1 files changed, 28 insertions, 6 deletions

diff --git a/src/video_core/shader/shader.cpp b/src/video_core/shader/shader.cpp
index 5e8930476..eb1db0778 100644
--- a/src/video_core/shader/shader.cpp
+++ b/src/video_core/shader/shader.cpp
@@ -32,6 +32,12 @@ namespace Shader {
 static std::unordered_map<u64, CompiledShader*> shader_map;
 static JitCompiler jit;
 static CompiledShader* jit_shader;
+
+static void ClearCache() {
+    shader_map.clear();
+    jit.Clear();
+    LOG_INFO(HW_GPU, "Shader JIT cache cleared");
+}
 #endif // ARCHITECTURE_x86_64
 
 void Setup(UnitState<false>& state) {
@@ -45,6 +51,12 @@ void Setup(UnitState<false>& state) {
         if (iter != shader_map.end()) {
             jit_shader = iter->second;
         } else {
+            // Check if remaining JIT code space is enough for at least one more (massive) shader
+            if (jit.GetSpaceLeft() < jit_shader_size) {
+                // If not, clear the cache of all previously compiled shaders
+                ClearCache();
+            }
+
             jit_shader = jit.Compile();
             shader_map.emplace(cache_key, jit_shader);
         }
@@ -54,7 +66,7 @@ void Setup(UnitState<false>& state) {
 
 void Shutdown() {
 #ifdef ARCHITECTURE_x86_64
-    shader_map.clear();
+    ClearCache();
 #endif // ARCHITECTURE_x86_64
 }
 
@@ -109,15 +121,23 @@ OutputVertex Run(UnitState<false>& state, const InputVertex& input, int num_attr
     OutputVertex ret;
     // TODO(neobrain): Under some circumstances, up to 16 attributes may be output. We need to
     // figure out what those circumstances are and enable the remaining outputs then.
-    for (int i = 0; i < 7; ++i) {
-        const auto& output_register_map = g_state.regs.vs_output_attributes[i]; // TODO: Don't hardcode VS here
+    unsigned index = 0;
+    for (unsigned i = 0; i < 7; ++i) {
+
+        if (index >= g_state.regs.vs_output_total)
+            break;
+
+        if ((g_state.regs.vs.output_mask & (1 << i)) == 0)
+            continue;
+
+        const auto& output_register_map = g_state.regs.vs_output_attributes[index]; // TODO: Don't hardcode VS here
 
         u32 semantics[4] = {
             output_register_map.map_x, output_register_map.map_y,
             output_register_map.map_z, output_register_map.map_w
         };
 
-        for (int comp = 0; comp < 4; ++comp) {
+        for (unsigned comp = 0; comp < 4; ++comp) {
             float24* out = ((float24*)&ret) + semantics[comp];
             if (semantics[comp] != Regs::VSOutputAttributes::INVALID) {
                 *out = state.registers.output[i][comp];
@@ -127,15 +147,17 @@ OutputVertex Run(UnitState<false>& state, const InputVertex& input, int num_attr
                 memset(out, 0, sizeof(*out));
             }
         }
+
+        index++;
     }
 
     // The hardware takes the absolute and saturates vertex colors like this, *before* doing interpolation
-    for (int i = 0; i < 4; ++i) {
+    for (unsigned i = 0; i < 4; ++i) {
         ret.color[i] = float24::FromFloat32(
             std::fmin(std::fabs(ret.color[i].ToFloat32()), 1.0f));
     }
 
-    LOG_TRACE(Render_Software, "Output vertex: pos(%.2f, %.2f, %.2f, %.2f), quat(%.2f, %.2f, %.2f, %.2f), "
+    LOG_TRACE(HW_GPU, "Output vertex: pos(%.2f, %.2f, %.2f, %.2f), quat(%.2f, %.2f, %.2f, %.2f), "
         "col(%.2f, %.2f, %.2f, %.2f), tc0(%.2f, %.2f), view(%.2f, %.2f, %.2f)",
         ret.pos.x.ToFloat32(), ret.pos.y.ToFloat32(), ret.pos.z.ToFloat32(), ret.pos.w.ToFloat32(),
         ret.quat.x.ToFloat32(), ret.quat.y.ToFloat32(), ret.quat.z.ToFloat32(), ret.quat.w.ToFloat32(),