The review of v2 changes touched issues which were addressed in a different way than planned in that review: 1. Context status processing While the review went towards finding common path to new preemption flag combinations and existing cases, I decided to split the two ways, because the !(status & GEN8_CTX_STATUS_COMPLETED_MASK) condition had to be skipped from the preemption flow. While pre-gen11 preemption always executed the preemptive context which guaranteed that flag set, nothing finishes execution in the new to-real-idle preemption. 2. Testing of the idle-to-idle preemption case It turns out the IGT test 'smoketest-render' from 'gem_exec_schedule' group triggers the idle-to-idle case. The test submits a lot of very short execs with changing priorities. Its run triggers preemption many times. and due to the short exec buffer, triggers idle-to-idle case as well, sooner or later. Tomasz Lis (1): drm/i915/gen11: Preempt-to-idle support in execlists. drivers/gpu/drm/i915/i915_drv.h | 2 + drivers/gpu/drm/i915/i915_gem_context.c | 5 +- drivers/gpu/drm/i915/i915_pci.c | 3 +- drivers/gpu/drm/i915/intel_device_info.h | 1 + drivers/gpu/drm/i915/intel_lrc.c | 115 ++++++++++++++++++++++--------- drivers/gpu/drm/i915/intel_lrc.h | 1 + 6 files changed, 92 insertions(+), 35 deletions(-) -- 2.7.4 _______________________________________________ Intel-gfx mailing list Intel-gfx@xxxxxxxxxxxxxxxxxxxxx https://lists.freedesktop.org/mailman/listinfo/intel-gfx
