All of lore.kernel.org
 help / color / mirror / Atom feed
From: Matt Roper <matthew.d.roper@intel.com>
To: intel-gfx@lists.freedesktop.org
Cc: dri-devel@lists.freedesktop.org
Subject: [PATCH v2 12/13] drm/i915/xehp: Add compute workarounds
Date: Mon, 28 Feb 2022 09:42:44 -0800	[thread overview]
Message-ID: <20220228174245.1569581-13-matthew.d.roper@intel.com> (raw)
In-Reply-To: <20220228174245.1569581-1-matthew.d.roper@intel.com>

Additional workarounds are required once we start exposing CCS engines.

Note that we have a number of workarounds that update registers in the
shared render/compute reset domain.  Historically we've just added such
registers to the RCS engine's workaround list.  But going forward we
should be more careful to place such workarounds on a wa_list for an
engine that definitely exists and is not fused off (e.g., a platform
with no RCS would never apply the RCS wa_list).  We'll keep
rcs_engine_wa_init() focused on RCS-specific workarounds that only need
to be applied if the RCS engine is present.  A separate
general_render_compute_wa_init() function will be used to define
workarounds that touch registers in the shared render/compute reset
domain and that we need to apply regardless of what render and/or
compute engines actually exist.  Any workarounds defined in this new
function will internally be added to the first present RCS or CCS
engine's workaround list to ensure they get applied (and only get
applied once rather than being needlessly re-applied several times).

Co-author: Srinivasan Shanmugam
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
---
 drivers/gpu/drm/i915/gt/intel_gt_regs.h     |  1 +
 drivers/gpu/drm/i915/gt/intel_lrc.c         |  6 +++
 drivers/gpu/drm/i915/gt/intel_workarounds.c | 47 +++++++++++++++++++++
 3 files changed, 54 insertions(+)

diff --git a/drivers/gpu/drm/i915/gt/intel_gt_regs.h b/drivers/gpu/drm/i915/gt/intel_gt_regs.h
index e629443e07ae..19cd34f24263 100644
--- a/drivers/gpu/drm/i915/gt/intel_gt_regs.h
+++ b/drivers/gpu/drm/i915/gt/intel_gt_regs.h
@@ -1060,6 +1060,7 @@
 #define   FLOW_CONTROL_ENABLE			REG_BIT(15)
 #define   UGM_BACKUP_MODE			REG_BIT(13)
 #define   MDQ_ARBITRATION_MODE			REG_BIT(12)
+#define   SYSTOLIC_DOP_CLOCK_GATING_DIS		REG_BIT(10)
 #define   PARTIAL_INSTRUCTION_SHOOTDOWN_DISABLE	REG_BIT(8)
 #define   STALL_DOP_GATING_DISABLE		REG_BIT(5)
 #define   THROTTLE_12_5				REG_GENMASK(4, 2)
diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
index d333400d29fe..e9ea576b96a4 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.c
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
@@ -1217,6 +1217,12 @@ gen12_emit_indirect_ctx_xcs(const struct intel_context *ce, u32 *cs)
 	cs = gen12_emit_timestamp_wa(ce, cs);
 	cs = gen12_emit_restore_scratch(ce, cs);
 
+	/* Wa_16013000631:dg2 */
+	if (IS_DG2_GRAPHICS_STEP(ce->engine->i915, G10, STEP_B0, STEP_C0) ||
+	    IS_DG2_G11(ce->engine->i915))
+		if (ce->engine->class == COMPUTE_CLASS)
+			cs = gen8_emit_pipe_control(cs, PIPE_CONTROL_INSTRUCTION_CACHE_INVALIDATE, 0);
+
 	return cs;
 }
 
diff --git a/drivers/gpu/drm/i915/gt/intel_workarounds.c b/drivers/gpu/drm/i915/gt/intel_workarounds.c
index 0471d475e680..0b9435d62808 100644
--- a/drivers/gpu/drm/i915/gt/intel_workarounds.c
+++ b/drivers/gpu/drm/i915/gt/intel_workarounds.c
@@ -1921,6 +1921,11 @@ static void dg2_whitelist_build(struct intel_engine_cs *engine)
 					  RING_FORCE_TO_NONPRIV_RANGE_4);
 
 		break;
+	case COMPUTE_CLASS:
+		/* Wa_16011157294:dg2_g10 */
+		if (IS_DG2_GRAPHICS_STEP(engine->i915, G10, STEP_A0, STEP_B0))
+			whitelist_reg(w, GEN9_CTX_PREEMPT_REG);
+		break;
 	default:
 		break;
 	}
@@ -2581,6 +2586,40 @@ xcs_engine_wa_init(struct intel_engine_cs *engine, struct i915_wa_list *wal)
 	}
 }
 
+/*
+ * The workarounds in this function apply to shared registers in
+ * the general render reset domain that aren't tied to a
+ * specific engine.  Since all render+compute engines get reset
+ * together, and the contents of these registers are lost during
+ * the shared render domain reset, we'll define such workarounds
+ * here and then add them to just a single RCS or CCS engine's
+ * workaround list (whichever engine has the XXXX flag).
+ */
+static void
+general_render_compute_wa_init(struct intel_engine_cs *engine, struct i915_wa_list *wal)
+{
+	struct drm_i915_private *i915 = engine->i915;
+
+	if (IS_XEHPSDV(i915)) {
+		/* Wa_1409954639 */
+		wa_masked_en(wal,
+			     GEN8_ROW_CHICKEN,
+			     SYSTOLIC_DOP_CLOCK_GATING_DIS);
+
+		/* Wa_1607196519 */
+		wa_masked_en(wal,
+			     GEN9_ROW_CHICKEN4,
+			     GEN12_DISABLE_GRF_CLEAR);
+
+		/* Wa_14010670810:xehpsdv */
+		wa_write_or(wal, XEHP_L3NODEARBCFG, XEHP_LNESPARE);
+
+		/* Wa_14010449647:xehpsdv */
+		wa_masked_en(wal, GEN7_HALF_SLICE_CHICKEN1,
+			     GEN7_PSD_SINGLE_PORT_DISPATCH_ENABLE);
+	}
+}
+
 static void
 engine_init_workarounds(struct intel_engine_cs *engine, struct i915_wa_list *wal)
 {
@@ -2589,6 +2628,14 @@ engine_init_workarounds(struct intel_engine_cs *engine, struct i915_wa_list *wal
 
 	engine_fake_wa_init(engine, wal);
 
+	/*
+	 * These are common workarounds that just need to applied
+	 * to a single RCS/CCS engine's workaround list since
+	 * they're reset as part of the general render domain reset.
+	 */
+	if (engine->class == RENDER_CLASS)
+		general_render_compute_wa_init(engine, wal);
+
 	if (engine->class == RENDER_CLASS)
 		rcs_engine_wa_init(engine, wal);
 	else
-- 
2.34.1


WARNING: multiple messages have this Message-ID (diff)
From: Matt Roper <matthew.d.roper@intel.com>
To: intel-gfx@lists.freedesktop.org
Cc: dri-devel@lists.freedesktop.org
Subject: [Intel-gfx] [PATCH v2 12/13] drm/i915/xehp: Add compute workarounds
Date: Mon, 28 Feb 2022 09:42:44 -0800	[thread overview]
Message-ID: <20220228174245.1569581-13-matthew.d.roper@intel.com> (raw)
In-Reply-To: <20220228174245.1569581-1-matthew.d.roper@intel.com>

Additional workarounds are required once we start exposing CCS engines.

Note that we have a number of workarounds that update registers in the
shared render/compute reset domain.  Historically we've just added such
registers to the RCS engine's workaround list.  But going forward we
should be more careful to place such workarounds on a wa_list for an
engine that definitely exists and is not fused off (e.g., a platform
with no RCS would never apply the RCS wa_list).  We'll keep
rcs_engine_wa_init() focused on RCS-specific workarounds that only need
to be applied if the RCS engine is present.  A separate
general_render_compute_wa_init() function will be used to define
workarounds that touch registers in the shared render/compute reset
domain and that we need to apply regardless of what render and/or
compute engines actually exist.  Any workarounds defined in this new
function will internally be added to the first present RCS or CCS
engine's workaround list to ensure they get applied (and only get
applied once rather than being needlessly re-applied several times).

Co-author: Srinivasan Shanmugam
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
---
 drivers/gpu/drm/i915/gt/intel_gt_regs.h     |  1 +
 drivers/gpu/drm/i915/gt/intel_lrc.c         |  6 +++
 drivers/gpu/drm/i915/gt/intel_workarounds.c | 47 +++++++++++++++++++++
 3 files changed, 54 insertions(+)

diff --git a/drivers/gpu/drm/i915/gt/intel_gt_regs.h b/drivers/gpu/drm/i915/gt/intel_gt_regs.h
index e629443e07ae..19cd34f24263 100644
--- a/drivers/gpu/drm/i915/gt/intel_gt_regs.h
+++ b/drivers/gpu/drm/i915/gt/intel_gt_regs.h
@@ -1060,6 +1060,7 @@
 #define   FLOW_CONTROL_ENABLE			REG_BIT(15)
 #define   UGM_BACKUP_MODE			REG_BIT(13)
 #define   MDQ_ARBITRATION_MODE			REG_BIT(12)
+#define   SYSTOLIC_DOP_CLOCK_GATING_DIS		REG_BIT(10)
 #define   PARTIAL_INSTRUCTION_SHOOTDOWN_DISABLE	REG_BIT(8)
 #define   STALL_DOP_GATING_DISABLE		REG_BIT(5)
 #define   THROTTLE_12_5				REG_GENMASK(4, 2)
diff --git a/drivers/gpu/drm/i915/gt/intel_lrc.c b/drivers/gpu/drm/i915/gt/intel_lrc.c
index d333400d29fe..e9ea576b96a4 100644
--- a/drivers/gpu/drm/i915/gt/intel_lrc.c
+++ b/drivers/gpu/drm/i915/gt/intel_lrc.c
@@ -1217,6 +1217,12 @@ gen12_emit_indirect_ctx_xcs(const struct intel_context *ce, u32 *cs)
 	cs = gen12_emit_timestamp_wa(ce, cs);
 	cs = gen12_emit_restore_scratch(ce, cs);
 
+	/* Wa_16013000631:dg2 */
+	if (IS_DG2_GRAPHICS_STEP(ce->engine->i915, G10, STEP_B0, STEP_C0) ||
+	    IS_DG2_G11(ce->engine->i915))
+		if (ce->engine->class == COMPUTE_CLASS)
+			cs = gen8_emit_pipe_control(cs, PIPE_CONTROL_INSTRUCTION_CACHE_INVALIDATE, 0);
+
 	return cs;
 }
 
diff --git a/drivers/gpu/drm/i915/gt/intel_workarounds.c b/drivers/gpu/drm/i915/gt/intel_workarounds.c
index 0471d475e680..0b9435d62808 100644
--- a/drivers/gpu/drm/i915/gt/intel_workarounds.c
+++ b/drivers/gpu/drm/i915/gt/intel_workarounds.c
@@ -1921,6 +1921,11 @@ static void dg2_whitelist_build(struct intel_engine_cs *engine)
 					  RING_FORCE_TO_NONPRIV_RANGE_4);
 
 		break;
+	case COMPUTE_CLASS:
+		/* Wa_16011157294:dg2_g10 */
+		if (IS_DG2_GRAPHICS_STEP(engine->i915, G10, STEP_A0, STEP_B0))
+			whitelist_reg(w, GEN9_CTX_PREEMPT_REG);
+		break;
 	default:
 		break;
 	}
@@ -2581,6 +2586,40 @@ xcs_engine_wa_init(struct intel_engine_cs *engine, struct i915_wa_list *wal)
 	}
 }
 
+/*
+ * The workarounds in this function apply to shared registers in
+ * the general render reset domain that aren't tied to a
+ * specific engine.  Since all render+compute engines get reset
+ * together, and the contents of these registers are lost during
+ * the shared render domain reset, we'll define such workarounds
+ * here and then add them to just a single RCS or CCS engine's
+ * workaround list (whichever engine has the XXXX flag).
+ */
+static void
+general_render_compute_wa_init(struct intel_engine_cs *engine, struct i915_wa_list *wal)
+{
+	struct drm_i915_private *i915 = engine->i915;
+
+	if (IS_XEHPSDV(i915)) {
+		/* Wa_1409954639 */
+		wa_masked_en(wal,
+			     GEN8_ROW_CHICKEN,
+			     SYSTOLIC_DOP_CLOCK_GATING_DIS);
+
+		/* Wa_1607196519 */
+		wa_masked_en(wal,
+			     GEN9_ROW_CHICKEN4,
+			     GEN12_DISABLE_GRF_CLEAR);
+
+		/* Wa_14010670810:xehpsdv */
+		wa_write_or(wal, XEHP_L3NODEARBCFG, XEHP_LNESPARE);
+
+		/* Wa_14010449647:xehpsdv */
+		wa_masked_en(wal, GEN7_HALF_SLICE_CHICKEN1,
+			     GEN7_PSD_SINGLE_PORT_DISPATCH_ENABLE);
+	}
+}
+
 static void
 engine_init_workarounds(struct intel_engine_cs *engine, struct i915_wa_list *wal)
 {
@@ -2589,6 +2628,14 @@ engine_init_workarounds(struct intel_engine_cs *engine, struct i915_wa_list *wal
 
 	engine_fake_wa_init(engine, wal);
 
+	/*
+	 * These are common workarounds that just need to applied
+	 * to a single RCS/CCS engine's workaround list since
+	 * they're reset as part of the general render domain reset.
+	 */
+	if (engine->class == RENDER_CLASS)
+		general_render_compute_wa_init(engine, wal);
+
 	if (engine->class == RENDER_CLASS)
 		rcs_engine_wa_init(engine, wal);
 	else
-- 
2.34.1


  parent reply	other threads:[~2022-02-28 17:43 UTC|newest]

Thread overview: 42+ messages / expand[flat|nested]  mbox.gz  Atom feed  top
2022-02-28 17:42 [PATCH v2 00/13] i915: Prepare for Xe_HP compute engines Matt Roper
2022-02-28 17:42 ` [Intel-gfx] " Matt Roper
2022-02-28 17:42 ` [PATCH v2 01/13] drm/i915/xehp: Define compute class and engine Matt Roper
2022-02-28 17:42   ` [Intel-gfx] " Matt Roper
2022-02-28 17:42 ` [PATCH v2 02/13] drm/i915/xehp: CCS shares the render reset domain Matt Roper
2022-02-28 17:42   ` [Intel-gfx] " Matt Roper
2022-02-28 17:42 ` [PATCH v2 03/13] drm/i915/xehp: Add Compute CS IRQ handlers Matt Roper
2022-02-28 17:42   ` [Intel-gfx] " Matt Roper
2022-02-28 17:42 ` [PATCH v2 04/13] drm/i915/xehp: compute engine pipe_control Matt Roper
2022-02-28 17:42   ` [Intel-gfx] " Matt Roper
2022-03-01  6:54   ` Lucas De Marchi
2022-03-01  6:54     ` [Intel-gfx] " Lucas De Marchi
2022-02-28 17:42 ` [PATCH v2 05/13] drm/i915/xehp: CCS should use RCS setup functions Matt Roper
2022-02-28 17:42   ` [Intel-gfx] " Matt Roper
2022-02-28 17:42 ` [Intel-gfx] [PATCH v2 06/13] drm/i915: Move context descriptor fields to intel_lrc.h Matt Roper
2022-02-28 17:42   ` Matt Roper
2022-03-01  6:57   ` [Intel-gfx] " Lucas De Marchi
2022-02-28 17:42 ` [Intel-gfx] [PATCH v2 07/13] drm/i915/xehp: Define context scheduling attributes in lrc descriptor Matt Roper
2022-02-28 17:42   ` Matt Roper
2022-02-28 17:42 ` [PATCH v2 08/13] drm/i915/xehp/guc: enable compute engine inside GuC Matt Roper
2022-02-28 17:42   ` [Intel-gfx] " Matt Roper
2022-03-01 18:55   ` Ceraolo Spurio, Daniele
2022-03-01 18:55     ` [Intel-gfx] " Ceraolo Spurio, Daniele
2022-02-28 17:42 ` [PATCH v2 09/13] drm/i915/xehp: Enable ccs/dual-ctx in RCU_MODE Matt Roper
2022-02-28 17:42   ` [Intel-gfx] " Matt Roper
2022-02-28 17:42 ` [Intel-gfx] [PATCH v2 10/13] drm/i915/xehp: Don't support parallel submission on compute / render Matt Roper
2022-02-28 17:42   ` Matt Roper
2022-03-01 19:04   ` Ceraolo Spurio, Daniele
2022-03-01 19:04     ` [Intel-gfx] " Ceraolo Spurio, Daniele
2022-02-28 17:42 ` [PATCH v2 11/13] drm/i915/xehp: handle fused off CCS engines Matt Roper
2022-02-28 17:42   ` [Intel-gfx] " Matt Roper
2022-03-01 22:54   ` Matt Roper
2022-03-01 22:54     ` [Intel-gfx] " Matt Roper
2022-02-28 17:42 ` Matt Roper [this message]
2022-02-28 17:42   ` [Intel-gfx] [PATCH v2 12/13] drm/i915/xehp: Add compute workarounds Matt Roper
2022-03-01 19:31   ` Ceraolo Spurio, Daniele
2022-02-28 17:42 ` [PATCH v2 13/13] drm/i915/xehpsdv: Move render/compute engine reset domains related workarounds Matt Roper
2022-02-28 17:42   ` [Intel-gfx] " Matt Roper
2022-02-28 22:48 ` [Intel-gfx] ✗ Fi.CI.CHECKPATCH: warning for i915: Prepare for Xe_HP compute engines Patchwork
2022-02-28 22:49 ` [Intel-gfx] ✗ Fi.CI.SPARSE: " Patchwork
2022-02-28 23:21 ` [Intel-gfx] ✓ Fi.CI.BAT: success " Patchwork
2022-03-01  9:38 ` [Intel-gfx] ✓ Fi.CI.IGT: " Patchwork

Reply instructions:

You may reply publicly to this message via plain-text email
using any one of the following methods:

* Save the following mbox file, import it into your mail client,
  and reply-to-all from there: mbox

  Avoid top-posting and favor interleaved quoting:
  https://en.wikipedia.org/wiki/Posting_style#Interleaved_style

* Reply using the --to, --cc, and --in-reply-to
  switches of git-send-email(1):

  git send-email \
    --in-reply-to=20220228174245.1569581-13-matthew.d.roper@intel.com \
    --to=matthew.d.roper@intel.com \
    --cc=dri-devel@lists.freedesktop.org \
    --cc=intel-gfx@lists.freedesktop.org \
    /path/to/YOUR_REPLY

  https://kernel.org/pub/software/scm/git/docs/git-send-email.html

* If your mail client supports setting the In-Reply-To header
  via mailto: links, try the mailto: link
Be sure your reply has a Subject: header at the top and a blank line before the message body.
This is an external index of several public inboxes,
see mirroring instructions on how to clone and mirror
all data and code used by this external index.