Add CommandBuffer abstraction with backend-specific downcasting

Command buffer creation and management was previously spread across
each backend's executeProgram() with no shared interface, making it
impossible to manage command buffers from backend-agnostic code. This
introduces a CommandBuffer base class on Device so that higher-level
code can create and pass around command buffers without knowing the
backend. Per-object allocator/pool ownership also prepares for future
async execution where multiple command buffers may be in-flight with
independent lifetimes.

- DX: DXCommandBuffer owns Allocator, CmdList, Fence, Event
- VK: VKCommandBuffer owns CmdPool, CmdBuffer; each submission
  creates a new CommandBuffer for independent lifetime management
- MTL: MTLCommandBuffer wraps MTL::CommandBuffer

Device::createCommandBuffer() returns Expected<unique_ptr<CommandBuffer>>
with a default "not supported" implementation.

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>

Author: MarijnS95

include/API/CommandBuffer.h

@@ -0,0 +1,44 @@
+//===- CommandBuffer.h - Offload Command Buffer API -----------------------===//
+//
+// Part of the LLVM Project, under the Apache License v2.0 with LLVM Exceptions.
+// See https://llvm.org/LICENSE.txt for license information.
+// SPDX-License-Identifier: Apache-2.0 WITH LLVM-exception
+//
+//===----------------------------------------------------------------------===//
+//
+//
+//===----------------------------------------------------------------------===//
+
+#ifndef OFFLOADTEST_API_COMMANDBUFFER_H
+#define OFFLOADTEST_API_COMMANDBUFFER_H
+
+#include "API/API.h"
+
+#include <cassert>
+
+namespace offloadtest {
+
+class CommandBuffer {
+  GPUAPI API;
+
+public:
+  explicit CommandBuffer(GPUAPI API) : API(API) {}
+  virtual ~CommandBuffer() = default;
+  CommandBuffer(const CommandBuffer &) = delete;
+  CommandBuffer &operator=(const CommandBuffer &) = delete;
+
+  GPUAPI getAPI() const { return API; }
+
+  template <typename T> T &as() {
+    assert(API == T::BackendAPI && "CommandBuffer backend mismatch");
+    return static_cast<T &>(*this);
+  }
+  template <typename T> const T &as() const {
+    assert(API == T::BackendAPI && "CommandBuffer backend mismatch");
+    return static_cast<const T &>(*this);
+  }
+};
+
+} // namespace offloadtest
+
+#endif // OFFLOADTEST_API_COMMANDBUFFER_H

include/API/Device.h

@@ -16,8 +16,10 @@
 
 #include "API/API.h"
 #include "API/Capabilities.h"
+#include "API/CommandBuffer.h"
 #include "llvm/ADT/StringRef.h"
 #include "llvm/ADT/iterator_range.h"
+#include "llvm/Support/Error.h"
 
 #include <memory>
 #include <string>
@@ -99,6 +101,12 @@ class Device {
                size_t SizeInBytes) = 0;
   virtual void printExtra(llvm::raw_ostream &OS) {}
 
+  virtual llvm::Expected<std::unique_ptr<CommandBuffer>> createCommandBuffer() {
+    return llvm::createStringError(
+        std::errc::not_supported,
+        "createCommandBuffer not implemented for this backend");
+  }
+
   virtual ~Device() = 0;
 
   llvm::StringRef getDescription() const { return Description; }