Home | History | Annotate | Download | only in x86
      1 /* Copyright (c) 2014, Cisco Systems, INC
      2    Written by XiangMingZhu WeiZhou MinPeng YanWang
      3 
      4    Redistribution and use in source and binary forms, with or without
      5    modification, are permitted provided that the following conditions
      6    are met:
      7 
      8    - Redistributions of source code must retain the above copyright
      9    notice, this list of conditions and the following disclaimer.
     10 
     11    - Redistributions in binary form must reproduce the above copyright
     12    notice, this list of conditions and the following disclaimer in the
     13    documentation and/or other materials provided with the distribution.
     14 
     15    THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
     16    ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
     17    LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
     18    A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER
     19    OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
     20    EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
     21    PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
     22    PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
     23    LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
     24    NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
     25    SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
     26 */
     27 
     28 #if !defined(X86CPU_H)
     29 # define X86CPU_H
     30 
     31 # if defined(OPUS_X86_MAY_HAVE_SSE)
     32 #  define MAY_HAVE_SSE(name) name ## _sse
     33 # else
     34 #  define MAY_HAVE_SSE(name) name ## _c
     35 # endif
     36 
     37 # if defined(OPUS_X86_MAY_HAVE_SSE2)
     38 #  define MAY_HAVE_SSE2(name) name ## _sse2
     39 # else
     40 #  define MAY_HAVE_SSE2(name) name ## _c
     41 # endif
     42 
     43 # if defined(OPUS_X86_MAY_HAVE_SSE4_1)
     44 #  define MAY_HAVE_SSE4_1(name) name ## _sse4_1
     45 # else
     46 #  define MAY_HAVE_SSE4_1(name) name ## _c
     47 # endif
     48 
     49 # if defined(OPUS_X86_MAY_HAVE_AVX)
     50 #  define MAY_HAVE_AVX(name) name ## _avx
     51 # else
     52 #  define MAY_HAVE_AVX(name) name ## _c
     53 # endif
     54 
     55 # if defined(OPUS_HAVE_RTCD)
     56 int opus_select_arch(void);
     57 # endif
     58 
     59 /*gcc appears to emit MOVDQA's to load the argument of an _mm_cvtepi8_epi32()
     60   or _mm_cvtepi16_epi32() when optimizations are disabled, even though the
     61   actual PMOVSXWD instruction takes an m32 or m64. Unlike a normal memory
     62   reference, these require 16-byte alignment and load a full 16 bytes (instead
     63   of 4 or 8), possibly reading out of bounds.
     64 
     65   We can insert an explicit MOVD or MOVQ using _mm_cvtsi32_si128() or
     66   _mm_loadl_epi64(), which should have the same semantics as an m32 or m64
     67   reference in the PMOVSXWD instruction itself, but gcc is not smart enough to
     68   optimize this out when optimizations ARE enabled.
     69 
     70   Clang, in contrast, requires us to do this always for _mm_cvtepi8_epi32
     71   (which is fair, since technically the compiler is always allowed to do the
     72   dereference before invoking the function implementing the intrinsic).
     73   However, it is smart enough to eliminate the extra MOVD instruction.
     74   For _mm_cvtepi16_epi32, it does the right thing, though does *not* optimize out
     75   the extra MOVQ if it's specified explicitly */
     76 
     77 # if defined(__clang__) || !defined(__OPTIMIZE__)
     78 #  define OP_CVTEPI8_EPI32_M32(x) \
     79  (_mm_cvtepi8_epi32(_mm_cvtsi32_si128(*(int *)(x))))
     80 # else
     81 #  define OP_CVTEPI8_EPI32_M32(x) \
     82  (_mm_cvtepi8_epi32(*(__m128i *)(x)))
     83 #endif
     84 
     85 /* similar reasoning about the instruction sequence as in the 32-bit macro above,
     86  */
     87 # if defined(__clang__) || !defined(__OPTIMIZE__)
     88 #  define OP_CVTEPI16_EPI32_M64(x) \
     89  (_mm_cvtepi16_epi32(_mm_loadl_epi64((__m128i *)(x))))
     90 # else
     91 #  define OP_CVTEPI16_EPI32_M64(x) \
     92  (_mm_cvtepi16_epi32(*(__m128i *)(x)))
     93 # endif
     94 
     95 #endif
     96