Move to using the FastMath from BasicBlocks

part of the big SC/XT refactor share project I'm chipping away
at and also aprt of surge-synthesizer/surge#6532

This will make this library depend on sst-basic-blocks (and that dependency
will get stronger as we move forward). Youc an auto-grab it with CPM if you
want.

CMakeLists.txt

@@ -29,6 +29,14 @@ option(SST_FILTERS_BUILD_EXAMPLES "Add targets for building and running sst-filt
 
 if (SST_FILTERS_BUILD_TESTS OR SST_FILTERS_BUILD_EXAMPLES)
     message(STATUS "Importing SIMDE with CPM")
+
+    if (NOT TARGET sst-basic-blocks)
+        CPMAddPackage(NAME sst-basic-blocks
+                GITHUB_REPOSITORY surge-synthesizer/sst-basic-blocks
+                GIT_TAG main
+                )
+    endif()
+
     CPMAddPackage(NAME simde
             GITHUB_REPOSITORY simd-everywhere/simde
             VERSION 0.7.2
@@ -37,10 +45,25 @@ if (SST_FILTERS_BUILD_TESTS OR SST_FILTERS_BUILD_EXAMPLES)
     target_include_directories(simde INTERFACE ${simde_SOURCE_DIR})
 endif ()
 
+if (SST_GET_BASIC_BLOCKS)
+    CPMAddPackage(NAME sst-basic-blocks
+            GITHUB_REPOSITORY surge-synthesizer/sst-basic-blocks
+            GIT_TAG main
+            )
+endif()
+
+if (NOT TARGET sst-basic-blocks)
+    message(FATAL_ERROR "sst-basic-blocks is not available in this context. Set SST_GET_BASIC_BLOCKS=1 or add it")
+else()
+    target_link_libraries(${PROJECT_NAME} INTERFACE sst-basic-blocks)
+endif()
+
 if (SST_FILTERS_BUILD_TESTS)
     add_subdirectory(tests)
 endif ()
 
 if(SST_FILTERS_BUILD_EXAMPLES)
     add_subdirectory(examples)
 endif()
+
+

include/sst/filters/CutoffWarp.h

@@ -4,7 +4,7 @@
 #include "QuadFilterUnit.h"
 #include "FilterCoefficientMaker.h"
 #include "sst/utilities/basic_dsp.h"
-#include "sst/utilities/FastMath.h"
+#include "sst/basic-blocks/dsp/FastMath.h"
 
 /**
  * This namespace contains an adaptation of the filter found at
@@ -89,7 +89,7 @@ static inline __m128 doNLFilter(const __m128 input, const __m128 a1, const __m12
         nf = ojd_waveshaper_ps(out);
         break;
     default: // SAT_TANH; the removed SAT_SINE and others are also caught here
-        nf = utilities::DSP::fasttanhSSEclamped(out);
+        nf = basic_blocks::dsp::fasttanhSSEclamped(out);
         break;
     }
 
@@ -136,8 +136,8 @@ void makeCoefficients(FilterCoefficientMaker<TuningProvider> *cm, float freq, fl
     const float normalisedFreq = 2.0f * clampedFrequency(freq, sampleRate, provider) / sampleRate;
     const float wc = (float)M_PI * normalisedFreq;
 
-    const float wsin = utilities::DSP::fastsin(wc);
-    const float wcos = utilities::DSP::fastcos(wc);
+    const float wsin = basic_blocks::dsp::fastsin(wc);
+    const float wcos = basic_blocks::dsp::fastcos(wc);
     const float alpha = wsin / (2.0f * q);
 
     // note we actually calculate the reciprocal of a0 because we only use a0 to normalize the

include/sst/filters/DiodeLadder.h

@@ -2,6 +2,7 @@
 #define SST_FILTERS_DIODELADDER_H
 
 #include "sst/utilities/basic_dsp.h"
+#include "sst/basic-blocks/dsp/FastMath.h"
 #include "QuadFilterUnit.h"
 #include "FilterCoefficientMaker.h"
 
@@ -79,7 +80,7 @@ void makeCoefficients(FilterCoefficientMaker<TuningProvider> *cm, float freq, fl
                       float sampleRate, float sampleRateInv, TuningProvider *provider)
 {
     const float wd = clampedFrequency(freq, sampleRate, provider) * 2.0f * (float)M_PI;
-    const float wa = (2.0f * sampleRate) * utilities::DSP::fasttan(wd * sampleRateInv * 0.5f);
+    const float wa = (2.0f * sampleRate) * basic_blocks::dsp::fasttan(wd * sampleRateInv * 0.5f);
     const float g = wa * sampleRateInv * 0.5f;
 
     const float G4 = 0.5f * g / (1.0f + g);

include/sst/filters/K35Filter.h

@@ -1,7 +1,7 @@
 #ifndef SST_FILTERS_K35FILTER_H
 #define SST_FILTERS_K35FILTER_H
 
-#include "sst/utilities/FastMath.h"
+#include "sst/basic-blocks/dsp/FastMath.h"
 #include "QuadFilterUnit.h"
 #include "FilterCoefficientMaker.h"
 
@@ -72,7 +72,7 @@ void makeCoefficients(FilterCoefficientMaker<TuningProvider> *cm, float freq, fl
     float C[n_cm_coeffs];
 
     const float wd = clampedFrequency(freq, sampleRate, provider) * 2.0f * (float)M_PI;
-    const float wa = (2.0f * sampleRate) * utilities::DSP::fasttan(wd * sampleRateInv * 0.5f);
+    const float wa = (2.0f * sampleRate) * basic_blocks::dsp::fasttan(wd * sampleRateInv * 0.5f);
     const float g = wa * sampleRateInv * 0.5f;
     const float gp1 = (1.0f + g); // g plus 1
     const float G = g / gp1;
@@ -120,7 +120,7 @@ inline __m128 process_lp(QuadFilterUnitState *__restrict f, __m128 input)
     const __m128 s35 = A(M(f->C[k35_lb], f->R[k35_2z]), M(f->C[k35_hb], f->R[k35_hz]));
     // alpha * (y1 + s35)
     const __m128 u_clean = M(f->C[k35_alpha], A(y1, s35));
-    const __m128 u_driven = utilities::DSP::fasttanhSSEclamped(M(u_clean, f->C[k35_saturation]));
+    const __m128 u_driven = basic_blocks::dsp::fasttanhSSEclamped(M(u_clean, f->C[k35_saturation]));
     const __m128 u =
         A(M(u_clean, f->C[k35_saturation_blend_inv]), M(u_driven, f->C[k35_saturation_blend]));
 
@@ -145,7 +145,7 @@ inline __m128 process_hp(QuadFilterUnitState *__restrict f, __m128 input)
 
     // mk * lpf2(u)
     const __m128 y_clean = M(f->C[k35_k], u);
-    const __m128 y_driven = utilities::DSP::fasttanhSSEclamped(M(y_clean, f->C[k35_saturation]));
+    const __m128 y_driven = basic_blocks::dsp::fasttanhSSEclamped(M(y_clean, f->C[k35_saturation]));
     const __m128 y =
         A(M(y_clean, f->C[k35_saturation_blend_inv]), M(y_driven, f->C[k35_saturation_blend]));
 

include/sst/filters/ResonanceWarp.h

@@ -4,7 +4,7 @@
 #include "QuadFilterUnit.h"
 #include "FilterCoefficientMaker.h"
 #include "sst/utilities/basic_dsp.h"
-#include "sst/utilities/FastMath.h"
+#include "sst/basic-blocks/dsp/FastMath.h"
 
 /**
  * This contains an adaptation of the filter found at
@@ -53,8 +53,8 @@ static inline __m128 doNLFilter(const __m128 input, const __m128 a1, const __m12
     switch (sat)
     {
     case SAT_TANH:
-        z1 = utilities::DSP::fasttanhSSEclamped(z1);
-        z2 = utilities::DSP::fasttanhSSEclamped(z2);
+        z1 = basic_blocks::dsp::fasttanhSSEclamped(z1);
+        z2 = basic_blocks::dsp::fasttanhSSEclamped(z2);
         break;
     default:
         z1 = utilities::softclip_ps(z1); // note, this is a bit different to Jatin's softclipper
@@ -98,8 +98,8 @@ void makeCoefficients(FilterCoefficientMaker<TuningProvider> *cm, float freq, fl
 
     const float wc = 2.0f * (float)M_PI * clampedFrequency(freq, sampleRate, provider) / sampleRate;
 
-    const float wsin = utilities::DSP::fastsin(wc);
-    const float wcos = utilities::DSP::fastcos(wc);
+    const float wsin = basic_blocks::dsp::fastsin(wc);
+    const float wcos = basic_blocks::dsp::fastcos(wc);
     const float alpha = wsin / (2.0f * q);
 
     // note we actually calculate the reciprocal of a0 because we only use a0 to normalize the

include/sst/filters/TriPoleFilter.h

@@ -3,6 +3,7 @@
 
 #include "QuadFilterUnit.h"
 #include "FilterCoefficientMaker.h"
+#include "sst/basic-blocks/dsp/FastMath.h"
 
 /**
  * This filter is an emulation of the "Threeler" VCF by
@@ -261,7 +262,7 @@ static inline __m128 res_func_ps(__m128 x)
     auto x_less_than = _mm_cmplt_ps(x_abs, F(max_val));
 
     auto y =
-        A(N(utilities::DSP::fastexpSSE(M(F(beta_exp), N(utilities::abs_ps(A(x, F(c))))))), F(bias));
+        A(N(basic_blocks::dsp::fastexpSSE(M(F(beta_exp), N(utilities::abs_ps(A(x, F(c))))))), F(bias));
     y = M(sign_ps(x), M(y, F(oneOverMult)));
 
     return _mm_or_ps(_mm_and_ps(x_less_than, M(x, F(oneOverMult))), _mm_andnot_ps(x_less_than, y));
@@ -274,7 +275,7 @@ static inline __m128 res_deriv_ps(__m128 x)
     auto x_abs = utilities::abs_ps(x);
     auto x_less_than = _mm_cmplt_ps(x_abs, F(max_val));
 
-    auto y = A(utilities::DSP::fastexpSSE(M(F(beta_exp), N(utilities::abs_ps(A(x, F(c)))))),
+    auto y = A(basic_blocks::dsp::fastexpSSE(M(F(beta_exp), N(utilities::abs_ps(A(x, F(c)))))),
                F(betaExpOverMult));
 
     return _mm_or_ps(_mm_and_ps(x_less_than, F(one)), _mm_andnot_ps(x_less_than, y));

include/sst/filters/VintageLadders.h

@@ -3,7 +3,7 @@
 
 #include "sst/utilities/globals.h"
 #include "sst/utilities/basic_dsp.h"
-#include "sst/utilities/FastMath.h"
+#include "sst/basic-blocks/dsp/FastMath.h"
 #include "QuadFilterUnit.h"
 #include "FilterCoefficientMaker.h"
 
@@ -365,7 +365,7 @@ inline __m128 process(QuadFilterUnitState *__restrict f, __m128 in)
         auto acr = A(M(mneg39364, fc2), A(M(m18409, fc), m09968));
 
         // auto tune = (1.0 - exp(-((2 * M_PI) * f * fcr))) / thermal;
-        auto tune = M(S(one, utilities::DSP::fastexpSSE(M(neg2pi, M(fr, fcr)))), oneoverthermal);
+        auto tune = M(S(one, basic_blocks::dsp::fastexpSSE(M(neg2pi, M(fr, fcr)))), oneoverthermal);
         // auto resquad = 4.0 * res * arc;
         auto resquad = M(four, M(res, acr));
 
@@ -380,7 +380,7 @@ inline __m128 process(QuadFilterUnitState *__restrict f, __m128 in)
 
         // delay[0] = stage[0] = delay[0] + tune * (tanh(input * thermal) - stageTanh[0]);
         f->R[h_stage + 0] =
-            A(f->R[h_delay + 0], M(tune, S(utilities::DSP::fasttanhSSEclamped(M(input, thermal)),
+            A(f->R[h_delay + 0], M(tune, S(basic_blocks::dsp::fasttanhSSEclamped(M(input, thermal)),
                                            f->R[h_stageTanh + 0])));
         f->R[h_delay + 0] = f->R[h_stage + 0];
 
@@ -391,11 +391,11 @@ inline __m128 process(QuadFilterUnitState *__restrict f, __m128 in)
 
             // stage[k] = delay[k] + tune * ((stageTanh[k-1] = tanh(input * thermal)) - (k != 3 ?
             // stageTanh[k] : tanh(delay[k] * thermal)));
-            f->R[h_stageTanh + k - 1] = utilities::DSP::fasttanhSSEclamped(M(input, thermal));
+            f->R[h_stageTanh + k - 1] = basic_blocks::dsp::fasttanhSSEclamped(M(input, thermal));
             f->R[h_stage + k] =
                 A(f->R[h_delay + k], M(tune, S(f->R[h_stageTanh + k - 1],
                                                (k != 3 ? f->R[h_stageTanh + k]
-                                                       : utilities::DSP::fasttanhSSEclamped(
+                                                       : basic_blocks::dsp::fasttanhSSEclamped(
                                                              M(f->R[h_delay + k], thermal))))));
 
             // delay[k] = stage[k];