Refactor the Command / DrawList System

Use Command Pool instead of always allocating.
This gives us e big performance diffrence on the 3ds
Fixed IDS of ui7 for now

include/pd/lithium/command.hpp

@@ -34,7 +34,7 @@ class Command {
   Command() = default;
   ~Command() = default;
 
-  PD_UNIQUE(Command);
+  PD_RAW(Command);
 
   Command& AddIdx(const u16& idx) {
     IndexBuffer.push_back(VertexBuffer.size() + idx);
@@ -52,6 +52,16 @@ class Command {
     return *this;
   }
 
+  void Clear() {
+    VertexBuffer.clear();
+    IndexBuffer.clear();
+    Index = 0;
+    Layer = 0;
+    Tex = 0;
+    ScissorOn = false;
+    ScissorRect = ivec4();
+  }
+
   std::vector<Vertex> VertexBuffer;
   std::vector<u16> IndexBuffer;
   ivec4 ScissorRect;
@@ -60,5 +70,71 @@ class Command {
   int Index;
   Texture::Ref Tex;
 };
+
+class CmdPool {
+ public:
+  CmdPool() {}
+  ~CmdPool() {}
+
+  Command::Ref NewCmd() {
+    if (pPoolIdx >= pPool.size()) {
+      Resize(pPool.size() + 128);
+    }
+    Command::Ref nu = pPool[pPoolIdx++];
+    nu->Layer = Layer;
+    nu->Index = pPoolIdx - 1;
+    return nu;
+  }
+
+  void Init(size_t initial_size) { Resize(initial_size); }
+
+  void Deinit() {
+    for (auto it : pPool) {
+      Command::Delete(it);
+    }
+    pPool.clear();
+  }
+
+  void Resize(size_t nulen) {
+    if (nulen <= pPool.size()) {
+      return;  // no idea yet
+    }
+    size_t oldlen = pPool.size();
+    pPool.resize(nulen);
+    for (size_t i = oldlen; i < pPool.size(); i++) {
+      pPool[i] = Command::New();
+    }
+  }
+
+  void Reset() {
+    for (u32 i = 0; i < pPoolIdx; i++) {
+      pPool[i]->Clear();
+    }
+    pPoolIdx = 0;
+  }
+
+  Command::Ref GetCmd(size_t idx) const { return pPool[idx]; }
+  Command::Ref GetCmd(size_t idx) { return pPool[idx]; }
+
+  size_t Size() const { return pPoolIdx; }
+  size_t Cap() const { return pPool.size(); }
+
+  void Merge(CmdPool& p) {
+    if (pPoolIdx + p.Size() > pPool.size()) {
+      Resize(pPoolIdx + p.Size());
+    }
+    for (size_t i = 0; i < p.Size(); i++) {
+      size_t idx = pPoolIdx++;
+      *pPool[idx] = *p.GetCmd(i);
+      pPool[idx]->Index = idx;
+    }
+    p.Reset();
+  }
+
+ private:
+  std::vector<Command::Ref> pPool;
+  u32 pPoolIdx = 0;
+  int Layer = 0;
+};
 }  // namespace Li
 }  // namespace PD