Files
third_party_mesa3d/src/mesa/drivers/dri/common/dri_util.c

1078 lines
31 KiB
C
Raw Normal View History

/* $XFree86: xc/lib/GL/dri/dri_util.c,v 1.7 2003/04/28 17:01:25 dawes Exp $ */
/**
* \file dri_util.c
* DRI utility functions.
*
* This module acts as glue between GLX and the actual hardware driver. A DRI
* driver doesn't really \e have to use any of this - it's optional. But, some
* useful stuff is done here that otherwise would have to be duplicated in most
* drivers.
*
* Basically, these utility functions take care of some of the dirty details of
* screen initialization, context creation, context binding, DRM setup, etc.
*
* These functions are compiled into each DRI driver so libGL.so knows nothing
* about them.
*/
#include <assert.h>
#include <stdarg.h>
#include <unistd.h>
#include <sys/mman.h>
#include <stdio.h>
2004-12-14 09:11:52 +00:00
#ifndef MAP_FAILED
#define MAP_FAILED ((void *)-1)
#endif
#include "imports.h"
#define None 0
#include "dri_util.h"
#include "drm_sarea.h"
#ifndef GLX_OML_sync_control
typedef GLboolean ( * PFNGLXGETMSCRATEOMLPROC) (__DRInativeDisplay *dpy, __DRIid drawable, int32_t *numerator, int32_t *denominator);
#endif
Fixes the glXGetProcAddress portion of the interface. Most of the functions that are currently obtained via glXGetProcAddress and all of the XF86DRI functions are replaced with a funciton table. This table will be passed to __driCreateNewScreen. One of the functions in the table is getProcAddress. This allows some loaders to expose functionality not in all loaders. This will be immediatly used for glxEnableExtension (formerly known to drivers as __glXScrEnableExtension). libGL (and in the future libglx) expose this function so that drivers can enable GLX extensions. libEGL should exposed eglEnableExtension to enable EGL extensions. The same function cannot be used for both because the extensions have different names and (possibly) different semantics. Drivers can optionally use one, both, or neither. The key parts are in the __DRIinterfaceMethodsRec structure in dri_interface.h. A pointer to one of these structures is passed into __driCreateNewScreen. Because of this, the version of the API is bumped to 20050725. Since the previous version(s) were never in a release, their existance is erased. I was actually a little surprised by how much code this cuts from the drivers. A lot of glXGetProcAddress calls disappear, and a lot of version checks go with them. Nice. The one thing I'm not sure of is removing __glXInitialize. For some reason that function was in the glXGetProcAddress table, but *nothing* in the Mesa tree used it. Did something with DRI conf. use this function? It seems odd...
2005-07-26 02:44:01 +00:00
/* This pointer *must* be set by the driver's __driCreateNewScreen funciton!
*/
const __DRIinterfaceMethods * dri_interface = NULL;
/**
* This is used in a couple of places that call \c driCreateNewDrawable.
*/
static const int empty_attribute_list[1] = { None };
/**
* Cached copy of the internal API version used by libGL and the client-side
* DRI driver.
*/
static int api_ver = 0;
/* forward declarations */
2005-09-04 22:55:57 +00:00
static int driQueryFrameTracking( __DRInativeDisplay *dpy, void *priv,
int64_t *sbc, int64_t *missedFrames,
float *lastMissedUsage, float *usage );
2005-09-04 22:55:57 +00:00
static void *driCreateNewDrawable(__DRInativeDisplay *dpy,
const __GLcontextModes *modes,
__DRIid draw, __DRIdrawable *pdraw,
int renderType, const int *attrs);
static void driDestroyDrawable(__DRInativeDisplay *dpy, void *drawablePrivate);
/**
* Print message to \c stderr if the \c LIBGL_DEBUG environment variable
* is set.
*
* Is called from the drivers.
*
* \param f \c printf like format string.
*/
void
__driUtilMessage(const char *f, ...)
{
va_list args;
if (getenv("LIBGL_DEBUG")) {
fprintf(stderr, "libGL error: \n");
va_start(args, f);
vfprintf(stderr, f, args);
va_end(args);
fprintf(stderr, "\n");
}
}
/*****************************************************************/
/** \name Drawable list management */
/*****************************************************************/
/*@{*/
static GLboolean __driAddDrawable(void *drawHash, __DRIdrawable *pdraw)
{
__DRIdrawablePrivate *pdp = (__DRIdrawablePrivate *)pdraw->private;
if (drmHashInsert(drawHash, pdp->draw, pdraw))
return GL_FALSE;
return GL_TRUE;
}
static __DRIdrawable *__driFindDrawable(void *drawHash, __DRIid draw)
{
int retcode;
__DRIdrawable *pdraw;
retcode = drmHashLookup(drawHash, draw, (void *)&pdraw);
if (retcode)
return NULL;
return pdraw;
}
/**
* Find drawables in the local hash that have been destroyed on the
* server.
*
* \param drawHash Hash-table containing all know drawables.
*/
static void __driGarbageCollectDrawables(void *drawHash)
{
__DRIid draw;
__DRInativeDisplay *dpy;
__DRIdrawable *pdraw;
if (drmHashFirst(drawHash, &draw, (void *)&pdraw) == 1) {
do {
__DRIdrawablePrivate *pdp = (__DRIdrawablePrivate *)pdraw->private;
dpy = pdp->driScreenPriv->display;
Fixes the glXGetProcAddress portion of the interface. Most of the functions that are currently obtained via glXGetProcAddress and all of the XF86DRI functions are replaced with a funciton table. This table will be passed to __driCreateNewScreen. One of the functions in the table is getProcAddress. This allows some loaders to expose functionality not in all loaders. This will be immediatly used for glxEnableExtension (formerly known to drivers as __glXScrEnableExtension). libGL (and in the future libglx) expose this function so that drivers can enable GLX extensions. libEGL should exposed eglEnableExtension to enable EGL extensions. The same function cannot be used for both because the extensions have different names and (possibly) different semantics. Drivers can optionally use one, both, or neither. The key parts are in the __DRIinterfaceMethodsRec structure in dri_interface.h. A pointer to one of these structures is passed into __driCreateNewScreen. Because of this, the version of the API is bumped to 20050725. Since the previous version(s) were never in a release, their existance is erased. I was actually a little surprised by how much code this cuts from the drivers. A lot of glXGetProcAddress calls disappear, and a lot of version checks go with them. Nice. The one thing I'm not sure of is removing __glXInitialize. For some reason that function was in the glXGetProcAddress table, but *nothing* in the Mesa tree used it. Did something with DRI conf. use this function? It seems odd...
2005-07-26 02:44:01 +00:00
if (! (*dri_interface->windowExists)(dpy, draw)) {
/* Destroy the local drawable data, if the drawable no
longer exists in the Xserver */
(*pdraw->destroyDrawable)(dpy, pdraw->private);
_mesa_free(pdraw);
}
} while (drmHashNext(drawHash, &draw, (void *)&pdraw) == 1);
}
}
/*@}*/
/*****************************************************************/
/** \name Context (un)binding functions */
/*****************************************************************/
/*@{*/
/**
* Unbind context.
*
* \param dpy the display handle.
* \param scrn the screen number.
* \param draw drawable.
* \param read Current reading drawable.
* \param gc context.
*
* \return \c GL_TRUE on success, or \c GL_FALSE on failure.
*
* \internal
* This function calls __DriverAPIRec::UnbindContext, and then decrements
* __DRIdrawablePrivateRec::refcount which must be non-zero for a successful
* return.
*
* While casting the opaque private pointers associated with the parameters
* into their respective real types it also assures they are not \c NULL.
*/
static GLboolean driUnbindContext(__DRInativeDisplay *dpy, int scrn,
__DRIid draw, __DRIid read,
__DRIcontext *ctx)
{
__DRIscreen *pDRIScreen;
__DRIdrawable *pdraw;
__DRIdrawable *pread;
__DRIcontextPrivate *pcp;
__DRIscreenPrivate *psp;
__DRIdrawablePrivate *pdp;
__DRIdrawablePrivate *prp;
/*
** Assume error checking is done properly in glXMakeCurrent before
** calling driUnbindContext.
*/
if (ctx == NULL || draw == None || read == None) {
/* ERROR!!! */
return GL_FALSE;
}
Fixes the glXGetProcAddress portion of the interface. Most of the functions that are currently obtained via glXGetProcAddress and all of the XF86DRI functions are replaced with a funciton table. This table will be passed to __driCreateNewScreen. One of the functions in the table is getProcAddress. This allows some loaders to expose functionality not in all loaders. This will be immediatly used for glxEnableExtension (formerly known to drivers as __glXScrEnableExtension). libGL (and in the future libglx) expose this function so that drivers can enable GLX extensions. libEGL should exposed eglEnableExtension to enable EGL extensions. The same function cannot be used for both because the extensions have different names and (possibly) different semantics. Drivers can optionally use one, both, or neither. The key parts are in the __DRIinterfaceMethodsRec structure in dri_interface.h. A pointer to one of these structures is passed into __driCreateNewScreen. Because of this, the version of the API is bumped to 20050725. Since the previous version(s) were never in a release, their existance is erased. I was actually a little surprised by how much code this cuts from the drivers. A lot of glXGetProcAddress calls disappear, and a lot of version checks go with them. Nice. The one thing I'm not sure of is removing __glXInitialize. For some reason that function was in the glXGetProcAddress table, but *nothing* in the Mesa tree used it. Did something with DRI conf. use this function? It seems odd...
2005-07-26 02:44:01 +00:00
pDRIScreen = (*dri_interface->getScreen)(dpy, scrn);
if ( (pDRIScreen == NULL) || (pDRIScreen->private == NULL) ) {
/* ERROR!!! */
return GL_FALSE;
}
psp = (__DRIscreenPrivate *)pDRIScreen->private;
pcp = (__DRIcontextPrivate *)ctx->private;
pdraw = __driFindDrawable(psp->drawHash, draw);
if (!pdraw) {
/* ERROR!!! */
return GL_FALSE;
}
pdp = (__DRIdrawablePrivate *)pdraw->private;
pread = __driFindDrawable(psp->drawHash, read);
if (!pread) {
/* ERROR!!! */
return GL_FALSE;
}
prp = (__DRIdrawablePrivate *)pread->private;
/* Let driver unbind drawable from context */
(*psp->DriverAPI.UnbindContext)(pcp);
if (pdp->refcount == 0) {
/* ERROR!!! */
return GL_FALSE;
}
pdp->refcount--;
if (prp != pdp) {
if (prp->refcount == 0) {
/* ERROR!!! */
return GL_FALSE;
}
prp->refcount--;
}
/* XXX this is disabled so that if we call SwapBuffers on an unbound
* window we can determine the last context bound to the window and
* use that context's lock. (BrianP, 2-Dec-2000)
*/
#if 0
/* Unbind the drawable */
pcp->driDrawablePriv = NULL;
pdp->driContextPriv = &psp->dummyContextPriv;
#endif
return GL_TRUE;
}
/**
* This function takes both a read buffer and a draw buffer. This is needed
* for \c glXMakeCurrentReadSGI or GLX 1.3's \c glXMakeContextCurrent
* function.
*
* \bug This function calls \c driCreateNewDrawable in two places with the
* \c renderType hard-coded to \c GLX_WINDOW_BIT. Some checking might
* be needed in those places when support for pbuffers and / or pixmaps
* is added. Is it safe to assume that the drawable is a window?
*/
static GLboolean DoBindContext(__DRInativeDisplay *dpy,
__DRIid draw, __DRIid read,
__DRIcontext *ctx, const __GLcontextModes * modes,
__DRIscreenPrivate *psp)
{
__DRIdrawable *pdraw;
__DRIdrawablePrivate *pdp;
__DRIdrawable *pread;
__DRIdrawablePrivate *prp;
__DRIcontextPrivate * const pcp = ctx->private;
/* Find the _DRIdrawable which corresponds to the writing drawable. */
pdraw = __driFindDrawable(psp->drawHash, draw);
if (!pdraw) {
/* Allocate a new drawable */
pdraw = (__DRIdrawable *)_mesa_malloc(sizeof(__DRIdrawable));
if (!pdraw) {
/* ERROR!!! */
return GL_FALSE;
}
/* Create a new drawable */
driCreateNewDrawable(dpy, modes, draw, pdraw, GLX_WINDOW_BIT,
empty_attribute_list);
if (!pdraw->private) {
/* ERROR!!! */
_mesa_free(pdraw);
return GL_FALSE;
}
}
pdp = (__DRIdrawablePrivate *) pdraw->private;
/* Find the _DRIdrawable which corresponds to the reading drawable. */
if (read == draw) {
/* read buffer == draw buffer */
prp = pdp;
}
else {
pread = __driFindDrawable(psp->drawHash, read);
if (!pread) {
/* Allocate a new drawable */
pread = (__DRIdrawable *)_mesa_malloc(sizeof(__DRIdrawable));
if (!pread) {
/* ERROR!!! */
return GL_FALSE;
}
/* Create a new drawable */
driCreateNewDrawable(dpy, modes, read, pread, GLX_WINDOW_BIT,
empty_attribute_list);
if (!pread->private) {
/* ERROR!!! */
_mesa_free(pread);
return GL_FALSE;
}
}
prp = (__DRIdrawablePrivate *) pread->private;
}
/* Bind the drawable to the context */
pcp->driDrawablePriv = pdp;
pdp->driContextPriv = pcp;
pdp->refcount++;
if ( pdp != prp ) {
prp->refcount++;
}
/*
** Now that we have a context associated with this drawable, we can
** initialize the drawable information if has not been done before.
*/
if (!pdp->pStamp || *pdp->pStamp != pdp->lastStamp) {
DRM_SPINLOCK(&psp->pSAREA->drawable_lock, psp->drawLockID);
__driUtilUpdateDrawableInfo(pdp);
DRM_SPINUNLOCK(&psp->pSAREA->drawable_lock, psp->drawLockID);
}
/* Call device-specific MakeCurrent */
(*psp->DriverAPI.MakeCurrent)(pcp, pdp, prp);
return GL_TRUE;
}
/**
* This function takes both a read buffer and a draw buffer. This is needed
* for \c glXMakeCurrentReadSGI or GLX 1.3's \c glXMakeContextCurrent
* function.
*/
static GLboolean driBindContext(__DRInativeDisplay *dpy, int scrn,
__DRIid draw, __DRIid read,
__DRIcontext * ctx)
{
__DRIscreen *pDRIScreen;
/*
** Assume error checking is done properly in glXMakeCurrent before
** calling driBindContext.
*/
if (ctx == NULL || draw == None || read == None) {
/* ERROR!!! */
return GL_FALSE;
}
Fixes the glXGetProcAddress portion of the interface. Most of the functions that are currently obtained via glXGetProcAddress and all of the XF86DRI functions are replaced with a funciton table. This table will be passed to __driCreateNewScreen. One of the functions in the table is getProcAddress. This allows some loaders to expose functionality not in all loaders. This will be immediatly used for glxEnableExtension (formerly known to drivers as __glXScrEnableExtension). libGL (and in the future libglx) expose this function so that drivers can enable GLX extensions. libEGL should exposed eglEnableExtension to enable EGL extensions. The same function cannot be used for both because the extensions have different names and (possibly) different semantics. Drivers can optionally use one, both, or neither. The key parts are in the __DRIinterfaceMethodsRec structure in dri_interface.h. A pointer to one of these structures is passed into __driCreateNewScreen. Because of this, the version of the API is bumped to 20050725. Since the previous version(s) were never in a release, their existance is erased. I was actually a little surprised by how much code this cuts from the drivers. A lot of glXGetProcAddress calls disappear, and a lot of version checks go with them. Nice. The one thing I'm not sure of is removing __glXInitialize. For some reason that function was in the glXGetProcAddress table, but *nothing* in the Mesa tree used it. Did something with DRI conf. use this function? It seems odd...
2005-07-26 02:44:01 +00:00
pDRIScreen = (*dri_interface->getScreen)(dpy, scrn);
if ( (pDRIScreen == NULL) || (pDRIScreen->private == NULL) ) {
/* ERROR!!! */
return GL_FALSE;
}
return DoBindContext( dpy, draw, read, ctx, ctx->mode,
(__DRIscreenPrivate *)pDRIScreen->private );
}
/*@}*/
/*****************************************************************/
/** \name Drawable handling functions */
/*****************************************************************/
/*@{*/
/**
* Update private drawable information.
*
* \param pdp pointer to the private drawable information to update.
*
* This function basically updates the __DRIdrawablePrivate struct's
Fixes the glXGetProcAddress portion of the interface. Most of the functions that are currently obtained via glXGetProcAddress and all of the XF86DRI functions are replaced with a funciton table. This table will be passed to __driCreateNewScreen. One of the functions in the table is getProcAddress. This allows some loaders to expose functionality not in all loaders. This will be immediatly used for glxEnableExtension (formerly known to drivers as __glXScrEnableExtension). libGL (and in the future libglx) expose this function so that drivers can enable GLX extensions. libEGL should exposed eglEnableExtension to enable EGL extensions. The same function cannot be used for both because the extensions have different names and (possibly) different semantics. Drivers can optionally use one, both, or neither. The key parts are in the __DRIinterfaceMethodsRec structure in dri_interface.h. A pointer to one of these structures is passed into __driCreateNewScreen. Because of this, the version of the API is bumped to 20050725. Since the previous version(s) were never in a release, their existance is erased. I was actually a little surprised by how much code this cuts from the drivers. A lot of glXGetProcAddress calls disappear, and a lot of version checks go with them. Nice. The one thing I'm not sure of is removing __glXInitialize. For some reason that function was in the glXGetProcAddress table, but *nothing* in the Mesa tree used it. Did something with DRI conf. use this function? It seems odd...
2005-07-26 02:44:01 +00:00
* cliprect information by calling \c __DRIinterfaceMethods::getDrawableInfo.
* This is usually called by the DRI_VALIDATE_DRAWABLE_INFO macro which
* compares the __DRIdrwablePrivate pStamp and lastStamp values. If
* the values are different that means we have to update the clipping
* info.
*/
void
__driUtilUpdateDrawableInfo(__DRIdrawablePrivate *pdp)
{
__DRIscreenPrivate *psp;
__DRIcontextPrivate *pcp = pdp->driContextPriv;
if (!pcp || (pdp != pcp->driDrawablePriv)) {
/* ERROR!!! */
return;
}
psp = pdp->driScreenPriv;
if (!psp) {
/* ERROR!!! */
return;
}
if (pdp->pClipRects) {
_mesa_free(pdp->pClipRects);
}
if (pdp->pBackClipRects) {
_mesa_free(pdp->pBackClipRects);
}
DRM_SPINUNLOCK(&psp->pSAREA->drawable_lock, psp->drawLockID);
if (!__driFindDrawable(psp->drawHash, pdp->draw) ||
Fixes the glXGetProcAddress portion of the interface. Most of the functions that are currently obtained via glXGetProcAddress and all of the XF86DRI functions are replaced with a funciton table. This table will be passed to __driCreateNewScreen. One of the functions in the table is getProcAddress. This allows some loaders to expose functionality not in all loaders. This will be immediatly used for glxEnableExtension (formerly known to drivers as __glXScrEnableExtension). libGL (and in the future libglx) expose this function so that drivers can enable GLX extensions. libEGL should exposed eglEnableExtension to enable EGL extensions. The same function cannot be used for both because the extensions have different names and (possibly) different semantics. Drivers can optionally use one, both, or neither. The key parts are in the __DRIinterfaceMethodsRec structure in dri_interface.h. A pointer to one of these structures is passed into __driCreateNewScreen. Because of this, the version of the API is bumped to 20050725. Since the previous version(s) were never in a release, their existance is erased. I was actually a little surprised by how much code this cuts from the drivers. A lot of glXGetProcAddress calls disappear, and a lot of version checks go with them. Nice. The one thing I'm not sure of is removing __glXInitialize. For some reason that function was in the glXGetProcAddress table, but *nothing* in the Mesa tree used it. Did something with DRI conf. use this function? It seems odd...
2005-07-26 02:44:01 +00:00
! (*dri_interface->getDrawableInfo)(pdp->display, pdp->screen, pdp->draw,
&pdp->index, &pdp->lastStamp,
&pdp->x, &pdp->y, &pdp->w, &pdp->h,
&pdp->numClipRects, &pdp->pClipRects,
&pdp->backX,
&pdp->backY,
&pdp->numBackClipRects,
&pdp->pBackClipRects )) {
/* Error -- eg the window may have been destroyed. Keep going
* with no cliprects.
*/
pdp->pStamp = &pdp->lastStamp; /* prevent endless loop */
pdp->numClipRects = 0;
pdp->pClipRects = NULL;
pdp->numBackClipRects = 0;
pdp->pBackClipRects = NULL;
}
else
pdp->pStamp = &(psp->pSAREA->drawableTable[pdp->index].stamp);
DRM_SPINLOCK(&psp->pSAREA->drawable_lock, psp->drawLockID);
}
/*@}*/
/*****************************************************************/
/** \name GLX callbacks */
/*****************************************************************/
/*@{*/
/**
* Swap buffers.
*
* \param dpy the display handle.
* \param drawablePrivate opaque pointer to the per-drawable private info.
*
* \internal
* This function calls __DRIdrawablePrivate::swapBuffers.
*
* Is called directly from glXSwapBuffers().
*/
static void driSwapBuffers( __DRInativeDisplay *dpy, void *drawablePrivate )
{
__DRIdrawablePrivate *dPriv = (__DRIdrawablePrivate *) drawablePrivate;
dPriv->swapBuffers(dPriv);
(void) dpy;
}
/**
* Called directly from a number of higher-level GLX functions.
*/
static int driGetMSC( void *screenPrivate, int64_t *msc )
{
__DRIscreenPrivate *sPriv = (__DRIscreenPrivate *) screenPrivate;
return sPriv->DriverAPI.GetMSC( sPriv, msc );
}
/**
* Called directly from a number of higher-level GLX functions.
*/
static int driGetSBC( __DRInativeDisplay *dpy, void *drawablePrivate, int64_t *sbc )
{
__DRIdrawablePrivate *dPriv = (__DRIdrawablePrivate *) drawablePrivate;
__DRIswapInfo sInfo;
int status;
status = dPriv->driScreenPriv->DriverAPI.GetSwapInfo( dPriv, & sInfo );
*sbc = sInfo.swap_count;
return status;
}
static int driWaitForSBC( __DRInativeDisplay * dpy, void *drawablePriv,
int64_t target_sbc,
int64_t * msc, int64_t * sbc )
{
__DRIdrawablePrivate *dPriv = (__DRIdrawablePrivate *) drawablePriv;
return dPriv->driScreenPriv->DriverAPI.WaitForSBC( dPriv, target_sbc,
msc, sbc );
}
static int driWaitForMSC( __DRInativeDisplay * dpy, void *drawablePriv,
int64_t target_msc,
int64_t divisor, int64_t remainder,
int64_t * msc, int64_t * sbc )
{
__DRIdrawablePrivate *dPriv = (__DRIdrawablePrivate *) drawablePriv;
__DRIswapInfo sInfo;
int status;
status = dPriv->driScreenPriv->DriverAPI.WaitForMSC( dPriv, target_msc,
divisor, remainder,
msc );
/* GetSwapInfo() may not be provided by the driver if GLX_SGI_video_sync
* is supported but GLX_OML_sync_control is not. Therefore, don't return
* an error value if GetSwapInfo() is not implemented.
*/
if ( status == 0
&& dPriv->driScreenPriv->DriverAPI.GetSwapInfo ) {
status = dPriv->driScreenPriv->DriverAPI.GetSwapInfo( dPriv, & sInfo );
*sbc = sInfo.swap_count;
}
return status;
}
static int64_t driSwapBuffersMSC( __DRInativeDisplay * dpy, void *drawablePriv,
int64_t target_msc,
int64_t divisor, int64_t remainder )
{
__DRIdrawablePrivate *dPriv = (__DRIdrawablePrivate *) drawablePriv;
return dPriv->driScreenPriv->DriverAPI.SwapBuffersMSC( dPriv, target_msc,
divisor,
remainder );
}
/**
* This is called via __DRIscreenRec's createNewDrawable pointer.
*/
static void *driCreateNewDrawable(__DRInativeDisplay *dpy,
const __GLcontextModes *modes,
__DRIid draw,
__DRIdrawable *pdraw,
int renderType,
const int *attrs)
{
Fixes the glXGetProcAddress portion of the interface. Most of the functions that are currently obtained via glXGetProcAddress and all of the XF86DRI functions are replaced with a funciton table. This table will be passed to __driCreateNewScreen. One of the functions in the table is getProcAddress. This allows some loaders to expose functionality not in all loaders. This will be immediatly used for glxEnableExtension (formerly known to drivers as __glXScrEnableExtension). libGL (and in the future libglx) expose this function so that drivers can enable GLX extensions. libEGL should exposed eglEnableExtension to enable EGL extensions. The same function cannot be used for both because the extensions have different names and (possibly) different semantics. Drivers can optionally use one, both, or neither. The key parts are in the __DRIinterfaceMethodsRec structure in dri_interface.h. A pointer to one of these structures is passed into __driCreateNewScreen. Because of this, the version of the API is bumped to 20050725. Since the previous version(s) were never in a release, their existance is erased. I was actually a little surprised by how much code this cuts from the drivers. A lot of glXGetProcAddress calls disappear, and a lot of version checks go with them. Nice. The one thing I'm not sure of is removing __glXInitialize. For some reason that function was in the glXGetProcAddress table, but *nothing* in the Mesa tree used it. Did something with DRI conf. use this function? It seems odd...
2005-07-26 02:44:01 +00:00
__DRIscreen * const pDRIScreen = (*dri_interface->getScreen)(dpy, modes->screen);
__DRIscreenPrivate *psp;
__DRIdrawablePrivate *pdp;
pdraw->private = NULL;
/* Since pbuffers are not yet supported, no drawable attributes are
* supported either.
*/
(void) attrs;
if ( (pDRIScreen == NULL) || (pDRIScreen->private == NULL) ) {
return NULL;
}
pdp = (__DRIdrawablePrivate *)_mesa_malloc(sizeof(__DRIdrawablePrivate));
if (!pdp) {
return NULL;
}
Fixes the glXGetProcAddress portion of the interface. Most of the functions that are currently obtained via glXGetProcAddress and all of the XF86DRI functions are replaced with a funciton table. This table will be passed to __driCreateNewScreen. One of the functions in the table is getProcAddress. This allows some loaders to expose functionality not in all loaders. This will be immediatly used for glxEnableExtension (formerly known to drivers as __glXScrEnableExtension). libGL (and in the future libglx) expose this function so that drivers can enable GLX extensions. libEGL should exposed eglEnableExtension to enable EGL extensions. The same function cannot be used for both because the extensions have different names and (possibly) different semantics. Drivers can optionally use one, both, or neither. The key parts are in the __DRIinterfaceMethodsRec structure in dri_interface.h. A pointer to one of these structures is passed into __driCreateNewScreen. Because of this, the version of the API is bumped to 20050725. Since the previous version(s) were never in a release, their existance is erased. I was actually a little surprised by how much code this cuts from the drivers. A lot of glXGetProcAddress calls disappear, and a lot of version checks go with them. Nice. The one thing I'm not sure of is removing __glXInitialize. For some reason that function was in the glXGetProcAddress table, but *nothing* in the Mesa tree used it. Did something with DRI conf. use this function? It seems odd...
2005-07-26 02:44:01 +00:00
if (!(*dri_interface->createDrawable)(dpy, modes->screen, draw, &pdp->hHWDrawable)) {
_mesa_free(pdp);
return NULL;
}
pdp->draw = draw;
pdp->pdraw = pdraw;
pdp->refcount = 0;
pdp->pStamp = NULL;
pdp->lastStamp = 0;
pdp->index = 0;
pdp->x = 0;
pdp->y = 0;
pdp->w = 0;
pdp->h = 0;
pdp->numClipRects = 0;
pdp->numBackClipRects = 0;
pdp->pClipRects = NULL;
pdp->pBackClipRects = NULL;
pdp->display = dpy;
pdp->screen = modes->screen;
psp = (__DRIscreenPrivate *)pDRIScreen->private;
pdp->driScreenPriv = psp;
pdp->driContextPriv = &psp->dummyContextPriv;
if (!(*psp->DriverAPI.CreateBuffer)(psp, pdp, modes,
renderType == GLX_PIXMAP_BIT)) {
Fixes the glXGetProcAddress portion of the interface. Most of the functions that are currently obtained via glXGetProcAddress and all of the XF86DRI functions are replaced with a funciton table. This table will be passed to __driCreateNewScreen. One of the functions in the table is getProcAddress. This allows some loaders to expose functionality not in all loaders. This will be immediatly used for glxEnableExtension (formerly known to drivers as __glXScrEnableExtension). libGL (and in the future libglx) expose this function so that drivers can enable GLX extensions. libEGL should exposed eglEnableExtension to enable EGL extensions. The same function cannot be used for both because the extensions have different names and (possibly) different semantics. Drivers can optionally use one, both, or neither. The key parts are in the __DRIinterfaceMethodsRec structure in dri_interface.h. A pointer to one of these structures is passed into __driCreateNewScreen. Because of this, the version of the API is bumped to 20050725. Since the previous version(s) were never in a release, their existance is erased. I was actually a little surprised by how much code this cuts from the drivers. A lot of glXGetProcAddress calls disappear, and a lot of version checks go with them. Nice. The one thing I'm not sure of is removing __glXInitialize. For some reason that function was in the glXGetProcAddress table, but *nothing* in the Mesa tree used it. Did something with DRI conf. use this function? It seems odd...
2005-07-26 02:44:01 +00:00
(void)(*dri_interface->destroyDrawable)(dpy, modes->screen, pdp->draw);
_mesa_free(pdp);
return NULL;
}
pdraw->private = pdp;
pdraw->destroyDrawable = driDestroyDrawable;
pdraw->swapBuffers = driSwapBuffers; /* called by glXSwapBuffers() */
pdraw->getSBC = driGetSBC;
pdraw->waitForSBC = driWaitForSBC;
pdraw->waitForMSC = driWaitForMSC;
pdraw->swapBuffersMSC = driSwapBuffersMSC;
pdraw->frameTracking = NULL;
pdraw->queryFrameTracking = driQueryFrameTracking;
/* This special default value is replaced with the configured
* default value when the drawable is first bound to a direct
* rendering context.
*/
pdraw->swap_interval = (unsigned)-1;
pdp->swapBuffers = psp->DriverAPI.SwapBuffers;
/* Add pdraw to drawable list */
if (!__driAddDrawable(psp->drawHash, pdraw)) {
/* ERROR!!! */
(*pdraw->destroyDrawable)(dpy, pdp);
_mesa_free(pdp);
pdp = NULL;
pdraw->private = NULL;
}
return (void *) pdp;
}
2005-09-04 22:55:57 +00:00
static __DRIdrawable *
driGetDrawable(__DRInativeDisplay *dpy, __DRIid draw, void *screenPrivate)
{
__DRIscreenPrivate *psp = (__DRIscreenPrivate *) screenPrivate;
/*
** Make sure this routine returns NULL if the drawable is not bound
** to a direct rendering context!
*/
return __driFindDrawable(psp->drawHash, draw);
}
2005-09-04 22:55:57 +00:00
static void
driDestroyDrawable(__DRInativeDisplay *dpy, void *drawablePrivate)
{
__DRIdrawablePrivate *pdp = (__DRIdrawablePrivate *) drawablePrivate;
__DRIscreenPrivate *psp = pdp->driScreenPriv;
int scrn = psp->myNum;
if (pdp) {
(*psp->DriverAPI.DestroyBuffer)(pdp);
Fixes the glXGetProcAddress portion of the interface. Most of the functions that are currently obtained via glXGetProcAddress and all of the XF86DRI functions are replaced with a funciton table. This table will be passed to __driCreateNewScreen. One of the functions in the table is getProcAddress. This allows some loaders to expose functionality not in all loaders. This will be immediatly used for glxEnableExtension (formerly known to drivers as __glXScrEnableExtension). libGL (and in the future libglx) expose this function so that drivers can enable GLX extensions. libEGL should exposed eglEnableExtension to enable EGL extensions. The same function cannot be used for both because the extensions have different names and (possibly) different semantics. Drivers can optionally use one, both, or neither. The key parts are in the __DRIinterfaceMethodsRec structure in dri_interface.h. A pointer to one of these structures is passed into __driCreateNewScreen. Because of this, the version of the API is bumped to 20050725. Since the previous version(s) were never in a release, their existance is erased. I was actually a little surprised by how much code this cuts from the drivers. A lot of glXGetProcAddress calls disappear, and a lot of version checks go with them. Nice. The one thing I'm not sure of is removing __glXInitialize. For some reason that function was in the glXGetProcAddress table, but *nothing* in the Mesa tree used it. Did something with DRI conf. use this function? It seems odd...
2005-07-26 02:44:01 +00:00
if ((*dri_interface->windowExists)(dpy, pdp->draw))
(void)(*dri_interface->destroyDrawable)(dpy, scrn, pdp->draw);
drmHashDelete(psp->drawHash, pdp->draw);
if (pdp->pClipRects) {
_mesa_free(pdp->pClipRects);
pdp->pClipRects = NULL;
}
if (pdp->pBackClipRects) {
_mesa_free(pdp->pBackClipRects);
pdp->pBackClipRects = NULL;
}
_mesa_free(pdp);
}
}
/*@}*/
/*****************************************************************/
/** \name Context handling functions */
/*****************************************************************/
/*@{*/
/**
* Destroy the per-context private information.
*
* \param dpy the display handle.
* \param scrn the screen number.
* \param contextPrivate opaque pointer to the per-drawable private info.
*
* \internal
* This function calls __DriverAPIRec::DestroyContext on \p contextPrivate, calls
* drmDestroyContext(), and finally frees \p contextPrivate.
*/
2005-09-04 22:55:57 +00:00
static void
driDestroyContext(__DRInativeDisplay *dpy, int scrn, void *contextPrivate)
{
__DRIcontextPrivate *pcp = (__DRIcontextPrivate *) contextPrivate;
if (pcp) {
(*pcp->driScreenPriv->DriverAPI.DestroyContext)(pcp);
__driGarbageCollectDrawables(pcp->driScreenPriv->drawHash);
Fixes the glXGetProcAddress portion of the interface. Most of the functions that are currently obtained via glXGetProcAddress and all of the XF86DRI functions are replaced with a funciton table. This table will be passed to __driCreateNewScreen. One of the functions in the table is getProcAddress. This allows some loaders to expose functionality not in all loaders. This will be immediatly used for glxEnableExtension (formerly known to drivers as __glXScrEnableExtension). libGL (and in the future libglx) expose this function so that drivers can enable GLX extensions. libEGL should exposed eglEnableExtension to enable EGL extensions. The same function cannot be used for both because the extensions have different names and (possibly) different semantics. Drivers can optionally use one, both, or neither. The key parts are in the __DRIinterfaceMethodsRec structure in dri_interface.h. A pointer to one of these structures is passed into __driCreateNewScreen. Because of this, the version of the API is bumped to 20050725. Since the previous version(s) were never in a release, their existance is erased. I was actually a little surprised by how much code this cuts from the drivers. A lot of glXGetProcAddress calls disappear, and a lot of version checks go with them. Nice. The one thing I'm not sure of is removing __glXInitialize. For some reason that function was in the glXGetProcAddress table, but *nothing* in the Mesa tree used it. Did something with DRI conf. use this function? It seems odd...
2005-07-26 02:44:01 +00:00
(void) (*dri_interface->destroyContext)(dpy, scrn, pcp->contextID);
_mesa_free(pcp);
}
}
/**
* Create the per-drawable private driver information.
*
* \param dpy The display handle.
* \param modes Mode used to create the new context.
* \param render_type Type of rendering target. \c GLX_RGBA is the only
* type likely to ever be supported for direct-rendering.
* \param sharedPrivate The shared context dependent methods or \c NULL if
* non-existent.
* \param pctx DRI context to receive the context dependent methods.
*
* \returns An opaque pointer to the per-context private information on
* success, or \c NULL on failure.
*
* \internal
* This function allocates and fills a __DRIcontextPrivateRec structure. It
* performs some device independent initialization and passes all the
* relevent information to __DriverAPIRec::CreateContext to create the
* context.
*
*/
static void *
driCreateNewContext(__DRInativeDisplay *dpy, const __GLcontextModes *modes,
int render_type, void *sharedPrivate, __DRIcontext *pctx)
{
__DRIscreen *pDRIScreen;
__DRIcontextPrivate *pcp;
__DRIcontextPrivate *pshare = (__DRIcontextPrivate *) sharedPrivate;
__DRIscreenPrivate *psp;
void * const shareCtx = (pshare != NULL) ? pshare->driverPrivate : NULL;
Fixes the glXGetProcAddress portion of the interface. Most of the functions that are currently obtained via glXGetProcAddress and all of the XF86DRI functions are replaced with a funciton table. This table will be passed to __driCreateNewScreen. One of the functions in the table is getProcAddress. This allows some loaders to expose functionality not in all loaders. This will be immediatly used for glxEnableExtension (formerly known to drivers as __glXScrEnableExtension). libGL (and in the future libglx) expose this function so that drivers can enable GLX extensions. libEGL should exposed eglEnableExtension to enable EGL extensions. The same function cannot be used for both because the extensions have different names and (possibly) different semantics. Drivers can optionally use one, both, or neither. The key parts are in the __DRIinterfaceMethodsRec structure in dri_interface.h. A pointer to one of these structures is passed into __driCreateNewScreen. Because of this, the version of the API is bumped to 20050725. Since the previous version(s) were never in a release, their existance is erased. I was actually a little surprised by how much code this cuts from the drivers. A lot of glXGetProcAddress calls disappear, and a lot of version checks go with them. Nice. The one thing I'm not sure of is removing __glXInitialize. For some reason that function was in the glXGetProcAddress table, but *nothing* in the Mesa tree used it. Did something with DRI conf. use this function? It seems odd...
2005-07-26 02:44:01 +00:00
pDRIScreen = (*dri_interface->getScreen)(dpy, modes->screen);
if ( (pDRIScreen == NULL) || (pDRIScreen->private == NULL) ) {
/* ERROR!!! */
return NULL;
}
psp = (__DRIscreenPrivate *)pDRIScreen->private;
pcp = (__DRIcontextPrivate *)_mesa_malloc(sizeof(__DRIcontextPrivate));
if (!pcp) {
return NULL;
}
Fixes the glXGetProcAddress portion of the interface. Most of the functions that are currently obtained via glXGetProcAddress and all of the XF86DRI functions are replaced with a funciton table. This table will be passed to __driCreateNewScreen. One of the functions in the table is getProcAddress. This allows some loaders to expose functionality not in all loaders. This will be immediatly used for glxEnableExtension (formerly known to drivers as __glXScrEnableExtension). libGL (and in the future libglx) expose this function so that drivers can enable GLX extensions. libEGL should exposed eglEnableExtension to enable EGL extensions. The same function cannot be used for both because the extensions have different names and (possibly) different semantics. Drivers can optionally use one, both, or neither. The key parts are in the __DRIinterfaceMethodsRec structure in dri_interface.h. A pointer to one of these structures is passed into __driCreateNewScreen. Because of this, the version of the API is bumped to 20050725. Since the previous version(s) were never in a release, their existance is erased. I was actually a little surprised by how much code this cuts from the drivers. A lot of glXGetProcAddress calls disappear, and a lot of version checks go with them. Nice. The one thing I'm not sure of is removing __glXInitialize. For some reason that function was in the glXGetProcAddress table, but *nothing* in the Mesa tree used it. Did something with DRI conf. use this function? It seems odd...
2005-07-26 02:44:01 +00:00
if (! (*dri_interface->createContext)(dpy, modes->screen, modes->fbconfigID,
&pcp->contextID, &pcp->hHWContext)) {
_mesa_free(pcp);
return NULL;
}
pcp->display = dpy;
pcp->driScreenPriv = psp;
pcp->driDrawablePriv = NULL;
/* When the first context is created for a screen, initialize a "dummy"
* context.
*/
if (!psp->dummyContextPriv.driScreenPriv) {
psp->dummyContextPriv.contextID = 0;
psp->dummyContextPriv.hHWContext = psp->pSAREA->dummy_context;
psp->dummyContextPriv.driScreenPriv = psp;
psp->dummyContextPriv.driDrawablePriv = NULL;
psp->dummyContextPriv.driverPrivate = NULL;
/* No other fields should be used! */
}
pctx->destroyContext = driDestroyContext;
pctx->bindContext = driBindContext;
pctx->unbindContext = driUnbindContext;
if ( !(*psp->DriverAPI.CreateContext)(modes, pcp, shareCtx) ) {
Fixes the glXGetProcAddress portion of the interface. Most of the functions that are currently obtained via glXGetProcAddress and all of the XF86DRI functions are replaced with a funciton table. This table will be passed to __driCreateNewScreen. One of the functions in the table is getProcAddress. This allows some loaders to expose functionality not in all loaders. This will be immediatly used for glxEnableExtension (formerly known to drivers as __glXScrEnableExtension). libGL (and in the future libglx) expose this function so that drivers can enable GLX extensions. libEGL should exposed eglEnableExtension to enable EGL extensions. The same function cannot be used for both because the extensions have different names and (possibly) different semantics. Drivers can optionally use one, both, or neither. The key parts are in the __DRIinterfaceMethodsRec structure in dri_interface.h. A pointer to one of these structures is passed into __driCreateNewScreen. Because of this, the version of the API is bumped to 20050725. Since the previous version(s) were never in a release, their existance is erased. I was actually a little surprised by how much code this cuts from the drivers. A lot of glXGetProcAddress calls disappear, and a lot of version checks go with them. Nice. The one thing I'm not sure of is removing __glXInitialize. For some reason that function was in the glXGetProcAddress table, but *nothing* in the Mesa tree used it. Did something with DRI conf. use this function? It seems odd...
2005-07-26 02:44:01 +00:00
(void) (*dri_interface->destroyContext)(dpy, modes->screen,
pcp->contextID);
_mesa_free(pcp);
return NULL;
}
__driGarbageCollectDrawables(pcp->driScreenPriv->drawHash);
return pcp;
}
/*@}*/
/*****************************************************************/
/** \name Screen handling functions */
/*****************************************************************/
/*@{*/
/**
* Destroy the per-screen private information.
*
* \param dpy the display handle.
* \param scrn the screen number.
* \param screenPrivate opaque pointer to the per-screen private information.
*
* \internal
* This function calls __DriverAPIRec::DestroyScreen on \p screenPrivate, calls
* drmClose(), and finally frees \p screenPrivate.
*/
static void driDestroyScreen(__DRInativeDisplay *dpy, int scrn, void *screenPrivate)
{
__DRIscreenPrivate *psp = (__DRIscreenPrivate *) screenPrivate;
if (psp) {
/* No interaction with the X-server is possible at this point. This
* routine is called after XCloseDisplay, so there is no protocol
* stream open to the X-server anymore.
*/
if (psp->DriverAPI.DestroyScreen)
(*psp->DriverAPI.DestroyScreen)(psp);
(void)drmUnmap((drmAddress)psp->pSAREA, SAREA_MAX);
(void)drmUnmap((drmAddress)psp->pFB, psp->fbSize);
_mesa_free(psp->pDevPriv);
(void)drmClose(psp->fd);
if ( psp->modes != NULL ) {
(*dri_interface->destroyContextModes)( psp->modes );
}
_mesa_free(psp);
}
}
/**
* Utility function used to create a new driver-private screen structure.
*
* \param dpy Display pointer
* \param scrn Index of the screen
* \param psc DRI screen data (not driver private)
* \param modes Linked list of known display modes. This list is, at a
* minimum, a list of modes based on the current display mode.
* These roughly match the set of available X11 visuals, but it
* need not be limited to X11! The calling libGL should create
* a list that will inform the driver of the current display
* mode (i.e., color buffer depth, depth buffer depth, etc.).
* \param ddx_version Version of the 2D DDX. This may not be meaningful for
* all drivers.
* \param dri_version Version of the "server-side" DRI.
* \param drm_version Version of the kernel DRM.
* \param frame_buffer Data describing the location and layout of the
* framebuffer.
* \param pSAREA Pointer the the SAREA.
* \param fd Device handle for the DRM.
* \param internal_api_version Version of the internal interface between the
* driver and libGL.
* \param driverAPI Driver API functions used by other routines in dri_util.c.
*
* \note
* There is no need to check the minimum API version in this function. Since
* the \c __driCreateNewScreen function is versioned, it is impossible for a
* loader that is too old to even load this driver.
*/
__DRIscreenPrivate *
__driUtilCreateNewScreen(__DRInativeDisplay *dpy, int scrn, __DRIscreen *psc,
__GLcontextModes * modes,
const __DRIversion * ddx_version,
const __DRIversion * dri_version,
const __DRIversion * drm_version,
const __DRIframebuffer * frame_buffer,
drm_sarea_t *pSAREA,
int fd,
int internal_api_version,
const struct __DriverAPIRec *driverAPI)
{
__DRIscreenPrivate *psp;
api_ver = internal_api_version;
psp = (__DRIscreenPrivate *)_mesa_malloc(sizeof(__DRIscreenPrivate));
if (!psp) {
return NULL;
}
/* Create the hash table */
psp->drawHash = drmHashCreate();
if ( psp->drawHash == NULL ) {
_mesa_free( psp );
return NULL;
}
psp->display = dpy;
psp->myNum = scrn;
psp->psc = psc;
psp->modes = modes;
/*
** NOT_DONE: This is used by the X server to detect when the client
** has died while holding the drawable lock. The client sets the
** drawable lock to this value.
*/
psp->drawLockID = 1;
psp->drmMajor = drm_version->major;
psp->drmMinor = drm_version->minor;
psp->drmPatch = drm_version->patch;
psp->ddxMajor = ddx_version->major;
psp->ddxMinor = ddx_version->minor;
psp->ddxPatch = ddx_version->patch;
psp->driMajor = dri_version->major;
psp->driMinor = dri_version->minor;
psp->driPatch = dri_version->patch;
/* install driver's callback functions */
memcpy( &psp->DriverAPI, driverAPI, sizeof(struct __DriverAPIRec) );
psp->pSAREA = pSAREA;
psp->pFB = frame_buffer->base;
psp->fbSize = frame_buffer->size;
psp->fbStride = frame_buffer->stride;
psp->fbWidth = frame_buffer->width;
psp->fbHeight = frame_buffer->height;
psp->devPrivSize = frame_buffer->dev_priv_size;
psp->pDevPriv = frame_buffer->dev_priv;
psp->fbBPP = psp->fbStride * 8 / frame_buffer->width;
psp->fd = fd;
/*
** Do not init dummy context here; actual initialization will be
** done when the first DRI context is created. Init screen priv ptr
** to NULL to let CreateContext routine that it needs to be inited.
*/
psp->dummyContextPriv.driScreenPriv = NULL;
psc->destroyScreen = driDestroyScreen;
psc->createNewDrawable = driCreateNewDrawable;
psc->getDrawable = driGetDrawable;
psc->getMSC = driGetMSC;
psc->createNewContext = driCreateNewContext;
if ( (psp->DriverAPI.InitDriver != NULL)
&& !(*psp->DriverAPI.InitDriver)(psp) ) {
_mesa_free( psp );
return NULL;
}
return psp;
}
/**
* Compare the current GLX API version with a driver supplied required version.
*
* The minimum required version is compared with the API version exported by
* the \c __glXGetInternalVersion function (in libGL.so).
*
* \param required_version Minimum required internal GLX API version.
* \return A tri-value return, as from strcmp is returned. A value less
* than, equal to, or greater than zero will be returned if the
* internal GLX API version is less than, equal to, or greater
* than \c required_version.
*
* \sa __glXGetInternalVersion().
*/
2005-02-14 09:27:38 +00:00
int driCompareGLXAPIVersion( GLint required_version )
{
if ( api_ver > required_version ) {
return 1;
}
else if ( api_ver == required_version ) {
return 0;
}
return -1;
}
static int
driQueryFrameTracking( __DRInativeDisplay * dpy, void * priv,
int64_t * sbc, int64_t * missedFrames,
float * lastMissedUsage, float * usage )
{
__DRIswapInfo sInfo;
int status;
int64_t ust;
__DRIdrawablePrivate * dpriv = (__DRIdrawablePrivate *) priv;
status = dpriv->driScreenPriv->DriverAPI.GetSwapInfo( dpriv, & sInfo );
if ( status == 0 ) {
*sbc = sInfo.swap_count;
*missedFrames = sInfo.swap_missed_count;
*lastMissedUsage = sInfo.swap_missed_usage;
Fixes the glXGetProcAddress portion of the interface. Most of the functions that are currently obtained via glXGetProcAddress and all of the XF86DRI functions are replaced with a funciton table. This table will be passed to __driCreateNewScreen. One of the functions in the table is getProcAddress. This allows some loaders to expose functionality not in all loaders. This will be immediatly used for glxEnableExtension (formerly known to drivers as __glXScrEnableExtension). libGL (and in the future libglx) expose this function so that drivers can enable GLX extensions. libEGL should exposed eglEnableExtension to enable EGL extensions. The same function cannot be used for both because the extensions have different names and (possibly) different semantics. Drivers can optionally use one, both, or neither. The key parts are in the __DRIinterfaceMethodsRec structure in dri_interface.h. A pointer to one of these structures is passed into __driCreateNewScreen. Because of this, the version of the API is bumped to 20050725. Since the previous version(s) were never in a release, their existance is erased. I was actually a little surprised by how much code this cuts from the drivers. A lot of glXGetProcAddress calls disappear, and a lot of version checks go with them. Nice. The one thing I'm not sure of is removing __glXInitialize. For some reason that function was in the glXGetProcAddress table, but *nothing* in the Mesa tree used it. Did something with DRI conf. use this function? It seems odd...
2005-07-26 02:44:01 +00:00
(*dri_interface->getUST)( & ust );
*usage = driCalculateSwapUsage( dpriv, sInfo.swap_ust, ust );
}
return status;
}
/**
* Calculate amount of swap interval used between GLX buffer swaps.
*
* The usage value, on the range [0,max], is the fraction of total swap
* interval time used between GLX buffer swaps is calculated.
*
* \f$p = t_d / (i * t_r)\f$
*
* Where \f$t_d\f$ is the time since the last GLX buffer swap, \f$i\f$ is the
* swap interval (as set by \c glXSwapIntervalSGI), and \f$t_r\f$ time
* required for a single vertical refresh period (as returned by \c
* glXGetMscRateOML).
*
* See the documentation for the GLX_MESA_swap_frame_usage extension for more
* details.
*
* \param dPriv Pointer to the private drawable structure.
* \return If less than a single swap interval time period was required
* between GLX buffer swaps, a number greater than 0 and less than
* 1.0 is returned. If exactly one swap interval time period is
* required, 1.0 is returned, and if more than one is required then
* a number greater than 1.0 will be returned.
*
* \sa glXSwapIntervalSGI glXGetMscRateOML
*
* \todo Instead of caching the \c glXGetMscRateOML function pointer, would it
* be possible to cache the sync rate?
*/
float
driCalculateSwapUsage( __DRIdrawablePrivate *dPriv, int64_t last_swap_ust,
int64_t current_ust )
{
int32_t n;
int32_t d;
int interval;
float usage = 1.0;
Fixes the glXGetProcAddress portion of the interface. Most of the functions that are currently obtained via glXGetProcAddress and all of the XF86DRI functions are replaced with a funciton table. This table will be passed to __driCreateNewScreen. One of the functions in the table is getProcAddress. This allows some loaders to expose functionality not in all loaders. This will be immediatly used for glxEnableExtension (formerly known to drivers as __glXScrEnableExtension). libGL (and in the future libglx) expose this function so that drivers can enable GLX extensions. libEGL should exposed eglEnableExtension to enable EGL extensions. The same function cannot be used for both because the extensions have different names and (possibly) different semantics. Drivers can optionally use one, both, or neither. The key parts are in the __DRIinterfaceMethodsRec structure in dri_interface.h. A pointer to one of these structures is passed into __driCreateNewScreen. Because of this, the version of the API is bumped to 20050725. Since the previous version(s) were never in a release, their existance is erased. I was actually a little surprised by how much code this cuts from the drivers. A lot of glXGetProcAddress calls disappear, and a lot of version checks go with them. Nice. The one thing I'm not sure of is removing __glXInitialize. For some reason that function was in the glXGetProcAddress table, but *nothing* in the Mesa tree used it. Did something with DRI conf. use this function? It seems odd...
2005-07-26 02:44:01 +00:00
if ( (*dri_interface->getMSCRate)( dPriv->display, dPriv->draw, &n, &d ) ) {
interval = (dPriv->pdraw->swap_interval != 0)
? dPriv->pdraw->swap_interval : 1;
/* We want to calculate
* (current_UST - last_swap_UST) / (interval * us_per_refresh). We get
* current_UST by calling __glXGetUST. last_swap_UST is stored in
* dPriv->swap_ust. interval has already been calculated.
*
* The only tricky part is us_per_refresh. us_per_refresh is
* 1000000 / MSC_rate. We know the MSC_rate is n / d. We can flip it
* around and say us_per_refresh = 1000000 * d / n. Since this goes in
* the denominator of the final calculation, we calculate
* (interval * 1000000 * d) and move n into the numerator.
*/
usage = (current_ust - last_swap_ust);
usage *= n;
usage /= (interval * d);
usage /= 1000000.0;
}
return usage;
}
/*@}*/