Files
clang-p2996/llvm/test/CodeGen/AMDGPU/optimize-negated-cond.ll
Jingu Kang ff68e43c81 [MachineLICM] Handle Subloops
It is a re-commit from reverted commit 3454cf67bd.

Following discussion on https://reviews.llvm.org/D154205, make MachineLICM pass
handle subloops with only visiting outermost loop's blocks once.

Differential Revision: https://reviews.llvm.org/D154205
2023-09-26 14:25:11 +01:00

78 lines
1.8 KiB
LLVM

; RUN: llc -march=amdgcn -verify-machineinstrs < %s | FileCheck -check-prefix=GCN %s
; GCN-LABEL: {{^}}negated_cond:
; GCN: .LBB0_2:
; GCN: v_cndmask_b32_e64
; GCN: v_cmp_ne_u32_e64
define amdgpu_kernel void @negated_cond(ptr addrspace(1) %arg1) {
bb:
br label %bb1
bb1:
%tmp1 = load i32, ptr addrspace(1) %arg1
%tmp2 = icmp eq i32 %tmp1, 0
br label %bb2
bb2:
%tmp3 = phi i32 [ 0, %bb1 ], [ %tmp6, %bb4 ]
%tmp4 = shl i32 %tmp3, 5
br i1 %tmp2, label %bb3, label %bb4
bb3:
%tmp5 = add i32 %tmp4, 1
br label %bb4
bb4:
%tmp6 = phi i32 [ %tmp5, %bb3 ], [ %tmp4, %bb2 ]
%gep = getelementptr inbounds i32, ptr addrspace(1) %arg1, i32 %tmp6
store i32 0, ptr addrspace(1) %gep
%tmp7 = icmp eq i32 %tmp6, 32
br i1 %tmp7, label %bb1, label %bb2
}
; GCN-LABEL: {{^}}negated_cond_dominated_blocks:
; GCN: s_cmp_lg_u32
; GCN: s_cselect_b64 [[CC1:[^,]+]], -1, 0
; GCN: s_branch [[BB1:.LBB[0-9]+_[0-9]+]]
; GCN: [[BB0:.LBB[0-9]+_[0-9]+]]
; GCN-NOT: v_cndmask_b32
; GCN-NOT: v_cmp
; GCN: [[BB1]]:
; GCN: s_mov_b64 vcc, [[CC1]]
; GCN: s_cbranch_vccz [[BB2:.LBB[0-9]+_[0-9]+]]
; GCN: s_mov_b64 vcc, exec
; GCN: s_cbranch_execnz [[BB0]]
; GCN: [[BB2]]:
define amdgpu_kernel void @negated_cond_dominated_blocks(ptr addrspace(1) %arg1) {
bb:
br label %bb2
bb2:
%tmp1 = load i32, ptr addrspace(1) %arg1
%tmp2 = icmp eq i32 %tmp1, 0
br label %bb4
bb3:
ret void
bb4:
%tmp3 = phi i32 [ 0, %bb2 ], [ %tmp7, %bb7 ]
%tmp4 = shl i32 %tmp3, 5
br i1 %tmp2, label %bb5, label %bb6
bb5:
%tmp5 = add i32 %tmp4, 1
br label %bb7
bb6:
%tmp6 = add i32 %tmp3, 1
br label %bb7
bb7:
%tmp7 = phi i32 [ %tmp5, %bb5 ], [ %tmp6, %bb6 ]
%gep = getelementptr inbounds i32, ptr addrspace(1) %arg1, i32 %tmp7
store i32 0, ptr addrspace(1) %gep
%tmp8 = icmp eq i32 %tmp7, 32
br i1 %tmp8, label %bb3, label %bb4
}