On 18/10/2019 03:50, Umesh Nerlige Ramappa wrote:
From: Lionel Landwerlin <[email protected]>
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.
v2: use LRI to modify context control (Chris)
v3: use MASKED_FIELD to program just the masked bits (Chris)
v4: reuse request created during emit_oa_config (Lionel)
Signed-off-by: Lionel Landwerlin <[email protected]>
Signed-off-by: Umesh Nerlige Ramappa <[email protected]>
Looks good :
Reviewed-by: Lionel Landwerlin <[email protected]>
---
drivers/gpu/drm/i915/gt/intel_lrc.h | 1 +
drivers/gpu/drm/i915/i915_perf.c | 35 ++++++++++++++++++++++++++++-
2 files changed, 35 insertions(+), 1 deletion(-)
diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.h
b/drivers/gpu/drm/i915/gt/intel_lrc.h
index 99dc576a4e25..b6daac712c9e 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)
diff --git a/drivers/gpu/drm/i915/i915_perf.c b/drivers/gpu/drm/i915/i915_perf.c
index 91707558a0f5..ce97af484a32 100644
--- a/drivers/gpu/drm/i915/i915_perf.c
+++ b/drivers/gpu/drm/i915/i915_perf.c
@@ -1862,14 +1862,36 @@ get_oa_vma(struct i915_perf_stream *stream, struct
i915_oa_config *oa_config)
return i915_vma_get(oa_bo->vma);
}
+static int gen12_emit_oar_config(struct i915_request *rq,
+ struct intel_context *ce,
+ bool enable)
+{
+ u32 *cs;
+
+ cs = intel_ring_begin(rq, 4);
+ if (IS_ERR(cs))
+ return PTR_ERR(cs);
+
+ *cs++ = MI_LOAD_REGISTER_IMM(1);
+ *cs++ =
i915_mmio_reg_offset(RING_CONTEXT_CONTROL(ce->engine->mmio_base));
+ *cs++ = _MASKED_FIELD(GEN12_CTX_CTRL_OAR_CONTEXT_ENABLE,
+ enable ? GEN12_CTX_CTRL_OAR_CONTEXT_ENABLE : 0);
+ *cs++ = MI_NOOP;
+
+ intel_ring_advance(rq, cs);
+
+ return 0;
+}
+
static int emit_oa_config(struct i915_perf_stream *stream,
struct intel_context *ce)
{
+ struct i915_oa_config *oa_config = stream->oa_config;
struct i915_request *rq;
struct i915_vma *vma;
int err;
- vma = get_oa_vma(stream, stream->oa_config);
+ vma = get_oa_vma(stream, oa_config);
if (IS_ERR(vma))
return PTR_ERR(vma);
@@ -1891,6 +1913,17 @@ static int emit_oa_config(struct i915_perf_stream *stream,
if (err)
goto err_add_request;
+ /*
+ * For Gen12, performance counters are context saved/restored.
+ * Only enable it for the context that requested this.
+ */
+ if (ce == stream->pinned_ctx && IS_GEN(stream->perf->i915, 12)) {
+ err = gen12_emit_oar_config(rq, ce, oa_config != NULL);
+ if (err)
+ goto err_add_request;
+
+ }
+
err = rq->engine->emit_bb_start(rq,
vma->node.start, 0,
I915_DISPATCH_SECURE);
_______________________________________________
Intel-gfx mailing list
[email protected]
https://lists.freedesktop.org/mailman/listinfo/intel-gfx
