wined3d: merge vertex & pixel GenerateProgramArbHW functions into baseshader.c

Jason Green jave27 at gmail.com
Fri May 5 00:03:30 CDT 2006


This patch "completes" the merge between vertex and pixel shaders for
generating the shader string for ARB_vertex_program and
fragment_program.  I ran numerous apps which use shaders and compared
their debug output for the d3d_shader channel.  Every shader's trace
was identical except for the function names.  Now, the ARB shader
traces have identical formats between pixel & vertex shaders as well. 
Plus, this will simplify the upcoming implementation of GLSL shaders. 
Once again, this patch should be a no-op.

This applies against Mike's latest (and probably last) git tree.
-------------- next part --------------
diff --git a/dlls/wined3d/baseshader.c b/dlls/wined3d/baseshader.c
index 6aed71b..50c1542 100644
--- a/dlls/wined3d/baseshader.c
+++ b/dlls/wined3d/baseshader.c
@@ -27,6 +27,8 @@
 
 WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader);
 
+#define GLNAME_REQUIRE_GLSL  ((const char *)1)
+
 int shader_addline(
     SHADER_BUFFER* buffer,  
     const char *format, ...) {
@@ -84,12 +86,13 @@ const SHADER_OPCODE* shader_get_opcode(
  *
  * Also note that this does not count the loop register
  * as an address register. */   
-
 void shader_get_registers_used(
     IWineD3DBaseShader *iface,
-    CONST DWORD* pToken, 
-    DWORD* tempsUsed, 
-    DWORD* texUsed) {
+    CONST DWORD* pToken) {
+
+    IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
+    DWORD* tempsUsed = &This->baseShader.temps_used;
+    DWORD* texUsed = &This->baseShader.textures_used;
 
     if (pToken == NULL)
         return;
@@ -225,4 +228,150 @@ void shader_program_dump_decl_usage(
     }
 }
 
+/** Shared code in order to generate the bulk of the shader string.
+    Use the shader_header_fct & shader_footer_fct to add strings
+    that are specific to pixel or vertex functions
+    NOTE: A description of how to parse tokens can be found at http://msdn.microsoft.com/library/default.asp?url=/library/en-us/graphics/hh/graphics/usermodedisplaydriver_shader_cc8e4e05-f5c3-4ec0-8853-8ce07c1551b2.xml.asp */
+void generate_base_shader(
+    IWineD3DBaseShader *iface,
+    SHADER_BUFFER* buffer,
+    SHADER_SPECIFIC_FCT shader_header_fct,
+    SHADER_SPECIFIC_FCT shader_footer_fct,
+    CONST DWORD* pFunction) {
+
+    IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
+    const DWORD *pToken = pFunction;
+    const SHADER_OPCODE *curOpcode = NULL;
+    DWORD i;
+
+    /* Initialize current parsing state */
+    This->baseShader.parse_state.current_row = 0;
+
+#if 0 /* FIXME: Use the buffer that is held by the device, this is ok since fixups 
+        will be skipped for software shaders
+        it also requires entering a critical section but cuts down the runtime footprint of 
+        wined3d and any memory fragmentation that may occur... */
+    if (This->device->fixupVertexBufferSize < SHADER_PGMSIZE) {
+        HeapFree(GetProcessHeap(), 0, This->fixupVertexBuffer);
+        This->fixupVertexBuffer = HeapAlloc(GetProcessHeap() , 0, SHADER_PGMSIZE);
+        This->fixupVertexBufferSize = SHADER_PGMSIZE;
+        This->fixupVertexBuffer[0] = 0;
+    }
+    buffer.buffer = This->device->fixupVertexBuffer;
+#else
+    buffer->buffer = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, SHADER_PGMSIZE);
+#endif
+    buffer->bsize = 0;
+    buffer->lineNo = 0;
+
+    /* First pass: figure out which temporary and texture registers are used */
+    shader_get_registers_used((IWineD3DBaseShader*)This, pToken);
+    TRACE("Texture/Address registers used: %#lx, Temp registers used %#lx\n", 
+          This->baseShader.textures_used, This->baseShader.temps_used);
+
+    /* TODO: check register usage against GL/Directx limits, and fail if they're exceeded
+        nUseAddressRegister < = GL_MAX_PROGRAM_ADDRESS_REGISTERS_AR
+        nUseTempRegister    <=  GL_MAX_PROGRAM_LOCAL_PARAMETERS_ARB
+    */
+
+    /* Add pixel/vertexshader specific header to the string */
+    shader_header_fct((IWineD3DBaseShader*)This, buffer);
+
+    /* Second pass, process opcodes */
+    if (NULL != pToken) {
+        while (D3DPS_END() != *pToken) {
+
+#if 0 
+/* For pixel and vertex shader versions 2_0 and later, bits 24 through 27 specify the size in WORDs of the instruction */            if (version >= 2) {
+                instructionSize = pToken & SIZEBITS >> 27;
+            }
+#endif
+
+            /* Skip version token */
+            if (shader_is_version_token(*pToken)) {
+                ++pToken;
+                continue;
+            }
+
+            /* Skip comment tokens */
+            if (shader_is_comment_token(*pToken)) {
+                DWORD comment_len = (*pToken & D3DSI_COMMENTSIZE_MASK) >> D3DSI_COMMENTSIZE_SHIFT;
+                ++pToken;
+                TRACE("#%s\n", (char*)pToken);
+                pToken += comment_len;
+                continue;
+            }
+
+            /* Read opcode */
+            curOpcode = shader_get_opcode((IWineD3DBaseShader*) This, *pToken);
+            ++pToken;
+
+            /* Unknown opcode and its parameters */
+            if (NULL == curOpcode) {
+                while (*pToken & 0x80000000) { /* TODO: Think of a sensible name for 0x80000000 */
+                    FIXME("unrecognized opcode: %08lx\n", *pToken);
+                    ++pToken;
+                }
+
+            /* Unhandled opcode */
+            } else if (GLNAME_REQUIRE_GLSL == curOpcode->glname) {
+
+                FIXME("Token %s requires greater functionality than "
+                    "Vertex or Fragment_Program_ARB supports\n", curOpcode->name);
+                pToken += curOpcode->num_params;
+
+            /* If a generator function is set, use it */
+            } else if (curOpcode->hw_fct != NULL) {
+
+                SHADER_OPCODE_ARG hw_arg;
+
+                hw_arg.shader = (IWineD3DBaseShader*) This;
+                hw_arg.opcode = curOpcode;
+                hw_arg.buffer = buffer;
+                if (curOpcode->num_params > 0) {
+                    hw_arg.dst = *pToken;
+
+                    /* FIXME: this does not account for relative address tokens */
+                    for (i = 1; i < curOpcode->num_params; i++)
+                       hw_arg.src[i-1] = *(pToken + i);
+                }
+
+                curOpcode->hw_fct(&hw_arg);
+                pToken += curOpcode->num_params;
+
+            } else {
+
+                TRACE("Found opcode D3D:%s GL:%s, PARAMS:%d, \n",
+                curOpcode->name, curOpcode->glname, curOpcode->num_params);
+
+                /* Build opcode for GL vertex_program */
+                switch (curOpcode->opcode) {
+                case D3DSIO_NOP:
+                    break;
+
+                default:
+                    FIXME("Can't handle opcode %s in hwShader\n", curOpcode->name);
+                    pToken += curOpcode->num_params;
+                }
+            }
+        }
+        /* TODO: What about result.depth? */
+
+        /* Add pixel/vertexshader specific footer to the string */
+        shader_footer_fct((IWineD3DBaseShader*)This, buffer);
+
+        shader_addline(buffer, "END\n");
+    }
+
+    /* finally null terminate the buffer */
+    buffer->buffer[buffer->bsize] = 0;
+    /*  Create the hw shader */
+
+    /* The program string sometimes gets too long for a normal TRACE */
+    TRACE("Generated program:\n");
+    if (TRACE_ON(d3d_shader)) {
+        fprintf(stderr, "%s\n", buffer->buffer);
+    }
+}
+
 /* TODO: Move other shared code here */
diff --git a/dlls/wined3d/pixelshader.c b/dlls/wined3d/pixelshader.c
index e6e5131..d0fd5a4 100644
--- a/dlls/wined3d/pixelshader.c
+++ b/dlls/wined3d/pixelshader.c
@@ -1290,170 +1290,71 @@ void pshader_hw_texm3x3spec(SHADER_OPCOD
     current_state.current_row = 0;
 }
 
-                    
-/* NOTE: A description of how to parse tokens can be found at http://msdn.microsoft.com/library/default.asp?url=/library/en-us/graphics/hh/graphics/usermodedisplaydriver_shader_cc8e4e05-f5c3-4ec0-8853-8ce07c1551b2.xml.asp */
-inline static VOID IWineD3DPixelShaderImpl_GenerateProgramArbHW(IWineD3DPixelShader *iface, CONST DWORD *pFunction) {
-    IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *)iface;
-    const DWORD *pToken = pFunction;
-    const SHADER_OPCODE *curOpcode = NULL;
-    DWORD i;
-    SHADER_BUFFER buffer;
-
-    /* Keep bitmaps of used temporary and texture registers */
-    DWORD tempsUsed, texUsed;
-
-    /* Initialize current parsing state */
-    This->baseShader.parse_state.current_row = 0;
-
-#if 0 /* FIXME: Use the buffer that is held by the device, this is ok since fixups will be skipped for software shaders
-        it also requires entering a critical section but cuts down the runtime footprint of wined3d and any memory fragmentation that may occur... */
-    if (This->device->fixupVertexBufferSize < SHADER_PGMSIZE) {
-        HeapFree(GetProcessHeap(), 0, This->fixupVertexBuffer);
-        This->fixupVertexBuffer = HeapAlloc(GetProcessHeap() , 0, SHADER_PGMSIZE);
-        This->fixupVertexBufferSize = SHADER_PGMSIZE;
-        This->fixupVertexBuffer[0] = 0;
-    }
-    buffer.buffer = This->device->fixupVertexBuffer;
-#else
-    buffer.buffer = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, SHADER_PGMSIZE);
-#endif
-    buffer.bsize = 0;
-    buffer.lineNo = 0;
-
-    /* FIXME: if jumps are used, use GLSL, else use ARB_fragment_program */
-    shader_addline(&buffer, "!!ARBfp1.0\n");
-
-    /* TODO: Think about using a first pass to work out what's required for the second pass. */
-    for(i = 0; i < WINED3D_PSHADER_MAX_CONSTANTS; i++)
-        This->constants[i] = 0;
-
-    /* First pass: figure out which temporary and texture registers are used */
-    shader_get_registers_used((IWineD3DBaseShader*) This, pToken, &tempsUsed, &texUsed);
-    TRACE("Texture registers used: %#lx, Temp registers used %#lx\n", texUsed, tempsUsed);
-
-    /* TODO: check register usage against GL/Directx limits, and fail if they're exceeded */
+/** Pixel Shader specific things that need to go into the header of the
+    ARB_FRAGMENT_PROGRAM string */
+void pixel_shader_header_arb(IWineD3DBaseShader* iface, SHADER_BUFFER* buffer) {
+
+    IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) iface;
+    DWORD tempsUsed = This->baseShader.temps_used;
+    DWORD texUsed = This->baseShader.textures_used;
+    int i;
 
+    shader_addline(buffer, "!!ARBfp1.0\n");
     /* Pre-declare registers */
     for(i = 0; i < This->baseShader.limits.texture; i++) {
         if (texUsed & (1 << i))
-            shader_addline(&buffer,"TEMP T%lu;\n", i);
+            shader_addline(buffer,"TEMP T%lu;\n", i);
     }
 
     for(i = 0; i < This->baseShader.limits.temporary; i++) {
         if (tempsUsed & (1 << i))
-             shader_addline(&buffer, "TEMP R%lu;\n", i);
+             shader_addline(buffer, "TEMP R%lu;\n", i);
     }
 
     /* Necessary for internal operations */
-    shader_addline(&buffer, "TEMP TMP;\n");
-    shader_addline(&buffer, "TEMP TMP2;\n");
-    shader_addline(&buffer, "TEMP TA;\n");
-    shader_addline(&buffer, "TEMP TB;\n");
-    shader_addline(&buffer, "TEMP TC;\n");
-    shader_addline(&buffer, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
-    shader_addline(&buffer, "PARAM coefmul = { 2, 4, 8, 16 };\n");
-    shader_addline(&buffer, "PARAM one = { 1.0, 1.0, 1.0, 1.0 };\n");
+    shader_addline(buffer, "TEMP TMP;\n");
+    shader_addline(buffer, "TEMP TMP2;\n");
+    shader_addline(buffer, "TEMP TA;\n");
+    shader_addline(buffer, "TEMP TB;\n");
+    shader_addline(buffer, "TEMP TC;\n");
+    shader_addline(buffer, "PARAM coefdiv = { 0.5, 0.25, 0.125, 0.0625 };\n");
+    shader_addline(buffer, "PARAM coefmul = { 2, 4, 8, 16 };\n");
+    shader_addline(buffer, "PARAM one = { 1.0, 1.0, 1.0, 1.0 };\n");
 
     /* Texture coordinate registers must be pre-loaded */
     for (i = 0; i < This->baseShader.limits.texture; i++) {
        if (texUsed & (1 << i))
-          shader_addline(&buffer, "MOV T%lu, fragment.texcoord[%lu];\n", i, i);
+          shader_addline(buffer, "MOV T%lu, fragment.texcoord[%lu];\n", i, i);
     }
+}
 
-    /* Second pass, process opcodes */
-    if (NULL != pToken) {
-        while (D3DPS_END() != *pToken) {
-#if 0 /* For pixel and vertex shader versions 2_0 and later, bits 24 through 27 specify the size in DWORDs of the instruction */
-            if (version >= 2) {
-                instructionSize = pToken & SIZEBITS >> 27;
-            }
-#endif
-
-            /* Skip version token */
-            if (shader_is_version_token(*pToken)) {
-                ++pToken;
-                continue;
-            }
-
-            /* Skip comment tokens */
-            if (shader_is_comment_token(*pToken)) {
-                DWORD comment_len = (*pToken & D3DSI_COMMENTSIZE_MASK) >> D3DSI_COMMENTSIZE_SHIFT;
-                ++pToken;
-                TRACE("#%s\n", (char*)pToken);
-                pToken += comment_len;
-                continue;
-            }
-
-            /* Read opcode */
-            curOpcode = shader_get_opcode((IWineD3DBaseShader*) This, *pToken);
-            ++pToken;
-
-            /* Unknown opcode and its parameters */
-            if (NULL == curOpcode) {
-                while (*pToken & 0x80000000) { /* TODO: Think of a sensible name for 0x80000000 */
-                    FIXME("unrecognized opcode: %08lx\n", *pToken);
-                    ++pToken;
-                }
-
-            /* Unhandled opcode */
-            } else if (GLNAME_REQUIRE_GLSL == curOpcode->glname) {
-
-                FIXME("Token %s requires greater functionality than "
-                    "Fragment_Progarm_ARB supports\n", curOpcode->name);
-                pToken += curOpcode->num_params;
-
-            /* If a generator function is set, use it */
-            } else if (curOpcode->hw_fct != NULL) {
-
-                SHADER_OPCODE_ARG hw_arg;
-
-                hw_arg.shader = (IWineD3DBaseShader*) This;
-                hw_arg.opcode = curOpcode;
-                hw_arg.buffer = &buffer;
-                if (curOpcode->num_params > 0) {
-                    hw_arg.dst = *pToken;
-
-                    /* FIXME: this does not account for relative address tokens */
-                    for (i = 1; i < curOpcode->num_params; i++)
-                       hw_arg.src[i-1] = *(pToken + i);
-                }
-
-                curOpcode->hw_fct(&hw_arg);
-                pToken += curOpcode->num_params;
-
-            } else {
-
-                TRACE("Found opcode D3D:%s GL:%s, PARAMS:%d, \n",
-                curOpcode->name, curOpcode->glname, curOpcode->num_params);
+/** Pixel Shader specific things that need to go into the footer of the
+    ARB_FRAGMENT_PROGRAM string */
+void pixel_shader_footer_arb(IWineD3DBaseShader* iface, SHADER_BUFFER* buffer) {
+    shader_addline(buffer, "MOV result.color, R0;\n");
+}
 
-                /* Build opcode for GL vertex_program */
-                switch (curOpcode->opcode) {
-                case D3DSIO_NOP:
-                    break;
+/** Generate a pixel shader string using either GL_FRAGMENT_PROGRAM_ARB
+    or GLSL and send it to the card */
+inline static VOID IWineD3DPixelShaderImpl_GenerateShader(
+    IWineD3DPixelShader *iface, 
+    CONST DWORD *pFunction) {
 
-                default:
-                    FIXME("Can't handle opcode %s in hwShader\n", curOpcode->name);
-                    pToken += curOpcode->num_params; 
-                }
-            }
-        }
-        /* TODO: What about result.depth? */
-        shader_addline(&buffer, "MOV result.color, R0;\n");
-        shader_addline(&buffer, "END\n");
-    }
+    IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *)iface;
+    SHADER_BUFFER buffer;
 
-    /* finally null terminate the buffer */
-    buffer.buffer[buffer.bsize] = 0;
+    /* TODO: Optionally, generate the GLSL shader instead */
     if (GL_SUPPORT(ARB_VERTEX_PROGRAM)) {
-        /*  Create the hw shader */
+        /*  Create the hw ARB shader */
 
-        /* The program string sometimes gets too long for a normal TRACE */
-        TRACE("Generated program:\n");
-        if (TRACE_ON(d3d_shader)) {
-            fprintf(stderr, "%s\n", buffer.buffer);
-        }
+        /** Call the base shader generation routine to generate the pixel shader string for us */
+        generate_base_shader( (IWineD3DBaseShader*) This, 
+            &buffer, 
+            pixel_shader_header_arb,
+            pixel_shader_footer_arb,
+            pFunction);
 
-        /* TODO: change to resource.glObjectHandel or something like that */
+        /* TODO: change to resource.glObjectHandle or something like that */
         GL_EXTCALL(glGenProgramsARB(1, &This->baseShader.prgId));
 
         TRACE("Creating a hw pixel shader, prg=%d\n", This->baseShader.prgId);
@@ -1462,7 +1363,7 @@ inline static VOID IWineD3DPixelShaderIm
         TRACE("Created hw pixel shader, prg=%d\n", This->baseShader.prgId);
         /* Create the program and check for errors */
         GL_EXTCALL(glProgramStringARB(GL_FRAGMENT_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, 
-           buffer.bsize, buffer.buffer));
+            buffer.bsize, buffer.buffer));
 
         if (glGetError() == GL_INVALID_OPERATION) {
             GLint errPos;
@@ -1726,7 +1627,7 @@ HRESULT WINAPI IWineD3DPixelShaderImpl_S
     if (NULL != pFunction  && wined3d_settings.vs_mode == VS_HW) {
         TRACE("(%p) : Generating hardware program\n", This);
 #if 1
-        IWineD3DPixelShaderImpl_GenerateProgramArbHW(iface, pFunction);
+        IWineD3DPixelShaderImpl_GenerateShader(iface, pFunction);
 #endif
     }
 
diff --git a/dlls/wined3d/vertexshader.c b/dlls/wined3d/vertexshader.c
index 3395803..db9245c 100644
--- a/dlls/wined3d/vertexshader.c
+++ b/dlls/wined3d/vertexshader.c
@@ -1118,164 +1118,84 @@ void vshader_hw_mnxn(SHADER_OPCODE_ARG* 
     }
 }
 
-/**
- * Function parser ...
- */
-
-inline static VOID IWineD3DVertexShaderImpl_GenerateProgramArbHW(IWineD3DVertexShader *iface, CONST DWORD* pFunction) {
-    IWineD3DVertexShaderImpl *This = (IWineD3DVertexShaderImpl *)iface;
-    const DWORD* pToken = pFunction;
-    const SHADER_OPCODE* curOpcode = NULL;
-    DWORD i;
-    SHADER_BUFFER buffer;
-    DWORD addressUsed, tempsUsed;
-
-#if 0 /* FIXME: Use the buffer that is held by the device, this is ok since fixups will be skipped for software shaders
-        it also requires entering a critical section but cuts down the runtime footprint of wined3d and any memory fragmentation that may occur... */
-    if (This->device->fixupVertexBufferSize < SHADER_PGMSIZE) {
-        HeapFree(GetProcessHeap(), 0, This->fixupVertexBuffer);
-        This->fixupVertexBuffer = HeapAlloc(GetProcessHeap() , 0, SHADER_PGMSIZE);
-        This->fixupVertexBufferSize = PGMSIZE;
-        This->fixupVertexBuffer[0] = 0;
-    }
-    buffer.buffer = This->device->fixupVertexBuffer;
-#endif
-    buffer.buffer = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, SHADER_PGMSIZE); 
-    buffer.bsize = 0;
-    buffer.lineNo = 0;
-
-    /* First pass: figure out which temporary and texture registers are used */
-    shader_get_registers_used((IWineD3DBaseShader*) This, pToken, &tempsUsed, &addressUsed);
-    TRACE("Address registers used: %#lx, Temp registers used %#lx\n", addressUsed, tempsUsed);
-
-    /* TODO: check register usage against GL/Directx limits, and fail if they're exceeded
-        nUseAddressRegister < = GL_MAX_PROGRAM_ADDRESS_REGISTERS_AR
-        nUseTempRegister    <=  GL_MAX_PROGRAM_LOCAL_PARAMETERS_ARB
-    */
+/** Vertex Shader specific things that need to go into the header of the
+    ARB_VERTEX_PROGRAM string */
+void vertex_shader_header_arb(IWineD3DBaseShader* iface, SHADER_BUFFER* buffer) {
+
+    IWineD3DVertexShaderImpl* This = (IWineD3DVertexShaderImpl*) iface;
+    DWORD tempsUsed = This->baseShader.temps_used;
+    DWORD addressUsed = This->baseShader.textures_used;
+    int i;
 
     /* Mesa supports only 95 constants */
     if (GL_VEND(MESA) || GL_VEND(WINE))
         This->baseShader.limits.constant_float = 
            min(95, This->baseShader.limits.constant_float);
 
-    /* FIXME: if jumps are used, use GLSL, else use ARB_vertex_program */
-    shader_addline(&buffer, "!!ARBvp1.0\n");
+    shader_addline(buffer, "!!ARBvp1.0\n");
 
     /* Pre-declare registers */
     for (i = 0; i < This->baseShader.limits.temporary; i++) {
         if (tempsUsed & (1 << i))
-            shader_addline(&buffer, "TEMP T%lu;\n", i);
+            shader_addline(buffer, "TEMP T%lu;\n", i);
     }
 
     for (i = 0; i < This->baseShader.limits.address; i++) {
         if (addressUsed & (1 << i))
-            shader_addline(&buffer, "ADDRESS A%ld;\n", i);
+            shader_addline(buffer, "ADDRESS A%ld;\n", i);
     }
 
     /* Why do we need to alias those? */
-    shader_addline(&buffer, "PARAM C[%d] = { program.env[0..%d] };\n",
+    shader_addline(buffer, "PARAM C[%d] = { program.env[0..%d] };\n",
         This->baseShader.limits.constant_float, 
-        This->baseShader.limits.constant_float - 1);            
-
-  /** second pass, now generate */
-  pToken = pFunction;
-
-  if (NULL != pToken) {
-    while (1) {
-
-      if (D3DVS_END() == *pToken)
-            break;
-
-      /* Skip version */
-      if (shader_is_version_token(*pToken)) {
-            ++pToken;
-            continue;
-      }
-
-      /* Skip comments */
-      if (shader_is_comment_token(*pToken)) {
-            DWORD comment_len = (*pToken & D3DSI_COMMENTSIZE_MASK) >> D3DSI_COMMENTSIZE_SHIFT;
-            ++pToken;
-            TRACE("#%s\n", (char*)pToken);
-            pToken += comment_len;
-            continue;
-      }
-
-      /* Read opcode */
-      curOpcode = shader_get_opcode((IWineD3DBaseShader*) This, *pToken);
-      ++pToken;
-
-      /* Unknown opcode and its parameters */
-      if (NULL == curOpcode) {
-        while (*pToken & 0x80000000) {
-            FIXME("unrecognized opcode: %08lx\n", *pToken);
-            ++pToken;
-        }
-
-      /* Unhandled opcode */
-      } else if (GLNAME_REQUIRE_GLSL == curOpcode->glname) {
-          
-            FIXME("Token %s requires greater functionality than "
-                 "Vertex_Program_ARB supports\n", curOpcode->name);
-            pToken += curOpcode->num_params;
-
-      /* If a generator function is set, use it */
-      } else if (curOpcode->hw_fct != NULL) {
-
-          SHADER_OPCODE_ARG hw_arg;
+        This->baseShader.limits.constant_float - 1);
+}
 
-          hw_arg.shader = (IWineD3DBaseShader*) This;
-          hw_arg.opcode = curOpcode;
-          hw_arg.buffer = &buffer;
-          if (curOpcode->num_params > 0) {
-              hw_arg.dst = *pToken;
-
-              /* FIXME: this does not account for relative address tokens */
-              for (i = 1; i < curOpcode->num_params; i++)
-                 hw_arg.src[i-1] = *(pToken + i);
-          }
+/** Vertex Shader specific things that need to go into the footer of the
+    ARB_VERTEX_PROGRAM string */
+void vertex_shader_footer_arb(IWineD3DBaseShader* iface, SHADER_BUFFER* buffer) {
+    /* Do nothing */
+}
 
-          curOpcode->hw_fct(&hw_arg);
-          pToken += curOpcode->num_params;
+/** Generate a vertex shader string using either GL_VERTEX_PROGRAM_ARB
+    or GLSL and send it to the card */
+inline static VOID IWineD3DVertexShaderImpl_GenerateShader(
+    IWineD3DVertexShader *iface, 
+    CONST DWORD *pFunction) {
 
-      /* Unrecognized or No-Op code */
-      } else {
+    IWineD3DVertexShaderImpl *This = (IWineD3DVertexShaderImpl *)iface;
+    SHADER_BUFFER buffer;
 
-        switch (curOpcode->opcode) {
-        case D3DSIO_NOP:
-            break;
-        default:
-            FIXME("Can't handle opcode %s in hwShader\n", curOpcode->name);
-            pToken += curOpcode->num_params;
+    /* TODO: Optionally, generate the GLSL shader instead */
+    if (GL_SUPPORT(ARB_VERTEX_PROGRAM)) {
+        /*  Create the hw ARB shader */
+
+        /** Call the base shader generation routine to generate the vertex shader string for us */
+        generate_base_shader( (IWineD3DBaseShader*) This, 
+            &buffer, 
+            vertex_shader_header_arb,
+            vertex_shader_footer_arb,
+            pFunction);
+
+        /* TODO: change to resource.glObjectHandle or something like that */
+        GL_EXTCALL(glGenProgramsARB(1, &This->baseShader.prgId));
+
+        TRACE("Creating a hw vertex shader, prg=%d\n", This->baseShader.prgId);
+        GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, This->baseShader.prgId));
+
+        TRACE("Created hw vertex shader, prg=%d\n", This->baseShader.prgId);
+        /* Create the program and check for errors */
+        GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, 
+            buffer.bsize, buffer.buffer));
+
+        if (glGetError() == GL_INVALID_OPERATION) {
+            GLint errPos;
+            glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &errPos);
+            FIXME("HW VertexShader Error at position %d: %s\n",
+                  errPos, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
+            This->baseShader.prgId = -1;
         }
-      }
     }
-    shader_addline(&buffer, "END\n"); 
-  }
-
-  /* finally null terminate the buffer */
-  buffer.buffer[buffer.bsize] = 0;
-
-  /* Check that Vertex Shaders are supported */
-  if (GL_SUPPORT(ARB_VERTEX_PROGRAM)) {
-      /*  Create the hw shader */
-      /* TODO: change to resource.glObjectHandel or something like that */
-      GL_EXTCALL(glGenProgramsARB(1, &This->baseShader.prgId));
-      TRACE("Creating a hw vertex shader, prg=%d\n", This->baseShader.prgId);
-      GL_EXTCALL(glBindProgramARB(GL_VERTEX_PROGRAM_ARB, This->baseShader.prgId));
-
-      /* Create the program and check for errors */
-      GL_EXTCALL(glProgramStringARB(GL_VERTEX_PROGRAM_ARB, GL_PROGRAM_FORMAT_ASCII_ARB, 
-          buffer.bsize, buffer.buffer));
-
-      if (glGetError() == GL_INVALID_OPERATION) {
-          GLint errPos;
-          glGetIntegerv(GL_PROGRAM_ERROR_POSITION_ARB, &errPos);
-          FIXME("HW VertexShader Error at position %d: %s\n",
-                errPos, debugstr_a((const char *)glGetString(GL_PROGRAM_ERROR_STRING_ARB)));
-          This->baseShader.prgId = -1;
-      }
-  }
 #if 1 /* if were using the data buffer of device then we don't need to free it */
   HeapFree(GetProcessHeap(), 0, buffer.buffer);
 #endif
@@ -1783,7 +1703,7 @@ HRESULT WINAPI IWineD3DVertexShaderImpl_
     /* Generate HW shader in needed */
     if (NULL != pFunction  && wined3d_settings.vs_mode == VS_HW) {
 #if 1
-        IWineD3DVertexShaderImpl_GenerateProgramArbHW(iface, pFunction);
+        IWineD3DVertexShaderImpl_GenerateShader(iface, pFunction);
 #endif
     }
 
diff --git a/dlls/wined3d/wined3d_private.h b/dlls/wined3d/wined3d_private.h
index af915b3..e68300f 100644
--- a/dlls/wined3d/wined3d_private.h
+++ b/dlls/wined3d/wined3d_private.h
@@ -1179,8 +1179,11 @@ enum vsConstantsEnum {
 };
 
 struct SHADER_OPCODE_ARG;
+struct SHADER_BUFFER;
+
 typedef void (*shader_fct_t)();
 typedef void (*SHADER_HANDLER) (struct SHADER_OPCODE_ARG*);
+typedef void (*SHADER_SPECIFIC_FCT) (IWineD3DBaseShader*, struct SHADER_BUFFER*);
 
 #define SHADER_PGMSIZE 65535
 typedef struct SHADER_BUFFER {
@@ -1236,14 +1239,19 @@ extern const SHADER_OPCODE* shader_get_o
 
 extern void shader_get_registers_used(
     IWineD3DBaseShader *iface,
-    CONST DWORD* pToken,
-    DWORD* tempsUsed,
-    DWORD* texUsed);
+    CONST DWORD* pToken);
 
 extern void shader_program_dump_decl_usage(
     DWORD dcl,
     DWORD param);
 
+extern void generate_base_shader(
+    IWineD3DBaseShader *iface,
+    SHADER_BUFFER* buffer,
+    SHADER_SPECIFIC_FCT shader_header_fct,
+    SHADER_SPECIFIC_FCT shader_footer_fct,
+    CONST DWORD* pFunction);
+
 inline static BOOL shader_is_version_token(DWORD token) {
     return 0xFFFF0000 == (token & 0xFFFF0000) || 
            0xFFFE0000 == (token & 0xFFFF0000);
@@ -1270,6 +1278,8 @@ typedef struct IWineD3DBaseShaderClass
     CONST SHADER_OPCODE             *shader_ins;
     CONST DWORD                     *function;
     UINT                            functionLength;
+    DWORD                           textures_used; /* Bitmap of either addresses or textures used */
+    DWORD                           temps_used;    /* Bitmap of temporary registers used */
     GLuint                          prgId;
 } IWineD3DBaseShaderClass;
 






More information about the wine-patches mailing list