Get Polygon offset fill to work.
[profile/ivi/mesa.git] / src / mesa / drivers / dri / r300 / r300_state.c
index f28c677..72dfe7a 100644 (file)
@@ -48,6 +48,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "swrast_setup/swrast_setup.h"
 #include "array_cache/acache.h"
 #include "tnl/tnl.h"
+#include "texformat.h"
 
 #include "radeon_ioctl.h"
 #include "radeon_state.h"
@@ -56,19 +57,459 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #include "r300_state.h"
 #include "r300_reg.h"
 #include "r300_program.h"
+#include "r300_emit.h"
+#include "r300_fixed_pipelines.h"
+
+
+static void r300AlphaFunc(GLcontext * ctx, GLenum func, GLfloat ref)
+{
+       r300ContextPtr rmesa = R300_CONTEXT(ctx);
+       int pp_misc = rmesa->hw.at.cmd[R300_AT_ALPHA_TEST];
+       GLubyte refByte;
+
+       CLAMPED_FLOAT_TO_UBYTE(refByte, ref);
+
+       R300_STATECHANGE(rmesa, at);
+
+       pp_misc &= ~(R300_ALPHA_TEST_OP_MASK | R300_REF_ALPHA_MASK);
+       pp_misc |= (refByte & R300_REF_ALPHA_MASK);
+
+       switch (func) {
+       case GL_NEVER:
+               pp_misc |= R300_ALPHA_TEST_FAIL;
+               break;
+       case GL_LESS:
+               pp_misc |= R300_ALPHA_TEST_LESS;
+               break;
+       case GL_EQUAL:
+               pp_misc |= R300_ALPHA_TEST_EQUAL;
+               break;
+       case GL_LEQUAL:
+               pp_misc |= R300_ALPHA_TEST_LEQUAL;
+               break;
+       case GL_GREATER:
+               pp_misc |= R300_ALPHA_TEST_GREATER;
+               break;
+       case GL_NOTEQUAL:
+               pp_misc |= R300_ALPHA_TEST_NEQUAL;
+               break;
+       case GL_GEQUAL:
+               pp_misc |= R300_ALPHA_TEST_GEQUAL;
+               break;
+       case GL_ALWAYS:
+               pp_misc |= R300_ALPHA_TEST_PASS;
+               break;
+       }
+
+       rmesa->hw.at.cmd[R300_AT_ALPHA_TEST] = pp_misc;
+}
+
+static void r300BlendColor(GLcontext * ctx, const GLfloat cf[4])
+{
+       GLubyte color[4];
+       r300ContextPtr rmesa = R300_CONTEXT(ctx);
+       fprintf(stderr, "%s:%s is not implemented yet. Fixme !\n", __FILE__, __FUNCTION__);
+       #if 0
+       R200_STATECHANGE(rmesa, ctx);
+       CLAMPED_FLOAT_TO_UBYTE(color[0], cf[0]);
+       CLAMPED_FLOAT_TO_UBYTE(color[1], cf[1]);
+       CLAMPED_FLOAT_TO_UBYTE(color[2], cf[2]);
+       CLAMPED_FLOAT_TO_UBYTE(color[3], cf[3]);
+       if (rmesa->radeon.radeonScreen->drmSupportsBlendColor)
+               rmesa->hw.ctx.cmd[CTX_RB3D_BLENDCOLOR] =
+                   radeonPackColor(4, color[0], color[1], color[2], color[3]);
+       #endif
+}
+
+/**
+ * Calculate the hardware blend factor setting.  This same function is used
+ * for source and destination of both alpha and RGB.
+ *
+ * \returns
+ * The hardware register value for the specified blend factor.  This value
+ * will need to be shifted into the correct position for either source or
+ * destination factor.
+ *
+ * \todo
+ * Since the two cases where source and destination are handled differently
+ * are essentially error cases, they should never happen.  Determine if these
+ * cases can be removed.
+ */
+static int blend_factor(GLenum factor, GLboolean is_src)
+{
+       int func;
+
+       switch (factor) {
+       case GL_ZERO:
+               func = R200_BLEND_GL_ZERO;
+               break;
+       case GL_ONE:
+               func = R200_BLEND_GL_ONE;
+               break;
+       case GL_DST_COLOR:
+               func = R200_BLEND_GL_DST_COLOR;
+               break;
+       case GL_ONE_MINUS_DST_COLOR:
+               func = R200_BLEND_GL_ONE_MINUS_DST_COLOR;
+               break;
+       case GL_SRC_COLOR:
+               func = R200_BLEND_GL_SRC_COLOR;
+               break;
+       case GL_ONE_MINUS_SRC_COLOR:
+               func = R200_BLEND_GL_ONE_MINUS_SRC_COLOR;
+               break;
+       case GL_SRC_ALPHA:
+               func = R200_BLEND_GL_SRC_ALPHA;
+               break;
+       case GL_ONE_MINUS_SRC_ALPHA:
+               func = R200_BLEND_GL_ONE_MINUS_SRC_ALPHA;
+               break;
+       case GL_DST_ALPHA:
+               func = R200_BLEND_GL_DST_ALPHA;
+               break;
+       case GL_ONE_MINUS_DST_ALPHA:
+               func = R200_BLEND_GL_ONE_MINUS_DST_ALPHA;
+               break;
+       case GL_SRC_ALPHA_SATURATE:
+               func =
+                   (is_src) ? R200_BLEND_GL_SRC_ALPHA_SATURATE :
+                   R200_BLEND_GL_ZERO;
+               break;
+       case GL_CONSTANT_COLOR:
+               func = R200_BLEND_GL_CONST_COLOR;
+               break;
+       case GL_ONE_MINUS_CONSTANT_COLOR:
+               func = R200_BLEND_GL_ONE_MINUS_CONST_COLOR;
+               break;
+       case GL_CONSTANT_ALPHA:
+               func = R200_BLEND_GL_CONST_ALPHA;
+               break;
+       case GL_ONE_MINUS_CONSTANT_ALPHA:
+               func = R200_BLEND_GL_ONE_MINUS_CONST_ALPHA;
+               break;
+       default:
+               func = (is_src) ? R200_BLEND_GL_ONE : R200_BLEND_GL_ZERO;
+       }
+       return func;
+}
+
+/**
+ * Sets both the blend equation and the blend function.
+ * This is done in a single
+ * function because some blend equations (i.e., \c GL_MIN and \c GL_MAX)
+ * change the interpretation of the blend function.
+ * Also, make sure that blend function and blend equation are set to their default
+ * value if color blending is not enabled, since at least blend equations GL_MIN
+ * and GL_FUNC_REVERSE_SUBTRACT will cause wrong results otherwise for
+ * unknown reasons.
+ */
+
+/* helper function */
+static void r300_set_blend_cntl(r300ContextPtr rmesa, int func, int eqn, int cbits, int funcA, int eqnA)
+{
+       GLuint new_ablend, new_cblend;
+
+       #if 0
+       fprintf(stderr, "eqnA=%08x funcA=%08x eqn=%08x func=%08x cbits=%08x\n", eqnA, funcA, eqn, func, cbits);
+       #endif
+       new_ablend = eqnA | funcA;
+       new_cblend = eqn | func;
+       if(funcA == func){
+               new_cblend |=  R300_BLEND_NO_SEPARATE;
+               }
+       new_cblend |= cbits;
+       
+       if((new_ablend != rmesa->hw.bld.cmd[R300_BLD_ABLEND])
+               || (new_cblend != rmesa->hw.bld.cmd[R300_BLD_CBLEND])){
+               R300_STATECHANGE(rmesa, bld);
+               rmesa->hw.bld.cmd[R300_BLD_ABLEND]=new_ablend;
+               rmesa->hw.bld.cmd[R300_BLD_CBLEND]=new_cblend;
+               }
+}
+
+static void r300_set_blend_state(GLcontext * ctx)
+{
+       r300ContextPtr rmesa = R300_CONTEXT(ctx);
+       #if 0
+       GLuint cntl = rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] &
+           ~(R300_ROP_ENABLE | R300_ALPHA_BLEND_ENABLE |
+             R300_SEPARATE_ALPHA_ENABLE);
+       #endif
+
+       int func = (R200_BLEND_GL_ONE << R200_SRC_BLEND_SHIFT) |
+           (R200_BLEND_GL_ZERO << R200_DST_BLEND_SHIFT);
+       int eqn = R200_COMB_FCN_ADD_CLAMP;
+       int funcA = (R200_BLEND_GL_ONE << R200_SRC_BLEND_SHIFT) |
+           (R200_BLEND_GL_ZERO << R200_DST_BLEND_SHIFT);
+       int eqnA = R200_COMB_FCN_ADD_CLAMP;
+
+
+       if (rmesa->radeon.radeonScreen->drmSupportsBlendColor) {
+               if (ctx->Color._LogicOpEnabled) {
+                       #if 0
+                       rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] =
+                           cntl | R300_ROP_ENABLE;
+                       #endif
+                       r300_set_blend_cntl(rmesa,
+                               func, eqn, 0,
+                               func, eqn);
+                       return;
+               } else if (ctx->Color.BlendEnabled) {
+                       #if 0
+                       rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] =
+                           cntl | R300_ALPHA_BLEND_ENABLE |
+                           R300_SEPARATE_ALPHA_ENABLE;
+                       #endif
+               } else {
+                       #if 0
+                       rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] = cntl;
+                       #endif
+                       r300_set_blend_cntl(rmesa,
+                               func, eqn, 0,
+                               func, eqn);
+                       return;
+               }
+       } else {
+               if (ctx->Color._LogicOpEnabled) {
+                       #if 0
+                       rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] =
+                           cntl | R300_ROP_ENABLE;
+                       rmesa->hw.ctx.cmd[CTX_RB3D_BLENDCNTL] = eqn | func;
+                       #endif
+                       return;
+               } else if (ctx->Color.BlendEnabled) {
+                       #if 0
+                       rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] =
+                           cntl | R300_ALPHA_BLEND_ENABLE;
+                       #endif
+               } else {
+                       #if 0
+                       rmesa->hw.ctx.cmd[CTX_RB3D_CNTL] = cntl;
+                       rmesa->hw.ctx.cmd[CTX_RB3D_BLENDCNTL] = eqn | func;
+                       #endif
+                       r300_set_blend_cntl(rmesa,
+                               func, eqn, 0,
+                               func, eqn);
+                       return;
+               }
+       }
+
+       func =
+           (blend_factor(ctx->Color.BlendSrcRGB, GL_TRUE) <<
+            R200_SRC_BLEND_SHIFT) | (blend_factor(ctx->Color.BlendDstRGB,
+                                                  GL_FALSE) <<
+                                     R200_DST_BLEND_SHIFT);
+
+       switch (ctx->Color.BlendEquationRGB) {
+       case GL_FUNC_ADD:
+               eqn = R300_COMB_FCN_ADD_CLAMP;
+               break;
+
+       case GL_FUNC_SUBTRACT:
+               eqn = R300_COMB_FCN_SUB_CLAMP;
+               break;
+
+       case GL_FUNC_REVERSE_SUBTRACT:
+               eqn = R200_COMB_FCN_RSUB_CLAMP;
+               break;
+
+       case GL_MIN:
+               eqn = R200_COMB_FCN_MIN;
+               func = (R200_BLEND_GL_ONE << R200_SRC_BLEND_SHIFT) |
+                   (R200_BLEND_GL_ONE << R200_DST_BLEND_SHIFT);
+               break;
+
+       case GL_MAX:
+               eqn = R200_COMB_FCN_MAX;
+               func = (R200_BLEND_GL_ONE << R200_SRC_BLEND_SHIFT) |
+                   (R200_BLEND_GL_ONE << R200_DST_BLEND_SHIFT);
+               break;
+
+       default:
+               fprintf(stderr,
+                       "[%s:%u] Invalid RGB blend equation (0x%04x).\n",
+                       __func__, __LINE__, ctx->Color.BlendEquationRGB);
+               return;
+       }
+
+       if (!rmesa->radeon.radeonScreen->drmSupportsBlendColor) {
+               #if 0
+               rmesa->hw.ctx.cmd[CTX_RB3D_BLENDCNTL] = eqn | func;
+               #endif
+               return;
+       }
+
+       funcA =
+           (blend_factor(ctx->Color.BlendSrcA, GL_TRUE) <<
+            R200_SRC_BLEND_SHIFT) | (blend_factor(ctx->Color.BlendDstA,
+                                                  GL_FALSE) <<
+                                     R200_DST_BLEND_SHIFT);
+
+       switch (ctx->Color.BlendEquationA) {
+       case GL_FUNC_ADD:
+               eqnA = R300_COMB_FCN_ADD_CLAMP;
+               break;
+
+       case GL_FUNC_SUBTRACT:
+               eqnA = R300_COMB_FCN_SUB_CLAMP;
+               break;
+
+       case GL_FUNC_REVERSE_SUBTRACT:
+               eqnA = R200_COMB_FCN_RSUB_CLAMP;
+               break;
+
+       case GL_MIN:
+               eqnA = R200_COMB_FCN_MIN;
+               funcA = (R200_BLEND_GL_ONE << R200_SRC_BLEND_SHIFT) |
+                   (R200_BLEND_GL_ONE << R200_DST_BLEND_SHIFT);
+               break;
+
+       case GL_MAX:
+               eqnA = R200_COMB_FCN_MAX;
+               funcA = (R200_BLEND_GL_ONE << R200_SRC_BLEND_SHIFT) |
+                   (R200_BLEND_GL_ONE << R200_DST_BLEND_SHIFT);
+               break;
+
+       default:
+               fprintf(stderr, "[%s:%u] Invalid A blend equation (0x%04x).\n",
+                       __func__, __LINE__, ctx->Color.BlendEquationA);
+               return;
+       }
+
+       r300_set_blend_cntl(rmesa,
+               func, eqn, R300_BLEND_UNKNOWN | R300_BLEND_ENABLE,
+               funcA, eqnA);
+       r300_set_blend_cntl(rmesa,
+               func, eqn, R300_BLEND_UNKNOWN | R300_BLEND_ENABLE,
+               funcA, eqnA);
+}
+
+static void r300BlendEquationSeparate(GLcontext * ctx,
+                                     GLenum modeRGB, GLenum modeA)
+{
+       r300_set_blend_state(ctx);
+}
+
+static void r300BlendFuncSeparate(GLcontext * ctx,
+                                 GLenum sfactorRGB, GLenum dfactorRGB,
+                                 GLenum sfactorA, GLenum dfactorA)
+{
+       r300_set_blend_state(ctx);
+}
+
+/**
+ * Update our tracked culling state based on Mesa's state.
+ */
+static void r300UpdateCulling(GLcontext* ctx)
+{
+       r300ContextPtr r300 = R300_CONTEXT(ctx);
+       uint32_t val = 0;
+
+       R300_STATECHANGE(r300, cul);
+       if (ctx->Polygon.CullFlag) {
+               if (ctx->Polygon.CullFaceMode == GL_FRONT_AND_BACK)
+                       val = R300_CULL_FRONT|R300_CULL_BACK;
+               else if (ctx->Polygon.CullFaceMode == GL_FRONT)
+                       val = R300_CULL_FRONT;
+               else
+                       val = R300_CULL_BACK;
+
+               if (ctx->Polygon.FrontFace == GL_CW)
+                       val |= R300_FRONT_FACE_CW;
+               else
+                       val |= R300_FRONT_FACE_CCW;             
+       }
+       r300->hw.cul.cmd[R300_CUL_CULL] = val;
+}
 
 
 /**
  * Handle glEnable()/glDisable().
+ *
+ * \note Mesa already filters redundant calls to glEnable/glDisable.
  */
 static void r300Enable(GLcontext* ctx, GLenum cap, GLboolean state)
 {
+       r300ContextPtr r300 = R300_CONTEXT(ctx);
+       uint32_t newval;
+
        if (RADEON_DEBUG & DEBUG_STATE)
                fprintf(stderr, "%s( %s = %s )\n", __FUNCTION__,
                        _mesa_lookup_enum_by_nr(cap),
                        state ? "GL_TRUE" : "GL_FALSE");
 
        switch (cap) {
+               /* Fast track this one...
+                */
+       case GL_TEXTURE_1D:
+       case GL_TEXTURE_2D:
+       case GL_TEXTURE_3D:
+               break;
+
+       case GL_ALPHA_TEST:
+               R200_STATECHANGE(r300, at);
+               if (state) {
+                       r300->hw.at.cmd[R300_AT_ALPHA_TEST] |=
+                           R300_ALPHA_TEST_ENABLE;
+               } else {
+                       r300->hw.at.cmd[R300_AT_ALPHA_TEST] |=
+                           ~R300_ALPHA_TEST_ENABLE;
+               }
+               break;
+
+       case GL_BLEND:
+       case GL_COLOR_LOGIC_OP:
+               r300_set_blend_state(ctx);
+               break;
+
+       case GL_DEPTH_TEST:
+               R300_STATECHANGE(r300, zs);
+
+               if (state) {
+                       if (ctx->Depth.Mask)
+                               newval = R300_RB3D_Z_TEST_AND_WRITE;
+                       else
+                               newval = R300_RB3D_Z_TEST;
+               } else
+                       newval = 0;
+
+               r300->hw.zs.cmd[R300_ZS_CNTL_0] = newval;
+               break;
+
+       case GL_STENCIL_TEST:
+
+               WARN_ONCE("Do not know how to enable stencil. Help me !\n");
+               
+               if (r300->state.hw_stencil) {
+                       //fprintf(stderr, "Stencil %s\n", state ? "enabled" : "disabled");
+                       R300_STATECHANGE(r300, zs);
+                       if (state) {
+                               r300->hw.zs.cmd[R300_ZS_CNTL_0] |=
+                                   R300_RB3D_STENCIL_ENABLE;
+                       } else {
+                               r300->hw.zs.cmd[R300_ZS_CNTL_0] &=
+                                   ~R300_RB3D_STENCIL_ENABLE;
+                       }
+               } else {
+                       FALLBACK(&r300->radeon, RADEON_FALLBACK_STENCIL, state);
+               }
+               break;
+
+       case GL_CULL_FACE:
+               r300UpdateCulling(ctx);
+               break;
+       case GL_POLYGON_OFFSET_FILL:
+               R300_STATECHANGE(r300, unk42B4);
+               if (state) {
+                       r300->hw.unk42B4.cmd[1] = (1<<1);
+               } else {
+                       r300->hw.unk42B4.cmd[1] = 0;
+               }
+               break;
+       case GL_VERTEX_PROGRAM_ARB:
+               //TCL_FALLBACK(rmesa->glCtx, R200_TCL_FALLBACK_TCL_DISABLE, state);
+       break;
+
        default:
                radeonEnable(ctx, cap, state);
                return;
@@ -77,6 +518,980 @@ static void r300Enable(GLcontext* ctx, GLenum cap, GLboolean state)
 
 
 /**
+ * Change the culling mode.
+ *
+ * \note Mesa already filters redundant calls to this function.
+ */
+static void r300CullFace(GLcontext* ctx, GLenum mode)
+{
+       (void)mode;
+
+       r300UpdateCulling(ctx);
+}
+
+
+/**
+ * Change the polygon orientation.
+ *
+ * \note Mesa already filters redundant calls to this function.
+ */
+static void r300FrontFace(GLcontext* ctx, GLenum mode)
+{
+       (void)mode;
+
+       r300UpdateCulling(ctx);
+}
+
+
+/**
+ * Change the depth testing function.
+ *
+ * \note Mesa already filters redundant calls to this function.
+ */
+static void r300DepthFunc(GLcontext* ctx, GLenum func)
+{
+       r300ContextPtr r300 = R300_CONTEXT(ctx);
+
+       R300_STATECHANGE(r300, zs);
+
+       r300->hw.zs.cmd[R300_ZS_CNTL_1] &= ~(R300_ZS_MASK << R300_RB3D_ZS1_DEPTH_FUNC_SHIFT);
+
+       switch(func) {
+       case GL_NEVER:
+               r300->hw.zs.cmd[R300_ZS_CNTL_1] |= R300_ZS_NEVER << R300_RB3D_ZS1_DEPTH_FUNC_SHIFT;
+               break;
+       case GL_LESS:
+               r300->hw.zs.cmd[R300_ZS_CNTL_1] |= R300_ZS_LESS << R300_RB3D_ZS1_DEPTH_FUNC_SHIFT;
+               break;
+       case GL_EQUAL:
+               r300->hw.zs.cmd[R300_ZS_CNTL_1] |= R300_ZS_EQUAL << R300_RB3D_ZS1_DEPTH_FUNC_SHIFT;
+               break;
+       case GL_LEQUAL:
+               r300->hw.zs.cmd[R300_ZS_CNTL_1] |= R300_ZS_LEQUAL << R300_RB3D_ZS1_DEPTH_FUNC_SHIFT;
+               break;
+       case GL_GREATER:
+               r300->hw.zs.cmd[R300_ZS_CNTL_1] |= R300_ZS_GREATER << R300_RB3D_ZS1_DEPTH_FUNC_SHIFT;
+               break;
+       case GL_NOTEQUAL:
+               r300->hw.zs.cmd[R300_ZS_CNTL_1] |= R300_ZS_NOTEQUAL << R300_RB3D_ZS1_DEPTH_FUNC_SHIFT;
+               break;
+       case GL_GEQUAL:
+               r300->hw.zs.cmd[R300_ZS_CNTL_1] |= R300_ZS_GEQUAL << R300_RB3D_ZS1_DEPTH_FUNC_SHIFT;
+               break;
+       case GL_ALWAYS:
+               r300->hw.zs.cmd[R300_ZS_CNTL_1] |= R300_ZS_ALWAYS << R300_RB3D_ZS1_DEPTH_FUNC_SHIFT;
+               break;
+       }
+
+}
+
+
+/**
+ * Enable/Disable depth writing.
+ *
+ * \note Mesa already filters redundant calls to this function.
+ */
+static void r300DepthMask(GLcontext* ctx, GLboolean mask)
+{
+       r300ContextPtr r300 = R300_CONTEXT(ctx);
+
+       if (!ctx->Depth.Test)
+               return;
+
+       R300_STATECHANGE(r300, zs);
+       r300->hw.zs.cmd[R300_ZS_CNTL_0] = mask
+               ? R300_RB3D_Z_TEST_AND_WRITE : R300_RB3D_Z_TEST;
+}
+
+
+/**
+ * Handle glColorMask()
+ */
+static void r300ColorMask(GLcontext* ctx,
+                         GLboolean r, GLboolean g, GLboolean b, GLboolean a)
+{
+       r300ContextPtr r300 = R300_CONTEXT(ctx);
+       int mask = (b << 0) | (g << 1) | (r << 2) | (a << 3);
+
+       if (mask != r300->hw.cmk.cmd[R300_CMK_COLORMASK]) {
+               R300_STATECHANGE(r300, cmk);
+               r300->hw.cmk.cmd[R300_CMK_COLORMASK] = mask;
+       }
+}
+
+/* =============================================================
+ * Point state
+ */
+static void r300PointSize(GLcontext * ctx, GLfloat size)
+{
+       r300ContextPtr r300 = R300_CONTEXT(ctx);
+
+       /* This might need fixing later */
+       R300_STATECHANGE(r300, vps);
+       r300->hw.vps.cmd[R300_VPS_POINTSIZE] = r300PackFloat32(1.0);
+}
+/* =============================================================
+ * Stencil
+ */
+
+ static int translate_stencil_func(int func)
+ {
+       switch (func) {
+       case GL_NEVER:
+                   return R300_ZS_NEVER;
+               break;
+       case GL_LESS:
+                   return R300_ZS_LESS;
+               break;
+       case GL_EQUAL:
+                   return R300_ZS_EQUAL;
+               break;
+       case GL_LEQUAL:
+                   return R300_ZS_LEQUAL;
+               break;
+       case GL_GREATER:
+                   return R300_ZS_GREATER;
+               break;
+       case GL_NOTEQUAL:
+                   return R300_ZS_NOTEQUAL;
+               break;
+       case GL_GEQUAL:
+                   return R300_ZS_GEQUAL;
+               break;
+       case GL_ALWAYS:
+                   return R300_ZS_ALWAYS;
+               break;
+       }
+ return 0;
+ }
+
+ static int translate_stencil_op(int op)
+{
+       switch (op) {
+       case GL_KEEP:
+                   return R300_ZS_KEEP;
+       case GL_ZERO:
+                   return R300_ZS_ZERO;
+       case GL_REPLACE:
+                   return R300_ZS_REPLACE;
+       case GL_INCR:
+                   return R300_ZS_INCR;
+       case GL_DECR:
+                   return R300_ZS_DECR;
+       case GL_INCR_WRAP_EXT:
+                   return R300_ZS_INCR_WRAP;
+       case GL_DECR_WRAP_EXT:
+                   return R300_ZS_DECR_WRAP;
+       case GL_INVERT:
+                   return R300_ZS_INVERT;
+       default:
+               WARN_ONCE("Do not know how to translate stencil op");
+               return R300_ZS_KEEP;
+       }
+}
+
+static void r300StencilFunc(GLcontext * ctx, GLenum func,
+                           GLint ref, GLuint mask)
+{
+       r300ContextPtr rmesa = R300_CONTEXT(ctx);
+       GLuint refmask = ((ctx->Stencil.Ref[0] << R300_RB3D_ZS2_STENCIL_REF_SHIFT) |
+                         (ctx->Stencil.
+                          ValueMask[0] << R300_RB3D_ZS2_STENCIL_MASK_SHIFT));
+       GLuint flag;
+
+       R200_STATECHANGE(rmesa, zs);
+
+       rmesa->hw.zs.cmd[R300_ZS_CNTL_1] &= ~(
+               (R300_ZS_MASK << R300_RB3D_ZS1_FRONT_FUNC_SHIFT)
+               | (R300_ZS_MASK << R300_RB3D_ZS1_BACK_FUNC_SHIFT));
+       rmesa->hw.zs.cmd[R300_ZS_CNTL_2] &=  ~((R300_ZS_MASK << R300_RB3D_ZS2_STENCIL_REF_SHIFT) |
+                                               (R300_ZS_MASK << R300_RB3D_ZS2_STENCIL_MASK_SHIFT));
+
+       flag = translate_stencil_func(ctx->Stencil.Function[0]);
+
+       rmesa->hw.zs.cmd[R300_ZS_CNTL_1] |= (flag << R300_RB3D_ZS1_FRONT_FUNC_SHIFT)
+                                         | (flag << R300_RB3D_ZS1_BACK_FUNC_SHIFT);
+       rmesa->hw.zs.cmd[R300_ZS_CNTL_2] |= refmask;
+}
+
+static void r300StencilMask(GLcontext * ctx, GLuint mask)
+{
+       r300ContextPtr rmesa = R300_CONTEXT(ctx);
+
+       R200_STATECHANGE(rmesa, zs);
+       rmesa->hw.zs.cmd[R300_ZS_CNTL_2]  &= ~(R300_ZS_MASK << R300_RB3D_ZS2_STENCIL_WRITE_MASK_SHIFT);
+       rmesa->hw.zs.cmd[R300_ZS_CNTL_2] |= ctx->Stencil.WriteMask[0] << R300_RB3D_ZS2_STENCIL_WRITE_MASK_SHIFT;
+}
+
+
+static void r300StencilOp(GLcontext * ctx, GLenum fail,
+                         GLenum zfail, GLenum zpass)
+{
+       r300ContextPtr rmesa = R300_CONTEXT(ctx);
+
+       R200_STATECHANGE(rmesa, zs);
+               /* It is easier to mask what's left.. */
+       rmesa->hw.zs.cmd[R300_ZS_CNTL_1] &= (R300_ZS_MASK << R300_RB3D_ZS1_DEPTH_FUNC_SHIFT);
+
+       rmesa->hw.zs.cmd[R300_ZS_CNTL_1] |=
+                (translate_stencil_op(ctx->Stencil.FailFunc[0]) << R300_RB3D_ZS1_FRONT_FAIL_OP_SHIFT)
+               |(translate_stencil_op(ctx->Stencil.ZFailFunc[0]) << R300_RB3D_ZS1_FRONT_ZFAIL_OP_SHIFT)
+               |(translate_stencil_op(ctx->Stencil.ZPassFunc[0]) << R300_RB3D_ZS1_FRONT_ZPASS_OP_SHIFT)
+               |(translate_stencil_op(ctx->Stencil.FailFunc[0]) << R300_RB3D_ZS1_BACK_FAIL_OP_SHIFT)
+               |(translate_stencil_op(ctx->Stencil.ZFailFunc[0]) << R300_RB3D_ZS1_BACK_ZFAIL_OP_SHIFT)
+               |(translate_stencil_op(ctx->Stencil.ZPassFunc[0]) << R300_RB3D_ZS1_BACK_ZPASS_OP_SHIFT);
+
+}
+
+static void r300ClearStencil(GLcontext * ctx, GLint s)
+{
+       r300ContextPtr rmesa = R300_CONTEXT(ctx);
+
+       /* Not sure whether this is correct.. */
+       R200_STATECHANGE(rmesa, zs);
+       rmesa->hw.zs.cmd[R300_ZS_CNTL_2] =
+           ((GLuint) ctx->Stencil.Clear |
+            (0xff << R200_STENCIL_MASK_SHIFT) |
+            (ctx->Stencil.WriteMask[0] << R200_STENCIL_WRITEMASK_SHIFT));
+}
+
+/* =============================================================
+ * Window position and viewport transformation
+ */
+
+/*
+ * To correctly position primitives:
+ */
+#define SUBPIXEL_X 0.125
+#define SUBPIXEL_Y 0.125
+
+void r300UpdateWindow(GLcontext * ctx)
+{
+       r300ContextPtr rmesa = R300_CONTEXT(ctx);
+       __DRIdrawablePrivate *dPriv = rmesa->radeon.dri.drawable;
+       GLfloat xoffset = dPriv ? (GLfloat) dPriv->x : 0;
+       GLfloat yoffset = dPriv ? (GLfloat) dPriv->y + dPriv->h : 0;
+       const GLfloat *v = ctx->Viewport._WindowMap.m;
+
+       GLfloat sx = v[MAT_SX];
+       GLfloat tx = v[MAT_TX] + xoffset + SUBPIXEL_X;
+       GLfloat sy = -v[MAT_SY];
+       GLfloat ty = (-v[MAT_TY]) + yoffset + SUBPIXEL_Y;
+       GLfloat sz = v[MAT_SZ] * rmesa->state.depth.scale;
+       GLfloat tz = v[MAT_TZ] * rmesa->state.depth.scale;
+
+       R300_FIREVERTICES(rmesa);
+       R300_STATECHANGE(rmesa, vpt);
+
+       rmesa->hw.vpt.cmd[R300_VPT_XSCALE]  = r300PackFloat32(sx);
+       rmesa->hw.vpt.cmd[R300_VPT_XOFFSET] = r300PackFloat32(tx);
+       rmesa->hw.vpt.cmd[R300_VPT_YSCALE]  = r300PackFloat32(sy);
+       rmesa->hw.vpt.cmd[R300_VPT_YOFFSET] = r300PackFloat32(ty);
+       rmesa->hw.vpt.cmd[R300_VPT_ZSCALE]  = r300PackFloat32(sz);
+       rmesa->hw.vpt.cmd[R300_VPT_ZOFFSET] = r300PackFloat32(tz);
+}
+
+static void r300Viewport(GLcontext * ctx, GLint x, GLint y,
+                        GLsizei width, GLsizei height)
+{
+       /* Don't pipeline viewport changes, conflict with window offset
+        * setting below.  Could apply deltas to rescue pipelined viewport
+        * values, or keep the originals hanging around.
+        */
+       R200_FIREVERTICES(R200_CONTEXT(ctx));
+       r300UpdateWindow(ctx);
+}
+
+static void r300DepthRange(GLcontext * ctx, GLclampd nearval, GLclampd farval)
+{
+       r300UpdateWindow(ctx);
+}
+
+/* =============================================================
+ * Polygon state
+ */
+
+static void r300PolygonOffset(GLcontext * ctx, GLfloat factor, GLfloat units)
+{
+       r300ContextPtr rmesa = R300_CONTEXT(ctx);
+       GLfloat constant = units * rmesa->state.depth.scale;
+
+/*    factor *= 2; */
+/*    constant *= 2; */
+
+/*    fprintf(stderr, "%s f:%f u:%f\n", __FUNCTION__, factor, constant); */
+
+       WARN_ONCE("ZBIAS registers locations might not be correct\n");
+
+       R200_STATECHANGE(rmesa, zbs);
+       rmesa->hw.zbs.cmd[R300_ZBS_FACTOR] = r300PackFloat32(factor);
+       rmesa->hw.zbs.cmd[R300_ZBS_CONSTANT] = r300PackFloat32(constant);
+}
+
+
+/* Routing and texture-related */
+
+void r300_setup_routing(GLcontext *ctx, GLboolean immediate)
+{
+       int i, count=0,reg=0;
+       GLuint dw, mask;
+       TNLcontext *tnl = TNL_CONTEXT(ctx);
+       struct vertex_buffer *VB = &tnl->vb;
+       r300ContextPtr r300 = R300_CONTEXT(ctx);
+
+
+       /* Stage 1 - input to VAP */
+
+       /* Assign register number automatically, retaining it in rmesa->state.reg */
+
+       /* Note: immediate vertex data includes all coordinates.
+       To save bandwidth use either VBUF or state-based vertex generation */
+
+       #define CONFIGURE_AOS(v, o, r, f) \
+               {\
+               if (RADEON_DEBUG & DEBUG_STATE)fprintf(stderr, "Enabling "#r "\n"); \
+               if(immediate){ \
+                       r300->state.aos[count].element_size=4; \
+                       r300->state.aos[count].stride=4; \
+                       r300->state.aos[count].ncomponents=4; \
+                       } else { \
+                       r300->state.aos[count].element_size=v->size; \
+                       r300->state.aos[count].stride=v->size; \
+                       r300->state.aos[count].ncomponents=v->size; \
+                       } \
+               r300->state.aos[count].offset=o; \
+               r300->state.aos[count].reg=reg; \
+               r300->state.aos[count].format=(f); \
+               r300->state.vap_reg.r=reg; \
+               count++; \
+               reg++; \
+               }
+
+               /* All offsets are 0 - for use by immediate mode.
+               Should change later to handle vertex buffers */
+       if(r300->current_vp){
+
+       /* VERT_ATTRIB_WEIGHT, VERT_ATTRIB_SIX, VERT_ATTRIB_SEVEN, VERT_ATTRIB_GENERIC0,
+          VERT_ATTRIB_GENERIC1, VERT_ATTRIB_GENERIC2, VERT_ATTRIB_GENERIC3 */
+       
+       if(r300->current_vp->inputs[VERT_ATTRIB_POS] != -1){
+               if(tnl->render_inputs & _TNL_BIT_POS){
+                       reg=r300->current_vp->inputs[VERT_ATTRIB_POS];
+                       CONFIGURE_AOS(VB->ObjPtr, 0, i_coords, AOS_FORMAT_FLOAT);
+               }else WARN_ONCE("vp expects pos but none was given\n");
+       }
+       if(r300->current_vp->inputs[VERT_ATTRIB_NORMAL] != -1){
+               if(tnl->render_inputs & _TNL_BIT_NORMAL){
+                       reg=r300->current_vp->inputs[VERT_ATTRIB_NORMAL];
+                       CONFIGURE_AOS(VB->NormalPtr, 0, i_normal, AOS_FORMAT_FLOAT);
+               }else WARN_ONCE("vp expects normal but none was given\n");
+       }
+       if(r300->current_vp->inputs[VERT_ATTRIB_COLOR0] != -1){
+               if(tnl->render_inputs & _TNL_BIT_COLOR0){
+                       reg=r300->current_vp->inputs[VERT_ATTRIB_COLOR0];
+                       CONFIGURE_AOS(VB->ColorPtr[0], 0, i_color[0], AOS_FORMAT_FLOAT_COLOR);
+               }else WARN_ONCE("vp expects primary color but none was given\n");
+       }
+       if(r300->current_vp->inputs[VERT_ATTRIB_COLOR1] != -1){
+               if(tnl->render_inputs & _TNL_BIT_COLOR1){
+                       reg=r300->current_vp->inputs[VERT_ATTRIB_COLOR1];
+                       CONFIGURE_AOS(VB->SecondaryColorPtr[0], 0, i_color[1], AOS_FORMAT_FLOAT_COLOR);
+               }else WARN_ONCE("vp expects secondary color but none was given\n");
+       }
+       if(r300->current_vp->inputs[VERT_ATTRIB_FOG] != -1){
+               if(tnl->render_inputs & _TNL_BIT_FOG){
+                       reg=r300->current_vp->inputs[VERT_ATTRIB_FOG];
+                       CONFIGURE_AOS(VB->FogCoordPtr, 0, i_fog, AOS_FORMAT_FLOAT);
+               }else WARN_ONCE("vp expects fog but none was given\n");
+       }
+       for(i=0;i < ctx->Const.MaxTextureUnits;i++) // tex 7 is last 
+               if(r300->current_vp->inputs[VERT_ATTRIB_TEX0+i] != -1){
+                       if(tnl->render_inputs & (_TNL_BIT_TEX0<<i)){
+                               reg=r300->current_vp->inputs[VERT_ATTRIB_TEX0+i];
+                               CONFIGURE_AOS(VB->TexCoordPtr[i], 0, i_tex[i], AOS_FORMAT_FLOAT);
+                       }else fprintf(stderr, "vp expects tex%d but none was given\n", i);
+               }
+#if 0
+       if((tnl->render_inputs & _TNL_BIT_INDEX))
+               CONFIGURE_AOS(VB->IndexPtr[0], 0, i_index, AOS_FORMAT_FLOAT);
+       
+       if((tnl->render_inputs & _TNL_BIT_POINTSIZE))
+               CONFIGURE_AOS(VB->PointSizePtr, 0, i_pointsize, AOS_FORMAT_FLOAT);
+#endif 
+       }else{
+       
+       if(tnl->render_inputs & _TNL_BIT_POS)
+               CONFIGURE_AOS(VB->ObjPtr, 0, i_coords, AOS_FORMAT_FLOAT);
+       if(tnl->render_inputs & _TNL_BIT_NORMAL)
+               CONFIGURE_AOS(VB->NormalPtr, 0, i_normal, AOS_FORMAT_FLOAT);
+       
+       if(tnl->render_inputs & _TNL_BIT_COLOR0)
+               CONFIGURE_AOS(VB->ColorPtr[0], 0, i_color[0], AOS_FORMAT_FLOAT_COLOR);
+       if(tnl->render_inputs & _TNL_BIT_COLOR1)
+               CONFIGURE_AOS(VB->SecondaryColorPtr[0], 0, i_color[1], AOS_FORMAT_FLOAT_COLOR);
+       
+       if(tnl->render_inputs & _TNL_BIT_FOG)
+               CONFIGURE_AOS(VB->FogCoordPtr, 0, i_fog, AOS_FORMAT_FLOAT);
+       
+       for(i=0;i < ctx->Const.MaxTextureUnits;i++)
+               if(tnl->render_inputs & (_TNL_BIT_TEX0<<i))
+                       CONFIGURE_AOS(VB->TexCoordPtr[i], 0, i_tex[i], AOS_FORMAT_FLOAT);
+       
+       if(tnl->render_inputs & _TNL_BIT_INDEX)
+               CONFIGURE_AOS(VB->IndexPtr[0], 0, i_index, AOS_FORMAT_FLOAT);
+       if(tnl->render_inputs & _TNL_BIT_POINTSIZE)
+               CONFIGURE_AOS(VB->PointSizePtr, 0, i_pointsize, AOS_FORMAT_FLOAT);
+       }
+       
+       r300->state.aos_count=count;
+
+       if (RADEON_DEBUG & DEBUG_STATE)
+               fprintf(stderr, "aos_count=%d\n", count);
+
+       if(count>R300_MAX_AOS_ARRAYS){
+               fprintf(stderr, "Aieee ! AOS array count exceeded !\n");
+               exit(-1);
+               }
+
+       /* Implement AOS */
+
+       /* setup INPUT_ROUTE */
+       R300_STATECHANGE(r300, vir[0]);
+       for(i=0;i+1<count;i+=2){
+               dw=(r300->state.aos[i].ncomponents-1)
+               | ((r300->state.aos[i].reg)<<8)
+               | (r300->state.aos[i].format<<14)
+               | (((r300->state.aos[i+1].ncomponents-1)
+               | ((r300->state.aos[i+1].reg)<<8)
+               | (r300->state.aos[i+1].format<<14))<<16);
+
+               if(i+2==count){
+                       dw|=(1<<(13+16));
+                       }
+               r300->hw.vir[0].cmd[R300_VIR_CNTL_0+(i>>1)]=dw;
+               }
+       if(count & 1){
+               dw=(r300->state.aos[count-1].ncomponents-1)
+               | (r300->state.aos[count-1].format<<14)
+               | ((r300->state.aos[count-1].reg)<<8)
+               | (1<<13);
+               r300->hw.vir[0].cmd[R300_VIR_CNTL_0+(count>>1)]=dw;
+               //fprintf(stderr, "vir0 dw=%08x\n", dw);
+               }
+       /* Set the rest of INPUT_ROUTE_0 to 0 */
+       //for(i=((count+1)>>1); i<8; i++)r300->hw.vir[0].cmd[R300_VIR_CNTL_0+i]=(0x0);
+       ((drm_r300_cmd_header_t*)r300->hw.vir[0].cmd)->unchecked_state.count = (count+1)>>1;
+
+
+       /* Mesa assumes that all missing components are from (0, 0, 0, 1) */
+       #define ALL_COMPONENTS ((R300_INPUT_ROUTE_SELECT_X<<R300_INPUT_ROUTE_X_SHIFT) \
+               | (R300_INPUT_ROUTE_SELECT_Y<<R300_INPUT_ROUTE_Y_SHIFT) \
+               | (R300_INPUT_ROUTE_SELECT_Z<<R300_INPUT_ROUTE_Z_SHIFT) \
+               | (R300_INPUT_ROUTE_SELECT_W<<R300_INPUT_ROUTE_W_SHIFT))
+
+       #define ALL_DEFAULT ((R300_INPUT_ROUTE_SELECT_ZERO<<R300_INPUT_ROUTE_X_SHIFT) \
+               | (R300_INPUT_ROUTE_SELECT_ZERO<<R300_INPUT_ROUTE_Y_SHIFT) \
+               | (R300_INPUT_ROUTE_SELECT_ZERO<<R300_INPUT_ROUTE_Z_SHIFT) \
+               | (R300_INPUT_ROUTE_SELECT_ONE<<R300_INPUT_ROUTE_W_SHIFT))
+
+       R300_STATECHANGE(r300, vir[1]);
+
+       for(i=0;i+1<count;i+=2){
+               /* do i first.. */
+               mask=(1<<(r300->state.aos[i].ncomponents*3))-1;
+               dw=(ALL_COMPONENTS & mask)
+               | (ALL_DEFAULT & ~mask)
+               | R300_INPUT_ROUTE_ENABLE;
+
+               /* i+1 */
+               mask=(1<<(r300->state.aos[i+1].ncomponents*3))-1;
+               dw|=(
+               (ALL_COMPONENTS & mask)
+               | (ALL_DEFAULT & ~mask)
+               | R300_INPUT_ROUTE_ENABLE
+               )<<16;
+
+               r300->hw.vir[1].cmd[R300_VIR_CNTL_0+(i>>1)]=dw;
+               }
+       if(count & 1){
+               mask=(1<<(r300->state.aos[count-1].ncomponents*3))-1;
+               dw=(ALL_COMPONENTS & mask)
+               | (ALL_DEFAULT & ~mask)
+               | R300_INPUT_ROUTE_ENABLE;
+               r300->hw.vir[1].cmd[R300_VIR_CNTL_0+(count>>1)]=dw;
+               //fprintf(stderr, "vir1 dw=%08x\n", dw);
+               }
+       /* Set the rest of INPUT_ROUTE_1 to 0 */
+       //for(i=((count+1)>>1); i<8; i++)r300->hw.vir[1].cmd[R300_VIR_CNTL_0+i]=0x0;
+       ((drm_r300_cmd_header_t*)r300->hw.vir[1].cmd)->unchecked_state.count = (count+1)>>1;
+
+       /* Set up input_cntl */
+
+       R300_STATECHANGE(r300, vic);
+       r300->hw.vic.cmd[R300_VIC_CNTL_0]=0x5555;  /* Hard coded value, no idea what it means */
+
+       r300->hw.vic.cmd[R300_VIC_CNTL_1]=R300_INPUT_CNTL_POS
+                                       | R300_INPUT_CNTL_COLOR;
+
+       for(i=0;i < ctx->Const.MaxTextureUnits;i++)
+               if(ctx->Texture.Unit[i].Enabled)
+                       r300->hw.vic.cmd[R300_VIC_CNTL_1]|=(R300_INPUT_CNTL_TC0<<i);
+
+       /* Stage 3: VAP output */
+       R300_STATECHANGE(r300, vof);
+       r300->hw.vof.cmd[R300_VOF_CNTL_0]=R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT
+                                       | R300_VAP_OUTPUT_VTX_FMT_0__COLOR_PRESENT;
+
+       r300->hw.vof.cmd[R300_VOF_CNTL_1]=0;
+       for(i=0;i < ctx->Const.MaxTextureUnits;i++)
+               if(ctx->Texture.Unit[i].Enabled)
+                       r300->hw.vof.cmd[R300_VOF_CNTL_1]|=(4<<(3*i));
+
+}
+
+static r300TexObj default_tex_obj={
+       filter:R300_TX_MAG_FILTER_LINEAR | R300_TX_MIN_FILTER_LINEAR,
+       pitch: 0x8000,
+       size: (0xff << R300_TX_WIDTHMASK_SHIFT)
+             | (0xff << R300_TX_HEIGHTMASK_SHIFT)
+             | (0x8 << R300_TX_SIZE_SHIFT),
+       format: 0x88a0c,
+       offset: 0x0,
+       unknown4: 0x0,
+       unknown5: 0x0
+       };
+
+       /* there is probably a system to these value, but, for now,
+          we just try by hand */
+
+static int inline translate_src(int src)
+{
+       switch (src) {
+       case GL_TEXTURE:
+               return 1;
+               break;
+       case GL_CONSTANT:
+               return 2;
+               break;
+       case GL_PRIMARY_COLOR:
+               return 3;
+               break;
+       case GL_PREVIOUS:
+               return 4;
+               break;
+       case GL_ZERO:
+               return 5;
+               break;
+       case GL_ONE:
+               return 6;
+               break;
+       default:
+               return 0;
+       }
+}
+
+/* r300 doesnt handle GL_CLAMP and GL_MIRROR_CLAMP_EXT correctly when filter is NEAREST.
+ * Since texwrap produces same results for GL_CLAMP and GL_CLAMP_TO_EDGE we use them instead.
+ * We need to recalculate wrap modes whenever filter mode is changed because someone might do:
+ * glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);
+ * glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP);
+ * glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
+ * Since r300 completely ignores R300_TX_CLAMP when either min or mag is nearest it cant handle
+ * combinations where only one of them is nearest.
+ */
+static unsigned long gen_fixed_filter(unsigned long f)
+{
+       unsigned long mag, min, needs_fixing=0;
+       //return f;
+       
+       /* We ignore MIRROR bit so we dont have to do everything twice */
+       if((f & ((7-1) << R300_TX_WRAP_S_SHIFT)) == (R300_TX_CLAMP << R300_TX_WRAP_S_SHIFT)){
+               needs_fixing |= 1;
+       }
+       if((f & ((7-1) << R300_TX_WRAP_T_SHIFT)) == (R300_TX_CLAMP << R300_TX_WRAP_T_SHIFT)){
+               needs_fixing |= 2;
+       }
+       if((f & ((7-1) << R300_TX_WRAP_Q_SHIFT)) == (R300_TX_CLAMP << R300_TX_WRAP_Q_SHIFT)){
+               needs_fixing |= 4;
+       }
+       
+       if(!needs_fixing)
+               return f;
+       
+       mag=f & R300_TX_MAG_FILTER_MASK;
+       min=f & R300_TX_MIN_FILTER_MASK;
+       
+       /* TODO: Check for anisto filters too */
+       if((mag != R300_TX_MAG_FILTER_NEAREST) && (min != R300_TX_MIN_FILTER_NEAREST))
+               return f;
+       
+       /* r300 cant handle these modes hence we force nearest to linear */
+       if((mag == R300_TX_MAG_FILTER_NEAREST) && (min != R300_TX_MIN_FILTER_NEAREST)){
+               f &= ~R300_TX_MAG_FILTER_NEAREST;
+               f |= R300_TX_MAG_FILTER_LINEAR;
+               return f;
+       }
+       
+       if((min == R300_TX_MIN_FILTER_NEAREST) && (mag != R300_TX_MAG_FILTER_NEAREST)){
+               f &= ~R300_TX_MIN_FILTER_NEAREST;
+               f |= R300_TX_MIN_FILTER_LINEAR;
+               return f;
+       }
+       
+       /* Both are nearest */
+       if(needs_fixing & 1){
+               f &= ~((7-1) << R300_TX_WRAP_S_SHIFT);
+               f |= R300_TX_CLAMP_TO_EDGE << R300_TX_WRAP_S_SHIFT;
+       }
+       if(needs_fixing & 2){
+               f &= ~((7-1) << R300_TX_WRAP_T_SHIFT);
+               f |= R300_TX_CLAMP_TO_EDGE << R300_TX_WRAP_T_SHIFT;
+       }
+       if(needs_fixing & 4){
+               f &= ~((7-1) << R300_TX_WRAP_Q_SHIFT);
+               f |= R300_TX_CLAMP_TO_EDGE << R300_TX_WRAP_Q_SHIFT;
+       }
+       return f;
+}
+
+void r300_setup_textures(GLcontext *ctx)
+{
+       int i, mtu;
+       struct r300_tex_obj *t;
+       r300ContextPtr r300 = R300_CONTEXT(ctx);
+       int max_texture_unit=-1; /* -1 translates into no setup costs for fields */
+       struct gl_texture_unit *texUnit;
+
+       R300_STATECHANGE(r300, txe);
+       R300_STATECHANGE(r300, tex.filter);
+       R300_STATECHANGE(r300, tex.unknown1);
+       R300_STATECHANGE(r300, tex.size);
+       R300_STATECHANGE(r300, tex.format);
+       R300_STATECHANGE(r300, tex.offset);
+       R300_STATECHANGE(r300, tex.unknown4);
+       R300_STATECHANGE(r300, tex.unknown5);
+       //R300_STATECHANGE(r300, tex.border_color);
+
+       r300->state.texture.tc_count=0;
+
+       r300->hw.txe.cmd[R300_TXE_ENABLE]=0x0;
+
+       mtu = r300->radeon.glCtx->Const.MaxTextureUnits;
+       if (RADEON_DEBUG & DEBUG_STATE)
+               fprintf(stderr, "mtu=%d\n", mtu);
+
+       if(mtu>R300_MAX_TEXTURE_UNITS){
+               fprintf(stderr, "Aiiee ! mtu=%d is greater than R300_MAX_TEXTURE_UNITS=%d\n",
+                       mtu, R300_MAX_TEXTURE_UNITS);
+               exit(-1);
+               }
+       for(i=0;i<mtu;i++){
+               if(ctx->Texture.Unit[i].Enabled){
+                       t=r300->state.texture.unit[i].texobj;
+                       //fprintf(stderr, "format=%08x\n", r300->state.texture.unit[i].format);
+                       r300->state.texture.tc_count++;
+                       if(t==NULL){
+                               fprintf(stderr, "Texture unit %d enabled, but corresponding texobj is NULL, using default object.\n", i);
+                               //exit(-1);
+                               t=&default_tex_obj;
+                               }
+                       //fprintf(stderr, "t->format=%08x\n", t->format);
+                       if (RADEON_DEBUG & DEBUG_STATE)
+                               fprintf(stderr, "Activating texture unit %d\n", i);
+                       max_texture_unit=i;
+                       r300->hw.txe.cmd[R300_TXE_ENABLE]|=(1<<i);
+
+                       r300->hw.tex.filter.cmd[R300_TEX_VALUE_0+i]=gen_fixed_filter(t->filter);
+                       /* No idea why linear filtered textures shake when puting random data */
+                       /*r300->hw.tex.unknown1.cmd[R300_TEX_VALUE_0+i]=(rand()%0xffffffff) & (~0x1fff);*/
+                       r300->hw.tex.size.cmd[R300_TEX_VALUE_0+i]=t->size;
+                       r300->hw.tex.format.cmd[R300_TEX_VALUE_0+i]=t->format;
+                       //fprintf(stderr, "t->format=%08x\n", t->format);
+                       r300->hw.tex.offset.cmd[R300_TEX_VALUE_0+i]=r300->radeon.radeonScreen->fbLocation+t->offset;
+                       r300->hw.tex.unknown4.cmd[R300_TEX_VALUE_0+i]=0x0;
+                       r300->hw.tex.unknown5.cmd[R300_TEX_VALUE_0+i]=0x0;
+                       //r300->hw.tex.border_color.cmd[R300_TEX_VALUE_0+i]=t->pp_border_color;
+                       }
+               }
+       ((drm_r300_cmd_header_t*)r300->hw.tex.filter.cmd)->unchecked_state.count = max_texture_unit+1;
+       ((drm_r300_cmd_header_t*)r300->hw.tex.unknown1.cmd)->unchecked_state.count = max_texture_unit+1;
+       ((drm_r300_cmd_header_t*)r300->hw.tex.size.cmd)->unchecked_state.count = max_texture_unit+1;
+       ((drm_r300_cmd_header_t*)r300->hw.tex.format.cmd)->unchecked_state.count = max_texture_unit+1;
+       ((drm_r300_cmd_header_t*)r300->hw.tex.offset.cmd)->unchecked_state.count = max_texture_unit+1;
+       ((drm_r300_cmd_header_t*)r300->hw.tex.unknown4.cmd)->unchecked_state.count = max_texture_unit+1;
+       ((drm_r300_cmd_header_t*)r300->hw.tex.unknown5.cmd)->unchecked_state.count = max_texture_unit+1;
+       //((drm_r300_cmd_header_t*)r300->hw.tex.border_color.cmd)->unchecked_state.count = max_texture_unit+1;
+
+       if (RADEON_DEBUG & DEBUG_STATE)
+               fprintf(stderr, "TX_ENABLE: %08x  max_texture_unit=%d\n", r300->hw.txe.cmd[R300_TXE_ENABLE], max_texture_unit);
+}
+
+void r300_setup_rs_unit(GLcontext *ctx)
+{
+       r300ContextPtr r300 = R300_CONTEXT(ctx);
+       int i;
+
+       /* This needs to be rewritten - it is a hack at best */
+
+       R300_STATECHANGE(r300, ri);
+       R300_STATECHANGE(r300, rc);
+       R300_STATECHANGE(r300, rr);
+
+       for(i = 1; i <= 8; ++i)
+               r300->hw.ri.cmd[i] = 0x00d10000;
+       r300->hw.ri.cmd[R300_RI_INTERP_1] |= R300_RS_INTERP_1_UNKNOWN;
+       r300->hw.ri.cmd[R300_RI_INTERP_2] |= R300_RS_INTERP_2_UNKNOWN;
+       r300->hw.ri.cmd[R300_RI_INTERP_3] |= R300_RS_INTERP_3_UNKNOWN;
+
+       #if 1
+       for(i = 2; i <= 8; ++i)
+               r300->hw.ri.cmd[i] |= 4;
+       #endif
+
+       for(i = 1; i <= 8; ++i)
+               r300->hw.rr.cmd[i] = 0;
+       /* textures enabled ? */
+       if(r300->state.texture.tc_count>0){
+
+               /* This code only really works with one set of texture coordinates */
+
+               /* The second constant is needed to get glxgears display anything .. */
+               r300->hw.rc.cmd[1] = R300_RS_CNTL_0_UNKNOWN_7
+                               | R300_RS_CNTL_0_UNKNOWN_18
+                               | (r300->state.texture.tc_count<<R300_RS_CNTL_TC_CNT_SHIFT);
+               r300->hw.rc.cmd[2] = 0xc0;
+
+
+               ((drm_r300_cmd_header_t*)r300->hw.rr.cmd)->unchecked_state.count = 1;
+               r300->hw.rr.cmd[R300_RR_ROUTE_0] = 0x24008;
+
+               } else {
+
+               /* The second constant is needed to get glxgears display anything .. */
+               r300->hw.rc.cmd[1] = R300_RS_CNTL_0_UNKNOWN_7 | R300_RS_CNTL_0_UNKNOWN_18;
+               r300->hw.rc.cmd[2] = 0;
+
+               ((drm_r300_cmd_header_t*)r300->hw.rr.cmd)->unchecked_state.count = 1;
+               r300->hw.rr.cmd[R300_RR_ROUTE_0] = 0x4000;
+
+               }
+}
+
+#define vpucount(ptr) (((drm_r300_cmd_header_t*)(ptr))->vpu.count)
+
+#define bump_vpu_count(ptr, new_count)   do{\
+       drm_r300_cmd_header_t* _p=((drm_r300_cmd_header_t*)(ptr));\
+       int _nc=(new_count)/4; \
+       if(_nc>_p->vpu.count)_p->vpu.count=_nc;\
+       }while(0)
+
+void static inline setup_vertex_shader_fragment(r300ContextPtr r300, int dest, struct r300_vertex_shader_fragment *vsf)
+{
+       int i;
+
+       if(vsf->length==0)return;
+
+       if(vsf->length & 0x3){
+               fprintf(stderr,"VERTEX_SHADER_FRAGMENT must have length divisible by 4\n");
+               exit(-1);
+               }
+
+       switch((dest>>8) & 0xf){
+       case 0:
+               R300_STATECHANGE(r300, vpi);
+               for(i=0;i<vsf->length;i++)
+                       r300->hw.vpi.cmd[R300_VPI_INSTR_0+i+4*(dest & 0xff)]=(vsf->body.d[i]);
+               bump_vpu_count(r300->hw.vpi.cmd, vsf->length+4*(dest & 0xff));
+               break;
+
+       case 2:
+               R300_STATECHANGE(r300, vpp);
+               for(i=0;i<vsf->length;i++)
+                       r300->hw.vpp.cmd[R300_VPP_PARAM_0+i+4*(dest & 0xff)]=(vsf->body.d[i]);
+               bump_vpu_count(r300->hw.vpp.cmd, vsf->length+4*(dest & 0xff));
+               break;
+       case 4:
+               R300_STATECHANGE(r300, vps);
+               for(i=0;i<vsf->length;i++)
+                       r300->hw.vps.cmd[1+i+4*(dest & 0xff)]=(vsf->body.d[i]);
+               bump_vpu_count(r300->hw.vps.cmd, vsf->length+4*(dest & 0xff));
+               break;
+       default:
+               fprintf(stderr, "%s:%s don't know how to handle dest %04x\n", __FILE__, __FUNCTION__, dest);
+               exit(-1);
+       }
+}
+
+void r300SetupVertexProgram(r300ContextPtr rmesa);
+
+void r300SetupVertexShader(r300ContextPtr rmesa)
+{
+       GLcontext* ctx = rmesa->radeon.glCtx;
+       
+       if(rmesa->current_vp){
+               r300SetupVertexProgram(rmesa);
+               return ;
+       }       
+       /* Reset state, in case we don't use something */
+       ((drm_r300_cmd_header_t*)rmesa->hw.vpp.cmd)->vpu.count = 0;
+       ((drm_r300_cmd_header_t*)rmesa->hw.vpi.cmd)->vpu.count = 0;
+       ((drm_r300_cmd_header_t*)rmesa->hw.vps.cmd)->vpu.count = 0;
+
+
+/* This needs to be replaced by vertex shader generation code */
+
+
+       /* textures enabled ? */
+       if(rmesa->state.texture.tc_count>0){
+               rmesa->state.vertex_shader=SINGLE_TEXTURE_VERTEX_SHADER;
+               } else {
+               rmesa->state.vertex_shader=FLAT_COLOR_VERTEX_SHADER;
+               }
+
+
+        rmesa->state.vertex_shader.matrix[0].length=16;
+        memcpy(rmesa->state.vertex_shader.matrix[0].body.f, ctx->_ModelProjectMatrix.m, 16*4);
+
+       setup_vertex_shader_fragment(rmesa, VSF_DEST_PROGRAM, &(rmesa->state.vertex_shader.program));
+
+       setup_vertex_shader_fragment(rmesa, VSF_DEST_MATRIX0, &(rmesa->state.vertex_shader.matrix[0]));
+       #if 0
+       setup_vertex_shader_fragment(rmesa, VSF_DEST_MATRIX1, &(rmesa->state.vertex_shader.matrix[0]));
+       setup_vertex_shader_fragment(rmesa, VSF_DEST_MATRIX2, &(rmesa->state.vertex_shader.matrix[0]));
+
+       setup_vertex_shader_fragment(rmesa, VSF_DEST_VECTOR0, &(rmesa->state.vertex_shader.vector[0]));
+       setup_vertex_shader_fragment(rmesa, VSF_DEST_VECTOR1, &(rmesa->state.vertex_shader.vector[1]));
+       #endif
+
+       #if 0
+       setup_vertex_shader_fragment(rmesa, VSF_DEST_UNKNOWN1, &(rmesa->state.vertex_shader.unknown1));
+       setup_vertex_shader_fragment(rmesa, VSF_DEST_UNKNOWN2, &(rmesa->state.vertex_shader.unknown2));
+       #endif
+
+       R300_STATECHANGE(rmesa, pvs);
+       rmesa->hw.pvs.cmd[R300_PVS_CNTL_1]=(rmesa->state.vertex_shader.program_start << R300_PVS_CNTL_1_PROGRAM_START_SHIFT)
+               | (rmesa->state.vertex_shader.unknown_ptr1 << R300_PVS_CNTL_1_UNKNOWN_SHIFT)
+               | (rmesa->state.vertex_shader.program_end << R300_PVS_CNTL_1_PROGRAM_END_SHIFT);
+       rmesa->hw.pvs.cmd[R300_PVS_CNTL_2]=(rmesa->state.vertex_shader.param_offset << R300_PVS_CNTL_2_PARAM_OFFSET_SHIFT)
+               | (rmesa->state.vertex_shader.param_count << R300_PVS_CNTL_2_PARAM_COUNT_SHIFT);
+       rmesa->hw.pvs.cmd[R300_PVS_CNTL_3]=(rmesa->state.vertex_shader.unknown_ptr2 << R300_PVS_CNTL_3_PROGRAM_UNKNOWN_SHIFT)
+       | (rmesa->state.vertex_shader.unknown_ptr3 << 0);
+
+       /* This is done for vertex shader fragments, but also needs to be done for vap_pvs,
+       so I leave it as a reminder */
+       #if 0
+       reg_start(R300_VAP_PVS_WAITIDLE,0);
+               e32(0x00000000);
+       #endif
+}
+
+void r300SetupVertexProgram(r300ContextPtr rmesa)
+{
+       GLcontext* ctx = rmesa->radeon.glCtx;
+       int inst_count;
+       int param_count;
+
+       /* Reset state, in case we don't use something */
+       ((drm_r300_cmd_header_t*)rmesa->hw.vpp.cmd)->vpu.count = 0;
+       ((drm_r300_cmd_header_t*)rmesa->hw.vpi.cmd)->vpu.count = 0;
+       ((drm_r300_cmd_header_t*)rmesa->hw.vps.cmd)->vpu.count = 0;
+
+       r300VertexProgUpdateParams(ctx, rmesa->current_vp);
+       
+       setup_vertex_shader_fragment(rmesa, VSF_DEST_PROGRAM, &(rmesa->current_vp->program));
+
+       setup_vertex_shader_fragment(rmesa, VSF_DEST_MATRIX0, &(rmesa->current_vp->params));
+       
+       #if 0
+       setup_vertex_shader_fragment(rmesa, VSF_DEST_UNKNOWN1, &(rmesa->state.vertex_shader.unknown1));
+       setup_vertex_shader_fragment(rmesa, VSF_DEST_UNKNOWN2, &(rmesa->state.vertex_shader.unknown2));
+       #endif
+       
+       inst_count=rmesa->current_vp->program.length/4 - 1;
+       param_count=rmesa->current_vp->params.length/4;
+                                       
+       R300_STATECHANGE(rmesa, pvs);
+       rmesa->hw.pvs.cmd[R300_PVS_CNTL_1]=(0 << R300_PVS_CNTL_1_PROGRAM_START_SHIFT)
+               | (inst_count/*0*/ << R300_PVS_CNTL_1_UNKNOWN_SHIFT)
+               | (inst_count << R300_PVS_CNTL_1_PROGRAM_END_SHIFT);
+       rmesa->hw.pvs.cmd[R300_PVS_CNTL_2]=(0 << R300_PVS_CNTL_2_PARAM_OFFSET_SHIFT)
+               | (param_count << R300_PVS_CNTL_2_PARAM_COUNT_SHIFT);
+       rmesa->hw.pvs.cmd[R300_PVS_CNTL_3]=(0/*rmesa->state.vertex_shader.unknown_ptr2*/ << R300_PVS_CNTL_3_PROGRAM_UNKNOWN_SHIFT)
+       | ((inst_count-rmesa->current_vp->t2rs) /*rmesa->state.vertex_shader.unknown_ptr3*/ << 0);
+
+       /* This is done for vertex shader fragments, but also needs to be done for vap_pvs,
+       so I leave it as a reminder */
+       #if 0
+       reg_start(R300_VAP_PVS_WAITIDLE,0);
+               e32(0x00000000);
+       #endif
+}
+
+void r300SetupPixelShader(r300ContextPtr rmesa)
+{
+int i,k;
+
+       /* This needs to be replaced by pixel shader generation code */
+
+       /* textures enabled ? */
+       if(rmesa->state.texture.tc_count>0){
+               rmesa->state.pixel_shader=SINGLE_TEXTURE_PIXEL_SHADER;
+               } else {
+               rmesa->state.pixel_shader=FLAT_COLOR_PIXEL_SHADER;
+               }
+
+       R300_STATECHANGE(rmesa, fpt);
+       for(i=0;i<rmesa->state.pixel_shader.program.tex.length;i++)
+               rmesa->hw.fpt.cmd[R300_FPT_INSTR_0+i]=rmesa->state.pixel_shader.program.tex.inst[i];
+       rmesa->hw.fpt.cmd[R300_FPT_CMD_0]=cmducs(R300_PFS_TEXI_0, rmesa->state.pixel_shader.program.tex.length);
+
+       #define OUTPUT_FIELD(st, reg, field)  \
+               R300_STATECHANGE(rmesa, st); \
+               for(i=0;i<rmesa->state.pixel_shader.program.alu.length;i++) \
+                       rmesa->hw.st.cmd[R300_FPI_INSTR_0+i]=rmesa->state.pixel_shader.program.alu.inst[i].field;\
+               rmesa->hw.st.cmd[R300_FPI_CMD_0]=cmducs(reg, rmesa->state.pixel_shader.program.alu.length);
+
+       OUTPUT_FIELD(fpi[0], R300_PFS_INSTR0_0, inst0);
+       OUTPUT_FIELD(fpi[1], R300_PFS_INSTR1_0, inst1);
+       OUTPUT_FIELD(fpi[2], R300_PFS_INSTR2_0, inst2);
+       OUTPUT_FIELD(fpi[3], R300_PFS_INSTR3_0, inst3);
+       #undef OUTPUT_FIELD
+
+       R300_STATECHANGE(rmesa, fp);
+       for(i=0;i<4;i++){
+               rmesa->hw.fp.cmd[R300_FP_NODE0+i]=
+               (rmesa->state.pixel_shader.program.node[i].alu_offset << R300_PFS_NODE_ALU_OFFSET_SHIFT)
+               | (rmesa->state.pixel_shader.program.node[i].alu_end  << R300_PFS_NODE_ALU_END_SHIFT)
+               | (rmesa->state.pixel_shader.program.node[i].tex_offset << R300_PFS_NODE_TEX_OFFSET_SHIFT)
+               | (rmesa->state.pixel_shader.program.node[i].tex_end  << R300_PFS_NODE_TEX_END_SHIFT)
+               | ( (i==3) ? R300_PFS_NODE_LAST_NODE : 0);
+               }
+
+               /*  PFS_CNTL_0 */
+       rmesa->hw.fp.cmd[R300_FP_CNTL0]=
+               (rmesa->state.pixel_shader.program.active_nodes-1)
+               | (rmesa->state.pixel_shader.program.first_node_has_tex<<3);
+               /* PFS_CNTL_1 */
+       rmesa->hw.fp.cmd[R300_FP_CNTL1]=rmesa->state.pixel_shader.program.temp_register_count;
+               /* PFS_CNTL_2 */
+       rmesa->hw.fp.cmd[R300_FP_CNTL2]=
+               (rmesa->state.pixel_shader.program.alu_offset << R300_PFS_CNTL_ALU_OFFSET_SHIFT)
+               | (rmesa->state.pixel_shader.program.alu_end << R300_PFS_CNTL_ALU_END_SHIFT)
+               | (rmesa->state.pixel_shader.program.tex_offset << R300_PFS_CNTL_TEX_OFFSET_SHIFT)
+               | (rmesa->state.pixel_shader.program.tex_end << R300_PFS_CNTL_TEX_END_SHIFT);
+
+       R300_STATECHANGE(rmesa, fpp);
+       for(i=0;i<rmesa->state.pixel_shader.param_length;i++){
+               rmesa->hw.fpp.cmd[R300_FPP_PARAM_0+4*i+0]=r300PackFloat32(rmesa->state.pixel_shader.param[i].x);
+               rmesa->hw.fpp.cmd[R300_FPP_PARAM_0+4*i+1]=r300PackFloat32(rmesa->state.pixel_shader.param[i].y);
+               rmesa->hw.fpp.cmd[R300_FPP_PARAM_0+4*i+2]=r300PackFloat32(rmesa->state.pixel_shader.param[i].z);
+               rmesa->hw.fpp.cmd[R300_FPP_PARAM_0+4*i+3]=r300PackFloat32(rmesa->state.pixel_shader.param[i].w);
+               }
+       rmesa->hw.fpp.cmd[R300_FPP_CMD_0]=cmducs(R300_PFS_PARAM_0_X, rmesa->state.pixel_shader.param_length);
+
+}
+
+/**
  * Called by Mesa after an internal state update.
  */
 static void r300InvalidateState(GLcontext * ctx, GLuint new_state)
@@ -99,45 +1514,69 @@ static void r300InvalidateState(GLcontext * ctx, GLuint new_state)
  */
 void r300ResetHwState(r300ContextPtr r300)
 {
+       GLcontext* ctx = r300->radeon.glCtx;
        int i;
 
        if (RADEON_DEBUG & DEBUG_STATE)
                fprintf(stderr, "%s\n", __FUNCTION__);
 
-       {
-               __DRIdrawablePrivate *dPriv = r300->radeon.dri.drawable;
-               GLfloat xoffset = dPriv ? (GLfloat) dPriv->x : 0;
-               GLfloat yoffset = dPriv ? (GLfloat) dPriv->y + dPriv->h : 0;
-               const GLfloat *v = r300->radeon.glCtx->Viewport._WindowMap.m;
-
-               r300->hw.vpt.cmd[R300_VPT_XSCALE] =
-                       r300PackFloat32(v[MAT_SX]);
-               r300->hw.vpt.cmd[R300_VPT_XOFFSET] =
-                       r300PackFloat32(v[MAT_TX] + xoffset);
-               r300->hw.vpt.cmd[R300_VPT_YSCALE] =
-                       r300PackFloat32(-v[MAT_SY]);
-               r300->hw.vpt.cmd[R300_VPT_YOFFSET] =
-                       r300PackFloat32(-v[MAT_TY] + yoffset);
-               r300->hw.vpt.cmd[R300_VPT_YSCALE] =
-                       r300PackFloat32(v[MAT_SZ]);
-               r300->hw.vpt.cmd[R300_VPT_YOFFSET] =
-                       r300PackFloat32(v[MAT_TZ]);
-       }
+               /* This is a place to initialize registers which
+                  have bitfields accessed by different functions
+                  and not all bits are used */
+       #if 0
+       r300->hw.zs.cmd[R300_ZS_CNTL_0] = 0;
+       r300->hw.zs.cmd[R300_ZS_CNTL_1] = 0;
+       r300->hw.zs.cmd[R300_ZS_CNTL_2] = 0xffff00;
+       #endif
 
-//BEGIN: TODO
-       r300->hw.unk2080.cmd[1] = 0x0030045A;
+               /* go and compute register values from GL state */
+
+       r300UpdateWindow(ctx);
+
+       r300ColorMask(ctx,
+               ctx->Color.ColorMask[RCOMP],
+               ctx->Color.ColorMask[GCOMP],
+               ctx->Color.ColorMask[BCOMP],
+               ctx->Color.ColorMask[ACOMP]);
+
+       r300Enable(ctx, GL_DEPTH_TEST, ctx->Depth.Test);
+       r300DepthMask(ctx, ctx->Depth.Mask);
+       r300DepthFunc(ctx, ctx->Depth.Func);
 
-       r300->hw.ovf.cmd[R300_OVF_FMT_0] = 0x00000003;
-       r300->hw.ovf.cmd[R300_OVF_FMT_1] = 0x00000000;
+       r300UpdateCulling(ctx);
 
-       r300->hw.unk20B0.cmd[1] = 0x0000040A;
-       r300->hw.unk20B0.cmd[2] = 0x00000008;
+       r300_setup_routing(ctx, GL_TRUE);
+
+       r300UpdateTextureState(ctx);
+       r300_setup_textures(ctx);
+       r300_setup_rs_unit(ctx);
+
+       r300SetupVertexShader(r300);
+       r300SetupPixelShader(r300);
+
+       r300_set_blend_state(ctx);
+       r300AlphaFunc(ctx, ctx->Color.AlphaFunc, ctx->Color.AlphaRef);
+
+               /* Initialize magic registers
+                TODO : learn what they really do, or get rid of
+                those we don't have to touch */
+       r300->hw.unk2080.cmd[1] = 0x0030045A;
+
+       r300->hw.vte.cmd[1] = R300_VPORT_X_SCALE_ENA
+                               | R300_VPORT_X_OFFSET_ENA
+                               | R300_VPORT_Y_SCALE_ENA
+                               | R300_VPORT_Y_OFFSET_ENA
+                               | R300_VPORT_Z_SCALE_ENA
+                               | R300_VPORT_Z_OFFSET_ENA
+                               | R300_VTX_W0_FMT;
+       r300->hw.vte.cmd[2] = 0x00000008;
 
        r300->hw.unk2134.cmd[1] = 0x00FFFFFF;
        r300->hw.unk2134.cmd[2] = 0x00000000;
 
        r300->hw.unk2140.cmd[1] = 0x00000000;
 
+       #if 0 /* Done in setup routing */
        ((drm_r300_cmd_header_t*)r300->hw.vir[0].cmd)->unchecked_state.count = 1;
        r300->hw.vir[0].cmd[1] = 0x21030003;
 
@@ -146,6 +1585,7 @@ void r300ResetHwState(r300ContextPtr r300)
 
        r300->hw.vic.cmd[R300_VIR_CNTL_0] = 0x00000001;
        r300->hw.vic.cmd[R300_VIR_CNTL_1] = 0x00000405;
+       #endif
 
        r300->hw.unk21DC.cmd[1] = 0xAAAAAAAA;
 
@@ -161,22 +1601,35 @@ void r300ResetHwState(r300ContextPtr r300)
        else
                r300->hw.unk2288.cmd[1] = R300_2288_RV350;
 
+       #if 0
+       r300->hw.vof.cmd[R300_VOF_CNTL_0] = R300_VAP_OUTPUT_VTX_FMT_0__POS_PRESENT
+                               | R300_VAP_OUTPUT_VTX_FMT_0__COLOR_PRESENT;
+       r300->hw.vof.cmd[R300_VOF_CNTL_1] = 0; /* no textures */
+
+
        r300->hw.pvs.cmd[R300_PVS_CNTL_1] = 0;
        r300->hw.pvs.cmd[R300_PVS_CNTL_2] = 0;
        r300->hw.pvs.cmd[R300_PVS_CNTL_3] = 0;
+       #endif
 
-       r300->hw.unk4008.cmd[1] = 0x00000007;
+       r300->hw.gb_enable.cmd[1] = R300_GB_POINT_STUFF_ENABLE
+               | R300_GB_LINE_STUFF_ENABLE
+               | R300_GB_TRIANGLE_STUFF_ENABLE;
 
-       r300->hw.unk4010.cmd[1] = 0x66666666;
-       r300->hw.unk4010.cmd[2] = 0x06666666;
+       r300->hw.gb_misc.cmd[R300_GB_MISC_MSPOS_0] = 0x66666666;
+       r300->hw.gb_misc.cmd[R300_GB_MISC_MSPOS_1] = 0x06666666;
        if (GET_CHIP(r300->radeon.radeonScreen) == RADEON_CHIP_R300)
-               r300->hw.unk4010.cmd[3] = 0x00000017;
+               r300->hw.gb_misc.cmd[R300_GB_MISC_TILE_CONFIG] = R300_GB_TILE_ENABLE
+                                                       | R300_GB_TILE_PIPE_COUNT_R300
+                                                       | R300_GB_TILE_SIZE_16;
        else
-               r300->hw.unk4010.cmd[3] = 0x00000011;
-       r300->hw.unk4010.cmd[4] = 0x00000000;
-       r300->hw.unk4010.cmd[5] = 0x00000000;
+               r300->hw.gb_misc.cmd[R300_GB_MISC_TILE_CONFIG] = R300_GB_TILE_ENABLE
+                                                       | R300_GB_TILE_PIPE_COUNT_RV300
+                                                       | R300_GB_TILE_SIZE_16;
+       r300->hw.gb_misc.cmd[R300_GB_MISC_SELECT] = 0x00000000;
+       r300->hw.gb_misc.cmd[R300_GB_MISC_AA_CONFIG] = 0x00000000; /* No antialiasing */
 
-       r300->hw.txe.cmd[R300_TXE_ENABLE] = 0;
+       //r300->hw.txe.cmd[R300_TXE_ENABLE] = 0;
 
        r300->hw.unk4200.cmd[1] = r300PackFloat32(0.0);
        r300->hw.unk4200.cmd[2] = r300PackFloat32(0.0);
@@ -209,27 +1662,20 @@ void r300ResetHwState(r300ContextPtr r300)
 
        r300->hw.unk42A0.cmd[1] = 0x00000000;
 
+       #if 0
        r300->hw.unk42B4.cmd[1] = 0x00000000;
-       r300->hw.unk42B4.cmd[2] = 0x00000000;
-
+       #endif
+       
        r300->hw.unk42C0.cmd[1] = 0x4B7FFFFF;
        r300->hw.unk42C0.cmd[2] = 0x00000000;
 
-       r300->hw.rc.cmd[1] = R300_RS_CNTL_0_UNKNOWN_7;
-       r300->hw.rc.cmd[2] = 0;
-
-       for(i = 1; i <= 8; ++i)
-               r300->hw.ri.cmd[i] = 0;
-
-       ((drm_r300_cmd_header_t*)r300->hw.rr.cmd)->unchecked_state.count = 1;
-       for(i = 1; i <= 8; ++i)
-               r300->hw.rr.cmd[1] = 0;
 
        r300->hw.unk43A4.cmd[1] = 0x0000001C;
        r300->hw.unk43A4.cmd[2] = 0x2DA49525;
 
        r300->hw.unk43E8.cmd[1] = 0x00FFFFFF;
 
+       #if 0
        r300->hw.fp.cmd[R300_FP_CNTL0] = 0;
        r300->hw.fp.cmd[R300_FP_CNTL1] = 0;
        r300->hw.fp.cmd[R300_FP_CNTL2] = 0;
@@ -237,6 +1683,7 @@ void r300ResetHwState(r300ContextPtr r300)
        r300->hw.fp.cmd[R300_FP_NODE1] = 0;
        r300->hw.fp.cmd[R300_FP_NODE2] = 0;
        r300->hw.fp.cmd[R300_FP_NODE3] = 0;
+       #endif
 
        r300->hw.unk46A4.cmd[1] = 0x00001B01;
        r300->hw.unk46A4.cmd[2] = 0x00001B0F;
@@ -244,6 +1691,7 @@ void r300ResetHwState(r300ContextPtr r300)
        r300->hw.unk46A4.cmd[4] = 0x00001B0F;
        r300->hw.unk46A4.cmd[5] = 0x00000001;
 
+       #if 0
        for(i = 1; i <= 64; ++i) {
                /* create NOP instructions */
                r300->hw.fpi[0].cmd[i] = FP_INSTRC(MAD, FP_ARGC(SRC0C_XYZ), FP_ARGC(ONE), FP_ARGC(ZERO));
@@ -251,6 +1699,7 @@ void r300ResetHwState(r300ContextPtr r300)
                r300->hw.fpi[2].cmd[i] = FP_INSTRA(MAD, FP_ARGA(SRC0A), FP_ARGA(ONE), FP_ARGA(ZERO));
                r300->hw.fpi[3].cmd[i] = FP_SELA(0,W,NO,FP_TMP(0),0,0);
        }
+       #endif
 
        r300->hw.unk4BC0.cmd[1] = 0;
 
@@ -258,16 +1707,19 @@ void r300ResetHwState(r300ContextPtr r300)
        r300->hw.unk4BC8.cmd[2] = 0;
        r300->hw.unk4BC8.cmd[3] = 0;
 
+       #if 0
        r300->hw.at.cmd[R300_AT_ALPHA_TEST] = 0;
+       #endif
 
+       r300->hw.at.cmd[R300_AT_UNKNOWN] = 0;
        r300->hw.unk4BD8.cmd[1] = 0;
 
        r300->hw.unk4E00.cmd[1] = 0;
 
+       #if 0
        r300->hw.bld.cmd[R300_BLD_CBLEND] = 0;
        r300->hw.bld.cmd[R300_BLD_ABLEND] = 0;
-
-       r300->hw.cmk.cmd[R300_CMK_COLORMASK] = 0xF;
+       #endif
 
        r300->hw.unk4E10.cmd[1] = 0;
        r300->hw.unk4E10.cmd[2] = 0;
@@ -291,16 +1743,17 @@ void r300ResetHwState(r300ContextPtr r300)
 
        r300->hw.unk4E88.cmd[1] = 0;
 
-       r300->hw.zc.cmd[R300_ZC_CNTL_0] = 0;
-       r300->hw.zc.cmd[R300_ZC_CNTL_1] = 0;
-
-       r300->hw.unk4F08.cmd[1] = 0x00FFFF00;
+       r300->hw.unk4EA0.cmd[1] = 0x00000000;
+       r300->hw.unk4EA0.cmd[2] = 0xffffffff;
 
        r300->hw.unk4F10.cmd[1] = 0x00000002; // depthbuffer format?
        r300->hw.unk4F10.cmd[2] = 0x00000000;
        r300->hw.unk4F10.cmd[3] = 0x00000003;
        r300->hw.unk4F10.cmd[4] = 0x00000000;
 
+       /* experiment a bit */
+       r300->hw.unk4F10.cmd[2] = 0x00000001; // depthbuffer format?
+
        r300->hw.zb.cmd[R300_ZB_OFFSET] =
                r300->radeon.radeonScreen->depthOffset +
                r300->radeon.radeonScreen->fbLocation;
@@ -315,6 +1768,7 @@ void r300ResetHwState(r300ContextPtr r300)
 
        r300->hw.unk4F54.cmd[1] = 0;
 
+       #if 0
        ((drm_r300_cmd_header_t*)r300->hw.vpi.cmd)->vpu.count = 0;
        for(i = 1; i < R300_VPI_CMDSIZE; i += 4) {
                /* MOV t0, t0 */
@@ -327,11 +1781,13 @@ void r300ResetHwState(r300ContextPtr r300)
        ((drm_r300_cmd_header_t*)r300->hw.vpp.cmd)->vpu.count = 0;
        for(i = 1; i < R300_VPP_CMDSIZE; ++i)
                r300->hw.vpp.cmd[i] = 0;
+       #endif
 
        r300->hw.vps.cmd[R300_VPS_ZERO_0] = 0;
        r300->hw.vps.cmd[R300_VPS_ZERO_1] = 0;
        r300->hw.vps.cmd[R300_VPS_POINTSIZE] = r300PackFloat32(1.0);
        r300->hw.vps.cmd[R300_VPS_ZERO_3] = 0;
+
 //END: TODO
 
        r300->hw.all_dirty = GL_TRUE;
@@ -346,8 +1802,34 @@ void r300ResetHwState(r300ContextPtr r300)
  */
 void r300InitState(r300ContextPtr r300)
 {
+       GLcontext *ctx = r300->radeon.glCtx;
+       GLuint depth_fmt;
+
        radeonInitState(&r300->radeon);
 
+       switch (ctx->Visual.depthBits) {
+       case 16:
+               r300->state.depth.scale = 1.0 / (GLfloat) 0xffff;
+               depth_fmt = R200_DEPTH_FORMAT_16BIT_INT_Z;
+               //r300->state.stencil.clear = 0x00000000;
+               break;
+       case 24:
+               r300->state.depth.scale = 1.0 / (GLfloat) 0xffffff;
+               depth_fmt = R200_DEPTH_FORMAT_24BIT_INT_Z;
+               //r300->state.stencil.clear = 0xff000000;
+               break;
+       default:
+               fprintf(stderr, "Error: Unsupported depth %d... exiting\n",
+                       ctx->Visual.depthBits);
+               exit(-1);
+       }
+
+       /* Only have hw stencil when depth buffer is 24 bits deep */
+       r300->state.hw_stencil = (ctx->Visual.stencilBits > 0 &&
+                                        ctx->Visual.depthBits == 24);
+
+       memset(&(r300->state.texture), 0, sizeof(r300->state.texture));
+
        r300ResetHwState(r300);
 }
 
@@ -360,6 +1842,28 @@ void r300InitStateFuncs(struct dd_function_table* functions)
        radeonInitStateFuncs(functions);
 
        functions->UpdateState = r300InvalidateState;
-       functions->Enable= r300Enable;
+       functions->AlphaFunc = r300AlphaFunc;
+       functions->BlendColor = r300BlendColor;
+       functions->BlendEquationSeparate = r300BlendEquationSeparate;
+       functions->BlendFuncSeparate = r300BlendFuncSeparate;
+       functions->Enable = r300Enable;
+       functions->ColorMask = r300ColorMask;
+       functions->DepthFunc = r300DepthFunc;
+       functions->DepthMask = r300DepthMask;
+       functions->CullFace = r300CullFace;
+       functions->FrontFace = r300FrontFace;
+
+       /* Stencil related */
+       functions->ClearStencil = r300ClearStencil;
+       functions->StencilFunc = r300StencilFunc;
+       functions->StencilMask = r300StencilMask;
+       functions->StencilOp = r300StencilOp;
+
+       /* Viewport related */
+       functions->Viewport = r300Viewport;
+       functions->DepthRange = r300DepthRange;
+       functions->PointSize = r300PointSize;
+       
+       
+       functions->PolygonOffset = r300PolygonOffset;
 }
-