diff --git a/examples/a2a3/tensormap_and_ringbuffer/paged_attention_unroll_manual_scope/README.md b/examples/a2a3/tensormap_and_ringbuffer/paged_attention_unroll_manual_scope/README.md index 4f896aa6b2..9c2def5681 100644 --- a/examples/a2a3/tensormap_and_ringbuffer/paged_attention_unroll_manual_scope/README.md +++ b/examples/a2a3/tensormap_and_ringbuffer/paged_attention_unroll_manual_scope/README.md @@ -34,8 +34,7 @@ Softmax becomes a two-pass over a wider buffer, which is why that kernel is rewritten rather than adjusted. `N_UNROLL` is a `#define` at the top of the orchestration source, currently -**64**. (The file's header comment still says `N_UNROLL=8`; the `#define` is -what takes effect.) +**64**. Dependencies are declared explicitly inside `PTO2_SCOPE(PTO2ScopeMode::MANUAL)`, using the primitive diff --git a/examples/a2a3/tensormap_and_ringbuffer/paged_attention_unroll_manual_scope/kernels/orchestration/paged_attention_orch.cpp b/examples/a2a3/tensormap_and_ringbuffer/paged_attention_unroll_manual_scope/kernels/orchestration/paged_attention_orch.cpp index 82bc89f374..dbeddf4574 100644 --- a/examples/a2a3/tensormap_and_ringbuffer/paged_attention_unroll_manual_scope/kernels/orchestration/paged_attention_orch.cpp +++ b/examples/a2a3/tensormap_and_ringbuffer/paged_attention_unroll_manual_scope/kernels/orchestration/paged_attention_orch.cpp @@ -9,7 +9,7 @@ * ----------------------------------------------------------------------------------------------------------- */ /** - * Paged Attention Orchestration Function V2 - N_UNROLL=8, 4 Tasks Per Group + * Paged Attention Orchestration Function V2 - 4 Tasks Per Group * * Batches up to N_UNROLL blocks per group. Each group submits exactly 4 tasks: * 1. QK matmul: qi @ K^T for n_blocks → sij_buf (q_tile, n_blocks * block_size) diff --git a/examples/a5/tensormap_and_ringbuffer/paged_attention_unroll_manual_scope/kernels/orchestration/paged_attention_orch.cpp b/examples/a5/tensormap_and_ringbuffer/paged_attention_unroll_manual_scope/kernels/orchestration/paged_attention_orch.cpp index 0db4b6c0d9..524b1f67c3 100644 --- a/examples/a5/tensormap_and_ringbuffer/paged_attention_unroll_manual_scope/kernels/orchestration/paged_attention_orch.cpp +++ b/examples/a5/tensormap_and_ringbuffer/paged_attention_unroll_manual_scope/kernels/orchestration/paged_attention_orch.cpp @@ -9,7 +9,7 @@ * ----------------------------------------------------------------------------------------------------------- */ /** - * Paged Attention Orchestration Function V2 - N_UNROLL=8, 4 Tasks Per Group + * Paged Attention Orchestration Function V2 - 4 Tasks Per Group * * Batches up to N_UNROLL blocks per group. Each group submits exactly 4 tasks: * 1. QK matmul: qi @ K^T for n_blocks -> sij_buf (q_tile, n_blocks * block_size) diff --git a/tests/st/a2a3/tensormap_and_ringbuffer/paged_attention_unroll/kernels/orchestration/paged_attention_orch.cpp b/tests/st/a2a3/tensormap_and_ringbuffer/paged_attention_unroll/kernels/orchestration/paged_attention_orch.cpp index 0978073d96..fb9d0feb69 100644 --- a/tests/st/a2a3/tensormap_and_ringbuffer/paged_attention_unroll/kernels/orchestration/paged_attention_orch.cpp +++ b/tests/st/a2a3/tensormap_and_ringbuffer/paged_attention_unroll/kernels/orchestration/paged_attention_orch.cpp @@ -9,7 +9,7 @@ * ----------------------------------------------------------------------------------------------------------- */ /** - * Paged Attention Orchestration Function V2 - N_UNROLL=8, 4 Tasks Per Group + * Paged Attention Orchestration Function V2 - 4 Tasks Per Group * * Batches up to N_UNROLL blocks per group. Each group submits exactly 4 tasks: * 1. QK matmul: qi @ K^T for n_blocks → sij_buf (q_tile, n_blocks * block_size) diff --git a/tests/st/a5/tensormap_and_ringbuffer/paged_attention_unroll/kernels/orchestration/paged_attention_orch.cpp b/tests/st/a5/tensormap_and_ringbuffer/paged_attention_unroll/kernels/orchestration/paged_attention_orch.cpp index 4fd91db8c9..02f3975fc0 100644 --- a/tests/st/a5/tensormap_and_ringbuffer/paged_attention_unroll/kernels/orchestration/paged_attention_orch.cpp +++ b/tests/st/a5/tensormap_and_ringbuffer/paged_attention_unroll/kernels/orchestration/paged_attention_orch.cpp @@ -9,7 +9,7 @@ * ----------------------------------------------------------------------------------------------------------- */ /** - * Paged Attention Orchestration Function V2 - N_UNROLL=8, 4 Tasks Per Group + * Paged Attention Orchestration Function V2 - 4 Tasks Per Group * * Batches up to N_UNROLL blocks per group. Each group submits exactly 4 tasks: * 1. QK matmul: qi @ K^T for n_blocks → sij_buf (q_tile, n_blocks * block_size)