/*
* GLSL pixel and vertex shader implementation
*
- * Copyright 2006 Jason Green
+ * Copyright 2006 Jason Green
* Copyright 2006-2007 Henri Verbeet
- * Copyright 2007-2008 Stefan Dösinger for CodeWeavers
+ * Copyright 2007-2008 Stefan Dösinger for CodeWeavers
+ * Copyright 2009 Henri Verbeet for CodeWeavers
*
* This library is free software; you can redistribute it and/or
* modify it under the terms of the GNU Lesser General Public
*/
#include "config.h"
+#include <limits.h>
#include <stdio.h>
#include "wined3d_private.h"
WINE_DEFAULT_DEBUG_CHANNEL(d3d_shader);
WINE_DECLARE_DEBUG_CHANNEL(d3d_constants);
WINE_DECLARE_DEBUG_CHANNEL(d3d_caps);
+WINE_DECLARE_DEBUG_CHANNEL(d3d);
#define GLINFO_LOCATION (*gl_info)
+#define WINED3D_GLSL_SAMPLE_PROJECTED 0x1
+#define WINED3D_GLSL_SAMPLE_RECT 0x2
+#define WINED3D_GLSL_SAMPLE_LOD 0x4
+#define WINED3D_GLSL_SAMPLE_GRAD 0x8
+
typedef struct {
char reg_name[150];
char mask_str[6];
typedef struct {
char reg_name[150];
- char param_str[100];
+ char param_str[200];
} glsl_src_param_t;
typedef struct {
DWORD coord_mask;
} glsl_sample_function_t;
+enum heap_node_op
+{
+ HEAP_NODE_TRAVERSE_LEFT,
+ HEAP_NODE_TRAVERSE_RIGHT,
+ HEAP_NODE_POP,
+};
+
+struct constant_entry
+{
+ unsigned int idx;
+ unsigned int version;
+};
+
+struct constant_heap
+{
+ struct constant_entry *entries;
+ unsigned int *positions;
+ unsigned int size;
+};
+
/* GLSL shader private data */
struct shader_glsl_priv {
- struct hash_table_t *glsl_program_lookup;
+ struct wined3d_shader_buffer shader_buffer;
+ struct wine_rb_tree program_lookup;
struct glsl_shader_prog_link *glsl_program;
- GLhandleARB depth_blt_glsl_program_id;
+ struct constant_heap vconst_heap;
+ struct constant_heap pconst_heap;
+ unsigned char *stack;
+ GLhandleARB depth_blt_program[tex_type_count];
+ UINT next_constant_version;
};
/* Struct to maintain data about a linked GLSL program */
struct glsl_shader_prog_link {
- struct list vshader_entry;
- struct list pshader_entry;
- GLhandleARB programId;
- GLhandleARB *vuniformF_locations;
- GLhandleARB *puniformF_locations;
- GLhandleARB vuniformI_locations[MAX_CONST_I];
- GLhandleARB puniformI_locations[MAX_CONST_I];
- GLhandleARB posFixup_location;
- GLhandleARB bumpenvmat_location[MAX_TEXTURES];
- GLhandleARB luminancescale_location[MAX_TEXTURES];
- GLhandleARB luminanceoffset_location[MAX_TEXTURES];
- GLhandleARB srgb_comparison_location;
- GLhandleARB srgb_mul_low_location;
- GLhandleARB ycorrection_location;
- GLenum vertex_color_clamp;
- GLhandleARB vshader;
- GLhandleARB pshader;
+ struct wine_rb_entry program_lookup_entry;
+ struct list vshader_entry;
+ struct list pshader_entry;
+ GLhandleARB programId;
+ GLint *vuniformF_locations;
+ GLint *puniformF_locations;
+ GLint vuniformI_locations[MAX_CONST_I];
+ GLint puniformI_locations[MAX_CONST_I];
+ GLint posFixup_location;
+ GLint np2Fixup_location;
+ GLint bumpenvmat_location[MAX_TEXTURES];
+ GLint luminancescale_location[MAX_TEXTURES];
+ GLint luminanceoffset_location[MAX_TEXTURES];
+ GLint ycorrection_location;
+ GLenum vertex_color_clamp;
+ IWineD3DVertexShader *vshader;
+ IWineD3DPixelShader *pshader;
+ struct vs_compile_args vs_args;
+ struct ps_compile_args ps_args;
+ UINT constant_version;
+ const struct ps_np2fixup_info *np2Fixup_info;
};
typedef struct {
- GLhandleARB vshader;
- GLhandleARB pshader;
+ IWineD3DVertexShader *vshader;
+ IWineD3DPixelShader *pshader;
+ struct ps_compile_args ps_args;
+ struct vs_compile_args vs_args;
} glsl_program_key_t;
+struct shader_glsl_ctx_priv {
+ const struct vs_compile_args *cur_vs_args;
+ const struct ps_compile_args *cur_ps_args;
+ struct ps_np2fixup_info *cur_np2fixup_info;
+};
+
+struct glsl_ps_compiled_shader
+{
+ struct ps_compile_args args;
+ struct ps_np2fixup_info np2fixup;
+ GLhandleARB prgId;
+};
+
+struct glsl_pshader_private
+{
+ struct glsl_ps_compiled_shader *gl_shaders;
+ UINT num_gl_shaders, shader_array_size;
+};
+
+struct glsl_vs_compiled_shader
+{
+ struct vs_compile_args args;
+ GLhandleARB prgId;
+};
+
+struct glsl_vshader_private
+{
+ struct glsl_vs_compiled_shader *gl_shaders;
+ UINT num_gl_shaders, shader_array_size;
+};
+
+/* Extract a line from the info log.
+ * Note that this modifies the source string. */
+static char *get_info_log_line(char **ptr)
+{
+ char *p, *q;
+
+ p = *ptr;
+ if (!(q = strstr(p, "\n")))
+ {
+ if (!*p) return NULL;
+ *ptr += strlen(p);
+ return p;
+ }
+ *q = '\0';
+ *ptr = q + 1;
+
+ return p;
+}
/** Prints the GLSL info log which will contain error messages if they exist */
-static void print_glsl_info_log(WineD3D_GL_Info *gl_info, GLhandleARB obj) {
+/* GL locking is done by the caller */
+static void print_glsl_info_log(const struct wined3d_gl_info *gl_info, GLhandleARB obj)
+{
int infologLength = 0;
char *infoLog;
- int i;
+ unsigned int i;
BOOL is_spam;
- const char *spam[] = {
+ static const char * const spam[] =
+ {
"Vertex shader was successfully compiled to run on hardware.\n", /* fglrx */
- "Fragment shader was successfully compiled to run on hardware.\n", /* fglrx */
+ "Fragment shader was successfully compiled to run on hardware.\n", /* fglrx, with \n */
+ "Fragment shader was successfully compiled to run on hardware.", /* fglrx, no \n */
"Fragment shader(s) linked, vertex shader(s) linked. \n ", /* fglrx, with \n */
"Fragment shader(s) linked, vertex shader(s) linked.", /* fglrx, no \n */
"Vertex shader(s) linked, no fragment shader(s) defined. \n ", /* fglrx, with \n */
"Vertex shader(s) linked, no fragment shader(s) defined.", /* fglrx, no \n */
- "Fragment shader was successfully compiled to run on hardware.\nWARNING: 0:1: extension 'GL_ARB_draw_buffers' is not supported",
- "Fragment shader(s) linked, no vertex shader(s) defined.", /* fglrx, no \n */
"Fragment shader(s) linked, no vertex shader(s) defined. \n ", /* fglrx, with \n */
- "WARNING: 0:2: extension 'GL_ARB_draw_buffers' is not supported\n" /* MacOS ati */
+ "Fragment shader(s) linked, no vertex shader(s) defined.", /* fglrx, no \n */
};
+ if (!TRACE_ON(d3d_shader) && !FIXME_ON(d3d_shader)) return;
+
GL_EXTCALL(glGetObjectParameterivARB(obj,
GL_OBJECT_INFO_LOG_LENGTH_ARB,
&infologLength));
* that if there are errors. */
if (infologLength > 1)
{
+ char *ptr, *line;
+
/* Fglrx doesn't terminate the string properly, but it tells us the proper length.
* So use HEAP_ZERO_MEMORY to avoid uninitialized bytes
*/
break;
}
}
- if(is_spam) {
- TRACE("Spam received from GLSL shader #%u: %s\n", obj, debugstr_a(infoLog));
- } else {
- FIXME("Error received from GLSL shader #%u: %s\n", obj, debugstr_a(infoLog));
+
+ ptr = infoLog;
+ if (is_spam)
+ {
+ TRACE("Spam received from GLSL shader #%u:\n", obj);
+ while ((line = get_info_log_line(&ptr))) TRACE(" %s\n", line);
+ }
+ else
+ {
+ FIXME("Error received from GLSL shader #%u:\n", obj);
+ while ((line = get_info_log_line(&ptr))) FIXME(" %s\n", line);
}
HeapFree(GetProcessHeap(), 0, infoLog);
}
/**
* Loads (pixel shader) samplers
*/
-static void shader_glsl_load_psamplers(
- WineD3D_GL_Info *gl_info,
- IWineD3DStateBlock* iface,
- GLhandleARB programId) {
-
- IWineD3DStateBlockImpl* stateBlock = (IWineD3DStateBlockImpl*) iface;
- GLhandleARB name_loc;
+/* GL locking is done by the caller */
+static void shader_glsl_load_psamplers(const struct wined3d_gl_info *gl_info,
+ DWORD *tex_unit_map, GLhandleARB programId)
+{
+ GLint name_loc;
int i;
char sampler_name[20];
snprintf(sampler_name, sizeof(sampler_name), "Psampler%d", i);
name_loc = GL_EXTCALL(glGetUniformLocationARB(programId, sampler_name));
if (name_loc != -1) {
- int mapped_unit = stateBlock->wineD3DDevice->texUnitMap[i];
- if (mapped_unit != -1 && mapped_unit < GL_LIMITS(fragment_samplers)) {
+ DWORD mapped_unit = tex_unit_map[i];
+ if (mapped_unit != WINED3D_UNMAPPED_STAGE && mapped_unit < gl_info->limits.fragment_samplers)
+ {
TRACE("Loading %s for texture %d\n", sampler_name, mapped_unit);
GL_EXTCALL(glUniform1iARB(name_loc, mapped_unit));
checkGLcall("glUniform1iARB");
}
}
-static void shader_glsl_load_vsamplers(WineD3D_GL_Info *gl_info, IWineD3DStateBlock* iface, GLhandleARB programId) {
- IWineD3DStateBlockImpl* stateBlock = (IWineD3DStateBlockImpl*) iface;
- GLhandleARB name_loc;
+/* GL locking is done by the caller */
+static void shader_glsl_load_vsamplers(const struct wined3d_gl_info *gl_info,
+ DWORD *tex_unit_map, GLhandleARB programId)
+{
+ GLint name_loc;
char sampler_name[20];
int i;
snprintf(sampler_name, sizeof(sampler_name), "Vsampler%d", i);
name_loc = GL_EXTCALL(glGetUniformLocationARB(programId, sampler_name));
if (name_loc != -1) {
- int mapped_unit = stateBlock->wineD3DDevice->texUnitMap[MAX_FRAGMENT_SAMPLERS + i];
- if (mapped_unit != -1 && mapped_unit < GL_LIMITS(combined_samplers)) {
+ DWORD mapped_unit = tex_unit_map[MAX_FRAGMENT_SAMPLERS + i];
+ if (mapped_unit != WINED3D_UNMAPPED_STAGE && mapped_unit < gl_info->limits.combined_samplers)
+ {
TRACE("Loading %s for texture %d\n", sampler_name, mapped_unit);
GL_EXTCALL(glUniform1iARB(name_loc, mapped_unit));
checkGLcall("glUniform1iARB");
}
}
-/**
- * Loads floating point constants (aka uniforms) into the currently set GLSL program.
- * When constant_list == NULL, it will load all the constants.
- */
-static void shader_glsl_load_constantsF(IWineD3DBaseShaderImpl* This, WineD3D_GL_Info *gl_info,
- unsigned int max_constants, float* constants, GLhandleARB *constant_locations,
- struct list *constant_list) {
- constants_entry *constant;
- local_constant* lconst;
- GLhandleARB tmp_loc;
- DWORD i, j, k;
- DWORD *idx;
-
- if (TRACE_ON(d3d_shader)) {
- LIST_FOR_EACH_ENTRY(constant, constant_list, constants_entry, entry) {
- idx = constant->idx;
- j = constant->count;
- while (j--) {
- i = *idx++;
- tmp_loc = constant_locations[i];
- if (tmp_loc != -1) {
- TRACE_(d3d_constants)("Loading constants %i: %f, %f, %f, %f\n", i,
- constants[i * 4 + 0], constants[i * 4 + 1],
- constants[i * 4 + 2], constants[i * 4 + 3]);
+/* GL locking is done by the caller */
+static inline void walk_constant_heap(const struct wined3d_gl_info *gl_info, const float *constants,
+ const GLint *constant_locations, const struct constant_heap *heap, unsigned char *stack, DWORD version)
+{
+ int stack_idx = 0;
+ unsigned int heap_idx = 1;
+ unsigned int idx;
+
+ if (heap->entries[heap_idx].version <= version) return;
+
+ idx = heap->entries[heap_idx].idx;
+ if (constant_locations[idx] != -1) GL_EXTCALL(glUniform4fvARB(constant_locations[idx], 1, &constants[idx * 4]));
+ stack[stack_idx] = HEAP_NODE_TRAVERSE_LEFT;
+
+ while (stack_idx >= 0)
+ {
+ /* Note that we fall through to the next case statement. */
+ switch(stack[stack_idx])
+ {
+ case HEAP_NODE_TRAVERSE_LEFT:
+ {
+ unsigned int left_idx = heap_idx << 1;
+ if (left_idx < heap->size && heap->entries[left_idx].version > version)
+ {
+ heap_idx = left_idx;
+ idx = heap->entries[heap_idx].idx;
+ if (constant_locations[idx] != -1)
+ GL_EXTCALL(glUniform4fvARB(constant_locations[idx], 1, &constants[idx * 4]));
+
+ stack[stack_idx++] = HEAP_NODE_TRAVERSE_RIGHT;
+ stack[stack_idx] = HEAP_NODE_TRAVERSE_LEFT;
+ break;
}
}
+
+ case HEAP_NODE_TRAVERSE_RIGHT:
+ {
+ unsigned int right_idx = (heap_idx << 1) + 1;
+ if (right_idx < heap->size && heap->entries[right_idx].version > version)
+ {
+ heap_idx = right_idx;
+ idx = heap->entries[heap_idx].idx;
+ if (constant_locations[idx] != -1)
+ GL_EXTCALL(glUniform4fvARB(constant_locations[idx], 1, &constants[idx * 4]));
+
+ stack[stack_idx++] = HEAP_NODE_POP;
+ stack[stack_idx] = HEAP_NODE_TRAVERSE_LEFT;
+ break;
+ }
+ }
+
+ case HEAP_NODE_POP:
+ {
+ heap_idx >>= 1;
+ --stack_idx;
+ break;
+ }
}
}
+ checkGLcall("walk_constant_heap()");
+}
- /* 1.X pshaders have the constants clamped to [-1;1] implicitly. */
- if(WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) == 1 &&
- shader_is_pshader_version(This->baseShader.hex_version)) {
- float lcl_const[4];
-
- LIST_FOR_EACH_ENTRY(constant, constant_list, constants_entry, entry) {
- idx = constant->idx;
- j = constant->count;
- while (j--) {
- i = *idx++;
- tmp_loc = constant_locations[i];
- if (tmp_loc != -1) {
- /* We found this uniform name in the program - go ahead and send the data */
- k = i * 4;
- if(constants[k + 0] < -1.0) lcl_const[0] = -1.0;
- else if(constants[k + 0] > 1.0) lcl_const[0] = 1.0;
- else lcl_const[0] = constants[k + 0];
- if(constants[k + 1] < -1.0) lcl_const[1] = -1.0;
- else if(constants[k + 1] > 1.0) lcl_const[1] = 1.0;
- else lcl_const[1] = constants[k + 1];
- if(constants[k + 2] < -1.0) lcl_const[2] = -1.0;
- else if(constants[k + 2] > 1.0) lcl_const[2] = 1.0;
- else lcl_const[2] = constants[k + 2];
- if(constants[k + 3] < -1.0) lcl_const[3] = -1.0;
- else if(constants[k + 3] > 1.0) lcl_const[3] = 1.0;
- else lcl_const[3] = constants[k + 3];
-
- GL_EXTCALL(glUniform4fvARB(tmp_loc, 1, lcl_const));
+/* GL locking is done by the caller */
+static inline void apply_clamped_constant(const struct wined3d_gl_info *gl_info, GLint location, const GLfloat *data)
+{
+ GLfloat clamped_constant[4];
+
+ if (location == -1) return;
+
+ clamped_constant[0] = data[0] < -1.0f ? -1.0f : data[0] > 1.0f ? 1.0f : data[0];
+ clamped_constant[1] = data[1] < -1.0f ? -1.0f : data[1] > 1.0f ? 1.0f : data[1];
+ clamped_constant[2] = data[2] < -1.0f ? -1.0f : data[2] > 1.0f ? 1.0f : data[2];
+ clamped_constant[3] = data[3] < -1.0f ? -1.0f : data[3] > 1.0f ? 1.0f : data[3];
+
+ GL_EXTCALL(glUniform4fvARB(location, 1, clamped_constant));
+}
+
+/* GL locking is done by the caller */
+static inline void walk_constant_heap_clamped(const struct wined3d_gl_info *gl_info, const float *constants,
+ const GLint *constant_locations, const struct constant_heap *heap, unsigned char *stack, DWORD version)
+{
+ int stack_idx = 0;
+ unsigned int heap_idx = 1;
+ unsigned int idx;
+
+ if (heap->entries[heap_idx].version <= version) return;
+
+ idx = heap->entries[heap_idx].idx;
+ apply_clamped_constant(gl_info, constant_locations[idx], &constants[idx * 4]);
+ stack[stack_idx] = HEAP_NODE_TRAVERSE_LEFT;
+
+ while (stack_idx >= 0)
+ {
+ /* Note that we fall through to the next case statement. */
+ switch(stack[stack_idx])
+ {
+ case HEAP_NODE_TRAVERSE_LEFT:
+ {
+ unsigned int left_idx = heap_idx << 1;
+ if (left_idx < heap->size && heap->entries[left_idx].version > version)
+ {
+ heap_idx = left_idx;
+ idx = heap->entries[heap_idx].idx;
+ apply_clamped_constant(gl_info, constant_locations[idx], &constants[idx * 4]);
+
+ stack[stack_idx++] = HEAP_NODE_TRAVERSE_RIGHT;
+ stack[stack_idx] = HEAP_NODE_TRAVERSE_LEFT;
+ break;
}
}
- }
- } else {
- LIST_FOR_EACH_ENTRY(constant, constant_list, constants_entry, entry) {
- idx = constant->idx;
- j = constant->count;
- while (j--) {
- i = *idx++;
- tmp_loc = constant_locations[i];
- if (tmp_loc != -1) {
- /* We found this uniform name in the program - go ahead and send the data */
- GL_EXTCALL(glUniform4fvARB(tmp_loc, 1, constants + (i * 4)));
+
+ case HEAP_NODE_TRAVERSE_RIGHT:
+ {
+ unsigned int right_idx = (heap_idx << 1) + 1;
+ if (right_idx < heap->size && heap->entries[right_idx].version > version)
+ {
+ heap_idx = right_idx;
+ idx = heap->entries[heap_idx].idx;
+ apply_clamped_constant(gl_info, constant_locations[idx], &constants[idx * 4]);
+
+ stack[stack_idx++] = HEAP_NODE_POP;
+ stack[stack_idx] = HEAP_NODE_TRAVERSE_LEFT;
+ break;
}
}
+
+ case HEAP_NODE_POP:
+ {
+ heap_idx >>= 1;
+ --stack_idx;
+ break;
+ }
}
}
- checkGLcall("glUniform4fvARB()");
+ checkGLcall("walk_constant_heap_clamped()");
+}
- if(!This->baseShader.load_local_constsF) {
+/* Loads floating point constants (aka uniforms) into the currently set GLSL program. */
+/* GL locking is done by the caller */
+static void shader_glsl_load_constantsF(IWineD3DBaseShaderImpl *This, const struct wined3d_gl_info *gl_info,
+ const float *constants, const GLint *constant_locations, const struct constant_heap *heap,
+ unsigned char *stack, UINT version)
+{
+ const local_constant *lconst;
+
+ /* 1.X pshaders have the constants clamped to [-1;1] implicitly. */
+ if (This->baseShader.reg_maps.shader_version.major == 1
+ && shader_is_pshader_version(This->baseShader.reg_maps.shader_version.type))
+ walk_constant_heap_clamped(gl_info, constants, constant_locations, heap, stack, version);
+ else
+ walk_constant_heap(gl_info, constants, constant_locations, heap, stack, version);
+
+ if (!This->baseShader.load_local_constsF)
+ {
TRACE("No need to load local float constants for this shader\n");
return;
}
- /* Load immediate constants */
- if (TRACE_ON(d3d_shader)) {
- LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
- tmp_loc = constant_locations[lconst->idx];
- if (tmp_loc != -1) {
- GLfloat* values = (GLfloat*)lconst->value;
- TRACE_(d3d_constants)("Loading local constants %i: %f, %f, %f, %f\n", lconst->idx,
- values[0], values[1], values[2], values[3]);
- }
- }
- }
/* Immediate constants are clamped to [-1;1] at shader creation time if needed */
- LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry) {
- tmp_loc = constant_locations[lconst->idx];
- if (tmp_loc != -1) {
- /* We found this uniform name in the program - go ahead and send the data */
- GL_EXTCALL(glUniform4fvARB(tmp_loc, 1, (GLfloat*)lconst->value));
- }
+ LIST_FOR_EACH_ENTRY(lconst, &This->baseShader.constantsF, local_constant, entry)
+ {
+ GLint location = constant_locations[lconst->idx];
+ /* We found this uniform name in the program - go ahead and send the data */
+ if (location != -1) GL_EXTCALL(glUniform4fvARB(location, 1, (const GLfloat *)lconst->value));
}
checkGLcall("glUniform4fvARB()");
}
-/**
- * Loads integer constants (aka uniforms) into the currently set GLSL program.
- * When @constants_set == NULL, it will load all the constants.
- */
-static void shader_glsl_load_constantsI(
- IWineD3DBaseShaderImpl* This,
- WineD3D_GL_Info *gl_info,
- GLhandleARB programId,
- GLhandleARB locations[MAX_CONST_I],
- unsigned max_constants,
- int* constants,
- BOOL* constants_set) {
-
- int i;
+/* Loads integer constants (aka uniforms) into the currently set GLSL program. */
+/* GL locking is done by the caller */
+static void shader_glsl_load_constantsI(IWineD3DBaseShaderImpl *This, const struct wined3d_gl_info *gl_info,
+ const GLint locations[MAX_CONST_I], const int *constants, WORD constants_set)
+{
+ unsigned int i;
struct list* ptr;
- for (i=0; i<max_constants; ++i) {
- if (NULL == constants_set || constants_set[i]) {
+ for (i = 0; constants_set; constants_set >>= 1, ++i)
+ {
+ if (!(constants_set & 1)) continue;
- TRACE_(d3d_constants)("Loading constants %i: %i, %i, %i, %i\n",
- i, constants[i*4], constants[i*4+1], constants[i*4+2], constants[i*4+3]);
+ TRACE_(d3d_constants)("Loading constants %u: %i, %i, %i, %i\n",
+ i, constants[i*4], constants[i*4+1], constants[i*4+2], constants[i*4+3]);
- /* We found this uniform name in the program - go ahead and send the data */
- GL_EXTCALL(glUniform4ivARB(locations[i], 1, &constants[i*4]));
- checkGLcall("glUniform4ivARB");
- }
+ /* We found this uniform name in the program - go ahead and send the data */
+ GL_EXTCALL(glUniform4ivARB(locations[i], 1, &constants[i*4]));
+ checkGLcall("glUniform4ivARB");
}
/* Load immediate constants */
ptr = list_head(&This->baseShader.constantsI);
while (ptr) {
- local_constant* lconst = LIST_ENTRY(ptr, struct local_constant, entry);
+ const struct local_constant *lconst = LIST_ENTRY(ptr, const struct local_constant, entry);
unsigned int idx = lconst->idx;
- GLint* values = (GLint*) lconst->value;
+ const GLint *values = (const GLint *)lconst->value;
TRACE_(d3d_constants)("Loading local constants %i: %i, %i, %i, %i\n", idx,
values[0], values[1], values[2], values[3]);
}
}
-/**
- * Loads boolean constants (aka uniforms) into the currently set GLSL program.
- * When @constants_set == NULL, it will load all the constants.
- */
-static void shader_glsl_load_constantsB(
- IWineD3DBaseShaderImpl* This,
- WineD3D_GL_Info *gl_info,
- GLhandleARB programId,
- unsigned max_constants,
- BOOL* constants,
- BOOL* constants_set) {
-
- GLhandleARB tmp_loc;
- int i;
+/* Loads boolean constants (aka uniforms) into the currently set GLSL program. */
+/* GL locking is done by the caller */
+static void shader_glsl_load_constantsB(IWineD3DBaseShaderImpl *This, const struct wined3d_gl_info *gl_info,
+ GLhandleARB programId, const BOOL *constants, WORD constants_set)
+{
+ GLint tmp_loc;
+ unsigned int i;
char tmp_name[8];
- char is_pshader = shader_is_pshader_version(This->baseShader.hex_version);
- const char* prefix = is_pshader? "PB":"VB";
+ const char *prefix;
struct list* ptr;
- for (i=0; i<max_constants; ++i) {
- if (NULL == constants_set || constants_set[i]) {
+ switch (This->baseShader.reg_maps.shader_version.type)
+ {
+ case WINED3D_SHADER_TYPE_VERTEX:
+ prefix = "VB";
+ break;
+
+ case WINED3D_SHADER_TYPE_GEOMETRY:
+ prefix = "GB";
+ break;
- TRACE_(d3d_constants)("Loading constants %i: %i;\n", i, constants[i]);
+ case WINED3D_SHADER_TYPE_PIXEL:
+ prefix = "PB";
+ break;
- /* TODO: Benchmark and see if it would be beneficial to store the
- * locations of the constants to avoid looking up each time */
- snprintf(tmp_name, sizeof(tmp_name), "%s[%i]", prefix, i);
- tmp_loc = GL_EXTCALL(glGetUniformLocationARB(programId, tmp_name));
- if (tmp_loc != -1) {
- /* We found this uniform name in the program - go ahead and send the data */
- GL_EXTCALL(glUniform1ivARB(tmp_loc, 1, &constants[i]));
- checkGLcall("glUniform1ivARB");
- }
+ default:
+ FIXME("Unknown shader type %#x.\n",
+ This->baseShader.reg_maps.shader_version.type);
+ prefix = "UB";
+ break;
+ }
+
+ /* TODO: Benchmark and see if it would be beneficial to store the
+ * locations of the constants to avoid looking up each time */
+ for (i = 0; constants_set; constants_set >>= 1, ++i)
+ {
+ if (!(constants_set & 1)) continue;
+
+ TRACE_(d3d_constants)("Loading constants %i: %i;\n", i, constants[i]);
+
+ /* TODO: Benchmark and see if it would be beneficial to store the
+ * locations of the constants to avoid looking up each time */
+ snprintf(tmp_name, sizeof(tmp_name), "%s[%i]", prefix, i);
+ tmp_loc = GL_EXTCALL(glGetUniformLocationARB(programId, tmp_name));
+ if (tmp_loc != -1)
+ {
+ /* We found this uniform name in the program - go ahead and send the data */
+ GL_EXTCALL(glUniform1ivARB(tmp_loc, 1, &constants[i]));
+ checkGLcall("glUniform1ivARB");
}
}
/* Load immediate constants */
ptr = list_head(&This->baseShader.constantsB);
while (ptr) {
- local_constant* lconst = LIST_ENTRY(ptr, struct local_constant, entry);
+ const struct local_constant *lconst = LIST_ENTRY(ptr, const struct local_constant, entry);
unsigned int idx = lconst->idx;
- GLint* values = (GLint*) lconst->value;
+ const GLint *values = (const GLint *)lconst->value;
TRACE_(d3d_constants)("Loading local constants %i: %i\n", idx, values[0]);
}
}
-
+static void reset_program_constant_version(struct wine_rb_entry *entry, void *context)
+{
+ WINE_RB_ENTRY_VALUE(entry, struct glsl_shader_prog_link, program_lookup_entry)->constant_version = 0;
+}
/**
- * Loads the app-supplied constants into the currently set GLSL program.
+ * Loads the texture dimensions for NP2 fixup into the currently set GLSL program.
*/
-static void shader_glsl_load_constants(
+/* GL locking is done by the caller (state handler) */
+static void shader_glsl_load_np2fixup_constants(
IWineD3DDevice* device,
char usePixelShader,
char useVertexShader) {
-
- IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) device;
- struct shader_glsl_priv *priv = (struct shader_glsl_priv *)deviceImpl->shader_priv;
- IWineD3DStateBlockImpl* stateBlock = deviceImpl->stateBlock;
- WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
-
- GLhandleARB *constant_locations;
- struct list *constant_list;
+
+ const IWineD3DDeviceImpl* deviceImpl = (const IWineD3DDeviceImpl*) device;
+ const struct glsl_shader_prog_link* prog = ((struct shader_glsl_priv *)(deviceImpl->shader_priv))->glsl_program;
+
+ if (!prog) {
+ /* No GLSL program set - nothing to do. */
+ return;
+ }
+
+ if (!usePixelShader) {
+ /* NP2 texcoord fixup is (currently) only done for pixelshaders. */
+ return;
+ }
+
+ if (prog->ps_args.np2_fixup && -1 != prog->np2Fixup_location) {
+ const struct wined3d_gl_info *gl_info = &deviceImpl->adapter->gl_info;
+ const IWineD3DStateBlockImpl* stateBlock = (const IWineD3DStateBlockImpl*) deviceImpl->stateBlock;
+ UINT i;
+ UINT fixup = prog->ps_args.np2_fixup;
+ GLfloat np2fixup_constants[4 * MAX_FRAGMENT_SAMPLERS];
+
+ for (i = 0; fixup; fixup >>= 1, ++i) {
+ const unsigned char idx = prog->np2Fixup_info->idx[i];
+ const IWineD3DBaseTextureImpl* const tex = (const IWineD3DBaseTextureImpl*) stateBlock->textures[i];
+ GLfloat* tex_dim = &np2fixup_constants[(idx >> 1) * 4];
+
+ if (!tex) {
+ FIXME("Nonexistent texture is flagged for NP2 texcoord fixup\n");
+ continue;
+ }
+
+ if (idx % 2) {
+ tex_dim[2] = tex->baseTexture.pow2Matrix[0]; tex_dim[3] = tex->baseTexture.pow2Matrix[5];
+ } else {
+ tex_dim[0] = tex->baseTexture.pow2Matrix[0]; tex_dim[1] = tex->baseTexture.pow2Matrix[5];
+ }
+ }
+
+ GL_EXTCALL(glUniform4fvARB(prog->np2Fixup_location, prog->np2Fixup_info->num_consts, np2fixup_constants));
+ }
+}
+
+/**
+ * Loads the app-supplied constants into the currently set GLSL program.
+ */
+/* GL locking is done by the caller (state handler) */
+static void shader_glsl_load_constants(const struct wined3d_context *context,
+ char usePixelShader, char useVertexShader)
+{
+ IWineD3DDeviceImpl *device = ((IWineD3DSurfaceImpl *)context->surface)->resource.device;
+ const struct wined3d_gl_info *gl_info = context->gl_info;
+ IWineD3DStateBlockImpl* stateBlock = device->stateBlock;
+ struct shader_glsl_priv *priv = device->shader_priv;
+
GLhandleARB programId;
struct glsl_shader_prog_link *prog = priv->glsl_program;
- unsigned int i;
+ UINT constant_version;
+ int i;
if (!prog) {
/* No GLSL program set - nothing to do. */
return;
}
programId = prog->programId;
+ constant_version = prog->constant_version;
if (useVertexShader) {
IWineD3DBaseShaderImpl* vshader = (IWineD3DBaseShaderImpl*) stateBlock->vertexShader;
- constant_locations = prog->vuniformF_locations;
- constant_list = &stateBlock->set_vconstantsF;
-
/* Load DirectX 9 float constants/uniforms for vertex shader */
- shader_glsl_load_constantsF(vshader, gl_info, GL_LIMITS(vshader_constantsF),
- stateBlock->vertexShaderConstantF, constant_locations, constant_list);
+ shader_glsl_load_constantsF(vshader, gl_info, stateBlock->vertexShaderConstantF,
+ prog->vuniformF_locations, &priv->vconst_heap, priv->stack, constant_version);
/* Load DirectX 9 integer constants/uniforms for vertex shader */
- shader_glsl_load_constantsI(vshader, gl_info, programId,
- prog->vuniformI_locations, MAX_CONST_I,
- stateBlock->vertexShaderConstantI,
- stateBlock->changed.vertexShaderConstantsI);
+ shader_glsl_load_constantsI(vshader, gl_info, prog->vuniformI_locations, stateBlock->vertexShaderConstantI,
+ stateBlock->changed.vertexShaderConstantsI & vshader->baseShader.reg_maps.integer_constants);
/* Load DirectX 9 boolean constants/uniforms for vertex shader */
- shader_glsl_load_constantsB(vshader, gl_info, programId, MAX_CONST_B,
- stateBlock->vertexShaderConstantB,
- stateBlock->changed.vertexShaderConstantsB);
+ shader_glsl_load_constantsB(vshader, gl_info, programId, stateBlock->vertexShaderConstantB,
+ stateBlock->changed.vertexShaderConstantsB & vshader->baseShader.reg_maps.boolean_constants);
/* Upload the position fixup params */
- GL_EXTCALL(glUniform4fvARB(prog->posFixup_location, 1, &deviceImpl->posFixup[0]));
+ GL_EXTCALL(glUniform4fvARB(prog->posFixup_location, 1, &device->posFixup[0]));
checkGLcall("glUniform4fvARB");
}
IWineD3DBaseShaderImpl* pshader = (IWineD3DBaseShaderImpl*) stateBlock->pixelShader;
- constant_locations = prog->puniformF_locations;
- constant_list = &stateBlock->set_pconstantsF;
-
/* Load DirectX 9 float constants/uniforms for pixel shader */
- shader_glsl_load_constantsF(pshader, gl_info, GL_LIMITS(pshader_constantsF),
- stateBlock->pixelShaderConstantF, constant_locations, constant_list);
+ shader_glsl_load_constantsF(pshader, gl_info, stateBlock->pixelShaderConstantF,
+ prog->puniformF_locations, &priv->pconst_heap, priv->stack, constant_version);
/* Load DirectX 9 integer constants/uniforms for pixel shader */
- shader_glsl_load_constantsI(pshader, gl_info, programId,
- prog->puniformI_locations, MAX_CONST_I,
- stateBlock->pixelShaderConstantI,
- stateBlock->changed.pixelShaderConstantsI);
+ shader_glsl_load_constantsI(pshader, gl_info, prog->puniformI_locations, stateBlock->pixelShaderConstantI,
+ stateBlock->changed.pixelShaderConstantsI & pshader->baseShader.reg_maps.integer_constants);
/* Load DirectX 9 boolean constants/uniforms for pixel shader */
- shader_glsl_load_constantsB(pshader, gl_info, programId, MAX_CONST_B,
- stateBlock->pixelShaderConstantB,
- stateBlock->changed.pixelShaderConstantsB);
+ shader_glsl_load_constantsB(pshader, gl_info, programId, stateBlock->pixelShaderConstantB,
+ stateBlock->changed.pixelShaderConstantsB & pshader->baseShader.reg_maps.boolean_constants);
/* Upload the environment bump map matrix if needed. The needsbumpmat member specifies the texture stage to load the matrix from.
* It can't be 0 for a valid texbem instruction.
*/
- for(i = 0; i < ((IWineD3DPixelShaderImpl *) pshader)->numbumpenvmatconsts; i++) {
- IWineD3DPixelShaderImpl *ps = (IWineD3DPixelShaderImpl *) pshader;
- int stage = ps->luminanceconst[i].texunit;
+ for(i = 0; i < MAX_TEXTURES; i++) {
+ const float *data;
- float *data = (float *) &stateBlock->textureState[(int) ps->bumpenvmatconst[i].texunit][WINED3DTSS_BUMPENVMAT00];
+ if(prog->bumpenvmat_location[i] == -1) continue;
+
+ data = (const float *)&stateBlock->textureState[i][WINED3DTSS_BUMPENVMAT00];
GL_EXTCALL(glUniformMatrix2fvARB(prog->bumpenvmat_location[i], 1, 0, data));
checkGLcall("glUniformMatrix2fvARB");
/* texbeml needs the luminance scale and offset too. If texbeml is used, needsbumpmat
* is set too, so we can check that in the needsbumpmat check
*/
- if(ps->baseShader.reg_maps.luminanceparams[stage]) {
- GLfloat *scale = (GLfloat *) &stateBlock->textureState[stage][WINED3DTSS_BUMPENVLSCALE];
- GLfloat *offset = (GLfloat *) &stateBlock->textureState[stage][WINED3DTSS_BUMPENVLOFFSET];
+ if(prog->luminancescale_location[i] != -1) {
+ const GLfloat *scale = (const GLfloat *)&stateBlock->textureState[i][WINED3DTSS_BUMPENVLSCALE];
+ const GLfloat *offset = (const GLfloat *)&stateBlock->textureState[i][WINED3DTSS_BUMPENVLOFFSET];
GL_EXTCALL(glUniform1fvARB(prog->luminancescale_location[i], 1, scale));
checkGLcall("glUniform1fvARB");
}
}
- if(((IWineD3DPixelShaderImpl *) pshader)->srgb_enabled &&
- !((IWineD3DPixelShaderImpl *) pshader)->srgb_mode_hardcoded) {
- float comparison[4];
- float mul_low[4];
-
- if(stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE]) {
- comparison[0] = srgb_cmp; comparison[1] = srgb_cmp;
- comparison[2] = srgb_cmp; comparison[3] = srgb_cmp;
-
- mul_low[0] = srgb_mul_low; mul_low[1] = srgb_mul_low;
- mul_low[2] = srgb_mul_low; mul_low[3] = srgb_mul_low;
- } else {
- comparison[0] = 1.0 / 0.0; comparison[1] = 1.0 / 0.0;
- comparison[2] = 1.0 / 0.0; comparison[3] = 1.0 / 0.0;
-
- mul_low[0] = 1.0; mul_low[1] = 1.0;
- mul_low[2] = 1.0; mul_low[3] = 1.0;
- }
-
- GL_EXTCALL(glUniform4fvARB(prog->srgb_comparison_location, 1, comparison));
- GL_EXTCALL(glUniform4fvARB(prog->srgb_mul_low_location, 1, mul_low));
- }
if(((IWineD3DPixelShaderImpl *) pshader)->vpos_uniform) {
float correction_params[4];
- if(deviceImpl->render_offscreen) {
- correction_params[0] = 0.0;
- correction_params[1] = 1.0;
+
+ if (context->render_offscreen)
+ {
+ correction_params[0] = 0.0f;
+ correction_params[1] = 1.0f;
} else {
/* position is window relative, not viewport relative */
- correction_params[0] = ((IWineD3DSurfaceImpl *) deviceImpl->render_targets[0])->currentDesc.Height;
- correction_params[1] = -1.0;
+ correction_params[0] = ((IWineD3DSurfaceImpl *)context->current_rt)->currentDesc.Height;
+ correction_params[1] = -1.0f;
}
GL_EXTCALL(glUniform4fvARB(prog->ycorrection_location, 1, correction_params));
}
}
+
+ if (priv->next_constant_version == UINT_MAX)
+ {
+ TRACE("Max constant version reached, resetting to 0.\n");
+ wine_rb_for_each_entry(&priv->program_lookup, reset_program_constant_version, NULL);
+ priv->next_constant_version = 1;
+ }
+ else
+ {
+ prog->constant_version = priv->next_constant_version++;
+ }
}
-/** Generate the variable & register declarations for the GLSL output target */
-static void shader_generate_glsl_declarations(
- IWineD3DBaseShader *iface,
- shader_reg_maps* reg_maps,
- SHADER_BUFFER* buffer,
- WineD3D_GL_Info* gl_info) {
+static inline void update_heap_entry(struct constant_heap *heap, unsigned int idx,
+ unsigned int heap_idx, DWORD new_version)
+{
+ struct constant_entry *entries = heap->entries;
+ unsigned int *positions = heap->positions;
+ unsigned int parent_idx;
+ while (heap_idx > 1)
+ {
+ parent_idx = heap_idx >> 1;
+
+ if (new_version <= entries[parent_idx].version) break;
+
+ entries[heap_idx] = entries[parent_idx];
+ positions[entries[parent_idx].idx] = heap_idx;
+ heap_idx = parent_idx;
+ }
+
+ entries[heap_idx].version = new_version;
+ entries[heap_idx].idx = idx;
+ positions[idx] = heap_idx;
+}
+
+static void shader_glsl_update_float_vertex_constants(IWineD3DDevice *iface, UINT start, UINT count)
+{
+ IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
+ struct shader_glsl_priv *priv = This->shader_priv;
+ struct constant_heap *heap = &priv->vconst_heap;
+ UINT i;
+
+ for (i = start; i < count + start; ++i)
+ {
+ if (!This->stateBlock->changed.vertexShaderConstantsF[i])
+ update_heap_entry(heap, i, heap->size++, priv->next_constant_version);
+ else
+ update_heap_entry(heap, i, heap->positions[i], priv->next_constant_version);
+ }
+}
+
+static void shader_glsl_update_float_pixel_constants(IWineD3DDevice *iface, UINT start, UINT count)
+{
+ IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
+ struct shader_glsl_priv *priv = This->shader_priv;
+ struct constant_heap *heap = &priv->pconst_heap;
+ UINT i;
+
+ for (i = start; i < count + start; ++i)
+ {
+ if (!This->stateBlock->changed.pixelShaderConstantsF[i])
+ update_heap_entry(heap, i, heap->size++, priv->next_constant_version);
+ else
+ update_heap_entry(heap, i, heap->positions[i], priv->next_constant_version);
+ }
+}
+
+static unsigned int vec4_varyings(DWORD shader_major, const struct wined3d_gl_info *gl_info)
+{
+ unsigned int ret = gl_info->limits.glsl_varyings / 4;
+ /* 4.0 shaders do not write clip coords because d3d10 does not support user clipplanes */
+ if(shader_major > 3) return ret;
+
+ /* 3.0 shaders may need an extra varying for the clip coord on some cards(mostly dx10 ones) */
+ if (gl_info->quirks & WINED3D_QUIRK_GLSL_CLIP_VARYING) ret -= 1;
+ return ret;
+}
+
+/** Generate the variable & register declarations for the GLSL output target */
+static void shader_generate_glsl_declarations(const struct wined3d_context *context,
+ struct wined3d_shader_buffer *buffer, IWineD3DBaseShader *iface,
+ const shader_reg_maps *reg_maps, struct shader_glsl_ctx_priv *ctx_priv)
+{
IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) iface;
IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *) This->baseShader.device;
- int i;
- unsigned int extra_constants_needed = 0;
- local_constant* lconst;
+ const struct ps_compile_args *ps_args = ctx_priv->cur_ps_args;
+ const struct wined3d_gl_info *gl_info = context->gl_info;
+ unsigned int i, extra_constants_needed = 0;
+ const local_constant *lconst;
+ DWORD map;
/* There are some minor differences between pixel and vertex shaders */
- char pshader = shader_is_pshader_version(This->baseShader.hex_version);
+ char pshader = shader_is_pshader_version(reg_maps->shader_version.type);
char prefix = pshader ? 'P' : 'V';
/* Prototype the subroutines */
- for (i = 0; i < This->baseShader.limits.label; i++) {
- if (reg_maps->labels[i])
- shader_addline(buffer, "void subroutine%u();\n", i);
+ for (i = 0, map = reg_maps->labels; map; map >>= 1, ++i)
+ {
+ if (map & 1) shader_addline(buffer, "void subroutine%u();\n", i);
}
/* Declare the constants (aka uniforms) */
if (This->baseShader.limits.constant_float > 0) {
- unsigned max_constantsF = min(This->baseShader.limits.constant_float,
- (pshader ? GL_LIMITS(pshader_constantsF) : GL_LIMITS(vshader_constantsF)));
+ unsigned max_constantsF;
+ /* Unless the shader uses indirect addressing, always declare the maximum array size and ignore that we need some
+ * uniforms privately. E.g. if GL supports 256 uniforms, and we need 2 for the pos fixup and immediate values, still
+ * declare VC[256]. If the shader needs more uniforms than we have it won't work in any case. If it uses less, the
+ * compiler will figure out which uniforms are really used and strip them out. This allows a shader to use c255 on
+ * a dx9 card, as long as it doesn't also use all the other constants.
+ *
+ * If the shader uses indirect addressing the compiler must assume that all declared uniforms are used. In this case,
+ * declare only the amount that we're assured to have.
+ *
+ * Thus we run into problems in these two cases:
+ * 1) The shader really uses more uniforms than supported
+ * 2) The shader uses indirect addressing, less constants than supported, but uses a constant index > #supported consts
+ */
+ if (pshader)
+ {
+ /* No indirect addressing here. */
+ max_constantsF = gl_info->limits.glsl_ps_float_constants;
+ }
+ else
+ {
+ if(This->baseShader.reg_maps.usesrelconstF) {
+ /* Subtract the other potential uniforms from the max available (bools, ints, and 1 row of projection matrix).
+ * Subtract another uniform for immediate values, which have to be loaded via uniform by the driver as well.
+ * The shader code only uses 0.5, 2.0, 1.0, 128 and -128 in vertex shader code, so one vec4 should be enough
+ * (Unfortunately the Nvidia driver doesn't store 128 and -128 in one float).
+ *
+ * Writing gl_ClipVertex requires one uniform for each clipplane as well.
+ */
+ max_constantsF = gl_info->limits.glsl_vs_float_constants - 3;
+ if(ctx_priv->cur_vs_args->clip_enabled)
+ {
+ max_constantsF -= gl_info->limits.clipplanes;
+ }
+ max_constantsF -= count_bits(This->baseShader.reg_maps.integer_constants);
+ /* Strictly speaking a bool only uses one scalar, but the nvidia(Linux) compiler doesn't pack them properly,
+ * so each scalar requires a full vec4. We could work around this by packing the booleans ourselves, but
+ * for now take this into account when calculating the number of available constants
+ */
+ max_constantsF -= count_bits(This->baseShader.reg_maps.boolean_constants);
+ /* Set by driver quirks in directx.c */
+ max_constantsF -= gl_info->reserved_glsl_constants;
+ }
+ else
+ {
+ max_constantsF = gl_info->limits.glsl_vs_float_constants;
+ }
+ }
+ max_constantsF = min(This->baseShader.limits.constant_float, max_constantsF);
shader_addline(buffer, "uniform vec4 %cC[%u];\n", prefix, max_constantsF);
}
- if (This->baseShader.limits.constant_int > 0)
+ /* Always declare the full set of constants, the compiler can remove the unused ones because d3d doesn't(yet)
+ * support indirect int and bool constant addressing. This avoids problems if the app uses e.g. i0 and i9.
+ */
+ if (This->baseShader.limits.constant_int > 0 && This->baseShader.reg_maps.integer_constants)
shader_addline(buffer, "uniform ivec4 %cI[%u];\n", prefix, This->baseShader.limits.constant_int);
- if (This->baseShader.limits.constant_bool > 0)
+ if (This->baseShader.limits.constant_bool > 0 && This->baseShader.reg_maps.boolean_constants)
shader_addline(buffer, "uniform bool %cB[%u];\n", prefix, This->baseShader.limits.constant_bool);
if(!pshader) {
* out. The nvidia driver only does that if the parameter is inout instead of out, hence the
* inout.
*/
- if(This->baseShader.hex_version >= WINED3DVS_VERSION(3, 0)) {
+ if (reg_maps->shader_version.major >= 3)
+ {
shader_addline(buffer, "void order_ps_input(in vec4[%u]);\n", MAX_REG_OUTPUT);
} else {
shader_addline(buffer, "void order_ps_input();\n");
}
} else {
- IWineD3DPixelShaderImpl *ps_impl = (IWineD3DPixelShaderImpl *) This;
-
- ps_impl->numbumpenvmatconsts = 0;
- for(i = 0; i < (sizeof(reg_maps->bumpmat) / sizeof(reg_maps->bumpmat[0])); i++) {
- if(!reg_maps->bumpmat[i]) {
- continue;
- }
+ for (i = 0, map = reg_maps->bumpmat; map; map >>= 1, ++i)
+ {
+ if (!(map & 1)) continue;
- ps_impl->bumpenvmatconst[(int) ps_impl->numbumpenvmatconsts].texunit = i;
shader_addline(buffer, "uniform mat2 bumpenvmat%d;\n", i);
- if(reg_maps->luminanceparams) {
- ps_impl->luminanceconst[(int) ps_impl->numbumpenvmatconsts].texunit = i;
+ if (reg_maps->luminanceparams & (1 << i))
+ {
shader_addline(buffer, "uniform float luminancescale%d;\n", i);
shader_addline(buffer, "uniform float luminanceoffset%d;\n", i);
extra_constants_needed++;
- } else {
- ps_impl->luminanceconst[(int) ps_impl->numbumpenvmatconsts].texunit = -1;
}
extra_constants_needed++;
- ps_impl->numbumpenvmatconsts++;
}
- if(device->stateBlock->renderState[WINED3DRS_SRGBWRITEENABLE]) {
- ps_impl->srgb_enabled = 1;
- if(This->baseShader.limits.constant_float + extra_constants_needed + 1 < GL_LIMITS(pshader_constantsF)) {
- shader_addline(buffer, "uniform vec4 srgb_mul_low;\n");
- shader_addline(buffer, "uniform vec4 srgb_comparison;\n");
- ps_impl->srgb_mode_hardcoded = 0;
- extra_constants_needed++;
- } else {
- ps_impl->srgb_mode_hardcoded = 1;
- shader_addline(buffer, "const vec4 srgb_mul_low = vec4(%f, %f, %f, %f);\n",
- srgb_mul_low, srgb_mul_low, srgb_mul_low, srgb_mul_low);
- shader_addline(buffer, "const vec4 srgb_comparison = vec4(%f, %f, %f, %f);\n",
- srgb_cmp, srgb_cmp, srgb_cmp, srgb_cmp);
- }
- } else {
- IWineD3DPixelShaderImpl *ps_impl = (IWineD3DPixelShaderImpl *) This;
-
- /* Do not write any srgb fixup into the shader to save shader size and processing time.
- * As a consequence, we can't toggle srgb write on without recompilation
- */
- ps_impl->srgb_enabled = 0;
- ps_impl->srgb_mode_hardcoded = 1;
+ if (ps_args->srgb_correction)
+ {
+ shader_addline(buffer, "const vec4 srgb_const0 = vec4(%.8e, %.8e, %.8e, %.8e);\n",
+ srgb_pow, srgb_mul_high, srgb_sub_high, srgb_mul_low);
+ shader_addline(buffer, "const vec4 srgb_const1 = vec4(%.8e, 0.0, 0.0, 0.0);\n",
+ srgb_cmp);
}
- if(reg_maps->vpos || reg_maps->usesdsy) {
- if(This->baseShader.limits.constant_float + extra_constants_needed + 1 < GL_LIMITS(pshader_constantsF)) {
+ if (reg_maps->vpos || reg_maps->usesdsy)
+ {
+ if (This->baseShader.limits.constant_float + extra_constants_needed
+ + 1 < gl_info->limits.glsl_ps_float_constants)
+ {
shader_addline(buffer, "uniform vec4 ycorrection;\n");
((IWineD3DPixelShaderImpl *) This)->vpos_uniform = 1;
extra_constants_needed++;
*/
FIXME("Cannot find a free uniform for vpos correction params\n");
shader_addline(buffer, "const vec4 ycorrection = vec4(%f, %f, 0.0, 0.0);\n",
- device->render_offscreen ? 0.0 : ((IWineD3DSurfaceImpl *) device->render_targets[0])->currentDesc.Height,
- device->render_offscreen ? 1.0 : -1.0);
+ context->render_offscreen ? 0.0f : ((IWineD3DSurfaceImpl *)device->render_targets[0])->currentDesc.Height,
+ context->render_offscreen ? 1.0f : -1.0f);
}
shader_addline(buffer, "vec4 vpos;\n");
}
}
- /* Declare texture samplers */
+ /* Declare texture samplers */
for (i = 0; i < This->baseShader.limits.sampler; i++) {
- if (reg_maps->samplers[i]) {
-
- DWORD stype = reg_maps->samplers[i] & WINED3DSP_TEXTURETYPE_MASK;
- switch (stype) {
-
+ if (reg_maps->sampler_type[i])
+ {
+ switch (reg_maps->sampler_type[i])
+ {
case WINED3DSTT_1D:
shader_addline(buffer, "uniform sampler1D %csampler%u;\n", prefix, i);
break;
break;
default:
shader_addline(buffer, "uniform unsupported_sampler %csampler%u;\n", prefix, i);
- FIXME("Unrecognized sampler type: %#x\n", stype);
+ FIXME("Unrecognized sampler type: %#x\n", reg_maps->sampler_type[i]);
break;
}
}
}
-
+
+ /* Declare uniforms for NP2 texcoord fixup:
+ * This is NOT done inside the loop that declares the texture samplers since the NP2 fixup code
+ * is currently only used for the GeforceFX series and when forcing the ARB_npot extension off.
+ * Modern cards just skip the code anyway, so put it inside a separate loop. */
+ if (pshader && ps_args->np2_fixup) {
+
+ struct ps_np2fixup_info* const fixup = ctx_priv->cur_np2fixup_info;
+ UINT cur = 0;
+
+ /* NP2/RECT textures in OpenGL use texcoords in the range [0,width]x[0,height]
+ * while D3D has them in the (normalized) [0,1]x[0,1] range.
+ * samplerNP2Fixup stores texture dimensions and is updated through
+ * shader_glsl_load_np2fixup_constants when the sampler changes. */
+
+ for (i = 0; i < This->baseShader.limits.sampler; ++i) {
+ if (reg_maps->sampler_type[i]) {
+ if (!(ps_args->np2_fixup & (1 << i))) continue;
+
+ if (WINED3DSTT_2D != reg_maps->sampler_type[i]) {
+ FIXME("Non-2D texture is flagged for NP2 texcoord fixup.\n");
+ continue;
+ }
+
+ fixup->idx[i] = cur++;
+ }
+ }
+
+ fixup->num_consts = (cur + 1) >> 1;
+ shader_addline(buffer, "uniform vec4 %csamplerNP2Fixup[%u];\n", prefix, fixup->num_consts);
+ }
+
/* Declare address variables */
- for (i = 0; i < This->baseShader.limits.address; i++) {
- if (reg_maps->address[i])
- shader_addline(buffer, "ivec4 A%d;\n", i);
+ for (i = 0, map = reg_maps->address; map; map >>= 1, ++i)
+ {
+ if (map & 1) shader_addline(buffer, "ivec4 A%u;\n", i);
}
/* Declare texture coordinate temporaries and initialize them */
- for (i = 0; i < This->baseShader.limits.texcoord; i++) {
- if (reg_maps->texcoord[i])
- shader_addline(buffer, "vec4 T%u = gl_TexCoord[%u];\n", i, i);
+ for (i = 0, map = reg_maps->texcoord; map; map >>= 1, ++i)
+ {
+ if (map & 1) shader_addline(buffer, "vec4 T%u = gl_TexCoord[%u];\n", i, i);
}
/* Declare input register varyings. Only pixel shader, vertex shaders have that declared in the
* helper function shader that is linked in at link time
*/
- if(pshader && This->baseShader.hex_version >= WINED3DPS_VERSION(3, 0)) {
- if(use_vs(device)) {
- shader_addline(buffer, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings) / 4);
+ if (pshader && reg_maps->shader_version.major >= 3)
+ {
+ if (use_vs(device->stateBlock))
+ {
+ shader_addline(buffer, "varying vec4 IN[%u];\n", vec4_varyings(reg_maps->shader_version.major, gl_info));
} else {
/* TODO: Write a replacement shader for the fixed function vertex pipeline, so this isn't needed.
* For fixed function vertex processing + 3.0 pixel shader we need a separate function in the
* pixel shader that reads the fixed function color into the packed input registers.
*/
- shader_addline(buffer, "vec4 IN[%u];\n", GL_LIMITS(glsl_varyings) / 4);
+ shader_addline(buffer, "vec4 IN[%u];\n", vec4_varyings(reg_maps->shader_version.major, gl_info));
}
}
}
/* Declare temporary variables */
- for(i = 0; i < This->baseShader.limits.temporary; i++) {
- if (reg_maps->temporary[i])
- shader_addline(buffer, "vec4 R%u;\n", i);
+ for (i = 0, map = reg_maps->temporary; map; map >>= 1, ++i)
+ {
+ if (map & 1) shader_addline(buffer, "vec4 R%u;\n", i);
}
/* Declare attributes */
- for (i = 0; i < This->baseShader.limits.attributes; i++) {
- if (reg_maps->attributes[i])
- shader_addline(buffer, "attribute vec4 attrib%i;\n", i);
+ if (reg_maps->shader_version.type == WINED3D_SHADER_TYPE_VERTEX)
+ {
+ for (i = 0, map = reg_maps->input_registers; map; map >>= 1, ++i)
+ {
+ if (map & 1) shader_addline(buffer, "attribute vec4 attrib%i;\n", i);
+ }
}
/* Declare loop registers aLx */
****************************************************************************/
/* Prototypes */
-static void shader_glsl_add_src_param(SHADER_OPCODE_ARG* arg, const DWORD param,
- const DWORD addr_token, DWORD mask, glsl_src_param_t *src_param);
+static void shader_glsl_add_src_param(const struct wined3d_shader_instruction *ins,
+ const struct wined3d_shader_src_param *wined3d_src, DWORD mask, glsl_src_param_t *glsl_src);
/** Used for opcode modifiers - They multiply the result by the specified amount */
static const char * const shift_glsl_tab[] = {
- "", /* 0 (none) */
- "2.0 * ", /* 1 (x2) */
- "4.0 * ", /* 2 (x4) */
- "8.0 * ", /* 3 (x8) */
- "16.0 * ", /* 4 (x16) */
- "32.0 * ", /* 5 (x32) */
- "", /* 6 (x64) */
- "", /* 7 (x128) */
- "", /* 8 (d256) */
- "", /* 9 (d128) */
- "", /* 10 (d64) */
- "", /* 11 (d32) */
- "0.0625 * ", /* 12 (d16) */
- "0.125 * ", /* 13 (d8) */
- "0.25 * ", /* 14 (d4) */
- "0.5 * " /* 15 (d2) */
+ "", /* 0 (none) */
+ "2.0 * ", /* 1 (x2) */
+ "4.0 * ", /* 2 (x4) */
+ "8.0 * ", /* 3 (x8) */
+ "16.0 * ", /* 4 (x16) */
+ "32.0 * ", /* 5 (x32) */
+ "", /* 6 (x64) */
+ "", /* 7 (x128) */
+ "", /* 8 (d256) */
+ "", /* 9 (d128) */
+ "", /* 10 (d64) */
+ "", /* 11 (d32) */
+ "0.0625 * ", /* 12 (d16) */
+ "0.125 * ", /* 13 (d8) */
+ "0.25 * ", /* 14 (d4) */
+ "0.5 * " /* 15 (d2) */
};
/* Generate a GLSL parameter that does the input modifier computation and return the input register/mask to use */
-static void shader_glsl_gen_modifier (
- const DWORD instr,
- const char *in_reg,
- const char *in_regswizzle,
- char *out_str) {
-
+static void shader_glsl_gen_modifier(DWORD src_modifier, const char *in_reg, const char *in_regswizzle, char *out_str)
+{
out_str[0] = 0;
-
- if (instr == WINED3DSIO_TEXKILL)
- return;
- switch (instr & WINED3DSP_SRCMOD_MASK) {
+ switch (src_modifier)
+ {
case WINED3DSPSM_DZ: /* Need to handle this in the instructions itself (texld & texcrd). */
case WINED3DSPSM_DW:
case WINED3DSPSM_NONE:
sprintf(out_str, "-abs(%s%s)", in_reg, in_regswizzle);
break;
default:
- FIXME("Unhandled modifier %u\n", (instr & WINED3DSP_SRCMOD_MASK));
+ FIXME("Unhandled modifier %u\n", src_modifier);
sprintf(out_str, "%s%s", in_reg, in_regswizzle);
}
}
/** Writes the GLSL variable name that corresponds to the register that the
* DX opcode parameter is trying to access */
-static void shader_glsl_get_register_name(
- const DWORD param,
- const DWORD addr_token,
- char* regstr,
- BOOL* is_color,
- SHADER_OPCODE_ARG* arg) {
-
+static void shader_glsl_get_register_name(const struct wined3d_shader_register *reg,
+ char *register_name, BOOL *is_color, const struct wined3d_shader_instruction *ins)
+{
/* oPos, oFog and oPts in D3D */
static const char * const hwrastout_reg_names[] = { "gl_Position", "gl_FogFragCoord", "gl_PointSize" };
- DWORD reg = param & WINED3DSP_REGNUM_MASK;
- DWORD regtype = shader_get_regtype(param);
- IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*) arg->shader;
+ IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *)ins->ctx->shader;
IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
- WineD3D_GL_Info* gl_info = &deviceImpl->adapter->gl_info;
-
- char pshader = shader_is_pshader_version(This->baseShader.hex_version);
- char tmpStr[150];
-
- *is_color = FALSE;
-
- switch (regtype) {
- case WINED3DSPR_TEMP:
- sprintf(tmpStr, "R%u", reg);
- break;
- case WINED3DSPR_INPUT:
- if (pshader) {
- /* Pixel shaders >= 3.0 */
- if (WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) >= 3) {
- DWORD in_count = GL_LIMITS(glsl_varyings) / 4;
-
- if (param & WINED3DSHADER_ADDRMODE_RELATIVE) {
+ const struct wined3d_gl_info *gl_info = &deviceImpl->adapter->gl_info;
+ char pshader = shader_is_pshader_version(This->baseShader.reg_maps.shader_version.type);
+
+ *is_color = FALSE;
+
+ switch (reg->type)
+ {
+ case WINED3DSPR_TEMP:
+ sprintf(register_name, "R%u", reg->idx);
+ break;
+
+ case WINED3DSPR_INPUT:
+ /* vertex shaders */
+ if (!pshader)
+ {
+ struct shader_glsl_ctx_priv *priv = ins->ctx->backend_data;
+ if (priv->cur_vs_args->swizzle_map & (1 << reg->idx)) *is_color = TRUE;
+ sprintf(register_name, "attrib%u", reg->idx);
+ break;
+ }
+
+ /* pixel shaders >= 3.0 */
+ if (This->baseShader.reg_maps.shader_version.major >= 3)
+ {
+ DWORD idx = ((IWineD3DPixelShaderImpl *)This)->input_reg_map[reg->idx];
+ unsigned int in_count = vec4_varyings(This->baseShader.reg_maps.shader_version.major, gl_info);
+
+ if (reg->rel_addr)
+ {
glsl_src_param_t rel_param;
- shader_glsl_add_src_param(arg, addr_token, 0, WINED3DSP_WRITEMASK_0, &rel_param);
+
+ shader_glsl_add_src_param(ins, reg->rel_addr, WINED3DSP_WRITEMASK_0, &rel_param);
/* Removing a + 0 would be an obvious optimization, but macos doesn't see the NOP
- * operation there
- */
- if(((IWineD3DPixelShaderImpl *) This)->input_reg_map[reg]) {
- if (((IWineD3DPixelShaderImpl *)This)->declared_in_count > in_count) {
- sprintf(tmpStr, "((%s + %u) > %d ? (%s + %u) > %d ? gl_SecondaryColor : gl_Color : IN[%s + %u])",
- rel_param.param_str, ((IWineD3DPixelShaderImpl *)This)->input_reg_map[reg], in_count - 1,
- rel_param.param_str, ((IWineD3DPixelShaderImpl *)This)->input_reg_map[reg], in_count,
- rel_param.param_str, ((IWineD3DPixelShaderImpl *)This)->input_reg_map[reg]);
- } else {
- sprintf(tmpStr, "IN[%s + %u]", rel_param.param_str, ((IWineD3DPixelShaderImpl *)This)->input_reg_map[reg]);
+ * operation there */
+ if (idx)
+ {
+ if (((IWineD3DPixelShaderImpl *)This)->declared_in_count > in_count)
+ {
+ sprintf(register_name,
+ "((%s + %u) > %d ? (%s + %u) > %d ? gl_SecondaryColor : gl_Color : IN[%s + %u])",
+ rel_param.param_str, idx, in_count - 1, rel_param.param_str, idx, in_count,
+ rel_param.param_str, idx);
}
- } else {
- if (((IWineD3DPixelShaderImpl *)This)->declared_in_count > in_count) {
- sprintf(tmpStr, "((%s) > %d ? (%s) > %d ? gl_SecondaryColor : gl_Color : IN[%s])",
- rel_param.param_str, in_count - 1,
- rel_param.param_str, in_count,
- rel_param.param_str);
- } else {
- sprintf(tmpStr, "IN[%s]", rel_param.param_str);
+ else
+ {
+ sprintf(register_name, "IN[%s + %u]", rel_param.param_str, idx);
}
}
- } else {
- DWORD idx = ((IWineD3DPixelShaderImpl *) This)->input_reg_map[reg];
- if (idx == in_count) {
- sprintf(tmpStr, "gl_Color");
- } else if (idx == in_count + 1) {
- sprintf(tmpStr, "gl_SecondaryColor");
- } else {
- sprintf(tmpStr, "IN[%u]", idx);
+ else
+ {
+ if (((IWineD3DPixelShaderImpl *)This)->declared_in_count > in_count)
+ {
+ sprintf(register_name, "((%s) > %d ? (%s) > %d ? gl_SecondaryColor : gl_Color : IN[%s])",
+ rel_param.param_str, in_count - 1, rel_param.param_str, in_count,
+ rel_param.param_str);
+ }
+ else
+ {
+ sprintf(register_name, "IN[%s]", rel_param.param_str);
+ }
}
}
- } else {
- if (reg==0)
- strcpy(tmpStr, "gl_Color");
else
- strcpy(tmpStr, "gl_SecondaryColor");
+ {
+ if (idx == in_count) sprintf(register_name, "gl_Color");
+ else if (idx == in_count + 1) sprintf(register_name, "gl_SecondaryColor");
+ else sprintf(register_name, "IN[%u]", idx);
+ }
}
- } else {
- if (vshader_input_is_color((IWineD3DVertexShader*) This, reg))
- *is_color = TRUE;
- sprintf(tmpStr, "attrib%u", reg);
- }
- break;
- case WINED3DSPR_CONST:
- {
- const char prefix = pshader? 'P':'V';
-
- /* Relative addressing */
- if (param & WINED3DSHADER_ADDRMODE_RELATIVE) {
-
- /* Relative addressing on shaders 2.0+ have a relative address token,
- * prior to that, it was hard-coded as "A0.x" because there's only 1 register */
- if (WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) >= 2) {
- glsl_src_param_t rel_param;
- shader_glsl_add_src_param(arg, addr_token, 0, WINED3DSP_WRITEMASK_0, &rel_param);
- if(reg) {
- sprintf(tmpStr, "%cC[%s + %u]", prefix, rel_param.param_str, reg);
- } else {
- sprintf(tmpStr, "%cC[%s]", prefix, rel_param.param_str);
- }
- } else {
- if(reg) {
- sprintf(tmpStr, "%cC[A0.x + %u]", prefix, reg);
- } else {
- sprintf(tmpStr, "%cC[A0.x]", prefix);
- }
- }
+ else
+ {
+ if (reg->idx == 0) strcpy(register_name, "gl_Color");
+ else strcpy(register_name, "gl_SecondaryColor");
+ break;
+ }
+ break;
- } else {
- if(shader_constant_is_local(This, reg)) {
- sprintf(tmpStr, "%cLC%u", prefix, reg);
- } else {
- sprintf(tmpStr, "%cC[%u]", prefix, reg);
+ case WINED3DSPR_CONST:
+ {
+ const char prefix = pshader ? 'P' : 'V';
+
+ /* Relative addressing */
+ if (reg->rel_addr)
+ {
+ glsl_src_param_t rel_param;
+ shader_glsl_add_src_param(ins, reg->rel_addr, WINED3DSP_WRITEMASK_0, &rel_param);
+ if (reg->idx) sprintf(register_name, "%cC[%s + %u]", prefix, rel_param.param_str, reg->idx);
+ else sprintf(register_name, "%cC[%s]", prefix, rel_param.param_str);
+ }
+ else
+ {
+ if (shader_constant_is_local(This, reg->idx))
+ sprintf(register_name, "%cLC%u", prefix, reg->idx);
+ else
+ sprintf(register_name, "%cC[%u]", prefix, reg->idx);
+ }
}
- }
+ break;
- break;
- }
- case WINED3DSPR_CONSTINT:
- if (pshader)
- sprintf(tmpStr, "PI[%u]", reg);
- else
- sprintf(tmpStr, "VI[%u]", reg);
- break;
- case WINED3DSPR_CONSTBOOL:
- if (pshader)
- sprintf(tmpStr, "PB[%u]", reg);
- else
- sprintf(tmpStr, "VB[%u]", reg);
- break;
- case WINED3DSPR_TEXTURE: /* case WINED3DSPR_ADDR: */
- if (pshader) {
- sprintf(tmpStr, "T%u", reg);
- } else {
- sprintf(tmpStr, "A%u", reg);
- }
- break;
- case WINED3DSPR_LOOP:
- sprintf(tmpStr, "aL%u", This->baseShader.cur_loop_regno - 1);
- break;
- case WINED3DSPR_SAMPLER:
- if (pshader)
- sprintf(tmpStr, "Psampler%u", reg);
- else
- sprintf(tmpStr, "Vsampler%u", reg);
- break;
- case WINED3DSPR_COLOROUT:
- if (reg >= GL_LIMITS(buffers)) {
- WARN("Write to render target %u, only %d supported\n", reg, 4);
- }
- if (GL_SUPPORT(ARB_DRAW_BUFFERS)) {
- sprintf(tmpStr, "gl_FragData[%u]", reg);
- } else { /* On older cards with GLSL support like the GeforceFX there's only one buffer. */
- sprintf(tmpStr, "gl_FragColor");
- }
- break;
- case WINED3DSPR_RASTOUT:
- sprintf(tmpStr, "%s", hwrastout_reg_names[reg]);
- break;
- case WINED3DSPR_DEPTHOUT:
- sprintf(tmpStr, "gl_FragDepth");
- break;
- case WINED3DSPR_ATTROUT:
- if (reg == 0) {
- sprintf(tmpStr, "gl_FrontColor");
- } else {
- sprintf(tmpStr, "gl_FrontSecondaryColor");
- }
- break;
- case WINED3DSPR_TEXCRDOUT:
- /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
- if (WINED3DSHADER_VERSION_MAJOR(This->baseShader.hex_version) >= 3)
- sprintf(tmpStr, "OUT[%u]", reg);
- else
- sprintf(tmpStr, "gl_TexCoord[%u]", reg);
- break;
- case WINED3DSPR_MISCTYPE:
- if (reg == 0) {
- /* vPos */
- sprintf(tmpStr, "vpos");
- } else if (reg == 1){
- /* Note that gl_FrontFacing is a bool, while vFace is
- * a float for which the sign determines front/back
- */
- sprintf(tmpStr, "(gl_FrontFacing ? 1.0 : -1.0)");
- } else {
- FIXME("Unhandled misctype register %d\n", reg);
- sprintf(tmpStr, "unrecognized_register");
- }
- break;
- default:
- FIXME("Unhandled register name Type(%d)\n", regtype);
- sprintf(tmpStr, "unrecognized_register");
- break;
+ case WINED3DSPR_CONSTINT:
+ if (pshader) sprintf(register_name, "PI[%u]", reg->idx);
+ else sprintf(register_name, "VI[%u]", reg->idx);
+ break;
+
+ case WINED3DSPR_CONSTBOOL:
+ if (pshader) sprintf(register_name, "PB[%u]", reg->idx);
+ else sprintf(register_name, "VB[%u]", reg->idx);
+ break;
+
+ case WINED3DSPR_TEXTURE: /* case WINED3DSPR_ADDR: */
+ if (pshader) sprintf(register_name, "T%u", reg->idx);
+ else sprintf(register_name, "A%u", reg->idx);
+ break;
+
+ case WINED3DSPR_LOOP:
+ sprintf(register_name, "aL%u", This->baseShader.cur_loop_regno - 1);
+ break;
+
+ case WINED3DSPR_SAMPLER:
+ if (pshader) sprintf(register_name, "Psampler%u", reg->idx);
+ else sprintf(register_name, "Vsampler%u", reg->idx);
+ break;
+
+ case WINED3DSPR_COLOROUT:
+ if (reg->idx >= gl_info->limits.buffers)
+ WARN("Write to render target %u, only %d supported.\n", reg->idx, gl_info->limits.buffers);
+
+ sprintf(register_name, "gl_FragData[%u]", reg->idx);
+ break;
+
+ case WINED3DSPR_RASTOUT:
+ sprintf(register_name, "%s", hwrastout_reg_names[reg->idx]);
+ break;
+
+ case WINED3DSPR_DEPTHOUT:
+ sprintf(register_name, "gl_FragDepth");
+ break;
+
+ case WINED3DSPR_ATTROUT:
+ if (reg->idx == 0) sprintf(register_name, "gl_FrontColor");
+ else sprintf(register_name, "gl_FrontSecondaryColor");
+ break;
+
+ case WINED3DSPR_TEXCRDOUT:
+ /* Vertex shaders >= 3.0: WINED3DSPR_OUTPUT */
+ if (This->baseShader.reg_maps.shader_version.major >= 3) sprintf(register_name, "OUT[%u]", reg->idx);
+ else sprintf(register_name, "gl_TexCoord[%u]", reg->idx);
+ break;
+
+ case WINED3DSPR_MISCTYPE:
+ if (reg->idx == 0)
+ {
+ /* vPos */
+ sprintf(register_name, "vpos");
+ }
+ else if (reg->idx == 1)
+ {
+ /* Note that gl_FrontFacing is a bool, while vFace is
+ * a float for which the sign determines front/back */
+ sprintf(register_name, "(gl_FrontFacing ? 1.0 : -1.0)");
+ }
+ else
+ {
+ FIXME("Unhandled misctype register %d\n", reg->idx);
+ sprintf(register_name, "unrecognized_register");
+ }
+ break;
+
+ case WINED3DSPR_IMMCONST:
+ switch (reg->immconst_type)
+ {
+ case WINED3D_IMMCONST_FLOAT:
+ sprintf(register_name, "%.8e", *(const float *)reg->immconst_data);
+ break;
+
+ case WINED3D_IMMCONST_FLOAT4:
+ sprintf(register_name, "vec4(%.8e, %.8e, %.8e, %.8e)",
+ *(const float *)®->immconst_data[0], *(const float *)®->immconst_data[1],
+ *(const float *)®->immconst_data[2], *(const float *)®->immconst_data[3]);
+ break;
+
+ default:
+ FIXME("Unhandled immconst type %#x\n", reg->immconst_type);
+ sprintf(register_name, "<unhandled_immconst_type %#x>", reg->immconst_type);
+ }
+ break;
+
+ default:
+ FIXME("Unhandled register name Type(%d)\n", reg->type);
+ sprintf(register_name, "unrecognized_register");
+ break;
}
+}
- strcat(regstr, tmpStr);
+static void shader_glsl_write_mask_to_str(DWORD write_mask, char *str)
+{
+ *str++ = '.';
+ if (write_mask & WINED3DSP_WRITEMASK_0) *str++ = 'x';
+ if (write_mask & WINED3DSP_WRITEMASK_1) *str++ = 'y';
+ if (write_mask & WINED3DSP_WRITEMASK_2) *str++ = 'z';
+ if (write_mask & WINED3DSP_WRITEMASK_3) *str++ = 'w';
+ *str = '\0';
}
/* Get the GLSL write mask for the destination register */
-static DWORD shader_glsl_get_write_mask(const DWORD param, char *write_mask) {
- char *ptr = write_mask;
- DWORD mask = param & WINED3DSP_WRITEMASK_ALL;
+static DWORD shader_glsl_get_write_mask(const struct wined3d_shader_dst_param *param, char *write_mask)
+{
+ DWORD mask = param->write_mask;
- if (shader_is_scalar(param)) {
+ if (shader_is_scalar(¶m->reg))
+ {
mask = WINED3DSP_WRITEMASK_0;
- } else {
- *ptr++ = '.';
- if (param & WINED3DSP_WRITEMASK_0) *ptr++ = 'x';
- if (param & WINED3DSP_WRITEMASK_1) *ptr++ = 'y';
- if (param & WINED3DSP_WRITEMASK_2) *ptr++ = 'z';
- if (param & WINED3DSP_WRITEMASK_3) *ptr++ = 'w';
+ *write_mask = '\0';
+ }
+ else
+ {
+ shader_glsl_write_mask_to_str(mask, write_mask);
}
-
- *ptr = '\0';
return mask;
}
return size;
}
-static void shader_glsl_get_swizzle(const DWORD param, BOOL fixup, DWORD mask, char *swizzle_str) {
+static void shader_glsl_swizzle_to_str(const DWORD swizzle, BOOL fixup, DWORD mask, char *str)
+{
/* For registers of type WINED3DDECLTYPE_D3DCOLOR, data is stored as "bgra",
* but addressed as "rgba". To fix this we need to swap the register's x
* and z components. */
- DWORD swizzle = (param & WINED3DSP_SWIZZLE_MASK) >> WINED3DSP_SWIZZLE_SHIFT;
const char *swizzle_chars = fixup ? "zyxw" : "xyzw";
- char *ptr = swizzle_str;
- if (!shader_is_scalar(param)) {
- *ptr++ = '.';
- /* swizzle bits fields: wwzzyyxx */
- if (mask & WINED3DSP_WRITEMASK_0) *ptr++ = swizzle_chars[swizzle & 0x03];
- if (mask & WINED3DSP_WRITEMASK_1) *ptr++ = swizzle_chars[(swizzle >> 2) & 0x03];
- if (mask & WINED3DSP_WRITEMASK_2) *ptr++ = swizzle_chars[(swizzle >> 4) & 0x03];
- if (mask & WINED3DSP_WRITEMASK_3) *ptr++ = swizzle_chars[(swizzle >> 6) & 0x03];
- }
+ *str++ = '.';
+ /* swizzle bits fields: wwzzyyxx */
+ if (mask & WINED3DSP_WRITEMASK_0) *str++ = swizzle_chars[swizzle & 0x03];
+ if (mask & WINED3DSP_WRITEMASK_1) *str++ = swizzle_chars[(swizzle >> 2) & 0x03];
+ if (mask & WINED3DSP_WRITEMASK_2) *str++ = swizzle_chars[(swizzle >> 4) & 0x03];
+ if (mask & WINED3DSP_WRITEMASK_3) *str++ = swizzle_chars[(swizzle >> 6) & 0x03];
+ *str = '\0';
+}
- *ptr = '\0';
+static void shader_glsl_get_swizzle(const struct wined3d_shader_src_param *param,
+ BOOL fixup, DWORD mask, char *swizzle_str)
+{
+ if (shader_is_scalar(¶m->reg))
+ *swizzle_str = '\0';
+ else
+ shader_glsl_swizzle_to_str(param->swizzle, fixup, mask, swizzle_str);
}
/* From a given parameter token, generate the corresponding GLSL string.
* Also, return the actual register name and swizzle in case the
* caller needs this information as well. */
-static void shader_glsl_add_src_param(SHADER_OPCODE_ARG* arg, const DWORD param,
- const DWORD addr_token, DWORD mask, glsl_src_param_t *src_param) {
+static void shader_glsl_add_src_param(const struct wined3d_shader_instruction *ins,
+ const struct wined3d_shader_src_param *wined3d_src, DWORD mask, glsl_src_param_t *glsl_src)
+{
BOOL is_color = FALSE;
char swizzle_str[6];
- src_param->reg_name[0] = '\0';
- src_param->param_str[0] = '\0';
+ glsl_src->reg_name[0] = '\0';
+ glsl_src->param_str[0] = '\0';
swizzle_str[0] = '\0';
- shader_glsl_get_register_name(param, addr_token, src_param->reg_name, &is_color, arg);
-
- shader_glsl_get_swizzle(param, is_color, mask, swizzle_str);
- shader_glsl_gen_modifier(param, src_param->reg_name, swizzle_str, src_param->param_str);
+ shader_glsl_get_register_name(&wined3d_src->reg, glsl_src->reg_name, &is_color, ins);
+ shader_glsl_get_swizzle(wined3d_src, is_color, mask, swizzle_str);
+ shader_glsl_gen_modifier(wined3d_src->modifiers, glsl_src->reg_name, swizzle_str, glsl_src->param_str);
}
/* From a given parameter token, generate the corresponding GLSL string.
* Also, return the actual register name and swizzle in case the
* caller needs this information as well. */
-static DWORD shader_glsl_add_dst_param(SHADER_OPCODE_ARG* arg, const DWORD param,
- const DWORD addr_token, glsl_dst_param_t *dst_param) {
+static DWORD shader_glsl_add_dst_param(const struct wined3d_shader_instruction *ins,
+ const struct wined3d_shader_dst_param *wined3d_dst, glsl_dst_param_t *glsl_dst)
+{
BOOL is_color = FALSE;
- dst_param->mask_str[0] = '\0';
- dst_param->reg_name[0] = '\0';
+ glsl_dst->mask_str[0] = '\0';
+ glsl_dst->reg_name[0] = '\0';
- shader_glsl_get_register_name(param, addr_token, dst_param->reg_name, &is_color, arg);
- return shader_glsl_get_write_mask(param, dst_param->mask_str);
+ shader_glsl_get_register_name(&wined3d_dst->reg, glsl_dst->reg_name, &is_color, ins);
+ return shader_glsl_get_write_mask(wined3d_dst, glsl_dst->mask_str);
}
/* Append the destination part of the instruction to the buffer, return the effective write mask */
-static DWORD shader_glsl_append_dst_ext(SHADER_BUFFER *buffer, SHADER_OPCODE_ARG *arg, const DWORD param) {
- glsl_dst_param_t dst_param;
+static DWORD shader_glsl_append_dst_ext(struct wined3d_shader_buffer *buffer,
+ const struct wined3d_shader_instruction *ins, const struct wined3d_shader_dst_param *dst)
+{
+ glsl_dst_param_t glsl_dst;
DWORD mask;
- int shift;
-
- mask = shader_glsl_add_dst_param(arg, param, arg->dst_addr, &dst_param);
- if(mask) {
- shift = (param & WINED3DSP_DSTSHIFT_MASK) >> WINED3DSP_DSTSHIFT_SHIFT;
- shader_addline(buffer, "%s%s = %s(", dst_param.reg_name, dst_param.mask_str, shift_glsl_tab[shift]);
- }
+ mask = shader_glsl_add_dst_param(ins, dst, &glsl_dst);
+ if (mask) shader_addline(buffer, "%s%s = %s(", glsl_dst.reg_name, glsl_dst.mask_str, shift_glsl_tab[dst->shift]);
return mask;
}
/* Append the destination part of the instruction to the buffer, return the effective write mask */
-static DWORD shader_glsl_append_dst(SHADER_BUFFER *buffer, SHADER_OPCODE_ARG *arg) {
- return shader_glsl_append_dst_ext(buffer, arg, arg->dst);
+static DWORD shader_glsl_append_dst(struct wined3d_shader_buffer *buffer, const struct wined3d_shader_instruction *ins)
+{
+ return shader_glsl_append_dst_ext(buffer, ins, &ins->dst[0]);
}
/** Process GLSL instruction modifiers */
-void shader_glsl_add_instruction_modifiers(SHADER_OPCODE_ARG* arg) {
-
- DWORD mask = arg->dst & WINED3DSP_DSTMOD_MASK;
-
- if (arg->opcode->dst_token && mask != 0) {
- glsl_dst_param_t dst_param;
+static void shader_glsl_add_instruction_modifiers(const struct wined3d_shader_instruction *ins)
+{
+ glsl_dst_param_t dst_param;
+ DWORD modifiers;
- shader_glsl_add_dst_param(arg, arg->dst, 0, &dst_param);
+ if (!ins->dst_count) return;
- if (mask & WINED3DSPDM_SATURATE) {
- /* _SAT means to clamp the value of the register to between 0 and 1 */
- shader_addline(arg->buffer, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param.reg_name,
- dst_param.mask_str, dst_param.reg_name, dst_param.mask_str);
- }
- if (mask & WINED3DSPDM_MSAMPCENTROID) {
- FIXME("_centroid modifier not handled\n");
- }
- if (mask & WINED3DSPDM_PARTIALPRECISION) {
- /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
- }
+ modifiers = ins->dst[0].modifiers;
+ if (!modifiers) return;
+
+ shader_glsl_add_dst_param(ins, &ins->dst[0], &dst_param);
+
+ if (modifiers & WINED3DSPDM_SATURATE)
+ {
+ /* _SAT means to clamp the value of the register to between 0 and 1 */
+ shader_addline(ins->ctx->buffer, "%s%s = clamp(%s%s, 0.0, 1.0);\n", dst_param.reg_name,
+ dst_param.mask_str, dst_param.reg_name, dst_param.mask_str);
}
-}
-static inline const char* shader_get_comp_op(
- const DWORD opcode) {
+ if (modifiers & WINED3DSPDM_MSAMPCENTROID)
+ {
+ FIXME("_centroid modifier not handled\n");
+ }
+
+ if (modifiers & WINED3DSPDM_PARTIALPRECISION)
+ {
+ /* MSDN says this modifier can be safely ignored, so that's what we'll do. */
+ }
+}
- DWORD op = (opcode & INST_CONTROLS_MASK) >> INST_CONTROLS_SHIFT;
+static inline const char *shader_get_comp_op(DWORD op)
+{
switch (op) {
case COMPARISON_GT: return ">";
case COMPARISON_EQ: return "==";
}
}
-static void shader_glsl_get_sample_function(DWORD sampler_type, BOOL projected, BOOL texrect, glsl_sample_function_t *sample_function) {
+static void shader_glsl_get_sample_function(DWORD sampler_type, DWORD flags, glsl_sample_function_t *sample_function)
+{
+ BOOL projected = flags & WINED3D_GLSL_SAMPLE_PROJECTED;
+ BOOL texrect = flags & WINED3D_GLSL_SAMPLE_RECT;
+ BOOL lod = flags & WINED3D_GLSL_SAMPLE_LOD;
+ BOOL grad = flags & WINED3D_GLSL_SAMPLE_GRAD;
+
/* Note that there's no such thing as a projected cube texture. */
switch(sampler_type) {
case WINED3DSTT_1D:
- sample_function->name = projected ? "texture1DProj" : "texture1D";
+ if(lod) {
+ sample_function->name = projected ? "texture1DProjLod" : "texture1DLod";
+ } else if(grad) {
+ sample_function->name = projected ? "texture1DProjGradARB" : "texture1DGradARB";
+ } else {
+ sample_function->name = projected ? "texture1DProj" : "texture1D";
+ }
sample_function->coord_mask = WINED3DSP_WRITEMASK_0;
break;
case WINED3DSTT_2D:
if(texrect) {
- sample_function->name = projected ? "texture2DRectProj" : "texture2DRect";
+ if(lod) {
+ sample_function->name = projected ? "texture2DRectProjLod" : "texture2DRectLod";
+ } else if(grad) {
+ /* What good are texrect grad functions? I don't know, but GL_EXT_gpu_shader4 defines them.
+ * There is no GL_ARB_shader_texture_lod spec yet, so I don't know if they're defined there
+ */
+ sample_function->name = projected ? "shadow2DRectProjGradARB" : "shadow2DRectGradARB";
+ } else {
+ sample_function->name = projected ? "texture2DRectProj" : "texture2DRect";
+ }
} else {
- sample_function->name = projected ? "texture2DProj" : "texture2D";
+ if(lod) {
+ sample_function->name = projected ? "texture2DProjLod" : "texture2DLod";
+ } else if(grad) {
+ sample_function->name = projected ? "texture2DProjGradARB" : "texture2DGradARB";
+ } else {
+ sample_function->name = projected ? "texture2DProj" : "texture2D";
+ }
}
sample_function->coord_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1;
break;
case WINED3DSTT_CUBE:
- sample_function->name = "textureCube";
+ if(lod) {
+ sample_function->name = "textureCubeLod";
+ } else if(grad) {
+ sample_function->name = "textureCubeGradARB";
+ } else {
+ sample_function->name = "textureCube";
+ }
sample_function->coord_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
break;
case WINED3DSTT_VOLUME:
- sample_function->name = projected ? "texture3DProj" : "texture3D";
+ if(lod) {
+ sample_function->name = projected ? "texture3DProjLod" : "texture3DLod";
+ } else if(grad) {
+ sample_function->name = projected ? "texture3DProjGradARB" : "texture3DGradARB";
+ } else {
+ sample_function->name = projected ? "texture3DProj" : "texture3D";
+ }
sample_function->coord_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
break;
default:
sample_function->name = "";
+ sample_function->coord_mask = 0;
FIXME("Unrecognized sampler type: %#x;\n", sampler_type);
break;
}
}
-static void shader_glsl_color_correction(SHADER_OPCODE_ARG* arg) {
- IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
- IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) shader->baseShader.device;
- WineD3D_GL_Info *gl_info = &deviceImpl->adapter->gl_info;
- glsl_dst_param_t dst_param;
- glsl_dst_param_t dst_param2;
- WINED3DFORMAT fmt;
- WINED3DFORMAT conversion_group;
- IWineD3DBaseTextureImpl *texture;
- DWORD mask, mask_size;
- UINT i;
- BOOL recorded = FALSE;
- DWORD sampler_idx;
- DWORD hex_version = shader->baseShader.hex_version;
+static void shader_glsl_append_fixup_arg(char *arguments, const char *reg_name,
+ BOOL sign_fixup, enum fixup_channel_source channel_source)
+{
+ switch(channel_source)
+ {
+ case CHANNEL_SOURCE_ZERO:
+ strcat(arguments, "0.0");
+ break;
- switch(arg->opcode->opcode) {
- case WINED3DSIO_TEX:
- if (hex_version < WINED3DPS_VERSION(2,0)) {
- sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
- } else {
- sampler_idx = arg->src[1] & WINED3DSP_REGNUM_MASK;
- }
+ case CHANNEL_SOURCE_ONE:
+ strcat(arguments, "1.0");
break;
- case WINED3DSIO_TEXLDL:
- FIXME("Add color fixup for vertex texture WINED3DSIO_TEXLDL\n");
- return;
+ case CHANNEL_SOURCE_X:
+ strcat(arguments, reg_name);
+ strcat(arguments, ".x");
+ break;
+
+ case CHANNEL_SOURCE_Y:
+ strcat(arguments, reg_name);
+ strcat(arguments, ".y");
+ break;
+
+ case CHANNEL_SOURCE_Z:
+ strcat(arguments, reg_name);
+ strcat(arguments, ".z");
+ break;
- case WINED3DSIO_TEXDP3TEX:
- case WINED3DSIO_TEXM3x3TEX:
- case WINED3DSIO_TEXM3x3SPEC:
- case WINED3DSIO_TEXM3x3VSPEC:
- case WINED3DSIO_TEXBEM:
- case WINED3DSIO_TEXREG2AR:
- case WINED3DSIO_TEXREG2GB:
- case WINED3DSIO_TEXREG2RGB:
- sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
+ case CHANNEL_SOURCE_W:
+ strcat(arguments, reg_name);
+ strcat(arguments, ".w");
break;
default:
- /* Not a texture sampling instruction, nothing to do */
- return;
- };
+ FIXME("Unhandled channel source %#x\n", channel_source);
+ strcat(arguments, "undefined");
+ break;
+ }
- texture = (IWineD3DBaseTextureImpl *) deviceImpl->stateBlock->textures[sampler_idx];
- if(texture) {
- fmt = texture->resource.format;
- conversion_group = texture->baseTexture.shader_conversion_group;
- } else {
- fmt = WINED3DFMT_UNKNOWN;
- conversion_group = WINED3DFMT_UNKNOWN;
+ if (sign_fixup) strcat(arguments, " * 2.0 - 1.0");
+}
+
+static void shader_glsl_color_correction(const struct wined3d_shader_instruction *ins, struct color_fixup_desc fixup)
+{
+ struct wined3d_shader_dst_param dst;
+ unsigned int mask_size, remaining;
+ glsl_dst_param_t dst_param;
+ char arguments[256];
+ DWORD mask;
+
+ mask = 0;
+ if (fixup.x_sign_fixup || fixup.x_source != CHANNEL_SOURCE_X) mask |= WINED3DSP_WRITEMASK_0;
+ if (fixup.y_sign_fixup || fixup.y_source != CHANNEL_SOURCE_Y) mask |= WINED3DSP_WRITEMASK_1;
+ if (fixup.z_sign_fixup || fixup.z_source != CHANNEL_SOURCE_Z) mask |= WINED3DSP_WRITEMASK_2;
+ if (fixup.w_sign_fixup || fixup.w_source != CHANNEL_SOURCE_W) mask |= WINED3DSP_WRITEMASK_3;
+ mask &= ins->dst[0].write_mask;
+
+ if (!mask) return; /* Nothing to do */
+
+ if (is_yuv_fixup(fixup))
+ {
+ enum yuv_fixup yuv_fixup = get_yuv_fixup(fixup);
+ FIXME("YUV fixup (%#x) not supported\n", yuv_fixup);
+ return;
}
- /* before doing anything, record the sampler with the format in the format conversion list,
- * but check if it's not there already
- */
- for(i = 0; i < shader->baseShader.num_sampled_samplers; i++) {
- if(shader->baseShader.sampled_samplers[i] == sampler_idx) {
- recorded = TRUE;
- break;
- }
+ mask_size = shader_glsl_get_write_mask_size(mask);
+
+ dst = ins->dst[0];
+ dst.write_mask = mask;
+ shader_glsl_add_dst_param(ins, &dst, &dst_param);
+
+ arguments[0] = '\0';
+ remaining = mask_size;
+ if (mask & WINED3DSP_WRITEMASK_0)
+ {
+ shader_glsl_append_fixup_arg(arguments, dst_param.reg_name, fixup.x_sign_fixup, fixup.x_source);
+ if (--remaining) strcat(arguments, ", ");
+ }
+ if (mask & WINED3DSP_WRITEMASK_1)
+ {
+ shader_glsl_append_fixup_arg(arguments, dst_param.reg_name, fixup.y_sign_fixup, fixup.y_source);
+ if (--remaining) strcat(arguments, ", ");
+ }
+ if (mask & WINED3DSP_WRITEMASK_2)
+ {
+ shader_glsl_append_fixup_arg(arguments, dst_param.reg_name, fixup.z_sign_fixup, fixup.z_source);
+ if (--remaining) strcat(arguments, ", ");
}
- if(!recorded) {
- shader->baseShader.sampled_samplers[shader->baseShader.num_sampled_samplers] = sampler_idx;
- shader->baseShader.num_sampled_samplers++;
- shader->baseShader.sampled_format[sampler_idx] = conversion_group;
+ if (mask & WINED3DSP_WRITEMASK_3)
+ {
+ shader_glsl_append_fixup_arg(arguments, dst_param.reg_name, fixup.w_sign_fixup, fixup.w_source);
+ if (--remaining) strcat(arguments, ", ");
}
- switch(fmt) {
- case WINED3DFMT_V8U8:
- case WINED3DFMT_V16U16:
- if(GL_SUPPORT(NV_TEXTURE_SHADER) ||
- (GL_SUPPORT(ATI_ENVMAP_BUMPMAP) && fmt == WINED3DFMT_V8U8)) {
- /* The 3rd channel returns 1.0 in d3d, but 0.0 in gl. Fix this while we're at it :-) */
- mask = shader_glsl_add_dst_param(arg, arg->dst, WINED3DSP_WRITEMASK_2, &dst_param);
- mask_size = shader_glsl_get_write_mask_size(mask);
- if(mask_size >= 3) {
- shader_addline(arg->buffer, "%s.%c = 1.0;\n", dst_param.reg_name, dst_param.mask_str[3]);
- }
- } else {
- /* Correct the sign, but leave the blue as it is - it was loaded correctly already */
- mask = shader_glsl_add_dst_param(arg, arg->dst,
- WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1,
- &dst_param);
- mask_size = shader_glsl_get_write_mask_size(mask);
- if(mask_size >= 2) {
- shader_addline(arg->buffer, "%s.%c%c = %s.%c%c * 2.0 - 1.0;\n",
- dst_param.reg_name, dst_param.mask_str[1], dst_param.mask_str[2],
- dst_param.reg_name, dst_param.mask_str[1], dst_param.mask_str[2]);
- } else if(mask_size == 1) {
- shader_addline(arg->buffer, "%s.%c = %s.%c * 2.0 - 1.0;\n", dst_param.reg_name, dst_param.mask_str[1],
- dst_param.reg_name, dst_param.mask_str[1]);
- }
- }
- break;
+ if (mask_size > 1)
+ {
+ shader_addline(ins->ctx->buffer, "%s%s = vec%u(%s);\n",
+ dst_param.reg_name, dst_param.mask_str, mask_size, arguments);
+ }
+ else
+ {
+ shader_addline(ins->ctx->buffer, "%s%s = %s;\n", dst_param.reg_name, dst_param.mask_str, arguments);
+ }
+}
- case WINED3DFMT_X8L8V8U8:
- if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
- /* Red and blue are the signed channels, fix them up; Blue(=L) is correct already,
- * and a(X) is always 1.0
- */
- mask = shader_glsl_add_dst_param(arg, arg->dst, WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &dst_param);
- mask_size = shader_glsl_get_write_mask_size(mask);
- if(mask_size >= 2) {
- shader_addline(arg->buffer, "%s.%c%c = %s.%c%c * 2.0 - 1.0;\n",
- dst_param.reg_name, dst_param.mask_str[1], dst_param.mask_str[2],
- dst_param.reg_name, dst_param.mask_str[1], dst_param.mask_str[2]);
- } else if(mask_size == 1) {
- shader_addline(arg->buffer, "%s.%c = %s.%c * 2.0 - 1.0;\n",
- dst_param.reg_name, dst_param.mask_str[1],
- dst_param.reg_name, dst_param.mask_str[1]);
- }
- }
- break;
+static void PRINTF_ATTR(8, 9) shader_glsl_gen_sample_code(const struct wined3d_shader_instruction *ins,
+ DWORD sampler, const glsl_sample_function_t *sample_function, DWORD swizzle,
+ const char *dx, const char *dy,
+ const char *bias, const char *coord_reg_fmt, ...)
+{
+ const char *sampler_base;
+ char dst_swizzle[6];
+ struct color_fixup_desc fixup;
+ BOOL np2_fixup = FALSE;
+ va_list args;
- case WINED3DFMT_L6V5U5:
- if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
- mask = shader_glsl_add_dst_param(arg, arg->dst, WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &dst_param);
- mask_size = shader_glsl_get_write_mask_size(mask);
- shader_glsl_add_dst_param(arg, arg->dst, WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_2, &dst_param2);
- if(mask_size >= 3) {
- /* Swap y and z (U and L), and do a sign conversion on x and the new y(V and U) */
- shader_addline(arg->buffer, "tmp0.g = %s.%c;\n",
- dst_param.reg_name, dst_param.mask_str[2]);
- shader_addline(arg->buffer, "%s.%c%c = %s.%c%c * 2.0 - 1.0;\n",
- dst_param.reg_name, dst_param.mask_str[2], dst_param.mask_str[1],
- dst_param2.reg_name, dst_param.mask_str[1], dst_param.mask_str[3]);
- shader_addline(arg->buffer, "%s.%c = tmp0.g;\n", dst_param.reg_name,
- dst_param.mask_str[3]);
- } else if(mask_size == 2) {
- /* This is bad: We have VL, but we need VU */
- FIXME("2 components sampled from a converted L6V5U5 texture\n");
- } else {
- shader_addline(arg->buffer, "%s.%c = %s.%c * 2.0 - 1.0;\n",
- dst_param.reg_name, dst_param.mask_str[1],
- dst_param2.reg_name, dst_param.mask_str[1]);
- }
- }
- break;
+ shader_glsl_swizzle_to_str(swizzle, FALSE, ins->dst[0].write_mask, dst_swizzle);
- case WINED3DFMT_Q8W8V8U8:
- if(!GL_SUPPORT(NV_TEXTURE_SHADER)) {
- /* Correct the sign in all channels. The writemask just applies as-is, no
- * need for checking the mask size
- */
- shader_glsl_add_dst_param(arg, arg->dst,
- WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 |
- WINED3DSP_WRITEMASK_2 | WINED3DSP_WRITEMASK_3,
- &dst_param);
- shader_addline(arg->buffer, "%s%s = %s%s * 2.0 - 1.0;\n", dst_param.reg_name, dst_param.mask_str,
- dst_param.reg_name, dst_param.mask_str);
- }
- break;
+ if (shader_is_pshader_version(ins->ctx->reg_maps->shader_version.type))
+ {
+ const struct shader_glsl_ctx_priv *priv = ins->ctx->backend_data;
+ fixup = priv->cur_ps_args->color_fixup[sampler];
+ sampler_base = "Psampler";
- case WINED3DFMT_ATI2N:
- /* GL_ATI_texture_compression_3dc returns the two channels as luminance-alpha,
- * which means the first one is replicated across .rgb, and the 2nd one is in
- * .a. We need the 2nd in .g
- *
- * GL_EXT_texture_compression_rgtc returns the values in .rg, however, they
- * are swapped compared to d3d. So swap red and green.
- */
- mask = shader_glsl_add_dst_param(arg, arg->dst, WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &dst_param);
- mask_size = shader_glsl_get_write_mask_size(mask);
- if(GL_SUPPORT(EXT_TEXTURE_COMPRESSION_RGTC)) {
- if(mask_size >= 2) {
- shader_addline(arg->buffer, "%s.%c%c = %s.%c%c;\n",
- dst_param.reg_name, dst_param.mask_str[1],
- dst_param.mask_str[2],
- dst_param.reg_name, dst_param.mask_str[2],
- dst_param.mask_str[1]);
- } else {
- FIXME("%u components sampled from a converted ATI2N texture\n", mask_size);
- }
+ if(priv->cur_ps_args->np2_fixup & (1 << sampler)) {
+ if(bias) {
+ FIXME("Biased sampling from NP2 textures is unsupported\n");
} else {
- if(mask_size == 4) {
- /* Swap y and z (U and L), and do a sign conversion on x and the new y(V and U) */
- shader_addline(arg->buffer, "%s.%c = %s.%c;\n",
- dst_param.reg_name, dst_param.mask_str[2],
- dst_param.reg_name, dst_param.mask_str[4]);
- } else if(mask_size == 1) {
- /* Nothing to do */
- } else {
- FIXME("%u components sampled from a converted ATI2N texture\n", mask_size);
- /* This is bad: We have .r[gb], but we need .ra */
- }
+ np2_fixup = TRUE;
}
- break;
+ }
+ } else {
+ sampler_base = "Vsampler";
+ fixup = COLOR_FIXUP_IDENTITY; /* FIXME: Vshader color fixup */
+ }
- /* stupid compiler */
- default:
- break;
+ shader_glsl_append_dst(ins->ctx->buffer, ins);
+
+ shader_addline(ins->ctx->buffer, "%s(%s%u, ", sample_function->name, sampler_base, sampler);
+
+ va_start(args, coord_reg_fmt);
+ shader_vaddline(ins->ctx->buffer, coord_reg_fmt, args);
+ va_end(args);
+
+ if(bias) {
+ shader_addline(ins->ctx->buffer, ", %s)%s);\n", bias, dst_swizzle);
+ } else {
+ if (np2_fixup) {
+ const struct shader_glsl_ctx_priv *priv = ins->ctx->backend_data;
+ const unsigned char idx = priv->cur_np2fixup_info->idx[sampler];
+
+ shader_addline(ins->ctx->buffer, " * PsamplerNP2Fixup[%u].%s)%s);\n", idx >> 1,
+ (idx % 2) ? "zw" : "xy", dst_swizzle);
+ } else if(dx && dy) {
+ shader_addline(ins->ctx->buffer, ", %s, %s)%s);\n", dx, dy, dst_swizzle);
+ } else {
+ shader_addline(ins->ctx->buffer, ")%s);\n", dst_swizzle);
+ }
+ }
+
+ if(!is_identity_fixup(fixup)) {
+ shader_glsl_color_correction(ins, fixup);
}
}
/*****************************************************************************
- *
* Begin processing individual instruction opcodes
- *
****************************************************************************/
/* Generate GLSL arithmetic functions (dst = src1 + src2) */
-void shader_glsl_arith(SHADER_OPCODE_ARG* arg) {
- CONST SHADER_OPCODE* curOpcode = arg->opcode;
- SHADER_BUFFER* buffer = arg->buffer;
+static void shader_glsl_arith(const struct wined3d_shader_instruction *ins)
+{
+ struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
glsl_src_param_t src0_param;
glsl_src_param_t src1_param;
DWORD write_mask;
char op;
/* Determine the GLSL operator to use based on the opcode */
- switch (curOpcode->opcode) {
- case WINED3DSIO_MUL: op = '*'; break;
- case WINED3DSIO_ADD: op = '+'; break;
- case WINED3DSIO_SUB: op = '-'; break;
+ switch (ins->handler_idx)
+ {
+ case WINED3DSIH_MUL: op = '*'; break;
+ case WINED3DSIH_ADD: op = '+'; break;
+ case WINED3DSIH_SUB: op = '-'; break;
default:
op = ' ';
- FIXME("Opcode %s not yet handled in GLSL\n", curOpcode->name);
+ FIXME("Opcode %#x not yet handled in GLSL\n", ins->handler_idx);
break;
}
- write_mask = shader_glsl_append_dst(buffer, arg);
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
- shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
+ write_mask = shader_glsl_append_dst(buffer, ins);
+ shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
+ shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
shader_addline(buffer, "%s %c %s);\n", src0_param.param_str, op, src1_param.param_str);
}
/* Process the WINED3DSIO_MOV opcode using GLSL (dst = src) */
-void shader_glsl_mov(SHADER_OPCODE_ARG* arg) {
- IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
- SHADER_BUFFER* buffer = arg->buffer;
+static void shader_glsl_mov(const struct wined3d_shader_instruction *ins)
+{
+ struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
glsl_src_param_t src0_param;
DWORD write_mask;
- write_mask = shader_glsl_append_dst(buffer, arg);
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
+ write_mask = shader_glsl_append_dst(buffer, ins);
+ shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
/* In vs_1_1 WINED3DSIO_MOV can write to the address register. In later
* shader versions WINED3DSIO_MOVA is used for this. */
- if ((WINED3DSHADER_VERSION_MAJOR(shader->baseShader.hex_version) == 1 &&
- !shader_is_pshader_version(shader->baseShader.hex_version) &&
- shader_get_regtype(arg->dst) == WINED3DSPR_ADDR)) {
+ if (ins->ctx->reg_maps->shader_version.major == 1
+ && !shader_is_pshader_version(ins->ctx->reg_maps->shader_version.type)
+ && ins->dst[0].reg.type == WINED3DSPR_ADDR)
+ {
/* This is a simple floor() */
unsigned int mask_size = shader_glsl_get_write_mask_size(write_mask);
if (mask_size > 1) {
} else {
shader_addline(buffer, "int(floor(%s)));\n", src0_param.param_str);
}
- } else if(arg->opcode->opcode == WINED3DSIO_MOVA) {
+ }
+ else if(ins->handler_idx == WINED3DSIH_MOVA)
+ {
/* We need to *round* to the nearest int here. */
unsigned int mask_size = shader_glsl_get_write_mask_size(write_mask);
if (mask_size > 1) {
}
/* Process the dot product operators DP3 and DP4 in GLSL (dst = dot(src0, src1)) */
-void shader_glsl_dot(SHADER_OPCODE_ARG* arg) {
- CONST SHADER_OPCODE* curOpcode = arg->opcode;
- SHADER_BUFFER* buffer = arg->buffer;
+static void shader_glsl_dot(const struct wined3d_shader_instruction *ins)
+{
+ struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
glsl_src_param_t src0_param;
glsl_src_param_t src1_param;
DWORD dst_write_mask, src_write_mask;
unsigned int dst_size = 0;
- dst_write_mask = shader_glsl_append_dst(buffer, arg);
+ dst_write_mask = shader_glsl_append_dst(buffer, ins);
dst_size = shader_glsl_get_write_mask_size(dst_write_mask);
/* dp3 works on vec3, dp4 on vec4 */
- if (curOpcode->opcode == WINED3DSIO_DP4) {
+ if (ins->handler_idx == WINED3DSIH_DP4)
+ {
src_write_mask = WINED3DSP_WRITEMASK_ALL;
} else {
src_write_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
}
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_write_mask, &src0_param);
- shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], src_write_mask, &src1_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], src_write_mask, &src0_param);
+ shader_glsl_add_src_param(ins, &ins->src[1], src_write_mask, &src1_param);
if (dst_size > 1) {
shader_addline(buffer, "vec%d(dot(%s, %s)));\n", dst_size, src0_param.param_str, src1_param.param_str);
/* Note that this instruction has some restrictions. The destination write mask
* can't contain the w component, and the source swizzles have to be .xyzw */
-void shader_glsl_cross(SHADER_OPCODE_ARG *arg) {
+static void shader_glsl_cross(const struct wined3d_shader_instruction *ins)
+{
DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
glsl_src_param_t src0_param;
glsl_src_param_t src1_param;
char dst_mask[6];
- shader_glsl_get_write_mask(arg->dst, dst_mask);
- shader_glsl_append_dst(arg->buffer, arg);
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
- shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], src_mask, &src1_param);
- shader_addline(arg->buffer, "cross(%s, %s)%s);\n", src0_param.param_str, src1_param.param_str, dst_mask);
+ shader_glsl_get_write_mask(&ins->dst[0], dst_mask);
+ shader_glsl_append_dst(ins->ctx->buffer, ins);
+ shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
+ shader_glsl_add_src_param(ins, &ins->src[1], src_mask, &src1_param);
+ shader_addline(ins->ctx->buffer, "cross(%s, %s)%s);\n", src0_param.param_str, src1_param.param_str, dst_mask);
}
/* Process the WINED3DSIO_POW instruction in GLSL (dst = |src0|^src1)
* Src0 and src1 are scalars. Note that D3D uses the absolute of src0, while
* GLSL uses the value as-is. */
-void shader_glsl_pow(SHADER_OPCODE_ARG *arg) {
- SHADER_BUFFER *buffer = arg->buffer;
+static void shader_glsl_pow(const struct wined3d_shader_instruction *ins)
+{
+ struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
glsl_src_param_t src0_param;
glsl_src_param_t src1_param;
DWORD dst_write_mask;
unsigned int dst_size;
- dst_write_mask = shader_glsl_append_dst(buffer, arg);
+ dst_write_mask = shader_glsl_append_dst(buffer, ins);
dst_size = shader_glsl_get_write_mask_size(dst_write_mask);
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
- shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0, &src1_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src0_param);
+ shader_glsl_add_src_param(ins, &ins->src[1], WINED3DSP_WRITEMASK_0, &src1_param);
if (dst_size > 1) {
shader_addline(buffer, "vec%d(pow(abs(%s), %s)));\n", dst_size, src0_param.param_str, src1_param.param_str);
/* Process the WINED3DSIO_LOG instruction in GLSL (dst = log2(|src0|))
* Src0 is a scalar. Note that D3D uses the absolute of src0, while
* GLSL uses the value as-is. */
-void shader_glsl_log(SHADER_OPCODE_ARG *arg) {
- SHADER_BUFFER *buffer = arg->buffer;
+static void shader_glsl_log(const struct wined3d_shader_instruction *ins)
+{
+ struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
glsl_src_param_t src0_param;
DWORD dst_write_mask;
unsigned int dst_size;
- dst_write_mask = shader_glsl_append_dst(buffer, arg);
+ dst_write_mask = shader_glsl_append_dst(buffer, ins);
dst_size = shader_glsl_get_write_mask_size(dst_write_mask);
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src0_param);
if (dst_size > 1) {
shader_addline(buffer, "vec%d(log2(abs(%s))));\n", dst_size, src0_param.param_str);
}
/* Map the opcode 1-to-1 to the GL code (arg->dst = instruction(src0, src1, ...) */
-void shader_glsl_map2gl(SHADER_OPCODE_ARG* arg) {
- CONST SHADER_OPCODE* curOpcode = arg->opcode;
- SHADER_BUFFER* buffer = arg->buffer;
+static void shader_glsl_map2gl(const struct wined3d_shader_instruction *ins)
+{
+ struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
glsl_src_param_t src_param;
const char *instruction;
- char arguments[256];
DWORD write_mask;
unsigned i;
/* Determine the GLSL function to use based on the opcode */
/* TODO: Possibly make this a table for faster lookups */
- switch (curOpcode->opcode) {
- case WINED3DSIO_MIN: instruction = "min"; break;
- case WINED3DSIO_MAX: instruction = "max"; break;
- case WINED3DSIO_ABS: instruction = "abs"; break;
- case WINED3DSIO_FRC: instruction = "fract"; break;
- case WINED3DSIO_NRM: instruction = "normalize"; break;
- case WINED3DSIO_LOGP:
- case WINED3DSIO_LOG: instruction = "log2"; break;
- case WINED3DSIO_EXP: instruction = "exp2"; break;
- case WINED3DSIO_SGN: instruction = "sign"; break;
- case WINED3DSIO_DSX: instruction = "dFdx"; break;
- case WINED3DSIO_DSY: instruction = "ycorrection.y * dFdy"; break;
+ switch (ins->handler_idx)
+ {
+ case WINED3DSIH_MIN: instruction = "min"; break;
+ case WINED3DSIH_MAX: instruction = "max"; break;
+ case WINED3DSIH_ABS: instruction = "abs"; break;
+ case WINED3DSIH_FRC: instruction = "fract"; break;
+ case WINED3DSIH_NRM: instruction = "normalize"; break;
+ case WINED3DSIH_EXP: instruction = "exp2"; break;
+ case WINED3DSIH_DSX: instruction = "dFdx"; break;
+ case WINED3DSIH_DSY: instruction = "ycorrection.y * dFdy"; break;
default: instruction = "";
- FIXME("Opcode %s not yet handled in GLSL\n", curOpcode->name);
+ FIXME("Opcode %#x not yet handled in GLSL\n", ins->handler_idx);
break;
}
- write_mask = shader_glsl_append_dst(buffer, arg);
+ write_mask = shader_glsl_append_dst(buffer, ins);
- arguments[0] = '\0';
- if (curOpcode->num_params > 0) {
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src_param);
- strcat(arguments, src_param.param_str);
- for (i = 2; i < curOpcode->num_params; ++i) {
- strcat(arguments, ", ");
- shader_glsl_add_src_param(arg, arg->src[i-1], arg->src_addr[i-1], write_mask, &src_param);
- strcat(arguments, src_param.param_str);
+ shader_addline(buffer, "%s(", instruction);
+
+ if (ins->src_count)
+ {
+ shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src_param);
+ shader_addline(buffer, "%s", src_param.param_str);
+ for (i = 1; i < ins->src_count; ++i)
+ {
+ shader_glsl_add_src_param(ins, &ins->src[i], write_mask, &src_param);
+ shader_addline(buffer, ", %s", src_param.param_str);
}
}
- shader_addline(buffer, "%s(%s));\n", instruction, arguments);
+ shader_addline(buffer, "));\n");
}
/** Process the WINED3DSIO_EXPP instruction in GLSL:
* For 2.0 shaders, just do this (honoring writemask and swizzle):
* dst = 2^src; (partial precision is allowed, but optional)
*/
-void shader_glsl_expp(SHADER_OPCODE_ARG* arg) {
- IWineD3DBaseShaderImpl *shader = (IWineD3DBaseShaderImpl *)arg->shader;
+static void shader_glsl_expp(const struct wined3d_shader_instruction *ins)
+{
glsl_src_param_t src_param;
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src_param);
- if (shader->baseShader.hex_version < WINED3DPS_VERSION(2,0)) {
+ if (ins->ctx->reg_maps->shader_version.major < 2)
+ {
char dst_mask[6];
- shader_addline(arg->buffer, "tmp0.x = exp2(floor(%s));\n", src_param.param_str);
- shader_addline(arg->buffer, "tmp0.y = %s - floor(%s);\n", src_param.param_str, src_param.param_str);
- shader_addline(arg->buffer, "tmp0.z = exp2(%s);\n", src_param.param_str);
- shader_addline(arg->buffer, "tmp0.w = 1.0;\n");
+ shader_addline(ins->ctx->buffer, "tmp0.x = exp2(floor(%s));\n", src_param.param_str);
+ shader_addline(ins->ctx->buffer, "tmp0.y = %s - floor(%s);\n", src_param.param_str, src_param.param_str);
+ shader_addline(ins->ctx->buffer, "tmp0.z = exp2(%s);\n", src_param.param_str);
+ shader_addline(ins->ctx->buffer, "tmp0.w = 1.0;\n");
- shader_glsl_append_dst(arg->buffer, arg);
- shader_glsl_get_write_mask(arg->dst, dst_mask);
- shader_addline(arg->buffer, "tmp0%s);\n", dst_mask);
+ shader_glsl_append_dst(ins->ctx->buffer, ins);
+ shader_glsl_get_write_mask(&ins->dst[0], dst_mask);
+ shader_addline(ins->ctx->buffer, "tmp0%s);\n", dst_mask);
} else {
DWORD write_mask;
unsigned int mask_size;
- write_mask = shader_glsl_append_dst(arg->buffer, arg);
+ write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
mask_size = shader_glsl_get_write_mask_size(write_mask);
if (mask_size > 1) {
- shader_addline(arg->buffer, "vec%d(exp2(%s)));\n", mask_size, src_param.param_str);
+ shader_addline(ins->ctx->buffer, "vec%d(exp2(%s)));\n", mask_size, src_param.param_str);
} else {
- shader_addline(arg->buffer, "exp2(%s));\n", src_param.param_str);
+ shader_addline(ins->ctx->buffer, "exp2(%s));\n", src_param.param_str);
}
}
}
/** Process the RCP (reciprocal or inverse) opcode in GLSL (dst = 1 / src) */
-void shader_glsl_rcp(SHADER_OPCODE_ARG* arg) {
+static void shader_glsl_rcp(const struct wined3d_shader_instruction *ins)
+{
glsl_src_param_t src_param;
DWORD write_mask;
unsigned int mask_size;
- write_mask = shader_glsl_append_dst(arg->buffer, arg);
+ write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
mask_size = shader_glsl_get_write_mask_size(write_mask);
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &src_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_3, &src_param);
if (mask_size > 1) {
- shader_addline(arg->buffer, "vec%d(1.0 / %s));\n", mask_size, src_param.param_str);
+ shader_addline(ins->ctx->buffer, "vec%d(1.0 / %s));\n", mask_size, src_param.param_str);
} else {
- shader_addline(arg->buffer, "1.0 / %s);\n", src_param.param_str);
+ shader_addline(ins->ctx->buffer, "1.0 / %s);\n", src_param.param_str);
}
}
-void shader_glsl_rsq(SHADER_OPCODE_ARG* arg) {
- SHADER_BUFFER* buffer = arg->buffer;
+static void shader_glsl_rsq(const struct wined3d_shader_instruction *ins)
+{
+ struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
glsl_src_param_t src_param;
DWORD write_mask;
unsigned int mask_size;
- write_mask = shader_glsl_append_dst(buffer, arg);
+ write_mask = shader_glsl_append_dst(buffer, ins);
mask_size = shader_glsl_get_write_mask_size(write_mask);
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &src_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_3, &src_param);
if (mask_size > 1) {
shader_addline(buffer, "vec%d(inversesqrt(%s)));\n", mask_size, src_param.param_str);
}
/** Process signed comparison opcodes in GLSL. */
-void shader_glsl_compare(SHADER_OPCODE_ARG* arg) {
+static void shader_glsl_compare(const struct wined3d_shader_instruction *ins)
+{
glsl_src_param_t src0_param;
glsl_src_param_t src1_param;
DWORD write_mask;
unsigned int mask_size;
- write_mask = shader_glsl_append_dst(arg->buffer, arg);
+ write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
mask_size = shader_glsl_get_write_mask_size(write_mask);
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
- shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
+ shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
if (mask_size > 1) {
const char *compare;
- switch(arg->opcode->opcode) {
- case WINED3DSIO_SLT: compare = "lessThan"; break;
- case WINED3DSIO_SGE: compare = "greaterThanEqual"; break;
+ switch(ins->handler_idx)
+ {
+ case WINED3DSIH_SLT: compare = "lessThan"; break;
+ case WINED3DSIH_SGE: compare = "greaterThanEqual"; break;
default: compare = "";
- FIXME("Can't handle opcode %s\n", arg->opcode->name);
+ FIXME("Can't handle opcode %#x\n", ins->handler_idx);
}
- shader_addline(arg->buffer, "vec%d(%s(%s, %s)));\n", mask_size, compare,
+ shader_addline(ins->ctx->buffer, "vec%d(%s(%s, %s)));\n", mask_size, compare,
src0_param.param_str, src1_param.param_str);
} else {
- switch(arg->opcode->opcode) {
- case WINED3DSIO_SLT:
+ switch(ins->handler_idx)
+ {
+ case WINED3DSIH_SLT:
/* Step(src0, src1) is not suitable here because if src0 == src1 SLT is supposed,
* to return 0.0 but step returns 1.0 because step is not < x
* An alternative is a bvec compare padded with an unused second component.
* issue. Playing with not() is not possible either because not() does not accept
* a scalar.
*/
- shader_addline(arg->buffer, "(%s < %s) ? 1.0 : 0.0);\n", src0_param.param_str, src1_param.param_str);
+ shader_addline(ins->ctx->buffer, "(%s < %s) ? 1.0 : 0.0);\n",
+ src0_param.param_str, src1_param.param_str);
break;
- case WINED3DSIO_SGE:
+ case WINED3DSIH_SGE:
/* Here we can use the step() function and safe a conditional */
- shader_addline(arg->buffer, "step(%s, %s));\n", src1_param.param_str, src0_param.param_str);
+ shader_addline(ins->ctx->buffer, "step(%s, %s));\n", src1_param.param_str, src0_param.param_str);
break;
default:
- FIXME("Can't handle opcode %s\n", arg->opcode->name);
+ FIXME("Can't handle opcode %#x\n", ins->handler_idx);
}
}
}
/** Process CMP instruction in GLSL (dst = src0 >= 0.0 ? src1 : src2), per channel */
-void shader_glsl_cmp(SHADER_OPCODE_ARG* arg) {
+static void shader_glsl_cmp(const struct wined3d_shader_instruction *ins)
+{
glsl_src_param_t src0_param;
glsl_src_param_t src1_param;
glsl_src_param_t src2_param;
char mask_char[6];
BOOL temp_destination = FALSE;
- if(shader_is_scalar(arg->src[0])) {
- write_mask = shader_glsl_append_dst(arg->buffer, arg);
+ if (shader_is_scalar(&ins->src[0].reg))
+ {
+ write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_ALL, &src0_param);
- shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
- shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_ALL, &src0_param);
+ shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
+ shader_glsl_add_src_param(ins, &ins->src[2], write_mask, &src2_param);
- shader_addline(arg->buffer, "%s >= 0.0 ? %s : %s);\n",
+ shader_addline(ins->ctx->buffer, "%s >= 0.0 ? %s : %s);\n",
src0_param.param_str, src1_param.param_str, src2_param.param_str);
} else {
- DWORD src0reg = arg->src[0] & WINED3DSP_REGNUM_MASK;
- DWORD src1reg = arg->src[1] & WINED3DSP_REGNUM_MASK;
- DWORD src2reg = arg->src[2] & WINED3DSP_REGNUM_MASK;
- DWORD src0regtype = shader_get_regtype(arg->src[0]);
- DWORD src1regtype = shader_get_regtype(arg->src[1]);
- DWORD src2regtype = shader_get_regtype(arg->src[2]);
- DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
- DWORD dstregtype = shader_get_regtype(arg->dst);
+ DWORD dst_mask = ins->dst[0].write_mask;
+ struct wined3d_shader_dst_param dst = ins->dst[0];
/* Cycle through all source0 channels */
for (i=0; i<4; i++) {
write_mask = 0;
/* Find the destination channels which use the current source0 channel */
for (j=0; j<4; j++) {
- if ( ((arg->src[0] >> (WINED3DSP_SWIZZLE_SHIFT + 2*j)) & 0x3) == i ) {
+ if (((ins->src[0].swizzle >> (2 * j)) & 0x3) == i)
+ {
write_mask |= WINED3DSP_WRITEMASK_0 << j;
cmp_channel = WINED3DSP_WRITEMASK_0 << j;
}
}
+ dst.write_mask = dst_mask & write_mask;
/* Splitting the cmp instruction up in multiple lines imposes a problem:
* The first lines may overwrite source parameters of the following lines.
* Deal with that by using a temporary destination register if needed
*/
- if((src0reg == dstreg && src0regtype == dstregtype) ||
- (src1reg == dstreg && src1regtype == dstregtype) ||
- (src2reg == dstreg && src2regtype == dstregtype)) {
-
- write_mask = shader_glsl_get_write_mask(arg->dst & (~WINED3DSP_SWIZZLE_MASK | write_mask), mask_char);
+ if ((ins->src[0].reg.idx == ins->dst[0].reg.idx
+ && ins->src[0].reg.type == ins->dst[0].reg.type)
+ || (ins->src[1].reg.idx == ins->dst[0].reg.idx
+ && ins->src[1].reg.type == ins->dst[0].reg.type)
+ || (ins->src[2].reg.idx == ins->dst[0].reg.idx
+ && ins->src[2].reg.type == ins->dst[0].reg.type))
+ {
+ write_mask = shader_glsl_get_write_mask(&dst, mask_char);
if (!write_mask) continue;
- shader_addline(arg->buffer, "tmp0%s = (", mask_char);
+ shader_addline(ins->ctx->buffer, "tmp0%s = (", mask_char);
temp_destination = TRUE;
} else {
- write_mask = shader_glsl_append_dst_ext(arg->buffer, arg, arg->dst & (~WINED3DSP_SWIZZLE_MASK | write_mask));
+ write_mask = shader_glsl_append_dst_ext(ins->ctx->buffer, ins, &dst);
if (!write_mask) continue;
}
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], cmp_channel, &src0_param);
- shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
- shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], cmp_channel, &src0_param);
+ shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
+ shader_glsl_add_src_param(ins, &ins->src[2], write_mask, &src2_param);
- shader_addline(arg->buffer, "%s >= 0.0 ? %s : %s);\n",
+ shader_addline(ins->ctx->buffer, "%s >= 0.0 ? %s : %s);\n",
src0_param.param_str, src1_param.param_str, src2_param.param_str);
}
if(temp_destination) {
- shader_glsl_get_write_mask(arg->dst, mask_char);
- shader_glsl_append_dst_ext(arg->buffer, arg, arg->dst);
- shader_addline(arg->buffer, "tmp0%s);\n", mask_char);
+ shader_glsl_get_write_mask(&ins->dst[0], mask_char);
+ shader_glsl_append_dst(ins->ctx->buffer, ins);
+ shader_addline(ins->ctx->buffer, "tmp0%s);\n", mask_char);
}
}
/** Process the CND opcode in GLSL (dst = (src0 > 0.5) ? src1 : src2) */
/* For ps 1.1-1.3, only a single component of src0 is used. For ps 1.4
* the compare is done per component of src0. */
-void shader_glsl_cnd(SHADER_OPCODE_ARG* arg) {
- IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
+static void shader_glsl_cnd(const struct wined3d_shader_instruction *ins)
+{
+ struct wined3d_shader_dst_param dst;
glsl_src_param_t src0_param;
glsl_src_param_t src1_param;
glsl_src_param_t src2_param;
DWORD write_mask, cmp_channel = 0;
unsigned int i, j;
+ DWORD dst_mask;
+ DWORD shader_version = WINED3D_SHADER_VERSION(ins->ctx->reg_maps->shader_version.major,
+ ins->ctx->reg_maps->shader_version.minor);
- if (shader->baseShader.hex_version < WINED3DPS_VERSION(1, 4)) {
- write_mask = shader_glsl_append_dst(arg->buffer, arg);
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
- shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
- shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
+ if (shader_version < WINED3D_SHADER_VERSION(1, 4))
+ {
+ write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
+ shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src0_param);
+ shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
+ shader_glsl_add_src_param(ins, &ins->src[2], write_mask, &src2_param);
/* Fun: The D3DSI_COISSUE flag changes the semantic of the cnd instruction for < 1.4 shaders */
- if(arg->opcode_token & WINED3DSI_COISSUE) {
- shader_addline(arg->buffer, "%s /* COISSUE! */);\n", src1_param.param_str);
+ if (ins->coissue)
+ {
+ shader_addline(ins->ctx->buffer, "%s /* COISSUE! */);\n", src1_param.param_str);
} else {
- shader_addline(arg->buffer, "%s > 0.5 ? %s : %s);\n",
+ shader_addline(ins->ctx->buffer, "%s > 0.5 ? %s : %s);\n",
src0_param.param_str, src1_param.param_str, src2_param.param_str);
}
return;
}
/* Cycle through all source0 channels */
+ dst_mask = ins->dst[0].write_mask;
+ dst = ins->dst[0];
for (i=0; i<4; i++) {
write_mask = 0;
/* Find the destination channels which use the current source0 channel */
for (j=0; j<4; j++) {
- if ( ((arg->src[0] >> (WINED3DSP_SWIZZLE_SHIFT + 2*j)) & 0x3) == i ) {
+ if (((ins->src[0].swizzle >> (2 * j)) & 0x3) == i)
+ {
write_mask |= WINED3DSP_WRITEMASK_0 << j;
cmp_channel = WINED3DSP_WRITEMASK_0 << j;
}
}
- write_mask = shader_glsl_append_dst_ext(arg->buffer, arg, arg->dst & (~WINED3DSP_SWIZZLE_MASK | write_mask));
+
+ dst.write_mask = dst_mask & write_mask;
+ write_mask = shader_glsl_append_dst_ext(ins->ctx->buffer, ins, &dst);
if (!write_mask) continue;
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], cmp_channel, &src0_param);
- shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
- shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], cmp_channel, &src0_param);
+ shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
+ shader_glsl_add_src_param(ins, &ins->src[2], write_mask, &src2_param);
- shader_addline(arg->buffer, "%s > 0.5 ? %s : %s);\n",
+ shader_addline(ins->ctx->buffer, "%s > 0.5 ? %s : %s);\n",
src0_param.param_str, src1_param.param_str, src2_param.param_str);
}
}
/** GLSL code generation for WINED3DSIO_MAD: Multiply the first 2 opcodes, then add the last */
-void shader_glsl_mad(SHADER_OPCODE_ARG* arg) {
+static void shader_glsl_mad(const struct wined3d_shader_instruction *ins)
+{
glsl_src_param_t src0_param;
glsl_src_param_t src1_param;
glsl_src_param_t src2_param;
DWORD write_mask;
- write_mask = shader_glsl_append_dst(arg->buffer, arg);
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
- shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
- shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
- shader_addline(arg->buffer, "(%s * %s) + %s);\n",
+ write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
+ shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
+ shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
+ shader_glsl_add_src_param(ins, &ins->src[2], write_mask, &src2_param);
+ shader_addline(ins->ctx->buffer, "(%s * %s) + %s);\n",
src0_param.param_str, src1_param.param_str, src2_param.param_str);
}
-/** Handles transforming all WINED3DSIO_M?x? opcodes for
- Vertex shaders to GLSL codes */
-void shader_glsl_mnxn(SHADER_OPCODE_ARG* arg) {
+/* Handles transforming all WINED3DSIO_M?x? opcodes for
+ Vertex shaders to GLSL codes */
+static void shader_glsl_mnxn(const struct wined3d_shader_instruction *ins)
+{
int i;
int nComponents = 0;
- SHADER_OPCODE_ARG tmpArg;
-
- memset(&tmpArg, 0, sizeof(SHADER_OPCODE_ARG));
+ struct wined3d_shader_dst_param tmp_dst = {{0}};
+ struct wined3d_shader_src_param tmp_src[2] = {{{0}}};
+ struct wined3d_shader_instruction tmp_ins;
+
+ memset(&tmp_ins, 0, sizeof(tmp_ins));
/* Set constants for the temporary argument */
- tmpArg.shader = arg->shader;
- tmpArg.buffer = arg->buffer;
- tmpArg.src[0] = arg->src[0];
- tmpArg.src_addr[0] = arg->src_addr[0];
- tmpArg.src_addr[1] = arg->src_addr[1];
- tmpArg.reg_maps = arg->reg_maps;
-
- switch(arg->opcode->opcode) {
- case WINED3DSIO_M4x4:
+ tmp_ins.ctx = ins->ctx;
+ tmp_ins.dst_count = 1;
+ tmp_ins.dst = &tmp_dst;
+ tmp_ins.src_count = 2;
+ tmp_ins.src = tmp_src;
+
+ switch(ins->handler_idx)
+ {
+ case WINED3DSIH_M4x4:
nComponents = 4;
- tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
+ tmp_ins.handler_idx = WINED3DSIH_DP4;
break;
- case WINED3DSIO_M4x3:
+ case WINED3DSIH_M4x3:
nComponents = 3;
- tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP4);
+ tmp_ins.handler_idx = WINED3DSIH_DP4;
break;
- case WINED3DSIO_M3x4:
+ case WINED3DSIH_M3x4:
nComponents = 4;
- tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
+ tmp_ins.handler_idx = WINED3DSIH_DP3;
break;
- case WINED3DSIO_M3x3:
+ case WINED3DSIH_M3x3:
nComponents = 3;
- tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
+ tmp_ins.handler_idx = WINED3DSIH_DP3;
break;
- case WINED3DSIO_M3x2:
+ case WINED3DSIH_M3x2:
nComponents = 2;
- tmpArg.opcode = shader_get_opcode(arg->shader, WINED3DSIO_DP3);
+ tmp_ins.handler_idx = WINED3DSIH_DP3;
break;
default:
break;
}
- for (i = 0; i < nComponents; i++) {
- tmpArg.dst = ((arg->dst) & ~WINED3DSP_WRITEMASK_ALL)|(WINED3DSP_WRITEMASK_0<<i);
- tmpArg.src[1] = arg->src[1]+i;
- shader_glsl_dot(&tmpArg);
+ tmp_dst = ins->dst[0];
+ tmp_src[0] = ins->src[0];
+ tmp_src[1] = ins->src[1];
+ for (i = 0; i < nComponents; ++i)
+ {
+ tmp_dst.write_mask = WINED3DSP_WRITEMASK_0 << i;
+ shader_glsl_dot(&tmp_ins);
+ ++tmp_src[1].reg.idx;
}
}
/**
- The LRP instruction performs a component-wise linear interpolation
+ The LRP instruction performs a component-wise linear interpolation
between the second and third operands using the first operand as the
blend factor. Equation: (dst = src2 + src0 * (src1 - src2))
This is equivalent to mix(src2, src1, src0);
*/
-void shader_glsl_lrp(SHADER_OPCODE_ARG* arg) {
+static void shader_glsl_lrp(const struct wined3d_shader_instruction *ins)
+{
glsl_src_param_t src0_param;
glsl_src_param_t src1_param;
glsl_src_param_t src2_param;
DWORD write_mask;
- write_mask = shader_glsl_append_dst(arg->buffer, arg);
+ write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], write_mask, &src0_param);
- shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], write_mask, &src1_param);
- shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], write_mask, &src2_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
+ shader_glsl_add_src_param(ins, &ins->src[1], write_mask, &src1_param);
+ shader_glsl_add_src_param(ins, &ins->src[2], write_mask, &src2_param);
- shader_addline(arg->buffer, "mix(%s, %s, %s));\n",
+ shader_addline(ins->ctx->buffer, "mix(%s, %s, %s));\n",
src2_param.param_str, src1_param.param_str, src0_param.param_str);
}
* dst.z = (src0.x > 0) ? ((src0.y > 0) ? pow(src0.y, src.w) : 0) : 0
* where src.w is clamped at +- 128
*/
-void shader_glsl_lit(SHADER_OPCODE_ARG* arg) {
+static void shader_glsl_lit(const struct wined3d_shader_instruction *ins)
+{
glsl_src_param_t src0_param;
glsl_src_param_t src1_param;
glsl_src_param_t src3_param;
char dst_mask[6];
- shader_glsl_append_dst(arg->buffer, arg);
- shader_glsl_get_write_mask(arg->dst, dst_mask);
+ shader_glsl_append_dst(ins->ctx->buffer, ins);
+ shader_glsl_get_write_mask(&ins->dst[0], dst_mask);
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_1, &src1_param);
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &src3_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src0_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_1, &src1_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_3, &src3_param);
/* The sdk specifies the instruction like this
* dst.x = 1.0;
* which sets dst.z to 0. If y > 0, but x = 0.0, we get pow(y * 0.0, power), which results in 0 too.
* if both x and y are > 0, we get pow(y * 1.0, power), as it is supposed to
*/
- shader_addline(arg->buffer, "vec4(1.0, max(%s, 0.0), pow(max(0.0, %s) * step(0.0, %s), clamp(%s, -128.0, 128.0)), 1.0)%s);\n",
- src0_param.param_str, src1_param.param_str, src0_param.param_str, src3_param.param_str, dst_mask);
+ shader_addline(ins->ctx->buffer,
+ "vec4(1.0, max(%s, 0.0), pow(max(0.0, %s) * step(0.0, %s), clamp(%s, -128.0, 128.0)), 1.0)%s);\n",
+ src0_param.param_str, src1_param.param_str, src0_param.param_str, src3_param.param_str, dst_mask);
}
/** Process the WINED3DSIO_DST instruction in GLSL:
* dst.z = src0.z
* dst.w = src1.w
*/
-void shader_glsl_dst(SHADER_OPCODE_ARG* arg) {
+static void shader_glsl_dst(const struct wined3d_shader_instruction *ins)
+{
glsl_src_param_t src0y_param;
glsl_src_param_t src0z_param;
glsl_src_param_t src1y_param;
glsl_src_param_t src1w_param;
char dst_mask[6];
- shader_glsl_append_dst(arg->buffer, arg);
- shader_glsl_get_write_mask(arg->dst, dst_mask);
+ shader_glsl_append_dst(ins->ctx->buffer, ins);
+ shader_glsl_get_write_mask(&ins->dst[0], dst_mask);
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_1, &src0y_param);
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_2, &src0z_param);
- shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_1, &src1y_param);
- shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_3, &src1w_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_1, &src0y_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_2, &src0z_param);
+ shader_glsl_add_src_param(ins, &ins->src[1], WINED3DSP_WRITEMASK_1, &src1y_param);
+ shader_glsl_add_src_param(ins, &ins->src[1], WINED3DSP_WRITEMASK_3, &src1w_param);
- shader_addline(arg->buffer, "vec4(1.0, %s * %s, %s, %s))%s;\n",
+ shader_addline(ins->ctx->buffer, "vec4(1.0, %s * %s, %s, %s))%s;\n",
src0y_param.param_str, src1y_param.param_str, src0z_param.param_str, src1w_param.param_str, dst_mask);
}
/** Process the WINED3DSIO_SINCOS instruction in GLSL:
* VS 2.0 requires that specific cosine and sine constants be passed to this instruction so the hardware
* can handle it. But, these functions are built-in for GLSL, so we can just ignore the last 2 params.
- *
+ *
* dst.x = cos(src0.?)
* dst.y = sin(src0.?)
* dst.z = dst.z
* dst.w = dst.w
*/
-void shader_glsl_sincos(SHADER_OPCODE_ARG* arg) {
+static void shader_glsl_sincos(const struct wined3d_shader_instruction *ins)
+{
glsl_src_param_t src0_param;
DWORD write_mask;
- write_mask = shader_glsl_append_dst(arg->buffer, arg);
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
+ write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
+ shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src0_param);
switch (write_mask) {
case WINED3DSP_WRITEMASK_0:
- shader_addline(arg->buffer, "cos(%s));\n", src0_param.param_str);
+ shader_addline(ins->ctx->buffer, "cos(%s));\n", src0_param.param_str);
break;
case WINED3DSP_WRITEMASK_1:
- shader_addline(arg->buffer, "sin(%s));\n", src0_param.param_str);
+ shader_addline(ins->ctx->buffer, "sin(%s));\n", src0_param.param_str);
break;
case (WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1):
- shader_addline(arg->buffer, "vec2(cos(%s), sin(%s)));\n", src0_param.param_str, src0_param.param_str);
+ shader_addline(ins->ctx->buffer, "vec2(cos(%s), sin(%s)));\n", src0_param.param_str, src0_param.param_str);
break;
default:
}
}
+/* sgn in vs_2_0 has 2 extra parameters(registers for temporary storage) which we don't use
+ * here. But those extra parameters require a dedicated function for sgn, since map2gl would
+ * generate invalid code
+ */
+static void shader_glsl_sgn(const struct wined3d_shader_instruction *ins)
+{
+ glsl_src_param_t src0_param;
+ DWORD write_mask;
+
+ write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
+ shader_glsl_add_src_param(ins, &ins->src[0], write_mask, &src0_param);
+
+ shader_addline(ins->ctx->buffer, "sign(%s));\n", src0_param.param_str);
+}
+
/** Process the WINED3DSIO_LOOP instruction in GLSL:
* Start a for() loop where src1.y is the initial value of aL,
* increment aL by src1.z for a total of src1.x iterations.
* Need to use a temporary variable for this operation.
*/
/* FIXME: I don't think nested loops will work correctly this way. */
-void shader_glsl_loop(SHADER_OPCODE_ARG* arg) {
+static void shader_glsl_loop(const struct wined3d_shader_instruction *ins)
+{
glsl_src_param_t src1_param;
- IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
- DWORD regtype = shader_get_regtype(arg->src[1]);
- DWORD reg = arg->src[1] & WINED3DSP_REGNUM_MASK;
+ IWineD3DBaseShaderImpl *shader = (IWineD3DBaseShaderImpl *)ins->ctx->shader;
const DWORD *control_values = NULL;
- local_constant *constant;
+ const local_constant *constant;
- shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_ALL, &src1_param);
+ shader_glsl_add_src_param(ins, &ins->src[1], WINED3DSP_WRITEMASK_ALL, &src1_param);
/* Try to hardcode the loop control parameters if possible. Direct3D 9 class hardware doesn't support real
* varying indexing, but Microsoft designed this feature for Shader model 2.x+. If the loop control is
* known at compile time, the GLSL compiler can unroll the loop, and replace indirect addressing with direct
* addressing.
*/
- if(regtype == WINED3DSPR_CONSTINT) {
+ if (ins->src[1].reg.type == WINED3DSPR_CONSTINT)
+ {
LIST_FOR_EACH_ENTRY(constant, &shader->baseShader.constantsI, local_constant, entry) {
- if(constant->idx == reg) {
+ if (constant->idx == ins->src[1].reg.idx)
+ {
control_values = constant->value;
break;
}
}
}
- if(control_values) {
- if(control_values[2] > 0) {
- shader_addline(arg->buffer, "for (aL%u = %d; aL%u < (%d * %d + %d); aL%u += %d) {\n",
- shader->baseShader.cur_loop_depth, control_values[1],
- shader->baseShader.cur_loop_depth, control_values[0], control_values[2], control_values[1],
- shader->baseShader.cur_loop_depth, control_values[2]);
- } else if(control_values[2] == 0) {
- shader_addline(arg->buffer, "for (aL%u = %d, tmpInt%u = 0; tmpInt%u < %d; tmpInt%u++) {\n",
- shader->baseShader.cur_loop_depth, control_values[1], shader->baseShader.cur_loop_depth,
- shader->baseShader.cur_loop_depth, control_values[0],
- shader->baseShader.cur_loop_depth);
- } else {
- shader_addline(arg->buffer, "for (aL%u = %d; aL%u > (%d * %d + %d); aL%u += %d) {\n",
- shader->baseShader.cur_loop_depth, control_values[1],
- shader->baseShader.cur_loop_depth, control_values[0], control_values[2], control_values[1],
- shader->baseShader.cur_loop_depth, control_values[2]);
+ if (control_values)
+ {
+ struct wined3d_shader_loop_control loop_control;
+ loop_control.count = control_values[0];
+ loop_control.start = control_values[1];
+ loop_control.step = (int)control_values[2];
+
+ if (loop_control.step > 0)
+ {
+ shader_addline(ins->ctx->buffer, "for (aL%u = %u; aL%u < (%u * %d + %u); aL%u += %d) {\n",
+ shader->baseShader.cur_loop_depth, loop_control.start,
+ shader->baseShader.cur_loop_depth, loop_control.count, loop_control.step, loop_control.start,
+ shader->baseShader.cur_loop_depth, loop_control.step);
+ }
+ else if (loop_control.step < 0)
+ {
+ shader_addline(ins->ctx->buffer, "for (aL%u = %u; aL%u > (%u * %d + %u); aL%u += %d) {\n",
+ shader->baseShader.cur_loop_depth, loop_control.start,
+ shader->baseShader.cur_loop_depth, loop_control.count, loop_control.step, loop_control.start,
+ shader->baseShader.cur_loop_depth, loop_control.step);
+ }
+ else
+ {
+ shader_addline(ins->ctx->buffer, "for (aL%u = %u, tmpInt%u = 0; tmpInt%u < %u; tmpInt%u++) {\n",
+ shader->baseShader.cur_loop_depth, loop_control.start, shader->baseShader.cur_loop_depth,
+ shader->baseShader.cur_loop_depth, loop_control.count,
+ shader->baseShader.cur_loop_depth);
}
} else {
- shader_addline(arg->buffer, "for (tmpInt%u = 0, aL%u = %s.y; tmpInt%u < %s.x; tmpInt%u++, aL%u += %s.z) {\n",
- shader->baseShader.cur_loop_depth, shader->baseShader.cur_loop_regno,
- src1_param.reg_name, shader->baseShader.cur_loop_depth, src1_param.reg_name,
- shader->baseShader.cur_loop_depth, shader->baseShader.cur_loop_regno, src1_param.reg_name);
+ shader_addline(ins->ctx->buffer,
+ "for (tmpInt%u = 0, aL%u = %s.y; tmpInt%u < %s.x; tmpInt%u++, aL%u += %s.z) {\n",
+ shader->baseShader.cur_loop_depth, shader->baseShader.cur_loop_regno,
+ src1_param.reg_name, shader->baseShader.cur_loop_depth, src1_param.reg_name,
+ shader->baseShader.cur_loop_depth, shader->baseShader.cur_loop_regno, src1_param.reg_name);
}
shader->baseShader.cur_loop_depth++;
shader->baseShader.cur_loop_regno++;
}
-void shader_glsl_end(SHADER_OPCODE_ARG* arg) {
- IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
+static void shader_glsl_end(const struct wined3d_shader_instruction *ins)
+{
+ IWineD3DBaseShaderImpl *shader = (IWineD3DBaseShaderImpl *)ins->ctx->shader;
- shader_addline(arg->buffer, "}\n");
+ shader_addline(ins->ctx->buffer, "}\n");
- if(arg->opcode->opcode == WINED3DSIO_ENDLOOP) {
+ if (ins->handler_idx == WINED3DSIH_ENDLOOP)
+ {
shader->baseShader.cur_loop_depth--;
shader->baseShader.cur_loop_regno--;
}
- if(arg->opcode->opcode == WINED3DSIO_ENDREP) {
+
+ if (ins->handler_idx == WINED3DSIH_ENDREP)
+ {
shader->baseShader.cur_loop_depth--;
}
}
-void shader_glsl_rep(SHADER_OPCODE_ARG* arg) {
- IWineD3DBaseShaderImpl* shader = (IWineD3DBaseShaderImpl*) arg->shader;
+static void shader_glsl_rep(const struct wined3d_shader_instruction *ins)
+{
+ IWineD3DBaseShaderImpl *shader = (IWineD3DBaseShaderImpl *)ins->ctx->shader;
glsl_src_param_t src0_param;
+ const DWORD *control_values = NULL;
+ const local_constant *constant;
+
+ /* Try to hardcode local values to help the GLSL compiler to unroll and optimize the loop */
+ if (ins->src[0].reg.type == WINED3DSPR_CONSTINT)
+ {
+ LIST_FOR_EACH_ENTRY(constant, &shader->baseShader.constantsI, local_constant, entry)
+ {
+ if (constant->idx == ins->src[0].reg.idx)
+ {
+ control_values = constant->value;
+ break;
+ }
+ }
+ }
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
- shader_addline(arg->buffer, "for (tmpInt%d = 0; tmpInt%d < %s; tmpInt%d++) {\n",
- shader->baseShader.cur_loop_depth, shader->baseShader.cur_loop_depth,
- src0_param.param_str, shader->baseShader.cur_loop_depth);
+ if(control_values) {
+ shader_addline(ins->ctx->buffer, "for (tmpInt%d = 0; tmpInt%d < %d; tmpInt%d++) {\n",
+ shader->baseShader.cur_loop_depth, shader->baseShader.cur_loop_depth,
+ control_values[0], shader->baseShader.cur_loop_depth);
+ } else {
+ shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src0_param);
+ shader_addline(ins->ctx->buffer, "for (tmpInt%d = 0; tmpInt%d < %s; tmpInt%d++) {\n",
+ shader->baseShader.cur_loop_depth, shader->baseShader.cur_loop_depth,
+ src0_param.param_str, shader->baseShader.cur_loop_depth);
+ }
shader->baseShader.cur_loop_depth++;
}
-void shader_glsl_if(SHADER_OPCODE_ARG* arg) {
+static void shader_glsl_if(const struct wined3d_shader_instruction *ins)
+{
glsl_src_param_t src0_param;
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
- shader_addline(arg->buffer, "if (%s) {\n", src0_param.param_str);
+ shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src0_param);
+ shader_addline(ins->ctx->buffer, "if (%s) {\n", src0_param.param_str);
}
-void shader_glsl_ifc(SHADER_OPCODE_ARG* arg) {
+static void shader_glsl_ifc(const struct wined3d_shader_instruction *ins)
+{
glsl_src_param_t src0_param;
glsl_src_param_t src1_param;
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
- shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0, &src1_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src0_param);
+ shader_glsl_add_src_param(ins, &ins->src[1], WINED3DSP_WRITEMASK_0, &src1_param);
- shader_addline(arg->buffer, "if (%s %s %s) {\n",
- src0_param.param_str, shader_get_comp_op(arg->opcode_token), src1_param.param_str);
+ shader_addline(ins->ctx->buffer, "if (%s %s %s) {\n",
+ src0_param.param_str, shader_get_comp_op(ins->flags), src1_param.param_str);
}
-void shader_glsl_else(SHADER_OPCODE_ARG* arg) {
- shader_addline(arg->buffer, "} else {\n");
+static void shader_glsl_else(const struct wined3d_shader_instruction *ins)
+{
+ shader_addline(ins->ctx->buffer, "} else {\n");
}
-void shader_glsl_break(SHADER_OPCODE_ARG* arg) {
- shader_addline(arg->buffer, "break;\n");
+static void shader_glsl_break(const struct wined3d_shader_instruction *ins)
+{
+ shader_addline(ins->ctx->buffer, "break;\n");
}
/* FIXME: According to MSDN the compare is done per component. */
-void shader_glsl_breakc(SHADER_OPCODE_ARG* arg) {
+static void shader_glsl_breakc(const struct wined3d_shader_instruction *ins)
+{
glsl_src_param_t src0_param;
glsl_src_param_t src1_param;
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0, &src0_param);
- shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0, &src1_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0, &src0_param);
+ shader_glsl_add_src_param(ins, &ins->src[1], WINED3DSP_WRITEMASK_0, &src1_param);
- shader_addline(arg->buffer, "if (%s %s %s) break;\n",
- src0_param.param_str, shader_get_comp_op(arg->opcode_token), src1_param.param_str);
+ shader_addline(ins->ctx->buffer, "if (%s %s %s) break;\n",
+ src0_param.param_str, shader_get_comp_op(ins->flags), src1_param.param_str);
}
-void shader_glsl_label(SHADER_OPCODE_ARG* arg) {
-
- DWORD snum = (arg->src[0]) & WINED3DSP_REGNUM_MASK;
- shader_addline(arg->buffer, "}\n");
- shader_addline(arg->buffer, "void subroutine%u () {\n", snum);
+static void shader_glsl_label(const struct wined3d_shader_instruction *ins)
+{
+ shader_addline(ins->ctx->buffer, "}\n");
+ shader_addline(ins->ctx->buffer, "void subroutine%u () {\n", ins->src[0].reg.idx);
}
-void shader_glsl_call(SHADER_OPCODE_ARG* arg) {
- DWORD snum = (arg->src[0]) & WINED3DSP_REGNUM_MASK;
- shader_addline(arg->buffer, "subroutine%u();\n", snum);
+static void shader_glsl_call(const struct wined3d_shader_instruction *ins)
+{
+ shader_addline(ins->ctx->buffer, "subroutine%u();\n", ins->src[0].reg.idx);
}
-void shader_glsl_callnz(SHADER_OPCODE_ARG* arg) {
+static void shader_glsl_callnz(const struct wined3d_shader_instruction *ins)
+{
glsl_src_param_t src1_param;
- DWORD snum = (arg->src[0]) & WINED3DSP_REGNUM_MASK;
- shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0, &src1_param);
- shader_addline(arg->buffer, "if (%s) subroutine%u();\n", src1_param.param_str, snum);
+ shader_glsl_add_src_param(ins, &ins->src[1], WINED3DSP_WRITEMASK_0, &src1_param);
+ shader_addline(ins->ctx->buffer, "if (%s) subroutine%u();\n", src1_param.param_str, ins->src[0].reg.idx);
+}
+
+static void shader_glsl_ret(const struct wined3d_shader_instruction *ins)
+{
+ /* No-op. The closing } is written when a new function is started, and at the end of the shader. This
+ * function only suppresses the unhandled instruction warning
+ */
}
/*********************************************
* Pixel Shader Specific Code begins here
********************************************/
-void pshader_glsl_tex(SHADER_OPCODE_ARG* arg) {
- IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
+static void shader_glsl_tex(const struct wined3d_shader_instruction *ins)
+{
+ IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *)ins->ctx->shader;
IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
- DWORD hex_version = This->baseShader.hex_version;
- char dst_swizzle[6];
+ DWORD shader_version = WINED3D_SHADER_VERSION(ins->ctx->reg_maps->shader_version.major,
+ ins->ctx->reg_maps->shader_version.minor);
glsl_sample_function_t sample_function;
- DWORD sampler_type;
+ DWORD sample_flags = 0;
+ WINED3DSAMPLER_TEXTURE_TYPE sampler_type;
DWORD sampler_idx;
- BOOL projected, texrect = FALSE;
- DWORD mask = 0;
-
- /* All versions have a destination register */
- shader_glsl_append_dst(arg->buffer, arg);
+ DWORD mask = 0, swizzle;
/* 1.0-1.4: Use destination register as sampler source.
* 2.0+: Use provided sampler source. */
- if (hex_version < WINED3DPS_VERSION(1,4)) {
- DWORD flags;
+ if (shader_version < WINED3D_SHADER_VERSION(2,0)) sampler_idx = ins->dst[0].reg.idx;
+ else sampler_idx = ins->src[1].reg.idx;
+ sampler_type = ins->ctx->reg_maps->sampler_type[sampler_idx];
- sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
- flags = deviceImpl->stateBlock->textureState[sampler_idx][WINED3DTSS_TEXTURETRANSFORMFLAGS];
+ if (shader_version < WINED3D_SHADER_VERSION(1,4))
+ {
+ DWORD flags = deviceImpl->stateBlock->textureState[sampler_idx][WINED3DTSS_TEXTURETRANSFORMFLAGS];
- if (flags & WINED3DTTFF_PROJECTED) {
- projected = TRUE;
+ /* Projected cube textures don't make a lot of sense, the resulting coordinates stay the same. */
+ if (flags & WINED3DTTFF_PROJECTED && sampler_type != WINED3DSTT_CUBE) {
+ sample_flags |= WINED3D_GLSL_SAMPLE_PROJECTED;
switch (flags & ~WINED3DTTFF_PROJECTED) {
case WINED3DTTFF_COUNT1: FIXME("WINED3DTTFF_PROJECTED with WINED3DTTFF_COUNT1?\n"); break;
case WINED3DTTFF_COUNT2: mask = WINED3DSP_WRITEMASK_1; break;
case WINED3DTTFF_COUNT4:
case WINED3DTTFF_DISABLE: mask = WINED3DSP_WRITEMASK_3; break;
}
- } else {
- projected = FALSE;
}
- } else if (hex_version < WINED3DPS_VERSION(2,0)) {
- DWORD src_mod = arg->src[0] & WINED3DSP_SRCMOD_MASK;
- sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
-
+ }
+ else if (shader_version < WINED3D_SHADER_VERSION(2,0))
+ {
+ DWORD src_mod = ins->src[0].modifiers;
+
if (src_mod == WINED3DSPSM_DZ) {
- projected = TRUE;
+ sample_flags |= WINED3D_GLSL_SAMPLE_PROJECTED;
mask = WINED3DSP_WRITEMASK_2;
} else if (src_mod == WINED3DSPSM_DW) {
- projected = TRUE;
+ sample_flags |= WINED3D_GLSL_SAMPLE_PROJECTED;
mask = WINED3DSP_WRITEMASK_3;
- } else {
- projected = FALSE;
}
} else {
- sampler_idx = arg->src[1] & WINED3DSP_REGNUM_MASK;
- if(arg->opcode_token & WINED3DSI_TEXLD_PROJECT) {
- /* ps 2.0 texldp instruction always divides by the fourth component. */
- projected = TRUE;
- mask = WINED3DSP_WRITEMASK_3;
- } else {
- projected = FALSE;
+ if (ins->flags & WINED3DSI_TEXLD_PROJECT)
+ {
+ /* ps 2.0 texldp instruction always divides by the fourth component. */
+ sample_flags |= WINED3D_GLSL_SAMPLE_PROJECTED;
+ mask = WINED3DSP_WRITEMASK_3;
}
}
if(deviceImpl->stateBlock->textures[sampler_idx] &&
IWineD3DBaseTexture_GetTextureDimensions(deviceImpl->stateBlock->textures[sampler_idx]) == GL_TEXTURE_RECTANGLE_ARB) {
- texrect = TRUE;
+ sample_flags |= WINED3D_GLSL_SAMPLE_RECT;
}
- sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
- shader_glsl_get_sample_function(sampler_type, projected, texrect, &sample_function);
+ shader_glsl_get_sample_function(sampler_type, sample_flags, &sample_function);
mask |= sample_function.coord_mask;
- if (hex_version < WINED3DPS_VERSION(2,0)) {
- shader_glsl_get_write_mask(arg->dst, dst_swizzle);
- } else {
- shader_glsl_get_swizzle(arg->src[1], FALSE, arg->dst, dst_swizzle);
- }
+ if (shader_version < WINED3D_SHADER_VERSION(2,0)) swizzle = WINED3DSP_NOSWIZZLE;
+ else swizzle = ins->src[1].swizzle;
/* 1.0-1.3: Use destination register as coordinate source.
1.4+: Use provided coordinate source register. */
- if (hex_version < WINED3DPS_VERSION(1,4)) {
+ if (shader_version < WINED3D_SHADER_VERSION(1,4))
+ {
char coord_mask[6];
- shader_glsl_get_write_mask(mask, coord_mask);
- shader_addline(arg->buffer, "%s(Psampler%u, T%u%s)%s);\n",
- sample_function.name, sampler_idx, sampler_idx, coord_mask, dst_swizzle);
+ shader_glsl_write_mask_to_str(mask, coord_mask);
+ shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, swizzle, NULL, NULL, NULL,
+ "T%u%s", sampler_idx, coord_mask);
} else {
glsl_src_param_t coord_param;
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], mask, &coord_param);
- if(arg->opcode_token & WINED3DSI_TEXLD_BIAS) {
+ shader_glsl_add_src_param(ins, &ins->src[0], mask, &coord_param);
+ if (ins->flags & WINED3DSI_TEXLD_BIAS)
+ {
glsl_src_param_t bias;
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &bias);
-
- shader_addline(arg->buffer, "%s(Psampler%u, %s, %s)%s);\n",
- sample_function.name, sampler_idx, coord_param.param_str,
- bias.param_str, dst_swizzle);
+ shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_3, &bias);
+ shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, swizzle, NULL, NULL, bias.param_str,
+ "%s", coord_param.param_str);
} else {
- shader_addline(arg->buffer, "%s(Psampler%u, %s)%s);\n",
- sample_function.name, sampler_idx, coord_param.param_str, dst_swizzle);
+ shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, swizzle, NULL, NULL, NULL,
+ "%s", coord_param.param_str);
}
}
}
-void shader_glsl_texldl(SHADER_OPCODE_ARG* arg) {
- IWineD3DBaseShaderImpl* This = (IWineD3DBaseShaderImpl*)arg->shader;
+static void shader_glsl_texldd(const struct wined3d_shader_instruction *ins)
+{
+ IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *)ins->ctx->shader;
IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
+ const struct wined3d_gl_info *gl_info = &deviceImpl->adapter->gl_info;
glsl_sample_function_t sample_function;
- glsl_src_param_t coord_param, lod_param;
- char dst_swizzle[6];
+ glsl_src_param_t coord_param, dx_param, dy_param;
+ DWORD sample_flags = WINED3D_GLSL_SAMPLE_GRAD;
DWORD sampler_type;
DWORD sampler_idx;
- BOOL texrect = FALSE;
+ DWORD swizzle = ins->src[1].swizzle;
+
+ if (!gl_info->supported[ARB_SHADER_TEXTURE_LOD])
+ {
+ FIXME("texldd used, but not supported by hardware. Falling back to regular tex\n");
+ return shader_glsl_tex(ins);
+ }
- shader_glsl_append_dst(arg->buffer, arg);
- shader_glsl_get_swizzle(arg->src[1], FALSE, arg->dst, dst_swizzle);
+ sampler_idx = ins->src[1].reg.idx;
+ sampler_type = ins->ctx->reg_maps->sampler_type[sampler_idx];
+ if(deviceImpl->stateBlock->textures[sampler_idx] &&
+ IWineD3DBaseTexture_GetTextureDimensions(deviceImpl->stateBlock->textures[sampler_idx]) == GL_TEXTURE_RECTANGLE_ARB) {
+ sample_flags |= WINED3D_GLSL_SAMPLE_RECT;
+ }
+
+ shader_glsl_get_sample_function(sampler_type, sample_flags, &sample_function);
+ shader_glsl_add_src_param(ins, &ins->src[0], sample_function.coord_mask, &coord_param);
+ shader_glsl_add_src_param(ins, &ins->src[2], sample_function.coord_mask, &dx_param);
+ shader_glsl_add_src_param(ins, &ins->src[3], sample_function.coord_mask, &dy_param);
- sampler_idx = arg->src[1] & WINED3DSP_REGNUM_MASK;
- sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
+ shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, swizzle, dx_param.param_str, dy_param.param_str, NULL,
+ "%s", coord_param.param_str);
+}
+
+static void shader_glsl_texldl(const struct wined3d_shader_instruction *ins)
+{
+ IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *)ins->ctx->shader;
+ IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
+ const struct wined3d_gl_info *gl_info = &deviceImpl->adapter->gl_info;
+ glsl_sample_function_t sample_function;
+ glsl_src_param_t coord_param, lod_param;
+ DWORD sample_flags = WINED3D_GLSL_SAMPLE_LOD;
+ DWORD sampler_type;
+ DWORD sampler_idx;
+ DWORD swizzle = ins->src[1].swizzle;
+
+ sampler_idx = ins->src[1].reg.idx;
+ sampler_type = ins->ctx->reg_maps->sampler_type[sampler_idx];
if(deviceImpl->stateBlock->textures[sampler_idx] &&
IWineD3DBaseTexture_GetTextureDimensions(deviceImpl->stateBlock->textures[sampler_idx]) == GL_TEXTURE_RECTANGLE_ARB) {
- texrect = TRUE;
+ sample_flags |= WINED3D_GLSL_SAMPLE_RECT;
}
- shader_glsl_get_sample_function(sampler_type, FALSE, texrect, &sample_function); shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], sample_function.coord_mask, &coord_param);
+ shader_glsl_get_sample_function(sampler_type, sample_flags, &sample_function);
+ shader_glsl_add_src_param(ins, &ins->src[0], sample_function.coord_mask, &coord_param);
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &lod_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_3, &lod_param);
- if (shader_is_pshader_version(This->baseShader.hex_version)) {
+ if (!gl_info->supported[ARB_SHADER_TEXTURE_LOD]
+ && shader_is_pshader_version(ins->ctx->reg_maps->shader_version.type))
+ {
/* The GLSL spec claims the Lod sampling functions are only supported in vertex shaders.
* However, they seem to work just fine in fragment shaders as well. */
- WARN("Using %sLod in fragment shader.\n", sample_function.name);
- shader_addline(arg->buffer, "%sLod(Psampler%u, %s, %s)%s);\n",
- sample_function.name, sampler_idx, coord_param.param_str, lod_param.param_str, dst_swizzle);
- } else {
- shader_addline(arg->buffer, "%sLod(Vsampler%u, %s, %s)%s);\n",
- sample_function.name, sampler_idx, coord_param.param_str, lod_param.param_str, dst_swizzle);
+ WARN("Using %s in fragment shader.\n", sample_function.name);
}
+ shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, swizzle, NULL, NULL, lod_param.param_str,
+ "%s", coord_param.param_str);
}
-void pshader_glsl_texcoord(SHADER_OPCODE_ARG* arg) {
-
+static void shader_glsl_texcoord(const struct wined3d_shader_instruction *ins)
+{
/* FIXME: Make this work for more than just 2D textures */
-
- IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
- SHADER_BUFFER* buffer = arg->buffer;
- DWORD hex_version = This->baseShader.hex_version;
- DWORD write_mask;
- char dst_mask[6];
+ struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
+ DWORD write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
- write_mask = shader_glsl_append_dst(arg->buffer, arg);
- shader_glsl_get_write_mask(write_mask, dst_mask);
+ if (!(ins->ctx->reg_maps->shader_version.major == 1 && ins->ctx->reg_maps->shader_version.minor == 4))
+ {
+ char dst_mask[6];
- if (hex_version != WINED3DPS_VERSION(1,4)) {
- DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
- shader_addline(buffer, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n", reg, dst_mask);
+ shader_glsl_get_write_mask(&ins->dst[0], dst_mask);
+ shader_addline(buffer, "clamp(gl_TexCoord[%u], 0.0, 1.0)%s);\n",
+ ins->dst[0].reg.idx, dst_mask);
} else {
- DWORD reg = arg->src[0] & WINED3DSP_REGNUM_MASK;
- DWORD src_mod = arg->src[0] & WINED3DSP_SRCMOD_MASK;
+ DWORD reg = ins->src[0].reg.idx;
+ DWORD src_mod = ins->src[0].modifiers;
char dst_swizzle[6];
- shader_glsl_get_swizzle(arg->src[0], FALSE, write_mask, dst_swizzle);
+ shader_glsl_get_swizzle(&ins->src[0], FALSE, write_mask, dst_swizzle);
if (src_mod == WINED3DSPSM_DZ) {
glsl_src_param_t div_param;
unsigned int mask_size = shader_glsl_get_write_mask_size(write_mask);
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_2, &div_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_2, &div_param);
if (mask_size > 1) {
shader_addline(buffer, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg, dst_swizzle, mask_size, div_param.param_str);
} else if (src_mod == WINED3DSPSM_DW) {
glsl_src_param_t div_param;
unsigned int mask_size = shader_glsl_get_write_mask_size(write_mask);
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_3, &div_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_3, &div_param);
if (mask_size > 1) {
shader_addline(buffer, "gl_TexCoord[%u]%s / vec%d(%s));\n", reg, dst_swizzle, mask_size, div_param.param_str);
/** Process the WINED3DSIO_TEXDP3TEX instruction in GLSL:
* Take a 3-component dot product of the TexCoord[dstreg] and src,
* then perform a 1D texture lookup from stage dstregnum, place into dst. */
-void pshader_glsl_texdp3tex(SHADER_OPCODE_ARG* arg) {
+static void shader_glsl_texdp3tex(const struct wined3d_shader_instruction *ins)
+{
glsl_src_param_t src0_param;
- char dst_mask[6];
glsl_sample_function_t sample_function;
- DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
+ DWORD sampler_idx = ins->dst[0].reg.idx;
DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
- DWORD sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
-
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
+ WINED3DSAMPLER_TEXTURE_TYPE sampler_type = ins->ctx->reg_maps->sampler_type[sampler_idx];
+ UINT mask_size;
- shader_glsl_append_dst(arg->buffer, arg);
- shader_glsl_get_write_mask(arg->dst, dst_mask);
+ shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
/* Do I have to take care about the projected bit? I don't think so, since the dp3 returns only one
* scalar, and projected sampling would require 4.
*
* It is a dependent read - not valid with conditional NP2 textures
*/
- shader_glsl_get_sample_function(sampler_type, FALSE, FALSE, &sample_function);
+ shader_glsl_get_sample_function(sampler_type, 0, &sample_function);
+ mask_size = shader_glsl_get_write_mask_size(sample_function.coord_mask);
- switch(count_bits(sample_function.coord_mask)) {
+ switch(mask_size)
+ {
case 1:
- shader_addline(arg->buffer, "%s(Psampler%u, dot(gl_TexCoord[%u].xyz, %s))%s);\n",
- sample_function.name, sampler_idx, sampler_idx, src0_param.param_str, dst_mask);
+ shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, WINED3DSP_NOSWIZZLE, NULL, NULL, NULL,
+ "dot(gl_TexCoord[%u].xyz, %s)", sampler_idx, src0_param.param_str);
break;
case 2:
- shader_addline(arg->buffer, "%s(Psampler%u, vec2(dot(gl_TexCoord[%u].xyz, %s), 0.0))%s);\n",
- sample_function.name, sampler_idx, sampler_idx, src0_param.param_str, dst_mask);
+ shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, WINED3DSP_NOSWIZZLE, NULL, NULL, NULL,
+ "vec2(dot(gl_TexCoord[%u].xyz, %s), 0.0)", sampler_idx, src0_param.param_str);
break;
case 3:
- shader_addline(arg->buffer, "%s(Psampler%u, vec3(dot(gl_TexCoord[%u].xyz, %s), 0.0, 0.0))%s);\n",
- sample_function.name, sampler_idx, sampler_idx, src0_param.param_str, dst_mask);
+ shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, WINED3DSP_NOSWIZZLE, NULL, NULL, NULL,
+ "vec3(dot(gl_TexCoord[%u].xyz, %s), 0.0, 0.0)", sampler_idx, src0_param.param_str);
break;
+
default:
- FIXME("Unexpected mask bitcount %d\n", count_bits(sample_function.coord_mask));
+ FIXME("Unexpected mask size %u\n", mask_size);
+ break;
}
}
/** Process the WINED3DSIO_TEXDP3 instruction in GLSL:
* Take a 3-component dot product of the TexCoord[dstreg] and src. */
-void pshader_glsl_texdp3(SHADER_OPCODE_ARG* arg) {
+static void shader_glsl_texdp3(const struct wined3d_shader_instruction *ins)
+{
glsl_src_param_t src0_param;
- DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
+ DWORD dstreg = ins->dst[0].reg.idx;
DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
DWORD dst_mask;
unsigned int mask_size;
- dst_mask = shader_glsl_append_dst(arg->buffer, arg);
+ dst_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
mask_size = shader_glsl_get_write_mask_size(dst_mask);
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
if (mask_size > 1) {
- shader_addline(arg->buffer, "vec%d(dot(T%u.xyz, %s)));\n", mask_size, dstreg, src0_param.param_str);
+ shader_addline(ins->ctx->buffer, "vec%d(dot(T%u.xyz, %s)));\n", mask_size, dstreg, src0_param.param_str);
} else {
- shader_addline(arg->buffer, "dot(T%u.xyz, %s));\n", dstreg, src0_param.param_str);
+ shader_addline(ins->ctx->buffer, "dot(T%u.xyz, %s));\n", dstreg, src0_param.param_str);
}
}
/** Process the WINED3DSIO_TEXDEPTH instruction in GLSL:
* Calculate the depth as dst.x / dst.y */
-void pshader_glsl_texdepth(SHADER_OPCODE_ARG* arg) {
+static void shader_glsl_texdepth(const struct wined3d_shader_instruction *ins)
+{
glsl_dst_param_t dst_param;
- shader_glsl_add_dst_param(arg, arg->dst, 0, &dst_param);
+ shader_glsl_add_dst_param(ins, &ins->dst[0], &dst_param);
/* Tests show that texdepth never returns anything below 0.0, and that r5.y is clamped to 1.0.
* Negative input is accepted, -0.25 / -0.5 returns 0.5. GL should clamp gl_FragDepth to [0;1], but
* too is irrelevant, since if x = 0, any y value < 1.0 (and > 1.0 is not allowed) results in a result
* >= 1.0 or < 0.0
*/
- shader_addline(arg->buffer, "gl_FragDepth = clamp((%s.x / min(%s.y, 1.0)), 0.0, 1.0);\n", dst_param.reg_name, dst_param.reg_name);
+ shader_addline(ins->ctx->buffer, "gl_FragDepth = clamp((%s.x / min(%s.y, 1.0)), 0.0, 1.0);\n",
+ dst_param.reg_name, dst_param.reg_name);
}
/** Process the WINED3DSIO_TEXM3X2DEPTH instruction in GLSL:
* Calculate tmp0.y = TexCoord[dstreg] . src.xyz; (tmp0.x has already been calculated)
* depth = (tmp0.y == 0.0) ? 1.0 : tmp0.x / tmp0.y
*/
-void pshader_glsl_texm3x2depth(SHADER_OPCODE_ARG* arg) {
+static void shader_glsl_texm3x2depth(const struct wined3d_shader_instruction *ins)
+{
DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
- DWORD dstreg = arg->dst & WINED3DSP_REGNUM_MASK;
+ DWORD dstreg = ins->dst[0].reg.idx;
glsl_src_param_t src0_param;
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
- shader_addline(arg->buffer, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg, src0_param.param_str);
- shader_addline(arg->buffer, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
+ shader_addline(ins->ctx->buffer, "tmp0.y = dot(T%u.xyz, %s);\n", dstreg, src0_param.param_str);
+ shader_addline(ins->ctx->buffer, "gl_FragDepth = (tmp0.y == 0.0) ? 1.0 : clamp(tmp0.x / tmp0.y, 0.0, 1.0);\n");
}
/** Process the WINED3DSIO_TEXM3X2PAD instruction in GLSL
* Calculate the 1st of a 2-row matrix multiplication. */
-void pshader_glsl_texm3x2pad(SHADER_OPCODE_ARG* arg) {
+static void shader_glsl_texm3x2pad(const struct wined3d_shader_instruction *ins)
+{
DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
- DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
- SHADER_BUFFER* buffer = arg->buffer;
+ DWORD reg = ins->dst[0].reg.idx;
+ struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
glsl_src_param_t src0_param;
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
shader_addline(buffer, "tmp0.x = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
}
/** Process the WINED3DSIO_TEXM3X3PAD instruction in GLSL
* Calculate the 1st or 2nd row of a 3-row matrix multiplication. */
-void pshader_glsl_texm3x3pad(SHADER_OPCODE_ARG* arg) {
-
- IWineD3DPixelShaderImpl* shader = (IWineD3DPixelShaderImpl*) arg->shader;
+static void shader_glsl_texm3x3pad(const struct wined3d_shader_instruction *ins)
+{
+ IWineD3DPixelShaderImpl *shader = (IWineD3DPixelShaderImpl *)ins->ctx->shader;
DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
- DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
- SHADER_BUFFER* buffer = arg->buffer;
+ DWORD reg = ins->dst[0].reg.idx;
+ struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
SHADER_PARSE_STATE* current_state = &shader->baseShader.parse_state;
glsl_src_param_t src0_param;
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
shader_addline(buffer, "tmp0.%c = dot(T%u.xyz, %s);\n", 'x' + current_state->current_row, reg, src0_param.param_str);
current_state->texcoord_w[current_state->current_row++] = reg;
}
-void pshader_glsl_texm3x2tex(SHADER_OPCODE_ARG* arg) {
+static void shader_glsl_texm3x2tex(const struct wined3d_shader_instruction *ins)
+{
DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
- DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
- SHADER_BUFFER* buffer = arg->buffer;
+ DWORD reg = ins->dst[0].reg.idx;
+ struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
glsl_src_param_t src0_param;
- char dst_mask[6];
+ WINED3DSAMPLER_TEXTURE_TYPE sampler_type = ins->ctx->reg_maps->sampler_type[reg];
+ glsl_sample_function_t sample_function;
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
shader_addline(buffer, "tmp0.y = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
- shader_glsl_append_dst(buffer, arg);
- shader_glsl_get_write_mask(arg->dst, dst_mask);
+ shader_glsl_get_sample_function(sampler_type, 0, &sample_function);
/* Sample the texture using the calculated coordinates */
- shader_addline(buffer, "texture2D(Psampler%u, tmp0.xy)%s);\n", reg, dst_mask);
+ shader_glsl_gen_sample_code(ins, reg, &sample_function, WINED3DSP_NOSWIZZLE, NULL, NULL, NULL, "tmp0.xy");
}
/** Process the WINED3DSIO_TEXM3X3TEX instruction in GLSL
* Perform the 3rd row of a 3x3 matrix multiply, then sample the texture using the calculated coordinates */
-void pshader_glsl_texm3x3tex(SHADER_OPCODE_ARG* arg) {
+static void shader_glsl_texm3x3tex(const struct wined3d_shader_instruction *ins)
+{
DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
glsl_src_param_t src0_param;
- char dst_mask[6];
- DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
- IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
+ DWORD reg = ins->dst[0].reg.idx;
+ IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *)ins->ctx->shader;
SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
- DWORD sampler_type = arg->reg_maps->samplers[reg] & WINED3DSP_TEXTURETYPE_MASK;
+ WINED3DSAMPLER_TEXTURE_TYPE sampler_type = ins->ctx->reg_maps->sampler_type[reg];
glsl_sample_function_t sample_function;
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
- shader_addline(arg->buffer, "tmp0.z = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
+ shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
+ shader_addline(ins->ctx->buffer, "tmp0.z = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
- shader_glsl_append_dst(arg->buffer, arg);
- shader_glsl_get_write_mask(arg->dst, dst_mask);
/* Dependent read, not valid with conditional NP2 */
- shader_glsl_get_sample_function(sampler_type, FALSE, FALSE, &sample_function);
+ shader_glsl_get_sample_function(sampler_type, 0, &sample_function);
/* Sample the texture using the calculated coordinates */
- shader_addline(arg->buffer, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function.name, reg, dst_mask);
+ shader_glsl_gen_sample_code(ins, reg, &sample_function, WINED3DSP_NOSWIZZLE, NULL, NULL, NULL, "tmp0.xyz");
current_state->current_row = 0;
}
/** Process the WINED3DSIO_TEXM3X3 instruction in GLSL
* Perform the 3rd row of a 3x3 matrix multiply */
-void pshader_glsl_texm3x3(SHADER_OPCODE_ARG* arg) {
+static void shader_glsl_texm3x3(const struct wined3d_shader_instruction *ins)
+{
DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
glsl_src_param_t src0_param;
char dst_mask[6];
- DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
- IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
+ DWORD reg = ins->dst[0].reg.idx;
+ IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *)ins->ctx->shader;
SHADER_PARSE_STATE* current_state = &This->baseShader.parse_state;
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
- shader_glsl_append_dst(arg->buffer, arg);
- shader_glsl_get_write_mask(arg->dst, dst_mask);
- shader_addline(arg->buffer, "vec4(tmp0.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg, src0_param.param_str, dst_mask);
+ shader_glsl_append_dst(ins->ctx->buffer, ins);
+ shader_glsl_get_write_mask(&ins->dst[0], dst_mask);
+ shader_addline(ins->ctx->buffer, "vec4(tmp0.xy, dot(T%u.xyz, %s), 1.0)%s);\n", reg, src0_param.param_str, dst_mask);
current_state->current_row = 0;
}
-/** Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
- * Peform the final texture lookup based on the previous 2 3x3 matrix multiplies */
-void pshader_glsl_texm3x3spec(SHADER_OPCODE_ARG* arg) {
-
- IWineD3DPixelShaderImpl* shader = (IWineD3DPixelShaderImpl*) arg->shader;
- DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
+/* Process the WINED3DSIO_TEXM3X3SPEC instruction in GLSL
+ * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
+static void shader_glsl_texm3x3spec(const struct wined3d_shader_instruction *ins)
+{
+ IWineD3DPixelShaderImpl *shader = (IWineD3DPixelShaderImpl *)ins->ctx->shader;
+ DWORD reg = ins->dst[0].reg.idx;
glsl_src_param_t src0_param;
glsl_src_param_t src1_param;
- char dst_mask[6];
- SHADER_BUFFER* buffer = arg->buffer;
+ struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
SHADER_PARSE_STATE* current_state = &shader->baseShader.parse_state;
- DWORD stype = arg->reg_maps->samplers[reg] & WINED3DSP_TEXTURETYPE_MASK;
+ WINED3DSAMPLER_TEXTURE_TYPE stype = ins->ctx->reg_maps->sampler_type[reg];
DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
glsl_sample_function_t sample_function;
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
- shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], src_mask, &src1_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
+ shader_glsl_add_src_param(ins, &ins->src[1], src_mask, &src1_param);
/* Perform the last matrix multiply operation */
shader_addline(buffer, "tmp0.z = dot(T%u.xyz, %s);\n", reg, src0_param.param_str);
/* Reflection calculation */
shader_addline(buffer, "tmp0.xyz = -reflect((%s), normalize(tmp0.xyz));\n", src1_param.param_str);
- shader_glsl_append_dst(buffer, arg);
- shader_glsl_get_write_mask(arg->dst, dst_mask);
/* Dependent read, not valid with conditional NP2 */
- shader_glsl_get_sample_function(stype, FALSE, FALSE, &sample_function);
+ shader_glsl_get_sample_function(stype, 0, &sample_function);
/* Sample the texture */
- shader_addline(buffer, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function.name, reg, dst_mask);
+ shader_glsl_gen_sample_code(ins, reg, &sample_function, WINED3DSP_NOSWIZZLE, NULL, NULL, NULL, "tmp0.xyz");
current_state->current_row = 0;
}
-/** Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
- * Peform the final texture lookup based on the previous 2 3x3 matrix multiplies */
-void pshader_glsl_texm3x3vspec(SHADER_OPCODE_ARG* arg) {
-
- IWineD3DPixelShaderImpl* shader = (IWineD3DPixelShaderImpl*) arg->shader;
- DWORD reg = arg->dst & WINED3DSP_REGNUM_MASK;
- SHADER_BUFFER* buffer = arg->buffer;
+/* Process the WINED3DSIO_TEXM3X3VSPEC instruction in GLSL
+ * Perform the final texture lookup based on the previous 2 3x3 matrix multiplies */
+static void shader_glsl_texm3x3vspec(const struct wined3d_shader_instruction *ins)
+{
+ IWineD3DPixelShaderImpl *shader = (IWineD3DPixelShaderImpl *)ins->ctx->shader;
+ DWORD reg = ins->dst[0].reg.idx;
+ struct wined3d_shader_buffer *buffer = ins->ctx->buffer;
SHADER_PARSE_STATE* current_state = &shader->baseShader.parse_state;
glsl_src_param_t src0_param;
- char dst_mask[6];
DWORD src_mask = WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1 | WINED3DSP_WRITEMASK_2;
- DWORD sampler_type = arg->reg_maps->samplers[reg] & WINED3DSP_TEXTURETYPE_MASK;
+ WINED3DSAMPLER_TEXTURE_TYPE sampler_type = ins->ctx->reg_maps->sampler_type[reg];
glsl_sample_function_t sample_function;
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], src_mask, &src0_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], src_mask, &src0_param);
/* Perform the last matrix multiply operation */
shader_addline(buffer, "tmp0.z = dot(vec3(T%u), vec3(%s));\n", reg, src0_param.param_str);
current_state->texcoord_w[0], current_state->texcoord_w[1], reg);
shader_addline(buffer, "tmp0.xyz = -reflect(tmp1.xyz, normalize(tmp0.xyz));\n");
- shader_glsl_append_dst(buffer, arg);
- shader_glsl_get_write_mask(arg->dst, dst_mask);
/* Dependent read, not valid with conditional NP2 */
- shader_glsl_get_sample_function(sampler_type, FALSE, FALSE, &sample_function);
+ shader_glsl_get_sample_function(sampler_type, 0, &sample_function);
/* Sample the texture using the calculated coordinates */
- shader_addline(buffer, "%s(Psampler%u, tmp0.xyz)%s);\n", sample_function.name, reg, dst_mask);
+ shader_glsl_gen_sample_code(ins, reg, &sample_function, WINED3DSP_NOSWIZZLE, NULL, NULL, NULL, "tmp0.xyz");
current_state->current_row = 0;
}
* Apply a fake bump map transform.
* texbem is pshader <= 1.3 only, this saves a few version checks
*/
-void pshader_glsl_texbem(SHADER_OPCODE_ARG* arg) {
- IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
+static void shader_glsl_texbem(const struct wined3d_shader_instruction *ins)
+{
+ IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *)ins->ctx->shader;
IWineD3DDeviceImpl* deviceImpl = (IWineD3DDeviceImpl*) This->baseShader.device;
- char dst_swizzle[6];
glsl_sample_function_t sample_function;
glsl_src_param_t coord_param;
- DWORD sampler_type;
+ WINED3DSAMPLER_TEXTURE_TYPE sampler_type;
DWORD sampler_idx;
DWORD mask;
DWORD flags;
char coord_mask[6];
- sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
+ sampler_idx = ins->dst[0].reg.idx;
flags = deviceImpl->stateBlock->textureState[sampler_idx][WINED3DTSS_TEXTURETRANSFORMFLAGS];
- sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
+ sampler_type = ins->ctx->reg_maps->sampler_type[sampler_idx];
/* Dependent read, not valid with conditional NP2 */
- shader_glsl_get_sample_function(sampler_type, FALSE, FALSE, &sample_function);
+ shader_glsl_get_sample_function(sampler_type, 0, &sample_function);
mask = sample_function.coord_mask;
- shader_glsl_get_write_mask(arg->dst, dst_swizzle);
-
- shader_glsl_get_write_mask(mask, coord_mask);
+ shader_glsl_write_mask_to_str(mask, coord_mask);
/* with projective textures, texbem only divides the static texture coord, not the displacement,
* so we can't let the GL handle this.
case WINED3DTTFF_COUNT4:
case WINED3DTTFF_DISABLE: div_mask = WINED3DSP_WRITEMASK_3; break;
}
- shader_glsl_get_write_mask(div_mask, coord_div_mask);
- shader_addline(arg->buffer, "T%u%s /= T%u%s;\n", sampler_idx, coord_mask, sampler_idx, coord_div_mask);
+ shader_glsl_write_mask_to_str(div_mask, coord_div_mask);
+ shader_addline(ins->ctx->buffer, "T%u%s /= T%u%s;\n", sampler_idx, coord_mask, sampler_idx, coord_div_mask);
}
- shader_glsl_append_dst(arg->buffer, arg);
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0|WINED3DSP_WRITEMASK_1, &coord_param);
- if(arg->opcode->opcode == WINED3DSIO_TEXBEML) {
+ shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &coord_param);
+
+ shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, WINED3DSP_NOSWIZZLE, NULL, NULL, NULL,
+ "T%u%s + vec4(bumpenvmat%d * %s, 0.0, 0.0)%s", sampler_idx, coord_mask, sampler_idx,
+ coord_param.param_str, coord_mask);
+
+ if (ins->handler_idx == WINED3DSIH_TEXBEML)
+ {
glsl_src_param_t luminance_param;
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_2, &luminance_param);
- shader_addline(arg->buffer, "(%s(Psampler%u, T%u%s + vec4(bumpenvmat%d * %s, 0.0, 0.0)%s )*(%s * luminancescale%d + luminanceoffset%d))%s);\n",
- sample_function.name, sampler_idx, sampler_idx, coord_mask, sampler_idx, coord_param.param_str, coord_mask,
- luminance_param.param_str, sampler_idx, sampler_idx, dst_swizzle);
- } else {
- shader_addline(arg->buffer, "%s(Psampler%u, T%u%s + vec4(bumpenvmat%d * %s, 0.0, 0.0)%s )%s);\n",
- sample_function.name, sampler_idx, sampler_idx, coord_mask, sampler_idx, coord_param.param_str, coord_mask, dst_swizzle);
+ glsl_dst_param_t dst_param;
+
+ shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_2, &luminance_param);
+ shader_glsl_add_dst_param(ins, &ins->dst[0], &dst_param);
+
+ shader_addline(ins->ctx->buffer, "%s%s *= (%s * luminancescale%d + luminanceoffset%d);\n",
+ dst_param.reg_name, dst_param.mask_str,
+ luminance_param.param_str, sampler_idx, sampler_idx);
}
}
-void pshader_glsl_bem(SHADER_OPCODE_ARG* arg) {
+static void shader_glsl_bem(const struct wined3d_shader_instruction *ins)
+{
glsl_src_param_t src0_param, src1_param;
- DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
+ DWORD sampler_idx = ins->dst[0].reg.idx;
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0|WINED3DSP_WRITEMASK_1, &src0_param);
- shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0|WINED3DSP_WRITEMASK_1, &src1_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &src0_param);
+ shader_glsl_add_src_param(ins, &ins->src[1], WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &src1_param);
- shader_glsl_append_dst(arg->buffer, arg);
- shader_addline(arg->buffer, "%s + bumpenvmat%d * %s);\n",
- src0_param.param_str, sampler_idx, src1_param.param_str);
+ shader_glsl_append_dst(ins->ctx->buffer, ins);
+ shader_addline(ins->ctx->buffer, "%s + bumpenvmat%d * %s);\n",
+ src0_param.param_str, sampler_idx, src1_param.param_str);
}
/** Process the WINED3DSIO_TEXREG2AR instruction in GLSL
* Sample 2D texture at dst using the alpha & red (wx) components of src as texture coordinates */
-void pshader_glsl_texreg2ar(SHADER_OPCODE_ARG* arg) {
-
+static void shader_glsl_texreg2ar(const struct wined3d_shader_instruction *ins)
+{
glsl_src_param_t src0_param;
- DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
- char dst_mask[6];
+ DWORD sampler_idx = ins->dst[0].reg.idx;
+ WINED3DSAMPLER_TEXTURE_TYPE sampler_type = ins->ctx->reg_maps->sampler_type[sampler_idx];
+ glsl_sample_function_t sample_function;
- shader_glsl_append_dst(arg->buffer, arg);
- shader_glsl_get_write_mask(arg->dst, dst_mask);
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_ALL, &src0_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_ALL, &src0_param);
- shader_addline(arg->buffer, "texture2D(Psampler%u, %s.wx)%s);\n", sampler_idx, src0_param.reg_name, dst_mask);
+ shader_glsl_get_sample_function(sampler_type, 0, &sample_function);
+ shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, WINED3DSP_NOSWIZZLE, NULL, NULL, NULL,
+ "%s.wx", src0_param.reg_name);
}
/** Process the WINED3DSIO_TEXREG2GB instruction in GLSL
* Sample 2D texture at dst using the green & blue (yz) components of src as texture coordinates */
-void pshader_glsl_texreg2gb(SHADER_OPCODE_ARG* arg) {
+static void shader_glsl_texreg2gb(const struct wined3d_shader_instruction *ins)
+{
glsl_src_param_t src0_param;
- DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
- char dst_mask[6];
+ DWORD sampler_idx = ins->dst[0].reg.idx;
+ WINED3DSAMPLER_TEXTURE_TYPE sampler_type = ins->ctx->reg_maps->sampler_type[sampler_idx];
+ glsl_sample_function_t sample_function;
- shader_glsl_append_dst(arg->buffer, arg);
- shader_glsl_get_write_mask(arg->dst, dst_mask);
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_ALL, &src0_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_ALL, &src0_param);
- shader_addline(arg->buffer, "texture2D(Psampler%u, %s.yz)%s);\n", sampler_idx, src0_param.reg_name, dst_mask);
+ shader_glsl_get_sample_function(sampler_type, 0, &sample_function);
+ shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, WINED3DSP_NOSWIZZLE, NULL, NULL, NULL,
+ "%s.yz", src0_param.reg_name);
}
/** Process the WINED3DSIO_TEXREG2RGB instruction in GLSL
* Sample texture at dst using the rgb (xyz) components of src as texture coordinates */
-void pshader_glsl_texreg2rgb(SHADER_OPCODE_ARG* arg) {
+static void shader_glsl_texreg2rgb(const struct wined3d_shader_instruction *ins)
+{
glsl_src_param_t src0_param;
- char dst_mask[6];
- DWORD sampler_idx = arg->dst & WINED3DSP_REGNUM_MASK;
- DWORD sampler_type = arg->reg_maps->samplers[sampler_idx] & WINED3DSP_TEXTURETYPE_MASK;
+ DWORD sampler_idx = ins->dst[0].reg.idx;
+ WINED3DSAMPLER_TEXTURE_TYPE sampler_type = ins->ctx->reg_maps->sampler_type[sampler_idx];
glsl_sample_function_t sample_function;
- shader_glsl_append_dst(arg->buffer, arg);
- shader_glsl_get_write_mask(arg->dst, dst_mask);
/* Dependent read, not valid with conditional NP2 */
- shader_glsl_get_sample_function(sampler_type, FALSE, FALSE, &sample_function);
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], sample_function.coord_mask, &src0_param);
+ shader_glsl_get_sample_function(sampler_type, 0, &sample_function);
+ shader_glsl_add_src_param(ins, &ins->src[0], sample_function.coord_mask, &src0_param);
- shader_addline(arg->buffer, "%s(Psampler%u, %s)%s);\n", sample_function.name, sampler_idx, src0_param.param_str, dst_mask);
+ shader_glsl_gen_sample_code(ins, sampler_idx, &sample_function, WINED3DSP_NOSWIZZLE, NULL, NULL, NULL,
+ "%s", src0_param.param_str);
}
/** Process the WINED3DSIO_TEXKILL instruction in GLSL.
* If any of the first 3 components are < 0, discard this pixel */
-void pshader_glsl_texkill(SHADER_OPCODE_ARG* arg) {
- IWineD3DPixelShaderImpl* This = (IWineD3DPixelShaderImpl*) arg->shader;
- DWORD hex_version = This->baseShader.hex_version;
+static void shader_glsl_texkill(const struct wined3d_shader_instruction *ins)
+{
glsl_dst_param_t dst_param;
/* The argument is a destination parameter, and no writemasks are allowed */
- shader_glsl_add_dst_param(arg, arg->dst, 0, &dst_param);
- if((hex_version >= WINED3DPS_VERSION(2,0))) {
+ shader_glsl_add_dst_param(ins, &ins->dst[0], &dst_param);
+ if (ins->ctx->reg_maps->shader_version.major >= 2)
+ {
/* 2.0 shaders compare all 4 components in texkill */
- shader_addline(arg->buffer, "if (any(lessThan(%s.xyzw, vec4(0.0)))) discard;\n", dst_param.reg_name);
+ shader_addline(ins->ctx->buffer, "if (any(lessThan(%s.xyzw, vec4(0.0)))) discard;\n", dst_param.reg_name);
} else {
/* 1.X shaders only compare the first 3 components, probably due to the nature of the texkill
* instruction as a tex* instruction, and phase, which kills all a / w components. Even if all
* 4 components are defined, only the first 3 are used
*/
- shader_addline(arg->buffer, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param.reg_name);
+ shader_addline(ins->ctx->buffer, "if (any(lessThan(%s.xyz, vec3(0.0)))) discard;\n", dst_param.reg_name);
}
}
/** Process the WINED3DSIO_DP2ADD instruction in GLSL.
* dst = dot2(src0, src1) + src2 */
-void pshader_glsl_dp2add(SHADER_OPCODE_ARG* arg) {
+static void shader_glsl_dp2add(const struct wined3d_shader_instruction *ins)
+{
glsl_src_param_t src0_param;
glsl_src_param_t src1_param;
glsl_src_param_t src2_param;
DWORD write_mask;
unsigned int mask_size;
- write_mask = shader_glsl_append_dst(arg->buffer, arg);
+ write_mask = shader_glsl_append_dst(ins->ctx->buffer, ins);
mask_size = shader_glsl_get_write_mask_size(write_mask);
- shader_glsl_add_src_param(arg, arg->src[0], arg->src_addr[0], WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &src0_param);
- shader_glsl_add_src_param(arg, arg->src[1], arg->src_addr[1], WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &src1_param);
- shader_glsl_add_src_param(arg, arg->src[2], arg->src_addr[2], WINED3DSP_WRITEMASK_0, &src2_param);
+ shader_glsl_add_src_param(ins, &ins->src[0], WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &src0_param);
+ shader_glsl_add_src_param(ins, &ins->src[1], WINED3DSP_WRITEMASK_0 | WINED3DSP_WRITEMASK_1, &src1_param);
+ shader_glsl_add_src_param(ins, &ins->src[2], WINED3DSP_WRITEMASK_0, &src2_param);
if (mask_size > 1) {
- shader_addline(arg->buffer, "vec%d(dot(%s, %s) + %s));\n", mask_size, src0_param.param_str, src1_param.param_str, src2_param.param_str);
+ shader_addline(ins->ctx->buffer, "vec%d(dot(%s, %s) + %s));\n",
+ mask_size, src0_param.param_str, src1_param.param_str, src2_param.param_str);
} else {
- shader_addline(arg->buffer, "dot(%s, %s) + %s);\n", src0_param.param_str, src1_param.param_str, src2_param.param_str);
+ shader_addline(ins->ctx->buffer, "dot(%s, %s) + %s);\n",
+ src0_param.param_str, src1_param.param_str, src2_param.param_str);
}
}
-static void pshader_glsl_input_pack(
- SHADER_BUFFER* buffer,
- semantic* semantics_in,
- IWineD3DPixelShader *iface) {
-
- unsigned int i;
- IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *) iface;
-
- for (i = 0; i < MAX_REG_INPUT; i++) {
-
- DWORD usage_token = semantics_in[i].usage;
- DWORD register_token = semantics_in[i].reg;
- DWORD usage, usage_idx;
- char reg_mask[6];
-
- /* Uninitialized */
- if (!usage_token) continue;
- usage = (usage_token & WINED3DSP_DCL_USAGE_MASK) >> WINED3DSP_DCL_USAGE_SHIFT;
- usage_idx = (usage_token & WINED3DSP_DCL_USAGEINDEX_MASK) >> WINED3DSP_DCL_USAGEINDEX_SHIFT;
- shader_glsl_get_write_mask(register_token, reg_mask);
-
- switch(usage) {
-
- case WINED3DDECLUSAGE_TEXCOORD:
- if(usage_idx < 8 && This->vertexprocessing == pretransformed) {
- shader_addline(buffer, "IN[%u]%s = gl_TexCoord[%u]%s;\n",
- This->input_reg_map[i], reg_mask, usage_idx, reg_mask);
- } else {
- shader_addline(buffer, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
- This->input_reg_map[i], reg_mask, reg_mask);
- }
- break;
-
- case WINED3DDECLUSAGE_COLOR:
- if (usage_idx == 0)
- shader_addline(buffer, "IN[%u]%s = vec4(gl_Color)%s;\n",
- This->input_reg_map[i], reg_mask, reg_mask);
- else if (usage_idx == 1)
- shader_addline(buffer, "IN[%u]%s = vec4(gl_SecondaryColor)%s;\n",
- This->input_reg_map[i], reg_mask, reg_mask);
- else
- shader_addline(buffer, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
- This->input_reg_map[i], reg_mask, reg_mask);
- break;
+static void shader_glsl_input_pack(IWineD3DPixelShader *iface, struct wined3d_shader_buffer *buffer,
+ const struct wined3d_shader_signature_element *input_signature, const struct shader_reg_maps *reg_maps,
+ enum vertexprocessing_mode vertexprocessing)
+{
+ unsigned int i;
+ IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *)iface;
+ WORD map = reg_maps->input_registers;
- default:
- shader_addline(buffer, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
- This->input_reg_map[i], reg_mask, reg_mask);
+ for (i = 0; map; map >>= 1, ++i)
+ {
+ const char *semantic_name;
+ UINT semantic_idx;
+ char reg_mask[6];
+
+ /* Unused */
+ if (!(map & 1)) continue;
+
+ semantic_name = input_signature[i].semantic_name;
+ semantic_idx = input_signature[i].semantic_idx;
+ shader_glsl_write_mask_to_str(input_signature[i].mask, reg_mask);
+
+ if (shader_match_semantic(semantic_name, WINED3DDECLUSAGE_TEXCOORD))
+ {
+ if (semantic_idx < 8 && vertexprocessing == pretransformed)
+ shader_addline(buffer, "IN[%u]%s = gl_TexCoord[%u]%s;\n",
+ This->input_reg_map[i], reg_mask, semantic_idx, reg_mask);
+ else
+ shader_addline(buffer, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
+ This->input_reg_map[i], reg_mask, reg_mask);
+ }
+ else if (shader_match_semantic(semantic_name, WINED3DDECLUSAGE_COLOR))
+ {
+ if (semantic_idx == 0)
+ shader_addline(buffer, "IN[%u]%s = vec4(gl_Color)%s;\n",
+ This->input_reg_map[i], reg_mask, reg_mask);
+ else if (semantic_idx == 1)
+ shader_addline(buffer, "IN[%u]%s = vec4(gl_SecondaryColor)%s;\n",
+ This->input_reg_map[i], reg_mask, reg_mask);
+ else
+ shader_addline(buffer, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
+ This->input_reg_map[i], reg_mask, reg_mask);
+ }
+ else
+ {
+ shader_addline(buffer, "IN[%u]%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
+ This->input_reg_map[i], reg_mask, reg_mask);
}
}
}
********************************************/
static void add_glsl_program_entry(struct shader_glsl_priv *priv, struct glsl_shader_prog_link *entry) {
- glsl_program_key_t *key;
+ glsl_program_key_t key;
- key = HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t));
- key->vshader = entry->vshader;
- key->pshader = entry->pshader;
+ key.vshader = entry->vshader;
+ key.pshader = entry->pshader;
+ key.vs_args = entry->vs_args;
+ key.ps_args = entry->ps_args;
- hash_table_put(priv->glsl_program_lookup, key, entry);
+ if (wine_rb_put(&priv->program_lookup, &key, &entry->program_lookup_entry) == -1)
+ {
+ ERR("Failed to insert program entry.\n");
+ }
}
static struct glsl_shader_prog_link *get_glsl_program_entry(struct shader_glsl_priv *priv,
- GLhandleARB vshader, GLhandleARB pshader) {
+ IWineD3DVertexShader *vshader, IWineD3DPixelShader *pshader, struct vs_compile_args *vs_args,
+ struct ps_compile_args *ps_args) {
+ struct wine_rb_entry *entry;
glsl_program_key_t key;
key.vshader = vshader;
key.pshader = pshader;
+ key.vs_args = *vs_args;
+ key.ps_args = *ps_args;
- return (struct glsl_shader_prog_link *)hash_table_get(priv->glsl_program_lookup, &key);
+ entry = wine_rb_get(&priv->program_lookup, &key);
+ return entry ? WINE_RB_ENTRY_VALUE(entry, struct glsl_shader_prog_link, program_lookup_entry) : NULL;
}
-static void delete_glsl_program_entry(struct shader_glsl_priv *priv, WineD3D_GL_Info *gl_info, struct glsl_shader_prog_link *entry) {
- glsl_program_key_t *key;
+/* GL locking is done by the caller */
+static void delete_glsl_program_entry(struct shader_glsl_priv *priv, const struct wined3d_gl_info *gl_info,
+ struct glsl_shader_prog_link *entry)
+{
+ glsl_program_key_t key;
- key = HeapAlloc(GetProcessHeap(), 0, sizeof(glsl_program_key_t));
- key->vshader = entry->vshader;
- key->pshader = entry->pshader;
- hash_table_remove(priv->glsl_program_lookup, key);
+ key.vshader = entry->vshader;
+ key.pshader = entry->pshader;
+ key.vs_args = entry->vs_args;
+ key.ps_args = entry->ps_args;
+ wine_rb_remove(&priv->program_lookup, &key);
GL_EXTCALL(glDeleteObjectARB(entry->programId));
if (entry->vshader) list_remove(&entry->vshader_entry);
HeapFree(GetProcessHeap(), 0, entry);
}
-static void handle_ps3_input(SHADER_BUFFER *buffer, semantic *semantics_in, semantic *semantics_out, WineD3D_GL_Info *gl_info, DWORD *map) {
+static void handle_ps3_input(struct wined3d_shader_buffer *buffer, const struct wined3d_gl_info *gl_info, const DWORD *map,
+ const struct wined3d_shader_signature_element *input_signature, const struct shader_reg_maps *reg_maps_in,
+ const struct wined3d_shader_signature_element *output_signature, const struct shader_reg_maps *reg_maps_out)
+{
unsigned int i, j;
- DWORD usage_token, usage_token_out;
- DWORD register_token, register_token_out;
- DWORD usage, usage_idx, usage_out, usage_idx_out;
+ const char *semantic_name_in, *semantic_name_out;
+ UINT semantic_idx_in, semantic_idx_out;
DWORD *set;
DWORD in_idx;
- DWORD in_count = GL_LIMITS(glsl_varyings) / 4;
+ unsigned int in_count = vec4_varyings(3, gl_info);
char reg_mask[6], reg_mask_out[6];
char destination[50];
+ WORD input_map, output_map;
set = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(*set) * (in_count + 2));
- if (!semantics_out) {
+ if (!output_signature)
+ {
/* Save gl_FrontColor & gl_FrontSecondaryColor before overwriting them. */
shader_addline(buffer, "vec4 front_color = gl_FrontColor;\n");
shader_addline(buffer, "vec4 front_secondary_color = gl_FrontSecondaryColor;\n");
}
- for(i = 0; i < MAX_REG_INPUT; i++) {
- usage_token = semantics_in[i].usage;
- if (!usage_token) continue;
+ input_map = reg_maps_in->input_registers;
+ for (i = 0; input_map; input_map >>= 1, ++i)
+ {
+ if (!(input_map & 1)) continue;
in_idx = map[i];
if (in_idx >= (in_count + 2)) {
FIXME("More input varyings declared than supported, expect issues\n");
continue;
- } else if(map[i] == -1) {
+ }
+ else if (map[i] == ~0U)
+ {
/* Declared, but not read register */
continue;
}
sprintf(destination, "IN[%u]", in_idx);
}
- register_token = semantics_in[i].reg;
-
- usage = (usage_token & WINED3DSP_DCL_USAGE_MASK) >> WINED3DSP_DCL_USAGE_SHIFT;
- usage_idx = (usage_token & WINED3DSP_DCL_USAGEINDEX_MASK) >> WINED3DSP_DCL_USAGEINDEX_SHIFT;
- set[map[i]] = shader_glsl_get_write_mask(register_token, reg_mask);
-
- if(!semantics_out) {
- switch(usage) {
- case WINED3DDECLUSAGE_COLOR:
- if (usage_idx == 0)
- shader_addline(buffer, "%s%s = front_color%s;\n",
- destination, reg_mask, reg_mask);
- else if (usage_idx == 1)
- shader_addline(buffer, "%s%s = front_secondary_color%s;\n",
- destination, reg_mask, reg_mask);
- else
- shader_addline(buffer, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
- destination, reg_mask, reg_mask);
- break;
-
- case WINED3DDECLUSAGE_TEXCOORD:
- if (usage_idx < 8) {
- shader_addline(buffer, "%s%s = gl_TexCoord[%u]%s;\n",
- destination, reg_mask, usage_idx, reg_mask);
- } else {
- shader_addline(buffer, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
- destination, reg_mask, reg_mask);
- }
- break;
-
- case WINED3DDECLUSAGE_FOG:
- shader_addline(buffer, "%s%s = vec4(gl_FogFragCoord, 0.0, 0.0, 0.0)%s;\n",
- destination, reg_mask, reg_mask);
- break;
-
- default:
+ semantic_name_in = input_signature[i].semantic_name;
+ semantic_idx_in = input_signature[i].semantic_idx;
+ set[map[i]] = input_signature[i].mask;
+ shader_glsl_write_mask_to_str(input_signature[i].mask, reg_mask);
+
+ if (!output_signature)
+ {
+ if (shader_match_semantic(semantic_name_in, WINED3DDECLUSAGE_COLOR))
+ {
+ if (semantic_idx_in == 0)
+ shader_addline(buffer, "%s%s = front_color%s;\n",
+ destination, reg_mask, reg_mask);
+ else if (semantic_idx_in == 1)
+ shader_addline(buffer, "%s%s = front_secondary_color%s;\n",
+ destination, reg_mask, reg_mask);
+ else
shader_addline(buffer, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
- destination, reg_mask, reg_mask);
+ destination, reg_mask, reg_mask);
+ }
+ else if (shader_match_semantic(semantic_name_in, WINED3DDECLUSAGE_TEXCOORD))
+ {
+ if (semantic_idx_in < 8)
+ {
+ shader_addline(buffer, "%s%s = gl_TexCoord[%u]%s;\n",
+ destination, reg_mask, semantic_idx_in, reg_mask);
+ }
+ else
+ {
+ shader_addline(buffer, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
+ destination, reg_mask, reg_mask);
+ }
+ }
+ else if (shader_match_semantic(semantic_name_in, WINED3DDECLUSAGE_FOG))
+ {
+ shader_addline(buffer, "%s%s = vec4(gl_FogFragCoord, 0.0, 0.0, 0.0)%s;\n",
+ destination, reg_mask, reg_mask);
+ }
+ else
+ {
+ shader_addline(buffer, "%s%s = vec4(0.0, 0.0, 0.0, 0.0)%s;\n",
+ destination, reg_mask, reg_mask);
}
} else {
BOOL found = FALSE;
- for(j = 0; j < MAX_REG_OUTPUT; j++) {
- usage_token_out = semantics_out[j].usage;
- if (!usage_token_out) continue;
- register_token_out = semantics_out[j].reg;
- usage_out = (usage_token_out & WINED3DSP_DCL_USAGE_MASK) >> WINED3DSP_DCL_USAGE_SHIFT;
- usage_idx_out = (usage_token_out & WINED3DSP_DCL_USAGEINDEX_MASK) >> WINED3DSP_DCL_USAGEINDEX_SHIFT;
- shader_glsl_get_write_mask(register_token_out, reg_mask_out);
+ output_map = reg_maps_out->output_registers;
+ for (j = 0; output_map; output_map >>= 1, ++j)
+ {
+ if (!(output_map & 1)) continue;
+
+ semantic_name_out = output_signature[j].semantic_name;
+ semantic_idx_out = output_signature[j].semantic_idx;
+ shader_glsl_write_mask_to_str(output_signature[j].mask, reg_mask_out);
- if(usage == usage_out &&
- usage_idx == usage_idx_out) {
+ if (semantic_idx_in == semantic_idx_out
+ && !strcmp(semantic_name_in, semantic_name_out))
+ {
shader_addline(buffer, "%s%s = OUT[%u]%s;\n",
- destination, reg_mask, j, reg_mask);
+ destination, reg_mask, j, reg_mask);
found = TRUE;
}
}
* input varyings are assigned above, if the optimizer works properly.
*/
for(i = 0; i < in_count + 2; i++) {
- if(set[i] != WINED3DSP_WRITEMASK_ALL) {
+ if (set[i] && set[i] != WINED3DSP_WRITEMASK_ALL)
+ {
unsigned int size = 0;
memset(reg_mask, 0, sizeof(reg_mask));
if(!(set[i] & WINED3DSP_WRITEMASK_0)) {
HeapFree(GetProcessHeap(), 0, set);
}
-static GLhandleARB generate_param_reorder_function(IWineD3DVertexShader *vertexshader,
- IWineD3DPixelShader *pixelshader,
- WineD3D_GL_Info *gl_info) {
+/* GL locking is done by the caller */
+static GLhandleARB generate_param_reorder_function(struct wined3d_shader_buffer *buffer,
+ IWineD3DVertexShader *vertexshader, IWineD3DPixelShader *pixelshader, const struct wined3d_gl_info *gl_info)
+{
GLhandleARB ret = 0;
IWineD3DVertexShaderImpl *vs = (IWineD3DVertexShaderImpl *) vertexshader;
IWineD3DPixelShaderImpl *ps = (IWineD3DPixelShaderImpl *) pixelshader;
IWineD3DDeviceImpl *device;
- DWORD vs_major = vs ? WINED3DSHADER_VERSION_MAJOR(vs->baseShader.hex_version) : 0;
- DWORD ps_major = ps ? WINED3DSHADER_VERSION_MAJOR(ps->baseShader.hex_version) : 0;
+ DWORD vs_major = vs->baseShader.reg_maps.shader_version.major;
+ DWORD ps_major = ps ? ps->baseShader.reg_maps.shader_version.major : 0;
unsigned int i;
- SHADER_BUFFER buffer;
- DWORD usage_token;
- DWORD register_token;
- DWORD usage, usage_idx, writemask;
+ const char *semantic_name;
+ UINT semantic_idx;
char reg_mask[6];
- semantic *semantics_out, *semantics_in;
+ const struct wined3d_shader_signature_element *output_signature;
- buffer.buffer = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, SHADER_PGMSIZE);
- buffer.bsize = 0;
- buffer.lineNo = 0;
- buffer.newline = TRUE;
+ shader_buffer_clear(buffer);
- shader_addline(&buffer, "#version 120\n");
+ shader_addline(buffer, "#version 120\n");
if(vs_major < 3 && ps_major < 3) {
/* That one is easy: The vertex shader writes to the builtin varyings, the pixel shader reads from them.
* Take care about the texcoord .w fixup though if we're using the fixed function fragment pipeline
*/
device = (IWineD3DDeviceImpl *) vs->baseShader.device;
- if((GLINFO_LOCATION).set_texcoord_w && ps_major == 0 && vs_major > 0 &&
- !device->frag_pipe->ffp_proj_control) {
- shader_addline(&buffer, "void order_ps_input() {\n");
+ if ((gl_info->quirks & WINED3D_QUIRK_SET_TEXCOORD_W)
+ && ps_major == 0 && vs_major > 0 && !device->frag_pipe->ffp_proj_control)
+ {
+ shader_addline(buffer, "void order_ps_input() {\n");
for(i = 0; i < min(8, MAX_REG_TEXCRD); i++) {
if(vs->baseShader.reg_maps.texcoord_mask[i] != 0 &&
vs->baseShader.reg_maps.texcoord_mask[i] != WINED3DSP_WRITEMASK_ALL) {
- shader_addline(&buffer, "gl_TexCoord[%u].w = 1.0;\n", i);
+ shader_addline(buffer, "gl_TexCoord[%u].w = 1.0;\n", i);
}
}
- shader_addline(&buffer, "}\n");
+ shader_addline(buffer, "}\n");
} else {
- shader_addline(&buffer, "void order_ps_input() { /* do nothing */ }\n");
+ shader_addline(buffer, "void order_ps_input() { /* do nothing */ }\n");
}
} else if(ps_major < 3 && vs_major >= 3) {
- /* The vertex shader writes to its own varyings, the pixel shader needs them in the builtin ones */
- semantics_out = vs->semantics_out;
-
- shader_addline(&buffer, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT);
- for(i = 0; i < MAX_REG_OUTPUT; i++) {
- usage_token = semantics_out[i].usage;
- if (!usage_token) continue;
- register_token = semantics_out[i].reg;
-
- usage = (usage_token & WINED3DSP_DCL_USAGE_MASK) >> WINED3DSP_DCL_USAGE_SHIFT;
- usage_idx = (usage_token & WINED3DSP_DCL_USAGEINDEX_MASK) >> WINED3DSP_DCL_USAGEINDEX_SHIFT;
- writemask = shader_glsl_get_write_mask(register_token, reg_mask);
-
- switch(usage) {
- case WINED3DDECLUSAGE_COLOR:
- if (usage_idx == 0)
- shader_addline(&buffer, "gl_FrontColor%s = OUT[%u]%s;\n", reg_mask, i, reg_mask);
- else if (usage_idx == 1)
- shader_addline(&buffer, "gl_FrontSecondaryColor%s = OUT[%u]%s;\n", reg_mask, i, reg_mask);
- break;
-
- case WINED3DDECLUSAGE_POSITION:
- shader_addline(&buffer, "gl_Position%s = OUT[%u]%s;\n", reg_mask, i, reg_mask);
- break;
-
- case WINED3DDECLUSAGE_TEXCOORD:
- if (usage_idx < 8) {
- if(!(GLINFO_LOCATION).set_texcoord_w || ps_major > 0) writemask |= WINED3DSP_WRITEMASK_3;
-
- shader_addline(&buffer, "gl_TexCoord[%u]%s = OUT[%u]%s;\n",
- usage_idx, reg_mask, i, reg_mask);
- if(!(writemask & WINED3DSP_WRITEMASK_3)) {
- shader_addline(&buffer, "gl_TexCoord[%u].w = 1.0;\n", usage_idx);
- }
- }
- break;
-
- case WINED3DDECLUSAGE_PSIZE:
- shader_addline(&buffer, "gl_PointSize = OUT[%u].x;\n", i);
- break;
+ WORD map = vs->baseShader.reg_maps.output_registers;
- case WINED3DDECLUSAGE_FOG:
- shader_addline(&buffer, "gl_FogFragCoord = OUT[%u].%c;\n", i, reg_mask[1]);
- break;
-
- default:
- break;
+ /* The vertex shader writes to its own varyings, the pixel shader needs them in the builtin ones */
+ output_signature = vs->baseShader.output_signature;
+
+ shader_addline(buffer, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT);
+ for (i = 0; map; map >>= 1, ++i)
+ {
+ DWORD write_mask;
+
+ if (!(map & 1)) continue;
+
+ semantic_name = output_signature[i].semantic_name;
+ semantic_idx = output_signature[i].semantic_idx;
+ write_mask = output_signature[i].mask;
+ shader_glsl_write_mask_to_str(write_mask, reg_mask);
+
+ if (shader_match_semantic(semantic_name, WINED3DDECLUSAGE_COLOR))
+ {
+ if (semantic_idx == 0)
+ shader_addline(buffer, "gl_FrontColor%s = OUT[%u]%s;\n", reg_mask, i, reg_mask);
+ else if (semantic_idx == 1)
+ shader_addline(buffer, "gl_FrontSecondaryColor%s = OUT[%u]%s;\n", reg_mask, i, reg_mask);
+ }
+ else if (shader_match_semantic(semantic_name, WINED3DDECLUSAGE_POSITION))
+ {
+ shader_addline(buffer, "gl_Position%s = OUT[%u]%s;\n", reg_mask, i, reg_mask);
+ }
+ else if (shader_match_semantic(semantic_name, WINED3DDECLUSAGE_TEXCOORD))
+ {
+ if (semantic_idx < 8)
+ {
+ if (!(gl_info->quirks & WINED3D_QUIRK_SET_TEXCOORD_W) || ps_major > 0)
+ write_mask |= WINED3DSP_WRITEMASK_3;
+
+ shader_addline(buffer, "gl_TexCoord[%u]%s = OUT[%u]%s;\n",
+ semantic_idx, reg_mask, i, reg_mask);
+ if (!(write_mask & WINED3DSP_WRITEMASK_3))
+ shader_addline(buffer, "gl_TexCoord[%u].w = 1.0;\n", semantic_idx);
+ }
+ }
+ else if (shader_match_semantic(semantic_name, WINED3DDECLUSAGE_PSIZE))
+ {
+ shader_addline(buffer, "gl_PointSize = OUT[%u].x;\n", i);
+ }
+ else if (shader_match_semantic(semantic_name, WINED3DDECLUSAGE_FOG))
+ {
+ shader_addline(buffer, "gl_FogFragCoord = OUT[%u].%c;\n", i, reg_mask[1]);
}
}
- shader_addline(&buffer, "}\n");
+ shader_addline(buffer, "}\n");
} else if(ps_major >= 3 && vs_major >= 3) {
- semantics_out = vs->semantics_out;
- semantics_in = ps->semantics_in;
+ WORD map = vs->baseShader.reg_maps.output_registers;
+
+ output_signature = vs->baseShader.output_signature;
/* This one is tricky: a 3.0 pixel shader reads from a 3.0 vertex shader */
- shader_addline(&buffer, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings) / 4);
- shader_addline(&buffer, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT);
+ shader_addline(buffer, "varying vec4 IN[%u];\n", vec4_varyings(3, gl_info));
+ shader_addline(buffer, "void order_ps_input(in vec4 OUT[%u]) {\n", MAX_REG_OUTPUT);
/* First, sort out position and point size. Those are not passed to the pixel shader */
- for(i = 0; i < MAX_REG_OUTPUT; i++) {
- usage_token = semantics_out[i].usage;
- if (!usage_token) continue;
- register_token = semantics_out[i].reg;
-
- usage = (usage_token & WINED3DSP_DCL_USAGE_MASK) >> WINED3DSP_DCL_USAGE_SHIFT;
- usage_idx = (usage_token & WINED3DSP_DCL_USAGEINDEX_MASK) >> WINED3DSP_DCL_USAGEINDEX_SHIFT;
- shader_glsl_get_write_mask(register_token, reg_mask);
-
- switch(usage) {
- case WINED3DDECLUSAGE_POSITION:
- shader_addline(&buffer, "gl_Position%s = OUT[%u]%s;\n", reg_mask, i, reg_mask);
- break;
+ for (i = 0; map; map >>= 1, ++i)
+ {
+ if (!(map & 1)) continue;
- case WINED3DDECLUSAGE_PSIZE:
- shader_addline(&buffer, "gl_PointSize = OUT[%u].x;\n", i);
- break;
+ semantic_name = output_signature[i].semantic_name;
+ shader_glsl_write_mask_to_str(output_signature[i].mask, reg_mask);
- default:
- break;
+ if (shader_match_semantic(semantic_name, WINED3DDECLUSAGE_POSITION))
+ {
+ shader_addline(buffer, "gl_Position%s = OUT[%u]%s;\n", reg_mask, i, reg_mask);
+ }
+ else if (shader_match_semantic(semantic_name, WINED3DDECLUSAGE_PSIZE))
+ {
+ shader_addline(buffer, "gl_PointSize = OUT[%u].x;\n", i);
}
}
/* Then, fix the pixel shader input */
- handle_ps3_input(&buffer, semantics_in, semantics_out, gl_info, ps->input_reg_map);
+ handle_ps3_input(buffer, gl_info, ps->input_reg_map, ps->baseShader.input_signature,
+ &ps->baseShader.reg_maps, output_signature, &vs->baseShader.reg_maps);
- shader_addline(&buffer, "}\n");
+ shader_addline(buffer, "}\n");
} else if(ps_major >= 3 && vs_major < 3) {
- semantics_in = ps->semantics_in;
-
- shader_addline(&buffer, "varying vec4 IN[%u];\n", GL_LIMITS(glsl_varyings) / 4);
- shader_addline(&buffer, "void order_ps_input() {\n");
+ shader_addline(buffer, "varying vec4 IN[%u];\n", vec4_varyings(3, gl_info));
+ shader_addline(buffer, "void order_ps_input() {\n");
/* The vertex shader wrote to the builtin varyings. There is no need to figure out position and
* point size, but we depend on the optimizers kindness to find out that the pixel shader doesn't
* read gl_TexCoord and gl_ColorX, otherwise we'll run out of varyings
*/
- handle_ps3_input(&buffer, semantics_in, NULL, gl_info, ps->input_reg_map);
- shader_addline(&buffer, "}\n");
+ handle_ps3_input(buffer, gl_info, ps->input_reg_map, ps->baseShader.input_signature,
+ &ps->baseShader.reg_maps, NULL, NULL);
+ shader_addline(buffer, "}\n");
} else {
ERR("Unexpected vertex and pixel shader version condition: vs: %d, ps: %d\n", vs_major, ps_major);
}
ret = GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB));
checkGLcall("glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB)");
- GL_EXTCALL(glShaderSourceARB(ret, 1, (const char**)&buffer.buffer, NULL));
- checkGLcall("glShaderSourceARB(ret, 1, (const char**)&buffer.buffer, NULL)");
+ GL_EXTCALL(glShaderSourceARB(ret, 1, (const char**)&buffer->buffer, NULL));
+ checkGLcall("glShaderSourceARB(ret, 1, &buffer->buffer, NULL)");
GL_EXTCALL(glCompileShaderARB(ret));
checkGLcall("glCompileShaderARB(ret)");
- HeapFree(GetProcessHeap(), 0, buffer.buffer);
return ret;
}
-static void hardcode_local_constants(IWineD3DBaseShaderImpl *shader, WineD3D_GL_Info *gl_info, GLhandleARB programId, char prefix) {
- local_constant* lconst;
- GLuint tmp_loc;
- float *value;
+/* GL locking is done by the caller */
+static void hardcode_local_constants(IWineD3DBaseShaderImpl *shader, const struct wined3d_gl_info *gl_info,
+ GLhandleARB programId, char prefix)
+{
+ const local_constant *lconst;
+ GLint tmp_loc;
+ const float *value;
char glsl_name[8];
LIST_FOR_EACH_ENTRY(lconst, &shader->baseShader.constantsF, local_constant, entry) {
- value = (float *) lconst->value;
+ value = (const float *)lconst->value;
snprintf(glsl_name, sizeof(glsl_name), "%cLC%u", prefix, lconst->idx);
tmp_loc = GL_EXTCALL(glGetUniformLocationARB(programId, glsl_name));
GL_EXTCALL(glUniform4fvARB(tmp_loc, 1, value));
}
- checkGLcall("Hardcoding local constants\n");
+ checkGLcall("Hardcoding local constants");
+}
+
+/* GL locking is done by the caller */
+static GLuint shader_glsl_generate_pshader(const struct wined3d_context *context,
+ struct wined3d_shader_buffer *buffer, IWineD3DPixelShaderImpl *This,
+ const struct ps_compile_args *args, struct ps_np2fixup_info *np2fixup_info)
+{
+ const struct shader_reg_maps *reg_maps = &This->baseShader.reg_maps;
+ const struct wined3d_gl_info *gl_info = context->gl_info;
+ CONST DWORD *function = This->baseShader.function;
+ struct shader_glsl_ctx_priv priv_ctx;
+
+ /* Create the hw GLSL shader object and assign it as the shader->prgId */
+ GLhandleARB shader_obj = GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB));
+
+ memset(&priv_ctx, 0, sizeof(priv_ctx));
+ priv_ctx.cur_ps_args = args;
+ priv_ctx.cur_np2fixup_info = np2fixup_info;
+
+ shader_addline(buffer, "#version 120\n");
+
+ if (gl_info->supported[ARB_SHADER_TEXTURE_LOD] && reg_maps->usestexldd)
+ {
+ shader_addline(buffer, "#extension GL_ARB_shader_texture_lod : enable\n");
+ }
+ if (gl_info->supported[ARB_TEXTURE_RECTANGLE])
+ {
+ /* The spec says that it doesn't have to be explicitly enabled, but the nvidia
+ * drivers write a warning if we don't do so
+ */
+ shader_addline(buffer, "#extension GL_ARB_texture_rectangle : enable\n");
+ }
+
+ /* Base Declarations */
+ shader_generate_glsl_declarations(context, buffer, (IWineD3DBaseShader *)This, reg_maps, &priv_ctx);
+
+ /* Pack 3.0 inputs */
+ if (reg_maps->shader_version.major >= 3 && args->vp_mode != vertexshader)
+ {
+ shader_glsl_input_pack((IWineD3DPixelShader *) This, buffer,
+ This->baseShader.input_signature, reg_maps, args->vp_mode);
+ }
+
+ /* Base Shader Body */
+ shader_generate_main((IWineD3DBaseShader *)This, buffer, reg_maps, function, &priv_ctx);
+
+ /* Pixel shaders < 2.0 place the resulting color in R0 implicitly */
+ if (reg_maps->shader_version.major < 2)
+ {
+ /* Some older cards like GeforceFX ones don't support multiple buffers, so also not gl_FragData */
+ shader_addline(buffer, "gl_FragData[0] = R0;\n");
+ }
+
+ if (args->srgb_correction)
+ {
+ shader_addline(buffer, "tmp0.xyz = pow(gl_FragData[0].xyz, vec3(srgb_const0.x));\n");
+ shader_addline(buffer, "tmp0.xyz = tmp0.xyz * vec3(srgb_const0.y) - vec3(srgb_const0.z);\n");
+ shader_addline(buffer, "tmp1.xyz = gl_FragData[0].xyz * vec3(srgb_const0.w);\n");
+ shader_addline(buffer, "bvec3 srgb_compare = lessThan(gl_FragData[0].xyz, vec3(srgb_const1.x));\n");
+ shader_addline(buffer, "gl_FragData[0].xyz = mix(tmp0.xyz, tmp1.xyz, vec3(srgb_compare));\n");
+ shader_addline(buffer, "gl_FragData[0] = clamp(gl_FragData[0], 0.0, 1.0);\n");
+ }
+ /* Pixel shader < 3.0 do not replace the fog stage.
+ * This implements linear fog computation and blending.
+ * TODO: non linear fog
+ * NOTE: gl_Fog.start and gl_Fog.end don't hold fog start s and end e but
+ * -1/(e-s) and e/(e-s) respectively.
+ */
+ if (reg_maps->shader_version.major < 3)
+ {
+ switch(args->fog) {
+ case FOG_OFF: break;
+ case FOG_LINEAR:
+ shader_addline(buffer, "float fogstart = -1.0 / (gl_Fog.end - gl_Fog.start);\n");
+ shader_addline(buffer, "float fogend = gl_Fog.end * -fogstart;\n");
+ shader_addline(buffer, "float Fog = clamp(gl_FogFragCoord * fogstart + fogend, 0.0, 1.0);\n");
+ shader_addline(buffer, "gl_FragData[0].xyz = mix(gl_Fog.color.xyz, gl_FragData[0].xyz, Fog);\n");
+ break;
+ case FOG_EXP:
+ /* Fog = e^(-gl_Fog.density * gl_FogFragCoord) */
+ shader_addline(buffer, "float Fog = exp(-gl_Fog.density * gl_FogFragCoord);\n");
+ shader_addline(buffer, "Fog = clamp(Fog, 0.0, 1.0);\n");
+ shader_addline(buffer, "gl_FragData[0].xyz = mix(gl_Fog.color.xyz, gl_FragData[0].xyz, Fog);\n");
+ break;
+ case FOG_EXP2:
+ /* Fog = e^(-(gl_Fog.density * gl_FogFragCoord)^2) */
+ shader_addline(buffer, "float Fog = exp(-gl_Fog.density * gl_Fog.density * gl_FogFragCoord * gl_FogFragCoord);\n");
+ shader_addline(buffer, "Fog = clamp(Fog, 0.0, 1.0);\n");
+ shader_addline(buffer, "gl_FragData[0].xyz = mix(gl_Fog.color.xyz, gl_FragData[0].xyz, Fog);\n");
+ break;
+ }
+ }
+
+ shader_addline(buffer, "}\n");
+
+ TRACE("Compiling shader object %u\n", shader_obj);
+ GL_EXTCALL(glShaderSourceARB(shader_obj, 1, (const char**)&buffer->buffer, NULL));
+ GL_EXTCALL(glCompileShaderARB(shader_obj));
+ print_glsl_info_log(gl_info, shader_obj);
+
+ /* Store the shader object */
+ return shader_obj;
+}
+
+/* GL locking is done by the caller */
+static GLuint shader_glsl_generate_vshader(const struct wined3d_context *context,
+ struct wined3d_shader_buffer *buffer, IWineD3DVertexShaderImpl *This,
+ const struct vs_compile_args *args)
+{
+ const struct shader_reg_maps *reg_maps = &This->baseShader.reg_maps;
+ const struct wined3d_gl_info *gl_info = context->gl_info;
+ CONST DWORD *function = This->baseShader.function;
+ struct shader_glsl_ctx_priv priv_ctx;
+
+ /* Create the hw GLSL shader program and assign it as the shader->prgId */
+ GLhandleARB shader_obj = GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB));
+
+ shader_addline(buffer, "#version 120\n");
+
+ memset(&priv_ctx, 0, sizeof(priv_ctx));
+ priv_ctx.cur_vs_args = args;
+
+ /* Base Declarations */
+ shader_generate_glsl_declarations(context, buffer, (IWineD3DBaseShader *)This, reg_maps, &priv_ctx);
+
+ /* Base Shader Body */
+ shader_generate_main((IWineD3DBaseShader*)This, buffer, reg_maps, function, &priv_ctx);
+
+ /* Unpack 3.0 outputs */
+ if (reg_maps->shader_version.major >= 3) shader_addline(buffer, "order_ps_input(OUT);\n");
+ else shader_addline(buffer, "order_ps_input();\n");
+
+ /* The D3DRS_FOGTABLEMODE render state defines if the shader-generated fog coord is used
+ * or if the fragment depth is used. If the fragment depth is used(FOGTABLEMODE != NONE),
+ * the fog frag coord is thrown away. If the fog frag coord is used, but not written by
+ * the shader, it is set to 0.0(fully fogged, since start = 1.0, end = 0.0)
+ */
+ if(args->fog_src == VS_FOG_Z) {
+ shader_addline(buffer, "gl_FogFragCoord = gl_Position.z;\n");
+ } else if (!reg_maps->fog) {
+ shader_addline(buffer, "gl_FogFragCoord = 0.0;\n");
+ }
+
+ /* Write the final position.
+ *
+ * OpenGL coordinates specify the center of the pixel while d3d coords specify
+ * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
+ * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
+ * contains 1.0 to allow a mad.
+ */
+ shader_addline(buffer, "gl_Position.y = gl_Position.y * posFixup.y;\n");
+ shader_addline(buffer, "gl_Position.xy += posFixup.zw * gl_Position.ww;\n");
+ if(args->clip_enabled) {
+ shader_addline(buffer, "gl_ClipVertex = gl_Position;\n");
+ }
+
+ /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
+ *
+ * Basically we want (in homogeneous coordinates) z = z * 2 - 1. However, shaders are run
+ * before the homogeneous divide, so we have to take the w into account: z = ((z / w) * 2 - 1) * w,
+ * which is the same as z = z * 2 - w.
+ */
+ shader_addline(buffer, "gl_Position.z = gl_Position.z * 2.0 - gl_Position.w;\n");
+
+ shader_addline(buffer, "}\n");
+
+ TRACE("Compiling shader object %u\n", shader_obj);
+ GL_EXTCALL(glShaderSourceARB(shader_obj, 1, (const char**)&buffer->buffer, NULL));
+ GL_EXTCALL(glCompileShaderARB(shader_obj));
+ print_glsl_info_log(gl_info, shader_obj);
+
+ return shader_obj;
+}
+
+static GLhandleARB find_glsl_pshader(const struct wined3d_context *context,
+ struct wined3d_shader_buffer *buffer, IWineD3DPixelShaderImpl *shader,
+ const struct ps_compile_args *args, const struct ps_np2fixup_info **np2fixup_info)
+{
+ UINT i;
+ DWORD new_size;
+ struct glsl_ps_compiled_shader *new_array;
+ struct glsl_pshader_private *shader_data;
+ struct ps_np2fixup_info *np2fixup = NULL;
+ GLhandleARB ret;
+
+ if (!shader->baseShader.backend_data)
+ {
+ shader->baseShader.backend_data = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(*shader_data));
+ if (!shader->baseShader.backend_data)
+ {
+ ERR("Failed to allocate backend data.\n");
+ return 0;
+ }
+ }
+ shader_data = shader->baseShader.backend_data;
+
+ /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
+ * so a linear search is more performant than a hashmap or a binary search
+ * (cache coherency etc)
+ */
+ for(i = 0; i < shader_data->num_gl_shaders; i++) {
+ if(memcmp(&shader_data->gl_shaders[i].args, args, sizeof(*args)) == 0) {
+ if(args->np2_fixup) *np2fixup_info = &shader_data->gl_shaders[i].np2fixup;
+ return shader_data->gl_shaders[i].prgId;
+ }
+ }
+
+ TRACE("No matching GL shader found for shader %p, compiling a new shader.\n", shader);
+ if(shader_data->shader_array_size == shader_data->num_gl_shaders) {
+ if (shader_data->num_gl_shaders)
+ {
+ new_size = shader_data->shader_array_size + max(1, shader_data->shader_array_size / 2);
+ new_array = HeapReAlloc(GetProcessHeap(), 0, shader_data->gl_shaders,
+ new_size * sizeof(*shader_data->gl_shaders));
+ } else {
+ new_array = HeapAlloc(GetProcessHeap(), 0, sizeof(*shader_data->gl_shaders));
+ new_size = 1;
+ }
+
+ if(!new_array) {
+ ERR("Out of memory\n");
+ return 0;
+ }
+ shader_data->gl_shaders = new_array;
+ shader_data->shader_array_size = new_size;
+ }
+
+ shader_data->gl_shaders[shader_data->num_gl_shaders].args = *args;
+
+ memset(&shader_data->gl_shaders[shader_data->num_gl_shaders].np2fixup, 0, sizeof(struct ps_np2fixup_info));
+ if (args->np2_fixup) np2fixup = &shader_data->gl_shaders[shader_data->num_gl_shaders].np2fixup;
+
+ pixelshader_update_samplers(&shader->baseShader.reg_maps,
+ ((IWineD3DDeviceImpl *)shader->baseShader.device)->stateBlock->textures);
+
+ shader_buffer_clear(buffer);
+ ret = shader_glsl_generate_pshader(context, buffer, shader, args, np2fixup);
+ shader_data->gl_shaders[shader_data->num_gl_shaders++].prgId = ret;
+ *np2fixup_info = np2fixup;
+
+ return ret;
+}
+
+static inline BOOL vs_args_equal(const struct vs_compile_args *stored, const struct vs_compile_args *new,
+ const DWORD use_map) {
+ if((stored->swizzle_map & use_map) != new->swizzle_map) return FALSE;
+ if((stored->clip_enabled) != new->clip_enabled) return FALSE;
+ return stored->fog_src == new->fog_src;
+}
+
+static GLhandleARB find_glsl_vshader(const struct wined3d_context *context,
+ struct wined3d_shader_buffer *buffer, IWineD3DVertexShaderImpl *shader,
+ const struct vs_compile_args *args)
+{
+ UINT i;
+ DWORD new_size;
+ struct glsl_vs_compiled_shader *new_array;
+ DWORD use_map = ((IWineD3DDeviceImpl *)shader->baseShader.device)->strided_streams.use_map;
+ struct glsl_vshader_private *shader_data;
+ GLhandleARB ret;
+
+ if (!shader->baseShader.backend_data)
+ {
+ shader->baseShader.backend_data = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(*shader_data));
+ if (!shader->baseShader.backend_data)
+ {
+ ERR("Failed to allocate backend data.\n");
+ return 0;
+ }
+ }
+ shader_data = shader->baseShader.backend_data;
+
+ /* Usually we have very few GL shaders for each d3d shader(just 1 or maybe 2),
+ * so a linear search is more performant than a hashmap or a binary search
+ * (cache coherency etc)
+ */
+ for(i = 0; i < shader_data->num_gl_shaders; i++) {
+ if(vs_args_equal(&shader_data->gl_shaders[i].args, args, use_map)) {
+ return shader_data->gl_shaders[i].prgId;
+ }
+ }
+
+ TRACE("No matching GL shader found for shader %p, compiling a new shader.\n", shader);
+
+ if(shader_data->shader_array_size == shader_data->num_gl_shaders) {
+ if (shader_data->num_gl_shaders)
+ {
+ new_size = shader_data->shader_array_size + max(1, shader_data->shader_array_size / 2);
+ new_array = HeapReAlloc(GetProcessHeap(), 0, shader_data->gl_shaders,
+ new_size * sizeof(*shader_data->gl_shaders));
+ } else {
+ new_array = HeapAlloc(GetProcessHeap(), 0, sizeof(*shader_data->gl_shaders));
+ new_size = 1;
+ }
+
+ if(!new_array) {
+ ERR("Out of memory\n");
+ return 0;
+ }
+ shader_data->gl_shaders = new_array;
+ shader_data->shader_array_size = new_size;
+ }
+
+ shader_data->gl_shaders[shader_data->num_gl_shaders].args = *args;
+
+ shader_buffer_clear(buffer);
+ ret = shader_glsl_generate_vshader(context, buffer, shader, args);
+ shader_data->gl_shaders[shader_data->num_gl_shaders++].prgId = ret;
+
+ return ret;
}
/** Sets the GLSL program ID for the given pixel and vertex shader combination.
* the program in the hash table. If it creates a program, it will link the
* given objects, too.
*/
-static void set_glsl_shader_program(IWineD3DDevice *iface, BOOL use_ps, BOOL use_vs) {
- IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
- struct shader_glsl_priv *priv = (struct shader_glsl_priv *)This->shader_priv;
- WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
- IWineD3DPixelShader *pshader = This->stateBlock->pixelShader;
- IWineD3DVertexShader *vshader = This->stateBlock->vertexShader;
+
+/* GL locking is done by the caller */
+static void set_glsl_shader_program(const struct wined3d_context *context,
+ IWineD3DDeviceImpl *device, BOOL use_ps, BOOL use_vs)
+{
+ IWineD3DVertexShader *vshader = use_vs ? device->stateBlock->vertexShader : NULL;
+ IWineD3DPixelShader *pshader = use_ps ? device->stateBlock->pixelShader : NULL;
+ const struct wined3d_gl_info *gl_info = context->gl_info;
+ struct shader_glsl_priv *priv = device->shader_priv;
struct glsl_shader_prog_link *entry = NULL;
GLhandleARB programId = 0;
GLhandleARB reorder_shader_id = 0;
- int i;
+ unsigned int i;
char glsl_name[8];
+ struct ps_compile_args ps_compile_args;
+ struct vs_compile_args vs_compile_args;
+
+ if (vshader) find_vs_compile_args((IWineD3DVertexShaderImpl *)vshader, device->stateBlock, &vs_compile_args);
+ if (pshader) find_ps_compile_args((IWineD3DPixelShaderImpl *)pshader, device->stateBlock, &ps_compile_args);
- GLhandleARB vshader_id = use_vs ? ((IWineD3DBaseShaderImpl*)vshader)->baseShader.prgId : 0;
- GLhandleARB pshader_id = use_ps ? ((IWineD3DBaseShaderImpl*)pshader)->baseShader.prgId : 0;
- entry = get_glsl_program_entry(priv, vshader_id, pshader_id);
+ entry = get_glsl_program_entry(priv, vshader, pshader, &vs_compile_args, &ps_compile_args);
if (entry) {
priv->glsl_program = entry;
return;
/* Create the entry */
entry = HeapAlloc(GetProcessHeap(), 0, sizeof(struct glsl_shader_prog_link));
entry->programId = programId;
- entry->vshader = vshader_id;
- entry->pshader = pshader_id;
+ entry->vshader = vshader;
+ entry->pshader = pshader;
+ entry->vs_args = vs_compile_args;
+ entry->ps_args = ps_compile_args;
+ entry->constant_version = 0;
+ entry->np2Fixup_info = NULL;
/* Add the hash table entry */
add_glsl_program_entry(priv, entry);
priv->glsl_program = entry;
/* Attach GLSL vshader */
- if (vshader_id) {
- int max_attribs = 16; /* TODO: Will this always be the case? It is at the moment... */
+ if (vshader)
+ {
+ GLhandleARB vshader_id = find_glsl_vshader(context, &priv->shader_buffer,
+ (IWineD3DVertexShaderImpl *)vshader, &vs_compile_args);
+ WORD map = ((IWineD3DBaseShaderImpl *)vshader)->baseShader.reg_maps.input_registers;
char tmp_name[10];
- reorder_shader_id = generate_param_reorder_function(vshader, pshader, gl_info);
+ reorder_shader_id = generate_param_reorder_function(&priv->shader_buffer, vshader, pshader, gl_info);
TRACE("Attaching GLSL shader object %u to program %u\n", reorder_shader_id, programId);
GL_EXTCALL(glAttachObjectARB(programId, reorder_shader_id));
checkGLcall("glAttachObjectARB");
* We have to do this here because we need to know the Program ID
* in order to make the bindings work, and it has to be done prior
* to linking the GLSL program. */
- for (i = 0; i < max_attribs; ++i) {
- if (((IWineD3DBaseShaderImpl*)vshader)->baseShader.reg_maps.attributes[i]) {
- snprintf(tmp_name, sizeof(tmp_name), "attrib%i", i);
- GL_EXTCALL(glBindAttribLocationARB(programId, i, tmp_name));
- }
+ for (i = 0; map; map >>= 1, ++i)
+ {
+ if (!(map & 1)) continue;
+
+ snprintf(tmp_name, sizeof(tmp_name), "attrib%u", i);
+ GL_EXTCALL(glBindAttribLocationARB(programId, i, tmp_name));
}
checkGLcall("glBindAttribLocationARB");
}
/* Attach GLSL pshader */
- if (pshader_id) {
+ if (pshader)
+ {
+ GLhandleARB pshader_id = find_glsl_pshader(context, &priv->shader_buffer,
+ (IWineD3DPixelShaderImpl *)pshader, &ps_compile_args, &entry->np2Fixup_info);
TRACE("Attaching GLSL shader object %u to program %u\n", pshader_id, programId);
GL_EXTCALL(glAttachObjectARB(programId, pshader_id));
checkGLcall("glAttachObjectARB");
/* Link the program */
TRACE("Linking GLSL shader program %u\n", programId);
GL_EXTCALL(glLinkProgramARB(programId));
- print_glsl_info_log(&GLINFO_LOCATION, programId);
+ print_glsl_info_log(gl_info, programId);
- entry->vuniformF_locations = HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB) * GL_LIMITS(vshader_constantsF));
- for (i = 0; i < GL_LIMITS(vshader_constantsF); ++i) {
+ entry->vuniformF_locations = HeapAlloc(GetProcessHeap(), 0,
+ sizeof(GLhandleARB) * gl_info->limits.glsl_vs_float_constants);
+ for (i = 0; i < gl_info->limits.glsl_vs_float_constants; ++i)
+ {
snprintf(glsl_name, sizeof(glsl_name), "VC[%i]", i);
entry->vuniformF_locations[i] = GL_EXTCALL(glGetUniformLocationARB(programId, glsl_name));
}
- for (i = 0; i < MAX_CONST_I; ++i) {
+ for (i = 0; i < MAX_CONST_I; ++i)
+ {
snprintf(glsl_name, sizeof(glsl_name), "VI[%i]", i);
entry->vuniformI_locations[i] = GL_EXTCALL(glGetUniformLocationARB(programId, glsl_name));
}
- entry->puniformF_locations = HeapAlloc(GetProcessHeap(), 0, sizeof(GLhandleARB) * GL_LIMITS(pshader_constantsF));
- for (i = 0; i < GL_LIMITS(pshader_constantsF); ++i) {
+ entry->puniformF_locations = HeapAlloc(GetProcessHeap(), 0,
+ sizeof(GLhandleARB) * gl_info->limits.glsl_ps_float_constants);
+ for (i = 0; i < gl_info->limits.glsl_ps_float_constants; ++i)
+ {
snprintf(glsl_name, sizeof(glsl_name), "PC[%i]", i);
entry->puniformF_locations[i] = GL_EXTCALL(glGetUniformLocationARB(programId, glsl_name));
}
- for (i = 0; i < MAX_CONST_I; ++i) {
+ for (i = 0; i < MAX_CONST_I; ++i)
+ {
snprintf(glsl_name, sizeof(glsl_name), "PI[%i]", i);
entry->puniformI_locations[i] = GL_EXTCALL(glGetUniformLocationARB(programId, glsl_name));
}
if(pshader) {
- for(i = 0; i < ((IWineD3DPixelShaderImpl*)pshader)->numbumpenvmatconsts; i++) {
- char name[32];
- sprintf(name, "bumpenvmat%d", ((IWineD3DPixelShaderImpl*)pshader)->bumpenvmatconst[i].texunit);
+ char name[32];
+
+ for(i = 0; i < MAX_TEXTURES; i++) {
+ sprintf(name, "bumpenvmat%u", i);
entry->bumpenvmat_location[i] = GL_EXTCALL(glGetUniformLocationARB(programId, name));
- sprintf(name, "luminancescale%d", ((IWineD3DPixelShaderImpl*)pshader)->luminanceconst[i].texunit);
+ sprintf(name, "luminancescale%u", i);
entry->luminancescale_location[i] = GL_EXTCALL(glGetUniformLocationARB(programId, name));
- sprintf(name, "luminanceoffset%d", ((IWineD3DPixelShaderImpl*)pshader)->luminanceconst[i].texunit);
+ sprintf(name, "luminanceoffset%u", i);
entry->luminanceoffset_location[i] = GL_EXTCALL(glGetUniformLocationARB(programId, name));
}
- }
+ if (ps_compile_args.np2_fixup) {
+ if (entry->np2Fixup_info) {
+ entry->np2Fixup_location = GL_EXTCALL(glGetUniformLocationARB(programId, "PsamplerNP2Fixup"));
+ } else {
+ FIXME("NP2 texcoord fixup needed for this pixelshader, but no fixup uniform found.\n");
+ }
+ }
+ }
entry->posFixup_location = GL_EXTCALL(glGetUniformLocationARB(programId, "posFixup"));
- entry->srgb_comparison_location = GL_EXTCALL(glGetUniformLocationARB(programId, "srgb_comparison"));
- entry->srgb_mul_low_location = GL_EXTCALL(glGetUniformLocationARB(programId, "srgb_mul_low"));
entry->ycorrection_location = GL_EXTCALL(glGetUniformLocationARB(programId, "ycorrection"));
checkGLcall("Find glsl program uniform locations");
- if (pshader && WINED3DSHADER_VERSION_MAJOR(((IWineD3DPixelShaderImpl *)pshader)->baseShader.hex_version) >= 3
- && ((IWineD3DPixelShaderImpl *)pshader)->declared_in_count > GL_LIMITS(glsl_varyings) / 4) {
+ if (pshader
+ && ((IWineD3DPixelShaderImpl *)pshader)->baseShader.reg_maps.shader_version.major >= 3
+ && ((IWineD3DPixelShaderImpl *)pshader)->declared_in_count > vec4_varyings(3, gl_info))
+ {
TRACE("Shader %d needs vertex color clamping disabled\n", programId);
entry->vertex_color_clamp = GL_FALSE;
} else {
* fixed function fragment processing setups. So once the program is linked these samplers
* won't change.
*/
- if(vshader_id) {
- /* Load vertex shader samplers */
- shader_glsl_load_vsamplers(gl_info, (IWineD3DStateBlock*)This->stateBlock, programId);
- }
- if(pshader_id) {
- /* Load pixel shader samplers */
- shader_glsl_load_psamplers(gl_info, (IWineD3DStateBlock*)This->stateBlock, programId);
- }
+ if (vshader) shader_glsl_load_vsamplers(gl_info, device->texUnitMap, programId);
+ if (pshader) shader_glsl_load_psamplers(gl_info, device->texUnitMap, programId);
/* If the local constants do not have to be loaded with the environment constants,
* load them now to have them hardcoded in the GLSL program. This saves some CPU cycles
}
}
-static GLhandleARB create_glsl_blt_shader(WineD3D_GL_Info *gl_info) {
+/* GL locking is done by the caller */
+static GLhandleARB create_glsl_blt_shader(const struct wined3d_gl_info *gl_info, enum tex_types tex_type)
+{
GLhandleARB program_id;
GLhandleARB vshader_id, pshader_id;
- const char *blt_vshader[] = {
+ static const char *blt_vshader[] =
+ {
"#version 120\n"
"void main(void)\n"
"{\n"
" gl_Position = gl_Vertex;\n"
" gl_FrontColor = vec4(1.0);\n"
- " gl_TexCoord[0].x = (gl_Vertex.x * 0.5) + 0.5;\n"
- " gl_TexCoord[0].y = (-gl_Vertex.y * 0.5) + 0.5;\n"
+ " gl_TexCoord[0] = gl_MultiTexCoord0;\n"
"}\n"
};
- const char *blt_pshader[] = {
+ static const char *blt_pshaders[tex_type_count] =
+ {
+ /* tex_1d */
+ NULL,
+ /* tex_2d */
"#version 120\n"
"uniform sampler2D sampler;\n"
"void main(void)\n"
"{\n"
" gl_FragDepth = texture2D(sampler, gl_TexCoord[0].xy).x;\n"
- "}\n"
+ "}\n",
+ /* tex_3d */
+ NULL,
+ /* tex_cube */
+ "#version 120\n"
+ "uniform samplerCube sampler;\n"
+ "void main(void)\n"
+ "{\n"
+ " gl_FragDepth = textureCube(sampler, gl_TexCoord[0].xyz).x;\n"
+ "}\n",
+ /* tex_rect */
+ "#version 120\n"
+ "#extension GL_ARB_texture_rectangle : enable\n"
+ "uniform sampler2DRect sampler;\n"
+ "void main(void)\n"
+ "{\n"
+ " gl_FragDepth = texture2DRect(sampler, gl_TexCoord[0].xy).x;\n"
+ "}\n",
};
+ if (!blt_pshaders[tex_type])
+ {
+ FIXME("tex_type %#x not supported\n", tex_type);
+ tex_type = tex_2d;
+ }
+
vshader_id = GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB));
GL_EXTCALL(glShaderSourceARB(vshader_id, 1, blt_vshader, NULL));
GL_EXTCALL(glCompileShaderARB(vshader_id));
pshader_id = GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB));
- GL_EXTCALL(glShaderSourceARB(pshader_id, 1, blt_pshader, NULL));
+ GL_EXTCALL(glShaderSourceARB(pshader_id, 1, &blt_pshaders[tex_type], NULL));
GL_EXTCALL(glCompileShaderARB(pshader_id));
program_id = GL_EXTCALL(glCreateProgramObjectARB());
GL_EXTCALL(glAttachObjectARB(program_id, pshader_id));
GL_EXTCALL(glLinkProgramARB(program_id));
- print_glsl_info_log(&GLINFO_LOCATION, program_id);
+ print_glsl_info_log(gl_info, program_id);
/* Once linked we can mark the shaders for deletion. They will be deleted once the program
* is destroyed
return program_id;
}
-static void shader_glsl_select(IWineD3DDevice *iface, BOOL usePS, BOOL useVS) {
- IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
- struct shader_glsl_priv *priv = (struct shader_glsl_priv *)This->shader_priv;
- WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
+/* GL locking is done by the caller */
+static void shader_glsl_select(const struct wined3d_context *context, BOOL usePS, BOOL useVS)
+{
+ IWineD3DDeviceImpl *device = ((IWineD3DSurfaceImpl *)context->surface)->resource.device;
+ const struct wined3d_gl_info *gl_info = context->gl_info;
+ struct shader_glsl_priv *priv = device->shader_priv;
GLhandleARB program_id = 0;
GLenum old_vertex_color_clamp, current_vertex_color_clamp;
old_vertex_color_clamp = priv->glsl_program ? priv->glsl_program->vertex_color_clamp : GL_FIXED_ONLY_ARB;
- if (useVS || usePS) set_glsl_shader_program(iface, usePS, useVS);
+ if (useVS || usePS) set_glsl_shader_program(context, device, usePS, useVS);
else priv->glsl_program = NULL;
current_vertex_color_clamp = priv->glsl_program ? priv->glsl_program->vertex_color_clamp : GL_FIXED_ONLY_ARB;
- if (old_vertex_color_clamp != current_vertex_color_clamp) {
- if (GL_SUPPORT(ARB_COLOR_BUFFER_FLOAT)) {
+ if (old_vertex_color_clamp != current_vertex_color_clamp)
+ {
+ if (gl_info->supported[ARB_COLOR_BUFFER_FLOAT])
+ {
GL_EXTCALL(glClampColorARB(GL_CLAMP_VERTEX_COLOR_ARB, current_vertex_color_clamp));
checkGLcall("glClampColorARB");
- } else {
+ }
+ else
+ {
FIXME("vertex color clamp needs to be changed, but extension not supported.\n");
}
}
if (program_id) TRACE("Using GLSL program %u\n", program_id);
GL_EXTCALL(glUseProgramObjectARB(program_id));
checkGLcall("glUseProgramObjectARB");
+
+ /* In case that NP2 texcoord fixup data is found for the selected program, trigger a reload of the
+ * constants. This has to be done because it can't be guaranteed that sampler() (from state.c) is
+ * called between selecting the shader and using it, which results in wrong fixup for some frames. */
+ if (priv->glsl_program && priv->glsl_program->np2Fixup_info)
+ {
+ shader_glsl_load_np2fixup_constants((IWineD3DDevice *)device, usePS, useVS);
+ }
}
-static void shader_glsl_select_depth_blt(IWineD3DDevice *iface) {
+/* GL locking is done by the caller */
+static void shader_glsl_select_depth_blt(IWineD3DDevice *iface, enum tex_types tex_type) {
IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
- WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
- struct shader_glsl_priv *priv = (struct shader_glsl_priv *) This->shader_priv;
- static GLhandleARB loc = -1;
-
- if (!priv->depth_blt_glsl_program_id) {
- priv->depth_blt_glsl_program_id = create_glsl_blt_shader(gl_info);
- loc = GL_EXTCALL(glGetUniformLocationARB(priv->depth_blt_glsl_program_id, "sampler"));
+ const struct wined3d_gl_info *gl_info = &This->adapter->gl_info;
+ struct shader_glsl_priv *priv = This->shader_priv;
+ GLhandleARB *blt_program = &priv->depth_blt_program[tex_type];
+
+ if (!*blt_program) {
+ GLint loc;
+ *blt_program = create_glsl_blt_shader(gl_info, tex_type);
+ loc = GL_EXTCALL(glGetUniformLocationARB(*blt_program, "sampler"));
+ GL_EXTCALL(glUseProgramObjectARB(*blt_program));
+ GL_EXTCALL(glUniform1iARB(loc, 0));
+ } else {
+ GL_EXTCALL(glUseProgramObjectARB(*blt_program));
}
-
- GL_EXTCALL(glUseProgramObjectARB(priv->depth_blt_glsl_program_id));
- GL_EXTCALL(glUniform1iARB(loc, 0));
}
+/* GL locking is done by the caller */
static void shader_glsl_deselect_depth_blt(IWineD3DDevice *iface) {
IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
- WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
- struct shader_glsl_priv *priv = (struct shader_glsl_priv *) This->shader_priv;
+ const struct wined3d_gl_info *gl_info = &This->adapter->gl_info;
+ struct shader_glsl_priv *priv = This->shader_priv;
GLhandleARB program_id;
program_id = priv->glsl_program ? priv->glsl_program->programId : 0;
checkGLcall("glUseProgramObjectARB");
}
-static void shader_glsl_cleanup(IWineD3DDevice *iface) {
- IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
- WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
- GL_EXTCALL(glUseProgramObjectARB(0));
-}
-
static void shader_glsl_destroy(IWineD3DBaseShader *iface) {
- struct list *linked_programs;
+ const struct list *linked_programs;
IWineD3DBaseShaderImpl *This = (IWineD3DBaseShaderImpl *) iface;
IWineD3DDeviceImpl *device = (IWineD3DDeviceImpl *)This->baseShader.device;
- struct shader_glsl_priv *priv = (struct shader_glsl_priv *)device->shader_priv;
- WineD3D_GL_Info *gl_info = &device->adapter->gl_info;
+ struct shader_glsl_priv *priv = device->shader_priv;
+ const struct wined3d_gl_info *gl_info;
+ IWineD3DPixelShaderImpl *ps = NULL;
+ IWineD3DVertexShaderImpl *vs = NULL;
+ struct wined3d_context *context;
/* Note: Do not use QueryInterface here to find out which shader type this is because this code
* can be called from IWineD3DBaseShader::Release
*/
- char pshader = shader_is_pshader_version(This->baseShader.hex_version);
+ char pshader = shader_is_pshader_version(This->baseShader.reg_maps.shader_version.type);
+
+ if(pshader) {
+ struct glsl_pshader_private *shader_data;
+ ps = (IWineD3DPixelShaderImpl *) This;
+ shader_data = ps->baseShader.backend_data;
+ if(!shader_data || shader_data->num_gl_shaders == 0)
+ {
+ HeapFree(GetProcessHeap(), 0, shader_data);
+ ps->baseShader.backend_data = NULL;
+ return;
+ }
+
+ context = context_acquire(device, NULL, CTXUSAGE_RESOURCELOAD);
+ gl_info = context->gl_info;
+
+ if (priv->glsl_program && (IWineD3DBaseShader *)priv->glsl_program->pshader == iface)
+ {
+ ENTER_GL();
+ shader_glsl_select(context, FALSE, FALSE);
+ LEAVE_GL();
+ }
+ } else {
+ struct glsl_vshader_private *shader_data;
+ vs = (IWineD3DVertexShaderImpl *) This;
+ shader_data = vs->baseShader.backend_data;
+ if(!shader_data || shader_data->num_gl_shaders == 0)
+ {
+ HeapFree(GetProcessHeap(), 0, shader_data);
+ vs->baseShader.backend_data = NULL;
+ return;
+ }
+
+ context = context_acquire(device, NULL, CTXUSAGE_RESOURCELOAD);
+ gl_info = context->gl_info;
+
+ if (priv->glsl_program && (IWineD3DBaseShader *)priv->glsl_program->vshader == iface)
+ {
+ ENTER_GL();
+ shader_glsl_select(context, FALSE, FALSE);
+ LEAVE_GL();
+ }
+ }
- if(This->baseShader.prgId == 0) return;
linked_programs = &This->baseShader.linked_programs;
TRACE("Deleting linked programs\n");
if (linked_programs->next) {
struct glsl_shader_prog_link *entry, *entry2;
+ ENTER_GL();
if(pshader) {
LIST_FOR_EACH_ENTRY_SAFE(entry, entry2, linked_programs, struct glsl_shader_prog_link, pshader_entry) {
delete_glsl_program_entry(priv, gl_info, entry);
delete_glsl_program_entry(priv, gl_info, entry);
}
}
+ LEAVE_GL();
}
- TRACE("Deleting shader object %u\n", This->baseShader.prgId);
- GL_EXTCALL(glDeleteObjectARB(This->baseShader.prgId));
- checkGLcall("glDeleteObjectARB");
- This->baseShader.prgId = 0;
- This->baseShader.is_compiled = FALSE;
-}
+ if(pshader) {
+ UINT i;
+ struct glsl_pshader_private *shader_data = ps->baseShader.backend_data;
+
+ ENTER_GL();
+ for(i = 0; i < shader_data->num_gl_shaders; i++) {
+ TRACE("deleting pshader %u\n", shader_data->gl_shaders[i].prgId);
+ GL_EXTCALL(glDeleteObjectARB(shader_data->gl_shaders[i].prgId));
+ checkGLcall("glDeleteObjectARB");
+ }
+ LEAVE_GL();
+ HeapFree(GetProcessHeap(), 0, shader_data->gl_shaders);
+ HeapFree(GetProcessHeap(), 0, shader_data);
+ ps->baseShader.backend_data = NULL;
+ } else {
+ UINT i;
+ struct glsl_vshader_private *shader_data = vs->baseShader.backend_data;
+
+ ENTER_GL();
+ for(i = 0; i < shader_data->num_gl_shaders; i++) {
+ TRACE("deleting vshader %u\n", shader_data->gl_shaders[i].prgId);
+ GL_EXTCALL(glDeleteObjectARB(shader_data->gl_shaders[i].prgId));
+ checkGLcall("glDeleteObjectARB");
+ }
+ LEAVE_GL();
+ HeapFree(GetProcessHeap(), 0, shader_data->gl_shaders);
+ HeapFree(GetProcessHeap(), 0, shader_data);
+ vs->baseShader.backend_data = NULL;
+ }
-static unsigned int glsl_program_key_hash(void *key) {
- glsl_program_key_t *k = (glsl_program_key_t *)key;
+ context_release(context);
+}
- unsigned int hash = k->vshader | k->pshader << 16;
- hash += ~(hash << 15);
- hash ^= (hash >> 10);
- hash += (hash << 3);
- hash ^= (hash >> 6);
- hash += ~(hash << 11);
- hash ^= (hash >> 16);
+static int glsl_program_key_compare(const void *key, const struct wine_rb_entry *entry)
+{
+ const glsl_program_key_t *k = key;
+ const struct glsl_shader_prog_link *prog = WINE_RB_ENTRY_VALUE(entry,
+ const struct glsl_shader_prog_link, program_lookup_entry);
+ int cmp;
- return hash;
-}
+ if (k->vshader > prog->vshader) return 1;
+ else if (k->vshader < prog->vshader) return -1;
-static BOOL glsl_program_key_compare(void *keya, void *keyb) {
- glsl_program_key_t *ka = (glsl_program_key_t *)keya;
- glsl_program_key_t *kb = (glsl_program_key_t *)keyb;
+ if (k->pshader > prog->pshader) return 1;
+ else if (k->pshader < prog->pshader) return -1;
- return ka->vshader == kb->vshader && ka->pshader == kb->pshader;
-}
+ if (k->vshader && (cmp = memcmp(&k->vs_args, &prog->vs_args, sizeof(prog->vs_args)))) return cmp;
+ if (k->pshader && (cmp = memcmp(&k->ps_args, &prog->ps_args, sizeof(prog->ps_args)))) return cmp;
-static HRESULT shader_glsl_alloc(IWineD3DDevice *iface) {
- IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
- struct shader_glsl_priv *priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct shader_glsl_priv));
- priv->glsl_program_lookup = hash_table_create(glsl_program_key_hash, glsl_program_key_compare);
- This->shader_priv = priv;
- return WINED3D_OK;
+ return 0;
}
-static void shader_glsl_free(IWineD3DDevice *iface) {
- IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
- WineD3D_GL_Info *gl_info = &This->adapter->gl_info;
- struct shader_glsl_priv *priv = (struct shader_glsl_priv *)This->shader_priv;
+static BOOL constant_heap_init(struct constant_heap *heap, unsigned int constant_count)
+{
+ SIZE_T size = (constant_count + 1) * sizeof(*heap->entries) + constant_count * sizeof(*heap->positions);
+ void *mem = HeapAlloc(GetProcessHeap(), 0, size);
- if(priv->depth_blt_glsl_program_id) {
- GL_EXTCALL(glDeleteObjectARB(priv->depth_blt_glsl_program_id));
+ if (!mem)
+ {
+ ERR("Failed to allocate memory\n");
+ return FALSE;
}
- hash_table_destroy(priv->glsl_program_lookup, NULL, NULL);
+ heap->entries = mem;
+ heap->entries[1].version = 0;
+ heap->positions = (unsigned int *)(heap->entries + constant_count + 1);
+ heap->size = 1;
- HeapFree(GetProcessHeap(), 0, This->shader_priv);
- This->shader_priv = NULL;
+ return TRUE;
}
-static BOOL shader_glsl_dirty_const(IWineD3DDevice *iface) {
- /* TODO: GL_EXT_bindable_uniform can be used to share constants across shaders */
- return FALSE;
+static void constant_heap_free(struct constant_heap *heap)
+{
+ HeapFree(GetProcessHeap(), 0, heap->entries);
}
-static void shader_glsl_generate_pshader(IWineD3DPixelShader *iface, SHADER_BUFFER *buffer) {
- IWineD3DPixelShaderImpl *This = (IWineD3DPixelShaderImpl *)iface;
- shader_reg_maps* reg_maps = &This->baseShader.reg_maps;
- CONST DWORD *function = This->baseShader.function;
- const char *fragcolor;
- WineD3D_GL_Info *gl_info = &((IWineD3DDeviceImpl *)This->baseShader.device)->adapter->gl_info;
-
- /* Create the hw GLSL shader object and assign it as the baseShader.prgId */
- GLhandleARB shader_obj = GL_EXTCALL(glCreateShaderObjectARB(GL_FRAGMENT_SHADER_ARB));
+static const struct wine_rb_functions wined3d_glsl_program_rb_functions =
+{
+ wined3d_rb_alloc,
+ wined3d_rb_realloc,
+ wined3d_rb_free,
+ glsl_program_key_compare,
+};
- shader_addline(buffer, "#version 120\n");
+static HRESULT shader_glsl_alloc(IWineD3DDevice *iface) {
+ IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
+ const struct wined3d_gl_info *gl_info = &This->adapter->gl_info;
+ struct shader_glsl_priv *priv = HeapAlloc(GetProcessHeap(), HEAP_ZERO_MEMORY, sizeof(struct shader_glsl_priv));
+ SIZE_T stack_size = wined3d_log2i(max(gl_info->limits.glsl_vs_float_constants,
+ gl_info->limits.glsl_ps_float_constants)) + 1;
- if (GL_SUPPORT(ARB_DRAW_BUFFERS)) {
- shader_addline(buffer, "#extension GL_ARB_draw_buffers : enable\n");
- }
- if (GL_SUPPORT(ARB_TEXTURE_RECTANGLE)) {
- /* The spec says that it doesn't have to be explicitly enabled, but the nvidia
- * drivers write a warning if we don't do so
- */
- shader_addline(buffer, "#extension GL_ARB_texture_rectangle : enable\n");
+ if (!shader_buffer_init(&priv->shader_buffer))
+ {
+ ERR("Failed to initialize shader buffer.\n");
+ goto fail;
}
- /* Base Declarations */
- shader_generate_glsl_declarations( (IWineD3DBaseShader*) This, reg_maps, buffer, &GLINFO_LOCATION);
-
- /* Pack 3.0 inputs */
- if (This->baseShader.hex_version >= WINED3DPS_VERSION(3,0)) {
-
- if(((IWineD3DDeviceImpl *) This->baseShader.device)->strided_streams.u.s.position_transformed) {
- This->vertexprocessing = pretransformed;
- pshader_glsl_input_pack(buffer, This->semantics_in, iface);
- } else if(!use_vs((IWineD3DDeviceImpl *) This->baseShader.device)) {
- This->vertexprocessing = fixedfunction;
- pshader_glsl_input_pack(buffer, This->semantics_in, iface);
- } else {
- This->vertexprocessing = vertexshader;
- }
+ priv->stack = HeapAlloc(GetProcessHeap(), 0, stack_size * sizeof(*priv->stack));
+ if (!priv->stack)
+ {
+ ERR("Failed to allocate memory.\n");
+ goto fail;
}
- /* Base Shader Body */
- shader_generate_main( (IWineD3DBaseShader*) This, buffer, reg_maps, function);
-
- /* Pixel shaders < 2.0 place the resulting color in R0 implicitly */
- if (This->baseShader.hex_version < WINED3DPS_VERSION(2,0)) {
- /* Some older cards like GeforceFX ones don't support multiple buffers, so also not gl_FragData */
- if(GL_SUPPORT(ARB_DRAW_BUFFERS))
- shader_addline(buffer, "gl_FragData[0] = R0;\n");
- else
- shader_addline(buffer, "gl_FragColor = R0;\n");
+ if (!constant_heap_init(&priv->vconst_heap, gl_info->limits.glsl_vs_float_constants))
+ {
+ ERR("Failed to initialize vertex shader constant heap\n");
+ goto fail;
}
- if(GL_SUPPORT(ARB_DRAW_BUFFERS)) {
- fragcolor = "gl_FragData[0]";
- } else {
- fragcolor = "gl_FragColor";
- }
- if(This->srgb_enabled) {
- shader_addline(buffer, "tmp0.xyz = pow(%s.xyz, vec3(%f, %f, %f)) * vec3(%f, %f, %f) - vec3(%f, %f, %f);\n",
- fragcolor, srgb_pow, srgb_pow, srgb_pow, srgb_mul_high, srgb_mul_high, srgb_mul_high,
- srgb_sub_high, srgb_sub_high, srgb_sub_high);
- shader_addline(buffer, "tmp1.xyz = %s.xyz * srgb_mul_low.xyz;\n", fragcolor);
- shader_addline(buffer, "%s.x = %s.x < srgb_comparison.x ? tmp1.x : tmp0.x;\n", fragcolor, fragcolor);
- shader_addline(buffer, "%s.y = %s.y < srgb_comparison.y ? tmp1.y : tmp0.y;\n", fragcolor, fragcolor);
- shader_addline(buffer, "%s.z = %s.z < srgb_comparison.z ? tmp1.z : tmp0.z;\n", fragcolor, fragcolor);
- shader_addline(buffer, "%s = clamp(%s, 0.0, 1.0);\n", fragcolor, fragcolor);
+ if (!constant_heap_init(&priv->pconst_heap, gl_info->limits.glsl_ps_float_constants))
+ {
+ ERR("Failed to initialize pixel shader constant heap\n");
+ goto fail;
}
- /* Pixel shader < 3.0 do not replace the fog stage.
- * This implements linear fog computation and blending.
- * TODO: non linear fog
- * NOTE: gl_Fog.start and gl_Fog.end don't hold fog start s and end e but
- * -1/(e-s) and e/(e-s) respectively.
- */
- if(This->baseShader.hex_version < WINED3DPS_VERSION(3,0)) {
- shader_addline(buffer, "float Fog = clamp(gl_FogFragCoord * gl_Fog.start + gl_Fog.end, 0.0, 1.0);\n");
- shader_addline(buffer, "%s.xyz = mix(gl_Fog.color.xyz, %s.xyz, Fog);\n", fragcolor, fragcolor);
+
+ if (wine_rb_init(&priv->program_lookup, &wined3d_glsl_program_rb_functions) == -1)
+ {
+ ERR("Failed to initialize rbtree.\n");
+ goto fail;
}
- shader_addline(buffer, "}\n");
+ priv->next_constant_version = 1;
- TRACE("Compiling shader object %u\n", shader_obj);
- GL_EXTCALL(glShaderSourceARB(shader_obj, 1, (const char**)&buffer->buffer, NULL));
- GL_EXTCALL(glCompileShaderARB(shader_obj));
- print_glsl_info_log(&GLINFO_LOCATION, shader_obj);
+ This->shader_priv = priv;
+ return WINED3D_OK;
- /* Store the shader object */
- This->baseShader.prgId = shader_obj;
+fail:
+ constant_heap_free(&priv->pconst_heap);
+ constant_heap_free(&priv->vconst_heap);
+ HeapFree(GetProcessHeap(), 0, priv->stack);
+ shader_buffer_free(&priv->shader_buffer);
+ HeapFree(GetProcessHeap(), 0, priv);
+ return E_OUTOFMEMORY;
}
-static void shader_glsl_generate_vshader(IWineD3DVertexShader *iface, SHADER_BUFFER *buffer) {
- IWineD3DVertexShaderImpl *This = (IWineD3DVertexShaderImpl *)iface;
- shader_reg_maps* reg_maps = &This->baseShader.reg_maps;
- CONST DWORD *function = This->baseShader.function;
- WineD3D_GL_Info *gl_info = &((IWineD3DDeviceImpl *)This->baseShader.device)->adapter->gl_info;
-
- /* Create the hw GLSL shader program and assign it as the baseShader.prgId */
- GLhandleARB shader_obj = GL_EXTCALL(glCreateShaderObjectARB(GL_VERTEX_SHADER_ARB));
-
- shader_addline(buffer, "#version 120\n");
-
- /* Base Declarations */
- shader_generate_glsl_declarations( (IWineD3DBaseShader*) This, reg_maps, buffer, &GLINFO_LOCATION);
-
- /* Base Shader Body */
- shader_generate_main( (IWineD3DBaseShader*) This, buffer, reg_maps, function);
+/* Context activation is done by the caller. */
+static void shader_glsl_free(IWineD3DDevice *iface) {
+ IWineD3DDeviceImpl *This = (IWineD3DDeviceImpl *)iface;
+ const struct wined3d_gl_info *gl_info = &This->adapter->gl_info;
+ struct shader_glsl_priv *priv = This->shader_priv;
+ int i;
- /* Unpack 3.0 outputs */
- if (This->baseShader.hex_version >= WINED3DVS_VERSION(3,0)) {
- shader_addline(buffer, "order_ps_input(OUT);\n");
- } else {
- shader_addline(buffer, "order_ps_input();\n");
+ ENTER_GL();
+ for (i = 0; i < tex_type_count; ++i)
+ {
+ if (priv->depth_blt_program[i])
+ {
+ GL_EXTCALL(glDeleteObjectARB(priv->depth_blt_program[i]));
+ }
}
+ LEAVE_GL();
- /* If this shader doesn't use fog copy the z coord to the fog coord so that we can use table fog */
- if (!reg_maps->fog)
- shader_addline(buffer, "gl_FogFragCoord = gl_Position.z;\n");
-
- /* Write the final position.
- *
- * OpenGL coordinates specify the center of the pixel while d3d coords specify
- * the corner. The offsets are stored in z and w in posFixup. posFixup.y contains
- * 1.0 or -1.0 to turn the rendering upside down for offscreen rendering. PosFixup.x
- * contains 1.0 to allow a mad.
- */
- shader_addline(buffer, "gl_Position.y = gl_Position.y * posFixup.y;\n");
- shader_addline(buffer, "gl_Position.xy += posFixup.zw * gl_Position.ww;\n");
-
- /* Z coord [0;1]->[-1;1] mapping, see comment in transform_projection in state.c
- *
- * Basically we want (in homogeneous coordinates) z = z * 2 - 1. However, shaders are run
- * before the homogeneous divide, so we have to take the w into account: z = ((z / w) * 2 - 1) * w,
- * which is the same as z = z * 2 - w.
- */
- shader_addline(buffer, "gl_Position.z = gl_Position.z * 2.0 - gl_Position.w;\n");
-
- shader_addline(buffer, "}\n");
+ wine_rb_destroy(&priv->program_lookup, NULL, NULL);
+ constant_heap_free(&priv->pconst_heap);
+ constant_heap_free(&priv->vconst_heap);
+ HeapFree(GetProcessHeap(), 0, priv->stack);
+ shader_buffer_free(&priv->shader_buffer);
- TRACE("Compiling shader object %u\n", shader_obj);
- GL_EXTCALL(glShaderSourceARB(shader_obj, 1, (const char**)&buffer->buffer, NULL));
- GL_EXTCALL(glCompileShaderARB(shader_obj));
- print_glsl_info_log(&GLINFO_LOCATION, shader_obj);
+ HeapFree(GetProcessHeap(), 0, This->shader_priv);
+ This->shader_priv = NULL;
+}
- /* Store the shader object */
- This->baseShader.prgId = shader_obj;
+static BOOL shader_glsl_dirty_const(IWineD3DDevice *iface) {
+ /* TODO: GL_EXT_bindable_uniform can be used to share constants across shaders */
+ return FALSE;
}
-static void shader_glsl_get_caps(WINED3DDEVTYPE devtype, WineD3D_GL_Info *gl_info, struct shader_caps *pCaps) {
+static void shader_glsl_get_caps(WINED3DDEVTYPE devtype,
+ const struct wined3d_gl_info *gl_info, struct shader_caps *pCaps)
+{
/* Nvidia Geforce6/7 or Ati R4xx/R5xx cards with GLSL support, support VS 3.0 but older Nvidia/Ati
- * models with GLSL support only support 2.0. In case of nvidia we can detect VS 2.0 support using
- * vs_nv_version which is based on NV_vertex_program.
+ * models with GLSL support only support 2.0. In case of nvidia we can detect VS 2.0 support based
+ * on the version of NV_vertex_program.
* For Ati cards there's no way using glsl (it abstracts the lowlevel info away) and also not
* using ARB_vertex_program. It is safe to assume that when a card supports pixel shader 2.0 it
* supports vertex shader 2.0 too and the way around. We can detect ps2.0 using the maximum number
* of native instructions, so use that here. For more info see the pixel shader versioning code below.
*/
- if((GLINFO_LOCATION.vs_nv_version == VS_VERSION_20) || (GLINFO_LOCATION.ps_arb_max_instructions <= 512))
+ if ((gl_info->supported[NV_VERTEX_PROGRAM2] && !gl_info->supported[NV_VERTEX_PROGRAM3])
+ || gl_info->limits.arb_ps_instructions <= 512)
pCaps->VertexShaderVersion = WINED3DVS_VERSION(2,0);
else
pCaps->VertexShaderVersion = WINED3DVS_VERSION(3,0);
TRACE_(d3d_caps)("Hardware vertex shader version %d.%d enabled (GLSL)\n", (pCaps->VertexShaderVersion >> 8) & 0xff, pCaps->VertexShaderVersion & 0xff);
- pCaps->MaxVertexShaderConst = GL_LIMITS(vshader_constantsF);
+ pCaps->MaxVertexShaderConst = gl_info->limits.glsl_vs_float_constants;
/* Older DX9-class videocards (GeforceFX / Radeon >9500/X*00) only support pixel shader 2.0/2.0a/2.0b.
* In OpenGL the extensions related to GLSL abstract lowlevel GL info away which is needed
* of instructions is 512 or less we have to do with ps2.0 hardware.
* NOTE: ps3.0 hardware requires 512 or more instructions but ati and nvidia offer 'enough' (1024 vs 4096) on their most basic ps3.0 hardware.
*/
- if((GLINFO_LOCATION.ps_nv_version == PS_VERSION_20) || (GLINFO_LOCATION.ps_arb_max_instructions <= 512))
+ if ((gl_info->supported[NV_FRAGMENT_PROGRAM] && !gl_info->supported[NV_FRAGMENT_PROGRAM2])
+ || gl_info->limits.arb_ps_instructions <= 512)
pCaps->PixelShaderVersion = WINED3DPS_VERSION(2,0);
else
pCaps->PixelShaderVersion = WINED3DPS_VERSION(3,0);
+ pCaps->MaxPixelShaderConst = gl_info->limits.glsl_ps_float_constants;
+
/* FIXME: The following line is card dependent. -8.0 to 8.0 is the
* Direct3D minimum requirement.
*
*/
pCaps->PixelShader1xMaxValue = 8.0;
TRACE_(d3d_caps)("Hardware pixel shader version %d.%d enabled (GLSL)\n", (pCaps->PixelShaderVersion >> 8) & 0xff, pCaps->PixelShaderVersion & 0xff);
+
+ pCaps->VSClipping = TRUE;
}
-static BOOL shader_glsl_conv_supported(WINED3DFORMAT fmt) {
- TRACE("Checking shader format support for format %s:", debug_d3dformat(fmt));
- switch(fmt) {
- case WINED3DFMT_V8U8:
- case WINED3DFMT_V16U16:
- case WINED3DFMT_X8L8V8U8:
- case WINED3DFMT_L6V5U5:
- case WINED3DFMT_Q8W8V8U8:
- case WINED3DFMT_ATI2N:
- TRACE("[OK]\n");
- return TRUE;
- default:
- TRACE("[FAILED\n");
- return FALSE;
+static BOOL shader_glsl_color_fixup_supported(struct color_fixup_desc fixup)
+{
+ if (TRACE_ON(d3d_shader) && TRACE_ON(d3d))
+ {
+ TRACE("Checking support for fixup:\n");
+ dump_color_fixup_desc(fixup);
}
+
+ /* We support everything except YUV conversions. */
+ if (!is_yuv_fixup(fixup))
+ {
+ TRACE("[OK]\n");
+ return TRUE;
+ }
+
+ TRACE("[FAILED]\n");
+ return FALSE;
+}
+
+static const SHADER_HANDLER shader_glsl_instruction_handler_table[WINED3DSIH_TABLE_SIZE] =
+{
+ /* WINED3DSIH_ABS */ shader_glsl_map2gl,
+ /* WINED3DSIH_ADD */ shader_glsl_arith,
+ /* WINED3DSIH_BEM */ shader_glsl_bem,
+ /* WINED3DSIH_BREAK */ shader_glsl_break,
+ /* WINED3DSIH_BREAKC */ shader_glsl_breakc,
+ /* WINED3DSIH_BREAKP */ NULL,
+ /* WINED3DSIH_CALL */ shader_glsl_call,
+ /* WINED3DSIH_CALLNZ */ shader_glsl_callnz,
+ /* WINED3DSIH_CMP */ shader_glsl_cmp,
+ /* WINED3DSIH_CND */ shader_glsl_cnd,
+ /* WINED3DSIH_CRS */ shader_glsl_cross,
+ /* WINED3DSIH_DCL */ NULL,
+ /* WINED3DSIH_DEF */ NULL,
+ /* WINED3DSIH_DEFB */ NULL,
+ /* WINED3DSIH_DEFI */ NULL,
+ /* WINED3DSIH_DP2ADD */ shader_glsl_dp2add,
+ /* WINED3DSIH_DP3 */ shader_glsl_dot,
+ /* WINED3DSIH_DP4 */ shader_glsl_dot,
+ /* WINED3DSIH_DST */ shader_glsl_dst,
+ /* WINED3DSIH_DSX */ shader_glsl_map2gl,
+ /* WINED3DSIH_DSY */ shader_glsl_map2gl,
+ /* WINED3DSIH_ELSE */ shader_glsl_else,
+ /* WINED3DSIH_ENDIF */ shader_glsl_end,
+ /* WINED3DSIH_ENDLOOP */ shader_glsl_end,
+ /* WINED3DSIH_ENDREP */ shader_glsl_end,
+ /* WINED3DSIH_EXP */ shader_glsl_map2gl,
+ /* WINED3DSIH_EXPP */ shader_glsl_expp,
+ /* WINED3DSIH_FRC */ shader_glsl_map2gl,
+ /* WINED3DSIH_IF */ shader_glsl_if,
+ /* WINED3DSIH_IFC */ shader_glsl_ifc,
+ /* WINED3DSIH_LABEL */ shader_glsl_label,
+ /* WINED3DSIH_LIT */ shader_glsl_lit,
+ /* WINED3DSIH_LOG */ shader_glsl_log,
+ /* WINED3DSIH_LOGP */ shader_glsl_log,
+ /* WINED3DSIH_LOOP */ shader_glsl_loop,
+ /* WINED3DSIH_LRP */ shader_glsl_lrp,
+ /* WINED3DSIH_M3x2 */ shader_glsl_mnxn,
+ /* WINED3DSIH_M3x3 */ shader_glsl_mnxn,
+ /* WINED3DSIH_M3x4 */ shader_glsl_mnxn,
+ /* WINED3DSIH_M4x3 */ shader_glsl_mnxn,
+ /* WINED3DSIH_M4x4 */ shader_glsl_mnxn,
+ /* WINED3DSIH_MAD */ shader_glsl_mad,
+ /* WINED3DSIH_MAX */ shader_glsl_map2gl,
+ /* WINED3DSIH_MIN */ shader_glsl_map2gl,
+ /* WINED3DSIH_MOV */ shader_glsl_mov,
+ /* WINED3DSIH_MOVA */ shader_glsl_mov,
+ /* WINED3DSIH_MUL */ shader_glsl_arith,
+ /* WINED3DSIH_NOP */ NULL,
+ /* WINED3DSIH_NRM */ shader_glsl_map2gl,
+ /* WINED3DSIH_PHASE */ NULL,
+ /* WINED3DSIH_POW */ shader_glsl_pow,
+ /* WINED3DSIH_RCP */ shader_glsl_rcp,
+ /* WINED3DSIH_REP */ shader_glsl_rep,
+ /* WINED3DSIH_RET */ shader_glsl_ret,
+ /* WINED3DSIH_RSQ */ shader_glsl_rsq,
+ /* WINED3DSIH_SETP */ NULL,
+ /* WINED3DSIH_SGE */ shader_glsl_compare,
+ /* WINED3DSIH_SGN */ shader_glsl_sgn,
+ /* WINED3DSIH_SINCOS */ shader_glsl_sincos,
+ /* WINED3DSIH_SLT */ shader_glsl_compare,
+ /* WINED3DSIH_SUB */ shader_glsl_arith,
+ /* WINED3DSIH_TEX */ shader_glsl_tex,
+ /* WINED3DSIH_TEXBEM */ shader_glsl_texbem,
+ /* WINED3DSIH_TEXBEML */ shader_glsl_texbem,
+ /* WINED3DSIH_TEXCOORD */ shader_glsl_texcoord,
+ /* WINED3DSIH_TEXDEPTH */ shader_glsl_texdepth,
+ /* WINED3DSIH_TEXDP3 */ shader_glsl_texdp3,
+ /* WINED3DSIH_TEXDP3TEX */ shader_glsl_texdp3tex,
+ /* WINED3DSIH_TEXKILL */ shader_glsl_texkill,
+ /* WINED3DSIH_TEXLDD */ shader_glsl_texldd,
+ /* WINED3DSIH_TEXLDL */ shader_glsl_texldl,
+ /* WINED3DSIH_TEXM3x2DEPTH */ shader_glsl_texm3x2depth,
+ /* WINED3DSIH_TEXM3x2PAD */ shader_glsl_texm3x2pad,
+ /* WINED3DSIH_TEXM3x2TEX */ shader_glsl_texm3x2tex,
+ /* WINED3DSIH_TEXM3x3 */ shader_glsl_texm3x3,
+ /* WINED3DSIH_TEXM3x3DIFF */ NULL,
+ /* WINED3DSIH_TEXM3x3PAD */ shader_glsl_texm3x3pad,
+ /* WINED3DSIH_TEXM3x3SPEC */ shader_glsl_texm3x3spec,
+ /* WINED3DSIH_TEXM3x3TEX */ shader_glsl_texm3x3tex,
+ /* WINED3DSIH_TEXM3x3VSPEC */ shader_glsl_texm3x3vspec,
+ /* WINED3DSIH_TEXREG2AR */ shader_glsl_texreg2ar,
+ /* WINED3DSIH_TEXREG2GB */ shader_glsl_texreg2gb,
+ /* WINED3DSIH_TEXREG2RGB */ shader_glsl_texreg2rgb,
+};
+
+static void shader_glsl_handle_instruction(const struct wined3d_shader_instruction *ins) {
+ SHADER_HANDLER hw_fct;
+
+ /* Select handler */
+ hw_fct = shader_glsl_instruction_handler_table[ins->handler_idx];
+
+ /* Unhandled opcode */
+ if (!hw_fct)
+ {
+ FIXME("Backend can't handle opcode %#x\n", ins->handler_idx);
+ return;
+ }
+ hw_fct(ins);
+
+ shader_glsl_add_instruction_modifiers(ins);
}
const shader_backend_t glsl_shader_backend = {
+ shader_glsl_handle_instruction,
shader_glsl_select,
shader_glsl_select_depth_blt,
shader_glsl_deselect_depth_blt,
+ shader_glsl_update_float_vertex_constants,
+ shader_glsl_update_float_pixel_constants,
shader_glsl_load_constants,
- shader_glsl_cleanup,
- shader_glsl_color_correction,
+ shader_glsl_load_np2fixup_constants,
shader_glsl_destroy,
shader_glsl_alloc,
shader_glsl_free,
shader_glsl_dirty_const,
- shader_glsl_generate_pshader,
- shader_glsl_generate_vshader,
shader_glsl_get_caps,
- shader_glsl_conv_supported,
+ shader_glsl_color_fixup_supported,
};
projects / reactos.git / blobdiff