User talk:Masterzorag: Difference between revisions
Masterzorag (talk | contribs) No edit summary |
Masterzorag (talk | contribs) (tested OpenCL on 3.2.55) |
||
Line 8: | Line 8: | ||
Latest tested kernels: | Latest tested kernels: | ||
* 3.2. | * 3.2.55 works fine | ||
<pre> | <pre> | ||
# ./perlin | # ./perlin |
Revision as of 17:16, 31 March 2014
As far as I know, I'm the only coding OpenCL on the Cell here, if someone want to test something be warned that due some spufs changes that ppc-kernel-devs are (maybe) trying to fix, latest 3.3/3.4/3.5 branches falls into 'possible circular locking dependency detected' and slowdown runtime.
- It's stable until 3.2 branch.
- Even disabling lock debugging it slowdowns without warnings, it happens even with OpenCL samples from IBM.
http://permalink.gmane.org/gmane.linux.ports.ppc.embedded/50547
Latest tested kernels:
- 3.2.55 works fine
# ./perlin OpenCL took 22.496168 seconds to compute 1000 frames. Pixel Rate = 46.611316 Mpixels/sec, Frame Rate = 44.452015 frames/sec Host code took 12.620616 seconds to compute 10 frames. Pixel Rate = 0.830844 Mpixels/sec, Frame Rate = 0.792354 frames/sec OpenCL provided a 56.101182 speedup
- 3.3.3/3.4.6/3.5.3 falls into 'possible circular locking dependency detected' and slowdown runtime
Here the slowdown effect:
# ./perlin OpenCL took 93.280273 seconds to compute 1000 frames. Pixel Rate = 11.241133 Mpixels/sec, Frame Rate = 10.720380 frames/sec Host code took 12.948244 seconds to compute 10 frames. Pixel Rate = 0.809821 Mpixels/sec, Frame Rate = 0.772305 frames/sec OpenCL provided a 13.881010 speedup
In this specific case time spent is 4x to do the same thing!
When program runs something is going weird, e.g. in my program I'm used to query an OpenCL builtin function to tell me how many available SPEs there are, and its reply 8.
Using spu_base.enum_shared=1 parameter it should reply 7, so seems that the issue is OpenCL related.
OtherOS region
OtherOS/OtherOS++ region is on HDD (ps3dd), we have new linux tools (ps3sed) and drivers.
To resize ps3da I've tried new ps3sed (manually), unsuccesfully: GameOS always detect corruption and redo its own things.
I've found a way to force resize on 4.46, no emer_init patch, no downgrading: GameOS respect standards.
I can now resize ps3da at arbitrary size.
Here I've forced ps3da to use 1216709344 sectors, this make me space for a ps3dd of about 16G.
After that GameOS do it own things, I've resetted ps3da to its real geometry (1250263728) and booted a new petitboot.
root@ps3-linux:~# dmesg | grep ps3disk [ 3.220526] ps3disk_init:601: registered block device major 254 [ 3.220549] ps3_system_bus_match:369: dev=6.0(sb_04), drv=6.0(ps3disk): match [ 3.220856] ps3disk sb_04: accessible region 0 start 0 size 1250263728 [ 3.220952] ps3disk sb_04: accessible region 1 start 32 size 1212515008 [ 3.221045] ps3disk sb_04: accessible region 2 start 1212515040 size 4194296 [ 3.221051] ps3disk sb_04: ps3stor_probe_access:133: 3 accessible regions found [ 3.227341] ps3disk sb_04: ps3da is a SAMSUNG HM641JI (610480 MiB total, 610480 MiB region) [ 3.229035] ps3disk sb_04: ps3db is a SAMSUNG HM641JI (610480 MiB total, 592048 MiB region) [ 3.230008] ps3disk sb_04: ps3dc is a SAMSUNG HM641JI (610480 MiB total, 2047 MiB region) root@ps3-linux:~# ps3sed print_region 3 0 0 1250263728 1 1 32 1212515008 8 2 1212515040 4194296 8 root@ps3-linux:~# create_hdd_region.sh INFO: device id 3 INFO: number of regions 3 INFO: total number of blocks 1250263728 INFO: last region start block 1212515040 INFO: last region number of blocks 4194296 INFO: new region start block 1216709344 INFO: new region number of blocks 33554376 INFO: new region id 3 root@ps3-linux:~# ps3sed print_region 3 0 0 1250263728 1 1 32 1212515008 8 2 1212515040 4194296 8 3 1216709344 33554376 1 root@ps3-linux:~# reboot && exit
Last number 1 is wrong, it says that last region has only one acl entry, we need to fix it at 8 entries:
- manually with ps3sed
- rebooting
Petitboot reboot finally detect a new ps3dd device, the fourth region, of (33554376 * 512 =) 17179840512 bytes.
All of this with a 3.10.26 kernel and new tools: no vflash hacking involved, linux on vflash7 is deprecated.
ps3vuart-tools
We miss some stuff from old ps3sm-utils, looking to port: temperature, get_tzpb, get_fan_policy and set_fan_policy to new ps3vuart-tools.
We need to enable some sort of fan control on petitboot now.