1 ; RUN: llc -march=r600 -mcpu=redwood < %s | FileCheck -check-prefix=R600 %s
2 ; RUN: llc -march=r600 -mcpu=SI -verify-machineinstrs < %s | FileCheck -check-prefix=SI %s
4 ; These tests check that fdiv is expanded correctly and also test that the
5 ; scheduler is scheduling the RECIP_IEEE and MUL_IEEE instructions in separate
8 ; FUNC-LABEL: @fdiv_f32
9 ; R600-DAG: RECIP_IEEE * T{{[0-9]+\.[XYZW]}}, KC0[3].Z
10 ; R600-DAG: RECIP_IEEE * T{{[0-9]+\.[XYZW]}}, KC0[3].Y
11 ; R600-DAG: MUL_IEEE {{\** *}}T{{[0-9]+\.[XYZW]}}, KC0[3].X, PS
12 ; R600-DAG: MUL_IEEE {{\** *}}T{{[0-9]+\.[XYZW]}}, KC0[2].W, PS
16 define void @fdiv_f32(float addrspace(1)* %out, float %a, float %b) {
18 %0 = fdiv float %a, %b
19 store float %0, float addrspace(1)* %out
25 ; FUNC-LABEL: @fdiv_v2f32
26 ; R600-DAG: RECIP_IEEE * T{{[0-9]+\.[XYZW]}}, KC0[3].Z
27 ; R600-DAG: RECIP_IEEE * T{{[0-9]+\.[XYZW]}}, KC0[3].Y
28 ; R600-DAG: MUL_IEEE {{\** *}}T{{[0-9]+\.[XYZW]}}, KC0[3].X, PS
29 ; R600-DAG: MUL_IEEE {{\** *}}T{{[0-9]+\.[XYZW]}}, KC0[2].W, PS
35 define void @fdiv_v2f32(<2 x float> addrspace(1)* %out, <2 x float> %a, <2 x float> %b) {
37 %0 = fdiv <2 x float> %a, %b
38 store <2 x float> %0, <2 x float> addrspace(1)* %out
42 ; FUNC-LABEL: @fdiv_v4f32
43 ; R600-DAG: RECIP_IEEE * T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
44 ; R600-DAG: RECIP_IEEE * T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
45 ; R600-DAG: RECIP_IEEE * T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
46 ; R600-DAG: RECIP_IEEE * T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}
47 ; R600-DAG: MUL_IEEE {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}, PS
48 ; R600-DAG: MUL_IEEE {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}, PS
49 ; R600-DAG: MUL_IEEE {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}, PS
50 ; R600-DAG: MUL_IEEE {{\** *}}T{{[0-9]+\.[XYZW], T[0-9]+\.[XYZW]}}, PS
60 define void @fdiv_v4f32(<4 x float> addrspace(1)* %out, <4 x float> addrspace(1)* %in) {
61 %b_ptr = getelementptr <4 x float> addrspace(1)* %in, i32 1
62 %a = load <4 x float> addrspace(1) * %in
63 %b = load <4 x float> addrspace(1) * %b_ptr
64 %result = fdiv <4 x float> %a, %b
65 store <4 x float> %result, <4 x float> addrspace(1)* %out