i915/gem_exec_balancer: Test parallel execbuf
PATCH 1/1 i915/gem_exec_balancer: Test parallel execbuf
Add basic parallel execbuf submission test which more or less just submits the same BB in loop a which does an atomic increment to a memory location. The memory location is checked at the end for the correct value. Different sections use various IOCTL options (e.g. fences, location of BBs, etc...).
In addition to above sections, an additional section ensure the ordering of parallel submission by submitting a spinning batch to 1 individual engine, submit a parallel execbuf to all engines instances within the class, verify none on parallel execbuf make to hardware, release spinner, and finally verify everything has completed.
--- include/drm-uapi/i915_drm.h | 136 ++++++++- lib/intel_ctx.c | 28 +- lib/intel_ctx.h | 2 + lib/intel_reg.h | 5 + tests/i915/gem_exec_balancer.c | 487 +++++++++++++++++++++++++++++++++ 5 files changed, 656 insertions(+), 2 deletions(-)
include/drm-uapi/i915_drm.h +135 -1
@@ -1824,6 +1824,7 @@ struct drm_i915_gem_context_param {* Extensions: * i915_context_engines_load_balance (I915_CONTEXT_ENGINES_EXT_LOAD_BALANCE) * i915_context_engines_bond (I915_CONTEXT_ENGINES_EXT_BOND) +* i915_context_engines_parallel_submit (I915_CONTEXT_ENGINES_EXT_PARALLEL_SUBMIT) */ #define I915_CONTEXT_PARAM_ENGINES 0xa
@@ -2104,10 +2105,137 @@ struct i915_context_engines_bond {* gem_execbuf(drm_fd, &execbuf); */ +/** +* struct i915_context_engines_parallel_submit - Configure engine for +* parallel submission. +* +* Setup a slot in the context engine map to allow multiple BBs to be submitted +* in a single execbuf IOCTL. Those BBs will then be scheduled to run on the GPU +* in parallel. Multiple hardware contexts are created internally in the i915 +* run these BBs. Once a slot is configured for N BBs only N BBs can be +* submitted in each execbuf IOCTL and this is implicit behavior e.g. The user +* doesn't tell the execbuf IOCTL there are N BBs, the execbuf IOCTL knows how +* many BBs there are based on the slot's configuration. The N BBs are the last +* N buffer objects or first N if I915_EXEC_BATCH_FIRST is set. +* +* The default placement behavior is to create implicit bonds between each +* context if each context maps to more than 1 physical engine (e.g. context is +* a virtual engine). Also we only allow contexts of same engine class and these +* contexts must be in logically contiguous order. Examples of the placement +* behavior described below. Lastly, the default is to not allow BBs to +* preempted mid BB rather insert coordinated preemption on all hardware +* contexts between each set of BBs. Flags may be added in the future to change +* both of these default behaviors. +* +* Returns -EINVAL if hardware context placement configuration is invalid or if +* the placement configuration isn't supported on the platform / submission +* interface. +* Returns -ENODEV if extension isn't supported on the platform / submission +* interface. +* +* .. code-block:: none +* +* Example 1 pseudo code: +* CS[X] = generic engine of same class, logical instance X +* INVALID = I915_ENGINE_CLASS_INVALID, I915_ENGINE_CLASS_INVALID_NONE +* set_engines(INVALID) +* set_parallel(engine_index=0, width=2, num_siblings=1, +* engines=CS[0],CS[1]) +* +* Results in the following valid placement: +* CS[0], CS[1] +* +* Example 2 pseudo code: +* CS[X] = generic engine of same class, logical instance X +* INVALID = I915_ENGINE_CLASS_INVALID, I915_ENGINE_CLASS_INVALID_NONE +* set_engines(INVALID) +* set_parallel(engine_index=0, width=2, num_siblings=2, +* engines=CS[0],CS[2],CS[1],CS[3]) +* +* Results in the following valid placements: +* CS[0], CS[1] +* CS[2], CS[3] +* +* This can also be thought of as 2 virtual engines described by 2-D array +* in the engines the field with bonds placed between each index of the +* virtual engines. e.g. CS[0] is bonded to CS[1], CS[2] is bonded to +* CS[3]. +* VE[0] = CS[0], CS[2] +* VE[1] = CS[1], CS[3] +* +* Example 3 pseudo code: +* CS[X] = generic engine of same class, logical instance X +* INVALID = I915_ENGINE_CLASS_INVALID, I915_ENGINE_CLASS_INVALID_NONE +* set_engines(INVALID) +* set_parallel(engine_index=0, width=2, num_siblings=2, +* engines=CS[0],CS[1],CS[1],CS[3]) +* +* Results in the following valid and invalid placements: +* CS[0], CS[1] +* CS[1], CS[3] - Not logical contiguous, return -EINVAL +*/ +struct i915_context_engines_parallel_submit { +/** +* @base: base user extension. +*/ +struct i915_user_extension base; + +/** +* @engine_index: slot for parallel engine +*/ +__u16 engine_index; + +/** +* @width: number of contexts per parallel engine +*/ +__u16 width; + +/** +* @num_siblings: number of siblings per context +*/ +__u16 num_siblings; + +/** +* @mbz16: reserved for future use; must be zero +*/ +__u16 mbz16; + +/** +* @flags: all undefined flags must be zero, currently not defined flags +*/ +__u64 flags; + +/** +* @mbz64: reserved for future use; must be zero +*/ +__u64 mbz64[3 ]; + +/** +* @engines: 2-d array of engine instances to configure parallel engine +* +* length = width (i) * num_siblings (j) +* index = j + i * num_siblings +*/ +struct i915_engine_class_instance engines[0 ]; + +} __packed; + +#define I915_DEFINE_CONTEXT_ENGINES_PARALLEL_SUBMIT (name__,N__ ) struct { \ + struct i915_user_extension base; \ + __u16 engine_index; \ + __u16 width; \ + __u16 num_siblings; \ + __u16 mbz16; \ + __u64 flags; \ + __u64 mbz64[3]; \ + struct i915_engine_class_instance engines[N__]; \ +} __attribute__((packed)) name__ +struct i915_context_param_engines {__u64 extensions;/* linked chain of extension blocks, 0 terminates */ #define I915_CONTEXT_ENGINES_EXT_LOAD_BALANCE 0/* see i915_context_engines_load_balance */ #define I915_CONTEXT_ENGINES_EXT_BOND 1/* see i915_context_engines_bond */ +#define I915_CONTEXT_ENGINES_EXT_PARALLEL_SUBMIT 2/* see i915_context_engines_parallel_submit */ struct i915_engine_class_instance engines[0 ]; } __attribute__((packed));
@@ -2726,14 +2854,20 @@ struct drm_i915_engine_info {/** @flags: Engine flags. */ __u64 flags; +#define I915_ENGINE_INFO_HAS_LOGICAL_INSTANCE (1 << 0)/** @capabilities: Capabilities of this engine. */ __u64 capabilities;#define I915_VIDEO_CLASS_CAPABILITY_HEVC (1 << 0)#define I915_VIDEO_AND_ENHANCE_CLASS_CAPABILITY_SFC (1 << 1) +/** @logical_instance: Logical instance of engine */ +__u16 logical_instance; +/** @rsvd1: Reserved fields. */ -__u64 rsvd1[4 ]; +__u16 rsvd1[3 ]; +/** @rsvd2: Reserved fields. */ +__u64 rsvd2[3 ]; };/**
lib/intel_ctx.c +27 -1
@@ -83,6 +83,7 @@ __context_create_cfg(int fd, const intel_ctx_cfg_t *cfg, uint32_t *ctx_id) {uint64_t ext_root =0 ;I915_DEFINE_CONTEXT_ENGINES_LOAD_BALANCE (balance,GEM_MAX_ENGINES ); +I915_DEFINE_CONTEXT_ENGINES_PARALLEL_SUBMIT (parallel,GEM_MAX_ENGINES );I915_DEFINE_CONTEXT_PARAM_ENGINES (engines,GEM_MAX_ENGINES );struct drm_i915_gem_context_create_ext_setparam engines_param, vm_param;struct drm_i915_gem_context_create_ext_setparam persist_param;
@@ -117,7 +118,29 @@ __context_create_cfg(int fd, const intel_ctx_cfg_t *cfg, uint32_t *ctx_id)unsigned num_logical_engines;memset (&engines,0 ,sizeof (engines)); -if (cfg->load_balance) { +if (cfg->parallel) { +memset (¶llel,0 ,sizeof (parallel)); + + num_logical_engines =1 ; + + parallel.base.name = +I915_CONTEXT_ENGINES_EXT_PARALLEL_SUBMIT ; + + engines.engines[0 ].engine_class = +I915_ENGINE_CLASS_INVALID ; + engines.engines[0 ].engine_instance = +I915_ENGINE_CLASS_INVALID_NONE ; + + parallel.num_siblings = cfg->num_engines; + parallel.width = cfg->width; +for (i =0 ; i < cfg->num_engines * cfg->width; i++) { +igt_assert_eq (cfg->engines[0 ].engine_class, + cfg->engines[i].engine_class); + parallel.engines[i] = cfg->engines[i]; + } + + engines.extensions =to_user_pointer (¶llel); + }else if (cfg->load_balance) {memset (&balance,0 ,sizeof (balance));/* In this case, the first engine is the virtual
@@ -127,6 +150,9 @@ __context_create_cfg(int fd, const intel_ctx_cfg_t *cfg, uint32_t *ctx_id)igt_assert (cfg->num_engines +1 <=GEM_MAX_ENGINES ); num_logical_engines = cfg->num_engines +1 ; + balance.base.name = +I915_CONTEXT_ENGINES_EXT_LOAD_BALANCE ; + engines.engines[0 ].engine_class =I915_ENGINE_CLASS_INVALID ; engines.engines[0].engine_instance =
lib/intel_ctx.h +2 -0
@@ -46,7 +46,9 @@ typedef struct intel_ctx_cfg {uint32_t vm;bool nopersist;bool load_balance; +bool parallel;unsigned int num_engines; +unsigned int width;struct i915_engine_class_instance engines[GEM_MAX_ENGINES ]; } intel_ctx_cfg_t;
lib/intel_reg.h +5 -0
@@ -2642,6 +2642,11 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.#define STATE3D_COLOR_FACTOR ((0x3<<29)|(0x1d<<24)|(0x01<<16)) +/* Atomics */ +#define MI_ATOMIC ((0x2f << 23) | 2) +#define MI_ATOMIC_INLINE_DATA (1 << 18) +#define MI_ATOMIC_ADD (0x7 << 8) +/* Batch */ #define MI_BATCH_BUFFER ((0x30 << 23) | 1)#define MI_BATCH_BUFFER_START (0x31 << 23)
tests/i915/gem_exec_balancer.c +487 -0
517 lines not on this page · raw message
Signature · 1 line
2.32.0
Patchwork
CI[igt-dev] ✓ Fi.CI.BAT: success for i915/gem_exec_balancer: Test parallel execbuf
== Series Details ==
== Series Details ==
Series: i915/gem_exec_balancer: Test parallel execbuf
URL : https://patchwork.freedesktop.org/series/96161/
State : success
== Summary ==
CI Bug Log - changes from CI_DRM_10773 -> IGTPW_6347
====================================================
Summary
-------
**SUCCESS**
No regressions found.
External URL: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/index.html
Possible new issues
-------------------
Here are the unknown changes that may have been introduced in IGTPW_6347:
### IGT changes ###
#### Suppressed ####
The following results come from untrusted machines, tests, or statuses. They do not affect the overall result.
* igt@kms_frontbuffer_tracking@basic:
- {fi-hsw-gt1}: [PASS][1] -> [DMESG-WARN][2]
[1]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/fi-hsw-gt1/igt@[email protected]
[2]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/fi-hsw-gt1/igt@[email protected]
Known issues
------------
Here are the changes found in IGTPW_6347 that come from known issues:
### IGT changes ###
#### Issues hit ####
* igt@core_hotunplug@unbind-rebind:
- fi-elk-e7500: [PASS][3] -> [FAIL][4] ([i915#3194])
[3]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/fi-elk-e7500/igt@[email protected]
[4]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/fi-elk-e7500/igt@[email protected]
* igt@gem_exec_fence@basic-busy@bcs0:
- fi-apl-guc: NOTRUN -> [SKIP][5] ([fdo#109271]) +1 similar issue
[5]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/fi-apl-guc/igt@gem_exec_fence@[email protected]
* igt@i915_hangman@error-state-basic:
- fi-apl-guc: NOTRUN -> [DMESG-WARN][6] ([i915#1610])
[6]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/fi-apl-guc/igt@[email protected]
* igt@i915_selftest@live@execlists:
- fi-bsw-nick: [PASS][7] -> [INCOMPLETE][8] ([i915#2940])
[7]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/fi-bsw-nick/igt@i915_selftest@[email protected]
[8]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/fi-bsw-nick/igt@i915_selftest@[email protected]
* igt@i915_selftest@live@hangcheck:
- fi-snb-2600: [PASS][9] -> [INCOMPLETE][10] ([i915#3921])
[9]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/fi-snb-2600/igt@i915_selftest@[email protected]
[10]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/fi-snb-2600/igt@i915_selftest@[email protected]
* igt@runner@aborted:
- fi-bsw-nick: NOTRUN -> [FAIL][11] ([fdo#109271] / [i915#1436] / [i915#3428] / [i915#4312])
[11]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/fi-bsw-nick/igt@[email protected]
- fi-apl-guc: NOTRUN -> [FAIL][12] ([i915#2426] / [i915#3363] / [i915#4312])
[12]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/fi-apl-guc/igt@[email protected]
#### Possible fixes ####
* igt@i915_pm_rpm@module-reload:
- fi-kbl-guc: [FAIL][13] ([i915#579]) -> [PASS][14]
[13]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/fi-kbl-guc/igt@[email protected]
[14]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/fi-kbl-guc/igt@[email protected]
* igt@i915_selftest@live@gt_heartbeat:
- fi-bdw-samus: [DMESG-FAIL][15] ([i915#541]) -> [PASS][16]
[15]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/fi-bdw-samus/igt@i915_selftest@live@gt_heartbeat.html
[16]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/fi-bdw-samus/igt@i915_selftest@live@gt_heartbeat.html
* igt@i915_selftest@live@hangcheck:
- {fi-hsw-gt1}: [DMESG-WARN][17] ([i915#3303]) -> [PASS][18]
[17]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/fi-hsw-gt1/igt@i915_selftest@[email protected]
[18]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/fi-hsw-gt1/igt@i915_selftest@[email protected]
* igt@kms_flip@basic-flip-vs-modeset@c-dp2:
- fi-cfl-8109u: [DMESG-WARN][19] ([i915#165]) -> [PASS][20] +2 similar issues
[19]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/fi-cfl-8109u/igt@kms_flip@[email protected]
[20]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/fi-cfl-8109u/igt@kms_flip@[email protected]
* igt@kms_flip@basic-plain-flip@c-dp2:
- fi-cfl-8109u: [DMESG-WARN][21] ([i915#165] / [i915#295]) -> [PASS][22] +4 similar issues
[21]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/fi-cfl-8109u/igt@kms_flip@[email protected]
[22]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/fi-cfl-8109u/igt@kms_flip@[email protected]
* igt@kms_frontbuffer_tracking@basic:
- fi-cml-u2: [DMESG-WARN][23] ([i915#4269]) -> [PASS][24]
[23]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/fi-cml-u2/igt@[email protected]
[24]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/fi-cml-u2/igt@[email protected]
#### Warnings ####
* igt@i915_pm_rpm@basic-rte:
- fi-kbl-guc: [SKIP][25] ([fdo#109271]) -> [FAIL][26] ([i915#3049])
[25]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/fi-kbl-guc/igt@[email protected]
[26]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/fi-kbl-guc/igt@[email protected]
* igt@kms_flip@basic-plain-flip@c-dp1:
- fi-cfl-8109u: [DMESG-WARN][27] ([i915#165] / [i915#295]) -> [FAIL][28] ([i915#4165])
[27]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/fi-cfl-8109u/igt@kms_flip@[email protected]
[28]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/fi-cfl-8109u/igt@kms_flip@[email protected]
* igt@kms_pipe_crc_basic@compare-crc-sanitycheck-pipe-b:
- fi-cfl-8109u: [DMESG-WARN][29] ([i915#165] / [i915#295]) -> [DMESG-WARN][30] ([i915#295]) +14 similar issues
[29]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/fi-cfl-8109u/igt@[email protected]
[30]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/fi-cfl-8109u/igt@[email protected]
{name}: This element is suppressed. This means it is ignored when computing
the status of the difference (SUCCESS, WARNING, or FAILURE).
[fdo#109271]: https://bugs.freedesktop.org/show_bug.cgi?id=109271
[i915#1436]: https://gitlab.freedesktop.org/drm/intel/issues/1436
[i915#1610]: https://gitlab.freedesktop.org/drm/intel/issues/1610
[i915#165]: https://gitlab.freedesktop.org/drm/intel/issues/165
[i915#2426]: https://gitlab.freedesktop.org/drm/intel/issues/2426
[i915#2940]: https://gitlab.freedesktop.org/drm/intel/issues/2940
[i915#295]: https://gitlab.freedesktop.org/drm/intel/issues/295
[i915#3049]: https://gitlab.freedesktop.org/drm/intel/issues/3049
[i915#3194]: https://gitlab.freedesktop.org/drm/intel/issues/3194
[i915#3303]: https://gitlab.freedesktop.org/drm/intel/issues/3303
[i915#3363]: https://gitlab.freedesktop.org/drm/intel/issues/3363
[i915#3428]: https://gitlab.freedesktop.org/drm/intel/issues/3428
[i915#3921]: https://gitlab.freedesktop.org/drm/intel/issues/3921
[i915#4165]: https://gitlab.freedesktop.org/drm/intel/issues/4165
[i915#4269]: https://gitlab.freedesktop.org/drm/intel/issues/4269
[i915#4312]: https://gitlab.freedesktop.org/drm/intel/issues/4312
[i915#541]: https://gitlab.freedesktop.org/drm/intel/issues/541
[i915#579]: https://gitlab.freedesktop.org/drm/intel/issues/579
Participating hosts (39 -> 37)
------------------------------
Additional (1): fi-apl-guc
Missing (3): fi-ctg-p8600 fi-bsw-cyan fi-hsw-4200u
Build changes
-------------
* CI: CI-20190529 -> None
* IGT: IGT_6258 -> IGTPW_6347
CI-20190529: 20190529
CI_DRM_10773: fa267509357bd9eb021c3d474fe0980cde18de62 @ git://anongit.freedesktop.org/gfx-ci/linux
IGTPW_6347: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/index.html
IGT_6258: 4c80c71d7dec29b6376846ae96bd04dc0b6e34d9 @ https://gitlab.freedesktop.org/drm/igt-gpu-tools.git
== Testlist changes ==
+igt@gem_exec_balancer@parallel
+igt@gem_exec_balancer@parallel-balancer
+igt@gem_exec_balancer@parallel-bb-first
+igt@gem_exec_balancer@parallel-contexts
+igt@gem_exec_balancer@parallel-keep-in-fence
+igt@gem_exec_balancer@parallel-keep-submit-fence
+igt@gem_exec_balancer@parallel-ordering
+igt@gem_exec_balancer@parallel-out-fence
== Logs ==
For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/index.html
Patchwork
CI[igt-dev] ✓ Fi.CI.IGT: success for i915/gem_exec_balancer: Test parallel execbuf
In reply to Matthew Brost== Series Details ==
== Series Details ==
Series: i915/gem_exec_balancer: Test parallel execbuf
URL : https://patchwork.freedesktop.org/series/96161/
State : success
== Summary ==
CI Bug Log - changes from CI_DRM_10773_full -> IGTPW_6347_full
====================================================
Summary
-------
**SUCCESS**
No regressions found.
External URL: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/index.html
Possible new issues
-------------------
Here are the unknown changes that may have been introduced in IGTPW_6347_full:
### IGT changes ###
#### Possible regressions ####
* {igt@gem_exec_balancer@parallel-contexts} (NEW):
- shard-tglb: NOTRUN -> [SKIP][1] +5 similar issues
[1]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb8/igt@[email protected]
* {igt@gem_exec_balancer@parallel-out-fence} (NEW):
- shard-iclb: NOTRUN -> [SKIP][2] +7 similar issues
[2]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-iclb7/igt@[email protected]
New tests
---------
New tests have been introduced between CI_DRM_10773_full and IGTPW_6347_full:
### New IGT tests (8) ###
* igt@gem_exec_balancer@parallel:
- Statuses : 6 skip(s)
- Exec time: [0.0] s
* igt@gem_exec_balancer@parallel-balancer:
- Statuses : 4 skip(s)
- Exec time: [0.0] s
* igt@gem_exec_balancer@parallel-bb-first:
- Statuses : 4 skip(s)
- Exec time: [0.0] s
* igt@gem_exec_balancer@parallel-contexts:
- Statuses : 6 skip(s)
- Exec time: [0.0] s
* igt@gem_exec_balancer@parallel-keep-in-fence:
- Statuses : 5 skip(s)
- Exec time: [0.0] s
* igt@gem_exec_balancer@parallel-keep-submit-fence:
- Statuses : 5 skip(s)
- Exec time: [0.0] s
* igt@gem_exec_balancer@parallel-ordering:
- Statuses : 5 skip(s)
- Exec time: [0.0] s
* igt@gem_exec_balancer@parallel-out-fence:
- Statuses : 5 skip(s)
- Exec time: [0.0] s
Known issues
------------
Here are the changes found in IGTPW_6347_full that come from known issues:
### IGT changes ###
#### Issues hit ####
* igt@feature_discovery@display-2x:
- shard-tglb: NOTRUN -> [SKIP][3] ([i915#1839])
[3]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb6/igt@[email protected]
* igt@gem_create@create-massive:
- shard-snb: NOTRUN -> [DMESG-WARN][4] ([i915#3002])
[4]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-snb5/igt@[email protected]
- shard-kbl: NOTRUN -> [DMESG-WARN][5] ([i915#3002])
[5]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-kbl3/igt@[email protected]
- shard-apl: NOTRUN -> [DMESG-WARN][6] ([i915#3002])
[6]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-apl1/igt@[email protected]
* igt@gem_ctx_persistence@legacy-engines-hostile-preempt:
- shard-snb: NOTRUN -> [SKIP][7] ([fdo#109271] / [i915#1099]) +4 similar issues
[7]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-snb2/igt@[email protected]
* igt@gem_eio@unwedge-stress:
- shard-glk: [PASS][8] -> [TIMEOUT][9] ([i915#2369])
[8]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/shard-glk7/igt@[email protected]
[9]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-glk8/igt@[email protected]
- shard-tglb: [PASS][10] -> [TIMEOUT][11] ([i915#2369] / [i915#3063] / [i915#3648])
[10]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/shard-tglb3/igt@[email protected]
[11]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb8/igt@[email protected]
* {igt@gem_exec_balancer@parallel-out-fence} (NEW):
- shard-glk: NOTRUN -> [SKIP][12] ([fdo#109271]) +42 similar issues
[12]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-glk6/igt@[email protected]
* igt@gem_exec_fair@basic-none@vcs1:
- shard-kbl: [PASS][13] -> [FAIL][14] ([i915#2842])
[13]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/shard-kbl4/igt@gem_exec_fair@[email protected]
[14]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-kbl2/igt@gem_exec_fair@[email protected]
* igt@gem_exec_fair@basic-pace-solo@rcs0:
- shard-glk: [PASS][15] -> [FAIL][16] ([i915#2842]) +1 similar issue
[15]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/shard-glk9/igt@gem_exec_fair@[email protected]
[16]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-glk1/igt@gem_exec_fair@[email protected]
* igt@gem_exec_fair@basic-pace@bcs0:
- shard-tglb: [PASS][17] -> [FAIL][18] ([i915#2842]) +1 similar issue
[17]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/shard-tglb8/igt@gem_exec_fair@[email protected]
[18]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb1/igt@gem_exec_fair@[email protected]
* igt@gem_exec_fair@basic-throttle@rcs0:
- shard-tglb: NOTRUN -> [FAIL][19] ([i915#2842])
[19]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb5/igt@gem_exec_fair@[email protected]
* igt@gem_exec_flush@basic-batch-kernel-default-cmd:
- shard-iclb: NOTRUN -> [SKIP][20] ([fdo#109313])
[20]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-iclb8/igt@[email protected]
- shard-tglb: NOTRUN -> [SKIP][21] ([fdo#109313])
[21]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb6/igt@[email protected]
* igt@gem_exec_params@no-bsd:
- shard-tglb: NOTRUN -> [SKIP][22] ([fdo#109283])
[22]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb7/igt@[email protected]
- shard-iclb: NOTRUN -> [SKIP][23] ([fdo#109283])
[23]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-iclb6/igt@[email protected]
* igt@gem_exec_params@secure-non-root:
- shard-tglb: NOTRUN -> [SKIP][24] ([fdo#112283])
[24]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb6/igt@[email protected]
* igt@gem_exec_whisper@basic-normal:
- shard-glk: [PASS][25] -> [DMESG-WARN][26] ([i915#118])
[25]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/shard-glk9/igt@[email protected]
[26]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-glk6/igt@[email protected]
* igt@gem_pxp@create-protected-buffer:
- shard-iclb: NOTRUN -> [SKIP][27] ([i915#4270])
[27]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-iclb7/igt@[email protected]
- shard-tglb: NOTRUN -> [SKIP][28] ([i915#4270])
[28]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb1/igt@[email protected]
* igt@gem_userptr_blits@dmabuf-sync:
- shard-apl: NOTRUN -> [SKIP][29] ([fdo#109271] / [i915#3323])
[29]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-apl6/igt@[email protected]
* igt@gem_userptr_blits@readonly-unsync:
- shard-tglb: NOTRUN -> [SKIP][30] ([i915#3297])
[30]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb8/igt@[email protected]
* igt@gem_userptr_blits@vma-merge:
- shard-kbl: NOTRUN -> [FAIL][31] ([i915#3318])
[31]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-kbl1/igt@[email protected]
* igt@gen7_exec_parse@bitmasks:
- shard-tglb: NOTRUN -> [SKIP][32] ([fdo#109289])
[32]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb7/igt@[email protected]
* igt@gen9_exec_parse@allowed-all:
- shard-kbl: [PASS][33] -> [DMESG-WARN][34] ([i915#1436] / [i915#716])
[33]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/shard-kbl2/igt@[email protected]
[34]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-kbl7/igt@[email protected]
* igt@gen9_exec_parse@shadow-peek:
- shard-tglb: NOTRUN -> [SKIP][35] ([i915#2856])
[35]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb2/igt@[email protected]
- shard-iclb: NOTRUN -> [SKIP][36] ([i915#2856])
[36]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-iclb2/igt@[email protected]
* igt@i915_pm_dc@dc9-dpms:
- shard-iclb: [PASS][37] -> [FAIL][38] ([i915#4275])
[37]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/shard-iclb4/igt@[email protected]
[38]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-iclb4/igt@[email protected]
* igt@i915_pm_rpm@dpms-non-lpsp:
- shard-tglb: NOTRUN -> [SKIP][39] ([fdo#111644] / [i915#1397] / [i915#2411])
[39]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb6/igt@[email protected]
* igt@i915_pm_sseu@full-enable:
- shard-tglb: NOTRUN -> [SKIP][40] ([fdo#109288])
[40]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb1/igt@[email protected]
* igt@i915_query@query-topology-unsupported:
- shard-tglb: NOTRUN -> [SKIP][41] ([fdo#109302])
[41]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb8/igt@[email protected]
* igt@i915_selftest@live@hangcheck:
- shard-snb: NOTRUN -> [INCOMPLETE][42] ([i915#3921])
[42]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-snb6/igt@i915_selftest@[email protected]
* igt@i915_suspend@fence-restore-tiled2untiled:
- shard-apl: [PASS][43] -> [DMESG-WARN][44] ([i915#180]) +1 similar issue
[43]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/shard-apl3/igt@[email protected]
[44]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-apl1/igt@[email protected]
* igt@i915_suspend@sysfs-reader:
- shard-tglb: [PASS][45] -> [INCOMPLETE][46] ([i915#456])
[45]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/shard-tglb5/igt@[email protected]
[46]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb7/igt@[email protected]
* igt@kms_big_fb@linear-8bpp-rotate-270:
- shard-tglb: NOTRUN -> [SKIP][47] ([fdo#111614]) +3 similar issues
[47]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb3/igt@[email protected]
* igt@kms_big_fb@x-tiled-16bpp-rotate-270:
- shard-iclb: NOTRUN -> [SKIP][48] ([fdo#110725] / [fdo#111614])
[48]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-iclb5/igt@[email protected]
* igt@kms_big_fb@yf-tiled-max-hw-stride-32bpp-rotate-0-hflip:
- shard-apl: NOTRUN -> [SKIP][49] ([fdo#109271] / [i915#3777]) +1 similar issue
[49]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-apl8/igt@[email protected]
* igt@kms_big_fb@yf-tiled-max-hw-stride-64bpp-rotate-0-hflip:
- shard-tglb: NOTRUN -> [SKIP][50] ([fdo#111615])
[50]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb2/igt@[email protected]
* igt@kms_big_joiner@invalid-modeset:
- shard-tglb: NOTRUN -> [SKIP][51] ([i915#2705])
[51]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb2/igt@[email protected]
* igt@kms_ccs@pipe-a-crc-sprite-planes-basic-y_tiled_gen12_mc_ccs:
- shard-glk: NOTRUN -> [SKIP][52] ([fdo#109271] / [i915#3886]) +4 similar issues
[52]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-glk2/igt@kms_ccs@pipe-a-crc-sprite-planes-basic-y_tiled_gen12_mc_ccs.html
- shard-iclb: NOTRUN -> [SKIP][53] ([fdo#109278] / [i915#3886]) +1 similar issue
[53]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-iclb8/igt@kms_ccs@pipe-a-crc-sprite-planes-basic-y_tiled_gen12_mc_ccs.html
- shard-tglb: NOTRUN -> [SKIP][54] ([i915#3689] / [i915#3886])
[54]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb1/igt@kms_ccs@pipe-a-crc-sprite-planes-basic-y_tiled_gen12_mc_ccs.html
* igt@kms_ccs@pipe-a-missing-ccs-buffer-y_tiled_gen12_rc_ccs_cc:
- shard-apl: NOTRUN -> [SKIP][55] ([fdo#109271] / [i915#3886]) +7 similar issues
[55]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-apl3/igt@kms_ccs@pipe-a-missing-ccs-buffer-y_tiled_gen12_rc_ccs_cc.html
* igt@kms_ccs@pipe-b-ccs-on-another-bo-y_tiled_gen12_rc_ccs:
- shard-iclb: NOTRUN -> [SKIP][56] ([fdo#109278]) +1 similar issue
[56]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-iclb8/igt@kms_ccs@pipe-b-ccs-on-another-bo-y_tiled_gen12_rc_ccs.html
* igt@kms_ccs@pipe-b-missing-ccs-buffer-y_tiled_gen12_rc_ccs_cc:
- shard-kbl: NOTRUN -> [SKIP][57] ([fdo#109271] / [i915#3886]) +10 similar issues
[57]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-kbl4/igt@kms_ccs@pipe-b-missing-ccs-buffer-y_tiled_gen12_rc_ccs_cc.html
* igt@kms_ccs@pipe-d-missing-ccs-buffer-yf_tiled_ccs:
- shard-tglb: NOTRUN -> [SKIP][58] ([i915#3689]) +8 similar issues
[58]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb8/igt@kms_ccs@pipe-d-missing-ccs-buffer-yf_tiled_ccs.html
* igt@kms_cdclk@mode-transition:
- shard-apl: NOTRUN -> [SKIP][59] ([fdo#109271]) +254 similar issues
[59]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-apl7/igt@[email protected]
* igt@kms_chamelium@vga-hpd:
- shard-apl: NOTRUN -> [SKIP][60] ([fdo#109271] / [fdo#111827]) +15 similar issues
[60]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-apl1/igt@[email protected]
* igt@kms_chamelium@vga-hpd-after-suspend:
- shard-glk: NOTRUN -> [SKIP][61] ([fdo#109271] / [fdo#111827])
[61]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-glk5/igt@[email protected]
* igt@kms_color_chamelium@pipe-a-ctm-0-25:
- shard-snb: NOTRUN -> [SKIP][62] ([fdo#109271] / [fdo#111827]) +15 similar issues
[62]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-snb6/igt@[email protected]
* igt@kms_color_chamelium@pipe-b-ctm-0-75:
- shard-tglb: NOTRUN -> [SKIP][63] ([fdo#109284] / [fdo#111827]) +5 similar issues
[63]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb1/igt@[email protected]
* igt@kms_color_chamelium@pipe-c-ctm-0-25:
- shard-kbl: NOTRUN -> [SKIP][64] ([fdo#109271] / [fdo#111827]) +18 similar issues
[64]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-kbl1/igt@[email protected]
* igt@kms_content_protection@srm:
- shard-apl: NOTRUN -> [TIMEOUT][65] ([i915#1319])
[65]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-apl8/igt@[email protected]
* igt@kms_content_protection@uevent:
- shard-apl: NOTRUN -> [FAIL][66] ([i915#2105])
[66]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-apl2/igt@[email protected]
* igt@kms_cursor_crc@pipe-c-cursor-512x170-offscreen:
- shard-tglb: NOTRUN -> [SKIP][67] ([fdo#109279] / [i915#3359]) +2 similar issues
[67]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb8/igt@[email protected]
* igt@kms_cursor_crc@pipe-c-cursor-max-size-onscreen:
- shard-tglb: NOTRUN -> [SKIP][68] ([i915#3359]) +3 similar issues
[68]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb1/igt@[email protected]
* igt@kms_cursor_crc@pipe-d-cursor-32x32-random:
- shard-tglb: NOTRUN -> [SKIP][69] ([i915#3319])
[69]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb2/igt@[email protected]
* igt@kms_cursor_edge_walk@pipe-d-128x128-right-edge:
- shard-snb: NOTRUN -> [SKIP][70] ([fdo#109271]) +409 similar issues
[70]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-snb2/igt@[email protected]
* igt@kms_cursor_legacy@cursorb-vs-flipa-atomic:
- shard-iclb: NOTRUN -> [SKIP][71] ([fdo#109274] / [fdo#109278])
[71]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-iclb2/igt@[email protected]
* igt@kms_cursor_legacy@flip-vs-cursor-atomic-transitions:
- shard-glk: [PASS][72] -> [FAIL][73] ([i915#2346])
[72]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/shard-glk8/igt@[email protected]
[73]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-glk7/igt@[email protected]
* igt@kms_dp_tiled_display@basic-test-pattern-with-chamelium:
- shard-tglb: NOTRUN -> [SKIP][74] ([i915#3528])
[74]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb5/igt@[email protected]
* igt@kms_flip@2x-plain-flip-ts-check-interruptible@ac-hdmi-a1-hdmi-a2:
- shard-glk: [PASS][75] -> [FAIL][76] ([i915#2122])
[75]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/shard-glk3/igt@kms_flip@[email protected]
[76]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-glk7/igt@kms_flip@[email protected]
* igt@kms_flip@flip-vs-suspend-interruptible@a-dp1:
- shard-kbl: [PASS][77] -> [DMESG-WARN][78] ([i915#180]) +4 similar issues
[77]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/shard-kbl2/igt@kms_flip@[email protected]
[78]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-kbl7/igt@kms_flip@[email protected]
* igt@kms_frontbuffer_tracking@fbc-2p-scndscrn-shrfb-msflip-blt:
- shard-tglb: NOTRUN -> [SKIP][79] ([fdo#111825]) +20 similar issues
[79]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb6/igt@[email protected]
- shard-iclb: NOTRUN -> [SKIP][80] ([fdo#109280]) +7 similar issues
[80]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-iclb4/igt@[email protected]
* igt@kms_frontbuffer_tracking@fbc-suspend:
- shard-apl: NOTRUN -> [DMESG-WARN][81] ([i915#180])
[81]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-apl8/igt@[email protected]
* igt@kms_frontbuffer_tracking@psr-suspend:
- shard-tglb: [PASS][82] -> [INCOMPLETE][83] ([i915#2411] / [i915#456])
[82]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/shard-tglb2/igt@[email protected]
[83]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb7/igt@[email protected]
* igt@kms_hdr@bpc-switch-suspend:
- shard-kbl: NOTRUN -> [DMESG-WARN][84] ([i915#180]) +6 similar issues
[84]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-kbl4/igt@[email protected]
* igt@kms_hdr@static-toggle-dpms:
- shard-tglb: NOTRUN -> [SKIP][85] ([i915#1187])
[85]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb5/igt@[email protected]
* igt@kms_pipe_crc_basic@disable-crc-after-crtc-pipe-d:
- shard-kbl: NOTRUN -> [SKIP][86] ([fdo#109271] / [i915#533]) +1 similar issue
[86]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-kbl6/igt@[email protected]
* igt@kms_pipe_crc_basic@suspend-read-crc-pipe-c:
- shard-tglb: [PASS][87] -> [INCOMPLETE][88] ([i915#2828] / [i915#456])
[87]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/shard-tglb1/igt@[email protected]
[88]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb7/igt@[email protected]
* igt@kms_plane@plane-panning-bottom-right-suspend@pipe-b-planes:
- shard-tglb: NOTRUN -> [INCOMPLETE][89] ([i915#456])
[89]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb7/igt@kms_plane@[email protected]
* igt@kms_plane_alpha_blend@pipe-a-alpha-7efc:
- shard-kbl: NOTRUN -> [FAIL][90] ([fdo#108145] / [i915#265]) +1 similar issue
[90]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-kbl3/igt@[email protected]
* igt@kms_plane_alpha_blend@pipe-a-alpha-transparent-fb:
- shard-kbl: NOTRUN -> [FAIL][91] ([i915#265])
[91]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-kbl6/igt@[email protected]
* igt@kms_plane_alpha_blend@pipe-b-constant-alpha-max:
- shard-apl: NOTRUN -> [FAIL][92] ([fdo#108145] / [i915#265]) +1 similar issue
[92]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-apl7/igt@[email protected]
* igt@kms_plane_alpha_blend@pipe-c-alpha-transparent-fb:
- shard-apl: NOTRUN -> [FAIL][93] ([i915#265])
[93]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-apl1/igt@[email protected]
* igt@kms_plane_lowres@pipe-a-tiling-y:
- shard-iclb: NOTRUN -> [SKIP][94] ([i915#3536])
[94]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-iclb3/igt@[email protected]
- shard-tglb: NOTRUN -> [SKIP][95] ([i915#3536])
[95]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb2/igt@[email protected]
* igt@kms_psr2_sf@overlay-plane-update-sf-dmg-area-2:
- shard-apl: NOTRUN -> [SKIP][96] ([fdo#109271] / [i915#658]) +4 similar issues
[96]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-apl3/igt@[email protected]
* igt@kms_psr2_sf@primary-plane-update-sf-dmg-area-1:
- shard-kbl: NOTRUN -> [SKIP][97] ([fdo#109271] / [i915#658]) +5 similar issues
[97]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-kbl2/igt@[email protected]
* igt@kms_psr2_sf@primary-plane-update-sf-dmg-area-5:
- shard-tglb: NOTRUN -> [SKIP][98] ([i915#2920]) +1 similar issue
[98]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb2/igt@[email protected]
* igt@kms_psr2_su@page_flip:
- shard-tglb: NOTRUN -> [SKIP][99] ([i915#1911])
[99]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb2/igt@kms_psr2_su@page_flip.html
* igt@kms_psr@psr2_cursor_plane_onoff:
- shard-tglb: NOTRUN -> [FAIL][100] ([i915#132] / [i915#3467])
[100]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb1/igt@kms_psr@psr2_cursor_plane_onoff.html
* igt@kms_psr@psr2_no_drrs:
- shard-iclb: [PASS][101] -> [SKIP][102] ([fdo#109441]) +2 similar issues
[101]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/shard-iclb2/igt@kms_psr@psr2_no_drrs.html
[102]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-iclb5/igt@kms_psr@psr2_no_drrs.html
* igt@kms_universal_plane@disable-primary-vs-flip-pipe-d:
- shard-kbl: NOTRUN -> [SKIP][103] ([fdo#109271]) +231 similar issues
[103]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-kbl1/igt@[email protected]
* igt@kms_vblank@pipe-d-wait-idle:
- shard-apl: NOTRUN -> [SKIP][104] ([fdo#109271] / [i915#533]) +2 similar issues
[104]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-apl1/igt@[email protected]
* igt@kms_writeback@writeback-check-output:
- shard-apl: NOTRUN -> [SKIP][105] ([fdo#109271] / [i915#2437]) +1 similar issue
[105]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-apl7/igt@[email protected]
* igt@nouveau_crc@pipe-d-ctx-flip-skip-current-frame:
- shard-tglb: NOTRUN -> [SKIP][106] ([i915#2530])
[106]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb2/igt@[email protected]
* igt@prime_nv_test@i915_import_cpu_mmap:
- shard-iclb: NOTRUN -> [SKIP][107] ([fdo#109291])
[107]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-iclb5/igt@prime_nv_test@i915_import_cpu_mmap.html
- shard-tglb: NOTRUN -> [SKIP][108] ([fdo#109291])
[108]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb8/igt@prime_nv_test@i915_import_cpu_mmap.html
* igt@prime_vgem@basic-userptr:
- shard-tglb: NOTRUN -> [SKIP][109] ([i915#3301])
[109]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb1/igt@[email protected]
* igt@prime_vgem@fence-write-hang:
- shard-tglb: NOTRUN -> [SKIP][110] ([fdo#109295])
[110]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb8/igt@[email protected]
* igt@sysfs_clients@fair-7:
- shard-apl: NOTRUN -> [SKIP][111] ([fdo#109271] / [i915#2994]) +4 similar issues
[111]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-apl8/igt@[email protected]
* igt@sysfs_clients@sema-10:
- shard-tglb: NOTRUN -> [SKIP][112] ([i915#2994]) +1 similar issue
[112]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb7/igt@[email protected]
- shard-kbl: NOTRUN -> [SKIP][113] ([fdo#109271] / [i915#2994])
[113]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-kbl1/igt@[email protected]
#### Possible fixes ####
* igt@gem_exec_fair@basic-pace@vecs0:
- shard-glk: [FAIL][114] ([i915#2842]) -> [PASS][115] +2 similar issues
[114]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/shard-glk3/igt@gem_exec_fair@[email protected]
[115]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-glk2/igt@gem_exec_fair@[email protected]
- shard-iclb: [FAIL][116] ([i915#2842]) -> [PASS][117]
[116]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/shard-iclb8/igt@gem_exec_fair@[email protected]
[117]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-iclb8/igt@gem_exec_fair@[email protected]
* igt@gem_exec_fair@basic-throttle@rcs0:
- shard-iclb: [FAIL][118] ([i915#2849]) -> [PASS][119]
[118]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/shard-iclb4/igt@gem_exec_fair@[email protected]
[119]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-iclb7/igt@gem_exec_fair@[email protected]
* igt@gem_huc_copy@huc-copy:
- shard-tglb: [SKIP][120] ([i915#2190]) -> [PASS][121]
[120]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/shard-tglb6/igt@[email protected]
[121]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb2/igt@[email protected]
* igt@i915_pm_dc@dc6-psr:
- shard-iclb: [FAIL][122] ([i915#454]) -> [PASS][123]
[122]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/shard-iclb6/igt@[email protected]
[123]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-iclb7/igt@[email protected]
* igt@i915_suspend@forcewake:
- shard-tglb: [INCOMPLETE][124] ([i915#2411] / [i915#456]) -> [PASS][125] +1 similar issue
[124]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/shard-tglb7/igt@[email protected]
[125]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb5/igt@[email protected]
* igt@kms_big_fb@linear-32bpp-rotate-180:
- shard-glk: [DMESG-WARN][126] ([i915#118]) -> [PASS][127]
[126]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/shard-glk6/igt@[email protected]
[127]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-glk5/igt@[email protected]
* igt@kms_cursor_crc@pipe-c-cursor-suspend:
- shard-kbl: [DMESG-WARN][128] ([i915#180]) -> [PASS][129] +5 similar issues
[128]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/shard-kbl6/igt@[email protected]
[129]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-kbl2/igt@[email protected]
* igt@kms_fbcon_fbt@psr-suspend:
- shard-tglb: [INCOMPLETE][130] ([i915#456]) -> [PASS][131]
[130]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/shard-tglb7/igt@[email protected]
[131]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb1/igt@[email protected]
* igt@kms_flip@flip-vs-expired-vblank-interruptible@a-dp1:
- shard-kbl: [FAIL][132] ([i915#79]) -> [PASS][133]
[132]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/shard-kbl6/igt@kms_flip@[email protected]
[133]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-kbl6/igt@kms_flip@[email protected]
* igt@kms_frontbuffer_tracking@fbc-suspend:
- shard-tglb: [INCOMPLETE][134] ([i915#2828] / [i915#456]) -> [PASS][135]
[134]: https://intel-gfx-ci.01.org/tree/drm-tip/CI_DRM_10773/shard-tglb7/igt@[email protected]
[135]: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/shard-tglb1/igt@kms_frontbuffer_trackin
== Logs ==
For more details see: https://intel-gfx-ci.01.org/tree/drm-tip/IGTPW_6347/index.html
include/drm-uapi/i915_drm.h
Quoted 25 lines · Matthew Brost
Add basic parallel execbuf submission test which more or less just submits the same BB in loop a which does an atomic increment to a memory location. The memory location is checked at the end for the correct value. Different sections use various IOCTL options (e.g. fences, location of BBs, etc...).
In addition to above sections, an additional section ensure the ordering of parallel submission by submitting a spinning batch to 1 individual engine, submit a parallel execbuf to all engines instances within the class, verify none on parallel execbuf make to hardware, release spinner, and finally verify everything has completed.
Signed-off-by Matthew Brost <[email protected]>--- include/drm-uapi/i915_drm.h | 136 ++++++++- lib/intel_ctx.c | 28 +- lib/intel_ctx.h | 2 + lib/intel_reg.h | 5 + tests/i915/gem_exec_balancer.c | 487 +++++++++++++++++++++++++++++++++ 5 files changed, 656 insertions(+), 2 deletions(-)include/drm-uapi/i915_drm.h +0
-0
The uapi file needs to be in sync with drm-next. If the changes have already reached drm-next then we should just have a separate patch doing the file sync, otherwise these defs must move to lib/i915/i915_drm_local.h
Quoted 181 lines
include/drm-uapi/i915_drm.h +135
-1@@ -1824,6 +1824,7 @@ struct drm_i915_gem_context_param {* Extensions: * i915_context_engines_load_balance (I915_CONTEXT_ENGINES_EXT_LOAD_BALANCE) * i915_context_engines_bond (I915_CONTEXT_ENGINES_EXT_BOND) +* i915_context_engines_parallel_submit (I915_CONTEXT_ENGINES_EXT_PARALLEL_SUBMIT) */ #define I915_CONTEXT_PARAM_ENGINES 0xa@@ -2104,10 +2105,137 @@ struct i915_context_engines_bond {* gem_execbuf(drm_fd, &execbuf); */ +/** +* struct i915_context_engines_parallel_submit - Configure engine for +* parallel submission. +* +* Setup a slot in the context engine map to allow multiple BBs to be submitted +* in a single execbuf IOCTL. Those BBs will then be scheduled to run on the GPU +* in parallel. Multiple hardware contexts are created internally in the i915 +* run these BBs. Once a slot is configured for N BBs only N BBs can be +* submitted in each execbuf IOCTL and this is implicit behavior e.g. The user +* doesn't tell the execbuf IOCTL there are N BBs, the execbuf IOCTL knows how +* many BBs there are based on the slot's configuration. The N BBs are the last +* N buffer objects or first N if I915_EXEC_BATCH_FIRST is set. +* +* The default placement behavior is to create implicit bonds between each +* context if each context maps to more than 1 physical engine (e.g. context is +* a virtual engine). Also we only allow contexts of same engine class and these +* contexts must be in logically contiguous order. Examples of the placement +* behavior described below. Lastly, the default is to not allow BBs to +* preempted mid BB rather insert coordinated preemption on all hardware +* contexts between each set of BBs. Flags may be added in the future to change +* both of these default behaviors. +* +* Returns -EINVAL if hardware context placement configuration is invalid or if +* the placement configuration isn't supported on the platform / submission +* interface. +* Returns -ENODEV if extension isn't supported on the platform / submission +* interface. +* +* .. code-block:: none +* +* Example 1 pseudo code: +* CS[X] = generic engine of same class, logical instance X +* INVALID = I915_ENGINE_CLASS_INVALID, I915_ENGINE_CLASS_INVALID_NONE +* set_engines(INVALID) +* set_parallel(engine_index=0, width=2, num_siblings=1, +* engines=CS[0],CS[1]) +* +* Results in the following valid placement: +* CS[0], CS[1] +* +* Example 2 pseudo code: +* CS[X] = generic engine of same class, logical instance X +* INVALID = I915_ENGINE_CLASS_INVALID, I915_ENGINE_CLASS_INVALID_NONE +* set_engines(INVALID) +* set_parallel(engine_index=0, width=2, num_siblings=2, +* engines=CS[0],CS[2],CS[1],CS[3]) +* +* Results in the following valid placements: +* CS[0], CS[1] +* CS[2], CS[3] +* +* This can also be thought of as 2 virtual engines described by 2-D array +* in the engines the field with bonds placed between each index of the +* virtual engines. e.g. CS[0] is bonded to CS[1], CS[2] is bonded to +* CS[3]. +* VE[0] = CS[0], CS[2] +* VE[1] = CS[1], CS[3] +* +* Example 3 pseudo code: +* CS[X] = generic engine of same class, logical instance X +* INVALID = I915_ENGINE_CLASS_INVALID, I915_ENGINE_CLASS_INVALID_NONE +* set_engines(INVALID) +* set_parallel(engine_index=0, width=2, num_siblings=2, +* engines=CS[0],CS[1],CS[1],CS[3]) +* +* Results in the following valid and invalid placements: +* CS[0], CS[1] +* CS[1], CS[3] - Not logical contiguous, return -EINVAL +*/ +struct i915_context_engines_parallel_submit { +/** +* @base: base user extension. +*/ +struct i915_user_extension base; + +/** +* @engine_index: slot for parallel engine +*/ +__u16 engine_index; + +/** +* @width: number of contexts per parallel engine +*/ +__u16 width; + +/** +* @num_siblings: number of siblings per context +*/ +__u16 num_siblings; + +/** +* @mbz16: reserved for future use; must be zero +*/ +__u16 mbz16; + +/** +* @flags: all undefined flags must be zero, currently not defined flags +*/ +__u64 flags; + +/** +* @mbz64: reserved for future use; must be zero +*/ +__u64 mbz64[3 ]; + +/** +* @engines: 2-d array of engine instances to configure parallel engine +* +* length = width (i) * num_siblings (j) +* index = j + i * num_siblings +*/ +struct i915_engine_class_instance engines[0 ]; + +} __packed; + +#define I915_DEFINE_CONTEXT_ENGINES_PARALLEL_SUBMIT (name__,N__ ) struct { \ + struct i915_user_extension base; \ + __u16 engine_index; \ + __u16 width; \ + __u16 num_siblings; \ + __u16 mbz16; \ + __u64 flags; \ + __u64 mbz64[3]; \ + struct i915_engine_class_instance engines[N__]; \ +} __attribute__((packed)) name__ +struct i915_context_param_engines {__u64 extensions;/* linked chain of extension blocks, 0 terminates */ #define I915_CONTEXT_ENGINES_EXT_LOAD_BALANCE 0/* see i915_context_engines_load_balance */ #define I915_CONTEXT_ENGINES_EXT_BOND 1/* see i915_context_engines_bond */ +#define I915_CONTEXT_ENGINES_EXT_PARALLEL_SUBMIT 2/* see i915_context_engines_parallel_submit */ struct i915_engine_class_instance engines[0 ]; } __attribute__((packed));@@ -2726,14 +2854,20 @@ struct drm_i915_engine_info {/** @flags: Engine flags. */ __u64 flags; +#define I915_ENGINE_INFO_HAS_LOGICAL_INSTANCE (1 << 0)/** @capabilities: Capabilities of this engine. */ __u64 capabilities;#define I915_VIDEO_CLASS_CAPABILITY_HEVC (1 << 0)#define I915_VIDEO_AND_ENHANCE_CLASS_CAPABILITY_SFC (1 << 1) +/** @logical_instance: Logical instance of engine */ +__u16 logical_instance; +/** @rsvd1: Reserved fields. */ -__u64 rsvd1[4 ]; +__u16 rsvd1[3 ]; +/** @rsvd2: Reserved fields. */ +__u64 rsvd2[3 ]; };/** lib/intel_ctx.c +1
-0@@ -83,6 +83,7 @@ __context_create_cfg(int fd, const intel_ctx_cfg_t *cfg, uint32_t *ctx_id) {uint64_t ext_root =0 ;I915_DEFINE_CONTEXT_ENGINES_LOAD_BALANCE (balance,GEM_MAX_ENGINES ); +I915_DEFINE_CONTEXT_ENGINES_PARALLEL_SUBMIT (parallel,GEM_MAX_ENGINES );I915_DEFINE_CONTEXT_PARAM_ENGINES (engines,GEM_MAX_ENGINES );struct drm_i915_gem_context_create_ext_setparam engines_param, vm_param;struct drm_i915_gem_context_create_ext_setparam persist_param;
lib/intel_ctx.c __context_create_cfg(int fd, const intel_ctx_cfg_t *cfg, uint32_t *ctx_id) · line 120
Quoted 3 lines
lib/intel_ctx.c +0
-0@@ -117,7 +118,29 @@ __context_create_cfg(int fd, const intel_ctx_cfg_t *cfg, uint32_t *ctx_id)unsigned num_logical_engines;memset (&engines,0 ,sizeof (engines));
Do we need an assert to make sure cfg->load_balance and cfg->parallel are not set at the same time?
Quoted 42 lines
lib/intel_ctx.c +26
-1-if (cfg->load_balance) { +if (cfg->parallel) { +memset (¶llel,0 ,sizeof (parallel)); + + num_logical_engines =1 ; + + parallel.base.name = +I915_CONTEXT_ENGINES_EXT_PARALLEL_SUBMIT ; + + engines.engines[0 ].engine_class = +I915_ENGINE_CLASS_INVALID ; + engines.engines[0 ].engine_instance = +I915_ENGINE_CLASS_INVALID_NONE ; + + parallel.num_siblings = cfg->num_engines; + parallel.width = cfg->width; +for (i =0 ; i < cfg->num_engines * cfg->width; i++) { +igt_assert_eq (cfg->engines[0 ].engine_class, + cfg->engines[i].engine_class); + parallel.engines[i] = cfg->engines[i]; + } + + engines.extensions =to_user_pointer (¶llel); + }else if (cfg->load_balance) {memset (&balance,0 ,sizeof (balance));/* In this case, the first engine is the virtual @@ -127,6 +150,9 @@ __context_create_cfg(int fd, const intel_ctx_cfg_t *cfg, uint32_t *ctx_id)igt_assert (cfg->num_engines +1 <=GEM_MAX_ENGINES ); num_logical_engines = cfg->num_engines +1 ; + balance.base.name = +I915_CONTEXT_ENGINES_EXT_LOAD_BALANCE ; + engines.engines[0 ].engine_class =I915_ENGINE_CLASS_INVALID ; engines.engines[0].engine_instance =lib/intel_ctx.h +0
-0
lib/intel_ctx.h typedef struct intel_ctx_cfg { · line 51
Quoted 6 lines
lib/intel_ctx.h +2
-0@@ -46,7 +46,9 @@ typedef struct intel_ctx_cfg {uint32_t vm;bool nopersist;bool load_balance; +bool parallel;unsigned int num_engines; +unsigned int width;
Given that width is only set when parallel is true, we could potentially have a single var (parallel_width?) and check for it being > 0 instead of checking the bool. Just a thought, not a blocker.
Quoted 44 lines
lib/intel_ctx.h +0
-0struct i915_engine_class_instance engines[GEM_MAX_ENGINES ]; } intel_ctx_cfg_t;lib/intel_reg.h +5
-0@@ -2642,6 +2642,11 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.#define STATE3D_COLOR_FACTOR ((0x3<<29)|(0x1d<<24)|(0x01<<16)) +/* Atomics */ +#define MI_ATOMIC ((0x2f << 23) | 2) +#define MI_ATOMIC_INLINE_DATA (1 << 18) +#define MI_ATOMIC_ADD (0x7 << 8) +/* Batch */ #define MI_BATCH_BUFFER ((0x30 << 23) | 1)#define MI_BATCH_BUFFER_START (0x31 << 23)tests/i915/gem_exec_balancer.c +10
-0@@ -25,6 +25,7 @@#include <sched.h> #include <sys/ioctl.h> #include <sys/signal.h> +#include <poll.h> #include "i915/gem.h" #include "i915/gem_create.h" @@ -56,6 +57,31 @@ static size_t sizeof_load_balance(int count)#define alloca0 (sz) ({ size_t sz__ = (sz); memset(alloca(sz__), 0, sz__); }) +static int +__i915_query (int fd,struct drm_i915_query *q) +{ +if (igt_ioctl (fd,DRM_IOCTL_I915_QUERY , q)) +return -errno; + +return 0 ; +} +
tests/i915/gem_exec_balancer.c static size_t sizeof_load_balance(int count) · line 78
Quoted 10 lines
tests/i915/gem_exec_balancer.c +10
-0+static int +__i915_query_items (int fd,struct drm_i915_query_item *items,uint32_t n_items) +{ +struct drm_i915_query q = { + .num_items = n_items, + .items_ptr =to_user_pointer (items), + }; + +return __i915_query (fd, &q); +}
Identical query helpers are implemented in a couple other places (lib/i915/intel_memory_region.c, tests/i915/i915_query.c), so I believe we have critical usage mass to move them to their own lib file.
Quoted 10 lines
tests/i915/gem_exec_balancer.c +6
-0+ +#define i915_query_items (fd, items, n_items) do { \ + igt_assert_eq(__i915_query_items(fd, items, n_items), 0); \ + errno = 0; \ + } while (0) +static bool has_class_instance (int i915,uint16_t class,uint16_t instance) {int fd;
tests/i915/gem_exec_balancer.c static void nohangcheck(int i915) · line 2781
Quoted 4 lines
tests/i915/gem_exec_balancer.c +1
-0@@ -2752,6 +2778,380 @@ static void nohangcheck(int i915)close (params); } +static void check_bo (int i915,uint32_t handle,unsigned int count,bool wait)
s/count/expected? you're not using that variable as a count, just as a value to compare against
Quoted 17 lines
tests/i915/gem_exec_balancer.c +17
-0+{ +uint32_t *map; + + map =gem_mmap__cpu (i915, handle,0 ,4096 ,PROT_READ ); +if (wait) +gem_set_domain (i915, handle,I915_GEM_DOMAIN_CPU , +I915_GEM_DOMAIN_CPU ); +igt_assert_eq (map[0 ], count); +munmap (map,4096 ); +} + +static struct drm_i915_query_engine_info *query_engine_info (int i915) +{ +struct drm_i915_query_engine_info *engines; +struct drm_i915_query_item item; + +#define QUERY_SIZE 0x4000
tests/i915/gem_exec_balancer.c static void nohangcheck(int i915) · line 2808
Quoted 10 lines
tests/i915/gem_exec_balancer.c +10
-0+ engines =malloc (QUERY_SIZE ); +igt_assert (engines); + +memset (engines,0 ,QUERY_SIZE ); +memset (&item,0 ,sizeof (item)); + item.query_id =DRM_I915_QUERY_ENGINE_INFO ; + item.data_ptr =to_user_pointer (engines); + item.length =QUERY_SIZE ; + +i915_query_items (i915, &item,1 );
There is an helper you can use for this query (__gem_query_engines)
Quoted 52 lines
tests/i915/gem_exec_balancer.c +52
-0+igt_assert (item.length >=0 ); +igt_assert (item.length <=QUERY_SIZE ); +#undef QUERY_SIZE + +return engines; +} + +/* This function only works if siblings contains all instances of a class */ +static void logical_sort_siblings (int i915, +struct i915_engine_class_instance *siblings, +unsigned int count) +{ +struct i915_engine_class_instance *sorted; +struct drm_i915_query_engine_info *engines; +unsigned int i, j; + + sorted =calloc (count,sizeof (*sorted)); +igt_assert (sorted); + + engines =query_engine_info (i915); + +for (j =0 ; j < count; ++j) { +for (i =0 ; i < engines->num_engines; ++i) { +if (siblings[j].engine_class == + engines->engines[i].engine.engine_class && + siblings[j].engine_instance == + engines->engines[i].engine.engine_instance) { +uint16_t logical_instance = + engines->engines[i].logical_instance; + +igt_assert (logical_instance < count); +igt_assert (!sorted[logical_instance].engine_class); +igt_assert (!sorted[logical_instance].engine_instance); + + sorted[logical_instance] = siblings[j]; +break ; + } + } +igt_assert (i != engines->num_engines); + } + +memcpy (siblings, sorted,sizeof (*sorted) * count); +free (sorted); +free (engines); +} + +#define PARALLEL_BB_FIRST (0x1 << 0) +#define PARALLEL_OUT_FENCE (0x1 << 1) +#define PARALLEL_IN_FENCE (0x1 << 2) +#define PARALLEL_SUBMIT_FENCE (0x1 << 3) +#define PARALLEL_CONTEXTS (0x1 << 4) +#define PARALLEL_VIRTUAL (0x1 << 5)
tests/i915/gem_exec_balancer.c static void nohangcheck(int i915) · line 2870
Quoted 10 lines
tests/i915/gem_exec_balancer.c +10
-0+ +static void parallel_thread (int i915,unsigned int flags, +struct i915_engine_class_instance *siblings, +unsigned int count,unsigned int bb_per_execbuf) +{ +const intel_ctx_t *ctx =NULL ; +int n, i, j, fence =0 ; +uint32_t batch[16 ]; +struct drm_i915_gem_execbuffer2 execbuf; +struct drm_i915_gem_exec_object2 obj[32 ];
Max num of objects is 32, do we need an assert that bb_per_execbuf <=31 to leave room for the target BO? Or is that overkill since we likely won't have that many engines?
Quoted 5 lines
tests/i915/gem_exec_balancer.c +5
-0+#define PARALLEL_BB_LOOP_COUNT 512 +const intel_ctx_t *ctxs[PARALLEL_BB_LOOP_COUNT ]; +uint32_t target_bo_idx =0 ; +uint32_t first_bb_idx =1 ; +intel_ctx_cfg_t cfg;
tests/i915/gem_exec_balancer.c static void nohangcheck(int i915) · line 2885
Quoted 10 lines
tests/i915/gem_exec_balancer.c +10
-0+ +if (flags &PARALLEL_BB_FIRST ) { + target_bo_idx = bb_per_execbuf; + first_bb_idx =0 ; + } + +memset (&cfg,0 ,sizeof (cfg)); +if (flags &PARALLEL_VIRTUAL ) { + cfg.parallel = true; + cfg.num_engines = count / bb_per_execbuf;
igt_assert (count >= bb_per_execbuf && count % bb_per_execbuf == 0) to make sure the provided values are fine?
tests/i915/gem_exec_balancer.c static void nohangcheck(int i915) · line 2896
Quoted 11 lines
tests/i915/gem_exec_balancer.c +11
-0+ cfg.width = bb_per_execbuf; + +for (i =0 ; i < cfg.width; ++i) +for (j =0 ; j < cfg.num_engines; ++j) +memcpy (cfg.engines + i * cfg.num_engines + j, + siblings + j * cfg.width + i, +sizeof (*siblings)); + }else { + cfg.parallel = true; + cfg.num_engines =1 ; + cfg.width = count;
Here the usage of count vs bb_per_execbuf gets a bit counfusing. AFAICS using width = count here only works if count = bb_per_execbuf , because in the loop below we only create bb_per_execbuf batches. Why not use bb_per_execbuf directly for consistency? That would also allow you to pull the base cfg out of the if statement here and just always use:
cfg.parallel = true;
cfg.num_engines = count / bb_per_execbuf;
cfg.width = bb_per_execbuf;
Because if count = bb_per_execbuf it resolves to the same values anyway.
Quoted 27 lines
tests/i915/gem_exec_balancer.c +27
-0+memcpy (cfg.engines, siblings,sizeof (*siblings) * count); + } + ctx =intel_ctx_create (i915, &cfg); + + i =0 ; + batch[i] =MI_ATOMIC |MI_ATOMIC_INLINE_DATA | +MI_ATOMIC_ADD ; +#define TARGET_BO_OFFSET (0x1 << 16) + batch[++i] =TARGET_BO_OFFSET ; + batch[++i] =0 ; + batch[++i] =1 ; + batch[++i] =MI_BATCH_BUFFER_END ; + +memset (obj,0 ,sizeof (obj)); + obj[target_bo_idx].offset =TARGET_BO_OFFSET ; + obj[target_bo_idx].flags =EXEC_OBJECT_PINNED |EXEC_OBJECT_WRITE ; + obj[target_bo_idx].handle =gem_create (i915,4096 ); + +for (i = first_bb_idx; i < bb_per_execbuf + first_bb_idx; ++i) { + obj[i].handle =gem_create (i915,4096 ); +gem_write (i915, obj[i].handle,0 , batch, +sizeof (batch)); + } + +memset (&execbuf,0 ,sizeof (execbuf)); + execbuf.buffers_ptr =to_user_pointer (obj); + execbuf.buffer_count = bb_per_execbuf +1 ;
tests/i915/gem_exec_balancer.c static void nohangcheck(int i915) · line 2933
Quoted 10 lines
tests/i915/gem_exec_balancer.c +10
-0+ execbuf.flags |=I915_EXEC_HANDLE_LUT ; +if (flags &PARALLEL_BB_FIRST ) + execbuf.flags |=I915_EXEC_BATCH_FIRST ; +if (flags &PARALLEL_OUT_FENCE ) + execbuf.flags |=I915_EXEC_FENCE_OUT ; + execbuf.buffers_ptr =to_user_pointer (obj); + execbuf.rsvd1 = ctx->id; + +for (n =0 ; n <PARALLEL_BB_LOOP_COUNT ; ++n) { +for (i =0 ; i < count / bb_per_execbuf; ++i ) {
As discussed offline, this internal loop doesn't do anything (we only ever cycle once) and should be removed.
Quoted 32 lines
tests/i915/gem_exec_balancer.c +32
-0+ execbuf.flags &= ~0x3full ; + execbuf.flags |= i; +gem_execbuf_wr (i915, &execbuf); + +if (flags &PARALLEL_OUT_FENCE ) { +igt_assert_eq (sync_fence_wait (execbuf.rsvd2 >>32 , +1000 ),0 ); +igt_assert_eq (sync_fence_status (execbuf.rsvd2 >>32 ),1 ); + +if (fence) +close (fence); + fence = execbuf.rsvd2 >>32 ; + +if (flags &PARALLEL_SUBMIT_FENCE ) { + execbuf.flags |= +I915_EXEC_FENCE_SUBMIT ; + execbuf.rsvd2 >>=32 ; + }else if (flags &PARALLEL_IN_FENCE ) { + execbuf.flags |= +I915_EXEC_FENCE_IN ; + execbuf.rsvd2 >>=32 ; + }else { + execbuf.rsvd2 =0 ; + } + } + +if (flags &PARALLEL_VIRTUAL ) +break ; + } + +if (flags &PARALLEL_CONTEXTS ) { + ctxs[n] = ctx;
tests/i915/gem_exec_balancer.c static void nohangcheck(int i915) · line 2975
Quoted 10 lines
tests/i915/gem_exec_balancer.c +10
-0+ ctx =intel_ctx_create (i915, &cfg); + execbuf.rsvd1 = ctx->id; + } + } +if (fence) +close (fence); + +check_bo (i915, obj[target_bo_idx].handle, flags &PARALLEL_VIRTUAL ? + bb_per_execbuf *PARALLEL_BB_LOOP_COUNT : + count *PARALLEL_BB_LOOP_COUNT , true);
same as above, can just use bb_per_execbuf unconditionally here
tests/i915/gem_exec_balancer.c static void nohangcheck(int i915) · line 2988
Quoted 13 lines
tests/i915/gem_exec_balancer.c +13
-0+ +intel_ctx_destroy (i915, ctx); +for (i =0 ; flags &PARALLEL_CONTEXTS && + i <PARALLEL_BB_LOOP_COUNT ; ++i) { +intel_ctx_destroy (i915, ctxs[i]); + } +for (i =0 ; i < bb_per_execbuf +1 ; ++i) +gem_close (i915, obj[i].handle); +} + +static void parallel (int i915,unsigned int flags) +{ +for (int class =0 ; class <32 ; class++) {
I think we usually avoid declaring variables inside the for loops statements, even if the recent C standards allow it, but not sure if we have an official style in this regard. there is multiple instance of this in this file.
Quoted 36 lines
tests/i915/gem_exec_balancer.c +36
-0+struct i915_engine_class_instance *siblings; +unsigned int count, bb_per_execbuf; + + siblings =list_engines (i915,1u << class, &count); +if (!siblings) +continue ; + +if (count <2 ) { +free (siblings); +continue ; + } + +logical_sort_siblings (i915, siblings, count); + bb_per_execbuf = count; + +parallel_thread (i915, flags, siblings, + count, bb_per_execbuf); + +free (siblings); + } +} + +static void parallel_balancer (int i915,unsigned int flags) +{ +for (int class =0 ; class <32 ; class++) { +struct i915_engine_class_instance *siblings; +unsigned int count; + + siblings =list_engines (i915,1u << class, &count); +if (!siblings) +continue ; + +if (count <4 ) { +free (siblings); +continue ; + }
tests/i915/gem_exec_balancer.c static void nohangcheck(int i915) · line 3034
Quoted 10 lines
tests/i915/gem_exec_balancer.c +10
-0+ +logical_sort_siblings (i915, siblings, count); + +for (unsigned int bb_per_execbuf =2 ;;) { + igt_fork(child, count / bb_per_execbuf) +parallel_thread (i915, + flags |PARALLEL_VIRTUAL , + siblings, + count, + bb_per_execbuf);
As a possible future improvement IMO it'd be nice to check that 2 parallel VEs are deployed to the HW at the same time. The test will currently pass even if they are serialized. Not a blocker.
tests/i915/gem_exec_balancer.c static void nohangcheck(int i915) · line 3038
Quoted 4 lines
tests/i915/gem_exec_balancer.c +4
-0+igt_waitchildren (); + +if (count / ++bb_per_execbuf <=1 ) +break ;
bikeshed: why not just put this in the if statement?
for (bb = 2; count / bb > 1 ; ++bb)
not a blocker.
tests/i915/gem_exec_balancer.c static void nohangcheck(int i915) · line 3050
Quoted 12 lines
tests/i915/gem_exec_balancer.c +12
-0+ } + +free (siblings); + } +} + +static bool fence_busy (int fence) +{ +return poll (&(struct pollfd ){fence,POLLIN },1 ,0 ) ==0 ; +} + +static void parallel_ordering (int i915,unsigned int flags)
A one-line comment about the test to describe what it's doing would help IMO.
Quoted 38 lines
tests/i915/gem_exec_balancer.c +38
-0+{ +for (int class =0 ; class <32 ; class++) { +const intel_ctx_t *ctx =NULL , *spin_ctx =NULL ; +struct i915_engine_class_instance *siblings; +unsigned int count; +int i =0 , fence =0 ; +uint32_t batch[16 ]; +struct drm_i915_gem_execbuffer2 execbuf; +struct drm_i915_gem_exec_object2 obj[32 ]; +igt_spin_t *spin; +intel_ctx_cfg_t cfg; + + siblings =list_engines (i915,1u << class, &count); +if (!siblings) +continue ; + +if (count <2 ) { +free (siblings); +continue ; + } + +logical_sort_siblings (i915, siblings, count); + +memset (&cfg,0 ,sizeof (cfg)); + cfg.parallel = true; + cfg.num_engines =1 ; + cfg.width = count; +memcpy (cfg.engines, siblings,sizeof (*siblings) * count); + + ctx =intel_ctx_create (i915, &cfg); + + batch[i] =MI_ATOMIC |MI_ATOMIC_INLINE_DATA | +MI_ATOMIC_ADD ; + batch[++i] =TARGET_BO_OFFSET ; + batch[++i] =0 ; + batch[++i] =1 ; + batch[++i] =MI_BATCH_BUFFER_END ; +
tests/i915/gem_exec_balancer.c static void nohangcheck(int i915) · line 3098
Quoted 10 lines
tests/i915/gem_exec_balancer.c +10
-0+memset (obj,0 ,sizeof (obj)); + obj[0 ].offset =TARGET_BO_OFFSET ; + obj[0 ].flags =EXEC_OBJECT_PINNED |EXEC_OBJECT_WRITE ; + obj[0 ].handle =gem_create (i915,4096 ); + +for (i =1 ; i < count +1 ; ++i) { + obj[i].handle =gem_create (i915,4096 ); +gem_write (i915, obj[i].handle,0 , batch, +sizeof (batch)); + }
The object setup code here is identical to the one in parallel_thread(), maybe move it to a common function?
Quoted 22 lines
tests/i915/gem_exec_balancer.c +22
-0+ +memset (&execbuf,0 ,sizeof (execbuf)); + execbuf.buffers_ptr =to_user_pointer (obj); + execbuf.buffer_count = count +1 ; + execbuf.flags |=I915_EXEC_HANDLE_LUT ; + execbuf.flags |=I915_EXEC_NO_RELOC ; + execbuf.flags |=I915_EXEC_FENCE_OUT ; + execbuf.buffers_ptr =to_user_pointer (obj); + execbuf.rsvd1 = ctx->id; + +/* Block parallel submission */ + spin_ctx =ctx_create_engines (i915, siblings, count); + spin = __igt_spin_new(i915, + .ctx = spin_ctx, + .engine = 0, + .flags =IGT_SPIN_FENCE_OUT | +IGT_SPIN_NO_PREEMPTION ); + +/* Wait for spinners to start */ +usleep (5 *10000 ); +igt_assert (fence_busy (spin->out_fence)); +
tests/i915/gem_exec_balancer.c static void nohangcheck(int i915) · line 3130
Quoted 10 lines
tests/i915/gem_exec_balancer.c +10
-0+/* Submit parallel execbuf */ +gem_execbuf_wr (i915, &execbuf); + fence = execbuf.rsvd2 >>32 ; + +/* +* Wait long enough for timeslcing to kick in but not +* preemption. Spinner + parallel execbuf should be +* active. +*/ +usleep (25 *10000 );
This is a pretty arbitrary number, what if the system has been set up with a longer timeslicing period (or none at all) and/or a shorter preemption timeout? IMO you should read those out of sysfs and tune the waits accordingly
Daniele
Quoted 128 lines
tests/i915/gem_exec_balancer.c +111
-0+igt_assert (fence_busy (spin->out_fence)); +igt_assert (fence_busy (fence)); +check_bo (i915, obj[0 ].handle,0 , false); + +/* +* End spinner and wait for spinner + parallel execbuf +* to compelte. +*/ +igt_spin_end (spin); +igt_assert_eq (sync_fence_wait (fence,1000 ),0 ); +igt_assert_eq (sync_fence_status (fence),1 ); +check_bo (i915, obj[0 ].handle, count, true); +close (fence); + +/* Clean up */ +intel_ctx_destroy (i915, ctx); +intel_ctx_destroy (i915, spin_ctx); +for (i =0 ; i < count +1 ; ++i) +gem_close (i915, obj[i].handle); +free (siblings); +igt_spin_free (i915, spin); + } +} +static bool has_persistence (int i915) {struct drm_i915_gem_context_param p = {@@ -2786,6 +3186,61 @@ static bool has_load_balancer(int i915)return err ==0 ; } +static bool has_logical_mapping (int i915) +{ +struct drm_i915_query_engine_info *engines; +unsigned int i; + + engines =query_engine_info (i915); + +for (i =0 ; i < engines->num_engines; ++i) +if (!(engines->engines[i].flags & +I915_ENGINE_INFO_HAS_LOGICAL_INSTANCE )) { +free (engines); +return false; + } + +free (engines); +return true; +} + +static bool has_parallel_execbuf (int i915) +{ +intel_ctx_cfg_t cfg = { + .parallel = true, + .num_engines =1 , + }; +const intel_ctx_t *ctx =NULL ; +int err; + +for (int class =0 ; class <32 ; class++) { +struct i915_engine_class_instance *siblings; +unsigned int count; + + siblings =list_engines (i915,1u << class, &count); +if (!siblings) +continue ; + +if (count <2 ) { +free (siblings); +continue ; + } + +logical_sort_siblings (i915, siblings, count); + + cfg.width = count; +memcpy (cfg.engines, siblings,sizeof (*siblings) * count); +free (siblings); + + err =__intel_ctx_create (i915, &cfg, &ctx); +intel_ctx_destroy (i915, ctx); + +return err ==0 ; + } + +return false; +} + igt_main {int i915 =-1 ;@@ -2886,6 +3341,38 @@ igt_mainigt_stop_hang_detector (); } + igt_subtest_group { + igt_fixture { +igt_require (has_logical_mapping (i915)); +igt_require (has_parallel_execbuf (i915)); + } + + igt_subtest("parallel-ordering") +parallel_ordering (i915,0 ); + + igt_subtest("parallel") +parallel (i915,0 ); + + igt_subtest("parallel-bb-first") +parallel (i915,PARALLEL_BB_FIRST ); + + igt_subtest("parallel-out-fence") +parallel (i915,PARALLEL_OUT_FENCE ); + + igt_subtest("parallel-keep-in-fence") +parallel (i915,PARALLEL_OUT_FENCE |PARALLEL_IN_FENCE ); + + igt_subtest("parallel-keep-submit-fence") +parallel (i915,PARALLEL_OUT_FENCE | +PARALLEL_SUBMIT_FENCE ); + + igt_subtest("parallel-contexts") +parallel (i915,PARALLEL_CONTEXTS ); + + igt_subtest("parallel-balancer") +parallel_balancer (i915,0 ); + } + igt_subtest_group {igt_hang_t hang;
Quoted 24 lines · Daniele Ceraolo Spurio
Quoted 21 lines · Matthew Brost
Add basic parallel execbuf submission test which more or less just submits the same BB in loop a which does an atomic increment to a memory location. The memory location is checked at the end for the correct value. Different sections use various IOCTL options (e.g. fences, location of BBs, etc...).
In addition to above sections, an additional section ensure the ordering of parallel submission by submitting a spinning batch to 1 individual engine, submit a parallel execbuf to all engines instances within the class, verify none on parallel execbuf make to hardware, release spinner, and finally verify everything has completed.
Signed-off-by Matthew Brost <[email protected]>--- include/drm-uapi/i915_drm.h | 136 ++++++++- lib/intel_ctx.c | 28 +- lib/intel_ctx.h | 2 + lib/intel_reg.h | 5 + tests/i915/gem_exec_balancer.c | 487 +++++++++++++++++++++++++++++++++ 5 files changed, 656 insertions(+), 2 deletions(-)
Quoted 9 lines
include/drm-uapi/i915_drm.h
Quoted 4 lines
include/drm-uapi/i915_drm.h +0
-0The uapi file needs to be in sync with drm-next. If the changes have already reached drm-next then we should just have a separate patch doing the file sync, otherwise these defs must move to
lib/i915/i915_drm_local.h
I think the uAPI changes have landed in drm-next. I'm going to guess the DIM script can sync header files for me. Let me look into this.
Quoted 171 lines
Quoted 171 lines
include/drm-uapi/i915_drm.h +135
-1@@ -1824,6 +1824,7 @@ struct drm_i915_gem_context_param {* Extensions: * i915_context_engines_load_balance (I915_CONTEXT_ENGINES_EXT_LOAD_BALANCE) * i915_context_engines_bond (I915_CONTEXT_ENGINES_EXT_BOND) +* i915_context_engines_parallel_submit (I915_CONTEXT_ENGINES_EXT_PARALLEL_SUBMIT) */ #define I915_CONTEXT_PARAM_ENGINES 0xa@@ -2104,10 +2105,137 @@ struct i915_context_engines_bond {* gem_execbuf(drm_fd, &execbuf); */ +/** +* struct i915_context_engines_parallel_submit - Configure engine for +* parallel submission. +* +* Setup a slot in the context engine map to allow multiple BBs to be submitted +* in a single execbuf IOCTL. Those BBs will then be scheduled to run on the GPU +* in parallel. Multiple hardware contexts are created internally in the i915 +* run these BBs. Once a slot is configured for N BBs only N BBs can be +* submitted in each execbuf IOCTL and this is implicit behavior e.g. The user +* doesn't tell the execbuf IOCTL there are N BBs, the execbuf IOCTL knows how +* many BBs there are based on the slot's configuration. The N BBs are the last +* N buffer objects or first N if I915_EXEC_BATCH_FIRST is set. +* +* The default placement behavior is to create implicit bonds between each +* context if each context maps to more than 1 physical engine (e.g. context is +* a virtual engine). Also we only allow contexts of same engine class and these +* contexts must be in logically contiguous order. Examples of the placement +* behavior described below. Lastly, the default is to not allow BBs to +* preempted mid BB rather insert coordinated preemption on all hardware +* contexts between each set of BBs. Flags may be added in the future to change +* both of these default behaviors. +* +* Returns -EINVAL if hardware context placement configuration is invalid or if +* the placement configuration isn't supported on the platform / submission +* interface. +* Returns -ENODEV if extension isn't supported on the platform / submission +* interface. +* +* .. code-block:: none +* +* Example 1 pseudo code: +* CS[X] = generic engine of same class, logical instance X +* INVALID = I915_ENGINE_CLASS_INVALID, I915_ENGINE_CLASS_INVALID_NONE +* set_engines(INVALID) +* set_parallel(engine_index=0, width=2, num_siblings=1, +* engines=CS[0],CS[1]) +* +* Results in the following valid placement: +* CS[0], CS[1] +* +* Example 2 pseudo code: +* CS[X] = generic engine of same class, logical instance X +* INVALID = I915_ENGINE_CLASS_INVALID, I915_ENGINE_CLASS_INVALID_NONE +* set_engines(INVALID) +* set_parallel(engine_index=0, width=2, num_siblings=2, +* engines=CS[0],CS[2],CS[1],CS[3]) +* +* Results in the following valid placements: +* CS[0], CS[1] +* CS[2], CS[3] +* +* This can also be thought of as 2 virtual engines described by 2-D array +* in the engines the field with bonds placed between each index of the +* virtual engines. e.g. CS[0] is bonded to CS[1], CS[2] is bonded to +* CS[3]. +* VE[0] = CS[0], CS[2] +* VE[1] = CS[1], CS[3] +* +* Example 3 pseudo code: +* CS[X] = generic engine of same class, logical instance X +* INVALID = I915_ENGINE_CLASS_INVALID, I915_ENGINE_CLASS_INVALID_NONE +* set_engines(INVALID) +* set_parallel(engine_index=0, width=2, num_siblings=2, +* engines=CS[0],CS[1],CS[1],CS[3]) +* +* Results in the following valid and invalid placements: +* CS[0], CS[1] +* CS[1], CS[3] - Not logical contiguous, return -EINVAL +*/ +struct i915_context_engines_parallel_submit { +/** +* @base: base user extension. +*/ +struct i915_user_extension base; + +/** +* @engine_index: slot for parallel engine +*/ +__u16 engine_index; + +/** +* @width: number of contexts per parallel engine +*/ +__u16 width; + +/** +* @num_siblings: number of siblings per context +*/ +__u16 num_siblings; + +/** +* @mbz16: reserved for future use; must be zero +*/ +__u16 mbz16; + +/** +* @flags: all undefined flags must be zero, currently not defined flags +*/ +__u64 flags; + +/** +* @mbz64: reserved for future use; must be zero +*/ +__u64 mbz64[3 ]; + +/** +* @engines: 2-d array of engine instances to configure parallel engine +* +* length = width (i) * num_siblings (j) +* index = j + i * num_siblings +*/ +struct i915_engine_class_instance engines[0 ]; + +} __packed; + +#define I915_DEFINE_CONTEXT_ENGINES_PARALLEL_SUBMIT (name__,N__ ) struct { \ + struct i915_user_extension base; \ + __u16 engine_index; \ + __u16 width; \ + __u16 num_siblings; \ + __u16 mbz16; \ + __u64 flags; \ + __u64 mbz64[3]; \ + struct i915_engine_class_instance engines[N__]; \ +} __attribute__((packed)) name__ +struct i915_context_param_engines {__u64 extensions;/* linked chain of extension blocks, 0 terminates */ #define I915_CONTEXT_ENGINES_EXT_LOAD_BALANCE 0/* see i915_context_engines_load_balance */ #define I915_CONTEXT_ENGINES_EXT_BOND 1/* see i915_context_engines_bond */ +#define I915_CONTEXT_ENGINES_EXT_PARALLEL_SUBMIT 2/* see i915_context_engines_parallel_submit */ struct i915_engine_class_instance engines[0 ]; } __attribute__((packed));@@ -2726,14 +2854,20 @@ struct drm_i915_engine_info {/** @flags: Engine flags. */ __u64 flags; +#define I915_ENGINE_INFO_HAS_LOGICAL_INSTANCE (1 << 0)/** @capabilities: Capabilities of this engine. */ __u64 capabilities;#define I915_VIDEO_CLASS_CAPABILITY_HEVC (1 << 0)#define I915_VIDEO_AND_ENHANCE_CLASS_CAPABILITY_SFC (1 << 1) +/** @logical_instance: Logical instance of engine */ +__u16 logical_instance; +/** @rsvd1: Reserved fields. */ -__u64 rsvd1[4 ]; +__u16 rsvd1[3 ]; +/** @rsvd2: Reserved fields. */ +__u64 rsvd2[3 ]; };/** lib/intel_ctx.c +1
-0@@ -83,6 +83,7 @@ __context_create_cfg(int fd, const intel_ctx_cfg_t *cfg, uint32_t *ctx_id) {uint64_t ext_root =0 ;I915_DEFINE_CONTEXT_ENGINES_LOAD_BALANCE (balance,GEM_MAX_ENGINES ); +I915_DEFINE_CONTEXT_ENGINES_PARALLEL_SUBMIT (parallel,GEM_MAX_ENGINES );I915_DEFINE_CONTEXT_PARAM_ENGINES (engines,GEM_MAX_ENGINES );
Quoted 9 lines
lib/intel_ctx.c __context_create_cfg(int fd, const intel_ctx_cfg_t *cfg, uint32_t *ctx_id) · line 119
Quoted 5 lines
lib/intel_ctx.c +0
-0struct drm_i915_gem_context_create_ext_setparam engines_param, vm_param;struct drm_i915_gem_context_create_ext_setparam persist_param;@@ -117,7 +118,29 @@ __context_create_cfg(int fd, const intel_ctx_cfg_t *cfg, uint32_t *ctx_id)unsigned num_logical_engines;memset (&engines,0 ,sizeof (engines));Do we need an assert to make sure
cfg->load_balanceandcfg->parallelare not set at the same time?
Do we need to - no. Would it be a good idea? Yes. Will do.
Quoted 42 lines
Quoted 42 lines
lib/intel_ctx.c +26
-1-if (cfg->load_balance) { +if (cfg->parallel) { +memset (¶llel,0 ,sizeof (parallel)); + + num_logical_engines =1 ; + + parallel.base.name = +I915_CONTEXT_ENGINES_EXT_PARALLEL_SUBMIT ; + + engines.engines[0 ].engine_class = +I915_ENGINE_CLASS_INVALID ; + engines.engines[0 ].engine_instance = +I915_ENGINE_CLASS_INVALID_NONE ; + + parallel.num_siblings = cfg->num_engines; + parallel.width = cfg->width; +for (i =0 ; i < cfg->num_engines * cfg->width; i++) { +igt_assert_eq (cfg->engines[0 ].engine_class, + cfg->engines[i].engine_class); + parallel.engines[i] = cfg->engines[i]; + } + + engines.extensions =to_user_pointer (¶llel); + }else if (cfg->load_balance) {memset (&balance,0 ,sizeof (balance));/* In this case, the first engine is the virtual @@ -127,6 +150,9 @@ __context_create_cfg(int fd, const intel_ctx_cfg_t *cfg, uint32_t *ctx_id)igt_assert (cfg->num_engines +1 <=GEM_MAX_ENGINES ); num_logical_engines = cfg->num_engines +1 ; + balance.base.name = +I915_CONTEXT_ENGINES_EXT_LOAD_BALANCE ; + engines.engines[0 ].engine_class =I915_ENGINE_CLASS_INVALID ; engines.engines[0].engine_instance =lib/intel_ctx.h +0
-0@@ -46,7 +46,9 @@ typedef struct intel_ctx_cfg {uint32_t vm;bool nopersist;
Quoted 9 lines
lib/intel_ctx.h typedef struct intel_ctx_cfg { · line 51
Quoted 4 lines
lib/intel_ctx.h +2
-0bool load_balance; +bool parallel;unsigned int num_engines; +unsigned int width;Given that width is only set when parallel is true, we could potentially have a single var (
parallel_width?) and check for it being > 0 instead of checking the bool. Just a thought, not a blocker.
The precedent seems to be just use a bool (e.g. nopersist, load_balance) are already in place.
Quoted 44 lines
Quoted 44 lines
lib/intel_ctx.h +0
-0struct i915_engine_class_instance engines[GEM_MAX_ENGINES ]; } intel_ctx_cfg_t;lib/intel_reg.h +5
-0@@ -2642,6 +2642,11 @@ SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.#define STATE3D_COLOR_FACTOR ((0x3<<29)|(0x1d<<24)|(0x01<<16)) +/* Atomics */ +#define MI_ATOMIC ((0x2f << 23) | 2) +#define MI_ATOMIC_INLINE_DATA (1 << 18) +#define MI_ATOMIC_ADD (0x7 << 8) +/* Batch */ #define MI_BATCH_BUFFER ((0x30 << 23) | 1)#define MI_BATCH_BUFFER_START (0x31 << 23)tests/i915/gem_exec_balancer.c +16
-0@@ -25,6 +25,7 @@#include <sched.h> #include <sys/ioctl.h> #include <sys/signal.h> +#include <poll.h> #include "i915/gem.h" #include "i915/gem_create.h" @@ -56,6 +57,31 @@ static size_t sizeof_load_balance(int count)#define alloca0 (sz) ({ size_t sz__ = (sz); memset(alloca(sz__), 0, sz__); }) +static int +__i915_query (int fd,struct drm_i915_query *q) +{ +if (igt_ioctl (fd,DRM_IOCTL_I915_QUERY , q)) +return -errno; + +return 0 ; +} + +static int +__i915_query_items (int fd,struct drm_i915_query_item *items,uint32_t n_items) +{ +struct drm_i915_query q = { + .num_items = n_items, + .items_ptr =to_user_pointer (items),
Quoted 9 lines
tests/i915/gem_exec_balancer.c static size_t sizeof_load_balance(int count) · line 78
Quoted 4 lines
tests/i915/gem_exec_balancer.c +4
-0+ }; + +return __i915_query (fd, &q); +}Identical query helpers are implemented in a couple other places (
lib/i915/intel_memory_region.c,tests/i915/i915_query.c), so I believe we have critical usage mass to move them to their own lib file.
Sure. That might be a follow up or just another patch in the series.
Quoted 8 lines
Quoted 8 lines
tests/i915/gem_exec_balancer.c +6
-0+ +#define i915_query_items (fd, items, n_items) do { \ + igt_assert_eq(__i915_query_items(fd, items, n_items), 0); \ + errno = 0; \ + } while (0) +static bool has_class_instance (int i915,uint16_t class,uint16_t instance) {
Quoted 9 lines
tests/i915/gem_exec_balancer.c static void nohangcheck(int i915) · line 2780
Quoted 5 lines
tests/i915/gem_exec_balancer.c +1
-0int fd;@@ -2752,6 +2778,380 @@ static void nohangcheck(int i915)close (params); } +static void check_bo (int i915,uint32_t handle,unsigned int count,bool wait)s/count/expected? you're not using that variable as a count, just as a value to compare against
expected is a better name. Will change.
Quoted 21 lines
Quoted 21 lines
tests/i915/gem_exec_balancer.c +21
-0+{ +uint32_t *map; + + map =gem_mmap__cpu (i915, handle,0 ,4096 ,PROT_READ ); +if (wait) +gem_set_domain (i915, handle,I915_GEM_DOMAIN_CPU , +I915_GEM_DOMAIN_CPU ); +igt_assert_eq (map[0 ], count); +munmap (map,4096 ); +} + +static struct drm_i915_query_engine_info *query_engine_info (int i915) +{ +struct drm_i915_query_engine_info *engines; +struct drm_i915_query_item item; + +#define QUERY_SIZE 0x4000 + engines =malloc (QUERY_SIZE ); +igt_assert (engines); + +memset (engines,0 ,QUERY_SIZE );
Quoted 9 lines
tests/i915/gem_exec_balancer.c static void nohangcheck(int i915) · line 2808
Quoted 6 lines
tests/i915/gem_exec_balancer.c +6
-0+memset (&item,0 ,sizeof (item)); + item.query_id =DRM_I915_QUERY_ENGINE_INFO ; + item.data_ptr =to_user_pointer (engines); + item.length =QUERY_SIZE ; + +i915_query_items (i915, &item,1 );There is an helper you can use for this query (
__gem_query_engines)
Yes, I think that will work.
Quoted 58 lines
Quoted 58 lines
tests/i915/gem_exec_balancer.c +58
-0+igt_assert (item.length >=0 ); +igt_assert (item.length <=QUERY_SIZE ); +#undef QUERY_SIZE + +return engines; +} + +/* This function only works if siblings contains all instances of a class */ +static void logical_sort_siblings (int i915, +struct i915_engine_class_instance *siblings, +unsigned int count) +{ +struct i915_engine_class_instance *sorted; +struct drm_i915_query_engine_info *engines; +unsigned int i, j; + + sorted =calloc (count,sizeof (*sorted)); +igt_assert (sorted); + + engines =query_engine_info (i915); + +for (j =0 ; j < count; ++j) { +for (i =0 ; i < engines->num_engines; ++i) { +if (siblings[j].engine_class == + engines->engines[i].engine.engine_class && + siblings[j].engine_instance == + engines->engines[i].engine.engine_instance) { +uint16_t logical_instance = + engines->engines[i].logical_instance; + +igt_assert (logical_instance < count); +igt_assert (!sorted[logical_instance].engine_class); +igt_assert (!sorted[logical_instance].engine_instance); + + sorted[logical_instance] = siblings[j]; +break ; + } + } +igt_assert (i != engines->num_engines); + } + +memcpy (siblings, sorted,sizeof (*sorted) * count); +free (sorted); +free (engines); +} + +#define PARALLEL_BB_FIRST (0x1 << 0) +#define PARALLEL_OUT_FENCE (0x1 << 1) +#define PARALLEL_IN_FENCE (0x1 << 2) +#define PARALLEL_SUBMIT_FENCE (0x1 << 3) +#define PARALLEL_CONTEXTS (0x1 << 4) +#define PARALLEL_VIRTUAL (0x1 << 5) + +static void parallel_thread (int i915,unsigned int flags, +struct i915_engine_class_instance *siblings, +unsigned int count,unsigned int bb_per_execbuf) +{ +const intel_ctx_t *ctx =NULL ;
Quoted 9 lines
tests/i915/gem_exec_balancer.c static void nohangcheck(int i915) · line 2870
Quoted 4 lines
tests/i915/gem_exec_balancer.c +4
-0+int n, i, j, fence =0 ; +uint32_t batch[16 ]; +struct drm_i915_gem_execbuffer2 execbuf; +struct drm_i915_gem_exec_object2 obj[32 ];Max num of objects is 32, do we need an assert that
bb_per_execbuf<=31 to leave room for the target BO? Or is that overkill since we likely won't have that many engines?
It can't hurt to future proof the code.
Quoted 10 lines
Quoted 10 lines
tests/i915/gem_exec_balancer.c +10
-0+#define PARALLEL_BB_LOOP_COUNT 512 +const intel_ctx_t *ctxs[PARALLEL_BB_LOOP_COUNT ]; +uint32_t target_bo_idx =0 ; +uint32_t first_bb_idx =1 ; +intel_ctx_cfg_t cfg; + +if (flags &PARALLEL_BB_FIRST ) { + target_bo_idx = bb_per_execbuf; + first_bb_idx =0 ; + }
Quoted 9 lines
tests/i915/gem_exec_balancer.c static void nohangcheck(int i915) · line 2885
Quoted 5 lines
tests/i915/gem_exec_balancer.c +5
-0+ +memset (&cfg,0 ,sizeof (cfg)); +if (flags &PARALLEL_VIRTUAL ) { + cfg.parallel = true; + cfg.num_engines = count / bb_per_execbuf;
igt_assert(count >=bb_per_execbuf&& count %bb_per_execbuf== 0) to make sure the provided values are fine?
Sure.
Quoted 18 lines
tests/i915/gem_exec_balancer.c static void nohangcheck(int i915) · line 2896
Quoted 11 lines
tests/i915/gem_exec_balancer.c +11
-0+ cfg.width = bb_per_execbuf; + +for (i =0 ; i < cfg.width; ++i) +for (j =0 ; j < cfg.num_engines; ++j) +memcpy (cfg.engines + i * cfg.num_engines + j, + siblings + j * cfg.width + i, +sizeof (*siblings)); + }else { + cfg.parallel = true; + cfg.num_engines =1 ; + cfg.width = count;Here the usage of count vs
bb_per_execbufgets a bit counfusing. AFAICS using width = count here only works if count =bb_per_execbuf, because in the loop below we only createbb_per_execbufbatches. Why not usebb_per_execbufdirectly for consistency? That would also allow you to pull the base cfg out of the if statement here and just always use:
Quoted 6 lines
cfg.parallel = true;
cfg.num_engines = count /bb_per_execbuf;
cfg.width =bb_per_execbuf;Because if count =
bb_per_execbufit resolves to the same values anyway.
I think that works.
Quoted 32 lines
Quoted 32 lines
tests/i915/gem_exec_balancer.c +32
-0+memcpy (cfg.engines, siblings,sizeof (*siblings) * count); + } + ctx =intel_ctx_create (i915, &cfg); + + i =0 ; + batch[i] =MI_ATOMIC |MI_ATOMIC_INLINE_DATA | +MI_ATOMIC_ADD ; +#define TARGET_BO_OFFSET (0x1 << 16) + batch[++i] =TARGET_BO_OFFSET ; + batch[++i] =0 ; + batch[++i] =1 ; + batch[++i] =MI_BATCH_BUFFER_END ; + +memset (obj,0 ,sizeof (obj)); + obj[target_bo_idx].offset =TARGET_BO_OFFSET ; + obj[target_bo_idx].flags =EXEC_OBJECT_PINNED |EXEC_OBJECT_WRITE ; + obj[target_bo_idx].handle =gem_create (i915,4096 ); + +for (i = first_bb_idx; i < bb_per_execbuf + first_bb_idx; ++i) { + obj[i].handle =gem_create (i915,4096 ); +gem_write (i915, obj[i].handle,0 , batch, +sizeof (batch)); + } + +memset (&execbuf,0 ,sizeof (execbuf)); + execbuf.buffers_ptr =to_user_pointer (obj); + execbuf.buffer_count = bb_per_execbuf +1 ; + execbuf.flags |=I915_EXEC_HANDLE_LUT ; +if (flags &PARALLEL_BB_FIRST ) + execbuf.flags |=I915_EXEC_BATCH_FIRST ; +if (flags &PARALLEL_OUT_FENCE ) + execbuf.flags |=I915_EXEC_FENCE_OUT ;
Quoted 9 lines
tests/i915/gem_exec_balancer.c static void nohangcheck(int i915) · line 2933
Quoted 5 lines
tests/i915/gem_exec_balancer.c +5
-0+ execbuf.buffers_ptr =to_user_pointer (obj); + execbuf.rsvd1 = ctx->id; + +for (n =0 ; n <PARALLEL_BB_LOOP_COUNT ; ++n) { +for (i =0 ; i < count / bb_per_execbuf; ++i ) {As discussed offline, this internal loop doesn't do anything (we only ever cycle once) and should be removed.
Agree, this stale code from a previous version of the test.
Quoted 36 lines
Quoted 36 lines
tests/i915/gem_exec_balancer.c +36
-0+ execbuf.flags &= ~0x3full ; + execbuf.flags |= i; +gem_execbuf_wr (i915, &execbuf); + +if (flags &PARALLEL_OUT_FENCE ) { +igt_assert_eq (sync_fence_wait (execbuf.rsvd2 >>32 , +1000 ),0 ); +igt_assert_eq (sync_fence_status (execbuf.rsvd2 >>32 ),1 ); + +if (fence) +close (fence); + fence = execbuf.rsvd2 >>32 ; + +if (flags &PARALLEL_SUBMIT_FENCE ) { + execbuf.flags |= +I915_EXEC_FENCE_SUBMIT ; + execbuf.rsvd2 >>=32 ; + }else if (flags &PARALLEL_IN_FENCE ) { + execbuf.flags |= +I915_EXEC_FENCE_IN ; + execbuf.rsvd2 >>=32 ; + }else { + execbuf.rsvd2 =0 ; + } + } + +if (flags &PARALLEL_VIRTUAL ) +break ; + } + +if (flags &PARALLEL_CONTEXTS ) { + ctxs[n] = ctx; + ctx =intel_ctx_create (i915, &cfg); + execbuf.rsvd1 = ctx->id; + } + }
Quoted 9 lines
tests/i915/gem_exec_balancer.c static void nohangcheck(int i915) · line 2975
Quoted 6 lines
tests/i915/gem_exec_balancer.c +6
-0+if (fence) +close (fence); + +check_bo (i915, obj[target_bo_idx].handle, flags &PARALLEL_VIRTUAL ? + bb_per_execbuf *PARALLEL_BB_LOOP_COUNT : + count *PARALLEL_BB_LOOP_COUNT , true);same as above, can just use
bb_per_execbufunconditionally here
Yep.
Quoted 10 lines
Quoted 10 lines
tests/i915/gem_exec_balancer.c +10
-0+ +intel_ctx_destroy (i915, ctx); +for (i =0 ; flags &PARALLEL_CONTEXTS && + i <PARALLEL_BB_LOOP_COUNT ; ++i) { +intel_ctx_destroy (i915, ctxs[i]); + } +for (i =0 ; i < bb_per_execbuf +1 ; ++i) +gem_close (i915, obj[i].handle); +} +
Quoted 9 lines
tests/i915/gem_exec_balancer.c static void nohangcheck(int i915) · line 2988
Quoted 3 lines
tests/i915/gem_exec_balancer.c +3
-0+static void parallel (int i915,unsigned int flags) +{ +for (int class =0 ; class <32 ; class++) {I think we usually avoid declaring variables inside the for loops statements, even if the recent C standards allow it, but not sure if we have an official style in this regard. there is multiple instance of this in this file.
Sure.
Quoted 42 lines
Quoted 42 lines
tests/i915/gem_exec_balancer.c +42
-0+struct i915_engine_class_instance *siblings; +unsigned int count, bb_per_execbuf; + + siblings =list_engines (i915,1u << class, &count); +if (!siblings) +continue ; + +if (count <2 ) { +free (siblings); +continue ; + } + +logical_sort_siblings (i915, siblings, count); + bb_per_execbuf = count; + +parallel_thread (i915, flags, siblings, + count, bb_per_execbuf); + +free (siblings); + } +} + +static void parallel_balancer (int i915,unsigned int flags) +{ +for (int class =0 ; class <32 ; class++) { +struct i915_engine_class_instance *siblings; +unsigned int count; + + siblings =list_engines (i915,1u << class, &count); +if (!siblings) +continue ; + +if (count <4 ) { +free (siblings); +continue ; + } + +logical_sort_siblings (i915, siblings, count); + +for (unsigned int bb_per_execbuf =2 ;;) { + igt_fork(child, count / bb_per_execbuf) + parallel_thread(i915,
Quoted 9 lines
tests/i915/gem_exec_balancer.c static void nohangcheck(int i915) · line 3034
Quoted 4 lines
tests/i915/gem_exec_balancer.c +4
-0+ flags |PARALLEL_VIRTUAL , + siblings, + count, + bb_per_execbuf);As a possible future improvement IMO it'd be nice to check that 2 parallel VEs are deployed to the HW at the same time. The test will currently pass even if they are serialized. Not a blocker.
We probably will need spinners to do this as with short running batches the 2 parallel VEs probably will never actually be on the hardware at the same time because they more less switch in and switch out at such a fast rate only 1 will actually be scheduled at a time. Agree this would be a good test to have though. Likely need a whole new functions / test sections to do this. Again this might be in a follow up or another patch later in this series.
Quoted 11 lines
tests/i915/gem_exec_balancer.c static void nohangcheck(int i915) · line 3038
Quoted 4 lines
tests/i915/gem_exec_balancer.c +4
-0+igt_waitchildren (); + +if (count / ++bb_per_execbuf <=1 ) +break ;bikeshed: why not just put this in the if statement?
for (bb = 2; count / bb > 1 ; ++bb)
not a blocker.
That works.
Quoted 7 lines
Quoted 7 lines
tests/i915/gem_exec_balancer.c +7
-0+ } + +free (siblings); + } +} + +static bool fence_busy (int fence)
Quoted 9 lines
tests/i915/gem_exec_balancer.c static void nohangcheck(int i915) · line 3050
Quoted 5 lines
tests/i915/gem_exec_balancer.c +5
-0+{ +return poll (&(struct pollfd ){fence,POLLIN },1 ,0 ) ==0 ; +} + +static void parallel_ordering (int i915,unsigned int flags)A one-line comment about the test to describe what it's doing would help IMO.
Sure. BTW, this section is fail with my current (on list) implementation (rather weak) for parallel submission with execlists.
Quoted 43 lines
Quoted 43 lines
tests/i915/gem_exec_balancer.c +43
-0+{ +for (int class =0 ; class <32 ; class++) { +const intel_ctx_t *ctx =NULL , *spin_ctx =NULL ; +struct i915_engine_class_instance *siblings; +unsigned int count; +int i =0 , fence =0 ; +uint32_t batch[16 ]; +struct drm_i915_gem_execbuffer2 execbuf; +struct drm_i915_gem_exec_object2 obj[32 ]; +igt_spin_t *spin; +intel_ctx_cfg_t cfg; + + siblings =list_engines (i915,1u << class, &count); +if (!siblings) +continue ; + +if (count <2 ) { +free (siblings); +continue ; + } + +logical_sort_siblings (i915, siblings, count); + +memset (&cfg,0 ,sizeof (cfg)); + cfg.parallel = true; + cfg.num_engines =1 ; + cfg.width = count; +memcpy (cfg.engines, siblings,sizeof (*siblings) * count); + + ctx =intel_ctx_create (i915, &cfg); + + batch[i] =MI_ATOMIC |MI_ATOMIC_INLINE_DATA | +MI_ATOMIC_ADD ; + batch[++i] =TARGET_BO_OFFSET ; + batch[++i] =0 ; + batch[++i] =1 ; + batch[++i] =MI_BATCH_BUFFER_END ; + +memset (obj,0 ,sizeof (obj)); + obj[0 ].offset =TARGET_BO_OFFSET ; + obj[0 ].flags =EXEC_OBJECT_PINNED |EXEC_OBJECT_WRITE ; + obj[0 ].handle =gem_create (i915,4096 ); +
Quoted 9 lines
tests/i915/gem_exec_balancer.c static void nohangcheck(int i915) · line 3098
Quoted 5 lines
tests/i915/gem_exec_balancer.c +5
-0+for (i =1 ; i < count +1 ; ++i) { + obj[i].handle =gem_create (i915,4096 ); +gem_write (i915, obj[i].handle,0 , batch, +sizeof (batch)); + }The object setup code here is identical to the one in
parallel_thread(), maybe move it to a common function?
Sure.
Quoted 29 lines
Quoted 29 lines
tests/i915/gem_exec_balancer.c +29
-0+ +memset (&execbuf,0 ,sizeof (execbuf)); + execbuf.buffers_ptr =to_user_pointer (obj); + execbuf.buffer_count = count +1 ; + execbuf.flags |=I915_EXEC_HANDLE_LUT ; + execbuf.flags |=I915_EXEC_NO_RELOC ; + execbuf.flags |=I915_EXEC_FENCE_OUT ; + execbuf.buffers_ptr =to_user_pointer (obj); + execbuf.rsvd1 = ctx->id; + +/* Block parallel submission */ + spin_ctx =ctx_create_engines (i915, siblings, count); + spin = __igt_spin_new(i915, + .ctx = spin_ctx, + .engine = 0, + .flags =IGT_SPIN_FENCE_OUT | +IGT_SPIN_NO_PREEMPTION ); + +/* Wait for spinners to start */ +usleep (5 *10000 ); +igt_assert (fence_busy (spin->out_fence)); + +/* Submit parallel execbuf */ +gem_execbuf_wr (i915, &execbuf); + fence = execbuf.rsvd2 >>32 ; + +/* +* Wait long enough for timeslcing to kick in but not +* preemption. Spinner + parallel execbuf should be
Quoted 9 lines
tests/i915/gem_exec_balancer.c static void nohangcheck(int i915) · line 3130
Quoted 3 lines
tests/i915/gem_exec_balancer.c +3
-0+* active. +*/ +usleep (25 *10000 );This is a pretty arbitrary number, what if the system has been set up with a longer timeslicing period (or none at all) and/or a shorter preemption timeout? IMO you should read those out of sysfs and tune the waits accordingly
I think we helpers for reading sysfs values, this should be easy enough to add. How about we wait for 5x the timeslice or something? If values are configured in way where that math doesn't work, this we just skip this test.
Matt
Quoted 131 lines
Daniele
Quoted 125 lines
tests/i915/gem_exec_balancer.c +111
-0+igt_assert (fence_busy (spin->out_fence)); +igt_assert (fence_busy (fence)); +check_bo (i915, obj[0 ].handle,0 , false); + +/* +* End spinner and wait for spinner + parallel execbuf +* to compelte. +*/ +igt_spin_end (spin); +igt_assert_eq (sync_fence_wait (fence,1000 ),0 ); +igt_assert_eq (sync_fence_status (fence),1 ); +check_bo (i915, obj[0 ].handle, count, true); +close (fence); + +/* Clean up */ +intel_ctx_destroy (i915, ctx); +intel_ctx_destroy (i915, spin_ctx); +for (i =0 ; i < count +1 ; ++i) +gem_close (i915, obj[i].handle); +free (siblings); +igt_spin_free (i915, spin); + } +} +static bool has_persistence (int i915) {struct drm_i915_gem_context_param p = {@@ -2786,6 +3186,61 @@ static bool has_load_balancer(int i915)return err ==0 ; } +static bool has_logical_mapping (int i915) +{ +struct drm_i915_query_engine_info *engines; +unsigned int i; + + engines =query_engine_info (i915); + +for (i =0 ; i < engines->num_engines; ++i) +if (!(engines->engines[i].flags & +I915_ENGINE_INFO_HAS_LOGICAL_INSTANCE )) { +free (engines); +return false; + } + +free (engines); +return true; +} + +static bool has_parallel_execbuf (int i915) +{ +intel_ctx_cfg_t cfg = { + .parallel = true, + .num_engines =1 , + }; +const intel_ctx_t *ctx =NULL ; +int err; + +for (int class =0 ; class <32 ; class++) { +struct i915_engine_class_instance *siblings; +unsigned int count; + + siblings =list_engines (i915,1u << class, &count); +if (!siblings) +continue ; + +if (count <2 ) { +free (siblings); +continue ; + } + +logical_sort_siblings (i915, siblings, count); + + cfg.width = count; +memcpy (cfg.engines, siblings,sizeof (*siblings) * count); +free (siblings); + + err =__intel_ctx_create (i915, &cfg, &ctx); +intel_ctx_destroy (i915, ctx); + +return err ==0 ; + } + +return false; +} + igt_main {int i915 =-1 ;@@ -2886,6 +3341,38 @@ igt_mainigt_stop_hang_detector (); } + igt_subtest_group { + igt_fixture { +igt_require (has_logical_mapping (i915)); +igt_require (has_parallel_execbuf (i915)); + } + + igt_subtest("parallel-ordering") +parallel_ordering (i915,0 ); + + igt_subtest("parallel") +parallel (i915,0 ); + + igt_subtest("parallel-bb-first") +parallel (i915,PARALLEL_BB_FIRST ); + + igt_subtest("parallel-out-fence") +parallel (i915,PARALLEL_OUT_FENCE ); + + igt_subtest("parallel-keep-in-fence") +parallel (i915,PARALLEL_OUT_FENCE |PARALLEL_IN_FENCE ); + + igt_subtest("parallel-keep-submit-fence") +parallel (i915,PARALLEL_OUT_FENCE | +PARALLEL_SUBMIT_FENCE ); + + igt_subtest("parallel-contexts") +parallel (i915,PARALLEL_CONTEXTS ); + + igt_subtest("parallel-balancer") +parallel_balancer (i915,0 ); + } + igt_subtest_group {igt_hang_t hang;