[Kernel] Integrate CUTLASS MoE kernel with PPLX

CMakeLists.txt

@@ -543,8 +543,8 @@ if(VLLM_GPU_LANG STREQUAL "CUDA")
   # CUTLASS MoE kernels
 
   # The MoE kernel cutlass_moe_mm requires CUDA 12.3 or later (and only works
-  # on Hopper). get_cutlass_moe_mm_data should only be compiled if it's possible
-  # to compile MoE kernels that use its output.
+  # on Hopper). get_cutlass_(pplx_)moe_mm_data should only be compiled
+  # if it's possible to compile MoE kernels that use its output.
   cuda_archs_loose_intersection(SCALED_MM_ARCHS "9.0a;10.0a" "${CUDA_ARCHS}")
   if(${CMAKE_CUDA_COMPILER_VERSION} VERSION_GREATER_EQUAL 12.3 AND SCALED_MM_ARCHS)
     set(SRCS "csrc/quantization/cutlass_w8a8/moe/grouped_mm_c3x.cu"

benchmarks/kernels/benchmark_grouped_gemm_cutlass.py

@@ -6,8 +6,8 @@
 
 from vllm import _custom_ops as ops
 from vllm.config import ParallelConfig, VllmConfig, set_current_vllm_config
+from vllm.model_executor.layers.fused_moe.cutlass_moe import cutlass_moe_fp8
 from vllm.model_executor.layers.fused_moe.fused_moe import (
-    cutlass_moe_fp8,
     fused_experts,
     fused_topk,
 )
@@ -69,18 +69,9 @@ def bench_run(
     w1_scale = torch.empty((num_experts, 1, 1), device="cuda", dtype=torch.float32)
     w2_scale = torch.empty((num_experts, 1, 1), device="cuda", dtype=torch.float32)
 
-    ab_strides1 = torch.full((num_experts,), k, device="cuda", dtype=torch.int64)
-    c_strides1 = torch.full((num_experts,), 2 * n, device="cuda", dtype=torch.int64)
-    ab_strides2 = torch.full((num_experts,), n, device="cuda", dtype=torch.int64)
-    c_strides2 = torch.full((num_experts,), k, device="cuda", dtype=torch.int64)
-
     for expert in range(num_experts):
         w1_q[expert], w1_scale[expert] = ops.scaled_fp8_quant(w1[expert])
         w2_q[expert], w2_scale[expert] = ops.scaled_fp8_quant(w2[expert])
-    w1_q_notransp = w1_q.clone()
-    w2_q_notransp = w2_q.clone()
-    w1_q = w1_q.transpose(1, 2)
-    w2_q = w2_q.transpose(1, 2)
 
     score = torch.randn((m, num_experts), device="cuda", dtype=dtype)
 
@@ -121,25 +112,17 @@ def run_cutlass_moe(
         w2_scale: torch.Tensor,
         topk_weights: torch.Tensor,
         topk_ids: torch.Tensor,
-        ab_strides1: torch.Tensor,
-        c_strides1: torch.Tensor,
-        ab_strides2: torch.Tensor,
-        c_strides2: torch.Tensor,
         num_repeats: int,
     ):
         for _ in range(num_repeats):
             cutlass_moe_fp8(
                 a,
                 w1,
                 w2,
-                w1_scale,
-                w2_scale,
                 topk_weights,
                 topk_ids,
-                ab_strides1,
-                c_strides1,
-                ab_strides2,
-                c_strides2,
+                w1_scale,
+                w2_scale,
                 a1_scale=a_scale,
             )
 
@@ -152,10 +135,6 @@ def run_cutlass_from_graph(
         w2_scale: torch.Tensor,
         topk_weights: torch.Tensor,
         topk_ids: torch.Tensor,
-        ab_strides1: torch.Tensor,
-        c_strides1: torch.Tensor,
-        ab_strides2: torch.Tensor,
-        c_strides2: torch.Tensor,
     ):
         with set_current_vllm_config(
             VllmConfig(parallel_config=ParallelConfig(pipeline_parallel_size=1))
@@ -164,14 +143,10 @@ def run_cutlass_from_graph(
                 a,
                 w1_q,
                 w2_q,
-                w1_scale,
-                w2_scale,
                 topk_weights,
                 topk_ids,
-                ab_strides1,
-                c_strides1,
-                ab_strides2,
-                c_strides2,
+                w1_scale,
+                w2_scale,
                 a1_scale=a_scale,
             )
 
@@ -217,10 +192,6 @@ def replay_graph(graph, num_repeats):
             w2_scale,
             topk_weights,
             topk_ids,
-            ab_strides1,
-            c_strides1,
-            ab_strides2,
-            c_strides2,
         )
     torch.cuda.synchronize()
 
@@ -229,8 +200,8 @@ def replay_graph(graph, num_repeats):
     with torch.cuda.graph(triton_graph, stream=triton_stream):
         run_triton_from_graph(
             a,
-            w1_q_notransp,
-            w2_q_notransp,
+            w1_q,
+            w2_q,
             topk_weights,
             topk_ids,
             w1_scale,
@@ -249,18 +220,12 @@ def replay_graph(graph, num_repeats):
         "w2": w2,
         "score": score,
         "topk": topk,
-        "w1_q_notransp": w1_q_notransp,
-        "w2_q_notransp": w2_q_notransp,
         # Cutlass params
         "a_scale": a_scale,
         "w1_q": w1_q,
         "w2_q": w2_q,
         "w1_scale": w1_scale,
         "w2_scale": w2_scale,
-        "ab_strides1": ab_strides1,
-        "c_strides1": c_strides1,
-        "ab_strides2": ab_strides2,
-        "c_strides2": c_strides2,
         # cuda graph params
         "cutlass_graph": cutlass_graph,
         "triton_graph": triton_graph,
@@ -278,8 +243,8 @@ def replay_graph(graph, num_repeats):
     # Warmup
     run_triton_moe(
         a,
-        w1_q_notransp,
-        w2_q_notransp,
+        w1_q,
+        w2_q,
         topk_weights,
         topk_ids,
         w1_scale,
@@ -290,7 +255,7 @@ def replay_graph(graph, num_repeats):
 
     results.append(
         benchmark.Timer(
-            stmt="run_triton_moe(a, w1_q_notransp, w2_q_notransp, topk_weights, topk_ids, w1_scale, w2_scale, a_scale, num_runs)",  # noqa: E501
+            stmt="run_triton_moe(a, w1_q, w2_q, topk_weights, topk_ids, w1_scale, w2_scale, a_scale, num_runs)",  # noqa: E501
             globals=globals,
             label=label,
             sub_label=sub_label,
@@ -321,16 +286,12 @@ def replay_graph(graph, num_repeats):
         w2_scale,
         topk_weights,
         topk_ids,
-        ab_strides1,
-        c_strides1,
-        ab_strides2,
-        c_strides2,
         num_warmup,
     )
 
     results.append(
         benchmark.Timer(
-            stmt="run_cutlass_moe(a, a_scale, w1_q, w2_q, w1_scale, w2_scale, topk_weights, topk_ids, ab_strides1, c_strides1, ab_strides2, c_strides2, num_runs)",  # noqa: E501
+            stmt="run_cutlass_moe(a, a_scale, w1_q, w2_q, w1_scale, w2_scale, topk_weights, topk_ids, num_runs)",  # noqa: E501
             globals=globals,
             label=label,
             sub_label=sub_label,

csrc/ops.h

@@ -231,7 +231,8 @@ void cutlass_moe_mm(
     torch::Tensor const& b_tensors, torch::Tensor const& a_scales,
     torch::Tensor const& b_scales, torch::Tensor const& expert_offsets,
     torch::Tensor const& problem_sizes, torch::Tensor const& a_strides,
-    torch::Tensor const& b_strides, torch::Tensor const& c_strides);
+    torch::Tensor const& b_strides, torch::Tensor const& c_strides,
+    bool per_act_token, bool per_out_ch);
 
 void cutlass_fp4_group_mm(
     torch::Tensor& output, const torch::Tensor& a, const torch::Tensor& b,
@@ -245,6 +246,12 @@ void get_cutlass_moe_mm_data(
     torch::Tensor& input_permutation, torch::Tensor& output_permutation,
     const int64_t num_experts, const int64_t n, const int64_t k);
 
+void get_cutlass_pplx_moe_mm_data(
+    torch::Tensor& expert_offsets, torch::Tensor& problem_sizes1,
+    torch::Tensor& problem_sizes2, const torch::Tensor& expert_num_tokens,
+    const torch::Tensor& non_zero_expert_idxs, const int64_t num_local_experts,
+    const int64_t padded_m, const int64_t n, const int64_t k);
+
 void cutlass_scaled_mm_azp(torch::Tensor& out, torch::Tensor const& a,
                            torch::Tensor const& b,
                            torch::Tensor const& a_scales,

csrc/quantization/cutlass_w8a8/moe/grouped_mm_c3x.cu

@@ -84,7 +84,8 @@ void run_cutlass_moe_mm_sm90(
     torch::Tensor const& b_tensors, torch::Tensor const& a_scales,
     torch::Tensor const& b_scales, torch::Tensor const& expert_offsets,
     torch::Tensor const& problem_sizes, torch::Tensor const& a_strides,
-    torch::Tensor const& b_strides, torch::Tensor const& c_strides) {
+    torch::Tensor const& b_strides, torch::Tensor const& c_strides,
+    bool per_act_token, bool per_out_ch) {
   TORCH_CHECK(a_tensors.size(0) > 0, "No input A tensors provided.");
   TORCH_CHECK(b_tensors.size(0) > 0, "No input B tensors provided.");
   TORCH_CHECK(out_tensors.size(0) > 0, "No output tensors provided.");
@@ -113,19 +114,23 @@ void run_cutlass_moe_mm_sm90(
   if (n >= 8192) {
     cutlass_group_gemm_caller<Cutlass3xGemmN8192>(
         out_tensors, a_tensors, b_tensors, a_scales, b_scales, expert_offsets,
-        problem_sizes, a_strides, b_strides, c_strides);
+        problem_sizes, a_strides, b_strides, c_strides, per_act_token,
+        per_out_ch);
   } else if (k >= 8192) {
     cutlass_group_gemm_caller<Cutlass3xGemmK8192>(
         out_tensors, a_tensors, b_tensors, a_scales, b_scales, expert_offsets,
-        problem_sizes, a_strides, b_strides, c_strides);
+        problem_sizes, a_strides, b_strides, c_strides, per_act_token,
+        per_out_ch);
   } else if (m <= 16) {
     cutlass_group_gemm_caller<Cutlass3xGemmM16>(
         out_tensors, a_tensors, b_tensors, a_scales, b_scales, expert_offsets,
-        problem_sizes, a_strides, b_strides, c_strides);
+        problem_sizes, a_strides, b_strides, c_strides, per_act_token,
+        per_out_ch);
   } else {
     cutlass_group_gemm_caller<Cutlass3xGemmDefault>(
         out_tensors, a_tensors, b_tensors, a_scales, b_scales, expert_offsets,
-        problem_sizes, a_strides, b_strides, c_strides);
+        problem_sizes, a_strides, b_strides, c_strides, per_act_token,
+        per_out_ch);
   }
 }
 
@@ -134,15 +139,18 @@ void dispatch_moe_mm_sm90(
     torch::Tensor const& b_tensors, torch::Tensor const& a_scales,
     torch::Tensor const& b_scales, torch::Tensor const& expert_offsets,
     torch::Tensor const& problem_sizes, torch::Tensor const& a_strides,
-    torch::Tensor const& b_strides, torch::Tensor const& c_strides) {
+    torch::Tensor const& b_strides, torch::Tensor const& c_strides,
+    bool per_act_token, bool per_out_ch) {
   if (out_tensors.dtype() == torch::kBFloat16) {
     run_cutlass_moe_mm_sm90<cutlass::float_e4m3_t, cutlass::bfloat16_t>(
         out_tensors, a_tensors, b_tensors, a_scales, b_scales, expert_offsets,
-        problem_sizes, a_strides, b_strides, c_strides);
+        problem_sizes, a_strides, b_strides, c_strides, per_act_token,
+        per_out_ch);
   } else {
     run_cutlass_moe_mm_sm90<cutlass::float_e4m3_t, cutlass::half_t>(
         out_tensors, a_tensors, b_tensors, a_scales, b_scales, expert_offsets,
-        problem_sizes, a_strides, b_strides, c_strides);
+        problem_sizes, a_strides, b_strides, c_strides, per_act_token,
+        per_out_ch);
   }
 }
 
@@ -153,8 +161,9 @@ void cutlass_moe_mm_sm90(
     torch::Tensor const& b_tensors, torch::Tensor const& a_scales,
     torch::Tensor const& b_scales, torch::Tensor const& expert_offsets,
     torch::Tensor const& problem_sizes, torch::Tensor const& a_strides,
-    torch::Tensor const& b_strides, torch::Tensor const& c_strides) {
+    torch::Tensor const& b_strides, torch::Tensor const& c_strides,
+    bool per_act_token, bool per_out_ch) {
   dispatch_moe_mm_sm90(out_tensors, a_tensors, b_tensors, a_scales, b_scales,
                        expert_offsets, problem_sizes, a_strides, b_strides,
-                       c_strides);
+                       c_strides, per_act_token, per_out_ch);
 }

csrc/quantization/cutlass_w8a8/moe/grouped_mm_c3x.cuh

@@ -76,17 +76,15 @@ void cutlass_group_gemm_caller(
     torch::Tensor const& b_tensors, torch::Tensor const& a_scales,
     torch::Tensor const& b_scales, torch::Tensor const& expert_offsets,
     torch::Tensor const& problem_sizes, torch::Tensor const& a_strides,
-    torch::Tensor const& b_strides, torch::Tensor const& c_strides) {
+    torch::Tensor const& b_strides, torch::Tensor const& c_strides,
+    bool per_act_token, bool per_out_ch) {
   using ElementAB = typename Gemm::ElementAB;
   using ElementD = typename Gemm::ElementD;
 
   int num_experts = static_cast<int>(expert_offsets.size(0));
   int k_size = a_tensors.size(1);
   int n_size = out_tensors.size(1);
 
-  bool per_act_token = a_scales.numel() != 1;
-  bool per_out_ch = b_scales.numel() != num_experts;
-
   auto stream = at::cuda::getCurrentCUDAStream(a_tensors.device().index());
 
   auto options_int =

csrc/quantization/cutlass_w8a8/moe/moe_data.cu

@@ -101,3 +101,35 @@ void get_cutlass_moe_mm_data_caller(
       static_cast<int32_t*>(atomic_buffer.data_ptr()), topk_ids.numel(),
       topk_ids.size(1));
 }
+
+__global__ void compute_pplx_data(
+    int32_t* expert_offsets, int32_t* problem_sizes1, int32_t* problem_sizes2,
+    const int32_t* __restrict__ expert_num_tokens,
+    const int64_t* __restrict__ non_zero_expert_idxs, const int padded_m,
+    const int n, const int k) {
+  int expert_idx_out = threadIdx.x;
+  int expert_idx_in = non_zero_expert_idxs[expert_idx_out];
+  expert_offsets[expert_idx_out] = expert_idx_in * padded_m;
+  problem_sizes1[expert_idx_out * 3] = expert_num_tokens[expert_idx_in];
+  problem_sizes1[expert_idx_out * 3 + 1] = 2 * n;
+  problem_sizes1[expert_idx_out * 3 + 2] = k;
+  problem_sizes2[expert_idx_out * 3] = expert_num_tokens[expert_idx_in];
+  problem_sizes2[expert_idx_out * 3 + 1] = k;
+  problem_sizes2[expert_idx_out * 3 + 2] = n;
+}
+
+void get_cutlass_pplx_moe_mm_data_caller(
+    torch::Tensor& expert_offsets, torch::Tensor& problem_sizes1,
+    torch::Tensor& problem_sizes2, const torch::Tensor& expert_num_tokens,
+    const torch::Tensor& non_zero_expert_idxs, const int64_t num_local_experts,
+    const int64_t padded_m, const int64_t n, const int64_t k) {
+  auto stream = at::cuda::getCurrentCUDAStream(expert_offsets.device().index());
+
+  compute_pplx_data<<<1, num_local_experts, 0, stream>>>(
+      static_cast<int32_t*>(expert_offsets.data_ptr()),
+      static_cast<int32_t*>(problem_sizes1.data_ptr()),
+      static_cast<int32_t*>(problem_sizes2.data_ptr()),
+      static_cast<const int32_t*>(expert_num_tokens.data_ptr()),
+      static_cast<const int64_t*>(non_zero_expert_idxs.data_ptr()), padded_m, n,
+      k);
+}

csrc/quantization/cutlass_w8a8/scaled_mm_entry.cu

@@ -36,7 +36,8 @@ void cutlass_moe_mm_sm90(
     torch::Tensor const& b_tensors, torch::Tensor const& a_scales,
     torch::Tensor const& b_scales, torch::Tensor const& expert_offsets,
     torch::Tensor const& problem_sizes, torch::Tensor const& a_strides,
-    torch::Tensor const& b_strides, torch::Tensor const& c_strides);
+    torch::Tensor const& b_strides, torch::Tensor const& c_strides,
+    bool per_act_token, bool per_out_ch);
 
 #endif
 
@@ -55,6 +56,12 @@ void get_cutlass_moe_mm_data_caller(
     torch::Tensor& problem_sizes1, torch::Tensor& problem_sizes2,
     torch::Tensor& input_permutation, torch::Tensor& output_permutation,
     const int64_t num_experts, const int64_t n, const int64_t k);
+
+void get_cutlass_pplx_moe_mm_data_caller(
+    torch::Tensor& expert_offsets, torch::Tensor& problem_sizes1,
+    torch::Tensor& problem_sizes2, const torch::Tensor& expert_num_tokens,
+    const torch::Tensor& non_zero_expert_idxs, const int64_t num_local_experts,
+    const int64_t padded_m, const int64_t n, const int64_t k);
 #endif
 
 void cutlass_scaled_mm_azp_sm75(torch::Tensor& c, torch::Tensor const& a,
@@ -206,12 +213,13 @@ void cutlass_moe_mm(
     torch::Tensor const& b_tensors, torch::Tensor const& a_scales,
     torch::Tensor const& b_scales, torch::Tensor const& expert_offsets,
     torch::Tensor const& problem_sizes, torch::Tensor const& a_strides,
-    torch::Tensor const& b_strides, torch::Tensor const& c_strides) {
+    torch::Tensor const& b_strides, torch::Tensor const& c_strides,
+    bool per_act_token, bool per_out_ch) {
   int32_t version_num = get_sm_version_num();
 #if defined ENABLE_CUTLASS_MOE_SM90 && ENABLE_CUTLASS_MOE_SM90
   cutlass_moe_mm_sm90(out_tensors, a_tensors, b_tensors, a_scales, b_scales,
                       expert_offsets, problem_sizes, a_strides, b_strides,
-                      c_strides);
+                      c_strides, per_act_token, per_out_ch);
   return;
 #endif
   TORCH_CHECK_NOT_IMPLEMENTED(
@@ -242,6 +250,28 @@ void get_cutlass_moe_mm_data(
       version_num, ". Required capability: 90");
 }
 
+void get_cutlass_pplx_moe_mm_data(
+    torch::Tensor& expert_offsets, torch::Tensor& problem_sizes1,
+    torch::Tensor& problem_sizes2, const torch::Tensor& expert_num_tokens,
+    const torch::Tensor& non_zero_expert_idxs, const int64_t num_local_experts,
+    const int64_t padded_m, const int64_t n, const int64_t k) {
+  // This function currently gets compiled only if we have a valid cutlass moe
+  // mm to run it for.
+  int32_t version_num = get_sm_version_num();
+#if (defined ENABLE_CUTLASS_MOE_SM90 && ENABLE_CUTLASS_MOE_SM90) || \
+    (defined ENABLE_SCALED_MM_SM100 && ENABLE_SCALED_MM_SM90)
+  get_cutlass_pplx_moe_mm_data_caller(
+      expert_offsets, problem_sizes1, problem_sizes2, expert_num_tokens,
+      non_zero_expert_idxs, num_local_experts, padded_m, n, k);
+  return;
+#endif
+  TORCH_CHECK_NOT_IMPLEMENTED(
+      false,
+      "No compiled get_cutlass_pplx_moe_mm_data: no cutlass_scaled_mm kernel "
+      "for CUDA device capability: ",
+      version_num, ". Required capability: 90");
+}
+
 void cutlass_scaled_mm_azp(torch::Tensor& c, torch::Tensor const& a,
                            torch::Tensor const& b,
                            torch::Tensor const& a_scales,