Fixed the stack overflow and removed the workaround for it

CMakeLists.txt

@@ -163,6 +163,6 @@ target_link_libraries(rpi_inmp441_kws
         CMSISDSP
         )
 
-#target_compile_definitions(rpi_inmp441_kws PRIVATE PICO_STACK_SIZE=0x1000)
+target_compile_definitions(rpi_inmp441_kws PRIVATE PICO_CORE1_STACK_SIZE=0x1000)
 pico_add_extra_outputs(rpi_inmp441_kws)
 

src/fast_grnn.c

@@ -163,9 +163,9 @@ void sha_rnn_get_max_logit(const sha_rnn_output_t input, float *max_logit, size_
 
 void sha_rnn_process(const sha_rnn_input_t input, float *max_logit, size_t *max_idx)
 {
-    float output[64] = {0.0f};
-    float output2[32] = {0.0f};
-    float output3[6] = {0.0f};
+    static float output[64] = {0.0f};
+    static float output2[32] = {0.0f};
+    static float output3[6] = {0.0f};
 
     sha_rnn_rnn0_process(input, output);
     sha_rnn_rnn1_process(output, output2);

src/fbank.c

@@ -271,8 +271,10 @@ static void preemphasis(float *input, size_t len)
 
 static void powspec(const float input[FRAME_LEN], float output[NUM_FFT_BINS])
 {
-    float32_t tmp[NUM_FFT] = {0.0f};
-    float32_t out[NUM_FFT];
+    static float32_t tmp[NUM_FFT];
+    static float32_t out[NUM_FFT];
+
+    memset(tmp, 0, sizeof(tmp));
 
     for (size_t i = 0; i < FRAME_LEN; i++)
     {

src/rpi_inmp441_kws.c

@@ -124,8 +124,6 @@ static void(core1_entry)(void)
     }
 }
 
-static uint32_t core1_stack[8 * 1024UL];
-
 static float fbins_out[SHARNN_BRICK_SIZE][NUM_FILT];
 
 int main()
@@ -147,7 +145,7 @@ int main()
     }
     printf("\n");
 
-    multicore_launch_core1_with_stack(core1_entry, core1_stack, sizeof(core1_stack));
+    multicore_launch_core1(core1_entry);
 
     fbank_init();