This test bb-slp-77.c extracted relies on the completely unrolling of the inner loop. However, for x86 in gcc.dg/vect/, loop is not unrolled and the inner loop is vectorized thus not triggering expected BB SLP
Also noticed that the "vectorizing stmts using SLP” count is different when I force the loop to unroll for x86. Thus, to keep it simple, moving the test to gcc.target/aarch64. Regression tested on aarch64-linux-gnu. Is this OK? Thanks, Kugan
0001-testsuite-Fix-bb-slp-77.c.patch
Description: 0001-testsuite-Fix-bb-slp-77.c.patch