Skip to content

CANN: add RoPE cache preload before ACL graph capture#20747

Merged
hipudding merged 1 commit intoggml-org:masterfrom
noemotiovon:warmup
Mar 23, 2026
Merged

CANN: add RoPE cache preload before ACL graph capture#20747
hipudding merged 1 commit intoggml-org:masterfrom
noemotiovon:warmup

Conversation

@noemotiovon
Copy link
Copy Markdown
Collaborator

@noemotiovon noemotiovon commented Mar 19, 2026

ACL graph capture disallows host-to-device memcpy and device memory malloc/free on the captured stream. Pre-load the RoPE cache before capture so that:

  • Host-to-device copies and allocations run on the non-captured stream
  • Cache metadata is populated and memory pool is warmed up
  • During capture, only on-device computations are recorded; host-side and allocation branches are skipped

Make sure to read the contributing guidelines before submitting a PR

@noemotiovon noemotiovon requested a review from a team as a code owner March 19, 2026 06:36
@noemotiovon
Copy link
Copy Markdown
Collaborator Author

noemotiovon commented Mar 19, 2026

Test Log:

🕒 06:27:55 ⬢ [Docker] ➜ ./bin/test-backend-ops test -b CANN0 -o ROPE
Testing 2 devices

ggml_backend_cann_context: device 0 execution mode is GRAPH (acl graph enabled)
Backend 1/2: CANN0
  Device description: Ascend910B4
  Device memory: 30196 MB (29812 MB free)

new_pool_for_device: device 0 use vmm pool
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,40,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,52,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,64,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[16,16,8192,1],n_dims=16,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[64,1,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[64,71,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[64,8,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[80,32,2,1],n_dims=20,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[80,32,2,1],n_dims=32,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[80,32,4,1],n_dims=32,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[80,32,2,1],n_dims=20,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[80,32,2,1],n_dims=32,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[80,32,4,1],n_dims=32,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[16,16,8192,1],n_dims=16,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,12,2,1],n_dims=128,mode=8,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,28,2,1],n_dims=128,mode=8,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,12,2,1],n_dims=20,mode=8,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,28,2,1],n_dims=32,mode=8,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,12,2,1],n_dims=128,mode=40,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,28,2,1],n_dims=128,mode=40,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,12,2,1],n_dims=20,mode=40,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,28,2,1],n_dims=32,mode=40,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[80,16,2,1],n_dims=80,mode=24,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,16,2,1],n_dims=128,mode=40,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[16,16,8192,1],n_dims=16,mode=40,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,40,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,52,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,64,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[16,16,8192,1],n_dims=16,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[64,1,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[64,71,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[64,8,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[80,32,2,1],n_dims=20,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[80,32,2,1],n_dims=32,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[80,32,4,1],n_dims=32,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[80,32,2,1],n_dims=20,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[80,32,2,1],n_dims=32,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[80,32,4,1],n_dims=32,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[16,16,8192,1],n_dims=16,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,12,2,1],n_dims=128,mode=8,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,28,2,1],n_dims=128,mode=8,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,12,2,1],n_dims=20,mode=8,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,28,2,1],n_dims=32,mode=8,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,12,2,1],n_dims=128,mode=40,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,28,2,1],n_dims=128,mode=40,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,12,2,1],n_dims=20,mode=40,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,28,2,1],n_dims=32,mode=40,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[80,16,2,1],n_dims=80,mode=24,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,16,2,1],n_dims=128,mode=40,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[16,16,8192,1],n_dims=16,mode=40,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,40,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,52,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,64,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[16,16,8192,1],n_dims=16,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[64,1,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[64,71,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[64,8,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[80,32,2,1],n_dims=20,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[80,32,2,1],n_dims=32,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[80,32,4,1],n_dims=32,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[80,32,2,1],n_dims=20,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[80,32,2,1],n_dims=32,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[80,32,4,1],n_dims=32,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[16,16,8192,1],n_dims=16,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,12,2,1],n_dims=128,mode=8,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,28,2,1],n_dims=128,mode=8,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,12,2,1],n_dims=20,mode=8,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,28,2,1],n_dims=32,mode=8,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,12,2,1],n_dims=128,mode=40,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,28,2,1],n_dims=128,mode=40,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,12,2,1],n_dims=20,mode=40,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,28,2,1],n_dims=32,mode=40,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[80,16,2,1],n_dims=80,mode=24,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,16,2,1],n_dims=128,mode=40,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[16,16,8192,1],n_dims=16,mode=40,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,40,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,52,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,64,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[16,16,8192,1],n_dims=16,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[64,1,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[64,71,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[64,8,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[80,32,2,1],n_dims=20,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[80,32,2,1],n_dims=32,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[80,32,4,1],n_dims=32,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[80,32,2,1],n_dims=20,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[80,32,2,1],n_dims=32,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[80,32,4,1],n_dims=32,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[16,16,8192,1],n_dims=16,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,12,2,1],n_dims=128,mode=8,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,28,2,1],n_dims=128,mode=8,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,12,2,1],n_dims=20,mode=8,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,28,2,1],n_dims=32,mode=8,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,12,2,1],n_dims=128,mode=40,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,28,2,1],n_dims=128,mode=40,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,12,2,1],n_dims=20,mode=40,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,28,2,1],n_dims=32,mode=40,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[80,16,2,1],n_dims=80,mode=24,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,16,2,1],n_dims=128,mode=40,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[16,16,8192,1],n_dims=16,mode=40,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.424500,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.424500,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.424500,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.424500,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.424500,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.424500,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.424500,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.424500,ff=1,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.424500,ff=0,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.424500,ff=0,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.424500,ff=0,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.424500,ff=0,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.424500,ff=1,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.424500,ff=1,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.000000,af=1.424500,ff=1,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.000000,af=1.424500,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.746500,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.746500,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.746500,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.746500,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.746500,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.746500,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.746500,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.746500,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.746500,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.746500,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.746500,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.746500,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.746500,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.746500,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.746500,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.746500,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.746500,af=1.424500,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.746500,af=1.424500,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.746500,af=1.424500,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.746500,af=1.424500,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.746500,af=1.424500,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.746500,af=1.424500,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.746500,af=1.424500,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.746500,af=1.424500,ff=1,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.746500,af=1.424500,ff=0,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.746500,af=1.424500,ff=0,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.746500,af=1.424500,ff=0,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.746500,af=1.424500,ff=0,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.746500,af=1.424500,ff=1,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.746500,af=1.424500,ff=1,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.000000,ef=0.746500,af=1.424500,ff=1,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.000000,ef=0.746500,af=1.424500,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.000000,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.000000,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.000000,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.000000,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.000000,af=1.424500,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.000000,af=1.424500,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.000000,af=1.424500,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.000000,af=1.424500,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.000000,af=1.424500,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.000000,af=1.424500,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.000000,af=1.424500,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.000000,af=1.424500,ff=1,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.000000,af=1.424500,ff=0,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.000000,af=1.424500,ff=0,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.000000,af=1.424500,ff=0,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.000000,af=1.424500,ff=0,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.000000,af=1.424500,ff=1,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.000000,af=1.424500,ff=1,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.000000,af=1.424500,ff=1,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.000000,af=1.424500,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.746500,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.746500,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.746500,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.746500,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.746500,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.746500,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.746500,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.746500,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.746500,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.746500,af=1.000000,ff=0,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.746500,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.746500,af=1.000000,ff=0,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.746500,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.746500,af=1.000000,ff=1,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.746500,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.746500,af=1.000000,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=0,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=0,inplace=0): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=1,inplace=0): OK
  ROPE(type=f16,ne_a=[64,128,2,1],n_dims=64,mode=2,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=1,inplace=0): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=0,inplace=1): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=1,inplace=1): OK
  ROPE(type=f32,ne_a=[128,32,2,3],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=1,inplace=1): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=0,inplace=1): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=1,inplace=1): OK
  ROPE(type=f32,ne_a=[128,32,2,3],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=1,inplace=1): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=2,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=0,inplace=1): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=2,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=1,inplace=1): OK
  ROPE(type=f32,ne_a=[128,32,2,3],n_dims=128,mode=2,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=1,inplace=1): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=2,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=0,inplace=1): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=2,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=1,inplace=1): OK
  ROPE(type=f32,ne_a=[128,32,2,3],n_dims=128,mode=2,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=1,inplace=1): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=8,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=0,inplace=1): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=8,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=1,inplace=1): OK
  ROPE(type=f32,ne_a=[128,32,2,3],n_dims=128,mode=8,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=1,inplace=1): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=8,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=0,inplace=1): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=8,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=1,inplace=1): OK
  ROPE(type=f32,ne_a=[128,32,2,3],n_dims=128,mode=8,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=1,inplace=1): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=40,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=0,inplace=1): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=40,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=1,inplace=1): OK
  ROPE(type=f32,ne_a=[128,32,2,3],n_dims=128,mode=40,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=1,inplace=1): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=40,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=0,inplace=1): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=40,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=1,inplace=1): OK
  ROPE(type=f32,ne_a=[128,32,2,3],n_dims=128,mode=40,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=1,inplace=1): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=24,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=0,inplace=1): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=24,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=1,inplace=1): OK
  ROPE(type=f32,ne_a=[128,32,2,3],n_dims=128,mode=24,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=1,inplace=1): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=24,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=0,inplace=1): OK
  ROPE(type=f32,ne_a=[128,32,2,1],n_dims=128,mode=24,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=1,inplace=1): OK
  ROPE(type=f32,ne_a=[128,32,2,3],n_dims=128,mode=24,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=1,inplace=1): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=0,inplace=1): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=1,inplace=1): OK
  ROPE(type=f16,ne_a=[128,32,2,3],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=1,inplace=1): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=0,inplace=1): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=1,inplace=1): OK
  ROPE(type=f16,ne_a=[128,32,2,3],n_dims=128,mode=0,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=1,inplace=1): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=2,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=0,inplace=1): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=2,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=1,inplace=1): OK
  ROPE(type=f16,ne_a=[128,32,2,3],n_dims=128,mode=2,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=1,inplace=1): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=2,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=0,inplace=1): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=2,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=1,inplace=1): OK
  ROPE(type=f16,ne_a=[128,32,2,3],n_dims=128,mode=2,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=1,inplace=1): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=8,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=0,inplace=1): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=8,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=1,inplace=1): OK
  ROPE(type=f16,ne_a=[128,32,2,3],n_dims=128,mode=8,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=1,inplace=1): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=8,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=0,inplace=1): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=8,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=1,inplace=1): OK
  ROPE(type=f16,ne_a=[128,32,2,3],n_dims=128,mode=8,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=1,inplace=1): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=40,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=0,inplace=1): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=40,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=1,inplace=1): OK
  ROPE(type=f16,ne_a=[128,32,2,3],n_dims=128,mode=40,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=1,inplace=1): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=40,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=0,inplace=1): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=40,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=1,inplace=1): OK
  ROPE(type=f16,ne_a=[128,32,2,3],n_dims=128,mode=40,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=1,inplace=1): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=24,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=0,inplace=1): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=24,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=1,inplace=1): OK
  ROPE(type=f16,ne_a=[128,32,2,3],n_dims=128,mode=24,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=0,v=1,inplace=1): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=24,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=0,inplace=1): OK
  ROPE(type=f16,ne_a=[128,32,2,1],n_dims=128,mode=24,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=1,inplace=1): OK
  ROPE(type=f16,ne_a=[128,32,2,3],n_dims=128,mode=24,n_ctx=512,fs=1.424500,ef=0.746500,af=1.424500,ff=1,v=1,inplace=1): OK
  288/288 tests passed
  Backend CANN0: OK
Backend 2/2: CPU
  Skipping
2/2 backends passed
OK

@github-actions github-actions Bot added ggml changes relating to the ggml tensor library for machine learning Ascend NPU issues specific to Ascend NPUs labels Mar 19, 2026
@noemotiovon
CANN: add RoPE cache preload before ACL graph capture
0fcc4cc 
ACL graph capture disallows host-to-device memcpy and device memory
malloc/free on the captured stream. Pre-load the RoPE cache before
capture so that:
- Host-to-device copies and allocations run on the non-captured stream
- Cache metadata is populated and memory pool is warmed up
- During capture, only on-device computations are recorded; host-side
  and allocation branches are skipped
@loci-dev loci-dev mentioned this pull request Mar 20, 2026
Open
hipudding
hipudding reviewed Mar 20, 2026
View reviewed changes
Comment thread ggml/src/ggml-cann/aclnn_ops.cpp
// Run the full cache init on the non-captured stream. This performs all
// host-to-device memcpy, aclrtMalloc/Free, and on-device computations
// so that the memory pool is warmed up and cache metadata is populated.
aclnn_rope_cache_init(ctx, dst, corr_dims, ext_factor, theta_scale, freq_scale, attn_factor, is_neox, sections,
Copy link
Copy Markdown
Contributor

@hipudding hipudding Mar 20, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

cache init do a lot of pre calculation, here we only need cache all cpu calculation results. You may refactor it later.

Copy link
Copy Markdown
Collaborator Author

@noemotiovon noemotiovon Mar 20, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yes, I took a shortcut here and directly reused the existing method. Later, I will customize the initialization process for the host data in RoPE.

Comment thread ggml/src/ggml-cann/common.h
}

if (node->op == GGML_OP_SCALE || node->op == GGML_OP_UNARY || node->op == GGML_OP_GLU) {
if (node->op == GGML_OP_SCALE || node->op == GGML_OP_UNARY || node->op == GGML_OP_GLU || node->op == GGML_OP_ROPE){
Copy link
Copy Markdown
Contributor

@hipudding hipudding Mar 20, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Why there is not a problem before since ROPE param is not stored.

Copy link
Copy Markdown
Collaborator Author

@noemotiovon noemotiovon Mar 20, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Previously, RoPE was not compared during graph matching because, in model inference, it’s unnecessary to compare the parameters since they are consistent. After adding it, RoPE can now pass precision tests in ACL graph mode.

@hipudding hipudding merged commit 07ff000 into ggml-org:master Mar 23, 2026
85 of 95 checks passed
Seunghhon pushed a commit to Seunghhon/llama.cpp that referenced this pull request Apr 26, 2026
@noemotiovon
CANN: add RoPE cache preload before ACL graph capture (ggml-org#20747)
e33da52 
ACL graph capture disallows host-to-device memcpy and device memory
malloc/free on the captured stream. Pre-load the RoPE cache before
capture so that:
- Host-to-device copies and allocations run on the non-captured stream
- Cache metadata is populated and memory pool is warmed up
- During capture, only on-device computations are recorded; host-side
  and allocation branches are skipped
rsenthilkumar6 pushed a commit to rsenthilkumar6/llama.cpp that referenced this pull request May 1, 2026
@noemotiovon @rsenthilkumar6
CANN: add RoPE cache preload before ACL graph capture (ggml-org#20747)
ac78341 
ACL graph capture disallows host-to-device memcpy and device memory
malloc/free on the captured stream. Pre-load the RoPE cache before
capture so that:
- Host-to-device copies and allocations run on the non-captured stream
- Cache metadata is populated and memory pool is warmed up
- During capture, only on-device computations are recorded; host-side
  and allocation branches are skipped
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@hipudding hipudding hipudding approved these changes

Assignees

No one assigned

Labels

Ascend NPU issues specific to Ascend NPUs ggml changes relating to the ggml tensor library for machine learning

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

2 participants

@noemotiovon @hipudding