mirror/glibc
2
0
Fork 0
mirror of git://sourceware.org/git/glibc.git synced 2024-12-09 04:11:27 +08:00
Code Issues Packages Projects Releases Wiki Activity
6c57d32048
glibc/manual/platform.texi
H.J. Lu ff6d62e9ed <sys/platform/x86.h>: Remove the C preprocessor magic 
In <sys/platform/x86.h>, define CPU features as enum instead of using
the C preprocessor magic to make it easier to wrap this functionality
in other languages.  Move the C preprocessor magic to internal header
for better GCC codegen when more than one features are checked in a
single expression as in x86-64 dl-hwcaps-subdirs.c.

1. Rename COMMON_CPUID_INDEX_XXX to CPUID_INDEX_XXX.
2. Move CPUID_INDEX_MAX to sysdeps/x86/include/cpu-features.h.
3. Remove struct cpu_features and __x86_get_cpu_features from
<sys/platform/x86.h>.
4. Add __x86_get_cpuid_feature_leaf to <sys/platform/x86.h> and put it
in libc.
5. Make __get_cpu_features() private to glibc.
6. Replace __x86_get_cpu_features(N) with __get_cpu_features().
7. Add _dl_x86_get_cpu_features to GLIBC_PRIVATE.
8. Use a single enum index for each CPU feature detection.
9. Pass the CPUID feature leaf to __x86_get_cpuid_feature_leaf.
10. Return zero struct cpuid_feature for the older glibc binary with a
smaller CPUID_INDEX_MAX [BZ #27104].
11. Inside glibc, use the C preprocessor magic so that cpu_features data
can be loaded just once leading to more compact code for glibc.

256 bits are used for each CPUID leaf.  Some leaves only contain a few
features.  We can add exceptions to such leaves.  But it will increase
code sizes and it is harder to provide backward/forward compatibilities
when new features are added to such leaves in the future.

When new leaves are added, _rtld_global_ro offsets will change which
leads to race condition during in-place updates. We may avoid in-place
updates by

1. Rename the old glibc.
2. Install the new glibc.
3. Remove the old glibc.

NB: A function, __x86_get_cpuid_feature_leaf , is used to avoid the copy
relocation issue with IFUNC resolver as shown in IFUNC resolver tests.
2021-01-21 05:58:17 -08:00

684 lines
16 KiB
Plaintext
Raw Blame History

@node Platform, Contributors, Maintenance, Top
@c %MENU% Describe all platform-specific facilities provided
@appendix Platform-specific facilities
@Theglibc{} can provide machine-specific functionality.
@menu
* PowerPC:: Facilities Specific to the PowerPC Architecture
* RISC-V:: Facilities Specific to the RISC-V Architecture
* X86:: Facilities Specific to the X86 Architecture
@end menu
@node PowerPC
@appendixsec PowerPC-specific Facilities
Facilities specific to PowerPC that are not specific to a particular
operating system are declared in @file{sys/platform/ppc.h}.
@deftypefun {uint64_t} __ppc_get_timebase (void)
@safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}}
Read the current value of the Time Base Register.
The @dfn{Time Base Register} is a 64-bit register that stores a monotonically
incremented value updated at a system-dependent frequency that may be
different from the processor frequency. More information is available in
@cite{Power ISA 2.06b - Book II - Section 5.2}.
@code{__ppc_get_timebase} uses the processor's time base facility directly
without requiring assistance from the operating system, so it is very
efficient.
@end deftypefun
@deftypefun {uint64_t} __ppc_get_timebase_freq (void)
@safety{@prelim{}@mtunsafe{@mtuinit{}}@asunsafe{@asucorrupt{:init}}@acunsafe{@acucorrupt{:init}}}
@c __ppc_get_timebase_freq=__get_timebase_freq @mtuinit @acsfd
@c __get_clockfreq @mtuinit @asucorrupt:init @acucorrupt:init @acsfd
@c the initialization of the static timebase_freq is not exactly
@c safe, because hp_timing_t cannot be atomically set up.
@c syscall:get_tbfreq ok
@c open dup @acsfd
@c read dup ok
@c memcpy dup ok
@c memmem dup ok
@c close dup @acsfd
Read the current frequency at which the Time Base Register is updated.
This frequency is not related to the processor clock or the bus clock.
It is also possible that this frequency is not constant. More information is
available in @cite{Power ISA 2.06b - Book II - Section 5.2}.
@end deftypefun
The following functions provide hints about the usage of resources that are
shared with other processors. They can be used, for example, if a program
waiting on a lock intends to divert the shared resources to be used by other
processors. More information is available in @cite{Power ISA 2.06b - Book II -
Section 3.2}.
@deftypefun {void} __ppc_yield (void)
@safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}}
Provide a hint that performance will probably be improved if shared resources
dedicated to the executing processor are released for use by other processors.
@end deftypefun
@deftypefun {void} __ppc_mdoio (void)
@safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}}
Provide a hint that performance will probably be improved if shared resources
dedicated to the executing processor are released until all outstanding storage
accesses to caching-inhibited storage have been completed.
@end deftypefun
@deftypefun {void} __ppc_mdoom (void)
@safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}}
Provide a hint that performance will probably be improved if shared resources
dedicated to the executing processor are released until all outstanding storage
accesses to cacheable storage for which the data is not in the cache have been
completed.
@end deftypefun
@deftypefun {void} __ppc_set_ppr_med (void)
@safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}}
Set the Program Priority Register to medium value (default).
The @dfn{Program Priority Register} (PPR) is a 64-bit register that controls
the program's priority. By adjusting the PPR value the programmer may
improve system throughput by causing the system resources to be used
more efficiently, especially in contention situations.
The three unprivileged states available are covered by the functions
@code{__ppc_set_ppr_med} (medium -- default), @code{__ppc_set_ppc_low} (low)
and @code{__ppc_set_ppc_med_low} (medium low). More information
available in @cite{Power ISA 2.06b - Book II - Section 3.1}.
@end deftypefun
@deftypefun {void} __ppc_set_ppr_low (void)
@safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}}
Set the Program Priority Register to low value.
@end deftypefun
@deftypefun {void} __ppc_set_ppr_med_low (void)
@safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}}
Set the Program Priority Register to medium low value.
@end deftypefun
Power ISA 2.07 extends the priorities that can be set to the Program Priority
Register (PPR). The following functions implement the new priority levels:
very low and medium high.
@deftypefun {void} __ppc_set_ppr_very_low (void)
@safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}}
Set the Program Priority Register to very low value.
@end deftypefun
@deftypefun {void} __ppc_set_ppr_med_high (void)
@safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}}
Set the Program Priority Register to medium high value. The medium high
priority is privileged and may only be set during certain time intervals by
problem-state programs. If the program priority is medium high when the time
interval expires or if an attempt is made to set the priority to medium high
when it is not allowed, the priority is set to medium.
@end deftypefun
@node RISC-V
@appendixsec RISC-V-specific Facilities
Cache management facilities specific to RISC-V systems that implement the Linux
ABI are declared in @file{sys/cachectl.h}.
@deftypefun {void} __riscv_flush_icache (void *@var{start}, void *@var{end}, unsigned long int @var{flags})
@safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}}
Enforce ordering between stores and instruction cache fetches. The range of
addresses over which ordering is enforced is specified by @var{start} and
@var{end}. The @var{flags} argument controls the extent of this ordering, with
the default behavior (a @var{flags} value of 0) being to enforce the fence on
all threads in the current process. Setting the
@code{SYS_RISCV_FLUSH_ICACHE_LOCAL} bit allows users to indicate that enforcing
ordering on only the current thread is necessary. All other flag bits are
reserved.
@end deftypefun
@node X86
@appendixsec X86-specific Facilities
Facilities specific to X86 that are not specific to a particular
operating system are declared in @file{sys/platform/x86.h}.
@deftypefun {const struct cpuid_feature *} __x86_get_cpuid_feature_leaf (unsigned int @var{leaf})
@safety{@prelim{}@mtsafe{}@assafe{}@acsafe{}}
Return a pointer to x86 CPU feature structure used by query macros for x86
CPU feature @var{leaf}.
@end deftypefun
@deftypefn Macro int HAS_CPU_FEATURE (@var{name})
This macro returns a nonzero value (true) if the processor has the feature
@var{name}.
@end deftypefn
@deftypefn Macro int CPU_FEATURE_USABLE (@var{name})
This macro returns a nonzero value (true) if the processor has the feature
@var{name} and the feature is supported by the operating system.
@end deftypefn
The supported processor features are:
@itemize @bullet
@item
@code{ACPI} -- Thermal Monitor and Software Controlled Clock Facilities.
@item
@code{ADX} -- ADX instruction extensions.
@item
@code{APIC} -- APIC On-Chip.
@item
@code{AES} -- The AES instruction extensions.
@item
@code{AESKLE} -- AES Key Locker instructions are enabled by OS.
@item
@code{AMX_BF16} -- Tile computational operations on bfloat16 numbers.
@item
@code{AMX_INT8} -- Tile computational operations on 8-bit numbers.
@item
@code{AMX_TILE} -- Tile architecture.
@item
@code{ARCH_CAPABILITIES} -- IA32_ARCH_CAPABILITIES MSR.
@item
@code{AVX} -- The AVX instruction extensions.
@item
@code{AVX2} -- The AVX2 instruction extensions.
@item
@code{AVX_VNNI} -- The AVX-VNNI instruction extensions.
@item
@code{AVX512_4FMAPS} -- The AVX512_4FMAPS instruction extensions.
@item
@code{AVX512_4VNNIW} -- The AVX512_4VNNIW instruction extensions.
@item
@code{AVX512_BF16} -- The AVX512_BF16 instruction extensions.
@item
@code{AVX512_BITALG} -- The AVX512_BITALG instruction extensions.
@item
@code{AVX512_FP16} -- The AVX512_FP16 instruction extensions.
@item
@code{AVX512_IFMA} -- The AVX512_IFMA instruction extensions.
@item
@code{AVX512_VBMI} -- The AVX512_VBMI instruction extensions.
@item
@code{AVX512_VBMI2} -- The AVX512_VBMI2 instruction extensions.
@item
@code{AVX512_VNNI} -- The AVX512_VNNI instruction extensions.
@item
@code{AVX512_VP2INTERSECT} -- The AVX512_VP2INTERSECT instruction
extensions.
@item
@code{AVX512_VPOPCNTDQ} -- The AVX512_VPOPCNTDQ instruction extensions.
@item
@code{AVX512BW} -- The AVX512BW instruction extensions.
@item
@code{AVX512CD} -- The AVX512CD instruction extensions.
@item
@code{AVX512ER} -- The AVX512ER instruction extensions.
@item
@code{AVX512DQ} -- The AVX512DQ instruction extensions.
@item
@code{AVX512F} -- The AVX512F instruction extensions.
@item
@code{AVX512PF} -- The AVX512PF instruction extensions.
@item
@code{AVX512VL} -- The AVX512VL instruction extensions.
@item
@code{BMI1} -- BMI1 instructions.
@item
@code{BMI2} -- BMI2 instructions.
@item
@code{CLDEMOTE} -- CLDEMOTE instruction.
@item
@code{CLFLUSHOPT} -- CLFLUSHOPT instruction.
@item
@code{CLFSH} -- CLFLUSH instruction.
@item
@code{CLWB} -- CLWB instruction.
@item
@code{CMOV} -- Conditional Move instructions.
@item
@code{CMPXCHG16B} -- CMPXCHG16B instruction.
@item
@code{CNXT_ID} -- L1 Context ID.
@item
@code{CORE_CAPABILITIES} -- IA32_CORE_CAPABILITIES MSR.
@item
@code{CX8} -- CMPXCHG8B instruction.
@item
@code{DCA} -- Data prefetch from a memory mapped device.
@item
@code{DE} -- Debugging Extensions.
@item
@code{DEPR_FPU_CS_DS} -- Deprecates FPU CS and FPU DS values.
@item
@code{DS} -- Debug Store.
@item
@code{DS_CPL} -- CPL Qualified Debug Store.
@item
@code{DTES64} -- 64-bit DS Area.
@item
@code{EIST} -- Enhanced Intel SpeedStep technology.
@item
@code{ENQCMD} -- Enqueue Stores instructions.
@item
@code{ERMS} -- Enhanced REP MOVSB/STOSB.
@item
@code{F16C} -- 16-bit floating-point conversion instructions.
@item
@code{FMA} -- FMA extensions using YMM state.
@item
@code{FMA4} -- FMA4 instruction extensions.
@item
@code{FPU} -- X87 Floating Point Unit On-Chip.
@item
@code{FSGSBASE} -- RDFSBASE/RDGSBASE/WRFSBASE/WRGSBASE instructions.
@item
@code{FSRCS} -- Fast Short REP CMP and SCA.
@item
@code{FSRM} -- Fast Short REP MOV.
@item
@code{FSRS} -- Fast Short REP STO.
@item
@code{FXSR} -- FXSAVE and FXRSTOR instructions.
@item
@code{FZLRM} -- Fast Zero-Length REP MOV.
@item
@code{GFNI} -- GFNI instruction extensions.
@item
@code{HLE} -- HLE instruction extensions.
@item
@code{HTT} -- Max APIC IDs reserved field is Valid.
@item
@code{HRESET} -- History reset.
@item
@code{HYBRID} -- Hybrid processor.
@item
@code{IBRS_IBPB} -- Indirect branch restricted speculation (IBRS) and
the indirect branch predictor barrier (IBPB).
@item
@code{IBT} -- Intel Indirect Branch Tracking instruction extensions.
@item
@code{INVARIANT_TSC} -- Invariant TSC.
@item
@code{INVPCID} -- INVPCID instruction.
@item
@code{KL} -- AES Key Locker instructions.
@item
@code{LAM} -- Linear Address Masking.
@item
@code{L1D_FLUSH} -- IA32_FLUSH_CMD MSR.
@item
@code{LAHF64_SAHF64} -- LAHF/SAHF available in 64-bit mode.
@item
@code{LM} -- Long mode.
@item
@code{LWP} -- Lightweight profiling.
@item
@code{LZCNT} -- LZCNT instruction.
@item
@code{MCA} -- Machine Check Architecture.
@item
@code{MCE} -- Machine Check Exception.
@item
@code{MD_CLEAR} -- MD_CLEAR.
@item
@code{MMX} -- Intel MMX Technology.
@item
@code{MONITOR} -- MONITOR/MWAIT instructions.
@item
@code{MOVBE} -- MOVBE instruction.
@item
@code{MOVDIRI} -- MOVDIRI instruction.
@item
@code{MOVDIR64B} -- MOVDIR64B instruction.
@item
@code{MPX} -- Intel Memory Protection Extensions.
@item
@code{MSR} -- Model Specific Registers RDMSR and WRMSR instructions.
@item
@code{MTRR} -- Memory Type Range Registers.
@item
@code{NX} -- No-execute page protection.
@item
@code{OSPKE} -- OS has set CR4.PKE to enable protection keys.
@item
@code{OSXSAVE} -- The OS has set CR4.OSXSAVE[bit 18] to enable
XSETBV/XGETBV instructions to access XCR0 and to support processor
extended state management using XSAVE/XRSTOR.
@item
@code{PAE} -- Physical Address Extension.
@item
@code{PAGE1GB} -- 1-GByte page.
@item
@code{PAT} -- Page Attribute Table.
@item
@code{PBE} -- Pending Break Enable.
@item
@code{PCID} -- Process-context identifiers.
@item
@code{PCLMULQDQ} -- PCLMULQDQ instruction.
@item
@code{PCONFIG} -- PCONFIG instruction.
@item
@code{PDCM} -- Perfmon and Debug Capability.
@item
@code{PGE} -- Page Global Bit.
@item
@code{PKS} -- Protection keys for supervisor-mode pages.
@item
@code{PKU} -- Protection keys for user-mode pages.
@item
@code{POPCNT} -- POPCNT instruction.
@item
@code{PREFETCHW} -- PREFETCHW instruction.
@item
@code{PREFETCHWT1} -- PREFETCHWT1 instruction.
@item
@code{PSE} -- Page Size Extension.
@item
@code{PSE_36} -- 36-Bit Page Size Extension.
@item
@code{PSN} -- Processor Serial Number.
@item
@code{RDPID} -- RDPID instruction.
@item
@code{RDRAND} -- RDRAND instruction.
@item
@code{RDSEED} -- RDSEED instruction.
@item
@code{RDT_A} -- Intel Resource Director Technology (Intel RDT) Allocation
capability.
@item
@code{RDT_M} -- Intel Resource Director Technology (Intel RDT) Monitoring
capability.
@item
@code{RDTSCP} -- RDTSCP instruction.
@item
@code{RTM} -- RTM instruction extensions.
@item
@code{SDBG} -- IA32_DEBUG_INTERFACE MSR for silicon debug.
@item
@code{SEP} -- SYSENTER and SYSEXIT instructions.
@item
@code{SERIALIZE} -- SERIALIZE instruction.
@item
@code{SGX} -- Intel Software Guard Extensions.
@item
@code{SGX_LC} -- SGX Launch Configuration.
@item
@code{SHA} -- SHA instruction extensions.
@item
@code{SHSTK} -- Intel Shadow Stack instruction extensions.
@item
@code{SMAP} -- Supervisor-Mode Access Prevention.
@item
@code{SMEP} -- Supervisor-Mode Execution Prevention.
@item
@code{SMX} -- Safer Mode Extensions.
@item
@code{SS} -- Self Snoop.
@item
@code{SSBD} -- Speculative Store Bypass Disable (SSBD).
@item
@code{SSE} -- Streaming SIMD Extensions.
@item
@code{SSE2} -- Streaming SIMD Extensions 2.
@item
@code{SSE3} -- Streaming SIMD Extensions 3.
@item
@code{SSE4_1} -- Streaming SIMD Extensions 4.1.
@item
@code{SSE4_2} -- Streaming SIMD Extensions 4.2.
@item
@code{SSE4A} -- SSE4A instruction extensions.
@item
@code{SSSE3} -- Supplemental Streaming SIMD Extensions 3.
@item
@code{STIBP} -- Single thread indirect branch predictors (STIBP).
@item
@code{SVM} -- Secure Virtual Machine.
@item
@code{SYSCALL_SYSRET} -- SYSCALL/SYSRET instructions.
@item
@code{TBM} -- Trailing bit manipulation instructions.
@item
@code{TM} -- Thermal Monitor.
@item
@code{TM2} -- Thermal Monitor 2.
@item
@code{TRACE} -- Intel Processor Trace.
@item
@code{TSC} -- Time Stamp Counter. RDTSC instruction.
@item
@code{TSC_ADJUST} -- IA32_TSC_ADJUST MSR.
@item
@code{TSC_DEADLINE} -- Local APIC timer supports one-shot operation
using a TSC deadline value.
@item
@code{TSXLDTRK} -- TSXLDTRK instructions.
@item
@code{UINTR} -- User interrupts.
@item
@code{UMIP} -- User-mode instruction prevention.
@item
@code{VAES} -- VAES instruction extensions.
@item
@code{VME} -- Virtual 8086 Mode Enhancements.
@item
@code{VMX} -- Virtual Machine Extensions.
@item
@code{VPCLMULQDQ} -- VPCLMULQDQ instruction.
@item
@code{WAITPKG} -- WAITPKG instruction extensions.
@item
@code{WBNOINVD} -- WBINVD/WBNOINVD instructions.
@item
@code{WIDE_KL} -- AES wide Key Locker instructions.
@item
@code{X2APIC} -- x2APIC.
@item
@code{XFD} -- Extended Feature Disable (XFD).
@item
@code{XGETBV_ECX_1} -- XGETBV with ECX = 1.
@item
@code{XOP} -- XOP instruction extensions.
@item
@code{XSAVE} -- The XSAVE/XRSTOR processor extended states feature, the
XSETBV/XGETBV instructions, and XCR0.
@item
@code{XSAVEC} -- XSAVEC instruction.
@item
@code{XSAVEOPT} -- XSAVEOPT instruction.
@item
@code{XSAVES} -- XSAVES/XRSTORS instructions.
@item
@code{XTPRUPDCTRL} -- xTPR Update Control.
@end itemize
You could query if a processor supports @code{AVX} with:
@smallexample
#include <sys/platform/x86.h>
int
support_avx (void)
@{
return HAS_CPU_FEATURE (AVX);
@}
@end smallexample
and if @code{AVX} is usable with:
@smallexample
#include <sys/platform/x86.h>
int
usable_avx (void)
@{
return CPU_FEATURE_USABLE (AVX);
@}
@end smallexample
Reference in New Issue View Git Blame Copy Permalink