moniere/CORDIC_Rotate_APFX
1
0
Fork 0
mirror of https://github.com/DrasLorus/CORDIC_Rotate_APFX.git synced 2024-11-14 17:13:27 +01:00
Code Issues Projects Releases Packages Wiki Activity
CORDIC_Rotate_APFX/sources/CCordicRotateRom/CCordicRotateRomTemplate.hpp

36 lines
1.3 KiB
C++
Raw Normal View History

Fix for GCC 9.4 support, formatting, and license addition.
2022-02-17 20:22:12 +01:00
/*
*
* Copyright 2022 Camille "DrasLorus" Monière.
*
* This file is part of CORDIC_Rotate_APFX.
*
* This program is free software: you can redistribute it and/or modify it under the terms of the GNU
* Lesser General Public License as published by the Free Software Foundation, either version 3 of
* the License, or (at your option) any later version.
*
* This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without
* even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
* Lesser General Public License for more details.
*
* You should have received a copy of the GNU Lesser General Public License along with this program.
* If not, see <https://www.gnu.org/licenses/>.
*
*/
Working AP template cordic
2022-02-11 18:17:13 +01:00
Big update, that compiles with Xilinx GCC 6.2 - Commit title implies possible Xilinx Vivado HLS 2019.1 support, thus support for e.g. all USRP Series 3 from Ettus. - Add a new CORDIC version, Rom based and meta-programmed via CMake features, to be even compiled with earlier GCC (not quite tested, but by changing constexpr to const and using gcc 4.6 -sdt=c++0x, it worked). - Class name modified, to be more explicit.
2022-02-18 21:33:59 +01:00
#ifndef C_CORDIC_ROTATE_ROM_TEMPLATE
#define C_CORDIC_ROTATE_ROM_TEMPLATE
enum rom_types {
Correct the name and improve widely - Fix the MC (Monte-Carlo) to the proper algoritm name, ML (maximum likelyhood) and remove HalfPi since the use of divider allow to theoretically support any pi / 2^k, k an integer. In reality, a too low rotation would require more stages than 7 but it is for futur improvements. - Make use of `divider` template to provide rotation grain finer than pi / 2. Validated (unit-tested) with pi / 4 with the same margins than pi / 2 (2% of error with floating scaling, 3% with fixed scaling). - Fix rom size which now use N_STAGES+1 bits instead of 8 regardless of N_STAGES. Simplify the cordic method implementation, which unexpectedly (and fortunately) improved its performance.
2022-03-14 14:07:10 +01:00
ml,
Big update, that compiles with Xilinx GCC 6.2 - Commit title implies possible Xilinx Vivado HLS 2019.1 support, thus support for e.g. all USRP Series 3 from Ettus. - Add a new CORDIC version, Rom based and meta-programmed via CMake features, to be even compiled with earlier GCC (not quite tested, but by changing constexpr to const and using gcc 4.6 -sdt=c++0x, it worked). - Class name modified, to be more explicit.
2022-02-18 21:33:59 +01:00
cst
};
Correct the name and improve widely - Fix the MC (Monte-Carlo) to the proper algoritm name, ML (maximum likelyhood) and remove HalfPi since the use of divider allow to theoretically support any pi / 2^k, k an integer. In reality, a too low rotation would require more stages than 7 but it is for futur improvements. - Make use of `divider` template to provide rotation grain finer than pi / 2. Validated (unit-tested) with pi / 4 with the same margins than pi / 2 (2% of error with floating scaling, 3% with fixed scaling). - Fix rom size which now use N_STAGES+1 bits instead of 8 regardless of N_STAGES. Simplify the cordic method implementation, which unexpectedly (and fortunately) improved its performance.
2022-03-14 14:07:10 +01:00
template <unsigned TIn_I, rom_types type, unsigned TIn_W, unsigned Tnb_stages, unsigned Tq, unsigned divider = 2>
Big update, that compiles with Xilinx GCC 6.2 - Commit title implies possible Xilinx Vivado HLS 2019.1 support, thus support for e.g. all USRP Series 3 from Ettus. - Add a new CORDIC version, Rom based and meta-programmed via CMake features, to be even compiled with earlier GCC (not quite tested, but by changing constexpr to const and using gcc 4.6 -sdt=c++0x, it worked). - Class name modified, to be more explicit.
2022-02-18 21:33:59 +01:00
class CCordicRotateRom {
static_assert(TIn_W > 0, "Inputs can't be on zero bits.");
static_assert(Tnb_stages < 8, "7 stages of CORDIC is the maximum supported.");
static_assert(Tnb_stages > 1, "2 stages of CORDIC is the minimum.");
Correct the name and improve widely - Fix the MC (Monte-Carlo) to the proper algoritm name, ML (maximum likelyhood) and remove HalfPi since the use of divider allow to theoretically support any pi / 2^k, k an integer. In reality, a too low rotation would require more stages than 7 but it is for futur improvements. - Make use of `divider` template to provide rotation grain finer than pi / 2. Validated (unit-tested) with pi / 4 with the same margins than pi / 2 (2% of error with floating scaling, 3% with fixed scaling). - Fix rom size which now use N_STAGES+1 bits instead of 8 regardless of N_STAGES. Simplify the cordic method implementation, which unexpectedly (and fortunately) improved its performance.
2022-03-14 14:07:10 +01:00
static_assert(((divider - 1) & divider) == 0, "divider must be a power of 2.");
Big update, that compiles with Xilinx GCC 6.2 - Commit title implies possible Xilinx Vivado HLS 2019.1 support, thus support for e.g. all USRP Series 3 from Ettus. - Add a new CORDIC version, Rom based and meta-programmed via CMake features, to be even compiled with earlier GCC (not quite tested, but by changing constexpr to const and using gcc 4.6 -sdt=c++0x, it worked). - Class name modified, to be more explicit.
2022-02-18 21:33:59 +01:00
};
#endif // C_CORDIC_ROTATE_ROM_TEMPLATE
Reference in a new issue Copy permalink