commit 2d651eb926
Author: H.J. Lu <hjl.tools@gmail.com>
Date: Fri Sep 18 07:55:14 2020 -0700
x86: Move x86 processor cache info to cpu_features
missed _SC_LEVEL1_ICACHE_LINESIZE.
1. Add level1_icache_linesize to struct cpu_features.
2. Initialize level1_icache_linesize by calling handle_intel,
handle_zhaoxin and handle_amd with _SC_LEVEL1_ICACHE_LINESIZE.
3. Return level1_icache_linesize for _SC_LEVEL1_ICACHE_LINESIZE.
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
1. Move x86 processor cache info to _dl_x86_cpu_features in ld.so.
2. Update tunable bounds with TUNABLE_SET_WITH_BOUNDS.
3. Move x86 cache info initialization to dl-cacheinfo.h and initialize
x86 cache info in init_cpu_features ().
4. Put x86 cache info for libc in cacheinfo.h, which is included in
libc-start.c in libc.a and is included in cacheinfo.c in libc.so.
Reviewed-by: Adhemerval Zanella <adhemerval.zanella@linaro.org>
I used these shell commands:
../glibc/scripts/update-copyrights $PWD/../gnulib/build-aux/update-copyright
(cd ../glibc && git commit -am"[this commit message]")
and then ignored the output, which consisted lines saying "FOO: warning:
copyright statement not found" for each of 6694 files FOO.
I then removed trailing white space from benchtests/bench-pthread-locks.c
and iconvdata/tst-iconv-big5-hkscs-to-2ucs4.c, to work around this
diagnostic from Savannah:
remote: *** pre-commit check failed ...
remote: *** error: lines with trailing whitespace found
remote: error: hook declined to update refs/heads/master
X86 CPU features in ld.so are initialized by init_cpu_features, which is
invoked by DL_PLATFORM_INIT from _dl_sysdep_start. But when ld.so is
loaded by static executable, DL_PLATFORM_INIT is never called. Also
x86 cache info in libc.o and libc.a is initialized by a constructor
which may be called too late. Since some fields in _rtld_global_ro
in ld.so are initialized by dynamic relocation, we can also initialize
x86 CPU features in _rtld_global_ro in ld.so and cache info in libc.so
by initializing dummy function pointers in ld.so and libc.so via IFUNC
relocation.
Key points:
1. IFUNC is always supported, independent of --enable-multi-arch or
--disable-multi-arch. Linker generates IFUNC relocations from input
IFUNC objects and ld.so performs IFUNC relocations.
2. There are no IFUNC dependencies in ld.so before dynamic relocation
have been performed,
3. The x86 CPU features in ld.so is initialized by DL_PLATFORM_INIT
in dynamic executable and by IFUNC relocation in dlopen in static
executable.
4. The x86 cache info in libc.o is initialized by IFUNC relocation.
5. In libc.a, both x86 CPU features and cache info are initialized from
ARCH_INIT_CPU_FEATURES, not by IFUNC relocation, before __libc_early_init
is called.
Note: _dl_x86_init_cpu_features can be called more than once from
DL_PLATFORM_INIT and during relocation in ld.so.
The __x86_shared_non_temporal_threshold determines when memcpy on x86
uses non_temporal stores to avoid pushing other data out of the last
level cache.
This patch proposes to revert the calculation change made by H.J. Lu's
patch of June 2, 2017.
H.J. Lu's patch selected a threshold suitable for a single thread
getting maximum performance. It was tuned using the single threaded
large memcpy micro benchmark on an 8 core processor. The last change
changes the threshold from using 3/4 of one thread's share of the
cache to using 3/4 of the entire cache of a multi-threaded system
before switching to non-temporal stores. Multi-threaded systems with
more than a few threads are server-class and typically have many
active threads. If one thread consumes 3/4 of the available cache for
all threads, it will cause other active threads to have data removed
from the cache. Two examples show the range of the effect. John
McCalpin's widely parallel Stream benchmark, which runs in parallel
and fetches data sequentially, saw a 20% slowdown with this patch on
an internal system test of 128 threads. This regression was discovered
when comparing OL8 performance to OL7. An example that compares
normal stores to non-temporal stores may be found at
https://vgatherps.github.io/2018-09-02-nontemporal/. A simple test
shows performance loss of 400 to 500% due to a failure to use
nontemporal stores. These performance losses are most likely to occur
when the system load is heaviest and good performance is critical.
The tunable x86_non_temporal_threshold can be used to override the
default for the knowledgable user who really wants maximum cache
allocation to a single thread in a multi-threaded system.
The manual entry for the tunable has been expanded to provide
more information about its purpose.
modified: sysdeps/x86/cacheinfo.c
modified: manual/tunables.texi
Support usable check for all CPU features with the following changes:
1. Change struct cpu_features to
struct cpuid_features
{
struct cpuid_registers cpuid;
struct cpuid_registers usable;
};
struct cpu_features
{
struct cpu_features_basic basic;
struct cpuid_features features[COMMON_CPUID_INDEX_MAX];
unsigned int preferred[PREFERRED_FEATURE_INDEX_MAX];
...
};
so that there is a usable bit for each cpuid bit.
2. After the cpuid bits have been initialized, copy the known bits to the
usable bits. EAX/EBX from INDEX_1 and EAX from INDEX_7 aren't used for
CPU feature detection.
3. Clear the usable bits which require OS support.
4. If the feature is supported by OS, copy its cpuid bit to its usable
bit.
5. Replace HAS_CPU_FEATURE and CPU_FEATURES_CPU_P with CPU_FEATURE_USABLE
and CPU_FEATURE_USABLE_P to check if a feature is usable.
6. Add DEPR_FPU_CS_DS for INDEX_7_EBX_13.
7. Unset MPX feature since it has been deprecated.
The results are
1. If the feature is known and doesn't requre OS support, its usable bit
is copied from the cpuid bit.
2. Otherwise, its usable bit is copied from the cpuid bit only if the
feature is known to supported by OS.
3. CPU_FEATURE_USABLE/CPU_FEATURE_USABLE_P are used to check if the
feature can be used.
4. HAS_CPU_FEATURE/CPU_FEATURE_CPU_P are used to check if CPU supports
the feature.
Since
commit c867597bff
Author: H.J. Lu <hjl.tools@gmail.com>
Date: Wed Jun 8 13:57:50 2016 -0700
X86-64: Remove previous default/SSE2/AVX2 memcpy/memmove
removed the only usage of __x86_prefetchw, we can remove the unused
__x86_prefetchw.
Add x86_rep_movsb_threshold and x86_rep_stosb_threshold to tunables
to update thresholds for "rep movsb" and "rep stosb" at run-time.
Note that the user specified threshold for "rep movsb" smaller than
the minimum threshold will be ignored.
Reviewed-by: Carlos O'Donell <carlos@redhat.com>
Commit a98dc92dd1 ("x86: Add cache
information support for Zhaoxin processors") introduced an unused
variable warning in the default i686-linux-gnu build:
In file included from ../sysdeps/i386/cacheinfo.c:3:
../sysdeps/x86/cacheinfo.c: In function 'init_cacheinfo':
../sysdeps/x86/cacheinfo.c:762:16: error: unused variable 'eax' [-Werror=unused-variable]
762 | unsigned int eax;
| ^~~
To obtain Zhaoxin CPU cache information, add a new function
handle_zhaoxin().
Add a new function get_common_cache_info() that extracts the code
in init_cacheinfo() to get the value of the variable shared, threads.
Add Zhaoxin branch in init_cacheinfo() for initializing variables,
such as __x86_shared_cache_size.
Extend CPUID support for all feature bits from CPUID. Add a new macro,
CPU_FEATURE_USABLE, which can be used to check if a feature is usable at
run-time, instead of HAS_CPU_FEATURE and HAS_ARCH_FEATURE.
Add COMMON_CPUID_INDEX_D_ECX_1, COMMON_CPUID_INDEX_80000007 and
COMMON_CPUID_INDEX_80000008 to check CPU feature bits in them.
Tested on i686 and x86-64 as well as using build-many-glibcs.py with
x86 targets.
* sysdeps/x86/cacheinfo.c (intel_check_word): Updated for
cpu_features_basic.
(__cache_sysconf): Likewise.
(init_cacheinfo): Likewise.
* sysdeps/x86/cpu-features.c (get_extended_indeces): Also
populate COMMON_CPUID_INDEX_80000007 and
COMMON_CPUID_INDEX_80000008.
(get_common_indices): Also populate COMMON_CPUID_INDEX_D_ECX_1.
Use CPU_FEATURES_CPU_P (cpu_features, XSAVEC) to check if
XSAVEC is available. Set the bit_arch_XXX_Usable bits.
(init_cpu_features): Use _Static_assert on
index_arch_Fast_Unaligned_Load.
__get_cpuid_registers and __get_arch_feature. Updated for
cpu_features_basic. Set stepping in cpu_features.
* sysdeps/x86/cpu-features.h: (FEATURE_INDEX_1): Changed to enum.
(FEATURE_INDEX_2): New.
(FEATURE_INDEX_MAX): Changed to enum.
(COMMON_CPUID_INDEX_D_ECX_1): New.
(COMMON_CPUID_INDEX_80000007): Likewise.
(COMMON_CPUID_INDEX_80000008): Likewise.
(cpuid_registers): Likewise.
(cpu_features_basic): Likewise.
(CPU_FEATURE_USABLE): Likewise.
(bit_arch_XXX_Usable): Likewise.
(cpu_features): Use cpuid_registers and cpu_features_basic.
(bit_arch_XXX): Reweritten.
(bit_cpu_XXX): Likewise.
(index_cpu_XXX): Likewise.
(reg_XXX): Likewise.
* sysdeps/x86/tst-get-cpu-features.c: Include <stdio.h> and
<support/check.h>.
(CHECK_CPU_FEATURE): New.
(CHECK_CPU_FEATURE_USABLE): Likewise.
(cpu_kinds): Likewise.
(do_test): Print vendor, family, model and stepping. Check
HAS_CPU_FEATURE and CPU_FEATURE_USABLE.
(TEST_FUNCTION): Removed.
Include <support/test-driver.c> instead of
"../../test-skeleton.c".
* sysdeps/x86_64/multiarch/sched_cpucount.c (__sched_cpucount):
Check POPCNT instead of POPCOUNT.
* sysdeps/x86_64/multiarch/test-multiarch.c (do_test): Likewise.
The current IFUNC selection is based on microbenchmarks in glibc. It
should give the best performance for most workloads. But other choices
may have better performance for a particular workload or on the hardware
which wasn't available at the selection was made. The environment
variable, GLIBC_TUNABLES=glibc.tune.ifunc=-xxx,yyy,-zzz...., can be used
to enable CPU/ARCH feature yyy, disable CPU/ARCH feature yyy and zzz,
where the feature name is case-sensitive and has to match the ones in
cpu-features.h. It can be used by glibc developers to override the
IFUNC selection to tune for a new processor or improve performance for
a particular workload. It isn't intended for normal end users.
NOTE: the IFUNC selection may change over time. Please check all
multiarch implementations when experimenting.
Also, GLIBC_TUNABLES=glibc.tune.x86_non_temporal_threshold=NUMBER is
provided to set threshold to use non temporal store to NUMBER,
GLIBC_TUNABLES=glibc.tune.x86_data_cache_size=NUMBER to set data cache
size, GLIBC_TUNABLES=glibc.tune.x86_shared_cache_size=NUMBER to set
shared cache size.
* elf/dl-tunables.list (tune): Add ifunc,
x86_non_temporal_threshold,
x86_data_cache_size and x86_shared_cache_size.
* manual/tunables.texi: Document glibc.tune.ifunc,
glibc.tune.x86_data_cache_size, glibc.tune.x86_shared_cache_size
and glibc.tune.x86_non_temporal_threshold.
* sysdeps/unix/sysv/linux/x86/dl-sysdep.c: New file.
* sysdeps/x86/cpu-tunables.c: Likewise.
* sysdeps/x86/cacheinfo.c
(init_cacheinfo): Check and get data cache size, shared cache
size and non temporal threshold from cpu_features.
* sysdeps/x86/cpu-features.c [HAVE_TUNABLES] (TUNABLE_NAMESPACE):
New.
[HAVE_TUNABLES] Include <unistd.h>.
[HAVE_TUNABLES] Include <elf/dl-tunables.h>.
[HAVE_TUNABLES] (TUNABLE_CALLBACK (set_ifunc)): Likewise.
[HAVE_TUNABLES] (init_cpu_features): Use TUNABLE_GET to set
IFUNC selection, data cache size, shared cache size and non
temporal threshold.
* sysdeps/x86/cpu-features.h (cpu_features): Add data_cache_size,
shared_cache_size and non_temporal_threshold.
Since cpu_features is available, use it instead of dl_x86_cpu_features.
* sysdeps/x86/cacheinfo.c (intel_check_word): Accept cpu_features
and use it instead of dl_x86_cpu_features.
(handle_intel): Replace maxidx with cpu_features. Pass
cpu_features to intel_check_word.
(__cache_sysconf): Pass cpu_features to handle_intel.
(init_cacheinfo): Likewise. Use cpu_features instead of
dl_x86_cpu_features.
__x86_shared_non_temporal_threshold was set to 6 times of per-core
shared cache size, based on the large memcpy micro benchmark in glibc
on a 8-core processor. For a processor with more than 8 cores, the
threshold is too low. Set __x86_shared_non_temporal_threshold to the
3/4 of the total shared cache size so that it is unchanged on 8-core
processors. On processors with less than 8 cores, the threshold is
lower.
* sysdeps/x86/cacheinfo.c (__x86_shared_non_temporal_threshold):
Set to the 3/4 of the total shared cache size.
Remove is_intel, is_amd and max_cpuid macros. Use __get_cpu_features
to get cpu_features instead.
* sysdeps/x86/cacheinfo.c (is_intel): Removed.
(is_amd): Likewise.
(max_cpuid): Likewise.
(__cache_sysconf): Use __get_cpu_features to get cpu_features.
(init_cacheinfo): Likewise.
Since the maximum CPUID level of older Intel CPUs is 1, change
handle_intel to return -1, instead of assert, when the maximum
CPUID level is less than 2.
[BZ #20647]
* sysdeps/x86/cacheinfo.c (handle_intel): Return -1 if the
maximum CPUID level is less than 2.
For Intel processors, when there are both L2 and L3 caches, SMT level
type should be ued to count number of available logical processors
sharing L2 cache. If there is only L2 cache, core level type should
be used to count number of available logical processors sharing L2
cache. Number of available logical processors sharing L2 cache should
be used for non-inclusive L2 and L3 caches.
* sysdeps/x86/cacheinfo.c (init_cacheinfo): Count number of
available logical processors with SMT level type sharing L2
cache for Intel processors.
L2 cache is shared by 2 cores on Knights Landing, which has 4 threads
per core:
https://en.wikipedia.org/wiki/Xeon_Phi#Knights_Landing
So L2 cache is shared by 8 threads on Knights Landing as reported by
CPUID. We should remove special L2 cache case for Knights Landing.
[BZ #18185]
* sysdeps/x86/cacheinfo.c (init_cacheinfo): Don't limit threads
sharing L2 cache to 2 for Knights Landing.
Intel CPUID with EAX == 11 returns:
ECX Bits 07 - 00: Level number. Same value in ECX input.
Bits 15 - 08: Level type.
^^^^^^^^^^^^^^^^^^^^^^^^ This is level type.
Bits 31 - 16: Reserved.
Intel processor level type mask should be 0xff00, not 0xff0.
[BZ #20119]
* sysdeps/x86/cacheinfo.c (init_cacheinfo): Correct Intel
processor level type mask for CPUID with EAX == 11.
Skip counting logical threads for Intel processors if the HTT bit is 0
which indicates there is only a single logical processor.
* sysdeps/x86/cacheinfo.c (init_cacheinfo): Skip counting
logical threads if the HTT bit is 0.
* sysdeps/x86/cpu-features.h (bit_cpu_HTT): New.
(index_cpu_HTT): Likewise.
(reg_HTT): Likewise.
Move sysdeps/x86_64/cacheinfo.c to sysdeps/x86. No code changes on x86
and x86_64.
* sysdeps/i386/cacheinfo.c: Include <sysdeps/x86/cacheinfo.c>
instead of <sysdeps/x86_64/cacheinfo.c>.
* sysdeps/x86_64/cacheinfo.c: Moved to ...
* sysdeps/x86/cacheinfo.c: Here.
