Starboard Module Reference: cpu_features.h

  • The content describes the SbCPUFeatures struct and its nested structs (SbCPUFeaturesARM, SbCPUFeaturesMIPS, SbCPUFeaturesX86), which hold detailed information about CPU architecture, features, and capabilities.

  • SbCPUFeatures contains fields like processor brand, cache line sizes, hardware capability bitmasks, and architecture-specific details.

  • Each architecture-specific struct (SbCPUFeaturesARM, SbCPUFeaturesMIPS, SbCPUFeaturesX86) has members that identify the processor's manufacturer, model, and supported instruction sets, such as NEON, AES, AVX, and SSE.

  • The SbCPUFeaturesGet function retrieves CPU features and populates an SbCPUFeatures struct, returning false if the CPU architecture is unknown and the struct is invalid.

Structs

SbCPUFeatures

Members

  • SbCPUFeaturesArchitecture architecture

    Architecture of the processor.

  • const char * brand

    Processor brand string retrievable by CPUID or from /proc/cpuinfo under the key "model name" or "Processor".

  • int32_t icache_line_size

    Processor cache line size in bytes of Level 1 instruction cache and data cache. Queried by sysconf(_SC_LEVEL1_ICACHE_LINESIZE) and sysconf(_SC_LEVEL1_DCACHE_LINESIZE), or from files /proc/cpuinfo, /proc/self/auxv, or CPUID with CLFLUSH instruction.

  • int32_t dcache_line_size

  • bool has_fpu

    Processor has floating-point unit on-chip.

  • uint32_t hwcap

    Bit-mask containing processor features flags. Queried by getauxval(AT_HWCAP) if it is supported.

  • uint32_t hwcap2

    Similar to hwcap. Queried by getauxval(AT_HWCAP2) if it is supported.

  • SbCPUFeaturesARM arm

    Processor features specific to each architecture. Set the appropriate SbCPUFeatures<ARCH_NAME> for the underlying architecture. Set the other SbCPUFeatures<OTHER_ARCH> to be invalid, meaning:

    • '-1' for signed integer fields

    • 'false' for feature flag fields

    • '0' for feature flag bitmasks

    • empty string for string fields

  • SbCPUFeaturesMIPS mips_arch

    The reason that the "_arch" suffix exists for mips is because on some platforms that use MIPS as the underlying architecture, "mips" is already defined as a macro.

  • SbCPUFeaturesX86 x86

SbCPUFeaturesARM

Members

  • int16_t implementer

    Processor implementer/implementor code. ARM is 0x41, NVIDIA is 0x4e, etc.

  • int16_t variant

    Processor variant number, indicating the major revision number.

  • int16_t revision

    Processor revision number, indicating the minor revision number.

  • int16_t architecture_generation

    Processor architecture generation number, indicating the generations (ARMv6-M, ARMv7, etc) within an architecture family. This field is called "Architecture" or "Constant" in the processor ID register.

  • int16_t part

    Processor part number, indicating Cortex-M0, Cortex-A8, etc.

  • bool has_neon

    ARM Advanced SIMD (NEON) vector instruction set extension.

  • bool has_thumb2

    Thumb-2 mode.

  • bool has_vfp

    VFP (SIMD vector floating point instructions).

  • bool has_vfp3

    VFP version 3

  • bool has_vfp4

    VFP version 4

  • bool has_vfp3_d32

    VFP version 3 with 32 D-registers.

  • bool has_idiva

    SDIV and UDIV hardware division in ARM mode.

  • bool has_aes

    Arm 64 feature flags

    AES instructions.

  • bool has_crc32

    CRC32 instructions.

  • bool has_sha1

    SHA-1 instructions.

  • bool has_sha2

    SHA-256 instructions.

  • bool has_pmull

    64-bit PMULL and PMULL2 instructions.

SbCPUFeaturesMIPS

Members

  • bool has_msa

    MIPS SIMD Architecture (MSA).

SbCPUFeaturesX86

Members

  • const char * vendor

    Processor vendor ID string, e.g. "GenuineIntel", "AuthenticAMD", etc

  • int16_t family

    Processor family ID

  • int16_t ext_family

    Processor extended family ID, needs to be examined only when the family ID is 0FH.

  • int16_t model

    Processor model ID

  • int16_t ext_model

    Processor extended model ID, needs to be examined only when the family ID is 06H or 0FH.

  • int16_t stepping

    Processor stepping ID, a product revision number

  • int16_t type

    Processor type ID

  • int32_t signature

    A raw form of collection of processor stepping, model, and family information

  • bool has_cmov

    Conditional Move Instructions (plus FCMOVcc, FCOMI with FPU).

  • bool has_mmx

    Multimedia extensions.

  • bool has_sse

    SSE (Streaming SIMD Extensions).

  • bool has_sse2

    SSE2 extensions.

  • bool has_tsc

    Time Stamp Counter.

  • bool has_sse3

    SSE3 extensions.

  • bool has_pclmulqdq

    PCLMULQDQ instruction.

  • bool has_ssse3

    Supplemental SSE3 extensions.

  • bool has_sse41

    SSE-4.1 extensions.

  • bool has_sse42

    SSE-4.2 extensions.

  • bool has_movbe

    MOVBE instruction.

  • bool has_popcnt

    POPCNT instruction.

  • bool has_osxsave

    XSAVE/XRSTOR/XGETBV/XSETBV instruction enabled in this OS.

  • bool has_avx

    AVX (Advanced Vector Extensions).

  • bool has_f16c

    16-bit FP conversions.

  • bool has_fma3

    Fused multiply-add.

  • bool has_aesni

    AES new instructions (AES-NI).

  • bool has_avx2

    AVX2 Extensions.

  • bool has_avx512f

    AVX-512 Foundation.

  • bool has_avx512dq

    AVX-512 DQ (Double/Quad granular) Instructions.

  • bool has_avx512ifma

    AVX-512 Integer Fused Multiply-Add instructions.

  • bool has_avx512pf

    AVX-512 Prefetch.

  • bool has_avx512er

    AVX-512 Exponential and Reciprocal.

  • bool has_avx512cd

    AVX-512 Conflict Detection.

  • bool has_avx512bw

    AVX-512 BW (Byte/Word granular) Instructions.

  • bool has_avx512vl

    AVX-512 VL (128/256 Vector Length) Extensions.

  • bool has_bmi1

    First group of bit manipulation extensions.

  • bool has_bmi2

    Second group of bit manipulation extensions.

  • bool has_lzcnt

    Bit manipulation instruction LZCNT.

  • bool has_sahf

    SAHF in long mode.

Functions

SbCPUFeaturesGet

Retrieve the underlying CPU features and place it in features, which must not be NULL.

If this function returns false, it means the CPU architecture is unknown and all fields in features are invalid.

Declaration

bool SbCPUFeaturesGet(SbCPUFeatures *features)