Update to biquad df2t f32 tests since the Neon version is disabled be…

…cause of problems.

So the test must use the scalar version.

Re-enabled all other tests.
Added Neon badge to README.

.github/workflows/runcpptest.yaml

@@ -1,10 +1,10 @@
 name: CPP Tests
-#on:
-#  workflow_dispatch:
-#  pull_request:
-#    branches: [main]
-#  push:
-#    branches: [main]
+on:
+  workflow_dispatch:
+  pull_request:
+    branches: [main]
+  push:
+    branches: [main]
 
 permissions: 
   actions: read

.github/workflows/runneontest.yaml

@@ -67,8 +67,8 @@ jobs:
 
           cmake -G "Ninja" .. 
         
-      - name: Setup tmate session
-        uses: mxschmitt/action-tmate@v3
+#      - name: Setup tmate session
+#        uses: mxschmitt/action-tmate@v3
 
       - name: Execute generic tests
         run: |

.github/workflows/runtest.yaml

@@ -1,10 +1,10 @@
 name: C Tests
-#on:
-#  workflow_dispatch:
-#  pull_request:
-#    branches: [main]
-#  push:
-#    branches: [main]
+on:
+  workflow_dispatch:
+  pull_request:
+    branches: [main]
+  push:
+    branches: [main]
 
 permissions: 
   actions: read

README.md

@@ -6,6 +6,8 @@
 
 [![CPP Tests](https://github.com/ARM-software/CMSIS-DSP/actions/workflows/runcpptest.yaml/badge.svg)](https://github.com/ARM-software/CMSIS-DSP/actions/workflows/runcpptest.yaml)
 
+[![Neon Tests](https://github.com/ARM-software/CMSIS-DSP/actions/workflows/runneontest.yaml/badge.svg)](https://github.com/ARM-software/CMSIS-DSP/actions/workflows/runneontest.yaml)
+
 ## About
 
 CMSIS-DSP is an optimized compute library for embedded systems (DSP is in the name for legacy reasons).

Testing/Source/Tests/BIQUADF32.cpp

@@ -10,8 +10,7 @@ Reference patterns are generated with
 a double precision computation.
 
 */
-#define REL_ERROR (1.2e-3)
-
+#define REL_ERROR (1.0e-4)
 
     void BIQUADF32::test_biquad_cascade_df1_ref()
     {
@@ -118,9 +117,12 @@ a double precision computation.
            The filter is initialized with the coefs, blockSize and numTaps.
 
            */
-#if !defined(ARM_MATH_NEON) 
+// Neon disabled because this filter has issues with Neon.
+// So test changed to use the scalar version
+//#if !defined(ARM_MATH_NEON) 
            arm_biquad_cascade_df2T_init_f32(&this->Sdf2T,3,coefsp,statep);
-#else
+#if 0
+//#else
            float32_t *vecCoefsPtr = vecCoefs.ptr();
 
            // Those Neon coefs must be computed from original coefs
@@ -244,11 +246,13 @@ a double precision computation.
         float32_t *statep = state.ptr();
         const int16_t *configsp = configs.ptr();
 
-#if !defined(ARM_MATH_NEON) 
+// Neon version has issues and is disabled.
+// So the test is not using Neon version
+//#if !defined(ARM_MATH_NEON) 
         const float32_t *coefsp = coefs.ptr();
-#else
-        float32_t *coefsp = coefs.ptr();
-#endif
+//#else
+//        float32_t *coefsp = coefs.ptr();
+//#endif
 
         const float32_t *inputp = inputs.ptr();
         float32_t *outp = output.ptr();
@@ -284,9 +288,10 @@ a double precision computation.
            The filter is initialized with the coefs, blockSize and numTaps.
 
            */
-#if !defined(ARM_MATH_NEON) 
+//#if !defined(ARM_MATH_NEON) 
            arm_biquad_cascade_df2T_init_f32(&this->Sdf2T,numStages,coefsp,statep);
-#else
+//#else
+#if 0
            float32_t *vecCoefsPtr = vecCoefs.ptr();
 
            // Those Neon coefs must be computed from original coefs