FreeCalypso > hg > gsm-codec-lib
diff libgsmefr/basic_op.h @ 347:1c514150c033
libgsmefr: first big perf opt: inline most basic_op.h functions
| author | Mychaela Falconia <falcon@freecalypso.org> |
|---|---|
| date | Sun, 21 Apr 2024 23:17:21 +0000 |
| parents | 38326102fc43 |
| children | ed8762eea8a1 |
line wrap: on
line diff
--- a/libgsmefr/basic_op.h Fri Apr 19 09:07:39 2024 +0000 +++ b/libgsmefr/basic_op.h Sun Apr 21 23:17:21 2024 +0000 @@ -1,62 +1,1196 @@ -/*___________________________________________________________________________ - | | - | Constants and Globals | - |___________________________________________________________________________| -*/ +/* + * This version of basic_op.h for GSM-EFR codec has been created at + * Themyscira Wireless with the aim of performance improvement. + * The original version contained only function declarations, + * resulting in a function call for every elementary operation; + * the present version implements most of these operations as inline + * functions. + * + * Elimination of Overflow flag: examination of EFR code reveals that + * both Carry and Overflow flags are accessed outside of basicop2.c + * only in g_pitch.c, as part of a code sequence that uses L_macNs. + * Since the Overflow flag is not checked anywhere else, we can eliminate + * its setting from all other elementary operation functions, saving + * some cycles and code size. + * + * Non-portability: we assume real target machines, not hypothetical + * ones. Specifically, we assume that >> does an arithmetic right shift + * (sign bit fill). + */ + +/* for the few functions that still use them */ extern __thread Flag Overflow; extern __thread Flag Carry; -#define MAX_32 (Word32)0x7fffffffL -#define MIN_32 (Word32)0x80000000L +#define MAX_32 (Word32)0x7fffffff +#define MIN_32 (Word32)0x80000000 #define MAX_16 (Word16)0x7fff #define MIN_16 (Word16)0x8000 +/* likely and unlikely conditional paths */ + +#define likely(x) __builtin_expect(!!(x), 1) +#define unlikely(x) __builtin_expect(!!(x), 0) + +/* Word16 operations */ + +/*___________________________________________________________________________ + | | + | Function Name : add | + | | + | Purpose : | + | | + | Performs the addition (var1+var2) with overflow control and saturation;| + | the 16 bit result is set at +32767 when overflow occurs or at -32768 | + | when underflow occurs. | + | | + | Complexity weight : 1 | + | | + | Inputs : | + | | + | var1 | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var1 <= 0x0000 7fff. | + | | + | var2 | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var1 <= 0x0000 7fff. | + | | + | Outputs : | + | | + | none | + | | + | Return Value : | + | | + | var_out | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var_out <= 0x0000 7fff. | + |___________________________________________________________________________| +*/ + +static inline Word16 add (Word16 var1, Word16 var2) +{ + Word32 L_sum; + + L_sum = (Word32) var1 + var2; + if (unlikely(L_sum > MAX_16)) + return MAX_16; + else if (unlikely(L_sum < MIN_16)) + return MIN_16; + else + return L_sum; +} + +/*___________________________________________________________________________ + | | + | Function Name : sub | + | | + | Purpose : | + | | + | Performs the subtraction (var1+var2) with overflow control and satu- | + | ration; the 16 bit result is set at +32767 when overflow occurs or at | + | -32768 when underflow occurs. | + | | + | Complexity weight : 1 | + | | + | Inputs : | + | | + | var1 | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var1 <= 0x0000 7fff. | + | | + | var2 | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var1 <= 0x0000 7fff. | + | | + | Outputs : | + | | + | none | + | | + | Return Value : | + | | + | var_out | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var_out <= 0x0000 7fff. | + |___________________________________________________________________________| +*/ + +static inline Word16 sub (Word16 var1, Word16 var2) +{ + Word32 L_diff; + + L_diff = (Word32) var1 - var2; + if (unlikely(L_diff > MAX_16)) + return MAX_16; + else if (unlikely(L_diff < MIN_16)) + return MIN_16; + else + return L_diff; +} + +/*___________________________________________________________________________ + | | + | Function Name : abs_s | + | | + | Purpose : | + | | + | Absolute value of var1; abs_s(-32768) = 32767. | + | | + | Complexity weight : 1 | + | | + | Inputs : | + | | + | var1 | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var1 <= 0x0000 7fff. | + | | + | Outputs : | + | | + | none | + | | + | Return Value : | + | | + | var_out | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0x0000 0000 <= var_out <= 0x0000 7fff. | + |___________________________________________________________________________| +*/ + +static inline Word16 abs_s (Word16 var1) +{ + Word16 var_out; + + if (unlikely(var1 == (Word16) 0X8000)) + { + var_out = MAX_16; + } + else + { + if (var1 < 0) + { + var_out = -var1; + } + else + { + var_out = var1; + } + } + return (var_out); +} + +/*___________________________________________________________________________ + | | + | Function Name : shl | + | | + | Purpose : | + | | + | Arithmetically shift the 16 bit input var1 left var2 positions.Zero fill| + | the var2 LSB of the result. If var2 is negative, arithmetically shift | + | var1 right by -var2 with sign extension. Saturate the result in case of | + | underflows or overflows. | + | | + | Complexity weight : 1 | + | | + | Inputs : | + | | + | var1 | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var1 <= 0x0000 7fff. | + | | + | var2 | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var1 <= 0x0000 7fff. | + | | + | Outputs : | + | | + | none | + | | + | Return Value : | + | | + | var_out | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var_out <= 0x0000 7fff. | + |___________________________________________________________________________| +*/ + +static inline Word16 shl (Word16 var1, Word16 var2) +{ + Word32 result; + + if (var2 < 0) { + var2 = -var2; + if (unlikely(var2 >= 15)) + return (var1 < 0) ? -1 : 0; + else + return var1 >> var2; + } else { + result = (Word32) var1 << var2; + if (var1 > 0) { + if (unlikely((var2 >= 15) || (result & 0xFFFF8000))) + return MAX_16; + else + return (Word16) result; + } else if (var1 < 0) { + if (unlikely((var2 >= 16) || ((result & 0xFFFF8000) != 0xFFFF8000))) + return MIN_16; + else + return (Word16) result; + } else + return 0; + } +} + /*___________________________________________________________________________ | | - | Prototypes for basic arithmetic operators | + | Function Name : shr | + | | + | Purpose : | + | | + | Arithmetically shift the 16 bit input var1 right var2 positions with | + | sign extension. If var2 is negative, arithmetically shift var1 left by | + | -var2 with sign extension. Saturate the result in case of underflows or | + | overflows. | + | | + | Complexity weight : 1 | + | | + | Inputs : | + | | + | var1 | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var1 <= 0x0000 7fff. | + | | + | var2 | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var1 <= 0x0000 7fff. | + | | + | Outputs : | + | | + | none | + | | + | Return Value : | + | | + | var_out | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var_out <= 0x0000 7fff. | + |___________________________________________________________________________| +*/ + +static inline Word16 shr (Word16 var1, Word16 var2) +{ + Word32 result; + + if (var2 < 0) { + var2 = -var2; + result = (Word32) var1 << var2; + if (var1 > 0) { + if (unlikely((var2 >= 15) || (result & 0xFFFF8000))) + return MAX_16; + else + return (Word16) result; + } else if (var1 < 0) { + if (unlikely((var2 >= 16) || ((result & 0xFFFF8000) != 0xFFFF8000))) + return MIN_16; + else + return (Word16) result; + } else + return 0; + } else { + if (unlikely(var2 >= 15)) + return (var1 < 0) ? -1 : 0; + else + return var1 >> var2; + } +} + +/*___________________________________________________________________________ + | | + | Function Name : mult | + | | + | Purpose : | + | | + | Performs the multiplication of var1 by var2 and gives a 16 bit result | + | which is scaled i.e.: | + | mult(var1,var2) = extract_l(L_shr((var1 times var2),15)) and | + | mult(-32768,-32768) = 32767. | + | | + | Complexity weight : 1 | + | | + | Inputs : | + | | + | var1 | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var1 <= 0x0000 7fff. | + | | + | var2 | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var1 <= 0x0000 7fff. | + | | + | Outputs : | + | | + | none | + | | + | Return Value : | + | | + | var_out | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var_out <= 0x0000 7fff. | + |___________________________________________________________________________| +*/ + +static inline Word16 mult (Word16 var1, Word16 var2) +{ + if (unlikely(var1 == MIN_16 && var2 == MIN_16)) + return MAX_16; + else + return ((Word32) var1 * var2) >> 15; +} + +/*___________________________________________________________________________ + | | + | Function Name : negate | + | | + | Purpose : | + | | + | Negate var1 with saturation, saturate in the case where input is -32768:| + | negate(var1) = sub(0,var1). | + | | + | Complexity weight : 1 | + | | + | Inputs : | + | | + | var1 | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var1 <= 0x0000 7fff. | + | | + | Outputs : | + | | + | none | + | | + | Return Value : | + | | + | var_out | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var_out <= 0x0000 7fff. | + |___________________________________________________________________________| +*/ + +static inline Word16 negate (Word16 var1) +{ + Word16 var_out; + + var_out = (var1 == MIN_16) ? MAX_16 : -var1; + return (var_out); +} + +/*___________________________________________________________________________ + | | + | Function Name : extract_h | + | | + | Purpose : | + | | + | Return the 16 MSB of L_var1. | + | | + | Complexity weight : 1 | + | | + | Inputs : | + | | + | L_var1 | + | 32 bit long signed integer (Word32 ) whose value falls in the | + | range : 0x8000 0000 <= L_var1 <= 0x7fff ffff. | + | | + | Outputs : | + | | + | none | + | | + | Return Value : | + | | + | var_out | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var_out <= 0x0000 7fff. | + |___________________________________________________________________________| +*/ + +static inline Word16 extract_h (Word32 L_var1) +{ + Word16 var_out; + + var_out = (Word16) (L_var1 >> 16); + return (var_out); +} + +/*___________________________________________________________________________ + | | + | Function Name : extract_l | + | | + | Purpose : | + | | + | Return the 16 LSB of L_var1. | + | | + | Complexity weight : 1 | + | | + | Inputs : | + | | + | L_var1 | + | 32 bit long signed integer (Word32 ) whose value falls in the | + | range : 0x8000 0000 <= L_var1 <= 0x7fff ffff. | + | | + | Outputs : | + | | + | none | + | | + | Return Value : | + | | + | var_out | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var_out <= 0x0000 7fff. | + |___________________________________________________________________________| +*/ + +static inline Word16 extract_l (Word32 L_var1) +{ + Word16 var_out; + + var_out = (Word16) L_var1; + return (var_out); +} + +/*___________________________________________________________________________ + | | + | Function Name : round | + | | + | Purpose : | + | | + | Round the lower 16 bits of the 32 bit input number into the MS 16 bits | + | with saturation. Shift the resulting bits right by 16 and return the 16 | + | bit number: | + | round(L_var1) = extract_h(L_add(L_var1,32768)) | + | | + | Complexity weight : 1 | + | | + | Inputs : | + | | + | L_var1 | + | 32 bit long signed integer (Word32 ) whose value falls in the | + | range : 0x8000 0000 <= L_var1 <= 0x7fff ffff. | + | | + | Outputs : | + | | + | none | + | | + | Return Value : | + | | + | var_out | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var_out <= 0x0000 7fff. | + |___________________________________________________________________________| +*/ + +static inline Word16 round (Word32 L_var1) +{ + if (unlikely(L_var1 >= 0x7fff8000)) + return MAX_16; + else + return (L_var1 + 0x8000) >> 16; +} + +/* Word32 operations */ + +/*___________________________________________________________________________ + | | + | Function Name : L_add | + | | + | Purpose : | + | | + | 32 bits addition of the two 32 bits variables (L_var1+L_var2) with | + | overflow control and saturation; the result is set at +2147483647 when | + | overflow occurs or at -2147483648 when underflow occurs. | + | | + | Complexity weight : 2 | + | | + | Inputs : | + | | + | L_var1 32 bit long signed integer (Word32) whose value falls in the | + | range : 0x8000 0000 <= L_var3 <= 0x7fff ffff. | + | | + | L_var2 32 bit long signed integer (Word32) whose value falls in the | + | range : 0x8000 0000 <= L_var3 <= 0x7fff ffff. | + | | + | Outputs : | + | | + | none | + | | + | Return Value : | + | | + | L_var_out | + | 32 bit long signed integer (Word32) whose value falls in the | + | range : 0x8000 0000 <= L_var_out <= 0x7fff ffff. | + |___________________________________________________________________________| +*/ + +static inline Word32 L_add (Word32 L_var1, Word32 L_var2) +{ + Word32 L_var_out; + + L_var_out = L_var1 + L_var2; + + if (((L_var1 ^ L_var2) & MIN_32) == 0) + { + if ((L_var_out ^ L_var1) & MIN_32) + { + L_var_out = (L_var1 < 0) ? MIN_32 : MAX_32; + } + } + return (L_var_out); +} + +/*___________________________________________________________________________ + | | + | Function Name : L_sub | + | | + | Purpose : | + | | + | 32 bits subtraction of the two 32 bits variables (L_var1-L_var2) with | + | overflow control and saturation; the result is set at +2147483647 when | + | overflow occurs or at -2147483648 when underflow occurs. | + | | + | Complexity weight : 2 | + | | + | Inputs : | + | | + | L_var1 32 bit long signed integer (Word32) whose value falls in the | + | range : 0x8000 0000 <= L_var3 <= 0x7fff ffff. | + | | + | L_var2 32 bit long signed integer (Word32) whose value falls in the | + | range : 0x8000 0000 <= L_var3 <= 0x7fff ffff. | + | | + | Outputs : | + | | + | none | + | | + | Return Value : | + | | + | L_var_out | + | 32 bit long signed integer (Word32) whose value falls in the | + | range : 0x8000 0000 <= L_var_out <= 0x7fff ffff. | |___________________________________________________________________________| */ -Word16 add (Word16 var1, Word16 var2); /* Short add, 1 */ -Word16 sub (Word16 var1, Word16 var2); /* Short sub, 1 */ -Word16 abs_s (Word16 var1); /* Short abs, 1 */ -Word16 shl (Word16 var1, Word16 var2); /* Short shift left, 1 */ -Word16 shr (Word16 var1, Word16 var2); /* Short shift right, 1 */ -Word16 mult (Word16 var1, Word16 var2); /* Short mult, 1 */ -Word32 L_mult (Word16 var1, Word16 var2); /* Long mult, 1 */ -Word16 negate (Word16 var1); /* Short negate, 1 */ -Word16 extract_h (Word32 L_var1); /* Extract high, 1 */ -Word16 extract_l (Word32 L_var1); /* Extract low, 1 */ -Word16 round (Word32 L_var1); /* Round, 1 */ -Word32 L_mac (Word32 L_var3, Word16 var1, Word16 var2); /* Mac, 1 */ -Word32 L_msu (Word32 L_var3, Word16 var1, Word16 var2); /* Msu, 1 */ +static inline Word32 L_sub (Word32 L_var1, Word32 L_var2) +{ + Word32 L_var_out; + + L_var_out = L_var1 - L_var2; + + if (((L_var1 ^ L_var2) & MIN_32) != 0) + { + if ((L_var_out ^ L_var1) & MIN_32) + { + L_var_out = (L_var1 < 0) ? MIN_32 : MAX_32; + } + } + return (L_var_out); +} + +/*___________________________________________________________________________ + | | + | Function Name : L_mult | + | | + | Purpose : | + | | + | L_mult is the 32 bit result of the multiplication of var1 times var2 | + | with one shift left i.e.: | + | L_mult(var1,var2) = L_shl((var1 times var2),1) and | + | L_mult(-32768,-32768) = 2147483647. | + | | + | Complexity weight : 1 | + | | + | Inputs : | + | | + | var1 | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var1 <= 0x0000 7fff. | + | | + | var2 | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var1 <= 0x0000 7fff. | + | | + | Outputs : | + | | + | none | + | | + | Return Value : | + | | + | L_var_out | + | 32 bit long signed integer (Word32) whose value falls in the | + | range : 0x8000 0000 <= L_var_out <= 0x7fff ffff. | + |___________________________________________________________________________| +*/ + +static inline Word32 L_mult (Word16 var1, Word16 var2) +{ + if (unlikely(var1 == MIN_16 && var2 == MIN_16)) { + Overflow = 1; + return MAX_32; + } else + return ((Word32) var1 * var2) << 1; +} + +/*___________________________________________________________________________ + | | + | Function Name : L_mac | + | | + | Purpose : | + | | + | Multiply var1 by var2 and shift the result left by 1. Add the 32 bit | + | result to L_var3 with saturation, return a 32 bit result: | + | L_mac(L_var3,var1,var2) = L_add(L_var3,L_mult(var1,var2)). | + | | + | Complexity weight : 1 | + | | + | Inputs : | + | | + | L_var3 32 bit long signed integer (Word32) whose value falls in the | + | range : 0x8000 0000 <= L_var3 <= 0x7fff ffff. | + | | + | var1 | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var1 <= 0x0000 7fff. | + | | + | var2 | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var1 <= 0x0000 7fff. | + | | + | Outputs : | + | | + | none | + | | + | Return Value : | + | | + | L_var_out | + | 32 bit long signed integer (Word32) whose value falls in the | + | range : 0x8000 0000 <= L_var_out <= 0x7fff ffff. | + |___________________________________________________________________________| +*/ + +static inline Word32 L_mac (Word32 L_var3, Word16 var1, Word16 var2) +{ + Word32 L_var_out; + Word32 L_product; + + L_product = L_mult (var1, var2); + L_var_out = L_add (L_var3, L_product); + return (L_var_out); +} + +/*___________________________________________________________________________ + | | + | Function Name : L_msu | + | | + | Purpose : | + | | + | Multiply var1 by var2 and shift the result left by 1. Subtract the 32 | + | bit result to L_var3 with saturation, return a 32 bit result: | + | L_msu(L_var3,var1,var2) = L_sub(L_var3,L_mult(var1,var2)). | + | | + | Complexity weight : 1 | + | | + | Inputs : | + | | + | L_var3 32 bit long signed integer (Word32) whose value falls in the | + | range : 0x8000 0000 <= L_var3 <= 0x7fff ffff. | + | | + | var1 | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var1 <= 0x0000 7fff. | + | | + | var2 | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var1 <= 0x0000 7fff. | + | | + | Outputs : | + | | + | none | + | | + | Return Value : | + | | + | L_var_out | + | 32 bit long signed integer (Word32) whose value falls in the | + | range : 0x8000 0000 <= L_var_out <= 0x7fff ffff. | + |___________________________________________________________________________| +*/ + +static inline Word32 L_msu (Word32 L_var3, Word16 var1, Word16 var2) +{ + Word32 L_var_out; + Word32 L_product; + + L_product = L_mult (var1, var2); + L_var_out = L_sub (L_var3, L_product); + return (L_var_out); +} + +/*___________________________________________________________________________ + | | + | Function Name : L_negate | + | | + | Purpose : | + | | + | Negate the 32 bit variable L_var1 with saturation; saturate in the case | + | where input is -2147483648 (0x8000 0000). | + | | + | Complexity weight : 2 | + | | + | Inputs : | + | | + | L_var1 32 bit long signed integer (Word32) whose value falls in the | + | range : 0x8000 0000 <= L_var3 <= 0x7fff ffff. | + | | + | Outputs : | + | | + | none | + | | + | Return Value : | + | | + | L_var_out | + | 32 bit long signed integer (Word32) whose value falls in the | + | range : 0x8000 0000 <= L_var_out <= 0x7fff ffff. | + |___________________________________________________________________________| +*/ + +static inline Word32 L_negate (Word32 L_var1) +{ + Word32 L_var_out; + + L_var_out = (L_var1 == MIN_32) ? MAX_32 : -L_var1; + return (L_var_out); +} + +/*___________________________________________________________________________ + | | + | Function Name : L_shl | + | | + | Purpose : | + | | + | Arithmetically shift the 32 bit input L_var1 left var2 positions. Zero | + | fill the var2 LSB of the result. If var2 is negative, arithmetically | + | shift L_var1 right by -var2 with sign extension. Saturate the result in | + | case of underflows or overflows. | + | | + | Complexity weight : 2 | + | | + | Inputs : | + | | + | L_var1 32 bit long signed integer (Word32) whose value falls in the | + | range : 0x8000 0000 <= L_var3 <= 0x7fff ffff. | + | | + | var2 | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var1 <= 0x0000 7fff. | + | | + | Outputs : | + | | + | none | + | | + | Return Value : | + | | + | L_var_out | + | 32 bit long signed integer (Word32) whose value falls in the | + | range : 0x8000 0000 <= L_var_out <= 0x7fff ffff. | + |___________________________________________________________________________| +*/ + +static inline Word32 L_shl (Word32 L_var1, Word16 var2) +{ + Word32 L_var_out; + + if (var2 < 0) { + var2 = -var2; + if (unlikely(var2 >= 31)) + L_var_out = (L_var1 < 0) ? -1 : 0; + else + L_var_out = L_var1 >> var2; + } else { + L_var_out = L_var1 << var2; + if (unlikely(L_var_out >> var2 != L_var1)) + L_var_out = (L_var1 >> 31) ^ MAX_32; + } + return (L_var_out); +} + +/*___________________________________________________________________________ + | | + | Function Name : L_shr | + | | + | Purpose : | + | | + | Arithmetically shift the 32 bit input L_var1 right var2 positions with | + | sign extension. If var2 is negative, arithmetically shift L_var1 left | + | by -var2 and zero fill the -var2 LSB of the result. Saturate the result | + | in case of underflows or overflows. | + | | + | Complexity weight : 2 | + | | + | Inputs : | + | | + | L_var1 32 bit long signed integer (Word32) whose value falls in the | + | range : 0x8000 0000 <= L_var3 <= 0x7fff ffff. | + | | + | var2 | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var1 <= 0x0000 7fff. | + | | + | Outputs : | + | | + | none | + | | + | Return Value : | + | | + | L_var_out | + | 32 bit long signed integer (Word32) whose value falls in the | + | range : 0x8000 0000 <= L_var_out <= 0x7fff ffff. | + |___________________________________________________________________________| +*/ + +static inline Word32 L_shr (Word32 L_var1, Word16 var2) +{ + Word32 L_var_out; + + if (var2 < 0) { + var2 = -var2; + L_var_out = L_var1 << var2; + if (unlikely(L_var_out >> var2 != L_var1)) + L_var_out = (L_var1 >> 31) ^ MAX_32; + } else { + if (unlikely(var2 >= 31)) + L_var_out = (L_var1 < 0) ? -1 : 0; + else + L_var_out = L_var1 >> var2; + } + return (L_var_out); +} + +/*___________________________________________________________________________ + | | + | Function Name : L_deposit_h | + | | + | Purpose : | + | | + | Deposit the 16 bit var1 into the 16 MS bits of the 32 bit output. The | + | 16 LS bits of the output are zeroed. | + | | + | Complexity weight : 2 | + | | + | Inputs : | + | | + | var1 | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var1 <= 0x0000 7fff. | + | | + | Outputs : | + | | + | none | + | | + | Return Value : | + | | + | L_var_out | + | 32 bit long signed integer (Word32) whose value falls in the | + | range : 0x8000 0000 <= var_out <= 0x7fff 0000. | + |___________________________________________________________________________| +*/ + +static inline Word32 L_deposit_h (Word16 var1) +{ + Word32 L_var_out; + + L_var_out = (Word32) var1 << 16; + return (L_var_out); +} + +/*___________________________________________________________________________ + | | + | Function Name : L_deposit_l | + | | + | Purpose : | + | | + | Deposit the 16 bit var1 into the 16 LS bits of the 32 bit output. The | + | 16 MS bits of the output are sign extended. | + | | + | Complexity weight : 2 | + | | + | Inputs : | + | | + | var1 | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var1 <= 0x0000 7fff. | + | | + | Outputs : | + | | + | none | + | | + | Return Value : | + | | + | L_var_out | + | 32 bit long signed integer (Word32) whose value falls in the | + | range : 0xFFFF 8000 <= var_out <= 0x0000 7fff. | + |___________________________________________________________________________| +*/ + +static inline Word32 L_deposit_l (Word16 var1) +{ + Word32 L_var_out; + + L_var_out = (Word32) var1; + return (L_var_out); +} + +/*___________________________________________________________________________ + | | + | Function Name : L_abs | + | | + | Purpose : | + | | + | Absolute value of L_var1; Saturate in case where the input is | + | -214783648 | + | | + | Complexity weight : 3 | + | | + | Inputs : | + | | + | L_var1 | + | 32 bit long signed integer (Word32) whose value falls in the | + | range : 0x8000 0000 <= var1 <= 0x7fff ffff. | + | | + | Outputs : | + | | + | none | + | | + | Return Value : | + | | + | L_var_out | + | 32 bit long signed integer (Word32) whose value falls in the | + | range : 0x0000 0000 <= var_out <= 0x7fff ffff. | + |___________________________________________________________________________| +*/ + +static inline Word32 L_abs (Word32 L_var1) +{ + Word32 L_var_out; + + if (L_var1 == MIN_32) + { + L_var_out = MAX_32; + } + else + { + if (L_var1 < 0) + { + L_var_out = -L_var1; + } + else + { + L_var_out = L_var1; + } + } + + return (L_var_out); +} + +/* additional ops */ + +/*___________________________________________________________________________ + | | + | Function Name : mult_r | + | | + | Purpose : | + | | + | Same as mult with rounding, i.e.: | + | mult_r(var1,var2) = extract_l(L_shr(((var1 * var2) + 16384),15)) and | + | mult_r(-32768,-32768) = 32767. | + | | + | Complexity weight : 2 | + | | + | Inputs : | + | | + | var1 | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var1 <= 0x0000 7fff. | + | | + | var2 | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var1 <= 0x0000 7fff. | + | | + | Outputs : | + | | + | none | + | | + | Return Value : | + | | + | var_out | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var_out <= 0x0000 7fff. | + |___________________________________________________________________________| +*/ + +static inline Word16 mult_r (Word16 var1, Word16 var2) +{ + Word16 var_out; + Word32 L_product_arr; + + L_product_arr = (Word32) var1 *(Word32) var2; /* product */ + L_product_arr += (Word32) 0x00004000L; /* round */ + L_product_arr &= (Word32) 0xffff8000L; + L_product_arr >>= 15; /* shift */ + + if (unlikely(L_product_arr > MAX_16)) + return MAX_16; + else if (unlikely(L_product_arr < MIN_16)) + return MIN_16; + else + return L_product_arr; +} + +/*___________________________________________________________________________ + | | + | Function Name : norm_s | + | | + | Purpose : | + | | + | Produces the number of left shift needed to normalize the 16 bit varia- | + | ble var1 for positive values on the interval with minimum of 16384 and | + | maximum of 32767, and for negative values on the interval with minimum | + | of -32768 and maximum of -16384; in order to normalize the result, the | + | following operation must be done : | + | norm_var1 = shl(var1,norm_s(var1)). | + | | + | Complexity weight : 15 | + | | + | Inputs : | + | | + | var1 | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0xffff 8000 <= var1 <= 0x0000 7fff. | + | | + | Outputs : | + | | + | none | + | | + | Return Value : | + | | + | var_out | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0x0000 0000 <= var_out <= 0x0000 000f. | + |___________________________________________________________________________| +*/ + +static inline Word16 norm_s (Word16 var1) +{ + Word16 var_out; + + if (var1 == 0) + { + var_out = 0; + } + else + { + if (var1 == (Word16) 0xffff) + { + var_out = 15; + } + else + { + if (var1 < 0) + { + var1 = ~var1; + } + for (var_out = 0; var1 < 0x4000; var_out++) + { + var1 <<= 1; + } + } + } + + return (var_out); +} + +/*___________________________________________________________________________ + | | + | Function Name : norm_l | + | | + | Purpose : | + | | + | Produces the number of left shifts needed to normalize the 32 bit varia-| + | ble L_var1 for positive values on the interval with minimum of | + | 1073741824 and maximum of 2147483647, and for negative values on the in-| + | terval with minimum of -2147483648 and maximum of -1073741824; in order | + | to normalize the result, the following operation must be done : | + | norm_L_var1 = L_shl(L_var1,norm_l(L_var1)). | + | | + | Complexity weight : 30 | + | | + | Inputs : | + | | + | L_var1 | + | 32 bit long signed integer (Word32) whose value falls in the | + | range : 0x8000 0000 <= var1 <= 0x7fff ffff. | + | | + | Outputs : | + | | + | none | + | | + | Return Value : | + | | + | var_out | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0x0000 0000 <= var_out <= 0x0000 001f. | + |___________________________________________________________________________| +*/ + +static inline Word16 norm_l (Word32 L_var1) +{ + Word16 var_out; + + if (L_var1 == 0) + { + var_out = 0; + } + else + { + if (L_var1 == (Word32) 0xffffffffL) + { + var_out = 31; + } + else + { + if (L_var1 < 0) + { + L_var1 = ~L_var1; + } + for (var_out = 0; L_var1 < (Word32) 0x40000000L; var_out++) + { + L_var1 <<= 1; + } + } + } + + return (var_out); +} + +/* ETSI functions that aren't inlined */ + Word32 L_macNs (Word32 L_var3, Word16 var1, Word16 var2); /* Mac without sat, 1 */ Word32 L_msuNs (Word32 L_var3, Word16 var1, Word16 var2); /* Msu without sat, 1 */ -Word32 L_add (Word32 L_var1, Word32 L_var2); /* Long add, 2 */ -Word32 L_sub (Word32 L_var1, Word32 L_var2); /* Long sub, 2 */ Word32 L_add_c (Word32 L_var1, Word32 L_var2); /* Long add with c, 2 */ Word32 L_sub_c (Word32 L_var1, Word32 L_var2); /* Long sub with c, 2 */ -Word32 L_negate (Word32 L_var1); /* Long negate, 2 */ -Word16 mult_r (Word16 var1, Word16 var2); /* Mult with round, 2 */ -Word32 L_shl (Word32 L_var1, Word16 var2); /* Long shift left, 2 */ -Word32 L_shr (Word32 L_var1, Word16 var2); /* Long shift right, 2*/ Word16 shr_r (Word16 var1, Word16 var2); /* Shift right with round, 2 */ Word16 mac_r (Word32 L_var3, Word16 var1, Word16 var2); /* Mac with rounding,2 */ Word16 msu_r (Word32 L_var3, Word16 var1, Word16 var2); /* Msu with rounding,2 */ -Word32 L_deposit_h (Word16 var1); /* 16 bit var1 -> MSB, 2 */ -Word32 L_deposit_l (Word16 var1); /* 16 bit var1 -> LSB, 2 */ - Word32 L_shr_r (Word32 L_var1, Word16 var2); /* Long shift right with round, 3 */ -Word32 L_abs (Word32 L_var1); /* Long abs, 3 */ Word32 L_sat (Word32 L_var1); /* Long saturation, 4 */ -Word16 norm_s (Word16 var1); /* Short norm, 15 */ Word16 div_s (Word16 var1, Word16 var2); /* Short division, 18 */ -Word16 norm_l (Word32 L_var1); /* Long norm, 30 */ -