Basic PowerPC Built-in Functions Available on all Configurations (Using the GNU Compiler Collection (GCC))

6.60.22.1 Basic PowerPC Built-in Functions Available on all Configurations

Built-in Function: void __builtin_cpu_init (void): This function is a nop on the PowerPC platform and is included solely to maintain API compatibility with the x86 builtins.

Built-in Function: int __builtin_cpu_is (const char *cpuname)

This function returns a value of 1 if the run-time CPU is of type cpuname and returns 0 otherwise

The __builtin_cpu_is function requires GLIBC 2.23 or newer which exports the hardware capability bits. GCC defines the macro __BUILTIN_CPU_SUPPORTS__ if the __builtin_cpu_supports built-in function is fully supported.

If GCC was configured to use a GLIBC before 2.23, the built-in function __builtin_cpu_is always returns a 0 and the compiler issues a warning.

The following CPU names can be detected:

‘power10’: IBM POWER10 Server CPU.
‘power9’: IBM POWER9 Server CPU.
‘power8’: IBM POWER8 Server CPU.
‘power7’: IBM POWER7 Server CPU.
‘power6x’: IBM POWER6 Server CPU (RAW mode).
‘power6’: IBM POWER6 Server CPU (Architected mode).
‘power5+’: IBM POWER5+ Server CPU.
‘power5’: IBM POWER5 Server CPU.
‘ppc970’: IBM 970 Server CPU (ie, Apple G5).
‘power4’: IBM POWER4 Server CPU.
‘ppca2’: IBM A2 64-bit Embedded CPU
‘ppc476’: IBM PowerPC 476FP 32-bit Embedded CPU.
‘ppc464’: IBM PowerPC 464 32-bit Embedded CPU.
‘ppc440’: PowerPC 440 32-bit Embedded CPU.
‘ppc405’: PowerPC 405 32-bit Embedded CPU.
‘ppc-cell-be’: IBM PowerPC Cell Broadband Engine Architecture CPU.

Here is an example:

#ifdef __BUILTIN_CPU_SUPPORTS__
  if (__builtin_cpu_is ("power8"))
    {
       do_power8 (); // POWER8 specific implementation.
    }
  else
#endif
    {
       do_generic (); // Generic implementation.
    }

Built-in Function: int __builtin_cpu_supports (const char *feature)

This function returns a value of 1 if the run-time CPU supports the HWCAP feature feature and returns 0 otherwise.

The __builtin_cpu_supports function requires GLIBC 2.23 or newer which exports the hardware capability bits. GCC defines the macro __BUILTIN_CPU_SUPPORTS__ if the __builtin_cpu_supports built-in function is fully supported.

If GCC was configured to use a GLIBC before 2.23, the built-in function __builtin_cpu_suports always returns a 0 and the compiler issues a warning.

The following features can be detected:

‘4xxmac’: 4xx CPU has a Multiply Accumulator.
‘altivec’: CPU has a SIMD/Vector Unit.
‘arch_2_05’: CPU supports ISA 2.05 (eg, POWER6)
‘arch_2_06’: CPU supports ISA 2.06 (eg, POWER7)
‘arch_2_07’: CPU supports ISA 2.07 (eg, POWER8)
‘arch_3_00’: CPU supports ISA 3.0 (eg, POWER9)
‘arch_3_1’: CPU supports ISA 3.1 (eg, POWER10)
‘archpmu’: CPU supports the set of compatible performance monitoring events.
‘booke’: CPU supports the Embedded ISA category.
‘cellbe’: CPU has a CELL broadband engine.
‘darn’: CPU supports the darn (deliver a random number) instruction.
‘dfp’: CPU has a decimal floating point unit.
‘dscr’: CPU supports the data stream control register.
‘ebb’: CPU supports event base branching.
‘efpdouble’: CPU has a SPE double precision floating point unit.
‘efpsingle’: CPU has a SPE single precision floating point unit.
‘fpu’: CPU has a floating point unit.
‘htm’: CPU has hardware transaction memory instructions.
‘htm-nosc’: Kernel aborts hardware transactions when a syscall is made.
‘htm-no-suspend’: CPU supports hardware transaction memory but does not support the tsuspend. instruction.
‘ic_snoop’: CPU supports icache snooping capabilities.
‘ieee128’: CPU supports 128-bit IEEE binary floating point instructions.
‘isel’: CPU supports the integer select instruction.
‘mma’: CPU supports the matrix-multiply assist instructions.
‘mmu’: CPU has a memory management unit.
‘notb’: CPU does not have a timebase (eg, 601 and 403gx).
‘pa6t’: CPU supports the PA Semi 6T CORE ISA.
‘power4’: CPU supports ISA 2.00 (eg, POWER4)
‘power5’: CPU supports ISA 2.02 (eg, POWER5)
‘power5+’: CPU supports ISA 2.03 (eg, POWER5+)
‘power6x’: CPU supports ISA 2.05 (eg, POWER6) extended opcodes mffgpr and mftgpr.
‘ppc32’: CPU supports 32-bit mode execution.
‘ppc601’: CPU supports the old POWER ISA (eg, 601)
‘ppc64’: CPU supports 64-bit mode execution.
‘ppcle’: CPU supports a little-endian mode that uses address swizzling.
‘scv’: Kernel supports system call vectored.
‘smt’: CPU support simultaneous multi-threading.
‘spe’: CPU has a signal processing extension unit.
‘tar’: CPU supports the target address register.
‘true_le’: CPU supports true little-endian mode.
‘ucache’: CPU has unified I/D cache.
‘vcrypto’: CPU supports the vector cryptography instructions.
‘vsx’: CPU supports the vector-scalar extension.

Here is an example:

#ifdef __BUILTIN_CPU_SUPPORTS__
  if (__builtin_cpu_supports ("fpu"))
    {
       asm("fadd %0,%1,%2" : "=d"(dst) : "d"(src1), "d"(src2));
    }
  else
#endif
    {
       dst = __fadd (src1, src2); // Software FP addition function.
    }

The following built-in functions are also available on all PowerPC processors:

uint64_t __builtin_ppc_get_timebase ();
unsigned long __builtin_ppc_mftb ();
double __builtin_unpack_ibm128 (__ibm128, int);
__ibm128 __builtin_pack_ibm128 (double, double);
double __builtin_mffs (void);
void __builtin_mtfsf (const int, double);
void __builtin_mtfsb0 (const int);
void __builtin_mtfsb1 (const int);
void __builtin_set_fpscr_rn (int);

The __builtin_ppc_get_timebase and __builtin_ppc_mftb functions generate instructions to read the Time Base Register. The __builtin_ppc_get_timebase function may generate multiple instructions and always returns the 64 bits of the Time Base Register. The __builtin_ppc_mftb function always generates one instruction and returns the Time Base Register value as an unsigned long, throwing away the most significant word on 32-bit environments. The __builtin_mffs return the value of the FPSCR register. Note, ISA 3.0 supports the __builtin_mffsl() which permits software to read the control and non-sticky status bits in the FSPCR without the higher latency associated with accessing the sticky status bits. The __builtin_mtfsf takes a constant 8-bit integer field mask and a double precision floating point argument and generates the mtfsf (extended mnemonic) instruction to write new values to selected fields of the FPSCR. The __builtin_mtfsb0 and __builtin_mtfsb1 take the bit to change as an argument. The valid bit range is between 0 and 31. The builtins map to the mtfsb0 and mtfsb1 instructions which take the argument and add 32. Hence these instructions only modify the FPSCR[32:63] bits by changing the specified bit to a zero or one respectively. The __builtin_set_fpscr_rn builtin allows changing both of the floating point rounding mode bits. The argument is a 2-bit value. The argument can either be a const int or stored in a variable. The builtin uses the ISA 3.0 instruction mffscrn if available, otherwise it reads the FPSCR, masks the current rounding mode bits out and OR’s in the new value.