AMD Evergreen Compute Support Lands Mainline

[agd5f]

The code supports EVERGREEN (radeon HD 5xxx).
You have NORTHERN ISLANDS (radeon HD 6xxx).

The code supports both evergreen and northern islands asics with the exception of cayman and trinity. Tom has preliminary code for those as well, but it hasn't been pushed yet.

[droidhacker]

I build all with this instruction, but all opencl-examples shows me

 /home/downloads/git/opencl-example/hello_world 
radeon: Failed to get PCI ID, error number -13
There are 1 platforms.
[B]clGetDeviceIDs() failed: CL_DEVICE_NOT_FOUND[/B]

I use hd6770 (juniper).
In what may be the problem?

libclc has same warnings at build time

LINK utils/prepare-builtins
LLVM-CC r600--/lib/workitem/get_global_id
LLVM-CC r600--/lib/workitem/get_global_size.cl.bc
LLVM-CC r600--/lib/convert.cl.bc
./generic/lib/convert.cl:69:19: warning: shift count >= width of type [-Wshift-count-overflow]
  return l > ((1UL<<32)-1) ? ((1UL<<32)-1) : l;
                  ^ ~~
./generic/lib/convert.cl:69:35: warning: shift count >= width of type [-Wshift-count-overflow]
  return l > ((1UL<<32)-1) ? ((1UL<<32)-1) : l;
                                  ^ ~~
2 warnings generated.
LLVM-CC r600--/lib/geometric/cross.cl.bc
LLVM-CC r600--/lib/geometric/dot.cl.bc
LLVM-CC r600--/lib/geometric/length.cl.bc

The code supports EVERGREEN (radeon HD 5xxx).
You have NORTHERN ISLANDS (radeon HD 6xxx).

[tstellar]

uname -a
Linux linux-gyo1 3.4.0-5-desktop #1 SMP PREEMPT Fri May 25 15:45:36 UTC 2012 (64b97cd) i686 i686 i386 GNU/Linux

drm-next have same problem

Ok, you should file a bug at bugs.freedesktop.org. You might have better success with an older kernel, it was working on my Juniper with 3.2, and I've been using 3.1 for most of my testing.

[Pontostroy]

What kernel version are you using?

uname -a
Linux linux-gyo1 3.4.0-5-desktop #1 SMP PREEMPT Fri May 25 15:45:36 UTC 2012 (64b97cd) i686 i686 i386 GNU/Linux

drm-next have same problem

[tstellar]

It works (almost).

/home/downloads/git/opencl-example/hello_world 
There are 1 platforms.
There are 1 GPU devices.
clCreateContext() succeeded.
clCreateCommandQueue() succeeded.
clCreateProgramWithSource() suceeded.
clBuildProgram() suceeded.
clCreateKernel() suceeded.
clCreateBuffer() succeeded.
clSetKernelArg() succeeded.
radeon: The kernel rejected CS, see dmesg for more information.
clEnqueueNDRangeKernel() suceeded.
clEnqueueReadBuffer() suceeded.
pi = 0.000000

[  424.341250] radeon 0000:01:00.0: evergreen_surface_check_linear_aligned:209 cb pitch 8 invalid must be aligned with 128
[  424.341254] radeon 0000:01:00.0: evergreen_cs_track_validate_cb:422 cb[0] invalid (0x00000000 0x00000000 0x00000010 0x04000134)
[  424.341257] radeon 0000:01:00.0: evergreen_packet3_check:2110 invalid cmd stream 628
[  424.341259] [drm:radeon_cs_ib_chunk] *ERROR* Invalid command stream !

What kernel version are you using?

[Pontostroy]

You need to run those programs as root if X is running.
.

It works (almost).

/home/downloads/git/opencl-example/hello_world 
There are 1 platforms.
There are 1 GPU devices.
clCreateContext() succeeded.
clCreateCommandQueue() succeeded.
clCreateProgramWithSource() suceeded.
clBuildProgram() suceeded.
clCreateKernel() suceeded.
clCreateBuffer() succeeded.
clSetKernelArg() succeeded.
radeon: The kernel rejected CS, see dmesg for more information.
clEnqueueNDRangeKernel() suceeded.
clEnqueueReadBuffer() suceeded.
pi = 0.000000

[  424.341250] radeon 0000:01:00.0: evergreen_surface_check_linear_aligned:209 cb pitch 8 invalid must be aligned with 128
[  424.341254] radeon 0000:01:00.0: evergreen_cs_track_validate_cb:422 cb[0] invalid (0x00000000 0x00000000 0x00000010 0x04000134)
[  424.341257] radeon 0000:01:00.0: evergreen_packet3_check:2110 invalid cmd stream 628
[  424.341259] [drm:radeon_cs_ib_chunk] *ERROR* Invalid command stream !

[bridgman]

The power management options you mentioned mostly sound obsolete -- all the PM code is in the kernel these days rather than the X driver. Take a look at the following link and see if that helps :

http://www.x.org/wiki/RadeonFeature#KMS_Power_Management_Options

The dynpm options still has problems on a lot of systems, so try the profile approach first. Either low or mid should reduce the heat -- mid is the best option if your VBIOS includes a "mid" profile.

[bug!]

Are you sure you had the proper drivers installed? I mean there is the case, where you have 64 bit system - then you have to install install 32 bit mesa alongside the 64 bit one in order to utilise hardware acceleration in wine.

You are indeed correct, I was missing the 32bit libs, my bad!
It works like a charm, dxdiag completes all Graphics related tests via. the Radeon Drivers perfectly.

The 64 Bit Snake has bitten me once again... I should have known this.

Now I'm off to tackle the Heat once more!

[Hirager]

Actually I got another question... when will the free Radeon Drivers be usable with Wine?
The last time I tried, I couldn't get 3D Acceleration working through wine, at least that's what DXDiag told me.

Are you sure you had the proper drivers installed? I mean there is the case, where you have 64 bit system - then you have to install install 32 bit mesa alongside the 64 bit one in order to utilise hardware acceleration in wine.

[bug!]

sup

I have a Mobility Radeon 5400 Series Card, will this also affect me?

I really can't wait to fully switch to the free Radeon Drivers...
My current situation is that the free Drivers keep my fans spinning on my Laptop due to heat generation, therefore I am using the Proprietary ones.
I have tried nearly everything to reduce the heat, but nothing ever worked (Options like ForceLowPowerMode/ClockGating/DynamicPM/etc).

I guess my main question is, what will it improve?

Actually I got another question... when will the free Radeon Drivers be usable with Wine?
The last time I tried, I couldn't get 3D Acceleration working through wine, at least that's what DXDiag told me.

Damn, this got me exited again!
Keep it up AMD!
;D