FreeCalypso > hg > gsm-codec-lib
diff libgsmefr/basicops.c @ 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 | libgsmefr/basicop2.c@38326102fc43 |
| children |
line wrap: on
line diff
--- /dev/null Thu Jan 01 00:00:00 1970 +0000 +++ b/libgsmefr/basicops.c Sun Apr 21 23:17:21 2024 +0000 @@ -0,0 +1,699 @@ +/* + * This C module is a stripped-down version made from ETSI basicop2.c, + * reduced to just those functions which we are not inlining. + */ + +/*___________________________________________________________________________ + | | + | Include-Files | + |___________________________________________________________________________| +*/ + +#include <stdint.h> +#include "typedef.h" +#include "namespace.h" +#include "basic_op.h" + +/*___________________________________________________________________________ + | | + | Functions | + |___________________________________________________________________________| +*/ + +/*___________________________________________________________________________ + | | + | Function Name : L_macNs | + | | + | Purpose : | + | | + | Multiply var1 by var2 and shift the result left by 1. Add the 32 bit | + | result to L_var3 without saturation, return a 32 bit result. Generate | + | carry and overflow values : | + | L_macNs(L_var3,var1,var2) = L_add_c(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. | + | | + | Caution : | + | | + | In some cases the Carry flag has to be cleared or set before using | + | operators which take into account its value. | + |___________________________________________________________________________| +*/ + +Word32 L_macNs (Word32 L_var3, Word16 var1, Word16 var2) +{ + Word32 L_var_out; + + L_var_out = L_mult (var1, var2); + L_var_out = L_add_c (L_var3, L_var_out); + return (L_var_out); +} + +/*___________________________________________________________________________ + | | + | Function Name : L_msuNs | + | | + | Purpose : | + | | + | Multiply var1 by var2 and shift the result left by 1. Subtract the 32 | + | bit result from L_var3 without saturation, return a 32 bit result. Ge- | + | nerate carry and overflow values : | + | L_msuNs(L_var3,var1,var2) = L_sub_c(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. | + | | + | Caution : | + | | + | In some cases the Carry flag has to be cleared or set before using | + | operators which take into account its value. | + |___________________________________________________________________________| +*/ + +Word32 L_msuNs (Word32 L_var3, Word16 var1, Word16 var2) +{ + Word32 L_var_out; + + L_var_out = L_mult (var1, var2); + L_var_out = L_sub_c (L_var3, L_var_out); + return (L_var_out); +} + +/*___________________________________________________________________________ + | | + | Function Name : L_add_c | + | | + | Purpose : | + | | + | Performs 32 bits addition of the two 32 bits variables (L_var1+L_var2+C)| + | with carry. No saturation. Generate carry and Overflow values. The car- | + | ry and overflow values are binary variables which can be tested and as- | + | signed values. | + | | + | 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. | + | | + | Caution : | + | | + | In some cases the Carry flag has to be cleared or set before using | + | operators which take into account its value. | + |___________________________________________________________________________| +*/ +Word32 L_add_c (Word32 L_var1, Word32 L_var2) +{ + Word32 L_var_out; + Word32 L_test; + Flag carry_int = 0; + + L_var_out = L_var1 + L_var2 + Carry; + + L_test = L_var1 + L_var2; + + if ((L_var1 > 0) && (L_var2 > 0) && (L_test < 0)) + { + Overflow = 1; + carry_int = 0; + } + else + { + if ((L_var1 < 0) && (L_var2 < 0)) + { + if (L_test >= 0) + { + Overflow = 1; + carry_int = 1; + } + else + { + Overflow = 0; + carry_int = 1; + } + } + else + { + if (((L_var1 ^ L_var2) < 0) && (L_test >= 0)) + { + Overflow = 0; + carry_int = 1; + } + else + { + Overflow = 0; + carry_int = 0; + } + } + } + + if (Carry) + { + if (L_test == MAX_32) + { + Overflow = 1; + Carry = carry_int; + } + else + { + if (L_test == (Word32) 0xFFFFFFFFL) + { + Carry = 1; + } + else + { + Carry = carry_int; + } + } + } + else + { + Carry = carry_int; + } + + return (L_var_out); +} + +/*___________________________________________________________________________ + | | + | Function Name : L_sub_c | + | | + | Purpose : | + | | + | Performs 32 bits subtraction of the two 32 bits variables with carry | + | (borrow) : L_var1-L_var2-C. No saturation. Generate carry and Overflow | + | values. The carry and overflow values are binary variables which can | + | be tested and assigned values. | + | | + | 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. | + | | + | Caution : | + | | + | In some cases the Carry flag has to be cleared or set before using | + | operators which take into account its value. | + |___________________________________________________________________________| +*/ + +Word32 L_sub_c (Word32 L_var1, Word32 L_var2) +{ + Word32 L_var_out; + Word32 L_test; + Flag carry_int = 0; + + if (Carry) + { + Carry = 0; + if (L_var2 != MIN_32) + { + L_var_out = L_add_c (L_var1, -L_var2); + } + else + { + L_var_out = L_var1 - L_var2; + if (L_var1 > 0L) + { + Overflow = 1; + Carry = 0; + } + } + } + else + { + L_var_out = L_var1 - L_var2 - (Word32) 0X00000001L; + L_test = L_var1 - L_var2; + + if ((L_test < 0) && (L_var1 > 0) && (L_var2 < 0)) + { + Overflow = 1; + carry_int = 0; + } + else if ((L_test > 0) && (L_var1 < 0) && (L_var2 > 0)) + { + Overflow = 1; + carry_int = 1; + } + else if ((L_test > 0) && ((L_var1 ^ L_var2) > 0)) + { + Overflow = 0; + carry_int = 1; + } + if (L_test == MIN_32) + { + Overflow = 1; + Carry = carry_int; + } + else + { + Carry = carry_int; + } + } + + return (L_var_out); +} + +/*___________________________________________________________________________ + | | + | Function Name : shr_r | + | | + | Purpose : | + | | + | Same as shr(var1,var2) but with rounding. Saturate the result in case of| + | underflows or overflows : | + | - If var2 is greater than zero : | + | if (sub(shl(shr(var1,var2),1),shr(var1,sub(var2,1)))) | + | is equal to zero | + | then | + | shr_r(var1,var2) = shr(var1,var2) | + | else | + | shr_r(var1,var2) = add(shr(var1,var2),1) | + | - If var2 is less than or equal to zero : | + | shr_r(var1,var2) = shr(var1,var2). | + | | + | 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. | + |___________________________________________________________________________| +*/ + +Word16 shr_r (Word16 var1, Word16 var2) +{ + Word16 var_out; + + if (var2 > 15) + { + var_out = 0; + } + else + { + var_out = shr (var1, var2); + + if (var2 > 0) + { + if ((var1 & ((Word16) 1 << (var2 - 1))) != 0) + { + var_out++; + } + } + } + return (var_out); +} + +/*___________________________________________________________________________ + | | + | Function Name : mac_r | + | | + | Purpose : | + | | + | Multiply var1 by var2 and shift the result left by 1. Add the 32 bit | + | result to L_var3 with saturation. Round the LS 16 bits of the result | + | into the MS 16 bits with saturation and shift the result right by 16. | + | Return a 16 bit result. | + | mac_r(L_var3,var1,var2) = round(L_mac(L_var3,var1,var2)) | + | | + | Complexity weight : 2 | + | | + | 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 : | + | | + | var_out | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0x0000 8000 <= L_var_out <= 0x0000 7fff. | + |___________________________________________________________________________| +*/ + +Word16 mac_r (Word32 L_var3, Word16 var1, Word16 var2) +{ + Word16 var_out; + + L_var3 = L_mac (L_var3, var1, var2); + L_var3 = L_add (L_var3, (Word32) 0x00008000L); + var_out = extract_h (L_var3); + return (var_out); +} + +/*___________________________________________________________________________ + | | + | Function Name : msu_r | + | | + | Purpose : | + | | + | Multiply var1 by var2 and shift the result left by 1. Subtract the 32 | + | bit result to L_var3 with saturation. Round the LS 16 bits of the res- | + | ult into the MS 16 bits with saturation and shift the result right by | + | 16. Return a 16 bit result. | + | msu_r(L_var3,var1,var2) = round(L_msu(L_var3,var1,var2)) | + | | + | Complexity weight : 2 | + | | + | 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 : | + | | + | var_out | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0x0000 8000 <= L_var_out <= 0x0000 7fff. | + |___________________________________________________________________________| +*/ + +Word16 msu_r (Word32 L_var3, Word16 var1, Word16 var2) +{ + Word16 var_out; + + L_var3 = L_msu (L_var3, var1, var2); + L_var3 = L_add (L_var3, (Word32) 0x00008000L); + var_out = extract_h (L_var3); + return (var_out); +} + +/*___________________________________________________________________________ + | | + | Function Name : L_shr_r | + | | + | Purpose : | + | | + | Same as L_shr(L_var1,var2) but with rounding. Saturate the result in | + | case of underflows or overflows : | + | - If var2 is greater than zero : | + | if (L_sub(L_shl(L_shr(L_var1,var2),1),L_shr(L_var1,sub(var2,1))))| + | is equal to zero | + | then | + | L_shr_r(L_var1,var2) = L_shr(L_var1,var2) | + | else | + | L_shr_r(L_var1,var2) = L_add(L_shr(L_var1,var2),1) | + | - If var2 is less than or equal to zero : | + | L_shr_r(L_var1,var2) = L_shr(L_var1,var2). | + | | + | Complexity weight : 3 | + | | + | Inputs : | + | | + | L_var1 | + | 32 bit long signed integer (Word32) whose value falls in the | + | range : 0x8000 0000 <= var1 <= 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 <= var_out <= 0x7fff ffff. | + |___________________________________________________________________________| +*/ + +Word32 L_shr_r (Word32 L_var1, Word16 var2) +{ + Word32 L_var_out; + + if (var2 > 31) + { + L_var_out = 0; + } + else + { + L_var_out = L_shr (L_var1, var2); + if (var2 > 0) + { + if ((L_var1 & ((Word32) 1 << (var2 - 1))) != 0) + { + L_var_out++; + } + } + } + return (L_var_out); +} + +/*___________________________________________________________________________ + | | + | Function Name : L_sat | + | | + | Purpose : | + | | + | 32 bit L_var1 is set to 2147483647 if an overflow occured or to | + | -2147483648 if an underflow occured on the most recent L_add_c, | + | L_sub_c, L_macNs or L_msuNs operations. The carry and overflow values | + | are binary values which can be tested and assigned values. | + | | + | Complexity weight : 4 | + | | + | 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 : 0x8000 0000 <= var_out <= 0x7fff ffff. | + |___________________________________________________________________________| +*/ + +Word32 L_sat (Word32 L_var1) +{ + Word32 L_var_out; + + L_var_out = L_var1; + + if (Overflow) + { + if (Carry) + { + L_var_out = MIN_32; + } + else + { + L_var_out = MAX_32; + } + + Carry = 0; + Overflow = 0; + } + return (L_var_out); +} + +/*___________________________________________________________________________ + | | + | Function Name : div_s | + | | + | Purpose : | + | | + | Produces a result which is the fractional integer division of var1 by | + | var2; var1 and var2 must be positive and var2 must be greater or equal | + | to var1; the result is positive (leading bit equal to 0) and truncated | + | to 16 bits. | + | If var1 = var2 then div(var1,var2) = 32767. | + | | + | Complexity weight : 18 | + | | + | Inputs : | + | | + | var1 | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0x0000 0000 <= var1 <= var2 and var2 != 0. | + | | + | var2 | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : var1 <= var2 <= 0x0000 7fff and var2 != 0. | + | | + | Outputs : | + | | + | none | + | | + | Return Value : | + | | + | var_out | + | 16 bit short signed integer (Word16) whose value falls in the | + | range : 0x0000 0000 <= var_out <= 0x0000 7fff. | + | It's a Q15 value (point between b15 and b14). | + |___________________________________________________________________________| +*/ + +Word16 div_s (Word16 var1, Word16 var2) +{ + Word16 var_out = 0; + Word16 iteration; + Word32 L_num; + Word32 L_denom; + +#if 0 + if ((var1 > var2) || (var1 < 0) || (var2 < 0)) + { + printf ("Division Error var1=%d var2=%d\n", var1, var2); + exit (0); + } + if (var2 == 0) + { + printf ("Division by 0, Fatal error \n"); + exit (0); + } +#endif + + if (var1 == 0) + { + var_out = 0; + } + else + { + if (var1 == var2) + { + var_out = MAX_16; + } + else + { + L_num = L_deposit_l (var1); + L_denom = L_deposit_l (var2); + + for (iteration = 0; iteration < 15; iteration++) + { + var_out <<= 1; + L_num <<= 1; + + if (L_num >= L_denom) + { + L_num = L_sub (L_num, L_denom); + var_out = add (var_out, 1); + } + } + } + } + + return (var_out); +}