[Intel-gfx] [PATCH 49/49] drm/i915: Cache last cmd descriptor when parsing

[Chris Wilson]

The cmd parser has the biggest impact on the BLT ring, because it is
relatively verbose composed to the other engines as the vertex data is
inline. It also typically has runs of repeating commands (again since
the vertex data is inline, it typically has sequences of XY_SETUP_BLT,
XY_SCANLINE_BLT..) We can easily reduce the impact of cmdparsing on
benchmarks by then caching the last descriptor and comparing it against
the next command header. To get maximum benefit, we also want to take
advantage of skipping a few validations and length determinations if the
header is unchanged between commands.

ivb i7-3720QM:
x11perf -dot: before 52.3M, after 124M (max 203M)
glxgears: before 7310 fps, after 7550 fps (max 7860 fps)

Signed-off-by: Chris Wilson <chris at chris-wilson.co.uk>
---
 drivers/gpu/drm/i915/i915_cmd_parser.c | 120 +++++++++++++++------------------
 1 file changed, 54 insertions(+), 66 deletions(-)

diff --git a/drivers/gpu/drm/i915/i915_cmd_parser.c b/drivers/gpu/drm/i915/i915_cmd_parser.c
index 60b30b4165d4..2843bce1b83c 100644
--- a/drivers/gpu/drm/i915/i915_cmd_parser.c
+++ b/drivers/gpu/drm/i915/i915_cmd_parser.c
@@ -754,6 +754,9 @@ void i915_cmd_parser_fini_ring(struct intel_engine_cs *ring)
 	fini_hash_table(ring);
 }
 
+/*
+ * Returns a pointer to a descriptor for the command specified by cmd_header.
+ */
 static const struct drm_i915_cmd_descriptor*
 find_cmd_in_table(struct intel_engine_cs *ring,
 		  u32 cmd_header)
@@ -773,37 +776,6 @@ find_cmd_in_table(struct intel_engine_cs *ring,
 	return NULL;
 }
 
-/*
- * Returns a pointer to a descriptor for the command specified by cmd_header.
- *
- * The caller must supply space for a default descriptor via the default_desc
- * parameter. If no descriptor for the specified command exists in the ring's
- * command parser tables, this function fills in default_desc based on the
- * ring's default length encoding and returns default_desc.
- */
-static const struct drm_i915_cmd_descriptor*
-find_cmd(struct intel_engine_cs *ring,
-	 u32 cmd_header,
-	 struct drm_i915_cmd_descriptor *default_desc)
-{
-	const struct drm_i915_cmd_descriptor *desc;
-	u32 mask;
-
-	desc = find_cmd_in_table(ring, cmd_header);
-	if (desc)
-		return desc;
-
-	mask = ring->get_cmd_length_mask(cmd_header);
-	if (!mask)
-		return NULL;
-
-	BUG_ON(!default_desc);
-	default_desc->flags = CMD_DESC_SKIP;
-	default_desc->length.mask = mask;
-
-	return default_desc;
-}
-
 static bool valid_reg(const u32 *table, int count, u32 addr)
 {
 	if (table && count != 0) {
@@ -844,17 +816,6 @@ static bool check_cmd(const struct intel_engine_cs *ring,
 		      const bool is_master,
 		      bool *oacontrol_set)
 {
-	if (desc->flags & CMD_DESC_REJECT) {
-		DRM_DEBUG_DRIVER("CMD: Rejected command: 0x%08X\n", *cmd);
-		return false;
-	}
-
-	if ((desc->flags & CMD_DESC_MASTER) && !is_master) {
-		DRM_DEBUG_DRIVER("CMD: Rejected master-only command: 0x%08X\n",
-				 *cmd);
-		return false;
-	}
-
 	if (desc->flags & CMD_DESC_REGISTER) {
 		u32 reg_addr = cmd[desc->reg.offset] & desc->reg.mask;
 
@@ -953,13 +914,14 @@ int i915_parse_cmds(struct intel_engine_cs *ring,
 		    bool is_master)
 {
 	u32 tmp[128];
+	struct drm_i915_cmd_descriptor default_desc;
 	struct sg_page_iter src_iter, dst_iter;
 	const struct drm_i915_cmd_descriptor *desc;
+	u32 last_cmd_header = 0;
 	int needs_clflush = 0;
 	void *src, *dst;
 	unsigned in, out;
 	u32 *buf, partial = 0, length;
-	struct drm_i915_cmd_descriptor default_desc = { 0 };
 	bool oacontrol_set = false; /* OACONTROL tracking. See check_cmd() */
 	int ret = 0;
 
@@ -1053,33 +1015,59 @@ int i915_parse_cmds(struct intel_engine_cs *ring,
 		}
 
 		do {
-			if (*cmd == MI_BATCH_BUFFER_END) {
-				ret = 0;
-				goto unmap_src;
-			}
+			if (*cmd != last_cmd_header) {
+				if (*cmd == MI_BATCH_BUFFER_END) {
+					ret = 0;
+					goto unmap_src;
+				}
 
-			desc = find_cmd(ring, *cmd, &default_desc);
-			if (!desc) {
-				DRM_DEBUG_DRIVER("CMD: Unrecognized command: 0x%08X\n",
-						 *cmd);
-				goto unmap_src;
-			}
+				desc = find_cmd_in_table(ring, *cmd);
+				if (desc) {
+					if (unlikely(desc->flags & CMD_DESC_REJECT)) {
+						DRM_DEBUG_DRIVER("CMD: Rejected command: 0x%08X\n", *cmd);
+						goto unmap_src;
+					}
+
+					if (unlikely((desc->flags & CMD_DESC_MASTER) && !is_master)) {
+						DRM_DEBUG_DRIVER("CMD: Rejected master-only command: 0x%08X\n",
+								 *cmd);
+						goto unmap_src;
+					}
+
+					/*
+					 * If the batch buffer contains a
+					 * chained batch, return an error that
+					 * tells the caller to abort and
+					 * dispatch the workload as a
+					 * non-secure batch.
+					 */
+					if (unlikely(desc->cmd.value == MI_BATCH_BUFFER_START)) {
+						ret = -EACCES;
+						goto unmap_src;
+					}
+
+					if (desc->flags & CMD_DESC_FIXED)
+						length = desc->length.fixed;
+					else
+						length = ((*cmd & desc->length.mask) + LENGTH_BIAS);
+				} else {
+					u32 mask = ring->get_cmd_length_mask(*cmd);
+					if (unlikely(!mask)) {
+						DRM_DEBUG_DRIVER("CMD: Unrecognized command: 0x%08X\n",
+								 *cmd);
+						goto unmap_src;
+					}
+
+					default_desc.flags = CMD_DESC_SKIP;
+					default_desc.length.mask = mask;
+					desc = &default_desc;
+
+					length = ((*cmd & mask) + LENGTH_BIAS);
+				}
 
-			/*
-			 * If the batch buffer contains a chained batch, return an
-			 * error that tells the caller to abort and dispatch the
-			 * workload as a non-secure batch.
-			 */
-			if (desc->cmd.value == MI_BATCH_BUFFER_START) {
-				ret = -EACCES;
-				goto unmap_src;
+				last_cmd_header = *cmd;
 			}
 
-			if (desc->flags & CMD_DESC_FIXED)
-				length = desc->length.fixed;
-			else
-				length = ((*cmd & desc->length.mask) + LENGTH_BIAS);
-
 			if (cmd + length > page_end) {
 				if (length + cmd > batch_end) {
 					DRM_DEBUG_DRIVER("CMD: Command length exceeds batch length: 0x%08X length=%u batchlen=%td\n",
-- 
2.1.4