Fix uninitialized variables in NEON FFT and matrix multiplication functions (#296)

Initialize the 'step' variable to 0 in both forward and backward FFT functions
to prevent compiler warnings about potentially uninitialized variables.

- arm_ne10_mixed_radix_fft_forward_float32_neon
- arm_ne10_mixed_radix_fft_backward_float32_neon

Initialize 'tw' to 'twiddles' at declaration in both forward and backward
FFT macros to suppress -Wmaybe-uninitialized errors.

Initialize 'temp' variable in arm_mat_mult_f64 NEON implementation to fix
uninitialized variable warning when used in vsetq_lane_f64() calls.

These changes fix compilation errors with GCC 14.3.0 which is stricter
about detecting potentially uninitialized variables.

Signed-off-by: Ryan McClelland <ryanmcclelland@meta.com>

Author: XenuIsWatching

Ne10/NE10_fft_float32.neonintrinsic.c

@@ -1214,7 +1214,7 @@ void arm_ne10_mixed_radix_fft_forward_float32_neon (
     ne10_int32_t fstride = S->factors[1];
     ne10_int32_t mstride = S->factors[3];
     ne10_int32_t first_radix = S->factors[2];
-    ne10_int32_t step, f_count;
+    ne10_int32_t step = 0, f_count;
     const ne10_fft_cpx_float32_t *src = input;
     ne10_fft_cpx_float32_t *in;
     ne10_fft_cpx_float32_t *dst = out;
@@ -1313,7 +1313,7 @@ void arm_ne10_mixed_radix_fft_backward_float32_neon (
     ne10_int32_t mstride = S->factors[3];
     ne10_int32_t first_radix = S->factors[2];
     ne10_int32_t nfft = fstride * first_radix;
-    ne10_int32_t step, f_count;
+    ne10_int32_t step = 0, f_count;
     const ne10_fft_cpx_float32_t *src = input;
     ne10_fft_cpx_float32_t *in;
     ne10_fft_cpx_float32_t *dst = out;

Ne10/NE10_fft_int16.neonintrinsic.c

@@ -1026,7 +1026,7 @@ static void arm_ne10_mixed_radix_fft_forward_int16_##scaled##_neon (ne10_fft_cpx
     ne10_fft_cpx_int16_t *Fout1; \
     ne10_fft_cpx_int16_t   *Fout_ls = Fout; \
     ne10_fft_cpx_int16_t   *Ftmp; \
-    const ne10_fft_cpx_int16_t   *tw; \
+    const ne10_fft_cpx_int16_t   *tw = twiddles; \
     const ne10_fft_cpx_int16_t *tw1; \
  \
     /* init fstride, mstride, N */ \
@@ -1111,7 +1111,7 @@ static void arm_ne10_mixed_radix_fft_backward_int16_##scaled##_neon (ne10_fft_cp
     ne10_fft_cpx_int16_t *Fout1; \
     ne10_fft_cpx_int16_t   *Fout_ls = Fout; \
     ne10_fft_cpx_int16_t   *Ftmp; \
-    const ne10_fft_cpx_int16_t   *tw;\
+    const ne10_fft_cpx_int16_t   *tw = twiddles;\
     const ne10_fft_cpx_int16_t *tw1; \
  \
     /* init fstride, mstride, N */ \

Ne10/NE10_fft_int32.neonintrinsic.c

@@ -1267,7 +1267,7 @@ static void arm_ne10_mixed_radix_fft_forward_int32_##scaled##_neon (ne10_fft_cpx
     ne10_fft_cpx_int32_t *Fout1; \
     ne10_fft_cpx_int32_t   *Fout_ls = Fout; \
     ne10_fft_cpx_int32_t   *Ftmp; \
-    const ne10_fft_cpx_int32_t   *tw;\
+    const ne10_fft_cpx_int32_t   *tw = twiddles;\
     const ne10_fft_cpx_int32_t *tw1; \
  \
     /* init fstride, mstride, N */ \
@@ -1353,7 +1353,7 @@ static void arm_ne10_mixed_radix_fft_backward_int32_##scaled##_neon (ne10_fft_cp
     ne10_fft_cpx_int32_t   *Fout1; \
     ne10_fft_cpx_int32_t   *Fout_ls = Fout; \
     ne10_fft_cpx_int32_t   *Ftmp; \
-    const ne10_fft_cpx_int32_t   *tw; \
+    const ne10_fft_cpx_int32_t   *tw = twiddles; \
     const  ne10_fft_cpx_int32_t *tw1; \
  \
     /* init fstride, mstride, N */ \

Source/MatrixFunctions/arm_mat_mult_f64.c

@@ -170,6 +170,7 @@ ARM_DSP_ATTRIBUTE arm_status arm_mat_mult_f64(
         acc5 = vdupq_n_f64(0.0);
         acc6 = vdupq_n_f64(0.0);
         acc7 = vdupq_n_f64(0.0);
+        temp = vdupq_n_f64(0.0);
 
         /* Compute 2 MACs simultaneously. */
         colCnt = numColsA >> 1U;
@@ -307,6 +308,7 @@ ARM_DSP_ATTRIBUTE arm_status arm_mat_mult_f64(
         pIn1 = pInA;
 
         acc0 = vdupq_n_f64(0.0);
+        temp = vdupq_n_f64(0.0);
 
         /* Compute 4 MACs simultaneously. */
         colCnt = numColsA >> 1U;