specifying whether only the exact supplied address is acceptable. If
false, the address is considered a hint only. It will be picked, if
available, otherwise a greater address is tried to be acquired, and
as last resort any address. This feature is only implemented for PPC.
It is needed since the preferred kernel text base address 0x80000000
might not be available (and actually isn't on my Mac mini).
* Fixed a bug in the PPC memory management code:
is_{virtual,physical}_allocated() were checking whether the given
range was completely contained by an existing range instead of
checking for intersection. As a consequence we could (and did) allocate
a range intersecting with already allocated ranges. The kernel segment
thus overwrote OF memory for instance.
* The ELF loader makes sure that it got both text and data segment of
the image to be loaded.
The PPC boot loader successfully loads kernel and modules now. Next
comes the hard part, I'm afraid.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@15708 a95241bf-73f2-0310-859d-f6bbb57e9c96
platform_allocation_region(). Since the callback stuff doesn't seem to
work, the Open Firmware didn't ask us to allocate memory in turn, and
thus we didn't know about the range that had been mapped and reused it
later.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@15699 a95241bf-73f2-0310-859d-f6bbb57e9c96
into a new frame_buffer_support.h
* added blend32 routine for blending a certain color with
a scanline in the frame buffer
* added "solid" versions of B_OP_ALPHA drawing with
B_ALPHA_OVERLAY alha function (blending on top of
a non-transparent background such as the frame buffer)
* implemented an optimized shortcut for alpha blended
FillRect() in Painter
* used the "packed" version of scanlines for shapes with
an outline thicker than 4 pixels (and filled shapes anyways),
this improves drawing speed when there are a few anti-aliased
pixels at the beginning of a scanline, then a solid fill and
some anti-aliased pixels at the end of the scanline. Such as
large letters.
To summarize: The alpha blending in Painter seems to be about
1.45 times faster than on BeOS R5 which benefits drawing large
shapes. For example, drawing a large alpha blended rounded rect
is 1.28 times faster on the Haiku app_server. On the other hand,
B_OP_COPY is quite tough to beat. It is currently 10 times faster
on R5. But a great deal seems to be caused by the Painter
rasterization algorithm itself, since commenting out the actual
drawing doesn't gain any speed.
The other useful experience I collected was that reading and
writing and over the PCI bus in the same loop really hurts
performance. It is actually faster (like 1.8 times!!) to allocate
a second buffer, read from frame buffer into that, doing the
blending at the same time, then writing the buffer back to the
screen.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@15698 a95241bf-73f2-0310-859d-f6bbb57e9c96
and thus, it could be used again which led to a crash.
Changed the free ranges list from a singly linked list to a doubly linked
list so that not all free ranges have to be searched for the one to be
freed anymore.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@15696 a95241bf-73f2-0310-859d-f6bbb57e9c96
the modified bit set were written back (that's the case when they weren't put into
the modified pages queue yet), they were enqueued into the active queue twice, messing
up the page lists with various effects - this little thing took me busy for the last
few days :-/
* Improved the "page" debugger command: it now prints out much more info, and also lets
you lookup pages via physical address.
* The dump commands are now using kprintf() instead of dprintf().
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@15692 a95241bf-73f2-0310-859d-f6bbb57e9c96
IP, and UDP, as well as a home brewn UDP based protocol, "remote disk",
which provides random access to a single remote file/device. The Open
Firmware flavored boot loader automatically initializes the net stack,
searches for a remote disk, and tries to boot from it, if the boot
device is a network device (e.g. when loading the boot loader via
TFTP).
This is quite nice for developing with a two-machine setup, since one
doesn't even need to install Haiku on the test machine anymore, but can
serve it directly from the development machine. When the networking
support in the kernel is working, this method could even be used to
fully boot, not just for loading kernel and initial modules.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@15689 a95241bf-73f2-0310-859d-f6bbb57e9c96
argument value of 0 should only free the old allocation.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@15688 a95241bf-73f2-0310-859d-f6bbb57e9c96
* bfs_inode must be packed, otherwise the structure is larger due to
different alignment requirements.
* The off_t array returned by bfs_inode::Values() is in on-disk
endianess. This was ignored in some places. Hope I squashed all
occurences.
Those fixes should probably be back-ported to the kernel add-on, too.
Will do that when I get to that point in the kernel, unless someone
beats me to it.
The PPC boot loader does at least find the kernel on a BFS volume now.
Haven't dared to let it be loaded yet.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@15686 a95241bf-73f2-0310-859d-f6bbb57e9c96
arch_cpu_user_TLB_invalidate(). Empty at the moment; will have a look at
that later. The PPC kernel builds again, now.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@15685 a95241bf-73f2-0310-859d-f6bbb57e9c96
of_call_method() which is actually an implementation of the
"call-method" client function.
* of_interpret() announced one less return value than actually needed.
Seemed to work anyway, though.
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@15683 a95241bf-73f2-0310-859d-f6bbb57e9c96
also added a few comments explaining what's going on. Japanese not updated?
git-svn-id: file:///srv/svn/repos/haiku/haiku/trunk@15667 a95241bf-73f2-0310-859d-f6bbb57e9c96
