Re: Screen snapshot example code posted
Re: Screen snapshot example code posted
- Subject: Re: Screen snapshot example code posted
- From: Mike Paquette <email@hidden>
- Date: Sat, 13 Aug 2005 11:46:46 -0700
On Aug 13, 2005, at 9:10 AM, Ben Haller wrote:
On the flip side, it's probably slow compared to what the core
video system could do itself; reading pixels one by one is not as
fast as getting down-and-dirty with cacheline reads and such, and
certainly isn't as fast as copying from one buffer on the video
card to another without involving the CPU at all, which is what the
system perhaps does (or should do, or someday will do) on at least
some accelerated video cards.
Absolutely! The best way to handle the abstraction of the hardware,
and get good performance, is probably to go with the OpenGL
software. Now, using OpenGL isn't exactly obvious, though, so some
sample code is in order.
Finally, it's uncomfortably dependent upon the layout of video
memory; Josh's code only works for 16-bit and 32-bit color, and
only with interleaved (non-planar) data, which is fine for now
AFAIK, but Apple has made no guarantees about sticking with only
those memory layouts in the future that I am aware of. It's
possible right now to switch the screen to 8-bit mode, and many
games do, but Josh's code wouldn't be able to take a snapshot of
the screen in that state, while my CopyBits-base code probably
would (haven't tried). And so forth.
Again, OpenGL can help here, as it includes the ability to convert
pixel formats, and can even deal with hardware-specific pixel reading
problems such as microtiling, or getting the correct framebuffer when
full screen buffer swapping (the GL version of page flipping) is in
effect. OpenGL will also use a DMA transfer for reading the
framebuffer content, where available.
OpenGL provides both a synchronous read path suitable for simple
screen captures using glReadPixels, and a fast asynchronous path
using the texture read mechanism, which is a bit more complex, but
works well for repeated capture operations, such as screencasting or
recording a movie from the screen.
Here's a simple mechanism that can be used to read a rectangle from a
display to a CGImageRef. It's more comment than code:
/*
* glGrab.c
*/
/*
IMPORTANT: This Apple software is supplied to you by Apple
Computer, Inc. ("Apple") in
consideration of your agreement to the following terms, and your
use, installation or modification
of this Apple software constitutes acceptance of these terms. If
you do not agree with these terms,
please do not use, install or modify this Apple software.
In consideration of your agreement to abide by the following terms,
and subject to these
terms, Apple grants you a personal, non-exclusive license, under
Apple\xd5s copyrights in
this original Apple software (the "Apple Software"), to use and
modify the Apple Software,
with or without modifications, in source and/or binary forms;
provided that if you redistribute
the Apple Software in its entirety and without modifications, you
must retain this notice and
the following text and disclaimers in all such redistributions of
the Apple Software.
Neither the name, trademarks, service marks or logos of Apple
Computer, Inc. may be used
to endorse or promote products derived from the Apple Software
without specific prior
written permission from Apple. Except as expressly tated in this
notice, no other rights or
licenses, express or implied, are granted by Apple herein,
including but not limited
to any patent rights that may be infringed by your derivative works
or by other works
in which the Apple Software may be incorporated.
The Apple Software is provided by Apple on an "AS IS" basis. APPLE
MAKES NO WARRANTIES,
EXPRESS OR IMPLIED, INCLUDING WITHOUT LIMITATION THE IMPLIED
WARRANTIES OF NON-INFRINGEMENT,
MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE, REGARDING THE
APPLE SOFTWARE OR ITS
USE AND OPERATION ALONE OR IN COMBINATION WITH YOUR PRODUCTS.
IN NO EVENT SHALL APPLE BE LIABLE FOR ANY SPECIAL, INDIRECT,
INCIDENTAL OR CONSEQUENTIAL
DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE
GOODS OR SERVICES; LOSS
OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) ARISING IN ANY
WAY OUT OF THE USE,
REPRODUCTION, MODIFICATION AND/OR DISTRIBUTION OF THE APPLE
SOFTWARE, HOWEVER CAUSED AND
WHETHER UNDER THEORY OF CONTRACT, TORT (INCLUDING NEGLIGENCE),
STRICT LIABILITY OR
OTHERWISE, EVEN IF APPLE HAS BEEN ADVISED OF THE POSSIBILITY OF
SUCH DAMAGE.
*/
#include <CoreFoundation/CoreFoundation.h>
#include <ApplicationServices/ApplicationServices.h>
#include <OpenGL/OpenGL.h>
#include <OpenGL/gl.h>
/*
* perform an in-place swap from Quadrant 1 to Quadrant III format
* (upside-down PostScript/GL to right side up QD/CG raster format)
* We do this in-place, which requires more copying, but will touch
* only half the pages. (Display grabs are BIG!)
*
* Pixel reformatting may optionally be done here if needed.
*/
static void swizzleBitmap(void * data, int rowBytes, int height)
{
int top, bottom;
void * buffer;
void * topP;
void * bottomP;
void * base;
top = 0;
bottom = height - 1;
base = data;
buffer = malloc(rowBytes);
while ( top < bottom )
{
topP = (void *)((top * rowBytes) + (intptr_t)base);
bottomP = (void *)((bottom * rowBytes) + (intptr_t)base);
/*
* Save and swap scanlines.
*
* This code does a simple in-place exchange with a temp
buffer.
* If you need to reformat the pixels, replace the first two
bcopy()
* calls with your own custom pixel reformatter.
*/
bcopy( topP, buffer, rowBytes );
bcopy( bottomP, topP, rowBytes );
bcopy( buffer, bottomP, rowBytes );
++top;
--bottom;
}
free( buffer );
}
/*
* Given a display ID and a rectangle on that display, generate a
CGImageRef
* containing the display contents.
*
* srcRect is display-origin relative.
*
* This function uses a full screen OpenGL read-only context.
* By using OpenGL, we can read the screen using a DMA transfer
* when it's in millions of colors mode, and we can correctly read
* a microtiled full screen OpenGL context, such as a game or full
* screen video display.
*
* Returns a CGImageRef. When you are done with the CGImageRef,
release it
* using CFRelease().
* Returns NULL on an error.
*/
CGImageRef grabViaOpenGL(CGDirectDisplayID display, CGRect srcRect)
{
CGContextRef bitmap;
CGImageRef image;
void * data;
long bytewidth;
GLint width, height;
long bytes;
CGColorSpaceRef cSpace = CGColorSpaceCreateWithName
(kCGColorSpaceGenericRGB);
CGLContextObj glContextObj;
CGLPixelFormatObj pixelFormatObj ;
long numPixelFormats ;
CGLPixelFormatAttribute attribs[] =
{
kCGLPFAFullScreen,
kCGLPFADisplayMask,
0, /* Display mask bit goes here */
0
} ;
if ( display == kCGNullDirectDisplay )
display = CGMainDisplayID();
attribs[2] = CGDisplayIDToOpenGLDisplayMask(display);
/* Build a full-screen GL context */
CGLChoosePixelFormat( attribs, &pixelFormatObj, &numPixelFormats );
if ( pixelFormatObj == NULL ) // No full screen context support
return NULL;
CGLCreateContext( pixelFormatObj, NULL, &glContextObj ) ;
CGLDestroyPixelFormat( pixelFormatObj ) ;
if ( glContextObj == NULL )
return NULL;
CGLSetCurrentContext( glContextObj ) ;
CGLSetFullScreen( glContextObj ) ;
glReadBuffer(GL_FRONT);
width = srcRect.size.width;
height = srcRect.size.height;
bytewidth = width * 4; // Assume 4 bytes/pixel
for now
bytewidth = (bytewidth + 3) & ~3; // Align to 4 bytes
bytes = bytewidth * height; // width * height
/* Build bitmap context */
data = malloc(height * bytewidth);
if ( data == NULL )
{
CGLSetCurrentContext( NULL );
CGLClearDrawable( glContextObj ); // disassociate from
full screen
CGLDestroyContext( glContextObj ); // and destroy the
context
return NULL;
}
bitmap = CGBitmapContextCreate(data, width, height, 8, bytewidth,
cSpace,
kCGImageAlphaNoneSkipFirst /* XRGB */);
CFRelease(cSpace);
/* Read framebuffer into our bitmap */
glFinish(); /* Finish all OpenGL commands */
glPixelStorei(GL_PACK_ALIGNMENT, 4); /* Force 4-byte
alignment */
glPixelStorei(GL_PACK_ROW_LENGTH, 0);
glPixelStorei(GL_PACK_SKIP_ROWS, 0);
glPixelStorei(GL_PACK_SKIP_PIXELS, 0);
/*
* Fetch the data in XRGB format, matching the bitmap context.
*/
glReadPixels((GLint)srcRect.origin.x, (GLint)srcRect.origin.y,
width, height,
GL_BGRA,
GL_UNSIGNED_INT_8_8_8_8_REV,
data);
/*
* glReadPixels generates a quadrant I raster, with origin in
the lower left
* This isn't a problem for signal processing routines such as
compressors,
* as they can simply use a negative 'advance' to move between
scanlines.
* CGImageRef and CGBitmapContext assume a quadrant III raster,
though, so we need to
* invert it. Pixel reformatting can also be done here.
*/
swizzleBitmap(data, bytewidth, height);
/* Make an image out of our bitmap; does a cheap vm_copy of the
bitmap */
image = CGBitmapContextCreateImage(bitmap);
/* Get rid of bitmap */
CFRelease(bitmap);
free(data);
/* Get rid of GL context */
CGLSetCurrentContext( NULL );
CGLClearDrawable( glContextObj ); // disassociate from full
screen
CGLDestroyContext( glContextObj ); // and destroy the context
/* Returned image has a reference count of 1 */
return image;
}
_______________________________________________
Do not post admin requests to the list. They will be ignored.
Cocoa-dev mailing list (email@hidden)
Help/Unsubscribe/Update your Subscription:
This email sent to email@hidden