#0: remove output offsetting from reader

ttnn/cpp/ttnn/operations/data_movement/pad/device/kernels/dataflow/reader_pad_dims_rm_sharded_stickwise.cpp

@@ -23,29 +23,25 @@ void kernel_main() {
     constexpr uint32_t unpadded_stick_step = get_compile_time_arg_val(7);
     constexpr uint32_t padded_stick_step = get_compile_time_arg_val(8);
 
-    auto output_offset_bytes = get_arg_val<uint32_t>(0);
     uint32_t input_shard_base_addr = get_write_ptr(input_shard_cb);
     uint32_t output_shard_base_addr = get_write_ptr(output_shard_cb);
-    // for the first stick, we need to add the front padding by adjusting where
-    // we start writing into the output shard
-    uint32_t output_shard_addr = output_shard_base_addr + output_offset_bytes;
 
     auto input_stick_ptr = reinterpret_cast<u8_l1_ptr>(input_shard_base_addr);
-    auto output_stick_ptr = reinterpret_cast<u8_l1_ptr>(output_shard_addr);
+    auto output_stick_ptr = reinterpret_cast<u8_l1_ptr>(output_shard_base_addr);
 
     // fill the sticks that aren't entirely padding with data from the input tensor
     for (uint32_t h = 0; h < unpadded_shard_height; h++) {
         cb_wait_front(output_shard_cb, 1);  // wait for writer to fill this stick with padding
 
-        // read the input stick into the padded output stick starting after the
-        // front padding
-
         // FIXME: this isn't aligned. we need to do a memcpy for now. we can try
         // to do a noc_async_read later on with a trick.
+        //
+        // currently small noc transfers are slow, but once runtime drops an
+        // optimization (upcoming as of 12/12/2024) this might be worth
+        // investigating.
 
-        // noc_async_read(output_stick_addr + W_front_pad_bytes, input_stick_addr, unpadded_stick_bytes);
-
-        // NOTE: memcpy is safe here because the input/output tensors have disjoint buffers.
+        // read the input stick into the padded output stick starting after the
+        // front padding
         for (uint32_t i = 0; i < unpadded_stick_bytes; i++) {
             output_stick_ptr[W_front_pad_bytes + i] = input_stick_ptr[i];
         }

ttnn/cpp/ttnn/operations/data_movement/pad/device/pad_program_factory.cpp

@@ -1696,26 +1696,8 @@ operation::ProgramWithCallbacks pad_rm_sharded_width_only(
         all_cores_padded,
         tt::tt_metal::WriterDataMovementConfig(writer_ct_args));
 
-    uint32_t first_shard_output_offset = 0;
-
-    auto [which_shard, shard_coord] = tensor_coord_to_shard_coord<ShardStrategy::HEIGHT>(
-        output_tensor_shape.logical_shape().view(), shard_spec_padded.shape, input_tensor_start.view());
-
-    TT_FATAL(which_shard == 0, "pad_rm_sharded_stickwise: input tensor start should be in the first shard.");
-    uint32_t h_in_shard = shard_coord[0];
-    first_shard_output_offset = h_in_shard * tt::round_up(padded_stick_bytes, dram_alignment_bytes);
-
-    auto all_cores_padded_vec = corerange_to_cores(all_cores_padded, std::nullopt, true);
-
-    // Set runtime args for all cores
-    for (const auto& core : all_cores_padded_vec) {
-        // First core gets first_shard_output_offset, others get 0
-        uint32_t offset = (core == *all_cores_padded_vec.begin()) ? first_shard_output_offset : 0;
-        tt::tt_metal::SetRuntimeArgs(program, reader_kernel_id, core, {offset});
-        tt::tt_metal::SetRuntimeArgs(program, writer_kernel_id, core, {});
-    }
-
-    // FIXME: need to update runtime args?
+    tt::tt_metal::SetRuntimeArgs(program, reader_kernel_id, all_cores_padded, {});
+    tt::tt_metal::SetRuntimeArgs(program, writer_kernel_id, all_cores_padded, {});
 
     auto override_runtime_args_callback = [
             input_shard_cb,