[ARM] pxa/dma: optimize irq handler loop

Reduce loop for dma irq handler callbacks to the minimum
required.

Since V1: included suggestion from Nicolas Pitre to improve
even further the loop.
Signed-off-by: Robert Jarzmik <robert.jarzmik@free.fr>
Signed-off-by: Eric Miao <eric.y.miao@gmail.com>

arch/arm/plat-pxa/dma.c

@@ -94,10 +94,12 @@ EXPORT_SYMBOL(pxa_free_dma);
 static irqreturn_t dma_irq_handler(int irq, void *dev_id)
 {
 	int i, dint = DINT;
+	struct dma_channel *channel;
 
-	for (i = 0; i < num_dma_channels; i++) {
-		if (dint & (1 << i)) {
-			struct dma_channel *channel = &dma_channels[i];
+	while (dint) {
+		i = __ffs(dint);
+		dint &= (dint - 1);
+		channel = &dma_channels[i];
 		if (channel->name && channel->irq_handler) {
 			channel->irq_handler(i, channel->data);
 		} else {
@@ -109,7 +111,6 @@ static irqreturn_t dma_irq_handler(int irq, void *dev_id)
 			DCSR(i) = DCSR_STARTINTR|DCSR_ENDINTR|DCSR_BUSERR;
 		}
 	}
-	}
 	return IRQ_HANDLED;
 }