diff options
Diffstat (limited to 'libFDK/include/arm')
| -rw-r--r-- | libFDK/include/arm/clz_arm.h | 59 | ||||
| -rw-r--r-- | libFDK/include/arm/cplx_mul.h | 153 | ||||
| -rw-r--r-- | libFDK/include/arm/fixmadd_arm.h | 97 | ||||
| -rw-r--r-- | libFDK/include/arm/fixmul_arm.h | 79 | ||||
| -rw-r--r-- | libFDK/include/arm/scale.h | 90 | ||||
| -rw-r--r-- | libFDK/include/arm/scramble.h | 98 |
6 files changed, 576 insertions, 0 deletions
diff --git a/libFDK/include/arm/clz_arm.h b/libFDK/include/arm/clz_arm.h new file mode 100644 index 0000000..d3c3e04 --- /dev/null +++ b/libFDK/include/arm/clz_arm.h @@ -0,0 +1,59 @@ +/*************************** Fraunhofer IIS FDK Tools ********************** + + (C) Copyright Fraunhofer IIS (2006) + All Rights Reserved + + Please be advised that this software and/or program delivery is + Confidential Information of Fraunhofer and subject to and covered by the + + Fraunhofer IIS Software Evaluation Agreement + between Google Inc. and Fraunhofer + effective and in full force since March 1, 2012. + + You may use this software and/or program only under the terms and + conditions described in the above mentioned Fraunhofer IIS Software + Evaluation Agreement. Any other and/or further use requires a separate agreement. + + + $Id$ + Author(s): + Description: fixed point intrinsics + + This software and/or program is protected by copyright law and international + treaties. Any reproduction or distribution of this software and/or program, + or any portion of it, may result in severe civil and criminal penalties, and + will be prosecuted to the maximum extent possible under law. + +******************************************************************************/ +#if defined(__arm__) + +#if defined(__GNUC__) && defined(__ARM_ARCH_5TE__) /* cppp replaced: elif */ + /* ARM gcc*/ + + #define FUNCTION_fixnormz_D + #define FUNCTION_fixnorm_D + + inline INT fixnormz_D(LONG value) + { + INT result; + asm("clz %0, %1 ": "=r"(result) : "r"(value) ); + return result; + } + + inline INT fixnorm_D(LONG value) + { + INT result; + if (value < 0) { + value = ~value; + } + if (value == 0) { + return 0; + } + result = fixnormz_D(value); + return result - 1; + } + +#endif /* arm toolchain */ + +#endif /* __arm__ */ + diff --git a/libFDK/include/arm/cplx_mul.h b/libFDK/include/arm/cplx_mul.h new file mode 100644 index 0000000..dba2a9f --- /dev/null +++ b/libFDK/include/arm/cplx_mul.h @@ -0,0 +1,153 @@ +/*************************** Fraunhofer IIS FDK Tools ********************** + + (C) Copyright Fraunhofer IIS (2006) + All Rights Reserved + + Please be advised that this software and/or program delivery is + Confidential Information of Fraunhofer and subject to and covered by the + + Fraunhofer IIS Software Evaluation Agreement + between Google Inc. and Fraunhofer + effective and in full force since March 1, 2012. + + You may use this software and/or program only under the terms and + conditions described in the above mentioned Fraunhofer IIS Software + Evaluation Agreement. Any other and/or further use requires a separate agreement. + + + $Id$ + Author(s): + Description: fixed point intrinsics + + This software and/or program is protected by copyright law and international + treaties. Any reproduction or distribution of this software and/or program, + or any portion of it, may result in severe civil and criminal penalties, and + will be prosecuted to the maximum extent possible under law. + +******************************************************************************/ + + +#if defined(__arm__) && defined(__GNUC__) /* cppp replaced: elif */ + +#if defined(__ARM_ARCH_5TE__) || defined(__ARM_ARCH_6__) + #define FUNCTION_cplxMultDiv2_32x16 + #define FUNCTION_cplxMultDiv2_32x16X2 + //#define FUNCTION_cplxMult_32x16 + //#define FUNCTION_cplxMult_32x16X2 +#endif + +#define FUNCTION_cplxMultDiv2_32x32X2 +//#define FUNCTION_cplxMult_32x32X2 + +#ifdef FUNCTION_cplxMultDiv2_32x16 +inline void cplxMultDiv2( FIXP_DBL *c_Re, + FIXP_DBL *c_Im, + const FIXP_DBL a_Re, + const FIXP_DBL a_Im, + const FIXP_SPK wpk ) +{ + LONG tmp1,tmp2; + const LONG w = wpk.w; + + asm("smulwt %0, %3, %4;\n" + "rsb %1,%0,#0;\n" + "smlawb %0, %2, %4, %1;\n" + "smulwt %1, %2, %4;\n" + "smlawb %1, %3, %4, %1;\n" + : "=&r"(tmp1), "=&r"(tmp2) + : "r"(a_Re), "r"(a_Im), "r"(w) + ); + + *c_Re = tmp1; + *c_Im = tmp2; +} +#endif /* FUNCTION_cplxMultDiv2_32x16 */ + +#ifdef FUNCTION_cplxMultDiv2_32x16X2 +inline void cplxMultDiv2( FIXP_DBL *c_Re, + FIXP_DBL *c_Im, + const FIXP_DBL a_Re, + const FIXP_DBL a_Im, + const FIXP_SGL b_Re, + const FIXP_SGL b_Im) +{ + LONG tmp1, tmp2; + + asm("smulwb %0, %3, %5;\n" /* %7 = -a_Im * b_Im */ + "rsb %1,%0,#0;\n" + "smlawb %0, %2, %4, %1;\n" /* tmp1 = a_Re * b_Re - a_Im * b_Im */ + "smulwb %1, %2, %5;\n" /* %7 = a_Re * b_Im */ + "smlawb %1, %3, %4, %1;\n" /* tmp2 = a_Im * b_Re + a_Re * b_Im */ + : "=&r"(tmp1), "=&r"(tmp2) + : "r"(a_Re), "r"(a_Im), "r"(b_Re), "r"(b_Im) + ); + + *c_Re = tmp1; + *c_Im = tmp2; +} +#endif /* FUNCTION_cplxMultDiv2_32x16X2 */ + +#ifdef FUNCTION_cplxMultAddDiv2_32x16X2 +inline void cplxMultAddDiv2( FIXP_DBL *c_Re, + FIXP_DBL *c_Im, + const FIXP_DBL a_Re, + const FIXP_DBL a_Im, + const FIXP_SGL b_Re, + const FIXP_SGL b_Im) +{ + LONG tmp1, tmp2; + + asm("smulwb %0, %3, %5;\n" + "rsb %1,%0,#0;\n" + "smlawb %0, %2, %4, %1;\n" + "smulwb %1, %2, %5;\n" + "smlawb %1, %3, %4, %1;\n" + : "=&r"(tmp1), "=&r"(tmp2) + : "r"(a_Re), "r"(a_Im), "r"(b_Re), "r"(b_Im) + ); + + *c_Re += tmp1; + *c_Im += tmp2; +} +#endif /* FUNCTION_cplxMultAddDiv2_32x16X2 */ + + +#ifdef FUNCTION_cplxMultDiv2_32x32X2 +inline void cplxMultDiv2( FIXP_DBL *c_Re, + FIXP_DBL *c_Im, + const FIXP_DBL a_Re, + const FIXP_DBL a_Im, + const FIXP_DBL b_Re, + const FIXP_DBL b_Im) +{ + LONG tmp1, tmp2; + +#ifdef __ARM_ARCH_6__ + asm( + "smmul %0, %2, %4;\n" /* tmp1 = a_Re * b_Re */ + "smmls %0, %3, %5, %0;\n" /* tmp1 -= a_Im * b_Im */ + "smmul %1, %2, %5;\n" /* tmp2 = a_Re * b_Im */ + "smmla %1, %3, %4, %1;\n" /* tmp2 += a_Im * b_Re */ + : "=&r"(tmp1), "=&r"(tmp2) + : "r"(a_Re), "r"(a_Im), "r"(b_Re), "r"(b_Im) + : "r0" + ); +#else + LONG discard; + asm( + "smull %2, %0, %7, %6;\n" /* tmp1 = -a_Im * b_Im */ + "smlal %2, %0, %3, %5;\n" /* tmp1 += a_Re * b_Re */ + "smull %2, %1, %3, %6;\n" /* tmp2 = a_Re * b_Im */ + "smlal %2, %1, %4, %5;\n" /* tmp2 += a_Im * b_Re */ + : "=&r"(tmp1), "=&r"(tmp2), "=&r"(discard) + : "r"(a_Re), "r"(a_Im), "r"(b_Re), "r"(b_Im), "r"(-a_Im) + ); + #endif + *c_Re = tmp1; + *c_Im = tmp2; +} +#endif /* FUNCTION_cplxMultDiv2_32x32X2 */ + + +#endif + diff --git a/libFDK/include/arm/fixmadd_arm.h b/libFDK/include/arm/fixmadd_arm.h new file mode 100644 index 0000000..15308f7 --- /dev/null +++ b/libFDK/include/arm/fixmadd_arm.h @@ -0,0 +1,97 @@ +/*************************** Fraunhofer IIS FDK Tools ********************** + + (C) Copyright Fraunhofer IIS (2006) + All Rights Reserved + + Please be advised that this software and/or program delivery is + Confidential Information of Fraunhofer and subject to and covered by the + + Fraunhofer IIS Software Evaluation Agreement + between Google Inc. and Fraunhofer + effective and in full force since March 1, 2012. + + You may use this software and/or program only under the terms and + conditions described in the above mentioned Fraunhofer IIS Software + Evaluation Agreement. Any other and/or further use requires a separate agreement. + + + $Id$ + Author(s): + Description: fixed point intrinsics + + This software and/or program is protected by copyright law and international + treaties. Any reproduction or distribution of this software and/or program, + or any portion of it, may result in severe civil and criminal penalties, and + will be prosecuted to the maximum extent possible under law. + +******************************************************************************/ +#if defined(__arm__) + + /* ############################################################################# */ + #if defined(__GNUC__) && defined(__arm__) && !defined(__SYMBIAN32__) /* cppp replaced: elif */ + /* ############################################################################# */ + /* ARM GNU GCC */ + + #define FUNCTION_fixmadddiv2_DD + + #ifdef __ARM_ARCH_6__ + inline FIXP_DBL fixmadddiv2_DD (FIXP_DBL x, const FIXP_DBL a, const FIXP_DBL b) { + INT result; + asm ("smmla %0, %1, %2, %3;\n" + : "=r" (result) + : "r" (a), "r" (b), "r"(x) ); + return result ; + } + #define FUNCTION_fixmsubdiv2_DD + inline FIXP_DBL fixmsubdiv2_DD (FIXP_DBL x, const FIXP_DBL a, const FIXP_DBL b) { + INT result; + asm ("smmls %0, %1, %2, %3;\n" + : "=r" (result) + : "r" (a), "r" (b), "r"(x) ); + return result ; + } + #else /* __ARM_ARCH_6__ */ + inline FIXP_DBL fixmadddiv2_DD (FIXP_DBL x, const FIXP_DBL a, const FIXP_DBL b) { + INT discard, result = x; + asm ("smlal %0, %1, %2, %3;\n" + : "=r" (discard), "+r" (result) + : "r" (a), "r" (b) ); + return result ; + } + #endif /* __ARM_ARCH_6__ */ + + #if defined(__ARM_ARCH_5TE__) || defined(__ARM_ARCH_6__) + + #define FUNCTION_fixmadddiv2_DS + + inline FIXP_DBL fixmadddiv2_DS (FIXP_DBL x, const FIXP_DBL a, const FIXP_SGL b) { + INT result; + asm("smlawb %0, %1, %2, %3 " + : "=r" (result) + : "r" (a), "r" (b), "r" (x) ); + return result ; + } + + #endif /* defined(__ARM_ARCH_5TE__) || defined(__ARM_ARCH_6__) */ + + #define FUNCTION_fixmadddiv2BitExact_DD + #define fixmadddiv2BitExact_DD(a, b, c) fixmadddiv2_DD(a, b, c) + + #define FUNCTION_fixmsubdiv2BitExact_DD + inline FIXP_DBL fixmsubdiv2BitExact_DD (FIXP_DBL x, const FIXP_DBL a, const FIXP_DBL b) { + return x - fixmuldiv2BitExact_DD(a, b); + } + + #define FUNCTION_fixmadddiv2BitExact_DS + #define fixmadddiv2BitExact_DS(a, b, c) fixmadddiv2_DS(a, b, c) + + #define FUNCTION_fixmsubdiv2BitExact_DS + inline FIXP_DBL fixmsubdiv2BitExact_DS (FIXP_DBL x, const FIXP_DBL a, const FIXP_SGL b) { + return x - fixmuldiv2BitExact_DS(a, b); + } + /* ############################################################################# */ + #endif /* toolchain */ + /* ############################################################################# */ + +#endif /* __arm__ */ + diff --git a/libFDK/include/arm/fixmul_arm.h b/libFDK/include/arm/fixmul_arm.h new file mode 100644 index 0000000..8cc448d --- /dev/null +++ b/libFDK/include/arm/fixmul_arm.h @@ -0,0 +1,79 @@ +/*************************** Fraunhofer IIS FDK Tools ********************** + + (C) Copyright Fraunhofer IIS (2006) + All Rights Reserved + + Please be advised that this software and/or program delivery is + Confidential Information of Fraunhofer and subject to and covered by the + + Fraunhofer IIS Software Evaluation Agreement + between Google Inc. and Fraunhofer + effective and in full force since March 1, 2012. + + You may use this software and/or program only under the terms and + conditions described in the above mentioned Fraunhofer IIS Software + Evaluation Agreement. Any other and/or further use requires a separate agreement. + + + $Id$ + Author(s): + Description: fixed point intrinsics + + This software and/or program is protected by copyright law and international + treaties. Any reproduction or distribution of this software and/or program, + or any portion of it, may result in severe civil and criminal penalties, and + will be prosecuted to the maximum extent possible under law. + +******************************************************************************/ +#if defined(__arm__) + +#if defined(__GNUC__) && defined(__arm__) /* cppp replaced: elif */ +/* ARM with GNU compiler */ + +#define FUNCTION_fixmuldiv2_DD + +#define FUNCTION_fixmuldiv2BitExact_DD +#define fixmuldiv2BitExact_DD(a,b) fixmuldiv2_DD(a,b) +#define FUNCTION_fixmulBitExact_DD +#define fixmulBitExact_DD(a,b) fixmul_DD(a,b) + +#define FUNCTION_fixmuldiv2BitExact_DS +#define fixmuldiv2BitExact_DS(a,b) fixmuldiv2_DS(a,b) + +#define FUNCTION_fixmulBitExact_DS +#define fixmulBitExact_DS(a,b) fixmul_DS(a,b) + +#if defined(__ARM_ARCH_6__) || defined(__TARGET_ARCH_7E_M) +inline INT fixmuldiv2_DD (const INT a, const INT b) +{ + INT result ; + __asm__ ("smmul %0, %1, %2" : "=r" (result) + : "r" (a), "r" (b)) ; + return result ; +} +#else +inline INT fixmuldiv2_DD (const INT a, const INT b) +{ + INT discard, result ; + __asm__ ("smull %0, %1, %2, %3" : "=&r" (discard), "=r" (result) + : "r" (a), "r" (b)) ; + return result ; +} +#endif + +#if defined(__ARM_ARCH_5TE__) || defined(__ARM_ARCH_6__) +#define FUNCTION_fixmuldiv2_SD +inline INT fixmuldiv2_SD (const SHORT a, const INT b) +{ + INT result ; + __asm__ ("smulwb %0, %1, %2" + : "=r" (result) + : "r" (b), "r" (a)) ; + return result ; +} +#endif + +#endif /* defined(__GNUC__) && defined(__arm__) */ + +#endif /* __arm__ */ + diff --git a/libFDK/include/arm/scale.h b/libFDK/include/arm/scale.h new file mode 100644 index 0000000..11d18fb --- /dev/null +++ b/libFDK/include/arm/scale.h @@ -0,0 +1,90 @@ +/******************************** Fraunhofer IIS *************************** + + (C) Copyright Fraunhofer IIS (2006) + All Rights Reserved + + Please be advised that this software and/or program delivery is + Confidential Information of Fraunhofer and subject to and covered by the + + Fraunhofer IIS Software Evaluation Agreement + between Google Inc. and Fraunhofer + effective and in full force since March 1, 2012. + + You may use this software and/or program only under the terms and + conditions described in the above mentioned Fraunhofer IIS Software + Evaluation Agreement. Any other and/or further use requires a separate agreement. + + + $Id: + Author(s): + Description: ARM scaling operations + + This software and/or program is protected by copyright law and international + treaties. Any reproduction or distribution of this software and/or program, + or any portion of it, may result in severe civil and criminal penalties, and + will be prosecuted to the maximum extent possible under law. + +******************************************************************************/ + +#if defined(__GNUC__) /* GCC Compiler */ /* cppp replaced: elif */ + +#if defined(__ARM_ARCH_6__) + +inline static INT shiftRightSat(INT src, int scale) +{ + INT result; + asm( + "ssat %0,%2,%0;\n" + + : "=&r"(result) + : "r"(src>>scale), "M"(SAMPLE_BITS) + ); + + return result; +} + + #define SATURATE_INT_PCM_RIGHT_SHIFT(src, scale) shiftRightSat(src, scale) + +inline static INT shiftLeftSat(INT src, int scale) +{ + INT result; + asm( + "ssat %0,%2,%0;\n" + + : "=&r"(result) + : "r"(src<<scale), "M"(SAMPLE_BITS) + ); + + return result; +} + + #define SATURATE_INT_PCM_LEFT_SHIFT(src, scale) shiftLeftSat(src, scale) + +#endif /* __ARM_ARCH_6__ */ + +#endif /* compiler selection */ + +#define FUNCTION_scaleValueInPlace +inline +void scaleValueInPlace(FIXP_DBL *value, /*!< Value */ + INT scalefactor /*!< Scalefactor */ + ) +{ + INT newscale; + if ((newscale = scalefactor) >= 0) + *value <<= newscale; + else + *value >>= -newscale; +} + + + #define SATURATE_RIGHT_SHIFT(src, scale, dBits) \ + ( (((LONG)(src) ^ ((LONG)(src) >> (DFRACT_BITS-1)))>>(scale)) > (LONG)(((1U)<<((dBits)-1))-1)) \ + ? ((LONG)(src) >> (DFRACT_BITS-1)) ^ (LONG)(((1U)<<((dBits)-1))-1) \ + : ((LONG)(src) >> (scale)) + + #define SATURATE_LEFT_SHIFT(src, scale, dBits) \ + ( ((LONG)(src) ^ ((LONG)(src) >> (DFRACT_BITS-1))) > ((LONG)(((1U)<<((dBits)-1))-1) >> (scale)) ) \ + ? ((LONG)(src) >> (DFRACT_BITS-1)) ^ (LONG)(((1U)<<((dBits)-1))-1) \ + : ((LONG)(src) << (scale)) + diff --git a/libFDK/include/arm/scramble.h b/libFDK/include/arm/scramble.h new file mode 100644 index 0000000..9143155 --- /dev/null +++ b/libFDK/include/arm/scramble.h @@ -0,0 +1,98 @@ +/*************************** Fraunhofer IIS FDK Tools ********************** + + (C) Copyright Fraunhofer IIS (2005) + All Rights Reserved + + Please be advised that this software and/or program delivery is + Confidential Information of Fraunhofer and subject to and covered by the + + Fraunhofer IIS Software Evaluation Agreement + between Google Inc. and Fraunhofer + effective and in full force since March 1, 2012. + + You may use this software and/or program only under the terms and + conditions described in the above mentioned Fraunhofer IIS Software + Evaluation Agreement. Any other and/or further use requires a separate agreement. + + + $Id$ + Author(s): + Description: bitreversal of input data + + This software and/or program is protected by copyright law and international + treaties. Any reproduction or distribution of this software and/or program, + or any portion of it, may result in severe civil and criminal penalties, and + will be prosecuted to the maximum extent possible under law. + +******************************************************************************/ + + + +#if defined(FUNCTION_scramble) +#if defined(__GNUC__) /* cppp replaced: elif */ + +#define FUNCTION_scramble + +#if defined(__ARM_ARCH_5TE__) +#define USE_LDRD_STRD /* LDRD requires 8 byte data alignment. */ +#endif + +inline void scramble(FIXP_DBL x [], INT n) { + FDK_ASSERT(!(((INT)x)&(ALIGNMENT_DEFAULT-1))); + asm("mov r2, #1;\n" /* r2(m) = 1; */ + "sub r3, %1, #1;\n" /* r3 = n-1; */ + "mov r4, #0;\n" /* r4(j) = 0; */ + +"scramble_m_loop%=:\n" /* { */ + "mov r5, %1;\n" /* r5(k) = 1; */ + +"scramble_k_loop%=:\n" /* { */ + "mov r5, r5, lsr #1;\n" /* k >>= 1; */ + "eor r4, r4, r5;\n" /* j ^=k; */ + "ands r10, r4, r5;\n" /* r10 = r4 & r5; */ + "beq scramble_k_loop%=;\n" /* } while (r10 == 0); */ + + "cmp r4, r2;\n" /* if (r4 < r2) break; */ + "bcc scramble_m_loop_end%=;\n" + +#ifdef USE_LDRD_STRD + "mov r5, r2, lsl #3;\n" /* m(r5) = r2*4*2 */ + "ldrd r10, [%0, r5];\n" /* r10 = x[r5], x7 = x[r5+1] */ + "mov r6, r4, lsl #3;\n" /* j(r6) = r4*4*2 */ + "ldrd r8, [%0, r6];\n" /* r8 = x[r6], r9 = x[r6+1]; */ + "strd r10, [%0, r6];\n" /* x[r6,r6+1] = r10,r11; */ + "strd r8, [%0, r5];\n" /* x[r5,r5+1] = r8,r9; */ +#else + "mov r5, r2, lsl #3;\n" /* m(r5) = r2*4*2 */ + "ldr r10, [%0, r5];\n" + "mov r6, r4, lsl #3;\n" /* j(r6) = r4*4*2 */ + "ldr r11, [%0, r6];\n" + + "str r10, [%0, r6];\n" + "str r11, [%0, r5];\n" + + "add r5, r5, #4;" + "ldr r10, [%0, r5];\n" + "add r6, r6, #4;" + "ldr r11, [%0, r6];\n" + "str r10, [%0, r6];\n" + "str r11, [%0, r5];\n" +#endif +"scramble_m_loop_end%=:\n" + "add r2, r2, #1;\n" /* r2++; */ + "cmp r2, r3;\n" + "bcc scramble_m_loop%=;\n" /* } while (r2(m) < r3(n-1)); */ + : + : "r"(x), "r"(n) +#ifdef USE_LDRD_STRD + : "r2","r3", "r4","r5", "r10","r11", "r8","r9", "r6" ); +#else + : "r2","r3", "r4","r5", "r10","r11", "r6" ); +#endif +} +#else +/* Force C implementation if no assembler version available. */ +#undef FUNCTION_scramble +#endif /* Toolchain selection. */ + +#endif /* defined(FUNCTION_scramble) */ |