diff --git a/misc/dreamcast/VertexDraw.S b/misc/dreamcast/VertexDraw.S
index 8248eb075..352e57c52 100644
--- a/misc/dreamcast/VertexDraw.S
+++ b/misc/dreamcast/VertexDraw.S
@@ -523,7 +523,7 @@ _ProcessVertexList:
 	bra     SUBMIT_LOOP
 	add    #-64,r15
 
-! Submits a PowerVR GPU command
+! Submits a PVR2 TA GPU command
 DO_CMD:
 	PushVertex REG_V0
 	bra     NEXT_ITER
@@ -543,7 +543,6 @@ SUBMIT_LOOP:
 	bf.s    DO_CMD    ! if (!T) goto DO_CMD
 ! PVR_CMD_VERTEX_EOL case
 	extu.b  r0,r1     ! EX, MASK = FLAGS & 0xFF (branch delay slot)
-	mov.l   r11,@r13  ! LS, CUR->flags = PVR_CMD_VERTEX_EOL
 
 ! Prepare and then jump to quad drawing function, based on quad clipflags
 	mova    .CASES,r0   ! LS, r0 = CASES
diff --git a/src/Graphics_Dreamcast.c b/src/Graphics_Dreamcast.c
index 9fc3d6054..d232a7e2c 100644
--- a/src/Graphics_Dreamcast.c
+++ b/src/Graphics_Dreamcast.c
@@ -60,7 +60,8 @@ static void InitGPU(void) {
 
 static void InitGLState(void) {
 	pvr_set_zclip(0.0f);
-	PVR_SET(PT_ALPHA_REF, 127); // define missing from KOS
+	PVR_SET(PT_ALPHA_REF, 127); // define missing from KOS    
+	//PVR_SET(PVR_SPANSORT_CFG, 0x0);
 
 	ALPHA_TEST_ENABLED = false;
 	CULLING_ENABLED    = false;
@@ -70,7 +71,7 @@ static void InitGLState(void) {
 	TEXTURES_ENABLED   = false;
 	FOG_ENABLED        = false;
 	
-	stateDirty        = true;
+	stateDirty       = true;
     listOP.list_type = PVR_LIST_OP_POLY;
     listPT.list_type = PVR_LIST_PT_POLY;
     listTR.list_type = PVR_LIST_TR_POLY;
@@ -683,7 +684,7 @@ static void SubmitList(AlignedVector* cmds) {
 	pvr_list_begin(cmds->list_type);
 	{
 		pvr_dr_init(&dr_state);
-		SceneListSubmit(cmds->data, cmds->size);
+		SceneListSubmit((Vertex*)cmds->data, cmds->size);
 		sq_wait();
 	}
 	pvr_list_finish();
diff --git a/third_party/gldc/src/gldc.h b/third_party/gldc/src/gldc.h
index 0dcbee0d4..9702e43f4 100644
--- a/third_party/gldc/src/gldc.h
+++ b/third_party/gldc/src/gldc.h
@@ -66,9 +66,6 @@ typedef struct {
     float w;
 } __attribute__ ((aligned (32))) Vertex;
 
-
-#define GL_FORCE_INLINE static __attribute__((always_inline)) inline
-
 typedef struct {
     uint32_t color; /* This is the PVR texture format */
     void     *data;
diff --git a/third_party/gldc/src/sh4.c b/third_party/gldc/src/sh4.c
index e2aeb4c66..5c1a17f0c 100644
--- a/third_party/gldc/src/sh4.c
+++ b/third_party/gldc/src/sh4.c
@@ -5,9 +5,10 @@
 
 #define PREFETCH(addr) __builtin_prefetch((addr))
 static volatile uint32_t* sq;
+#define GL_FORCE_INLINE __attribute__((always_inline)) inline
 
 // calculates 1/sqrt(x)
-GL_FORCE_INLINE float sh4_fsrra(float x) {
+static GL_FORCE_INLINE float sh4_fsrra(float x) {
   asm volatile ("fsrra %[value]\n"
   : [value] "+f" (x) // outputs (r/w to FPU register)
   : // no inputs
@@ -16,11 +17,11 @@ GL_FORCE_INLINE float sh4_fsrra(float x) {
   return x;
 }
 
-GL_FORCE_INLINE float _glFastInvert(float x) {
+static GL_FORCE_INLINE float _glFastInvert(float x) {
     return sh4_fsrra(x * x);
 }
 
-GL_FORCE_INLINE void _glPerspectiveDivideVertex(Vertex* vertex) {
+static GL_FORCE_INLINE void _glPerspectiveDivideVertex(Vertex* vertex) {
     const float f = _glFastInvert(vertex->w);
 
     /* Convert to NDC (viewport already applied) */
@@ -45,7 +46,6 @@ static inline void _glPushHeaderOrVertex(Vertex* v)  {
 
 extern void ClipEdge(const Vertex* const v1, const Vertex* const v2, Vertex* vout);
 
-#define SPAN_SORT_CFG 0x005F8030
 #define V0_VIS (1 << 0)
 #define V1_VIS (1 << 1)
 #define V2_VIS (1 << 2)
@@ -371,11 +371,9 @@ static void SubmitClipped(Vertex* v0, Vertex* v1, Vertex* v2, Vertex* v3, uint8_
 }
 
 extern void ProcessVertexList(Vertex* v3, int n, void* sq_addr);
-void SceneListSubmit(Vertex* v3, int n) {
-    PVR_SET(SPAN_SORT_CFG, 0x0);
 
+void SceneListSubmit(Vertex* v3, int n) {
 	sq = (uint32_t*)MEM_AREA_SQ_BASE;
-	uint8_t visible_mask = 0;
 
     for(int i = 0; i < n; ++i, ++v3) 
 	{
@@ -395,8 +393,7 @@ void SceneListSubmit(Vertex* v3, int n) {
         Vertex* const v1 = v3 - 2;
         Vertex* const v2 = v3 - 1;
 
-        visible_mask = v3->flags & 0xFF;
-        v3->flags &= ~0xFF;
+        uint8_t visible_mask = v3->flags & 0xFF;
         
         // Stats gathering found that when testing a 64x64x64 sized world, at most
         //   ~400-500 triangles needed clipping
diff --git a/third_party/gldc/src/state.c b/third_party/gldc/src/state.c
index d587a0c91..21fb94f87 100644
--- a/third_party/gldc/src/state.c
+++ b/third_party/gldc/src/state.c
@@ -88,10 +88,10 @@ void apply_poly_header(pvr_poly_hdr_t* dst, int list_type) {
     dst->mode1 |= (depth_write << PVR_TA_PM1_DEPTHWRITE_SHIFT) & PVR_TA_PM1_DEPTHWRITE_MASK;
     dst->mode1 |= (txr_enable  << PVR_TA_PM1_TXRENABLE_SHIFT)  & PVR_TA_PM1_TXRENABLE_MASK;
 
-    dst->mode2  = (blend_src       << PVR_TA_PM2_SRCBLEND_SHIFT) & PVR_TA_PM2_SRCBLEND_MASK;
-    dst->mode2 |= (blend_dst       << PVR_TA_PM2_DSTBLEND_SHIFT) & PVR_TA_PM2_DSTBLEND_MASK;
-    dst->mode2 |= (gen_fog_type    << PVR_TA_PM2_FOG_SHIFT)      & PVR_TA_PM2_FOG_MASK;
-    dst->mode2 |= (gen_alpha       << PVR_TA_PM2_ALPHA_SHIFT)    & PVR_TA_PM2_ALPHA_MASK;
+    dst->mode2  = (blend_src    << PVR_TA_PM2_SRCBLEND_SHIFT) & PVR_TA_PM2_SRCBLEND_MASK;
+    dst->mode2 |= (blend_dst    << PVR_TA_PM2_DSTBLEND_SHIFT) & PVR_TA_PM2_DSTBLEND_MASK;
+    dst->mode2 |= (gen_fog_type << PVR_TA_PM2_FOG_SHIFT)      & PVR_TA_PM2_FOG_MASK;
+    dst->mode2 |= (gen_alpha    << PVR_TA_PM2_ALPHA_SHIFT)    & PVR_TA_PM2_ALPHA_MASK;
 
     if (txr_enable == PVR_TEXTURE_DISABLE) {
         dst->mode3 = 0;