Add GXSetProjectionFull extension

include/dolphin/gx/GXAurora.h

@@ -23,6 +23,12 @@ extern "C" {
  */
 #define GX_AURORA_LOAD_SCISSOR_RENDER 0x0002
 
+/**
+ * Loads a full 4x4 projection matrix, bypassing GXSetProjection's 6-parameter
+ * hardware encoding. Must be followed by sixteen f32 values in row-major order.
+ */
+#define GX_AURORA_LOAD_PROJECTION_FULL 0x0003
+
 /**
  * Aurora equivalent of CP_REG_ARRAYBASE_ID: sets the base address and size of a vertex array.
  * This command must be followed by a 64-bit memory address, 32-bit size, and 1-byte little-endian flag.
@@ -144,6 +150,11 @@ void GXSetScissorRender(u32 left, u32 top, u32 wd, u32 ht);
 
 void GX2SetPolygonOffset(f32 mFrontOffset, f32 mFrontScale, f32 mBackOffset, f32 mBackScale, f32 mClamp);
 
+/**
+ * Load an arbitrary 4x4 projection matrix, avoiding the 6-parameter hardware encoding.
+ */
+void GXSetProjectionFull(const void* mtx);
+
 /**
  * Create an offscreen framebuffer and switch rendering to it.
  * All subsequent GX rendering will target this framebuffer until GXRestoreFrameBuffer() is called.

lib/dolphin/gx/GXAurora.cpp

@@ -66,6 +66,14 @@ void GXSetScissorRender(u32 left, u32 top, u32 wd, u32 ht) {
   GX_WRITE_U32(ht);
 }
 
+void GXSetProjectionFull(const void* mtx) {
+  const f32* values = reinterpret_cast<const f32*>(mtx);
+  GX_WRITE_AURORA(GX_AURORA_LOAD_PROJECTION_FULL);
+  for (int i = 0; i < 16; ++i) {
+    GX_WRITE_F32(values[i]);
+  }
+}
+
 void GX2SetPolygonOffset(f32 mFrontOffset, f32 mFrontScale, f32 mBackOffset, f32 mBackScale, f32 mClamp) {
   GX_WRITE_AURORA(GX2_SET_POLYGON_OFFSET);
   GX_WRITE_F32(mFrontOffset);

lib/gx/command_processor.cpp

@@ -1741,6 +1741,16 @@ void handle_aurora(const u8* data, u32& pos, u32 size, bool bigEndian) {
     const int32_t height = static_cast<int32_t>(read_u32(data + pos, bigEndian));
     pos += 4;
     set_render_scissor({left, top, width, height});
+  } else if (subCmd == GX_AURORA_LOAD_PROJECTION_FULL) {
+    CHECK(pos + 64 <= size, "GX_AURORA_LOAD_PROJECTION_FULL read overrun");
+    auto& proj = g_gxState.proj;
+    for (int r = 0; r < 4; ++r) {
+      for (int c = 0; c < 4; ++c) {
+        proj[r][c] = read_f32(data + pos, bigEndian);
+        pos += 4;
+      }
+    }
+    g_gxState.stateDirty = true;
   } else if (subCmd >= GX_AURORA_LOAD_ARRAYBASE && subCmd <= (GX_AURORA_LOAD_ARRAYBASE | 0x0f)) {
     CHECK(pos + 13 <= size, "GX_AURORA_LOAD_ARRAYBASE read overrun");
     u32 attrIdx = subCmd - GX_AURORA_LOAD_ARRAYBASE + GX_VA_POS;