microblaze: Do loop unrolling for optimized memset implementation

Align implementation with memcpy and memmove where also remaining bytes are
copied via final switch case instead of using simple implementations which
loop. But this alignment has much stronger reason and definitely aligning
implementation is not the key point here. It is just good to have in mind
that the same technique is used already there.

In GCC 10, now -ftree-loop-distribute-patterns optimization is on at O2.
This optimization causes GCC to convert the while loop in memset.c into a
call to memset.
So this optimization is transforming a loop in a memset/memcpy into a call
to the function itself. This makes the memset implementation as recursive.
"-freestanding" option will disable the built-in library function but it
has been added in generic library implementation.

In default microblaze kernel defconfig we have CONFIG_OPT_LIB_FUNCTION
enabled so it will always pick optimized version of memset which is target
specific so we are replacing the while() loop with switch case to avoid
recursive memset call.

Issue with freestanding was already discussed in connection to commit
33d0f96ffd73 ("lib/string.c: Use freestanding environment") and also this
is topic in glibc and gcc.
https://gcc.gnu.org/bugzilla/show_bug.cgi?id=56888
http://patchwork.ozlabs.org/project/glibc/patch/20191121021040.14554-1-sandra@codesourcery.com/

Signed-off-by: Michal Simek <michal.simek@xilinx.com>
Signed-off-by: Mahesh Bodapati <mbodapat@xilinx.com>
Link: https://lore.kernel.org/r/10a432e269a6d3349cf458e4f5792522779cba0d.1645797329.git.michal.simek@xilinx.com

diff --git a/arch/microblaze/lib/memset.c b/arch/microblaze/lib/memset.c
index 615a2f8..7c2352d 100644 (file)
--- a/arch/microblaze/lib/memset.c
+++ b/arch/microblaze/lib/memset.c
@@ -74,8 +74,19 @@ void *memset(void *v_src, int c, __kernel_size_t n)
        }
 
        /* Simple, byte oriented memset or the rest of count. */
-       while (n--)
+       switch (n) {
+       case 3:
                *src++ = c;
+               fallthrough;
+       case 2:
+               *src++ = c;
+               fallthrough;
+       case 1:
+               *src++ = c;
+               break;
+       default:
+               break;
+       }
 
        return v_src;
 }