devoptab: cache all reads to max read throughput, now as fast as normal sd card reads, including compressed zips (80MiB/s).

sphaira/include/utils/devoptab.hpp

@@ -19,6 +19,4 @@ void UmountNsp(const fs::FsPath& mount);
 Result MountXci(fs::Fs* fs, const fs::FsPath& path, fs::FsPath& out_path);
 void UmountXci(const fs::FsPath& mount);
 
-bool fix_path(const char* str, char* out);
-
 } // namespace sphaira::devoptab

sphaira/include/utils/devoptab_common.hpp

@@ -0,0 +1,34 @@
+#pragma once
+
+#include "yati/source/file.hpp"
+#include <memory>
+
+namespace sphaira::devoptab::common {
+
+// buffers data in 512k chunks to maximise throughput.
+// not suitable if random access >= 512k is common.
+// if that is needed, see the LRU cache varient used for fatfs.
+struct BufferedData final : yati::source::Base {
+    static constexpr inline u64 CHUNK_SIZE = 1024 * 512;
+
+    BufferedData(std::unique_ptr<yati::source::Base>&& _source, u64 _size)
+    : source{std::forward<decltype(_source)>(_source)}
+    , capacity{_size} {
+
+    }
+
+    Result Read(void *buf, s64 off, s64 size);
+    Result Read(void* buf, s64 off, s64 size, u64* bytes_read) override;
+
+private:
+    std::unique_ptr<yati::source::Base> source;
+    const u64 capacity;
+
+    u64 m_off{};
+    u64 m_size{};
+    u8 m_data[CHUNK_SIZE]{};
+};
+
+bool fix_path(const char* str, char* out);
+
+} // namespace sphaira::devoptab::common