[BugFix][Kernel] Fix Illegal memory access in causal_conv1d in H100 (vllm-project#9838)

Signed-off-by: mzusman <mor.zusmann@gmail.com>
Signed-off-by: s.kochetkov <s.m.kochetkov@tcsbank.ru>

Author: mzusman

csrc/mamba/causal_conv1d/causal_conv1d.cu

@@ -418,6 +418,31 @@ void causal_conv1d_fwd_kernel(ConvParamsBase params) {
             typename Ktraits::BlockStoreT(smem_store).Store(out, out_vals_store, seqlen - chunk * kChunkSize);
         }
         out += kChunkSize;
+
+        int final_state_position =  ((seqlen - (kWidth - 1)) - (n_chunks - 1) * kChunkSize);
+        // in case the final state is separated between the last "smem_exchange" and 
+        // and the one before it (chunk = n_chunks - 1 and chunk = n_chunks - 2), 
+        // (which occurs when `final_state_position` is a non-positivie index)
+        // we load the correct data from smem_exchange from both chunks, the last chunk iteration and the one before it
+        if (final_state_position < 0 && seqlen > kWidth){
+            input_t vals_load[kNElts] = {0};
+            if ((chunk == n_chunks - 2) && (tidx == kNThreads - 1)){
+                // chunk = n_chunks - 2, a segment of the final state sits in the last index
+                reinterpret_cast<vec_t *>(vals_load)[0] = smem_exchange[kNThreads - 1];
+                #pragma unroll
+                for (int w = 0; w < -final_state_position; ++w){
+                    conv_states[w] = vals_load[kNElts + final_state_position + w];
+                }
+            }
+            if ((chunk == n_chunks - 1) && tidx == 0){
+                // chunk = n_chunks - 1, the second segment of the final state first positions
+                reinterpret_cast<vec_t *>(vals_load)[0] = smem_exchange[0];
+                for (int w = -final_state_position; w < kWidth - 1; ++w){
+                    conv_states[w] = vals_load[w + final_state_position];
+                }
+                return;
+            }
+        }
     }
     // Final state is stored in the smem_exchange last token slot,
     // in case seqlen < kWidth, we would need to take the final state from the 
@@ -446,9 +471,14 @@ void causal_conv1d_fwd_kernel(ConvParamsBase params) {
         }
         else {
             // in case the final state is in between the threads data
-            reinterpret_cast<vec_t *>(x_vals_load)[1] = smem_exchange[last_thread + 1];
-            reinterpret_cast<vec_t *>(x_vals_load)[0] = smem_exchange[last_thread];
             const int offset = ((seqlen - (kWidth - 1)) % (kNElts));
+            if ((offset + kWidth - 2) >= kNElts && (last_thread + 1 < kNThreads)){
+                // In case last_thread == kNThreads - 1, accessing last_thread + 1 will result in a 
+                // illegal access error on H100.
+                // Therefore, we access last_thread + 1, only if the final state data sits there
+                reinterpret_cast<vec_t *>(x_vals_load)[1] = smem_exchange[last_thread + 1];
+            }
+            reinterpret_cast<vec_t *>(x_vals_load)[0] = smem_exchange[last_thread];
             #pragma unroll
             for (int w = 0; w < kWidth - 1; ++w){
                 conv_states[w] = x_vals_load[offset + w ];

tests/kernels/test_causal_conv1d.py

@@ -151,7 +151,7 @@ def causal_conv1d_opcheck_fn(x: torch.Tensor,
 @pytest.mark.parametrize("has_bias", [True])
 @pytest.mark.parametrize("width", [4])
 @pytest.mark.parametrize(
-    'seqlen', [1, 8, 16, 32, 64, 128, 256, 512, 784, 1024, 2048, 4096])
+    'seqlen', [1, 8, 16, 32, 64, 128, 256, 512, 784, 1024, 1025, 2048, 4096])
 @pytest.mark.parametrize('dim', [64])
 @pytest.mark.parametrize('batch', [1])
 def test_causal_conv1d(batch, dim, seqlen, width, has_bias, silu_activation,
@@ -420,7 +420,10 @@ def test_causal_conv1d_varlen(with_padding, dim, seqlen, width, has_bias,
 
     unpadded_out = out[:, :out_ref_tensor.shape[-1]]
     assert torch.allclose(unpadded_out, out_ref_tensor, rtol=rtol, atol=atol)
-    assert torch.allclose(final_states, final_states_ref, rtol=rtol, atol=atol)
+    assert torch.allclose(final_states[state_indices],
+                          final_states_ref[state_indices],
+                          rtol=rtol,
+                          atol=atol)
 
     causal_conv1d_opcheck_fn(x.squeeze(0), weight, bias, cumsum.cuda(),
                              padded_state_indices, has_initial_states,

tests/kernels/test_mamba_ssm.py

@@ -555,7 +555,7 @@ def test_selective_state_update_with_batch_indices(with_padding, dim, dstate,
     device = "cuda"
     rtol, atol = (3e-4, 1e-3) if itype == torch.float32 else (5e-3, 1e-2)
     if itype == torch.bfloat16:
-        rtol, atol = 7e-2, 7e-2
+        rtol, atol = 1e-1, 1e-1
         if torch.version.hip:
             atol *= 2
     # set seed
@@ -610,8 +610,8 @@ def test_selective_state_update_with_batch_indices(with_padding, dim, dstate,
                                          dt_bias=dt_bias,
                                          dt_softplus=True)
 
-    print("Output diff max", (out - out_ref[0]).max())
-    print("Output diff mean", (out - out_ref[0]).mean())
+    print("Output diff max", (out[:batch_size] - out_ref).max())
+    print("Output diff mean", (out[:batch_size] - out_ref).mean())
     print("Output state diff max", (state[state_indices, :] - state_ref).max())
     print("Output state diff mean",
           (state[state_indices, :] - state_ref).mean())