#0: add comments

tenstorrent · Dec 16, 2024 · cc2301b · cc2301b
1 parent 1b1fcf3
commit cc2301b
Show file tree

Hide file tree

Showing 3 changed files with 3 additions and 0 deletions.
diff --git a/...ovement/permute/device/kernels/dataflow/reader_permute_interleaved_rm_blocked_generic.cpp b/...ovement/permute/device/kernels/dataflow/reader_permute_interleaved_rm_blocked_generic.cpp
@@ -27,6 +27,7 @@ void kernel_main() {
     uint32_t end_block = get_arg_val<uint32_t>(2);
 
     // Input shape and strides (excluding W dimension and measured in rows, not bytes)
+    // start at runtime arg 3 since address/start_block/end_block make up the first 3 args
     uint32_t input_shape[N], src_strides[N];
     for (uint32_t i = 3; i < N + 3; i++) {
         input_shape[i - 3] = get_arg_val<uint32_t>(i);

diff --git a/...ovement/permute/device/kernels/dataflow/writer_permute_interleaved_rm_blocked_generic.cpp b/...ovement/permute/device/kernels/dataflow/writer_permute_interleaved_rm_blocked_generic.cpp
@@ -49,6 +49,7 @@ void kernel_main() {
     const InterleavedAddrGen<dst_is_dram> s0 = {.bank_base_address = dst_addr, .page_size = output_tensor_page_size};
 
     // Input shape, permutation, and destination strides
+    // start at runtime arg 3 since address/start_block/end_block make up the first 3 args
     uint32_t input_shape[N], perm[N], dest_strides[N];
     for (uint32_t i = 3; i < N + 3; i++) {
         input_shape[i - 3] = get_arg_val<uint32_t>(i);

diff --git a/..._movement/permute/device/kernels/dataflow/writer_permute_interleaved_rm_row_invariant.cpp b/..._movement/permute/device/kernels/dataflow/writer_permute_interleaved_rm_row_invariant.cpp
@@ -17,6 +17,7 @@ void kernel_main() {
 
     const InterleavedAddrGen<dst_is_dram> s0 = {.bank_base_address = dst_addr, .page_size = page_size};
 
+    // start at runtime arg 3 since address/start_block/end_block make up the first 3 args
     uint32_t input_shape[N], perm[N], dest_strides[N];
     for (uint32_t i = 3; i < N + 3; i++) {
         input_shape[i - 3] = get_arg_val<uint32_t>(i);