Julien Vermillard
2010-Mar-25 15:51 UTC
[CELT-dev] Blackfin inline assembly for fixed math
Here some Blackfin inline assembly, mainly picked and adapted from speex. It's helps a little on my BF537 eval board. Julien -------- /** @file fixed_bfin.h @brief Fixed-point operations for the ADI BF5xx DSP family */ /* Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: - Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. - Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. - Neither the name of the Xiph.org Foundation nor the names of its contributors may be used to endorse or promote products derived from this software without specific prior written permission. THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS ``AS IS'' AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL THE FOUNDATION OR CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. */ #ifndef FIXED_BFIN_H #define FIXED_BFIN_H #undef MAX16 static inline celt_int16 MAX16(celt_int16 a, celt_int16 b) { celt_int32 res; __asm__ ( "%1 = %1.L (X);\n\t" "%2 = %2.L (X);\n\t" "%0 = MAX(%1,%2);" : "=d" (res) : "%d" (a), "d" (b) ); return res; } #undef MIN16 static inline celt_int16 MIN16(celt_int16 a, celt_int16 b) { celt_int32 res; __asm__ ( "%1 = %1.L (X);\n\t" "%2 = %2.L (X);\n\t" "%0 = MIN(%1,%2);" : "=d" (res) : "%d" (a), "d" (b) ); return res; } #undef MAX32 static inline celt_int32 MAX32(celt_int32 a, celt_int32 b) { celt_int32 res; __asm__ ( "%0 = MAX(%1,%2);" : "=d" (res) : "%d" (a), "d" (b) ); return res; } #undef MIN32 static inline celt_int32 MIN32(celt_int32 a, celt_int32 b) { celt_int32 res; __asm__ ( "%0 = MIN(%1,%2);" : "=d" (res) : "%d" (a), "d" (b) ); return res; } // ???? #undef MULT16_32_Q15 static inline celt_int32 MULT16_32_Q15(celt_int16 a, celt_int32 b) { celt_int32 res; __asm__ ( "A1 = %2.L*%1.L (M);\n\t" "A1 = A1 >>> 15;\n\t" "%0 = (A1 += %2.L*%1.H) ;\n\t" : "=&W" (res), "=&d" (b) : "d" (a), "1" (b) : "A1" ); return res; } #undef MULT16_32_Q14 static inline celt_int32 MULT16_32_Q14(celt_int16 a, celt_int16 b) { celt_int32 res; __asm__ ( "%2 <<= 1;\n\t" "A1 = %1.L*%2.L (M);\n\t" "A1 = A1 >>> 15;\n\t" "%0 = (A1 += %1.L*%2.H);\n\t" : "=W" (res), "=d" (a), "=d" (b) : "1" (a), "2" (b) : "A1" ); return res; } #undef MAC16_32_Q15 static inline celt_int32 MAC16_32_Q15(celt_int32 c, celt_int16 a, celt_int32 b) { celt_int32 res; __asm__ ( "A1 = %2.L*%1.L (M);\n\t" "A1 = A1 >>> 15;\n\t" "%0 = (A1 += %2.L*%1.H);\n\t" "%0 = %0 + %4;\n\t" : "=&W" (res), "=&d" (b) : "d" (a), "1" (b), "d" (c) : "A1" ); return res; } #undef PDIV32_16 static inline celt_int16 PDIV32_16(celt_int32 a, celt_int16 b) { celt_int32 res, bb; bb = b; a += b>>1; __asm__ ( "P0 = 15;\n\t" "R0 = %1;\n\t" "R1 = %2;\n\t" //"R0 = R0 + R1;\n\t" "R0 <<= 1;\n\t" "DIVS (R0, R1);\n\t" "LOOP divide%= LC0 = P0;\n\t" "LOOP_BEGIN divide%=;\n\t" "DIVQ (R0, R1);\n\t" "LOOP_END divide%=;\n\t" "R0 = R0.L;\n\t" "%0 = R0;\n\t" : "=m" (res) : "m" (a), "m" (bb) : "P0", "R0", "R1", "cc"); return res; } #undef DIV32_16 static inline celt_int16 DIV32_16(celt_int32 a, celt_int16 b) { celt_int32 res, bb; bb = b; // Make the roundinf consistent with the C version // (do we need to do that?) if (a<0) a += (b-1); __asm__ ( "P0 = 15;\n\t" "R0 = %1;\n\t" "R1 = %2;\n\t" "R0 <<= 1;\n\t" "DIVS (R0, R1);\n\t" "LOOP divide%= LC0 = P0;\n\t" "LOOP_BEGIN divide%=;\n\t" "DIVQ (R0, R1);\n\t" "LOOP_END divide%=;\n\t" "R0 = R0.L;\n\t" "%0 = R0;\n\t" : "=m" (res) : "m" (a), "m" (bb) : "P0", "R0", "R1", "cc"); return res; } #undef MULT16_16 static inline celt_int32 MULT16_16(celt_int16 a, celt_int16 b) { int res; __asm__ ( "%0 = %1.l*%2.l (is);\n\t" : "=W" (res) : "d" (a), "d" (b) : "A1"); return res; } #endif /* FIXED_BFIN_H */ -- Julien Vermillard Archean Technologies http://www.archean.fr -------------- next part -------------- A non-text attachment was scrubbed... Name: signature.asc Type: application/pgp-signature Size: 197 bytes Desc: not available Url : http://lists.xiph.org/pipermail/opus/attachments/20100325/71daaec9/attachment-0002.pgp
Possibly Parallel Threads
- [PATCH] Blackfin: cleanup astat/cc/hardware loop asm clobbers
- [PATCH] Eliminate the ec_int32 and ec_uint32 typedefs.
- Compile Speex for Blackfin in VisualDsp
- fixed point code
- [LLVMdev] why LoopUnswitch pass does not constant fold conditional branch and merge blocks