mfd: Micro-optimization on twl4030 IRQ handler
__ffs() will be far faster than the for loop used. Signed-off-by: Felipe Balbi <balbi@ti.com> Signed-off-by: Samuel Ortiz <sameo@linux.intel.com>
This commit is contained in:
parent
f01b1f90bf
commit
5a903090e7
|
@ -293,7 +293,6 @@ static unsigned twl4030_irq_base;
|
|||
*/
|
||||
static irqreturn_t handle_twl4030_pih(int irq, void *devid)
|
||||
{
|
||||
int module_irq;
|
||||
irqreturn_t ret;
|
||||
u8 pih_isr;
|
||||
|
||||
|
@ -304,12 +303,13 @@ static irqreturn_t handle_twl4030_pih(int irq, void *devid)
|
|||
return IRQ_NONE;
|
||||
}
|
||||
|
||||
/* these handlers deal with the relevant SIH irq status */
|
||||
for (module_irq = twl4030_irq_base;
|
||||
pih_isr;
|
||||
pih_isr >>= 1, module_irq++) {
|
||||
if (pih_isr & 0x1)
|
||||
handle_nested_irq(module_irq);
|
||||
while (pih_isr) {
|
||||
unsigned long pending = __ffs(pih_isr);
|
||||
unsigned int irq;
|
||||
|
||||
pih_isr &= ~BIT(pending);
|
||||
irq = pending + twl4030_irq_base;
|
||||
handle_nested_irq(irq);
|
||||
}
|
||||
|
||||
return IRQ_HANDLED;
|
||||
|
|
Loading…
Reference in New Issue