Chris@69: /* Copyright (C) 2007-2009 Xiph.Org Foundation
Chris@69:    Copyright (C) 2003-2008 Jean-Marc Valin
Chris@69:    Copyright (C) 2007-2008 CSIRO
Chris@69:    Copyright (C) 2013      Parrot */
Chris@69: /*
Chris@69:    Redistribution and use in source and binary forms, with or without
Chris@69:    modification, are permitted provided that the following conditions
Chris@69:    are met:
Chris@69: 
Chris@69:    - Redistributions of source code must retain the above copyright
Chris@69:    notice, this list of conditions and the following disclaimer.
Chris@69: 
Chris@69:    - Redistributions in binary form must reproduce the above copyright
Chris@69:    notice, this list of conditions and the following disclaimer in the
Chris@69:    documentation and/or other materials provided with the distribution.
Chris@69: 
Chris@69:    THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS
Chris@69:    ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT
Chris@69:    LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR
Chris@69:    A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER
Chris@69:    OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL,
Chris@69:    EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
Chris@69:    PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR
Chris@69:    PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
Chris@69:    LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING
Chris@69:    NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS
Chris@69:    SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
Chris@69: */
Chris@69: 
Chris@69: #ifndef FIXED_ARMv5E_H
Chris@69: #define FIXED_ARMv5E_H
Chris@69: 
Chris@69: #include "fixed_armv4.h"
Chris@69: 
Chris@69: /** 16x32 multiplication, followed by a 16-bit shift right. Results fits in 32 bits */
Chris@69: #undef MULT16_32_Q16
Chris@69: static OPUS_INLINE opus_val32 MULT16_32_Q16_armv5e(opus_val16 a, opus_val32 b)
Chris@69: {
Chris@69:   int res;
Chris@69:   __asm__(
Chris@69:       "#MULT16_32_Q16\n\t"
Chris@69:       "smulwb %0, %1, %2\n\t"
Chris@69:       : "=r"(res)
Chris@69:       : "r"(b),"r"(a)
Chris@69:   );
Chris@69:   return res;
Chris@69: }
Chris@69: #define MULT16_32_Q16(a, b) (MULT16_32_Q16_armv5e(a, b))
Chris@69: 
Chris@69: 
Chris@69: /** 16x32 multiplication, followed by a 15-bit shift right. Results fits in 32 bits */
Chris@69: #undef MULT16_32_Q15
Chris@69: static OPUS_INLINE opus_val32 MULT16_32_Q15_armv5e(opus_val16 a, opus_val32 b)
Chris@69: {
Chris@69:   int res;
Chris@69:   __asm__(
Chris@69:       "#MULT16_32_Q15\n\t"
Chris@69:       "smulwb %0, %1, %2\n\t"
Chris@69:       : "=r"(res)
Chris@69:       : "r"(b), "r"(a)
Chris@69:   );
Chris@69:   return SHL32(res,1);
Chris@69: }
Chris@69: #define MULT16_32_Q15(a, b) (MULT16_32_Q15_armv5e(a, b))
Chris@69: 
Chris@69: 
Chris@69: /** 16x32 multiply, followed by a 15-bit shift right and 32-bit add.
Chris@69:     b must fit in 31 bits.
Chris@69:     Result fits in 32 bits. */
Chris@69: #undef MAC16_32_Q15
Chris@69: static OPUS_INLINE opus_val32 MAC16_32_Q15_armv5e(opus_val32 c, opus_val16 a,
Chris@69:  opus_val32 b)
Chris@69: {
Chris@69:   int res;
Chris@69:   __asm__(
Chris@69:       "#MAC16_32_Q15\n\t"
Chris@69:       "smlawb %0, %1, %2, %3;\n"
Chris@69:       : "=r"(res)
Chris@69:       : "r"(SHL32(b,1)), "r"(a), "r"(c)
Chris@69:   );
Chris@69:   return res;
Chris@69: }
Chris@69: #define MAC16_32_Q15(c, a, b) (MAC16_32_Q15_armv5e(c, a, b))
Chris@69: 
Chris@69: /** 16x32 multiply, followed by a 16-bit shift right and 32-bit add.
Chris@69:     Result fits in 32 bits. */
Chris@69: #undef MAC16_32_Q16
Chris@69: static OPUS_INLINE opus_val32 MAC16_32_Q16_armv5e(opus_val32 c, opus_val16 a,
Chris@69:  opus_val32 b)
Chris@69: {
Chris@69:   int res;
Chris@69:   __asm__(
Chris@69:       "#MAC16_32_Q16\n\t"
Chris@69:       "smlawb %0, %1, %2, %3;\n"
Chris@69:       : "=r"(res)
Chris@69:       : "r"(b), "r"(a), "r"(c)
Chris@69:   );
Chris@69:   return res;
Chris@69: }
Chris@69: #define MAC16_32_Q16(c, a, b) (MAC16_32_Q16_armv5e(c, a, b))
Chris@69: 
Chris@69: /** 16x16 multiply-add where the result fits in 32 bits */
Chris@69: #undef MAC16_16
Chris@69: static OPUS_INLINE opus_val32 MAC16_16_armv5e(opus_val32 c, opus_val16 a,
Chris@69:  opus_val16 b)
Chris@69: {
Chris@69:   int res;
Chris@69:   __asm__(
Chris@69:       "#MAC16_16\n\t"
Chris@69:       "smlabb %0, %1, %2, %3;\n"
Chris@69:       : "=r"(res)
Chris@69:       : "r"(a), "r"(b), "r"(c)
Chris@69:   );
Chris@69:   return res;
Chris@69: }
Chris@69: #define MAC16_16(c, a, b) (MAC16_16_armv5e(c, a, b))
Chris@69: 
Chris@69: /** 16x16 multiplication where the result fits in 32 bits */
Chris@69: #undef MULT16_16
Chris@69: static OPUS_INLINE opus_val32 MULT16_16_armv5e(opus_val16 a, opus_val16 b)
Chris@69: {
Chris@69:   int res;
Chris@69:   __asm__(
Chris@69:       "#MULT16_16\n\t"
Chris@69:       "smulbb %0, %1, %2;\n"
Chris@69:       : "=r"(res)
Chris@69:       : "r"(a), "r"(b)
Chris@69:   );
Chris@69:   return res;
Chris@69: }
Chris@69: #define MULT16_16(a, b) (MULT16_16_armv5e(a, b))
Chris@69: 
Chris@69: #ifdef OPUS_ARM_INLINE_MEDIA
Chris@69: 
Chris@69: #undef SIG2WORD16
Chris@69: static OPUS_INLINE opus_val16 SIG2WORD16_armv6(opus_val32 x)
Chris@69: {
Chris@69:    celt_sig res;
Chris@69:    __asm__(
Chris@69:        "#SIG2WORD16\n\t"
Chris@69:        "ssat %0, #16, %1, ASR #12\n\t"
Chris@69:        : "=r"(res)
Chris@69:        : "r"(x+2048)
Chris@69:    );
Chris@69:    return EXTRACT16(res);
Chris@69: }
Chris@69: #define SIG2WORD16(x) (SIG2WORD16_armv6(x))
Chris@69: 
Chris@69: #endif /* OPUS_ARM_INLINE_MEDIA */
Chris@69: 
Chris@69: #endif