The optimization is achieved by following techniques:
> data alignment [gain from aligned memory access on read/write]
> POWER7 gains performance with loop unrolling/unwinding
[gain by reduction of branch penalty].
> zero padding done by calling optimized memset
This patch changes de default symbol redirection for internal call of
memcpy, memset, memchr, and strlen to the IFUNC resolved ones. The
performance improvement is noticeable in algorithms that uses these
symbols extensible, like the regex functions.
This patch fixes some powerpc32 and powerpc64 builds with
--disable-multi-arch option along with different --with-cpu=powerN.
It cleanups the Implies directories by removing the multiarch
folder for non multiarch config and also fixing two assembly
implementations: powerpc64/power7/strncat.S that is calling the
wrong strlen; and power8/fpu/s_isnan.S that misses the hidden_def and
weak_alias directives.
This patch add an optimized strpbrk for POWER7 by using a different
algorithm than default implementation: it constructs a table based on
the 'accept' argument and use this table to check for any occurance on
the input string. The idea is similar as x86_64 uses.
For PowerPC some tunings were added, such as unroll loops and memory
clear using VSX instructions.
This patch add a optimized strcspn for POWER7 by using a different
algorithm than default implementation: it constructs a table based on
the 'accept' argument and use this table to check for any occurance
on the input string. The idea is similar as x86_64 uses.
For PowerPC some tunings were added, such as unroll loops and align
stack memory to table to 16 bytes (so VSX clean can ran without
alignment issues).
The roundl assembly implementation
(sysdeps/powerpc/powerpc64/fpu/s_roundl.S)
returns wrong results for some inputs where first double is a exact
integer and the precision is determined by second long double.
Checking on implementation comments and history, I am very confident the
assembly implementation was based on a version before commit
5c68d40169 that fixes BZ#2423 (Errors in
long double (ldbl-128ibm) rounding functions in glibc-2.4).
By just removing the implementation and make the build select
sysdeps/ieee754/ldbl-128ibm/s_roundl.c instead fixes the failing math.
This fixes 16707.
The nearbyintl assembly implementation
(sysdeps/powerpc/powerpc64/fpu/s_nearbyintl.S)
returns wrong results for some inputs where first double is a exact
integer and the precision is determined by second long double.
Checking on implementation comments and history, I am very confident the
assembly implementation was based on a version before commit
5c68d40169 that fixes BZ#2423 (Errors in
long double (ldbl-128ibm) rounding functions in glibc-2.4).
By just removing the implementation and make the build select
sysdeps/ieee754/ldbl-128ibm/s_nearbyintl.c instead fixes the failing
math.
Fixes BZ#16706.
The ceill assembly implementation (sysdeps/powerpc/powerpc64/fpu/s_ceill.S)
returns wrong results for some inputs where first double is a exact
integer and the precision is determined by second long double.
Checking on implementation comments and history, I am very confident the
assembly implementation was based on a version before commit
5c68d40169 that fixes BZ#2423 (Errors in
long double (ldbl-128ibm) rounding functions in glibc-2.4).
By just removing the implementation and make the build select
sysdeps/ieee754/ldbl-128ibm/s_ceill.c instead fixes the failing math.
Fixes BZ#16701.
This patch fixes an issue for powerpc64[le] static build where __bzero
is definied in multiple places (memset-ppc64.o and bzero.o). It is now
defined only in bzero.o and memset-ppc64.o only defined __bzero_ppc for
both dynamic and static library.
Fixes BZ#16683.
The optimization is achieved by following techniques:
> hashing of needle.
> hashing avoids scanning of duplicate entries in needle across the string.
> initializing the hash table with Vector instructions (VSX) by quadword access.
> unrolling when scanning for character in string across hash table.
The optimization is achieved by following techniques:
1. Doubleword aligned memory access and compares using
cmpb instruction.
2. Loop unrolling for byte load/store.
3. CPU pre-fetch to avoid cache miss.
This patch optimizes strrchr() for ppc64. It uses aligned memory
access along with cmpb instruction and CPU prefetch to avoid
cache misses for speed improvement.
This patch add a optimized llround/llroundf implementation for POWER8
using the new Move From VSR Doubleword instruction to gains some
cycles from FP to GRP register move.
This patch add a optimized llrint/llrintf implementation for POWER8
using the new Move From VSR Doubleword instruction to gains some
cycles from FP to GRP register move.
This patch add a optimized finite/finitef implementation for POWER8
using the new Move From VSR Doubleword instruction to gains some
cycles from FP to GRP register move.
This patch add a optimized isinf/isinff implementation for POWER8
using the new Move From VSR Doubleword instruction to gains some
cycles from FP to GRP register move.
This patch add a optimized isnan/isnanf implementation for POWER8
using the new Move From VSR Doubleword instruction to gains some
cycles from FP to GRP register move.
The truncl assembly implementation (sysdeps/powerpc/powerpc64/fpu/s_truncl.S)
returns wrong results for some inputs where first double is a exact integer
and the precision is determined by second long double.
Checking on implementation comments and history, I am very confident the
assembly implementation was based on a version before commit
5c68d40169 that fixes BZ#2423 (Errors in
long double (ldbl-128ibm) rounding functions in glibc-2.4).
By just removing the implementation and make the build select
sysdeps/ieee754/ldbl-128ibm/s_truncl.c instead it fixes tgammal
issues regarding wrong result sign.
For PPC64, all the wrappers at sysdeps are superfluous: they are
basically the same implementation from math/w_sqrt.c with the
'#ifdef _IEEE_LIBM'. And the power4 version just force the 'fsqrt'
instruction utilization with an inline assembly, which is already
handled by math_private.h __ieee754_sqrt implementation.
This patch add optimized __mpn_addmul, __mpn_addsub, __mpn_lshift, and
__mpn_mul_1 implementations for PowerPC64. They are originally from GMP
with adjustments for GLIBC.
This patch add static probes for setjmp/longjmp in the way gdb expects,fixing
the gdb.base/longjmp.exp gdb testcases.
It changes the symbol_name and use macros to to avoid change the probe names
and ending up adding more logic on GDB (since with the expected name
GDB work seamlessly).
The ELFv2 ABI changes the calling convention by passing and returning
structures in registers in more cases than the old ABI:
http://gcc.gnu.org/ml/gcc-patches/2013-11/msg01145.htmlhttp://gcc.gnu.org/ml/gcc-patches/2013-11/msg01147.html
For the most part, this does not affect glibc, since glibc assembler
files do not use structure parameters / return values. However, one
place is affected: the LD_AUDIT interface provides a structure to
the audit routine that contains all registers holding function
argument and return values for the intercepted PLT call.
Since the new ABI now sometimes uses registers to return values
that were never used for this purpose in the old ABI, this structure
has to be extended. To force audit routines to be modified for the
new ABI if necessary, the patch defines v2 variants of the la_ppc64
types and routines.
In addition, the patch contains two unrelated changes to the
PLT trampoline routines: it fixes a bug where FPR return values
were stored in the wrong place, and it removes the unnecessary
save/restore of CR.
This updates glibc for the changes in the ELFv2 relating to the
stack frame layout. These are described in more detail here:
http://gcc.gnu.org/ml/gcc-patches/2013-11/msg01149.htmlhttp://gcc.gnu.org/ml/gcc-patches/2013-11/msg01146.html
Specifically, the "compiler and linker doublewords" were removed,
which has the effect that the save slot for the TOC register is
now at offset 24 rather than 40 to the stack pointer.
In addition, a function may now no longer necessarily assume that
its caller has set up a 64-byte register save area its use.
To address the first change, the patch goes through all assembler
files and replaces immediate offsets in instructions accessing the
ABI-defined stack slots by symbolic offsets. Those already were
defined in ucontext_i.sym and used in some of the context routines,
but that doesn't really seem like the right place for those defines.
The patch instead defines those symbolic offsets in sysdeps.h,
in two variants for the old and new ABI, and uses them systematically
in all assembler files, not just the context routines.
The second change only affected a few assembler files that used
the save area to temporarily store some registers. In those
cases where this happens within a leaf function, this patch
changes the code to store those registers to the "red zone"
below the stack pointer. Otherwise, the functions already allocate
a stack frame, and the patch changes them to add extra space in
these frames as temporary space for the ELFv2 ABI.
This is a follow-on to the previous patch to support the ELFv2 ABI in the
dynamic loader, split off into its own patch since it is just an optional
optimization.
In the ELFv2 ABI, most functions define both a global and a local entry
point; the local entry requires r2 to be already set up by the caller
to point to the callee's TOC; while the global entry does not require
the caller to know about the callee's TOC, but it needs to set up r12
to the callee's entry point address.
Now, when setting up a PLT slot, the dynamic linker will usually need
to enter the target function's global entry point. However, if the
linker can prove that the target function is in the same DSO as the
PLT slot itself, and the whole DSO only uses a single TOC (which the
linker will let ld.so know via a DT_PPC64_OPT entry), then it is
possible to actually enter the local entry point address into the
PLT slot, for a slight improvement in performance.
Note that this uncovered a problem on the first call via _dl_runtime_resolve,
because that routine neglected to restore the caller's TOC before calling
the target function for the first time, since it assumed that function
would always reload its own TOC anyway ...
This patch adds support for the ELFv2 ABI feature to remove function
descriptors. See this GCC patch for in-depth discussion:
http://gcc.gnu.org/ml/gcc-patches/2013-11/msg01141.html
This mostly involves two types of changes: updating assembler source
files to the new logic, and updating the dynamic loader.
After the refactoring in the previous patch, most of the assembler source
changes can be handled simply by providing ELFv2 versions of the
macros in sysdep.h. One somewhat non-obvious change is in __GI__setjmp:
this used to "fall through" to the immediately following __setjmp ENTRY
point. This is no longer safe in the ELFv2 since ENTRY defines both
a global and a local entry point, and you cannot simply fall through
to a global entry point as it requires r12 to be set up.
Also, makecontext needs to be updated to set up registers according to
the new ABI for calling into the context's start routine.
The dynamic linker changes mostly consist of removing special code
to handle function descriptors. We also need to support the new PLT
and glink format used by the the ELFv2 linker, see:
https://sourceware.org/ml/binutils/2013-10/msg00376.html
In addition, the dynamic linker now verifies that the dynamic libraries
it loads match its own ABI.
The hack in VDSO_IFUNC_RET to "synthesize" a function descriptor
for vDSO routines is also no longer necessary for ELFv2.
This is the first patch to support the new ELFv2 ABI in glibc.
As preparation, this patch simply refactors some of the powerpc64 assembler
code to move all code related to creating function descriptors (.opd section)
or using function descriptors (function pointer call) into a central place
in sysdep.h.
Note that most locations creating .opd entries were already using macros
in sysdep.h, this patch simply extends this to the remaining places.
No relevant change in generated code expected.
This patch updates glibc in accordance with the binutils patch checked in here:
https://sourceware.org/ml/binutils/2013-10/msg00372.html
This changes the various R_PPC64_..._HI and _HA relocations to report
32-bit overflows. The motivation is that existing uses of @h / @ha
are to build up 32-bit offsets (for the "medium model" TOC access
that GCC now defaults to), and we'd really like to see failures at
link / load time rather than silent truncations.
For those rare cases where a modifier is needed to build up a 64-bit
constant, new relocations _HIGH / _HIGHA are supported.
The patch also fixes a bug in overflow checking for the R_PPC64_ADDR30
and R_PPC64_ADDR32 relocations.
Autoconf has been deprecating configure.in for quite a long time.
Rename all our configure.in and preconfigure.in files to .ac.
Signed-off-by: Mike Frysinger <vapier@gentoo.org>
This patch intends to unify both strcpy and stpcpy implementationsi
for PPC64 and PPC64/POWER7. The idead default powerpc64 implementation
is to provide both doubleword and word aligned memory access.
For PPC64/POWER7 is also provide doubleword and word memory access,
remove the branch hints, use the cmpb instruction for compare
doubleword/words, and add an optimization for inputs of same alignment.
* sysdeps/powerpc/powerpc32/dl-machine.c (__process_machine_rela):
Use stdint types in rather than __attribute__((mode())).
* sysdeps/powerpc/powerpc64/dl-machine.h (elf_machine_rela): Likewise.
http://sourceware.org/ml/libc-alpha/2013-08/msg00105.html
Like strnlen, memchr and memrchr had a number of defects fixed by this
patch as well as adding little-endian support. The first one I
noticed was that the entry to the main loop needlessly checked for
"are we done yet?" when we know the size is large enough that we can't
be done. The second defect I noticed was that the main loop count was
wrong, which in turn meant that the small loop needed to handle an
extra word. Thirdly, there is nothing to say that the string can't
wrap around zero, except of course that we'd normally hit a segfault
on trying to read from address zero. Fixing that simplified a number
of places:
- /* Are we done already? */
- addi r9,r8,8
- cmpld r9,r7
- bge L(null)
becomes
+ cmpld r8,r7
+ beqlr
However, the exit gets an extra test because I test for being on the
last word then if so whether the byte offset is less than the end.
Overall, the change is a win.
Lastly, memrchr used the wrong cache hint.
* sysdeps/powerpc/powerpc64/power7/memchr.S: Replace rlwimi with
insrdi. Make better use of reg selection to speed exit slightly.
Schedule entry path a little better. Remove useless "are we done"
checks on entry to main loop. Handle wrapping around zero address.
Correct main loop count. Handle single left-over word from main
loop inline rather than by using loop_small. Remove extra word
case in loop_small caused by wrong loop count. Add little-endian
support.
* sysdeps/powerpc/powerpc32/power7/memchr.S: Likewise.
* sysdeps/powerpc/powerpc64/power7/memrchr.S: Likewise. Use proper
cache hint.
* sysdeps/powerpc/powerpc32/power7/memrchr.S: Likewise.
* sysdeps/powerpc/powerpc64/power7/rawmemchr.S: Add little-endian
support. Avoid rlwimi.
* sysdeps/powerpc/powerpc32/power7/rawmemchr.S: Likewise.
http://sourceware.org/ml/libc-alpha/2013-08/msg00104.html
One of the things I noticed when looking at power7 timing is that rlwimi
is cracked and the two resulting insns have a register dependency.
That makes it a little slower than the equivalent rldimi.
* sysdeps/powerpc/powerpc64/memset.S: Replace rlwimi with
insrdi. Formatting.
* sysdeps/powerpc/powerpc64/power4/memset.S: Likewise.
* sysdeps/powerpc/powerpc64/power6/memset.S: Likewise.
* sysdeps/powerpc/powerpc64/power7/memset.S: Likewise.
* sysdeps/powerpc/powerpc32/power4/memset.S: Likewise.
* sysdeps/powerpc/powerpc32/power6/memset.S: Likewise.
* sysdeps/powerpc/powerpc32/power7/memset.S: Likewise.
http://sourceware.org/ml/libc-alpha/2013-08/msg00103.html
LIttle-endian support for memcpy. I spent some time cleaning up the
64-bit power7 memcpy, in order to avoid the extra alignment traps
power7 takes for little-endian. It probably would have been better
to copy the linux kernel version of memcpy.
* sysdeps/powerpc/powerpc32/power4/memcpy.S: Add little endian support.
* sysdeps/powerpc/powerpc32/power6/memcpy.S: Likewise.
* sysdeps/powerpc/powerpc32/power7/memcpy.S: Likewise.
* sysdeps/powerpc/powerpc32/power7/mempcpy.S: Likewise.
* sysdeps/powerpc/powerpc64/memcpy.S: Likewise.
* sysdeps/powerpc/powerpc64/power4/memcpy.S: Likewise.
* sysdeps/powerpc/powerpc64/power6/memcpy.S: Likewise.
* sysdeps/powerpc/powerpc64/power7/memcpy.S: Likewise.
* sysdeps/powerpc/powerpc64/power7/mempcpy.S: Likewise. Make better
use of regs. Use power7 mtocrf. Tidy function tails.
http://sourceware.org/ml/libc-alpha/2013-08/msg00102.html
This is a rather large patch due to formatting and renaming. The
formatting changes were to make it possible to compare power7 and
power4 versions of memcmp. Using different register defines came
about while I was wrestling with the code, trying to find spare
registers at one stage. I found it much simpler if we refer to a reg
by the same name throughout a function, so it's better if short-term
multiple use regs like rTMP are referred to using their register
number. I made the cr field usage changes when attempting to reload
rWORDn regs in the exit path to byte swap before comparing when
little-endian. That proved a bad idea due to the pipelining involved
in the main loop; Offsets to reload the regs were different first
time around the loop.. Anyway, I left the cr field usage changes in
place for consistency.
Aside from these more-or-less cosmetic changes, I fixed a number of
places where an early exit path restores regs unnecessarily, removed
some dead code, and optimised one or two exits.
* sysdeps/powerpc/powerpc64/power7/memcmp.S: Add little-endian support.
Formatting. Consistently use rXXX register defines or rN defines.
Use early exit labels that avoid restoring unused non-volatile regs.
Make cr field use more consistent with rWORDn compares. Rename
regs used as shift registers for unaligned loop, using rN defines
for short lifetime/multiple use regs.
* sysdeps/powerpc/powerpc64/power4/memcmp.S: Likewise.
* sysdeps/powerpc/powerpc32/power7/memcmp.S: Likewise. Exit with
addi 1,1,64 to pop stack frame. Simplify return value code.
* sysdeps/powerpc/powerpc32/power4/memcmp.S: Likewise.
http://sourceware.org/ml/libc-alpha/2013-08/msg00101.html
Adds little-endian support to optimised strchr assembly. I've also
tweaked the big-endian code a little. In power7/strchr.S there's a
check in the tail of the function that we didn't match 0 before
finding a c match, done by comparing leading zero counts. It's just
as valid, and quicker, to compare the raw output from cmpb.
Another little tweak is to use rldimi/insrdi in place of rlwimi for
the power7 strchr functions. Since rlwimi is cracked, it is a few
cycles slower. rldimi can be used on the 32-bit power7 functions
too.
* sysdeps/powerpc/powerpc64/power7/strchr.S (strchr): Add little-endian
support. Correct typos, formatting. Optimize tail. Use insrdi
rather than rlwimi.
* sysdeps/powerpc/powerpc32/power7/strchr.S: Likewise.
* sysdeps/powerpc/powerpc64/power7/strchrnul.S (__strchrnul): Add
little-endian support. Correct typos.
* sysdeps/powerpc/powerpc32/power7/strchrnul.S: Likewise. Use insrdi
rather than rlwimi.
* sysdeps/powerpc/powerpc64/strchr.S (rTMP4, rTMP5): Define. Use
in loop and entry code to keep "and." results.
(strchr): Add little-endian support. Comment. Move cntlzd
earlier in tail.
* sysdeps/powerpc/powerpc32/strchr.S: Likewise.
http://sourceware.org/ml/libc-alpha/2013-08/msg00100.html
The strcpy changes for little-endian are quite straight-forward, just
a matter of rotating the last word differently.
I'll note that the powerpc64 version of stpcpy is just begging to be
converted to use 64-bit loads and stores..
* sysdeps/powerpc/powerpc64/strcpy.S: Add little-endian support:
* sysdeps/powerpc/powerpc32/strcpy.S: Likewise.
* sysdeps/powerpc/powerpc64/stpcpy.S: Likewise.
* sysdeps/powerpc/powerpc32/stpcpy.S: Likewise.
http://sourceware.org/ml/libc-alpha/2013-08/msg00099.html
More little-endian support. I leave the main strcmp loops unchanged,
(well, except for renumbering rTMP to something other than r0 since
it's needed in an addi insn) and modify the tail for little-endian.
I noticed some of the big-endian tail code was a little untidy so have
cleaned that up too.
* sysdeps/powerpc/powerpc64/strcmp.S (rTMP2): Define as r0.
(rTMP): Define as r11.
(strcmp): Add little-endian support. Optimise tail.
* sysdeps/powerpc/powerpc32/strcmp.S: Similarly.
* sysdeps/powerpc/powerpc64/strncmp.S: Likewise.
* sysdeps/powerpc/powerpc32/strncmp.S: Likewise.
* sysdeps/powerpc/powerpc64/power4/strncmp.S: Likewise.
* sysdeps/powerpc/powerpc32/power4/strncmp.S: Likewise.
* sysdeps/powerpc/powerpc64/power7/strncmp.S: Likewise.
* sysdeps/powerpc/powerpc32/power7/strncmp.S: Likewise.
http://sourceware.org/ml/libc-alpha/2013-08/msg00098.html
The existing strnlen code has a number of defects, so this patch is more
than just adding little-endian support. The changes here are similar to
those for memchr.
* sysdeps/powerpc/powerpc64/power7/strnlen.S (strnlen): Add
little-endian support. Remove unnecessary "are we done" tests.
Handle "s" wrapping around zero and extremely large "size".
Correct main loop count. Handle single left-over word from main
loop inline rather than by using small_loop. Correct comments.
Delete "zero" tail, use "end_max" instead.
* sysdeps/powerpc/powerpc32/power7/strnlen.S: Likewise.
http://sourceware.org/ml/libc-alpha/2013-08/msg00097.html
This is the first of nine patches adding little-endian support to the
existing optimised string and memory functions. I did spend some
time with a power7 simulator looking at cycle by cycle behaviour for
memchr, but most of these patches have not been run on cpu simulators
to check that we are going as fast as possible. I'm sure PowerPC can
do better. However, the little-endian support mostly leaves main
loops unchanged, so I'm banking on previous authors having done a
good job on big-endian.. As with most code you stare at long enough,
I found some improvements for big-endian too.
Little-endian support for strlen. Like most of the string functions,
I leave the main word or multiple-word loops substantially unchanged,
just needing to modify the tail.
Removing the branch in the power7 functions is just a tidy. .align
produces a branch anyway. Modifying regs in the non-power7 functions
is to suit the new little-endian tail.
* sysdeps/powerpc/powerpc64/power7/strlen.S (strlen): Add little-endian
support. Don't branch over align.
* sysdeps/powerpc/powerpc32/power7/strlen.S: Likewise.
* sysdeps/powerpc/powerpc64/strlen.S (strlen): Add little-endian support.
Rearrange tmp reg use to suit. Comment.
* sysdeps/powerpc/powerpc32/strlen.S: Likewise.
http://sourceware.org/ml/libc-alpha/2013-08/msg00090.html
This patch fixes symbol versioning in setjmp/longjmp. The existing
code uses raw versions, which results in wrong symbol versioning when
you want to build glibc with a base version of 2.19 for LE.
Note that the merging the 64-bit and 32-bit versions in novmx-lonjmp.c
and pt-longjmp.c doesn't result in GLIBC_2.0 versions for 64-bit, due
to the base in shlib_versions.
* sysdeps/powerpc/longjmp.c: Use proper symbol versioning macros.
* sysdeps/powerpc/novmx-longjmp.c: Likewise.
* sysdeps/powerpc/powerpc32/bsd-_setjmp.S: Likewise.
* sysdeps/powerpc/powerpc32/bsd-setjmp.S: Likewise.
* sysdeps/powerpc/powerpc32/fpu/__longjmp.S: Likewise.
* sysdeps/powerpc/powerpc32/fpu/setjmp.S: Likewise.
* sysdeps/powerpc/powerpc32/mcount.c: Likewise.
* sysdeps/powerpc/powerpc32/setjmp.S: Likewise.
* sysdeps/powerpc/powerpc64/setjmp.S: Likewise.
* nptl/sysdeps/unix/sysv/linux/powerpc/pt-longjmp.c: Likewise.
http://sourceware.org/ml/libc-alpha/2013-08/msg00089.html
Little-endian fixes for setjmp/longjmp. When writing these I noticed
the setjmp code corrupts the non volatile VMX registers when using an
unaligned buffer. Anton fixed this, and also simplified it quite a
bit.
The current code uses boilerplate for the case where we want to store
16 bytes to an unaligned address. For that we have to do a
read/modify/write of two aligned 16 byte quantities. In our case we
are storing a bunch of back to back data (consective VMX registers),
and only the start and end of the region need the read/modify/write.
[BZ #15723]
* sysdeps/powerpc/jmpbuf-offsets.h: Comment fix.
* sysdeps/powerpc/powerpc32/fpu/__longjmp-common.S: Correct
_dl_hwcap access for little-endian.
* sysdeps/powerpc/powerpc32/fpu/setjmp-common.S: Likewise. Don't
destroy vmx regs when saving unaligned.
* sysdeps/powerpc/powerpc64/__longjmp-common.S: Correct CR load.
* sysdeps/powerpc/powerpc64/setjmp-common.S: Likewise CR save. Don't
destroy vmx regs when saving unaligned.
The pointer guard used for pointer mangling was not initialized for
static applications resulting in the security feature being disabled.
The pointer guard is now correctly initialized to a random value for
static applications. Existing static applications need to be
recompiled to take advantage of the fix.
The test tst-ptrguard1-static and tst-ptrguard1 add regression
coverage to ensure the pointer guards are sufficiently random
and initialized to a default value.
