On 11/21/13 04:20, Richard Biener wrote:
This moves another fold-const.c folding to the GIMPLE level.
In PR59058 it was noticed we fail to optimize
Well, you duplicated the optimization, you didn't move it :(
Of course the latter takes a lot more time to verify that nothing regresses.
Anyway,
This moves another fold-const.c folding to the GIMPLE level.
In PR59058 it was noticed we fail to optimize
vect_vec_iv_.16_57 = VIEW_CONVERT_EXPR(vect_vec_iv_.15_55);
vect_b.17_58 = VIEW_CONVERT_EXPR(vect_vec_iv_.16_57);
Bootstrapped and tested on x86_64-unknown-linux-gnu, applied.
Richard.
