[<prev] [next>] [<thread-prev] [thread-next>] [day] [month] [year] [list]
Message-Id: <20210824133351.88179-3-andriy.shevchenko@linux.intel.com>
Date: Tue, 24 Aug 2021 16:33:51 +0300
From: Andy Shevchenko <andriy.shevchenko@...ux.intel.com>
To: Mauro Carvalho Chehab <mchehab+huawei@...nel.org>,
Sakari Ailus <sakari.ailus@...ux.intel.com>,
Andy Shevchenko <andriy.shevchenko@...ux.intel.com>,
linux-media@...r.kernel.org, linux-kernel@...r.kernel.org
Cc: Yong Zhi <yong.zhi@...el.com>, Bingbu Cao <bingbu.cao@...el.com>,
Dan Scally <djrscally@...il.com>,
Tianshu Qiu <tian.shu.qiu@...el.com>,
Mauro Carvalho Chehab <mchehab@...nel.org>
Subject: [RFT, PATCH v1 3/3] media: ipu3-cio2: Replace custom implementation of rotate()
rotate() is more efficient than custom implementation.
Replace the latter by the former.
Signed-off-by: Andy Shevchenko <andriy.shevchenko@...ux.intel.com>
---
This should be a copy'n'paste of the algorithm with a slight difference that
it should copy by 4 or 8 bytes at a time. Nonetheless it has to be tested.
Hence, RFT. (Obviously no hurry with this, we are close to release)
drivers/media/pci/intel/ipu3/ipu3-cio2-main.c | 59 ++-----------------
1 file changed, 5 insertions(+), 54 deletions(-)
diff --git a/drivers/media/pci/intel/ipu3/ipu3-cio2-main.c b/drivers/media/pci/intel/ipu3/ipu3-cio2-main.c
index 8bcba168cc57..0fd6040d2f2d 100644
--- a/drivers/media/pci/intel/ipu3/ipu3-cio2-main.c
+++ b/drivers/media/pci/intel/ipu3/ipu3-cio2-main.c
@@ -21,6 +21,7 @@
#include <linux/pfn.h>
#include <linux/pm_runtime.h>
#include <linux/property.h>
+#include <linux/sort.h>
#include <linux/vmalloc.h>
#include <media/v4l2-ctrls.h>
#include <media/v4l2-device.h>
@@ -1877,56 +1878,6 @@ static int __maybe_unused cio2_runtime_resume(struct device *dev)
return 0;
}
-/*
- * Helper function to advance all the elements of a circular buffer by "start"
- * positions
- */
-static void arrange(void *ptr, size_t elem_size, size_t elems, size_t start)
-{
- struct {
- size_t begin, end;
- } arr[2] = {
- { 0, start - 1 },
- { start, elems - 1 },
- };
-
-#define CHUNK_SIZE(a) ((a)->end - (a)->begin + 1)
-
- /* Loop as long as we have out-of-place entries */
- while (CHUNK_SIZE(&arr[0]) && CHUNK_SIZE(&arr[1])) {
- size_t size0, i;
-
- /*
- * Find the number of entries that can be arranged on this
- * iteration.
- */
- size0 = min(CHUNK_SIZE(&arr[0]), CHUNK_SIZE(&arr[1]));
-
- /* Swap the entries in two parts of the array. */
- for (i = 0; i < size0; i++) {
- u8 *d = ptr + elem_size * (arr[1].begin + i);
- u8 *s = ptr + elem_size * (arr[0].begin + i);
- size_t j;
-
- for (j = 0; j < elem_size; j++)
- swap(d[j], s[j]);
- }
-
- if (CHUNK_SIZE(&arr[0]) > CHUNK_SIZE(&arr[1])) {
- /* The end of the first array remains unarranged. */
- arr[0].begin += size0;
- } else {
- /*
- * The first array is fully arranged so we proceed
- * handling the next one.
- */
- arr[0].begin = arr[1].begin;
- arr[0].end = arr[1].begin + size0 - 1;
- arr[1].begin += size0;
- }
- }
-}
-
static void cio2_fbpt_rearrange(struct cio2_device *cio2, struct cio2_queue *q)
{
unsigned int i, j;
@@ -1940,10 +1891,10 @@ static void cio2_fbpt_rearrange(struct cio2_device *cio2, struct cio2_queue *q)
return;
if (j) {
- arrange(q->fbpt, sizeof(struct cio2_fbpt_entry) * CIO2_MAX_LOPS,
- CIO2_MAX_BUFFERS, j);
- arrange(q->bufs, sizeof(struct cio2_buffer *),
- CIO2_MAX_BUFFERS, j);
+ rotate(q->fbpt, sizeof(struct cio2_fbpt_entry) * CIO2_MAX_LOPS,
+ CIO2_MAX_BUFFERS, j, NULL);
+ rotate(q->bufs, sizeof(struct cio2_buffer *),
+ CIO2_MAX_BUFFERS, j, NULL);
}
/*
--
2.32.0
Powered by blists - more mailing lists