Commit Graph

159708 Commits

Author SHA1 Message Date
Matt Sealey
c84310addb cpufreq: allow building without cpu frequency drivers 2011-09-09 16:49:21 -05:00
Matt Sealey
e5fe366470 gpu: maximum apertures = 2 now. clarify flags debug (multiple ways -> %x) 2011-09-05 12:11:03 -05:00
Matt Sealey
ef06068ea1 gpu: correct varargs debug (memdesc is a pointer NOT static!) to fix ridiculous sizes reported for sizebytes/offsetbytes in read/write operations.
Also add scatterlist and gsl_property_t debugging so the thing is more readable as a whole..
2011-09-04 15:02:04 -05:00
Matt Sealey
2c2be6f567 gpu: move files out of common/ since there's no need for a seperate directory. This also means we can make drivers/mxc/amd-gpu/gsl_*.i to see preprocessed output and get hints as to what is being compiled in and what the final code is like
additional minor changes:
 * rename PAGE_SIZE register MH_PAGE_SIZE in yamato/22/yamato_genreg.h (although this is an unused debug file so far)
 * remove references to stdio files in gsl_debug_pm4.c (will come in handy later)
2011-08-31 11:22:26 -05:00
Matt Sealey
13678b44c9 gpu: disabling the clip mask should improve performance but it actually causes the depth buffer to act weird because I don't think our userspace/kernel combination handles it properly.. 2011-08-30 15:27:37 -05:00
Matt Sealey
bc30f49241 gpu: don't clear the Z160 global on context destroy, do it on device close instead (this is part of a Freescale patch ENGR00142363 for WindowsCE but makes a hell of a lot o srense for Linux too) 2011-08-30 13:33:11 -05:00
Matt Sealey
c9f5504a1a gpu: enable HAL debug
gpu: pull in some command stream differences from Qualcomm's MSM kernel (might fix some intermittent GPU hangs)
display: enable quick testing of non-mmu operation (hangs X right now)
2011-08-30 12:38:48 -05:00
Matt Sealey
cdd8d1a03b gpu: move hal abstraction to the same place as the other files 2011-08-30 08:15:24 -05:00
Matt Sealey
aac2fc75d3 gpu driver: remove most of the remaining "abstraction" for OS mutex and so on
* mutex, events, sleeps, process id unabstracted
 * defines for function types (inline, extern) unabstracted
 * defines for debug assert change name (KOS_ASSERT -> DEBUG_ASSERT) and left disabled
 * driver now warns, if logging is enabled, that interruptible mutex locks are interrupted (which never happens in reality but it fixes a warning)
 * remove fine grained locking "feature" since it's not enabled (and Qualcomm's public driver does not even implement it)
 * refine mutex initialization where it's still present - don't try and allocate mutex structures where they're always used, just make them parts of the structure (struct mutex * -> struct mutex)
 * whitespace police
2011-08-29 14:55:29 -05:00
Matt Sealey
63da19a022 gpu os abstraction: more unused functions for memory leak checking now gone 2011-08-26 13:02:03 -05:00
Matt Sealey
9d43088ce9 gpu os abstraction: kos_malloc/kos_free -> kmalloc/kfree. Remove SYMBIAN and WIN32 code 2011-08-26 12:58:10 -05:00
Matt Sealey
f1a9436eac gpu os abstraction: remove _LINUX switches and all other OS code 2011-08-26 12:02:51 -05:00
Matt Sealey
8aff22cac1 gpu os abstraction: remove some wince/symbian/brew stuff and atoi/strtoul abstractions (unused) 2011-08-26 11:16:25 -05:00
Matt Sealey
64b3530cc9 gpu os abstraction: kos_thread_create/destroy was only used for !LINUX, so just get rid of it 2011-08-26 11:08:56 -05:00
Matt Sealey
a0adc5279b __int64 custom typedef --> standard __s64 linux/type.h 2011-08-26 11:01:22 -05:00
Matt Sealey
b45c20854c gpu os abstraction: remove stats api (unused) 2011-08-26 10:52:21 -05:00
Matt Sealey
72edabd161 gpu os abstraction: strcpy, strncpy, strcat, strcmp, strncmp, strlen all using Linux ones now 2011-08-26 10:40:35 -05:00
Matt Sealey
b564e29f61 gpu os abstraction: remove physical memory API (unused) 2011-08-26 10:37:02 -05:00
Matt Sealey
90a20cdd1e gpu os abstraction: missed an include for msleep 2011-08-26 10:34:01 -05:00
Matt Sealey
f2b2a01e64 gpu os abstraction: remove redundant cross-process memory allocators (makes no sense in a kernel and not used anyway) 2011-08-26 10:32:41 -05:00
Matt Sealey
92212608f6 gpu os abstraction: kos_sleep -> msleep 2011-08-26 10:29:29 -05:00
Matt Sealey
55395acb61 gpu os abstraction: remove unused functions for system information 2011-08-26 10:27:46 -05:00
Matt Sealey
263baa8817 gpu os abstraction: start with memset, memcmp, memcpy as low hanging fruit 2011-08-26 10:23:31 -05:00
Matt Sealey
b9cf657c71 gpu: logging abstraction cleanups. don't re-use PAGE_SIZE in mmu block definition (rename to MH_PAGE_SIZE) 2011-08-25 13:09:21 -05:00
Matt Sealey
0ce707fff2 gpu logging: fix more unsigned int -> uint (and %d -> %u), fix a few value debug mishaps (0x08x -> 0x%08x) 2011-08-25 12:46:25 -05:00
Matt Sealey
863664804f gsl memory manager: whitespace policing mostly, break pooled functions out for better readability (not compiled yet). Some other stuff that isn't important right now. 2011-08-25 11:06:24 -05:00
Matt Sealey
9876e46062 fb: scale triple buffers down to double buffers to reduce likelihood of extra memory allocations 2011-08-25 10:49:22 -05:00
Matt Sealey
d82a8f4074 bump reserved memory for all platforms to 16MB - reduction of 16MB from 'top, increase of 8MB from 'book. 16MB is enough to hold a double buffered 1080p
screen at 32bit. We will simply drop from 3*yres to 2*yres if we're pushing the limit in the framebuffer driver.
2011-08-25 10:37:01 -05:00
Matt Sealey
76589e9bac GPU debug: pid=%d -> pid=0x%08x, device_id=%d -> %D, for coherent debug. Also shortened some "unsigned int" markings to "uint" to be more concise. 2011-08-25 09:42:46 -05:00
Matt Sealey
2b537c8496 amd-gpu: extremely chatty logging, add CONFIG_MXC_AMD_GPU_LOGGING=y to config to enable. 2011-08-24 17:30:20 -05:00
Matt Sealey
f64e1efb63 disable hal debug output since it's not needed right now 2011-08-11 09:28:44 -05:00
Matt Sealey
a0ae9134ba Reduce number of apertures needed for GPU memory access by refining aperture setup to use PHYS and MMU/EMEM
apertures instead of PHYS and EMEM1/MMU and EMEM2. Modify setup to utilize all of reserved memory for PHYS
aperture instead of splitting. Now, a 32MB (in platform fixup, mx51_efikamx.c) reserved block is utilized as
the PHYS, or an 8MB chunk is carved out of coherent memory space (DMA allocation) if reservation is disabled.
Normal memory usage for a desktop with browser, office and media player is well over 64MB but the Xorg driver
will evict unused pixmaps where necessary to keep the current performance pixmaps in place.
2011-08-11 09:22:55 -05:00
Matt Sealey
c5de0bc6cf refine GSL memory sizes in reserved region to 24MB for physical and 8MB other 2011-08-04 16:17:02 -05:00
Matt Sealey
fe784e276c Bump to .25 for release-prep 2011-08-02 19:32:44 -05:00
Matt Sealey
a940a406a3 Reinstate the 32MB reserved area for GPU since DMA allocations of large sizes just don't work.
Correct EMEM2 and PHYS area sizes for GPU to 16MB each (may need refinement)
Bump max zone order anyway
2011-08-02 17:08:32 -05:00
Matt Sealey
e91d512070 whitespace police 2011-08-01 08:24:37 -05:00
Matt Sealey
ca17b53a2d fb: re-enable second framebuffer 2011-08-01 08:22:02 -05:00
Matt Sealey
099ebbe1f9 fb: only change alignment for non-overlay vars. fixes stride issues on v4lsink. 2011-07-30 16:33:10 -05:00
Matt Sealey
be72671768 siihdmi: when GNOME decides that it should "put display to sleep" it sends the POWERDOWN blank type and we were dutifully powering the SII9022 down to D2 state. Unfortunately without resetting the chip there's very little way to bring it back (and certainly "unblank" does not do what is necessary to do so). So, just live with a standard screen blank for now. 2011-07-29 18:15:03 -05:00
Matt Sealey
5f019c06ae fb: map reserved memory as writecombine to match non-reserved allocations 2011-07-29 12:45:05 -05:00
Kirill A. Shutemov
eca801dcf0 ARM: 5701/1: ARM: copy_page.S: take into account the size of the cache line
Optimized version of copy_page() was written with assumption that cache
line size is 32 bytes. On Cortex-A8 cache line size is 64 bytes.

This patch tries to generalize copy_page() to work with any cache line
size if cache line size is multiple of 16 and page size is multiple of
two cache line size.

After this optimization we've got ~25% speedup on OMAP3(tested in
userspace).

There is test for kernelspace which trigger copy-on-write after fork():

 #include <stdlib.h>
 #include <string.h>
 #include <unistd.h>

 #define BUF_SIZE (10000*4096)
 #define NFORK 200

 int main(int argc, char **argv)
 {
         char *buf = malloc(BUF_SIZE);
         int i;

         memset(buf, 0, BUF_SIZE);

         for(i = 0; i < NFORK; i++) {
                 if (fork()) {
                         wait(NULL);
                 } else {
                         int j;

                         for(j = 0; j < BUF_SIZE; j+= 4096)
                                 buf[j] = (j & 0xFF) + 1;
                         break;
                 }
         }

         free(buf);
         return 0;
 }

Before optimization this test takes ~66 seconds, after optimization
takes ~56 seconds.

Signed-off-by: Siarhei Siamashka <siarhei.siamashka@nokia.com>
Signed-off-by: Kirill A. Shutemov <kirill@shutemov.name>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2011-07-28 19:36:56 -05:00
Laura Abbott
721fc6f779 ARM: 6945/1: Add unwinding support for division functions
The software division functions never had unwinding annotations
added. Currently, when a division by zero occurs the backtrace shown
will stop at Ldiv0 or some completely unrelated function. Add
unwinding annotations in hopes of getting a more useful backtrace
when a division by zero occurs.

Signed-off-by: Laura Abbott <lauraa@codeaurora.org>
Acked-by: Dave Martin <dave.martin@linaro.org>
Signed-off-by: Russell King <rmk+kernel@arm.linux.org.uk>
2011-07-28 19:28:06 -05:00
Matt Sealey
272436b5e1 siihdmi: don't turn the chip off just because there's no receiver sense for quirky monitors like the Samsung SyncMaster E1920 2011-07-28 19:03:22 -05:00
Matt Sealey
8bf0ded04f gpu: use ilog2 instead of custom int-to-float messing around (backported from Qualcomm's GSL) 2011-07-28 19:02:50 -05:00
Matt Sealey
4ffd877f1b display: go back to reserving a framebuffer so the base address doesn't move around on boot and make X go crazy.
8MB on Smartbook, 32MB on Smarttop. Dynamically allocating the framebuffer for 1920x1080-32 didn't work with the
dynamic method anyway (sigh) due to memory fragmentation.
Also, align framebuffer stride to 32 bytes to fix Z430 acceleration in X
2011-07-28 17:59:54 -05:00
Matt Sealey
46cf9c5817 siihdmi: don't use a timeout to enter the hotplug event handler, fixes some twitchy displays (Samsung SyncMaster E1920 in particular) 2011-07-28 17:58:50 -05:00
Matt Sealey
e507d6b161 Revert "Revert "bump to .24 for release""
This reverts commit e8d079d4a5.
2011-07-20 19:50:22 -05:00
Matt Sealey
cc8deb53f4 Revert "ENGR00131389: remove "area->vm_ops->open" calling in imx-pcm.c to support IRAM playback"
This reverts commit 4dda9e941d.
2011-07-20 19:49:47 -05:00
Matt Sealey
e8d079d4a5 Revert "bump to .24 for release"
This reverts commit 89ac8f6c13.
2011-07-20 19:49:22 -05:00
Matt Sealey
89ac8f6c13 bump to .24 for release 2011-07-15 19:50:05 -05:00