remove buggy and slow neonv8 kernel

as noticed by argilo when fixing the integer generation, that kernel was buggy. It seems compilers are better at building byte-swapping code than people writing SIMD intrinsics, so falling back on generic doesn't hurt Signed-off-by: Marcus Müller <[email protected]>
gnuradio · Oct 23, 2023 · 34f4575 · 34f4575
1 parent e853e9b
commit 34f4575
Show file tree

Hide file tree

Showing 2 changed files with 0 additions and 47 deletions.
diff --git a/kernels/volk/volk_64u_byteswap.h b/kernels/volk/volk_64u_byteswap.h
@@ -229,41 +229,6 @@ static inline void volk_64u_byteswap_a_ssse3(uint64_t* intsToSwap,
 #endif /* LV_HAVE_SSSE3 */
 
 
-#ifdef LV_HAVE_NEONV8
-#include <arm_neon.h>
-
-static inline void volk_64u_byteswap_neonv8(uint64_t* intsToSwap, unsigned int num_points)
-{
-    uint32_t* inputPtr = (uint32_t*)intsToSwap;
-    const unsigned int n4points = num_points / 4;
-    uint8x16x2_t input;
-    uint8x16_t idx = { 7, 6, 5, 4, 3, 2, 1, 0, 15, 14, 13, 12, 11, 10, 9, 8 };
-
-    unsigned int number = 0;
-    for (number = 0; number < n4points; ++number) {
-        __VOLK_PREFETCH(inputPtr + 8);
-        input = vld2q_u8((uint8_t*)inputPtr);
-        input.val[0] = vqtbl1q_u8(input.val[0], idx);
-        input.val[1] = vqtbl1q_u8(input.val[1], idx);
-        vst2q_u8((uint8_t*)inputPtr, input);
-
-        inputPtr += 8;
-    }
-
-    for (number = n4points * 4; number < num_points; ++number) {
-        uint32_t output1 = *inputPtr;
-        uint32_t output2 = inputPtr[1];
-
-        output1 = (((output1 >> 24) & 0xff) | ((output1 >> 8) & 0x0000ff00) |
-                   ((output1 << 8) & 0x00ff0000) | ((output1 << 24) & 0xff000000));
-        output2 = (((output2 >> 24) & 0xff) | ((output2 >> 8) & 0x0000ff00) |
-                   ((output2 << 8) & 0x00ff0000) | ((output2 << 24) & 0xff000000));
-
-        *inputPtr++ = output2;
-        *inputPtr++ = output1;
-    }
-}
-#else
 #ifdef LV_HAVE_NEON
 #include <arm_neon.h>
 
@@ -318,7 +283,6 @@ static inline void volk_64u_byteswap_neon(uint64_t* intsToSwap, unsigned int num
     }
 }
 #endif /* LV_HAVE_NEON */
-#endif
 
 #endif /* INCLUDED_volk_64u_byteswap_u_H */
 #ifndef INCLUDED_volk_64u_byteswap_a_H

diff --git a/kernels/volk/volk_64u_byteswappuppet_64u.h b/kernels/volk/volk_64u_byteswappuppet_64u.h
@@ -26,16 +26,6 @@ static inline void volk_64u_byteswappuppet_64u_generic(uint64_t* output,
 }
 #endif
 
-#ifdef LV_HAVE_NEONV8
-static inline void volk_64u_byteswappuppet_64u_neonv8(uint64_t* output,
-                                                      uint64_t* intsToSwap,
-                                                      unsigned int num_points)
-{
-
-    volk_64u_byteswap_neonv8((uint64_t*)intsToSwap, num_points);
-    memcpy((void*)output, (void*)intsToSwap, num_points * sizeof(uint64_t));
-}
-#else
 #ifdef LV_HAVE_NEON
 static inline void volk_64u_byteswappuppet_64u_neon(uint64_t* output,
                                                     uint64_t* intsToSwap,
@@ -46,7 +36,6 @@ static inline void volk_64u_byteswappuppet_64u_neon(uint64_t* output,
     memcpy((void*)output, (void*)intsToSwap, num_points * sizeof(uint64_t));
 }
 #endif
-#endif
 
 #ifdef LV_HAVE_SSE2
 static inline void volk_64u_byteswappuppet_64u_u_sse2(uint64_t* output,