Speed up sha512 on x86

Enable avx version on amd, since sha-ni sha512 doesn't exist

Change-Id: I2ea4b74c7995ccae6b673adefc4b3e25d515e321
Reviewed-on: https://boringssl-review.googlesource.com/c/boringssl/+/73567
Reviewed-by: David Benjamin <davidben@google.com>
Commit-Queue: David Benjamin <davidben@google.com>

crypto/fipsmodule/sha/internal.h

diff --git a/crypto/fipsmodule/sha/internal.h b/crypto/fipsmodule/sha/internal.h
index d2a4269..24a9230 100644
--- a/crypto/fipsmodule/sha/internal.h
+++ b/crypto/fipsmodule/sha/internal.h
@@ -196,9 +196,7 @@
 
 #define SHA512_ASM_AVX
 OPENSSL_INLINE int sha512_avx_capable(void) {
-  // Pre-Zen AMD CPUs had slow SHLD/SHRD; Zen added the SHA extension; see the
-  // discussion in sha1-586.pl.
-  return CRYPTO_is_AVX_capable() && CRYPTO_is_intel_cpu();
+  return CRYPTO_is_AVX_capable();
 }
 void sha512_block_data_order_avx(uint64_t state[8], const uint8_t *data,
                                  size_t num);