summaryrefslogtreecommitdiff
path: root/include/asm-generic
diff options
context:
space:
mode:
authorArd Biesheuvel <ardb@kernel.org>2022-02-05 18:23:45 +0300
committerHerbert Xu <herbert@gondor.apana.org.au>2022-02-11 12:39:39 +0300
commit297565aa22cfa80ab0f88c3569693aea0b6afb6d (patch)
tree86c452349612ec00b52d83c78900bcf45b6bbd8d /include/asm-generic
parente8bf24bd439da1ee7f37c2b03f44c6ad37c0c8c0 (diff)
downloadlinux-297565aa22cfa80ab0f88c3569693aea0b6afb6d.tar.xz
lib/xor: make xor prototypes more friendly to compiler vectorization
Modern compilers are perfectly capable of extracting parallelism from the XOR routines, provided that the prototypes reflect the nature of the input accurately, in particular, the fact that the input vectors are expected not to overlap. This is not documented explicitly, but is implied by the interchangeability of the various C routines, some of which use temporary variables while others don't: this means that these routines only behave identically for non-overlapping inputs. So let's decorate these input vectors with the __restrict modifier, which informs the compiler that there is no overlap. While at it, make the input-only vectors pointer-to-const as well. Tested-by: Nathan Chancellor <nathan@kernel.org> Signed-off-by: Ard Biesheuvel <ardb@kernel.org> Reviewed-by: Nick Desaulniers <ndesaulniers@google.com> Link: https://github.com/ClangBuiltLinux/linux/issues/563 Signed-off-by: Herbert Xu <herbert@gondor.apana.org.au>
Diffstat (limited to 'include/asm-generic')
-rw-r--r--include/asm-generic/xor.h84
1 files changed, 56 insertions, 28 deletions
diff --git a/include/asm-generic/xor.h b/include/asm-generic/xor.h
index b62a2a56a4d4..44509d48fca2 100644
--- a/include/asm-generic/xor.h
+++ b/include/asm-generic/xor.h
@@ -8,7 +8,8 @@
#include <linux/prefetch.h>
static void
-xor_8regs_2(unsigned long bytes, unsigned long *p1, unsigned long *p2)
+xor_8regs_2(unsigned long bytes, unsigned long * __restrict p1,
+ const unsigned long * __restrict p2)
{
long lines = bytes / (sizeof (long)) / 8;
@@ -27,8 +28,9 @@ xor_8regs_2(unsigned long bytes, unsigned long *p1, unsigned long *p2)
}
static void
-xor_8regs_3(unsigned long bytes, unsigned long *p1, unsigned long *p2,
- unsigned long *p3)
+xor_8regs_3(unsigned long bytes, unsigned long * __restrict p1,
+ const unsigned long * __restrict p2,
+ const unsigned long * __restrict p3)
{
long lines = bytes / (sizeof (long)) / 8;
@@ -48,8 +50,10 @@ xor_8regs_3(unsigned long bytes, unsigned long *p1, unsigned long *p2,
}
static void
-xor_8regs_4(unsigned long bytes, unsigned long *p1, unsigned long *p2,
- unsigned long *p3, unsigned long *p4)
+xor_8regs_4(unsigned long bytes, unsigned long * __restrict p1,
+ const unsigned long * __restrict p2,
+ const unsigned long * __restrict p3,
+ const unsigned long * __restrict p4)
{
long lines = bytes / (sizeof (long)) / 8;
@@ -70,8 +74,11 @@ xor_8regs_4(unsigned long bytes, unsigned long *p1, unsigned long *p2,
}
static void
-xor_8regs_5(unsigned long bytes, unsigned long *p1, unsigned long *p2,
- unsigned long *p3, unsigned long *p4, unsigned long *p5)
+xor_8regs_5(unsigned long bytes, unsigned long * __restrict p1,
+ const unsigned long * __restrict p2,
+ const unsigned long * __restrict p3,
+ const unsigned long * __restrict p4,
+ const unsigned long * __restrict p5)
{
long lines = bytes / (sizeof (long)) / 8;
@@ -93,7 +100,8 @@ xor_8regs_5(unsigned long bytes, unsigned long *p1, unsigned long *p2,
}
static void
-xor_32regs_2(unsigned long bytes, unsigned long *p1, unsigned long *p2)
+xor_32regs_2(unsigned long bytes, unsigned long * __restrict p1,
+ const unsigned long * __restrict p2)
{
long lines = bytes / (sizeof (long)) / 8;
@@ -129,8 +137,9 @@ xor_32regs_2(unsigned long bytes, unsigned long *p1, unsigned long *p2)
}
static void
-xor_32regs_3(unsigned long bytes, unsigned long *p1, unsigned long *p2,
- unsigned long *p3)
+xor_32regs_3(unsigned long bytes, unsigned long * __restrict p1,
+ const unsigned long * __restrict p2,
+ const unsigned long * __restrict p3)
{
long lines = bytes / (sizeof (long)) / 8;
@@ -175,8 +184,10 @@ xor_32regs_3(unsigned long bytes, unsigned long *p1, unsigned long *p2,
}
static void
-xor_32regs_4(unsigned long bytes, unsigned long *p1, unsigned long *p2,
- unsigned long *p3, unsigned long *p4)
+xor_32regs_4(unsigned long bytes, unsigned long * __restrict p1,
+ const unsigned long * __restrict p2,
+ const unsigned long * __restrict p3,
+ const unsigned long * __restrict p4)
{
long lines = bytes / (sizeof (long)) / 8;
@@ -230,8 +241,11 @@ xor_32regs_4(unsigned long bytes, unsigned long *p1, unsigned long *p2,
}
static void
-xor_32regs_5(unsigned long bytes, unsigned long *p1, unsigned long *p2,
- unsigned long *p3, unsigned long *p4, unsigned long *p5)
+xor_32regs_5(unsigned long bytes, unsigned long * __restrict p1,
+ const unsigned long * __restrict p2,
+ const unsigned long * __restrict p3,
+ const unsigned long * __restrict p4,
+ const unsigned long * __restrict p5)
{
long lines = bytes / (sizeof (long)) / 8;
@@ -294,7 +308,8 @@ xor_32regs_5(unsigned long bytes, unsigned long *p1, unsigned long *p2,
}
static void
-xor_8regs_p_2(unsigned long bytes, unsigned long *p1, unsigned long *p2)
+xor_8regs_p_2(unsigned long bytes, unsigned long * __restrict p1,
+ const unsigned long * __restrict p2)
{
long lines = bytes / (sizeof (long)) / 8 - 1;
prefetchw(p1);
@@ -320,8 +335,9 @@ xor_8regs_p_2(unsigned long bytes, unsigned long *p1, unsigned long *p2)
}
static void
-xor_8regs_p_3(unsigned long bytes, unsigned long *p1, unsigned long *p2,
- unsigned long *p3)
+xor_8regs_p_3(unsigned long bytes, unsigned long * __restrict p1,
+ const unsigned long * __restrict p2,
+ const unsigned long * __restrict p3)
{
long lines = bytes / (sizeof (long)) / 8 - 1;
prefetchw(p1);
@@ -350,8 +366,10 @@ xor_8regs_p_3(unsigned long bytes, unsigned long *p1, unsigned long *p2,
}
static void
-xor_8regs_p_4(unsigned long bytes, unsigned long *p1, unsigned long *p2,
- unsigned long *p3, unsigned long *p4)
+xor_8regs_p_4(unsigned long bytes, unsigned long * __restrict p1,
+ const unsigned long * __restrict p2,
+ const unsigned long * __restrict p3,
+ const unsigned long * __restrict p4)
{
long lines = bytes / (sizeof (long)) / 8 - 1;
@@ -384,8 +402,11 @@ xor_8regs_p_4(unsigned long bytes, unsigned long *p1, unsigned long *p2,
}
static void
-xor_8regs_p_5(unsigned long bytes, unsigned long *p1, unsigned long *p2,
- unsigned long *p3, unsigned long *p4, unsigned long *p5)
+xor_8regs_p_5(unsigned long bytes, unsigned long * __restrict p1,
+ const unsigned long * __restrict p2,
+ const unsigned long * __restrict p3,
+ const unsigned long * __restrict p4,
+ const unsigned long * __restrict p5)
{
long lines = bytes / (sizeof (long)) / 8 - 1;
@@ -421,7 +442,8 @@ xor_8regs_p_5(unsigned long bytes, unsigned long *p1, unsigned long *p2,
}
static void
-xor_32regs_p_2(unsigned long bytes, unsigned long *p1, unsigned long *p2)
+xor_32regs_p_2(unsigned long bytes, unsigned long * __restrict p1,
+ const unsigned long * __restrict p2)
{
long lines = bytes / (sizeof (long)) / 8 - 1;
@@ -466,8 +488,9 @@ xor_32regs_p_2(unsigned long bytes, unsigned long *p1, unsigned long *p2)
}
static void
-xor_32regs_p_3(unsigned long bytes, unsigned long *p1, unsigned long *p2,
- unsigned long *p3)
+xor_32regs_p_3(unsigned long bytes, unsigned long * __restrict p1,
+ const unsigned long * __restrict p2,
+ const unsigned long * __restrict p3)
{
long lines = bytes / (sizeof (long)) / 8 - 1;
@@ -523,8 +546,10 @@ xor_32regs_p_3(unsigned long bytes, unsigned long *p1, unsigned long *p2,
}
static void
-xor_32regs_p_4(unsigned long bytes, unsigned long *p1, unsigned long *p2,
- unsigned long *p3, unsigned long *p4)
+xor_32regs_p_4(unsigned long bytes, unsigned long * __restrict p1,
+ const unsigned long * __restrict p2,
+ const unsigned long * __restrict p3,
+ const unsigned long * __restrict p4)
{
long lines = bytes / (sizeof (long)) / 8 - 1;
@@ -591,8 +616,11 @@ xor_32regs_p_4(unsigned long bytes, unsigned long *p1, unsigned long *p2,
}
static void
-xor_32regs_p_5(unsigned long bytes, unsigned long *p1, unsigned long *p2,
- unsigned long *p3, unsigned long *p4, unsigned long *p5)
+xor_32regs_p_5(unsigned long bytes, unsigned long * __restrict p1,
+ const unsigned long * __restrict p2,
+ const unsigned long * __restrict p3,
+ const unsigned long * __restrict p4,
+ const unsigned long * __restrict p5)
{
long lines = bytes / (sizeof (long)) / 8 - 1;