Merge branch 'mesa_7_7_branch'
Conflicts:
src/gallium/auxiliary/draw/draw_context.c
src/gallium/auxiliary/draw/draw_pt_fetch_shade_pipeline.c
src/gallium/auxiliary/pipebuffer/Makefile
src/gallium/auxiliary/pipebuffer/SConscript
src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c
src/gallium/auxiliary/tgsi/tgsi_scan.c
src/gallium/drivers/i915/i915_surface.c
src/gallium/drivers/i915/i915_texture.c
src/gallium/drivers/llvmpipe/lp_setup.c
src/gallium/drivers/llvmpipe/lp_tex_sample_c.c
src/gallium/drivers/llvmpipe/lp_texture.c
src/gallium/drivers/softpipe/sp_prim_vbuf.c
src/gallium/state_trackers/xorg/xorg_dri2.c
src/gallium/winsys/drm/intel/gem/intel_drm_api.c
src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c
src/gallium/winsys/drm/radeon/core/radeon_drm.c
src/gallium/winsys/drm/vmware/core/vmw_screen_dri.c
src/mesa/state_tracker/st_cb_clear.c
diff --git a/docs/lists.html b/docs/lists.html
index 5227fbd..9c17a9f 100644
--- a/docs/lists.html
+++ b/docs/lists.html
@@ -13,36 +13,41 @@
</p>
<ul>
-<li><a href="https://lists.sourceforge.net/lists/listinfo/mesa3d-announce"
-target="_parent">mesa3d-announce</a> - announcements of new Mesa
-versions are sent to this list.
-</li>
-<br>
<li><a href="https://lists.sourceforge.net/lists/listinfo/mesa3d-users"
-target="_parent">mesa3d-users</a> - intended for users of the Mesa and DRI.
-Newbie questions are appropriate, but please try the general OpenGL
+target="_parent">mesa3d-users</a> - intended for end-users of Mesa and DRI
+drivers. Newbie questions are OK, but please try the general OpenGL
resources and Mesa/DRI documentation first.
</li>
<br>
<li><a href="https://lists.sourceforge.net/lists/listinfo/mesa3d-dev"
-target="_parent">mesa3d-dev</a> - for discussion of Mesa and Direct Rendering
-Infrastructure development. Not for beginners.
+target="_parent">mesa3d-dev</a> - for Mesa, Gallium and DRI development
+discussion. Not for beginners.
</li>
<br>
<li><a href="http://lists.freedesktop.org/mailman/listinfo/mesa-commit"
target="_parent">mesa-commit</a> - relays git check-in messages
(for developers).
+In general, people should not post to this list.
+</li>
<br>
-Note: the old mesa3d-cvs list is no longer in use.
+<li><a href="https://lists.sourceforge.net/lists/listinfo/mesa3d-announce"
+target="_parent">mesa3d-announce</a> - announcements of new Mesa
+versions are sent to this list. Very low traffic.
</li>
</ul>
+<p>
+Follow the links above for list archives.
+</p>
+
<p>For mailing lists about Direct Rendering Modules (drm) in Linux/BSD
-kernels, see <a href="http://dri.freedesktop.org/wiki/MailingLists">wiki</a>.
+kernels, see the
+<a href="http://dri.freedesktop.org/wiki/MailingLists" target="_parent">
+DRI wiki</a>.
+</p>
<p>
-<b>Notice</b>: non-member posts to any of these lists will be automatically
-rejected.
+<b>Notice</b>: You must subscribe to these lists in order to post to them.
</p>
diff --git a/progs/trivial/tri-fbo-tex-mip.c b/progs/trivial/tri-fbo-tex-mip.c
index df4725c..2e8fb74 100644
--- a/progs/trivial/tri-fbo-tex-mip.c
+++ b/progs/trivial/tri-fbo-tex-mip.c
@@ -6,7 +6,6 @@
#include <assert.h>
#include <stdio.h>
#include <stdlib.h>
-#include <math.h>
/* For debug */
diff --git a/progs/xdemos/glxgears.c b/progs/xdemos/glxgears.c
index 088f25a..2993c82 100644
--- a/progs/xdemos/glxgears.c
+++ b/progs/xdemos/glxgears.c
@@ -771,6 +771,7 @@
glDeleteLists(gear1, 1);
glDeleteLists(gear2, 1);
glDeleteLists(gear3, 1);
+ glXMakeCurrent(dpy, None, NULL);
glXDestroyContext(dpy, ctx);
XDestroyWindow(dpy, win);
XCloseDisplay(dpy);
diff --git a/src/gallium/auxiliary/Makefile b/src/gallium/auxiliary/Makefile
index 8f937e3..9b6babc 100644
--- a/src/gallium/auxiliary/Makefile
+++ b/src/gallium/auxiliary/Makefile
@@ -48,12 +48,10 @@
draw/draw_vs_sse.c \
indices/u_indices_gen.c \
indices/u_unfilled_gen.c \
- pipebuffer/pb_buffer_fenced.c \
pipebuffer/pb_buffer_malloc.c \
pipebuffer/pb_bufmgr_alt.c \
pipebuffer/pb_bufmgr_cache.c \
pipebuffer/pb_bufmgr_debug.c \
- pipebuffer/pb_bufmgr_fenced.c \
pipebuffer/pb_bufmgr_mm.c \
pipebuffer/pb_bufmgr_ondemand.c \
pipebuffer/pb_bufmgr_pool.c \
diff --git a/src/gallium/auxiliary/draw/draw_context.c b/src/gallium/auxiliary/draw/draw_context.c
index e90dfc5..fb1bc05 100644
--- a/src/gallium/auxiliary/draw/draw_context.c
+++ b/src/gallium/auxiliary/draw/draw_context.c
@@ -34,7 +34,6 @@
#include "util/u_memory.h"
#include "util/u_math.h"
#include "draw_context.h"
-#include "draw_vbuf.h"
#include "draw_vs.h"
#include "draw_gs.h"
#include "draw_pt.h"
diff --git a/src/gallium/auxiliary/draw/draw_pipe.c b/src/gallium/auxiliary/draw/draw_pipe.c
index 1c6d657..11d6485 100644
--- a/src/gallium/auxiliary/draw/draw_pipe.c
+++ b/src/gallium/auxiliary/draw/draw_pipe.c
@@ -106,10 +106,9 @@
-
-
-
-
+/**
+ * Build primitive to render a point with vertex at v0.
+ */
static void do_point( struct draw_context *draw,
const char *v0 )
{
@@ -123,6 +122,10 @@
}
+/**
+ * Build primitive to render a line with vertices at v0, v1.
+ * \param flags bitmask of DRAW_PIPE_EDGE_x, DRAW_PIPE_RESET_STIPPLE
+ */
static void do_line( struct draw_context *draw,
ushort flags,
const char *v0,
@@ -139,6 +142,10 @@
}
+/**
+ * Build primitive to render a triangle with vertices at v0, v1, v2.
+ * \param flags bitmask of DRAW_PIPE_EDGE_x, DRAW_PIPE_RESET_STIPPLE
+ */
static void do_triangle( struct draw_context *draw,
ushort flags,
char *v0,
@@ -157,7 +164,10 @@
}
-
+/*
+ * Set up macros for draw_pt_decompose.h template code.
+ * This code uses vertex indexes / elements.
+ */
#define QUAD(i0,i1,i2,i3) \
do_triangle( draw, \
( DRAW_PIPE_RESET_STIPPLE | \
@@ -175,16 +185,16 @@
#define TRIANGLE(flags,i0,i1,i2) \
do_triangle( draw, \
- elts[i0], /* flags */ \
+ elts[i0], /* flags */ \
verts + stride * (elts[i0] & ~DRAW_PIPE_FLAG_MASK), \
- verts + stride * elts[i1], \
- verts + stride * elts[i2])
+ verts + stride * (elts[i1] & ~DRAW_PIPE_FLAG_MASK), \
+ verts + stride * (elts[i2] & ~DRAW_PIPE_FLAG_MASK) );
#define LINE(flags,i0,i1) \
do_line( draw, \
- elts[i0], \
+ elts[i0], \
verts + stride * (elts[i0] & ~DRAW_PIPE_FLAG_MASK), \
- verts + stride * elts[i1])
+ verts + stride * (elts[i1] & ~DRAW_PIPE_FLAG_MASK) );
#define POINT(i0) \
do_point( draw, \
@@ -213,7 +223,9 @@
-/* Code to run the pipeline on a fairly arbitary collection of vertices.
+/**
+ * Code to run the pipeline on a fairly arbitary collection of vertices.
+ * For drawing indexed primitives.
*
* Vertex headers must be pre-initialized with the
* UNDEFINED_VERTEX_ID, this code will cause that id to become
@@ -243,6 +255,12 @@
draw->pipeline.vertex_count = 0;
}
+
+
+/*
+ * Set up macros for draw_pt_decompose.h template code.
+ * This code is for non-indexed rendering (no elts).
+ */
#define QUAD(i0,i1,i2,i3) \
do_triangle( draw, \
( DRAW_PIPE_RESET_STIPPLE | \
@@ -293,6 +311,10 @@
#include "draw_pt_decompose.h"
+
+/*
+ * For drawing non-indexed primitives.
+ */
void draw_pipeline_run_linear( struct draw_context *draw,
unsigned prim,
struct vertex_header *vertices,
diff --git a/src/gallium/auxiliary/draw/draw_pt.c b/src/gallium/auxiliary/draw/draw_pt.c
index a5ddec5..f5ed32d 100644
--- a/src/gallium/auxiliary/draw/draw_pt.c
+++ b/src/gallium/auxiliary/draw/draw_pt.c
@@ -33,7 +33,6 @@
#include "draw/draw_context.h"
#include "draw/draw_private.h"
#include "draw/draw_pt.h"
-#include "draw/draw_vs.h"
#include "tgsi/tgsi_dump.h"
#include "util/u_math.h"
#include "util/u_prim.h"
diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch.c b/src/gallium/auxiliary/draw/draw_pt_fetch.c
index 55e7a7b..252be50 100644
--- a/src/gallium/auxiliary/draw/draw_pt_fetch.c
+++ b/src/gallium/auxiliary/draw/draw_pt_fetch.c
@@ -30,7 +30,6 @@
#include "draw/draw_context.h"
#include "draw/draw_private.h"
#include "draw/draw_vbuf.h"
-#include "draw/draw_vertex.h"
#include "draw/draw_pt.h"
#include "translate/translate.h"
#include "translate/translate_cache.h"
diff --git a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c
index 734c05f..c5dfbcf 100644
--- a/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c
+++ b/src/gallium/auxiliary/draw/draw_pt_fetch_shade_emit.c
@@ -40,7 +40,6 @@
#include "draw/draw_pt.h"
#include "draw/draw_vs.h"
-#include "translate/translate.h"
struct fetch_shade_emit;
diff --git a/src/gallium/auxiliary/draw/draw_pt_post_vs.c b/src/gallium/auxiliary/draw/draw_pt_post_vs.c
index 5515182..9728d5c 100644
--- a/src/gallium/auxiliary/draw/draw_pt_post_vs.c
+++ b/src/gallium/auxiliary/draw/draw_pt_post_vs.c
@@ -30,7 +30,6 @@
#include "draw/draw_context.h"
#include "draw/draw_private.h"
#include "draw/draw_vbuf.h"
-#include "draw/draw_vertex.h"
#include "draw/draw_pt.h"
struct pt_post_vs {
diff --git a/src/gallium/auxiliary/draw/draw_vs_varient.c b/src/gallium/auxiliary/draw/draw_vs_varient.c
index 9f40030..b87a465 100644
--- a/src/gallium/auxiliary/draw/draw_vs_varient.c
+++ b/src/gallium/auxiliary/draw/draw_vs_varient.c
@@ -38,7 +38,6 @@
#include "draw/draw_vertex.h"
#include "draw/draw_vs.h"
#include "translate/translate.h"
-#include "translate/translate_cache.h"
/* A first pass at incorporating vertex fetch/emit functionality into
*/
diff --git a/src/gallium/auxiliary/pipebuffer/Makefile b/src/gallium/auxiliary/pipebuffer/Makefile
new file mode 100644
index 0000000..21d25d2
--- /dev/null
+++ b/src/gallium/auxiliary/pipebuffer/Makefile
@@ -0,0 +1,18 @@
+TOP = ../../../..
+include $(TOP)/configs/current
+
+LIBNAME = pipebuffer
+
+C_SOURCES = \
+ pb_buffer_fenced.c \
+ pb_buffer_malloc.c \
+ pb_bufmgr_alt.c \
+ pb_bufmgr_cache.c \
+ pb_bufmgr_debug.c \
+ pb_bufmgr_mm.c \
+ pb_bufmgr_ondemand.c \
+ pb_bufmgr_pool.c \
+ pb_bufmgr_slab.c \
+ pb_validate.c
+
+include ../../Makefile.template
diff --git a/src/gallium/auxiliary/pipebuffer/SConscript b/src/gallium/auxiliary/pipebuffer/SConscript
new file mode 100644
index 0000000..a074a55
--- /dev/null
+++ b/src/gallium/auxiliary/pipebuffer/SConscript
@@ -0,0 +1,18 @@
+Import('*')
+
+pipebuffer = env.ConvenienceLibrary(
+ target = 'pipebuffer',
+ source = [
+ 'pb_buffer_fenced.c',
+ 'pb_buffer_malloc.c',
+ 'pb_bufmgr_alt.c',
+ 'pb_bufmgr_cache.c',
+ 'pb_bufmgr_debug.c',
+ 'pb_bufmgr_mm.c',
+ 'pb_bufmgr_ondemand.c',
+ 'pb_bufmgr_pool.c',
+ 'pb_bufmgr_slab.c',
+ 'pb_validate.c',
+ ])
+
+auxiliaries.insert(0, pipebuffer)
diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c
index a4b78f1..1ac424d 100644
--- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c
+++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.c
@@ -1,6 +1,7 @@
/**************************************************************************
*
* Copyright 2007-2009 VMware, Inc.
+ * Copyright 2007-2010 VMware, Inc.
* All Rights Reserved.
*
* Permission is hereby granted, free of charge, to any person obtaining a
@@ -28,9 +29,9 @@
/**
* \file
* Implementation of fenced buffers.
- *
- * \author Jose Fonseca <jrfonseca-at-tungstengraphics-dot-com>
- * \author Thomas Hellström <thomas-at-tungstengraphics-dot-com>
+ *
+ * \author Jose Fonseca <jfonseca-at-vmware-dot-com>
+ * \author Thomas Hellström <thellstrom-at-vmware-dot-com>
*/
@@ -50,6 +51,7 @@
#include "pb_buffer.h"
#include "pb_buffer_fenced.h"
+#include "pb_bufmgr.h"
@@ -59,32 +61,79 @@
#define SUPER(__derived) (&(__derived)->base)
-struct fenced_buffer_list
+struct fenced_manager
{
- pipe_mutex mutex;
-
+ struct pb_manager base;
+ struct pb_manager *provider;
struct pb_fence_ops *ops;
-
- pb_size numDelayed;
- struct list_head delayed;
-
-#ifdef DEBUG
- pb_size numUnfenced;
+
+ /**
+ * Maximum buffer size that can be safely allocated.
+ */
+ pb_size max_buffer_size;
+
+ /**
+ * Maximum cpu memory we can allocate before we start waiting for the
+ * GPU to idle.
+ */
+ pb_size max_cpu_total_size;
+
+ /**
+ * Following members are mutable and protected by this mutex.
+ */
+ pipe_mutex mutex;
+
+ /**
+ * Fenced buffer list.
+ *
+ * All fenced buffers are placed in this listed, ordered from the oldest
+ * fence to the newest fence.
+ */
+ struct list_head fenced;
+ pb_size num_fenced;
+
struct list_head unfenced;
-#endif
+ pb_size num_unfenced;
+
+ /**
+ * How much temporary CPU memory is being used to hold unvalidated buffers.
+ */
+ pb_size cpu_total_size;
};
/**
+ * Fenced buffer.
+ *
* Wrapper around a pipe buffer which adds fencing and reference counting.
*/
struct fenced_buffer
{
- struct pb_buffer base;
-
- struct pb_buffer *buffer;
+ /*
+ * Immutable members.
+ */
- /* FIXME: protect access with mutex */
+ struct pb_buffer base;
+ struct fenced_manager *mgr;
+
+ /*
+ * Following members are mutable and protected by fenced_manager::mutex.
+ */
+
+ struct list_head head;
+
+ /**
+ * Buffer with storage.
+ */
+ struct pb_buffer *buffer;
+ pb_size size;
+ struct pb_desc desc;
+
+ /**
+ * Temporary CPU storage data. Used when there isn't enough GPU memory to
+ * store the buffer.
+ */
+ void *data;
/**
* A bitmask of PIPE_BUFFER_USAGE_CPU/GPU_READ/WRITE describing the current
@@ -93,15 +142,22 @@
unsigned flags;
unsigned mapcount;
+
struct pb_validate *vl;
unsigned validation_flags;
- struct pipe_fence_handle *fence;
- struct list_head head;
- struct fenced_buffer_list *list;
+ struct pipe_fence_handle *fence;
};
+static INLINE struct fenced_manager *
+fenced_manager(struct pb_manager *mgr)
+{
+ assert(mgr);
+ return (struct fenced_manager *)mgr;
+}
+
+
static INLINE struct fenced_buffer *
fenced_buffer(struct pb_buffer *buf)
{
@@ -110,221 +166,569 @@
}
-static INLINE void
-_fenced_buffer_add(struct fenced_buffer *fenced_buf)
-{
- struct fenced_buffer_list *fenced_list = fenced_buf->list;
+static void
+fenced_buffer_destroy_cpu_storage_locked(struct fenced_buffer *fenced_buf);
- assert(pipe_is_referenced(&fenced_buf->base.base.reference));
- assert(fenced_buf->flags & PIPE_BUFFER_USAGE_GPU_READ_WRITE);
- assert(fenced_buf->fence);
+static enum pipe_error
+fenced_buffer_create_cpu_storage_locked(struct fenced_manager *fenced_mgr,
+ struct fenced_buffer *fenced_buf);
-#ifdef DEBUG
- LIST_DEL(&fenced_buf->head);
- assert(fenced_list->numUnfenced);
- --fenced_list->numUnfenced;
-#endif
- LIST_ADDTAIL(&fenced_buf->head, &fenced_list->delayed);
- ++fenced_list->numDelayed;
-}
+static void
+fenced_buffer_destroy_gpu_storage_locked(struct fenced_buffer *fenced_buf);
+
+static enum pipe_error
+fenced_buffer_create_gpu_storage_locked(struct fenced_manager *fenced_mgr,
+ struct fenced_buffer *fenced_buf,
+ boolean wait);
+
+static enum pipe_error
+fenced_buffer_copy_storage_to_gpu_locked(struct fenced_buffer *fenced_buf);
+
+static enum pipe_error
+fenced_buffer_copy_storage_to_cpu_locked(struct fenced_buffer *fenced_buf);
/**
- * Actually destroy the buffer.
+ * Dump the fenced buffer list.
+ *
+ * Useful to understand failures to allocate buffers.
*/
-static INLINE void
-_fenced_buffer_destroy(struct fenced_buffer *fenced_buf)
+static void
+fenced_manager_dump_locked(struct fenced_manager *fenced_mgr)
{
- struct fenced_buffer_list *fenced_list = fenced_buf->list;
-
- assert(!pipe_is_referenced(&fenced_buf->base.base.reference));
- assert(!fenced_buf->fence);
#ifdef DEBUG
+ struct pb_fence_ops *ops = fenced_mgr->ops;
+ struct list_head *curr, *next;
+ struct fenced_buffer *fenced_buf;
+
+ debug_printf("%10s %7s %8s %7s %10s %s\n",
+ "buffer", "size", "refcount", "storage", "fence", "signalled");
+
+ curr = fenced_mgr->unfenced.next;
+ next = curr->next;
+ while(curr != &fenced_mgr->unfenced) {
+ fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head);
+ assert(!fenced_buf->fence);
+ debug_printf("%10p %7u %8u %7s\n",
+ (void *) fenced_buf,
+ fenced_buf->base.base.size,
+ p_atomic_read(&fenced_buf->base.base.reference.count),
+ fenced_buf->buffer ? "gpu" : (fenced_buf->data ? "cpu" : "none"));
+ curr = next;
+ next = curr->next;
+ }
+
+ curr = fenced_mgr->fenced.next;
+ next = curr->next;
+ while(curr != &fenced_mgr->fenced) {
+ int signaled;
+ fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head);
+ assert(fenced_buf->buffer);
+ signaled = ops->fence_signalled(ops, fenced_buf->fence, 0);
+ debug_printf("%10p %7u %8u %7s %10p %s\n",
+ (void *) fenced_buf,
+ fenced_buf->base.base.size,
+ p_atomic_read(&fenced_buf->base.base.reference.count),
+ "gpu",
+ (void *) fenced_buf->fence,
+ signaled == 0 ? "y" : "n");
+ curr = next;
+ next = curr->next;
+ }
+#else
+ (void)fenced_mgr;
+#endif
+}
+
+
+static INLINE void
+fenced_buffer_destroy_locked(struct fenced_manager *fenced_mgr,
+ struct fenced_buffer *fenced_buf)
+{
+ assert(!pipe_is_referenced(&fenced_buf->base.base.reference));
+
+ assert(!fenced_buf->fence);
assert(fenced_buf->head.prev);
assert(fenced_buf->head.next);
LIST_DEL(&fenced_buf->head);
- assert(fenced_list->numUnfenced);
- --fenced_list->numUnfenced;
-#else
- (void)fenced_list;
-#endif
- pb_reference(&fenced_buf->buffer, NULL);
+ assert(fenced_mgr->num_unfenced);
+ --fenced_mgr->num_unfenced;
+
+ fenced_buffer_destroy_gpu_storage_locked(fenced_buf);
+ fenced_buffer_destroy_cpu_storage_locked(fenced_buf);
+
FREE(fenced_buf);
}
+/**
+ * Add the buffer to the fenced list.
+ *
+ * Reference count should be incremented before calling this function.
+ */
static INLINE void
-_fenced_buffer_remove(struct fenced_buffer_list *fenced_list,
- struct fenced_buffer *fenced_buf)
+fenced_buffer_add_locked(struct fenced_manager *fenced_mgr,
+ struct fenced_buffer *fenced_buf)
{
- struct pb_fence_ops *ops = fenced_list->ops;
-
+ assert(pipe_is_referenced(&fenced_buf->base.base.reference));
+ assert(fenced_buf->flags & PIPE_BUFFER_USAGE_GPU_READ_WRITE);
assert(fenced_buf->fence);
- assert(fenced_buf->list == fenced_list);
-
- ops->fence_reference(ops, &fenced_buf->fence, NULL);
- fenced_buf->flags &= ~PIPE_BUFFER_USAGE_GPU_READ_WRITE;
-
- assert(fenced_buf->head.prev);
- assert(fenced_buf->head.next);
-
- LIST_DEL(&fenced_buf->head);
- assert(fenced_list->numDelayed);
- --fenced_list->numDelayed;
-
-#ifdef DEBUG
- LIST_ADDTAIL(&fenced_buf->head, &fenced_list->unfenced);
- ++fenced_list->numUnfenced;
-#endif
-
- /**
- * FIXME!!!
- */
- if(!pipe_is_referenced(&fenced_buf->base.base.reference))
- _fenced_buffer_destroy(fenced_buf);
+ p_atomic_inc(&fenced_buf->base.base.reference.count);
+
+ LIST_DEL(&fenced_buf->head);
+ assert(fenced_mgr->num_unfenced);
+ --fenced_mgr->num_unfenced;
+ LIST_ADDTAIL(&fenced_buf->head, &fenced_mgr->fenced);
+ ++fenced_mgr->num_fenced;
}
-static INLINE enum pipe_error
-_fenced_buffer_finish(struct fenced_buffer *fenced_buf)
+/**
+ * Remove the buffer from the fenced list, and potentially destroy the buffer
+ * if the reference count reaches zero.
+ *
+ * Returns TRUE if the buffer was detroyed.
+ */
+static INLINE boolean
+fenced_buffer_remove_locked(struct fenced_manager *fenced_mgr,
+ struct fenced_buffer *fenced_buf)
{
- struct fenced_buffer_list *fenced_list = fenced_buf->list;
- struct pb_fence_ops *ops = fenced_list->ops;
+ struct pb_fence_ops *ops = fenced_mgr->ops;
+
+ assert(fenced_buf->fence);
+ assert(fenced_buf->mgr == fenced_mgr);
+
+ ops->fence_reference(ops, &fenced_buf->fence, NULL);
+ fenced_buf->flags &= ~PIPE_BUFFER_USAGE_GPU_READ_WRITE;
+
+ assert(fenced_buf->head.prev);
+ assert(fenced_buf->head.next);
+
+ LIST_DEL(&fenced_buf->head);
+ assert(fenced_mgr->num_fenced);
+ --fenced_mgr->num_fenced;
+
+ LIST_ADDTAIL(&fenced_buf->head, &fenced_mgr->unfenced);
+ ++fenced_mgr->num_unfenced;
+
+ if (p_atomic_dec_zero(&fenced_buf->base.base.reference.count)) {
+ fenced_buffer_destroy_locked(fenced_mgr, fenced_buf);
+ return TRUE;
+ }
+
+ return FALSE;
+}
+
+
+/**
+ * Wait for the fence to expire, and remove it from the fenced list.
+ *
+ * This function will release and re-aquire the mutex, so any copy of mutable
+ * state must be discarded after calling it.
+ */
+static INLINE enum pipe_error
+fenced_buffer_finish_locked(struct fenced_manager *fenced_mgr,
+ struct fenced_buffer *fenced_buf)
+{
+ struct pb_fence_ops *ops = fenced_mgr->ops;
+ enum pipe_error ret = PIPE_ERROR;
#if 0
debug_warning("waiting for GPU");
#endif
+ assert(pipe_is_referenced(&fenced_buf->base.base.reference));
assert(fenced_buf->fence);
+
if(fenced_buf->fence) {
- if(ops->fence_finish(ops, fenced_buf->fence, 0) != 0) {
- return PIPE_ERROR;
+ struct pipe_fence_handle *fence = NULL;
+ int finished;
+ boolean proceed;
+
+ ops->fence_reference(ops, &fence, fenced_buf->fence);
+
+ pipe_mutex_unlock(fenced_mgr->mutex);
+
+ finished = ops->fence_finish(ops, fenced_buf->fence, 0);
+
+ pipe_mutex_lock(fenced_mgr->mutex);
+
+ assert(pipe_is_referenced(&fenced_buf->base.base.reference));
+
+ /*
+ * Only proceed if the fence object didn't change in the meanwhile.
+ * Otherwise assume the work has been already carried out by another
+ * thread that re-aquired the lock before us.
+ */
+ proceed = fence == fenced_buf->fence ? TRUE : FALSE;
+
+ ops->fence_reference(ops, &fence, NULL);
+
+ if(proceed && finished == 0) {
+ /*
+ * Remove from the fenced list
+ */
+
+ boolean destroyed;
+
+ destroyed = fenced_buffer_remove_locked(fenced_mgr, fenced_buf);
+
+ /* TODO: remove consequents buffers with the same fence? */
+
+ assert(!destroyed);
+
+ fenced_buf->flags &= ~PIPE_BUFFER_USAGE_GPU_READ_WRITE;
+
+ ret = PIPE_OK;
}
- /* Remove from the fenced list */
- /* TODO: remove consequents */
- _fenced_buffer_remove(fenced_list, fenced_buf);
}
- fenced_buf->flags &= ~PIPE_BUFFER_USAGE_GPU_READ_WRITE;
+ return ret;
+}
+
+
+/**
+ * Remove as many fenced buffers from the fenced list as possible.
+ *
+ * Returns TRUE if at least one buffer was removed.
+ */
+static boolean
+fenced_manager_check_signalled_locked(struct fenced_manager *fenced_mgr,
+ boolean wait)
+{
+ struct pb_fence_ops *ops = fenced_mgr->ops;
+ struct list_head *curr, *next;
+ struct fenced_buffer *fenced_buf;
+ struct pb_buffer *pb_buf;
+ struct pipe_fence_handle *prev_fence = NULL;
+ boolean ret = FALSE;
+
+ curr = fenced_mgr->fenced.next;
+ next = curr->next;
+ while(curr != &fenced_mgr->fenced) {
+ fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head);
+
+ if(fenced_buf->fence != prev_fence) {
+ int signaled;
+
+ if (wait) {
+ signaled = ops->fence_finish(ops, fenced_buf->fence, 0);
+
+ /*
+ * Don't return just now. Instead preemptively check if the
+ * following buffers' fences already expired, without further waits.
+ */
+ wait = FALSE;
+ }
+ else {
+ signaled = ops->fence_signalled(ops, fenced_buf->fence, 0);
+ }
+
+ if (signaled != 0) {
+ return ret;
+ }
+
+ prev_fence = fenced_buf->fence;
+ }
+ else {
+ /* This buffer's fence object is identical to the previous buffer's
+ * fence object, so no need to check the fence again.
+ */
+ assert(ops->fence_signalled(ops, fenced_buf->fence, 0) == 0);
+ }
+
+ fenced_buffer_remove_locked(fenced_mgr, fenced_buf);
+
+ ret = TRUE;
+
+ curr = next;
+ next = curr->next;
+ }
+
+ return ret;
+}
+
+
+/**
+ * Try to free some GPU memory by backing it up into CPU memory.
+ *
+ * Returns TRUE if at least one buffer was freed.
+ */
+static boolean
+fenced_manager_free_gpu_storage_locked(struct fenced_manager *fenced_mgr)
+{
+ struct list_head *curr, *next;
+ struct fenced_buffer *fenced_buf;
+
+ curr = fenced_mgr->unfenced.next;
+ next = curr->next;
+ while(curr != &fenced_mgr->unfenced) {
+ fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head);
+
+ /*
+ * We can only move storage if the buffer is not mapped and not
+ * validated.
+ */
+ if(fenced_buf->buffer &&
+ !fenced_buf->mapcount &&
+ !fenced_buf->vl) {
+ enum pipe_error ret;
+
+ ret = fenced_buffer_create_cpu_storage_locked(fenced_mgr, fenced_buf);
+ if(ret == PIPE_OK) {
+ ret = fenced_buffer_copy_storage_to_cpu_locked(fenced_buf);
+ if(ret == PIPE_OK) {
+ fenced_buffer_destroy_gpu_storage_locked(fenced_buf);
+ return TRUE;
+ }
+ fenced_buffer_destroy_cpu_storage_locked(fenced_buf);
+ }
+ }
+
+ curr = next;
+ next = curr->next;
+ }
+
+ return FALSE;
+}
+
+
+/**
+ * Destroy CPU storage for this buffer.
+ */
+static void
+fenced_buffer_destroy_cpu_storage_locked(struct fenced_buffer *fenced_buf)
+{
+ if(fenced_buf->data) {
+ align_free(fenced_buf->data);
+ fenced_buf->data = NULL;
+ assert(fenced_buf->mgr->cpu_total_size >= fenced_buf->size);
+ fenced_buf->mgr->cpu_total_size -= fenced_buf->size;
+ }
+}
+
+
+/**
+ * Create CPU storage for this buffer.
+ */
+static enum pipe_error
+fenced_buffer_create_cpu_storage_locked(struct fenced_manager *fenced_mgr,
+ struct fenced_buffer *fenced_buf)
+{
+ assert(!fenced_buf->data);
+ if(fenced_buf->data)
+ return PIPE_OK;
+
+ if (fenced_mgr->cpu_total_size + fenced_buf->size > fenced_mgr->max_cpu_total_size)
+ return PIPE_ERROR_OUT_OF_MEMORY;
+
+ fenced_buf->data = align_malloc(fenced_buf->size, fenced_buf->desc.alignment);
+ if(!fenced_buf->data)
+ return PIPE_ERROR_OUT_OF_MEMORY;
+
+ fenced_mgr->cpu_total_size += fenced_buf->size;
+
return PIPE_OK;
}
/**
- * Free as many fenced buffers from the list head as possible.
+ * Destroy the GPU storage.
*/
static void
-_fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list,
- int wait)
+fenced_buffer_destroy_gpu_storage_locked(struct fenced_buffer *fenced_buf)
{
- struct pb_fence_ops *ops = fenced_list->ops;
- struct list_head *curr, *next;
- struct fenced_buffer *fenced_buf;
- struct pb_buffer *pb_buf;
- struct pipe_fence_handle *prev_fence = NULL;
-
- curr = fenced_list->delayed.next;
- next = curr->next;
- while(curr != &fenced_list->delayed) {
- fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head);
-
- if(fenced_buf->fence != prev_fence) {
- int signaled;
- if (wait)
- signaled = ops->fence_finish(ops, fenced_buf->fence, 0);
- else
- signaled = ops->fence_signalled(ops, fenced_buf->fence, 0);
- if (signaled != 0)
- break;
- prev_fence = fenced_buf->fence;
- }
- else {
- assert(ops->fence_signalled(ops, fenced_buf->fence, 0) == 0);
- }
-
- _fenced_buffer_remove(fenced_list, fenced_buf);
-
- curr = next;
- next = curr->next;
+ if(fenced_buf->buffer) {
+ pb_reference(&fenced_buf->buffer, NULL);
}
}
+/**
+ * Try to create GPU storage for this buffer.
+ *
+ * This function is a shorthand around pb_manager::create_buffer for
+ * fenced_buffer_create_gpu_storage_locked()'s benefit.
+ */
+static INLINE boolean
+fenced_buffer_try_create_gpu_storage_locked(struct fenced_manager *fenced_mgr,
+ struct fenced_buffer *fenced_buf)
+{
+ struct pb_manager *provider = fenced_mgr->provider;
+
+ assert(!fenced_buf->buffer);
+
+ fenced_buf->buffer = provider->create_buffer(fenced_mgr->provider,
+ fenced_buf->size,
+ &fenced_buf->desc);
+ return fenced_buf->buffer ? TRUE : FALSE;
+}
+
+
+/**
+ * Create GPU storage for this buffer.
+ */
+static enum pipe_error
+fenced_buffer_create_gpu_storage_locked(struct fenced_manager *fenced_mgr,
+ struct fenced_buffer *fenced_buf,
+ boolean wait)
+{
+ assert(!fenced_buf->buffer);
+
+ /*
+ * Check for signaled buffers before trying to allocate.
+ */
+ fenced_manager_check_signalled_locked(fenced_mgr, FALSE);
+
+ fenced_buffer_try_create_gpu_storage_locked(fenced_mgr, fenced_buf);
+
+ /*
+ * Keep trying while there is some sort of progress:
+ * - fences are expiring,
+ * - or buffers are being being swapped out from GPU memory into CPU memory.
+ */
+ while(!fenced_buf->buffer &&
+ (fenced_manager_check_signalled_locked(fenced_mgr, FALSE) ||
+ fenced_manager_free_gpu_storage_locked(fenced_mgr))) {
+ fenced_buffer_try_create_gpu_storage_locked(fenced_mgr, fenced_buf);
+ }
+
+ if(!fenced_buf->buffer && wait) {
+ /*
+ * Same as before, but this time around, wait to free buffers if
+ * necessary.
+ */
+ while(!fenced_buf->buffer &&
+ (fenced_manager_check_signalled_locked(fenced_mgr, TRUE) ||
+ fenced_manager_free_gpu_storage_locked(fenced_mgr))) {
+ fenced_buffer_try_create_gpu_storage_locked(fenced_mgr, fenced_buf);
+ }
+ }
+
+ if(!fenced_buf->buffer) {
+ if(0)
+ fenced_manager_dump_locked(fenced_mgr);
+
+ /* give up */
+ return PIPE_ERROR_OUT_OF_MEMORY;
+ }
+
+ return PIPE_OK;
+}
+
+
+static enum pipe_error
+fenced_buffer_copy_storage_to_gpu_locked(struct fenced_buffer *fenced_buf)
+{
+ uint8_t *map;
+
+ assert(fenced_buf->data);
+ assert(fenced_buf->buffer);
+
+ map = pb_map(fenced_buf->buffer, PIPE_BUFFER_USAGE_CPU_WRITE);
+ if(!map)
+ return PIPE_ERROR;
+
+ memcpy(map, fenced_buf->data, fenced_buf->size);
+
+ pb_unmap(fenced_buf->buffer);
+
+ return PIPE_OK;
+}
+
+
+static enum pipe_error
+fenced_buffer_copy_storage_to_cpu_locked(struct fenced_buffer *fenced_buf)
+{
+ const uint8_t *map;
+
+ assert(fenced_buf->data);
+ assert(fenced_buf->buffer);
+
+ map = pb_map(fenced_buf->buffer, PIPE_BUFFER_USAGE_CPU_READ);
+ if(!map)
+ return PIPE_ERROR;
+
+ memcpy(fenced_buf->data, map, fenced_buf->size);
+
+ pb_unmap(fenced_buf->buffer);
+
+ return PIPE_OK;
+}
+
+
static void
fenced_buffer_destroy(struct pb_buffer *buf)
{
- struct fenced_buffer *fenced_buf = fenced_buffer(buf);
- struct fenced_buffer_list *fenced_list = fenced_buf->list;
+ struct fenced_buffer *fenced_buf = fenced_buffer(buf);
+ struct fenced_manager *fenced_mgr = fenced_buf->mgr;
- pipe_mutex_lock(fenced_list->mutex);
assert(!pipe_is_referenced(&fenced_buf->base.base.reference));
- if (fenced_buf->fence) {
- struct pb_fence_ops *ops = fenced_list->ops;
- if(ops->fence_signalled(ops, fenced_buf->fence, 0) == 0) {
- struct list_head *curr, *prev;
- curr = &fenced_buf->head;
- prev = curr->prev;
- do {
- fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head);
- assert(ops->fence_signalled(ops, fenced_buf->fence, 0) == 0);
- _fenced_buffer_remove(fenced_list, fenced_buf);
- curr = prev;
- prev = curr->prev;
- } while (curr != &fenced_list->delayed);
- }
- else {
- /* delay destruction */
- }
- }
- else {
- _fenced_buffer_destroy(fenced_buf);
- }
- pipe_mutex_unlock(fenced_list->mutex);
+
+ pipe_mutex_lock(fenced_mgr->mutex);
+
+ fenced_buffer_destroy_locked(fenced_mgr, fenced_buf);
+
+ pipe_mutex_unlock(fenced_mgr->mutex);
}
static void *
-fenced_buffer_map(struct pb_buffer *buf,
+fenced_buffer_map(struct pb_buffer *buf,
unsigned flags)
{
struct fenced_buffer *fenced_buf = fenced_buffer(buf);
- struct fenced_buffer_list *fenced_list = fenced_buf->list;
- struct pb_fence_ops *ops = fenced_list->ops;
- void *map;
+ struct fenced_manager *fenced_mgr = fenced_buf->mgr;
+ struct pb_fence_ops *ops = fenced_mgr->ops;
+ void *map = NULL;
+
+ pipe_mutex_lock(fenced_mgr->mutex);
assert(!(flags & PIPE_BUFFER_USAGE_GPU_READ_WRITE));
-
- /* Serialize writes */
- if((fenced_buf->flags & PIPE_BUFFER_USAGE_GPU_WRITE) ||
- ((fenced_buf->flags & PIPE_BUFFER_USAGE_GPU_READ) && (flags & PIPE_BUFFER_USAGE_CPU_WRITE))) {
- if(flags & PIPE_BUFFER_USAGE_DONTBLOCK) {
- /* Don't wait for the GPU to finish writing */
- if(ops->fence_signalled(ops, fenced_buf->fence, 0) == 0)
- _fenced_buffer_remove(fenced_list, fenced_buf);
- else
- return NULL;
+
+ /*
+ * Serialize writes.
+ */
+ while((fenced_buf->flags & PIPE_BUFFER_USAGE_GPU_WRITE) ||
+ ((fenced_buf->flags & PIPE_BUFFER_USAGE_GPU_READ) &&
+ (flags & PIPE_BUFFER_USAGE_CPU_WRITE))) {
+
+ /*
+ * Don't wait for the GPU to finish accessing it, if blocking is forbidden.
+ */
+ if((flags & PIPE_BUFFER_USAGE_DONTBLOCK) &&
+ ops->fence_signalled(ops, fenced_buf->fence, 0) == 0) {
+ goto done;
}
- else {
- /* Wait for the GPU to finish writing */
- _fenced_buffer_finish(fenced_buf);
+
+ if (flags & PIPE_BUFFER_USAGE_UNSYNCHRONIZED) {
+ break;
}
+
+ /*
+ * Wait for the GPU to finish accessing. This will release and re-acquire
+ * the mutex, so all copies of mutable state must be discarded.
+ */
+ fenced_buffer_finish_locked(fenced_mgr, fenced_buf);
}
-#if 0
- /* Check for CPU write access (read is OK) */
- if(fenced_buf->flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE) {
- /* this is legal -- just for debugging */
- debug_warning("concurrent CPU writes");
+ if(fenced_buf->buffer) {
+ map = pb_map(fenced_buf->buffer, flags);
}
-#endif
-
- map = pb_map(fenced_buf->buffer, flags);
+ else {
+ assert(fenced_buf->data);
+ map = fenced_buf->data;
+ }
+
if(map) {
++fenced_buf->mapcount;
fenced_buf->flags |= flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE;
}
+done:
+ pipe_mutex_unlock(fenced_mgr->mutex);
+
return map;
}
@@ -333,13 +737,20 @@
fenced_buffer_unmap(struct pb_buffer *buf)
{
struct fenced_buffer *fenced_buf = fenced_buffer(buf);
+ struct fenced_manager *fenced_mgr = fenced_buf->mgr;
+
+ pipe_mutex_lock(fenced_mgr->mutex);
+
assert(fenced_buf->mapcount);
if(fenced_buf->mapcount) {
- pb_unmap(fenced_buf->buffer);
+ if (fenced_buf->buffer)
+ pb_unmap(fenced_buf->buffer);
--fenced_buf->mapcount;
if(!fenced_buf->mapcount)
fenced_buf->flags &= ~PIPE_BUFFER_USAGE_CPU_READ_WRITE;
}
+
+ pipe_mutex_unlock(fenced_mgr->mutex);
}
@@ -349,48 +760,72 @@
unsigned flags)
{
struct fenced_buffer *fenced_buf = fenced_buffer(buf);
+ struct fenced_manager *fenced_mgr = fenced_buf->mgr;
enum pipe_error ret;
-
+
+ pipe_mutex_lock(fenced_mgr->mutex);
+
if(!vl) {
/* invalidate */
fenced_buf->vl = NULL;
fenced_buf->validation_flags = 0;
- return PIPE_OK;
+ ret = PIPE_OK;
+ goto done;
}
-
+
assert(flags & PIPE_BUFFER_USAGE_GPU_READ_WRITE);
assert(!(flags & ~PIPE_BUFFER_USAGE_GPU_READ_WRITE));
flags &= PIPE_BUFFER_USAGE_GPU_READ_WRITE;
- /* Buffer cannot be validated in two different lists */
- if(fenced_buf->vl && fenced_buf->vl != vl)
- return PIPE_ERROR_RETRY;
-
-#if 0
- /* Do not validate if buffer is still mapped */
- if(fenced_buf->flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE) {
- /* TODO: wait for the thread that mapped the buffer to unmap it */
- return PIPE_ERROR_RETRY;
+ /* Buffer cannot be validated in two different lists */
+ if(fenced_buf->vl && fenced_buf->vl != vl) {
+ ret = PIPE_ERROR_RETRY;
+ goto done;
}
- /* Final sanity checking */
- assert(!(fenced_buf->flags & PIPE_BUFFER_USAGE_CPU_READ_WRITE));
- assert(!fenced_buf->mapcount);
-#endif
if(fenced_buf->vl == vl &&
(fenced_buf->validation_flags & flags) == flags) {
/* Nothing to do -- buffer already validated */
- return PIPE_OK;
+ ret = PIPE_OK;
+ goto done;
}
-
+
+ /*
+ * Create and update GPU storage.
+ */
+ if(!fenced_buf->buffer) {
+ assert(!fenced_buf->mapcount);
+
+ ret = fenced_buffer_create_gpu_storage_locked(fenced_mgr, fenced_buf, TRUE);
+ if(ret != PIPE_OK) {
+ goto done;
+ }
+
+ ret = fenced_buffer_copy_storage_to_gpu_locked(fenced_buf);
+ if(ret != PIPE_OK) {
+ fenced_buffer_destroy_gpu_storage_locked(fenced_buf);
+ goto done;
+ }
+
+ if(fenced_buf->mapcount) {
+ debug_printf("warning: validating a buffer while it is still mapped\n");
+ }
+ else {
+ fenced_buffer_destroy_cpu_storage_locked(fenced_buf);
+ }
+ }
+
ret = pb_validate(fenced_buf->buffer, vl, flags);
if (ret != PIPE_OK)
- return ret;
-
+ goto done;
+
fenced_buf->vl = vl;
fenced_buf->validation_flags |= flags;
-
- return PIPE_OK;
+
+done:
+ pipe_mutex_unlock(fenced_mgr->mutex);
+
+ return ret;
}
@@ -398,36 +833,37 @@
fenced_buffer_fence(struct pb_buffer *buf,
struct pipe_fence_handle *fence)
{
- struct fenced_buffer *fenced_buf;
- struct fenced_buffer_list *fenced_list;
- struct pb_fence_ops *ops;
+ struct fenced_buffer *fenced_buf = fenced_buffer(buf);
+ struct fenced_manager *fenced_mgr = fenced_buf->mgr;
+ struct pb_fence_ops *ops = fenced_mgr->ops;
- fenced_buf = fenced_buffer(buf);
- fenced_list = fenced_buf->list;
- ops = fenced_list->ops;
-
- if(fence == fenced_buf->fence) {
- /* Nothing to do */
- return;
+ pipe_mutex_lock(fenced_mgr->mutex);
+
+ assert(pipe_is_referenced(&fenced_buf->base.base.reference));
+ assert(fenced_buf->buffer);
+
+ if(fence != fenced_buf->fence) {
+ assert(fenced_buf->vl);
+ assert(fenced_buf->validation_flags);
+
+ if (fenced_buf->fence) {
+ boolean destroyed;
+ destroyed = fenced_buffer_remove_locked(fenced_mgr, fenced_buf);
+ assert(!destroyed);
+ }
+ if (fence) {
+ ops->fence_reference(ops, &fenced_buf->fence, fence);
+ fenced_buf->flags |= fenced_buf->validation_flags;
+ fenced_buffer_add_locked(fenced_mgr, fenced_buf);
+ }
+
+ pb_fence(fenced_buf->buffer, fence);
+
+ fenced_buf->vl = NULL;
+ fenced_buf->validation_flags = 0;
}
- assert(fenced_buf->vl);
- assert(fenced_buf->validation_flags);
-
- pipe_mutex_lock(fenced_list->mutex);
- if (fenced_buf->fence)
- _fenced_buffer_remove(fenced_list, fenced_buf);
- if (fence) {
- ops->fence_reference(ops, &fenced_buf->fence, fence);
- fenced_buf->flags |= fenced_buf->validation_flags;
- _fenced_buffer_add(fenced_buf);
- }
- pipe_mutex_unlock(fenced_list->mutex);
-
- pb_fence(fenced_buf->buffer, fence);
-
- fenced_buf->vl = NULL;
- fenced_buf->validation_flags = 0;
+ pipe_mutex_unlock(fenced_mgr->mutex);
}
@@ -437,11 +873,29 @@
pb_size *offset)
{
struct fenced_buffer *fenced_buf = fenced_buffer(buf);
- pb_get_base_buffer(fenced_buf->buffer, base_buf, offset);
+ struct fenced_manager *fenced_mgr = fenced_buf->mgr;
+
+ pipe_mutex_lock(fenced_mgr->mutex);
+
+ /*
+ * This should only be called when the buffer is validated. Typically
+ * when processing relocations.
+ */
+ assert(fenced_buf->vl);
+ assert(fenced_buf->buffer);
+
+ if(fenced_buf->buffer)
+ pb_get_base_buffer(fenced_buf->buffer, base_buf, offset);
+ else {
+ *base_buf = buf;
+ *offset = 0;
+ }
+
+ pipe_mutex_unlock(fenced_mgr->mutex);
}
-static const struct pb_vtbl
+static const struct pb_vtbl
fenced_buffer_vtbl = {
fenced_buffer_destroy,
fenced_buffer_map,
@@ -452,147 +906,166 @@
};
-struct pb_buffer *
-fenced_buffer_create(struct fenced_buffer_list *fenced_list,
- struct pb_buffer *buffer)
+/**
+ * Wrap a buffer in a fenced buffer.
+ */
+static struct pb_buffer *
+fenced_bufmgr_create_buffer(struct pb_manager *mgr,
+ pb_size size,
+ const struct pb_desc *desc)
{
- struct fenced_buffer *buf;
-
- if(!buffer)
- return NULL;
-
- buf = CALLOC_STRUCT(fenced_buffer);
- if(!buf) {
- pb_reference(&buffer, NULL);
- return NULL;
- }
-
- pipe_reference_init(&buf->base.base.reference, 1);
- buf->base.base.alignment = buffer->base.alignment;
- buf->base.base.usage = buffer->base.usage;
- buf->base.base.size = buffer->base.size;
-
- buf->base.vtbl = &fenced_buffer_vtbl;
- buf->buffer = buffer;
- buf->list = fenced_list;
-
-#ifdef DEBUG
- pipe_mutex_lock(fenced_list->mutex);
- LIST_ADDTAIL(&buf->head, &fenced_list->unfenced);
- ++fenced_list->numUnfenced;
- pipe_mutex_unlock(fenced_list->mutex);
-#endif
-
- return &buf->base;
-}
-
-
-struct fenced_buffer_list *
-fenced_buffer_list_create(struct pb_fence_ops *ops)
-{
- struct fenced_buffer_list *fenced_list;
-
- fenced_list = CALLOC_STRUCT(fenced_buffer_list);
- if (!fenced_list)
- return NULL;
-
- fenced_list->ops = ops;
-
- LIST_INITHEAD(&fenced_list->delayed);
- fenced_list->numDelayed = 0;
-
-#ifdef DEBUG
- LIST_INITHEAD(&fenced_list->unfenced);
- fenced_list->numUnfenced = 0;
-#endif
-
- pipe_mutex_init(fenced_list->mutex);
-
- return fenced_list;
-}
-
-
-void
-fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list,
- int wait)
-{
- pipe_mutex_lock(fenced_list->mutex);
- _fenced_buffer_list_check_free(fenced_list, wait);
- pipe_mutex_unlock(fenced_list->mutex);
-}
-
-
-#ifdef DEBUG
-void
-fenced_buffer_list_dump(struct fenced_buffer_list *fenced_list)
-{
- struct pb_fence_ops *ops = fenced_list->ops;
- struct list_head *curr, *next;
+ struct fenced_manager *fenced_mgr = fenced_manager(mgr);
struct fenced_buffer *fenced_buf;
+ enum pipe_error ret;
- pipe_mutex_lock(fenced_list->mutex);
+ /*
+ * Don't stall the GPU, waste time evicting buffers, or waste memory
+ * trying to create a buffer that will most likely never fit into the
+ * graphics aperture.
+ */
+ if(size > fenced_mgr->max_buffer_size) {
+ goto no_buffer;
+ }
- debug_printf("%10s %7s %7s %10s %s\n",
- "buffer", "size", "refcount", "fence", "signalled");
-
- curr = fenced_list->unfenced.next;
- next = curr->next;
- while(curr != &fenced_list->unfenced) {
- fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head);
- assert(!fenced_buf->fence);
- debug_printf("%10p %7u %7u\n",
- (void *) fenced_buf,
- fenced_buf->base.base.size,
- p_atomic_read(&fenced_buf->base.base.reference.count));
- curr = next;
- next = curr->next;
+ fenced_buf = CALLOC_STRUCT(fenced_buffer);
+ if(!fenced_buf)
+ goto no_buffer;
+
+ pipe_reference_init(&fenced_buf->base.base.reference, 1);
+ fenced_buf->base.base.alignment = desc->alignment;
+ fenced_buf->base.base.usage = desc->usage;
+ fenced_buf->base.base.size = size;
+ fenced_buf->size = size;
+ fenced_buf->desc = *desc;
+
+ fenced_buf->base.vtbl = &fenced_buffer_vtbl;
+ fenced_buf->mgr = fenced_mgr;
+
+ pipe_mutex_lock(fenced_mgr->mutex);
+
+ /*
+ * Try to create GPU storage without stalling,
+ */
+ ret = fenced_buffer_create_gpu_storage_locked(fenced_mgr, fenced_buf, FALSE);
+
+ /*
+ * Attempt to use CPU memory to avoid stalling the GPU.
+ */
+ if(ret != PIPE_OK) {
+ ret = fenced_buffer_create_cpu_storage_locked(fenced_mgr, fenced_buf);
}
-
- curr = fenced_list->delayed.next;
- next = curr->next;
- while(curr != &fenced_list->delayed) {
- int signaled;
- fenced_buf = LIST_ENTRY(struct fenced_buffer, curr, head);
- signaled = ops->fence_signalled(ops, fenced_buf->fence, 0);
- debug_printf("%10p %7u %7u %10p %s\n",
- (void *) fenced_buf,
- fenced_buf->base.base.size,
- p_atomic_read(&fenced_buf->base.base.reference.count),
- (void *) fenced_buf->fence,
- signaled == 0 ? "y" : "n");
- curr = next;
- next = curr->next;
+
+ /*
+ * Create GPU storage, waiting for some to be available.
+ */
+ if(ret != PIPE_OK) {
+ ret = fenced_buffer_create_gpu_storage_locked(fenced_mgr, fenced_buf, TRUE);
}
-
- pipe_mutex_unlock(fenced_list->mutex);
+
+ /*
+ * Give up.
+ */
+ if(ret != PIPE_OK) {
+ goto no_storage;
+ }
+
+ assert(fenced_buf->buffer || fenced_buf->data);
+
+ LIST_ADDTAIL(&fenced_buf->head, &fenced_mgr->unfenced);
+ ++fenced_mgr->num_unfenced;
+ pipe_mutex_unlock(fenced_mgr->mutex);
+
+ return &fenced_buf->base;
+
+no_storage:
+ pipe_mutex_unlock(fenced_mgr->mutex);
+ FREE(fenced_buf);
+no_buffer:
+ return NULL;
}
-#endif
-void
-fenced_buffer_list_destroy(struct fenced_buffer_list *fenced_list)
+static void
+fenced_bufmgr_flush(struct pb_manager *mgr)
{
- pipe_mutex_lock(fenced_list->mutex);
+ struct fenced_manager *fenced_mgr = fenced_manager(mgr);
+
+ pipe_mutex_lock(fenced_mgr->mutex);
+ while(fenced_manager_check_signalled_locked(fenced_mgr, TRUE))
+ ;
+ pipe_mutex_unlock(fenced_mgr->mutex);
+
+ assert(fenced_mgr->provider->flush);
+ if(fenced_mgr->provider->flush)
+ fenced_mgr->provider->flush(fenced_mgr->provider);
+}
+
+
+static void
+fenced_bufmgr_destroy(struct pb_manager *mgr)
+{
+ struct fenced_manager *fenced_mgr = fenced_manager(mgr);
+
+ pipe_mutex_lock(fenced_mgr->mutex);
/* Wait on outstanding fences */
- while (fenced_list->numDelayed) {
- pipe_mutex_unlock(fenced_list->mutex);
+ while (fenced_mgr->num_fenced) {
+ pipe_mutex_unlock(fenced_mgr->mutex);
#if defined(PIPE_OS_LINUX) || defined(PIPE_OS_BSD) || defined(PIPE_OS_SOLARIS)
sched_yield();
#endif
- _fenced_buffer_list_check_free(fenced_list, 1);
- pipe_mutex_lock(fenced_list->mutex);
+ pipe_mutex_lock(fenced_mgr->mutex);
+ while(fenced_manager_check_signalled_locked(fenced_mgr, TRUE))
+ ;
}
#ifdef DEBUG
- /*assert(!fenced_list->numUnfenced);*/
+ /*assert(!fenced_mgr->num_unfenced);*/
#endif
-
- pipe_mutex_unlock(fenced_list->mutex);
-
- fenced_list->ops->destroy(fenced_list->ops);
-
- FREE(fenced_list);
+
+ pipe_mutex_unlock(fenced_mgr->mutex);
+ pipe_mutex_destroy(fenced_mgr->mutex);
+
+ if(fenced_mgr->provider)
+ fenced_mgr->provider->destroy(fenced_mgr->provider);
+
+ fenced_mgr->ops->destroy(fenced_mgr->ops);
+
+ FREE(fenced_mgr);
}
+struct pb_manager *
+fenced_bufmgr_create(struct pb_manager *provider,
+ struct pb_fence_ops *ops,
+ pb_size max_buffer_size,
+ pb_size max_cpu_total_size)
+{
+ struct fenced_manager *fenced_mgr;
+
+ if(!provider)
+ return NULL;
+
+ fenced_mgr = CALLOC_STRUCT(fenced_manager);
+ if (!fenced_mgr)
+ return NULL;
+
+ fenced_mgr->base.destroy = fenced_bufmgr_destroy;
+ fenced_mgr->base.create_buffer = fenced_bufmgr_create_buffer;
+ fenced_mgr->base.flush = fenced_bufmgr_flush;
+
+ fenced_mgr->provider = provider;
+ fenced_mgr->ops = ops;
+ fenced_mgr->max_buffer_size = max_buffer_size;
+ fenced_mgr->max_cpu_total_size = max_cpu_total_size;
+
+ LIST_INITHEAD(&fenced_mgr->fenced);
+ fenced_mgr->num_fenced = 0;
+
+ LIST_INITHEAD(&fenced_mgr->unfenced);
+ fenced_mgr->num_unfenced = 0;
+
+ pipe_mutex_init(fenced_mgr->mutex);
+
+ return &fenced_mgr->base;
+}
diff --git a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.h b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.h
index 034ca1e..0372f81 100644
--- a/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.h
+++ b/src/gallium/auxiliary/pipebuffer/pb_buffer_fenced.h
@@ -98,43 +98,6 @@
};
-/**
- * Create a fenced buffer list.
- *
- * See also fenced_bufmgr_create for a more convenient way to use this.
- */
-struct fenced_buffer_list *
-fenced_buffer_list_create(struct pb_fence_ops *ops);
-
-
-/**
- * Walk the fenced buffer list to check and free signalled buffers.
- */
-void
-fenced_buffer_list_check_free(struct fenced_buffer_list *fenced_list,
- int wait);
-
-
-#ifdef DEBUG
-void
-fenced_buffer_list_dump(struct fenced_buffer_list *fenced_list);
-#endif
-
-
-void
-fenced_buffer_list_destroy(struct fenced_buffer_list *fenced_list);
-
-
-/**
- * Wrap a buffer in a fenced buffer.
- *
- * NOTE: this will not increase the buffer reference count.
- */
-struct pb_buffer *
-fenced_buffer_create(struct fenced_buffer_list *fenced,
- struct pb_buffer *buffer);
-
-
#ifdef __cplusplus
}
#endif
diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h
index 8c8d713..0666991 100644
--- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h
+++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr.h
@@ -175,7 +175,9 @@
*/
struct pb_manager *
fenced_bufmgr_create(struct pb_manager *provider,
- struct pb_fence_ops *ops);
+ struct pb_fence_ops *ops,
+ pb_size max_buffer_size,
+ pb_size max_cpu_total_size);
struct pb_manager *
diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c
index 6e3214c..8f74180 100644
--- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c
+++ b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_debug.c
@@ -371,6 +371,9 @@
struct pb_desc real_desc;
pb_size real_size;
+ assert(size);
+ assert(desc->alignment);
+
buf = CALLOC_STRUCT(pb_debug_buffer);
if(!buf)
return NULL;
diff --git a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c b/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c
deleted file mode 100644
index 97dd142..0000000
--- a/src/gallium/auxiliary/pipebuffer/pb_bufmgr_fenced.c
+++ /dev/null
@@ -1,152 +0,0 @@
-/**************************************************************************
- *
- * Copyright 2006 Tungsten Graphics, Inc., Bismarck, ND., USA
- * 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 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
- * THE COPYRIGHT HOLDERS, AUTHORS 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.
- *
- * The above copyright notice and this permission notice (including the
- * next paragraph) shall be included in all copies or substantial portions
- * of the Software.
- *
- *
- **************************************************************************/
-
-/**
- * \file
- * A buffer manager that wraps buffers in fenced buffers.
- *
- * \author Jose Fonseca <jrfonseca@tungstengraphics.dot.com>
- */
-
-
-#include "util/u_debug.h"
-#include "util/u_memory.h"
-
-#include "pb_buffer.h"
-#include "pb_buffer_fenced.h"
-#include "pb_bufmgr.h"
-
-
-struct fenced_pb_manager
-{
- struct pb_manager base;
-
- struct pb_manager *provider;
-
- struct fenced_buffer_list *fenced_list;
-};
-
-
-static INLINE struct fenced_pb_manager *
-fenced_pb_manager(struct pb_manager *mgr)
-{
- assert(mgr);
- return (struct fenced_pb_manager *)mgr;
-}
-
-
-static struct pb_buffer *
-fenced_bufmgr_create_buffer(struct pb_manager *mgr,
- pb_size size,
- const struct pb_desc *desc)
-{
- struct fenced_pb_manager *fenced_mgr = fenced_pb_manager(mgr);
- struct pb_buffer *buf;
- struct pb_buffer *fenced_buf;
-
- /* check for free buffers before allocating new ones */
- fenced_buffer_list_check_free(fenced_mgr->fenced_list, 0);
-
- buf = fenced_mgr->provider->create_buffer(fenced_mgr->provider, size, desc);
- if(!buf) {
- /* try harder to get a buffer */
- fenced_buffer_list_check_free(fenced_mgr->fenced_list, 1);
-
- buf = fenced_mgr->provider->create_buffer(fenced_mgr->provider, size, desc);
- if(!buf) {
-#if 0
- fenced_buffer_list_dump(fenced_mgr->fenced_list);
-#endif
-
- /* give up */
- return NULL;
- }
- }
-
- fenced_buf = fenced_buffer_create(fenced_mgr->fenced_list, buf);
- if(!fenced_buf) {
- pb_reference(&buf, NULL);
- }
-
- return fenced_buf;
-}
-
-
-static void
-fenced_bufmgr_flush(struct pb_manager *mgr)
-{
- struct fenced_pb_manager *fenced_mgr = fenced_pb_manager(mgr);
-
- fenced_buffer_list_check_free(fenced_mgr->fenced_list, TRUE);
-
- assert(fenced_mgr->provider->flush);
- if(fenced_mgr->provider->flush)
- fenced_mgr->provider->flush(fenced_mgr->provider);
-}
-
-
-static void
-fenced_bufmgr_destroy(struct pb_manager *mgr)
-{
- struct fenced_pb_manager *fenced_mgr = fenced_pb_manager(mgr);
-
- fenced_buffer_list_destroy(fenced_mgr->fenced_list);
-
- if(fenced_mgr->provider)
- fenced_mgr->provider->destroy(fenced_mgr->provider);
-
- FREE(fenced_mgr);
-}
-
-
-struct pb_manager *
-fenced_bufmgr_create(struct pb_manager *provider,
- struct pb_fence_ops *ops)
-{
- struct fenced_pb_manager *fenced_mgr;
-
- if(!provider)
- return NULL;
-
- fenced_mgr = CALLOC_STRUCT(fenced_pb_manager);
- if (!fenced_mgr)
- return NULL;
-
- fenced_mgr->base.destroy = fenced_bufmgr_destroy;
- fenced_mgr->base.create_buffer = fenced_bufmgr_create_buffer;
- fenced_mgr->base.flush = fenced_bufmgr_flush;
-
- fenced_mgr->provider = provider;
- fenced_mgr->fenced_list = fenced_buffer_list_create(ops);
- if(!fenced_mgr->fenced_list) {
- FREE(fenced_mgr);
- return NULL;
- }
-
- return &fenced_mgr->base;
-}
diff --git a/src/gallium/auxiliary/pipebuffer/pb_validate.c b/src/gallium/auxiliary/pipebuffer/pb_validate.c
index ce40c0c..903afc7 100644
--- a/src/gallium/auxiliary/pipebuffer/pb_validate.c
+++ b/src/gallium/auxiliary/pipebuffer/pb_validate.c
@@ -39,7 +39,6 @@
#include "util/u_debug.h"
#include "pb_buffer.h"
-#include "pb_buffer_fenced.h"
#include "pb_validate.h"
diff --git a/src/gallium/auxiliary/tgsi/tgsi_scan.c b/src/gallium/auxiliary/tgsi/tgsi_scan.c
index a6cc773..b9be8dc 100644
--- a/src/gallium/auxiliary/tgsi/tgsi_scan.c
+++ b/src/gallium/auxiliary/tgsi/tgsi_scan.c
@@ -101,12 +101,10 @@
src->Register.File == TGSI_FILE_SYSTEM_VALUE) {
const int ind = src->Register.Index;
if (info->input_semantic_name[ind] == TGSI_SEMANTIC_FOG) {
- if (src->Register.SwizzleX == TGSI_SWIZZLE_X) {
- info->uses_fogcoord = TRUE;
- }
- else if (src->Register.SwizzleX == TGSI_SWIZZLE_Y) {
- info->uses_frontfacing = TRUE;
- }
+ info->uses_fogcoord = TRUE;
+ }
+ else if (info->input_semantic_name[ind] == TGSI_SEMANTIC_FACE) {
+ info->uses_frontfacing = TRUE;
}
}
}
diff --git a/src/gallium/auxiliary/util/u_blit.c b/src/gallium/auxiliary/util/u_blit.c
index 9725890..236f1e4 100644
--- a/src/gallium/auxiliary/util/u_blit.c
+++ b/src/gallium/auxiliary/util/u_blit.c
@@ -226,8 +226,8 @@
offset = get_next_slot( ctx );
- pipe_buffer_write(ctx->pipe->screen, ctx->vbuf,
- offset, sizeof(ctx->vertices), ctx->vertices);
+ pipe_buffer_write_nooverlap(ctx->pipe->screen, ctx->vbuf,
+ offset, sizeof(ctx->vertices), ctx->vertices);
return offset;
}
diff --git a/src/gallium/auxiliary/util/u_gen_mipmap.c b/src/gallium/auxiliary/util/u_gen_mipmap.c
index 7602379..5426c91 100644
--- a/src/gallium/auxiliary/util/u_gen_mipmap.c
+++ b/src/gallium/auxiliary/util/u_gen_mipmap.c
@@ -1411,8 +1411,8 @@
offset = get_next_slot( ctx );
- pipe_buffer_write(ctx->pipe->screen, ctx->vbuf,
- offset, sizeof(ctx->vertices), ctx->vertices);
+ pipe_buffer_write_nooverlap(ctx->pipe->screen, ctx->vbuf,
+ offset, sizeof(ctx->vertices), ctx->vertices);
return offset;
}
diff --git a/src/gallium/auxiliary/util/u_upload_mgr.c b/src/gallium/auxiliary/util/u_upload_mgr.c
index 975ee89..55a6537 100644
--- a/src/gallium/auxiliary/util/u_upload_mgr.c
+++ b/src/gallium/auxiliary/util/u_upload_mgr.c
@@ -85,7 +85,9 @@
map = pipe_buffer_map_range(screen, buf, offset, size,
PIPE_BUFFER_USAGE_CPU_WRITE |
- PIPE_BUFFER_USAGE_FLUSH_EXPLICIT);
+ PIPE_BUFFER_USAGE_FLUSH_EXPLICIT |
+ PIPE_BUFFER_USAGE_DISCARD |
+ PIPE_BUFFER_USAGE_UNSYNCHRONIZED);
if (map == NULL)
return PIPE_ERROR_OUT_OF_MEMORY;
diff --git a/src/gallium/drivers/i915/i915_clear.c b/src/gallium/drivers/i915/i915_clear.c
index 90530f2..0d0859f 100644
--- a/src/gallium/drivers/i915/i915_clear.c
+++ b/src/gallium/drivers/i915/i915_clear.c
@@ -32,7 +32,6 @@
#include "util/u_clear.h"
#include "i915_context.h"
-#include "i915_state.h"
/**
diff --git a/src/gallium/drivers/i915/i915_context.c b/src/gallium/drivers/i915/i915_context.c
index 89feead..4c5ff37 100644
--- a/src/gallium/drivers/i915/i915_context.c
+++ b/src/gallium/drivers/i915/i915_context.c
@@ -29,12 +29,9 @@
#include "i915_state.h"
#include "i915_screen.h"
#include "i915_batch.h"
-#include "i915_texture.h"
-#include "i915_reg.h"
#include "draw/draw_context.h"
#include "pipe/p_defines.h"
-#include "pipe/internal/p_winsys_screen.h"
#include "pipe/p_inlines.h"
#include "util/u_memory.h"
#include "pipe/p_screen.h"
diff --git a/src/gallium/drivers/i915/i915_debug.c b/src/gallium/drivers/i915/i915_debug.c
index c6e6d6f..237654d 100644
--- a/src/gallium/drivers/i915/i915_debug.c
+++ b/src/gallium/drivers/i915/i915_debug.c
@@ -29,7 +29,6 @@
#include "i915_context.h"
#include "i915_debug.h"
#include "i915_batch.h"
-#include "pipe/internal/p_winsys_screen.h"
#include "util/u_debug.h"
diff --git a/src/gallium/drivers/i915/i915_debug_fp.c b/src/gallium/drivers/i915/i915_debug_fp.c
index 9c5b117..f9c40d8 100644
--- a/src/gallium/drivers/i915/i915_debug_fp.c
+++ b/src/gallium/drivers/i915/i915_debug_fp.c
@@ -29,7 +29,6 @@
#include "i915_reg.h"
#include "i915_debug.h"
#include "pipe/internal/p_winsys_screen.h"
-#include "util/u_memory.h"
static void
diff --git a/src/gallium/drivers/i915/i915_state.c b/src/gallium/drivers/i915/i915_state.c
index 0fab6e1..06949c1 100644
--- a/src/gallium/drivers/i915/i915_state.c
+++ b/src/gallium/drivers/i915/i915_state.c
@@ -30,7 +30,6 @@
#include "draw/draw_context.h"
-#include "pipe/internal/p_winsys_screen.h"
#include "pipe/p_inlines.h"
#include "util/u_math.h"
#include "util/u_memory.h"
@@ -38,7 +37,6 @@
#include "i915_context.h"
#include "i915_reg.h"
-#include "i915_state.h"
#include "i915_state_inlines.h"
#include "i915_fpc.h"
diff --git a/src/gallium/drivers/i915/i915_state_derived.c b/src/gallium/drivers/i915/i915_state_derived.c
index 03dd509..f5b0e9f 100644
--- a/src/gallium/drivers/i915/i915_state_derived.c
+++ b/src/gallium/drivers/i915/i915_state_derived.c
@@ -33,7 +33,6 @@
#include "i915_context.h"
#include "i915_state.h"
#include "i915_reg.h"
-#include "i915_fpc.h"
diff --git a/src/gallium/drivers/i915/i915_state_sampler.c b/src/gallium/drivers/i915/i915_state_sampler.c
index cbac417..e5c6d87 100644
--- a/src/gallium/drivers/i915/i915_state_sampler.c
+++ b/src/gallium/drivers/i915/i915_state_sampler.c
@@ -27,7 +27,6 @@
#include "pipe/p_context.h"
#include "pipe/p_state.h"
-#include "util/u_memory.h"
#include "i915_state_inlines.h"
#include "i915_context.h"
diff --git a/src/gallium/drivers/i915/i915_surface.c b/src/gallium/drivers/i915/i915_surface.c
index c693eb3..13684aa 100644
--- a/src/gallium/drivers/i915/i915_surface.c
+++ b/src/gallium/drivers/i915/i915_surface.c
@@ -27,14 +27,10 @@
#include "i915_context.h"
#include "i915_blit.h"
-#include "i915_state.h"
#include "pipe/p_defines.h"
#include "pipe/p_inlines.h"
-#include "pipe/p_inlines.h"
#include "pipe/internal/p_winsys_screen.h"
#include "util/u_format.h"
-#include "util/u_tile.h"
-#include "util/u_rect.h"
/* Assumes all values are within bounds -- no checking at this level -
diff --git a/src/gallium/drivers/i915/i915_texture.c b/src/gallium/drivers/i915/i915_texture.c
index 50a9e19..441bc4f 100644
--- a/src/gallium/drivers/i915/i915_texture.c
+++ b/src/gallium/drivers/i915/i915_texture.c
@@ -41,7 +41,6 @@
#include "i915_context.h"
#include "i915_texture.h"
-#include "i915_debug.h"
#include "i915_screen.h"
#include "intel_winsys.h"
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_alpha.c b/src/gallium/drivers/llvmpipe/lp_bld_alpha.c
index 2b4bc5c..7245730 100644
--- a/src/gallium/drivers/llvmpipe/lp_bld_alpha.c
+++ b/src/gallium/drivers/llvmpipe/lp_bld_alpha.c
@@ -35,7 +35,6 @@
#include "lp_bld_type.h"
#include "lp_bld_const.h"
-#include "lp_bld_arit.h"
#include "lp_bld_logic.h"
#include "lp_bld_flow.h"
#include "lp_bld_debug.h"
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_arit.c b/src/gallium/drivers/llvmpipe/lp_bld_arit.c
index eea6b5d..1aee9b3 100644
--- a/src/gallium/drivers/llvmpipe/lp_bld_arit.c
+++ b/src/gallium/drivers/llvmpipe/lp_bld_arit.c
@@ -56,7 +56,6 @@
#include "lp_bld_intr.h"
#include "lp_bld_logic.h"
#include "lp_bld_pack.h"
-#include "lp_bld_debug.h"
#include "lp_bld_arit.h"
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_blend_soa.c b/src/gallium/drivers/llvmpipe/lp_bld_blend_soa.c
index 9511299..5d5ca7a 100644
--- a/src/gallium/drivers/llvmpipe/lp_bld_blend_soa.c
+++ b/src/gallium/drivers/llvmpipe/lp_bld_blend_soa.c
@@ -71,7 +71,6 @@
#include "pipe/p_state.h"
#include "lp_bld_type.h"
-#include "lp_bld_const.h"
#include "lp_bld_arit.h"
#include "lp_bld_blend.h"
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_conv.c b/src/gallium/drivers/llvmpipe/lp_bld_conv.c
index 9935209..ebf554c 100644
--- a/src/gallium/drivers/llvmpipe/lp_bld_conv.c
+++ b/src/gallium/drivers/llvmpipe/lp_bld_conv.c
@@ -63,11 +63,9 @@
#include "util/u_debug.h"
#include "util/u_math.h"
-#include "util/u_cpu_detect.h"
#include "lp_bld_type.h"
#include "lp_bld_const.h"
-#include "lp_bld_intr.h"
#include "lp_bld_arit.h"
#include "lp_bld_pack.h"
#include "lp_bld_conv.h"
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_format_aos.c b/src/gallium/drivers/llvmpipe/lp_bld_format_aos.c
index 10e82f1..dfa080b 100644
--- a/src/gallium/drivers/llvmpipe/lp_bld_format_aos.c
+++ b/src/gallium/drivers/llvmpipe/lp_bld_format_aos.c
@@ -38,7 +38,6 @@
#include "lp_bld_type.h"
#include "lp_bld_const.h"
-#include "lp_bld_logic.h"
#include "lp_bld_swizzle.h"
#include "lp_bld_format.h"
diff --git a/src/gallium/drivers/llvmpipe/lp_bld_tgsi_soa.c b/src/gallium/drivers/llvmpipe/lp_bld_tgsi_soa.c
index fb1eda4..85e3b1b 100644
--- a/src/gallium/drivers/llvmpipe/lp_bld_tgsi_soa.c
+++ b/src/gallium/drivers/llvmpipe/lp_bld_tgsi_soa.c
@@ -47,13 +47,11 @@
#include "tgsi/tgsi_exec.h"
#include "lp_bld_type.h"
#include "lp_bld_const.h"
-#include "lp_bld_intr.h"
#include "lp_bld_arit.h"
#include "lp_bld_logic.h"
#include "lp_bld_swizzle.h"
#include "lp_bld_flow.h"
#include "lp_bld_tgsi.h"
-#include "lp_bld_debug.h"
#define LP_MAX_TEMPS 256
diff --git a/src/gallium/drivers/llvmpipe/lp_draw_arrays.c b/src/gallium/drivers/llvmpipe/lp_draw_arrays.c
index c152b44..a031619 100644
--- a/src/gallium/drivers/llvmpipe/lp_draw_arrays.c
+++ b/src/gallium/drivers/llvmpipe/lp_draw_arrays.c
@@ -33,8 +33,6 @@
#include "pipe/p_defines.h"
#include "pipe/p_context.h"
-#include "pipe/internal/p_winsys_screen.h"
-#include "pipe/p_inlines.h"
#include "util/u_prim.h"
#include "lp_buffer.h"
diff --git a/src/gallium/drivers/llvmpipe/lp_flush.c b/src/gallium/drivers/llvmpipe/lp_flush.c
index cd8381f..6c81012 100644
--- a/src/gallium/drivers/llvmpipe/lp_flush.c
+++ b/src/gallium/drivers/llvmpipe/lp_flush.c
@@ -37,8 +37,6 @@
#include "lp_surface.h"
#include "lp_state.h"
#include "lp_tile_cache.h"
-#include "lp_tex_cache.h"
-#include "lp_winsys.h"
void
diff --git a/src/gallium/drivers/llvmpipe/lp_jit.c b/src/gallium/drivers/llvmpipe/lp_jit.c
index 4ef0783..9e0118c 100644
--- a/src/gallium/drivers/llvmpipe/lp_jit.c
+++ b/src/gallium/drivers/llvmpipe/lp_jit.c
@@ -39,7 +39,6 @@
#include "util/u_cpu_detect.h"
#include "lp_screen.h"
#include "lp_bld_intr.h"
-#include "lp_bld_misc.h"
#include "lp_jit.h"
diff --git a/src/gallium/drivers/llvmpipe/lp_state_fs.c b/src/gallium/drivers/llvmpipe/lp_state_fs.c
index 9f4bbef..70e2a08 100644
--- a/src/gallium/drivers/llvmpipe/lp_state_fs.c
+++ b/src/gallium/drivers/llvmpipe/lp_state_fs.c
@@ -62,7 +62,6 @@
#include "util/u_memory.h"
#include "util/u_format.h"
#include "util/u_debug_dump.h"
-#include "pipe/internal/p_winsys_screen.h"
#include "pipe/p_shader_tokens.h"
#include "draw/draw_context.h"
#include "tgsi/tgsi_dump.h"
@@ -85,7 +84,6 @@
#include "lp_context.h"
#include "lp_buffer.h"
#include "lp_state.h"
-#include "lp_quad.h"
#include "lp_tex_sample.h"
#include "lp_debug.h"
diff --git a/src/gallium/drivers/llvmpipe/lp_state_surface.c b/src/gallium/drivers/llvmpipe/lp_state_surface.c
index e37ff04..3b08b0d 100644
--- a/src/gallium/drivers/llvmpipe/lp_state_surface.c
+++ b/src/gallium/drivers/llvmpipe/lp_state_surface.c
@@ -30,7 +30,6 @@
#include "lp_context.h"
#include "lp_state.h"
-#include "lp_surface.h"
#include "lp_tile_cache.h"
#include "draw/draw_context.h"
diff --git a/src/gallium/drivers/llvmpipe/lp_state_vertex.c b/src/gallium/drivers/llvmpipe/lp_state_vertex.c
index 1a17631..57ac25e 100644
--- a/src/gallium/drivers/llvmpipe/lp_state_vertex.c
+++ b/src/gallium/drivers/llvmpipe/lp_state_vertex.c
@@ -31,7 +31,6 @@
#include "lp_context.h"
#include "lp_state.h"
-#include "lp_surface.h"
#include "draw/draw_context.h"
diff --git a/src/gallium/drivers/llvmpipe/lp_test_blend.c b/src/gallium/drivers/llvmpipe/lp_test_blend.c
index 6c29e8d..7b65bab 100644
--- a/src/gallium/drivers/llvmpipe/lp_test_blend.c
+++ b/src/gallium/drivers/llvmpipe/lp_test_blend.c
@@ -38,7 +38,6 @@
#include "lp_bld_type.h"
-#include "lp_bld_arit.h"
#include "lp_bld_blend.h"
#include "lp_bld_debug.h"
#include "lp_test.h"
diff --git a/src/gallium/drivers/llvmpipe/lp_tex_cache.c b/src/gallium/drivers/llvmpipe/lp_tex_cache.c
index a6d9a2c..8094625 100644
--- a/src/gallium/drivers/llvmpipe/lp_tex_cache.c
+++ b/src/gallium/drivers/llvmpipe/lp_tex_cache.c
@@ -38,7 +38,6 @@
#include "util/u_format.h"
#include "util/u_math.h"
#include "lp_context.h"
-#include "lp_surface.h"
#include "lp_texture.h"
#include "lp_tex_cache.h"
diff --git a/src/gallium/drivers/llvmpipe/lp_tex_sample_c.c b/src/gallium/drivers/llvmpipe/lp_tex_sample_c.c
new file mode 100644
index 0000000..d1f5d95
--- /dev/null
+++ b/src/gallium/drivers/llvmpipe/lp_tex_sample_c.c
@@ -0,0 +1,1712 @@
+/**************************************************************************
+ *
+ * Copyright 2007 Tungsten Graphics, Inc., Cedar Park, Texas.
+ * All Rights Reserved.
+ * Copyright 2008 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 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.
+ *
+ **************************************************************************/
+
+/**
+ * Texture sampling
+ *
+ * Authors:
+ * Brian Paul
+ */
+
+#include "lp_context.h"
+#include "lp_quad.h"
+#include "lp_texture.h"
+#include "lp_tex_sample.h"
+#include "lp_tex_cache.h"
+#include "pipe/p_context.h"
+#include "pipe/p_defines.h"
+#include "pipe/p_shader_tokens.h"
+#include "util/u_math.h"
+#include "util/u_memory.h"
+
+
+
+/*
+ * Note, the FRAC macro has to work perfectly. Otherwise you'll sometimes
+ * see 1-pixel bands of improperly weighted linear-filtered textures.
+ * The tests/texwrap.c demo is a good test.
+ * Also note, FRAC(x) doesn't truly return the fractional part of x for x < 0.
+ * Instead, if x < 0 then FRAC(x) = 1 - true_frac(x).
+ */
+#define FRAC(f) ((f) - util_ifloor(f))
+
+
+/**
+ * Linear interpolation macro
+ */
+static INLINE float
+lerp(float a, float v0, float v1)
+{
+ return v0 + a * (v1 - v0);
+}
+
+
+/**
+ * Do 2D/biliner interpolation of float values.
+ * v00, v10, v01 and v11 are typically four texture samples in a square/box.
+ * a and b are the horizontal and vertical interpolants.
+ * It's important that this function is inlined when compiled with
+ * optimization! If we find that's not true on some systems, convert
+ * to a macro.
+ */
+static INLINE float
+lerp_2d(float a, float b,
+ float v00, float v10, float v01, float v11)
+{
+ const float temp0 = lerp(a, v00, v10);
+ const float temp1 = lerp(a, v01, v11);
+ return lerp(b, temp0, temp1);
+}
+
+
+/**
+ * As above, but 3D interpolation of 8 values.
+ */
+static INLINE float
+lerp_3d(float a, float b, float c,
+ float v000, float v100, float v010, float v110,
+ float v001, float v101, float v011, float v111)
+{
+ const float temp0 = lerp_2d(a, b, v000, v100, v010, v110);
+ const float temp1 = lerp_2d(a, b, v001, v101, v011, v111);
+ return lerp(c, temp0, temp1);
+}
+
+
+
+/**
+ * If A is a signed integer, A % B doesn't give the right value for A < 0
+ * (in terms of texture repeat). Just casting to unsigned fixes that.
+ */
+#define REMAINDER(A, B) ((unsigned) (A) % (unsigned) (B))
+
+
+/**
+ * Apply texture coord wrapping mode and return integer texture indexes
+ * for a vector of four texcoords (S or T or P).
+ * \param wrapMode PIPE_TEX_WRAP_x
+ * \param s the incoming texcoords
+ * \param size the texture image size
+ * \param icoord returns the integer texcoords
+ * \return integer texture index
+ */
+static INLINE void
+nearest_texcoord_4(unsigned wrapMode, const float s[4], unsigned size,
+ int icoord[4])
+{
+ uint ch;
+ switch (wrapMode) {
+ case PIPE_TEX_WRAP_REPEAT:
+ /* s limited to [0,1) */
+ /* i limited to [0,size-1] */
+ for (ch = 0; ch < 4; ch++) {
+ int i = util_ifloor(s[ch] * size);
+ icoord[ch] = REMAINDER(i, size);
+ }
+ return;
+ case PIPE_TEX_WRAP_CLAMP:
+ /* s limited to [0,1] */
+ /* i limited to [0,size-1] */
+ for (ch = 0; ch < 4; ch++) {
+ if (s[ch] <= 0.0F)
+ icoord[ch] = 0;
+ else if (s[ch] >= 1.0F)
+ icoord[ch] = size - 1;
+ else
+ icoord[ch] = util_ifloor(s[ch] * size);
+ }
+ return;
+ case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
+ {
+ /* s limited to [min,max] */
+ /* i limited to [0, size-1] */
+ const float min = 1.0F / (2.0F * size);
+ const float max = 1.0F - min;
+ for (ch = 0; ch < 4; ch++) {
+ if (s[ch] < min)
+ icoord[ch] = 0;
+ else if (s[ch] > max)
+ icoord[ch] = size - 1;
+ else
+ icoord[ch] = util_ifloor(s[ch] * size);
+ }
+ }
+ return;
+ case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
+ {
+ /* s limited to [min,max] */
+ /* i limited to [-1, size] */
+ const float min = -1.0F / (2.0F * size);
+ const float max = 1.0F - min;
+ for (ch = 0; ch < 4; ch++) {
+ if (s[ch] <= min)
+ icoord[ch] = -1;
+ else if (s[ch] >= max)
+ icoord[ch] = size;
+ else
+ icoord[ch] = util_ifloor(s[ch] * size);
+ }
+ }
+ return;
+ case PIPE_TEX_WRAP_MIRROR_REPEAT:
+ {
+ const float min = 1.0F / (2.0F * size);
+ const float max = 1.0F - min;
+ for (ch = 0; ch < 4; ch++) {
+ const int flr = util_ifloor(s[ch]);
+ float u;
+ if (flr & 1)
+ u = 1.0F - (s[ch] - (float) flr);
+ else
+ u = s[ch] - (float) flr;
+ if (u < min)
+ icoord[ch] = 0;
+ else if (u > max)
+ icoord[ch] = size - 1;
+ else
+ icoord[ch] = util_ifloor(u * size);
+ }
+ }
+ return;
+ case PIPE_TEX_WRAP_MIRROR_CLAMP:
+ for (ch = 0; ch < 4; ch++) {
+ /* s limited to [0,1] */
+ /* i limited to [0,size-1] */
+ const float u = fabsf(s[ch]);
+ if (u <= 0.0F)
+ icoord[ch] = 0;
+ else if (u >= 1.0F)
+ icoord[ch] = size - 1;
+ else
+ icoord[ch] = util_ifloor(u * size);
+ }
+ return;
+ case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE:
+ {
+ /* s limited to [min,max] */
+ /* i limited to [0, size-1] */
+ const float min = 1.0F / (2.0F * size);
+ const float max = 1.0F - min;
+ for (ch = 0; ch < 4; ch++) {
+ const float u = fabsf(s[ch]);
+ if (u < min)
+ icoord[ch] = 0;
+ else if (u > max)
+ icoord[ch] = size - 1;
+ else
+ icoord[ch] = util_ifloor(u * size);
+ }
+ }
+ return;
+ case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER:
+ {
+ /* s limited to [min,max] */
+ /* i limited to [0, size-1] */
+ const float min = -1.0F / (2.0F * size);
+ const float max = 1.0F - min;
+ for (ch = 0; ch < 4; ch++) {
+ const float u = fabsf(s[ch]);
+ if (u < min)
+ icoord[ch] = -1;
+ else if (u > max)
+ icoord[ch] = size;
+ else
+ icoord[ch] = util_ifloor(u * size);
+ }
+ }
+ return;
+ default:
+ assert(0);
+ }
+}
+
+
+/**
+ * Used to compute texel locations for linear sampling for four texcoords.
+ * \param wrapMode PIPE_TEX_WRAP_x
+ * \param s the texcoords
+ * \param size the texture image size
+ * \param icoord0 returns first texture indexes
+ * \param icoord1 returns second texture indexes (usually icoord0 + 1)
+ * \param w returns blend factor/weight between texture indexes
+ * \param icoord returns the computed integer texture coords
+ */
+static INLINE void
+linear_texcoord_4(unsigned wrapMode, const float s[4], unsigned size,
+ int icoord0[4], int icoord1[4], float w[4])
+{
+ uint ch;
+
+ switch (wrapMode) {
+ case PIPE_TEX_WRAP_REPEAT:
+ for (ch = 0; ch < 4; ch++) {
+ float u = s[ch] * size - 0.5F;
+ icoord0[ch] = REMAINDER(util_ifloor(u), size);
+ icoord1[ch] = REMAINDER(icoord0[ch] + 1, size);
+ w[ch] = FRAC(u);
+ }
+ break;;
+ case PIPE_TEX_WRAP_CLAMP:
+ for (ch = 0; ch < 4; ch++) {
+ float u = CLAMP(s[ch], 0.0F, 1.0F);
+ u = u * size - 0.5f;
+ icoord0[ch] = util_ifloor(u);
+ icoord1[ch] = icoord0[ch] + 1;
+ w[ch] = FRAC(u);
+ }
+ break;;
+ case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
+ for (ch = 0; ch < 4; ch++) {
+ float u = CLAMP(s[ch], 0.0F, 1.0F);
+ u = u * size - 0.5f;
+ icoord0[ch] = util_ifloor(u);
+ icoord1[ch] = icoord0[ch] + 1;
+ if (icoord0[ch] < 0)
+ icoord0[ch] = 0;
+ if (icoord1[ch] >= (int) size)
+ icoord1[ch] = size - 1;
+ w[ch] = FRAC(u);
+ }
+ break;;
+ case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
+ {
+ const float min = -1.0F / (2.0F * size);
+ const float max = 1.0F - min;
+ for (ch = 0; ch < 4; ch++) {
+ float u = CLAMP(s[ch], min, max);
+ u = u * size - 0.5f;
+ icoord0[ch] = util_ifloor(u);
+ icoord1[ch] = icoord0[ch] + 1;
+ w[ch] = FRAC(u);
+ }
+ }
+ break;;
+ case PIPE_TEX_WRAP_MIRROR_REPEAT:
+ for (ch = 0; ch < 4; ch++) {
+ const int flr = util_ifloor(s[ch]);
+ float u;
+ if (flr & 1)
+ u = 1.0F - (s[ch] - (float) flr);
+ else
+ u = s[ch] - (float) flr;
+ u = u * size - 0.5F;
+ icoord0[ch] = util_ifloor(u);
+ icoord1[ch] = icoord0[ch] + 1;
+ if (icoord0[ch] < 0)
+ icoord0[ch] = 0;
+ if (icoord1[ch] >= (int) size)
+ icoord1[ch] = size - 1;
+ w[ch] = FRAC(u);
+ }
+ break;;
+ case PIPE_TEX_WRAP_MIRROR_CLAMP:
+ for (ch = 0; ch < 4; ch++) {
+ float u = fabsf(s[ch]);
+ if (u >= 1.0F)
+ u = (float) size;
+ else
+ u *= size;
+ u -= 0.5F;
+ icoord0[ch] = util_ifloor(u);
+ icoord1[ch] = icoord0[ch] + 1;
+ w[ch] = FRAC(u);
+ }
+ break;;
+ case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_EDGE:
+ for (ch = 0; ch < 4; ch++) {
+ float u = fabsf(s[ch]);
+ if (u >= 1.0F)
+ u = (float) size;
+ else
+ u *= size;
+ u -= 0.5F;
+ icoord0[ch] = util_ifloor(u);
+ icoord1[ch] = icoord0[ch] + 1;
+ if (icoord0[ch] < 0)
+ icoord0[ch] = 0;
+ if (icoord1[ch] >= (int) size)
+ icoord1[ch] = size - 1;
+ w[ch] = FRAC(u);
+ }
+ break;;
+ case PIPE_TEX_WRAP_MIRROR_CLAMP_TO_BORDER:
+ {
+ const float min = -1.0F / (2.0F * size);
+ const float max = 1.0F - min;
+ for (ch = 0; ch < 4; ch++) {
+ float u = fabsf(s[ch]);
+ if (u <= min)
+ u = min * size;
+ else if (u >= max)
+ u = max * size;
+ else
+ u *= size;
+ u -= 0.5F;
+ icoord0[ch] = util_ifloor(u);
+ icoord1[ch] = icoord0[ch] + 1;
+ w[ch] = FRAC(u);
+ }
+ }
+ break;;
+ default:
+ assert(0);
+ }
+}
+
+
+/**
+ * For RECT textures / unnormalized texcoords
+ * Only a subset of wrap modes supported.
+ */
+static INLINE void
+nearest_texcoord_unnorm_4(unsigned wrapMode, const float s[4], unsigned size,
+ int icoord[4])
+{
+ uint ch;
+ switch (wrapMode) {
+ case PIPE_TEX_WRAP_CLAMP:
+ for (ch = 0; ch < 4; ch++) {
+ int i = util_ifloor(s[ch]);
+ icoord[ch]= CLAMP(i, 0, (int) size-1);
+ }
+ return;
+ case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
+ /* fall-through */
+ case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
+ for (ch = 0; ch < 4; ch++) {
+ icoord[ch]= util_ifloor( CLAMP(s[ch], 0.5F, (float) size - 0.5F) );
+ }
+ return;
+ default:
+ assert(0);
+ }
+}
+
+
+/**
+ * For RECT textures / unnormalized texcoords.
+ * Only a subset of wrap modes supported.
+ */
+static INLINE void
+linear_texcoord_unnorm_4(unsigned wrapMode, const float s[4], unsigned size,
+ int icoord0[4], int icoord1[4], float w[4])
+{
+ uint ch;
+ switch (wrapMode) {
+ case PIPE_TEX_WRAP_CLAMP:
+ for (ch = 0; ch < 4; ch++) {
+ /* Not exactly what the spec says, but it matches NVIDIA output */
+ float u = CLAMP(s[ch] - 0.5F, 0.0f, (float) size - 1.0f);
+ icoord0[ch] = util_ifloor(u);
+ icoord1[ch] = icoord0[ch] + 1;
+ w[ch] = FRAC(u);
+ }
+ return;
+ case PIPE_TEX_WRAP_CLAMP_TO_EDGE:
+ /* fall-through */
+ case PIPE_TEX_WRAP_CLAMP_TO_BORDER:
+ for (ch = 0; ch < 4; ch++) {
+ float u = CLAMP(s[ch], 0.5F, (float) size - 0.5F);
+ u -= 0.5F;
+ icoord0[ch] = util_ifloor(u);
+ icoord1[ch] = icoord0[ch] + 1;
+ if (icoord1[ch] > (int) size - 1)
+ icoord1[ch] = size - 1;
+ w[ch] = FRAC(u);
+ }
+ break;
+ default:
+ assert(0);
+ }
+}
+
+
+static unsigned
+choose_cube_face(float rx, float ry, float rz, float *newS, float *newT)
+{
+ /*
+ major axis
+ direction target sc tc ma
+ ---------- ------------------------------- --- --- ---
+ +rx TEXTURE_CUBE_MAP_POSITIVE_X_EXT -rz -ry rx
+ -rx TEXTURE_CUBE_MAP_NEGATIVE_X_EXT +rz -ry rx
+ +ry TEXTURE_CUBE_MAP_POSITIVE_Y_EXT +rx +rz ry
+ -ry TEXTURE_CUBE_MAP_NEGATIVE_Y_EXT +rx -rz ry
+ +rz TEXTURE_CUBE_MAP_POSITIVE_Z_EXT +rx -ry rz
+ -rz TEXTURE_CUBE_MAP_NEGATIVE_Z_EXT -rx -ry rz
+ */
+ const float arx = fabsf(rx), ary = fabsf(ry), arz = fabsf(rz);
+ unsigned face;
+ float sc, tc, ma;
+
+ if (arx > ary && arx > arz) {
+ if (rx >= 0.0F) {
+ face = PIPE_TEX_FACE_POS_X;
+ sc = -rz;
+ tc = -ry;
+ ma = arx;
+ }
+ else {
+ face = PIPE_TEX_FACE_NEG_X;
+ sc = rz;
+ tc = -ry;
+ ma = arx;
+ }
+ }
+ else if (ary > arx && ary > arz) {
+ if (ry >= 0.0F) {
+ face = PIPE_TEX_FACE_POS_Y;
+ sc = rx;
+ tc = rz;
+ ma = ary;
+ }
+ else {
+ face = PIPE_TEX_FACE_NEG_Y;
+ sc = rx;
+ tc = -rz;
+ ma = ary;
+ }
+ }
+ else {
+ if (rz > 0.0F) {
+ face = PIPE_TEX_FACE_POS_Z;
+ sc = rx;
+ tc = -ry;
+ ma = arz;
+ }
+ else {
+ face = PIPE_TEX_FACE_NEG_Z;
+ sc = -rx;
+ tc = -ry;
+ ma = arz;
+ }
+ }
+
+ *newS = ( sc / ma + 1.0F ) * 0.5F;
+ *newT = ( tc / ma + 1.0F ) * 0.5F;
+
+ return face;
+}
+
+
+/**
+ * Examine the quad's texture coordinates to compute the partial
+ * derivatives w.r.t X and Y, then compute lambda (level of detail).
+ *
+ * This is only done for fragment shaders, not vertex shaders.
+ */
+static float
+compute_lambda(struct tgsi_sampler *tgsi_sampler,
+ const float s[QUAD_SIZE],
+ const float t[QUAD_SIZE],
+ const float p[QUAD_SIZE],
+ float lodbias)
+{
+ const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler);
+ const struct pipe_texture *texture = samp->texture;
+ const struct pipe_sampler_state *sampler = samp->sampler;
+ float rho, lambda;
+
+ if (samp->processor == TGSI_PROCESSOR_VERTEX)
+ return lodbias;
+
+ assert(sampler->normalized_coords);
+
+ assert(s);
+ {
+ float dsdx = s[QUAD_BOTTOM_RIGHT] - s[QUAD_BOTTOM_LEFT];
+ float dsdy = s[QUAD_TOP_LEFT] - s[QUAD_BOTTOM_LEFT];
+ dsdx = fabsf(dsdx);
+ dsdy = fabsf(dsdy);
+ rho = MAX2(dsdx, dsdy) * texture->width[0];
+ }
+ if (t) {
+ float dtdx = t[QUAD_BOTTOM_RIGHT] - t[QUAD_BOTTOM_LEFT];
+ float dtdy = t[QUAD_TOP_LEFT] - t[QUAD_BOTTOM_LEFT];
+ float max;
+ dtdx = fabsf(dtdx);
+ dtdy = fabsf(dtdy);
+ max = MAX2(dtdx, dtdy) * texture->height[0];
+ rho = MAX2(rho, max);
+ }
+ if (p) {
+ float dpdx = p[QUAD_BOTTOM_RIGHT] - p[QUAD_BOTTOM_LEFT];
+ float dpdy = p[QUAD_TOP_LEFT] - p[QUAD_BOTTOM_LEFT];
+ float max;
+ dpdx = fabsf(dpdx);
+ dpdy = fabsf(dpdy);
+ max = MAX2(dpdx, dpdy) * texture->depth[0];
+ rho = MAX2(rho, max);
+ }
+
+ lambda = util_fast_log2(rho);
+ lambda += lodbias + sampler->lod_bias;
+ lambda = CLAMP(lambda, sampler->min_lod, sampler->max_lod);
+
+ return lambda;
+}
+
+
+/**
+ * Do several things here:
+ * 1. Compute lambda from the texcoords, if needed
+ * 2. Determine if we're minifying or magnifying
+ * 3. If minifying, choose mipmap levels
+ * 4. Return image filter to use within mipmap images
+ * \param level0 Returns first mipmap level to sample from
+ * \param level1 Returns second mipmap level to sample from
+ * \param levelBlend Returns blend factor between levels, in [0,1]
+ * \param imgFilter Returns either the min or mag filter, depending on lambda
+ */
+static void
+choose_mipmap_levels(struct tgsi_sampler *tgsi_sampler,
+ const float s[QUAD_SIZE],
+ const float t[QUAD_SIZE],
+ const float p[QUAD_SIZE],
+ float lodbias,
+ unsigned *level0, unsigned *level1, float *levelBlend,
+ unsigned *imgFilter)
+{
+ const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler);
+ const struct pipe_texture *texture = samp->texture;
+ const struct pipe_sampler_state *sampler = samp->sampler;
+
+ if (sampler->min_mip_filter == PIPE_TEX_MIPFILTER_NONE) {
+ /* no mipmap selection needed */
+ *level0 = *level1 = CLAMP((int) sampler->min_lod,
+ 0, (int) texture->last_level);
+
+ if (sampler->min_img_filter != sampler->mag_img_filter) {
+ /* non-mipmapped texture, but still need to determine if doing
+ * minification or magnification.
+ */
+ float lambda = compute_lambda(tgsi_sampler, s, t, p, lodbias);
+ if (lambda <= 0.0) {
+ *imgFilter = sampler->mag_img_filter;
+ }
+ else {
+ *imgFilter = sampler->min_img_filter;
+ }
+ }
+ else {
+ *imgFilter = sampler->mag_img_filter;
+ }
+ }
+ else {
+ float lambda = compute_lambda(tgsi_sampler, s, t, p, lodbias);
+
+ if (lambda <= 0.0) { /* XXX threshold depends on the filter */
+ /* magnifying */
+ *imgFilter = sampler->mag_img_filter;
+ *level0 = *level1 = 0;
+ }
+ else {
+ /* minifying */
+ *imgFilter = sampler->min_img_filter;
+
+ /* choose mipmap level(s) and compute the blend factor between them */
+ if (sampler->min_mip_filter == PIPE_TEX_MIPFILTER_NEAREST) {
+ /* Nearest mipmap level */
+ const int lvl = (int) (lambda + 0.5);
+ *level0 =
+ *level1 = CLAMP(lvl, 0, (int) texture->last_level);
+ }
+ else {
+ /* Linear interpolation between mipmap levels */
+ const int lvl = (int) lambda;
+ *level0 = CLAMP(lvl, 0, (int) texture->last_level);
+ *level1 = CLAMP(lvl + 1, 0, (int) texture->last_level);
+ *levelBlend = FRAC(lambda); /* blending weight between levels */
+ }
+ }
+ }
+}
+
+
+/**
+ * Get a texel from a texture, using the texture tile cache.
+ *
+ * \param face the cube face in 0..5
+ * \param level the mipmap level
+ * \param x the x coord of texel within 2D image
+ * \param y the y coord of texel within 2D image
+ * \param z which slice of a 3D texture
+ * \param rgba the quad to put the texel/color into
+ * \param j which element of the rgba quad to write to
+ *
+ * XXX maybe move this into lp_tile_cache.c and merge with the
+ * lp_get_cached_tile_tex() function. Also, get 4 texels instead of 1...
+ */
+static void
+get_texel_quad_2d(const struct tgsi_sampler *tgsi_sampler,
+ unsigned face, unsigned level, int x, int y,
+ const uint8_t *out[4])
+{
+ const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler);
+
+ const struct llvmpipe_cached_tex_tile *tile
+ = lp_get_cached_tex_tile(samp->cache,
+ tex_tile_address(x, y, 0, face, level));
+
+ y %= TEX_TILE_SIZE;
+ x %= TEX_TILE_SIZE;
+
+ out[0] = &tile->color[y ][x ][0];
+ out[1] = &tile->color[y ][x+1][0];
+ out[2] = &tile->color[y+1][x ][0];
+ out[3] = &tile->color[y+1][x+1][0];
+}
+
+static INLINE const uint8_t *
+get_texel_2d_ptr(const struct tgsi_sampler *tgsi_sampler,
+ unsigned face, unsigned level, int x, int y)
+{
+ const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler);
+
+ const struct llvmpipe_cached_tex_tile *tile
+ = lp_get_cached_tex_tile(samp->cache,
+ tex_tile_address(x, y, 0, face, level));
+
+ y %= TEX_TILE_SIZE;
+ x %= TEX_TILE_SIZE;
+
+ return &tile->color[y][x][0];
+}
+
+
+static void
+get_texel_quad_2d_mt(const struct tgsi_sampler *tgsi_sampler,
+ unsigned face, unsigned level,
+ int x0, int y0,
+ int x1, int y1,
+ const uint8_t *out[4])
+{
+ unsigned i;
+
+ for (i = 0; i < 4; i++) {
+ unsigned tx = (i & 1) ? x1 : x0;
+ unsigned ty = (i >> 1) ? y1 : y0;
+
+ out[i] = get_texel_2d_ptr( tgsi_sampler, face, level, tx, ty );
+ }
+}
+
+static void
+get_texel(const struct tgsi_sampler *tgsi_sampler,
+ unsigned face, unsigned level, int x, int y, int z,
+ float rgba[NUM_CHANNELS][QUAD_SIZE], unsigned j)
+{
+ const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler);
+ const struct pipe_texture *texture = samp->texture;
+ const struct pipe_sampler_state *sampler = samp->sampler;
+
+ if (x < 0 || x >= (int) texture->width[level] ||
+ y < 0 || y >= (int) texture->height[level] ||
+ z < 0 || z >= (int) texture->depth[level]) {
+ rgba[0][j] = sampler->border_color[0];
+ rgba[1][j] = sampler->border_color[1];
+ rgba[2][j] = sampler->border_color[2];
+ rgba[3][j] = sampler->border_color[3];
+ }
+ else {
+ const unsigned tx = x % TEX_TILE_SIZE;
+ const unsigned ty = y % TEX_TILE_SIZE;
+ const struct llvmpipe_cached_tex_tile *tile;
+
+ tile = lp_get_cached_tex_tile(samp->cache,
+ tex_tile_address(x, y, z, face, level));
+
+ rgba[0][j] = ubyte_to_float(tile->color[ty][tx][0]);
+ rgba[1][j] = ubyte_to_float(tile->color[ty][tx][1]);
+ rgba[2][j] = ubyte_to_float(tile->color[ty][tx][2]);
+ rgba[3][j] = ubyte_to_float(tile->color[ty][tx][3]);
+ if (0)
+ {
+ debug_printf("Get texel %f %f %f %f from %s\n",
+ rgba[0][j], rgba[1][j], rgba[2][j], rgba[3][j],
+ pf_name(texture->format));
+ }
+ }
+}
+
+
+/**
+ * Compare texcoord 'p' (aka R) against texture value 'rgba[0]'
+ * When we sampled the depth texture, the depth value was put into all
+ * RGBA channels. We look at the red channel here.
+ * \param rgba quad of (depth) texel values
+ * \param p texture 'P' components for four pixels in quad
+ * \param j which pixel in the quad to test [0..3]
+ */
+static INLINE void
+shadow_compare(const struct pipe_sampler_state *sampler,
+ float rgba[NUM_CHANNELS][QUAD_SIZE],
+ const float p[QUAD_SIZE],
+ uint j)
+{
+ int k;
+ switch (sampler->compare_func) {
+ case PIPE_FUNC_LESS:
+ k = p[j] < rgba[0][j];
+ break;
+ case PIPE_FUNC_LEQUAL:
+ k = p[j] <= rgba[0][j];
+ break;
+ case PIPE_FUNC_GREATER:
+ k = p[j] > rgba[0][j];
+ break;
+ case PIPE_FUNC_GEQUAL:
+ k = p[j] >= rgba[0][j];
+ break;
+ case PIPE_FUNC_EQUAL:
+ k = p[j] == rgba[0][j];
+ break;
+ case PIPE_FUNC_NOTEQUAL:
+ k = p[j] != rgba[0][j];
+ break;
+ case PIPE_FUNC_ALWAYS:
+ k = 1;
+ break;
+ case PIPE_FUNC_NEVER:
+ k = 0;
+ break;
+ default:
+ k = 0;
+ assert(0);
+ break;
+ }
+
+ /* XXX returning result for default GL_DEPTH_TEXTURE_MODE = GL_LUMINANCE */
+ rgba[0][j] = rgba[1][j] = rgba[2][j] = (float) k;
+ rgba[3][j] = 1.0F;
+}
+
+
+/**
+ * As above, but do four z/texture comparisons.
+ */
+static INLINE void
+shadow_compare4(const struct pipe_sampler_state *sampler,
+ float rgba[NUM_CHANNELS][QUAD_SIZE],
+ const float p[QUAD_SIZE])
+{
+ int j, k0, k1, k2, k3;
+ float val;
+
+ /* compare four texcoords vs. four texture samples */
+ switch (sampler->compare_func) {
+ case PIPE_FUNC_LESS:
+ k0 = p[0] < rgba[0][0];
+ k1 = p[1] < rgba[0][1];
+ k2 = p[2] < rgba[0][2];
+ k3 = p[3] < rgba[0][3];
+ break;
+ case PIPE_FUNC_LEQUAL:
+ k0 = p[0] <= rgba[0][0];
+ k1 = p[1] <= rgba[0][1];
+ k2 = p[2] <= rgba[0][2];
+ k3 = p[3] <= rgba[0][3];
+ break;
+ case PIPE_FUNC_GREATER:
+ k0 = p[0] > rgba[0][0];
+ k1 = p[1] > rgba[0][1];
+ k2 = p[2] > rgba[0][2];
+ k3 = p[3] > rgba[0][3];
+ break;
+ case PIPE_FUNC_GEQUAL:
+ k0 = p[0] >= rgba[0][0];
+ k1 = p[1] >= rgba[0][1];
+ k2 = p[2] >= rgba[0][2];
+ k3 = p[3] >= rgba[0][3];
+ break;
+ case PIPE_FUNC_EQUAL:
+ k0 = p[0] == rgba[0][0];
+ k1 = p[1] == rgba[0][1];
+ k2 = p[2] == rgba[0][2];
+ k3 = p[3] == rgba[0][3];
+ break;
+ case PIPE_FUNC_NOTEQUAL:
+ k0 = p[0] != rgba[0][0];
+ k1 = p[1] != rgba[0][1];
+ k2 = p[2] != rgba[0][2];
+ k3 = p[3] != rgba[0][3];
+ break;
+ case PIPE_FUNC_ALWAYS:
+ k0 = k1 = k2 = k3 = 1;
+ break;
+ case PIPE_FUNC_NEVER:
+ k0 = k1 = k2 = k3 = 0;
+ break;
+ default:
+ k0 = k1 = k2 = k3 = 0;
+ assert(0);
+ break;
+ }
+
+ /* convert four pass/fail values to an intensity in [0,1] */
+ val = 0.25F * (k0 + k1 + k2 + k3);
+
+ /* XXX returning result for default GL_DEPTH_TEXTURE_MODE = GL_LUMINANCE */
+ for (j = 0; j < 4; j++) {
+ rgba[0][j] = rgba[1][j] = rgba[2][j] = val;
+ rgba[3][j] = 1.0F;
+ }
+}
+
+
+
+static void
+lp_get_samples_2d_linear_repeat_POT(struct tgsi_sampler *tgsi_sampler,
+ const float s[QUAD_SIZE],
+ const float t[QUAD_SIZE],
+ const float p[QUAD_SIZE],
+ float lodbias,
+ float rgba[NUM_CHANNELS][QUAD_SIZE])
+{
+ const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler);
+ unsigned j;
+ unsigned level = samp->level;
+ unsigned xpot = 1 << (samp->xpot - level);
+ unsigned ypot = 1 << (samp->ypot - level);
+ unsigned xmax = (xpot - 1) & (TEX_TILE_SIZE - 1); /* MIN2(TEX_TILE_SIZE, xpot) - 1; */
+ unsigned ymax = (ypot - 1) & (TEX_TILE_SIZE - 1); /* MIN2(TEX_TILE_SIZE, ypot) - 1; */
+
+ for (j = 0; j < QUAD_SIZE; j++) {
+ int c;
+
+ float u = s[j] * xpot - 0.5F;
+ float v = t[j] * ypot - 0.5F;
+
+ int uflr = util_ifloor(u);
+ int vflr = util_ifloor(v);
+
+ float xw = u - (float)uflr;
+ float yw = v - (float)vflr;
+
+ int x0 = uflr & (xpot - 1);
+ int y0 = vflr & (ypot - 1);
+
+ const uint8_t *tx[4];
+
+
+ /* Can we fetch all four at once:
+ */
+ if (x0 < xmax && y0 < ymax)
+ {
+ get_texel_quad_2d(tgsi_sampler, 0, level, x0, y0, tx);
+ }
+ else
+ {
+ unsigned x1 = (x0 + 1) & (xpot - 1);
+ unsigned y1 = (y0 + 1) & (ypot - 1);
+ get_texel_quad_2d_mt(tgsi_sampler, 0, level,
+ x0, y0, x1, y1, tx);
+ }
+
+
+ /* interpolate R, G, B, A */
+ for (c = 0; c < 4; c++) {
+ rgba[c][j] = lerp_2d(xw, yw,
+ ubyte_to_float(tx[0][c]), ubyte_to_float(tx[1][c]),
+ ubyte_to_float(tx[2][c]), ubyte_to_float(tx[3][c]));
+ }
+ }
+}
+
+
+static void
+lp_get_samples_2d_nearest_repeat_POT(struct tgsi_sampler *tgsi_sampler,
+ const float s[QUAD_SIZE],
+ const float t[QUAD_SIZE],
+ const float p[QUAD_SIZE],
+ float lodbias,
+ float rgba[NUM_CHANNELS][QUAD_SIZE])
+{
+ const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler);
+ unsigned j;
+ unsigned level = samp->level;
+ unsigned xpot = 1 << (samp->xpot - level);
+ unsigned ypot = 1 << (samp->ypot - level);
+
+ for (j = 0; j < QUAD_SIZE; j++) {
+ int c;
+
+ float u = s[j] * xpot;
+ float v = t[j] * ypot;
+
+ int uflr = util_ifloor(u);
+ int vflr = util_ifloor(v);
+
+ int x0 = uflr & (xpot - 1);
+ int y0 = vflr & (ypot - 1);
+
+ const uint8_t *out = get_texel_2d_ptr(tgsi_sampler, 0, level, x0, y0);
+
+ for (c = 0; c < 4; c++) {
+ rgba[c][j] = ubyte_to_float(out[c]);
+ }
+ }
+}
+
+
+static void
+lp_get_samples_2d_nearest_clamp_POT(struct tgsi_sampler *tgsi_sampler,
+ const float s[QUAD_SIZE],
+ const float t[QUAD_SIZE],
+ const float p[QUAD_SIZE],
+ float lodbias,
+ float rgba[NUM_CHANNELS][QUAD_SIZE])
+{
+ const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler);
+ unsigned j;
+ unsigned level = samp->level;
+ unsigned xpot = 1 << (samp->xpot - level);
+ unsigned ypot = 1 << (samp->ypot - level);
+
+ for (j = 0; j < QUAD_SIZE; j++) {
+ int c;
+
+ float u = s[j] * xpot;
+ float v = t[j] * ypot;
+
+ int x0, y0;
+ const uint8_t *out;
+
+ x0 = util_ifloor(u);
+ if (x0 < 0)
+ x0 = 0;
+ else if (x0 > xpot - 1)
+ x0 = xpot - 1;
+
+ y0 = util_ifloor(v);
+ if (y0 < 0)
+ y0 = 0;
+ else if (y0 > ypot - 1)
+ y0 = ypot - 1;
+
+ out = get_texel_2d_ptr(tgsi_sampler, 0, level, x0, y0);
+
+ for (c = 0; c < 4; c++) {
+ rgba[c][j] = ubyte_to_float(out[c]);
+ }
+ }
+}
+
+
+static void
+lp_get_samples_2d_linear_mip_linear_repeat_POT(struct tgsi_sampler *tgsi_sampler,
+ const float s[QUAD_SIZE],
+ const float t[QUAD_SIZE],
+ const float p[QUAD_SIZE],
+ float lodbias,
+ float rgba[NUM_CHANNELS][QUAD_SIZE])
+{
+ struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler);
+ const struct pipe_texture *texture = samp->texture;
+ int level0;
+ float lambda;
+
+ lambda = compute_lambda(tgsi_sampler, s, t, p, lodbias);
+ level0 = (int)lambda;
+
+ if (lambda < 0.0) {
+ samp->level = 0;
+ lp_get_samples_2d_linear_repeat_POT( tgsi_sampler,
+ s, t, p, 0, rgba );
+ }
+ else if (level0 >= texture->last_level) {
+ samp->level = texture->last_level;
+ lp_get_samples_2d_linear_repeat_POT( tgsi_sampler,
+ s, t, p, 0, rgba );
+ }
+ else {
+ float levelBlend = lambda - level0;
+ float rgba0[4][4];
+ float rgba1[4][4];
+ int c,j;
+
+ samp->level = level0;
+ lp_get_samples_2d_linear_repeat_POT( tgsi_sampler,
+ s, t, p, 0, rgba0 );
+
+ samp->level = level0+1;
+ lp_get_samples_2d_linear_repeat_POT( tgsi_sampler,
+ s, t, p, 0, rgba1 );
+
+ for (j = 0; j < QUAD_SIZE; j++) {
+ for (c = 0; c < 4; c++) {
+ rgba[c][j] = lerp(levelBlend, rgba0[c][j], rgba1[c][j]);
+ }
+ }
+ }
+}
+
+/**
+ * Common code for sampling 1D/2D/cube textures.
+ * Could probably extend for 3D...
+ */
+static void
+lp_get_samples_2d_common(struct tgsi_sampler *tgsi_sampler,
+ const float s[QUAD_SIZE],
+ const float t[QUAD_SIZE],
+ const float p[QUAD_SIZE],
+ float lodbias,
+ float rgba[NUM_CHANNELS][QUAD_SIZE],
+ const unsigned faces[4])
+{
+ const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler);
+ const struct pipe_texture *texture = samp->texture;
+ const struct pipe_sampler_state *sampler = samp->sampler;
+ unsigned level0, level1, j, imgFilter;
+ int width, height;
+ float levelBlend = 0.0F;
+
+ choose_mipmap_levels(tgsi_sampler, s, t, p,
+ lodbias,
+ &level0, &level1, &levelBlend, &imgFilter);
+
+ assert(sampler->normalized_coords);
+
+ width = texture->width[level0];
+ height = texture->height[level0];
+
+ assert(width > 0);
+
+ switch (imgFilter) {
+ case PIPE_TEX_FILTER_NEAREST:
+ {
+ int x[4], y[4];
+ nearest_texcoord_4(sampler->wrap_s, s, width, x);
+ nearest_texcoord_4(sampler->wrap_t, t, height, y);
+
+ for (j = 0; j < QUAD_SIZE; j++) {
+ get_texel(tgsi_sampler, faces[j], level0, x[j], y[j], 0, rgba, j);
+ if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
+ shadow_compare(sampler, rgba, p, j);
+ }
+
+ if (level0 != level1) {
+ /* get texels from second mipmap level and blend */
+ float rgba2[4][4];
+ unsigned c;
+ x[j] /= 2;
+ y[j] /= 2;
+ get_texel(tgsi_sampler, faces[j], level1, x[j], y[j], 0,
+ rgba2, j);
+ if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE){
+ shadow_compare(sampler, rgba2, p, j);
+ }
+
+ for (c = 0; c < NUM_CHANNELS; c++) {
+ rgba[c][j] = lerp(levelBlend, rgba[c][j], rgba2[c][j]);
+ }
+ }
+ }
+ }
+ break;
+ case PIPE_TEX_FILTER_LINEAR:
+ case PIPE_TEX_FILTER_ANISO:
+ {
+ int x0[4], y0[4], x1[4], y1[4];
+ float xw[4], yw[4]; /* weights */
+
+ linear_texcoord_4(sampler->wrap_s, s, width, x0, x1, xw);
+ linear_texcoord_4(sampler->wrap_t, t, height, y0, y1, yw);
+
+ for (j = 0; j < QUAD_SIZE; j++) {
+ float tx[4][4]; /* texels */
+ int c;
+ get_texel(tgsi_sampler, faces[j], level0, x0[j], y0[j], 0, tx, 0);
+ get_texel(tgsi_sampler, faces[j], level0, x1[j], y0[j], 0, tx, 1);
+ get_texel(tgsi_sampler, faces[j], level0, x0[j], y1[j], 0, tx, 2);
+ get_texel(tgsi_sampler, faces[j], level0, x1[j], y1[j], 0, tx, 3);
+ if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
+ shadow_compare4(sampler, tx, p);
+ }
+
+ /* interpolate R, G, B, A */
+ for (c = 0; c < 4; c++) {
+ rgba[c][j] = lerp_2d(xw[j], yw[j],
+ tx[c][0], tx[c][1],
+ tx[c][2], tx[c][3]);
+ }
+
+ if (level0 != level1) {
+ /* get texels from second mipmap level and blend */
+ float rgba2[4][4];
+
+ /* XXX: This is incorrect -- will often end up with (x0
+ * == x1 && y0 == y1), meaning that we fetch the same
+ * texel four times and linearly interpolate between
+ * identical values. The correct approach would be to
+ * call linear_texcoord again for the second level.
+ */
+ x0[j] /= 2;
+ y0[j] /= 2;
+ x1[j] /= 2;
+ y1[j] /= 2;
+ get_texel(tgsi_sampler, faces[j], level1, x0[j], y0[j], 0, tx, 0);
+ get_texel(tgsi_sampler, faces[j], level1, x1[j], y0[j], 0, tx, 1);
+ get_texel(tgsi_sampler, faces[j], level1, x0[j], y1[j], 0, tx, 2);
+ get_texel(tgsi_sampler, faces[j], level1, x1[j], y1[j], 0, tx, 3);
+ if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE){
+ shadow_compare4(sampler, tx, p);
+ }
+
+ /* interpolate R, G, B, A */
+ for (c = 0; c < 4; c++) {
+ rgba2[c][j] = lerp_2d(xw[j], yw[j],
+ tx[c][0], tx[c][1], tx[c][2], tx[c][3]);
+ }
+
+ for (c = 0; c < NUM_CHANNELS; c++) {
+ rgba[c][j] = lerp(levelBlend, rgba[c][j], rgba2[c][j]);
+ }
+ }
+ }
+ }
+ break;
+ default:
+ assert(0);
+ }
+}
+
+
+static INLINE void
+lp_get_samples_1d(struct tgsi_sampler *sampler,
+ const float s[QUAD_SIZE],
+ const float t[QUAD_SIZE],
+ const float p[QUAD_SIZE],
+ float lodbias,
+ float rgba[NUM_CHANNELS][QUAD_SIZE])
+{
+ static const unsigned faces[4] = {0, 0, 0, 0};
+ static const float tzero[4] = {0, 0, 0, 0};
+ lp_get_samples_2d_common(sampler, s, tzero, NULL,
+ lodbias, rgba, faces);
+}
+
+
+static INLINE void
+lp_get_samples_2d(struct tgsi_sampler *sampler,
+ const float s[QUAD_SIZE],
+ const float t[QUAD_SIZE],
+ const float p[QUAD_SIZE],
+ float lodbias,
+ float rgba[NUM_CHANNELS][QUAD_SIZE])
+{
+ static const unsigned faces[4] = {0, 0, 0, 0};
+ lp_get_samples_2d_common(sampler, s, t, p,
+ lodbias, rgba, faces);
+}
+
+
+static INLINE void
+lp_get_samples_3d(struct tgsi_sampler *tgsi_sampler,
+ const float s[QUAD_SIZE],
+ const float t[QUAD_SIZE],
+ const float p[QUAD_SIZE],
+ float lodbias,
+ float rgba[NUM_CHANNELS][QUAD_SIZE])
+{
+ const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler);
+ const struct pipe_texture *texture = samp->texture;
+ const struct pipe_sampler_state *sampler = samp->sampler;
+ /* get/map pipe_surfaces corresponding to 3D tex slices */
+ unsigned level0, level1, j, imgFilter;
+ int width, height, depth;
+ float levelBlend;
+ const uint face = 0;
+
+ choose_mipmap_levels(tgsi_sampler, s, t, p,
+ lodbias,
+ &level0, &level1, &levelBlend, &imgFilter);
+
+ assert(sampler->normalized_coords);
+
+ width = texture->width[level0];
+ height = texture->height[level0];
+ depth = texture->depth[level0];
+
+ assert(width > 0);
+ assert(height > 0);
+ assert(depth > 0);
+
+ switch (imgFilter) {
+ case PIPE_TEX_FILTER_NEAREST:
+ {
+ int x[4], y[4], z[4];
+ nearest_texcoord_4(sampler->wrap_s, s, width, x);
+ nearest_texcoord_4(sampler->wrap_t, t, height, y);
+ nearest_texcoord_4(sampler->wrap_r, p, depth, z);
+ for (j = 0; j < QUAD_SIZE; j++) {
+ get_texel(tgsi_sampler, face, level0, x[j], y[j], z[j], rgba, j);
+ if (level0 != level1) {
+ /* get texels from second mipmap level and blend */
+ float rgba2[4][4];
+ unsigned c;
+ x[j] /= 2;
+ y[j] /= 2;
+ z[j] /= 2;
+ get_texel(tgsi_sampler, face, level1, x[j], y[j], z[j], rgba2, j);
+ for (c = 0; c < NUM_CHANNELS; c++) {
+ rgba[c][j] = lerp(levelBlend, rgba2[c][j], rgba[c][j]);
+ }
+ }
+ }
+ }
+ break;
+ case PIPE_TEX_FILTER_LINEAR:
+ case PIPE_TEX_FILTER_ANISO:
+ {
+ int x0[4], x1[4], y0[4], y1[4], z0[4], z1[4];
+ float xw[4], yw[4], zw[4]; /* interpolation weights */
+ linear_texcoord_4(sampler->wrap_s, s, width, x0, x1, xw);
+ linear_texcoord_4(sampler->wrap_t, t, height, y0, y1, yw);
+ linear_texcoord_4(sampler->wrap_r, p, depth, z0, z1, zw);
+
+ for (j = 0; j < QUAD_SIZE; j++) {
+ int c;
+ float tx0[4][4], tx1[4][4];
+ get_texel(tgsi_sampler, face, level0, x0[j], y0[j], z0[j], tx0, 0);
+ get_texel(tgsi_sampler, face, level0, x1[j], y0[j], z0[j], tx0, 1);
+ get_texel(tgsi_sampler, face, level0, x0[j], y1[j], z0[j], tx0, 2);
+ get_texel(tgsi_sampler, face, level0, x1[j], y1[j], z0[j], tx0, 3);
+ get_texel(tgsi_sampler, face, level0, x0[j], y0[j], z1[j], tx1, 0);
+ get_texel(tgsi_sampler, face, level0, x1[j], y0[j], z1[j], tx1, 1);
+ get_texel(tgsi_sampler, face, level0, x0[j], y1[j], z1[j], tx1, 2);
+ get_texel(tgsi_sampler, face, level0, x1[j], y1[j], z1[j], tx1, 3);
+
+ /* interpolate R, G, B, A */
+ for (c = 0; c < 4; c++) {
+ rgba[c][j] = lerp_3d(xw[j], yw[j], zw[j],
+ tx0[c][0], tx0[c][1],
+ tx0[c][2], tx0[c][3],
+ tx1[c][0], tx1[c][1],
+ tx1[c][2], tx1[c][3]);
+ }
+
+ if (level0 != level1) {
+ /* get texels from second mipmap level and blend */
+ float rgba2[4][4];
+ x0[j] /= 2;
+ y0[j] /= 2;
+ z0[j] /= 2;
+ x1[j] /= 2;
+ y1[j] /= 2;
+ z1[j] /= 2;
+ get_texel(tgsi_sampler, face, level1, x0[j], y0[j], z0[j], tx0, 0);
+ get_texel(tgsi_sampler, face, level1, x1[j], y0[j], z0[j], tx0, 1);
+ get_texel(tgsi_sampler, face, level1, x0[j], y1[j], z0[j], tx0, 2);
+ get_texel(tgsi_sampler, face, level1, x1[j], y1[j], z0[j], tx0, 3);
+ get_texel(tgsi_sampler, face, level1, x0[j], y0[j], z1[j], tx1, 0);
+ get_texel(tgsi_sampler, face, level1, x1[j], y0[j], z1[j], tx1, 1);
+ get_texel(tgsi_sampler, face, level1, x0[j], y1[j], z1[j], tx1, 2);
+ get_texel(tgsi_sampler, face, level1, x1[j], y1[j], z1[j], tx1, 3);
+
+ /* interpolate R, G, B, A */
+ for (c = 0; c < 4; c++) {
+ rgba2[c][j] = lerp_3d(xw[j], yw[j], zw[j],
+ tx0[c][0], tx0[c][1],
+ tx0[c][2], tx0[c][3],
+ tx1[c][0], tx1[c][1],
+ tx1[c][2], tx1[c][3]);
+ }
+
+ /* blend mipmap levels */
+ for (c = 0; c < NUM_CHANNELS; c++) {
+ rgba[c][j] = lerp(levelBlend, rgba[c][j], rgba2[c][j]);
+ }
+ }
+ }
+ }
+ break;
+ default:
+ assert(0);
+ }
+}
+
+
+static void
+lp_get_samples_cube(struct tgsi_sampler *sampler,
+ const float s[QUAD_SIZE],
+ const float t[QUAD_SIZE],
+ const float p[QUAD_SIZE],
+ float lodbias,
+ float rgba[NUM_CHANNELS][QUAD_SIZE])
+{
+ unsigned faces[QUAD_SIZE], j;
+ float ssss[4], tttt[4];
+ for (j = 0; j < QUAD_SIZE; j++) {
+ faces[j] = choose_cube_face(s[j], t[j], p[j], ssss + j, tttt + j);
+ }
+ lp_get_samples_2d_common(sampler, ssss, tttt, NULL,
+ lodbias, rgba, faces);
+}
+
+
+static void
+lp_get_samples_rect(struct tgsi_sampler *tgsi_sampler,
+ const float s[QUAD_SIZE],
+ const float t[QUAD_SIZE],
+ const float p[QUAD_SIZE],
+ float lodbias,
+ float rgba[NUM_CHANNELS][QUAD_SIZE])
+{
+ const struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler);
+ const struct pipe_texture *texture = samp->texture;
+ const struct pipe_sampler_state *sampler = samp->sampler;
+ const uint face = 0;
+ unsigned level0, level1, j, imgFilter;
+ int width, height;
+ float levelBlend;
+
+ choose_mipmap_levels(tgsi_sampler, s, t, p,
+ lodbias,
+ &level0, &level1, &levelBlend, &imgFilter);
+
+ /* texture RECTS cannot be mipmapped */
+ assert(level0 == level1);
+
+ width = texture->width[level0];
+ height = texture->height[level0];
+
+ assert(width > 0);
+
+ switch (imgFilter) {
+ case PIPE_TEX_FILTER_NEAREST:
+ {
+ int x[4], y[4];
+ nearest_texcoord_unnorm_4(sampler->wrap_s, s, width, x);
+ nearest_texcoord_unnorm_4(sampler->wrap_t, t, height, y);
+ for (j = 0; j < QUAD_SIZE; j++) {
+ get_texel(tgsi_sampler, face, level0, x[j], y[j], 0, rgba, j);
+ if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
+ shadow_compare(sampler, rgba, p, j);
+ }
+ }
+ }
+ break;
+ case PIPE_TEX_FILTER_LINEAR:
+ case PIPE_TEX_FILTER_ANISO:
+ {
+ int x0[4], y0[4], x1[4], y1[4];
+ float xw[4], yw[4]; /* weights */
+ linear_texcoord_unnorm_4(sampler->wrap_s, s, width, x0, x1, xw);
+ linear_texcoord_unnorm_4(sampler->wrap_t, t, height, y0, y1, yw);
+ for (j = 0; j < QUAD_SIZE; j++) {
+ float tx[4][4]; /* texels */
+ int c;
+ get_texel(tgsi_sampler, face, level0, x0[j], y0[j], 0, tx, 0);
+ get_texel(tgsi_sampler, face, level0, x1[j], y0[j], 0, tx, 1);
+ get_texel(tgsi_sampler, face, level0, x0[j], y1[j], 0, tx, 2);
+ get_texel(tgsi_sampler, face, level0, x1[j], y1[j], 0, tx, 3);
+ if (sampler->compare_mode == PIPE_TEX_COMPARE_R_TO_TEXTURE) {
+ shadow_compare4(sampler, tx, p);
+ }
+ for (c = 0; c < 4; c++) {
+ rgba[c][j] = lerp_2d(xw[j], yw[j],
+ tx[c][0], tx[c][1], tx[c][2], tx[c][3]);
+ }
+ }
+ }
+ break;
+ default:
+ assert(0);
+ }
+}
+
+
+/**
+ * Error condition handler
+ */
+static INLINE void
+lp_get_samples_null(struct tgsi_sampler *tgsi_sampler,
+ const float s[QUAD_SIZE],
+ const float t[QUAD_SIZE],
+ const float p[QUAD_SIZE],
+ float lodbias,
+ float rgba[NUM_CHANNELS][QUAD_SIZE])
+{
+ int i,j;
+
+ for (i = 0; i < 4; i++)
+ for (j = 0; j < 4; j++)
+ rgba[i][j] = 1.0;
+}
+
+/**
+ * Called via tgsi_sampler::get_samples() when using a sampler for the
+ * first time. Determine the actual sampler function, link it in and
+ * call it.
+ */
+void
+lp_get_samples(struct tgsi_sampler *tgsi_sampler,
+ const float s[QUAD_SIZE],
+ const float t[QUAD_SIZE],
+ const float p[QUAD_SIZE],
+ float lodbias,
+ float rgba[NUM_CHANNELS][QUAD_SIZE])
+{
+ struct lp_shader_sampler *samp = lp_shader_sampler(tgsi_sampler);
+ const struct pipe_texture *texture = samp->texture;
+ const struct pipe_sampler_state *sampler = samp->sampler;
+
+ /* Default to the 'undefined' case:
+ */
+ tgsi_sampler->get_samples = lp_get_samples_null;
+
+ if (!texture) {
+ assert(0); /* is this legal?? */
+ goto out;
+ }
+
+ if (!sampler->normalized_coords) {
+ assert (texture->target == PIPE_TEXTURE_2D);
+ tgsi_sampler->get_samples = lp_get_samples_rect;
+ goto out;
+ }
+
+ switch (texture->target) {
+ case PIPE_TEXTURE_1D:
+ tgsi_sampler->get_samples = lp_get_samples_1d;
+ break;
+ case PIPE_TEXTURE_2D:
+ tgsi_sampler->get_samples = lp_get_samples_2d;
+ break;
+ case PIPE_TEXTURE_3D:
+ tgsi_sampler->get_samples = lp_get_samples_3d;
+ break;
+ case PIPE_TEXTURE_CUBE:
+ tgsi_sampler->get_samples = lp_get_samples_cube;
+ break;
+ default:
+ assert(0);
+ break;
+ }
+
+ /* Do this elsewhere:
+ */
+ samp->xpot = util_unsigned_logbase2( samp->texture->width[0] );
+ samp->ypot = util_unsigned_logbase2( samp->texture->height[0] );
+
+ /* Try to hook in a faster sampler. Ultimately we'll have to
+ * code-generate these. Luckily most of this looks like it is
+ * orthogonal state within the sampler.
+ */
+ if (texture->target == PIPE_TEXTURE_2D &&
+ sampler->min_img_filter == sampler->mag_img_filter &&
+ sampler->wrap_s == sampler->wrap_t &&
+ sampler->compare_mode == FALSE &&
+ sampler->normalized_coords)
+ {
+ if (sampler->min_mip_filter == PIPE_TEX_MIPFILTER_NONE) {
+ samp->level = CLAMP((int) sampler->min_lod,
+ 0, (int) texture->last_level);
+
+ if (sampler->wrap_s == PIPE_TEX_WRAP_REPEAT) {
+ switch (sampler->min_img_filter) {
+ case PIPE_TEX_FILTER_NEAREST:
+ tgsi_sampler->get_samples = lp_get_samples_2d_nearest_repeat_POT;
+ break;
+ case PIPE_TEX_FILTER_LINEAR:
+ tgsi_sampler->get_samples = lp_get_samples_2d_linear_repeat_POT;
+ break;
+ default:
+ break;
+ }
+ }
+ else if (sampler->wrap_s == PIPE_TEX_WRAP_CLAMP) {
+ switch (sampler->min_img_filter) {
+ case PIPE_TEX_FILTER_NEAREST:
+ tgsi_sampler->get_samples = lp_get_samples_2d_nearest_clamp_POT;
+ break;
+ default:
+ break;
+ }
+ }
+ }
+ else if (sampler->min_mip_filter == PIPE_TEX_MIPFILTER_LINEAR) {
+ if (sampler->wrap_s == PIPE_TEX_WRAP_REPEAT) {
+ switch (sampler->min_img_filter) {
+ case PIPE_TEX_FILTER_LINEAR:
+ tgsi_sampler->get_samples = lp_get_samples_2d_linear_mip_linear_repeat_POT;
+ break;
+ default:
+ break;
+ }
+ }
+ }
+ }
+ else if (0) {
+ _debug_printf("target %d/%d min_mip %d/%d min_img %d/%d wrap %d/%d compare %d/%d norm %d/%d\n",
+ texture->target, PIPE_TEXTURE_2D,
+ sampler->min_mip_filter, PIPE_TEX_MIPFILTER_NONE,
+ sampler->min_img_filter, sampler->mag_img_filter,
+ sampler->wrap_s, sampler->wrap_t,
+ sampler->compare_mode, FALSE,
+ sampler->normalized_coords, TRUE);
+ }
+
+out:
+ tgsi_sampler->get_samples( tgsi_sampler, s, t, p, lodbias, rgba );
+}
+
+
+void PIPE_CDECL
+lp_fetch_texel_soa( struct tgsi_sampler **samplers,
+ uint32_t unit,
+ float *store )
+{
+ struct tgsi_sampler *sampler = samplers[unit];
+
+#if 0
+ uint j;
+
+ debug_printf("%s sampler: %p (%p) store: %p\n",
+ __FUNCTION__,
+ sampler, *sampler,
+ store );
+
+ debug_printf("lodbias %f\n", store[12]);
+
+ for (j = 0; j < 4; j++)
+ debug_printf("sample %d texcoord %f %f\n",
+ j,
+ store[0+j],
+ store[4+j]);
+#endif
+
+ {
+ float rgba[NUM_CHANNELS][QUAD_SIZE];
+ sampler->get_samples(sampler,
+ &store[0],
+ &store[4],
+ &store[8],
+ 0.0f, /*store[12], lodbias */
+ rgba);
+ memcpy(store, rgba, sizeof rgba);
+ }
+
+#if 0
+ for (j = 0; j < 4; j++)
+ debug_printf("sample %d result %f %f %f %f\n",
+ j,
+ store[0+j],
+ store[4+j],
+ store[8+j],
+ store[12+j]);
+#endif
+}
+
+
+#include "lp_bld_type.h"
+#include "lp_bld_intr.h"
+#include "lp_bld_tgsi.h"
+
+
+struct lp_c_sampler_soa
+{
+ struct lp_build_sampler_soa base;
+
+ LLVMValueRef context_ptr;
+
+ LLVMValueRef samplers_ptr;
+
+ /** Coords/texels store */
+ LLVMValueRef store_ptr;
+};
+
+
+static void
+lp_c_sampler_soa_destroy(struct lp_build_sampler_soa *sampler)
+{
+ FREE(sampler);
+}
+
+
+static void
+lp_c_sampler_soa_emit_fetch_texel(struct lp_build_sampler_soa *_sampler,
+ LLVMBuilderRef builder,
+ struct lp_type type,
+ unsigned unit,
+ unsigned num_coords,
+ const LLVMValueRef *coords,
+ LLVMValueRef lodbias,
+ LLVMValueRef *texel)
+{
+ struct lp_c_sampler_soa *sampler = (struct lp_c_sampler_soa *)_sampler;
+ LLVMTypeRef vec_type = LLVMTypeOf(coords[0]);
+ LLVMValueRef args[3];
+ unsigned i;
+
+ if(!sampler->samplers_ptr)
+ sampler->samplers_ptr = lp_jit_context_samplers(builder, sampler->context_ptr);
+
+ if(!sampler->store_ptr)
+ sampler->store_ptr = LLVMBuildArrayAlloca(builder,
+ vec_type,
+ LLVMConstInt(LLVMInt32Type(), 4, 0),
+ "texel_store");
+
+ for (i = 0; i < num_coords; i++) {
+ LLVMValueRef index = LLVMConstInt(LLVMInt32Type(), i, 0);
+ LLVMValueRef coord_ptr = LLVMBuildGEP(builder, sampler->store_ptr, &index, 1, "");
+ LLVMBuildStore(builder, coords[i], coord_ptr);
+ }
+
+ args[0] = sampler->samplers_ptr;
+ args[1] = LLVMConstInt(LLVMInt32Type(), unit, 0);
+ args[2] = sampler->store_ptr;
+
+ lp_build_intrinsic(builder, "fetch_texel", LLVMVoidType(), args, 3);
+
+ for (i = 0; i < NUM_CHANNELS; ++i) {
+ LLVMValueRef index = LLVMConstInt(LLVMInt32Type(), i, 0);
+ LLVMValueRef texel_ptr = LLVMBuildGEP(builder, sampler->store_ptr, &index, 1, "");
+ texel[i] = LLVMBuildLoad(builder, texel_ptr, "");
+ }
+}
+
+
+struct lp_build_sampler_soa *
+lp_c_sampler_soa_create(LLVMValueRef context_ptr)
+{
+ struct lp_c_sampler_soa *sampler;
+
+ sampler = CALLOC_STRUCT(lp_c_sampler_soa);
+ if(!sampler)
+ return NULL;
+
+ sampler->base.destroy = lp_c_sampler_soa_destroy;
+ sampler->base.emit_fetch_texel = lp_c_sampler_soa_emit_fetch_texel;
+ sampler->context_ptr = context_ptr;
+
+ return &sampler->base;
+}
+
diff --git a/src/gallium/drivers/llvmpipe/lp_tex_sample_llvm.c b/src/gallium/drivers/llvmpipe/lp_tex_sample_llvm.c
index d2a6ae2..5138ccf7 100644
--- a/src/gallium/drivers/llvmpipe/lp_tex_sample_llvm.c
+++ b/src/gallium/drivers/llvmpipe/lp_tex_sample_llvm.c
@@ -44,7 +44,6 @@
#include "pipe/p_shader_tokens.h"
#include "lp_bld_debug.h"
#include "lp_bld_type.h"
-#include "lp_bld_intr.h"
#include "lp_bld_sample.h"
#include "lp_bld_tgsi.h"
#include "lp_state.h"
diff --git a/src/gallium/drivers/llvmpipe/lp_texture.c b/src/gallium/drivers/llvmpipe/lp_texture.c
index 2c13502..039539d 100644
--- a/src/gallium/drivers/llvmpipe/lp_texture.c
+++ b/src/gallium/drivers/llvmpipe/lp_texture.c
@@ -42,7 +42,6 @@
#include "lp_context.h"
#include "lp_state.h"
#include "lp_texture.h"
-#include "lp_tex_cache.h"
#include "lp_screen.h"
#include "lp_winsys.h"
diff --git a/src/gallium/drivers/llvmpipe/lp_tile_cache.c b/src/gallium/drivers/llvmpipe/lp_tile_cache.c
index 7a1ecf5..971d933 100644
--- a/src/gallium/drivers/llvmpipe/lp_tile_cache.c
+++ b/src/gallium/drivers/llvmpipe/lp_tile_cache.c
@@ -38,8 +38,6 @@
#include "util/u_tile.h"
#include "util/u_rect.h"
#include "lp_context.h"
-#include "lp_surface.h"
-#include "lp_texture.h"
#include "lp_tile_soa.h"
#include "lp_tile_cache.h"
diff --git a/src/gallium/drivers/softpipe/sp_context.c b/src/gallium/drivers/softpipe/sp_context.c
index 8e01793..73e075f 100644
--- a/src/gallium/drivers/softpipe/sp_context.c
+++ b/src/gallium/drivers/softpipe/sp_context.c
@@ -43,7 +43,6 @@
#include "sp_surface.h"
#include "sp_tile_cache.h"
#include "sp_tex_tile_cache.h"
-#include "sp_texture.h"
#include "sp_winsys.h"
#include "sp_query.h"
diff --git a/src/gallium/drivers/softpipe/sp_flush.c b/src/gallium/drivers/softpipe/sp_flush.c
index 75dac81..e8952bf 100644
--- a/src/gallium/drivers/softpipe/sp_flush.c
+++ b/src/gallium/drivers/softpipe/sp_flush.c
@@ -34,11 +34,9 @@
#include "draw/draw_context.h"
#include "sp_flush.h"
#include "sp_context.h"
-#include "sp_surface.h"
#include "sp_state.h"
#include "sp_tile_cache.h"
#include "sp_tex_tile_cache.h"
-#include "sp_winsys.h"
void
diff --git a/src/gallium/drivers/softpipe/sp_prim_vbuf.c b/src/gallium/drivers/softpipe/sp_prim_vbuf.c
index 5812d1e..98c08ea 100644
--- a/src/gallium/drivers/softpipe/sp_prim_vbuf.c
+++ b/src/gallium/drivers/softpipe/sp_prim_vbuf.c
@@ -526,7 +526,8 @@
sp_vbuf_destroy(struct vbuf_render *vbr)
{
struct softpipe_vbuf_render *cvbr = softpipe_vbuf_render(vbr);
- align_free(cvbr->vertex_buffer);
+ if(cvbr->vertex_buffer)
+ align_free(cvbr->vertex_buffer);
sp_setup_destroy_context(cvbr->setup);
FREE(cvbr);
}
diff --git a/src/gallium/drivers/softpipe/sp_quad_blend.c b/src/gallium/drivers/softpipe/sp_quad_blend.c
index d9babe8..3b8c2d5 100644
--- a/src/gallium/drivers/softpipe/sp_quad_blend.c
+++ b/src/gallium/drivers/softpipe/sp_quad_blend.c
@@ -35,7 +35,6 @@
#include "util/u_memory.h"
#include "sp_context.h"
#include "sp_quad.h"
-#include "sp_surface.h"
#include "sp_tile_cache.h"
#include "sp_quad_pipe.h"
diff --git a/src/gallium/drivers/softpipe/sp_quad_depth_test.c b/src/gallium/drivers/softpipe/sp_quad_depth_test.c
index 0ca86c4..a981775 100644
--- a/src/gallium/drivers/softpipe/sp_quad_depth_test.c
+++ b/src/gallium/drivers/softpipe/sp_quad_depth_test.c
@@ -30,11 +30,11 @@
*/
#include "pipe/p_defines.h"
+#include "util/u_format.h"
#include "util/u_memory.h"
#include "tgsi/tgsi_scan.h"
#include "sp_context.h"
#include "sp_quad.h"
-#include "sp_surface.h"
#include "sp_quad_pipe.h"
#include "sp_tile_cache.h"
#include "sp_state.h" /* for sp_fragment_shader */
@@ -651,6 +651,20 @@
+/** helper to get number of Z buffer bits */
+static unsigned
+get_depth_bits(struct quad_stage *qs)
+{
+ struct pipe_surface *zsurf = qs->softpipe->framebuffer.zsbuf;
+ if (zsurf)
+ return util_format_get_component_bits(zsurf->format,
+ UTIL_FORMAT_COLORSPACE_ZS, 0);
+ else
+ return 0;
+}
+
+
+
static void
depth_test_quads_fallback(struct quad_stage *qs,
struct quad_header *quads[],
@@ -666,7 +680,7 @@
nr = alpha_test_quads(qs, quads, nr);
}
- if (qs->softpipe->framebuffer.zsbuf &&
+ if (get_depth_bits(qs) > 0 &&
(qs->softpipe->depth_stencil->depth.enabled ||
qs->softpipe->depth_stencil->stencil[0].enabled)) {
@@ -884,7 +898,7 @@
boolean alpha = qs->softpipe->depth_stencil->alpha.enabled;
- boolean depth = (qs->softpipe->framebuffer.zsbuf &&
+ boolean depth = (get_depth_bits(qs) > 0 &&
qs->softpipe->depth_stencil->depth.enabled);
unsigned depthfunc = qs->softpipe->depth_stencil->depth.func;
@@ -895,7 +909,6 @@
boolean occlusion = qs->softpipe->active_query_count;
-
if (!alpha &&
!depth &&
!stencil) {
diff --git a/src/gallium/drivers/softpipe/sp_quad_fs.c b/src/gallium/drivers/softpipe/sp_quad_fs.c
index 1e7533d..e799df1 100644
--- a/src/gallium/drivers/softpipe/sp_quad_fs.c
+++ b/src/gallium/drivers/softpipe/sp_quad_fs.c
@@ -45,8 +45,6 @@
#include "sp_state.h"
#include "sp_quad.h"
#include "sp_quad_pipe.h"
-#include "sp_texture.h"
-#include "sp_tex_sample.h"
struct quad_shade_stage
diff --git a/src/gallium/drivers/softpipe/sp_setup.c b/src/gallium/drivers/softpipe/sp_setup.c
index 3da7536..f6c3a2b 100644
--- a/src/gallium/drivers/softpipe/sp_setup.c
+++ b/src/gallium/drivers/softpipe/sp_setup.c
@@ -41,7 +41,6 @@
#include "draw/draw_private.h"
#include "draw/draw_vertex.h"
#include "pipe/p_shader_tokens.h"
-#include "pipe/p_thread.h"
#include "util/u_math.h"
#include "util/u_memory.h"
diff --git a/src/gallium/drivers/softpipe/sp_state_surface.c b/src/gallium/drivers/softpipe/sp_state_surface.c
index f615410..3946678 100644
--- a/src/gallium/drivers/softpipe/sp_state_surface.c
+++ b/src/gallium/drivers/softpipe/sp_state_surface.c
@@ -30,7 +30,6 @@
#include "sp_context.h"
#include "sp_state.h"
-#include "sp_surface.h"
#include "sp_tile_cache.h"
#include "draw/draw_context.h"
diff --git a/src/gallium/drivers/softpipe/sp_state_vertex.c b/src/gallium/drivers/softpipe/sp_state_vertex.c
index 46b6991..b491d92 100644
--- a/src/gallium/drivers/softpipe/sp_state_vertex.c
+++ b/src/gallium/drivers/softpipe/sp_state_vertex.c
@@ -31,7 +31,6 @@
#include "sp_context.h"
#include "sp_state.h"
-#include "sp_surface.h"
#include "draw/draw_context.h"
diff --git a/src/gallium/drivers/softpipe/sp_tex_tile_cache.c b/src/gallium/drivers/softpipe/sp_tex_tile_cache.c
index e50a76a..50242d5 100644
--- a/src/gallium/drivers/softpipe/sp_tex_tile_cache.c
+++ b/src/gallium/drivers/softpipe/sp_tex_tile_cache.c
@@ -37,7 +37,6 @@
#include "util/u_tile.h"
#include "util/u_math.h"
#include "sp_context.h"
-#include "sp_surface.h"
#include "sp_texture.h"
#include "sp_tex_tile_cache.h"
diff --git a/src/gallium/drivers/softpipe/sp_texture.c b/src/gallium/drivers/softpipe/sp_texture.c
index 45e80c5..a5fff91 100644
--- a/src/gallium/drivers/softpipe/sp_texture.c
+++ b/src/gallium/drivers/softpipe/sp_texture.c
@@ -38,7 +38,6 @@
#include "util/u_memory.h"
#include "sp_context.h"
-#include "sp_state.h"
#include "sp_texture.h"
#include "sp_screen.h"
#include "sp_winsys.h"
diff --git a/src/gallium/drivers/svga/svga_draw_arrays.c b/src/gallium/drivers/svga/svga_draw_arrays.c
index 75492df..6b6ebc9 100644
--- a/src/gallium/drivers/svga/svga_draw_arrays.c
+++ b/src/gallium/drivers/svga/svga_draw_arrays.c
@@ -26,7 +26,6 @@
#include "svga_cmd.h"
#include "pipe/p_inlines.h"
-#include "util/u_prim.h"
#include "indices/u_indices.h"
#include "svga_hw_reg.h"
diff --git a/src/gallium/drivers/svga/svga_draw_elements.c b/src/gallium/drivers/svga/svga_draw_elements.c
index 167d817..022b444 100644
--- a/src/gallium/drivers/svga/svga_draw_elements.c
+++ b/src/gallium/drivers/svga/svga_draw_elements.c
@@ -24,7 +24,6 @@
**********************************************************/
#include "pipe/p_inlines.h"
-#include "util/u_prim.h"
#include "util/u_upload_mgr.h"
#include "indices/u_indices.h"
diff --git a/src/gallium/drivers/svga/svga_pipe_blend.c b/src/gallium/drivers/svga/svga_pipe_blend.c
index 855d228..3ad3f97 100644
--- a/src/gallium/drivers/svga/svga_pipe_blend.c
+++ b/src/gallium/drivers/svga/svga_pipe_blend.c
@@ -29,7 +29,6 @@
#include "util/u_memory.h"
#include "svga_context.h"
-#include "svga_state.h"
#include "svga_hw_reg.h"
diff --git a/src/gallium/drivers/svga/svga_pipe_constants.c b/src/gallium/drivers/svga/svga_pipe_constants.c
index ca2c7c4..93022f3 100644
--- a/src/gallium/drivers/svga/svga_pipe_constants.c
+++ b/src/gallium/drivers/svga/svga_pipe_constants.c
@@ -30,9 +30,6 @@
#include "tgsi/tgsi_parse.h"
#include "svga_context.h"
-#include "svga_state.h"
-#include "svga_hw_reg.h"
-#include "svga_cmd.h"
/***********************************************************************
* Constant buffers
diff --git a/src/gallium/drivers/svga/svga_pipe_depthstencil.c b/src/gallium/drivers/svga/svga_pipe_depthstencil.c
index df636c0..34e60cb 100644
--- a/src/gallium/drivers/svga/svga_pipe_depthstencil.c
+++ b/src/gallium/drivers/svga/svga_pipe_depthstencil.c
@@ -29,7 +29,6 @@
#include "util/u_memory.h"
#include "svga_context.h"
-#include "svga_state.h"
#include "svga_hw_reg.h"
diff --git a/src/gallium/drivers/svga/svga_pipe_draw.c b/src/gallium/drivers/svga/svga_pipe_draw.c
index 0f24ef4..4e0c499 100644
--- a/src/gallium/drivers/svga/svga_pipe_draw.c
+++ b/src/gallium/drivers/svga/svga_pipe_draw.c
@@ -33,7 +33,6 @@
#include "svga_hw_reg.h"
#include "svga_context.h"
#include "svga_screen.h"
-#include "svga_winsys.h"
#include "svga_draw.h"
#include "svga_state.h"
#include "svga_swtnl.h"
diff --git a/src/gallium/drivers/svga/svga_pipe_flush.c b/src/gallium/drivers/svga/svga_pipe_flush.c
index 0becb07..3eb1033 100644
--- a/src/gallium/drivers/svga/svga_pipe_flush.c
+++ b/src/gallium/drivers/svga/svga_pipe_flush.c
@@ -28,13 +28,8 @@
#include "svga_screen_texture.h"
#include "svga_context.h"
#include "svga_winsys.h"
-#include "svga_draw.h"
#include "svga_debug.h"
-#include "svga_hw_reg.h"
-
-
-
static void svga_flush( struct pipe_context *pipe,
unsigned flags,
diff --git a/src/gallium/drivers/svga/svga_pipe_fs.c b/src/gallium/drivers/svga/svga_pipe_fs.c
index 5f1213e..32f07fb 100644
--- a/src/gallium/drivers/svga/svga_pipe_fs.c
+++ b/src/gallium/drivers/svga/svga_pipe_fs.c
@@ -32,11 +32,9 @@
#include "svga_screen.h"
#include "svga_context.h"
-#include "svga_state.h"
#include "svga_tgsi.h"
#include "svga_hw_reg.h"
#include "svga_cmd.h"
-#include "svga_draw.h"
#include "svga_debug.h"
diff --git a/src/gallium/drivers/svga/svga_pipe_misc.c b/src/gallium/drivers/svga/svga_pipe_misc.c
index 58cb1e6..8cf1f2e 100644
--- a/src/gallium/drivers/svga/svga_pipe_misc.c
+++ b/src/gallium/drivers/svga/svga_pipe_misc.c
@@ -27,12 +27,6 @@
#include "svga_context.h"
#include "svga_screen_texture.h"
-#include "svga_state.h"
-#include "svga_winsys.h"
-
-#include "svga_hw_reg.h"
-
-
static void svga_set_scissor_state( struct pipe_context *pipe,
diff --git a/src/gallium/drivers/svga/svga_pipe_query.c b/src/gallium/drivers/svga/svga_pipe_query.c
index 01336b0..08283e3 100644
--- a/src/gallium/drivers/svga/svga_pipe_query.c
+++ b/src/gallium/drivers/svga/svga_pipe_query.c
@@ -32,7 +32,6 @@
#include "svga_screen.h"
#include "svga_screen_buffer.h"
#include "svga_winsys.h"
-#include "svga_draw.h"
#include "svga_debug.h"
diff --git a/src/gallium/drivers/svga/svga_pipe_rasterizer.c b/src/gallium/drivers/svga/svga_pipe_rasterizer.c
index b03f8eb..9ea11aa 100644
--- a/src/gallium/drivers/svga/svga_pipe_rasterizer.c
+++ b/src/gallium/drivers/svga/svga_pipe_rasterizer.c
@@ -30,7 +30,6 @@
#include "util/u_memory.h"
#include "svga_context.h"
-#include "svga_state.h"
#include "svga_hw_reg.h"
diff --git a/src/gallium/drivers/svga/svga_pipe_sampler.c b/src/gallium/drivers/svga/svga_pipe_sampler.c
index 460a101..161c66d 100644
--- a/src/gallium/drivers/svga/svga_pipe_sampler.c
+++ b/src/gallium/drivers/svga/svga_pipe_sampler.c
@@ -32,9 +32,6 @@
#include "svga_context.h"
#include "svga_screen_texture.h"
-#include "svga_state.h"
-
-#include "svga_hw_reg.h"
#include "svga_debug.h"
diff --git a/src/gallium/drivers/svga/svga_pipe_vertex.c b/src/gallium/drivers/svga/svga_pipe_vertex.c
index 42f290d..0bf43fa 100644
--- a/src/gallium/drivers/svga/svga_pipe_vertex.c
+++ b/src/gallium/drivers/svga/svga_pipe_vertex.c
@@ -32,10 +32,6 @@
#include "svga_screen.h"
#include "svga_screen_buffer.h"
#include "svga_context.h"
-#include "svga_state.h"
-#include "svga_winsys.h"
-
-#include "svga_hw_reg.h"
static void svga_set_vertex_buffers(struct pipe_context *pipe,
diff --git a/src/gallium/drivers/svga/svga_pipe_vs.c b/src/gallium/drivers/svga/svga_pipe_vs.c
index 7e6ab57..c4ac530 100644
--- a/src/gallium/drivers/svga/svga_pipe_vs.c
+++ b/src/gallium/drivers/svga/svga_pipe_vs.c
@@ -33,7 +33,6 @@
#include "svga_screen.h"
#include "svga_context.h"
-#include "svga_state.h"
#include "svga_tgsi.h"
#include "svga_hw_reg.h"
#include "svga_cmd.h"
diff --git a/src/gallium/drivers/svga/svga_screen.c b/src/gallium/drivers/svga/svga_screen.c
index fc1b3c9..cd1ed7b 100644
--- a/src/gallium/drivers/svga/svga_screen.c
+++ b/src/gallium/drivers/svga/svga_screen.c
@@ -33,10 +33,8 @@
#include "svga_screen.h"
#include "svga_screen_texture.h"
#include "svga_screen_buffer.h"
-#include "svga_cmd.h"
#include "svga_debug.h"
-#include "svga_hw_reg.h"
#include "svga3d_shaderdefs.h"
@@ -393,8 +391,6 @@
pipe_mutex_init(svgascreen->tex_mutex);
pipe_mutex_init(svgascreen->swc_mutex);
- LIST_INITHEAD(&svgascreen->cached_buffers);
-
svga_screen_cache_init(svgascreen);
return screen;
diff --git a/src/gallium/drivers/svga/svga_screen.h b/src/gallium/drivers/svga/svga_screen.h
index b94ca7f..a009b60 100644
--- a/src/gallium/drivers/svga/svga_screen.h
+++ b/src/gallium/drivers/svga/svga_screen.h
@@ -68,12 +68,6 @@
pipe_mutex tex_mutex;
pipe_mutex swc_mutex; /* Protects the use of swc and dirty_buffers */
- /**
- * List of buffers with cached GMR. Ordered from the most recently used to
- * the least recently used
- */
- struct list_head cached_buffers;
-
struct svga_host_surface_cache cache;
};
diff --git a/src/gallium/drivers/svga/svga_screen_buffer.c b/src/gallium/drivers/svga/svga_screen_buffer.c
index 58a1aba..430a697 100644
--- a/src/gallium/drivers/svga/svga_screen_buffer.c
+++ b/src/gallium/drivers/svga/svga_screen_buffer.c
@@ -113,68 +113,9 @@
if(sbuf->hw.buf) {
sws->buffer_destroy(sws, sbuf->hw.buf);
sbuf->hw.buf = NULL;
- assert(sbuf->head.prev && sbuf->head.next);
- LIST_DEL(&sbuf->head);
-#ifdef DEBUG
- sbuf->head.next = sbuf->head.prev = NULL;
-#endif
}
}
-static INLINE enum pipe_error
-svga_buffer_backup(struct svga_screen *ss, struct svga_buffer *sbuf)
-{
- if (sbuf->hw.buf && sbuf->hw.num_ranges) {
- void *src;
-
- if (!sbuf->swbuf)
- sbuf->swbuf = align_malloc(sbuf->base.size, sbuf->base.alignment);
- if (!sbuf->swbuf)
- return PIPE_ERROR_OUT_OF_MEMORY;
-
- src = ss->sws->buffer_map(ss->sws, sbuf->hw.buf,
- PIPE_BUFFER_USAGE_CPU_READ);
- if (!src)
- return PIPE_ERROR;
-
- memcpy(sbuf->swbuf, src, sbuf->base.size);
- ss->sws->buffer_unmap(ss->sws, sbuf->hw.buf);
- }
-
- return PIPE_OK;
-}
-
-/**
- * Try to make GMR space available by freeing the hardware storage of
- * unmapped
- */
-boolean
-svga_buffer_free_cached_hw_storage(struct svga_screen *ss)
-{
- struct list_head *curr;
- struct svga_buffer *sbuf;
- enum pipe_error ret = PIPE_OK;
-
- curr = ss->cached_buffers.prev;
-
- /* free the least recently used buffer's hw storage which is not mapped */
- do {
- if(curr == &ss->cached_buffers)
- return FALSE;
-
- sbuf = LIST_ENTRY(struct svga_buffer, curr, head);
-
- curr = curr->prev;
- if (sbuf->map.count == 0)
- ret = svga_buffer_backup(ss, sbuf);
-
- } while(sbuf->map.count != 0 || ret != PIPE_OK);
-
- svga_buffer_destroy_hw_storage(ss, sbuf);
-
- return TRUE;
-}
-
struct svga_winsys_buffer *
svga_winsys_buffer_create( struct svga_screen *ss,
unsigned alignment,
@@ -195,12 +136,6 @@
svga_screen_flush(ss, NULL);
buf = sws->buffer_create(sws, alignment, usage, size);
- SVGA_DBG(DEBUG_DMA|DEBUG_PERF, "evicting buffers to find %d bytes GMR\n",
- size);
-
- /* Try evicing all buffer storage */
- while(!buf && svga_buffer_free_cached_hw_storage(ss))
- buf = sws->buffer_create(sws, alignment, usage, size);
}
return buf;
@@ -226,8 +161,6 @@
return PIPE_ERROR_OUT_OF_MEMORY;
assert(!sbuf->needs_flush);
- assert(!sbuf->head.prev && !sbuf->head.next);
- LIST_ADD(&sbuf->head, &ss->cached_buffers);
}
return PIPE_OK;
@@ -311,7 +244,6 @@
svga_buffer_upload_flush(struct svga_context *svga,
struct svga_buffer *sbuf)
{
- struct svga_screen *ss = svga_screen(svga->pipe.screen);
SVGA3dCopyBox *boxes;
unsigned i;
@@ -348,13 +280,16 @@
assert(sbuf->head.prev && sbuf->head.next);
LIST_DEL(&sbuf->head);
+#ifdef DEBUG
+ sbuf->head.next = sbuf->head.prev = NULL;
+#endif
sbuf->needs_flush = FALSE;
- /* XXX: do we care about cached_buffers any more ?*/
- LIST_ADD(&sbuf->head, &ss->cached_buffers);
sbuf->hw.svga = NULL;
sbuf->hw.boxes = NULL;
+ sbuf->host_written = TRUE;
+
/* Decrement reference count */
pipe_reference(&(sbuf->base.reference), NULL);
sbuf = NULL;
@@ -437,17 +372,17 @@
}
else {
if(!sbuf->hw.buf) {
- struct svga_winsys_surface *handle = sbuf->handle;
-
if(svga_buffer_create_hw_storage(ss, sbuf) != PIPE_OK)
return NULL;
/* Populate the hardware storage if the host surface pre-existed */
- if((usage & PIPE_BUFFER_USAGE_CPU_READ) && handle) {
+ if(sbuf->host_written) {
SVGA3dSurfaceDMAFlags flags;
enum pipe_error ret;
struct pipe_fence_handle *fence = NULL;
+ assert(sbuf->handle);
+
SVGA_DBG(DEBUG_DMA|DEBUG_PERF, "dma from sid %p (buffer), bytes %u - %u\n",
sbuf->handle, 0, sbuf->base.size);
@@ -478,17 +413,6 @@
sws->fence_reference(sws, &fence, NULL);
}
}
- else {
- if((usage & PIPE_BUFFER_USAGE_CPU_READ) && !sbuf->needs_flush) {
- /* We already had the hardware storage but we would have to issue
- * a download if we hadn't, so move the buffer to the begginning
- * of the LRU list.
- */
- assert(sbuf->head.prev && sbuf->head.next);
- LIST_DEL(&sbuf->head);
- LIST_ADD(&sbuf->head, &ss->cached_buffers);
- }
- }
map = sws->buffer_map(sws, sbuf->hw.buf, usage);
}
@@ -572,10 +496,8 @@
assert(!sbuf->needs_flush);
- if(sbuf->handle) {
- SVGA_DBG(DEBUG_DMA, "release sid %p sz %d\n", sbuf->handle, sbuf->base.size);
- svga_screen_surface_destroy(ss, &sbuf->key, &sbuf->handle);
- }
+ if(sbuf->handle)
+ svga_buffer_destroy_host_surface(ss, sbuf);
if(sbuf->hw.buf)
svga_buffer_destroy_hw_storage(ss, sbuf);
@@ -595,6 +517,9 @@
struct svga_screen *ss = svga_screen(screen);
struct svga_buffer *sbuf;
+ assert(size);
+ assert(alignment);
+
sbuf = CALLOC_STRUCT(svga_buffer);
if(!sbuf)
goto error1;
@@ -755,8 +680,7 @@
assert(sbuf->hw.svga == svga);
sbuf->needs_flush = TRUE;
- assert(sbuf->head.prev && sbuf->head.next);
- LIST_DEL(&sbuf->head);
+ assert(!sbuf->head.prev && !sbuf->head.next);
LIST_ADDTAIL(&sbuf->head, &svga->dirty_buffers);
}
diff --git a/src/gallium/drivers/svga/svga_screen_buffer.h b/src/gallium/drivers/svga/svga_screen_buffer.h
index 5d7af5a..448ac10 100644
--- a/src/gallium/drivers/svga/svga_screen_buffer.h
+++ b/src/gallium/drivers/svga/svga_screen_buffer.h
@@ -135,6 +135,11 @@
*/
struct svga_winsys_surface *handle;
+ /**
+ * Whether the host has been ever written.
+ */
+ boolean host_written;
+
struct {
unsigned count;
boolean writing;
@@ -178,9 +183,6 @@
void
svga_context_flush_buffers(struct svga_context *svga);
-boolean
-svga_buffer_free_cached_hw_storage(struct svga_screen *ss);
-
struct svga_winsys_buffer *
svga_winsys_buffer_create(struct svga_screen *ss,
unsigned alignment,
diff --git a/src/gallium/drivers/svga/svga_screen_texture.c b/src/gallium/drivers/svga/svga_screen_texture.c
index 2224c2d..0d69007 100644
--- a/src/gallium/drivers/svga/svga_screen_texture.c
+++ b/src/gallium/drivers/svga/svga_screen_texture.c
@@ -306,11 +306,19 @@
tex->key.numFaces = 1;
}
+ tex->key.cachable = 1;
+
if(templat->tex_usage & PIPE_TEXTURE_USAGE_SAMPLER)
tex->key.flags |= SVGA3D_SURFACE_HINT_TEXTURE;
- if(templat->tex_usage & PIPE_TEXTURE_USAGE_PRIMARY)
+ if(templat->tex_usage & PIPE_TEXTURE_USAGE_DISPLAY_TARGET) {
+ tex->key.cachable = 0;
+ }
+
+ if(templat->tex_usage & PIPE_TEXTURE_USAGE_PRIMARY) {
tex->key.flags |= SVGA3D_SURFACE_HINT_SCANOUT;
+ tex->key.cachable = 0;
+ }
/*
* XXX: Never pass the SVGA3D_SURFACE_HINT_RENDERTARGET hint. Mesa cannot
@@ -333,8 +341,6 @@
if(tex->key.format == SVGA3D_FORMAT_INVALID)
goto error2;
- tex->key.cachable = 1;
-
SVGA_DBG(DEBUG_DMA, "surface_create for texture\n", tex->handle);
tex->handle = svga_screen_surface_create(svgascreen, &tex->key);
if (tex->handle)
@@ -416,6 +422,62 @@
}
+struct pipe_texture *
+svga_screen_texture_wrap_surface(struct pipe_screen *screen,
+ struct pipe_texture *base,
+ enum SVGA3dSurfaceFormat format,
+ struct svga_winsys_surface *srf)
+{
+ struct svga_texture *tex;
+ assert(screen);
+
+ /* Only supports one type */
+ if (base->target != PIPE_TEXTURE_2D ||
+ base->last_level != 0 ||
+ base->depth0 != 1) {
+ return NULL;
+ }
+
+ if (!srf)
+ return NULL;
+
+ if (svga_translate_format(base->format) != format) {
+ unsigned f1 = svga_translate_format(base->format);
+ unsigned f2 = format;
+
+ /* It's okay for XRGB and ARGB or depth with/out stencil to get mixed up */
+ if ( !( (f1 == SVGA3D_X8R8G8B8 && f2 == SVGA3D_A8R8G8B8) ||
+ (f1 == SVGA3D_A8R8G8B8 && f2 == SVGA3D_X8R8G8B8) ||
+ (f1 == SVGA3D_Z_D24X8 && f2 == SVGA3D_Z_D24S8) ) ) {
+ debug_printf("%s wrong format %u != %u\n", __FUNCTION__, f1, f2);
+ return NULL;
+ }
+ }
+
+ tex = CALLOC_STRUCT(svga_texture);
+ if (!tex)
+ return NULL;
+
+ tex->base = *base;
+
+
+ if (format == 1)
+ tex->base.format = PIPE_FORMAT_X8R8G8B8_UNORM;
+ else if (format == 2)
+ tex->base.format = PIPE_FORMAT_A8R8G8B8_UNORM;
+
+ pipe_reference_init(&tex->base.reference, 1);
+ tex->base.screen = screen;
+
+ SVGA_DBG(DEBUG_DMA, "wrap surface sid %p\n", srf);
+
+ tex->key.cachable = 0;
+ tex->handle = srf;
+
+ return &tex->base;
+}
+
+
static void
svga_texture_destroy(struct pipe_texture *pt)
{
diff --git a/src/gallium/drivers/svga/svga_state_framebuffer.c b/src/gallium/drivers/svga/svga_state_framebuffer.c
index cfdcae4..eda1aef 100644
--- a/src/gallium/drivers/svga/svga_state_framebuffer.c
+++ b/src/gallium/drivers/svga/svga_state_framebuffer.c
@@ -32,8 +32,6 @@
#include "svga_cmd.h"
#include "svga_debug.h"
-#include "svga_hw_reg.h"
-
/***********************************************************************
* Hardware state update
diff --git a/src/gallium/drivers/svga/svga_state_rss.c b/src/gallium/drivers/svga/svga_state_rss.c
index 8b6803a..2f9adae 100644
--- a/src/gallium/drivers/svga/svga_state_rss.c
+++ b/src/gallium/drivers/svga/svga_state_rss.c
@@ -31,9 +31,6 @@
#include "svga_state.h"
#include "svga_cmd.h"
-#include "svga_hw_reg.h"
-
-
struct rs_queue {
unsigned rs_count;
diff --git a/src/gallium/drivers/svga/svga_state_tss.c b/src/gallium/drivers/svga/svga_state_tss.c
index b313794..b3c9687 100644
--- a/src/gallium/drivers/svga/svga_state_tss.c
+++ b/src/gallium/drivers/svga/svga_state_tss.c
@@ -33,8 +33,6 @@
#include "svga_state.h"
#include "svga_cmd.h"
-#include "svga_hw_reg.h"
-
void svga_cleanup_tss_binding(struct svga_context *svga)
{
diff --git a/src/gallium/drivers/svga/svga_swtnl_backend.c b/src/gallium/drivers/svga/svga_swtnl_backend.c
index b4f757a..aafb3e2 100644
--- a/src/gallium/drivers/svga/svga_swtnl_backend.c
+++ b/src/gallium/drivers/svga/svga_swtnl_backend.c
@@ -31,7 +31,6 @@
#include "pipe/p_inlines.h"
#include "util/u_math.h"
#include "util/u_memory.h"
-#include "util/u_simple_shaders.h"
#include "svga_context.h"
#include "svga_state.h"
@@ -87,13 +86,13 @@
if (!svga_render->vbuf) {
svga_render->vbuf_size = MAX2(size, svga_render->vbuf_alloc_size);
svga_render->vbuf = pipe_buffer_create(screen,
- 0,
+ 16,
PIPE_BUFFER_USAGE_VERTEX,
svga_render->vbuf_size);
if(!svga_render->vbuf) {
svga_context_flush(svga, NULL);
svga_render->vbuf = pipe_buffer_create(screen,
- 0,
+ 16,
PIPE_BUFFER_USAGE_VERTEX,
svga_render->vbuf_size);
assert(svga_render->vbuf);
@@ -123,7 +122,9 @@
char *ptr = (char*)pipe_buffer_map(screen,
svga_render->vbuf,
PIPE_BUFFER_USAGE_CPU_WRITE |
- PIPE_BUFFER_USAGE_FLUSH_EXPLICIT);
+ PIPE_BUFFER_USAGE_FLUSH_EXPLICIT |
+ PIPE_BUFFER_USAGE_DISCARD |
+ PIPE_BUFFER_USAGE_UNSYNCHRONIZED);
return ptr + svga_render->vbuf_offset;
}
@@ -259,14 +260,14 @@
if (!svga_render->ibuf) {
svga_render->ibuf_size = MAX2(size, svga_render->ibuf_alloc_size);
svga_render->ibuf = pipe_buffer_create(screen,
- 0,
+ 2,
PIPE_BUFFER_USAGE_VERTEX,
svga_render->ibuf_size);
svga_render->ibuf_offset = 0;
}
- pipe_buffer_write(screen, svga_render->ibuf,
- svga_render->ibuf_offset, 2 * nr_indices, indices);
+ pipe_buffer_write_nooverlap(screen, svga_render->ibuf,
+ svga_render->ibuf_offset, 2 * nr_indices, indices);
/* off to hardware */
diff --git a/src/gallium/drivers/svga/svga_swtnl_draw.c b/src/gallium/drivers/svga/svga_swtnl_draw.c
index 7655121..0ae58c7 100644
--- a/src/gallium/drivers/svga/svga_swtnl_draw.c
+++ b/src/gallium/drivers/svga/svga_swtnl_draw.c
@@ -27,7 +27,6 @@
#include "draw/draw_vbuf.h"
#include "pipe/p_inlines.h"
#include "pipe/p_state.h"
-#include "util/u_memory.h"
#include "svga_context.h"
#include "svga_swtnl.h"
diff --git a/src/gallium/drivers/svga/svga_swtnl_state.c b/src/gallium/drivers/svga/svga_swtnl_state.c
index 94b6ccc..fe03e20 100644
--- a/src/gallium/drivers/svga/svga_swtnl_state.c
+++ b/src/gallium/drivers/svga/svga_swtnl_state.c
@@ -27,7 +27,6 @@
#include "draw/draw_vbuf.h"
#include "pipe/p_inlines.h"
#include "pipe/p_state.h"
-#include "util/u_memory.h"
#include "svga_context.h"
#include "svga_swtnl.h"
diff --git a/src/gallium/drivers/svga/svga_tgsi_decl_sm20.c b/src/gallium/drivers/svga/svga_tgsi_decl_sm20.c
index 23b3ace..1ae9906 100644
--- a/src/gallium/drivers/svga/svga_tgsi_decl_sm20.c
+++ b/src/gallium/drivers/svga/svga_tgsi_decl_sm20.c
@@ -29,9 +29,6 @@
#include "util/u_memory.h"
#include "svga_tgsi_emit.h"
-#include "svga_context.h"
-
-
static boolean ps20_input( struct svga_shader_emitter *emit,
diff --git a/src/gallium/drivers/svga/svga_tgsi_decl_sm30.c b/src/gallium/drivers/svga/svga_tgsi_decl_sm30.c
index d1c7336..43fc0d3 100644
--- a/src/gallium/drivers/svga/svga_tgsi_decl_sm30.c
+++ b/src/gallium/drivers/svga/svga_tgsi_decl_sm30.c
@@ -29,7 +29,6 @@
#include "util/u_memory.h"
#include "svga_tgsi_emit.h"
-#include "svga_context.h"
static boolean translate_vs_ps_semantic( struct tgsi_declaration_semantic semantic,
unsigned *usage,
diff --git a/src/gallium/drivers/svga/svga_winsys.h b/src/gallium/drivers/svga/svga_winsys.h
index 59f299c..27b99fe 100644
--- a/src/gallium/drivers/svga/svga_winsys.h
+++ b/src/gallium/drivers/svga/svga_winsys.h
@@ -296,4 +296,10 @@
struct pipe_buffer **buffer,
unsigned *stride);
+struct pipe_texture *
+svga_screen_texture_wrap_surface(struct pipe_screen *screen,
+ struct pipe_texture *base,
+ enum SVGA3dSurfaceFormat format,
+ struct svga_winsys_surface *srf);
+
#endif /* SVGA_WINSYS_H_ */
diff --git a/src/gallium/drivers/trace/tr_drm.c b/src/gallium/drivers/trace/tr_drm.c
index 48d1c40..e7ca3a8 100644
--- a/src/gallium/drivers/trace/tr_drm.c
+++ b/src/gallium/drivers/trace/tr_drm.c
@@ -173,6 +173,7 @@
if (!tr_api)
goto error;
+ tr_api->base.driver_name = api->driver_name;
tr_api->base.create_screen = trace_drm_create_screen;
tr_api->base.create_context = trace_drm_create_context;
tr_api->base.texture_from_shared_handle = trace_drm_texture_from_shared_handle;
diff --git a/src/gallium/include/pipe/p_inlines.h b/src/gallium/include/pipe/p_inlines.h
index 5fbd62a..72f5c1d 100644
--- a/src/gallium/include/pipe/p_inlines.h
+++ b/src/gallium/include/pipe/p_inlines.h
@@ -63,13 +63,6 @@
if(screen->buffer_map_range) {
unsigned offset = 0;
unsigned length = buf->size;
-
- /* XXX: Actually we should be using/detecting DISCARD
- * instead of assuming that WRITE implies discard */
- if((usage & PIPE_BUFFER_USAGE_CPU_WRITE) &&
- !(usage & PIPE_BUFFER_USAGE_DISCARD))
- usage |= PIPE_BUFFER_USAGE_CPU_READ;
-
return screen->buffer_map_range(screen, buf, offset, length, usage);
}
else
@@ -126,7 +119,39 @@
map = pipe_buffer_map_range(screen, buf, offset, size,
PIPE_BUFFER_USAGE_CPU_WRITE |
- PIPE_BUFFER_USAGE_FLUSH_EXPLICIT);
+ PIPE_BUFFER_USAGE_FLUSH_EXPLICIT |
+ PIPE_BUFFER_USAGE_DISCARD);
+ assert(map);
+ if(map) {
+ memcpy((uint8_t *)map + offset, data, size);
+ pipe_buffer_flush_mapped_range(screen, buf, offset, size);
+ pipe_buffer_unmap(screen, buf);
+ }
+}
+
+/**
+ * Special case for writing non-overlapping ranges.
+ *
+ * We can avoid GPU/CPU synchronization when writing range that has never
+ * been written before.
+ */
+static INLINE void
+pipe_buffer_write_nooverlap(struct pipe_screen *screen,
+ struct pipe_buffer *buf,
+ unsigned offset, unsigned size,
+ const void *data)
+{
+ void *map;
+
+ assert(offset < buf->size);
+ assert(offset + size <= buf->size);
+ assert(size);
+
+ map = pipe_buffer_map_range(screen, buf, offset, size,
+ PIPE_BUFFER_USAGE_CPU_WRITE |
+ PIPE_BUFFER_USAGE_FLUSH_EXPLICIT |
+ PIPE_BUFFER_USAGE_DISCARD |
+ PIPE_BUFFER_USAGE_UNSYNCHRONIZED);
assert(map);
if(map) {
memcpy((uint8_t *)map + offset, data, size);
diff --git a/src/gallium/include/state_tracker/drm_api.h b/src/gallium/include/state_tracker/drm_api.h
index bb92892..b248a81 100644
--- a/src/gallium/include/state_tracker/drm_api.h
+++ b/src/gallium/include/state_tracker/drm_api.h
@@ -31,6 +31,11 @@
const char *name;
/**
+ * Kernel driver name, as accepted by drmOpenByName.
+ */
+ const char *driver_name;
+
+ /**
* Special buffer functions
*/
/*@{*/
diff --git a/src/gallium/state_trackers/dri/dri_context.c b/src/gallium/state_trackers/dri/dri_context.c
index f2e5f3f..07f0554 100644
--- a/src/gallium/state_trackers/dri/dri_context.c
+++ b/src/gallium/state_trackers/dri/dri_context.c
@@ -101,6 +101,12 @@
{
struct dri_context *ctx = dri_context(cPriv);
+ /* note: we are freeing values and nothing more because
+ * driParseConfigFiles allocated values only - the rest
+ * is owned by screen optionCache.
+ */
+ FREE(ctx->optionCache.values);
+
/* No particular reason to wait for command completion before
* destroying a context, but it is probably worthwhile flushing it
* to avoid having to add code elsewhere to cope with flushing a
diff --git a/src/gallium/state_trackers/dri/dri_drawable.c b/src/gallium/state_trackers/dri/dri_drawable.c
index 0fdfa96..28fd8de 100644
--- a/src/gallium/state_trackers/dri/dri_drawable.c
+++ b/src/gallium/state_trackers/dri/dri_drawable.c
@@ -123,11 +123,12 @@
struct dri_drawable *drawable = dri_drawable(dPriv);
struct pipe_surface *surface = NULL;
- struct pipe_screen *screen = dri_screen(drawable->sPriv)->pipe_screen;
+ struct dri_screen *st_screen = dri_screen(drawable->sPriv);
+ struct pipe_screen *screen = st_screen->pipe_screen;
__DRIbuffer *buffers = NULL;
__DRIscreen *dri_screen = drawable->sPriv;
__DRIdrawable *dri_drawable = drawable->dPriv;
- struct drm_api *api = ((struct dri_screen*)(dri_screen->private))->api;
+ struct drm_api *api = st_screen->api;
boolean have_depth = FALSE;
int i, count;
@@ -180,7 +181,9 @@
switch (buffers[i].attachment) {
case __DRI_BUFFER_FRONT_LEFT:
- continue;
+ if (!st_screen->auto_fake_front)
+ continue;
+ /* fallthrough */
case __DRI_BUFFER_FAKE_FRONT_LEFT:
index = ST_SURFACE_FRONT_LEFT;
format = drawable->color_format;
@@ -373,8 +376,8 @@
/* TODO incase of double buffer visual, delay fake creation */
i = 0;
drawable->attachments[i++] = __DRI_BUFFER_FRONT_LEFT;
- drawable->attachments[i++] = __DRI_BUFFER_FAKE_FRONT_LEFT;
-
+ if (!screen->auto_fake_front)
+ drawable->attachments[i++] = __DRI_BUFFER_FAKE_FRONT_LEFT;
if (visual->doubleBufferMode)
drawable->attachments[i++] = __DRI_BUFFER_BACK_LEFT;
if (visual->depthBits && visual->stencilBits)
diff --git a/src/gallium/state_trackers/dri/dri_screen.c b/src/gallium/state_trackers/dri/dri_screen.c
index d8c0543..cdc8eb1 100644
--- a/src/gallium/state_trackers/dri/dri_screen.c
+++ b/src/gallium/state_trackers/dri/dri_screen.c
@@ -292,6 +292,8 @@
{
struct dri_screen *screen;
struct drm_create_screen_arg arg;
+ const __DRIdri2LoaderExtension *dri2_ext =
+ sPriv->dri2.loader;
screen = CALLOC_STRUCT(dri_screen);
if (!screen)
@@ -317,6 +319,9 @@
driParseOptionInfo(&screen->optionCache,
__driConfigOptions, __driNConfigOptions);
+ screen->auto_fake_front = dri2_ext->base.version >= 3 &&
+ dri2_ext->getBuffersWithFormat != NULL;
+
return dri_fill_in_modes(screen, 32);
fail:
return NULL;
@@ -326,8 +331,18 @@
dri_destroy_screen(__DRIscreen * sPriv)
{
struct dri_screen *screen = dri_screen(sPriv);
+ int i;
screen->pipe_screen->destroy(screen->pipe_screen);
+
+ for (i = 0; i < (1 << screen->optionCache.tableSize); ++i) {
+ FREE(screen->optionCache.info[i].name);
+ FREE(screen->optionCache.info[i].ranges);
+ }
+
+ FREE(screen->optionCache.info);
+ FREE(screen->optionCache.values);
+
FREE(screen);
sPriv->private = NULL;
}
diff --git a/src/gallium/state_trackers/dri/dri_screen.h b/src/gallium/state_trackers/dri/dri_screen.h
index 03387a0..75a0ee4 100644
--- a/src/gallium/state_trackers/dri/dri_screen.h
+++ b/src/gallium/state_trackers/dri/dri_screen.h
@@ -59,6 +59,7 @@
struct pipe_screen *pipe_screen;
boolean d_depth_bits_last;
boolean sd_depth_bits_last;
+ boolean auto_fake_front;
};
/** cast wrapper */
diff --git a/src/gallium/state_trackers/wgl/stw_pixelformat.c b/src/gallium/state_trackers/wgl/stw_pixelformat.c
index 54cc361..7d4c243 100644
--- a/src/gallium/state_trackers/wgl/stw_pixelformat.c
+++ b/src/gallium/state_trackers/wgl/stw_pixelformat.c
@@ -95,8 +95,6 @@
{ PIPE_FORMAT_Z24X8_UNORM, {24, 0} },
{ PIPE_FORMAT_X8Z24_UNORM, {24, 0} },
{ PIPE_FORMAT_Z16_UNORM, {16, 0} },
- /* pure stencil */
- { PIPE_FORMAT_S8_UNORM, { 0, 8} },
/* combined depth-stencil */
{ PIPE_FORMAT_S8Z24_UNORM, {24, 8} },
{ PIPE_FORMAT_Z24S8_UNORM, {24, 8} }
@@ -220,7 +218,8 @@
const struct stw_pf_color_info *color = &stw_pf_color[j];
if(!screen->is_format_supported(screen, color->format, PIPE_TEXTURE_2D,
- PIPE_TEXTURE_USAGE_RENDER_TARGET, 0))
+ PIPE_TEXTURE_USAGE_RENDER_TARGET |
+ PIPE_TEXTURE_USAGE_DISPLAY_TARGET, 0))
continue;
for(k = 0; k < Elements(stw_pf_doublebuffer); ++k) {
diff --git a/src/gallium/state_trackers/xorg/xorg_composite.c b/src/gallium/state_trackers/xorg/xorg_composite.c
index 1c248a6..0324441 100644
--- a/src/gallium/state_trackers/xorg/xorg_composite.c
+++ b/src/gallium/state_trackers/xorg/xorg_composite.c
@@ -4,10 +4,7 @@
#include "xorg_exa_tgsi.h"
#include "cso_cache/cso_context.h"
-#include "util/u_draw_quad.h"
-#include "util/u_math.h"
-#include "pipe/p_inlines.h"
/*XXX also in Xrender.h but the including it here breaks compilition */
#define XFixedToDouble(f) (((double) (f)) / 65536.)
diff --git a/src/gallium/state_trackers/xorg/xorg_dri2.c b/src/gallium/state_trackers/xorg/xorg_dri2.c
index fd82f4f..e6a89c7 100644
--- a/src/gallium/state_trackers/xorg/xorg_dri2.c
+++ b/src/gallium/state_trackers/xorg/xorg_dri2.c
@@ -44,9 +44,12 @@
#include "util/u_rect.h"
/* Make all the #if cases in the code esier to read */
-/* XXX can it be set to 1? */
#ifndef DRI2INFOREC_VERSION
-#define DRI2INFOREC_VERSION 0
+#define DRI2INFOREC_VERSION 1
+#endif
+
+#if DRI2INFOREC_VERSION == 2
+static Bool set_format_in_do_create_buffer;
#endif
typedef struct {
@@ -147,7 +150,9 @@
buffer->driverPrivate = private;
buffer->flags = 0; /* not tiled */
#if DRI2INFOREC_VERSION == 2
- ((DRI2Buffer2Ptr)buffer)->format = 0;
+ /* ABI forwards/backwards compatibility */
+ if (set_format_in_do_create_buffer)
+ ((DRI2Buffer2Ptr)buffer)->format = 0;
#elif DRI2INFOREC_VERSION >= 3
buffer->format = 0;
#endif
@@ -211,7 +216,9 @@
xfree(buffer);
}
-#else /* DRI2INFOREC_VERSION < 2 */
+#endif /* DRI2INFOREC_VERSION >= 2 */
+
+#if DRI2INFOREC_VERSION <= 2
static DRI2BufferPtr
dri2_create_buffers(DrawablePtr pDraw, unsigned int *attachments, int count)
@@ -261,7 +268,7 @@
}
}
-#endif /* DRI2INFOREC_VERSION >= 2 */
+#endif /* DRI2INFOREC_VERSION <= 2 */
static void
dri2_copy_region(DrawablePtr pDraw, RegionPtr pRegion,
@@ -369,12 +376,17 @@
ScrnInfoPtr pScrn = xf86Screens[pScreen->myNum];
modesettingPtr ms = modesettingPTR(pScrn);
DRI2InfoRec dri2info;
+ int major, minor;
-#if DRI2INFOREC_VERSION >= 2
+ if (xf86LoaderCheckSymbol("DRI2Version")) {
+ DRI2Version(&major, &minor);
+ } else {
+ /* Assume version 1.0 */
+ major = 1;
+ minor = 0;
+ }
+
dri2info.version = DRI2INFOREC_VERSION;
-#else
- dri2info.version = 1;
-#endif
dri2info.fd = ms->fd;
dri2info.driverName = pScrn->driverName;
@@ -383,7 +395,22 @@
#if DRI2INFOREC_VERSION >= 2
dri2info.CreateBuffer = dri2_create_buffer;
dri2info.DestroyBuffer = dri2_destroy_buffer;
-#else
+#endif
+
+ /* For X servers in the 1.6.x series there where two DRI2 version.
+ * This allows us to build one binary that works on both servers.
+ */
+#if DRI2INFOREC_VERSION == 2
+ if (minor == 0) {
+ set_format_in_do_create_buffer = FALSE;
+ dri2info.CreateBuffers = dri2_create_buffers;
+ dri2info.DestroyBuffers = dri2_destroy_buffers;
+ } else
+ set_format_in_do_create_buffer = FALSE;
+#endif
+
+ /* For version 1 set these unconditionaly. */
+#if DRI2INFOREC_VERSION == 1
dri2info.CreateBuffers = dri2_create_buffers;
dri2info.DestroyBuffers = dri2_destroy_buffers;
#endif
diff --git a/src/gallium/state_trackers/xorg/xorg_driver.c b/src/gallium/state_trackers/xorg/xorg_driver.c
index b02fe68..41bfcd0 100644
--- a/src/gallium/state_trackers/xorg/xorg_driver.c
+++ b/src/gallium/state_trackers/xorg/xorg_driver.c
@@ -45,7 +45,6 @@
#include "miscstruct.h"
#include "dixstruct.h"
#include "xf86xv.h"
-#include <X11/extensions/Xv.h>
#ifndef XSERVER_LIBPCIACCESS
#error "libpciaccess needed"
#endif
@@ -206,16 +205,41 @@
ms->PciInfo->dev, ms->PciInfo->func
);
- ms->fd = drmOpen(NULL, BusID);
- if (ms->fd < 0)
- return FALSE;
+ ms->api = drm_api_create();
+ ms->fd = drmOpen(ms->api ? ms->api->driver_name : NULL, BusID);
+ xfree(BusID);
+
+ if (ms->fd >= 0)
+ return TRUE;
+
+ if (ms->api && ms->api->destroy)
+ ms->api->destroy(ms->api);
+
+ ms->api = NULL;
+
+ return FALSE;
}
return TRUE;
}
static Bool
+drv_close_drm(ScrnInfoPtr pScrn)
+{
+ modesettingPtr ms = modesettingPTR(pScrn);
+
+ if (ms->api && ms->api->destroy)
+ ms->api->destroy(ms->api);
+ ms->api = NULL;
+
+ drmClose(ms->fd);
+ ms->fd = -1;
+
+ return TRUE;
+}
+
+static Bool
drv_init_resource_management(ScrnInfoPtr pScrn)
{
modesettingPtr ms = modesettingPTR(pScrn);
@@ -229,7 +253,6 @@
if (ms->screen || ms->kms)
return TRUE;
- ms->api = drm_api_create();
if (ms->api) {
ms->screen = ms->api->create_screen(ms->api, ms->fd, NULL);
@@ -269,10 +292,6 @@
}
ms->screen = NULL;
- if (ms->api && ms->api->destroy)
- ms->api->destroy(ms->api);
- ms->api = NULL;
-
#ifdef HAVE_LIBKMS
if (ms->kms)
kms_destroy(&ms->kms);
@@ -823,8 +842,7 @@
drv_close_resource_management(pScrn);
- drmClose(ms->fd);
- ms->fd = -1;
+ drv_close_drm(pScrn);
pScrn->vtSema = FALSE;
pScreen->CloseScreen = ms->CloseScreen;
diff --git a/src/gallium/state_trackers/xorg/xorg_exa.c b/src/gallium/state_trackers/xorg/xorg_exa.c
index d9432ba..c91dee7 100644
--- a/src/gallium/state_trackers/xorg/xorg_exa.c
+++ b/src/gallium/state_trackers/xorg/xorg_exa.c
@@ -41,7 +41,6 @@
#include "pipe/p_format.h"
#include "pipe/p_context.h"
#include "pipe/p_state.h"
-#include "pipe/p_inlines.h"
#include "util/u_format.h"
#include "util/u_rect.h"
diff --git a/src/gallium/state_trackers/xorg/xorg_exa_tgsi.c b/src/gallium/state_trackers/xorg/xorg_exa_tgsi.c
index bed17ca..3e5e6bd 100644
--- a/src/gallium/state_trackers/xorg/xorg_exa_tgsi.c
+++ b/src/gallium/state_trackers/xorg/xorg_exa_tgsi.c
@@ -6,11 +6,9 @@
#include "pipe/p_format.h"
#include "pipe/p_context.h"
#include "pipe/p_state.h"
-#include "pipe/p_inlines.h"
#include "pipe/p_shader_tokens.h"
#include "util/u_memory.h"
-#include "util/u_simple_shaders.h"
#include "tgsi/tgsi_ureg.h"
diff --git a/src/gallium/state_trackers/xorg/xorg_output.c b/src/gallium/state_trackers/xorg/xorg_output.c
index 251f331..13c3fb9 100644
--- a/src/gallium/state_trackers/xorg/xorg_output.c
+++ b/src/gallium/state_trackers/xorg/xorg_output.c
@@ -49,8 +49,6 @@
#include <X11/extensions/dpms.h>
#endif
-#include "X11/Xatom.h"
-
#include "xorg_tracker.h"
static char *output_enum_list[] = {
diff --git a/src/gallium/state_trackers/xorg/xorg_xv.c b/src/gallium/state_trackers/xorg/xorg_xv.c
index 5bf0e94..7bcf77e 100644
--- a/src/gallium/state_trackers/xorg/xorg_xv.c
+++ b/src/gallium/state_trackers/xorg/xorg_xv.c
@@ -11,7 +11,6 @@
#include "cso_cache/cso_context.h"
#include "pipe/p_screen.h"
-#include "pipe/p_inlines.h"
#include "util/u_format.h"
diff --git a/src/gallium/winsys/drm/intel/gem/intel_drm_api.c b/src/gallium/winsys/drm/intel/gem/intel_drm_api.c
index 450ae09..8c8176e 100644
--- a/src/gallium/winsys/drm/intel/gem/intel_drm_api.c
+++ b/src/gallium/winsys/drm/intel/gem/intel_drm_api.c
@@ -196,6 +196,7 @@
struct drm_api intel_drm_api =
{
.name = "i915",
+ .driver_name = "i915",
.create_context = intel_drm_create_context,
.create_screen = intel_drm_create_screen,
.texture_from_shared_handle = intel_drm_texture_from_shared_handle,
diff --git a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c
index 4b2c6a1..c9f39d8 100644
--- a/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c
+++ b/src/gallium/winsys/drm/nouveau/drm/nouveau_drm_api.c
@@ -255,6 +255,7 @@
struct drm_api drm_api_hooks = {
.name = "nouveau",
+ .driver_name = "nouveau",
.create_screen = nouveau_drm_create_screen,
.create_context = nouveau_drm_create_context,
.texture_from_shared_handle = nouveau_drm_pt_from_name,
diff --git a/src/gallium/winsys/drm/radeon/core/radeon_drm.c b/src/gallium/winsys/drm/radeon/core/radeon_drm.c
index 9552f0a..bff6fdc 100644
--- a/src/gallium/winsys/drm/radeon/core/radeon_drm.c
+++ b/src/gallium/winsys/drm/radeon/core/radeon_drm.c
@@ -270,6 +270,7 @@
struct drm_api drm_api_hooks = {
.name = "radeon",
+ .driver_name = "radeon",
.create_screen = radeon_create_screen,
.create_context = radeon_create_context,
.texture_from_shared_handle = radeon_texture_from_shared_handle,
diff --git a/src/gallium/winsys/drm/vmware/core/vmw_context.c b/src/gallium/winsys/drm/vmware/core/vmw_context.c
index b699758..b5fd4f5 100644
--- a/src/gallium/winsys/drm/vmware/core/vmw_context.c
+++ b/src/gallium/winsys/drm/vmware/core/vmw_context.c
@@ -41,9 +41,18 @@
#define VMW_COMMAND_SIZE (64*1024)
#define VMW_SURFACE_RELOCS (1024)
+#define VMW_REGION_RELOCS (512)
#define VMW_MUST_FLUSH_STACK 8
+struct vmw_region_relocation
+{
+ struct SVGAGuestPtr *where;
+ struct pb_buffer *buffer;
+ /* TODO: put offset info inside where */
+ uint32 offset;
+};
+
struct vmw_svga_winsys_context
{
struct svga_winsys_context base;
@@ -69,10 +78,31 @@
uint32_t staged;
uint32_t reserved;
} surface;
+
+ struct {
+ struct vmw_region_relocation relocs[VMW_REGION_RELOCS];
+ uint32_t size;
+ uint32_t used;
+ uint32_t staged;
+ uint32_t reserved;
+ } region;
struct pb_validate *validate;
uint32_t last_fence;
+
+ /**
+ * The amount of GMR that is referred by the commands currently batched
+ * in the context.
+ */
+ uint32_t seen_regions;
+
+ /**
+ * Whether this context should fail to reserve more commands, not because it
+ * ran out of command space, but because a substantial ammount of GMR was
+ * referred.
+ */
+ boolean preemptive_flush;
};
@@ -96,6 +126,19 @@
ret = pb_validate_validate(vswc->validate);
assert(ret == PIPE_OK);
if(ret == PIPE_OK) {
+
+ /* Apply relocations */
+ for(i = 0; i < vswc->region.used; ++i) {
+ struct vmw_region_relocation *reloc = &vswc->region.relocs[i];
+ struct SVGAGuestPtr ptr;
+
+ if(!vmw_gmr_bufmgr_region_ptr(reloc->buffer, &ptr))
+ assert(0);
+
+ ptr.offset += reloc->offset;
+
+ *reloc->where = ptr;
+ }
if (vswc->command.used)
vmw_ioctl_command(vswc->vws,
@@ -121,9 +164,18 @@
vswc->surface.used = 0;
vswc->surface.reserved = 0;
+ for(i = 0; i < vswc->region.used + vswc->region.staged; ++i) {
+ pb_reference(&vswc->region.relocs[i].buffer, NULL);
+ }
+
+ vswc->region.used = 0;
+ vswc->region.reserved = 0;
+
#ifdef DEBUG
vswc->must_flush = FALSE;
#endif
+ vswc->preemptive_flush = FALSE;
+ vswc->seen_regions = 0;
if(pfence)
*pfence = fence;
@@ -151,8 +203,10 @@
if(nr_bytes > vswc->command.size)
return NULL;
- if(vswc->command.used + nr_bytes > vswc->command.size ||
- vswc->surface.used + nr_relocs > vswc->surface.size) {
+ if(vswc->preemptive_flush ||
+ vswc->command.used + nr_bytes > vswc->command.size ||
+ vswc->surface.used + nr_relocs > vswc->surface.size ||
+ vswc->region.used + nr_relocs > vswc->region.size) {
#ifdef DEBUG
vswc->must_flush = TRUE;
debug_backtrace_capture(vswc->must_flush_stack, 1,
@@ -163,11 +217,14 @@
assert(vswc->command.used + nr_bytes <= vswc->command.size);
assert(vswc->surface.used + nr_relocs <= vswc->surface.size);
-
+ assert(vswc->region.used + nr_relocs <= vswc->region.size);
+
vswc->command.reserved = nr_bytes;
vswc->surface.reserved = nr_relocs;
vswc->surface.staged = 0;
-
+ vswc->region.reserved = nr_relocs;
+ vswc->region.staged = 0;
+
return vswc->command.buffer + vswc->command.used;
}
@@ -206,20 +263,41 @@
unsigned flags)
{
struct vmw_svga_winsys_context *vswc = vmw_svga_winsys_context(swc);
- struct SVGAGuestPtr ptr;
- struct pb_buffer *buf = vmw_pb_buffer(buffer);
+ struct vmw_region_relocation *reloc;
enum pipe_error ret;
+
+ assert(vswc->region.staged < vswc->region.reserved);
- if(!vmw_gmr_bufmgr_region_ptr(buf, &ptr))
- assert(0);
+ reloc = &vswc->region.relocs[vswc->region.used + vswc->region.staged];
+ reloc->where = where;
+ pb_reference(&reloc->buffer, vmw_pb_buffer(buffer));
+ reloc->offset = offset;
- ptr.offset += offset;
+ ++vswc->region.staged;
- *where = ptr;
-
- ret = pb_validate_add_buffer(vswc->validate, buf, flags);
+ ret = pb_validate_add_buffer(vswc->validate, reloc->buffer, flags);
/* TODO: Update pipebuffer to reserve buffers and not fail here */
assert(ret == PIPE_OK);
+
+ /*
+ * Flush preemptively the FIFO commands to keep the GMR working set within
+ * the GMR pool size.
+ *
+ * This is necessary for applications like SPECviewperf that generate huge
+ * amounts of immediate vertex data, so that we don't pile up too much of
+ * that vertex data neither in the guest nor in the host.
+ *
+ * Note that in the current implementation if a region is referred twice in
+ * a command stream, it will be accounted twice. We could detect repeated
+ * regions and count only once, but there is no incentive to do that, since
+ * regions are typically short-lived; always referred in a single command;
+ * and at the worst we just flush the commands a bit sooner, which for the
+ * SVGA virtual device it's not a performance issue since flushing commands
+ * to the FIFO won't cause flushing in the host.
+ */
+ vswc->seen_regions += reloc->buffer->base.size;
+ if(vswc->seen_regions >= VMW_GMR_POOL_SIZE/2)
+ vswc->preemptive_flush = TRUE;
}
@@ -238,6 +316,12 @@
vswc->surface.used += vswc->surface.staged;
vswc->surface.staged = 0;
vswc->surface.reserved = 0;
+
+ assert(vswc->region.staged <= vswc->region.reserved);
+ assert(vswc->region.used + vswc->region.staged <= vswc->region.size);
+ vswc->region.used += vswc->region.staged;
+ vswc->region.staged = 0;
+ vswc->region.reserved = 0;
}
@@ -246,6 +330,11 @@
{
struct vmw_svga_winsys_context *vswc = vmw_svga_winsys_context(swc);
unsigned i;
+
+ for(i = 0; i < vswc->region.used; ++i) {
+ pb_reference(&vswc->region.relocs[i].buffer, NULL);
+ }
+
for(i = 0; i < vswc->surface.used; ++i) {
p_atomic_dec(&vswc->surface.handles[i]->validated);
vmw_svga_winsys_surface_reference(&vswc->surface.handles[i], NULL);
@@ -279,6 +368,7 @@
vswc->command.size = VMW_COMMAND_SIZE;
vswc->surface.size = VMW_SURFACE_RELOCS;
+ vswc->region.size = VMW_REGION_RELOCS;
vswc->validate = pb_validate_create();
if(!vswc->validate) {
diff --git a/src/gallium/winsys/drm/vmware/core/vmw_screen.h b/src/gallium/winsys/drm/vmware/core/vmw_screen.h
index a875107..f1d6986 100644
--- a/src/gallium/winsys/drm/vmware/core/vmw_screen.h
+++ b/src/gallium/winsys/drm/vmware/core/vmw_screen.h
@@ -40,6 +40,10 @@
#include "svga_winsys.h"
+
+#define VMW_GMR_POOL_SIZE (16*1024*1024)
+
+
struct pb_manager;
struct vmw_region;
diff --git a/src/gallium/winsys/drm/vmware/core/vmw_screen_dri.c b/src/gallium/winsys/drm/vmware/core/vmw_screen_dri.c
index 4f5ccea..8be9d74 100644
--- a/src/gallium/winsys/drm/vmware/core/vmw_screen_dri.c
+++ b/src/gallium/winsys/drm/vmware/core/vmw_screen_dri.c
@@ -85,6 +85,23 @@
struct pipe_screen *screen;
struct dri1_create_screen_arg *dri1;
+ if (!arg || arg->mode == DRM_CREATE_NORMAL) {
+ struct dri1_api_version drm_ver;
+ drmVersionPtr ver;
+
+ ver = drmGetVersion(fd);
+ if (ver == NULL)
+ return NULL;
+
+ drm_ver.major = ver->version_major;
+ drm_ver.minor = ver->version_minor;
+
+ drmFreeVersion(ver);
+ if (!vmw_dri1_check_version(&drm_ver, &drm_required,
+ &drm_compat, "vmwgfx drm driver"))
+ return NULL;
+ }
+
if (arg != NULL) {
switch (arg->mode) {
case DRM_CREATE_NORMAL:
@@ -220,22 +237,19 @@
vmw_svga_winsys_surface_reference(&vsrf, NULL);
}
-/**
- * FIXME: We'd probably want to cache these buffers in the
- * screen, based on handle.
- */
-
-static struct pipe_buffer *
-vmw_drm_buffer_from_handle(struct drm_api *drm_api,
- struct pipe_screen *screen,
- const char *name,
- unsigned handle)
+static struct pipe_texture *
+vmw_drm_texture_from_handle(struct drm_api *drm_api,
+ struct pipe_screen *screen,
+ struct pipe_texture *templat,
+ const char *name,
+ unsigned stride,
+ unsigned handle)
{
struct vmw_svga_winsys_surface *vsrf;
struct svga_winsys_surface *ssrf;
struct vmw_winsys_screen *vws =
vmw_winsys_screen(svga_winsys_screen(screen));
- struct pipe_buffer *buf;
+ struct pipe_texture *tex;
union drm_vmw_surface_reference_arg arg;
struct drm_vmw_surface_arg *req = &arg.req;
struct drm_vmw_surface_create_req *rep = &arg.rep;
@@ -282,43 +296,28 @@
pipe_reference_init(&vsrf->refcnt, 1);
p_atomic_set(&vsrf->validated, 0);
+ vsrf->screen = vws;
vsrf->sid = handle;
ssrf = svga_winsys_surface(vsrf);
- buf = svga_screen_buffer_wrap_surface(screen, rep->format, ssrf);
- if (!buf)
+ tex = svga_screen_texture_wrap_surface(screen, templat, rep->format, ssrf);
+ if (!tex)
vmw_svga_winsys_surface_reference(&vsrf, NULL);
- return buf;
+ return tex;
out_mip:
vmw_ioctl_surface_destroy(vws, handle);
return NULL;
}
-static struct pipe_texture *
-vmw_drm_texture_from_handle(struct drm_api *drm_api,
- struct pipe_screen *screen,
- struct pipe_texture *templat,
- const char *name,
- unsigned stride,
- unsigned handle)
-{
- struct pipe_buffer *buffer;
- buffer = vmw_drm_buffer_from_handle(drm_api, screen, name, handle);
-
- if (!buffer)
- return NULL;
-
- return screen->texture_blanket(screen, templat, &stride, buffer);
-}
-
static boolean
-vmw_drm_handle_from_buffer(struct drm_api *drm_api,
+vmw_drm_handle_from_texture(struct drm_api *drm_api,
struct pipe_screen *screen,
- struct pipe_buffer *buffer,
+ struct pipe_texture *texture,
+ unsigned *stride,
unsigned *handle)
{
struct svga_winsys_surface *surface =
- svga_screen_buffer_get_winsys_surface(buffer);
+ svga_screen_texture_get_winsys_surface(texture);
struct vmw_svga_winsys_surface *vsrf;
if (!surface)
@@ -326,25 +325,13 @@
vsrf = vmw_svga_winsys_surface(surface);
*handle = vsrf->sid;
+ *stride = pf_get_nblocksx(&texture->block, texture->width[0]) *
+ texture->block.size;
+
vmw_svga_winsys_surface_reference(&vsrf, NULL);
return TRUE;
}
-static boolean
-vmw_drm_handle_from_texture(struct drm_api *drm_api,
- struct pipe_screen *screen,
- struct pipe_texture *texture,
- unsigned *stride,
- unsigned *handle)
-{
- struct pipe_buffer *buffer;
-
- if (!svga_screen_buffer_from_texture(texture, &buffer, stride))
- return FALSE;
-
- return vmw_drm_handle_from_buffer(drm_api, screen, buffer, handle);
-}
-
static struct pipe_context*
vmw_drm_create_context(struct drm_api *drm_api,
struct pipe_screen *screen)
@@ -359,6 +346,7 @@
static struct drm_api vmw_drm_api_hooks = {
.name = "vmwgfx",
+ .driver_name = "vmwgfx",
.create_screen = vmw_drm_create_screen,
.create_context = vmw_drm_create_context,
.texture_from_shared_handle = vmw_drm_texture_from_handle,
diff --git a/src/gallium/winsys/drm/vmware/core/vmw_screen_pools.c b/src/gallium/winsys/drm/vmware/core/vmw_screen_pools.c
index b1c24b0..b9823d7 100644
--- a/src/gallium/winsys/drm/vmware/core/vmw_screen_pools.c
+++ b/src/gallium/winsys/drm/vmware/core/vmw_screen_pools.c
@@ -53,14 +53,32 @@
goto error;
vws->pools.gmr_mm = mm_bufmgr_create(vws->pools.gmr,
- 16*1024*1024,
+ VMW_GMR_POOL_SIZE,
12 /* 4096 alignment */);
if(!vws->pools.gmr_mm)
goto error;
+ /*
+ * GMR buffers are typically shortlived, but it's possible that at a given
+ * instance a buffer is mapped. So to avoid stalling we tell pipebuffer to
+ * forbid creation of buffers beyond half the GMR pool size,
+ *
+ * XXX: It is unclear weather we want to limit the total amount of temporary
+ * malloc memory used to backup unvalidated GMR buffers. On one hand it is
+ * preferrable to fail an allocation than exhausting the guest memory with
+ * temporary data, but on the other hand it is possible that a stupid
+ * application creates large vertex buffers and does not use them for a long
+ * time -- since the svga pipe driver only emits the DMA uploads when a
+ * buffer is used for drawing this would effectively disabling swapping GMR
+ * buffers to memory. So far, the preemptively flush already seems to keep
+ * total allocated memory within relatively small numbers, so we don't
+ * limit.
+ */
vws->pools.gmr_fenced = fenced_bufmgr_create(
vws->pools.gmr_mm,
- vmw_fence_ops_create(vws));
+ vmw_fence_ops_create(vws),
+ VMW_GMR_POOL_SIZE/2,
+ ~0);
#ifdef DEBUG
vws->pools.gmr_fenced = pb_debug_manager_create(vws->pools.gmr_fenced,
diff --git a/src/gallium/winsys/xlib/xlib_softpipe.c b/src/gallium/winsys/xlib/xlib_softpipe.c
index f7c0099..49dbf44 100644
--- a/src/gallium/winsys/xlib/xlib_softpipe.c
+++ b/src/gallium/winsys/xlib/xlib_softpipe.c
@@ -204,6 +204,14 @@
{
struct xm_buffer *oldBuf = xm_buffer(buf);
+ /*
+ * Note oldBuf->data may point to one of three things:
+ * 1. XShm shared memory image data
+ * 2. User-provided (wrapped) memory, see xm_user_buffer_create()
+ * 3. Regular, malloc'd memory
+ * We need to be careful with freeing that data now.
+ */
+
if (oldBuf->data) {
#ifdef USE_XSHM
if (oldBuf->shminfo.shmid >= 0) {
@@ -213,12 +221,19 @@
oldBuf->shminfo.shmid = -1;
oldBuf->shminfo.shmaddr = (char *) -1;
}
- else
-#endif
- {
- if (!oldBuf->userBuffer) {
- align_free(oldBuf->data);
+
+ if (oldBuf->tempImage) {
+ if (oldBuf->data == oldBuf->tempImage->data) {
+ /* oldBuf->data points at the xshm memory which we'll now free */
+ oldBuf->data = NULL;
}
+ XDestroyImage(oldBuf->tempImage);
+ }
+#endif
+
+ if (oldBuf->data && !oldBuf->userBuffer) {
+ /* this was regular malloc'd memory */
+ align_free(oldBuf->data);
}
oldBuf->data = NULL;
diff --git a/src/glx/x11/glxcurrent.c b/src/glx/x11/glxcurrent.c
index fae1bd9..50de7d6 100644
--- a/src/glx/x11/glxcurrent.c
+++ b/src/glx/x11/glxcurrent.c
@@ -162,6 +162,7 @@
__glXSetCurrentContext(&dummyContext);
#ifdef GLX_DIRECT_RENDERING
_glapi_set_dispatch(NULL); /* no-op functions */
+ _glapi_set_context(NULL);
#endif
}
diff --git a/src/mesa/drivers/dri/common/spantmp2.h b/src/mesa/drivers/dri/common/spantmp2.h
index 447f3d1..c152226 100644
--- a/src/mesa/drivers/dri/common/spantmp2.h
+++ b/src/mesa/drivers/dri/common/spantmp2.h
@@ -400,7 +400,7 @@
# define READ_RGBA( rgba, _x, _y ) \
do { \
GLuint p = GET_VALUE(_x, _y); \
- *((uint32_t *) rgba) = (t << 8) | 0xff; \
+ *((uint32_t *) rgba) = (p << 8) | 0xff; \
} while (0)
# else
# define READ_RGBA( rgba, _x, _y ) \
diff --git a/src/mesa/drivers/dri/ffb/ffb_bitmap.c b/src/mesa/drivers/dri/ffb/ffb_bitmap.c
index 611afdd..b71a552 100644
--- a/src/mesa/drivers/dri/ffb/ffb_bitmap.c
+++ b/src/mesa/drivers/dri/ffb/ffb_bitmap.c
@@ -30,7 +30,6 @@
#include "ffb_lock.h"
#include "ffb_bitmap.h"
#include "swrast/swrast.h"
-#include "main/image.h"
#include "main/macros.h"
/* Compute ceiling of integer quotient of A divided by B: */
diff --git a/src/mesa/drivers/dri/ffb/ffb_clear.c b/src/mesa/drivers/dri/ffb/ffb_clear.c
index dfe60f3..aa3fa0a 100644
--- a/src/mesa/drivers/dri/ffb/ffb_clear.c
+++ b/src/mesa/drivers/dri/ffb/ffb_clear.c
@@ -26,15 +26,12 @@
*/
#include "main/mtypes.h"
-#include "main/extensions.h"
#include "main/mm.h"
#include "ffb_dd.h"
#include "ffb_span.h"
-#include "ffb_depth.h"
#include "ffb_context.h"
#include "ffb_vb.h"
-#include "ffb_tris.h"
#include "ffb_clear.h"
#include "ffb_lock.h"
diff --git a/src/mesa/drivers/dri/ffb/ffb_dd.c b/src/mesa/drivers/dri/ffb/ffb_dd.c
index cf83b91..91b6d31 100644
--- a/src/mesa/drivers/dri/ffb/ffb_dd.c
+++ b/src/mesa/drivers/dri/ffb/ffb_dd.c
@@ -27,13 +27,9 @@
#include "main/mtypes.h"
#include "main/mm.h"
-#include "main/extensions.h"
#include "ffb_dd.h"
#include "ffb_span.h"
-#include "ffb_depth.h"
#include "ffb_context.h"
-#include "ffb_vb.h"
-#include "ffb_tris.h"
#include "ffb_clear.h"
#include "ffb_lock.h"
diff --git a/src/mesa/drivers/dri/ffb/ffb_depth.c b/src/mesa/drivers/dri/ffb/ffb_depth.c
index 5d509ff..d19385b 100644
--- a/src/mesa/drivers/dri/ffb/ffb_depth.c
+++ b/src/mesa/drivers/dri/ffb/ffb_depth.c
@@ -26,7 +26,6 @@
*/
#include "main/mtypes.h"
-#include "swrast/swrast.h"
#include "ffb_dd.h"
#include "ffb_span.h"
#include "ffb_context.h"
diff --git a/src/mesa/drivers/dri/ffb/ffb_lines.c b/src/mesa/drivers/dri/ffb/ffb_lines.c
index 19dff50..6dca4ed 100644
--- a/src/mesa/drivers/dri/ffb/ffb_lines.c
+++ b/src/mesa/drivers/dri/ffb/ffb_lines.c
@@ -27,15 +27,11 @@
#include "main/mtypes.h"
#include "main/mm.h"
-#include "main/extensions.h"
#include "ffb_dd.h"
#include "ffb_span.h"
-#include "ffb_depth.h"
#include "ffb_context.h"
#include "ffb_vb.h"
#include "ffb_lines.h"
-#include "ffb_tris.h"
-#include "ffb_lock.h"
#undef FFB_LINE_TRACE
diff --git a/src/mesa/drivers/dri/ffb/ffb_points.c b/src/mesa/drivers/dri/ffb/ffb_points.c
index 9c37a47..5bf4f8f 100644
--- a/src/mesa/drivers/dri/ffb/ffb_points.c
+++ b/src/mesa/drivers/dri/ffb/ffb_points.c
@@ -30,8 +30,6 @@
#include "ffb_context.h"
#include "ffb_vb.h"
#include "ffb_points.h"
-#include "ffb_tris.h"
-#include "ffb_lock.h"
#undef FFB_POINT_TRACE
diff --git a/src/mesa/drivers/dri/ffb/ffb_span.c b/src/mesa/drivers/dri/ffb/ffb_span.c
index 8ec33a1..61901cc 100644
--- a/src/mesa/drivers/dri/ffb/ffb_span.c
+++ b/src/mesa/drivers/dri/ffb/ffb_span.c
@@ -31,8 +31,6 @@
#include "ffb_context.h"
#include "ffb_lock.h"
-#include "swrast/swrast.h"
-
#define DBG 0
#define HW_LOCK() \
diff --git a/src/mesa/drivers/dri/ffb/ffb_state.c b/src/mesa/drivers/dri/ffb/ffb_state.c
index 6f8a46d..c09d2fe 100644
--- a/src/mesa/drivers/dri/ffb/ffb_state.c
+++ b/src/mesa/drivers/dri/ffb/ffb_state.c
@@ -27,8 +27,6 @@
#include "main/mtypes.h"
#include "main/colormac.h"
-#include "main/mm.h"
-#include "main/extensions.h"
#include "main/enums.h"
#include "vbo/vbo.h"
@@ -39,12 +37,9 @@
#include "ffb_dd.h"
#include "ffb_span.h"
-#include "ffb_depth.h"
#include "ffb_context.h"
-#include "ffb_vb.h"
#include "ffb_tris.h"
#include "ffb_state.h"
-#include "ffb_lock.h"
#undef STATE_TRACE
diff --git a/src/mesa/drivers/dri/ffb/ffb_stencil.c b/src/mesa/drivers/dri/ffb/ffb_stencil.c
index ce8ef43..10cdfbc 100644
--- a/src/mesa/drivers/dri/ffb/ffb_stencil.c
+++ b/src/mesa/drivers/dri/ffb/ffb_stencil.c
@@ -32,8 +32,6 @@
#include "ffb_stencil.h"
#include "ffb_lock.h"
-#include "swrast/swrast.h"
-
#undef STENCIL_TRACE
static void FFBWriteStencilSpan( GLcontext *ctx,
diff --git a/src/mesa/drivers/dri/ffb/ffb_vb.c b/src/mesa/drivers/dri/ffb/ffb_vb.c
index f9c6fd1..ca8ffb2 100644
--- a/src/mesa/drivers/dri/ffb/ffb_vb.c
+++ b/src/mesa/drivers/dri/ffb/ffb_vb.c
@@ -30,8 +30,6 @@
#include "ffb_vb.h"
#include "main/imports.h"
#include "tnl/t_context.h"
-#include "swrast_setup/swrast_setup.h"
-#include "math/m_translate.h"
#undef VB_DEBUG
diff --git a/src/mesa/drivers/dri/ffb/ffb_xmesa.c b/src/mesa/drivers/dri/ffb/ffb_xmesa.c
index 88285f4..6a84651 100644
--- a/src/mesa/drivers/dri/ffb/ffb_xmesa.c
+++ b/src/mesa/drivers/dri/ffb/ffb_xmesa.c
@@ -28,7 +28,6 @@
#include "ffb_xmesa.h"
#include "main/context.h"
#include "main/framebuffer.h"
-#include "main/matrix.h"
#include "main/renderbuffer.h"
#include "main/simple_list.h"
#include "main/imports.h"
@@ -52,7 +51,6 @@
#include "ffb_lines.h"
#include "ffb_points.h"
#include "ffb_state.h"
-#include "ffb_tex.h"
#include "ffb_lock.h"
#include "ffb_vtxfmt.h"
#include "ffb_bitmap.h"
diff --git a/src/mesa/drivers/dri/i810/i810context.c b/src/mesa/drivers/dri/i810/i810context.c
index bd9cfe5..34e3460 100644
--- a/src/mesa/drivers/dri/i810/i810context.c
+++ b/src/mesa/drivers/dri/i810/i810context.c
@@ -34,10 +34,7 @@
#include "main/glheader.h"
#include "main/context.h"
-#include "main/matrix.h"
#include "main/simple_list.h"
-#include "main/extensions.h"
-#include "main/framebuffer.h"
#include "main/imports.h"
#include "main/points.h"
diff --git a/src/mesa/drivers/dri/i810/i810render.c b/src/mesa/drivers/dri/i810/i810render.c
index 1d98e00..b543d4f 100644
--- a/src/mesa/drivers/dri/i810/i810render.c
+++ b/src/mesa/drivers/dri/i810/i810render.c
@@ -44,7 +44,6 @@
#include "i810context.h"
#include "i810tris.h"
-#include "i810state.h"
#include "i810vb.h"
#include "i810ioctl.h"
diff --git a/src/mesa/drivers/dri/i810/i810screen.c b/src/mesa/drivers/dri/i810/i810screen.c
index 2a30782..476c801 100644
--- a/src/mesa/drivers/dri/i810/i810screen.c
+++ b/src/mesa/drivers/dri/i810/i810screen.c
@@ -36,8 +36,6 @@
#include "main/imports.h"
#include "main/context.h"
#include "main/framebuffer.h"
-#include "main/fbobject.h"
-#include "main/matrix.h"
#include "main/renderbuffer.h"
#include "main/simple_list.h"
#include "utils.h"
@@ -48,8 +46,6 @@
#include "i810state.h"
#include "i810tex.h"
#include "i810span.h"
-#include "i810tris.h"
-#include "i810ioctl.h"
#include "GL/internal/dri_interface.h"
diff --git a/src/mesa/drivers/dri/i810/i810state.c b/src/mesa/drivers/dri/i810/i810state.c
index 642245c..0c68e12 100644
--- a/src/mesa/drivers/dri/i810/i810state.c
+++ b/src/mesa/drivers/dri/i810/i810state.c
@@ -20,8 +20,6 @@
#include "i810context.h"
#include "i810state.h"
#include "i810tex.h"
-#include "i810vb.h"
-#include "i810tris.h"
#include "i810ioctl.h"
diff --git a/src/mesa/drivers/dri/i810/i810tex.c b/src/mesa/drivers/dri/i810/i810tex.c
index e764644..2ccb956 100644
--- a/src/mesa/drivers/dri/i810/i810tex.c
+++ b/src/mesa/drivers/dri/i810/i810tex.c
@@ -33,7 +33,6 @@
#include "main/colormac.h"
#include "main/texobj.h"
#include "main/mm.h"
-#include "swrast/swrast.h"
#include "texmem.h"
@@ -42,7 +41,6 @@
#include "i810context.h"
#include "i810tex.h"
-#include "i810state.h"
#include "i810ioctl.h"
diff --git a/src/mesa/drivers/dri/i810/i810texmem.c b/src/mesa/drivers/dri/i810/i810texmem.c
index d93afbf..bb426a4 100644
--- a/src/mesa/drivers/dri/i810/i810texmem.c
+++ b/src/mesa/drivers/dri/i810/i810texmem.c
@@ -35,7 +35,6 @@
#include "i810_dri.h"
#include "i810context.h"
#include "i810tex.h"
-#include "i810state.h"
#include "i810ioctl.h"
diff --git a/src/mesa/drivers/dri/i810/i810vb.c b/src/mesa/drivers/dri/i810/i810vb.c
index 30890dc..09a7722 100644
--- a/src/mesa/drivers/dri/i810/i810vb.c
+++ b/src/mesa/drivers/dri/i810/i810vb.c
@@ -38,7 +38,6 @@
#include "i810context.h"
#include "i810vb.h"
#include "i810ioctl.h"
-#include "i810tris.h"
#include "i810state.h"
diff --git a/src/mesa/drivers/dri/mach64/mach64_context.c b/src/mesa/drivers/dri/mach64/mach64_context.c
index 3b4ef7f..11bce31 100644
--- a/src/mesa/drivers/dri/mach64/mach64_context.c
+++ b/src/mesa/drivers/dri/mach64/mach64_context.c
@@ -33,8 +33,6 @@
#include "main/context.h"
#include "main/simple_list.h"
#include "main/imports.h"
-#include "main/matrix.h"
-#include "main/extensions.h"
#include "swrast/swrast.h"
#include "swrast_setup/swrast_setup.h"
diff --git a/src/mesa/drivers/dri/mach64/mach64_dd.c b/src/mesa/drivers/dri/mach64/mach64_dd.c
index e400e9a..ca713e2 100644
--- a/src/mesa/drivers/dri/mach64/mach64_dd.c
+++ b/src/mesa/drivers/dri/mach64/mach64_dd.c
@@ -31,12 +31,9 @@
#include "mach64_context.h"
#include "mach64_ioctl.h"
-#include "mach64_state.h"
-#include "mach64_vb.h"
#include "mach64_dd.h"
#include "main/context.h"
-#include "main/framebuffer.h"
#include "utils.h"
diff --git a/src/mesa/drivers/dri/mach64/mach64_lock.c b/src/mesa/drivers/dri/mach64/mach64_lock.c
index 8653c77..1a95a8f 100644
--- a/src/mesa/drivers/dri/mach64/mach64_lock.c
+++ b/src/mesa/drivers/dri/mach64/mach64_lock.c
@@ -32,7 +32,6 @@
#include "mach64_context.h"
#include "mach64_state.h"
#include "mach64_lock.h"
-#include "mach64_tex.h"
#include "drirenderbuffer.h"
#if DEBUG_LOCKING
diff --git a/src/mesa/drivers/dri/mach64/mach64_screen.c b/src/mesa/drivers/dri/mach64/mach64_screen.c
index 1ed3b0b..5cbfb856 100644
--- a/src/mesa/drivers/dri/mach64/mach64_screen.c
+++ b/src/mesa/drivers/dri/mach64/mach64_screen.c
@@ -31,8 +31,6 @@
#include "mach64_context.h"
#include "mach64_ioctl.h"
-#include "mach64_tris.h"
-#include "mach64_vb.h"
#include "mach64_span.h"
#include "main/context.h"
diff --git a/src/mesa/drivers/dri/mach64/mach64_span.c b/src/mesa/drivers/dri/mach64/mach64_span.c
index b4ba2a4..0c52c0c 100644
--- a/src/mesa/drivers/dri/mach64/mach64_span.c
+++ b/src/mesa/drivers/dri/mach64/mach64_span.c
@@ -31,7 +31,6 @@
#include "mach64_context.h"
#include "mach64_ioctl.h"
-#include "mach64_state.h"
#include "mach64_span.h"
#include "swrast/swrast.h"
diff --git a/src/mesa/drivers/dri/mach64/mach64_state.c b/src/mesa/drivers/dri/mach64/mach64_state.c
index df7cbc8..b9093b5 100644
--- a/src/mesa/drivers/dri/mach64/mach64_state.c
+++ b/src/mesa/drivers/dri/mach64/mach64_state.c
@@ -36,7 +36,6 @@
#include "mach64_vb.h"
#include "mach64_tex.h"
-#include "main/context.h"
#include "main/enums.h"
#include "main/colormac.h"
#include "swrast/swrast.h"
@@ -44,8 +43,6 @@
#include "tnl/tnl.h"
#include "swrast_setup/swrast_setup.h"
-#include "tnl/t_pipeline.h"
-
/* =============================================================
* Alpha blending
diff --git a/src/mesa/drivers/dri/mach64/mach64_tex.c b/src/mesa/drivers/dri/mach64/mach64_tex.c
index 6627d3c..1bce967 100644
--- a/src/mesa/drivers/dri/mach64/mach64_tex.c
+++ b/src/mesa/drivers/dri/mach64/mach64_tex.c
@@ -31,13 +31,8 @@
#include "mach64_context.h"
#include "mach64_ioctl.h"
-#include "mach64_state.h"
-#include "mach64_vb.h"
-#include "mach64_tris.h"
#include "mach64_tex.h"
-#include "main/context.h"
-#include "main/macros.h"
#include "main/simple_list.h"
#include "main/enums.h"
#include "main/texstore.h"
diff --git a/src/mesa/drivers/dri/mach64/mach64_texmem.c b/src/mesa/drivers/dri/mach64/mach64_texmem.c
index b97e9ee..46cee43 100644
--- a/src/mesa/drivers/dri/mach64/mach64_texmem.c
+++ b/src/mesa/drivers/dri/mach64/mach64_texmem.c
@@ -37,10 +37,7 @@
#include "main/imports.h"
#include "mach64_context.h"
-#include "mach64_state.h"
#include "mach64_ioctl.h"
-#include "mach64_vb.h"
-#include "mach64_tris.h"
#include "mach64_tex.h"
diff --git a/src/mesa/drivers/dri/mach64/mach64_texstate.c b/src/mesa/drivers/dri/mach64/mach64_texstate.c
index df0a09a..adf774e 100644
--- a/src/mesa/drivers/dri/mach64/mach64_texstate.c
+++ b/src/mesa/drivers/dri/mach64/mach64_texstate.c
@@ -36,8 +36,6 @@
#include "mach64_context.h"
#include "mach64_ioctl.h"
-#include "mach64_state.h"
-#include "mach64_vb.h"
#include "mach64_tris.h"
#include "mach64_tex.h"
diff --git a/src/mesa/drivers/dri/mach64/mach64_vb.c b/src/mesa/drivers/dri/mach64/mach64_vb.c
index e58812e..00da835 100644
--- a/src/mesa/drivers/dri/mach64/mach64_vb.c
+++ b/src/mesa/drivers/dri/mach64/mach64_vb.c
@@ -42,7 +42,6 @@
#include "mach64_vb.h"
#include "mach64_ioctl.h"
#include "mach64_tris.h"
-#include "mach64_state.h"
#define MACH64_TEX1_BIT 0x1
diff --git a/src/mesa/drivers/dri/r200/r200_state_init.c b/src/mesa/drivers/dri/r200/r200_state_init.c
index 6c5a0b7..80b08dc 100644
--- a/src/mesa/drivers/dri/r200/r200_state_init.c
+++ b/src/mesa/drivers/dri/r200/r200_state_init.c
@@ -698,7 +698,8 @@
uint32_t dwords = atom->check(ctx, atom);
int i = atom->idx;
radeonTexObj *t = r200->state.texture.unit[i].texobj;
- if (!r200->state.texture.unit[i].unitneeded)
+
+ if (!r200->state.texture.unit[i].unitneeded && !(dwords <= atom->cmd_size))
dwords -= 4;
BEGIN_BATCH_NO_AUTOSTATE(dwords);
diff --git a/src/mesa/drivers/dri/r600/r600_texstate.c b/src/mesa/drivers/dri/r600/r600_texstate.c
index 124365b..3289d89 100644
--- a/src/mesa/drivers/dri/r600/r600_texstate.c
+++ b/src/mesa/drivers/dri/r600/r600_texstate.c
@@ -85,6 +85,7 @@
CLEARfield(t->SQ_TEX_RESOURCE4, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
CLEARfield(t->SQ_TEX_RESOURCE4, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
CLEARfield(t->SQ_TEX_RESOURCE4, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+ CLEARbit(t->SQ_TEX_RESOURCE4, SQ_TEX_RESOURCE_WORD4_0__FORCE_DEGAMMA_bit);
SETfield(t->SQ_TEX_RESOURCE4, SQ_FORMAT_COMP_UNSIGNED,
FORMAT_COMP_X_shift, FORMAT_COMP_X_mask);
@@ -95,6 +96,11 @@
SETfield(t->SQ_TEX_RESOURCE4, SQ_FORMAT_COMP_UNSIGNED,
FORMAT_COMP_W_shift, FORMAT_COMP_W_mask);
+ CLEARbit(t->SQ_TEX_RESOURCE0, TILE_TYPE_bit);
+ SETfield(t->SQ_TEX_RESOURCE0, ARRAY_LINEAR_GENERAL,
+ SQ_TEX_RESOURCE_WORD0_0__TILE_MODE_shift,
+ SQ_TEX_RESOURCE_WORD0_0__TILE_MODE_mask);
+
switch (mesa_format) /* This is mesa format. */
{
case MESA_FORMAT_RGBA8888:
@@ -158,6 +164,32 @@
SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_W,
SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
break;
+ case MESA_FORMAT_XRGB8888:
+ SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8_8_8,
+ SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+ SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Z,
+ SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+ SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Y,
+ SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+ SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+ SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+ SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_1,
+ SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+ break;
+ case MESA_FORMAT_XRGB8888_REV:
+ SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8_8_8,
+ SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
+
+ SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_1,
+ SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_X_mask);
+ SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_Z,
+ SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Y_mask);
+ SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_W,
+ SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_Z_mask);
+ SETfield(t->SQ_TEX_RESOURCE4, SQ_SEL_X,
+ SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_shift, SQ_TEX_RESOURCE_WORD4_0__DST_SEL_W_mask);
+ break;
case MESA_FORMAT_ARGB8888_REV:
SETfield(t->SQ_TEX_RESOURCE1, FMT_8_8_8_8,
SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_shift, SQ_TEX_RESOURCE_WORD1_0__DATA_FORMAT_mask);
@@ -515,6 +547,10 @@
case MESA_FORMAT_Z24_S8:
case MESA_FORMAT_Z32:
case MESA_FORMAT_S8:
+ SETbit(t->SQ_TEX_RESOURCE0, TILE_TYPE_bit);
+ SETfield(t->SQ_TEX_RESOURCE0, ARRAY_1D_TILED_THIN1,
+ SQ_TEX_RESOURCE_WORD0_0__TILE_MODE_shift,
+ SQ_TEX_RESOURCE_WORD0_0__TILE_MODE_mask);
switch (mesa_format) {
case MESA_FORMAT_Z16:
SETfield(t->SQ_TEX_RESOURCE1, FMT_16,
diff --git a/src/mesa/drivers/dri/r600/r700_state.c b/src/mesa/drivers/dri/r600/r700_state.c
index a795f62..20e8afe 100644
--- a/src/mesa/drivers/dri/r600/r700_state.c
+++ b/src/mesa/drivers/dri/r600/r700_state.c
@@ -1580,9 +1580,9 @@
SETbit(r700->sq_config.SQ_CONFIG.u32All, DX9_CONSTS_bit);
SETbit(r700->sq_config.SQ_CONFIG.u32All, ALU_INST_PREFER_VECTOR_bit);
SETfield(r700->sq_config.SQ_CONFIG.u32All, ps_prio, PS_PRIO_shift, PS_PRIO_mask);
- SETfield(r700->sq_config.SQ_CONFIG.u32All, ps_prio, VS_PRIO_shift, VS_PRIO_mask);
- SETfield(r700->sq_config.SQ_CONFIG.u32All, ps_prio, GS_PRIO_shift, GS_PRIO_mask);
- SETfield(r700->sq_config.SQ_CONFIG.u32All, ps_prio, ES_PRIO_shift, ES_PRIO_mask);
+ SETfield(r700->sq_config.SQ_CONFIG.u32All, vs_prio, VS_PRIO_shift, VS_PRIO_mask);
+ SETfield(r700->sq_config.SQ_CONFIG.u32All, gs_prio, GS_PRIO_shift, GS_PRIO_mask);
+ SETfield(r700->sq_config.SQ_CONFIG.u32All, es_prio, ES_PRIO_shift, ES_PRIO_mask);
r700->sq_config.SQ_GPR_RESOURCE_MGMT_1.u32All = 0;
SETfield(r700->sq_config.SQ_GPR_RESOURCE_MGMT_1.u32All, num_ps_gprs, NUM_PS_GPRS_shift, NUM_PS_GPRS_mask);
diff --git a/src/mesa/drivers/dri/unichrome/via_context.c b/src/mesa/drivers/dri/unichrome/via_context.c
index d17a160..9da96bd 100644
--- a/src/mesa/drivers/dri/unichrome/via_context.c
+++ b/src/mesa/drivers/dri/unichrome/via_context.c
@@ -33,10 +33,7 @@
#include "main/glheader.h"
#include "main/context.h"
#include "main/formats.h"
-#include "main/matrix.h"
-#include "main/state.h"
#include "main/simple_list.h"
-#include "main/extensions.h"
#include "main/framebuffer.h"
#include "main/renderbuffer.h"
diff --git a/src/mesa/drivers/dri/unichrome/via_ioctl.c b/src/mesa/drivers/dri/unichrome/via_ioctl.c
index 8d4edfa..c9a31f3 100644
--- a/src/mesa/drivers/dri/unichrome/via_ioctl.c
+++ b/src/mesa/drivers/dri/unichrome/via_ioctl.c
@@ -34,7 +34,6 @@
#include "via_context.h"
#include "via_tris.h"
#include "via_ioctl.h"
-#include "via_state.h"
#include "via_fb.h"
#include "via_3d_reg.h"
diff --git a/src/mesa/drivers/dri/unichrome/via_render.c b/src/mesa/drivers/dri/unichrome/via_render.c
index f676cc1..896c43d 100644
--- a/src/mesa/drivers/dri/unichrome/via_render.c
+++ b/src/mesa/drivers/dri/unichrome/via_render.c
@@ -37,7 +37,6 @@
#include "via_context.h"
#include "via_tris.h"
-#include "via_state.h"
#include "via_ioctl.h"
/*
diff --git a/src/mesa/drivers/dri/unichrome/via_screen.c b/src/mesa/drivers/dri/unichrome/via_screen.c
index 2cfb983..8c91c93 100644
--- a/src/mesa/drivers/dri/unichrome/via_screen.c
+++ b/src/mesa/drivers/dri/unichrome/via_screen.c
@@ -30,17 +30,13 @@
#include "main/context.h"
#include "main/framebuffer.h"
#include "main/renderbuffer.h"
-#include "main/matrix.h"
#include "main/simple_list.h"
#include "vblank.h"
#include "via_state.h"
#include "via_tex.h"
#include "via_span.h"
-#include "via_tris.h"
-#include "via_ioctl.h"
#include "via_screen.h"
-#include "via_fb.h"
#include "via_dri.h"
#include "GL/internal/dri_interface.h"
diff --git a/src/mesa/drivers/dri/unichrome/via_state.c b/src/mesa/drivers/dri/unichrome/via_state.c
index e6e5526..f7029b9 100644
--- a/src/mesa/drivers/dri/unichrome/via_state.c
+++ b/src/mesa/drivers/dri/unichrome/via_state.c
@@ -35,7 +35,6 @@
#include "via_context.h"
#include "via_state.h"
#include "via_tex.h"
-#include "via_tris.h"
#include "via_ioctl.h"
#include "via_3d_reg.h"
@@ -44,8 +43,6 @@
#include "tnl/tnl.h"
#include "swrast_setup/swrast_setup.h"
-#include "tnl/t_pipeline.h"
-
static GLuint ROP[16] = {
HC_HROP_BLACK, /* GL_CLEAR 0 */
diff --git a/src/mesa/drivers/dri/unichrome/via_tex.c b/src/mesa/drivers/dri/unichrome/via_tex.c
index 24924d2..917f975 100644
--- a/src/mesa/drivers/dri/unichrome/via_tex.c
+++ b/src/mesa/drivers/dri/unichrome/via_tex.c
@@ -37,14 +37,12 @@
#include "main/mipmap.h"
#include "main/mm.h"
#include "main/simple_list.h"
-#include "main/texcompress.h"
#include "main/texobj.h"
#include "main/texstore.h"
#include "via_context.h"
#include "via_fb.h"
#include "via_tex.h"
-#include "via_state.h"
#include "via_ioctl.h"
#include "via_3d_reg.h"
diff --git a/src/mesa/drivers/dri/unichrome/via_texcombine.c b/src/mesa/drivers/dri/unichrome/via_texcombine.c
index b646897..f87ba07 100644
--- a/src/mesa/drivers/dri/unichrome/via_texcombine.c
+++ b/src/mesa/drivers/dri/unichrome/via_texcombine.c
@@ -38,7 +38,6 @@
#include "main/enums.h"
#include "via_context.h"
-#include "via_state.h"
#include "via_tex.h"
#include "via_3d_reg.h"
diff --git a/src/mesa/main/attrib.c b/src/mesa/main/attrib.c
index 0641b98..3fbdba2 100644
--- a/src/mesa/main/attrib.c
+++ b/src/mesa/main/attrib.c
@@ -33,7 +33,6 @@
#include "bufferobj.h"
#include "clear.h"
#include "colormac.h"
-#include "colortab.h"
#include "context.h"
#include "depth.h"
#include "enable.h"
diff --git a/src/mesa/main/blend.c b/src/mesa/main/blend.c
index b8170dd..de60031 100644
--- a/src/mesa/main/blend.c
+++ b/src/mesa/main/blend.c
@@ -35,7 +35,6 @@
#include "enums.h"
#include "macros.h"
#include "mtypes.h"
-#include "glapi/glapitable.h"
/**
diff --git a/src/mesa/main/buffers.c b/src/mesa/main/buffers.c
index 97f0659..fb30b59 100644
--- a/src/mesa/main/buffers.c
+++ b/src/mesa/main/buffers.c
@@ -35,8 +35,6 @@
#include "colormac.h"
#include "context.h"
#include "enums.h"
-#include "fbobject.h"
-#include "state.h"
#define BAD_MASK ~0u
diff --git a/src/mesa/main/convolve.c b/src/mesa/main/convolve.c
index 8db3e79..5ed93e0 100644
--- a/src/mesa/main/convolve.c
+++ b/src/mesa/main/convolve.c
@@ -38,7 +38,6 @@
#include "context.h"
#include "image.h"
#include "mtypes.h"
-#include "pixel.h"
#include "state.h"
#include "glapi/dispatch.h"
diff --git a/src/mesa/main/debug.c b/src/mesa/main/debug.c
index a42113e..9bad834 100644
--- a/src/mesa/main/debug.c
+++ b/src/mesa/main/debug.c
@@ -26,7 +26,6 @@
#include "mtypes.h"
#include "attrib.h"
#include "colormac.h"
-#include "context.h"
#include "enums.h"
#include "formats.h"
#include "hash.h"
@@ -35,7 +34,6 @@
#include "get.h"
#include "pixelstore.h"
#include "readpix.h"
-#include "texgetimage.h"
#include "texobj.h"
@@ -54,7 +52,7 @@
"GL_QUAD_STRIP",
"GL_POLYGON",
"outside begin/end",
- "inside unkown primitive",
+ "inside unknown primitive",
"unknown state"
};
diff --git a/src/mesa/main/depthstencil.c b/src/mesa/main/depthstencil.c
index 193c7f8..49946a6 100644
--- a/src/mesa/main/depthstencil.c
+++ b/src/mesa/main/depthstencil.c
@@ -25,7 +25,6 @@
#include "glheader.h"
#include "imports.h"
#include "context.h"
-#include "fbobject.h"
#include "formats.h"
#include "mtypes.h"
#include "depthstencil.h"
diff --git a/src/mesa/main/dlist.c b/src/mesa/main/dlist.c
index 21a8216..7db686c 100644
--- a/src/mesa/main/dlist.c
+++ b/src/mesa/main/dlist.c
@@ -35,46 +35,25 @@
#include "api_loopback.h"
#include "config.h"
#include "mfeatures.h"
-#include "attrib.h"
-#include "blend.h"
-#include "buffers.h"
#if FEATURE_ARB_vertex_buffer_object
#include "bufferobj.h"
#endif
#include "arrayobj.h"
-#include "clip.h"
-#include "colortab.h"
#include "context.h"
-#include "convolve.h"
-#include "depth.h"
#include "dlist.h"
-#include "enable.h"
#include "enums.h"
#include "eval.h"
-#include "extensions.h"
-#include "feedback.h"
#include "framebuffer.h"
-#include "get.h"
#include "glapi/glapi.h"
#include "hash.h"
-#include "histogram.h"
#include "image.h"
#include "light.h"
-#include "lines.h"
#include "dlist.h"
#include "macros.h"
-#include "matrix.h"
-#include "pixel.h"
-#include "points.h"
-#include "polygon.h"
#include "queryobj.h"
-#include "state.h"
-#include "texobj.h"
#include "teximage.h"
-#include "texstate.h"
#include "mtypes.h"
#include "varray.h"
-#include "vtxfmt.h"
#if FEATURE_ARB_vertex_program || FEATURE_ARB_fragment_program
#include "shader/arbprogram.h"
#include "shader/program.h"
diff --git a/src/mesa/main/drawpix.c b/src/mesa/main/drawpix.c
index 5d4b53a..0afd47b 100644
--- a/src/mesa/main/drawpix.c
+++ b/src/mesa/main/drawpix.c
@@ -30,7 +30,6 @@
#include "enums.h"
#include "feedback.h"
#include "framebuffer.h"
-#include "image.h"
#include "readpix.h"
#include "state.h"
#include "glapi/dispatch.h"
diff --git a/src/mesa/main/enable.c b/src/mesa/main/enable.c
index cd6e881..f5c88a6 100644
--- a/src/mesa/main/enable.c
+++ b/src/mesa/main/enable.c
@@ -32,7 +32,6 @@
#include "context.h"
#include "enable.h"
#include "light.h"
-#include "macros.h"
#include "simple_list.h"
#include "mtypes.h"
#include "enums.h"
diff --git a/src/mesa/main/fbobject.c b/src/mesa/main/fbobject.c
index 4da245a..0556f16 100644
--- a/src/mesa/main/fbobject.c
+++ b/src/mesa/main/fbobject.c
@@ -40,13 +40,10 @@
#include "framebuffer.h"
#include "hash.h"
#include "macros.h"
-#include "mipmap.h"
#include "renderbuffer.h"
#include "state.h"
#include "teximage.h"
#include "texobj.h"
-#include "texstore.h"
-#include "texstate.h"
/** Set this to 1 to help debug FBO incompleteness problems */
diff --git a/src/mesa/main/formats.c b/src/mesa/main/formats.c
index 5983f00..d0c9c00 100644
--- a/src/mesa/main/formats.c
+++ b/src/mesa/main/formats.c
@@ -27,7 +27,6 @@
#include "imports.h"
#include "formats.h"
#include "config.h"
-#include "texstore.h"
/**
diff --git a/src/mesa/main/image.c b/src/mesa/main/image.c
index fc278bb..81993e7 100644
--- a/src/mesa/main/image.c
+++ b/src/mesa/main/image.c
@@ -37,7 +37,6 @@
#include "image.h"
#include "imports.h"
#include "macros.h"
-#include "pixel.h"
/**
diff --git a/src/mesa/main/lines.c b/src/mesa/main/lines.c
index 81d0d33..cc63a75 100644
--- a/src/mesa/main/lines.c
+++ b/src/mesa/main/lines.c
@@ -25,10 +25,8 @@
#include "glheader.h"
#include "context.h"
-#include "depth.h"
#include "lines.h"
#include "macros.h"
-#include "texstate.h"
#include "mtypes.h"
diff --git a/src/mesa/main/mipmap.c b/src/mesa/main/mipmap.c
index 7350c7a..77cd1d4 100644
--- a/src/mesa/main/mipmap.c
+++ b/src/mesa/main/mipmap.c
@@ -30,7 +30,6 @@
#include "imports.h"
#include "formats.h"
#include "mipmap.h"
-#include "texcompress.h"
#include "teximage.h"
#include "texstore.h"
#include "image.h"
diff --git a/src/mesa/main/pixel.c b/src/mesa/main/pixel.c
index 3820ebd..f6f9c11 100644
--- a/src/mesa/main/pixel.c
+++ b/src/mesa/main/pixel.c
@@ -32,7 +32,6 @@
#include "bufferobj.h"
#include "colormac.h"
#include "context.h"
-#include "image.h"
#include "macros.h"
#include "pixel.h"
#include "mtypes.h"
diff --git a/src/mesa/main/pixelstore.c b/src/mesa/main/pixelstore.c
index 6a641f8..ec585ef 100644
--- a/src/mesa/main/pixelstore.c
+++ b/src/mesa/main/pixelstore.c
@@ -30,10 +30,7 @@
#include "glheader.h"
#include "bufferobj.h"
-#include "colormac.h"
#include "context.h"
-#include "image.h"
-#include "macros.h"
#include "pixelstore.h"
#include "mtypes.h"
diff --git a/src/mesa/main/points.c b/src/mesa/main/points.c
index dcaeccd..e743a2e 100644
--- a/src/mesa/main/points.c
+++ b/src/mesa/main/points.c
@@ -32,7 +32,6 @@
#include "context.h"
#include "macros.h"
#include "points.h"
-#include "texstate.h"
#include "mtypes.h"
diff --git a/src/mesa/main/polygon.c b/src/mesa/main/polygon.c
index 376a87a..dcde675 100644
--- a/src/mesa/main/polygon.c
+++ b/src/mesa/main/polygon.c
@@ -34,7 +34,6 @@
#include "context.h"
#include "image.h"
#include "enums.h"
-#include "macros.h"
#include "polygon.h"
#include "mtypes.h"
diff --git a/src/mesa/main/state.c b/src/mesa/main/state.c
index f10e6b0..904ad01 100644
--- a/src/mesa/main/state.c
+++ b/src/mesa/main/state.c
@@ -48,7 +48,6 @@
#include "texenvprogram.h"
#include "texobj.h"
#include "texstate.h"
-#include "viewport.h"
static void
diff --git a/src/mesa/main/texcompress.c b/src/mesa/main/texcompress.c
index a4f1926..cff6de8 100644
--- a/src/mesa/main/texcompress.c
+++ b/src/mesa/main/texcompress.c
@@ -35,10 +35,7 @@
#include "colormac.h"
#include "context.h"
#include "formats.h"
-#include "image.h"
-#include "mipmap.h"
#include "texcompress.h"
-#include "texstore.h"
/**
diff --git a/src/mesa/main/texformat.c b/src/mesa/main/texformat.c
index 1a374e7..096945a 100644
--- a/src/mesa/main/texformat.c
+++ b/src/mesa/main/texformat.c
@@ -35,8 +35,6 @@
#include "context.h"
#include "texcompress.h"
-#include "texcompress_fxt1.h"
-#include "texcompress_s3tc.h"
#include "texformat.h"
diff --git a/src/mesa/main/texgetimage.c b/src/mesa/main/texgetimage.c
index d7943ea..66d01c1 100644
--- a/src/mesa/main/texgetimage.c
+++ b/src/mesa/main/texgetimage.c
@@ -35,10 +35,8 @@
#include "context.h"
#include "formats.h"
#include "image.h"
-#include "texcompress.h"
#include "texgetimage.h"
#include "teximage.h"
-#include "texstate.h"
diff --git a/src/mesa/main/teximage.c b/src/mesa/main/teximage.c
index b946f3c..76273b0 100644
--- a/src/mesa/main/teximage.c
+++ b/src/mesa/main/teximage.c
@@ -46,7 +46,6 @@
#include "texfetch.h"
#include "teximage.h"
#include "texstate.h"
-#include "texstore.h"
#include "mtypes.h"
diff --git a/src/mesa/main/texobj.c b/src/mesa/main/texobj.c
index 7f0a246..9db9581 100644
--- a/src/mesa/main/texobj.c
+++ b/src/mesa/main/texobj.c
@@ -38,7 +38,6 @@
#include "imports.h"
#include "macros.h"
#include "teximage.h"
-#include "texstate.h"
#include "texobj.h"
#include "mtypes.h"
#include "shader/prog_instruction.h"
diff --git a/src/mesa/main/texparam.c b/src/mesa/main/texparam.c
index d917e21..c4f2495 100644
--- a/src/mesa/main/texparam.c
+++ b/src/mesa/main/texparam.c
@@ -33,7 +33,6 @@
#include "main/glheader.h"
#include "main/colormac.h"
#include "main/context.h"
-#include "main/enums.h"
#include "main/formats.h"
#include "main/macros.h"
#include "main/texcompress.h"
diff --git a/src/mesa/main/texstate.c b/src/mesa/main/texstate.c
index c735e18..25333d8 100644
--- a/src/mesa/main/texstate.c
+++ b/src/mesa/main/texstate.c
@@ -35,11 +35,9 @@
#include "context.h"
#include "enums.h"
#include "macros.h"
-#include "texcompress.h"
#include "texobj.h"
#include "teximage.h"
#include "texstate.h"
-#include "texenvprogram.h"
#include "mtypes.h"
diff --git a/src/mesa/main/texstore.c b/src/mesa/main/texstore.c
index 792c831..fcd0a56 100644
--- a/src/mesa/main/texstore.c
+++ b/src/mesa/main/texstore.c
@@ -263,7 +263,7 @@
map[ZERO] = ZERO;
map[ONE] = ONE;
-/*
+#if 0
_mesa_printf("from %x/%s to %x/%s map %d %d %d %d %d %d\n",
inFormat, _mesa_lookup_enum_by_nr(inFormat),
outFormat, _mesa_lookup_enum_by_nr(outFormat),
@@ -273,7 +273,7 @@
map[3],
map[4],
map[5]);
-*/
+#endif
}
diff --git a/src/mesa/main/vtxfmt.c b/src/mesa/main/vtxfmt.c
index c9eea8a..0dd3e5e 100644
--- a/src/mesa/main/vtxfmt.c
+++ b/src/mesa/main/vtxfmt.c
@@ -28,11 +28,9 @@
#include "glheader.h"
#include "api_arrayelt.h"
-#include "api_loopback.h"
#include "context.h"
#include "imports.h"
#include "mtypes.h"
-#include "state.h"
#include "vtxfmt.h"
#include "eval.h"
#include "dlist.h"
diff --git a/src/mesa/shader/arbprogparse.c b/src/mesa/shader/arbprogparse.c
index 3e22a8a..bdd26b7 100644
--- a/src/mesa/shader/arbprogparse.c
+++ b/src/mesa/shader/arbprogparse.c
@@ -54,10 +54,8 @@
#include "main/glheader.h"
#include "main/imports.h"
#include "main/context.h"
-#include "main/macros.h"
#include "main/mtypes.h"
#include "arbprogparse.h"
-#include "program.h"
#include "programopt.h"
#include "prog_parameter.h"
#include "prog_statevars.h"
diff --git a/src/mesa/shader/nvvertparse.c b/src/mesa/shader/nvvertparse.c
index 8574016..baff765 100644
--- a/src/mesa/shader/nvvertparse.c
+++ b/src/mesa/shader/nvvertparse.c
@@ -40,7 +40,6 @@
#include "main/glheader.h"
#include "main/context.h"
#include "main/imports.h"
-#include "main/macros.h"
#include "nvprogram.h"
#include "nvvertparse.h"
#include "prog_instruction.h"
diff --git a/src/mesa/shader/prog_execute.c b/src/mesa/shader/prog_execute.c
index 7f03452..7ea792f 100644
--- a/src/mesa/shader/prog_execute.c
+++ b/src/mesa/shader/prog_execute.c
@@ -38,7 +38,6 @@
#include "main/glheader.h"
#include "main/colormac.h"
#include "main/context.h"
-#include "program.h"
#include "prog_execute.h"
#include "prog_instruction.h"
#include "prog_parameter.h"
diff --git a/src/mesa/shader/prog_statevars.c b/src/mesa/shader/prog_statevars.c
index 058d4bb..3a446fd 100644
--- a/src/mesa/shader/prog_statevars.c
+++ b/src/mesa/shader/prog_statevars.c
@@ -31,7 +31,6 @@
#include "main/glheader.h"
#include "main/context.h"
-#include "main/hash.h"
#include "main/imports.h"
#include "main/macros.h"
#include "main/mtypes.h"
diff --git a/src/mesa/shader/shader_api.c b/src/mesa/shader/shader_api.c
index 453cd39..e522d70 100644
--- a/src/mesa/shader/shader_api.c
+++ b/src/mesa/shader/shader_api.c
@@ -39,10 +39,8 @@
#include "main/glheader.h"
#include "main/context.h"
#include "main/hash.h"
-#include "main/macros.h"
#include "shader/program.h"
#include "shader/prog_parameter.h"
-#include "shader/prog_print.h"
#include "shader/prog_statevars.h"
#include "shader/prog_uniform.h"
#include "shader/shader_api.h"
diff --git a/src/mesa/shader/slang/slang_builtin.c b/src/mesa/shader/slang/slang_builtin.c
index e580950..0a9f0b9 100644
--- a/src/mesa/shader/slang/slang_builtin.c
+++ b/src/mesa/shader/slang/slang_builtin.c
@@ -36,7 +36,6 @@
#include "shader/prog_parameter.h"
#include "shader/prog_statevars.h"
#include "shader/slang/slang_ir.h"
-#include "shader/slang/slang_emit.h"
#include "shader/slang/slang_builtin.h"
diff --git a/src/mesa/shader/slang/slang_compile.c b/src/mesa/shader/slang/slang_compile.c
index 6499cfc..63d10f4 100644
--- a/src/mesa/shader/slang/slang_compile.c
+++ b/src/mesa/shader/slang/slang_compile.c
@@ -40,14 +40,11 @@
#include "slang_codegen.h"
#include "slang_compile.h"
#include "slang_storage.h"
-#include "slang_emit.h"
#include "slang_log.h"
#include "slang_mem.h"
#include "slang_vartable.h"
#include "slang_simplify.h"
-#include "slang_print.h"
-
/*
* This is a straightforward implementation of the slang front-end
* compiler. Lots of error-checking functionality is missing but
diff --git a/src/mesa/shader/slang/slang_emit.c b/src/mesa/shader/slang/slang_emit.c
index ce3f6ab..c9ecbd2 100644
--- a/src/mesa/shader/slang/slang_emit.c
+++ b/src/mesa/shader/slang/slang_emit.c
@@ -38,7 +38,6 @@
#include "main/imports.h"
#include "main/context.h"
-#include "main/macros.h"
#include "shader/program.h"
#include "shader/prog_instruction.h"
#include "shader/prog_parameter.h"
diff --git a/src/mesa/shader/slang/slang_link.c b/src/mesa/shader/slang/slang_link.c
index ed27821..21497b3 100644
--- a/src/mesa/shader/slang/slang_link.c
+++ b/src/mesa/shader/slang/slang_link.c
@@ -31,7 +31,6 @@
#include "main/imports.h"
#include "main/context.h"
-#include "main/hash.h"
#include "main/macros.h"
#include "shader/program.h"
#include "shader/prog_instruction.h"
diff --git a/src/mesa/shader/slang/slang_log.c b/src/mesa/shader/slang/slang_log.c
index d7d2b4f..4f6b854 100644
--- a/src/mesa/shader/slang/slang_log.c
+++ b/src/mesa/shader/slang/slang_log.c
@@ -24,7 +24,6 @@
*/
#include "main/imports.h"
-#include "main/context.h"
#include "slang_log.h"
#include "slang_utility.h"
diff --git a/src/mesa/state_tracker/st_atom_framebuffer.c b/src/mesa/state_tracker/st_atom_framebuffer.c
index 8ca4335..8d045f2 100644
--- a/src/mesa/state_tracker/st_atom_framebuffer.c
+++ b/src/mesa/state_tracker/st_atom_framebuffer.c
@@ -37,7 +37,6 @@
#include "st_public.h"
#include "st_texture.h"
#include "pipe/p_context.h"
-#include "pipe/p_inlines.h"
#include "cso_cache/cso_context.h"
#include "util/u_rect.h"
#include "util/u_math.h"
diff --git a/src/mesa/state_tracker/st_atom_pixeltransfer.c b/src/mesa/state_tracker/st_atom_pixeltransfer.c
index 6a5854e..0b2e3f5 100644
--- a/src/mesa/state_tracker/st_atom_pixeltransfer.c
+++ b/src/mesa/state_tracker/st_atom_pixeltransfer.c
@@ -43,7 +43,6 @@
#include "st_context.h"
#include "st_format.h"
-#include "st_program.h"
#include "st_texture.h"
#include "st_inlines.h"
diff --git a/src/mesa/state_tracker/st_atom_sampler.c b/src/mesa/state_tracker/st_atom_sampler.c
index 7b84a86..9d63f1c 100644
--- a/src/mesa/state_tracker/st_atom_sampler.c
+++ b/src/mesa/state_tracker/st_atom_sampler.c
@@ -37,7 +37,6 @@
#include "st_context.h"
#include "st_cb_texture.h"
#include "st_atom.h"
-#include "st_program.h"
#include "pipe/p_context.h"
#include "pipe/p_defines.h"
diff --git a/src/mesa/state_tracker/st_atom_shader.c b/src/mesa/state_tracker/st_atom_shader.c
index 176f3ea..181f88a 100644
--- a/src/mesa/state_tracker/st_atom_shader.c
+++ b/src/mesa/state_tracker/st_atom_shader.c
@@ -51,7 +51,6 @@
#include "st_atom.h"
#include "st_program.h"
#include "st_atom_shader.h"
-#include "st_mesa_to_tgsi.h"
@@ -79,7 +78,7 @@
stfp->num_input_slots = numIn;
- assert(stfp->Base.Base.NumInstructions > 1);
+ assert(stfp->Base.Base.NumInstructions > 0);
st_translate_fragment_program(st, stfp, stfp->input_to_slot);
}
diff --git a/src/mesa/state_tracker/st_cb_accum.c b/src/mesa/state_tracker/st_cb_accum.c
index da7b97d..798081e 100644
--- a/src/mesa/state_tracker/st_cb_accum.c
+++ b/src/mesa/state_tracker/st_cb_accum.c
@@ -38,9 +38,7 @@
#include "st_context.h"
#include "st_cb_accum.h"
#include "st_cb_fbo.h"
-#include "st_draw.h"
#include "st_public.h"
-#include "st_format.h"
#include "st_texture.h"
#include "st_inlines.h"
#include "pipe/p_context.h"
diff --git a/src/mesa/state_tracker/st_cb_bitmap.c b/src/mesa/state_tracker/st_cb_bitmap.c
index 1bdeacc..d1b35f7 100644
--- a/src/mesa/state_tracker/st_cb_bitmap.c
+++ b/src/mesa/state_tracker/st_cb_bitmap.c
@@ -34,9 +34,7 @@
#include "main/image.h"
#include "main/bufferobj.h"
#include "main/macros.h"
-#include "main/texformat.h"
#include "shader/program.h"
-#include "shader/prog_parameter.h"
#include "shader/prog_print.h"
#include "st_context.h"
@@ -44,15 +42,12 @@
#include "st_atom_constbuf.h"
#include "st_program.h"
#include "st_cb_bitmap.h"
-#include "st_cb_program.h"
-#include "st_mesa_to_tgsi.h"
#include "st_texture.h"
#include "st_inlines.h"
#include "pipe/p_context.h"
#include "pipe/p_defines.h"
#include "pipe/p_inlines.h"
-#include "util/u_tile.h"
#include "util/u_draw_quad.h"
#include "util/u_simple_shaders.h"
#include "shader/prog_instruction.h"
@@ -386,11 +381,11 @@
}
/* put vertex data into vbuf */
- st_no_flush_pipe_buffer_write(st,
- st->bitmap.vbuf,
- st->bitmap.vbuf_slot * sizeof st->bitmap.vertices,
- sizeof st->bitmap.vertices,
- st->bitmap.vertices);
+ st_no_flush_pipe_buffer_write_nooverlap(st,
+ st->bitmap.vbuf,
+ st->bitmap.vbuf_slot * sizeof st->bitmap.vertices,
+ sizeof st->bitmap.vertices,
+ st->bitmap.vertices);
return st->bitmap.vbuf_slot++ * sizeof st->bitmap.vertices;
}
diff --git a/src/mesa/state_tracker/st_cb_blit.c b/src/mesa/state_tracker/st_cb_blit.c
index 563615e..65aa2a2 100644
--- a/src/mesa/state_tracker/st_cb_blit.c
+++ b/src/mesa/state_tracker/st_cb_blit.c
@@ -33,14 +33,10 @@
#include "main/imports.h"
#include "main/image.h"
#include "main/macros.h"
-#include "main/texformat.h"
#include "shader/program.h"
-#include "shader/prog_parameter.h"
-#include "shader/prog_print.h"
#include "st_context.h"
#include "st_texture.h"
-#include "st_program.h"
#include "st_cb_blit.h"
#include "st_cb_fbo.h"
diff --git a/src/mesa/state_tracker/st_cb_clear.c b/src/mesa/state_tracker/st_cb_clear.c
index 192d765..398587a 100644
--- a/src/mesa/state_tracker/st_cb_clear.c
+++ b/src/mesa/state_tracker/st_cb_clear.c
@@ -42,10 +42,8 @@
#include "st_cb_accum.h"
#include "st_cb_clear.h"
#include "st_cb_fbo.h"
-#include "st_draw.h"
#include "st_program.h"
#include "st_public.h"
-#include "st_mesa_to_tgsi.h"
#include "st_inlines.h"
#include "pipe/p_context.h"
@@ -53,7 +51,6 @@
#include "pipe/p_state.h"
#include "pipe/p_defines.h"
#include "util/u_format.h"
-#include "util/u_pack_color.h"
#include "util/u_simple_shaders.h"
#include "util/u_draw_quad.h"
@@ -166,10 +163,10 @@
}
/* put vertex data into vbuf */
- st_no_flush_pipe_buffer_write(st, st->clear.vbuf,
- st->clear.vbuf_slot * sizeof(st->clear.vertices),
- sizeof(st->clear.vertices),
- st->clear.vertices);
+ st_no_flush_pipe_buffer_write_nooverlap(st, st->clear.vbuf,
+ st->clear.vbuf_slot * sizeof(st->clear.vertices),
+ sizeof(st->clear.vertices),
+ st->clear.vertices);
/* draw */
util_draw_vertex_buffer(pipe,
diff --git a/src/mesa/state_tracker/st_cb_drawpixels.c b/src/mesa/state_tracker/st_cb_drawpixels.c
index 7c66426..4cec259 100644
--- a/src/mesa/state_tracker/st_cb_drawpixels.c
+++ b/src/mesa/state_tracker/st_cb_drawpixels.c
@@ -36,24 +36,18 @@
#include "main/macros.h"
#include "main/texformat.h"
#include "main/texstore.h"
-#include "main/state.h"
#include "shader/program.h"
-#include "shader/prog_parameter.h"
#include "shader/prog_print.h"
#include "st_debug.h"
#include "st_context.h"
#include "st_atom.h"
#include "st_atom_constbuf.h"
-#include "st_draw.h"
#include "st_program.h"
#include "st_cb_drawpixels.h"
#include "st_cb_readpixels.h"
#include "st_cb_fbo.h"
-#include "st_cb_texture.h"
-#include "st_draw.h"
#include "st_format.h"
-#include "st_mesa_to_tgsi.h"
#include "st_texture.h"
#include "st_inlines.h"
diff --git a/src/mesa/state_tracker/st_cb_fbo.c b/src/mesa/state_tracker/st_cb_fbo.c
index 45ce34a..f7350ef 100644
--- a/src/mesa/state_tracker/st_cb_fbo.c
+++ b/src/mesa/state_tracker/st_cb_fbo.c
@@ -44,7 +44,6 @@
#include "pipe/p_screen.h"
#include "st_context.h"
#include "st_cb_fbo.h"
-#include "st_cb_texture.h"
#include "st_format.h"
#include "st_public.h"
#include "st_texture.h"
diff --git a/src/mesa/state_tracker/st_cb_feedback.c b/src/mesa/state_tracker/st_cb_feedback.c
index 93f7145..17261f8 100644
--- a/src/mesa/state_tracker/st_cb_feedback.c
+++ b/src/mesa/state_tracker/st_cb_feedback.c
@@ -45,14 +45,11 @@
#include "vbo/vbo.h"
#include "st_context.h"
-#include "st_atom.h"
#include "st_draw.h"
#include "st_cb_feedback.h"
-#include "st_cb_bufferobjects.h"
#include "pipe/p_context.h"
#include "pipe/p_defines.h"
-#include "cso_cache/cso_cache.h"
#include "draw/draw_context.h"
#include "draw/draw_pipe.h"
diff --git a/src/mesa/state_tracker/st_cb_program.c b/src/mesa/state_tracker/st_cb_program.c
index 8c276f8..5138e59 100644
--- a/src/mesa/state_tracker/st_cb_program.c
+++ b/src/mesa/state_tracker/st_cb_program.c
@@ -36,7 +36,6 @@
#include "shader/prog_instruction.h"
#include "shader/prog_parameter.h"
#include "shader/program.h"
-#include "shader/programopt.h"
#include "shader/shader_api.h"
#include "cso_cache/cso_context.h"
diff --git a/src/mesa/state_tracker/st_cb_queryobj.c b/src/mesa/state_tracker/st_cb_queryobj.c
index 10629e9..2281d10 100644
--- a/src/mesa/state_tracker/st_cb_queryobj.c
+++ b/src/mesa/state_tracker/st_cb_queryobj.c
@@ -41,7 +41,6 @@
#include "pipe/p_defines.h"
#include "st_context.h"
#include "st_cb_queryobj.h"
-#include "st_public.h"
static struct gl_query_object *
diff --git a/src/mesa/state_tracker/st_cb_rasterpos.c b/src/mesa/state_tracker/st_cb_rasterpos.c
index d82b2a2..42a1377 100644
--- a/src/mesa/state_tracker/st_cb_rasterpos.c
+++ b/src/mesa/state_tracker/st_cb_rasterpos.c
@@ -47,7 +47,6 @@
#include "st_draw.h"
#include "draw/draw_context.h"
#include "draw/draw_pipe.h"
-#include "shader/prog_instruction.h"
#include "vbo/vbo.h"
diff --git a/src/mesa/state_tracker/st_cb_readpixels.c b/src/mesa/state_tracker/st_cb_readpixels.c
index 6fa7bb6..8eb825a 100644
--- a/src/mesa/state_tracker/st_cb_readpixels.c
+++ b/src/mesa/state_tracker/st_cb_readpixels.c
@@ -45,10 +45,8 @@
#include "st_debug.h"
#include "st_context.h"
-#include "st_cb_bitmap.h"
#include "st_cb_readpixels.h"
#include "st_cb_fbo.h"
-#include "st_format.h"
#include "st_public.h"
#include "st_texture.h"
#include "st_inlines.h"
diff --git a/src/mesa/state_tracker/st_cb_strings.c b/src/mesa/state_tracker/st_cb_strings.c
index bb931f1..f22c536 100644
--- a/src/mesa/state_tracker/st_cb_strings.c
+++ b/src/mesa/state_tracker/st_cb_strings.c
@@ -33,7 +33,6 @@
#include "main/glheader.h"
#include "main/macros.h"
-#include "main/version.h"
#include "pipe/p_context.h"
#include "pipe/p_screen.h"
#include "st_context.h"
diff --git a/src/mesa/state_tracker/st_cb_texture.c b/src/mesa/state_tracker/st_cb_texture.c
index f01053c..b94bfc8 100644
--- a/src/mesa/state_tracker/st_cb_texture.c
+++ b/src/mesa/state_tracker/st_cb_texture.c
@@ -31,15 +31,14 @@
#include "main/convolve.h"
#endif
#include "main/enums.h"
+#include "main/fbobject.h"
#include "main/formats.h"
#include "main/image.h"
#include "main/imports.h"
#include "main/macros.h"
#include "main/mipmap.h"
-#include "main/pixel.h"
#include "main/texcompress.h"
#include "main/texfetch.h"
-#include "main/texformat.h"
#include "main/texgetimage.h"
#include "main/teximage.h"
#include "main/texobj.h"
@@ -1368,33 +1367,64 @@
}
+
+/**
+ * If the format of the src renderbuffer and the format of the dest
+ * texture are compatible (in terms of blitting), return a TGSI writemask
+ * to be used during the blit.
+ * If the src/dest are incompatible, return 0.
+ */
static unsigned
-compatible_src_dst_formats(const struct gl_renderbuffer *src,
+compatible_src_dst_formats(GLcontext *ctx,
+ const struct gl_renderbuffer *src,
const struct gl_texture_image *dst)
{
- const GLenum srcFormat = _mesa_get_format_base_format(src->Format);
- const GLenum dstLogicalFormat = _mesa_get_format_base_format(dst->TexFormat);
+ /* Get logical base formats for the src and dest.
+ * That is, use the user-requested formats and not the actual, device-
+ * chosen formats.
+ * For example, the user may have requested an A8 texture but the
+ * driver may actually be using an RGBA texture format. When we
+ * copy/blit to that texture, we only want to copy the Alpha channel
+ * and not the RGB channels.
+ *
+ * Similarly, when the src FBO was created an RGB format may have been
+ * requested but the driver actually chose an RGBA format. In that case,
+ * we don't want to copy the undefined Alpha channel to the dest texture
+ * (it should be 1.0).
+ */
+ const GLenum srcFormat = _mesa_base_fbo_format(ctx, src->InternalFormat);
+ const GLenum dstFormat = _mesa_base_tex_format(ctx, dst->InternalFormat);
- if (srcFormat == dstLogicalFormat) {
+ /**
+ * XXX when we have red-only and red/green renderbuffers we'll need
+ * to add more cases here (or implement a general-purpose routine that
+ * queries the existance of the R,G,B,A channels in the src and dest).
+ */
+ if (srcFormat == dstFormat) {
/* This is the same as matching_base_formats, which should
* always pass, as it did previously.
*/
return TGSI_WRITEMASK_XYZW;
}
- else if (srcFormat == GL_RGBA &&
- dstLogicalFormat == GL_RGB) {
- /* Add a single special case to cope with RGBA->RGB transfers,
- * setting A to 1.0 to cope with situations where the RGB
- * destination is actually stored as RGBA.
+ else if (srcFormat == GL_RGB && dstFormat == GL_RGBA) {
+ /* Make sure that A in the dest is 1. The actual src format
+ * may be RGBA and have undefined A values.
*/
- return TGSI_WRITEMASK_XYZ; /* A ==> 1.0 */
+ return TGSI_WRITEMASK_XYZ;
+ }
+ else if (srcFormat == GL_RGBA && dstFormat == GL_RGB) {
+ /* Make sure that A in the dest is 1. The actual dst format
+ * may be RGBA and will need A=1 to provide proper alpha values
+ * when sampled later.
+ */
+ return TGSI_WRITEMASK_XYZ;
}
else {
if (ST_DEBUG & DEBUG_FALLBACK)
debug_printf("%s failed for src %s, dst %s\n",
__FUNCTION__,
_mesa_lookup_enum_by_nr(srcFormat),
- _mesa_lookup_enum_by_nr(dstLogicalFormat));
+ _mesa_lookup_enum_by_nr(dstFormat));
/* Otherwise fail.
*/
@@ -1505,7 +1535,7 @@
matching_base_formats =
(_mesa_get_format_base_format(strb->Base.Format) ==
_mesa_get_format_base_format(texImage->TexFormat));
- format_writemask = compatible_src_dst_formats(&strb->Base, texImage);
+ format_writemask = compatible_src_dst_formats(ctx, &strb->Base, texImage);
if (ctx->_ImageTransferState == 0x0) {
diff --git a/src/mesa/state_tracker/st_cb_viewport.c b/src/mesa/state_tracker/st_cb_viewport.c
index ab11c5b..b29191a 100644
--- a/src/mesa/state_tracker/st_cb_viewport.c
+++ b/src/mesa/state_tracker/st_cb_viewport.c
@@ -27,14 +27,11 @@
#include "main/glheader.h"
#include "st_context.h"
-#include "st_public.h"
#include "st_cb_viewport.h"
#include "pipe/p_context.h"
-#include "pipe/p_inlines.h"
#include "pipe/p_state.h"
#include "pipe/p_defines.h"
-#include "pipe/internal/p_winsys_screen.h"
static void st_viewport(GLcontext * ctx, GLint x, GLint y,
diff --git a/src/mesa/state_tracker/st_context.c b/src/mesa/state_tracker/st_context.c
index 9e6ce30..a62ff24 100644
--- a/src/mesa/state_tracker/st_context.c
+++ b/src/mesa/state_tracker/st_context.c
@@ -27,11 +27,6 @@
#include "main/imports.h"
#include "main/context.h"
-#include "main/extensions.h"
-#include "main/matrix.h"
-#include "main/buffers.h"
-#include "main/scissor.h"
-#include "main/viewport.h"
#include "vbo/vbo.h"
#include "shader/shader_api.h"
#include "glapi/glapi.h"
@@ -69,7 +64,6 @@
#include "st_program.h"
#include "pipe/p_context.h"
#include "draw/draw_context.h"
-#include "cso_cache/cso_cache.h"
#include "cso_cache/cso_context.h"
diff --git a/src/mesa/state_tracker/st_draw_feedback.c b/src/mesa/state_tracker/st_draw_feedback.c
index a05d6dd..dd27a1e 100644
--- a/src/mesa/state_tracker/st_draw_feedback.c
+++ b/src/mesa/state_tracker/st_draw_feedback.c
@@ -28,7 +28,6 @@
#include "main/imports.h"
#include "main/image.h"
#include "main/macros.h"
-#include "shader/prog_uniform.h"
#include "vbo/vbo.h"
diff --git a/src/mesa/state_tracker/st_extensions.c b/src/mesa/state_tracker/st_extensions.c
index 2a5fb27..89a16c1 100644
--- a/src/mesa/state_tracker/st_extensions.c
+++ b/src/mesa/state_tracker/st_extensions.c
@@ -28,7 +28,6 @@
#include "main/imports.h"
#include "main/context.h"
-#include "main/extensions.h"
#include "main/macros.h"
#include "pipe/p_context.h"
diff --git a/src/mesa/state_tracker/st_format.c b/src/mesa/state_tracker/st_format.c
index d00b67a..3ffc2ae 100644
--- a/src/mesa/state_tracker/st_format.c
+++ b/src/mesa/state_tracker/st_format.c
@@ -35,7 +35,6 @@
#include "main/imports.h"
#include "main/context.h"
#include "main/texstore.h"
-#include "main/texformat.h"
#include "main/enums.h"
#include "main/macros.h"
@@ -288,6 +287,8 @@
return MESA_FORMAT_XRGB8888;
case PIPE_FORMAT_B8G8R8A8_UNORM:
return MESA_FORMAT_ARGB8888_REV;
+ case PIPE_FORMAT_B8G8R8X8_UNORM:
+ return MESA_FORMAT_XRGB8888_REV;
case PIPE_FORMAT_A1R5G5B5_UNORM:
return MESA_FORMAT_ARGB1555;
case PIPE_FORMAT_A4R4G4B4_UNORM:
diff --git a/src/mesa/state_tracker/st_framebuffer.c b/src/mesa/state_tracker/st_framebuffer.c
index a5d1ae3..835142e 100644
--- a/src/mesa/state_tracker/st_framebuffer.c
+++ b/src/mesa/state_tracker/st_framebuffer.c
@@ -30,15 +30,11 @@
#include "main/buffers.h"
#include "main/context.h"
#include "main/framebuffer.h"
-#include "main/matrix.h"
#include "main/renderbuffer.h"
-#include "main/scissor.h"
-#include "main/viewport.h"
#include "st_context.h"
#include "st_cb_fbo.h"
#include "st_public.h"
#include "pipe/p_defines.h"
-#include "pipe/p_context.h"
struct st_framebuffer *
diff --git a/src/mesa/state_tracker/st_gen_mipmap.c b/src/mesa/state_tracker/st_gen_mipmap.c
index 2c283d4..3823a59 100644
--- a/src/mesa/state_tracker/st_gen_mipmap.c
+++ b/src/mesa/state_tracker/st_gen_mipmap.c
@@ -46,9 +46,7 @@
#include "st_debug.h"
#include "st_context.h"
-#include "st_draw.h"
#include "st_gen_mipmap.h"
-#include "st_program.h"
#include "st_texture.h"
#include "st_cb_texture.h"
#include "st_inlines.h"
diff --git a/src/mesa/state_tracker/st_inlines.h b/src/mesa/state_tracker/st_inlines.h
index a41cfeb..dccc46f 100644
--- a/src/mesa/state_tracker/st_inlines.h
+++ b/src/mesa/state_tracker/st_inlines.h
@@ -126,6 +126,16 @@
}
static INLINE void
+st_no_flush_pipe_buffer_write_nooverlap(struct st_context *st,
+ struct pipe_buffer *buf,
+ unsigned int offset,
+ unsigned int size,
+ const void * data)
+{
+ pipe_buffer_write_nooverlap(st->pipe->screen, buf, offset, size, data);
+}
+
+static INLINE void
st_cond_flush_pipe_buffer_read(struct st_context *st,
struct pipe_buffer *buf,
unsigned int offset,
diff --git a/src/mesa/swrast/s_depth.c b/src/mesa/swrast/s_depth.c
index 0b6bb7e..ac5dae2 100644
--- a/src/mesa/swrast/s_depth.c
+++ b/src/mesa/swrast/s_depth.c
@@ -497,17 +497,24 @@
return passed;
}
-/* Apply ARB_depth_clamp to span of fragments. */
+
+
+/**
+ * Clamp fragment Z values to the depth near/far range (glDepthRange()).
+ * This is used when GL_ARB_depth_clamp/GL_DEPTH_CLAMP is turned on.
+ * In that case, vertexes are not clipped against the near/far planes
+ * so rasterization will produce fragment Z values outside the usual
+ * [0,1] range.
+ */
void
_swrast_depth_clamp_span( GLcontext *ctx, SWspan *span )
{
struct gl_framebuffer *fb = ctx->DrawBuffer;
- struct gl_renderbuffer *rb = fb->_DepthBuffer;
const GLuint count = span->end;
- GLuint *zValues = span->array->z;
- GLuint min, max;
+ GLint *zValues = (GLint *) span->array->z; /* sign change */
+ GLint min, max;
GLfloat min_f, max_f;
- int i;
+ GLuint i;
if (ctx->Viewport.Near < ctx->Viewport.Far) {
min_f = ctx->Viewport.Near;
@@ -517,15 +524,21 @@
max_f = ctx->Viewport.Near;
}
- if (rb->DataType == GL_UNSIGNED_SHORT) {
- CLAMPED_FLOAT_TO_USHORT(min, min_f);
- CLAMPED_FLOAT_TO_USHORT(max, max_f);
- } else {
- assert(rb->DataType == GL_UNSIGNED_INT);
- min = FLOAT_TO_UINT(min_f);
- max = FLOAT_TO_UINT(max_f);
- }
+ /* Convert floating point values in [0,1] to device Z coordinates in
+ * [0, DepthMax].
+ * ex: If the the Z buffer has 24 bits, DepthMax = 0xffffff.
+ *
+ * XXX this all falls apart if we have 31 or more bits of Z because
+ * the triangle rasterization code produces unsigned Z values. Negative
+ * vertex Z values come out as large fragment Z uints.
+ */
+ min = (GLint) (min_f * fb->_DepthMaxF);
+ max = (GLint) (max_f * fb->_DepthMaxF);
+ if (max < 0)
+ max = 0x7fffffff; /* catch over flow for 30-bit z */
+ /* Note that we do the comparisons here using signed integers.
+ */
for (i = 0; i < count; i++) {
if (zValues[i] < min)
zValues[i] = min;
diff --git a/src/mesa/swrast/s_span.c b/src/mesa/swrast/s_span.c
index 4ea9547..874a37b 100644
--- a/src/mesa/swrast/s_span.c
+++ b/src/mesa/swrast/s_span.c
@@ -1316,6 +1316,13 @@
ASSERT(span->end <= MAX_WIDTH);
+ /* Depth bounds test */
+ if (ctx->Depth.BoundsTest && fb->Visual.depthBits > 0) {
+ if (!_swrast_depth_bounds_test(ctx, span)) {
+ return;
+ }
+ }
+
#ifdef DEBUG
/* Make sure all fragments are within window bounds */
if (span->arrayMask & SPAN_XY) {