Giuseppe Bilotta [Wed, 5 Nov 2014 21:42:03 +0000 (22:42 +0100)]
Build with debug and all warnings enabled
Giuseppe Bilotta [Wed, 5 Nov 2014 21:41:40 +0000 (22:41 +0100)]
Traits for device partition
Giuseppe Bilotta [Tue, 4 Nov 2014 20:14:37 +0000 (21:14 +0100)]
Further progress in dev info traits
Giuseppe Bilotta [Tue, 4 Nov 2014 19:29:01 +0000 (20:29 +0100)]
Device Topology (NV) traits
Giuseppe Bilotta [Tue, 4 Nov 2014 18:54:00 +0000 (19:54 +0100)]
Introduce output-mode specific device info
Giuseppe Bilotta [Tue, 4 Nov 2014 18:46:58 +0000 (19:46 +0100)]
Device topology AMD traits
Giuseppe Bilotta [Tue, 4 Nov 2014 18:25:45 +0000 (19:25 +0100)]
Hack to pass devtype to the device info checks
Giuseppe Bilotta [Tue, 4 Nov 2014 18:18:00 +0000 (19:18 +0100)]
Device type as traits (almost)
Giuseppe Bilotta [Tue, 4 Nov 2014 18:01:13 +0000 (19:01 +0100)]
Device extensions via traits
Giuseppe Bilotta [Tue, 4 Nov 2014 17:32:22 +0000 (18:32 +0100)]
Cleanup extensions treatment
Giuseppe Bilotta [Tue, 4 Nov 2014 17:24:17 +0000 (18:24 +0100)]
Start working on traits approach for device info
Giuseppe Bilotta [Tue, 4 Nov 2014 13:28:51 +0000 (14:28 +0100)]
Preliminary change to ease switching device info over to traits
Giuseppe Bilotta [Tue, 4 Nov 2014 11:45:15 +0000 (12:45 +0100)]
Raw vs human output modes
Giuseppe Bilotta [Mon, 3 Nov 2014 21:10:42 +0000 (22:10 +0100)]
Conditional platform property check
Giuseppe Bilotta [Mon, 3 Nov 2014 19:05:56 +0000 (20:05 +0100)]
Convert platform info to traits mechanism
Giuseppe Bilotta [Mon, 27 Oct 2014 20:45:32 +0000 (21:45 +0100)]
Add missing header to ms_support.h
Giuseppe Bilotta [Mon, 27 Oct 2014 20:33:57 +0000 (21:33 +0100)]
Support older OpenCL 1.2 headers
Giuseppe Bilotta [Mon, 27 Oct 2014 20:09:31 +0000 (21:09 +0100)]
Initial support for Microsoft's C compiler
Giuseppe Bilotta [Mon, 27 Oct 2014 20:04:03 +0000 (21:04 +0100)]
Put format macro inclusion/definition in own header
Giuseppe Bilotta [Mon, 27 Oct 2014 19:53:33 +0000 (20:53 +0100)]
Tweaks for easier MSVC support
Giuseppe Bilotta [Mon, 27 Oct 2014 18:32:58 +0000 (19:32 +0100)]
Properly support multi-type devices
Giuseppe Bilotta [Sun, 26 Oct 2014 23:00:39 +0000 (00:00 +0100)]
cl_amd_svm support
Giuseppe Bilotta [Sun, 26 Oct 2014 22:48:10 +0000 (23:48 +0100)]
Handle unspecified partition types
Some versions of the AMD driver claim the GPU supports device partition
(to as many subdevices as there are CUs), but then they don't specify
any supported partition type.
Giuseppe Bilotta [Sun, 26 Oct 2014 22:32:46 +0000 (23:32 +0100)]
Up clinfo version in manpage
Giuseppe Bilotta [Sun, 26 Oct 2014 22:32:01 +0000 (23:32 +0100)]
Document preliminary 2.0 support
Giuseppe Bilotta [Sun, 26 Oct 2014 22:29:03 +0000 (23:29 +0100)]
Introduce support for 2.0 properties
Giuseppe Bilotta [Sun, 26 Oct 2014 21:03:24 +0000 (22:03 +0100)]
Add 2.0 core properties to ext.h
Because headers aren't in widespread circulation yet.
Giuseppe Bilotta [Sun, 26 Oct 2014 20:56:09 +0000 (21:56 +0100)]
Cleanup 1.2 device detection and add 2.0
Giuseppe Bilotta [Sun, 26 Oct 2014 20:18:04 +0000 (21:18 +0100)]
Add forgotten CL_DEVICE_VENDOR_ID
Giuseppe Bilotta [Sun, 26 Oct 2014 07:42:11 +0000 (08:42 +0100)]
minor manpage tuning
Giuseppe Bilotta [Thu, 16 Oct 2014 07:06:05 +0000 (09:06 +0200)]
Skip leading whitespaces in strings
Giuseppe Bilotta [Thu, 16 Oct 2014 06:55:38 +0000 (08:55 +0200)]
Add PCI-E to NVIDIA's device topology output
It's not needed (NVIDIA doesn't expose other forms of topology),
but this way it's consistent with AMD's.
Giuseppe Bilotta [Thu, 16 Oct 2014 06:53:02 +0000 (08:53 +0200)]
Cleanup memory presentation
No whitespace in parenthesis, show KiB etc instead of KB etc, promote
exact magnitude orders to the next (e.g. 1024KiB => 1MiB).
Giuseppe Bilotta [Mon, 29 Sep 2014 06:05:13 +0000 (08:05 +0200)]
Single-precision is part of core
It has been objected that while the other floating-point precision
capabilities always have something next to them (either n/a or the
extension through which they are supported), single-precision doesn't.
Since it's always supported because it's part of core, just mark it as
such, to make it consistent with the other, and with things such as the
support for device partition.
Giuseppe Bilotta [Sun, 28 Sep 2014 04:12:56 +0000 (06:12 +0200)]
write_only and read_only are only for images
Giuseppe Bilotta [Sat, 27 Sep 2014 10:23:36 +0000 (12:23 +0200)]
Up clinfo version in manpage
Giuseppe Bilotta [Sat, 27 Sep 2014 07:08:10 +0000 (09:08 +0200)]
Dump build log on stderr in case of failure
... during work-group size detection. This should make it easier to
debug this situation.
Giuseppe Bilotta [Fri, 26 Sep 2014 20:42:41 +0000 (22:42 +0200)]
cl_APPLE_fp64_basic_ops
Giuseppe Bilotta [Fri, 26 Sep 2014 20:37:56 +0000 (22:37 +0200)]
On Darwin, use -framework instead of -l
Giuseppe Bilotta [Fri, 26 Sep 2014 20:30:06 +0000 (22:30 +0200)]
Don't rely on external cl_ext
A lot of the stuff present in Khronos headers is missing on Mac OS X,
anyway, so we'd need to define it ourselves anyway.
magnum [Sat, 20 Sep 2014 23:51:35 +0000 (01:51 +0200)]
Translate nvidia's 'devfunc' to device and function.
magnum [Sat, 20 Sep 2014 23:27:00 +0000 (01:27 +0200)]
Fix a bug where PCI id's of 0x80 or higher got sign extended in output
due to being cl_char in AMD's struct. 86:00.0 was shown as
ffffff86:00.0
Giuseppe Bilotta [Mon, 22 Sep 2014 16:23:36 +0000 (18:23 +0200)]
More switchover to REPORT_ERROR
Giuseppe Bilotta [Mon, 22 Sep 2014 14:57:32 +0000 (16:57 +0200)]
Use REPORT_ERROR instead of CHECK_ERROR for non-fatals
Giuseppe Bilotta [Mon, 22 Sep 2014 14:27:01 +0000 (16:27 +0200)]
Ensure better pretty-printing of param names
Giuseppe Bilotta [Mon, 22 Sep 2014 14:15:37 +0000 (16:15 +0200)]
Cosmetic fix
Giuseppe Bilotta [Sun, 21 Sep 2014 12:04:23 +0000 (14:04 +0200)]
Older cl_ext don't have the CL_DEVICE_*_QCOM ext
Giuseppe Bilotta [Fri, 22 Aug 2014 15:56:35 +0000 (17:56 +0200)]
Up clinfo version in man page
Giuseppe Bilotta [Fri, 22 Aug 2014 15:53:45 +0000 (17:53 +0200)]
Support cl_khr_icd extension
Giuseppe Bilotta [Fri, 22 Aug 2014 08:03:30 +0000 (10:03 +0200)]
Support cl_qcom_ext_host_ptr extension
Giuseppe Bilotta [Fri, 22 Aug 2014 07:37:44 +0000 (09:37 +0200)]
Make it easier to compile clinfo for people with less recent 1.2 OpenCL header files
Giuseppe Bilotta [Wed, 13 Aug 2014 15:44:06 +0000 (17:44 +0200)]
Shift a few fields around
Giuseppe Bilotta [Wed, 13 Aug 2014 15:38:04 +0000 (17:38 +0200)]
Found some extra NVIDIA properties
Giuseppe Bilotta [Wed, 13 Aug 2014 03:44:30 +0000 (05:44 +0200)]
Document cl_khr_spir support
Giuseppe Bilotta [Tue, 12 Aug 2014 10:58:25 +0000 (12:58 +0200)]
Show supported SPIR versions for cl_khr_spir devices
Giuseppe Bilotta [Thu, 31 Jul 2014 19:50:22 +0000 (21:50 +0200)]
Flush stdout/err at strategic places
Giuseppe Bilotta [Wed, 9 Jul 2014 20:57:36 +0000 (22:57 +0200)]
man page cleanup
Giuseppe Bilotta [Sun, 22 Jun 2014 21:18:29 +0000 (23:18 +0200)]
Add support for AMD device topology property
Giuseppe Bilotta [Fri, 20 Jun 2014 09:09:32 +0000 (11:09 +0200)]
Support cl_altera_device_temperature extension
Giuseppe Bilotta [Fri, 9 May 2014 06:05:41 +0000 (08:05 +0200)]
Remove double space in man page
Giuseppe Bilotta [Tue, 29 Apr 2014 21:08:17 +0000 (23:08 +0200)]
Refactor MEM_PARAM parenthetical print
So it can be used for the AMD Free global memory extension
without code duplication.
Giuseppe Bilotta [Tue, 29 Apr 2014 21:03:07 +0000 (23:03 +0200)]
Be more careful about handling cl_ types
Particularly, cl_ulong might not be a (host) long, since cl_ulong
is guaranteed to be 64-bit wide, while on some 32-bit architectures
the long type is only 32-bit wide. So rely on the inttypes.h definitions
and macros on the assumption that cl_ulong is an uint64_t.
While we're at it, use the correct printf format macros for other
types.
Giuseppe Bilotta [Tue, 29 Apr 2014 20:23:49 +0000 (22:23 +0200)]
Fix param order in calloc()
Giuseppe Bilotta [Mon, 28 Apr 2014 15:02:07 +0000 (17:02 +0200)]
Tune handling of no-dev platforms
In case no devices are present, just skip getting the device IDs.
This ensure more consistent output (specifically, no-dev platforms
don't get stuck together with the following platform).
Also add a missing CHECK_ERROR when actually getting the IDs.
Giuseppe Bilotta [Tue, 22 Apr 2014 06:12:35 +0000 (08:12 +0200)]
Allocate with calloc, not malloc
Make sure the memory is initialized to 0, so that untouched array values
have sensible values (for example, when no devices are present in a
platform, `num_devs + p` could be untouched, and we want it to be zero
in that case).
Giuseppe Bilotta [Mon, 21 Apr 2014 19:30:07 +0000 (21:30 +0200)]
Better solution to handle no-dev platforms
This should also work fine when no platforms has devices available.
Giuseppe Bilotta [Tue, 15 Apr 2014 21:17:36 +0000 (23:17 +0200)]
Allow a platform to have no devices
Giuseppe Bilotta [Mon, 10 Mar 2014 20:35:08 +0000 (21:35 +0100)]
Handle errors in getWGsizes gracefully
Giuseppe Bilotta [Mon, 10 Mar 2014 19:45:25 +0000 (20:45 +0100)]
Intel extension to partition by name
Giuseppe Bilotta [Mon, 22 Jul 2013 21:51:30 +0000 (23:51 +0200)]
Forgot max samplers property
Giuseppe Bilotta [Tue, 9 Jul 2013 18:32:49 +0000 (20:32 +0200)]
Fix formatting string warnings
Giuseppe Bilotta [Sun, 16 Jun 2013 14:29:05 +0000 (16:29 +0200)]
Also show max number of sub-devices
Giuseppe Bilotta [Thu, 13 Jun 2013 11:39:47 +0000 (13:39 +0200)]
ext.h cleanup and expansion
Giuseppe Bilotta [Tue, 11 Jun 2013 17:58:06 +0000 (19:58 +0200)]
Intel local thread execution extension
Giuseppe Bilotta [Tue, 11 Jun 2013 17:47:35 +0000 (19:47 +0200)]
Define vendor extensions
They might be not defined in the official headers.
Giuseppe Bilotta [Tue, 11 Jun 2013 17:17:47 +0000 (19:17 +0200)]
Show hex for unknown flags
Giuseppe Bilotta [Sun, 9 Jun 2013 17:21:30 +0000 (19:21 +0200)]
Fix excess buffer->strbuf replacement
Giuseppe Bilotta [Sat, 8 Jun 2013 09:29:48 +0000 (11:29 +0200)]
Fix extension name in man page too
Giuseppe Bilotta [Sat, 8 Jun 2013 09:16:26 +0000 (11:16 +0200)]
Fix extension name
Giuseppe Bilotta [Fri, 7 Jun 2013 18:48:03 +0000 (20:48 +0200)]
Minor tuning
Giuseppe Bilotta [Fri, 7 Jun 2013 18:44:17 +0000 (20:44 +0200)]
Whitespace
Giuseppe Bilotta [Fri, 7 Jun 2013 18:34:29 +0000 (20:34 +0200)]
OpenCL C Version
Giuseppe Bilotta [Fri, 7 Jun 2013 18:31:50 +0000 (20:31 +0200)]
CL_FP_CORRECTLY_ROUNDED_DIVIDE_SQRT floating-point config flag
Giuseppe Bilotta [Fri, 7 Jun 2013 18:20:22 +0000 (20:20 +0200)]
Better wording
Giuseppe Bilotta [Fri, 7 Jun 2013 18:16:28 +0000 (20:16 +0200)]
Man typo
Giuseppe Bilotta [Fri, 7 Jun 2013 18:16:13 +0000 (20:16 +0200)]
cl_khr_image2D_buffer extension
Giuseppe Bilotta [Fri, 7 Jun 2013 18:07:30 +0000 (20:07 +0200)]
Missed a couple formatting cleanups
Giuseppe Bilotta [Fri, 7 Jun 2013 15:38:32 +0000 (17:38 +0200)]
Present the timer offset in a nicer way
In seconds, it gives the "start time" of the device. Display this as a
date.
Giuseppe Bilotta [Fri, 7 Jun 2013 11:58:02 +0000 (13:58 +0200)]
Formatting cleanup
Giuseppe Bilotta [Fri, 7 Jun 2013 11:34:00 +0000 (13:34 +0200)]
More string sanitation
Giuseppe Bilotta [Fri, 7 Jun 2013 10:08:55 +0000 (12:08 +0200)]
Rename GET_STRINGX to GET_STRING
Giuseppe Bilotta [Fri, 7 Jun 2013 10:07:52 +0000 (12:07 +0200)]
Get rid of single-param GET_STRING, part 1
Start by only using GET_STRINGX all around.
Giuseppe Bilotta [Fri, 7 Jun 2013 10:02:06 +0000 (12:02 +0200)]
Start refactoring
Giuseppe Bilotta [Fri, 7 Jun 2013 06:17:15 +0000 (08:17 +0200)]
Man page
Giuseppe Bilotta [Fri, 7 Jun 2013 05:13:13 +0000 (07:13 +0200)]
Preferred work group size multiple
Giuseppe Bilotta [Thu, 6 Jun 2013 15:20:33 +0000 (17:20 +0200)]
AMD device attributes (plus NV warp size)
Giuseppe Bilotta [Thu, 6 Jun 2013 15:04:36 +0000 (17:04 +0200)]
Atomic counters EXT
Giuseppe Bilotta [Thu, 6 Jun 2013 14:59:07 +0000 (16:59 +0200)]
All OpenCL 1.2 properties
Giuseppe Bilotta [Thu, 6 Jun 2013 14:51:22 +0000 (16:51 +0200)]
Device partition/fission
Giuseppe Bilotta [Thu, 6 Jun 2013 11:15:48 +0000 (13:15 +0200)]
README and LICENSE