chore: clean comments and add requirement installation instructions in README

Author: cenxi.lx

README.md

@@ -70,6 +70,8 @@ dq, dk, dv, db, dg, dh0 = chunk_gated_delta_rule_bwd(
 ## Tests
 
 ```bash
+pip install flash_linear_attention==0.5.0 # require flash linear attention for comparison
+
 cd tests
 python test_gdr.py --set develop
 python test_gdr.py --set varlen --num-heads 32
@@ -90,6 +92,8 @@ Specifically, the forward (FWD) benchmarks measure single-kernel latency for dif
 More detail in [./benchmark/benchmark_results_H200.txt].
 
 ```bash
+pip install flash_linear_attention==0.5.0 flashinfer-python==0.6.9 # require flash linear attention and flashinfer for comparison
+
 cd benchmark
 python bench_gated_delta_rule.py
 ```

flash_qla/ops/gated_delta_rule/chunk/hopper/fused_fwd.py

@@ -154,18 +154,6 @@ def tilelang_fused_chunk_gdr_fwd_kernel(
             bar_4 = T.alloc_barrier(arrive_count=128)
             bar_5 = T.alloc_barrier(arrive_count=416)
 
-            # T.annotate_layout({
-            #     q_shared: tilelang.layout.make_swizzled_layout(q_shared[0, :, :]).expand([2]),
-            #     k_shared: tilelang.layout.make_swizzled_layout(k_shared[0, :, :]).expand([2]),
-            #     v_shared: tilelang.layout.make_swizzled_layout(v_shared[0, :, :]).expand([2]),
-            #     a_shared: tilelang.layout.make_swizzled_layout(a_shared[0, :, :]).expand([2]),
-            #     o_shared: tilelang.layout.make_swizzled_layout(o_shared),
-            #     h_shared: tilelang.layout.make_swizzled_layout(h_shared),
-            #     vd_shared: tilelang.layout.make_swizzled_layout(vd_shared),
-            #     vn_shared: tilelang.layout.make_swizzled_layout(vn_shared),
-            #     p_shared: tilelang.layout.make_swizzled_layout(p_shared),
-            # })
-
             T.use_swizzle(10)
 
             tx = T.get_thread_binding()