[Metaschedule] Support rocm and spirv

Author: masahi

python/tvm/meta_schedule/tune.py

@@ -412,7 +412,7 @@ def _sch_rules(sch_rules: Optional[FnScheduleRule], target: Target) -> List[Sche
         # pylint: disable=protected-access
         if target.kind.name == "llvm":
             return DefaultLLVM._sch_rules()
-        if target.kind.name == "cuda":
+        if target.kind.name in ["cuda", "rocm", "vulkan"]:
             return DefaultCUDA._sch_rules()
         # pylint: enable=protected-access
         raise ValueError(f"Unsupported target: {target}")
@@ -426,7 +426,7 @@ def _postproc(postproc: Optional[FnPostproc], target: Target) -> List[Postproc]:
         # pylint: disable=protected-access
         if target.kind.name == "llvm":
             return DefaultLLVM._postproc()
-        if target.kind.name == "cuda":
+        if target.kind.name in ["cuda", "rocm", "vulkan"]:
             return DefaultCUDA._postproc()
         # pylint: enable=protected-access
         raise ValueError(f"Unsupported target: {target}")
@@ -445,7 +445,7 @@ def _mutator_probs(
         # pylint: disable=protected-access
         if target.kind.name == "llvm":
             return DefaultLLVM._mutator_probs()
-        if target.kind.name == "cuda":
+        if target.kind.name in ["cuda", "rocm", "vulkan"]:
             return DefaultCUDA._mutator_probs()
         # pylint: enable=protected-access
         raise ValueError(f"Unsupported target: {target}")

src/target/target_kind.cc

@@ -308,7 +308,9 @@ TVM_REGISTER_TARGET_KIND("rocm", kDLROCM)
     .add_attr_option<String>("mtriple")
     .add_attr_option<Bool>("system-lib")
     .add_attr_option<Integer>("max_num_threads", Integer(256))
+    .add_attr_option<Integer>("max_threads_per_block", Integer(256))
     .add_attr_option<Integer>("thread_warp_size", Integer(64))
+    .add_attr_option<Integer>("max_shared_memory_per_block", Integer(64000))
     .set_default_keys({"rocm", "gpu"})
     .set_attrs_preprocessor(UpdateROCmAttrs);
 
@@ -349,6 +351,7 @@ TVM_REGISTER_TARGET_KIND("vulkan", kDLVulkan)
     .add_attr_option<Integer>("supported_subgroup_operations")
     // Physical device limits
     .add_attr_option<Integer>("max_num_threads", Integer(256))
+    .add_attr_option<Integer>("max_threads_per_block", Integer(256))
     .add_attr_option<Integer>("thread_warp_size", Integer(1))
     .add_attr_option<Integer>("max_block_size_x")
     .add_attr_option<Integer>("max_block_size_y")