 +51,7 @@#define DRM_AMDGPU_GEM_OP 0x10#define DRM_AMDGPU_GEM_USERPTR 0x11#define DRM_AMDGPU_WAIT_FENCES 0x12 +#define DRM_AMDGPU_VM 0x13/* hybrid specific ioctls */ #define DRM_AMDGPU_SEM 0x5b@@ -71,6 +72,7 @@#define DRM_IOCTL_AMDGPU_GEM_OP DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_GEM_OP, struct drm_amdgpu_gem_op)#define DRM_IOCTL_AMDGPU_GEM_USERPTR DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_GEM_USERPTR, struct drm_amdgpu_gem_userptr)#define DRM_IOCTL_AMDGPU_WAIT_FENCES DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_WAIT_FENCES, union drm_amdgpu_wait_fences) +#define DRM_IOCTL_AMDGPU_VM DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_VM, union drm_amdgpu_vm)/* hybrid specific ioctls */ #define DRM_IOCTL_AMDGPU_GEM_DGMA DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_GEM_DGMA, struct drm_amdgpu_gem_dgma)@@ -212,6 +214,24 @@ struct drm_amdgpu_ctx_in {union drm_amdgpu_ctx_out out; }; +/* vm ioctl */ +#define AMDGPU_VM_OP_RESERVE_VMID 1 +struct drm_amdgpu_vm_in { +/** AMDGPU_VM_OP_* */ +__u32 op; +__u32 flags; +}; + +struct drm_amdgpu_vm_out { +/** For future use, no flags defined so far */ +__u64 flags;
include/uapi/drm/amdgpu_drm.h struct drm_amdgpu_ctx_in { · line 237
Quoted 10 lines
include/uapi/drm/amdgpu_drm.h +7
-0+}; + +union drm_amdgpu_vm { +struct drm_amdgpu_vm_in in; +struct drm_amdgpu_vm_out out; +}; +/* sem related */ #define AMDGPU_SEM_OP_CREATE_SEM 1#define AMDGPU_SEM_OP_WAIT_SEM 2
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
A message that is not in the archive (sent off-list or not yet received): [email protected]
Re: [PATCH 1/6] drm/amdgpu: add vm ioctl
Am 26.04.2017 um 09:09 schrieb Zhang, Jerry (Junwei):
Quoted 43 lines · Chunming Zhou
Quoted 42 lines
It will be used for reserving vmid.
Change-Id Ib7169ea999690c8e82d0dcbccdd2d97760c0270aSigned-off-by Chunming Zhou <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 16 ++++++++++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 1 + include/uapi/drm/amdgpu_drm.h | 20 ++++++++++++++++++++ 4 files changed, 38 insertions(+)drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c +1
-0@@ -1051,6 +1051,7 @@ int amdgpu_get_vblank_timestamp_kms(struct drm_device *dev, unsigned int pipe, const struct drm_ioctl_desc amdgpu_ioctls_kms[] = {DRM_IOCTL_DEF_DRV (AMDGPU_GEM_CREATE , amdgpu_gem_create_ioctl,DRM_AUTH |DRM_RENDER_ALLOW ),DRM_IOCTL_DEF_DRV (AMDGPU_CTX , amdgpu_ctx_ioctl,DRM_AUTH |DRM_RENDER_ALLOW ), + DRM_IOCTL_DEF_DRV(AMDGPU_VM, amdgpu_vm_ioctl,DRM_AUTH|DRM_RENDER_ALLOW),
DRM_IOCTL_DEF_DRV(AMDGPU_BO_LIST,amdgpu_bo_list_ioctl,
DRM_AUTH|DRM_RENDER_ALLOW),/* KMS */ DRM_IOCTL_DEF_DRV(AMDGPU_GEM_MMAP, amdgpu_gem_mmap_ioctl,DRM_AUTH|DRM_RENDER_ALLOW),
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +2
-0@@ -2358,3 +2358,19 @@ void amdgpu_vm_manager_fini(struct amdgpu_device *adev) } } } + +int amdgpu_vm_ioctl(struct drm_device *dev, void *data, struct
drm_file*filp)
+{
Quoted 8 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c void amdgpu_vm_manager_fini(struct amdgpu_device *adev) · line 2369
Quoted 6 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +6
-0+union drm_amdgpu_vm *args = data; +struct amdgpu_device *adev = dev->dev_private; +struct amdgpu_fpriv *fpriv = filp->driver_priv; + +switch (args->in.op) { +case AMDGPU_VM_OP_RESERVE_VMID :How do you think to add an op to release the reserved vmid as well?
Yeah, that's a really good point.
Additional to Jerrys comment you either need to reorder the patches or return an error here as long as the implementation is missing.
Otherwise you just ignore th IOCTL till it is really implemented and that isn't really an option.
Christian.
Quoted 70 lines
Jerry
Quoted 68 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +7
-0+break ; +default : +return -EINVAL ; + } + +return 0 ; +}drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +1
-0@@ -247,5 +247,6 @@ int amdgpu_vm_bo_clear_mappings(structamdgpu_device *adev, void amdgpu_vm_bo_rmv(struct amdgpu_device *adev,struct amdgpu_bo_va *bo_va);void amdgpu_vm_adjust_size (struct amdgpu_device *adev,uint64_t vm_size); +int amdgpu_vm_ioctl(struct drm_device *dev, void *data, struct
drm_file*filp);#endifinclude/uapi/drm/amdgpu_drm.h +2
-0@@ -51,6 +51,7 @@#define DRM_AMDGPU_GEM_OP 0x10#define DRM_AMDGPU_GEM_USERPTR 0x11#define DRM_AMDGPU_WAIT_FENCES 0x12 +#define DRM_AMDGPU_VM 0x13/* hybrid specific ioctls */ #define DRM_AMDGPU_SEM 0x5b@@ -71,6 +72,7 @@#define DRM_IOCTL_AMDGPU_GEM_OP DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_GEM_OP, struct drm_amdgpu_gem_op)#define DRM_IOCTL_AMDGPU_GEM_USERPTR DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_GEM_USERPTR, struct drm_amdgpu_gem_userptr)#define DRM_IOCTL_AMDGPU_WAIT_FENCES DRM_IOWR(DRM_COMMAND_BASE + DRM_AMDGPU_WAIT_FENCES, union drm_amdgpu_wait_fences) +#define DRM_IOCTL_AMDGPU_VM DRM_IOWR(DRM_COMMAND_BASE +
DRM_AMDGPU_VM, uniondrm_amdgpu_vm)/* hybrid specific ioctls */
#defineDRM_IOCTL_AMDGPU_GEM_DGMADRM_IOWR(DRM_COMMAND_BASE +DRM_AMDGPU_GEM_DGMA, structdrm_amdgpu_gem_dgma)include/uapi/drm/amdgpu_drm.h +18
-0@@ -212,6 +214,24 @@ struct drm_amdgpu_ctx_in {union drm_amdgpu_ctx_out out; }; +/* vm ioctl */ +#define AMDGPU_VM_OP_RESERVE_VMID 1 +struct drm_amdgpu_vm_in { +/** AMDGPU_VM_OP_* */ +__u32 op; +__u32 flags; +}; + +struct drm_amdgpu_vm_out { +/** For future use, no flags defined so far */ +__u64 flags; +}; + +union drm_amdgpu_vm { +struct drm_amdgpu_vm_in in; +struct drm_amdgpu_vm_out out; +}; +
Quoted 8 lines
include/uapi/drm/amdgpu_drm.h struct drm_amdgpu_ctx_in { · line 237
Quoted 4 lines
include/uapi/drm/amdgpu_drm.h +0
-0/* sem related */ #define AMDGPU_SEM_OP_CREATE_SEM 1#define AMDGPU_SEM_OP_WAIT_SEM 2_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 2/6 drm/amdgpu: add dedicated vmid field in vm struct
In reply to a message not in the archive--- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 17 ++++++++++++++++- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 2 ++ 2 files changed, 18 insertions(+), 1 deletion(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +16 -1
@@ -2144,10 +2144,12 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,unsigned ring_instance;struct amdgpu_ring *ring;struct amd_sched_rq *rq; -int r; +int r, i; vm->va =RB_ROOT ; vm->client_id =atomic64_inc_return (&adev->vm_manager.client_counter); +for (i =0 ; i <AMDGPU_MAX_VMHUBS ; i++) + vm->dedicated_vmid[i] =NULL ;spin_lock_init (&vm->status_lock);INIT_LIST_HEAD (&vm->invalidated);INIT_LIST_HEAD (&vm->cleared);
@@ -2250,6 +2252,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) {struct amdgpu_bo_va_mapping *mapping, *tmp;bool prt_fini_needed = !!adev->gart.gart_funcs->set_prt; +int i;if (vm->is_kfd_vm) { struct amdgpu_vm_id_manager *id_mgr =
@@ -2292,6 +2295,18 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)amdgpu_vm_free_levels (&vm->root);fence_put (vm->last_dir_update); +for (i =0 ; i <AMDGPU_MAX_VMHUBS ; i++) { +struct amdgpu_vm_id_manager *id_mgr = + &adev->vm_manager.id_mgr[i]; + +mutex_lock (&id_mgr->lock); +if (vm->dedicated_vmid[i]) { +list_add (&vm->dedicated_vmid[i]->list, + &id_mgr->ids_lru); + vm->dedicated_vmid[i] =NULL ; + } +mutex_unlock (&id_mgr->lock); + } }/**
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +2 -0
@@ -122,6 +122,8 @@ struct amdgpu_vm {/* client id */ u64 client_id; +/* dedicated vmid */ +struct amdgpu_vm_id *dedicated_vmid[AMDGPU_MAX_VMHUBS ];/* each VM will map on CSA */ struct amdgpu_bo_va *csa_bo_va;
Signature · 6 lines
1.9.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
A message that is not in the archive (sent off-list or not yet received): 1493013460-13344-3-git-send-email-David1.Zhou-5C7GfCeVMHo@public.gmane.org
Re: [PATCH 2/6] drm/amdgpu: add dedicated vmid field in vm struct
Quoted 12 lines · Chunming Zhou
Change-Id Id728e20366c8a1ae90d4e901dc80e136e2a613bbSigned-off-by Chunming Zhou <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 17 ++++++++++++++++- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 2 ++ 2 files changed, 18 insertions(+), 1 deletion(-)drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +0
-0
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, · line 2152
Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +3
-1@@ -2144,10 +2144,12 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm,unsigned ring_instance;struct amdgpu_ring *ring;struct amd_sched_rq *rq; -int r; +int r, i; vm->va =RB_ROOT ; vm->client_id =atomic64_inc_return (&adev->vm_manager.client_counter); +for (i =0 ; i <AMDGPU_MAX_VMHUBS ; i++) + vm->dedicated_vmid[i] =NULL ;
Maybe it's better to give it a consistent name as resv_vmid, or anything like that.
Jerry
Quoted 36 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +13
-0spin_lock_init (&vm->status_lock);INIT_LIST_HEAD (&vm->invalidated);INIT_LIST_HEAD (&vm->cleared);@@ -2250,6 +2252,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) {struct amdgpu_bo_va_mapping *mapping, *tmp;bool prt_fini_needed = !!adev->gart.gart_funcs->set_prt; +int i;if (vm->is_kfd_vm) { struct amdgpu_vm_id_manager *id_mgr =@@ -2292,6 +2295,18 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)amdgpu_vm_free_levels (&vm->root);fence_put (vm->last_dir_update); +for (i =0 ; i <AMDGPU_MAX_VMHUBS ; i++) { +struct amdgpu_vm_id_manager *id_mgr = + &adev->vm_manager.id_mgr[i]; + +mutex_lock (&id_mgr->lock); +if (vm->dedicated_vmid[i]) { +list_add (&vm->dedicated_vmid[i]->list, + &id_mgr->ids_lru); + vm->dedicated_vmid[i] =NULL ; + } +mutex_unlock (&id_mgr->lock); + } }/** drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +0
-0
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h struct amdgpu_vm { · line 129
Quoted 8 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +2
-0@@ -122,6 +122,8 @@ struct amdgpu_vm {/* client id */ u64 client_id; +/* dedicated vmid */ +struct amdgpu_vm_id *dedicated_vmid[AMDGPU_MAX_VMHUBS ];/* each VM will map on CSA */ struct amdgpu_bo_va *csa_bo_va;
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
A message that is not in the archive (sent off-list or not yet received): [email protected]
Re: [PATCH 2/6] drm/amdgpu: add dedicated vmid field in vm struct
Am 26.04.2017 um 09:09 schrieb Zhang, Jerry (Junwei):
Quoted 20 lines · Chunming Zhou
Quoted 19 lines
Change-Id Id728e20366c8a1ae90d4e901dc80e136e2a613bbSigned-off-by Chunming Zhou <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 17 ++++++++++++++++- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 2 ++ 2 files changed, 18 insertions(+), 1 deletion(-)drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +1
-1@@ -2144,10 +2144,12 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, unsigned ring_instance;struct amdgpu_ring *ring;struct amd_sched_rq *rq; -int r; +int r, i;
Quoted 7 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, · line 2152
Quoted 5 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +2
-0vm->va =RB_ROOT ; vm->client_id =atomic64_inc_return (&adev->vm_manager.client_counter); +for (i =0 ; i <AMDGPU_MAX_VMHUBS ; i++) + vm->dedicated_vmid[i] =NULL ;Maybe it's better to give it a consistent name as
resv_vmid, or anything like that.
Yes, agree. And if I'm not completely mistaken we still should only apply that to the GFX hub on Vega10.
Christian.
Quoted 40 lines
Jerry
Quoted 38 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +0
-0spin_lock_init (&vm->status_lock);INIT_LIST_HEAD(&vm->invalidated); INIT_LIST_HEAD(&vm->cleared);drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +1
-0@@ -2250,6 +2252,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) {struct amdgpu_bo_va_mapping *mapping, *tmp;bool prt_fini_needed = !!adev->gart.gart_funcs->set_prt; +int i;if (vm->is_kfd_vm) {struct
amdgpu_vm_id_manager*id_mgr=drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +12
-0@@ -2292,6 +2295,18 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)amdgpu_vm_free_levels (&vm->root);fence_put (vm->last_dir_update); +for (i =0 ; i <AMDGPU_MAX_VMHUBS ; i++) { +struct amdgpu_vm_id_manager *id_mgr = + &adev->vm_manager.id_mgr[i]; + +mutex_lock (&id_mgr->lock); +if (vm->dedicated_vmid[i]) { +list_add (&vm->dedicated_vmid[i]->list, + &id_mgr->ids_lru); + vm->dedicated_vmid[i] =NULL ; + } +mutex_unlock (&id_mgr->lock); + } }/**
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +0
-0@@ -122,6 +122,8 @@ struct amdgpu_vm {
Quoted 12 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h struct amdgpu_vm { · line 129
Quoted 8 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +2
-0/* client id */ u64 client_id; +/* dedicated vmid */ +struct amdgpu_vm_id *dedicated_vmid[AMDGPU_MAX_VMHUBS ];/* each VM will map on CSA */ struct amdgpu_bo_va *csa_bo_va;_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
A message that is not in the archive (sent off-list or not yet received): 291734a9-5306-f3e3-219c-c53cbf433358-ANTagKRnAhcb1SvskN2V4Q@public.gmane.org
Re: [PATCH 2/6] drm/amdgpu: add dedicated vmid field in vm struct
Quoted 29 lines · Christian König
Am 26.04.2017 um 09:09 schrieb Zhang, Jerry (Junwei):
Quoted 20 lines · Chunming Zhou
Quoted 19 lines
Change-Id Id728e20366c8a1ae90d4e901dc80e136e2a613bbSigned-off-by Chunming Zhou <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 17 ++++++++++++++++- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 2 ++ 2 files changed, 18 insertions(+), 1 deletion(-)drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +1
-1@@ -2144,10 +2144,12 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, unsigned ring_instance;struct amdgpu_ring *ring;struct amd_sched_rq *rq; -int r; +int r, i;Quoted 7 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, · line 2152
Quoted 5 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +2
-0vm->va =RB_ROOT ; vm->client_id =atomic64_inc_return (&adev->vm_manager.client_counter); +for (i =0 ; i <AMDGPU_MAX_VMHUBS ; i++) + vm->dedicated_vmid[i] =NULL ;Maybe it's better to give it a consistent name as
resv_vmid, or anything like that.
Quoted 1 line
Yes, agree.
I think the reserved vmid is dedicated to this vm, I don't know where this name doesn't make sense.
Quoted 1 line
And if I'm not completely mistaken we still should only apply that to the GFX hub on Vega10.
David Mao required mmhub as well. IIRC, we don't have necessary to argue more on this.
Regards,
David Zhou
Quoted 57 lines
Christian.
Quoted 52 lines
Jerry
Quoted 38 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +0
-0spin_lock_init (&vm->status_lock);INIT_LIST_HEAD(&vm->invalidated); INIT_LIST_HEAD(&vm->cleared);drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +1
-0@@ -2250,6 +2252,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) {struct amdgpu_bo_va_mapping *mapping, *tmp;bool prt_fini_needed = !!adev->gart.gart_funcs->set_prt; +int i;if (vm->is_kfd_vm) {struct
amdgpu_vm_id_manager*id_mgr=drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +12
-0@@ -2292,6 +2295,18 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)amdgpu_vm_free_levels (&vm->root);fence_put (vm->last_dir_update); +for (i =0 ; i <AMDGPU_MAX_VMHUBS ; i++) { +struct amdgpu_vm_id_manager *id_mgr = + &adev->vm_manager.id_mgr[i]; + +mutex_lock (&id_mgr->lock); +if (vm->dedicated_vmid[i]) { +list_add (&vm->dedicated_vmid[i]->list, + &id_mgr->ids_lru); + vm->dedicated_vmid[i] =NULL ; + } +mutex_unlock (&id_mgr->lock); + } }/**
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +0
-0drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h struct amdgpu_vm { · line 129
Quoted 8 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +2
-0@@ -122,6 +122,8 @@ struct amdgpu_vm {/* client id */ u64 client_id; +/* dedicated vmid */ +struct amdgpu_vm_id *dedicated_vmid[AMDGPU_MAX_VMHUBS ];/* each VM will map on CSA */ struct amdgpu_bo_va *csa_bo_va;_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
A message that is not in the archive (sent off-list or not yet received): [email protected]
Re: [PATCH 2/6] drm/amdgpu: add dedicated vmid field in vm struct
Am 26.04.2017 um 11:05 schrieb zhoucm1:
Quoted 32 lines · Christian König
Quoted 29 lines
Am 26.04.2017 um 09:09 schrieb Zhang, Jerry (Junwei):
Quoted 27 lines · Chunming Zhou
Quoted 14 lines
Change-Id Id728e20366c8a1ae90d4e901dc80e136e2a613bbSigned-off-by Chunming Zhou <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 17 ++++++++++++++++- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 2 ++ 2 files changed, 18 insertions(+), 1 deletion(-)drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +0
-0@@ -2144,10 +2144,12 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, unsigned ring_instance;drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c int amdgpu_vm_init(struct amdgpu_device · line 2154
Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +3
-1struct amdgpu_ring *ring;struct amd_sched_rq *rq; -int r; +int r, i; vm->va =RB_ROOT ; vm->client_id =atomic64_inc_return (&adev->vm_manager.client_counter); +for (i =0 ; i <AMDGPU_MAX_VMHUBS ; i++) + vm->dedicated_vmid[i] =NULL ;Maybe it's better to give it a consistent name as
resv_vmid, or anything like that.
Quoted 5 lines
Quoted 1 line
Yes, agree.
I think the reserved vmid is dedicated to this vm, I don't know where this name doesn't make sense.
Quoted 1 line
And if I'm not completely mistaken we still should only apply that to the GFX hub on Vega10.
David Mao required mmhub as well. IIRC, we don't have necessary to argue more on this.
I think we still have. There is no technical reason why we should use the reserved/dedicated VM for other engines than the one involved in the SQ trace.
So I would say we should limit this to GFX and Compute jobs and only allocate the dedicated VMID for those.
Regards,
Christian.
Quoted 61 lines
Regards,
David ZhouQuoted 57 lines
Christian.
Quoted 52 lines
Jerry
Quoted 38 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +0
-0spin_lock_init (&vm->status_lock);INIT_LIST_HEAD(&vm->invalidated); INIT_LIST_HEAD(&vm->cleared);drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +1
-0@@ -2250,6 +2252,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) {struct amdgpu_bo_va_mapping *mapping, *tmp;bool prt_fini_needed = !!adev->gart.gart_funcs->set_prt; +int i;if (vm->is_kfd_vm) {struct
amdgpu_vm_id_manager*id_mgr=drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +12
-0@@ -2292,6 +2295,18 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)amdgpu_vm_free_levels (&vm->root);fence_put (vm->last_dir_update); +for (i =0 ; i <AMDGPU_MAX_VMHUBS ; i++) { +struct amdgpu_vm_id_manager *id_mgr = + &adev->vm_manager.id_mgr[i]; + +mutex_lock (&id_mgr->lock); +if (vm->dedicated_vmid[i]) { +list_add (&vm->dedicated_vmid[i]->list, + &id_mgr->ids_lru); + vm->dedicated_vmid[i] =NULL ; + } +mutex_unlock (&id_mgr->lock); + } }/**
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +0
-0drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h struct amdgpu_vm { · line 129
Quoted 8 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +2
-0@@ -122,6 +122,8 @@ struct amdgpu_vm {/* client id */ u64 client_id; +/* dedicated vmid */ +struct amdgpu_vm_id *dedicated_vmid[AMDGPU_MAX_VMHUBS ];/* each VM will map on CSA */ struct amdgpu_bo_va *csa_bo_va;_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
Quoted 4 lines
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
A message that is not in the archive (sent off-list or not yet received): 05c4fcec-9dae-24ac-4a5d-26e0fd5bc148-ANTagKRnAhcb1SvskN2V4Q@public.gmane.org
Re: [PATCH 2/6] drm/amdgpu: add dedicated vmid field in vm struct
Quoted 33 lines · Christian König
Am 26.04.2017 um 11:05 schrieb zhoucm1:
Quoted 32 lines · Christian König
Quoted 29 lines
Am 26.04.2017 um 09:09 schrieb Zhang, Jerry (Junwei):
Quoted 27 lines · Chunming Zhou
Quoted 14 lines
Change-Id Id728e20366c8a1ae90d4e901dc80e136e2a613bbSigned-off-by Chunming Zhou <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 17 ++++++++++++++++- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 2 ++ 2 files changed, 18 insertions(+), 1 deletion(-)drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +0
-0@@ -2144,10 +2144,12 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, unsigned ring_instance;drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c int amdgpu_vm_init(struct amdgpu_device · line 2154
Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +3
-1struct amdgpu_ring *ring;struct amd_sched_rq *rq; -int r; +int r, i; vm->va =RB_ROOT ; vm->client_id =atomic64_inc_return (&adev->vm_manager.client_counter); +for (i =0 ; i <AMDGPU_MAX_VMHUBS ; i++) + vm->dedicated_vmid[i] =NULL ;Maybe it's better to give it a consistent name as
resv_vmid, or anything like that.
Quoted 7 lines
Quoted 5 lines
Quoted 1 line
Yes, agree.
I think the reserved vmid is dedicated to this vm, I don't know where this name doesn't make sense.
Quoted 1 line
And if I'm not completely mistaken we still should only apply that to the GFX hub on Vega10.
David Mao required mmhub as well. IIRC, we don't have necessary to argue more on this.
I think we still have. There is no technical reason why we should use the reserved/dedicated VM for other engines than the one involved in the SQ trace.
How about removing SQ trace reason? is it ok just that one process wants to do experiment for other purpose? :)
Regards,
David Zhou
Quoted 73 lines
So I would say we should limit this to GFX and Compute jobs and only allocate the dedicated VMID for those.
Regards,
Christian.Quoted 65 lines
Regards,
David ZhouQuoted 57 lines
Christian.
Quoted 52 lines
Jerry
Quoted 38 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +0
-0spin_lock_init (&vm->status_lock);INIT_LIST_HEAD(&vm->invalidated); INIT_LIST_HEAD(&vm->cleared);drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +1
-0@@ -2250,6 +2252,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) {struct amdgpu_bo_va_mapping *mapping, *tmp;bool prt_fini_needed = !!adev->gart.gart_funcs->set_prt; +int i;if (vm->is_kfd_vm) {struct
amdgpu_vm_id_manager*id_mgr=drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +12
-0@@ -2292,6 +2295,18 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)amdgpu_vm_free_levels (&vm->root);fence_put (vm->last_dir_update); +for (i =0 ; i <AMDGPU_MAX_VMHUBS ; i++) { +struct amdgpu_vm_id_manager *id_mgr = + &adev->vm_manager.id_mgr[i]; + +mutex_lock (&id_mgr->lock); +if (vm->dedicated_vmid[i]) { +list_add (&vm->dedicated_vmid[i]->list, + &id_mgr->ids_lru); + vm->dedicated_vmid[i] =NULL ; + } +mutex_unlock (&id_mgr->lock); + } }/**
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +0
-0drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h struct amdgpu_vm { · line 129
Quoted 8 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +2
-0@@ -122,6 +122,8 @@ struct amdgpu_vm {/* client id */ u64 client_id; +/* dedicated vmid */ +struct amdgpu_vm_id *dedicated_vmid[AMDGPU_MAX_VMHUBS ];/* each VM will map on CSA */ struct amdgpu_bo_va *csa_bo_va;_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
A message that is not in the archive (sent off-list or not yet received): [email protected]
Re: [PATCH 2/6] drm/amdgpu: add dedicated vmid field in vm struct
Am 26.04.2017 um 11:14 schrieb zhoucm1:
Quoted 36 lines · Christian König
Quoted 33 lines
Am 26.04.2017 um 11:05 schrieb zhoucm1:
Quoted 32 lines · Christian König
Quoted 29 lines
Am 26.04.2017 um 09:09 schrieb Zhang, Jerry (Junwei):
Quoted 27 lines · Chunming Zhou
Quoted 14 lines
Change-Id Id728e20366c8a1ae90d4e901dc80e136e2a613bbSigned-off-by Chunming Zhou <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 17 ++++++++++++++++- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 2 ++ 2 files changed, 18 insertions(+), 1 deletion(-)drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +0
-0@@ -2144,10 +2144,12 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, unsigned ring_instance;drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c int amdgpu_vm_init(struct amdgpu_device · line 2154
Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +3
-1struct amdgpu_ring *ring;struct amd_sched_rq *rq; -int r; +int r, i; vm->va =RB_ROOT ; vm->client_id =atomic64_inc_return (&adev->vm_manager.client_counter); +for (i =0 ; i <AMDGPU_MAX_VMHUBS ; i++) + vm->dedicated_vmid[i] =NULL ;Maybe it's better to give it a consistent name as
resv_vmid, or anything like that.
Quoted 8 lines
Quoted 7 lines
Quoted 5 lines
Quoted 1 line
Yes, agree.
I think the reserved vmid is dedicated to this vm, I don't know where this name doesn't make sense.
Quoted 1 line
And if I'm not completely mistaken we still should only apply that to the GFX hub on Vega10.
David Mao required mmhub as well. IIRC, we don't have necessary to argue more on this.
I think we still have. There is no technical reason why we should use the reserved/dedicated VM for other engines than the one involved in the SQ trace.
How about removing SQ trace reason? is it ok just that one process wants to do experiment for other purpose? :)
In this case I would clearly NAK the whole approach.
Regards,
Christian.
Quoted 77 lines
Regards,
David ZhouQuoted 73 lines
So I would say we should limit this to GFX and Compute jobs and only allocate the dedicated VMID for those.
Regards,
Christian.Quoted 65 lines
Regards,
David ZhouQuoted 57 lines
Christian.
Quoted 52 lines
Jerry
Quoted 38 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +0
-0spin_lock_init (&vm->status_lock);INIT_LIST_HEAD(&vm->invalidated); INIT_LIST_HEAD(&vm->cleared);drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +1
-0@@ -2250,6 +2252,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) {struct amdgpu_bo_va_mapping *mapping, *tmp;bool prt_fini_needed = !!adev->gart.gart_funcs->set_prt; +int i;if (vm->is_kfd_vm) {struct
amdgpu_vm_id_manager*id_mgr=drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +12
-0@@ -2292,6 +2295,18 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)amdgpu_vm_free_levels (&vm->root);fence_put (vm->last_dir_update); +for (i =0 ; i <AMDGPU_MAX_VMHUBS ; i++) { +struct amdgpu_vm_id_manager *id_mgr = + &adev->vm_manager.id_mgr[i]; + +mutex_lock (&id_mgr->lock); +if (vm->dedicated_vmid[i]) { +list_add (&vm->dedicated_vmid[i]->list, + &id_mgr->ids_lru); + vm->dedicated_vmid[i] =NULL ; + } +mutex_unlock (&id_mgr->lock); + } }/**
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +0
-0drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h struct amdgpu_vm { · line 129
Quoted 8 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +2
-0@@ -122,6 +122,8 @@ struct amdgpu_vm {/* client id */ u64 client_id; +/* dedicated vmid */ +struct amdgpu_vm_id *dedicated_vmid[AMDGPU_MAX_VMHUBS ];/* each VM will map on CSA */ struct amdgpu_bo_va *csa_bo_va;_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
A message that is not in the archive (sent off-list or not yet received): f41de8f1-11d9-a727-c34f-d218b08f9047-ANTagKRnAhcb1SvskN2V4Q@public.gmane.org
Re: [PATCH 2/6] drm/amdgpu: add dedicated vmid field in vm struct
Quoted 40 lines · Christian König
Am 26.04.2017 um 11:14 schrieb zhoucm1:
Quoted 39 lines · Christian König
Quoted 36 lines
Am 26.04.2017 um 11:05 schrieb zhoucm1:
Quoted 35 lines · Christian König
Quoted 29 lines
Am 26.04.2017 um 09:09 schrieb Zhang, Jerry (Junwei):
Quoted 20 lines · Chunming Zhou
Quoted 19 lines
Change-Id Id728e20366c8a1ae90d4e901dc80e136e2a613bbSigned-off-by Chunming Zhou <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 17 ++++++++++++++++- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 2 ++ 2 files changed, 18 insertions(+), 1 deletion(-)drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +1
-1@@ -2144,10 +2144,12 @@ int amdgpu_vm_init(struct amdgpu_device *adev, struct amdgpu_vm *vm, unsigned ring_instance;struct amdgpu_ring *ring;struct amd_sched_rq *rq; -int r; +int r, i;Quoted 7 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c int amdgpu_vm_init(struct amdgpu_device · line 2154
Quoted 5 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +2
-0vm->va =RB_ROOT ; vm->client_id =atomic64_inc_return (&adev->vm_manager.client_counter); +for (i =0 ; i <AMDGPU_MAX_VMHUBS ; i++) + vm->dedicated_vmid[i] =NULL ;Maybe it's better to give it a consistent name as
resv_vmid, or anything like that.Quoted 1 line
Yes, agree.
I think the reserved vmid is dedicated to this vm, I don't know where this name doesn't make sense.
Quoted 7 lines
Quoted 5 lines
Quoted 4 lines
Quoted 2 lines
Quoted 1 line
And if I'm not completely mistaken we still should only apply that to the GFX hub on Vega10.
David Mao required mmhub as well. IIRC, we don't have necessary to argue more on this.
I think we still have. There is no technical reason why we should use the reserved/dedicated VM for other engines than the one involved in the SQ trace.
How about removing SQ trace reason? is it ok just that one process wants to do experiment for other purpose? :)
In this case I would clearly NAK the whole approach.
Which makes me very difficult to talk between you and David Mao.
Regards,
David Zhou
Quoted 82 lines
Regards,
Christian.Quoted 77 lines
Regards,
David ZhouQuoted 73 lines
So I would say we should limit this to GFX and Compute jobs and only allocate the dedicated VMID for those.
Regards,
Christian.Quoted 65 lines
Regards,
David ZhouQuoted 57 lines
Christian.
Quoted 52 lines
Jerry
Quoted 38 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +0
-0spin_lock_init (&vm->status_lock);INIT_LIST_HEAD(&vm->invalidated); INIT_LIST_HEAD(&vm->cleared);drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +1
-0@@ -2250,6 +2252,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) {struct amdgpu_bo_va_mapping *mapping, *tmp;bool prt_fini_needed = !!adev->gart.gart_funcs->set_prt; +int i;if (vm->is_kfd_vm) {struct
amdgpu_vm_id_manager*id_mgr=drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +12
-0@@ -2292,6 +2295,18 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)amdgpu_vm_free_levels (&vm->root);fence_put (vm->last_dir_update); +for (i =0 ; i <AMDGPU_MAX_VMHUBS ; i++) { +struct amdgpu_vm_id_manager *id_mgr = + &adev->vm_manager.id_mgr[i]; + +mutex_lock (&id_mgr->lock); +if (vm->dedicated_vmid[i]) { +list_add (&vm->dedicated_vmid[i]->list, + &id_mgr->ids_lru); + vm->dedicated_vmid[i] =NULL ; + } +mutex_unlock (&id_mgr->lock); + } }/**
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +0
-0drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h struct amdgpu_vm { · line 129
Quoted 8 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +2
-0@@ -122,6 +122,8 @@ struct amdgpu_vm {/* client id */ u64 client_id; +/* dedicated vmid */ +struct amdgpu_vm_id *dedicated_vmid[AMDGPU_MAX_VMHUBS ];/* each VM will map on CSA */ struct amdgpu_bo_va *csa_bo_va;_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 3/6 drm/amdgpu: reserve vmid by vm ioctl
In reply to a message not in the archive--- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 56 ++++++++++++++++++++++++++++++++++ 1 file changed, 56 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +56 -0
@@ -397,6 +397,17 @@ static bool amdgpu_vm_had_gpu_reset(struct amdgpu_device *adev,atomic_read (&adev->gpu_reset_counter); } +static bool amdgpu_vm_dedicated_vmid_ready (struct amdgpu_vm *vm) +{ +unsigned vmhub; + +for (vmhub =0 ; vmhub <AMDGPU_MAX_VMHUBS ; vmhub++) { +if (!vm->dedicated_vmid[vmhub]) +return false; + } +return true; +} +/** * amdgpu_vm_grab_id - allocate the next free VMID *
@@ -546,6 +557,45 @@ int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring,return r; } +static int amdgpu_vm_alloc_dedicated_vmid (struct amdgpu_device *adev, +struct amdgpu_vm *vm) +{ +struct amdgpu_vm_id_manager *id_mgr; +struct amdgpu_vm_id *idle; +unsigned vmhub; +int r; + +for (vmhub =0 ; vmhub <AMDGPU_MAX_VMHUBS ; vmhub++) { + id_mgr = &adev->vm_manager.id_mgr[vmhub]; + +mutex_lock (&id_mgr->lock); +/* Select the first entry VMID */ + idle = list_first_entry(&id_mgr->ids_lru, struct amdgpu_vm_id, + list); +list_del_init (&idle->list); + vm->dedicated_vmid[vmhub] = idle; +mutex_unlock (&id_mgr->lock); + + r =amdgpu_sync_wait (&idle->active); +if (r) + goto err; + } + +return 0 ; +err: +for (vmhub =0 ; vmhub <AMDGPU_MAX_VMHUBS ; vmhub++) { + id_mgr = &adev->vm_manager.id_mgr[vmhub]; + +mutex_lock (&id_mgr->lock); +if (vm->dedicated_vmid[vmhub]) +list_add (&vm->dedicated_vmid[vmhub]->list, + &id_mgr->ids_lru); + vm->dedicated_vmid[vmhub] =NULL ; +mutex_unlock (&id_mgr->lock); + } +return r; +} +static bool amdgpu_vm_ring_has_compute_vm_bug (struct amdgpu_ring *ring) {struct amdgpu_device *adev = ring->adev;
@@ -2379,9 +2429,15 @@ int amdgpu_vm_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)union drm_amdgpu_vm *args = data;struct amdgpu_device *adev = dev->dev_private;struct amdgpu_fpriv *fpriv = filp->driver_priv; +int r;switch (args->in.op) {case AMDGPU_VM_OP_RESERVE_VMID : +if (!amdgpu_vm_dedicated_vmid_ready (&fpriv->vm)) { + r =amdgpu_vm_alloc_dedicated_vmid (adev, &fpriv->vm); +if (r) +return r; + }break ;default :return -EINVAL ;
Signature · 6 lines
1.9.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
A message that is not in the archive (sent off-list or not yet received): 1493013460-13344-4-git-send-email-David1.Zhou-5C7GfCeVMHo@public.gmane.org
Re: [PATCH 3/6] drm/amdgpu: reserve vmid by vm ioctl
Quoted 11 lines · Chunming Zhou
Change-Id I5f80dc39dc9d44660a96a2b710b0dbb4d3b9039dSigned-off-by Chunming Zhou <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 56 ++++++++++++++++++++++++++++++++++ 1 file changed, 56 insertions(+)drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +0
-0
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c static bool amdgpu_vm_had_gpu_reset(struct amdgpu_device *adev, · line 406
Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +7
-0@@ -397,6 +397,17 @@ static bool amdgpu_vm_had_gpu_reset(struct amdgpu_device *adev,atomic_read (&adev->gpu_reset_counter); } +static bool amdgpu_vm_dedicated_vmid_ready (struct amdgpu_vm *vm) +{ +unsigned vmhub; + +for (vmhub =0 ; vmhub <AMDGPU_MAX_VMHUBS ; vmhub++) { +if (!vm->dedicated_vmid[vmhub]) +return false;
Just note:
Seemingly for pre-gmcv9, it will always allocate one more dedicated/reserved vmid for the 2nd un-used vmhub. anyway it will not be used either.
Jerry
Quoted 54 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +43
-0+ } +return true; +} +/** * amdgpu_vm_grab_id - allocate the next free VMID * @@ -546,6 +557,45 @@ int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring,return r; } +static int amdgpu_vm_alloc_dedicated_vmid (struct amdgpu_device *adev, +struct amdgpu_vm *vm) +{ +struct amdgpu_vm_id_manager *id_mgr; +struct amdgpu_vm_id *idle; +unsigned vmhub; +int r; + +for (vmhub =0 ; vmhub <AMDGPU_MAX_VMHUBS ; vmhub++) { + id_mgr = &adev->vm_manager.id_mgr[vmhub]; + +mutex_lock (&id_mgr->lock); +/* Select the first entry VMID */ + idle = list_first_entry(&id_mgr->ids_lru, struct amdgpu_vm_id, + list); +list_del_init (&idle->list); + vm->dedicated_vmid[vmhub] = idle; +mutex_unlock (&id_mgr->lock); + + r =amdgpu_sync_wait (&idle->active); +if (r) + goto err; + } + +return 0 ; +err: +for (vmhub =0 ; vmhub <AMDGPU_MAX_VMHUBS ; vmhub++) { + id_mgr = &adev->vm_manager.id_mgr[vmhub]; + +mutex_lock (&id_mgr->lock); +if (vm->dedicated_vmid[vmhub]) +list_add (&vm->dedicated_vmid[vmhub]->list, + &id_mgr->ids_lru); + vm->dedicated_vmid[vmhub] =NULL ; +mutex_unlock (&id_mgr->lock); + } +return r; +} +static bool amdgpu_vm_ring_has_compute_vm_bug (struct amdgpu_ring *ring) {struct amdgpu_device *adev = ring->adev;
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c int amdgpu_vm_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) · line 2437
Quoted 9 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +3
-0@@ -2379,9 +2429,15 @@ int amdgpu_vm_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)union drm_amdgpu_vm *args = data;struct amdgpu_device *adev = dev->dev_private;struct amdgpu_fpriv *fpriv = filp->driver_priv; +int r;switch (args->in.op) {case AMDGPU_VM_OP_RESERVE_VMID : +if (!amdgpu_vm_dedicated_vmid_ready (&fpriv->vm)) { + r =amdgpu_vm_alloc_dedicated_vmid (adev, &fpriv->vm);
Could you change the return value in this ready checking func? Usually we can easily get to know this kind of things. if (sth_ready())
do_sth();
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c int amdgpu_vm_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) · line 2443
Quoted 7 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +3
-0+if (r) +return r; + }break ;default :return -EINVAL ;
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
A message that is not in the archive (sent off-list or not yet received): [email protected]
Re: [PATCH 3/6] drm/amdgpu: reserve vmid by vm ioctl
Am 26.04.2017 um 09:09 schrieb Zhang, Jerry (Junwei):
Quoted 19 lines · Chunming Zhou
Quoted 16 lines
Change-Id I5f80dc39dc9d44660a96a2b710b0dbb4d3b9039dSigned-off-by Chunming Zhou <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 56 ++++++++++++++++++++++++++++++++++ 1 file changed, 56 insertions(+)drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +0
-0@@ -397,6 +397,17 @@ static bool amdgpu_vm_had_gpu_reset(struct amdgpu_device *adev,atomic_read (&adev->gpu_reset_counter); }
Quoted 11 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c static bool amdgpu_vm_had_gpu_reset(struct amdgpu_device *adev, · line 406
Quoted 7 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +7
-0+static bool amdgpu_vm_dedicated_vmid_ready (struct amdgpu_vm *vm) +{ +unsigned vmhub; + +for (vmhub =0 ; vmhub <AMDGPU_MAX_VMHUBS ; vmhub++) { +if (!vm->dedicated_vmid[vmhub]) +return false;Just note:
Seemingly for pre-gmcv9, it will always allocate one more dedicated/reserved vmid for the 2nd un-used vmhub. anyway it will not be used either.
It's even worse. IIRC we don't initialize the 2nd hub on pre-gmcv9. So that could work with uninitialized data.
Christian.
Quoted 82 lines
Jerry
Quoted 56 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +4
-0+ } +return true; +} +/** * amdgpu_vm_grab_id - allocate the next free VMID *drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +39
-0@@ -546,6 +557,45 @@ int amdgpu_vm_grab_id(struct amdgpu_vm *vm,struct amdgpu_ring *ring, return r; } +static int amdgpu_vm_alloc_dedicated_vmid (struct amdgpu_device *adev, +struct amdgpu_vm *vm) +{ +struct amdgpu_vm_id_manager *id_mgr; +struct amdgpu_vm_id *idle; +unsigned vmhub; +int r; + +for (vmhub =0 ; vmhub <AMDGPU_MAX_VMHUBS ; vmhub++) { + id_mgr = &adev->vm_manager.id_mgr[vmhub]; + +mutex_lock (&id_mgr->lock); +/* Select the first entry VMID */ + idle = list_first_entry(&id_mgr->ids_lru, struct amdgpu_vm_id, + list); +list_del_init (&idle->list); + vm->dedicated_vmid[vmhub] = idle; +mutex_unlock (&id_mgr->lock); + + r =amdgpu_sync_wait (&idle->active); +if (r) + goto err; + } + +return 0 ; +err: +for (vmhub =0 ; vmhub <AMDGPU_MAX_VMHUBS ; vmhub++) { + id_mgr = &adev->vm_manager.id_mgr[vmhub]; + +mutex_lock (&id_mgr->lock); +if (vm->dedicated_vmid[vmhub]) +list_add (&vm->dedicated_vmid[vmhub]->list, + &id_mgr->ids_lru); + vm->dedicated_vmid[vmhub] =NULL ; +mutex_unlock (&id_mgr->lock); + } +return r; +} +static bool amdgpu_vm_ring_has_compute_vm_bug (struct amdgpu_ring *ring){ struct amdgpu_device *adev = ring->adev;drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c int amdgpu_vm_ioctl(struct drm_device *dev, · line 2438
Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +3
-0@@ -2379,9 +2429,15 @@ int amdgpu_vm_ioctl(struct drm_device *dev, void *data, struct drm_file *filp) union drm_amdgpu_vm *args = data;struct amdgpu_device *adev = dev->dev_private;struct amdgpu_fpriv *fpriv = filp->driver_priv; +int r;switch (args->in.op) {case AMDGPU_VM_OP_RESERVE_VMID : +if (!amdgpu_vm_dedicated_vmid_ready (&fpriv->vm)) { + r =amdgpu_vm_alloc_dedicated_vmid (adev, &fpriv->vm);Could you change the return value in this ready checking func? Usually we can easily get to know this kind of things. if (
sth_ready())do_sth();Quoted 7 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +3
-0+if (r) +return r; + }break ;default :return -EINVAL;
Quoted 4 lines
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
A message that is not in the archive (sent off-list or not yet received): 238b3d86-743a-4ca9-c008-92ae4e1ec79c-ANTagKRnAhcb1SvskN2V4Q@public.gmane.org
Re: [PATCH 3/6] drm/amdgpu: reserve vmid by vm ioctl
Quoted 34 lines · Christian König
Am 26.04.2017 um 09:09 schrieb Zhang, Jerry (Junwei):
Quoted 20 lines · Chunming Zhou
Quoted 17 lines
Change-Id I5f80dc39dc9d44660a96a2b710b0dbb4d3b9039dSigned-off-by Chunming Zhou <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 56 ++++++++++++++++++++++++++++++++++ 1 file changed, 56 insertions(+)drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +0
-0@@ -397,6 +397,17 @@ static bool amdgpu_vm_had_gpu_reset(struct amdgpu_device *adev,atomic_read (&adev->gpu_reset_counter); }Quoted 12 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c static bool amdgpu_vm_had_gpu_reset(struct amdgpu_device *adev, · line 406
Quoted 7 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +7
-0+static bool amdgpu_vm_dedicated_vmid_ready (struct amdgpu_vm *vm) +{ +unsigned vmhub; + +for (vmhub =0 ; vmhub <AMDGPU_MAX_VMHUBS ; vmhub++) { +if (!vm->dedicated_vmid[vmhub]) +return false;Just note:
Seemingly for pre-gmcv9, it will always allocate one more dedicated/reserved vmid for the 2nd un-used vmhub.
anyway it will not be used either.
Quoted 2 lines
It's even worse. IIRC we don't initialize the 2nd hub on pre-gmcv9. So that could work with uninitialized data.
I saw AMDGPU_MAX_VMHUBS is defined 2 for all ASICs.
So the 2nd dedicated_vmid here will be 0 always, returning false. Right?
Additionally, in amdgpu_vm_manager_init() the id_mgr[i] is initialized in loop of AMDGPU_MAX_VMHUBS as well. Thus I though both vmhub are initialized.
Any missing, please correct me.
Thanks.
Jerry
Quoted 90 lines
Christian.
Quoted 85 lines
Jerry
Quoted 55 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +4
-0+ } +return true; +} +/** * amdgpu_vm_grab_id - allocate the next free VMID *drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +39
-0@@ -546,6 +557,45 @@ int amdgpu_vm_grab_id(struct amdgpu_vm *vm, structamdgpu_ring *ring, return r; } +static int amdgpu_vm_alloc_dedicated_vmid (struct amdgpu_device *adev, +struct amdgpu_vm *vm) +{ +struct amdgpu_vm_id_manager *id_mgr; +struct amdgpu_vm_id *idle; +unsigned vmhub; +int r; + +for (vmhub =0 ; vmhub <AMDGPU_MAX_VMHUBS ; vmhub++) { + id_mgr = &adev->vm_manager.id_mgr[vmhub]; + +mutex_lock (&id_mgr->lock); +/* Select the first entry VMID */ + idle = list_first_entry(&id_mgr->ids_lru, struct amdgpu_vm_id, + list); +list_del_init (&idle->list); + vm->dedicated_vmid[vmhub] = idle; +mutex_unlock (&id_mgr->lock); + + r =amdgpu_sync_wait (&idle->active); +if (r) + goto err; + } + +return 0 ; +err: +for (vmhub =0 ; vmhub <AMDGPU_MAX_VMHUBS ; vmhub++) { + id_mgr = &adev->vm_manager.id_mgr[vmhub]; + +mutex_lock (&id_mgr->lock); +if (vm->dedicated_vmid[vmhub]) +list_add (&vm->dedicated_vmid[vmhub]->list, + &id_mgr->ids_lru); + vm->dedicated_vmid[vmhub] =NULL ; +mutex_unlock (&id_mgr->lock); + } +return r; +} +static bool amdgpu_vm_ring_has_compute_vm_bug (struct amdgpu_ring *ring) {struct amdgpu_device *adev = ring->adev;drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c int amdgpu_vm_ioctl(struct drm_device *dev, void · line 2438
Quoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +3
-0@@ -2379,9 +2429,15 @@ int amdgpu_vm_ioctl(struct drm_device *dev, void *data, struct drm_file *filp)union drm_amdgpu_vm *args = data;struct amdgpu_device *adev = dev->dev_private;struct amdgpu_fpriv *fpriv = filp->driver_priv; +int r;switch (args->in.op) {case AMDGPU_VM_OP_RESERVE_VMID : +if (!amdgpu_vm_dedicated_vmid_ready (&fpriv->vm)) { + r =amdgpu_vm_alloc_dedicated_vmid (adev, &fpriv->vm);Could you change the return value in this ready checking func? Usually we can easily get to know this kind of things. if (
sth_ready())do_sth();Quoted 7 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +3
-0+if (r) +return r; + }break ;default :return -EINVAL;_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
A message that is not in the archive (sent off-list or not yet received): [email protected]
Re: [PATCH 3/6] drm/amdgpu: reserve vmid by vm ioctl
Am 26.04.2017 um 11:04 schrieb Zhang, Jerry (Junwei):
Quoted 38 lines · Christian König
Quoted 34 lines
Am 26.04.2017 um 09:09 schrieb Zhang, Jerry (Junwei):
Quoted 20 lines · Chunming Zhou
Quoted 17 lines
Change-Id I5f80dc39dc9d44660a96a2b710b0dbb4d3b9039dSigned-off-by Chunming Zhou <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 56 ++++++++++++++++++++++++++++++++++ 1 file changed, 56 insertions(+)drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +0
-0@@ -397,6 +397,17 @@ static bool amdgpu_vm_had_gpu_reset(struct amdgpu_device *adev,atomic_read (&adev->gpu_reset_counter); }Quoted 12 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c static bool amdgpu_vm_had_gpu_reset(struct · line 407
Quoted 7 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +7
-0+static bool amdgpu_vm_dedicated_vmid_ready (struct amdgpu_vm *vm) +{ +unsigned vmhub; + +for (vmhub =0 ; vmhub <AMDGPU_MAX_VMHUBS ; vmhub++) { +if (!vm->dedicated_vmid[vmhub]) +return false;Just note:
Seemingly for pre-gmcv9, it will always allocate one more dedicated/reserved vmid for the 2nd un-used vmhub.
anyway it will not be used either.Quoted 2 lines
It's even worse. IIRC we don't initialize the 2nd hub on pre-gmcv9. So that could work with uninitialized data.
Quoted 8 lines
I saw
AMDGPU_MAX_VMHUBSis defined 2 for all ASICs.
So the 2nddedicated_vmidhere will be 0 always, returning false. Right?Additionally, in
amdgpu_vm_manager_init()the id_mgr[i] is initialized in loop ofAMDGPU_MAX_VMHUBSas well. Thus I though both vmhub are initialized.
Any missing, please correct me.
Thanks.
Indeed you are right we do always initialize all HUBs here, so that should work but is still a bit awkward.
Christian.
Quoted 97 lines
Jerry
Quoted 94 lines
Christian.
Quoted 89 lines
Jerry
Quoted 63 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +4
-0+ } +return true; +} +/** * amdgpu_vm_grab_id - allocate the next free VMID *drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +0
-0@@ -546,6 +557,45 @@ int amdgpu_vm_grab_id(struct amdgpu_vm *vm,struct
amdgpu_ring*ring,return r;}
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +14
-0+static int amdgpu_vm_alloc_dedicated_vmid (struct amdgpu_device *adev, +struct amdgpu_vm *vm) +{ +struct amdgpu_vm_id_manager *id_mgr; +struct amdgpu_vm_id *idle; +unsigned vmhub; +int r; + +for (vmhub =0 ; vmhub <AMDGPU_MAX_VMHUBS ; vmhub++) { + id_mgr = &adev->vm_manager.id_mgr[vmhub]; + +mutex_lock (&id_mgr->lock); +/* Select the first entry VMID */ + idle = list_first_entry(&id_mgr->ids_lru, struct
amdgpu_vm_id,drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +25
-0+ list); +list_del_init (&idle->list); + vm->dedicated_vmid[vmhub] = idle; +mutex_unlock (&id_mgr->lock); + + r =amdgpu_sync_wait (&idle->active); +if (r) + goto err; + } + +return 0 ; +err: +for (vmhub =0 ; vmhub <AMDGPU_MAX_VMHUBS ; vmhub++) { + id_mgr = &adev->vm_manager.id_mgr[vmhub]; + +mutex_lock (&id_mgr->lock); +if (vm->dedicated_vmid[vmhub]) +list_add (&vm->dedicated_vmid[vmhub]->list, + &id_mgr->ids_lru); + vm->dedicated_vmid[vmhub] =NULL ; +mutex_unlock (&id_mgr->lock); + } +return r; +} + static bool amdgpu_vm_ring_has_compute_vm_bug(struct amdgpu_ring*ring)
{ struct amdgpu_device *adev = ring->adev;drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +0
-0@@ -2379,9 +2429,15 @@ int amdgpu_vm_ioctl(struct drm_device *dev,void
*data, structdrm_file*filp)union drm_amdgpu_vm *args = data; struct amdgpu_device *adev = dev->dev_private; struct amdgpu_fpriv *fpriv = filp->driver_priv;drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
Quoted 6 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +3
-0+int r;switch (args->in.op) {case AMDGPU_VM_OP_RESERVE_VMID : +if (!amdgpu_vm_dedicated_vmid_ready (&fpriv->vm)) { + r =amdgpu_vm_alloc_dedicated_vmid (adev, &fpriv->vm);Could you change the return value in this ready checking func? Usually we can easily get to know this kind of things. if (
sth_ready())do_sth();drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c
Quoted 7 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +3
-0+if (r) +return r; + }break ;default :return -EINVAL ;_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 4/6 drm/amdgpu: add limitation for dedicated vm number v2
In reply to a message not in the archivev2: move #define to amdgpu_vm.h
--- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 9 +++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 2 ++ 4 files changed, 13 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu.h +1 -0
@@ -1591,6 +1591,7 @@ struct amdgpu_device {struct amdgpu_dummy_page dummy_page;struct amdgpu_vm_manager vm_manager;struct amdgpu_vmhub vmhub[AMDGPU_MAX_VMHUBS ]; +atomic_t reserved_vmid;/* memory management */ struct amdgpu_mman mman;
drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +1 -0
@@ -1889,6 +1889,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, adev->vm_manager.vm_pte_num_rings =0 ; adev->gart.gart_funcs =NULL ; adev->fence_context =kcl_fence_context_alloc (AMDGPU_MAX_RINGS ); +atomic_set (&adev->reserved_vmid,0 ); adev->smc_rreg = &amdgpu_invalid_rreg; adev->smc_wreg = &amdgpu_invalid_wreg;
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +9 -0
@@ -565,6 +565,10 @@ static int amdgpu_vm_alloc_dedicated_vmid(struct amdgpu_device *adev,unsigned vmhub;int r; +if (atomic_read (&adev->reserved_vmid) >=AMDGPU_VM_MAX_RESERVED_VMID ) { +DRM_ERROR ("Over limitation of reserved vmid\n" ); +return -EINVAL ; + }for (vmhub =0 ; vmhub <AMDGPU_MAX_VMHUBS ; vmhub++) { id_mgr = &adev->vm_manager.id_mgr[vmhub];
@@ -580,6 +584,7 @@ static int amdgpu_vm_alloc_dedicated_vmid(struct amdgpu_device *adev,if (r) goto err; } +atomic_inc (&adev->reserved_vmid);return 0 ; err:
@@ -2302,6 +2307,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) {struct amdgpu_bo_va_mapping *mapping, *tmp;bool prt_fini_needed = !!adev->gart.gart_funcs->set_prt; +bool dedicated = false;int i;if (vm->is_kfd_vm) {
@@ -2354,9 +2360,12 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)list_add (&vm->dedicated_vmid[i]->list, &id_mgr->ids_lru); vm->dedicated_vmid[i] =NULL ; + dedicated = true; }mutex_unlock (&id_mgr->lock); } +if (dedicated) +atomic_dec (&adev->reserved_vmid); }/**
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +2 -0
@@ -83,6 +83,8 @@/* hardcode that limit for now */ #define AMDGPU_VA_RESERVED_SIZE (8 << 20) +/* max vmids dedicated for process */ +#define AMDGPU_VM_MAX_RESERVED_VMID 2struct amdgpu_vm_pt {struct amdgpu_bo *bo;
Signature · 6 lines
1.9.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
A message that is not in the archive (sent off-list or not yet received): 1493013460-13344-5-git-send-email-David1.Zhou-5C7GfCeVMHo@public.gmane.org
Re: [PATCH 4/6] drm/amdgpu: add limitation for dedicated vm number v2
Quoted 16 lines · Chunming Zhou
v2: move #define to
amdgpu_vm.hChange-Id Ie5958cf6dbdc1c8278e61d9158483472d6f5c6e3Signed-off-by Chunming Zhou <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 9 +++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 2 ++ 4 files changed, 13 insertions(+)drivers/gpu/drm/amd/amdgpu/amdgpu.h +0
-0
drivers/gpu/drm/amd/amdgpu/amdgpu.h struct amdgpu_device { · line 1594
Quoted 4 lines
drivers/gpu/drm/amd/amdgpu/amdgpu.h +1
-0@@ -1591,6 +1591,7 @@ struct amdgpu_device {struct amdgpu_dummy_page dummy_page;struct amdgpu_vm_manager vm_manager;struct amdgpu_vmhub vmhub[AMDGPU_MAX_VMHUBS ]; +atomic_t reserved_vmid;
Perhaps it's more reasonable to belongs to vm_manager.
Jerry
Quoted 65 lines
drivers/gpu/drm/amd/amdgpu/amdgpu.h +0
-0/* memory management */ struct amdgpu_mman mman;drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +1
-0@@ -1889,6 +1889,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, adev->vm_manager.vm_pte_num_rings =0 ; adev->gart.gart_funcs =NULL ; adev->fence_context =kcl_fence_context_alloc (AMDGPU_MAX_RINGS ); +atomic_set (&adev->reserved_vmid,0 ); adev->smc_rreg = &amdgpu_invalid_rreg; adev->smc_wreg = &amdgpu_invalid_wreg;drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +9
-0@@ -565,6 +565,10 @@ static int amdgpu_vm_alloc_dedicated_vmid(struct amdgpu_device *adev,unsigned vmhub;int r; +if (atomic_read (&adev->reserved_vmid) >=AMDGPU_VM_MAX_RESERVED_VMID ) { +DRM_ERROR ("Over limitation of reserved vmid\n" ); +return -EINVAL ; + }for (vmhub =0 ; vmhub <AMDGPU_MAX_VMHUBS ; vmhub++) { id_mgr = &adev->vm_manager.id_mgr[vmhub];@@ -580,6 +584,7 @@ static int amdgpu_vm_alloc_dedicated_vmid(struct amdgpu_device *adev,if (r) goto err; } +atomic_inc (&adev->reserved_vmid);return 0 ; err:@@ -2302,6 +2307,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) {struct amdgpu_bo_va_mapping *mapping, *tmp;bool prt_fini_needed = !!adev->gart.gart_funcs->set_prt; +bool dedicated = false;int i;if (vm->is_kfd_vm) {@@ -2354,9 +2360,12 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)list_add (&vm->dedicated_vmid[i]->list, &id_mgr->ids_lru); vm->dedicated_vmid[i] =NULL ; + dedicated = true; }mutex_unlock (&id_mgr->lock); } +if (dedicated) +atomic_dec (&adev->reserved_vmid); }/** drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +0
-0
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h line 90
Quoted 8 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +2
-0@@ -83,6 +83,8 @@/* hardcode that limit for now */ #define AMDGPU_VA_RESERVED_SIZE (8 << 20) +/* max vmids dedicated for process */ +#define AMDGPU_VM_MAX_RESERVED_VMID 2struct amdgpu_vm_pt {struct amdgpu_bo *bo;
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
A message that is not in the archive (sent off-list or not yet received): [email protected]
Re: [PATCH 4/6] drm/amdgpu: add limitation for dedicated vm number v2
Am 26.04.2017 um 09:10 schrieb Zhang, Jerry (Junwei):
Quoted 12 lines · Chunming Zhou
Quoted 11 lines
v2: move #define to
amdgpu_vm.hChange-Id Ie5958cf6dbdc1c8278e61d9158483472d6f5c6e3Signed-off-by Chunming Zhou <[email protected]>--- drivers/gpu/drm/amd/amdgpu/amdgpu.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 9 +++++++++ drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 2 ++ 4 files changed, 13 insertions(+)
Quoted 11 lines
drivers/gpu/drm/amd/amdgpu/amdgpu.h struct amdgpu_device { · line 1594
Quoted 9 lines
drivers/gpu/drm/amd/amdgpu/amdgpu.h +1
-0@@ -1591,6 +1591,7 @@ struct amdgpu_device {struct amdgpu_dummy_page dummy_page;struct amdgpu_vm_manager vm_manager;struct amdgpu_vmhub vmhub[AMDGPU_MAX_VMHUBS ]; +atomic_t reserved_vmid;Perhaps it's more reasonable to belongs to
vm_manager.
Yes, agree.
Quoted 35 lines
Jerry
Quoted 32 lines
drivers/gpu/drm/amd/amdgpu/amdgpu.h +0
-0/* memory management */ struct amdgpu_mman mman;drivers/gpu/drm/amd/amdgpu/amdgpu_device.c +1
-0@@ -1889,6 +1889,7 @@ int amdgpu_device_init(struct amdgpu_device *adev, adev->vm_manager.vm_pte_num_rings =0 ; adev->gart.gart_funcs =NULL ; adev->fence_context =kcl_fence_context_alloc (AMDGPU_MAX_RINGS ); +atomic_set (&adev->reserved_vmid,0 ); adev->smc_rreg = &amdgpu_invalid_rreg; adev->smc_wreg = &amdgpu_invalid_wreg;drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +4
-0@@ -565,6 +565,10 @@ static int amdgpu_vm_alloc_dedicated_vmid(structamdgpu_device *adev, unsigned vmhub;int r; +if (atomic_read (&adev->reserved_vmid) >=AMDGPU_VM_MAX_RESERVED_VMID ) { +DRM_ERROR ("Over limitation of reserved vmid\n" ); +return -EINVAL ; + }for (vmhub =0 ; vmhub <AMDGPU_MAX_VMHUBS ; vmhub++) {id_mgr = &adev->vm_manager.id_mgr[vmhub];
Quoted 6 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c static int amdgpu_vm_alloc_dedicated_vmid(struct · line 588
Quoted 6 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +1
-0@@ -580,6 +584,7 @@ static int amdgpu_vm_alloc_dedicated_vmid(struct amdgpu_device *adev, if (r) goto err; } +atomic_inc (&adev->reserved_vmid);
That's not atomic at all, two threads could try to reserve a VMID at the same time and race.
You need to use atomic_inc_return for this before allocating the IDs and properly clean up if allocating them fails.
Quoted 32 lines
Quoted 32 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +1
-0return 0 ; err:@@ -2302,6 +2307,7 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm) {struct amdgpu_bo_va_mapping *mapping, *tmp;bool prt_fini_needed = !!adev->gart.gart_funcs->set_prt; +bool dedicated = false;int i;if (vm->is_kfd_vm) {drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +3
-0@@ -2354,9 +2360,12 @@ void amdgpu_vm_fini(struct amdgpu_device *adev, struct amdgpu_vm *vm)list_add (&vm->dedicated_vmid[i]->list, &id_mgr->ids_lru); vm->dedicated_vmid[i] =NULL ; + dedicated = true; }mutex_unlock (&id_mgr->lock); } +if (dedicated) +atomic_dec (&adev->reserved_vmid); }/**
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +0
-0@@ -83,6 +83,8 @@
Quoted 4 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h line 87
Quoted 4 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +2
-0/* hardcode that limit for now */ #define AMDGPU_VA_RESERVED_SIZE (8 << 20) +/* max vmids dedicated for process */ +#define AMDGPU_VM_MAX_RESERVED_VMID 2
This should be 1 instead of 2, or did I miss something?
Regards,
Christian.
Quoted 8 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h
Quoted 4 lines
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h +0
-0struct amdgpu_vm_pt {struct amdgpu_bo *bo;_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 5/6 drm/amdgpu: implement grab dedicated vmid V2
In reply to a message not in the archivev2: move sync waiting only when flush needs
--- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 61 ++++++++++++++++++++++++++++++++++ 1 file changed, 61 insertions(+)
drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c +61 -0
@@ -408,6 +408,63 @@ static bool amdgpu_vm_dedicated_vmid_ready(struct amdgpu_vm *vm)return true; } +static int amdgpu_vm_grab_dedicated_vmid (struct amdgpu_vm *vm, +struct amdgpu_ring *ring, +struct amdgpu_sync *sync, +struct fence *fence, +struct amdgpu_job *job) +{ +struct amdgpu_device *adev = ring->adev; +unsigned vmhub = ring->funcs->vmhub; +struct amdgpu_vm_id *id = vm->dedicated_vmid[vmhub]; +struct amdgpu_vm_id_manager *id_mgr = &adev->vm_manager.id_mgr[vmhub]; +struct fence *updates = sync->last_vm_update; +int r =0 ; +struct fence *flushed, *tmp; +bool needs_flush = false; + +mutex_lock (&id_mgr->lock); +if (amdgpu_vm_had_gpu_reset (adev, id)) + needs_flush = true; + + flushed = id->flushed_updates; +if (updates && (!flushed || updates->context != flushed->context || +fence_is_later (updates, flushed))) + needs_flush = true; +if (needs_flush) { + tmp =amdgpu_sync_get_fence (&id->active); +if (tmp) { + r =amdgpu_sync_fence (adev, sync, tmp); +fence_put (tmp); +mutex_unlock (&id_mgr->lock); +return r; + } + } + +/* Good we can use this VMID. Remember this submission as +* user of the VMID. +*/ + r =amdgpu_sync_fence (ring->adev, &id->active, fence); +if (r) + goto out; + +if (updates && (!flushed || updates->context != flushed->context || +fence_is_later (updates, flushed))) { +fence_put (id->flushed_updates); + id->flushed_updates =fence_get (updates); + } + id->pd_gpu_addr = job->vm_pd_addr; + id->current_gpu_reset_count =atomic_read (&adev->gpu_reset_counter); +atomic64_set (&id->owner, vm->client_id); + job->vm_needs_flush = needs_flush; + + job->vm_id = id - id_mgr->ids; +trace_amdgpu_vm_grab_id (vm, ring, job); +out: +mutex_unlock (&id_mgr->lock); +return r; +} +/** * amdgpu_vm_grab_id - allocate the next free VMID *
@@ -432,6 +489,10 @@ int amdgpu_vm_grab_id(struct amdgpu_vm *vm, struct amdgpu_ring *ring,unsigned i;int r =0 ; +if (amdgpu_vm_dedicated_vmid_ready (vm)) +return amdgpu_vm_grab_dedicated_vmid (vm, ring, sync, + fence, job); + fences =kmalloc_array (sizeof (void *), id_mgr->num_ids,GFP_KERNEL );if (!fences)return -ENOMEM ;
Signature · 6 lines
1.9.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
PATCH 6/6 drm/amdgpu: bump module verion for reserved vmid
In reply to a message not in the archive--- drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 3 ++- 1 file changed, 2 insertions(+), 1 deletion(-)
drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c +2 -1
@@ -64,9 +64,10 @@* - 3.12.0 - Add query for double offchip LDS buffers * - 3.13.0 - Add PRT support * - 3.14.0 - Fix race in amdgpu_ctx_get_fence() and note new functionality +* - 3.15.0 - Add reserved vmid support */ #define KMS_DRIVER_MAJOR 3 -#define KMS_DRIVER_MINOR 14 +#define KMS_DRIVER_MINOR 15#define KMS_DRIVER_PATCHLEVEL 0int amdgpu_vram_limit =0 ;
Signature · 6 lines
1.9.1 _______________________________________________ amd-gfx mailing list [email protected] https://lists.freedesktop.org/mailman/listinfo/amd-gfx
ping... anyone can give the review?
Thanks,
David Zhou
Quoted 31 lines · Chunming Zhou
The current kernel implementation, which grabs the idle VMID from pool when emitting the job may:
The back-to-back submission from one process could use different VMID. The submission to different queues from single process could use different VMID
It works well in most case but cannot work for the SQ thread trace capture.
The VMID for the submission that set the {SQTT}_BASE, which refers to the address of the trace buffer, is stored in shader engine.
If the profiling application have to use different VMIDs to submit IBs in its life cycle:
Some trace is not captured since it actually uses different VMID to submit jobs.
Some part of captured trace may come from different application since they are accidentally uses the owner’s VMID to submit jobs.V2:
1. address Christian's comments:
a. drop context flags for tag process, instead, add vm ioctl. b. change order of patches.
c. sync waiting only when vm flush needs.2. address Alex's comments;
bump module versionChunming Zhou (6):
drm/amdgpu: add vm ioctl
drm/amdgpu: add dedicated vmid field in vm struct drm/amdgpu: reserve vmid by vm ioctl drm/amdgpu: add limitation for dedicated vm number v2 drm/amdgpu: implement grab dedicated vmid V2 drm/amdgpu: bump module verion for reserved vmid
Quoted 9 lines
drivers/gpu/drm/amd/amdgpu/amdgpu.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 3 +- drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 159 ++++++++++++++++++++++++++++- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 5 + include/uapi/drm/amdgpu_drm.h | 20 ++++ 7 files changed, 188 insertions(+), 2 deletions(-)
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
A message that is not in the archive (sent off-list or not yet received): [email protected]
On my TODO list, but behind about two or three other items I need to tackle first.
Christian.
Am 25.04.2017 um 11:07 schrieb zhoucm1:
Quoted 52 lines
ping... anyone can give the review?
Thanks,
David ZhouQuoted 35 lines · Chunming Zhou
The current kernel implementation, which grabs the idle VMID from pool when emitting the job may:
The back-to-back submission from one process could use different
VMID.
The submission to different queues from single process could use
different VMIDIt works well in most case but cannot work for the SQ thread trace capture.
The VMID for the submission that set the {SQTT}_BASE, which refers to the address of the trace buffer, is stored in shader engine.
If the profiling application have to use different VMIDs to submit IBs in its life cycle:
Some trace is not captured since it actually uses different VMID
to submit jobs.
Some part of captured trace may come from different application
since they are accidentally uses the owner’s VMID to submit jobs.V2:
1. address Christian's comments:
a. drop context flags for tag process, instead, add vm ioctl. b. change order of patches.
c. sync waiting only when vm flush needs.2. address Alex's comments;
bump module versionChunming Zhou (6):
drm/amdgpu: add vm ioctl
drm/amdgpu: add dedicated vmid field in vm struct drm/amdgpu: reserve vmid by vm ioctl drm/amdgpu: add limitation for dedicated vm number v2 drm/amdgpu: implement grab dedicated vmid V2 drm/amdgpu: bump module verion for reserved vmidQuoted 10 lines
drivers/gpu/drm/amd/amdgpu/amdgpu.h | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_device.c | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_drv.c | 3 +- drivers/gpu/drm/amd/amdgpu/amdgpu_kms.c | 1 + drivers/gpu/drm/amd/amdgpu/amdgpu_vm.c | 159 ++++++++++++++++++++++++++++- drivers/gpu/drm/amd/amdgpu/amdgpu_vm.h | 5 + include/uapi/drm/amdgpu_drm.h | 20 ++++ 7 files changed, 188 insertions(+), 2 deletions(-)
Quoted 4 lines
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx
_______________________________________________
amd-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/amd-gfx