about summary refs log tree commit diff
path: root/nptl/tst-rwlock2a.c
diff options
context:
space:
mode:
authorStefan Liebler <stli@linux.ibm.com>2018-12-18 13:57:08 +0100
committerStefan Liebler <stli@linux.ibm.com>2018-12-18 13:57:08 +0100
commitcdd927d98cc38acf55e1c6594b5c9451df8f239f (patch)
tree442bb8c8fa5eb23ade648bbc99db0614cbfe56b3 /nptl/tst-rwlock2a.c
parent2ee1bc57ab50737ee2ab88c4d796b90e08b4bf93 (diff)
downloadglibc-cdd927d98cc38acf55e1c6594b5c9451df8f239f.tar.gz
glibc-cdd927d98cc38acf55e1c6594b5c9451df8f239f.tar.xz
glibc-cdd927d98cc38acf55e1c6594b5c9451df8f239f.zip
S390: Add z13 memmove ifunc variant.
This patch introduces a z13 specific ifunc variant for memmove.
As the common code implementation, it checks if we can copy from
the beginning to the end - with z196 memcpy implementation - or
if we have to copy from the end to the beginning.
The latter case is done by using vector load/store instructions.

If vector instructions are not available, the common-code is
used as fallback.  Therefore it is implemented in memmove-c with
a different name.
Furthermore the ifunc logic decides if we need the common-code
implementation at all.  If vector instructions are supported
due to the minimum architecture level set we can skip the
common-code ifunc variant.

ChangeLog:

	* sysdeps/s390/Makefile (sysdep_routines): Add memmove-c.
	* sysdeps/s390/ifunc-memcpy.h (HAVE_MEMMOVE_IFUNC,
	HAVE_MEMMOVE_IFUNC_AND_VX_SUPPORT, MEMMOVE_DEFAULT,
	HAVE_MEMMOVE_C, MEMMOVE_C,  HAVE_MEMMOVE_Z13, MEMMOVE_Z13):
	New defines.
	* sysdeps/s390/memcpy-z900.S: Add z13 memmove implementation.
	* sysdeps/s390/memmove-c.c: New file.
	* sysdeps/s390/memmove.c: Likewise.
	* sysdeps/s390/multiarch/ifunc-impl-list.c
	(__libc_ifunc_impl_list): Add ifunc variants for memmove.
Diffstat (limited to 'nptl/tst-rwlock2a.c')
0 files changed, 0 insertions, 0 deletions