https://gcc.gnu.org/bugzilla/show_bug.cgi?id=47860
Andrew Pinski <pinskia at gcc dot gnu.org> changed:
What |Removed |Added
----------------------------------------------------------------------------
Ever confirmed|0 |1
Status|UNCONFIRMED |NEW
Last reconfirmed| |2021-08-16
--- Comment #6 from Andrew Pinski <pinskia at gcc dot gnu.org> ---
Confirmed, ICC is able to vectorize this loop even without AVX (GCC can do the
vectorize the loop currently with AVX).
movdqa %xmm0, %xmm11 #10.11
lea 1(%r14), %r15d #9.31
movups (%rdx,%r15,8), %xmm9 #9.27
movups (%rcx,%r14,8), %xmm10 #10.24
cmpltpd %xmm1, %xmm10 #10.24
pxor %xmm2, %xmm10 #10.24
movmskpd %xmm10, %r15d #10.24
testl %r15d, %r15d #10.24
je ..B1.14 # Prob 50% #10.24
# LOE rax rdx rcx rbx rsi rdi ebp r8d r9d r10d
r11d r12d r13d r14d xmm0 xmm1 xmm2 xmm3 xmm4 xmm5 xmm6 xmm7 xmm8 xmm9 xmm10
xmm11
..B1.13: # Preds ..B1.12
# Execution count [1.25e+01]
pshufd $8, %xmm10, %xmm11 #10.24
movaps %xmm9, %xmm8 #5.21
pand %xmm6, %xmm11 #10.24
# LOE rax rdx rcx rbx rsi rdi ebp r8d r9d r10d
r11d r12d r13d r14d xmm0 xmm1 xmm2 xmm3 xmm4 xmm5 xmm6 xmm7 xmm8 xmm11
..B1.14: # Pre
