Patchwork powerpc/spufs: improve search of node for contexts with SPU affinity

login
register
mail settings
Submitter Andre Detsch
Date Oct. 20, 2008, 3:27 p.m.
Message ID <200810201327.03078.adetsch@br.ibm.com>
Download mbox | patch
Permalink /patch/5074/
State Accepted
Headers show

Comments

Andre Detsch - Oct. 20, 2008, 3:27 p.m.
This patch improves redability of the code responsible for trying to find
a node with enough SPUs not committed to other affinity gangs.

An additional check is also added, to avoid taking into account gangs that
have no SPU affinity.

Signed-off-by: Andre Detsch <adetsch@br.ibm.com>

Patch

Index: spufs/arch/powerpc/platforms/cell/spufs/sched.c
===================================================================
--- spufs.orig/arch/powerpc/platforms/cell/spufs/sched.c
+++ spufs/arch/powerpc/platforms/cell/spufs/sched.c
@@ -312,6 +312,15 @@  static struct spu *aff_ref_location(stru
 	 */
 	node = cpu_to_node(raw_smp_processor_id());
 	for (n = 0; n < MAX_NUMNODES; n++, node++) {
+		/*
+		 * "available_spus" counts how many spus are not potentially
+		 * going to be used by other affinity gangs whose reference
+		 * context is already in place. Although this code seeks to
+		 * avoid having affinity gangs with a summed amount of
+		 * contexts bigger than the amount of spus in the node,
+		 * this may happen sporadically. In this case, available_spus
+		 * becomes negative, which is harmless.
+		 */
 		int available_spus;

 		node = (node < MAX_NUMNODES) ? node : 0;
@@ -321,12 +330,10 @@  static struct spu *aff_ref_location(stru
 		available_spus = 0;
 		mutex_lock(&cbe_spu_info[node].list_mutex);
 		list_for_each_entry(spu, &cbe_spu_info[node].spus, cbe_list) {
-			if (spu->ctx && spu->ctx->gang
-					&& spu->ctx->aff_offset == 0)
-				available_spus -=
-					(spu->ctx->gang->contexts - 1);
-			else
-				available_spus++;
+			if (spu->ctx && spu->ctx->gang && !spu->ctx->aff_offset
+					&& spu->ctx->gang->aff_ref_spu)
+				available_spus -= spu->ctx->gang->contexts;
+			available_spus++;
 		}
 		if (available_spus < ctx->gang->contexts) {
 			mutex_unlock(&cbe_spu_info[node].list_mutex);