Looks like a Poly1305 issue still:

./configure --enable-chacha --enable-poly1305 --enable-debug --enable-intelasm --disable-aesgcm && make
./wolfcrypt/test/testwolfcrypt
Chacha test passed!
POLY1305 test failed!
error = -3614

./wolfcrypt/benchmark/benchmark
wolfCrypt Benchmark (block bytes 1048576, min 1.0 sec each)
RNG 100 MB took 1.021 seconds, 97.984 MB/s Cycles per byte = 34.96
AES-Enc 845 MB took 1.000 seconds, 844.768 MB/s Cycles per byte = 4.05
AES-Dec 5855 MB took 1.000 seconds, 5854.789 MB/s Cycles per byte = 0.59
CHACHA 3215 MB took 1.001 seconds, 3211.548 MB/s Cycles per byte = 1.07

Program received signal SIGSEGV, Segmentation fault.
0x000000000043a271 in poly1305_blocks_avx2 (ctx=0x7fffffffdf50, m=0x7ffff73fd010 "\223\032\005\004K\311\033T\266\222\220U"(@߉\O\002\245\356\300\276jIc3\035\202\223\215\373c", bytes=1048576)
at wolfcrypt/src/poly1305.c:741
741 asm volatile (
(gdb) bt
#0 0x000000000043a271 in poly1305_blocks_avx2 (ctx=0x7fffffffdf50, m=0x7ffff73fd010 "\223\032\005\004K\311\033T\266\222\220U"(@߉\O\002\245\356\300\276jIc3\035\202\223\215\373c", bytes=1048576)
at wolfcrypt/src/poly1305.c:741
#1 0x000000000043cbb7 in wc_Poly1305Update (ctx=0x7fffffffdf50, m=0x7ffff73fd010 "\223\032\005\004K\311\033T\266\222\220U"(@߉\O\002\245\356\300\276jIc3\035\202\223\215\373c", bytes=1048576)
at wolfcrypt/src/poly1305.c:1602
#2 0x0000000000444a70 in calculateAuthTag (inAuthKey=0x7fffffffe210 "\021\231g-\027\224 \370\342W\vTtO\375s'", inAAD=0x0, inAADLen=0,
inCiphertext=0x7ffff73fd010 "\223\032\005\004K\311\033T\266\222\220U"(@߉\O\002\245\356\300\276jIc3\035\202\223\215\373c", inCiphertextLen=1048576, outAuthTag=0x7fffffffe270 "")
at wolfcrypt/src/chacha20_poly1305.c:218
#3 0x0000000000444790 in wc_ChaCha20Poly1305_Encrypt (
inKey=0x476930 <bench_key_buf> "\001#Eg\211\253\315\357\376\336\272\230vT2\020\211\253\315\357\001#Eg\001#Eg\211\253\315\357\022\064Vx\220\253\315\357\001\001\001\001\001\001\001\001\021!1AQaq\201MB",
inIV=0x476950 <bench_iv_buf> "\022\064Vx\220\253\315\357\001\001\001\001\001\001\001\001\021!1AQaq\201MB", inAAD=0x0, inAADLen=0,
inPlaintext=0x7ffff7ee8010 "\314YA\005\342}\217\364\001\217\036@\302V\024\064\314YA\005\342}\217\364\001\217\036@\302V\024\064\314YA\005\342}\217\364\001\217\036@\302V\024\064\314YA\005\342}\217\364\001\217\036@\302V\024\064\314YA\005\342}\217\364\001\217\036@\302V\024\064\314YA\005\342}\217\364\001\217\036@\302V\024\064\314YA\005\342}\217\364\001\217\036@\302V\024\064\314YA\005\342}\217\364\001\217\036@\302V\024\064\314YA\005\342}\217\364\001\217\036@\302V\024\064\314YA\005\342}\217\364\001\217\036@\302V\024\064\314YA\005\342}\217\364\001\217\036@\302V\024\064\314YA\005\342}\217\364\001\217\036@\302V\024\064\314YA\005\342}\217", <incomplete sequence \364>..., inPlaintextLen=1048576, outCiphertext=0x7ffff73fd010 "\223\032\005\004K\311\033T\266\222\220U"(@߉\O\002\245\356\300\276jIc3\035\202\223\215\373c",
outAuthTag=0x7fffffffe270 "") at wolfcrypt/src/chacha20_poly1305.c:100
#4 0x0000000000402380 in bench_chacha20_poly1305_aead () at wolfcrypt/benchmark/benchmark.c:1667
#5 0x00000000004018ad in benchmarks_do (args=0x0) at wolfcrypt/benchmark/benchmark.c:721
#6 0x0000000000401b55 in benchmark_test (args=0x0) at wolfcrypt/benchmark/benchmark.c:1058
#7 0x0000000000405413 in main (argc=1, argv=0x7fffffffe418) at wolfcrypt/benchmark/benchmark.c:3811

Fixed the issue!

Poly1305 now passes on CentOS / GCC 7.1 and Mac Clang using AVX/AVX2 speedups.