drm/i915/pvc: Add SSEU changes

PVC splits the mask of enabled DSS over two registers.  It also changes
the meaning of the EU fuse register such that each bit represents a
single EU rather than a pair of EUs.

Signed-off-by: Matt Roper <[email protected]>
Acked-by: Tvrtko Ursulin <[email protected]>
Reviewed-by: Balasubramani Vivekanandan <[email protected]>
Link: https://patchwork.freedesktop.org/patch/msgid/[email protected]

Author: mattrope

drivers/gpu/drm/i915/gt/intel_gt_regs.h

@@ -561,6 +561,7 @@
 #define   GEN11_GT_VEBOX_DISABLE_MASK		(0x0f << GEN11_GT_VEBOX_DISABLE_SHIFT)
 
 #define GEN12_GT_COMPUTE_DSS_ENABLE		_MMIO(0x9144)
+#define XEHPC_GT_COMPUTE_DSS_ENABLE_EXT		_MMIO(0x9148)
 
 #define GEN6_UCGCTL1				_MMIO(0x9400)
 #define   GEN6_GAMUNIT_CLOCK_GATE_DISABLE	(1 << 22)

drivers/gpu/drm/i915/gt/intel_sseu.c

@@ -210,27 +210,44 @@ static void xehp_sseu_info_init(struct intel_gt *gt)
 	struct intel_uncore *uncore = gt->uncore;
 	u16 eu_en = 0;
 	u8 eu_en_fuse;
+	int num_compute_regs, num_geometry_regs;
 	int eu;
 
+	if (IS_PONTEVECCHIO(gt->i915)) {
+		num_geometry_regs = 0;
+		num_compute_regs = 2;
+	} else {
+		num_geometry_regs = 1;
+		num_compute_regs = 1;
+	}
+
 	/*
 	 * The concept of slice has been removed in Xe_HP.  To be compatible
 	 * with prior generations, assume a single slice across the entire
 	 * device. Then calculate out the DSS for each workload type within
 	 * that software slice.
 	 */
-	intel_sseu_set_info(sseu, 1, 32, 16);
+	intel_sseu_set_info(sseu, 1,
+			    32 * max(num_geometry_regs, num_compute_regs),
+			    16);
 	sseu->has_xehp_dss = 1;
 
-	xehp_load_dss_mask(uncore, &sseu->geometry_subslice_mask, 1,
+	xehp_load_dss_mask(uncore, &sseu->geometry_subslice_mask,
+			   num_geometry_regs,
 			   GEN12_GT_GEOMETRY_DSS_ENABLE);
-	xehp_load_dss_mask(uncore, &sseu->compute_subslice_mask, 1,
-			   GEN12_GT_COMPUTE_DSS_ENABLE);
+	xehp_load_dss_mask(uncore, &sseu->compute_subslice_mask,
+			   num_compute_regs,
+			   GEN12_GT_COMPUTE_DSS_ENABLE,
+			   XEHPC_GT_COMPUTE_DSS_ENABLE_EXT);
 
 	eu_en_fuse = intel_uncore_read(uncore, XEHP_EU_ENABLE) & XEHP_EU_ENA_MASK;
 
-	for (eu = 0; eu < sseu->max_eus_per_subslice / 2; eu++)
-		if (eu_en_fuse & BIT(eu))
-			eu_en |= BIT(eu * 2) | BIT(eu * 2 + 1);
+	if (HAS_ONE_EU_PER_FUSE_BIT(gt->i915))
+		eu_en = eu_en_fuse;
+	else
+		for (eu = 0; eu < sseu->max_eus_per_subslice / 2; eu++)
+			if (eu_en_fuse & BIT(eu))
+				eu_en |= BIT(eu * 2) | BIT(eu * 2 + 1);
 
 	xehp_compute_sseu_info(sseu, eu_en);
 }

drivers/gpu/drm/i915/gt/intel_sseu.h

@@ -33,7 +33,7 @@ struct drm_printer;
  * Maximum number of 32-bit registers used by hardware to express the
  * enabled/disabled subslices.
  */
-#define I915_MAX_SS_FUSE_REGS	1
+#define I915_MAX_SS_FUSE_REGS	2
 #define I915_MAX_SS_FUSE_BITS	(I915_MAX_SS_FUSE_REGS * 32)
 
 /* Maximum number of EUs that can exist within a subslice or DSS. */

drivers/gpu/drm/i915/i915_drv.h

@@ -1422,6 +1422,8 @@ IS_SUBPLATFORM(const struct drm_i915_private *i915,
 
 #define HAS_3D_PIPELINE(i915)	(INTEL_INFO(i915)->has_3d_pipeline)
 
+#define HAS_ONE_EU_PER_FUSE_BIT(i915)	(INTEL_INFO(i915)->has_one_eu_per_fuse_bit)
+
 /* i915_gem.c */
 void i915_gem_init_early(struct drm_i915_private *dev_priv);
 void i915_gem_cleanup_early(struct drm_i915_private *dev_priv);

drivers/gpu/drm/i915/i915_pci.c

@@ -1090,7 +1090,8 @@ static const struct intel_device_info ats_m_info = {
 	XE_HP_FEATURES, \
 	.dma_mask_size = 52, \
 	.has_3d_pipeline = 0, \
-	.has_l3_ccs_read = 1
+	.has_l3_ccs_read = 1, \
+	.has_one_eu_per_fuse_bit = 1
 
 __maybe_unused
 static const struct intel_device_info pvc_info = {

drivers/gpu/drm/i915/intel_device_info.h

@@ -158,6 +158,7 @@ enum intel_ppgtt_type {
 	func(has_logical_ring_elsq); \
 	func(has_media_ratio_mode); \
 	func(has_mslices); \
+	func(has_one_eu_per_fuse_bit); \
 	func(has_pooled_eu); \
 	func(has_pxp); \
 	func(has_rc6); \