vshl() will call the correct implementation

This commit is contained in:
Konstantinos Margaritis 2025-06-05 20:05:38 +00:00
parent 213ac75e0e
commit 7d9a543100

View File

@ -265,19 +265,7 @@ const u8 *shuftiDoubleExecReal(m128 mask1_lo, m128 mask1_hi, m128 mask2_lo, m128
d = ROUNDUP_PTR(d, S); d = ROUNDUP_PTR(d, S);
ptrdiff_t offset = d - buf; ptrdiff_t offset = d - buf;
first_char_mask.print8("inout_c1"); first_char_mask.print8("inout_c1");
if constexpr (S == 16) { first_char_mask = first_char_mask.vshl(S - offset);
first_char_mask = first_char_mask.vshl_128(S - offset);
}
#ifdef HAVE_SIMD_256_BITS
else if constexpr (S == 32) {
first_char_mask = first_char_mask.vshl_256(S - offset);
}
#endif
#ifdef HAVE_SIMD_512_BITS
else if constexpr (S == 64) {
first_char_mask = first_char_mask.vshl_512(S - offset);
}
#endif
first_char_mask.print8("inout_c1 shifted"); first_char_mask.print8("inout_c1 shifted");
} }