swr/swl instructions used for word aligning were wrong. This
made memset() terribly broken. I can't imagine how it went
uncaught for soooo long. Spotted by Solomon Peachy.
I run unit tests for alignments 0,1,2,3
size 1, 2, 3, 4, 5, 63, 64, 65, 127, 128, 129;
and fill pattern 0x00 and other (since 0 is special case in this
implementation).
Change-Id: I513a10734335fe97734c10ab5a6c3e3fb3f4687a
On Classic, IRAM1 (second 128Kb of a total of 256KB available IRAM) is
slower than DRAM. Codecs that actually are using regions of IRAM1 runs
faster when DRAM is used, so IRAM1 is disabled and only IRAM0 remains
enabled: 48KB for core and 80KB for codecs/plugins.
The next test_codec results shows how decode time is decreased:
file boosted unboosted
*.ra ~1.5% ~0.5%
*.mpc ~21% ~4.5%
*.ogg ~0.5% ~0%
nero_he*.m4a ~8% ~1%
nero*.m4a ~25% ~7%
wmapro*.wma ~4.5% ~0%
wma*.wma ~25% ~7%
In addition there is a small power save when IRAM1 HW is disabled.
Change-Id: I102adee11458e82037f23076d5d5956e23235de8
For s5l8701, s5l8702 and as3525v2 targets.
The crash occurs when USB is inserted or extracted while the playlist
is being loaded or updated (it could take a few seconds for huge
playlists), at this point all buflib memory is allocated and not freed
before usb_starage_init_connection() is executed.
This workaround mitigates this panic by using static memory for USB
buffers, so this memory cannot be used for other tasks, in addition
the problem still persist when playlist load is 'paused' by USB
insertion and then updated after USB extraction.
Change-Id: Iff1db5a949361fd543e0b494924d1f2906c84b5e
Cleaned up, rebased, and forward-ported from the xvortex fork.
(original credit to vsoftster@gmail.com)
Change-Id: Ibcc023a0271ea81e901450a88317708c2683236d
Signed-off-by: Solomon Peachy <pizza@shaftnet.org>
Allow user to select cpu undervolt
There have been quite a few issues across the SANSA AMS line related
to CPU undervolting while most players show greatly increased runtime
some crash.
Rather than constanly upping the voltage we now have a
setting with a safe value for all players and the option for lower voltages
I plan to add a few other options here later such as disk
timings and maybe some other clocks/experimental settings
Added: Disk Low speed option for AS3525v2 devices cuts
frequency to 12 MHz from 24 MHz
Added: Disk Low speed option for AS3525v1 devices cuts
frequency to 15.5 MHz from 31 MHz
Added: I2c Low Speed AS3525 devices, should be bigger improvement for v1 devices
Fixed: Debug menu for AS3525v2 No SDSLOT frequency,
Showed IDE freq though it is unused
Added: DBOP and SSP underclocking affects display on v1/v2 respectively
Fixed: debug menu now has SSP frequency, and SSP_CPSR
Update: made settings menu more generic
Update: cleaned up code
Added: Clip v1 & Fuze v1 didn't have HAVE_ADJUSTABLE_CPU_VOLTAGE.
not sure why but, waiting on testing to confirm
Added: C200v2 and E200v2 devices and HAVE_ADJUSTABLE_CPU_VOLTAGE.
Fixed: v1 devices don't like display timing set lower (dbop)
v1 devices don't have a divider set for ssp (causes divide by 0)
Fixed: ClipZip display lags with Max SSP divider changed from 0xFE to 0x32
Fixed: v1 devices didn't work properly with highspeed sd cards
Added code from http://gerrit.rockbox.org/r/#/c/1704/
Added powersave and IDE interface enable/disable
Added: V2 devices now have powersave enabled on sd interface
Update: cleaned up code, lang defines, added manual entries
Update ssp clock mechanism added calculated ssp divider to clipzip
Update turn display clock off when clip+ turns off display
Fixed: clipzip wrong register for SSP clock
Change-Id: I04137682243be92f0f8d8bf1cfa54fbb1965559b
TODO: add other players?
I previously noticed that manually clearing the framebuffer while scroll
function was active caused lua to crash
I could reproduce in sim and on device but I thought using the plugin
supplied rb->lcd_clear_screen was immune to this issue
Unfortunately some devices exhibit this behavior with the plugin function
as well
This patch adds rb->lcd_scroll_stop() before lcd_clear_screen at lua start-up
and to the supplied include file lcd.lua
Change-Id: I9800145e5c834ea27df5db5f1bca50b0d40faa49
Saves 100+ bytes (50 of it in iram), saves a bit of power
Internal LCD clock decreased but with added efficiency of drawing routines
loses only around 2 Hz on the scanrate (~75Hz) while fps is slightly increased
Column offsets are now calculated outside the loops saving a few instructions
Passing a LCD_NOP command after lcd_update turns off Data/Cmd# gpio
saving a bit more power
Added a function lcd_write_cmd_triple() that allows 3 commands to be sent at once
when enabled with LCD_USE_FIFO_FOR_COMMANDS it sends them back to back without
checking FIFO status in between or sending to thhe D/C# Gpio.
Makes an assumption about the FIFO being large enough to
accept 3 commands after being emptied which should be the case on the
clipv1, clipv2, clipplus. I have only enabled it for the clip plus
as thats the only device I have to test it on.
On clip+ the SSP clock is now turned off when screen is off
Change-Id: Ib5fd24697bfe4ac8b8ee017361e789e4a7910d21
On the clip zip most debug menu items get cut off and there is
no way to read most of the debug menu items.
This patch makes the menu button scroll the text 1 character
to the right with each press and the center (select) button
re-aligns the text
Adds SSP frequency(v2) & register
Adds SD slot frequency(v2)
Change-Id: If4705d6790e25061931ca654062e22fc2e0a6f16
I2c controller needs to be enabled in order to read CSPR0, CSPR1
registers function sets CGU_I2C_AUDIO_MASTER_CLOCK_ENABLE
and only clears if it wasn't previously enabled
Use divider set in register to calculate frequency rather than
hard coded divider
Change-Id: I54ecc0c1859e906c00f4c2ae8ae2424a4619df98
Last commit was just a test to see if it work this one cleans it up
a bit and should be a bit faster
Change-Id: Ifdff5c5b78bcc6889506de607193246beccdde6b
Some devices(1-bit / 2-bit displays) have packed bit formats that
need to be unpacked in order to work on them at a pixel level.
This caused a few issues on 1 & 2-bit devices:
Greatly Oversized data arrays for bitmaps
Improper handling of native image data
Framebuffer data was near unusable without jumping through hoops
Conversion between native addressing and per pixel addressing
incurs extra overhead but it is much faster to do it
on the 'C' side rather than in lua.
Not to mention the advantage of a unified interface for the end programer
-------------------------------------------------------------------
Adds a sane way to access each pixel of image data
Adds:
--------------------------------------------------------------------
img:clear([color],[x1],[y1],[x2],[y2])
(set whole image or a portion to a particular value)
--------------------------------------------------------------------
img:invert([x1],[y1],[x2],[y2])
(inverts whole image or a portion)
--------------------------------------------------------------------
img:marshal([x1],[y1],[x2],[y2],[funct])
(calls funct for each point defined by rect of x1,y1 x2,y2
returns value and allows setting value of each point return
nil to terminate early)
--------------------------------------------------------------------
img:points([x1],[y1],[x2],[y2],[dx],[dy])
(returns iterator function that steps delta-x and delta-y pixels each call
returns value of pixel each call but doesn't allow setting to a new value
compare to lua pairs method)
--------------------------------------------------------------------
img:copy(src,[x1],[y1],[x2],[y2],[w],[h],[clip][operation][clr/funct])
(copies all or part of an image -- straight copy or special ops
optionally calls funct for each point defined by rect of
x1, y1, w, h and x2, y2, w, h for dest and src images
returns value of dst and src and allows setting value of
each point return nil to terminate early)
--------------------------------------------------------------------
img:line(x1, y1, x2, y2, color)
--------------------------------------------------------------------
img:ellipse(x1, y1, x2, y2, color, [fillcolor]
--------------------------------------------------------------------
Fixed handling of 2-bit vertical integrated screens
Added direct element access for saving / restoring native image etc.
Added more data to tostring() handler and a way to access individual items
Added equals method to see if two variables reference the same image address
(doesn't check if two separate images contain the same 'picture')
Optimized get and set routines
Fixed out of bound x coord access shifting to next line
Added lua include files to expose new functionality
Finished image saving routine
Static allocation of set_viewport struct faster + saves ram over dynamic
Cleaned up code
Fixed pixel get/set for 1/2 bit devices
Fixed handling for 24-bit devices (32?)
-------------------------------------------------------------------------
Example lua script to follow on forums
-------------------------------------------------------------------------
Change-Id: I8a9ff0ff72aacf4b1662767ccb2b312fc355239c
Some devices(1-bit / 2-bit displays) have packed bit formats that
need to be unpacked in order to work on them at a pixel level.
This caused a few issues on 1 & 2-bit devices:
Greatly Oversized data arrays for bitmaps
Improper handling of native image data
Framebuffer data was near unusable without jumping through hoops
Conversion between native addressing and per pixel addressing
incurs extra overhead but it is much faster to do it
on the 'C' side rather than in lua.
Not to mention the advantage of a unified interface for the end programer
-------------------------------------------------------------------
Adds a sane way to access each pixel of image data
Adds:
--------------------------------------------------------------------
img:clear([color],[x1],[y1],[x2],[y2])
(set whole image or a portion to a particular value)
--------------------------------------------------------------------
img:invert([x1],[y1],[x2],[y2])
(inverts whole image or a portion)
--------------------------------------------------------------------
img:marshal([x1],[y1],[x2],[y2],[funct])
(calls funct for each point defined by rect of x1,y1 x2,y2
returns value and allows setting value of each point return
nil to terminate early)
--------------------------------------------------------------------
img:points([x1],[y1],[x2],[y2],[dx],[dy])
(returns iterator function that steps delta-x and delta-y pixels each call
returns value of pixel each call but doesn't allow setting to a new value
compare to lua pairs method)
--------------------------------------------------------------------
img:copy(src,[x1],[y1],[x2],[y2],[w],[h],[clip][operation][clr/funct])
(copies all or part of an image -- straight copy or special ops
optionally calls funct for each point defined by rect of
x1, y1, w, h and x2, y2, w, h for dest and src images
returns value of dst and src and allows setting value of
each point return nil to terminate early)
--------------------------------------------------------------------
img:line(x1, y1, x2, y2, color)
--------------------------------------------------------------------
img:ellipse(x1, y1, x2, y2, color, [fillcolor]
--------------------------------------------------------------------
Fixed handling of 2-bit vertical integrated screens
Added direct element access for saving / restoring native image etc.
Added more data to tostring() handler and a way to access individual items
Added equals method to see if two variables reference the same image address
(doesn't check if two separate images contain the same 'picture')
Optimized get and set routines
Fixed out of bound x coord access shifting to next line
Added lua include files to expose new functionality
Finished image saving routine
Static allocation of set_viewport struct faster + saves ram over dynamic
Cleaned up code
Fixed pixel get/set for 1/2 bit devices
-------------------------------------------------------------------------
Example lua script to follow on forums
-------------------------------------------------------------------------
Change-Id: I7b9c1fd699442fb683760f781021091786c18509
Also removes the sd_enable() function call. It was only used in
the debug screen on AMSv1 and not used at all on AMS v2.
For v1,obtain debug info in a struture passed to a dedicated
debug info function so that enabling and disabling the controller
isn't racy.
Change-Id: I7c44693bc2df5a1f16168b05b3abfe622f9584ce
This brings the source to upstream commit 506b073 (though I have made some
extra commits on top of that). Notably this includes a fix for a double-free
bug that I myself introduced upstream.
Change-Id: I02671586bbc34d63e05398ee971271fed42538cf
