ARB prog parser: Compile lexer as C++

This is in anticipation of generating GLSL IR from the parser.

The C++ rules for enums vs. ints are just plain broken.

Makefile

@@ -5,7 +5,10 @@ TOP = .
 SUBDIRS = src
 
 
+# The git command below generates an empty string when we're not
+# building in a GIT tree (i.e., building from a release tarball).
 default: $(TOP)/configs/current
+	@$(TOP)/bin/extract_git_sha1
 	@for dir in $(SUBDIRS) ; do \
 		if [ -d $$dir ] ; then \
 			(cd $$dir && $(MAKE)) || exit 1 ; \
@@ -199,6 +202,7 @@ MAIN_FILES = \
 	$(DIRECTORY)/aclocal.m4						\
 	$(DIRECTORY)/bin/config.guess					\
 	$(DIRECTORY)/bin/config.sub					\
+	$(DIRECTORY)/bin/extract_git_sha1				\
 	$(DIRECTORY)/bin/install-sh					\
 	$(DIRECTORY)/bin/mklib						\
 	$(DIRECTORY)/bin/minstall					\

bin/extract_git_sha1

@@ -0,0 +1,10 @@
+#!/bin/sh
+touch src/mesa/main/git_sha1.h
+if which git > /dev/null; then
+    # Extract the 7-digit "short" SHA1 for the current HEAD, convert
+    # it to a string, and wrap it in a #define.  This is used in
+    # src/mesa/main/version.c to put the GIT SHA1 in the GL_VERSION string.
+    git log -n 1 --oneline |\
+	sed 's/^\([^ ]*\) .*/#define MESA_GIT_SHA1 "git-\1"/' \
+	> src/mesa/main/git_sha1.h
+fi

configs/linux-dri

@@ -63,7 +63,7 @@ GALLIUM_WINSYS_DIRS = sw sw/xlib drm/vmware drm/intel drm/i965
 GALLIUM_TARGET_DIRS = 
 GALLIUM_STATE_TRACKERS_DIRS = egl
 
-DRI_DIRS = i810 i915 i965 mach64 mga r128 r200 r300 radeon \
+DRI_DIRS = i810 i915 i965 mach64 mga r128 r200 r300 r600 radeon \
 	savage sis tdfx unichrome swrast
 
 INTEL_LIBS = `pkg-config --libs libdrm_intel`

configure.ac

@@ -729,6 +729,27 @@ else
     enable_xcb=no
 fi
 
+dnl Direct rendering or just indirect rendering
+case "$host_os" in
+gnu*)
+    dnl Disable by default on GNU/Hurd
+    driglx_direct_default="no"
+    ;;
+cygwin*)
+    dnl Disable by default on cygwin
+    driglx_direct_default="no"
+    ;;
+*)
+    driglx_direct_default="yes"
+    ;;
+esac
+AC_ARG_ENABLE([driglx-direct],
+    [AS_HELP_STRING([--disable-driglx-direct],
+        [enable direct rendering in GLX and EGL for DRI \
+            @<:@default=auto@:>@])],
+    [driglx_direct="$enableval"],
+    [driglx_direct="$driglx_direct_default"])
+
 dnl
 dnl libGL configuration per driver
 dnl
@@ -762,12 +783,17 @@ dri|no) # these checks are still desired when there is no mesa_driver
         AC_MSG_ERROR([Can't use static libraries for DRI drivers])
     fi
 
-    # Check for libdrm
-    PKG_CHECK_MODULES([LIBDRM], [libdrm >= $LIBDRM_REQUIRED])
-    PKG_CHECK_MODULES([DRI2PROTO], [dri2proto >= $DRI2PROTO_REQUIRED])
     PKG_CHECK_MODULES([GLPROTO], [glproto >= $GLPROTO_REQUIRED])
-    GL_PC_REQ_PRIV="libdrm >= $LIBDRM_REQUIRED dri2proto >= $DRI2PROTO_REQUIRED glproto >= $GLPROTO_REQUIRED"
-    DRI_PC_REQ_PRIV="libdrm >= $LIBDRM_REQUIRED"
+    GL_PC_REQ_PRIV="glproto >= $GLPROTO_REQUIRED"
+    DRI_PC_REQ_PRIV=""
+
+    if test x"$driglx_direct" = xyes; then
+        # Check for libdrm
+        PKG_CHECK_MODULES([LIBDRM], [libdrm >= $LIBDRM_REQUIRED])
+        PKG_CHECK_MODULES([DRI2PROTO], [dri2proto >= $DRI2PROTO_REQUIRED])
+        GL_PC_REQ_PRIV="$GL_PC_REQ_PRIV libdrm >= $LIBDRM_REQUIRED dri2proto >= $DRI2PROTO_REQUIRED"
+        DRI_PC_REQ_PRIV="libdrm >= $LIBDRM_REQUIRED"
+    fi
 
     # find the DRI deps for libGL
     if test "$x11_pkgconfig" = yes; then
@@ -910,12 +936,6 @@ AC_ARG_WITH([dri-searchpath],
     [DRI_DRIVER_SEARCH_DIR="$withval"],
     [DRI_DRIVER_SEARCH_DIR='${DRI_DRIVER_INSTALL_DIR}'])
 AC_SUBST([DRI_DRIVER_SEARCH_DIR])
-dnl Direct rendering or just indirect rendering
-AC_ARG_ENABLE([driglx-direct],
-    [AS_HELP_STRING([--disable-driglx-direct],
-        [enable direct rendering in GLX and EGL for DRI @<:@default=enabled@:>@])],
-    [driglx_direct="$enableval"],
-    [driglx_direct="yes"])
 dnl Which drivers to build - default is chosen by platform
 AC_ARG_WITH([dri-drivers],
     [AS_HELP_STRING([--with-dri-drivers@<:@=DIRS...@:>@],
@@ -1006,6 +1026,16 @@ if test "$mesa_driver" = dri -o "$mesa_driver" = no; then
             DEFINES="$DEFINES -DGLX_DIRECT_RENDERING"
         fi
         ;;
+    cygwin*)
+        DEFINES="$DEFINES -DUSE_EXTERNAL_DXTN_LIB=1 -DIN_DRI_DRIVER"
+        DEFINES="$DEFINES -DGLX_INDIRECT_RENDERING"
+        if test "x$driglx_direct" = xyes; then
+            DEFINES="$DEFINES -DGLX_DIRECT_RENDERING"
+        fi
+        if test "x$DRI_DIRS" = "xyes"; then
+            DRI_DIRS="swrast"
+        fi
+        ;;
     esac
 
     # default drivers

docs/ARB_color_buffer_float.txt

@@ -0,0 +1,4 @@
+Known issues in the ARB_color_buffer_float implementation:
+- Rendering to multiple render targets, some fixed-point, some floating-point, with FIXED_ONLY fragment clamping and polygon smooth enabled may write incorrect values to the fixed point buffers (depends on spec interpretation)
+- For fragment programs with ARB_fog_* options, colors are clamped before fog application regardless of the fragment clamping setting (this depends on spec interpretation)
+

docs/GL3.txt

@@ -15,7 +15,7 @@ GL 3.0:
 GLSL changes (GL_EXT_gpu_shader4, etc)                not started
 Conditional rendering (GL_NV_conditional_render)      DONE (swrast & softpipe)
 Map buffer subranges (GL_ARB_map_buffer_range)        DONE
-Clamping controls (GL_ARB_color_buffer_float)         BRANCH ~mareko/mesa floating2
+Clamping controls (GL_ARB_color_buffer_float)         DONE
 Float textures, renderbuffers (GL_ARB_texture_float)  BRANCH ~mareko/mesa floating2
 GL_EXT_packed_float                                   not started
 GL_EXT_texture_shared_exponent                        not started
@@ -53,7 +53,7 @@ Primitive restart (GL_NV_primitive_restart)           DONE (gallium)
 Texture buffer objs (GL_ARB_texture_buffer_object)    not started
 Rectangular textures (GL_ARB_texture_rectangle)       DONE
 Uniform buffer objs (GL_ARB_uniform_buffer_object)    not started
-Signed normalized textures (GL_EXT_texture_snorm)     ~50% done
+Signed normalized textures (GL_EXT_texture_snorm)     DONE (gallium)
 
 
 GL 3.2:
@@ -77,7 +77,7 @@ GLSL 3.30                                             not started
 GL_ARB_blend_func_extended                            not started
 GL_ARB_explicit_attrib_location                       DONE (swrast, i915, i965)
 GL_ARB_occlusion_query2                               DONE (swrast, gallium)
-GL_ARB_sampler_objects                                not started
+GL_ARB_sampler_objects                                DONE (gallium)
 GL_ARB_texture_rgb10_a2ui                             not started
 GL_ARB_texture_swizzle                                DONE (same as EXT version)
 GL_ARB_timer_query                                    DONE (only Xlib sw driver)

docs/MESA_agp_offset.spec

@@ -67,7 +67,7 @@ Additions to Chapter 3 the GLX 1.4 Specification (Functions and Errors)
     On "PC" computers, AGP memory can be allocated with glXAllocateMemoryNV
     and freed with glXFreeMemoryNV.  Sometimes it's useful to know where a
     block of AGP memory is located with respect to the start of the AGP
-    aperature.  The function
+    aperture.  The function
 
         GLuint glXGetAGPOffsetMESA( const GLvoid *pointer )
 

docs/MESA_drm_image.spec

@@ -24,7 +24,7 @@ Number
 
 Dependencies
 
-    Reguires EGL 1.4 or later.  This extension is written against the
+    Requires EGL 1.4 or later.  This extension is written against the
     wording of the EGL 1.4 specification.
 
     EGL_KHR_base_image is required.
@@ -134,7 +134,7 @@ Issues
         RESOLVED: The eglQueryImage function has been proposed often,
         but it goes against the EGLImage design.  EGLImages are opaque
         handles to a 2D array of pixels, which can be passed between
-        client APIs.  By referenceing an EGLImage in a client API, the
+        client APIs.  By referencing an EGLImage in a client API, the
         EGLImage target (a texture, a renderbuffer or such) can be
         used to query the attributes of the EGLImage.  We don't have a
         full client API for creating and querying DRM buffers, though,

docs/MESA_pack_invert.spec

@@ -48,7 +48,7 @@ IP Status
 
 Issues
 
-    1. Should we also defined UNPACK_INVERT_MESA for glDrawPixels, etc?
+    1. Should we also define UNPACK_INVERT_MESA for glDrawPixels, etc?
 
     Resolved:  No, we're only concerned with pixel packing.  There are other
     solutions for inverting images when using glDrawPixels (negative Y pixel

docs/MESA_release_buffers.spec

@@ -31,12 +31,12 @@ Overview
 
     Mesa's implementation of GLX is entirely implemented on the client side.
     Therefore, Mesa cannot immediately detect when an X window or pixmap is
-    destroyed in order to free any ancilliary data associated with the window
+    destroyed in order to free any ancillary data associated with the window
     or pixmap.
 
     The glxMesaReleaseBuffers() function can be used to explicitly indicate
-    when the back color buffer, depth buffer, stencil buffer, and/or accum-
-    ulation buffer associated with a drawable can be freed.
+    when the back color buffer, depth buffer, stencil buffer, and/or accumu-
+    lation buffer associated with a drawable can be freed.
 
 IP Status
 
@@ -60,7 +60,7 @@ Additions to Chapter 3 of the GLX 1.3 Specification (Functions and Errors)
 
 	Bool glXReleaseBuffersMESA( Display *dpy, GLXDrawable d );
 
-    causes all software ancilliary buffers (back buffer, depth, stencil,
+    causes all software ancillary buffers (back buffer, depth, stencil,
     accum, etc) associated with the named drawable to be immediately
     deallocated.  True is returned if <d> is a valid Mesa GLX drawable,
     else False is returned.  After calling glXReleaseBuffersMESA, the

docs/MESA_resize_buffers.spec

@@ -35,7 +35,7 @@ Overview
 
     Mesa's glViewport command queries the current window size and updates
     its internal data structors accordingly.  This normally works fine
-    since most applications call glViewport in responce to window size
+    since most applications call glViewport in response to window size
     changes.
 
     In some situations, however, the application may not call glViewport
@@ -65,7 +65,7 @@ Additions to the AGL/GLX/WGL Specifications
 
 Errors
 
-    INVALID_OPERATION is generated if ResizeBuffersMESA is called betweeen
+    INVALID_OPERATION is generated if glResizeBuffersMESA is called between
     Begin and End.
 
 New State

docs/MESA_swap_control.spec

@@ -88,8 +88,8 @@ Additions to the GLX 1.3 Specification
     this is typically the time required to display both the even and odd 
     fields of a frame of video data.
 
-    If <interval> is set to a value of 0, buffer swaps are not synchron-
-    ized to a video frame.  The <interval> value is silently clamped to
+    If <interval> is set to a value of 0, buffer swaps are not synchro-
+    nized to a video frame.  The <interval> value is silently clamped to
     the maximum implementation-dependent value supported before being
     stored.
 

docs/MESA_texture_array.spec

@@ -328,7 +328,7 @@ Additions to Chapter 3 of the OpenGL 2.0 Specification (Rasterization)
 
     -- Section 3.8.10 "Texture Completeness"
 
-       Change the second paragaph (page 177) to say (spec changes identical
+       Change the second paragraph (page 177) to say (spec changes identical
        to EXT_texture_array):
 
        "For one-, two-, or three-dimensional textures and one- or
@@ -443,10 +443,10 @@ Additions to Chapter 3 of the OpenGL 2.0 Specification (Rasterization)
        Change second through fourth paragraphs (page 188) to say:
 
        "Let D_t be the depth texture value, in the range [0, 1].  For
-       texture lookups from one- and two-dimesional, rectangle, and
+       texture lookups from one- and two-dimensional, rectangle, and
        one-dimensional array targets, let R be the interpolated <r>
        texture coordinate, clamped to the range [0, 1].  For texture lookups
-       from two-dimesional array texture targets, let R be the interpolated
+       from two-dimensional array texture targets, let R be the interpolated
        <q> texture coordinate, clamped to the range [0, 1].  Then the
        effective texture value L_t, I_t, or A_t is computed as follows:
 
@@ -604,7 +604,7 @@ Additions to Chapter 5 of the OpenGL 2.0 Specification (Special Functions)
 
     -- Section 5.4 "Display Lists"
 
-       Change the first paragraphi on page 242 to say (spec changes
+       Change the first paragraph on page 242 to say (spec changes
        identical to EXT_texture_array):
 
        "TexImage3D, TexImage2D, TexImage1D, Histogram, and ColorTable are

docs/MESA_texture_signed_rgba.spec

@@ -44,7 +44,7 @@ Overview
     ARB_texture_compression_rgtc introduces some compressed red and
     red_green signed formats but no uncompressed ones, which might
     still be useful. NV_texture_shader adds signed texture formats,
-    but also a lot of functionality which has been superceded by fragment
+    but also a lot of functionality which has been superseded by fragment
     shaders.
     It is usually possible to get the same functionality
     using a unsigned format by doing scale and bias in a shader, but this
@@ -105,7 +105,7 @@ Issues
        This is also different to what NV_texture_shader used.
        The above mapping should be considered the reference, but there
        is some leeway so other mappings are allowed for implementations which
-       cannot do this. Particulary the mapping given in NV_texture_shader or
+       cannot do this. Particularly the mapping given in NV_texture_shader or
        the standard OpenGL byte/float mapping is considered acceptable too, as
        might be a mapping which represents -1.0 by -128, 0.0 by 0 and 1.0 by
        127 (that is, uses different scale factors for negative and positive

docs/MESA_window_pos.spec

@@ -88,7 +88,7 @@ Additions to Chapter 2 of the OpenGL 1.2 Specification (OpenGL Operation)
       WindosPos4MESA takes four values indicating x, y, z, and w.
       WindowPos3MESA (or WindowPos2MESA) is analaguos, but sets only
       x, y, and z with w implicitly set to 1 (or only x and y with z
-      implicititly set to 0 and w implicitly set to 1).
+      implicitly set to 0 and w implicitly set to 1).
 
       WindowPosMESA operates like RasterPos except that the current modelview
       matrix, projection matrix and viewport parameters are ignored and the
@@ -108,7 +108,7 @@ GLX Protocol
 
 Errors
 
-    INVALID_OPERATION is generated if WindowPosMESA is called betweeen
+    INVALID_OPERATION is generated if WindowPosMESA is called between
     Begin and End.
 
 New State

docs/MESA_ycbcr_texture.spec

@@ -26,7 +26,7 @@ Number
 Dependencies
 
     OpenGL 1.0 or later is required
-    This extensions is written against the OpenGL 1.4 Specification.
+    This extension is written against the OpenGL 1.4 Specification.
     NV_texture_rectangle effects the definition of this extension.
 
 Overview
@@ -104,7 +104,7 @@ Additions to Chapter 3 of the OpenGL 1.4 Specification (Rasterization)
 
 
     In section 3.6.4, Rasterization of Pixel Rectangles, on page 104,
-    add the following to Table 3.12 (Packed pixel fiedl assignments):
+    add the following to Table 3.12 (Packed pixel field assignments):
 
                        First       Second     Third      Fourth
     Format             Element     Element    Element    Element

docs/WL_bind_wayland_display.spec

@@ -25,7 +25,7 @@ Number
 
 Dependencies
 
-    Reguires EGL 1.4 or later.  This extension is written against the
+    Requires EGL 1.4 or later.  This extension is written against the
     wording of the EGL 1.4 specification.
 
     EGL_KHR_base_image is required.
@@ -40,9 +40,9 @@ Overview
     provides a new target for eglCreateImageKHR, to create an EGLImage
     from a wl_buffer
 
-    Adding a implementation specific wayland interface, allows the
+    Adding an implementation specific wayland interface, allows the
     EGL implementation to define specific wayland requests and events,
-    needed for buffer sharing in a EGL wayland platform.
+    needed for buffer sharing in an EGL wayland platform.
 
 IP Status
 

docs/contents.html

@@ -62,6 +62,7 @@ a:visited {
 <LI><A HREF="perf.html" target="MainFrame">Performance Tips</A>
 <LI><A HREF="extensions.html" target="MainFrame">Mesa Extensions</A>
 <LI><A HREF="mangling.html" target="MainFrame">Function Name Mangling</A>
+<LI><A href="llvmpipe.html" target="MainFrame">Gallium llvmpipe driver</A>
 </ul>
 
 <b>Developer Topics</b>

docs/download.html

@@ -21,76 +21,52 @@ When a new release is coming, release candidates (betas) may be found
 
 
 <p>
-Mesa is distributed in several parts:
+The Mesa package is named MesaLib-x.y.z.{tar.bz2, tar.gz, zip} where x.y.z
+is the version.  There are three types of compressed archives.
 </p>
-<ul>
-<li><b>MesaLib-x.y.z</b> - the main Mesa library source code, drivers
-    and documentation.
-</li>
-<li><b>MesaDemos-x.y.z</b> - OpenGL demonstration and test programs.
-    Most of the programs require GLUT (either the
-    <a href="http://www.opengl.org/resources/libraries/glut"
-    target="_parent">original GLUT by Mark Kilgard</a> or
-    <a href="http://freeglut.sourceforge.net" target="_parent">freeglut</a> or
-    <a href="http://openglut.sourceforge.net" target="_parent">OpenGLUT</a>).
-</li>
-<li><b>MesaGLUT-x.y.z</b> - Mark Kilgard's GLUT, easily compiled and used
-    with Mesa.  Plus, other implementation of GLUT for DOS, OS/2, BeOS, etc.
-</li>
-</ul>
-
 <p>
-If you're not interested in running the demos, you'll only need the first
-package.
+There's also the MesaGLUT-x.y.z.{tar.bz2, tar.gz, zip} packages which
+contain Mark Kilgard's GLUT library.
+This is optional.
+Most Linux distributions include an implementation of GLUT (such as freeglut).
+</p>
+<p>
+In the past, the Mesa demos collection was distributed as
+MesaDemos-x.y.z.{tar.bz2, tar.gz, zip}.
+Now, the
+<a href="ftp://ftp.freedesktop.org/pub/mesa/demos/" target="_parent">
+Mesa demos</a> are distributed separately.
 </p>
 
-<p>
-If you're new to this and not sure what you're doing, grab all three packages.
-</p>
-
-<p>
-The packages are available in .tar.gz, .tar.bz2 and .zip formats.
-Other sites might offer additional package formats.
-</p>
 
 <H1>Unpacking</H1>
 
 <p>
-All the packages should be in the same directory prior to unpacking.
+To unpack .tar.gz files:
 </p>
-
-<ul>
-<li>To unpack .tar.gz files:
 <pre>
-	tar zxf MesaLib-X.Y.tar.gz
-	tar zxf MesaDemos-X.Y.tar.gz
-	tar zxf MesaGLUT-X.Y.tar.gz
+	tar zxf MesaLib-x.y.z.tar.gz
 </pre>
 or
 <pre>
-	gzcat MesaLib-X.Y.tar.gz | tar xf -
-	gzcat MesaDemos-X.Y.tar.gz | tar xf -
-	gzcat MesaGLUT-X.Y.tar.gz | tar xf -
+	gzcat MesaLib-x.y.z.tar.gz | tar xf -
 </pre>
 or
 <pre>
-	gunzip MesaLib-X.Y.tar.gz ; tar xf MesaLib-X.Y.tar
-	gunzip MesaDemos-X.Y.tar.gz ; tar xf MesaDemos-X.Y.tar
-	gunzip MesaGLUT-X.Y.tar.gz ; tar xf MesaGLUT-X.Y.tar
+	gunzip MesaLib-x.y.z.tar.gz ; tar xf MesaLib-x.y.z.tar
 </pre>
-<li>To unpack .tar.bz2 files:
+<p>
+To unpack .tar.bz2 files:
+</p>
 <pre>
-	bunzip2 -c MesaLib-X.Y.tar.gz | tar xf -
-	bunzip2 -c MesaDemos-X.Y.tar.gz | tar xf -
-	bunzip2 -c MesaGLUT-X.Y.tar.gz | tar xf -
+	bunzip2 -c MesaLib-x.y.z.tar.gz | tar xf -
 </pre>
-<li>To unpack .zip files:
+<p>
+To unpack .zip files:
+</p>
 <pre>
-	unzip MesaLib-X.Y.zip
-	unzip MesaDemos-X.Y.zip
-	unzip MesaGLUT-X.Y.zip
+	unzip MesaLib-x.y.z.zip
 </pre>
-</ul>
 
 
 <h1>Contents</h1>
@@ -106,22 +82,13 @@ bin/		- shell scripts for making shared libraries, etc
 docs/		- documentation
 src/		- source code for libraries
 src/mesa	- sources for the main Mesa library and device drivers
+src/gallium     - sources for Gallium and Gallium drivers
 src/glu		- libGLU source code
 src/glx		- sources for building libGL with full GLX and DRI support
 src/glw		- Xt/Motif/OpenGL widget code
 </pre>
 
-If you downloaded and unpacked the MesaDemos.X.Y package:
-
-<pre>
-progs/demos	- original Mesa demos
-progs/xdemos	- GLX OpenGL/Mesa demos
-progs/redbook	- examples from the OpenGL Programming Guide
-progs/samples	- examples from SGI
-progs/images/	- image files
-</pre>
-
-If you downloaded and unpacked the MesaGLUT.X.Y package:
+If you downloaded and unpacked the MesaGLUT.x.y.z package:
 <pre>
 src/glut	- GLUT library source code
 </pre>

docs/envvars.html

@@ -49,6 +49,15 @@ Setting this variable automatically sets the MESA_TEX_PROG variable as well.
 <li>MESA_EXTENSION_OVERRIDE - can be used to enable/disable extensions.
 A value such as "GL_EXT_foo -GL_EXT_bar" will enable the GL_EXT_foo extension
 and disable the GL_EXT_bar extension.
+<li>MESA_EXTENSION_MAX_YEAR - The GL_EXTENSIONS string returned by Mesa is sorted
+by extension year.
+If this variable is set to year X, only extensions defined on or before year
+X will be reported.
+This is to work-around a bug in some games where the extension string is
+copied into a fixed-size buffer without truncating.
+If the extension string is too long, the buffer overrun can cause the game
+to crash.
+This is a work-around for that.
 <li>MESA_GLSL - <a href="shading.html#envvars">shading language compiler options</a>
 </ul>
 

docs/install.html

@@ -12,16 +12,16 @@
 <ol>
 <li><a href="#unix-x11">Unix / X11</a>
   <ul>
-  <li><a href="#prereq">Prerequisites for DRI and hardware acceleration</a>
+  <li><a href="#prereq-general">General prerequisites for building</a>
+  <li><a href="#prereq-dri">Prerequisites for DRI and hardware acceleration</a>
   <li><a href="#autoconf">Building with autoconf</a>
   <li><a href="#traditional">Building with traditional Makefiles</a>
   <li><a href="#libs">The Libraries</a>
-  <li><a href="#demos">Running the demos
   <li><a href="#install">Installing the header and library files
   <li><a href="#pkg-config">Building OpenGL programs with pkg-config
   </ul>
 <li><a href="#windows">Windows</a>
-<li><a href="#scons">SCons</a>
+<li><a href="#scons">Building with SCons</a>
 <li><a href="#other">Other</a>
 </ol>
 <br>
@@ -31,8 +31,22 @@
 <H2>1. Unix/X11 Compilation and Installation</H1>
 
 
-<a name="prereq">
-<h3>1.1 Prerequisites for DRI and hardware acceleration</h3>
+<a name="prereq-general">
+<h3>1.1 General prerequisites for building</h3>
+
+<ul>
+<li>lex / yacc - for building the GLSL compiler.
+On Linux systems, flex and bison are used.
+Versions 2.5.35 and 2.4.1, respectively, (or later) should work.
+</li>
+<li>python - Python is needed for building the Gallium components.
+Version 2.6.4 or later should work.
+</li>
+</ul>
+
+
+<a name="prereq-dri">
+<h3>1.2 Prerequisites for DRI and hardware acceleration</h3>
 
 <p>
 The following are required for DRI-based hardware acceleration with Mesa:
@@ -49,7 +63,7 @@ version 2.4.15 or later
 
 
 <a name="autoconf">
-<h3>1.2 Building with Autoconf</h3>
+<h3>1.3 Building with Autoconf</h3>
 
 <p>
 Mesa may be <a href="autoconf.html">built using autoconf</a>.
@@ -59,7 +73,7 @@ If that fails the traditional Mesa build system is available.
 
 
 <a name="traditional">
-<h3>1.3 Building with traditional Makefiles</h3>
+<h3>1.4 Building with traditional Makefiles</h3>
 
 <p>
 The traditional Mesa build system is based on a collection of pre-defined
@@ -126,7 +140,7 @@ Later, if you want to rebuild for a different configuration run
 
 
 <a name="libs">
-<h3>1.4 The libraries</h3>
+<h3>1.5 The libraries</h3>
 
 <p>
 When compilation has finished, look in the top-level <code>lib/</code>
@@ -185,81 +199,11 @@ If you built the DRI hardware drivers, you'll also see the DRI drivers:
 -rwxr-xr-x   1 brian users 10997120 Jul 21 12:13 unichrome_dri.so
 </pre>
 
-
-<a name="demos">
-<h3>1.5 Running the demos</h3>
-
 <p>
-If you downloaded/unpacked the MesaDemos-x.y.z.tar.gz archive or
-obtained Mesa from CVS, the <b>progs/</b> directory will contain a
-bunch of demonstration programs.
+If you built with Gallium support, look in lib/gallium/ for Gallium-based
+versions of libGL and device drivers.
 </p>
 
-<p>
-Before running a demo, you'll probably have to set two environment variables
-to indicate where the libraries are located.  For example:
-<p>
-<blockquote>
-<b>cd lib/</b>
-<br>
-<b>export LD_LIBRARY_PATH=${PWD}</b>
-<br>
-<b>export LIBGL_DRIVERS_PATH=${PWD}</b> (if using DRI drivers)
-</blockquote>
-
-<p>
-Next, change to the Mesa/demos/ directory:
-</p>
-<blockquote>
-<b>cd ../progs/demos</b>
-</blockquote>
-
-<p>
-Run a demo such as gears:
-</p>
-<blockquote>
-<b>./gears</b>
-</blockquote>
-
-<p>
-If this doesn't work, try the <b>Mesa/progs/xdemos/glxinfo</b> program
-and see that it prints the expected Mesa version number.
-</p>
-
-<p>
-If you're using Linux or a similar OS, verify that the demo program is
-being linked with the proper library files:
-</p>
-<blockquote>
-<b>ldd gears</b>
-</blockquote>
-
-<p>
-You should see something like this:
-</p>
-<pre>
-        libglut.so.3 => /home/brian/Mesa/lib/libglut.so.3 (0x40013000)
-        libGLU.so.1 => /home/brian/Mesa/lib/libGLU.so.1 (0x40051000)
-        libGL.so.1 => /home/brian/Mesa/lib/libGL.so.1 (0x400e0000)
-        libc.so.6 => /lib/i686/libc.so.6 (0x42000000)
-        libm.so.6 => /lib/i686/libm.so.6 (0x403da000)
-        libX11.so.6 => /usr/X11R6/lib/libX11.so.6 (0x403fc000)
-        libXmu.so.6 => /usr/X11R6/lib/libXmu.so.6 (0x404da000)
-        libXt.so.6 => /usr/X11R6/lib/libXt.so.6 (0x404f1000)
-        libXi.so.6 => /usr/X11R6/lib/libXi.so.6 (0x40543000)
-        libstdc++.so.5 => /usr/lib/libstdc++.so.5 (0x4054b000)
-        libgcc_s.so.1 => /lib/libgcc_s.so.1 (0x405fd000)
-        libXext.so.6 => /usr/X11R6/lib/libXext.so.6 (0x40605000)
-        libpthread.so.0 => /lib/i686/libpthread.so.0 (0x40613000)
-        /lib/ld-linux.so.2 => /lib/ld-linux.so.2 (0x40000000)
-        libdl.so.2 => /lib/libdl.so.2 (0x40644000)
-        libSM.so.6 => /usr/X11R6/lib/libSM.so.6 (0x40647000)
-        libICE.so.6 => /usr/X11R6/lib/libICE.so.6 (0x40650000)
-</pre>
-
-<p>
-Retrace your steps if this doesn't look right.
-</p>
 
 
 <a name="install">

docs/llvmpipe.html

@@ -1,61 +1,116 @@
-LLVMPIPE -- a fork of softpipe that employs LLVM for code generation.
+<HTML>
+
+<TITLE>llvmpipe</TITLE>
+
+<link rel="stylesheet" type="text/css" href="mesa.css"></head>
+
+<BODY>
+
+<H1>Introduction</H1>
+
+<p>
+The Gallium llvmpipe driver is a software rasterizer that uses LLVM to
+do runtime code generation.
+Shaders, point/line/triangle rasterization and vertex processing are
+implemented with LLVM IR which is translated to x86 or x86-64 machine
+code.
+Also, the driver is multithreaded to take advantage of multiple CPU cores
+(up to 8 at this time).
+It's the fastest software rasterizer for Mesa.
+</p>
 
 
-Requirements
-============
+<h1>Requirements</h1>
 
- - A x86 or amd64 processor.  64bit mode is preferred.
- 
+<dl>
+<dt>An x86 or amd64 processor.  64-bit mode is preferred.</dt>
+<dd>
+   <p>
    Support for sse2 is strongly encouraged.  Support for ssse3, and sse4.1 will
    yield the most efficient code.  The less features the CPU has the more
    likely is that you ran into underperforming, buggy, or incomplete code.  
-   
+   </p>
+   <p>
    See /proc/cpuinfo to know what your CPU supports.
- 
- - LLVM. Version 2.8 recommended. 2.6 or later required.
- 
+   </p>
+</dd>
+<dt>LLVM. Version 2.8 recommended. 2.6 or later required.</dt>
+<dd>
+   <p>
+   <b>NOTE</b>: LLVM 2.8 and earlier will not work on systems that support the
+   Intel AVX extensions (e.g. Sandybridge).  LLVM's code generator will
+   fail when trying to emit AVX instructions.  This was fixed in LLVM 2.9.
+   </p>
+   <p>
    For Linux, on a recent Debian based distribution do:
- 
+   </p>
+<pre>
      aptitude install llvm-dev
+</pre>
+   For a RPM-based distribution do:
+   </p>
+<pre>
+     yum install llvm-devel
+</pre>
 
+   <p>
    For Windows download pre-built MSVC 9.0 or MinGW binaries from
    http://people.freedesktop.org/~jrfonseca/llvm/ and set the LLVM environment
    variable to the extracted path.
+   </p>
 
+   <p>
    For MSVC there are two set of binaries: llvm-x.x-msvc32mt.7z and
    llvm-x.x-msvc32mtd.7z .
+   </p>
 
+   <p>
    You have to set the LLVM=/path/to/llvm-x.x-msvc32mtd env var when passing
    debug=yes to scons, and LLVM=/path/to/llvm-x.x-msvc32mt when building with
    debug=no. This is necessary as LLVM builds as static library so the chosen
    MS CRT must match.
+   </p>
+</dd>
 
- - scons (optional)
+<dt>scons (optional)</dt>
+</dl>
 
 
-Building
-========
+
+<h1>Building</h1>
 
 To build everything on Linux invoke scons as:
 
+<pre>
   scons build=debug libgl-xlib
+</pre>
 
 Alternatively, you can build it with GNU make, if you prefer, by invoking it as
 
+<pre>
   make linux-llvm
+</pre>
 
 but the rest of these instructions assume that scons is used.
 
 For windows is everything the except except the winsys:
 
+<pre>
   scons build=debug libgl-gdi
+</pre>
 
-Using
-=====
+
+<h1>Using</h1>
 
 On Linux, building will create a drop-in alternative for libGL.so into
 
+<pre>
   build/foo/gallium/targets/libgl-xlib/libGL.so
+</pre>
+or
+<pre>
+  lib/gallium/libGL.so
+</pre>
 
 To use it set the LD_LIBRARY_PATH environment variable accordingly.
 
@@ -68,19 +123,20 @@ replacing the native ICD driver, but it's quite an advanced usage, so if you
 need to ask, don't even try it.
 
 
-Profiling
-=========
+<h1>Profiling</h1>
 
 To profile llvmpipe you should pass the options
 
+<pre>
   scons build=profile <same-as-before>
+</pre>
 
 This will ensure that frame pointers are used both in C and JIT functions, and
 that no tail call optimizations are done by gcc.
 
-
 To better profile JIT code you'll need to build LLVM with oprofile integration.
 
+<pre>
   ./configure \
       --prefix=$install_dir \
       --enable-optimized \
@@ -92,43 +148,57 @@ To better profile JIT code you'll need to build LLVM with oprofile integration.
   make -C "$build_dir" install
 
   find "$install_dir/lib" -iname '*.a' -print0 | xargs -0 strip --strip-debug
+</pre>
 
 The you should define
 
+<pre>
   export LLVM=/path/to/llvm-2.6-profile
+</pre>
 
 and rebuild.
 
 
-Unit testing
-============
+<h1>Unit testing</h1>
 
+<p>
 Building will also create several unit tests in
 build/linux-???-debug/gallium/drivers/llvmpipe:
+</p>
 
- - lp_test_blend: blending
- - lp_test_conv: SIMD vector conversion
- - lp_test_format: pixel unpacking/packing
+</ul>
+<li> lp_test_blend: blending
+<li> lp_test_conv: SIMD vector conversion
+<li> lp_test_format: pixel unpacking/packing
+</ul>
 
+<p>
 Some of this tests can output results and benchmarks to a tab-separated-file
 for posterior analysis, e.g.:
-
+</p>
+<pre>
   build/linux-x86_64-debug/gallium/drivers/llvmpipe/lp_test_blend -o blend.tsv
+</pre>
 
 
-Development Notes
-=================
+<h1>Development Notes</h1>
 
-- When looking to this code by the first time start in lp_state_fs.c, and 
+<ul>
+<li>
+  When looking to this code by the first time start in lp_state_fs.c, and 
   then skim through the lp_bld_* functions called in there, and the comments
   at the top of the lp_bld_*.c functions.  
-
-- The driver-independent parts of the LLVM / Gallium code are found in
+</li>
+<li>
+  The driver-independent parts of the LLVM / Gallium code are found in
   src/gallium/auxiliary/gallivm/.  The filenames and function prefixes
   need to be renamed from "lp_bld_" to something else though.
-
-- We use LLVM-C bindings for now. They are not documented, but follow the C++
+</li>
+<li>
+  We use LLVM-C bindings for now. They are not documented, but follow the C++
   interfaces very closely, and appear to be complete enough for code
   generation. See 
   http://npcontemplation.blogspot.com/2008/06/secret-of-llvm-c-bindings.html
   for a stand-alone example.  See the llvm-c/Core.h file for reference.
+</li>
+</ul>

docs/news.html

@@ -11,6 +11,13 @@
 <H1>News</H1>
 
 
+<h2>April 6, 2011</h2>
+
+<p>
+<a href="relnotes-7.10.2.html">Mesa 7.10.2</a> is released.  This is a bug
+fix release release.
+</p>
+
 <h2>March 2, 2011</h2>
 
 <p>

docs/relnotes-7.10.2.html

@@ -0,0 +1,206 @@
+<HTML>
+
+<head>
+<TITLE>Mesa Release Notes</TITLE>
+<link rel="stylesheet" type="text/css" href="mesa.css">
+<meta http-equiv="content-type" content="text/html; charset=utf-8" />
+</head>
+
+<BODY>
+
+<body bgcolor="#eeeeee">
+
+<H1>Mesa 7.10.2 Release Notes / April 6, 2011</H1>
+
+<p>
+Mesa 7.10.2 is a bug fix release which fixes bugs found since the 7.10 release.
+</p>
+<p>
+Mesa 7.10.2 implements the OpenGL 2.1 API, but the version reported by
+glGetString(GL_VERSION) depends on the particular driver being used.
+Some drivers don't support all the features required in OpenGL 2.1.
+</p>
+<p>
+See the <a href="install.html">Compiling/Installing page</a> for prerequisites
+for DRI hardware acceleration.
+</p>
+
+
+<h2>MD5 checksums</h2>
+<pre>
+2f9f444265534a2cfd9a99d1a8291089  MesaLib-7.10.2.tar.gz
+f5de82852f1243f42cc004039e10b771  MesaLib-7.10.2.tar.bz2
+47836e37bab6fcafe3ac90c9544ba0e9  MesaLib-7.10.2.zip
+175120325828f313621cc5bc6c504803  MesaGLUT-7.10.2.tar.gz
+8c71d273f5f8d6c5eda4ffc39e0fe03e  MesaGLUT-7.10.2.tar.bz2
+03036c8efe7b791a90fa0f2c41b43f43  MesaGLUT-7.10.2.zip
+</pre>
+
+
+<h2>New features</h2>
+<p>None.</p>
+
+<h2>Bug fixes</h2>
+<p>This list is likely incomplete.</p>
+<ul>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=29172">Bug 29172</a> - Arrandale - Pill Popper Pops Pills</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=31159">Bug 31159</a> - shadow problem in 0ad game</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=32688">Bug 32688</a> - [RADEON:KMS:R300G] some games have a wireframe or outline visible</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=32949">Bug 32949</a> - [glsl wine] Need for Speed renders incorrectly with GLSL enabled</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=34203">Bug 34203</a> - [GLSL] fail to call long chains across shaders</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=34376">Bug 34376</a> - [GLSL] allowing assignment to unsized array
+  <ul>
+    <li>The commit message incorrectly
+    lists <a href="https://bugs.freedesktop.org/show_bug.cgi?id=34367">bug
+    34367</a>.</li>
+  </ul>
+</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=34370">Bug 34370</a> - [GLSL] "i&lt;5 &amp;&amp; i&lt;4" in for loop fails</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=34374">Bug 34374</a> - [GLSL] fail to redeclare an array using initializer</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=35073">Bug 35073</a> - [GM45] Alpha test is broken when rendering to FBO with no color attachment</li>
+
+<li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=35483">Bug 35483</a> - util_blit_pixels_writemask: crash in line 322 of src/gallium/auxiliary/util/u_blit.c</li>
+
+<!-- <li><a href="https://bugs.freedesktop.org/show_bug.cgi?id=">Bug </a> - </li> -->
+
+</ul>
+
+
+<h2>Changes</h2>
+<p>The full set of changes can be viewed by using the following GIT command:</p>
+
+<pre>
+  git log mesa-7.10.1..mesa-7.10.2
+</pre>
+
+<p><em>Note:</em> Reverted commits and the reverts are not included in this list.</p>
+
+<p>Alex Deucher (2):
+<ul>
+  <li>r600c: add new ontario pci ids</li>
+  <li>r600g: add some additional ontario pci ids</li>
+</ul></p>
+
+<p>Benjamin Franzke (1):
+<ul>
+  <li>st/dri: Fix surfaceless gl using contexts with previous bound surfaces</li>
+</ul></p>
+
+<p>Brian Paul (9):
+<ul>
+  <li>docs: pull 7.9.2 release notes into 7.10 branch</li>
+  <li>docs: update news.html with 7.10.1 and 7.9.2 releases</li>
+  <li>docs: fill in 7.10.1 release data</li>
+  <li>docs: add, fix release notes links</li>
+  <li>docs: update info about Mesa packaging/contents</li>
+  <li>docs: update prerequisites, remove old demo info</li>
+  <li>mesa: Guard against null pointer deref in fbo validation</li>
+  <li>st/mesa: Apply LOD bias from correct texture unit</li>
+  <li>glsl: silence warning in printf() with a cast</li>
+</ul></p>
+
+<p>Chad Versace (1):
+<ul>
+  <li>i965: Fix tex_swizzle when depth mode is GL_RED</li>
+</ul></p>
+
+<p>Dave Airlie (1):
+<ul>
+  <li>r600: don't close fd on failed load</li>
+</ul></p>
+
+<p>Eric Anholt (2):
+<ul>
+  <li>i965: Apply a workaround for the Ironlake "vertex flashing".</li>
+  <li>i965: Fix alpha testing when there is no color buffer in the FBO.</li>
+</ul></p>
+
+<p>Fabian Bieler (1):
+<ul>
+  <li>st/mesa: Apply LOD from texture object</li>
+</ul></p>
+
+<p>Henri Verbeet (1):
+<ul>
+  <li>st/mesa: Validate state before doing blits.</li>
+</ul></p>
+
+<p>Ian Romanick (13):
+<ul>
+  <li>docs: Add 7.10.1 md5sums</li>
+  <li>glsl: Refactor AST-to-HIR code handling variable initializers</li>
+  <li>glsl: Refactor AST-to-HIR code handling variable redeclarations</li>
+  <li>glsl: Process redeclarations before initializers</li>
+  <li>glsl: Function signatures cannot have NULL return type</li>
+  <li>glsl: Add several function / call related validations</li>
+  <li>linker: Add imported functions to the linked IR</li>
+  <li>glsl: Use insert_before for lists instead of open coding it</li>
+  <li>glsl: Only allow unsized array assignment in an initializer</li>
+  <li>glcpp: Refresh autogenerated lexer files</li>
+  <li>docs: Initial bits of 7.10.2 release notes</li>
+  <li>mesa: set version string to 7.10.2</li>
+  <li>mesa: Remove nonexistant files from _FILES lists</li>
+</ul></p>
+
+<p>Jerome Glisse (1):
+<ul>
+  <li>r600g: move user fence into base radeon structure</li>
+</ul></p>
+
+<p>José Fonseca (2):
+<ul>
+  <li>mesa: Fix typo glGet*v(GL_TEXTURE_COORD_ARRAY_*).</li>
+  <li>mesa: More glGet* fixes.</li>
+</ul></p>
+
+<p>Kenneth Graunke (4):
+<ul>
+  <li>glcpp: Rework lexer to use a SKIP state rather than REJECT.</li>
+  <li>glcpp: Remove trailing contexts from #if rules.</li>
+  <li>i965/fs: Fix linear gl_Color interpolation on pre-gen6 hardware.</li>
+  <li>glsl: Accept precision qualifiers on sampler types, but only in ES.</li>
+</ul></p>
+
+<p>Marek Olšák (15):
+<ul>
+  <li>st/mesa: fix crash when DrawBuffer-&gt;_ColorDrawBuffers[0] is NULL</li>
+  <li>st/mesa: fail to alloc a renderbuffer if st_choose_renderbuffer_format fails</li>
+  <li>r300/compiler: fix the saturate modifier when applied to TEX instructions</li>
+  <li>r300/compiler: fix translating the src negate bits in pair_translate</li>
+  <li>r300/compiler: Abs doesn't cancel Negate (in the conversion to native swizzles)</li>
+  <li>r300/compiler: TEX instructions don't support negation on source arguments</li>
+  <li>r300/compiler: do not set TEX_IGNORE_UNCOVERED on r500</li>
+  <li>r300/compiler: saturate Z before the shadow comparison</li>
+  <li>r300/compiler: fix equal and notequal shadow compare functions</li>
+  <li>r300/compiler: remove unused variables</li>
+  <li>st/mesa: fix crash when using both user and vbo buffers with the same stride</li>
+  <li>r300g: fix alpha-test with no colorbuffer</li>
+  <li>r300g: tell the GLSL compiler to lower the continue opcode</li>
+  <li>r300/compiler: propagate SaturateMode down to the result of shadow comparison</li>
+  <li>r300/compiler: apply the texture swizzle to shadow pass and fail values too</li>
+</ul></p>
+
+<p>Michel Dänzer (1):
+<ul>
+  <li>Use proper source row stride when getting depth/stencil texels.</li>
+</ul></p>
+
+<p>Tom Stellard (4):
+<ul>
+  <li>r300/compiler: Use a 4-bit writemask in pair instructions</li>
+  <li>prog_optimize: Fix reallocating registers for shaders with loops</li>
+  <li>r300/compiler: Fix vertex shader MAD instructions with constant swizzles</li>
+  <li>r300/compiler: Don't try to convert RGB to Alpha in full instructions</li>
+</ul></p>
+
+</body>
+</html>

docs/relnotes-7.11.html

@@ -41,6 +41,7 @@ tbd
 <li>GL_ARB_texture_compression_rgtc (gallium r600, swrast)
 <li>GL_ARB_draw_buffers_blend (gallium)
 <li>GL_EXT_texture_sRGB_decode (gallium drivers, swrast, i965)
+<li>GL_ARB_sampler_objects (gallium drivers)
 </ul>
 
 

docs/relnotes.html

@@ -14,6 +14,7 @@ The release notes summarize what's new or changed in each Mesa release.
 
 <UL>
 <LI><A HREF="relnotes-7.11.html">7.11 release notes</A>
+<LI><A HREF="relnotes-7.10.2.html">7.10.2 release notes</A>
 <LI><A HREF="relnotes-7.10.1.html">7.10.1 release notes</A>
 <LI><A HREF="relnotes-7.10.html">7.10 release notes</A>
 <LI><A HREF="relnotes-7.9.2.html">7.9.2 release notes</A>

include/GL/gl.h

@@ -655,6 +655,8 @@ typedef double		GLclampd;	/* double precision float in [0,1] */
 #define GL_TEXTURE_ENV_COLOR			0x2201
 #define GL_TEXTURE_GEN_S			0x0C60
 #define GL_TEXTURE_GEN_T			0x0C61
+#define GL_TEXTURE_GEN_R			0x0C62
+#define GL_TEXTURE_GEN_Q			0x0C63
 #define GL_TEXTURE_GEN_MODE			0x2500
 #define GL_TEXTURE_BORDER_COLOR			0x1004
 #define GL_TEXTURE_WIDTH			0x1000
@@ -685,8 +687,6 @@ typedef double		GLclampd;	/* double precision float in [0,1] */
 #define GL_T					0x2001
 #define GL_R					0x2002
 #define GL_Q					0x2003
-#define GL_TEXTURE_GEN_R			0x0C62
-#define GL_TEXTURE_GEN_Q			0x0C63
 
 /* Utility */
 #define GL_VENDOR				0x1F00

scons/python.py

@@ -1,72 +0,0 @@
-"""gallium
-
-Frontend-tool for Gallium3D architecture.
-
-"""
-
-# 
-# Copyright 2008 Tungsten Graphics, Inc., Cedar Park, Texas.
-# All Rights Reserved.
-# 
-# Permission is hereby granted, free of charge, to any person obtaining a
-# copy of this software and associated documentation files (the
-# "Software"), to deal in the Software without restriction, including
-# without limitation the rights to use, copy, modify, merge, publish,
-# distribute, sub license, and/or sell copies of the Software, and to
-# permit persons to whom the Software is furnished to do so, subject to
-# the following conditions:
-# 
-# The above copyright notice and this permission notice (including the
-# next paragraph) shall be included in all copies or substantial portions
-# of the Software.
-# 
-# THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
-# OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
-# MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
-# IN NO EVENT SHALL TUNGSTEN GRAPHICS AND/OR ITS SUPPLIERS BE LIABLE FOR
-# ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
-# TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
-# SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
-# 
-
-
-import sys
-import distutils.sysconfig
-import os.path
-
-
-def generate(env):
-    # See http://www.scons.org/wiki/PythonExtensions
-    
-    if sys.platform in ['win32']:
-        python_root = sys.prefix
-        python_version = '%u%u' % sys.version_info[:2]
-        python_include = os.path.join(python_root, 'include')
-        python_libs = os.path.join(python_root, 'libs')
-        python_lib = os.path.join(python_libs, 'python' + python_version + '.lib')
-        
-        env.Append(CPPPATH = [python_include])
-        env.Append(LIBPATH = [python_libs])
-        env.Append(LIBS = ['python' + python_version + '.lib'])
-        env.Replace(SHLIBPREFIX = '')
-        env.Replace(SHLIBSUFFIX = '.pyd')
-        
-        # XXX; python25_d.lib is not included in Python for windows, and 
-        # we'll get missing symbols unless we undefine _DEBUG
-        cppdefines = env['CPPDEFINES']
-        cppdefines = [define for define in cppdefines if define != '_DEBUG']
-        env.Replace(CPPDEFINES = cppdefines)
-        env.AppendUnique(CPPFLAGS = ['/U_DEBUG'])
-        env.AppendUnique(LINKFLAGS = ['/nodefaultlib:python25_d.lib'])
-    else:
-        #env.ParseConfig('python-config --cflags --ldflags --libs')
-        env.AppendUnique(CPPPATH = [distutils.sysconfig.get_python_inc()])
-        env.Replace(SHLIBPREFIX = '')
-        env.Replace(SHLIBSUFFIX = distutils.sysconfig.get_config_vars()['SO'])
-
-    # for debugging
-    #print env.Dump()
-
-
-def exists(env):
-    return 1

src/egl/drivers/dri2/egl_dri2.c

@@ -462,6 +462,12 @@ dri2_create_screen(_EGLDisplay *disp)
       }
    }
 
+   if (dri2_dpy->image) {
+      disp->Extensions.MESA_drm_image = EGL_TRUE;
+      disp->Extensions.KHR_image_base = EGL_TRUE;
+      disp->Extensions.KHR_gl_renderbuffer_image = EGL_TRUE;
+   }
+
    return EGL_TRUE;
 
  cleanup_dri_screen:

src/egl/drivers/dri2/platform_drm.c

@@ -708,11 +708,6 @@ dri2_initialize_drm(_EGLDriver *drv, _EGLDisplay *disp)
    for (i = 0; dri2_dpy->driver_configs[i]; i++)
       dri2_add_config(disp, dri2_dpy->driver_configs[i], i + 1, 0, 0, NULL);
 
-   disp->Extensions.MESA_drm_image = EGL_TRUE;
-   disp->Extensions.KHR_image_base = EGL_TRUE;
-   disp->Extensions.KHR_gl_renderbuffer_image = EGL_TRUE;
-   disp->Extensions.KHR_gl_texture_2D_image = EGL_TRUE;
-
 #ifdef HAVE_WAYLAND_PLATFORM
    disp->Extensions.WL_bind_wayland_display = EGL_TRUE;
 #endif

src/egl/drivers/dri2/platform_wayland.c

@@ -429,7 +429,7 @@ wayland_create_buffer(struct dri2_egl_surface *dri2_surf, __DRIbuffer *buffer)
 }
 
 static void
-wayland_frame_callback(void *data, uint32_t time)
+wayland_frame_callback(struct wl_surface *surface, void *data, uint32_t time)
 {
    struct dri2_egl_surface *dri2_surf = data;
 
@@ -459,7 +459,8 @@ dri2_swap_buffers(_EGLDriver *drv, _EGLDisplay *disp, _EGLSurface *draw)
 
    dri2_surf->block_swap_buffers = EGL_TRUE;
    wl_display_frame_callback(dri2_dpy->wl_dpy->display,
-	 wayland_frame_callback, dri2_surf);
+	 		     dri2_surf->wl_win->surface,
+			     wayland_frame_callback, dri2_surf);
 
    if (dri2_surf->type == DRI2_WINDOW_SURFACE) {
       pointer_swap(
@@ -677,11 +678,7 @@ dri2_initialize_wayland(_EGLDriver *drv, _EGLDisplay *disp)
 		      EGL_WINDOW_BIT | EGL_PIXMAP_BIT, NULL);
 
 
-   disp->Extensions.MESA_drm_image = EGL_TRUE;
-   disp->Extensions.KHR_image_base = EGL_TRUE;
    disp->Extensions.KHR_image_pixmap = EGL_TRUE;
-   disp->Extensions.KHR_gl_renderbuffer_image = EGL_TRUE;
-   disp->Extensions.KHR_gl_texture_2D_image = EGL_TRUE;
 
    disp->Extensions.WL_bind_wayland_display = EGL_TRUE;
    dri2_dpy->authenticate = dri2_wayland_authenticate;

src/egl/drivers/dri2/platform_x11.c

@@ -231,7 +231,7 @@ dri2_create_surface(_EGLDriver *drv, _EGLDisplay *disp, EGLint type,
    if (dri2_dpy->dri2) {
       xcb_dri2_create_drawable (dri2_dpy->conn, dri2_surf->drawable);
    } else {
-      swrastCreateDrawable(dri2_dpy, dri2_surf, _eglGetConfigKey(conf, EGL_DEPTH_SIZE));
+      swrastCreateDrawable(dri2_dpy, dri2_surf, _eglGetConfigKey(conf, EGL_BUFFER_SIZE));
    }
 
    if (type != EGL_PBUFFER_BIT) {
@@ -990,7 +990,7 @@ dri2_initialize_x11_dri2(_EGLDriver *drv, _EGLDisplay *disp)
    if (!dri2_load_driver(disp))
       goto cleanup_conn;
 
-   dri2_dpy->fd = open(dri2_dpy->device_name, O_RDWR);
+   dri2_dpy->fd = open(dri2_dpy->device_name, O_RDWR | O_CLOEXEC);
    if (dri2_dpy->fd == -1) {
       _eglLog(_EGL_WARNING,
 	      "DRI2: could not open %s (%s)", dri2_dpy->device_name,
@@ -1030,11 +1030,7 @@ dri2_initialize_x11_dri2(_EGLDriver *drv, _EGLDisplay *disp)
 	 goto cleanup_configs;
    }
 
-   disp->Extensions.MESA_drm_image = EGL_TRUE;
-   disp->Extensions.KHR_image_base = EGL_TRUE;
    disp->Extensions.KHR_image_pixmap = EGL_TRUE;
-   disp->Extensions.KHR_gl_renderbuffer_image = EGL_TRUE;
-   disp->Extensions.KHR_gl_texture_2D_image = EGL_TRUE;
    disp->Extensions.NOK_swap_region = EGL_TRUE;
    disp->Extensions.NOK_texture_from_pixmap = EGL_TRUE;
 

src/egl/wayland/wayland-drm/Makefile

@@ -12,7 +12,8 @@ wayland_drm_OBJECTS = $(wayland_drm_SOURCES:.c=.o)
 
 wayland_drm_INCLUDES = \
 		$(WAYLAND_CFLAGS) \
-		-I$(TOP)/src/egl/main
+		-I$(TOP)/src/egl/main \
+		-I$(TOP)/include
 
 # Generate protocol sources
 prefix=$(shell pkg-config --variable=prefix wayland-server)

src/gallium/SConscript

@@ -53,7 +53,6 @@ if env['drm']:
 # Needed by some state trackers
 SConscript('winsys/sw/null/SConscript')
 
-SConscript('state_trackers/python/SConscript')
 if env['platform'] != 'embedded':
     SConscript('state_trackers/vega/SConscript')
     SConscript('state_trackers/egl/SConscript')

src/gallium/auxiliary/Makefile

@@ -107,6 +107,7 @@ C_SOURCES = \
 	util/u_caps.c \
 	util/u_cpu_detect.c \
 	util/u_dl.c \
+	util/u_draw.c \
 	util/u_draw_quad.c \
 	util/u_format.c \
 	util/u_format_other.c \

src/gallium/auxiliary/SConscript

@@ -154,6 +154,7 @@ source = [
     'util/u_dump_defines.c',
     'util/u_dump_state.c',
     'util/u_dl.c',
+    'util/u_draw.c',
     'util/u_draw_quad.c',
     'util/u_format.c',
     'util/u_format_other.c',

src/gallium/auxiliary/draw/draw_context.c

@@ -751,7 +751,7 @@ void
 draw_set_mapped_texture(struct draw_context *draw,
                         unsigned sampler_idx,
                         uint32_t width, uint32_t height, uint32_t depth,
-                        uint32_t last_level,
+                        uint32_t first_level, uint32_t last_level,
                         uint32_t row_stride[PIPE_MAX_TEXTURE_LEVELS],
                         uint32_t img_stride[PIPE_MAX_TEXTURE_LEVELS],
                         const void *data[PIPE_MAX_TEXTURE_LEVELS])
@@ -760,7 +760,7 @@ draw_set_mapped_texture(struct draw_context *draw,
    if(draw->llvm)
       draw_llvm_set_mapped_texture(draw,
                                 sampler_idx,
-                                width, height, depth, last_level,
+                                width, height, depth, first_level, last_level,
                                 row_stride, img_stride, data);
 #endif
 }

src/gallium/auxiliary/draw/draw_context.h

@@ -123,7 +123,7 @@ void
 draw_set_mapped_texture(struct draw_context *draw,
                         unsigned sampler_idx,
                         uint32_t width, uint32_t height, uint32_t depth,
-                        uint32_t last_level,
+                        uint32_t first_level, uint32_t last_level,
                         uint32_t row_stride[PIPE_MAX_TEXTURE_LEVELS],
                         uint32_t img_stride[PIPE_MAX_TEXTURE_LEVELS],
                         const void *data[PIPE_MAX_TEXTURE_LEVELS]);

src/gallium/auxiliary/draw/draw_llvm.c

@@ -106,6 +106,7 @@ create_jit_texture_type(struct gallivm_state *gallivm)
    elem_types[DRAW_JIT_TEXTURE_WIDTH]  =
    elem_types[DRAW_JIT_TEXTURE_HEIGHT] =
    elem_types[DRAW_JIT_TEXTURE_DEPTH] =
+   elem_types[DRAW_JIT_TEXTURE_FIRST_LEVEL] =
    elem_types[DRAW_JIT_TEXTURE_LAST_LEVEL] = int32_type;
    elem_types[DRAW_JIT_TEXTURE_ROW_STRIDE] =
    elem_types[DRAW_JIT_TEXTURE_IMG_STRIDE] =
@@ -136,6 +137,9 @@ create_jit_texture_type(struct gallivm_state *gallivm)
    LP_CHECK_MEMBER_OFFSET(struct draw_jit_texture, depth,
                           target, texture_type,
                           DRAW_JIT_TEXTURE_DEPTH);
+   LP_CHECK_MEMBER_OFFSET(struct draw_jit_texture, first_level,
+                          target, texture_type,
+                          DRAW_JIT_TEXTURE_FIRST_LEVEL);
    LP_CHECK_MEMBER_OFFSET(struct draw_jit_texture, last_level,
                           target, texture_type,
                           DRAW_JIT_TEXTURE_LAST_LEVEL);
@@ -438,7 +442,8 @@ generate_vs(struct draw_llvm *llvm,
             const LLVMValueRef (*inputs)[NUM_CHANNELS],
             LLVMValueRef system_values_array,
             LLVMValueRef context_ptr,
-            struct lp_build_sampler_soa *draw_sampler)
+            struct lp_build_sampler_soa *draw_sampler,
+            boolean clamp_vertex_color)
 {
    const struct tgsi_token *tokens = llvm->draw->vs.vertex_shader->state.tokens;
    struct lp_type vs_type;
@@ -474,6 +479,30 @@ generate_vs(struct draw_llvm *llvm,
                      outputs,
                      sampler,
                      &llvm->draw->vs.vertex_shader->info);
+
+   if(clamp_vertex_color)
+   {
+      LLVMValueRef out;
+      unsigned chan, attrib;
+      struct lp_build_context bld;
+      struct tgsi_shader_info* info = &llvm->draw->vs.vertex_shader->info;
+      lp_build_context_init(&bld, llvm->gallivm, vs_type);
+
+      for (attrib = 0; attrib < info->num_outputs; ++attrib) {
+         for(chan = 0; chan < NUM_CHANNELS; ++chan) {
+            if(outputs[attrib][chan]) {
+               switch (info->output_semantic_name[attrib]) {
+               case TGSI_SEMANTIC_COLOR:
+               case TGSI_SEMANTIC_BCOLOR:
+                  out = LLVMBuildLoad(builder, outputs[attrib][chan], "");
+                  out = lp_build_clamp(&bld, out, bld.zero, bld.one);
+                  LLVMBuildStore(builder, out, outputs[attrib][chan]);
+                  break;
+               }
+            }
+         }
+      }
+   }
 }
 
 #if DEBUG_STORE
@@ -1235,7 +1264,8 @@ draw_llvm_generate(struct draw_llvm *llvm, struct draw_llvm_variant *variant)
                   ptr_aos,
                   system_values_array,
                   context_ptr,
-                  sampler);
+                  sampler,
+                  variant->key.clamp_vertex_color);
 
       /* store original positions in clip before further manipulation */
       store_clip(gallivm, io, outputs);
@@ -1446,7 +1476,8 @@ draw_llvm_generate_elts(struct draw_llvm *llvm, struct draw_llvm_variant *varian
                   ptr_aos,
                   system_values_array,
                   context_ptr,
-                  sampler);
+                  sampler,
+                  variant->key.clamp_vertex_color);
 
       /* store original positions in clip before further manipulation */
       store_clip(gallivm, io, outputs);
@@ -1524,6 +1555,8 @@ draw_llvm_make_variant_key(struct draw_llvm *llvm, char *store)
 
    key = (struct draw_llvm_variant_key *)store;
 
+   key->clamp_vertex_color = llvm->draw->rasterizer->clamp_vertex_color; /**/
+
    /* Presumably all variants of the shader should have the same
     * number of vertex elements - ie the number of shader inputs.
     */
@@ -1566,7 +1599,7 @@ void
 draw_llvm_set_mapped_texture(struct draw_context *draw,
                              unsigned sampler_idx,
                              uint32_t width, uint32_t height, uint32_t depth,
-                             uint32_t last_level,
+                             uint32_t first_level, uint32_t last_level,
                              uint32_t row_stride[PIPE_MAX_TEXTURE_LEVELS],
                              uint32_t img_stride[PIPE_MAX_TEXTURE_LEVELS],
                              const void *data[PIPE_MAX_TEXTURE_LEVELS])
@@ -1582,9 +1615,10 @@ draw_llvm_set_mapped_texture(struct draw_context *draw,
    jit_tex->width = width;
    jit_tex->height = height;
    jit_tex->depth = depth;
+   jit_tex->first_level = first_level;
    jit_tex->last_level = last_level;
 
-   for (j = 0; j <= last_level; j++) {
+   for (j = first_level; j <= last_level; j++) {
       jit_tex->data[j] = data[j];
       jit_tex->row_stride[j] = row_stride[j];
       jit_tex->img_stride[j] = img_stride[j];

src/gallium/auxiliary/draw/draw_llvm.h

@@ -50,6 +50,7 @@ struct draw_jit_texture
    uint32_t width;
    uint32_t height;
    uint32_t depth;
+   uint32_t first_level;
    uint32_t last_level;
    uint32_t row_stride[PIPE_MAX_TEXTURE_LEVELS];
    uint32_t img_stride[PIPE_MAX_TEXTURE_LEVELS];
@@ -64,6 +65,7 @@ enum {
    DRAW_JIT_TEXTURE_WIDTH = 0,
    DRAW_JIT_TEXTURE_HEIGHT,
    DRAW_JIT_TEXTURE_DEPTH,
+   DRAW_JIT_TEXTURE_FIRST_LEVEL,
    DRAW_JIT_TEXTURE_LAST_LEVEL,
    DRAW_JIT_TEXTURE_ROW_STRIDE,
    DRAW_JIT_TEXTURE_IMG_STRIDE,
@@ -162,6 +164,7 @@ struct draw_llvm_variant_key
 {
    unsigned nr_vertex_elements:8;
    unsigned nr_samplers:8;
+   unsigned clamp_vertex_color:1;
    unsigned clip_xy:1;
    unsigned clip_z:1;
    unsigned clip_user:1;
@@ -169,7 +172,7 @@ struct draw_llvm_variant_key
    unsigned bypass_viewport:1;
    unsigned need_edgeflags:1;
    unsigned nr_planes:4;
-   unsigned pad:6;
+   unsigned pad:5;
 
    /* Variable number of vertex elements:
     */
@@ -295,7 +298,7 @@ void
 draw_llvm_set_mapped_texture(struct draw_context *draw,
                              unsigned sampler_idx,
                              uint32_t width, uint32_t height, uint32_t depth,
-                             uint32_t last_level,
+                             uint32_t first_level, uint32_t last_level,
                              uint32_t row_stride[PIPE_MAX_TEXTURE_LEVELS],
                              uint32_t img_stride[PIPE_MAX_TEXTURE_LEVELS],
                              const void *data[PIPE_MAX_TEXTURE_LEVELS]);

src/gallium/auxiliary/draw/draw_llvm_sample.c

@@ -144,6 +144,7 @@ draw_llvm_texture_member(const struct lp_sampler_dynamic_state *base,
 DRAW_LLVM_TEXTURE_MEMBER(width,      DRAW_JIT_TEXTURE_WIDTH, TRUE)
 DRAW_LLVM_TEXTURE_MEMBER(height,     DRAW_JIT_TEXTURE_HEIGHT, TRUE)
 DRAW_LLVM_TEXTURE_MEMBER(depth,      DRAW_JIT_TEXTURE_DEPTH, TRUE)
+DRAW_LLVM_TEXTURE_MEMBER(first_level,DRAW_JIT_TEXTURE_FIRST_LEVEL, TRUE)
 DRAW_LLVM_TEXTURE_MEMBER(last_level, DRAW_JIT_TEXTURE_LAST_LEVEL, TRUE)
 DRAW_LLVM_TEXTURE_MEMBER(row_stride, DRAW_JIT_TEXTURE_ROW_STRIDE, FALSE)
 DRAW_LLVM_TEXTURE_MEMBER(img_stride, DRAW_JIT_TEXTURE_IMG_STRIDE, FALSE)
@@ -209,6 +210,7 @@ draw_llvm_sampler_soa_create(const struct lp_sampler_static_state *static_state,
    sampler->dynamic_state.base.width = draw_llvm_texture_width;
    sampler->dynamic_state.base.height = draw_llvm_texture_height;
    sampler->dynamic_state.base.depth = draw_llvm_texture_depth;
+   sampler->dynamic_state.base.first_level = draw_llvm_texture_first_level;
    sampler->dynamic_state.base.last_level = draw_llvm_texture_last_level;
    sampler->dynamic_state.base.row_stride = draw_llvm_texture_row_stride;
    sampler->dynamic_state.base.img_stride = draw_llvm_texture_img_stride;

src/gallium/auxiliary/draw/draw_private.h

@@ -146,6 +146,14 @@ struct draw_context
       struct pipe_vertex_buffer vertex_buffer[PIPE_MAX_ATTRIBS];
       unsigned nr_vertex_buffers;
 
+      /*
+       * This is the largest legal index value for the current set of
+       * bound vertex buffers.  Regardless of any other consideration,
+       * all vertex lookups need to be clamped to 0..max_index to
+       * prevent out-of-bound access.
+       */
+      unsigned max_index;
+
       struct pipe_vertex_element vertex_element[PIPE_MAX_ATTRIBS];
       unsigned nr_vertex_elements;
 

src/gallium/auxiliary/draw/draw_pt.c

@@ -470,6 +470,17 @@ draw_vbo(struct draw_context *draw,
    if (0)
       draw_print_arrays(draw, info->mode, info->start, MIN2(info->count, 20));
 
+   draw->pt.max_index = util_draw_max_index(draw->pt.vertex_buffer,
+                                            draw->pt.nr_vertex_buffers,
+                                            draw->pt.vertex_element,
+                                            draw->pt.nr_vertex_elements,
+                                            info);
+
+   /*
+    * TODO: We could use draw->pt.max_index to further narrow
+    * the min_index/max_index hints given by the state tracker.
+    */
+
    for (instance = 0; instance < info->instance_count; instance++) {
       draw->instance_id = instance + info->start_instance;
 

src/gallium/auxiliary/draw/draw_pt_fetch.c

@@ -139,7 +139,7 @@ void draw_pt_fetch_run( struct pt_fetch *fetch,
 			    ((char *)draw->pt.user.vbuffer[i] + 
 			     draw->pt.vertex_buffer[i].buffer_offset),
 			    draw->pt.vertex_buffer[i].stride,
-			    draw->pt.user.max_index);
+			    draw->pt.max_index);
    }
 
    translate->run_elts( translate,

src/gallium/auxiliary/draw/draw_pt_fetch_emit.c

@@ -186,7 +186,7 @@ static void fetch_emit_prepare( struct draw_pt_middle_end *middle,
                                   ((char *)draw->pt.user.vbuffer[i] + 
                                    draw->pt.vertex_buffer[i].buffer_offset),
                                   draw->pt.vertex_buffer[i].stride,
-                                  draw->pt.user.max_index);
+                                  draw->pt.max_index);
    }
 
    *max_vertices = (draw->render->max_vertex_buffer_bytes / 

src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c

@@ -169,7 +169,7 @@ static void fse_prepare( struct draw_pt_middle_end *middle,
                                ((const ubyte *) draw->pt.user.vbuffer[i] + 
                                 draw->pt.vertex_buffer[i].buffer_offset),
                               draw->pt.vertex_buffer[i].stride,
-                              draw->pt.user.max_index );
+                              draw->pt.max_index );
    }
 
    *max_vertices = (draw->render->max_vertex_buffer_bytes / 

src/gallium/auxiliary/draw/draw_pt_vsplit.c

@@ -85,7 +85,12 @@ vsplit_flush_cache(struct vsplit_frontend *vsplit, unsigned flags)
 static INLINE void
 vsplit_add_cache(struct vsplit_frontend *vsplit, unsigned fetch)
 {
-   unsigned hash = fetch % MAP_SIZE;
+   struct draw_context *draw = vsplit->draw;
+   unsigned hash;
+
+   fetch = MIN2(fetch, draw->pt.max_index);
+
+   hash = fetch % MAP_SIZE;
 
    if (vsplit->cache.fetches[hash] != fetch) {
       /* update cache */

src/gallium/auxiliary/draw/draw_pt_vsplit_tmp.h

@@ -47,14 +47,18 @@ CONCAT(vsplit_primitive_, ELT_TYPE)(struct vsplit_frontend *vsplit,
    const ushort *draw_elts = NULL;
    unsigned i;
 
+   ib += istart;
+
    /* use the ib directly */
    if (min_index == 0 && sizeof(ib[0]) == sizeof(draw_elts[0])) {
       if (icount > vsplit->max_vertices)
          return FALSE;
 
       for (i = 0; i < icount; i++) {
-         ELT_TYPE idx = ib[istart + i];
-         assert(idx >= min_index && idx <= max_index);
+         ELT_TYPE idx = ib[i];
+            if (idx < min_index || idx > max_index) {
+            debug_printf("warning: index out of range\n");
+         }
       }
       draw_elts = (const ushort *) ib;
    }
@@ -83,17 +87,21 @@ CONCAT(vsplit_primitive_, ELT_TYPE)(struct vsplit_frontend *vsplit,
    if (!draw_elts) {
       if (min_index == 0) {
          for (i = 0; i < icount; i++) {
-            ELT_TYPE idx = ib[istart + i];
+            ELT_TYPE idx = ib[i];
 
-            assert(idx >= min_index && idx <= max_index);
+            if (idx < min_index || idx > max_index) {
+               debug_printf("warning: index out of range\n");
+	    }
             vsplit->draw_elts[i] = (ushort) idx;
          }
       }
       else {
          for (i = 0; i < icount; i++) {
-            ELT_TYPE idx = ib[istart + i];
+            ELT_TYPE idx = ib[i];
 
-            assert(idx >= min_index && idx <= max_index);
+            if (idx < min_index || idx > max_index) {
+               debug_printf("warning: index out of range\n");
+	    }
             vsplit->draw_elts[i] = (ushort) (idx - min_index);
          }
       }

src/gallium/auxiliary/draw/draw_vs.c

@@ -104,10 +104,17 @@ draw_create_vertex_shader(struct draw_context *draw,
    }
 
    if (!draw->pt.middle.llvm) {
+#if 0
+/* these paths don't support vertex clamping
+ * TODO: either add it, or remove them completely
+ * use LLVM instead if you want performance
+ * use exec instead if you want debugging/more correctness
+ */
 #if defined(PIPE_ARCH_X86)
       vs = draw_create_vs_sse( draw, shader );
 #elif defined(PIPE_ARCH_PPC)
       vs = draw_create_vs_ppc( draw, shader );
+#endif
 #endif
    }
 #if HAVE_LLVM

src/gallium/auxiliary/draw/draw_vs_exec.c

@@ -95,6 +95,7 @@ vs_exec_run_linear( struct draw_vertex_shader *shader,
    struct tgsi_exec_machine *machine = evs->machine;
    unsigned int i, j;
    unsigned slot;
+   boolean clamp_vertex_color = shader->draw->rasterizer->clamp_vertex_color;
 
    tgsi_exec_set_constant_buffers(machine, PIPE_MAX_CONSTANT_BUFFERS,
                                   constants, const_size);
@@ -151,11 +152,22 @@ vs_exec_run_linear( struct draw_vertex_shader *shader,
        */
       for (j = 0; j < max_vertices; j++) {
          for (slot = 0; slot < shader->info.num_outputs; slot++) {
-            output[slot][0] = machine->Outputs[slot].xyzw[0].f[j];
-            output[slot][1] = machine->Outputs[slot].xyzw[1].f[j];
-            output[slot][2] = machine->Outputs[slot].xyzw[2].f[j];
-            output[slot][3] = machine->Outputs[slot].xyzw[3].f[j];
-
+            unsigned name = shader->info.output_semantic_name[slot];
+            if(clamp_vertex_color &&
+                  (name == TGSI_SEMANTIC_COLOR || name == TGSI_SEMANTIC_BCOLOR))
+            {
+               output[slot][0] = CLAMP(machine->Outputs[slot].xyzw[0].f[j], 0.0f, 1.0f);
+               output[slot][1] = CLAMP(machine->Outputs[slot].xyzw[1].f[j], 0.0f, 1.0f);
+               output[slot][2] = CLAMP(machine->Outputs[slot].xyzw[2].f[j], 0.0f, 1.0f);
+               output[slot][3] = CLAMP(machine->Outputs[slot].xyzw[3].f[j], 0.0f, 1.0f);
+            }
+            else
+            {
+               output[slot][0] = machine->Outputs[slot].xyzw[0].f[j];
+               output[slot][1] = machine->Outputs[slot].xyzw[1].f[j];
+               output[slot][2] = machine->Outputs[slot].xyzw[2].f[j];
+               output[slot][3] = machine->Outputs[slot].xyzw[3].f[j];
+            }
          }
 
 #if 0

src/gallium/auxiliary/gallivm/lp_bld_debug.cpp

@@ -32,7 +32,12 @@
 #include <llvm/Target/TargetInstrInfo.h>
 #include <llvm/Support/raw_ostream.h>
 #include <llvm/Support/MemoryObject.h>
+
+#if HAVE_LLVM >= 0x0209
+#include <llvm/Support/Host.h>
+#else
 #include <llvm/System/Host.h>
+#endif
 
 #if HAVE_LLVM >= 0x0207
 #include <llvm/MC/MCDisassembler.h>
@@ -202,9 +207,17 @@ lp_disassemble(const void* func)
    }
 
    raw_debug_ostream Out;
+   TargetMachine *TM = T->createTargetMachine(Triple, "");
 
+#if HAVE_LLVM >= 0x0300
+   unsigned int AsmPrinterVariant = AsmInfo->getAssemblerDialect();
+#else
    int AsmPrinterVariant = AsmInfo->getAssemblerDialect();
-#if HAVE_LLVM >= 0x0208
+#endif
+#if HAVE_LLVM >= 0x0300
+   OwningPtr<MCInstPrinter> Printer(
+         T->createMCInstPrinter(*TM, AsmPrinterVariant, *AsmInfo));
+#elif HAVE_LLVM >= 0x0208
    OwningPtr<MCInstPrinter> Printer(
          T->createMCInstPrinter(AsmPrinterVariant, *AsmInfo));
 #else
@@ -216,8 +229,6 @@ lp_disassemble(const void* func)
       return;
    }
 
-   TargetMachine *TM = T->createTargetMachine(Triple, "");
-
    const TargetInstrInfo *TII = TM->getInstrInfo();
 
    /*

src/gallium/auxiliary/gallivm/lp_bld_sample.c

@@ -51,6 +51,10 @@
  */
 #define BRILINEAR_FACTOR 2
 
+static LLVMValueRef
+lp_build_minify(struct lp_build_context *bld,
+                LLVMValueRef base_size,
+                LLVMValueRef level);
 
 /**
  * Does the given texture wrap mode allow sampling the texture border color?
@@ -184,9 +188,11 @@ lp_sampler_static_state(struct lp_sampler_static_state *state,
  */
 static LLVMValueRef
 lp_build_rho(struct lp_build_sample_context *bld,
+             unsigned unit,
              const LLVMValueRef ddx[4],
              const LLVMValueRef ddy[4])
 {
+   struct lp_build_context *int_size_bld = &bld->int_size_bld;
    struct lp_build_context *float_size_bld = &bld->float_size_bld;
    struct lp_build_context *float_bld = &bld->float_bld;
    const unsigned dims = bld->dims;
@@ -198,8 +204,9 @@ lp_build_rho(struct lp_build_sample_context *bld,
    LLVMValueRef dsdx, dsdy, dtdx, dtdy, drdx, drdy;
    LLVMValueRef rho_x, rho_y;
    LLVMValueRef rho_vec;
-   LLVMValueRef float_size;
+   LLVMValueRef int_size, float_size;
    LLVMValueRef rho;
+   LLVMValueRef first_level, first_level_vec;
 
    dsdx = ddx[0];
    dsdy = ddy[0];
@@ -235,7 +242,11 @@ lp_build_rho(struct lp_build_sample_context *bld,
 
    rho_vec = lp_build_max(float_size_bld, rho_x, rho_y);
 
-   float_size = lp_build_int_to_float(float_size_bld, bld->int_size);
+   first_level = bld->dynamic_state->first_level(bld->dynamic_state,
+                                                 bld->gallivm, unit);
+   first_level_vec = lp_build_broadcast_scalar(&bld->int_size_bld, first_level);
+   int_size = lp_build_minify(int_size_bld, bld->int_size, first_level_vec);
+   float_size = lp_build_int_to_float(float_size_bld, int_size);
 
    rho_vec = lp_build_mul(float_size_bld, rho_vec, float_size);
 
@@ -442,7 +453,7 @@ lp_build_lod_selector(struct lp_build_sample_context *bld,
       else {
          LLVMValueRef rho;
 
-         rho = lp_build_rho(bld, ddx, ddy);
+         rho = lp_build_rho(bld, unit, ddx, ddy);
 
          /*
           * Compute lod = log2(rho)
@@ -542,18 +553,18 @@ lp_build_nearest_mip_level(struct lp_build_sample_context *bld,
                            LLVMValueRef *level_out)
 {
    struct lp_build_context *int_bld = &bld->int_bld;
-   LLVMValueRef last_level, level;
-
-   LLVMValueRef zero = lp_build_const_int32(bld->gallivm, 0);
+   LLVMValueRef first_level, last_level, level;
 
+   first_level = bld->dynamic_state->first_level(bld->dynamic_state,
+                                                 bld->gallivm, unit);
    last_level = bld->dynamic_state->last_level(bld->dynamic_state,
                                                bld->gallivm, unit);
 
    /* convert float lod to integer */
-   level = lod_ipart;
+   level = lp_build_add(int_bld, lod_ipart, first_level);
 
    /* clamp level to legal range of levels */
-   *level_out = lp_build_clamp(int_bld, level, zero, last_level);
+   *level_out = lp_build_clamp(int_bld, level, first_level, last_level);
 }
 
 
@@ -573,39 +584,42 @@ lp_build_linear_mip_levels(struct lp_build_sample_context *bld,
    LLVMBuilderRef builder = bld->gallivm->builder;
    struct lp_build_context *int_bld = &bld->int_bld;
    struct lp_build_context *float_bld = &bld->float_bld;
-   LLVMValueRef last_level;
+   LLVMValueRef first_level, last_level;
    LLVMValueRef clamp_min;
    LLVMValueRef clamp_max;
 
-   *level0_out = lod_ipart;
-   *level1_out = lp_build_add(int_bld, lod_ipart, int_bld->one);
+   first_level = bld->dynamic_state->first_level(bld->dynamic_state,
+                                                 bld->gallivm, unit);
+
+   *level0_out = lp_build_add(int_bld, lod_ipart, first_level);
+   *level1_out = lp_build_add(int_bld, *level0_out, int_bld->one);
 
    last_level = bld->dynamic_state->last_level(bld->dynamic_state,
                                                bld->gallivm, unit);
 
    /*
-    * Clamp both lod_ipart and lod_ipart + 1 to [0, last_level], with the
-    * minimum number of comparisons, and zeroing lod_fpart in the extreme
+    * Clamp both *level0_out and *level1_out to [first_level, last_level], with
+    * the minimum number of comparisons, and zeroing lod_fpart in the extreme
     * ends in the process.
     */
 
-   /* lod_ipart < 0 */
+   /* *level0_out < first_level */
    clamp_min = LLVMBuildICmp(builder, LLVMIntSLT,
-                             lod_ipart, int_bld->zero,
-                             "clamp_lod_to_zero");
+                             *level0_out, first_level,
+                             "clamp_lod_to_first");
 
    *level0_out = LLVMBuildSelect(builder, clamp_min,
-                                 int_bld->zero, *level0_out, "");
+                                 first_level, *level0_out, "");
 
    *level1_out = LLVMBuildSelect(builder, clamp_min,
-                                 int_bld->zero, *level1_out, "");
+                                 first_level, *level1_out, "");
 
    *lod_fpart_inout = LLVMBuildSelect(builder, clamp_min,
                                       float_bld->zero, *lod_fpart_inout, "");
 
-   /* lod_ipart >= last_level */
+   /* *level0_out >= last_level */
    clamp_max = LLVMBuildICmp(builder, LLVMIntSGE,
-                             lod_ipart, last_level,
+                             *level0_out, last_level,
                              "clamp_lod_to_last");
 
    *level0_out = LLVMBuildSelect(builder, clamp_max,

src/gallium/auxiliary/gallivm/lp_bld_sample.h

@@ -120,6 +120,12 @@ struct lp_sampler_dynamic_state
              struct gallivm_state *gallivm,
              unsigned unit);
 
+   /** Obtain the first mipmap level (base level) (returns int32) */
+   LLVMValueRef
+   (*first_level)( const struct lp_sampler_dynamic_state *state,
+                   struct gallivm_state *gallivm,
+                   unsigned unit);
+
    /** Obtain the number of mipmap levels minus one (returns int32) */
    LLVMValueRef
    (*last_level)( const struct lp_sampler_dynamic_state *state,

src/gallium/auxiliary/gallivm/lp_bld_sample_aos.c

@@ -939,6 +939,7 @@ lp_build_sample_aos(struct lp_build_sample_context *bld,
    LLVMValueRef unswizzled[4];
    LLVMValueRef face_ddx[4], face_ddy[4];
    struct lp_build_context h16_bld;
+   LLVMValueRef first_level;
    LLVMValueRef i32t_zero = lp_build_const_int32(bld->gallivm, 0);
 
    /* we only support the common/simple wrap modes at this time */
@@ -1008,7 +1009,9 @@ lp_build_sample_aos(struct lp_build_sample_context *bld,
          lp_build_nearest_mip_level(bld, unit, lod_ipart, &ilevel0);
       }
       else {
-         ilevel0 = i32t_zero;
+         first_level = bld->dynamic_state->first_level(bld->dynamic_state,
+                                                       bld->gallivm, unit);
+         ilevel0 = first_level;
       }
       break;
    case PIPE_TEX_MIPFILTER_NEAREST:
@@ -1065,7 +1068,7 @@ lp_build_sample_aos(struct lp_build_sample_context *bld,
          lp_build_sample_mipmap(bld, 
                                 mag_filter, PIPE_TEX_MIPFILTER_NONE,
                                 s, t, r,
-                                i32t_zero, NULL, NULL,
+                                ilevel0, NULL, NULL,
                                 packed_lo, packed_hi);
       }
       lp_build_endif(&if_ctx);

src/gallium/auxiliary/gallivm/lp_bld_sample_soa.c

@@ -943,6 +943,7 @@ lp_build_sample_general(struct lp_build_sample_context *bld,
    LLVMValueRef ilevel0, ilevel1 = NULL;
    LLVMValueRef face_ddx[4], face_ddy[4];
    LLVMValueRef texels[4];
+   LLVMValueRef first_level;
    LLVMValueRef i32t_zero = lp_build_const_int32(bld->gallivm, 0);
    unsigned chan;
 
@@ -1009,7 +1010,9 @@ lp_build_sample_general(struct lp_build_sample_context *bld,
          lp_build_nearest_mip_level(bld, unit, lod_ipart, &ilevel0);
       }
       else {
-         ilevel0 = i32t_zero;
+         first_level = bld->dynamic_state->first_level(bld->dynamic_state,
+                                                       bld->gallivm, unit);
+         ilevel0 = first_level;
       }
       break;
    case PIPE_TEX_MIPFILTER_NEAREST:
@@ -1068,7 +1071,7 @@ lp_build_sample_general(struct lp_build_sample_context *bld,
          lp_build_sample_mipmap(bld, unit,
                                 mag_filter, PIPE_TEX_MIPFILTER_NONE,
                                 s, t, r,
-                                i32t_zero, NULL, NULL,
+                                ilevel0, NULL, NULL,
                                 texels);
       }
       lp_build_endif(&if_ctx);

src/gallium/auxiliary/rtasm/rtasm_x86sse.c

@@ -726,6 +726,16 @@ void x86_movzx16(struct x86_function *p, struct x86_reg dst, struct x86_reg src
    emit_modrm(p, dst, src);
 }
 
+void x86_cmovcc( struct x86_function *p,
+                 struct x86_reg dst,
+                 struct x86_reg src,
+                 enum x86_cc cc)
+{
+   DUMP_RRI( dst, src, cc );
+   emit_2ub( p, 0x0f, 0x40 + cc );
+   emit_modrm( p, dst, src );
+}
+
 void x86_xor( struct x86_function *p,
 	      struct x86_reg dst,
 	      struct x86_reg src )

src/gallium/auxiliary/rtasm/rtasm_x86sse.h

@@ -309,6 +309,7 @@ void sse_movmskps( struct x86_function *p, struct x86_reg dst, struct x86_reg sr
 
 void x86_add( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void x86_and( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
+void x86_cmovcc( struct x86_function *p, struct x86_reg dst, struct x86_reg src, enum x86_cc cc );
 void x86_cmp( struct x86_function *p, struct x86_reg dst, struct x86_reg src );
 void x86_dec( struct x86_function *p, struct x86_reg reg );
 void x86_inc( struct x86_function *p, struct x86_reg reg );

src/gallium/auxiliary/translate/translate_sse.c

@@ -1067,6 +1067,8 @@ static boolean init_inputs( struct translate_sse *p,
       struct translate_buffer *buffer = &p->buffer[varient->buffer_index];
 
       if (!index_size || varient->instance_divisor) {
+         struct x86_reg buf_max_index = x86_make_disp(p->machine_EDI,
+                                                     get_offset(p, &buffer->max_index));
          struct x86_reg buf_stride   = x86_make_disp(p->machine_EDI,
                                                      get_offset(p, &buffer->stride));
          struct x86_reg buf_ptr      = x86_make_disp(p->machine_EDI,
@@ -1100,14 +1102,17 @@ static boolean init_inputs( struct translate_sse *p,
             x86_mov(p->func, tmp_EAX, elt);
          }
 
-         /*
-          * TODO: Respect translate_buffer::max_index.
+         /* Clamp to max_index
           */
+         x86_cmp(p->func, tmp_EAX, buf_max_index);
+         x86_cmovcc(p->func, tmp_EAX, buf_max_index, cc_AE);
 
          x86_imul(p->func, tmp_EAX, buf_stride);
          x64_rexw(p->func);
          x86_add(p->func, tmp_EAX, buf_base_ptr);
 
+         x86_cmp(p->func, p->count_EBP, p->tmp_EAX);
+
 
          /* In the linear case, keep the buffer pointer instead of the
           * index number.
@@ -1163,6 +1168,10 @@ static struct x86_reg get_buffer_ptr( struct translate_sse *p,
          x86_make_disp(p->machine_EDI,
                        get_offset(p, &p->buffer[varient->buffer_index].base_ptr));
 
+      struct x86_reg buf_max_index =
+         x86_make_disp(p->machine_EDI,
+                       get_offset(p, &p->buffer[varient->buffer_index].max_index));
+
 
 
       /* Calculate pointer to current attrib:
@@ -1179,6 +1188,12 @@ static struct x86_reg get_buffer_ptr( struct translate_sse *p,
          x86_mov(p->func, ptr, elt);
          break;
       }
+
+      /* Clamp to max_index
+       */
+      x86_cmp(p->func, ptr, buf_max_index);
+      x86_cmovcc(p->func, ptr, buf_max_index, cc_AE);
+
       x86_imul(p->func, ptr, buf_stride);
       x64_rexw(p->func);
       x86_add(p->func, ptr, buf_base_ptr);

src/gallium/auxiliary/util/u_cpu_detect.c

@@ -270,13 +270,15 @@ util_cpu_detect(void)
              util_cpu_caps.x86_cpu_type = 8 + ((regs2[0] >> 20) & 255); /* use extended family (P4, IA64) */
 
          /* general feature flags */
-         util_cpu_caps.has_tsc    = (regs2[3] & (1 << 8  )) >>  8; /* 0x0000010 */
-         util_cpu_caps.has_mmx    = (regs2[3] & (1 << 23 )) >> 23; /* 0x0800000 */
-         util_cpu_caps.has_sse    = (regs2[3] & (1 << 25 )) >> 25; /* 0x2000000 */
-         util_cpu_caps.has_sse2   = (regs2[3] & (1 << 26 )) >> 26; /* 0x4000000 */
-         util_cpu_caps.has_sse3   = (regs2[2] & (1));          /* 0x0000001 */
-         util_cpu_caps.has_ssse3  = (regs2[2] & (1 << 9 )) >> 9;   /* 0x0000020 */
-         util_cpu_caps.has_sse4_1 = (regs2[2] & (1 << 19)) >> 19;
+         util_cpu_caps.has_tsc    = (regs2[3] >>  8) & 1; /* 0x0000010 */
+         util_cpu_caps.has_mmx    = (regs2[3] >> 23) & 1; /* 0x0800000 */
+         util_cpu_caps.has_sse    = (regs2[3] >> 25) & 1; /* 0x2000000 */
+         util_cpu_caps.has_sse2   = (regs2[3] >> 26) & 1; /* 0x4000000 */
+         util_cpu_caps.has_sse3   = (regs2[2] >>  0) & 1; /* 0x0000001 */
+         util_cpu_caps.has_ssse3  = (regs2[2] >>  9) & 1; /* 0x0000020 */
+         util_cpu_caps.has_sse4_1 = (regs2[2] >> 19) & 1;
+         util_cpu_caps.has_sse4_2 = (regs2[2] >> 20) & 1;
+         util_cpu_caps.has_avx    = (regs2[2] >> 28) & 1;
          util_cpu_caps.has_mmx2   = util_cpu_caps.has_sse; /* SSE cpus supports mmxext too */
 
          cacheline = ((regs2[1] >> 8) & 0xFF) * 8;
@@ -290,10 +292,10 @@ util_cpu_detect(void)
 
          cpuid(0x80000001, regs2);
 
-         util_cpu_caps.has_mmx  |= (regs2[3] & (1 << 23 )) >> 23; /* 0x0800000 */
-         util_cpu_caps.has_mmx2 |= (regs2[3] & (1 << 22 )) >> 22; /* 0x400000 */
-         util_cpu_caps.has_3dnow    = (regs2[3] & (1 << 31 )) >> 31; /* 0x80000000 */
-         util_cpu_caps.has_3dnow_ext = (regs2[3] & (1 << 30 )) >> 30;
+         util_cpu_caps.has_mmx  |= (regs2[3] >> 23) & 1;
+         util_cpu_caps.has_mmx2 |= (regs2[3] >> 22) & 1;
+         util_cpu_caps.has_3dnow = (regs2[3] >> 31) & 1;
+         util_cpu_caps.has_3dnow_ext = (regs2[3] >> 30) & 1;
       }
 
       if (regs[0] >= 0x80000006) {
@@ -329,6 +331,8 @@ util_cpu_detect(void)
       debug_printf("util_cpu_caps.has_sse3 = %u\n", util_cpu_caps.has_sse3);
       debug_printf("util_cpu_caps.has_ssse3 = %u\n", util_cpu_caps.has_ssse3);
       debug_printf("util_cpu_caps.has_sse4_1 = %u\n", util_cpu_caps.has_sse4_1);
+      debug_printf("util_cpu_caps.has_sse4_2 = %u\n", util_cpu_caps.has_sse4_2);
+      debug_printf("util_cpu_caps.has_avx = %u\n", util_cpu_caps.has_avx);
       debug_printf("util_cpu_caps.has_3dnow = %u\n", util_cpu_caps.has_3dnow);
       debug_printf("util_cpu_caps.has_3dnow_ext = %u\n", util_cpu_caps.has_3dnow_ext);
       debug_printf("util_cpu_caps.has_altivec = %u\n", util_cpu_caps.has_altivec);

src/gallium/auxiliary/util/u_cpu_detect.h

@@ -53,6 +53,8 @@ struct util_cpu_caps {
    unsigned has_sse3:1;
    unsigned has_ssse3:1;
    unsigned has_sse4_1:1;
+   unsigned has_sse4_2:1;
+   unsigned has_avx:1;
    unsigned has_3dnow:1;
    unsigned has_3dnow_ext:1;
    unsigned has_altivec:1;

src/gallium/auxiliary/util/u_double_list.h

@@ -39,6 +39,7 @@
 
 
 #include <stddef.h>
+#include "pipe/p_compiler.h"
 
 
 struct list_head
@@ -47,50 +48,56 @@ struct list_head
     struct list_head *next;
 };
 
+static INLINE void list_inithead(struct list_head *item)
+{
+    item->prev = item;
+    item->next = item;
+}
 
-#define LIST_INITHEAD(__item)			\
-  do {						\
-    (__item)->prev = (__item);			\
-    (__item)->next = (__item);			\
-  } while (0)
+static INLINE void list_add(struct list_head *item, struct list_head *list)
+{
+    item->prev = list;
+    item->next = list->next;
+    list->next->prev = item;
+    list->next = item;
+}
 
-#define LIST_ADD(__item, __list)		\
-  do {						\
-    (__item)->prev = (__list);			\
-    (__item)->next = (__list)->next;		\
-    (__list)->next->prev = (__item);		\
-    (__list)->next = (__item);			\
-  } while (0)
+static INLINE void list_addtail(struct list_head *item, struct list_head *list)
+{
+    item->next = list;
+    item->prev = list->prev;
+    list->prev->next = item;
+    list->prev = item;
+}
 
-#define LIST_ADDTAIL(__item, __list)		\
-  do {						\
-    (__item)->next = (__list);			\
-    (__item)->prev = (__list)->prev;		\
-    (__list)->prev->next = (__item);		\
-    (__list)->prev = (__item);			\
-  } while(0)
+static INLINE void list_replace(struct list_head *from, struct list_head *to)
+{
+    to->prev = from->prev;
+    to->next = from->next;
+    from->next->prev = to;
+    from->prev->next = to;
+}
 
-#define LIST_REPLACE(__from, __to)		\
-  do {						\
-    (__to)->prev = (__from)->prev;		\
-    (__to)->next = (__from)->next;		\
-    (__from)->next->prev = (__to);		\
-    (__from)->prev->next = (__to);		\
-  } while (0)
+static INLINE void list_del(struct list_head *item)
+{
+    item->prev->next = item->next;
+    item->next->prev = item->prev;
+}
 
-#define LIST_DEL(__item)			\
-  do {						\
-    (__item)->prev->next = (__item)->next;	\
-    (__item)->next->prev = (__item)->prev;	\
-  } while(0)
+static INLINE void list_delinit(struct list_head *item)
+{
+    item->prev->next = item->next;
+    item->next->prev = item->prev;
+    item->next = item;
+    item->prev = item;
+}
 
-#define LIST_DELINIT(__item)			\
-  do {						\
-    (__item)->prev->next = (__item)->next;	\
-    (__item)->next->prev = (__item)->prev;	\
-    (__item)->next = (__item);			\
-    (__item)->prev = (__item);			\
-  } while(0)
+#define LIST_INITHEAD(__item) list_inithead(__item)
+#define LIST_ADD(__item, __list) list_add(__item, __list)
+#define LIST_ADDTAIL(__item, __list) list_addtail(__item, __list)
+#define LIST_REPLACE(__from, __to) list_replace(__from, __to)
+#define LIST_DEL(__item) list_del(__item)
+#define LIST_DELINIT(__item) list_delinit(__item)
 
 #define LIST_ENTRY(__type, __item, __field)   \
     ((__type *)(((char *)(__item)) - offsetof(__type, __field)))
@@ -114,4 +121,21 @@ struct list_head
 	storage = container_of(pos->member.next, pos, member);	\
 	&pos->member != (head);						\
 	pos = storage, storage = container_of(storage->member.next, storage, member))
+
+#define LIST_FOR_EACH_ENTRY_SAFE_REV(pos, storage, head, member)	\
+   for (pos = container_of((head)->prev, pos, member),			\
+	storage = container_of(pos->member.prev, pos, member);		\
+	&pos->member != (head);						\
+	pos = storage, storage = container_of(storage->member.prev, storage, member))
+
+#define LIST_FOR_EACH_ENTRY_FROM(pos, start, head, member)		\
+   for (pos = container_of((start), pos, member);			\
+	&pos->member != (head);						\
+	pos = container_of(pos->member.next, pos, member))
+
+#define LIST_FOR_EACH_ENTRY_FROM_REV(pos, start, head, member)		\
+   for (pos = container_of((start), pos, member);			\
+	&pos->member != (head);						\
+	pos = container_of(pos->member.prev, pos, member))
+
 #endif /*_U_DOUBLE_LIST_H_*/

src/gallium/auxiliary/util/u_draw.c

@@ -0,0 +1,99 @@
+/**************************************************************************
+ *
+ * Copyright 2011 VMware, Inc.
+ * All Rights Reserved.
+ *
+ * Permission is hereby granted, free of charge, to any person obtaining a
+ * copy of this software and associated documentation files (the
+ * "Software"), to deal in the Software without restriction, including
+ * without limitation the rights to use, copy, modify, merge, publish,
+ * distribute, sub license, and/or sell copies of the Software, and to
+ * permit persons to whom the Software is furnished to do so, subject to
+ * the following conditions:
+ *
+ * The above copyright notice and this permission notice (including the
+ * next paragraph) shall be included in all copies or substantial portions
+ * of the Software.
+ *
+ * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS
+ * OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF
+ * MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NON-INFRINGEMENT.
+ * IN NO EVENT SHALL VMWARE AND/OR ITS SUPPLIERS BE LIABLE FOR
+ * ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT,
+ * TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE
+ * SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
+ *
+ **************************************************************************/
+
+
+#include "util/u_debug.h"
+#include "util/u_math.h"
+#include "util/u_format.h"
+#include "util/u_draw.h"
+
+
+/**
+ * Returns the largest legal index value for the current set of bound vertex
+ * buffers.  Regardless of any other consideration, all vertex lookups need to
+ * be clamped to 0..max_index to prevent an out-of-bound access.
+ */
+unsigned
+util_draw_max_index(
+      const struct pipe_vertex_buffer *vertex_buffers,
+      unsigned nr_vertex_buffers,
+      const struct pipe_vertex_element *vertex_elements,
+      unsigned nr_vertex_elements,
+      const struct pipe_draw_info *info)
+{
+   unsigned max_index;
+   unsigned i;
+
+   max_index = ~0;
+   for (i = 0; i < nr_vertex_elements; i++) {
+      const struct pipe_vertex_element *element =
+         &vertex_elements[i];
+      const struct pipe_vertex_buffer *buffer =
+         &vertex_buffers[element->vertex_buffer_index];
+      unsigned buffer_size;
+      const struct util_format_description *format_desc;
+      unsigned format_size;
+
+      assert(buffer->buffer->height0 == 1);
+      assert(buffer->buffer->depth0 == 1);
+      buffer_size = buffer->buffer->width0;
+
+      format_desc = util_format_description(element->src_format);
+      assert(format_desc->block.width == 1);
+      assert(format_desc->block.height == 1);
+      assert(format_desc->block.bits % 8 == 0);
+      format_size = format_desc->block.bits/8;
+
+      assert(buffer_size - buffer->buffer_offset <= buffer_size);
+      buffer_size -= buffer->buffer_offset;
+
+      assert(buffer_size - element->src_offset <= buffer_size);
+      buffer_size -= element->src_offset;
+
+      assert(buffer_size - format_size <= buffer_size);
+      buffer_size -= format_size;
+
+      if (buffer->stride != 0) {
+         unsigned buffer_max_index;
+
+         buffer_max_index = buffer_size / buffer->stride;
+
+         if (element->instance_divisor == 0) {
+            /* Per-vertex data */
+            max_index = MIN2(max_index, buffer_max_index);
+         }
+         else {
+            /* Per-instance data. Simply make sure the state tracker didn't
+             * request more instances than those that fit in the buffer */
+            assert((info->start_instance + info->instance_count)/element->instance_divisor
+                   <= (buffer_max_index + 1));
+         }
+      }
+   }
+
+   return max_index;
+}

src/gallium/auxiliary/util/u_draw.h

@@ -34,6 +34,11 @@
 #include "pipe/p_state.h"
 
 
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+
 static INLINE void
 util_draw_init_info(struct pipe_draw_info *info)
 {
@@ -136,4 +141,18 @@ util_draw_range_elements(struct pipe_context *pipe,
    pipe->draw_vbo(pipe, &info);
 }
 
+
+unsigned
+util_draw_max_index(
+      const struct pipe_vertex_buffer *vertex_buffers,
+      unsigned nr_vertex_buffers,
+      const struct pipe_vertex_element *vertex_elements,
+      unsigned nr_vertex_elements,
+      const struct pipe_draw_info *info);
+
+
+#ifdef __cplusplus
+}
 #endif
+
+#endif /* !U_DRAW_H */

src/gallium/auxiliary/util/u_format.csv

@@ -88,6 +88,14 @@ PIPE_FORMAT_L16_UNORM             , plain, 1, 1, un16,     ,     ,     , xxx1, r
 PIPE_FORMAT_A16_UNORM             , plain, 1, 1, un16,     ,     ,     , 000x, rgb
 PIPE_FORMAT_I16_UNORM             , plain, 1, 1, un16,     ,     ,     , xxxx, rgb
 PIPE_FORMAT_L16A16_UNORM          , plain, 1, 1, un16, un16,     ,     , xxxy, rgb
+PIPE_FORMAT_A8_SNORM              , plain, 1, 1, sn8 ,     ,     ,     , 000x, rgb
+PIPE_FORMAT_L8_SNORM              , plain, 1, 1, sn8 ,     ,     ,     , xxx1, rgb
+PIPE_FORMAT_L8A8_SNORM            , plain, 1, 1, sn8 , sn8 ,     ,     , xxxy, rgb
+PIPE_FORMAT_I8_SNORM              , plain, 1, 1, sn8 ,     ,     ,     , xxxx, rgb
+PIPE_FORMAT_A16_SNORM             , plain, 1, 1, sn16,     ,     ,     , 000x, rgb
+PIPE_FORMAT_L16_SNORM             , plain, 1, 1, sn16,     ,     ,     , xxx1, rgb
+PIPE_FORMAT_L16A16_SNORM          , plain, 1, 1, sn16, sn16,     ,     , xxxy, rgb
+PIPE_FORMAT_I16_SNORM             , plain, 1, 1, sn16,     ,     ,     , xxxx, rgb
 
 # SRGB formats
 PIPE_FORMAT_L8_SRGB               , plain, 1, 1, un8 ,     ,     ,     , xxx1, srgb 

src/gallium/auxiliary/util/u_vbuf_mgr.c

@@ -499,7 +499,7 @@ void u_vbuf_mgr_set_vertex_buffers(struct u_vbuf_mgr *mgrb,
       /* Update the maximum index. */
       mgr->b.max_index =
             MIN2(mgr->b.max_index,
-                 (vb->buffer->width0 - vb->buffer_offset) / vb->stride);
+                 (vb->buffer->width0 - vb->buffer_offset) / vb->stride - 1);
    }
 
    for (; i < mgr->b.nr_real_vertex_buffers; i++) {
@@ -572,7 +572,11 @@ void u_vbuf_mgr_draw_begin(struct u_vbuf_mgr *mgrb,
    int min_index, max_index;
 
    min_index = info->min_index - info->index_bias;
-   max_index = info->max_index - info->index_bias;
+   if (info->max_index == ~0) {
+      max_index = mgr->b.max_index;
+   } else {
+      max_index = MIN2(info->max_index - info->index_bias, mgr->b.max_index);
+   }
 
    /* Translate vertices with non-native layouts or formats. */
    if (mgr->incompatible_vb_layout || mgr->ve->incompatible_layout) {

src/gallium/auxiliary/util/u_vbuf_mgr.h

@@ -53,7 +53,7 @@ struct u_vbuf_mgr {
    int nr_real_vertex_buffers;
 
    /* Precomputed max_index for hardware vertex buffers. */
-   int max_index;
+   unsigned max_index;
 
    /* This uploader can optionally be used by the driver.
     *

src/gallium/docs/source/cso/rasterizer.rst

@@ -7,6 +7,30 @@ The rasterizer state controls the rendering of points, lines and triangles.
 Attributes include polygon culling state, line width, line stipple,
 multisample state, scissoring and flat/smooth shading.
 
+Linkage
+
+clamp_vertex_color
+^^^^^^^^^^^^^^^^^^
+
+If set, TGSI_SEMANTIC_COLOR registers are clamped to the [0, 1] range after
+the execution of the vertex shader, before being passed to the geometry
+shader or fragment shader.
+
+OpenGL: glClampColor(GL_CLAMP_VERTEX_COLOR) in GL 3.0 or GL_ARB_color_buffer_float
+
+D3D11: seems always disabled
+
+clamp_fragment_color
+^^^^^^^^^^^^^^^^^^^^
+
+Controls whether TGSI_SEMANTIC_COLOR outputs of the fragment shader
+are clamped to [0, 1].
+
+OpenGL: glClampColor(GL_CLAMP_FRAGMENT_COLOR) in GL 3.0 or ARB_color_buffer_float
+
+D3D11: seems always disabled
+
+
 Shading
 -------
 

src/gallium/drivers/llvmpipe/lp_jit.c

@@ -54,6 +54,7 @@ lp_jit_create_types(struct llvmpipe_context *lp)
       elem_types[LP_JIT_TEXTURE_WIDTH]  =
       elem_types[LP_JIT_TEXTURE_HEIGHT] =
       elem_types[LP_JIT_TEXTURE_DEPTH] =
+      elem_types[LP_JIT_TEXTURE_FIRST_LEVEL] =
       elem_types[LP_JIT_TEXTURE_LAST_LEVEL] =  LLVMInt32TypeInContext(lc);
       elem_types[LP_JIT_TEXTURE_ROW_STRIDE] =
       elem_types[LP_JIT_TEXTURE_IMG_STRIDE] =
@@ -81,6 +82,9 @@ lp_jit_create_types(struct llvmpipe_context *lp)
       LP_CHECK_MEMBER_OFFSET(struct lp_jit_texture, depth,
                              gallivm->target, texture_type,
                              LP_JIT_TEXTURE_DEPTH);
+      LP_CHECK_MEMBER_OFFSET(struct lp_jit_texture, first_level,
+                             gallivm->target, texture_type,
+                             LP_JIT_TEXTURE_FIRST_LEVEL);
       LP_CHECK_MEMBER_OFFSET(struct lp_jit_texture, last_level,
                              gallivm->target, texture_type,
                              LP_JIT_TEXTURE_LAST_LEVEL);

src/gallium/drivers/llvmpipe/lp_jit.h

@@ -50,6 +50,7 @@ struct lp_jit_texture
    uint32_t width;
    uint32_t height;
    uint32_t depth;
+   uint32_t first_level;
    uint32_t last_level;
    uint32_t row_stride[LP_MAX_TEXTURE_LEVELS];
    uint32_t img_stride[LP_MAX_TEXTURE_LEVELS];
@@ -66,6 +67,7 @@ enum {
    LP_JIT_TEXTURE_WIDTH = 0,
    LP_JIT_TEXTURE_HEIGHT,
    LP_JIT_TEXTURE_DEPTH,
+   LP_JIT_TEXTURE_FIRST_LEVEL,
    LP_JIT_TEXTURE_LAST_LEVEL,
    LP_JIT_TEXTURE_ROW_STRIDE,
    LP_JIT_TEXTURE_IMG_STRIDE,

src/gallium/drivers/llvmpipe/lp_screen.c

@@ -166,6 +166,7 @@ llvmpipe_get_param(struct pipe_screen *screen, enum pipe_cap param)
       return 0;
    case PIPE_CAP_TGSI_INSTANCEID:
    case PIPE_CAP_VERTEX_ELEMENT_INSTANCE_DIVISOR:
+   case PIPE_CAP_MIXED_COLORBUFFER_FORMATS:
       return 1;
    default:
       return 0;

src/gallium/drivers/llvmpipe/lp_setup.c

@@ -653,6 +653,7 @@ lp_setup_set_fragment_sampler_views(struct lp_setup_context *setup,
          jit_tex->width = tex->width0;
          jit_tex->height = tex->height0;
          jit_tex->depth = tex->depth0;
+         jit_tex->first_level = view->u.tex.first_level;
          jit_tex->last_level = tex->last_level;
 
          /* We're referencing the texture's internal data, so save a
@@ -663,7 +664,7 @@ lp_setup_set_fragment_sampler_views(struct lp_setup_context *setup,
          if (!lp_tex->dt) {
             /* regular texture - setup array of mipmap level pointers */
             int j;
-            for (j = 0; j <= tex->last_level; j++) {
+            for (j = view->u.tex.first_level; j <= tex->last_level; j++) {
                jit_tex->data[j] =
                   llvmpipe_get_texture_image_all(lp_tex, j, LP_TEX_USAGE_READ,
                                                  LP_TEX_LAYOUT_LINEAR);
@@ -677,6 +678,7 @@ lp_setup_set_fragment_sampler_views(struct lp_setup_context *setup,
                   jit_tex->width = TILE_SIZE/8;
                   jit_tex->height = TILE_SIZE/8;
                   jit_tex->depth = 1;
+                  jit_tex->first_level = 0;
                   jit_tex->last_level = 0;
                   jit_tex->row_stride[j] = 0;
                   jit_tex->img_stride[j] = 0;

src/gallium/drivers/llvmpipe/lp_state_sampler.c

@@ -269,7 +269,7 @@ llvmpipe_prepare_vertex_sampling(struct llvmpipe_context *lp,
          if (!lp_tex->dt) {
             /* regular texture - setup array of mipmap level pointers */
             int j;
-            for (j = 0; j <= tex->last_level; j++) {
+            for (j = view->u.tex.first_level; j <= tex->last_level; j++) {
                data[j] =
                   llvmpipe_get_texture_image_all(lp_tex, j, LP_TEX_USAGE_READ,
                                                  LP_TEX_LAYOUT_LINEAR);
@@ -293,7 +293,7 @@ llvmpipe_prepare_vertex_sampling(struct llvmpipe_context *lp,
          draw_set_mapped_texture(lp->draw,
                                  i,
                                  tex->width0, tex->height0, tex->depth0,
-                                 tex->last_level,
+                                 view->u.tex.first_level, tex->last_level,
                                  row_stride, img_stride, data);
       }
    }

src/gallium/drivers/llvmpipe/lp_tex_sample.c

@@ -149,6 +149,7 @@ lp_llvm_texture_member(const struct lp_sampler_dynamic_state *base,
 LP_LLVM_TEXTURE_MEMBER(width,      LP_JIT_TEXTURE_WIDTH, TRUE)
 LP_LLVM_TEXTURE_MEMBER(height,     LP_JIT_TEXTURE_HEIGHT, TRUE)
 LP_LLVM_TEXTURE_MEMBER(depth,      LP_JIT_TEXTURE_DEPTH, TRUE)
+LP_LLVM_TEXTURE_MEMBER(first_level, LP_JIT_TEXTURE_FIRST_LEVEL, TRUE)
 LP_LLVM_TEXTURE_MEMBER(last_level, LP_JIT_TEXTURE_LAST_LEVEL, TRUE)
 LP_LLVM_TEXTURE_MEMBER(row_stride, LP_JIT_TEXTURE_ROW_STRIDE, FALSE)
 LP_LLVM_TEXTURE_MEMBER(img_stride, LP_JIT_TEXTURE_IMG_STRIDE, FALSE)
@@ -219,6 +220,7 @@ lp_llvm_sampler_soa_create(const struct lp_sampler_static_state *static_state,
    sampler->dynamic_state.base.width = lp_llvm_texture_width;
    sampler->dynamic_state.base.height = lp_llvm_texture_height;
    sampler->dynamic_state.base.depth = lp_llvm_texture_depth;
+   sampler->dynamic_state.base.first_level = lp_llvm_texture_first_level;
    sampler->dynamic_state.base.last_level = lp_llvm_texture_last_level;
    sampler->dynamic_state.base.row_stride = lp_llvm_texture_row_stride;
    sampler->dynamic_state.base.img_stride = lp_llvm_texture_img_stride;

src/gallium/drivers/noop/noop_pipe.c

@@ -337,6 +337,7 @@ static int noop_get_param(struct pipe_screen* pscreen, enum pipe_cap param)
 	case PIPE_CAP_TEXTURE_SHADOW_MAP:
 	case PIPE_CAP_TEXTURE_SWIZZLE:
 	case PIPE_CAP_BLEND_EQUATION_SEPARATE:
+	case PIPE_CAP_MIXED_COLORBUFFER_FORMATS:
 
 	  return 1;
 	case PIPE_CAP_DUAL_SOURCE_BLEND:

src/gallium/drivers/nouveau/nouveau_screen.c

@@ -12,6 +12,7 @@
 #include <errno.h>
 
 #include "nouveau/nouveau_bo.h"
+#include "nouveau/nouveau_mm.h"
 #include "nouveau_winsys.h"
 #include "nouveau_screen.h"
 #include "nouveau_fence.h"

src/gallium/drivers/nv50/nv50_3d.xml.h

@@ -558,7 +558,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #define NV50_3D_UNK0F8C						0x00000f8c
 
-#define NV50_3D_UNK0F90						0x00000f90
+#define NV50_3D_COLOR_MASK_COMMON				0x00000f90
 
 #define NV50_3D_UNK0F94						0x00000f94
 
@@ -1007,7 +1007,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define NV50_3D_TEX_CACHE_CTL_UNK1__MASK			0x00000030
 #define NV50_3D_TEX_CACHE_CTL_UNK1__SHIFT			4
 
-#define NV50_3D_UNK133C						0x0000133c
+#define NV50_3D_BLEND_SEPARATE_ALPHA				0x0000133c
 
 #define NV50_3D_BLEND_EQUATION_RGB				0x00001340
 #define NV50_3D_BLEND_EQUATION_RGB_FUNC_ADD			0x00008006
@@ -1033,7 +1033,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #define NV50_3D_BLEND_FUNC_DST_ALPHA				0x00001358
 
-#define NV50_3D_UNK135C						0x0000135c
+#define NV50_3D_BLEND_ENABLE_COMMON				0x0000135c
 
 #define NV50_3D_BLEND_ENABLE(i0)			       (0x00001360 + 0x4*(i0))
 #define NV50_3D_BLEND_ENABLE__ESIZE				0x00000004
@@ -1673,7 +1673,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define NV50_3D_MAP_SEMANTIC_0_BFC0_ID__SHIFT			8
 #define NV50_3D_MAP_SEMANTIC_0_COLR_NR__MASK			0x00ff0000
 #define NV50_3D_MAP_SEMANTIC_0_COLR_NR__SHIFT			16
-#define NV50_3D_MAP_SEMANTIC_0_CLMP_EN				0xff000000
+#define NV50_3D_MAP_SEMANTIC_0_CLMP_EN				0x01000000
 
 #define NV50_3D_MAP_SEMANTIC_1					0x00001908
 #define NV50_3D_MAP_SEMANTIC_1_CLIP_START__MASK			0x000000ff
@@ -1706,7 +1706,7 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define NV50_3D_CULL_FACE_BACK					0x00000405
 #define NV50_3D_CULL_FACE_FRONT_AND_BACK			0x00000408
 
-#define NV50_3D_UNK1924						0x00001924
+#define NV50_3D_LINE_LAST_PIXEL					0x00001924
 
 #define NVA3_3D_FP_MULTISAMPLE					0x00001928
 #define NVA3_3D_FP_MULTISAMPLE_EXPORT_SAMPLE_MASK		0x00000001

src/gallium/drivers/nv50/nv50_context.c

@@ -46,6 +46,17 @@ nv50_flush(struct pipe_context *pipe,
    FIRE_RING(screen->channel);
 }
 
+static void
+nv50_texture_barrier(struct pipe_context *pipe)
+{
+   struct nouveau_channel *chan = nv50_context(pipe)->screen->base.channel;
+
+   BEGIN_RING(chan, RING_3D(SERIALIZE), 1);
+   OUT_RING  (chan, 0);
+   BEGIN_RING(chan, RING_3D(TEX_CACHE_CTL), 1);
+   OUT_RING  (chan, 0x20);
+}
+
 void
 nv50_default_flush_notify(struct nouveau_channel *chan)
 {
@@ -125,6 +136,7 @@ nv50_create(struct pipe_screen *pscreen, void *priv)
    pipe->clear = nv50_clear;
 
    pipe->flush = nv50_flush;
+   pipe->texture_barrier = nv50_texture_barrier;
 
    if (!screen->cur_ctx)
       screen->cur_ctx = nv50;

src/gallium/drivers/nv50/nv50_context.h

@@ -1,7 +1,6 @@
 #ifndef __NV50_CONTEXT_H__
 #define __NV50_CONTEXT_H__
 
-#include <stdio.h>
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_state.h"
@@ -13,6 +12,7 @@
 
 #include "draw/draw_vertex.h"
 
+#include "nv50_debug.h"
 #include "nv50_winsys.h"
 #include "nv50_stateobj.h"
 #include "nv50_screen.h"
@@ -26,15 +26,6 @@
 #include "nv50_3d.xml.h"
 #include "nv50_2d.xml.h"
 
-#define NOUVEAU_ERR(fmt, args...) \
-   fprintf(stderr, "%s:%d -  "fmt, __FUNCTION__, __LINE__, ##args);
-
-#ifdef NOUVEAU_DEBUG
-# define NOUVEAU_DBG(args...) printf(args);
-#else
-# define NOUVEAU_DBG(args...)
-#endif
-
 #define NV50_NEW_BLEND        (1 << 0)
 #define NV50_NEW_RASTERIZER   (1 << 1)
 #define NV50_NEW_ZSA          (1 << 2)
@@ -80,6 +71,8 @@ struct nv50_context {
       uint32_t instance_elts; /* bitmask of per-instance elements */
       uint32_t instance_base;
       uint32_t interpolant_ctrl;
+      uint32_t semantic_color;
+      uint32_t semantic_psize;
       int32_t index_bias;
       boolean prim_restart;
       boolean point_sprite;
@@ -183,7 +176,7 @@ void nv50_fragprog_validate(struct nv50_context *);
 void nv50_fp_linkage_validate(struct nv50_context *);
 void nv50_gp_linkage_validate(struct nv50_context *);
 void nv50_constbufs_validate(struct nv50_context *);
-void nv50_sprite_coords_validate(struct nv50_context *);
+void nv50_validate_derived_rs(struct nv50_context *);
 
 /* nv50_state.c */
 extern void nv50_init_state_functions(struct nv50_context *);

src/gallium/drivers/nv50/nv50_debug.h

@@ -0,0 +1,25 @@
+
+#ifndef __NV50_DEBUG_H__
+#define __NV50_DEBUG_H__
+
+#include <stdio.h>
+
+#include "util/u_debug.h"
+
+#define NV50_DEBUG_MISC       0x0001
+#define NV50_DEBUG_SHADER     0x0100
+#define NV50_DEBUG_PROG_IR    0x0200
+#define NV50_DEBUG_PROG_RA    0x0400
+#define NV50_DEBUG_PROG_CFLOW 0x0800
+#define NV50_DEBUG_PROG_ALL   0x1f00
+
+#define NV50_DEBUG 0
+
+#define NOUVEAU_ERR(fmt, args...)                                 \
+   fprintf(stderr, "%s:%d - "fmt, __FUNCTION__, __LINE__, ##args)
+
+#define NV50_DBGMSG(ch, args...)           \
+   if ((NV50_DEBUG) & (NV50_DEBUG_##ch))        \
+      debug_printf(args)
+
+#endif /* __NV50_DEBUG_H__ */

src/gallium/drivers/nv50/nv50_pc.c

@@ -20,8 +20,6 @@
  * SOFTWARE.
  */
 
-/* #define NV50PC_DEBUG */
-
 #include "nv50_pc.h"
 #include "nv50_program.h"
 
@@ -180,6 +178,7 @@ nv50_op_can_write_flags(uint opcode)
    switch (opcode) { /* obvious ones like KIL, CALL, etc. not included */
    case NV_OP_PHI:
    case NV_OP_MOV:
+   case NV_OP_SELECT:
    case NV_OP_LINTERP:
    case NV_OP_PINTERP:
    case NV_OP_LDA:
@@ -367,7 +366,7 @@ nv_print_program(struct nv_pc *pc)
          nv_print_function(pc->root[i]);
 }
 
-#ifdef NV50PC_DEBUG
+#if NV50_DEBUG & NV50_DEBUG_PROG_CFLOW
 static void
 nv_do_print_cfgraph(struct nv_pc *pc, FILE *f, struct nv_basic_block *b)
 {
@@ -425,7 +424,7 @@ nv_print_cfgraph(struct nv_pc *pc, const char *filepath, int subr)
 
    fclose(f);
 }
-#endif
+#endif /* NV50_DEBUG_PROG_CFLOW */
 
 static INLINE void
 nvcg_show_bincode(struct nv_pc *pc)
@@ -446,7 +445,7 @@ nv50_emit_program(struct nv_pc *pc)
    uint32_t *code = pc->emit;
    int n;
 
-   NV50_DBGMSG("emitting program: size = %u\n", pc->bin_size);
+   NV50_DBGMSG(SHADER, "emitting program: size = %u\n", pc->bin_size);
 
    for (n = 0; n < pc->num_blocks; ++n) {
       struct nv_instruction *i;
@@ -472,7 +471,7 @@ nv50_emit_program(struct nv_pc *pc)
    pc->emit = code;
    code[pc->bin_size / 4 - 1] |= 1;
 
-#ifdef NV50PC_DEBUG
+#if NV50_DEBUG & NV50_DEBUG_SHADER
    nvcg_show_bincode(pc);
 #endif
 
@@ -500,7 +499,7 @@ nv50_generate_code(struct nv50_translation_info *ti)
    ret = nv50_tgsi_to_nc(pc, ti);
    if (ret)
       goto out;
-#ifdef NV50PC_DEBUG
+#if NV50_DEBUG & NV50_DEBUG_PROG_IR
    nv_print_program(pc);
 #endif
 
@@ -510,7 +509,7 @@ nv50_generate_code(struct nv50_translation_info *ti)
    ret = nv_pc_exec_pass0(pc);
    if (ret)
       goto out;
-#ifdef NV50PC_DEBUG
+#if NV50_DEBUG & NV50_DEBUG_PROG_IR
    nv_print_program(pc);
 #endif
 
@@ -518,7 +517,7 @@ nv50_generate_code(struct nv50_translation_info *ti)
    ret = nv_pc_exec_pass1(pc);
    if (ret)
       goto out;
-#ifdef NV50PC_DEBUG
+#if NV50_DEBUG & NV50_DEBUG_PROG_CFLOW
    nv_print_program(pc);
    nv_print_cfgraph(pc, "nv50_shader_cfgraph.dot", 0);
 #endif
@@ -552,7 +551,7 @@ nv50_generate_code(struct nv50_translation_info *ti)
 
    ti->p->uses_lmem = ti->store_to_memory;
 
-   NV50_DBGMSG("SHADER TRANSLATION - %s\n", ret ? "failure" : "success");
+   NV50_DBGMSG(SHADER, "SHADER TRANSLATION - %s\n", ret ? "failed" : "success");
 
 out:
    nv_pc_free_refs(pc);
@@ -624,6 +623,9 @@ nvbb_insert_tail(struct nv_basic_block *b, struct nv_instruction *i)
 
    i->bb = b;
    b->num_instructions++;
+
+   if (i->prev && i->prev->is_terminator)
+      nv_nvi_permute(i->prev, i);
 }
 
 void
@@ -669,7 +671,7 @@ nv_nvi_delete(struct nv_instruction *nvi)
 
    if (nvi == b->phi) {
       if (nvi->opcode != NV_OP_PHI)
-         NV50_DBGMSG("NOTE: b->phi points to non-PHI instruction\n");
+         NV50_DBGMSG(PROG_IR, "NOTE: b->phi points to non-PHI instruction\n");
 
       assert(!nvi->prev);
       if (!nvi->next || nvi->next->opcode != NV_OP_PHI)

src/gallium/drivers/nv50/nv50_pc.h

@@ -23,13 +23,7 @@
 #ifndef __NV50_COMPILER_H__
 #define __NV50_COMPILER_H__
 
-#define NV50PC_DEBUG
-
-#ifdef NV50PC_DEBUG
-# define NV50_DBGMSG(args...) debug_printf(args)
-#else
-# define NV50_DBGMSG(args...)
-#endif
+#include "nv50_debug.h"
 
 #include "pipe/p_defines.h"
 #include "util/u_inlines.h"
@@ -228,6 +222,8 @@ struct nv_ref {
    ubyte flags; /* not used yet */
 };
 
+#define NV_REF_FLAG_REGALLOC_PRIV (1 << 0)
+
 struct nv_basic_block;
 
 struct nv_instruction {
@@ -263,6 +259,15 @@ struct nv_instruction {
    ubyte quadop;
 };
 
+static INLINE int
+nvi_vector_size(struct nv_instruction *nvi)
+{
+   int i;
+   assert(nvi);
+   for (i = 0; i < 4 && nvi->def[i]; ++i);
+   return i;
+}
+
 #define CFG_EDGE_FORWARD     0
 #define CFG_EDGE_BACK        1
 #define CFG_EDGE_LOOP_ENTER  2

src/gallium/drivers/nv50/nv50_pc_optimize.c

@@ -20,8 +20,6 @@
  * SOFTWARE.
  */
 
-/* #define NV50PC_DEBUG */
-
 #include "nv50_pc.h"
 
 #define DESCEND_ARBITRARY(j, f)                                 \
@@ -116,7 +114,7 @@ nvi_isnop(struct nv_instruction *nvi)
       return FALSE;
 
    if (nvi->src[0]->value->join->reg.id < 0) {
-      NV50_DBGMSG("nvi_isnop: orphaned value detected\n");
+      NV50_DBGMSG(PROG_IR, "nvi_isnop: orphaned value detected\n");
       return TRUE;
    }
 
@@ -201,7 +199,7 @@ nv_pc_pass_pre_emission(void *priv, struct nv_basic_block *b)
    }
 
    if (!b->entry) {
-      NV50_DBGMSG("block %p is now empty\n", b);
+      NV50_DBGMSG(PROG_IR, "block %p is now empty\n", b);
    } else
    if (!b->exit->is_long) {
       assert(n32);
@@ -240,7 +238,7 @@ nv_pc_exec_pass2(struct nv_pc *pc)
 {
    int i, ret;
 
-   NV50_DBGMSG("preparing %u blocks for emission\n", pc->num_blocks);
+   NV50_DBGMSG(PROG_IR, "preparing %u blocks for emission\n", pc->num_blocks);
 
    pc->num_blocks = 0; /* will reorder bb_list */
 
@@ -966,7 +964,8 @@ nv_pass_flatten(struct nv_pass *ctx, struct nv_basic_block *b)
 
    if (bb_is_if_else_endif(b)) {
 
-      NV50_DBGMSG("pass_flatten: IF/ELSE/ENDIF construct at BB:%i\n", b->id);
+      NV50_DBGMSG(PROG_IR,
+                  "pass_flatten: IF/ELSE/ENDIF construct at BB:%i\n", b->id);
 
       for (n0 = 0, nvi = b->out[0]->entry; nvi; nvi = nvi->next, ++n0)
          if (!nv50_nvi_can_predicate(nvi))
@@ -975,7 +974,7 @@ nv_pass_flatten(struct nv_pass *ctx, struct nv_basic_block *b)
          for (n1 = 0, nvi = b->out[1]->entry; nvi; nvi = nvi->next, ++n1)
             if (!nv50_nvi_can_predicate(nvi))
                break;
-#ifdef NV50PC_DEBUG
+#if NV50_DEBUG & NV50_DEBUG_PROG_IR
          if (nvi) {
             debug_printf("cannot predicate: "); nv_print_instruction(nvi);
          }

src/gallium/drivers/nv50/nv50_pc_regalloc.c

@@ -20,11 +20,11 @@
  * SOFTWARE.
  */
 
-/* #define NV50PC_DEBUG */
-
-/* #define NV50_RA_DEBUG_LIVEI */
-/* #define NV50_RA_DEBUG_LIVE_SETS */
-/* #define NV50_RA_DEBUG_JOIN */
+#if NV50_DEBUG & NV50_DEBUG_PROG_RA
+# define NV50_RA_DEBUG_LIVEI
+# define NV50_RA_DEBUG_LIVE_SETS
+# define NV50_RA_DEBUG_JOIN
+#endif
 
 #include "nv50_context.h"
 #include "nv50_pc.h"
@@ -32,14 +32,39 @@
 #include "util/u_simple_list.h"
 
 #define NUM_REGISTER_FILES 4
+#define MAX_REGISTER_COUNT 256
 
 struct register_set {
    struct nv_pc *pc;
 
    uint32_t last[NUM_REGISTER_FILES];
-   uint32_t bits[NUM_REGISTER_FILES][8];
+   uint32_t bits[NUM_REGISTER_FILES][(MAX_REGISTER_COUNT + 31) / 32];
 };
 
+/* using OR because a set bit means occupied/unavailable, aliasing is allowed */
+static void
+intersect_register_sets(struct register_set *dst,
+                        struct register_set *src1, struct register_set *src2)
+{
+   int i, j;
+
+   for (i = 0; i < NUM_REGISTER_FILES; ++i) {
+      for (j = 0; j < (MAX_REGISTER_COUNT + 31) / 32; ++j)
+         dst->bits[i][j] = src1->bits[i][j] | src2->bits[i][j];
+   }
+}
+
+static void
+mask_register_set(struct register_set *set, uint32_t mask, uint32_t umask)
+{
+   int i, j;
+
+   for (i = 0; i < NUM_REGISTER_FILES; ++i) {
+      for (j = 0; j < (MAX_REGISTER_COUNT + 31) / 32; ++j)
+         set->bits[i][j] = (set->bits[i][j] | mask) & umask;
+   }
+}
+
 struct nv_pc_pass {
    struct nv_pc *pc;
 
@@ -61,11 +86,15 @@ ranges_coalesce(struct nv_range *range)
    }
 }
 
+/* @return: TRUE if @new_range can be freed (i.e. was not reused) */
 static boolean
 add_range_ex(struct nv_value *val, int bgn, int end, struct nv_range *new_range)
 {
    struct nv_range *range, **nextp = &val->livei;
 
+   if (bgn == end) /* [a, a) is invalid / empty */
+      return TRUE;
+
    for (range = val->livei; range; range = range->next) {
       if (end < range->bgn)
          break; /* insert before */
@@ -251,6 +280,8 @@ reg_occupy(struct register_set *set, struct nv_value *val)
    id <<= s;
    m = (1 << (1 << s)) - 1;
 
+   assert(s >= 0); /* XXX: remove me */
+
    set->bits[f][id / 32] |= m << (id % 32);
 
    if (set->pc->max_reg[f] < id)
@@ -286,15 +317,12 @@ join_allowed(struct nv_pc_pass *ctx, struct nv_value *a, struct nv_value *b)
    if (a->join->reg.id == b->join->reg.id)
       return TRUE;
 
-#if 1
    /* either a or b or both have been assigned */
 
    if (a->join->reg.id >= 0 && b->join->reg.id >= 0)
       return FALSE;
    else
    if (b->join->reg.id >= 0) {
-      if (a->join->reg.id >= 0)
-         return FALSE;
       val = a;
       a = b;
       b = val;
@@ -309,8 +337,6 @@ join_allowed(struct nv_pc_pass *ctx, struct nv_value *a, struct nv_value *b)
          return FALSE;
    }
    return TRUE;
-#endif
-   return FALSE;
 }
 
 static INLINE void
@@ -336,14 +362,14 @@ do_join_values(struct nv_pc_pass *ctx, struct nv_value *a, struct nv_value *b)
    assert(b->join == a->join);
 }
 
-static INLINE void
+static INLINE boolean
 try_join_values(struct nv_pc_pass *ctx, struct nv_value *a, struct nv_value *b)
 {
    if (!join_allowed(ctx, a, b)) {
 #ifdef NV50_RA_DEBUG_JOIN
       debug_printf("cannot join %i to %i: not allowed\n", b->n, a->n);
 #endif
-      return;
+      return FALSE;
    }
    if (livei_have_overlap(a->join, b->join)) {
 #ifdef NV50_RA_DEBUG_JOIN
@@ -351,10 +377,27 @@ try_join_values(struct nv_pc_pass *ctx, struct nv_value *a, struct nv_value *b)
       livei_print(a);
       livei_print(b);
 #endif
-      return;
+      return FALSE;
    }
 
    do_join_values(ctx, a, b);
+
+   return TRUE;
+}
+
+static void
+join_values_nofail(struct nv_pc_pass *ctx,
+                   struct nv_value *a, struct nv_value *b, boolean type_only)
+{
+   if (type_only) {
+      assert(join_allowed(ctx, a, b));
+      do_join_values(ctx, a, b);
+   } else {
+      boolean ok = try_join_values(ctx, a, b);
+      if (!ok) {
+         NOUVEAU_ERR("failed to coalesce values\n");
+      }
+   }
 }
 
 static INLINE boolean
@@ -369,20 +412,32 @@ need_new_else_block(struct nv_basic_block *b, struct nv_basic_block *p)
    return (b->num_in > 1) && (n == 2);
 }
 
+/* Look for the @phi's operand whose definition reaches @b. */
 static int
 phi_opnd_for_bb(struct nv_instruction *phi, struct nv_basic_block *b,
                 struct nv_basic_block *tb)
 {
+   struct nv_ref *srci, *srcj;
    int i, j;
 
-   for (j = -1, i = 0; i < 4 && phi->src[i]; ++i) {
-      if (!nvbb_reachable_by(b, phi->src[i]->value->insn->bb, tb))
+   for (j = -1, i = 0; i < 6 && phi->src[i]; ++i) {
+      srci = phi->src[i];
+      /* if already replaced, check with original source first */
+      if (srci->flags & NV_REF_FLAG_REGALLOC_PRIV)
+         srci = srci->value->insn->src[0];
+      if (!nvbb_reachable_by(b, srci->value->insn->bb, NULL))
          continue;
       /* NOTE: back-edges are ignored by the reachable-by check */
-      if (j < 0 || !nvbb_reachable_by(phi->src[j]->value->insn->bb,
-                                      phi->src[i]->value->insn->bb, tb))
+      if (j < 0 || !nvbb_reachable_by(srcj->value->insn->bb,
+                                      srci->value->insn->bb, NULL)) {
          j = i;
+         srcj = srci;
+      }
    }
+   if (j >= 0 && nvbb_reachable_by(b, phi->def[0]->insn->bb, NULL))
+      if (!nvbb_reachable_by(srcj->value->insn->bb,
+                             phi->def[0]->insn->bb, NULL))
+         j = -1;
    return j;
 }
 
@@ -429,16 +484,21 @@ pass_generate_phi_movs(struct nv_pc_pass *ctx, struct nv_basic_block *b)
       ctx->pc->current_block = pn;
 
       for (i = b->phi; i && i->opcode == NV_OP_PHI; i = i->next) {
-         if ((j = phi_opnd_for_bb(i, p, b)) < 0)
-            continue;
-         val = i->src[j]->value;
+         j = phi_opnd_for_bb(i, p, b);
 
-         if (i->src[j]->flags) {
-            val = val->insn->src[0]->value;
-            while (j < 4 && i->src[j])
-               ++j;
-            assert(j < 4);
+         if (j < 0) {
+            val = i->def[0];
+         } else {
+            val = i->src[j]->value;
+            if (i->src[j]->flags & NV_REF_FLAG_REGALLOC_PRIV) {
+               j = -1;
+               /* use original value, we already encountered & replaced it */
+               val = val->insn->src[0]->value;
+            }
          }
+         if (j < 0) /* need an additional source ? */
+            for (j = 0; j < 5 && i->src[j] && i->src[j]->value != val; ++j);
+         assert(j < 5);
 
          ni = new_instruction(ctx->pc, NV_OP_MOV);
 
@@ -452,11 +512,13 @@ pass_generate_phi_movs(struct nv_pc_pass *ctx, struct nv_basic_block *b)
 
          nv_reference(ctx->pc, &i->src[j], ni->def[0]);
 
-         i->src[j]->flags = 1;
+         i->src[j]->flags |= NV_REF_FLAG_REGALLOC_PRIV;
       }
 
       if (pn != p && pn->exit) {
-         ctx->pc->current_block = b->in[n ? 0 : 1];
+         assert(!b->in[!n]->exit || b->in[!n]->exit->is_terminator);
+         /* insert terminator (branch to ENDIF) in new else block */
+         ctx->pc->current_block = pn;
          ni = new_instruction(ctx->pc, NV_OP_BRA);
          ni->target = b;
          ni->is_terminator = 1;
@@ -470,45 +532,50 @@ pass_generate_phi_movs(struct nv_pc_pass *ctx, struct nv_basic_block *b)
    return 0;
 }
 
+#define JOIN_MASK_PHI    (1 << 0)
+#define JOIN_MASK_SELECT (1 << 1)
+#define JOIN_MASK_MOV    (1 << 2)
+#define JOIN_MASK_TEX    (1 << 3)
+
 static int
-pass_join_values(struct nv_pc_pass *ctx, int iter)
+pass_join_values(struct nv_pc_pass *ctx, unsigned mask)
 {
    int c, n;
 
    for (n = 0; n < ctx->num_insns; ++n) {
-      struct nv_instruction *i = ctx->insns[n];
+      struct nv_instruction *nvi, *i = ctx->insns[n];
 
       switch (i->opcode) {
       case NV_OP_PHI:
-         if (iter != 2)
+         if (!(mask & JOIN_MASK_PHI))
             break;
-         for (c = 0; c < 4 && i->src[c]; ++c)
-            try_join_values(ctx, i->def[0], i->src[c]->value);
+         for (c = 0; c < 5 && i->src[c]; ++c)
+            join_values_nofail(ctx, i->def[0], i->src[c]->value, FALSE);
          break;
       case NV_OP_MOV:
-         if ((iter == 2) && i->src[0]->value->insn &&
-             !nv_is_vector_op(i->src[0]->value->join->insn->opcode))
+         if (!(mask & JOIN_MASK_MOV))
+            break;
+         nvi = i->src[0]->value->join->insn;
+         if (nvi && !nv_is_vector_op(nvi->opcode))
             try_join_values(ctx, i->def[0], i->src[0]->value);
          break;
       case NV_OP_SELECT:
-         if (iter != 1)
+         if (!(mask & JOIN_MASK_SELECT))
             break;
-         for (c = 0; c < 4 && i->src[c]; ++c) {
-            assert(join_allowed(ctx, i->def[0], i->src[c]->value));
-            do_join_values(ctx, i->def[0], i->src[c]->value);
-         }
+         for (c = 0; c < 5 && i->src[c]; ++c)
+            join_values_nofail(ctx, i->def[0], i->src[c]->value, TRUE);
          break;
       case NV_OP_TEX:
       case NV_OP_TXB:
       case NV_OP_TXL:
       case NV_OP_TXQ:
-         if (iter)
+         if (!(mask & JOIN_MASK_TEX))
             break;
-         for (c = 0; c < 4; ++c) {
-            if (!i->src[c])
-               break;
-            do_join_values(ctx, i->def[c], i->src[c]->value);
-         }
+         /* This should work without conflicts because we always generate
+          * extra MOVs for the sources of a TEX.
+          */
+         for (c = 0; c < 4 && i->src[c]; ++c)
+            join_values_nofail(ctx, i->def[c], i->src[c]->value, TRUE);
          break;
       default:
          break;
@@ -643,15 +710,15 @@ static void collect_live_values(struct nv_basic_block *b, const int n)
 {
    int i;
 
-   if (b->out[0]) {
-      if (b->out[1]) { /* what to do about back-edges ? */
+   if (b->out[0] && b->out_kind[0] != CFG_EDGE_FAKE) {
+      if (b->out[1] && b->out_kind[1] != CFG_EDGE_FAKE) {
          for (i = 0; i < n; ++i)
             b->live_set[i] = b->out[0]->live_set[i] | b->out[1]->live_set[i];
       } else {
          memcpy(b->live_set, b->out[0]->live_set, n * sizeof(uint32_t));
       }
    } else
-   if (b->out[1]) {
+   if (b->out[1] && b->out_kind[1] != CFG_EDGE_FAKE) {
       memcpy(b->live_set, b->out[1]->live_set, n * sizeof(uint32_t));
    } else {
       memset(b->live_set, 0, n * sizeof(uint32_t));
@@ -770,8 +837,8 @@ insert_ordered_tail(struct nv_value *list, struct nv_value *nval)
    struct nv_value *elem;
 
    for (elem = list->prev;
-	elem != list && elem->livei->bgn > nval->livei->bgn;
-	elem = elem->prev);
+        elem != list && elem->livei->bgn > nval->livei->bgn;
+        elem = elem->prev);
    /* now elem begins before or at the same time as val */
 
    nval->prev = elem;
@@ -780,44 +847,49 @@ insert_ordered_tail(struct nv_value *list, struct nv_value *nval)
    elem->next = nval;
 }
 
+static void
+collect_register_values(struct nv_pc_pass *ctx, struct nv_value *head,
+                        boolean assigned_only)
+{
+   struct nv_value *val;
+   int k, n;
+
+   make_empty_list(head);
+
+   for (n = 0; n < ctx->num_insns; ++n) {
+      struct nv_instruction *i = ctx->insns[n];
+
+      /* for joined values, only the representative will have livei != NULL */
+      for (k = 0; k < 4; ++k) {
+         if (i->def[k] && i->def[k]->livei)
+            if (!assigned_only || i->def[k]->reg.id >= 0)
+               insert_ordered_tail(head, i->def[k]);
+      }
+      if (i->flags_def && i->flags_def->livei)
+         if (!assigned_only || i->flags_def->reg.id >= 0)
+            insert_ordered_tail(head, i->flags_def);
+   }
+
+   for (val = head->next; val != head->prev; val = val->next) {
+      assert(val->join == val);
+      assert(val->livei->bgn <= val->next->livei->bgn);
+   }
+}
+
 static int
 pass_linear_scan(struct nv_pc_pass *ctx, int iter)
 {
-   struct nv_instruction *i;
    struct register_set f, free;
-   int k, n;
    struct nv_value *cur, *val, *tmp[2];
    struct nv_value active, inactive, handled, unhandled;
 
    make_empty_list(&active);
    make_empty_list(&inactive);
    make_empty_list(&handled);
-   make_empty_list(&unhandled);
 
    nv50_ctor_register_set(ctx->pc, &free);
 
-   /* joined values should have range = NULL and thus not be added;
-    * also, fixed memory values won't be added because they're not
-    * def'd, just used
-    */
-   for (n = 0; n < ctx->num_insns; ++n) {
-      i = ctx->insns[n];
-
-      for (k = 0; k < 4; ++k) {
-         if (i->def[k] && i->def[k]->livei)
-            insert_ordered_tail(&unhandled, i->def[k]);
-         else
-         if (0 && i->def[k])
-            debug_printf("skipping def'd value %i: no livei\n", i->def[k]->n);
-      }
-      if (i->flags_def && i->flags_def->livei)
-         insert_ordered_tail(&unhandled, i->flags_def);
-   }
-
-   for (val = unhandled.next; val != unhandled.prev; val = val->next) {
-      assert(val->join == val);
-      assert(val->livei->bgn <= val->next->livei->bgn);
-   }
+   collect_register_values(ctx, &unhandled, FALSE);
 
    foreach_s(cur, tmp[0], &unhandled) {
       remove_from_list(cur);
@@ -854,13 +926,7 @@ pass_linear_scan(struct nv_pc_pass *ctx, int iter)
             reg_occupy(&f, val);
 
       if (cur->reg.id < 0) {
-         boolean mem = FALSE;
-
-         if (nv_is_vector_op(cur->insn->opcode))
-            mem = !reg_assign(&f, &cur->insn->def[0], 4);
-         else
-         if (iter)
-            mem = !reg_assign(&f, &cur, 1);
+         boolean mem = !reg_assign(&f, &cur, 1);
 
          if (mem) {
             NOUVEAU_ERR("out of registers\n");
@@ -874,13 +940,80 @@ pass_linear_scan(struct nv_pc_pass *ctx, int iter)
    return 0;
 }
 
+/* Allocate values defined by instructions such as TEX, which have to be
+ * assigned to consecutive registers.
+ * Linear scan doesn't really work here since the values can have different
+ * live intervals.
+ */
+static int
+pass_allocate_constrained_values(struct nv_pc_pass *ctx)
+{
+   struct nv_value regvals, *val;
+   struct nv_instruction *i;
+   struct nv_value *defs[4];
+   struct register_set regs[4];
+   int n, vsize, c;
+   uint32_t mask;
+   boolean mem;
+
+   collect_register_values(ctx, &regvals, TRUE);
+
+   for (n = 0; n < ctx->num_insns; ++n) {
+      i = ctx->insns[n];
+      vsize = nvi_vector_size(i);
+      if (!(vsize > 1))
+         continue;
+      assert(vsize <= 4);
+      for (c = 0; c < vsize; ++c)
+         defs[c] = i->def[c]->join;
+
+      if (defs[0]->reg.id >= 0) {
+         for (c = 1; c < vsize; ++c)
+            assert(defs[c]->reg.id >= 0);
+         continue;
+      }
+
+      /* Compute registers available for this "vector" of consecutive registers.
+       * Each value (component) has its own independent live interval.
+       */
+      for (c = 0; c < vsize; ++c) {
+         nv50_ctor_register_set(ctx->pc, &regs[c]);
+
+         foreach(val, &regvals) {
+            if (val->reg.id >= 0 && livei_have_overlap(val, defs[c]))
+               reg_occupy(&regs[c], val);
+         }
+         /* Only 32 bit GPRs will be allocated here, but register set
+          * granularity for GPRs is 16 bit.
+          */
+         mask = 0x03030303;
+         if (vsize == 2) /* granularity is 2 and not 4 */
+            mask |= 0x03030303 << 4;
+         mask_register_set(&regs[c], 0, mask << (c * 2));
+
+         if (defs[c]->livei)
+            insert_ordered_tail(&regvals, defs[c]);
+      }
+      for (c = 1; c < vsize; ++c)
+         intersect_register_sets(&regs[0], &regs[0], &regs[c]);
+
+      mem = !reg_assign(&regs[0], &defs[0], vsize);
+
+      if (mem) {
+         NOUVEAU_ERR("out of registers\n");
+         abort();
+      }
+   }
+   return 0;
+}
+
 static int
 nv_pc_pass1(struct nv_pc *pc, struct nv_basic_block *root)
 {
    struct nv_pc_pass *ctx;
    int i, ret;
 
-   NV50_DBGMSG("REGISTER ALLOCATION - entering\n");
+   NV50_DBGMSG(PROG_RA, "REGISTER ALLOCATION - entering\n");
 
    ctx = CALLOC_STRUCT(nv_pc_pass);
    if (!ctx)
@@ -923,16 +1056,16 @@ nv_pc_pass1(struct nv_pc *pc, struct nv_basic_block *root)
       livei_print(&pc->values[i]);
 #endif
 
-   ret = pass_join_values(ctx, 0);
+   ret = pass_join_values(ctx, JOIN_MASK_PHI);
    if (ret)
       goto out;
-   ret = pass_linear_scan(ctx, 0);
+   ret = pass_join_values(ctx, JOIN_MASK_SELECT | JOIN_MASK_TEX);
    if (ret)
       goto out;
-   ret = pass_join_values(ctx, 1);
+   ret = pass_join_values(ctx, JOIN_MASK_MOV);
    if (ret)
       goto out;
-   ret = pass_join_values(ctx, 2);
+   ret = pass_allocate_constrained_values(ctx);
    if (ret)
       goto out;
    ret = pass_linear_scan(ctx, 1);
@@ -942,7 +1075,7 @@ nv_pc_pass1(struct nv_pc *pc, struct nv_basic_block *root)
    for (i = 0; i < pc->num_values; ++i)
       livei_release(&pc->values[i]);
 
-   NV50_DBGMSG("REGISTER ALLOCATION - leaving\n");
+   NV50_DBGMSG(PROG_RA, "REGISTER ALLOCATION - leaving\n");
 
 out:
    FREE(ctx->insns);

src/gallium/drivers/nv50/nv50_program.c

@@ -20,8 +20,6 @@
  * SOFTWARE.
  */
 
-/* #define NV50_PROGRAM_DEBUG */
-
 #include "nv50_program.h"
 #include "nv50_pc.h"
 #include "nv50_context.h"
@@ -486,7 +484,7 @@ nv50_fragprog_prepare(struct nv50_translation_info *ti)
       ++nintp;
    }
 
-   p->fp.colors = (1 << 24) | 4; /* CLAMP, FFC0_ID = 4 */
+   p->fp.colors = 4 << NV50_3D_MAP_SEMANTIC_0_FFC0_ID__SHIFT; /* after HPOS */
 
    for (i = 0; i < p->in_nr; ++i) {
       int j = p->in[i].id;
@@ -564,7 +562,7 @@ nv50_prog_scan(struct nv50_translation_info *ti)
 
    tgsi_scan_shader(p->pipe.tokens, &ti->scan);
 
-#ifdef NV50_PROGRAM_DEBUG
+#if NV50_DEBUG & NV50_DEBUG_SHADER
    tgsi_dump(p->pipe.tokens, 0);
 #endif
 

src/gallium/drivers/nv50/nv50_screen.c

@@ -108,6 +108,8 @@ nv50_screen_get_param(struct pipe_screen *pscreen, enum pipe_cap param)
       return 1;
    case PIPE_CAP_MAX_RENDER_TARGETS:
       return 8;
+   case PIPE_CAP_FRAGMENT_COLOR_CLAMP_CONTROL:
+      return 1;
    case PIPE_CAP_TIMER_QUERY:
    case PIPE_CAP_OCCLUSION_QUERY:
       return 1;
@@ -129,6 +131,7 @@ nv50_screen_get_param(struct pipe_screen *pscreen, enum pipe_cap param)
    case PIPE_CAP_PRIMITIVE_RESTART:
    case PIPE_CAP_TGSI_INSTANCEID:
    case PIPE_CAP_VERTEX_ELEMENT_INSTANCE_DIVISOR:
+   case PIPE_CAP_MIXED_COLORBUFFER_FORMATS:
       return 1;
    default:
       NOUVEAU_ERR("unknown PIPE_CAP %d\n", param);
@@ -417,6 +420,10 @@ nv50_screen_create(struct pipe_winsys *ws, struct nouveau_device *dev)
    OUT_RING  (chan, NV50_3D_MULTISAMPLE_MODE_MS1);
    BEGIN_RING(chan, RING_3D(MULTISAMPLE_CTRL), 1);
    OUT_RING  (chan, 0);
+   BEGIN_RING(chan, RING_3D(LINE_LAST_PIXEL), 1);
+   OUT_RING  (chan, 0);
+   BEGIN_RING(chan, RING_3D(BLEND_SEPARATE_ALPHA), 1);
+   OUT_RING  (chan, 1);
 
    BEGIN_RING(chan, RING_3D(SCREEN_Y_CONTROL), 1);
    OUT_RING  (chan, 0);

src/gallium/drivers/nv50/nv50_shader_state.c

@@ -226,7 +226,7 @@ nv50_gmtyprog_validate(struct nv50_context *nv50)
    OUT_RING  (chan, gp->code_base);
 }
 
-void
+static void
 nv50_sprite_coords_validate(struct nv50_context *nv50)
 {
    struct nouveau_channel *chan = nv50->screen->base.channel;
@@ -282,6 +282,39 @@ nv50_sprite_coords_validate(struct nv50_context *nv50)
    OUT_RINGp (chan, pntc, 8);
 }
 
+/* Validate state derived from shaders and the rasterizer cso. */
+void
+nv50_validate_derived_rs(struct nv50_context *nv50)
+{
+   struct nouveau_channel *chan = nv50->screen->base.channel;
+   uint32_t color, psize;
+
+   nv50_sprite_coords_validate(nv50);
+
+   if (nv50->dirty & NV50_NEW_FRAGPROG)
+      return;
+   psize = nv50->state.semantic_psize & ~NV50_3D_MAP_SEMANTIC_3_PTSZ_EN__MASK;
+   color = nv50->state.semantic_color & ~NV50_3D_MAP_SEMANTIC_0_CLMP_EN;
+
+   if (nv50->rast->pipe.clamp_vertex_color)
+      color |= NV50_3D_MAP_SEMANTIC_0_CLMP_EN;
+
+   if (color != nv50->state.semantic_color) {
+      nv50->state.semantic_color = color;
+      BEGIN_RING(chan, RING_3D(MAP_SEMANTIC_0), 1);
+      OUT_RING  (chan, color);
+   }
+
+   if (nv50->rast->pipe.point_size_per_vertex)
+      psize |= NV50_3D_MAP_SEMANTIC_3_PTSZ_EN__MASK;
+
+   if (psize != nv50->state.semantic_psize) {
+      nv50->state.semantic_psize = psize;
+      BEGIN_RING(chan, RING_3D(MAP_SEMANTIC_3), 1);
+      OUT_RING  (chan, psize);
+   }
+}
+
 static int
 nv50_vec4_map(uint8_t *map, int mid, uint32_t lin[4],
               struct nv50_varying *in, struct nv50_varying *out)
@@ -372,6 +405,9 @@ nv50_fp_linkage_validate(struct nv50_context *nv50)
       map[m++] = vp->vp.psiz;
    }
 
+   if (nv50->rast->pipe.clamp_vertex_color)
+      colors |= NV50_3D_MAP_SEMANTIC_0_CLMP_EN;
+
    n = (m + 3) / 4;
    assert(m <= 64);
 
@@ -404,6 +440,9 @@ nv50_fp_linkage_validate(struct nv50_context *nv50)
 
    nv50->state.interpolant_ctrl = interp;
 
+   nv50->state.semantic_color = colors;
+   nv50->state.semantic_psize = psiz;
+
    BEGIN_RING(chan, RING_3D(NOPERSPECTIVE_BITMAP(0)), 4);
    OUT_RINGp (chan, lin, 4);
 

src/gallium/drivers/nv50/nv50_state.c

@@ -97,8 +97,14 @@ nv50_blend_state_create(struct pipe_context *pipe,
 
    so->pipe = *cso;
 
-   SB_BEGIN_3D(so, BLEND_ENABLE(0), 8);
+   SB_BEGIN_3D(so, COLOR_MASK_COMMON, 1);
+   SB_DATA    (so, !cso->independent_blend_enable);
+
+   SB_BEGIN_3D(so, BLEND_ENABLE_COMMON, 1);
+   SB_DATA    (so, !cso->independent_blend_enable);
+
    if (cso->independent_blend_enable) {
+      SB_BEGIN_3D(so, BLEND_ENABLE(0), 8);
       for (i = 0; i < 8; ++i) {
          SB_DATA(so, cso->rt[i].blend_enable);
          if (cso->rt[i].blend_enable)
@@ -121,8 +127,8 @@ nv50_blend_state_create(struct pipe_context *pipe,
          }
       }
    } else {
-      for (i = 0; i < 8; ++i)
-         SB_DATA(so, cso->rt[0].blend_enable);
+      SB_BEGIN_3D(so, BLEND_ENABLE(0), 1);
+      SB_DATA    (so, cso->rt[0].blend_enable);
    }
 
    if (emit_common_func) {
@@ -145,17 +151,16 @@ nv50_blend_state_create(struct pipe_context *pipe,
       SB_DATA    (so, 0);
    }
 
-   SB_BEGIN_3D(so, COLOR_MASK(0), 8);
    if (cso->independent_blend_enable) {
+      SB_BEGIN_3D(so, COLOR_MASK(0), 8);
       for (i = 0; i < 8; ++i)
          SB_DATA(so, nv50_colormask(cso->rt[i].colormask));
    } else {
-      uint32_t cmask = nv50_colormask(cso->rt[0].colormask);
-      for (i = 0; i < 8; ++i)
-         SB_DATA(so, cmask);
+      SB_BEGIN_3D(so, COLOR_MASK(0), 1);
+      SB_DATA    (so, nv50_colormask(cso->rt[0].colormask));
    }
 
-   assert(so->size < (sizeof(so->state) / sizeof(so->state[0])));
+   assert(so->size <= (sizeof(so->state) / sizeof(so->state[0])));
    return so;
 }
 
@@ -174,6 +179,7 @@ nv50_blend_state_delete(struct pipe_context *pipe, void *hwcso)
    FREE(hwcso);
 }
 
+/* NOTE: ignoring line_last_pixel, using FALSE (set on screen init) */
 static void *
 nv50_rasterizer_state_create(struct pipe_context *pipe,
                              const struct pipe_rasterizer_state *cso)
@@ -198,6 +204,9 @@ nv50_rasterizer_state_create(struct pipe_context *pipe,
    SB_BEGIN_3D(so, VERTEX_TWO_SIDE_ENABLE, 1);
    SB_DATA    (so, cso->light_twoside);
 
+   SB_BEGIN_3D(so, FRAG_COLOR_CLAMP_EN, 1);
+   SB_DATA    (so, cso->clamp_fragment_color ? 0x11111111 : 0x00000000);
+
    SB_BEGIN_3D(so, LINE_WIDTH, 1);
    SB_DATA    (so, fui(cso->line_width));
    SB_BEGIN_3D(so, LINE_SMOOTH_ENABLE, 1);
@@ -258,7 +267,7 @@ nv50_rasterizer_state_create(struct pipe_context *pipe,
       SB_DATA    (so, fui(cso->offset_units * 2.0f));
    }
 
-   assert(so->size < (sizeof(so->state) / sizeof(so->state[0])));
+   assert(so->size <= (sizeof(so->state) / sizeof(so->state[0])));
    return (void *)so;
 }
 
@@ -337,7 +346,7 @@ nv50_zsa_state_create(struct pipe_context *pipe,
       SB_DATA    (so, 0);
    }
 
-   assert(so->size < (sizeof(so->state) / sizeof(so->state[0])));
+   assert(so->size <= (sizeof(so->state) / sizeof(so->state[0])));
    return (void *)so;
 }
 

src/gallium/drivers/nv50/nv50_state_validate.c

@@ -309,7 +309,7 @@ static struct state_validate {
     { nv50_fp_linkage_validate,    NV50_NEW_FRAGPROG | NV50_NEW_VERTPROG |
                                    NV50_NEW_GMTYPROG },
     { nv50_gp_linkage_validate,    NV50_NEW_GMTYPROG | NV50_NEW_VERTPROG },
-    { nv50_sprite_coords_validate, NV50_NEW_FRAGPROG | NV50_NEW_RASTERIZER |
+    { nv50_validate_derived_rs,    NV50_NEW_FRAGPROG | NV50_NEW_RASTERIZER |
                                    NV50_NEW_VERTPROG | NV50_NEW_GMTYPROG },
     { nv50_constbufs_validate,     NV50_NEW_CONSTBUF },
     { nv50_validate_textures,      NV50_NEW_TEXTURES },

src/gallium/drivers/nv50/nv50_stateobj.h

@@ -21,13 +21,13 @@
 struct nv50_blend_stateobj {
    struct pipe_blend_state pipe;
    int size;
-   uint32_t state[78];
+   uint32_t state[82]; // TODO: allocate less if !independent_blend_enable
 };
 
 struct nv50_rasterizer_stateobj {
    struct pipe_rasterizer_state pipe;
    int size;
-   uint32_t state[40];
+   uint32_t state[42];
 };
 
 struct nv50_zsa_stateobj {

src/gallium/drivers/nv50/nv50_tgsi_to_nc.c

@@ -20,8 +20,6 @@
  * SOFTWARE.
  */
 
-/* #define NV50_TGSI2NC_DEBUG */
-
 #include <unistd.h>
 
 #include "nv50_context.h"
@@ -213,7 +211,7 @@ static INLINE void
 bld_warn_uninitialized(struct bld_context *bld, int kind,
                        struct bld_value_stack *stk, struct nv_basic_block *b)
 {
-#ifdef NV50_TGSI2NC_DEBUG
+#if NV50_DEBUG & NV50_DEBUG_PROG_IR
    long i = (stk - &bld->tvs[0][0]) / 4;
    long c = (stk - &bld->tvs[0][0]) & 3;
 
@@ -273,6 +271,12 @@ fetch_by_bb(struct bld_value_stack *stack,
          fetch_by_bb(stack, vals, n, b->in[i]);
 }
 
+static INLINE boolean
+nvbb_is_terminated(struct nv_basic_block *bb)
+{
+   return bb->exit && bb->exit->is_terminator;
+}
+
 static INLINE struct nv_value *
 bld_load_imm_u32(struct bld_context *bld, uint32_t u);
 
@@ -1556,7 +1560,7 @@ bld_instruction(struct bld_context *bld,
    int c;
    uint opcode = translate_opcode(insn->Instruction.Opcode);
 
-#ifdef NV50_TGSI2NC_DEBUG
+#if NV50_DEBUG & NV50_DEBUG_PROG_IR
    debug_printf("bld_instruction:"); tgsi_dump_instruction(insn, 1);
 #endif
 	
@@ -1727,8 +1731,7 @@ bld_instruction(struct bld_context *bld,
    {
       struct nv_basic_block *b = new_basic_block(bld->pc);
 
-      if (bld->pc->current_block->exit &&
-          !bld->pc->current_block->exit->is_terminator)
+      if (!nvbb_is_terminated(bld->pc->current_block))
          bld_flow(bld, NV_OP_BRA, NV_CC_TR, NULL, b, FALSE);
 
       --bld->cond_lvl;
@@ -1800,7 +1803,8 @@ bld_instruction(struct bld_context *bld,
    {
       struct nv_basic_block *bb = bld->loop_bb[bld->loop_lvl - 1];
 
-      bld_flow(bld, NV_OP_BRA, NV_CC_TR, NULL, bb, FALSE);
+      if (!nvbb_is_terminated(bld->pc->current_block))
+         bld_flow(bld, NV_OP_BRA, NV_CC_TR, NULL, bb, FALSE);
 
       nvbb_attach_block(bld->pc->current_block, bb, CFG_EDGE_BACK);
 

src/gallium/drivers/nvc0/nvc0_3d.xml.h

@@ -82,6 +82,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #define NVC0_3D_SERIALIZE					0x00000110
 
+#define NVC0_3D_LINE_WIDTH_SEPARATE				0x0000020c
+
 #define NVC0_3D_EARLY_FRAGMENT_TESTS				0x00000210
 
 #define NVC0_3D_MEM_BARRIER					0x0000021c
@@ -342,10 +344,14 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #define NVC0_3D_VERTEX_RUNOUT_ADDRESS_LOW			0x00000f88
 
+#define NVC0_3D_COLOR_MASK_COMMON				0x00000f90
+
 #define NVC0_3D_DEPTH_BOUNDS(i0)			       (0x00000f9c + 0x4*(i0))
 #define NVC0_3D_DEPTH_BOUNDS__ESIZE				0x00000004
 #define NVC0_3D_DEPTH_BOUNDS__LEN				0x00000002
 
+#define NVC0_3D_RT_SEPARATE_FRAG_DATA				0x00000fac
+
 #define NVC0_3D_MSAA_MASK(i0)				       (0x00000fbc + 0x4*(i0))
 #define NVC0_3D_MSAA_MASK__ESIZE				0x00000004
 #define NVC0_3D_MSAA_MASK__LEN					0x00000004
@@ -561,6 +567,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define NVC0_3D_TEX_CACHE_CTL_UNK1__MASK			0x00000030
 #define NVC0_3D_TEX_CACHE_CTL_UNK1__SHIFT			4
 
+#define NVC0_3D_BLEND_SEPARATE_ALPHA				0x0000133c
+
 #define NVC0_3D_BLEND_EQUATION_RGB				0x00001340
 #define NVC0_3D_BLEND_EQUATION_RGB_FUNC_ADD			0x00008006
 #define NVC0_3D_BLEND_EQUATION_RGB_MIN				0x00008007
@@ -583,6 +591,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 
 #define NVC0_3D_BLEND_FUNC_DST_ALPHA				0x00001358
 
+#define NVC0_3D_BLEND_ENABLE_COMMON				0x0000135c
+
 #define NVC0_3D_BLEND_ENABLE(i0)			       (0x00001360 + 0x4*(i0))
 #define NVC0_3D_BLEND_ENABLE__ESIZE				0x00000004
 #define NVC0_3D_BLEND_ENABLE__LEN				0x00000008
@@ -651,7 +661,9 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define NVC0_3D_SCREEN_Y_CONTROL_Y_NEGATE			0x00000001
 #define NVC0_3D_SCREEN_Y_CONTROL_TRIANGLE_RAST_FLIP		0x00000010
 
-#define NVC0_3D_LINE_WIDTH					0x000013b0
+#define NVC0_3D_LINE_WIDTH_SMOOTH				0x000013b0
+
+#define NVC0_3D_LINE_WIDTH_ALIASED				0x000013b4
 
 #define NVC0_3D_GP_VERTEX_OUTPUT_COUNT				0x00001420
 #define NVC0_3D_GP_VERTEX_OUTPUT_COUNT__MIN			0x00000001
@@ -977,6 +989,8 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define NVC0_3D_CULL_FACE_BACK					0x00000405
 #define NVC0_3D_CULL_FACE_FRONT_AND_BACK			0x00000408
 
+#define NVC0_3D_LINE_LAST_PIXEL					0x00001924
+
 #define NVC0_3D_VIEWPORT_TRANSFORM_EN				0x0000192c
 
 #define NVC0_3D_VIEW_VOLUME_CLIP_CTRL				0x0000193c
@@ -1220,8 +1234,6 @@ WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.
 #define NVC0_3D_TFB_VARYING_LOCS__ESIZE				0x00000004
 #define NVC0_3D_TFB_VARYING_LOCS__LEN				0x00000020
 
-#define NVC0_3D_COLOR_MASK_BROADCAST				0x00003808
-
 #define NVC0_3D_VERTEX_ARRAY_SELECT				0x00003820
 
 #define NVC0_3D_BLEND_ENABLES					0x00003858

src/gallium/drivers/nvc0/nvc0_context.c

@@ -46,6 +46,15 @@ nvc0_flush(struct pipe_context *pipe,
    FIRE_RING(screen->channel);
 }
 
+static void
+nvc0_texture_barrier(struct pipe_context *pipe)
+{
+   struct nouveau_channel *chan = nvc0_context(pipe)->screen->base.channel;
+
+   IMMED_RING(chan, RING_3D(SERIALIZE), 0);
+   IMMED_RING(chan, RING_3D(TEX_CACHE_CTL), 0);
+}
+
 static void
 nvc0_context_unreference_resources(struct nvc0_context *nvc0)
 {
@@ -128,6 +137,7 @@ nvc0_create(struct pipe_screen *pscreen, void *priv)
    pipe->clear = nvc0_clear;
 
    pipe->flush = nvc0_flush;
+   pipe->texture_barrier = nvc0_texture_barrier;
 
    if (!screen->cur_ctx)
       screen->cur_ctx = nvc0;

src/gallium/drivers/nvc0/nvc0_context.h

@@ -1,7 +1,6 @@
 #ifndef __NVC0_CONTEXT_H__
 #define __NVC0_CONTEXT_H__
 
-#include <stdio.h>
 #include "pipe/p_context.h"
 #include "pipe/p_defines.h"
 #include "pipe/p_state.h"
@@ -13,6 +12,7 @@
 
 #include "draw/draw_vertex.h"
 
+#include "nv50/nv50_debug.h"
 #include "nvc0_winsys.h"
 #include "nvc0_stateobj.h"
 #include "nvc0_screen.h"
@@ -26,15 +26,6 @@
 #include "nvc0_2d.xml.h"
 #include "nvc0_m2mf.xml.h"
 
-#define NOUVEAU_ERR(fmt, args...) \
-   fprintf(stderr, "%s:%d -  "fmt, __FUNCTION__, __LINE__, ##args);
-
-#ifdef NOUVEAU_DEBUG
-# define NOUVEAU_DBG(args...) printf(args);
-#else
-# define NOUVEAU_DBG(args...)
-#endif
-
 #define NVC0_NEW_BLEND        (1 << 0)
 #define NVC0_NEW_RASTERIZER   (1 << 1)
 #define NVC0_NEW_ZSA          (1 << 2)

src/gallium/drivers/nvc0/nvc0_graph_macros.h

@@ -68,19 +68,6 @@ static const uint32_t nvc0_9097_vertex_array_select[] =
    0x00001841, /* 0x0b: send $r3 */
 };
 
-static const uint32_t nvc0_9097_color_mask_brdc[] =
-{
-   0x05a00021, /* maddr [NVC0_3D_COLOR_MASK(0), increment = 4] */
-   0x00000841, /* send $r1 */
-   0x00000841, /* send $r1 */
-   0x00000841, /* send $r1 */
-   0x00000841, /* send $r1 */
-   0x00000841, /* send $r1 */
-   0x00000841, /* send $r1 */
-   0x000008c1, /* exit send $r1 */
-   0x00000841, /* send $r1 */
-};
-
 /*
  * [GL_POLYGON_MODE_FRONT] = arg;
  *

src/gallium/drivers/nvc0/nvc0_pc.c

@@ -20,8 +20,6 @@
  * SOFTWARE.
  */
 
-#define NOUVEAU_DEBUG 1
-
 #include "nvc0_pc.h"
 #include "nvc0_program.h"
 
@@ -262,7 +260,7 @@ nvc0_print_program(struct nv_pc *pc)
          nvc0_print_function(pc->root[i]);
 }
 
-#if NOUVEAU_DEBUG > 1
+#if NV50_DEBUG & NV50_DEBUG_PROG_CFLOW
 static void
 nv_do_print_cfgraph(struct nv_pc *pc, FILE *f, struct nv_basic_block *b)
 {
@@ -327,7 +325,7 @@ nvc0_pc_print_binary(struct nv_pc *pc)
 {
    unsigned i;
 
-   NOUVEAU_DBG("nvc0_pc_print_binary(%u ops)\n", pc->emit_size / 8);
+   NV50_DBGMSG(SHADER, "nvc0_pc_print_binary(%u ops)\n", pc->emit_size / 8);
 
    for (i = 0; i < pc->emit_size / 4; i += 2) {
       debug_printf("0x%08x ", pc->emit[i + 0]);
@@ -344,7 +342,7 @@ nvc0_emit_program(struct nv_pc *pc)
    uint32_t *code = pc->emit;
    int n;
 
-   NOUVEAU_DBG("emitting program: size = %u\n", pc->emit_size);
+   NV50_DBGMSG(SHADER, "emitting program: size = %u\n", pc->emit_size);
 
    pc->emit_pos = 0;
    for (n = 0; n < pc->num_blocks; ++n) {
@@ -365,11 +363,10 @@ nvc0_emit_program(struct nv_pc *pc)
 
    pc->emit = code;
 
-#ifdef NOUVEAU_DEBUG
+#if NV50_DEBUG & NV50_DEBUG_SHADER
    nvc0_pc_print_binary(pc);
-#else
-   debug_printf("not printing binary\n");
 #endif
+
    return 0;
 }
 
@@ -396,7 +393,7 @@ nvc0_generate_code(struct nvc0_translation_info *ti)
    ret = nvc0_tgsi_to_nc(pc, ti);
    if (ret)
       goto out;
-#if NOUVEAU_DEBUG > 1
+#if NV50_DEBUG & NV50_DEBUG_PROG_IR
    nvc0_print_program(pc);
 #endif
 
@@ -406,7 +403,7 @@ nvc0_generate_code(struct nvc0_translation_info *ti)
    ret = nvc0_pc_exec_pass0(pc);
    if (ret)
       goto out;
-#ifdef NOUVEAU_DEBUG
+#if NV50_DEBUG & NV50_DEBUG_PROG_IR
    nvc0_print_program(pc);
 #endif
 
@@ -414,7 +411,7 @@ nvc0_generate_code(struct nvc0_translation_info *ti)
    ret = nvc0_pc_exec_pass1(pc);
    if (ret)
       goto out;
-#if NOUVEAU_DEBUG > 1
+#if NV50_DEBUG & NV50_DEBUG_PROG_CFLOW
    nvc0_print_program(pc);
    nv_print_cfgraph(pc, "nvc0_shader_cfgraph.dot", 0);
 #endif
@@ -444,7 +441,7 @@ nvc0_generate_code(struct nvc0_translation_info *ti)
    ti->prog->relocs = pc->reloc_entries;
    ti->prog->num_relocs = pc->num_relocs;
 
-   NOUVEAU_DBG("SHADER TRANSLATION - %s\n", ret ? "failure" : "success");
+   NV50_DBGMSG(SHADER, "SHADER TRANSLATION - %s\n", ret ? "failed" : "success");
 
 out:
    nv_pc_free_refs(pc);
@@ -573,7 +570,7 @@ nvc0_insn_delete(struct nv_instruction *nvi)
 
    if (nvi == b->phi) {
       if (nvi->opcode != NV_OP_PHI)
-         NOUVEAU_DBG("NOTE: b->phi points to non-PHI instruction\n");
+         NV50_DBGMSG(PROG_IR, "NOTE: b->phi points to non-PHI instruction\n");
 
       assert(!nvi->prev);
       if (!nvi->next || nvi->next->opcode != NV_OP_PHI)

src/gallium/drivers/nvc0/nvc0_pc.h

@@ -23,20 +23,7 @@
 #ifndef __NVC0_COMPILER_H__
 #define __NVC0_COMPILER_H__
 
-#include <stdio.h>
-
-#ifndef NOUVEAU_DBG
-#ifdef NOUVEAU_DEBUG
-# define NOUVEAU_DBG(args...) debug_printf(args);
-#else
-# define NOUVEAU_DBG(args...)
-#endif
-#endif
-
-#ifndef NOUVEAU_ERR
-#define NOUVEAU_ERR(fmt, args...) \
-   fprintf(stderr, "%s:%d -  "fmt, __FUNCTION__, __LINE__, ##args);
-#endif
+#include "nv50/nv50_debug.h"
 
 #include "pipe/p_defines.h"
 #include "util/u_inlines.h"

src/gallium/drivers/nvc0/nvc0_pc_emit.c

@@ -715,6 +715,10 @@ emit_interp(struct nv_pc *pc, struct nv_instruction *i)
    if (i->opcode == NV_OP_PINTERP) {
       pc->emit[0] |= 0x040;
       SID(pc, i->src[1], 26);
+
+      if (i->src[0]->value->reg.address >= 0x280 &&
+          i->src[0]->value->reg.address <= 0x29c)
+         pc->emit[0] |= 0x080; /* XXX: ? */
    } else {
       SID(pc, NULL, 26);
    }
@@ -875,7 +879,9 @@ emit_st(struct nv_pc *pc, struct nv_instruction *i)
 void
 nvc0_emit_instruction(struct nv_pc *pc, struct nv_instruction *i)
 {
+#if NV50_DEBUG & NV50_DEBUG_SHADER
    debug_printf("EMIT: "); nvc0_print_instruction(i);
+#endif
 
    switch (i->opcode) {
    case NV_OP_VFETCH:

src/gallium/drivers/nvc0/nvc0_pc_optimize.c

@@ -115,7 +115,7 @@ inst_is_noop(struct nv_instruction *nvi)
       return FALSE;
 
    if (nvi->src[0]->value->join->reg.id < 0) {
-      NOUVEAU_DBG("inst_is_noop: orphaned value detected\n");
+      NV50_DBGMSG(PROG_IR, "inst_is_noop: orphaned value detected\n");
       return TRUE;
    }
 
@@ -178,7 +178,7 @@ nv_pc_pass_pre_emission(void *priv, struct nv_basic_block *b)
    }
    pc->emit_size += b->emit_size;
 
-#ifdef NOUVEAU_DEBUG
+#if NV50_DEBUG & NV50_DEBUG_PROG_IR
    if (!b->entry)
       debug_printf("BB:%i is now empty\n", b->id);
    else
@@ -206,7 +206,7 @@ nvc0_pc_exec_pass2(struct nv_pc *pc)
 {
    int i, ret;
 
-   NOUVEAU_DBG("preparing %u blocks for emission\n", pc->num_blocks);
+   NV50_DBGMSG(PROG_IR, "preparing %u blocks for emission\n", pc->num_blocks);
 
    pc->num_blocks = 0; /* will reorder bb_list */
 

src/gallium/drivers/nvc0/nvc0_pc_regalloc.c

@@ -20,11 +20,11 @@
  * SOFTWARE.
  */
 
-#define NOUVEAU_DEBUG 1
-
-/* #define NVC0_RA_DEBUG_LIVEI */
-/* #define NVC0_RA_DEBUG_LIVE_SETS */
-/* #define NVC0_RA_DEBUG_JOIN */
+#if NV50_DEBUG & NV50_DEBUG_PROG_RA
+# define NVC0_RA_DEBUG_LIVEI
+# define NVC0_RA_DEBUG_LIVE_SETS
+# define NVC0_RA_DEBUG_JOIN
+#endif
 
 #include "nvc0_pc.h"
 #include "util/u_simple_list.h"
@@ -504,7 +504,9 @@ pass_generate_phi_movs(struct nv_pc_pass *ctx, struct nv_basic_block *b)
       }
 
       if (pn != p && pn->exit) {
-         ctx->pc->current_block = b->in[n ? 0 : 1];
+         assert(!b->in[!n]->exit || b->in[!n]->exit->terminator);
+         /* insert terminator (branch to ENDIF) in new else block */
+         ctx->pc->current_block = pn;
          ni = new_instruction(ctx->pc, NV_OP_BRA);
          ni->target = b;
          ni->terminator = 1;
@@ -965,7 +967,7 @@ nv_pc_pass1(struct nv_pc *pc, struct nv_basic_block *root)
    struct nv_pc_pass *ctx;
    int i, ret;
 
-   NOUVEAU_DBG("REGISTER ALLOCATION - entering\n");
+   NV50_DBGMSG(PROG_RA, "REGISTER ALLOCATION - entering\n");
 
    ctx = CALLOC_STRUCT(nv_pc_pass);
    if (!ctx)
@@ -1031,7 +1033,7 @@ nv_pc_pass1(struct nv_pc *pc, struct nv_basic_block *root)
    for (i = 0; i < pc->num_values; ++i)
       livei_release(&pc->values[i]);
 
-   NOUVEAU_DBG("REGISTER ALLOCATION - leaving\n");
+   NV50_DBGMSG(PROG_RA, "REGISTER ALLOCATION - leaving\n");
 
 out:
    FREE(ctx->insns);

src/gallium/drivers/nvc0/nvc0_program.c

@@ -23,8 +23,6 @@
 #include "pipe/p_shader_tokens.h"
 #include "pipe/p_defines.h"
 
-#define NOUVEAU_DEBUG
-
 #include "tgsi/tgsi_parse.h"
 #include "tgsi/tgsi_util.h"
 #include "tgsi/tgsi_dump.h"
@@ -224,6 +222,9 @@ nvc0_interp_mode(const struct tgsi_full_declaration *decl)
    else
    if (decl->Declaration.Interpolate == TGSI_INTERPOLATE_PERSPECTIVE)
       mode = NVC0_INTERP_PERSPECTIVE;
+   else
+   if (decl->Declaration.Semantic && decl->Semantic.Name == TGSI_SEMANTIC_COLOR)
+      mode = NVC0_INTERP_PERSPECTIVE;
    else
       mode = NVC0_INTERP_LINEAR;
 
@@ -574,7 +575,7 @@ nvc0_prog_scan(struct nvc0_translation_info *ti)
    int ret;
    unsigned i;
 
-#ifdef NOUVEAU_DEBUG
+#if NV50_DEBUG & NV50_DEBUG_SHADER
    tgsi_dump(prog->pipe.tokens, 0);
 #endif
 
@@ -691,12 +692,12 @@ nvc0_program_translate(struct nvc0_program *prog)
    if (ret)
       NOUVEAU_ERR("shader translation failed\n");
 
-   {
-      unsigned i;
-      for (i = 0; i < sizeof(prog->hdr) / sizeof(prog->hdr[0]); ++i)
-         debug_printf("HDR[%02lx] = 0x%08x\n",
-                      i * sizeof(prog->hdr[0]), prog->hdr[i]);
-   }
+#if NV50_DEBUG & NV50_DEBUG_SHADER
+   unsigned i;
+   for (i = 0; i < sizeof(prog->hdr) / sizeof(prog->hdr[0]); ++i)
+      debug_printf("HDR[%02lx] = 0x%08x\n",
+                   i * sizeof(prog->hdr[0]), prog->hdr[i]);
+#endif
 
 out:
    if (ti->immd32)