[Mesa-dev] [PATCH 19/25] gallium/i915: handle query_renderer caps
Emil Velikov
emil.l.velikov at gmail.com
Sat Feb 22 05:22:14 PST 2014
On 22/02/14 03:33, Stéphane Marchesin wrote:
> On Fri, Feb 21, 2014 at 7:04 PM, Emil Velikov <emil.l.velikov at gmail.com>wrote:
>
>> Implementation is a verbatim copy from the classic driver.
>>
>> This introduces a driver dependency on libdrm-intel, as the winsys does not
>> cache the aperture size of the device.
>>
>
> Usually if you have to add calls into libdrm, you add them to the winsys,
> and then add a new entry point to the winsys which you call... If you want
> I can do it.
>
> I've been ensuring that the only place calling into libdrm was the winsys
> so far. Whether or not we remove the sw winsys, I think it's a healthy
> thing to do.
>
Agreed, keeping OS/platform specifics within the winsys is the way to do it.
Should I store the device aperture size within i915_winsys or is there a
more suitable place for it ? The current implementation is essentially a
vtable for the drm/sw callbacks.
-Emil
>
>
>>
>> Cc: Stephane Marchesin <stephane.marchesin at gmail.com>
>> Signed-off-by: Emil Velikov <emil.l.velikov at gmail.com>
>> ---
>>
>> NOTE: This _will_ fail on i915/sw due to the lack of fd on the
>> particular winsys.
>>
>> Stephane,
>> Perhaps we can take a look at i915/sw, its users and future.
>>
>
> I'm not sure if there are any i915/sw users; I certainly don't test/use the
> sw winsys. I just keep it around in case someone does. If someone wants to
> speak up, now is the time.
>
> Stéphane
>
>
>> I can vision (but I'm not planing to pursue) some cleanups
>> if the software winsys is retired.
>>
>> Cheers
>> -Emil
>>
>> src/gallium/drivers/i915/Makefile.am | 3 ++-
>> src/gallium/drivers/i915/SConscript | 2 ++
>> src/gallium/drivers/i915/i915_screen.c | 39
>> ++++++++++++++++++++++++++++++++++
>> 3 files changed, 43 insertions(+), 1 deletion(-)
>>
>> diff --git a/src/gallium/drivers/i915/Makefile.am
>> b/src/gallium/drivers/i915/Makefile.am
>> index a4a3e86..67936ca 100644
>> --- a/src/gallium/drivers/i915/Makefile.am
>> +++ b/src/gallium/drivers/i915/Makefile.am
>> @@ -24,7 +24,8 @@ include Makefile.sources
>> include $(top_srcdir)/src/gallium/Automake.inc
>>
>> AM_CFLAGS = \
>> - $(GALLIUM_DRIVER_CFLAGS)
>> + $(GALLIUM_DRIVER_CFLAGS) \
>> + $(INTEL_CFLAGS)
>>
>> noinst_LTLIBRARIES = libi915.la
>>
>> diff --git a/src/gallium/drivers/i915/SConscript
>> b/src/gallium/drivers/i915/SConscript
>> index 22de67d..a04517e 100644
>> --- a/src/gallium/drivers/i915/SConscript
>> +++ b/src/gallium/drivers/i915/SConscript
>> @@ -2,6 +2,8 @@ Import('*')
>>
>> env = env.Clone()
>>
>> +env.PkgUseModules('DRM_INTEL')
>> +
>> i915 = env.ConvenienceLibrary(
>> target = 'i915',
>> source = env.ParseSourceList('Makefile.sources', 'C_SOURCES')
>> diff --git a/src/gallium/drivers/i915/i915_screen.c
>> b/src/gallium/drivers/i915/i915_screen.c
>> index 1f55a06..3970dae 100644
>> --- a/src/gallium/drivers/i915/i915_screen.c
>> +++ b/src/gallium/drivers/i915/i915_screen.c
>> @@ -26,6 +26,7 @@
>>
>> **************************************************************************/
>>
>>
>> +#include <intel_bufmgr.h>
>> #include "draw/draw_context.h"
>> #include "util/u_format.h"
>> #include "util/u_format_s3tc.h"
>> @@ -40,6 +41,7 @@
>> #include "i915_resource.h"
>> #include "i915_winsys.h"
>> #include "i915_public.h"
>> +#include <../winsys/i915/drm/i915_drm_winsys.h>
>>
>>
>> /*
>> @@ -281,6 +283,43 @@ i915_get_param(struct pipe_screen *screen, enum
>> pipe_cap cap)
>> case PIPE_CAP_MAX_GL_ES2_VERSION:
>> return 20;
>>
>> + case PIPE_CAP_VENDOR_ID:
>> + return 0x8086;
>> + case PIPE_CAP_DEVICE_ID:
>> + return is->iws->pci_id;
>> + case PIPE_CAP_ACCELERATED:
>> + return 1;
>> + case PIPE_CAP_VIDEO_MEMORY: {
>> + /* Once a batch uses more than 75% of the maximum mappable size, we
>> + * assume that there's some fragmentation, and we start doing extra
>> + * flushing, etc. That's the big cliff apps will care about.
>> + */
>> + struct i915_drm_winsys *idws =(struct i915_drm_winsys *)is->iws;
>> + size_t aper_size;
>> + size_t mappable_size;
>> +
>> + drm_intel_get_aperture_sizes(idws->fd, &mappable_size, &aper_size);
>> +
>> + const unsigned gpu_mappable_megabytes =
>> + (aper_size / (1024 * 1024)) * 3 / 4;
>> +
>> + const long system_memory_pages = sysconf(_SC_PHYS_PAGES);
>> + const long system_page_size = sysconf(_SC_PAGE_SIZE);
>> +
>> + if (system_memory_pages <= 0 || system_page_size <= 0)
>> + return 0;
>> +
>> + const uint64_t system_memory_bytes = (uint64_t) system_memory_pages
>> + * (uint64_t) system_page_size;
>> +
>> + const unsigned system_memory_megabytes =
>> + (unsigned) (system_memory_bytes / (1024 * 1024));
>> +
>> + return MIN2(system_memory_megabytes, gpu_mappable_megabytes);
>> + }
>> + case PIPE_CAP_UMA:
>> + return 1;
>> +
>> default:
>> debug_printf("%s: Unknown cap %u.\n", __FUNCTION__, cap);
>> return 0;
>> --
>> 1.9.0
>>
>>
>
More information about the mesa-dev
mailing list