Now that the necessary GuC-based hardware workarounds have landed, we're finally ready to actually enable compute engines for use by userspace. All of the "under-the-hood" heavy lifting already landed a while back in other series so all that remains now is to add I915_ENGINE_CLASS_COMPUTE to the uapi enum and add the CCS engines to the engine lists for the Xe_HP SDV and DG2.
Userspace (Mesa) is linked in the ABI patch. Existing IGT tests (e.g., i915_hangman) provide test coverage for general engine behavior since compute engines should follow the same general rules as other engines. We've also recently added some additional subtests like igt@gem_reset_stats@shared-reset-domain to cover the user-visible impacts of the compute engines sharing the same hardware reset domain as the render engine.
v2: - Update TLB invalidation register for compute engines and move it to a separate patch since it isn't related to the new uapi. (Tvrtko, Prathap) - Move new kerneldoc for pre-existing engine classes to a separate patch. (Andi) - Drop the compute UMD merge request link for now because it also included some additional multi-tile uapi that we're not ready to upstream just yet. Even if they don't have a disentangled MR ready for reference, we still have the Mesa MR as a key userspace consumer. (Tvrtko)
Cc: Lucas De Marchi lucas.demarchi@intel.com Cc: Tvrtko Ursulin tvrtko.ursulin@linux.intel.com
Daniele Ceraolo Spurio (1): drm/i915: Xe_HP SDV and DG2 have up to 4 CCS engines
Matt Roper (3): drm/i915/uapi: Add kerneldoc for engine class enum drm/i915/xehp: Add register for compute engine's MMIO-based TLB invalidation drm/i915/xehp: Add compute engine ABI
drivers/gpu/drm/i915/gt/intel_engine_user.c | 2 +- drivers/gpu/drm/i915/gt/intel_gt.c | 1 + drivers/gpu/drm/i915/gt/intel_gt_regs.h | 1 + drivers/gpu/drm/i915/i915_drm_client.c | 1 + drivers/gpu/drm/i915/i915_drm_client.h | 2 +- drivers/gpu/drm/i915/i915_pci.c | 6 +- include/uapi/drm/i915_drm.h | 62 +++++++++++++++++++-- 7 files changed, 65 insertions(+), 10 deletions(-)
We'll be adding a new type of engine soon. Let's document the existing engine classes first to help make it clear what each type of engine is used for.
Cc: Andi Shyti andi.shyti@linux.intel.com Signed-off-by: Matt Roper matthew.d.roper@intel.com --- include/uapi/drm/i915_drm.h | 53 ++++++++++++++++++++++++++++++++----- 1 file changed, 47 insertions(+), 6 deletions(-)
diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h index 35ca528803fd..ec000fc6c879 100644 --- a/include/uapi/drm/i915_drm.h +++ b/include/uapi/drm/i915_drm.h @@ -154,21 +154,62 @@ enum i915_mocs_table_index { I915_MOCS_CACHED, };
-/* +/** + * enum drm_i915_gem_engine_class - uapi engine type enumeration + * * Different engines serve different roles, and there may be more than one - * engine serving each role. enum drm_i915_gem_engine_class provides a - * classification of the role of the engine, which may be used when requesting - * operations to be performed on a certain subset of engines, or for providing - * information about that group. + * engine serving each role. This enum provides a classification of the role + * of the engine, which may be used when requesting operations to be performed + * on a certain subset of engines, or for providing information about that + * group. */ enum drm_i915_gem_engine_class { + /** + * @I915_ENGINE_CLASS_RENDER: + * + * Render engines support instructions used for 3D, Compute (GPGPU), + * and programmable media workloads. These instructions fetch data and + * dispatch individual work items to threads that operate in parallel. + * The threads run small programs (called "kernels" or "shaders") on + * the GPU's execution units (EUs). + */ I915_ENGINE_CLASS_RENDER = 0, + + /** + * @I915_ENGINE_CLASS_COPY: + * + * Copy engines (also referred to as "blitters") support instructions + * that move blocks of data from one location in memory to another, + * or that fill a specified location of memory with fixed data. + * Copy engines can perform pre-defined logical or bitwise operations + * on the source, destination, or pattern data. + */ I915_ENGINE_CLASS_COPY = 1, + + /** + * @I915_ENGINE_CLASS_VIDEO: + * + * Video engines (also referred to as "bit stream decode" (BSD) or + * "vdbox") support instructions that perform fixed-function media + * decode and encode. + */ I915_ENGINE_CLASS_VIDEO = 2, + + /** + * @I915_ENGINE_CLASS_VIDEO_ENHANCE: + * + * Video enhancement engines (also referred to as "vebox") support + * instructions related to image enhancement. + */ I915_ENGINE_CLASS_VIDEO_ENHANCE = 3,
- /* should be kept compact */ + /* Values in this enum should be kept compact. */
+ /** + * @I915_ENGINE_CLASS_INVALID: + * + * Placeholder value to represent an invalid engine class assignment. + */ I915_ENGINE_CLASS_INVALID = -1 };
Hi Matt,
On Wed, Apr 27, 2022 at 09:19:23PM -0700, Matt Roper wrote:
We'll be adding a new type of engine soon. Let's document the existing engine classes first to help make it clear what each type of engine is used for.
Cc: Andi Shyti andi.shyti@linux.intel.com Signed-off-by: Matt Roper matthew.d.roper@intel.com
Reviewed-by: Andi Shyti andi.shyti@linux.intel.com
Thank you for splitting this patch, Andi
Compute engines have a separate register that the driver should use to perform MMIO-based TLB invalidation.
Note that the term "context" in this register's bspec description is used to refer to the engine instance (in the same way "context" is used on bspec 46167).
Bspec: 43930 Cc: Prathap Kumar Valsan prathap.kumar.valsan@intel.com Cc: Tvrtko Ursulin tvrtko.ursulin@linux.intel.com Signed-off-by: Matt Roper matthew.d.roper@intel.com --- drivers/gpu/drm/i915/gt/intel_gt.c | 1 + drivers/gpu/drm/i915/gt/intel_gt_regs.h | 1 + 2 files changed, 2 insertions(+)
diff --git a/drivers/gpu/drm/i915/gt/intel_gt.c b/drivers/gpu/drm/i915/gt/intel_gt.c index 92394f13b42f..53307ca0eed0 100644 --- a/drivers/gpu/drm/i915/gt/intel_gt.c +++ b/drivers/gpu/drm/i915/gt/intel_gt.c @@ -1175,6 +1175,7 @@ void intel_gt_invalidate_tlbs(struct intel_gt *gt) [VIDEO_DECODE_CLASS] = GEN12_VD_TLB_INV_CR, [VIDEO_ENHANCEMENT_CLASS] = GEN12_VE_TLB_INV_CR, [COPY_ENGINE_CLASS] = GEN12_BLT_TLB_INV_CR, + [COMPUTE_CLASS] = GEN12_COMPCTX_TLB_INV_CR, }; struct drm_i915_private *i915 = gt->i915; struct intel_uncore *uncore = gt->uncore; diff --git a/drivers/gpu/drm/i915/gt/intel_gt_regs.h b/drivers/gpu/drm/i915/gt/intel_gt_regs.h index a39718a40cc3..a0a49c16babd 100644 --- a/drivers/gpu/drm/i915/gt/intel_gt_regs.h +++ b/drivers/gpu/drm/i915/gt/intel_gt_regs.h @@ -1007,6 +1007,7 @@ #define GEN12_VD_TLB_INV_CR _MMIO(0xcedc) #define GEN12_VE_TLB_INV_CR _MMIO(0xcee0) #define GEN12_BLT_TLB_INV_CR _MMIO(0xcee4) +#define GEN12_COMPCTX_TLB_INV_CR _MMIO(0xcf04)
#define GEN12_MERT_MOD_CTRL _MMIO(0xcf28) #define RENDER_MOD_CTRL _MMIO(0xcf2c)
On 28/04/2022 05:19, Matt Roper wrote:
Compute engines have a separate register that the driver should use to perform MMIO-based TLB invalidation.
Note that the term "context" in this register's bspec description is used to refer to the engine instance (in the same way "context" is used on bspec 46167).
Bspec: 43930 Cc: Prathap Kumar Valsan prathap.kumar.valsan@intel.com Cc: Tvrtko Ursulin tvrtko.ursulin@linux.intel.com Signed-off-by: Matt Roper matthew.d.roper@intel.com
Happy even a blind chicken (me) managed to pick on this piece of correctness. :)
Acked-by: Tvrtko Ursulin tvrtko.ursulin@intel.com
Prathap please r-b since you were the authoritative source in this case.
Regards,
Tvrtko
drivers/gpu/drm/i915/gt/intel_gt.c | 1 + drivers/gpu/drm/i915/gt/intel_gt_regs.h | 1 + 2 files changed, 2 insertions(+)
diff --git a/drivers/gpu/drm/i915/gt/intel_gt.c b/drivers/gpu/drm/i915/gt/intel_gt.c index 92394f13b42f..53307ca0eed0 100644 --- a/drivers/gpu/drm/i915/gt/intel_gt.c +++ b/drivers/gpu/drm/i915/gt/intel_gt.c @@ -1175,6 +1175,7 @@ void intel_gt_invalidate_tlbs(struct intel_gt *gt) [VIDEO_DECODE_CLASS] = GEN12_VD_TLB_INV_CR, [VIDEO_ENHANCEMENT_CLASS] = GEN12_VE_TLB_INV_CR, [COPY_ENGINE_CLASS] = GEN12_BLT_TLB_INV_CR,
}; struct drm_i915_private *i915 = gt->i915; struct intel_uncore *uncore = gt->uncore;[COMPUTE_CLASS] = GEN12_COMPCTX_TLB_INV_CR,
diff --git a/drivers/gpu/drm/i915/gt/intel_gt_regs.h b/drivers/gpu/drm/i915/gt/intel_gt_regs.h index a39718a40cc3..a0a49c16babd 100644 --- a/drivers/gpu/drm/i915/gt/intel_gt_regs.h +++ b/drivers/gpu/drm/i915/gt/intel_gt_regs.h @@ -1007,6 +1007,7 @@ #define GEN12_VD_TLB_INV_CR _MMIO(0xcedc) #define GEN12_VE_TLB_INV_CR _MMIO(0xcee0) #define GEN12_BLT_TLB_INV_CR _MMIO(0xcee4) +#define GEN12_COMPCTX_TLB_INV_CR _MMIO(0xcf04)
#define GEN12_MERT_MOD_CTRL _MMIO(0xcf28) #define RENDER_MOD_CTRL _MMIO(0xcf2c)
On Wed, Apr 27, 2022 at 09:19:24PM -0700, Matt Roper wrote:
Compute engines have a separate register that the driver should use to perform MMIO-based TLB invalidation.
Note that the term "context" in this register's bspec description is used to refer to the engine instance (in the same way "context" is used on bspec 46167).
Bspec: 43930 Cc: Prathap Kumar Valsan prathap.kumar.valsan@intel.com Cc: Tvrtko Ursulin tvrtko.ursulin@linux.intel.com Signed-off-by: Matt Roper matthew.d.roper@intel.com
drivers/gpu/drm/i915/gt/intel_gt.c | 1 + drivers/gpu/drm/i915/gt/intel_gt_regs.h | 1 + 2 files changed, 2 insertions(+)
diff --git a/drivers/gpu/drm/i915/gt/intel_gt.c b/drivers/gpu/drm/i915/gt/intel_gt.c index 92394f13b42f..53307ca0eed0 100644 --- a/drivers/gpu/drm/i915/gt/intel_gt.c +++ b/drivers/gpu/drm/i915/gt/intel_gt.c @@ -1175,6 +1175,7 @@ void intel_gt_invalidate_tlbs(struct intel_gt *gt) [VIDEO_DECODE_CLASS] = GEN12_VD_TLB_INV_CR, [VIDEO_ENHANCEMENT_CLASS] = GEN12_VE_TLB_INV_CR, [COPY_ENGINE_CLASS] = GEN12_BLT_TLB_INV_CR,
}; struct drm_i915_private *i915 = gt->i915; struct intel_uncore *uncore = gt->uncore;[COMPUTE_CLASS] = GEN12_COMPCTX_TLB_INV_CR,
diff --git a/drivers/gpu/drm/i915/gt/intel_gt_regs.h b/drivers/gpu/drm/i915/gt/intel_gt_regs.h index a39718a40cc3..a0a49c16babd 100644 --- a/drivers/gpu/drm/i915/gt/intel_gt_regs.h +++ b/drivers/gpu/drm/i915/gt/intel_gt_regs.h @@ -1007,6 +1007,7 @@ #define GEN12_VD_TLB_INV_CR _MMIO(0xcedc) #define GEN12_VE_TLB_INV_CR _MMIO(0xcee0) #define GEN12_BLT_TLB_INV_CR _MMIO(0xcee4) +#define GEN12_COMPCTX_TLB_INV_CR _MMIO(0xcf04)
#define GEN12_MERT_MOD_CTRL _MMIO(0xcf28) #define RENDER_MOD_CTRL _MMIO(0xcf2c)
Reviewed-by: Prathap Kumar Valsan prathap.kumar.valsan@intel.com
-- 2.35.1
We're now ready to start exposing compute engines to userspace.
v2: - Move kerneldoc for other engine classes to a separate patch. (Andi)
Cc: Daniele Ceraolo Spurio daniele.ceraolospurio@intel.com Cc: Tvrtko Ursulin tvrtko.ursulin@linux.intel.com Cc: Vinay Belgaumkar vinay.belgaumkar@intel.com Cc: Jordan Justen jordan.l.justen@intel.com Cc: Szymon Morek szymon.morek@intel.com UMD (mesa): https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14395 Signed-off-by: Matt Roper matthew.d.roper@intel.com --- drivers/gpu/drm/i915/gt/intel_engine_user.c | 2 +- drivers/gpu/drm/i915/i915_drm_client.c | 1 + drivers/gpu/drm/i915/i915_drm_client.h | 2 +- include/uapi/drm/i915_drm.h | 9 +++++++++ 4 files changed, 12 insertions(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_user.c b/drivers/gpu/drm/i915/gt/intel_engine_user.c index 0f6cd96b459f..46a174f8aa00 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine_user.c +++ b/drivers/gpu/drm/i915/gt/intel_engine_user.c @@ -47,7 +47,7 @@ static const u8 uabi_classes[] = { [COPY_ENGINE_CLASS] = I915_ENGINE_CLASS_COPY, [VIDEO_DECODE_CLASS] = I915_ENGINE_CLASS_VIDEO, [VIDEO_ENHANCEMENT_CLASS] = I915_ENGINE_CLASS_VIDEO_ENHANCE, - /* TODO: Add COMPUTE_CLASS mapping once ABI is available */ + [COMPUTE_CLASS] = I915_ENGINE_CLASS_COMPUTE, };
static int engine_cmp(void *priv, const struct list_head *A, diff --git a/drivers/gpu/drm/i915/i915_drm_client.c b/drivers/gpu/drm/i915/i915_drm_client.c index 475a6f824cad..18d38cb59923 100644 --- a/drivers/gpu/drm/i915/i915_drm_client.c +++ b/drivers/gpu/drm/i915/i915_drm_client.c @@ -81,6 +81,7 @@ static const char * const uabi_class_names[] = { [I915_ENGINE_CLASS_COPY] = "copy", [I915_ENGINE_CLASS_VIDEO] = "video", [I915_ENGINE_CLASS_VIDEO_ENHANCE] = "video-enhance", + [I915_ENGINE_CLASS_COMPUTE] = "compute", };
static u64 busy_add(struct i915_gem_context *ctx, unsigned int class) diff --git a/drivers/gpu/drm/i915/i915_drm_client.h b/drivers/gpu/drm/i915/i915_drm_client.h index 5f5b02b01ba0..f796c5e8e060 100644 --- a/drivers/gpu/drm/i915/i915_drm_client.h +++ b/drivers/gpu/drm/i915/i915_drm_client.h @@ -13,7 +13,7 @@
#include "gt/intel_engine_types.h"
-#define I915_LAST_UABI_ENGINE_CLASS I915_ENGINE_CLASS_VIDEO_ENHANCE +#define I915_LAST_UABI_ENGINE_CLASS I915_ENGINE_CLASS_COMPUTE
struct drm_i915_private;
diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h index ec000fc6c879..a2def7b27009 100644 --- a/include/uapi/drm/i915_drm.h +++ b/include/uapi/drm/i915_drm.h @@ -203,6 +203,15 @@ enum drm_i915_gem_engine_class { */ I915_ENGINE_CLASS_VIDEO_ENHANCE = 3,
+ /** + * @I915_ENGINE_CLASS_COMPUTE: + * + * Compute engines support a subset of the instructions available + * on render engines: compute engines support Compute (GPGPU) and + * programmable media workloads, but do not support the 3D pipeline. + */ + I915_ENGINE_CLASS_COMPUTE = 4, + /* Values in this enum should be kept compact. */
/**
On 28/04/2022 05:19, Matt Roper wrote:
We're now ready to start exposing compute engines to userspace.
v2:
- Move kerneldoc for other engine classes to a separate patch. (Andi)
Cc: Daniele Ceraolo Spurio daniele.ceraolospurio@intel.com Cc: Tvrtko Ursulin tvrtko.ursulin@linux.intel.com Cc: Vinay Belgaumkar vinay.belgaumkar@intel.com Cc: Jordan Justen jordan.l.justen@intel.com Cc: Szymon Morek szymon.morek@intel.com UMD (mesa): https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14395
If in doubt cut it out. :) Works for me.
The rest looks complete.
Acked-by: Tvrtko Ursulin tvrtko.ursulin@intel.com
Anyone cares to smoke test and update intel_gpu_top to display a nice name?
Regards,
Tvrtko
Signed-off-by: Matt Roper matthew.d.roper@intel.com
drivers/gpu/drm/i915/gt/intel_engine_user.c | 2 +- drivers/gpu/drm/i915/i915_drm_client.c | 1 + drivers/gpu/drm/i915/i915_drm_client.h | 2 +- include/uapi/drm/i915_drm.h | 9 +++++++++ 4 files changed, 12 insertions(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/i915/gt/intel_engine_user.c b/drivers/gpu/drm/i915/gt/intel_engine_user.c index 0f6cd96b459f..46a174f8aa00 100644 --- a/drivers/gpu/drm/i915/gt/intel_engine_user.c +++ b/drivers/gpu/drm/i915/gt/intel_engine_user.c @@ -47,7 +47,7 @@ static const u8 uabi_classes[] = { [COPY_ENGINE_CLASS] = I915_ENGINE_CLASS_COPY, [VIDEO_DECODE_CLASS] = I915_ENGINE_CLASS_VIDEO, [VIDEO_ENHANCEMENT_CLASS] = I915_ENGINE_CLASS_VIDEO_ENHANCE,
- /* TODO: Add COMPUTE_CLASS mapping once ABI is available */
[COMPUTE_CLASS] = I915_ENGINE_CLASS_COMPUTE, };
static int engine_cmp(void *priv, const struct list_head *A,
diff --git a/drivers/gpu/drm/i915/i915_drm_client.c b/drivers/gpu/drm/i915/i915_drm_client.c index 475a6f824cad..18d38cb59923 100644 --- a/drivers/gpu/drm/i915/i915_drm_client.c +++ b/drivers/gpu/drm/i915/i915_drm_client.c @@ -81,6 +81,7 @@ static const char * const uabi_class_names[] = { [I915_ENGINE_CLASS_COPY] = "copy", [I915_ENGINE_CLASS_VIDEO] = "video", [I915_ENGINE_CLASS_VIDEO_ENHANCE] = "video-enhance",
[I915_ENGINE_CLASS_COMPUTE] = "compute", };
static u64 busy_add(struct i915_gem_context *ctx, unsigned int class)
diff --git a/drivers/gpu/drm/i915/i915_drm_client.h b/drivers/gpu/drm/i915/i915_drm_client.h index 5f5b02b01ba0..f796c5e8e060 100644 --- a/drivers/gpu/drm/i915/i915_drm_client.h +++ b/drivers/gpu/drm/i915/i915_drm_client.h @@ -13,7 +13,7 @@
#include "gt/intel_engine_types.h"
-#define I915_LAST_UABI_ENGINE_CLASS I915_ENGINE_CLASS_VIDEO_ENHANCE +#define I915_LAST_UABI_ENGINE_CLASS I915_ENGINE_CLASS_COMPUTE
struct drm_i915_private;
diff --git a/include/uapi/drm/i915_drm.h b/include/uapi/drm/i915_drm.h index ec000fc6c879..a2def7b27009 100644 --- a/include/uapi/drm/i915_drm.h +++ b/include/uapi/drm/i915_drm.h @@ -203,6 +203,15 @@ enum drm_i915_gem_engine_class { */ I915_ENGINE_CLASS_VIDEO_ENHANCE = 3,
/**
* @I915_ENGINE_CLASS_COMPUTE:
*
* Compute engines support a subset of the instructions available
* on render engines: compute engines support Compute (GPGPU) and
* programmable media workloads, but do not support the 3D pipeline.
*/
I915_ENGINE_CLASS_COMPUTE = 4,
/* Values in this enum should be kept compact. */
/**
Hi Matt,
On Wed, Apr 27, 2022 at 09:19:25PM -0700, Matt Roper wrote:
We're now ready to start exposing compute engines to userspace.
v2:
- Move kerneldoc for other engine classes to a separate patch. (Andi)
Cc: Daniele Ceraolo Spurio daniele.ceraolospurio@intel.com Cc: Tvrtko Ursulin tvrtko.ursulin@linux.intel.com Cc: Vinay Belgaumkar vinay.belgaumkar@intel.com Cc: Jordan Justen jordan.l.justen@intel.com Cc: Szymon Morek szymon.morek@intel.com UMD (mesa): https://gitlab.freedesktop.org/mesa/mesa/-/merge_requests/14395 Signed-off-by: Matt Roper matthew.d.roper@intel.com
looks good,
Reviewed-by: Andi Shyti andi.shyti@linux.intel.com
Thanks, Andi
From: Daniele Ceraolo Spurio daniele.ceraolospurio@intel.com
Cc: Vinay Belgaumkar vinay.belgaumkar@intel.com Signed-off-by: Daniele Ceraolo Spurio daniele.ceraolospurio@intel.com Signed-off-by: Matt Roper matthew.d.roper@intel.com Reviewed-by: Matt Roper matthew.d.roper@intel.com Reviewed-by: Andi Shyti andi.shyti@linux.intel.com --- drivers/gpu/drm/i915/i915_pci.c | 6 ++++-- 1 file changed, 4 insertions(+), 2 deletions(-)
diff --git a/drivers/gpu/drm/i915/i915_pci.c b/drivers/gpu/drm/i915/i915_pci.c index b60492826478..7739d6c33481 100644 --- a/drivers/gpu/drm/i915/i915_pci.c +++ b/drivers/gpu/drm/i915/i915_pci.c @@ -1037,7 +1037,8 @@ static const struct intel_device_info xehpsdv_info = { BIT(RCS0) | BIT(BCS0) | BIT(VECS0) | BIT(VECS1) | BIT(VECS2) | BIT(VECS3) | BIT(VCS0) | BIT(VCS1) | BIT(VCS2) | BIT(VCS3) | - BIT(VCS4) | BIT(VCS5) | BIT(VCS6) | BIT(VCS7), + BIT(VCS4) | BIT(VCS5) | BIT(VCS6) | BIT(VCS7) | + BIT(CCS0) | BIT(CCS1) | BIT(CCS2) | BIT(CCS3), .require_force_probe = 1, };
@@ -1056,7 +1057,8 @@ static const struct intel_device_info xehpsdv_info = { .platform_engine_mask = \ BIT(RCS0) | BIT(BCS0) | \ BIT(VECS0) | BIT(VECS1) | \ - BIT(VCS0) | BIT(VCS2) + BIT(VCS0) | BIT(VCS2) | \ + BIT(CCS0) | BIT(CCS1) | BIT(CCS2) | BIT(CCS3)
static const struct intel_device_info dg2_info = { DG2_FEATURES,
I did some light testing with our anvil (Vulkan) and iris (OpenGL) Mesa drivers after applying these patches on top of drm-tip tagged intel/CI_DRM_11574. All the unit tests that I tried passed. I also ran the gl_manhattan31 benchmark which used the compute engine for iris compute shader ops.
Series:
Reviewed-by: Jordan Justen jordan.l.justen@intel.com Tested-by: Jordan Justen jordan.l.justen@intel.com
-Jordan
On 2022-04-27 21:19:22, Matt Roper wrote:
Now that the necessary GuC-based hardware workarounds have landed, we're finally ready to actually enable compute engines for use by userspace. All of the "under-the-hood" heavy lifting already landed a while back in other series so all that remains now is to add I915_ENGINE_CLASS_COMPUTE to the uapi enum and add the CCS engines to the engine lists for the Xe_HP SDV and DG2.
Userspace (Mesa) is linked in the ABI patch. Existing IGT tests (e.g., i915_hangman) provide test coverage for general engine behavior since compute engines should follow the same general rules as other engines. We've also recently added some additional subtests like igt@gem_reset_stats@shared-reset-domain to cover the user-visible impacts of the compute engines sharing the same hardware reset domain as the render engine.
v2:
- Update TLB invalidation register for compute engines and move it to a separate patch since it isn't related to the new uapi. (Tvrtko, Prathap)
- Move new kerneldoc for pre-existing engine classes to a separate patch. (Andi)
- Drop the compute UMD merge request link for now because it also included some additional multi-tile uapi that we're not ready to upstream just yet. Even if they don't have a disentangled MR ready for reference, we still have the Mesa MR as a key userspace consumer. (Tvrtko)
Cc: Lucas De Marchi lucas.demarchi@intel.com Cc: Tvrtko Ursulin tvrtko.ursulin@linux.intel.com
Daniele Ceraolo Spurio (1): drm/i915: Xe_HP SDV and DG2 have up to 4 CCS engines
Matt Roper (3): drm/i915/uapi: Add kerneldoc for engine class enum drm/i915/xehp: Add register for compute engine's MMIO-based TLB invalidation drm/i915/xehp: Add compute engine ABI
drivers/gpu/drm/i915/gt/intel_engine_user.c | 2 +- drivers/gpu/drm/i915/gt/intel_gt.c | 1 + drivers/gpu/drm/i915/gt/intel_gt_regs.h | 1 + drivers/gpu/drm/i915/i915_drm_client.c | 1 + drivers/gpu/drm/i915/i915_drm_client.h | 2 +- drivers/gpu/drm/i915/i915_pci.c | 6 +- include/uapi/drm/i915_drm.h | 62 +++++++++++++++++++-- 7 files changed, 65 insertions(+), 10 deletions(-)
-- 2.35.1
dri-devel@lists.freedesktop.org