From: Lionel Landwerlin <lionel.g.landwerlin@xxxxxxxxx>
We want this so we can preempt performance queries and keep the system
responsive even when long running queries are ongoing. We avoid doing
it for all contexts.
Signed-off-by: Lionel Landwerlin <lionel.g.landwerlin@xxxxxxxxx>
---
drivers/gpu/drm/i915/gt/intel_lrc.c | 23 +++++---
drivers/gpu/drm/i915/gt/intel_lrc.h | 3 +
drivers/gpu/drm/i915/i915_perf.c | 89 +++++++++++++++++++++++++++++
3 files changed, 107 insertions(+), 8 deletions(-)
diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
index 9d8eaf8edaab..1bab671dc7a0 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.c
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
@@ -1218,6 +1218,19 @@ static bool can_merge_rq(const struct i915_request *prev,
return true;
}
+u32 intel_lrc_make_ctx_control(const struct intel_engine_cs *engine)
+{
+ u32 value =
+ _MASKED_BIT_DISABLE(CTX_CTRL_ENGINE_CTX_RESTORE_INHIBIT) |
+ _MASKED_BIT_ENABLE(CTX_CTRL_INHIBIT_SYN_CTX_SWITCH);
+
+ if (INTEL_GEN(engine->i915) < 11)
+ value |= _MASKED_BIT_DISABLE(CTX_CTRL_ENGINE_CTX_SAVE_INHIBIT |
+ CTX_CTRL_RS_CTX_ENABLE);
+
+ return value;
+}
+
static void virtual_update_register_offsets(u32 *regs,
struct intel_engine_cs *engine)
{
@@ -2124,6 +2137,7 @@ __execlists_update_reg_state(const struct intel_context *ce,
i915_oa_init_reg_state(ce, engine);
}
+ regs[CTX_CONTEXT_CONTROL] |= intel_lrc_make_ctx_control(engine);
}
static int
@@ -3628,14 +3642,7 @@ static void init_common_reg_state(u32 * const regs,
const struct intel_engine_cs *engine,
const struct intel_ring *ring)
{
- regs[CTX_CONTEXT_CONTROL] =
- _MASKED_BIT_DISABLE(CTX_CTRL_ENGINE_CTX_RESTORE_INHIBIT) |
- _MASKED_BIT_ENABLE(CTX_CTRL_INHIBIT_SYN_CTX_SWITCH);
- if (INTEL_GEN(engine->i915) < 11)
- regs[CTX_CONTEXT_CONTROL] |=
- _MASKED_BIT_DISABLE(CTX_CTRL_ENGINE_CTX_SAVE_INHIBIT |
- CTX_CTRL_RS_CTX_ENABLE);
-
+ regs[CTX_CONTEXT_CONTROL] = intel_lrc_make_ctx_control(engine);
regs[CTX_RING_BUFFER_CONTROL] = RING_CTL_SIZE(ring->size) | RING_VALID;
regs[CTX_BB_STATE] = RING_BB_PPGTT;
}
diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.h b/drivers/gpu/drm/i915/gt/intel_lrc.h
index 99dc576a4e25..6b2b196f09e7 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.h
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.h
@@ -43,6 +43,7 @@ struct intel_engine_cs;
#define CTX_CTRL_ENGINE_CTX_RESTORE_INHIBIT (1 << 0)
#define CTX_CTRL_RS_CTX_ENABLE (1 << 1)
#define CTX_CTRL_ENGINE_CTX_SAVE_INHIBIT (1 << 2)
+#define GEN12_CTX_CTRL_OAR_CONTEXT_ENABLE (1 << 8)
#define RING_CONTEXT_STATUS_PTR(base) _MMIO((base) + 0x3a0)
#define RING_EXECLIST_SQ_CONTENTS(base) _MMIO((base) + 0x510)
#define RING_EXECLIST_CONTROL(base) _MMIO((base) + 0x550)
@@ -145,4 +146,6 @@ struct intel_engine_cs *
intel_virtual_engine_get_sibling(struct intel_engine_cs *engine,
unsigned int sibling);
+u32 intel_lrc_make_ctx_control(const struct intel_engine_cs *engine);
+
#endif /* _INTEL_LRC_H_ */
diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c
index df3b6976ba5b..d0ddd37dc078 100644
--- a/drivers/gpu/drm/i915/i915_perf.c
+++ b/drivers/gpu/drm/i915/i915_perf.c
@@ -2208,6 +2208,70 @@ static int gen8_configure_context(struct i915_gem_context *ctx,
return err;
}
+static int gen12_emit_oar_config(struct intel_context *ce, bool enable)
+{
+ struct i915_request *rq;
+ u32 *cs, offset;
+ int err = 0;
+
+ rq = i915_request_create(ce);
+ if (IS_ERR(rq))
+ return PTR_ERR(rq);
+
+ cs = intel_ring_begin(rq, 4);
+ if (IS_ERR(cs)) {
+ err = PTR_ERR(cs);
+ goto out;
+ }
+
+ offset = i915_ggtt_offset(ce->state) + LRC_STATE_PN * PAGE_SIZE;
+
+ *cs++ = MI_STORE_DWORD_IMM_GEN4 | MI_USE_GGTT;
+ *cs++ = offset + CTX_CONTEXT_CONTROL * sizeof(u32);
+ *cs++ = 0;
+ *cs++ = intel_lrc_make_ctx_control(ce->engine) |
+ (enable ?
+ _MASKED_BIT_ENABLE(GEN12_CTX_CTRL_OAR_CONTEXT_ENABLE) :
+ _MASKED_BIT_DISABLE(GEN12_CTX_CTRL_OAR_CONTEXT_ENABLE));
+
+ intel_ring_advance(rq, cs);
+
+out:
+ i915_request_add(rq);
+
+ return err;
+}
+
+static int gen12_configure_context_oar(struct i915_gem_context *ctx,
+ bool enable)
+{
+ struct i915_gem_engines_iter it;
+ struct intel_context *ce;
+ int err = 0;
+
+ for_each_gem_engine(ce, i915_gem_context_lock_engines(ctx), it) {
+ GEM_BUG_ON(ce == ce->engine->kernel_context);
+
+ if (ce->engine->class != RENDER_CLASS)
+ continue;
+
+ err = intel_context_lock_pinned(ce);
+ if (err)
+ break;
+
+ /* Otherwise OA settings will be set upon first use */
+ if (intel_context_is_pinned(ce))
+ err = gen12_emit_oar_config(ce, enable);
+
+ intel_context_unlock_pinned(ce);
+ if (err)
+ break;
+ }
+ i915_gem_context_unlock_engines(ctx);
+
+ return err;
+}
+
/*
* Manages updating the per-context aspects of the OA stream
* configuration across all contexts.
@@ -2313,6 +2377,17 @@ static int lrc_configure_all_contexts(struct i915_perf_stream *stream,
return err;
}
+ /*
+ * For Gen12, performance counters are context
+ * saved/restored. Only enable it for the context that
+ * requested this.
+ */
+ if (ctx == stream->ctx && IS_GEN(i915, 12)) {
+ err = gen12_configure_context_oar(ctx, oa_config != NULL);
+ if (err)
+ return err;
+ }
+
spin_lock(&i915->gem.contexts.lock);
list_safe_reset_next(ctx, cn, link);
i915_gem_context_put(ctx);
@@ -2782,6 +2857,20 @@ void i915_oa_init_reg_state(const struct intel_context *ce,
stream = engine->i915->perf.exclusive_stream;
if (stream)
gen8_update_reg_state_unlocked(ce, stream);
+
+ /*
+ * Enable context save & restore of performance counters for
+ * the OAR unit only on the context selected for performance
+ * queries.
+ */
+ if (IS_GEN(engine->i915, 12)) {
+ u32 *regs = ce->lrc_reg_state;
+
+ regs[CTX_CONTEXT_CONTROL] |=
+ (stream && ce->gem_context == stream->ctx) ?
+ _MASKED_BIT_ENABLE(GEN12_CTX_CTRL_OAR_CONTEXT_ENABLE) :
+ _MASKED_BIT_DISABLE(GEN12_CTX_CTRL_OAR_CONTEXT_ENABLE);
+ }
}
/**
--
2.20.1
_______________________________________________
Intel-gfx mailing list
Intel-gfx@xxxxxxxxxxxxxxxxxxxxx
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
