panthema / 2013 / parallel-string-sorting / parallel-string-sorting-0.5 / minitbb / tbb / machine / gcc_ia32_common.h (Download File)
/*
    Copyright 2005-2012 Intel Corporation.  All Rights Reserved.

    This file is part of Threading Building Blocks.

    Threading Building Blocks is free software; you can redistribute it
    and/or modify it under the terms of the GNU General Public License
    version 2 as published by the Free Software Foundation.

    Threading Building Blocks is distributed in the hope that it will be
    useful, but WITHOUT ANY WARRANTY; without even the implied warranty
    of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
    GNU General Public License for more details.

    You should have received a copy of the GNU General Public License
    along with Threading Building Blocks; if not, write to the Free Software
    Foundation, Inc., 51 Franklin St, Fifth Floor, Boston, MA  02110-1301  USA

    As a special exception, you may use this file as part of a free software
    library without restriction.  Specifically, if other files instantiate
    templates or use macros or inline functions from this file, or you compile
    this file and link it with other files to produce an executable, this
    file does not by itself cause the resulting executable to be covered by
    the GNU General Public License.  This exception does not however
    invalidate any other reasons why the executable file might be covered by
    the GNU General Public License.
*/

#ifndef __TBB_machine_gcc_ia32_common_H
#define __TBB_machine_gcc_ia32_common_H

//TODO: Add a higher-level function, e.g. tbb::interal::log2(), into tbb_stddef.h, which
//uses __TBB_Log2 and contains the assert and remove the assert from here and all other
//platform-specific headers.
//TODO: Check if use of gcc intrinsic gives a better chance for cross call optimizations
static inline intptr_t __TBB_machine_lg( uintptr_t x ) {
    __TBB_ASSERT(x, "__TBB_Log2(0) undefined");
    uintptr_t j;
    __asm__ ("bsr %1,%0" : "=r"(j) : "r"(x));
    return j;
}
#define __TBB_Log2(V)  __TBB_machine_lg(V)

#ifndef __TBB_Pause
//TODO: check if raising a ratio of pause instructions to loop control instructions
//(via e.g. loop unrolling) gives any benefit for HT.  E.g, the current implementation
//does about 2 CPU-consuming instructions for every pause instruction.  Perhaps for
//high pause counts it should use an unrolled loop to raise the ratio, and thus free
//up more integer cycles for the other hyperthread.  On the other hand, if the loop is
//unrolled too far, it won't fit in the core's loop cache, and thus take away
//instruction decode slots from the other hyperthread.

//TODO: check if use of gcc __builtin_ia32_pause intrinsic gives a "some how" better performing code
static inline void __TBB_machine_pause( int32_t delay ) {
    for (int32_t i = 0; i < delay; i++) {
       __asm__ __volatile__("pause;");
    }
    return;
}
#define __TBB_Pause(V) __TBB_machine_pause(V)
#endif /* !__TBB_Pause */

// API to retrieve/update FPU control setting
#ifndef __TBB_CPU_CTL_ENV_PRESENT
#define __TBB_CPU_CTL_ENV_PRESENT 1

struct __TBB_cpu_ctl_env_t {
    int     mxcsr;
    short   x87cw;
};
inline void __TBB_get_cpu_ctl_env ( __TBB_cpu_ctl_env_t* ctl ) {
#if __TBB_ICC_12_0_INL_ASM_FSTCW_BROKEN
    __TBB_cpu_ctl_env_t loc_ctl;
    __asm__ __volatile__ (
            "stmxcsr %0\n\t"
            "fstcw %1"
            : "=m"(loc_ctl.mxcsr), "=m"(loc_ctl.x87cw)
    );
    *ctl = loc_ctl;
#else
    __asm__ __volatile__ (
            "stmxcsr %0\n\t"
            "fstcw %1"
            : "=m"(ctl->mxcsr), "=m"(ctl->x87cw)
    );
#endif
}
inline void __TBB_set_cpu_ctl_env ( const __TBB_cpu_ctl_env_t* ctl ) {
    __asm__ __volatile__ (
            "ldmxcsr %0\n\t"
            "fldcw %1"
            : : "m"(ctl->mxcsr), "m"(ctl->x87cw)
    );
}
#endif /* !__TBB_CPU_CTL_ENV_PRESENT */

#endif /* __TBB_machine_gcc_ia32_common_H */