drm/i915/xehp: Define compute class and engine
authorMatt Roper <matthew.d.roper@intel.com>
Tue, 1 Mar 2022 23:15:37 +0000 (15:15 -0800)
committerMatt Roper <matthew.d.roper@intel.com>
Wed, 2 Mar 2022 14:45:16 +0000 (06:45 -0800)
Introduce a Compute Command Streamer (CCS), which has access to
the media and GPGPU pipelines (but not the 3D pipeline).

To begin with, define the compute class/engine common functions, based
on the existing render ones.

v2:
 - Add kerneldoc for drm_i915_gem_engine_class since we're adding a new
   element to it.  (Daniel)
 - Make engine class <-> guc class converters use lookup tables to make
   it more clear/explicit how the IDs map.  (Tvrtko)

v3:
 - Don't update uapi for now; we'll just include the driver-internal
   changes for the time being.

Bspec: 46167, 45544
Original-author: Michel Thierry
Cc: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Cc: Tvrtko Ursulin <tvrtko.ursulin@linux.intel.com>
Cc: Vinay Belgaumkar <vinay.belgaumkar@intel.com>
Signed-off-by: Rodrigo Vivi <rodrigo.vivi@intel.com>
Signed-off-by: Daniele Ceraolo Spurio <daniele.ceraolospurio@intel.com>
Signed-off-by: Aravind Iddamsetty <aravind.iddamsetty@intel.com>
Signed-off-by: Matt Roper <matthew.d.roper@intel.com>
Reviewed-by: Tvrtko Ursulin <tvrtko.ursulin@intel.com> #v1
Link: https://patchwork.freedesktop.org/patch/msgid/20220301231549.1817978-2-matthew.d.roper@intel.com
drivers/gpu/drm/i915/gt/intel_engine_cs.c
drivers/gpu/drm/i915/gt/intel_engine_types.h
drivers/gpu/drm/i915/gt/intel_engine_user.c
drivers/gpu/drm/i915/gt/intel_gt_regs.h
drivers/gpu/drm/i915/gt/uc/intel_guc_fwif.h
drivers/gpu/drm/i915/i915_reg.h

index e855c80..3190b7b 100644 (file)
@@ -156,6 +156,34 @@ static const struct engine_info intel_engines[] = {
                        { .graphics_ver = 12, .base = XEHP_VEBOX4_RING_BASE }
                },
        },
+       [CCS0] = {
+               .class = COMPUTE_CLASS,
+               .instance = 0,
+               .mmio_bases = {
+                       { .graphics_ver = 12, .base = GEN12_COMPUTE0_RING_BASE }
+               }
+       },
+       [CCS1] = {
+               .class = COMPUTE_CLASS,
+               .instance = 1,
+               .mmio_bases = {
+                       { .graphics_ver = 12, .base = GEN12_COMPUTE1_RING_BASE }
+               }
+       },
+       [CCS2] = {
+               .class = COMPUTE_CLASS,
+               .instance = 2,
+               .mmio_bases = {
+                       { .graphics_ver = 12, .base = GEN12_COMPUTE2_RING_BASE }
+               }
+       },
+       [CCS3] = {
+               .class = COMPUTE_CLASS,
+               .instance = 3,
+               .mmio_bases = {
+                       { .graphics_ver = 12, .base = GEN12_COMPUTE3_RING_BASE }
+               }
+       },
 };
 
 /**
index 36365bd..f4533cc 100644 (file)
@@ -33,7 +33,8 @@
 #define VIDEO_ENHANCEMENT_CLASS        2
 #define COPY_ENGINE_CLASS      3
 #define OTHER_CLASS            4
-#define MAX_ENGINE_CLASS       4
+#define COMPUTE_CLASS          5
+#define MAX_ENGINE_CLASS       5
 #define MAX_ENGINE_INSTANCE    7
 
 #define I915_MAX_SLICES        3
@@ -95,6 +96,7 @@ struct i915_ctx_workarounds {
 
 #define I915_MAX_VCS   8
 #define I915_MAX_VECS  4
+#define I915_MAX_CCS   4
 
 /*
  * Engine IDs definitions.
@@ -117,6 +119,11 @@ enum intel_engine_id {
        VECS2,
        VECS3,
 #define _VECS(n) (VECS0 + (n))
+       CCS0,
+       CCS1,
+       CCS2,
+       CCS3,
+#define _CCS(n) (CCS0 + (n))
        I915_NUM_ENGINES
 #define INVALID_ENGINE ((enum intel_engine_id)-1)
 };
index 9ce85a8..b8c9b6b 100644 (file)
@@ -47,6 +47,7 @@ static const u8 uabi_classes[] = {
        [COPY_ENGINE_CLASS] = I915_ENGINE_CLASS_COPY,
        [VIDEO_DECODE_CLASS] = I915_ENGINE_CLASS_VIDEO,
        [VIDEO_ENHANCEMENT_CLASS] = I915_ENGINE_CLASS_VIDEO_ENHANCE,
+       /* TODO: Add COMPUTE_CLASS mapping once ABI is available */
 };
 
 static int engine_cmp(void *priv, const struct list_head *A,
@@ -139,6 +140,7 @@ const char *intel_engine_class_repr(u8 class)
                [COPY_ENGINE_CLASS] = "bcs",
                [VIDEO_DECODE_CLASS] = "vcs",
                [VIDEO_ENHANCEMENT_CLASS] = "vecs",
+               [COMPUTE_CLASS] = "ccs",
        };
 
        if (class >= ARRAY_SIZE(uabi_names) || !uabi_names[class])
@@ -162,6 +164,7 @@ static int legacy_ring_idx(const struct legacy_ring *ring)
                [COPY_ENGINE_CLASS] = { BCS0, 1 },
                [VIDEO_DECODE_CLASS] = { VCS0, I915_MAX_VCS },
                [VIDEO_ENHANCEMENT_CLASS] = { VECS0, I915_MAX_VECS },
+               [COMPUTE_CLASS] = { CCS0, I915_MAX_CCS },
        };
 
        if (GEM_DEBUG_WARN_ON(ring->class >= ARRAY_SIZE(map)))
@@ -190,7 +193,7 @@ static void add_legacy_ring(struct legacy_ring *ring,
 void intel_engines_driver_register(struct drm_i915_private *i915)
 {
        struct legacy_ring ring = {};
-       u8 uabi_instances[4] = {};
+       u8 uabi_instances[5] = {};
        struct list_head *it, *next;
        struct rb_node **p, *prev;
        LIST_HEAD(engines);
index d752db5..530807b 100644 (file)
 #define   GEN11_KCR                            (19)
 #define   GEN11_GTPM                           (16)
 #define   GEN11_BCS                            (15)
+#define   GEN12_CCS3                           (7)
+#define   GEN12_CCS2                           (6)
+#define   GEN12_CCS1                           (5)
+#define   GEN12_CCS0                           (4)
 #define   GEN11_RCS0                           (0)
 #define   GEN11_VECS(x)                                (31 - (x))
 #define   GEN11_VCS(x)                         (x)
index 6a4612a..4b300b6 100644 (file)
@@ -46,8 +46,8 @@
 #define GUC_VIDEO_CLASS                        1
 #define GUC_VIDEOENHANCE_CLASS         2
 #define GUC_BLITTER_CLASS              3
-#define GUC_RESERVED_CLASS             4
-#define GUC_LAST_ENGINE_CLASS          GUC_RESERVED_CLASS
+#define GUC_COMPUTE_CLASS              4
+#define GUC_LAST_ENGINE_CLASS          GUC_COMPUTE_CLASS
 #define GUC_MAX_ENGINE_CLASSES         16
 #define GUC_MAX_INSTANCES_PER_CLASS    32
 
@@ -156,23 +156,37 @@ FIELD_PREP(HOST2GUC_PC_SLPC_REQUEST_MSG_1_EVENT_ID, id) | \
 FIELD_PREP(HOST2GUC_PC_SLPC_REQUEST_MSG_1_EVENT_ARGC, c) \
 )
 
+/* the GuC arrays don't include OTHER_CLASS */
+static u8 engine_class_guc_class_map[] = {
+       [RENDER_CLASS]            = GUC_RENDER_CLASS,
+       [COPY_ENGINE_CLASS]       = GUC_BLITTER_CLASS,
+       [VIDEO_DECODE_CLASS]      = GUC_VIDEO_CLASS,
+       [VIDEO_ENHANCEMENT_CLASS] = GUC_VIDEOENHANCE_CLASS,
+       [COMPUTE_CLASS]           = GUC_COMPUTE_CLASS,
+};
+
+static u8 guc_class_engine_class_map[] = {
+       [GUC_RENDER_CLASS]       = RENDER_CLASS,
+       [GUC_BLITTER_CLASS]      = COPY_ENGINE_CLASS,
+       [GUC_VIDEO_CLASS]        = VIDEO_DECODE_CLASS,
+       [GUC_VIDEOENHANCE_CLASS] = VIDEO_ENHANCEMENT_CLASS,
+       [GUC_COMPUTE_CLASS]      = COMPUTE_CLASS,
+};
+
 static inline u8 engine_class_to_guc_class(u8 class)
 {
-       BUILD_BUG_ON(GUC_RENDER_CLASS != RENDER_CLASS);
-       BUILD_BUG_ON(GUC_BLITTER_CLASS != COPY_ENGINE_CLASS);
-       BUILD_BUG_ON(GUC_VIDEO_CLASS != VIDEO_DECODE_CLASS);
-       BUILD_BUG_ON(GUC_VIDEOENHANCE_CLASS != VIDEO_ENHANCEMENT_CLASS);
+       BUILD_BUG_ON(ARRAY_SIZE(engine_class_guc_class_map) != MAX_ENGINE_CLASS + 1);
        GEM_BUG_ON(class > MAX_ENGINE_CLASS || class == OTHER_CLASS);
 
-       return class;
+       return engine_class_guc_class_map[class];
 }
 
 static inline u8 guc_class_to_engine_class(u8 guc_class)
 {
+       BUILD_BUG_ON(ARRAY_SIZE(guc_class_engine_class_map) != GUC_LAST_ENGINE_CLASS + 1);
        GEM_BUG_ON(guc_class > GUC_LAST_ENGINE_CLASS);
-       GEM_BUG_ON(guc_class == GUC_RESERVED_CLASS);
 
-       return guc_class;
+       return guc_class_engine_class_map[guc_class];
 }
 
 /* Work item for submitting workloads into work queue of GuC. */
index cfd5696..4da10e1 100644 (file)
 #define GEN11_VEBOX2_RING_BASE         0x1d8000
 #define XEHP_VEBOX3_RING_BASE          0x1e8000
 #define XEHP_VEBOX4_RING_BASE          0x1f8000
+#define GEN12_COMPUTE0_RING_BASE       0x1a000
+#define GEN12_COMPUTE1_RING_BASE       0x1c000
+#define GEN12_COMPUTE2_RING_BASE       0x1e000
+#define GEN12_COMPUTE3_RING_BASE       0x26000
 #define BLT_RING_BASE          0x22000