* [PATCH] intel: Export GT config attributes @ 2014-12-18 18:12 jeff.mcgee 2014-12-18 18:12 ` jeff.mcgee 2015-01-07 23:46 ` Jeff McGee 0 siblings, 2 replies; 5+ messages in thread From: jeff.mcgee @ 2014-12-18 18:12 UTC (permalink / raw) To: dri-devel From: Jeff McGee <jeff.mcgee@intel.com> The motivation for this change is that fusing can be used to create multiple slice, subslice, and EU configuration within the same PCI ID. CHV is the first such device to do this and thus make an ID-based lookup table approach unreliable. The best solution is for the kernel to determine the precise config from fuse registers and share the required information with userspace. Moving to this approach has the added benefit of reducing the number of static parameters that userspace must maintain for current and future devices. The associated kernel patch has been posted to intel-gfx list (no archive link yet). I'm trying to collect feedback from userspace which I hope will be interested in adopting some form of this interface. I am starting with these 4 attributes, but we can add or remove them according to the userspace needs. Please provide any comments. Thanks Jeff McGee (1): intel: Export GT config attributes include/drm/i915_drm.h | 4 +++ intel/intel_bufmgr.h | 5 ++++ intel/intel_bufmgr_gem.c | 63 ++++++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 72 insertions(+) -- 2.2.0 _______________________________________________ dri-devel mailing list dri-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/dri-devel ^ permalink raw reply [flat|nested] 5+ messages in thread
* [PATCH] intel: Export GT config attributes 2014-12-18 18:12 [PATCH] intel: Export GT config attributes jeff.mcgee @ 2014-12-18 18:12 ` jeff.mcgee 2015-01-09 2:56 ` Zhenyu Wang 2015-01-07 23:46 ` Jeff McGee 1 sibling, 1 reply; 5+ messages in thread From: jeff.mcgee @ 2014-12-18 18:12 UTC (permalink / raw) To: dri-devel From: Jeff McGee <jeff.mcgee@intel.com> Update kernel interface with new I915_GETPARAM ioctl entries for slice total, subslice total, EU total, and threads per EU. Add a wrapping function for each parameter. The motivation for this change is that fusing can be used to create multiple slice, subslice, and EU configuration within the same PCI ID. CHV is the first such device to do this and thus make an ID-based lookup table approach unreliable. The best solution is for the kernel to determine the precise config from fuse registers and share the required information with userspace. Moving to this approach has the added benefit of reducing the number of static parameters that userspace must maintain for current and future devices. For: VIZ-4636 Signed-off-by: Jeff McGee <jeff.mcgee@intel.com> --- include/drm/i915_drm.h | 4 +++ intel/intel_bufmgr.h | 5 ++++ intel/intel_bufmgr_gem.c | 63 ++++++++++++++++++++++++++++++++++++++++++++++++ 3 files changed, 72 insertions(+) diff --git a/include/drm/i915_drm.h b/include/drm/i915_drm.h index 15dd01d..be38adf 100644 --- a/include/drm/i915_drm.h +++ b/include/drm/i915_drm.h @@ -340,6 +340,10 @@ typedef struct drm_i915_irq_wait { #define I915_PARAM_HAS_EXEC_HANDLE_LUT 26 #define I915_PARAM_HAS_WT 27 #define I915_PARAM_CMD_PARSER_VERSION 28 +#define I915_PARAM_SLICE_TOTAL 30 +#define I915_PARAM_SUBSLICE_TOTAL 31 +#define I915_PARAM_EU_TOTAL 32 +#define I915_PARAM_THREADS_PER_EU 33 typedef struct drm_i915_getparam { int param; diff --git a/intel/intel_bufmgr.h b/intel/intel_bufmgr.h index be83a56..90e535d 100644 --- a/intel/intel_bufmgr.h +++ b/intel/intel_bufmgr.h @@ -264,6 +264,11 @@ int drm_intel_get_reset_stats(drm_intel_context *ctx, uint32_t *active, uint32_t *pending); +int drm_intel_get_slice_total(int fd, unsigned int *slice_total); +int drm_intel_get_subslice_total(int fd, unsigned int *subslice_total); +int drm_intel_get_eu_total(int fd, unsigned int *eu_total); +int drm_intel_get_threads_per_eu(int fd, unsigned int *threads_per_eu); + /** @{ Compatibility defines to keep old code building despite the symbol rename * from dri_* to drm_intel_* */ diff --git a/intel/intel_bufmgr_gem.c b/intel/intel_bufmgr_gem.c index 14e92c9..e0b13e7 100644 --- a/intel/intel_bufmgr_gem.c +++ b/intel/intel_bufmgr_gem.c @@ -3293,6 +3293,69 @@ drm_intel_reg_read(drm_intel_bufmgr *bufmgr, return ret; } +drm_public int +drm_intel_get_slice_total(int fd, unsigned int *slice_total) +{ + drm_i915_getparam_t gp; + int ret; + + VG_CLEAR(gp); + gp.value = (int*)slice_total; + gp.param = I915_PARAM_SLICE_TOTAL; + ret = drmIoctl(fd, DRM_IOCTL_I915_GETPARAM, &gp); + if (ret) + return -errno; + + return 0; +} + +drm_public int +drm_intel_get_subslice_total(int fd, unsigned int *subslice_total) +{ + drm_i915_getparam_t gp; + int ret; + + VG_CLEAR(gp); + gp.value = (int*)subslice_total; + gp.param = I915_PARAM_SUBSLICE_TOTAL; + ret = drmIoctl(fd, DRM_IOCTL_I915_GETPARAM, &gp); + if (ret) + return -errno; + + return 0; +} + +drm_public int +drm_intel_get_eu_total(int fd, unsigned int *eu_total) +{ + drm_i915_getparam_t gp; + int ret; + + VG_CLEAR(gp); + gp.value = (int*)eu_total; + gp.param = I915_PARAM_EU_TOTAL; + ret = drmIoctl(fd, DRM_IOCTL_I915_GETPARAM, &gp); + if (ret) + return -errno; + + return 0; +} + +drm_public int +drm_intel_get_threads_per_eu(int fd, unsigned int *threads_per_eu) +{ + drm_i915_getparam_t gp; + int ret; + + VG_CLEAR(gp); + gp.value = (int*)threads_per_eu; + gp.param = I915_PARAM_THREADS_PER_EU; + ret = drmIoctl(fd, DRM_IOCTL_I915_GETPARAM, &gp); + if (ret) + return -errno; + + return 0; +} /** * Annotate the given bo for use in aub dumping. -- 2.2.0 _______________________________________________ dri-devel mailing list dri-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/dri-devel ^ permalink raw reply related [flat|nested] 5+ messages in thread
* Re: [PATCH] intel: Export GT config attributes 2014-12-18 18:12 ` jeff.mcgee @ 2015-01-09 2:56 ` Zhenyu Wang 2015-01-09 16:10 ` Jeff McGee 0 siblings, 1 reply; 5+ messages in thread From: Zhenyu Wang @ 2015-01-09 2:56 UTC (permalink / raw) To: jeff.mcgee; +Cc: intel-gfx, dri-devel [-- Attachment #1.1: Type: text/plain, Size: 3676 bytes --] On 2014.12.18 12:12:33 -0600, jeff.mcgee@intel.com wrote: > diff --git a/include/drm/i915_drm.h b/include/drm/i915_drm.h > index 15dd01d..be38adf 100644 > --- a/include/drm/i915_drm.h > +++ b/include/drm/i915_drm.h > @@ -340,6 +340,10 @@ typedef struct drm_i915_irq_wait { > #define I915_PARAM_HAS_EXEC_HANDLE_LUT 26 > #define I915_PARAM_HAS_WT 27 > #define I915_PARAM_CMD_PARSER_VERSION 28 > +#define I915_PARAM_SLICE_TOTAL 30 > +#define I915_PARAM_SUBSLICE_TOTAL 31 > +#define I915_PARAM_EU_TOTAL 32 > +#define I915_PARAM_THREADS_PER_EU 33 > > typedef struct drm_i915_getparam { > int param; > diff --git a/intel/intel_bufmgr.h b/intel/intel_bufmgr.h > index be83a56..90e535d 100644 > --- a/intel/intel_bufmgr.h > +++ b/intel/intel_bufmgr.h > @@ -264,6 +264,11 @@ int drm_intel_get_reset_stats(drm_intel_context *ctx, > uint32_t *active, > uint32_t *pending); > > +int drm_intel_get_slice_total(int fd, unsigned int *slice_total); > +int drm_intel_get_subslice_total(int fd, unsigned int *subslice_total); > +int drm_intel_get_eu_total(int fd, unsigned int *eu_total); > +int drm_intel_get_threads_per_eu(int fd, unsigned int *threads_per_eu); > + I think we normally use drm_intel_bufmgr in API instead of fd directly. And this config info can be initialized when bufmgr init, so later request can get info without calling ioctl everytime. Also please cc intel-gfx list for any intel related patch. thanks. > /** @{ Compatibility defines to keep old code building despite the symbol rename > * from dri_* to drm_intel_* > */ > diff --git a/intel/intel_bufmgr_gem.c b/intel/intel_bufmgr_gem.c > index 14e92c9..e0b13e7 100644 > --- a/intel/intel_bufmgr_gem.c > +++ b/intel/intel_bufmgr_gem.c > @@ -3293,6 +3293,69 @@ drm_intel_reg_read(drm_intel_bufmgr *bufmgr, > return ret; > } > > +drm_public int > +drm_intel_get_slice_total(int fd, unsigned int *slice_total) > +{ > + drm_i915_getparam_t gp; > + int ret; > + > + VG_CLEAR(gp); > + gp.value = (int*)slice_total; > + gp.param = I915_PARAM_SLICE_TOTAL; > + ret = drmIoctl(fd, DRM_IOCTL_I915_GETPARAM, &gp); > + if (ret) > + return -errno; > + > + return 0; > +} > + > +drm_public int > +drm_intel_get_subslice_total(int fd, unsigned int *subslice_total) > +{ > + drm_i915_getparam_t gp; > + int ret; > + > + VG_CLEAR(gp); > + gp.value = (int*)subslice_total; > + gp.param = I915_PARAM_SUBSLICE_TOTAL; > + ret = drmIoctl(fd, DRM_IOCTL_I915_GETPARAM, &gp); > + if (ret) > + return -errno; > + > + return 0; > +} > + > +drm_public int > +drm_intel_get_eu_total(int fd, unsigned int *eu_total) > +{ > + drm_i915_getparam_t gp; > + int ret; > + > + VG_CLEAR(gp); > + gp.value = (int*)eu_total; > + gp.param = I915_PARAM_EU_TOTAL; > + ret = drmIoctl(fd, DRM_IOCTL_I915_GETPARAM, &gp); > + if (ret) > + return -errno; > + > + return 0; > +} > + > +drm_public int > +drm_intel_get_threads_per_eu(int fd, unsigned int *threads_per_eu) > +{ > + drm_i915_getparam_t gp; > + int ret; > + > + VG_CLEAR(gp); > + gp.value = (int*)threads_per_eu; > + gp.param = I915_PARAM_THREADS_PER_EU; > + ret = drmIoctl(fd, DRM_IOCTL_I915_GETPARAM, &gp); > + if (ret) > + return -errno; > + > + return 0; > +} > > /** > * Annotate the given bo for use in aub dumping. > -- > 2.2.0 > > _______________________________________________ > dri-devel mailing list > dri-devel@lists.freedesktop.org > http://lists.freedesktop.org/mailman/listinfo/dri-devel -- Open Source Technology Center, Intel ltd. $gpg --keyserver wwwkeys.pgp.net --recv-keys 4D781827 [-- Attachment #1.2: Digital signature --] [-- Type: application/pgp-signature, Size: 181 bytes --] [-- Attachment #2: Type: text/plain, Size: 159 bytes --] _______________________________________________ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx ^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] intel: Export GT config attributes 2015-01-09 2:56 ` Zhenyu Wang @ 2015-01-09 16:10 ` Jeff McGee 0 siblings, 0 replies; 5+ messages in thread From: Jeff McGee @ 2015-01-09 16:10 UTC (permalink / raw) To: Zhenyu Wang; +Cc: intel-gfx, dri-devel On Fri, Jan 09, 2015 at 10:56:16AM +0800, Zhenyu Wang wrote: > On 2014.12.18 12:12:33 -0600, jeff.mcgee@intel.com wrote: > > diff --git a/include/drm/i915_drm.h b/include/drm/i915_drm.h > > index 15dd01d..be38adf 100644 > > --- a/include/drm/i915_drm.h > > +++ b/include/drm/i915_drm.h > > @@ -340,6 +340,10 @@ typedef struct drm_i915_irq_wait { > > #define I915_PARAM_HAS_EXEC_HANDLE_LUT 26 > > #define I915_PARAM_HAS_WT 27 > > #define I915_PARAM_CMD_PARSER_VERSION 28 > > +#define I915_PARAM_SLICE_TOTAL 30 > > +#define I915_PARAM_SUBSLICE_TOTAL 31 > > +#define I915_PARAM_EU_TOTAL 32 > > +#define I915_PARAM_THREADS_PER_EU 33 > > > > typedef struct drm_i915_getparam { > > int param; > > diff --git a/intel/intel_bufmgr.h b/intel/intel_bufmgr.h > > index be83a56..90e535d 100644 > > --- a/intel/intel_bufmgr.h > > +++ b/intel/intel_bufmgr.h > > @@ -264,6 +264,11 @@ int drm_intel_get_reset_stats(drm_intel_context *ctx, > > uint32_t *active, > > uint32_t *pending); > > > > +int drm_intel_get_slice_total(int fd, unsigned int *slice_total); > > +int drm_intel_get_subslice_total(int fd, unsigned int *subslice_total); > > +int drm_intel_get_eu_total(int fd, unsigned int *eu_total); > > +int drm_intel_get_threads_per_eu(int fd, unsigned int *threads_per_eu); > > + > > I think we normally use drm_intel_bufmgr in API instead of fd directly. > And this config info can be initialized when bufmgr init, so later request > can get info without calling ioctl everytime. > > Also please cc intel-gfx list for any intel related patch. > > thanks. > The implementation doesn't require a drm_intel_bufmgr instance to query and return these values, so I think it is best to ask for just the drm file descriptor from client. I see this approach also in use by drm_intel_get_aperture_sizes function. Let me know if there is a strong reason from client perspective to use drm_intel_bufmgr instead of fd. I thought about caching these values during bufmgr init, but decided there was no benefit. Some clients won't need these at all. And those that do should be calling the query functions once - these values are unchanging. Jeff > > /** @{ Compatibility defines to keep old code building despite the symbol rename > > * from dri_* to drm_intel_* > > */ > > diff --git a/intel/intel_bufmgr_gem.c b/intel/intel_bufmgr_gem.c > > index 14e92c9..e0b13e7 100644 > > --- a/intel/intel_bufmgr_gem.c > > +++ b/intel/intel_bufmgr_gem.c > > @@ -3293,6 +3293,69 @@ drm_intel_reg_read(drm_intel_bufmgr *bufmgr, > > return ret; > > } > > > > +drm_public int > > +drm_intel_get_slice_total(int fd, unsigned int *slice_total) > > +{ > > + drm_i915_getparam_t gp; > > + int ret; > > + > > + VG_CLEAR(gp); > > + gp.value = (int*)slice_total; > > + gp.param = I915_PARAM_SLICE_TOTAL; > > + ret = drmIoctl(fd, DRM_IOCTL_I915_GETPARAM, &gp); > > + if (ret) > > + return -errno; > > + > > + return 0; > > +} > > + > > +drm_public int > > +drm_intel_get_subslice_total(int fd, unsigned int *subslice_total) > > +{ > > + drm_i915_getparam_t gp; > > + int ret; > > + > > + VG_CLEAR(gp); > > + gp.value = (int*)subslice_total; > > + gp.param = I915_PARAM_SUBSLICE_TOTAL; > > + ret = drmIoctl(fd, DRM_IOCTL_I915_GETPARAM, &gp); > > + if (ret) > > + return -errno; > > + > > + return 0; > > +} > > + > > +drm_public int > > +drm_intel_get_eu_total(int fd, unsigned int *eu_total) > > +{ > > + drm_i915_getparam_t gp; > > + int ret; > > + > > + VG_CLEAR(gp); > > + gp.value = (int*)eu_total; > > + gp.param = I915_PARAM_EU_TOTAL; > > + ret = drmIoctl(fd, DRM_IOCTL_I915_GETPARAM, &gp); > > + if (ret) > > + return -errno; > > + > > + return 0; > > +} > > + > > +drm_public int > > +drm_intel_get_threads_per_eu(int fd, unsigned int *threads_per_eu) > > +{ > > + drm_i915_getparam_t gp; > > + int ret; > > + > > + VG_CLEAR(gp); > > + gp.value = (int*)threads_per_eu; > > + gp.param = I915_PARAM_THREADS_PER_EU; > > + ret = drmIoctl(fd, DRM_IOCTL_I915_GETPARAM, &gp); > > + if (ret) > > + return -errno; > > + > > + return 0; > > +} > > > > /** > > * Annotate the given bo for use in aub dumping. > > -- > > 2.2.0 > > > > _______________________________________________ > > dri-devel mailing list > > dri-devel@lists.freedesktop.org > > http://lists.freedesktop.org/mailman/listinfo/dri-devel > > -- > Open Source Technology Center, Intel ltd. > > $gpg --keyserver wwwkeys.pgp.net --recv-keys 4D781827 _______________________________________________ Intel-gfx mailing list Intel-gfx@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/intel-gfx ^ permalink raw reply [flat|nested] 5+ messages in thread
* Re: [PATCH] intel: Export GT config attributes 2014-12-18 18:12 [PATCH] intel: Export GT config attributes jeff.mcgee 2014-12-18 18:12 ` jeff.mcgee @ 2015-01-07 23:46 ` Jeff McGee 1 sibling, 0 replies; 5+ messages in thread From: Jeff McGee @ 2015-01-07 23:46 UTC (permalink / raw) To: dri-devel Link to the archived drm/i915 patches corresponding to this change: http://lists.freedesktop.org/archives/intel-gfx/2014-December/057817.html _______________________________________________ dri-devel mailing list dri-devel@lists.freedesktop.org http://lists.freedesktop.org/mailman/listinfo/dri-devel ^ permalink raw reply [flat|nested] 5+ messages in thread
end of thread, other threads:[~2015-01-09 16:10 UTC | newest] Thread overview: 5+ messages (download: mbox.gz follow: Atom feed -- links below jump to the message on this page -- 2014-12-18 18:12 [PATCH] intel: Export GT config attributes jeff.mcgee 2014-12-18 18:12 ` jeff.mcgee 2015-01-09 2:56 ` Zhenyu Wang 2015-01-09 16:10 ` Jeff McGee 2015-01-07 23:46 ` Jeff McGee
This is a public inbox, see mirroring instructions for how to clone and mirror all data and code used for this inbox