1; RUN: llc -mtriple=amdgcn--amdpal -mcpu=gfx900 -verify-machineinstrs <%s | FileCheck -check-prefixes=GCN %s 2; 3; This test checks that we have the correct fold for zext(cc1) - zext(cc2). 4; 5; GCN-LABEL: sub_zext_zext: 6; GCN: ds_read_b32 [[VAL:v[0-9]+]], 7; GCN: v_cmp_lt_f32{{.*}} vcc, 0, [[VAL]] 8; GCN: v_cndmask_{{.*}} [[ZEXTCC1:v[0-9]+]], 0, 1, vcc 9; GCN: v_cmp_gt_f32{{.*}} vcc, 0, [[VAL]] 10; GCN: v_subbrev{{.*}} {{v[0-9]+}}, vcc, 0, [[ZEXTCC1]], vcc 11; 12; Before the reversion that this test is attached to, the compiler commuted 13; the operands to the sub and used different logic to select the addc/subc 14; instruction: 15; sub zext (setcc), x => addcarry 0, x, setcc 16; sub sext (setcc), x => subcarry 0, x, setcc 17; 18; ... but that is bogus. I believe it is not possible to fold those commuted 19; patterns into any form of addcarry or subcarry. 20 21define amdgpu_cs float @sub_zext_zext() { 22.entry: 23 24 %t519 = load float, float addrspace(3)* null 25 26 %t524 = fcmp ogt float %t519, 0.000000e+00 27 %t525 = fcmp olt float %t519, 0.000000e+00 28 %t526 = zext i1 %t524 to i32 29 %t527 = zext i1 %t525 to i32 30 %t528 = sub nsw i32 %t526, %t527 31 %t529 = sitofp i32 %t528 to float 32 ret float %t529 33} 34 35