gl_common: move things used by vo_opengl_old.c only to vo_opengl_old.c

Having this in gl_common is confusing.

1 files changed, 1241 insertions, 0 deletions

diff --git a/video/out/vo_opengl_old.c b/video/out/vo_opengl_old.c
index d010daacb9..8c32676e1d 100644
--- a/video/out/vo_opengl_old.c
+++ b/video/out/vo_opengl_old.c
@@ -42,6 +42,7 @@
 #include "gl_osd.h"
 #include "aspect.h"
 #include "video/memcpy_pic.h"
+#include "pnm_loader.h"
 
 //for gl_priv.use_yuv
 #define MASK_ALL_YUV (~(1 << YUV_CONVERSION_NONE))
@@ -110,6 +111,1246 @@ struct gl_priv {
     unsigned int slice_height;
 };
 
+static int glFindFormat(uint32_t format, int have_texture_rg, int *bpp,
+                        GLint *gl_texfmt, GLenum *gl_format, GLenum *gl_type);
+static void glCreateClearTex(GL *gl, GLenum target, GLenum fmt, GLenum format,
+                             GLenum type, GLint filter, int w, int h,
+                             unsigned char val);
+static int glCreatePPMTex(GL *gl, GLenum target, GLenum fmt, GLint filter,
+                          FILE *f, int *width, int *height, int *maxval);
+static void glDrawTex(GL *gl, GLfloat x, GLfloat y, GLfloat w, GLfloat h,
+                      GLfloat tx, GLfloat ty, GLfloat tw, GLfloat th,
+                      int sx, int sy, int rect_tex, int is_yv12, int flip);
+static int loadGPUProgram(GL *gl, GLenum target, char *prog);
+//! do not use YUV conversion, this should always stay 0
+#define YUV_CONVERSION_NONE 0
+//! use nVidia specific register combiners for YUV conversion
+//! implementation has been removed
+#define YUV_CONVERSION_COMBINERS 1
+//! use a fragment program for YUV conversion
+#define YUV_CONVERSION_FRAGMENT 2
+//! use a fragment program for YUV conversion with gamma using POW
+#define YUV_CONVERSION_FRAGMENT_POW 3
+//! use a fragment program with additional table lookup for YUV conversion
+#define YUV_CONVERSION_FRAGMENT_LOOKUP 4
+//! use ATI specific register combiners ("fragment program")
+#define YUV_CONVERSION_COMBINERS_ATI 5
+//! use a fragment program with 3D table lookup for YUV conversion
+#define YUV_CONVERSION_FRAGMENT_LOOKUP3D 6
+//! use ATI specific "text" register combiners ("fragment program")
+#define YUV_CONVERSION_TEXT_FRAGMENT 7
+//! use normal bilinear scaling for textures
+#define YUV_SCALER_BILIN 0
+//! use higher quality bicubic scaling for textures
+#define YUV_SCALER_BICUB 1
+//! use cubic scaling in X and normal linear scaling in Y direction
+#define YUV_SCALER_BICUB_X 2
+//! use cubic scaling without additional lookup texture
+#define YUV_SCALER_BICUB_NOTEX 3
+#define YUV_SCALER_UNSHARP 4
+#define YUV_SCALER_UNSHARP2 5
+//! mask for conversion type
+#define YUV_CONVERSION_MASK 0xF
+//! mask for scaler type
+#define YUV_SCALER_MASK 0xF
+//! shift value for luminance scaler type
+#define YUV_LUM_SCALER_SHIFT 8
+//! shift value for chrominance scaler type
+#define YUV_CHROM_SCALER_SHIFT 12
+//! extract conversion out of type
+#define YUV_CONVERSION(t) ((t) & YUV_CONVERSION_MASK)
+//! extract luminance scaler out of type
+#define YUV_LUM_SCALER(t) (((t) >> YUV_LUM_SCALER_SHIFT) & YUV_SCALER_MASK)
+//! extract chrominance scaler out of type
+#define YUV_CHROM_SCALER(t) (((t) >> YUV_CHROM_SCALER_SHIFT) & YUV_SCALER_MASK)
+#define SET_YUV_CONVERSION(c)   ((c) & YUV_CONVERSION_MASK)
+#define SET_YUV_LUM_SCALER(s)   (((s) & YUV_SCALER_MASK) << YUV_LUM_SCALER_SHIFT)
+#define SET_YUV_CHROM_SCALER(s) (((s) & YUV_SCALER_MASK) << YUV_CHROM_SCALER_SHIFT)
+//! returns whether the yuv conversion supports large brightness range etc.
+static inline int glYUVLargeRange(int conv)
+{
+    switch (conv) {
+    case YUV_CONVERSION_NONE:
+    case YUV_CONVERSION_COMBINERS_ATI:
+    case YUV_CONVERSION_FRAGMENT_LOOKUP3D:
+    case YUV_CONVERSION_TEXT_FRAGMENT:
+        return 0;
+    }
+    return 1;
+}
+typedef struct {
+    GLenum target;
+    int type;
+    struct mp_csp_params csp_params;
+    int texw;
+    int texh;
+    int chrom_texw;
+    int chrom_texh;
+    float filter_strength;
+    float noise_strength;
+} gl_conversion_params_t;
+
+static int glAutodetectYUVConversion(GL *gl);
+static void glSetupYUVConversion(GL *gl, gl_conversion_params_t *params);
+static void glEnableYUVConversion(GL *gl, GLenum target, int type);
+static void glDisableYUVConversion(GL *gl, GLenum target, int type);
+
+//! always return this format as internal texture format in glFindFormat
+#define TEXTUREFORMAT_ALWAYS GL_RGB8
+#undef TEXTUREFORMAT_ALWAYS
+
+/**
+ * \brief find the OpenGL settings coresponding to format.
+ *
+ * All parameters may be NULL.
+ * \param fmt MPlayer format to analyze.
+ * \param dummy reserved
+ * \param gl_texfmt [OUT] internal texture format that fits the
+ * image format, not necessarily the best for performance.
+ * \param gl_format [OUT] OpenGL format for this image format.
+ * \param gl_type [OUT] OpenGL type for this image format.
+ * \return 1 if format is supported by OpenGL, 0 if not.
+ * \ingroup gltexture
+ */
+static int glFindFormat(uint32_t fmt, int have_texture_rg, int *dummy,
+                        GLint *gl_texfmt, GLenum *gl_format, GLenum *gl_type)
+{
+    int supported = 1;
+    GLenum dummy2;
+    GLint dummy3;
+    if (!gl_texfmt)
+        gl_texfmt = &dummy3;
+    if (!gl_format)
+        gl_format = &dummy2;
+    if (!gl_type)
+        gl_type = &dummy2;
+
+    struct mp_imgfmt_desc desc = mp_imgfmt_get_desc(fmt);
+    if (desc.flags & MP_IMGFLAG_YUV_P) {
+        // reduce the possible cases a bit
+        if (desc.plane_bits > 8)
+            fmt = IMGFMT_420P16;
+        else
+            fmt = IMGFMT_420P;
+    }
+
+    *gl_texfmt = 3;
+    switch (fmt) {
+    case IMGFMT_RGB48:
+        *gl_format = GL_RGB;
+        *gl_type = GL_UNSIGNED_SHORT;
+        break;
+    case IMGFMT_RGB24:
+        *gl_format = GL_RGB;
+        *gl_type = GL_UNSIGNED_BYTE;
+        break;
+    case IMGFMT_RGBA:
+        *gl_texfmt = 4;
+        *gl_format = GL_RGBA;
+        *gl_type = GL_UNSIGNED_BYTE;
+        break;
+    case IMGFMT_420P16:
+        supported = 0; // no native YUV support
+        *gl_texfmt = have_texture_rg ? GL_R16 : GL_LUMINANCE16;
+        *gl_format = have_texture_rg ? GL_RED : GL_LUMINANCE;
+        *gl_type = GL_UNSIGNED_SHORT;
+        break;
+    case IMGFMT_420P:
+        supported = 0; // no native YV12 support
+    case IMGFMT_Y8:
+        *gl_texfmt = 1;
+        *gl_format = GL_LUMINANCE;
+        *gl_type = GL_UNSIGNED_BYTE;
+        break;
+    case IMGFMT_UYVY:
+        *gl_texfmt = GL_YCBCR_MESA;
+        *gl_format = GL_YCBCR_MESA;
+        *gl_type = fmt == IMGFMT_UYVY ? GL_UNSIGNED_SHORT_8_8 : GL_UNSIGNED_SHORT_8_8_REV;
+        break;
+#if 0
+    // we do not support palettized formats, although the format the
+    // swscale produces works
+    case IMGFMT_RGB8:
+        *gl_format = GL_RGB;
+        *gl_type = GL_UNSIGNED_BYTE_2_3_3_REV;
+        break;
+#endif
+    case IMGFMT_RGB15:
+        *gl_format = GL_RGBA;
+        *gl_type = GL_UNSIGNED_SHORT_1_5_5_5_REV;
+        break;
+    case IMGFMT_RGB16:
+        *gl_format = GL_RGB;
+        *gl_type = GL_UNSIGNED_SHORT_5_6_5_REV;
+        break;
+#if 0
+    case IMGFMT_BGR8:
+        // special case as red and blue have a different number of bits.
+        // GL_BGR and GL_UNSIGNED_BYTE_3_3_2 isn't supported at least
+        // by nVidia drivers, and in addition would give more bits to
+        // blue than to red, which isn't wanted
+        *gl_format = GL_RGB;
+        *gl_type = GL_UNSIGNED_BYTE_3_3_2;
+        break;
+#endif
+    case IMGFMT_BGR15:
+        *gl_format = GL_BGRA;
+        *gl_type = GL_UNSIGNED_SHORT_1_5_5_5_REV;
+        break;
+    case IMGFMT_BGR16:
+        *gl_format = GL_RGB;
+        *gl_type = GL_UNSIGNED_SHORT_5_6_5;
+        break;
+    case IMGFMT_BGR24:
+        *gl_format = GL_BGR;
+        *gl_type = GL_UNSIGNED_BYTE;
+        break;
+    case IMGFMT_BGRA:
+        *gl_texfmt = 4;
+        *gl_format = GL_BGRA;
+        *gl_type = GL_UNSIGNED_BYTE;
+        break;
+    default:
+        *gl_texfmt = 4;
+        *gl_format = GL_RGBA;
+        *gl_type = GL_UNSIGNED_BYTE;
+        supported = 0;
+    }
+#ifdef TEXTUREFORMAT_ALWAYS
+    *gl_texfmt = TEXTUREFORMAT_ALWAYS;
+#endif
+    return supported;
+}
+
+/**
+ * \brief create a texture and set some defaults
+ * \param target texture taget, usually GL_TEXTURE_2D
+ * \param fmt internal texture format
+ * \param format texture host data format
+ * \param type texture host data type
+ * \param filter filter used for scaling, e.g. GL_LINEAR
+ * \param w texture width
+ * \param h texture height
+ * \param val luminance value to fill texture with
+ * \ingroup gltexture
+ */
+static void glCreateClearTex(GL *gl, GLenum target, GLenum fmt, GLenum format,
+                             GLenum type, GLint filter, int w, int h,
+                             unsigned char val)
+{
+    GLfloat fval = (GLfloat)val / 255.0;
+    GLfloat border[4] = {
+        fval, fval, fval, fval
+    };
+    int stride;
+    char *init;
+    if (w == 0)
+        w = 1;
+    if (h == 0)
+        h = 1;
+    stride = w * glFmt2bpp(format, type);
+    if (!stride)
+        return;
+    init = malloc(stride * h);
+    memset(init, val, stride * h);
+    glAdjustAlignment(gl, stride);
+    gl->PixelStorei(GL_UNPACK_ROW_LENGTH, w);
+    gl->TexImage2D(target, 0, fmt, w, h, 0, format, type, init);
+    gl->TexParameterf(target, GL_TEXTURE_PRIORITY, 1.0);
+    gl->TexParameteri(target, GL_TEXTURE_MIN_FILTER, filter);
+    gl->TexParameteri(target, GL_TEXTURE_MAG_FILTER, filter);
+    gl->TexParameteri(target, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
+    gl->TexParameteri(target, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
+    // Border texels should not be used with CLAMP_TO_EDGE
+    // We set a sane default anyway.
+    gl->TexParameterfv(target, GL_TEXTURE_BORDER_COLOR, border);
+    free(init);
+}
+
+static GLint detect_hqtexfmt(GL *gl)
+{
+    const char *extensions = (const char *)gl->GetString(GL_EXTENSIONS);
+    if (strstr(extensions, "_texture_float"))
+        return GL_RGB32F;
+    else if (strstr(extensions, "NV_float_buffer"))
+        return GL_FLOAT_RGB32_NV;
+    return GL_RGB16;
+}
+
+/**
+ * \brief creates a texture from a PPM file
+ * \param target texture taget, usually GL_TEXTURE_2D
+ * \param fmt internal texture format, 0 for default
+ * \param filter filter used for scaling, e.g. GL_LINEAR
+ * \param f file to read PPM from
+ * \param width [out] width of texture
+ * \param height [out] height of texture
+ * \param maxval [out] maxval value from PPM file
+ * \return 0 on error, 1 otherwise
+ * \ingroup gltexture
+ */
+static int glCreatePPMTex(GL *gl, GLenum target, GLenum fmt, GLint filter,
+                          FILE *f, int *width, int *height, int *maxval)
+{
+    int w, h, m, bpp;
+    GLenum type;
+    uint8_t *data = read_pnm(f, &w, &h, &bpp, &m);
+    GLint hqtexfmt = detect_hqtexfmt(gl);
+    if (!data || (bpp != 3 && bpp != 6)) {
+        free(data);
+        return 0;
+    }
+    if (!fmt) {
+        fmt = bpp == 6 ? hqtexfmt : 3;
+        if (fmt == GL_FLOAT_RGB32_NV && target != GL_TEXTURE_RECTANGLE)
+            fmt = GL_RGB16;
+    }
+    type = bpp == 6 ? GL_UNSIGNED_SHORT : GL_UNSIGNED_BYTE;
+    glCreateClearTex(gl, target, fmt, GL_RGB, type, filter, w, h, 0);
+    glUploadTex(gl, target, GL_RGB, type,
+                data, w * bpp, 0, 0, w, h, 0);
+    free(data);
+    if (width)
+        *width = w;
+    if (height)
+        *height = h;
+    if (maxval)
+        *maxval = m;
+    return 1;
+}
+
+
+/**
+ * \brief Setup ATI version of register combiners for YUV to RGB conversion.
+ * \param csp_params parameters used for colorspace conversion
+ * \param text if set use the GL_ATI_text_fragment_shader API as
+ *             used on OS X.
+ */
+static void glSetupYUVFragmentATI(GL *gl, struct mp_csp_params *csp_params,
+                                  int text)
+{
+    GLint i;
+    float yuv2rgb[3][4];
+
+    gl->GetIntegerv(GL_MAX_TEXTURE_UNITS, &i);
+    if (i < 3)
+        mp_msg(MSGT_VO, MSGL_ERR,
+               "[gl] 3 texture units needed for YUV combiner (ATI) support (found %i)\n", i);
+
+    mp_get_yuv2rgb_coeffs(csp_params, yuv2rgb);
+    for (i = 0; i < 3; i++) {
+        int j;
+        yuv2rgb[i][3] -= -0.5 * (yuv2rgb[i][1] + yuv2rgb[i][2]);
+        for (j = 0; j < 4; j++) {
+            yuv2rgb[i][j] *= 0.125;
+            yuv2rgb[i][j] += 0.5;
+            if (yuv2rgb[i][j] > 1)
+                yuv2rgb[i][j] = 1;
+            if (yuv2rgb[i][j] < 0)
+                yuv2rgb[i][j] = 0;
+        }
+    }
+    if (text == 0) {
+        GLfloat c0[4] = { yuv2rgb[0][0], yuv2rgb[1][0], yuv2rgb[2][0] };
+        GLfloat c1[4] = { yuv2rgb[0][1], yuv2rgb[1][1], yuv2rgb[2][1] };
+        GLfloat c2[4] = { yuv2rgb[0][2], yuv2rgb[1][2], yuv2rgb[2][2] };
+        GLfloat c3[4] = { yuv2rgb[0][3], yuv2rgb[1][3], yuv2rgb[2][3] };
+        if (!gl->BeginFragmentShader || !gl->EndFragmentShader ||
+            !gl->SetFragmentShaderConstant || !gl->SampleMap ||
+            !gl->ColorFragmentOp2 || !gl->ColorFragmentOp3) {
+            mp_msg(MSGT_VO, MSGL_FATAL, "[gl] Combiner (ATI) functions missing!\n");
+            return;
+        }
+        gl->GetIntegerv(GL_NUM_FRAGMENT_REGISTERS_ATI, &i);
+        if (i < 3)
+            mp_msg(MSGT_VO, MSGL_ERR,
+                   "[gl] 3 registers needed for YUV combiner (ATI) support (found %i)\n", i);
+        gl->BeginFragmentShader();
+        gl->SetFragmentShaderConstant(GL_CON_0_ATI, c0);
+        gl->SetFragmentShaderConstant(GL_CON_1_ATI, c1);
+        gl->SetFragmentShaderConstant(GL_CON_2_ATI, c2);
+        gl->SetFragmentShaderConstant(GL_CON_3_ATI, c3);
+        gl->SampleMap(GL_REG_0_ATI, GL_TEXTURE0, GL_SWIZZLE_STR_ATI);
+        gl->SampleMap(GL_REG_1_ATI, GL_TEXTURE1, GL_SWIZZLE_STR_ATI);
+        gl->SampleMap(GL_REG_2_ATI, GL_TEXTURE2, GL_SWIZZLE_STR_ATI);
+        gl->ColorFragmentOp2(GL_MUL_ATI, GL_REG_1_ATI, GL_NONE, GL_NONE,
+                             GL_REG_1_ATI, GL_NONE, GL_BIAS_BIT_ATI,
+                             GL_CON_1_ATI, GL_NONE, GL_BIAS_BIT_ATI);
+        gl->ColorFragmentOp3(GL_MAD_ATI, GL_REG_2_ATI, GL_NONE, GL_NONE,
+                             GL_REG_2_ATI, GL_NONE, GL_BIAS_BIT_ATI,
+                             GL_CON_2_ATI, GL_NONE, GL_BIAS_BIT_ATI,
+                             GL_REG_1_ATI, GL_NONE, GL_NONE);
+        gl->ColorFragmentOp3(GL_MAD_ATI, GL_REG_0_ATI, GL_NONE, GL_NONE,
+                             GL_REG_0_ATI, GL_NONE, GL_NONE,
+                             GL_CON_0_ATI, GL_NONE, GL_BIAS_BIT_ATI,
+                             GL_REG_2_ATI, GL_NONE, GL_NONE);
+        gl->ColorFragmentOp2(GL_ADD_ATI, GL_REG_0_ATI, GL_NONE, GL_8X_BIT_ATI,
+                             GL_REG_0_ATI, GL_NONE, GL_NONE,
+                             GL_CON_3_ATI, GL_NONE, GL_BIAS_BIT_ATI);
+        gl->EndFragmentShader();
+    } else {
+        static const char template[] =
+            "!!ATIfs1.0\n"
+            "StartConstants;\n"
+            "  CONSTANT c0 = {%e, %e, %e};\n"
+            "  CONSTANT c1 = {%e, %e, %e};\n"
+            "  CONSTANT c2 = {%e, %e, %e};\n"
+            "  CONSTANT c3 = {%e, %e, %e};\n"
+            "EndConstants;\n"
+            "StartOutputPass;\n"
+            "  SampleMap r0, t0.str;\n"
+            "  SampleMap r1, t1.str;\n"
+            "  SampleMap r2, t2.str;\n"
+            "  MUL r1.rgb, r1.bias, c1.bias;\n"
+            "  MAD r2.rgb, r2.bias, c2.bias, r1;\n"
+            "  MAD r0.rgb, r0, c0.bias, r2;\n"
+            "  ADD r0.rgb.8x, r0, c3.bias;\n"
+            "EndPass;\n";
+        char buffer[512];
+        snprintf(buffer, sizeof(buffer), template,
+                 yuv2rgb[0][0], yuv2rgb[1][0], yuv2rgb[2][0],
+                 yuv2rgb[0][1], yuv2rgb[1][1], yuv2rgb[2][1],
+                 yuv2rgb[0][2], yuv2rgb[1][2], yuv2rgb[2][2],
+                 yuv2rgb[0][3], yuv2rgb[1][3], yuv2rgb[2][3]);
+        mp_msg(MSGT_VO, MSGL_DBG2, "[gl] generated fragment program:\n%s\n",
+               buffer);
+        loadGPUProgram(gl, GL_TEXT_FRAGMENT_SHADER_ATI, buffer);
+    }
+}
+
+// Replace all occurances of variables named "$"+name (e.g. $foo) in *text with
+// replace, and return the result. *text must have been allocated with talloc.
+static void replace_var_str(char **text, const char *name, const char *replace)
+{
+    size_t namelen = strlen(name);
+    char *nextvar = *text;
+    void *parent = talloc_parent(*text);
+    for (;;) {
+        nextvar = strchr(nextvar, '$');
+        if (!nextvar)
+            break;
+        char *until = nextvar;
+        nextvar++;
+        if (strncmp(nextvar, name, namelen) != 0)
+            continue;
+        nextvar += namelen;
+        // try not to replace prefixes of other vars (e.g. $foo vs. $foo_bar)
+        char term = nextvar[0];
+        if (isalnum(term) || term == '_')
+            continue;
+        int prelength = until - *text;
+        int postlength = nextvar - *text;
+        char *n = talloc_asprintf(parent, "%.*s%s%s", prelength, *text, replace,
+                                  nextvar);
+        talloc_free(*text);
+        *text = n;
+        nextvar = *text + postlength;
+    }
+}
+
+static void replace_var_float(char **text, const char *name, float replace)
+{
+    char *s = talloc_asprintf(NULL, "%e", replace);
+    replace_var_str(text, name, s);
+    talloc_free(s);
+}
+
+static void replace_var_char(char **text, const char *name, char replace)
+{
+    char s[2] = { replace, '\0' };
+    replace_var_str(text, name, s);
+}
+
+// Append template to *text. Possibly initialize *text if it's NULL.
+static void append_template(char **text, const char* template)
+{
+    if (!text)
+        *text = talloc_strdup(NULL, template);
+    else
+        *text = talloc_strdup_append(*text, template);
+}
+
+/**
+ * \brief helper function for gen_spline_lookup_tex
+ * \param x subpixel-position ((0,1) range) to calculate weights for
+ * \param dst where to store transformed weights, must provide space for 4 GLfloats
+ *
+ * calculates the weights and stores them after appropriate transformation
+ * for the scaler fragment program.
+ */
+static void store_weights(float x, GLfloat *dst)
+{
+    float w0 = (((-1 * x + 3) * x - 3) * x + 1) / 6;
+    float w1 = (((3 * x - 6) * x + 0) * x + 4) / 6;
+    float w2 = (((-3 * x + 3) * x + 3) * x + 1) / 6;
+    float w3 = (((1 * x + 0) * x + 0) * x + 0) / 6;
+    *dst++ = 1 + x - w1 / (w0 + w1);
+    *dst++ = 1 - x + w3 / (w2 + w3);
+    *dst++ = w0 + w1;
+    *dst++ = 0;
+}
+
+//! to avoid artefacts this should be rather large
+#define LOOKUP_BSPLINE_RES (2 * 1024)
+/**
+ * \brief creates the 1D lookup texture needed for fast higher-order filtering
+ * \param unit texture unit to attach texture to
+ */
+static void gen_spline_lookup_tex(GL *gl, GLenum unit)
+{
+    GLfloat *tex = calloc(4 * LOOKUP_BSPLINE_RES, sizeof(*tex));
+    GLfloat *tp = tex;
+    int i;
+    for (i = 0; i < LOOKUP_BSPLINE_RES; i++) {
+        float x = (float)(i + 0.5) / LOOKUP_BSPLINE_RES;
+        store_weights(x, tp);
+        tp += 4;
+    }
+    store_weights(0, tex);
+    store_weights(1, &tex[4 * (LOOKUP_BSPLINE_RES - 1)]);
+    gl->ActiveTexture(unit);
+    gl->TexImage1D(GL_TEXTURE_1D, 0, GL_RGBA16, LOOKUP_BSPLINE_RES, 0, GL_RGBA,
+                   GL_FLOAT, tex);
+    gl->TexParameterf(GL_TEXTURE_1D, GL_TEXTURE_PRIORITY, 1.0);
+    gl->TexParameteri(GL_TEXTURE_1D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
+    gl->TexParameteri(GL_TEXTURE_1D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
+    gl->TexParameteri(GL_TEXTURE_1D, GL_TEXTURE_WRAP_S, GL_REPEAT);
+    gl->ActiveTexture(GL_TEXTURE0);
+    free(tex);
+}
+
+#define NOISE_RES 2048
+
+/**
+ * \brief creates the 1D lookup texture needed to generate pseudo-random numbers.
+ * \param unit texture unit to attach texture to
+ */
+static void gen_noise_lookup_tex(GL *gl, GLenum unit) {
+    GLfloat *tex = calloc(NOISE_RES, sizeof(*tex));
+    uint32_t lcg = 0x79381c11;
+    int i;
+    for (i = 0; i < NOISE_RES; i++)
+        tex[i] = (double)i / (NOISE_RES - 1);
+    for (i = 0; i < NOISE_RES - 1; i++) {
+        int remain = NOISE_RES - i;
+        int idx = i + (lcg >> 16) % remain;
+        GLfloat tmp = tex[i];
+        tex[i] = tex[idx];
+        tex[idx] = tmp;
+        lcg = lcg * 1664525 + 1013904223;
+    }
+    gl->ActiveTexture(unit);
+    gl->TexImage1D(GL_TEXTURE_1D, 0, 1, NOISE_RES, 0, GL_RED, GL_FLOAT, tex);
+    gl->TexParameterf(GL_TEXTURE_1D, GL_TEXTURE_PRIORITY, 1.0);
+    gl->TexParameteri(GL_TEXTURE_1D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
+    gl->TexParameteri(GL_TEXTURE_1D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
+    gl->TexParameteri(GL_TEXTURE_1D, GL_TEXTURE_WRAP_S, GL_REPEAT);
+    gl->ActiveTexture(GL_TEXTURE0);
+    free(tex);
+}
+
+#define SAMPLE(dest, coord, texture) \
+    "TEX textemp, " coord ", " texture ", $tex_type;\n" \
+    "MOV " dest ", textemp.r;\n"
+
+static const char bilin_filt_template[] =
+    SAMPLE("yuv.$out_comp","fragment.texcoord[$in_tex]","texture[$in_tex]");
+
+#define BICUB_FILT_MAIN \
+    /* first y-interpolation */ \
+    "ADD coord, fragment.texcoord[$in_tex].xyxy, cdelta.xyxw;\n" \
+    "ADD coord2, fragment.texcoord[$in_tex].xyxy, cdelta.zyzw;\n" \
+    SAMPLE("a.r","coord.xyxy","texture[$in_tex]") \
+    SAMPLE("a.g","coord.zwzw","texture[$in_tex]") \
+    /* second y-interpolation */ \
+    SAMPLE("b.r","coord2.xyxy","texture[$in_tex]") \
+    SAMPLE("b.g","coord2.zwzw","texture[$in_tex]") \
+    "LRP a.b, parmy.b, a.rrrr, a.gggg;\n" \
+    "LRP a.a, parmy.b, b.rrrr, b.gggg;\n" \
+    /* x-interpolation */ \
+    "LRP yuv.$out_comp, parmx.b, a.bbbb, a.aaaa;\n"
+
+static const char bicub_filt_template_2D[] =
+    "MAD coord.xy, fragment.texcoord[$in_tex], {$texw, $texh}, {0.5, 0.5};\n"
+    "TEX parmx, coord.x, texture[$texs], 1D;\n"
+    "MUL cdelta.xz, parmx.rrgg, {-$ptw, 0, $ptw, 0};\n"
+    "TEX parmy, coord.y, texture[$texs], 1D;\n"
+    "MUL cdelta.yw, parmy.rrgg, {0, -$pth, 0, $pth};\n"
+    BICUB_FILT_MAIN;
+
+static const char bicub_filt_template_RECT[] =
+    "ADD coord, fragment.texcoord[$in_tex], {0.5, 0.5};\n"
+    "TEX parmx, coord.x, texture[$texs], 1D;\n"
+    "MUL cdelta.xz, parmx.rrgg, {-1, 0, 1, 0};\n"
+    "TEX parmy, coord.y, texture[$texs], 1D;\n"
+    "MUL cdelta.yw, parmy.rrgg, {0, -1, 0, 1};\n"
+    BICUB_FILT_MAIN;
+
+#define CALCWEIGHTS(t, s) \
+    "MAD "t ", {-0.5, 0.1666, 0.3333, -0.3333}, "s ", {1, 0, -0.5, 0.5};\n" \
+    "MAD "t ", "t ", "s ", {0, 0, -0.5, 0.5};\n" \
+    "MAD "t ", "t ", "s ", {-0.6666, 0, 0.8333, 0.1666};\n" \
+    "RCP a.x, "t ".z;\n" \
+    "RCP a.y, "t ".w;\n" \
+    "MAD "t ".xy, "t ".xyxy, a.xyxy, {1, 1, 0, 0};\n" \
+    "ADD "t ".x, "t ".xxxx, "s ";\n" \
+    "SUB "t ".y, "t ".yyyy, "s ";\n"
+
+static const char bicub_notex_filt_template_2D[] =
+    "MAD coord.xy, fragment.texcoord[$in_tex], {$texw, $texh}, {0.5, 0.5};\n"
+    "FRC coord.xy, coord.xyxy;\n"
+    CALCWEIGHTS("parmx", "coord.xxxx")
+    "MUL cdelta.xz, parmx.rrgg, {-$ptw, 0, $ptw, 0};\n"
+    CALCWEIGHTS("parmy", "coord.yyyy")
+    "MUL cdelta.yw, parmy.rrgg, {0, -$pth, 0, $pth};\n"
+    BICUB_FILT_MAIN;
+
+static const char bicub_notex_filt_template_RECT[] =
+    "ADD coord, fragment.texcoord[$in_tex], {0.5, 0.5};\n"
+    "FRC coord.xy, coord.xyxy;\n"
+    CALCWEIGHTS("parmx", "coord.xxxx")
+    "MUL cdelta.xz, parmx.rrgg, {-1, 0, 1, 0};\n"
+    CALCWEIGHTS("parmy", "coord.yyyy")
+    "MUL cdelta.yw, parmy.rrgg, {0, -1, 0, 1};\n"
+    BICUB_FILT_MAIN;
+
+#define BICUB_X_FILT_MAIN \
+    "ADD coord.xy, fragment.texcoord[$in_tex].xyxy, cdelta.xyxy;\n" \
+    "ADD coord2.xy, fragment.texcoord[$in_tex].xyxy, cdelta.zyzy;\n" \
+    SAMPLE("a.r","coord","texture[$in_tex]") \
+    SAMPLE("b.r","coord2","texture[$in_tex]") \
+    /* x-interpolation */ \
+    "LRP yuv.$out_comp, parmx.b, a.rrrr, b.rrrr;\n"
+
+static const char bicub_x_filt_template_2D[] =
+    "MAD coord.x, fragment.texcoord[$in_tex], {$texw}, {0.5};\n"
+    "TEX parmx, coord, texture[$texs], 1D;\n"
+    "MUL cdelta.xyz, parmx.rrgg, {-$ptw, 0, $ptw};\n"
+    BICUB_X_FILT_MAIN;
+
+static const char bicub_x_filt_template_RECT[] =
+    "ADD coord.x, fragment.texcoord[$in_tex], {0.5};\n"
+    "TEX parmx, coord, texture[$texs], 1D;\n"
+    "MUL cdelta.xyz, parmx.rrgg, {-1, 0, 1};\n"
+    BICUB_X_FILT_MAIN;
+
+static const char unsharp_filt_template[] =
+    "PARAM dcoord$out_comp = {$ptw_05, $pth_05, $ptw_05, -$pth_05};\n"
+    "ADD coord, fragment.texcoord[$in_tex].xyxy, dcoord$out_comp;\n"
+    "SUB coord2, fragment.texcoord[$in_tex].xyxy, dcoord$out_comp;\n"
+    SAMPLE("a.r","fragment.texcoord[$in_tex]","texture[$in_tex]")
+    SAMPLE("b.r","coord.xyxy","texture[$in_tex]")
+    SAMPLE("b.g","coord.zwzw","texture[$in_tex]")
+    "ADD b.r, b.r, b.g;\n"
+    SAMPLE("b.b","coord2.xyxy","texture[$in_tex]")
+    SAMPLE("b.g","coord2.zwzw","texture[$in_tex]")
+    "DP3 b, b, {0.25, 0.25, 0.25};\n"
+    "SUB b.r, a.r, b.r;\n"
+    "MAD textemp.r, b.r, {$strength}, a.r;\n"
+    "MOV yuv.$out_comp, textemp.r;\n";
+
+static const char unsharp_filt_template2[] =
+    "PARAM dcoord$out_comp = {$ptw_12, $pth_12, $ptw_12, -$pth_12};\n"
+    "PARAM dcoord2$out_comp = {$ptw_15, 0, 0, $pth_15};\n"
+    "ADD coord, fragment.texcoord[$in_tex].xyxy, dcoord$out_comp;\n"
+    "SUB coord2, fragment.texcoord[$in_tex].xyxy, dcoord$out_comp;\n"
+    SAMPLE("a.r","fragment.texcoord[$in_tex]","texture[$in_tex]")
+    SAMPLE("b.r","coord.xyxy","texture[$in_tex]")
+    SAMPLE("b.g","coord.zwzw","texture[$in_tex]")
+    "ADD b.r, b.r, b.g;\n"
+    SAMPLE("b.b","coord2.xyxy","texture[$in_tex]")
+    SAMPLE("b.g","coord2.zwzw","texture[$in_tex]")
+    "ADD b.r, b.r, b.b;\n"
+    "ADD b.a, b.r, b.g;\n"
+    "ADD coord, fragment.texcoord[$in_tex].xyxy, dcoord2$out_comp;\n"
+    "SUB coord2, fragment.texcoord[$in_tex].xyxy, dcoord2$out_comp;\n"
+    SAMPLE("b.r","coord.xyxy","texture[$in_tex]")
+    SAMPLE("b.g","coord.zwzw","texture[$in_tex]")
+    "ADD b.r, b.r, b.g;\n"
+    SAMPLE("b.b","coord2.xyxy","texture[$in_tex]")
+    SAMPLE("b.g","coord2.zwzw","texture[$in_tex]")
+    "DP4 b.r, b, {-0.1171875, -0.1171875, -0.1171875, -0.09765625};\n"
+    "MAD b.r, a.r, {0.859375}, b.r;\n"
+    "MAD textemp.r, b.r, {$strength}, a.r;\n"
+    "MOV yuv.$out_comp, textemp.r;\n";
+
+static const char yuv_prog_template[] =
+    "PARAM ycoef = {$cm11, $cm21, $cm31};\n"
+    "PARAM ucoef = {$cm12, $cm22, $cm32};\n"
+    "PARAM vcoef = {$cm13, $cm23, $cm33};\n"
+    "PARAM offsets = {$cm14, $cm24, $cm34};\n"
+    "TEMP res;\n"
+    "MAD res.rgb, yuv.rrrr, ycoef, offsets;\n"
+    "MAD res.rgb, yuv.gggg, ucoef, res;\n"
+    "MAD res.rgb, yuv.bbbb, vcoef, res;\n";
+
+static const char yuv_pow_prog_template[] =
+    "PARAM ycoef = {$cm11, $cm21, $cm31};\n"
+    "PARAM ucoef = {$cm12, $cm22, $cm32};\n"
+    "PARAM vcoef = {$cm13, $cm23, $cm33};\n"
+    "PARAM offsets = {$cm14, $cm24, $cm34};\n"
+    "PARAM gamma = {$gamma_r, $gamma_g, $gamma_b};\n"
+    "TEMP res;\n"
+    "MAD res.rgb, yuv.rrrr, ycoef, offsets;\n"
+    "MAD res.rgb, yuv.gggg, ucoef, res;\n"
+    "MAD_SAT res.rgb, yuv.bbbb, vcoef, res;\n"
+    "POW res.r, res.r, gamma.r;\n"
+    "POW res.g, res.g, gamma.g;\n"
+    "POW res.b, res.b, gamma.b;\n";
+
+static const char yuv_lookup_prog_template[] =
+    "PARAM ycoef = {$cm11, $cm21, $cm31, 0};\n"
+    "PARAM ucoef = {$cm12, $cm22, $cm32, 0};\n"
+    "PARAM vcoef = {$cm13, $cm23, $cm33, 0};\n"
+    "PARAM offsets = {$cm14, $cm24, $cm34, 0.125};\n"
+    "TEMP res;\n"
+    "MAD res, yuv.rrrr, ycoef, offsets;\n"
+    "MAD res.rgb, yuv.gggg, ucoef, res;\n"
+    "MAD res.rgb, yuv.bbbb, vcoef, res;\n"
+    "TEX res.r, res.raaa, texture[$conv_tex0], 2D;\n"
+    "ADD res.a, res.a, 0.25;\n"
+    "TEX res.g, res.gaaa, texture[$conv_tex0], 2D;\n"
+    "ADD res.a, res.a, 0.25;\n"
+    "TEX res.b, res.baaa, texture[$conv_tex0], 2D;\n";
+
+static const char yuv_lookup3d_prog_template[] =
+    "TEMP res;\n"
+    "TEX res, yuv, texture[$conv_tex0], 3D;\n";
+
+static const char noise_filt_template[] =
+    "MUL coord.xy, fragment.texcoord[0], {$noise_sx, $noise_sy};\n"
+    "TEMP rand;\n"
+    "TEX rand.r, coord.x, texture[$noise_filt_tex], 1D;\n"
+    "ADD rand.r, rand.r, coord.y;\n"
+    "TEX rand.r, rand.r, texture[$noise_filt_tex], 1D;\n"
+    "MAD res.rgb, rand.rrrr, {$noise_str, $noise_str, $noise_str}, res;\n";
+
+/**
+ * \brief creates and initializes helper textures needed for scaling texture read
+ * \param scaler scaler type to create texture for
+ * \param texu contains next free texture unit number
+ * \param texs texture unit ids for the scaler are stored in this array
+ */
+static void create_scaler_textures(GL *gl, int scaler, int *texu, char *texs)
+{
+    switch (scaler) {
+    case YUV_SCALER_BILIN:
+    case YUV_SCALER_BICUB_NOTEX:
+    case YUV_SCALER_UNSHARP:
+    case YUV_SCALER_UNSHARP2:
+        break;
+    case YUV_SCALER_BICUB:
+    case YUV_SCALER_BICUB_X:
+        texs[0] = (*texu)++;
+        gen_spline_lookup_tex(gl, GL_TEXTURE0 + texs[0]);
+        texs[0] += '0';
+        break;
+    default:
+        mp_msg(MSGT_VO, MSGL_ERR, "[gl] unknown scaler type %i\n", scaler);
+    }
+}
+
+//! resolution of texture for gamma lookup table
+#define LOOKUP_RES 512
+//! resolution for 3D yuv->rgb conversion lookup table
+#define LOOKUP_3DRES 32
+/**
+ * \brief creates and initializes helper textures needed for yuv conversion
+ * \param params struct containing parameters like brightness, gamma, ...
+ * \param texu contains next free texture unit number
+ * \param texs texture unit ids for the conversion are stored in this array
+ */
+static void create_conv_textures(GL *gl, gl_conversion_params_t *params,
+                                 int *texu, char *texs)
+{
+    unsigned char *lookup_data = NULL;
+    int conv = YUV_CONVERSION(params->type);
+    switch (conv) {
+    case YUV_CONVERSION_FRAGMENT:
+    case YUV_CONVERSION_FRAGMENT_POW:
+        break;
+    case YUV_CONVERSION_FRAGMENT_LOOKUP:
+        texs[0] = (*texu)++;
+        gl->ActiveTexture(GL_TEXTURE0 + texs[0]);
+        lookup_data = malloc(4 * LOOKUP_RES);
+        mp_gen_gamma_map(lookup_data, LOOKUP_RES, params->csp_params.rgamma);
+        mp_gen_gamma_map(&lookup_data[LOOKUP_RES], LOOKUP_RES,
+                         params->csp_params.ggamma);
+        mp_gen_gamma_map(&lookup_data[2 * LOOKUP_RES], LOOKUP_RES,
+                         params->csp_params.bgamma);
+        glCreateClearTex(gl, GL_TEXTURE_2D, GL_LUMINANCE8, GL_LUMINANCE,
+                         GL_UNSIGNED_BYTE, GL_LINEAR, LOOKUP_RES, 4, 0);
+        glUploadTex(gl, GL_TEXTURE_2D, GL_LUMINANCE, GL_UNSIGNED_BYTE,
+                    lookup_data, LOOKUP_RES, 0, 0, LOOKUP_RES, 4, 0);
+        gl->ActiveTexture(GL_TEXTURE0);
+        texs[0] += '0';
+        break;
+    case YUV_CONVERSION_FRAGMENT_LOOKUP3D:
+    {
+        int sz = LOOKUP_3DRES + 2; // texture size including borders
+        if (!gl->TexImage3D) {
+            mp_msg(MSGT_VO, MSGL_ERR, "[gl] Missing 3D texture function!\n");
+            break;
+        }
+        texs[0] = (*texu)++;
+        gl->ActiveTexture(GL_TEXTURE0 + texs[0]);
+        lookup_data = malloc(3 * sz * sz * sz);
+        mp_gen_yuv2rgb_map(&params->csp_params, lookup_data, LOOKUP_3DRES);
+        glAdjustAlignment(gl, sz);
+        gl->PixelStorei(GL_UNPACK_ROW_LENGTH, 0);
+        gl->TexImage3D(GL_TEXTURE_3D, 0, 3, sz, sz, sz, 1,
+                       GL_RGB, GL_UNSIGNED_BYTE, lookup_data);
+        gl->TexParameterf(GL_TEXTURE_3D, GL_TEXTURE_PRIORITY, 1.0);
+        gl->TexParameteri(GL_TEXTURE_3D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
+        gl->TexParameteri(GL_TEXTURE_3D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
+        gl->TexParameteri(GL_TEXTURE_3D, GL_TEXTURE_WRAP_S, GL_CLAMP);
+        gl->TexParameteri(GL_TEXTURE_3D, GL_TEXTURE_WRAP_T, GL_CLAMP);
+        gl->TexParameteri(GL_TEXTURE_3D, GL_TEXTURE_WRAP_R, GL_CLAMP);
+        gl->ActiveTexture(GL_TEXTURE0);
+        texs[0] += '0';
+    }
+    break;
+    default:
+        mp_msg(MSGT_VO, MSGL_ERR, "[gl] unknown conversion type %i\n", conv);
+    }
+    free(lookup_data);
+}
+
+/**
+ * \brief adds a scaling texture read at the current fragment program position
+ * \param scaler type of scaler to insert
+ * \param prog pointer to fragment program so far
+ * \param texs array containing the texture unit identifiers for this scaler
+ * \param in_tex texture unit the scaler should read from
+ * \param out_comp component of the yuv variable the scaler stores the result in
+ * \param rect if rectangular (pixel) adressing should be used for in_tex
+ * \param texw width of the in_tex texture
+ * \param texh height of the in_tex texture
+ * \param strength strength of filter effect if the scaler does some kind of filtering
+ */
+static void add_scaler(int scaler, char **prog, char *texs,
+                       char in_tex, char out_comp, int rect, int texw, int texh,
+                       double strength)
+{
+    const char *ttype = rect ? "RECT" : "2D";
+    const float ptw = rect ? 1.0 : 1.0 / texw;
+    const float pth = rect ? 1.0 : 1.0 / texh;
+    switch (scaler) {
+    case YUV_SCALER_BILIN:
+        append_template(prog, bilin_filt_template);
+        break;
+    case YUV_SCALER_BICUB:
+        if (rect)
+            append_template(prog, bicub_filt_template_RECT);
+        else
+            append_template(prog, bicub_filt_template_2D);
+        break;
+    case YUV_SCALER_BICUB_X:
+        if (rect)
+            append_template(prog, bicub_x_filt_template_RECT);
+        else
+            append_template(prog, bicub_x_filt_template_2D);
+        break;
+    case YUV_SCALER_BICUB_NOTEX:
+        if (rect)
+            append_template(prog, bicub_notex_filt_template_RECT);
+        else
+            append_template(prog, bicub_notex_filt_template_2D);
+        break;
+    case YUV_SCALER_UNSHARP:
+        append_template(prog, unsharp_filt_template);
+        break;
+    case YUV_SCALER_UNSHARP2:
+        append_template(prog, unsharp_filt_template2);
+        break;
+    }
+
+    replace_var_char(prog, "texs", texs[0]);
+    replace_var_char(prog, "in_tex", in_tex);
+    replace_var_char(prog, "out_comp", out_comp);
+    replace_var_str(prog, "tex_type", ttype);
+    replace_var_float(prog, "texw", texw);
+    replace_var_float(prog, "texh", texh);
+    replace_var_float(prog, "ptw", ptw);
+    replace_var_float(prog, "pth", pth);
+
+    // this is silly, not sure if that couldn't be in the shader source instead
+    replace_var_float(prog, "ptw_05", ptw * 0.5);
+    replace_var_float(prog, "pth_05", pth * 0.5);
+    replace_var_float(prog, "ptw_15", ptw * 1.5);
+    replace_var_float(prog, "pth_15", pth * 1.5);
+    replace_var_float(prog, "ptw_12", ptw * 1.2);
+    replace_var_float(prog, "pth_12", pth * 1.2);
+
+    replace_var_float(prog, "strength", strength);
+}
+
+static const struct {
+    const char *name;
+    GLenum cur;
+    GLenum max;
+} progstats[] = {
+    {"instructions", 0x88A0, 0x88A1},
+    {"native instructions", 0x88A2, 0x88A3},
+    {"temporaries", 0x88A4, 0x88A5},
+    {"native temporaries", 0x88A6, 0x88A7},
+    {"parameters", 0x88A8, 0x88A9},
+    {"native parameters", 0x88AA, 0x88AB},
+    {"attribs", 0x88AC, 0x88AD},
+    {"native attribs", 0x88AE, 0x88AF},
+    {"ALU instructions", 0x8805, 0x880B},
+    {"TEX instructions", 0x8806, 0x880C},
+    {"TEX indirections", 0x8807, 0x880D},
+    {"native ALU instructions", 0x8808, 0x880E},
+    {"native TEX instructions", 0x8809, 0x880F},
+    {"native TEX indirections", 0x880A, 0x8810},
+    {NULL, 0, 0}
+};
+
+/**
+ * \brief load the specified GPU Program
+ * \param target program target to load into, only GL_FRAGMENT_PROGRAM is tested
+ * \param prog program string
+ * \return 1 on success, 0 otherwise
+ */
+static int loadGPUProgram(GL *gl, GLenum target, char *prog)
+{
+    int i;
+    GLint cur = 0, max = 0, err = 0;
+    if (!gl->ProgramString) {
+        mp_msg(MSGT_VO, MSGL_ERR, "[gl] Missing GPU program function\n");
+        return 0;