https://gcc.gnu.org/bugzilla/show_bug.cgi?id=113326
Bug ID: 113326 Summary: Optimize vector shift with constant delta on shifting-count operand Product: gcc Version: unknown Status: UNCONFIRMED Severity: normal Priority: P3 Component: tree-optimization Assignee: unassigned at gcc dot gnu.org Reporter: fxue at os dot amperecomputing.com Target Milestone: --- For shift by induction variable, loop vectorization could generate a series of vector shifts, whose shifting-count operands are advanced by constant delta. We could use result of the first vector shift as base, and transform others to be vector shift on the result by scalar delta, which would eliminate non-uniform shift vector count constant, and thereby save a load on most architectures. int test(int array[16]); int foo(int value) { int array[16]; for (int i = 0; i < 16; i++) array[i] = value >> i; return test(array); } // Current vect_value = { value, value, value, value }; vect_shift_0 = vect_value >> { 0, 1, 2, 3 }; vect_shift_1 = vect_value >> { 4, 5, 6, 7 }; vect_shift_2 = vect_value >> { 8, 9, 10, 11 }; vect_shift_3 = vect_value >> { 12, 13, 14, 15 }; // To be optimized to vect_value = { value, value, value, value }; vect_shift_0 = vect_value >> { 0, 1, 2, 3 }; vect_shift_1 = vect_shift_0 >> { 4, 4, 4, 4 }; vect_shift_2 = vect_shift_0 >> { 8, 8, 8, 8 }; vect_shift_3 = vect_shift_0 >> { 12, 12, 12, 12 };