[TMA] Bugfix when a shared buffer is both issued with tma store and tma load (#857)

- Updated `init_desc_arg_map` to use `Var` as the key instead of `String` in `lower_hopper_intrin.cc`.
- Enhanced `func_call_args` method in `TLCUDASourceWrapper` to accept additional parameters for better argument mapping.
- Added assertions to ensure consistency between function parameters and arguments during kernel launches.
- Modified `generate_tma_descriptor_args` to utilize a mapping of variable names for TMA descriptor initialization.

src/transform/lower_hopper_intrin.cc

@@ -25,7 +25,7 @@ public:
     PrimFuncNode *fptr = f.CopyOnWrite();
     LowerHopperIntrin substituter(disable_shuffle_elect);
     fptr->body = substituter.VisitStmt(f->body);
-    Map<String, Array<PrimExpr>> init_desc_arg_map;
+    Map<Var, Array<PrimExpr>> init_desc_arg_map;
     for (const auto &[call, var] : substituter.desc_map_) {
       // Should allocate 128 bytes for TensorMap on stack
       Call alloc_desc = Call(DataType::Handle(), builtin::tvm_stack_alloca(),
@@ -46,7 +46,7 @@ public:
           Call(DataType::Handle(), builtin::tvm_call_packed(), init_desc_args);
       fptr->body =
           LetStmt(var, alloc_desc, SeqStmt({Evaluate(init_desc), fptr->body}));
-      init_desc_arg_map.Set(var->name_hint, init_desc_args);
+      init_desc_arg_map.Set(var, init_desc_args);
     }
     f = WithAttr(std::move(f), "tma_descriptor_args", init_desc_arg_map);
     return f;

tilelang/jit/adapter/wrapper.py

@@ -8,6 +8,7 @@ from .utils import (match_declare_kernel, match_declare_kernel_cpu, is_cuda_targ
 import re
 import logging
 import textwrap
+from tvm.tir.stmt_functor import post_order_visit
 
 PREDEF_ATTRIBUTE_SET_DYNAMIC_MEMORY = """
     cudaError_t result_{0} = cudaFuncSetAttribute({0}, cudaFuncAttributeMaxDynamicSharedMemorySize, {1});
@@ -260,7 +261,11 @@ class TLCUDASourceWrapper(object):
         # Format the function arguments for declaration
         def_args = ", ".join([f"{arg['type']} {arg['name']}" for arg in function_args])
 
-        def func_call_args(s, function_args, desc_name_map: Optional[Dict[str, str]] = None):
+        def func_call_args(s,
+                           function_args,
+                           function_params,
+                           desc_name_map: Optional[Dict[str, str]] = None,
+                           desc_name_var_map: Optional[Dict[str, tvm.tir.Var]] = None):
             # Extract the function call arguments matching the function definition
             def maybe_desc(name: str, matches: List[str], i: int):
                 match = matches[i]
@@ -280,8 +285,15 @@ class TLCUDASourceWrapper(object):
             call_args = []
             for i, match in enumerate(matches):
                 for arg in function_args:
-                    if arg["name"] == match or maybe_desc(arg["name"], matches, i):
+                    if arg["name"] == match:
                         call_args.append(match)
+                    elif maybe_desc(arg["name"], matches, i):
+                        call_args.append(match)
+                        assert len(call_args) <= len(
+                            function_params
+                        ), f"Function {function_name} has {len(function_params)} parameters, but {len(call_args)} arguments"
+                        desc_name_var_map[match] = function_params[len(call_args) - 1]
+
             return call_args
 
         has_l2_persistent_map = False
@@ -294,10 +306,12 @@ class TLCUDASourceWrapper(object):
         if has_l2_persistent_map:
             kernel_launch_code += L2_PERSISTENT_MAP_CREATE_HANDLE
         desc_name_map: Dict[str, str] = {}
+        desc_name_var_map: Dict[str, tvm.tir.Var] = {}
         for function_name, function_info in function_informations.items():
             block_info = function_info["block_info"]
             grid_info = function_info["grid_info"]
             dynamic_smem_buf = function_info["dynamic_smem_buf"]
+            function_params = function_info["function_params"]
 
             # Find the location of the global kernel function in the code
             index = match_declare_kernel(code, function_name + "(")
@@ -321,7 +335,11 @@ class TLCUDASourceWrapper(object):
             kernel_launch_code += init_l2_persistent_map
 
             if self.use_cooperative_groups[function_name]:
-                args_list = func_call_args(declaration, function_args, desc_name_map)
+                args_list = func_call_args(declaration, function_args, function_params,
+                                           desc_name_map, desc_name_var_map)
+                assert len(function_params) == len(
+                    args_list
+                ), f"Function {function_name} has {len(function_params)} parameters, but {len(args_list)} arguments"
                 args_array = [f"(void*)&{arg}" for arg in args_list]
                 call_args = f"\tvoid* {function_name}_args[] = {{{', '.join(args_array)}}};\n"
                 kernel_launch_code += call_args
@@ -329,14 +347,20 @@ class TLCUDASourceWrapper(object):
                 kernel_launch_code += "\tTILELANG_CHECK(cudaLaunchCooperativeKernel((void*){}, {}, {}, {}, {}, stream));\n".format(
                     function_name, grid_str, block_str, function_name + "_args", smem_str)
             else:
-                call_args = ", ".join(func_call_args(declaration, function_args, desc_name_map))
+                args_list = func_call_args(declaration, function_args, function_params,
+                                           desc_name_map, desc_name_var_map)
+                assert len(function_params) == len(
+                    args_list
+                ), f"Function {function_name} has {len(function_params)} parameters, but {len(args_list)} arguments"
+                call_args = ", ".join(args_list)
                 kernel_launch_code += "\t{}<<<{}, {}, {}, stream>>>({});\n".format(
                     function_name, grid_str, block_str, smem_str, call_args)
                 kernel_launch_code += "\tTILELANG_CHECK_LAST_ERROR(\"{}\");\n".format(function_name)
             if has_l2_persistent_map:
                 kernel_launch_code += L2_PERSISTENT_MAP_RESET_HANDLE
 
-        init_tma_descriptor_args = self.generate_tma_descriptor_args(desc_name_map)
+        init_tma_descriptor_args = self.generate_tma_descriptor_args(desc_name_map,
+                                                                     desc_name_var_map)
         kernel_launch_code = init_tma_descriptor_args + kernel_launch_code
 
         # Wrap the kernel dispatch logic in an external C function
@@ -362,15 +386,17 @@ class TLCUDASourceWrapper(object):
 
         return init_l2_persistent_map
 
-    def generate_tma_descriptor_args(self, desc_name_map: Dict[str, str]) -> str:
+    def generate_tma_descriptor_args(self, desc_name_map: Dict[str, str],
+                                     desc_name_var_map: Dict[str, tvm.tir.Var]) -> str:
         tma_descripter_init = ""
         if self.tma_descriptor_args is None:
             return tma_descripter_init
+        for handle_name, _ in desc_name_map.items():
+            assert handle_name in desc_name_var_map, f"Handle name {handle_name} not found in desc_name_var_map"
+            desc_var = desc_name_var_map[handle_name]
 
-        for handle_name, name in desc_name_map.items():
-            desc_name = name + "_desc"
-            assert desc_name in self.tma_descriptor_args, f"TMA descriptor {desc_name} not found in {self.tma_descriptor_args}"
-            args = self.tma_descriptor_args[desc_name]
+            assert desc_var in self.tma_descriptor_args, f"TMA descriptor {desc_var} not found in {self.tma_descriptor_args}"
+            args = self.tma_descriptor_args[desc_var]
             # Skip __tvm_tensormap_create_tiled
             if len(args) < 3:
                 raise ValueError(
@@ -536,12 +562,35 @@ class TLCUDASourceWrapper(object):
             # Do not update function with dispatch host function
             if (function_name not in self.block_info) or (function_name not in self.grid_info):
                 continue
+            assert function_name in self.device_mod, f"Function {function_name} not found in device module"
+            device_func = self.device_mod[function_name]
+            kernel_params_cnt = len(device_func.params)
+            function_params: List[str] = None
+
+            def visitor(node, fn=function_name, param_cnt=kernel_params_cnt):
+                nonlocal function_params
+                if isinstance(node, tvm.tir.Call):
+                    if not (hasattr(node, "op") and
+                            node.op == tvm.ir.Op.get("tir.tvm_call_packed")):
+                        return
+                    args = node.args
+                    if not args or args[0] != fn:
+                        return
+                    if len(args) < 1 + param_cnt:
+                        raise AssertionError(
+                            "tvm_call_packed should have at least 1 argument and match device function parameters"
+                        )
+                    function_params = args[1:1 + param_cnt]
+
+            post_order_visit(self.host_func.body, visitor)
+            assert function_params is not None, "function_params should not be None"
 
             function_informations[function_name] = {
                 "function_name": function_name,
                 "block_info": self.block_info[function_name],
                 "grid_info": self.grid_info[function_name],
                 "dynamic_smem_buf": self.dynamic_smem_buf[function_name],
+                "function_params": function_params,
             }
 
         # Create the host function wrapper for the CUDA kernel
@@ -579,6 +628,19 @@ class TLCUDASourceWrapper(object):
                     return function
             raise ValueError("Cannot find primary function in the module.")
 
+    @property
+    def host_func(self):
+        if len(self.host_mod.get_global_vars()) == 1:
+            return self.host_mod[self.host_mod.get_global_vars()[0]]
+        elif "main" in self.host_mod:
+            return self.host_mod["main"]
+        else:
+            for _, function in self.host_mod.functions.items():
+                attr = function.attrs
+                if "tir.is_global_func" in attr and attr["tir.is_global_func"]:
+                    return function
+            raise ValueError("Cannot find primary function in the module.")
+
 
 class TLNVRTCSourceWrapper(TLCUDASourceWrapper):
     """
@@ -636,7 +698,6 @@ class TLNVRTCSourceWrapper(TLCUDASourceWrapper):
                 function_args.append({"name": dyn_sym, "type": "ctypes.c_int"})
 
         function_args.append(self.get_stream_type())
-
         # Format the function arguments for declaration
         def_args = ", ".join([f"{arg['name']}" for arg in function_args])