sna: Align uploads to start on page boundaries

This reduces the number of loops and restarts required in the kernel.

Signed-off-by: Chris Wilson <chris@chris-wilson.co.uk>

4 files changed, 9 insertions, 6 deletions

diff --git a/src/sna/compiler.h b/src/sna/compiler.h
index 62f51f0e..75e691ad 100644
--- a/src/sna/compiler.h
+++ b/src/sna/compiler.h
@@ -39,6 +39,7 @@
 #define pure __attribute__((pure))
 #define __packed__ __attribute__((__packed__))
 #define flatten __attribute__((flatten))
+#define page_aligned __attribute__((aligned(4096)))
 #else
 #define likely(expr) (expr)
 #define unlikely(expr) (expr)
@@ -50,6 +51,7 @@
 #define pure
 #define __packed__
 #define flatten
+#define page_aligned
 #endif
 
 #define HAS_GCC(major, minor) defined(__GNUC__) && (__GNUC__ > (major) || __GNUC__ == (major) && __GNUC_MINOR__ >= (minor))
diff --git a/src/sna/kgem.h b/src/sna/kgem.h
index f2b1c980..e4862921 100644
--- a/src/sna/kgem.h
+++ b/src/sna/kgem.h
@@ -194,9 +194,9 @@ struct kgem {
 	void (*retire)(struct kgem *kgem);
 	void (*expire)(struct kgem *kgem);
 
-	uint32_t batch[64*1024-8];
-	struct drm_i915_gem_exec_object2 exec[256];
-	struct drm_i915_gem_relocation_entry reloc[4096];
+	uint32_t batch[64*1024-8] page_aligned;
+	struct drm_i915_gem_exec_object2 exec[256] page_aligned;
+	struct drm_i915_gem_relocation_entry reloc[4096] page_aligned;
 	uint16_t reloc__self[256];
 
 #ifdef DEBUG_MEMORY
diff --git a/src/sna/sna.h b/src/sna/sna.h
index 61f26722..20981b3b 100644
--- a/src/sna/sna.h
+++ b/src/sna/sna.h
@@ -198,6 +198,8 @@ enum {
 };
 
 struct sna {
+	struct kgem kgem;
+
 	ScrnInfoPtr scrn;
 
 	unsigned flags;
@@ -289,7 +291,6 @@ struct sna {
 	InputHandlerProc uevent_handler;
 #endif
 
-	struct kgem kgem;
 	struct sna_render render;
 
 #if DEBUG_MEMORY
diff --git a/src/sna/sna_driver.c b/src/sna/sna_driver.c
index 5c8ba2db..1cfba5b6 100644
--- a/src/sna/sna_driver.c
+++ b/src/sna/sna_driver.c
@@ -515,10 +515,10 @@ static Bool sna_pre_init(ScrnInfoPtr scrn, int flags)
 	sna_selftest();
 
 	if (((uintptr_t)scrn->driverPrivate) & 1) {
-		sna = xnfcalloc(sizeof(struct sna), 1);
-		if (sna == NULL)
+		if (posix_memalign((void **)&sna, 4096, sizeof(*sna)))
 			return FALSE;
 
+		memset(sna, 0, sizeof(*sna)); /* should be unnecessary */
 		sna->info = (void *)((uintptr_t)scrn->driverPrivate & ~1);
 		scrn->driverPrivate = sna;