2015-12-06 10:36:14

by Zhaoxiu Zeng

Subject: [PATCH 04/10] omap_hwspinlock: Replace "hweight_long(i & 0xf) != 1" with "!is_power_of_2(i & 0xf)"

From: Zeng Zhaoxiu <[email protected]>

Signed-off-by: Zeng Zhaoxiu <[email protected]>
---
drivers/hwspinlock/omap_hwspinlock.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/hwspinlock/omap_hwspinlock.c b/drivers/hwspinlock/omap_hwspinlock.c
index ad2f8ca..1848a4c 100644
--- a/drivers/hwspinlock/omap_hwspinlock.c
+++ b/drivers/hwspinlock/omap_hwspinlock.c
@@ -29,6 +29,7 @@
#include <linux/hwspinlock.h>
#include <linux/of.h>
#include <linux/platform_device.h>
+#include <linux/log2.h>

#include "hwspinlock_internal.h"

@@ -125,7 +126,7 @@ static int omap_hwspinlock_probe(struct platform_device *pdev)
goto iounmap_base;

/* one of the four lsb's must be set, and nothing else */
- if (hweight_long(i & 0xf) != 1 || i > 8) {
+ if (!is_power_of_2(i & 0xf) || i > 8) {
ret = -EINVAL;
goto iounmap_base;
}
--
2.5.0

2015-12-07 11:08:46

by Ohad Ben Cohen

Subject: Re: [PATCH 04/10] omap_hwspinlock: Replace "hweight_long(i & 0xf) != 1" with "!is_power_of_2(i & 0xf)"

Hi,

On Sun, Dec 6, 2015 at 12:33 PM, Zhaoxiu Zeng <[email protected]> wrote:
>
> From: Zeng Zhaoxiu <[email protected]>
>
> Signed-off-by: Zeng Zhaoxiu <[email protected]>

Please explain why do you think we should make this change.

Btw, the original code used is_power_of_2, but we thought hweight is
more explicit so it was adopted.

Thanks,
Ohad.

2015-12-07 15:04:47

by Zhaoxiu Zeng

Subject: Re: [PATCH 04/10] omap_hwspinlock: Replace "hweight_long(i & 0xf) != 1" with "!is_power_of_2(i & 0xf)"

在 2015/12/7 19:08, Ohad Ben-Cohen 写道:
> Hi,
>
> On Sun, Dec 6, 2015 at 12:33 PM, Zhaoxiu Zeng <[email protected]> wrote:
>>
>> From: Zeng Zhaoxiu <[email protected]>
>>
>> Signed-off-by: Zeng Zhaoxiu <[email protected]>
>
> Please explain why do you think we should make this change.

is_power_of_2 is simple, and faster than "hweightN(x) == 1" on most architectures.
And the "& 0xf" operation is unnecessary, we will check whether or not greater than 8 behind.

>
> Btw, the original code used is_power_of_2, but we thought hweight is
> more explicit so it was adopted.
>
> Thanks,
> Ohad.
>

---
drivers/hwspinlock/omap_hwspinlock.c | 3 ++-
1 file changed, 2 insertions(+), 1 deletion(-)

diff --git a/drivers/hwspinlock/omap_hwspinlock.c b/drivers/hwspinlock/omap_hwspinlock.c
index ad2f8ca..1848a4c 100644
--- a/drivers/hwspinlock/omap_hwspinlock.c
+++ b/drivers/hwspinlock/omap_hwspinlock.c
@@ -29,6 +29,7 @@
#include <linux/hwspinlock.h>
#include <linux/of.h>
#include <linux/platform_device.h>
+#include <linux/log2.h>

#include "hwspinlock_internal.h"

@@ -125,7 +126,7 @@ static int omap_hwspinlock_probe(struct platform_device *pdev)
goto iounmap_base;

/* one of the four lsb's must be set, and nothing else */
- if (hweight_long(i & 0xf) != 1 || i > 8) {
+ if (!is_power_of_2(i) || i > 8) {
ret = -EINVAL;
goto iounmap_base;
}
--

2015-12-07 16:17:23

by Ohad Ben Cohen

Subject: Re: [PATCH 04/10] omap_hwspinlock: Replace "hweight_long(i & 0xf) != 1" with "!is_power_of_2(i & 0xf)"

On Mon, Dec 7, 2015 at 5:03 PM, zhaoxiu.zeng <[email protected]> wrote:
> is_power_of_2 is simple, and faster than "hweightN(x) == 1" on most architectures.

Thanks. I'm not sure that speed is a major concern here, since this
code executes only once during the lifetime of the driver. Readability
is probably more important.