gsdx-ogl: implement the wildhack on the GPU

Likely much faster for opengl and much easier to implement

Note: hopefully UserHacks_round_sprite_offset will replace it

plugins/GSdx/GSDeviceOGL.cpp

@@ -659,6 +659,7 @@ GLuint GSDeviceOGL::CompileVS(VSSelector sel)
 		+ format("#define VS_LOGZ %d\n", sel.logz)
 		+ format("#define VS_TME %d\n", sel.tme)
 		+ format("#define VS_FST %d\n", sel.fst)
+		+ format("#define VS_WILDHACK %d\n", sel.wildhack)
 		;
 
 	return m_shader->Compile("tfx.glsl", "vs_main", GL_VERTEX_SHADER, tfx_glsl, macro);
@@ -1063,16 +1064,6 @@ void GSDeviceOGL::IASetVertexBuffer(const void* vertices, size_t count)
 	m_va->UploadVB(vertices, count);
 }
 
-bool GSDeviceOGL::IAMapVertexBuffer(void** vertex, size_t stride, size_t count)
-{
-	return m_va->MapVB(vertex, count);
-}
-
-void GSDeviceOGL::IAUnmapVertexBuffer()
-{
-	m_va->UnmapVB();
-}
-
 void GSDeviceOGL::IASetIndexBuffer(const void* index, size_t count)
 {
 	m_va->UploadIB(index, count);

plugins/GSdx/GSDeviceOGL.h

@@ -273,6 +273,7 @@ class GSDeviceOGL : public GSDevice
 		{
 			struct
 			{
+				uint32 wildhack:1;
 				uint32 bppz:2;
 				uint32 logz:1;
 				// Next param will be handle by subroutine
@@ -283,12 +284,12 @@ class GSDeviceOGL : public GSDevice
 			uint32 key;
 		};
 
-		operator uint32() {return key & 0x3f;}
+		operator uint32() {return key & 0x7f;}
 
 		VSSelector() : key(0) {}
 		VSSelector(uint32 k) : key(k) {}
 
-		static uint32 size() { return 1 << 5; }
+		static uint32 size() { return 1 << 6; }
 	};
 
 	__aligned(struct, 32) PSConstantBuffer
@@ -516,7 +517,7 @@ class GSDeviceOGL : public GSDevice
 		float bf; // blend factor
 	} m_state;
 
-	GLuint m_vs[1<<5];
+	GLuint m_vs[1<<6];
 	GLuint m_gs;
 	GLuint m_ps_ss[1<<3];
 	GSDepthStencilOGL* m_om_dss[1<<6];
@@ -599,8 +600,6 @@ class GSDeviceOGL : public GSDevice
 
 	void IASetPrimitiveTopology(GLenum topology);
 	void IASetVertexBuffer(const void* vertices, size_t count);
-	bool IAMapVertexBuffer(void** vertex, size_t stride, size_t count);
-	void IAUnmapVertexBuffer();
 	void IASetIndexBuffer(const void* index, size_t count);
 
 	void PSSetShaderResource(GLuint sr);

plugins/GSdx/GSRendererOGL.cpp

@@ -122,28 +122,7 @@ void GSRendererOGL::SetupIA()
 	if (!GLLoader::found_geometry_shader)
 		EmulateGS();
 
-	void* ptr = NULL;
+	dev->IASetVertexBuffer(m_vertex.buff, m_vertex.next);
-
-	if(UserHacks_WildHack && !isPackedUV_HackFlag) {
-		// FIXME: why not put it on the Vertex shader
-		if(dev->IAMapVertexBuffer(&ptr, sizeof(GSVertex), m_vertex.next))
-		{
-			GSVector4i::storent(ptr, m_vertex.buff, sizeof(GSVertex) * m_vertex.next);
-
-			GSVertex* RESTRICT d = (GSVertex*)ptr;
-
-			for(unsigned int i = 0; i < m_vertex.next; i++)
-			{
-				if(PRIM->TME && PRIM->FST) d[i].UV &= 0x3FEF3FEF;
-			}
-
-			dev->IAUnmapVertexBuffer();
-		}
-	} else {
-		// By default use the common path (in case it can be made faster)
-		dev->IASetVertexBuffer(m_vertex.buff, m_vertex.next);
-	}
-
 	dev->IASetIndexBuffer(m_index.buff, m_index.tail);
 
 	GLenum t = 0;
@@ -284,6 +263,7 @@ void GSRendererOGL::DrawPrims(GSTexture* rt, GSTexture* ds, GSTextureCache::Sour
 	vs_sel.tme = PRIM->TME;
 	vs_sel.fst = PRIM->FST;
 	vs_sel.logz = m_logz ? 1 : 0;
+	vs_sel.wildhack = (UserHacks_WildHack && !isPackedUV_HackFlag) ? 1 : 0;
 
 	// The real GS appears to do no masking based on the Z buffer format and writing larger Z values
 	// than the buffer supports seems to be an error condition on the real GS, causing it to crash.

plugins/GSdx/GSVertexArrayOGL.h

@@ -328,10 +328,6 @@ public:
 #endif
 	}
 
-	bool MapVB(void **pointer, size_t count) { return m_vb->Map(pointer, count); }
-
-	void UnmapVB() { m_vb->Unmap(); }
-
 	~GSVertexBufferStateOGL()
 	{
 		gl_DeleteVertexArrays(1, &m_va);

plugins/GSdx/res/glsl_source.h

@@ -359,14 +359,14 @@ static const char* merge_glsl =
 	"void ps_main0()\n"
 	"{\n"
 	"    vec4 c = texture(TextureSampler, PSin_t);\n"
-	"	c.a = min(c.a * 2.0, 1.0);\n"
+	"    c.a = min(c.a * 2.0, 1.0);\n"
 	"    SV_Target0 = c;\n"
 	"}\n"
 	"\n"
 	"void ps_main1()\n"
 	"{\n"
 	"    vec4 c = texture(TextureSampler, PSin_t);\n"
-	"	c.a = BGColor.a;\n"
+	"    c.a = BGColor.a;\n"
 	"    SV_Target0 = c;\n"
 	"}\n"
 	"\n"
@@ -458,8 +458,8 @@ static const char* tfx_glsl =
 	"#define FMT_16 2\n"
 	"#define FMT_PAL 4 /* flag bit */\n"
 	"\n"
-	"// APITRACE_DEBUG allows to force pixel output to easily detect \n"
+	"// APITRACE_DEBUG allows to force pixel output to easily detect\n"
-	"// the fragment computed by primitive \n"
+	"// the fragment computed by primitive\n"
 	"#define APITRACE_DEBUG 0\n"
 	"// TEX_COORD_DEBUG output the uv coordinate as color. It is useful\n"
 	"// to detect bad sampling due to upscaling\n"
@@ -582,7 +582,11 @@ static const char* tfx_glsl =
 	"    {\n"
 	"        if(VS_FST != 0)\n"
 	"        {\n"
-	"            VSout_t.xy = vec2(i_uv) * TextureScale;\n"
+	"            if (VS_WILDHACK == 1) {\n"
+	"                VSout_t.xy = vec2(i_uv &  uvec2(0x3FEF, 0x3FEF)) * TextureScale;\n"
+	"            } else {\n"
+	"                VSout_t.xy = vec2(i_uv) * TextureScale;\n"
+	"            }\n"
 	"            VSout_t.w = 1.0f;\n"
 	"        }\n"
 	"        else\n"

plugins/GSdx/res/merge.glsl

@@ -33,14 +33,14 @@ layout(binding = 0) uniform sampler2D TextureSampler;
 void ps_main0()
 {
     vec4 c = texture(TextureSampler, PSin_t);
-	c.a = min(c.a * 2.0, 1.0);
+    c.a = min(c.a * 2.0, 1.0);
     SV_Target0 = c;
 }
 
 void ps_main1()
 {
     vec4 c = texture(TextureSampler, PSin_t);
-	c.a = BGColor.a;
+    c.a = BGColor.a;
     SV_Target0 = c;
 }
 

plugins/GSdx/res/tfx.glsl

@@ -131,7 +131,11 @@ void texture_coord()
     {
         if(VS_FST != 0)
         {
-            VSout_t.xy = vec2(i_uv) * TextureScale;
+            if (VS_WILDHACK == 1) {
+                VSout_t.xy = vec2(i_uv &  uvec2(0x3FEF, 0x3FEF)) * TextureScale;
+            } else {
+                VSout_t.xy = vec2(i_uv) * TextureScale;
+            }
             VSout_t.w = 1.0f;
         }
         else