xref: /llvm-project/llvm/test/CodeGen/AArch64/aarch64-sme2-asm.ll (revision 62baf21daa377c4ec1a641b26931063c1117d262)
1; RUN: llc < %s -mtriple aarch64-none-linux-gnu -mattr=+sme2 -force-streaming -stop-after=finalize-isel | FileCheck %s
2
3define void @UphPNR(target("aarch64.svcount") %predcnt) {
4entry:
5; CHECK:  %0:ppr = COPY $p0
6; CHECK:  STR_PXI %0, %stack.0.predcnt.addr, 0 :: (store (<vscale x 1 x s16>) into %ir.predcnt.addr)
7; CHECK:  %1:pnr_p8to15 = COPY %0
8; CHECK:  INLINEASM &"ld1w {z0.s,z1.s,z2.s,z3.s}, $0/z, [x10]", 1 /* sideeffect attdialect */, {{[0-9]+}} /* reguse:PNR_p8to15 */, %1
9; CHECK:  RET_ReallyLR
10  %predcnt.addr = alloca target("aarch64.svcount"), align 2
11  store target("aarch64.svcount") %predcnt, ptr %predcnt.addr, align 2
12  %0 = load target("aarch64.svcount"), ptr %predcnt.addr, align 2
13  call void asm sideeffect "ld1w {z0.s,z1.s,z2.s,z3.s}, $0/z, [x10]", "@3Uph"(target("aarch64.svcount") %0)
14  ret void
15}
16
17define void @UpaPNR(target("aarch64.svcount") %predcnt) {
18entry:
19; CHECK:  %0:ppr = COPY $p0
20; CHECK:  STR_PXI %0, %stack.0.predcnt.addr, 0 :: (store (<vscale x 1 x s16>) into %ir.predcnt.addr)
21; CHECK:  %1:pnr = COPY %0
22; CHECK:  INLINEASM &"ld1w {z0.s,z1.s,z2.s,z3.s}, $0/z, [x10]", 1 /* sideeffect attdialect */, {{[0-9]+}} /* reguse:PNR */, %1
23; CHECK:  RET_ReallyLR
24  %predcnt.addr = alloca target("aarch64.svcount"), align 2
25  store target("aarch64.svcount") %predcnt, ptr %predcnt.addr, align 2
26  %0 = load target("aarch64.svcount"), ptr %predcnt.addr, align 2
27  call void asm sideeffect "ld1w {z0.s,z1.s,z2.s,z3.s}, $0/z, [x10]", "@3Upa"(target("aarch64.svcount") %0)
28  ret void
29}
30
31define void @UplPNR(target("aarch64.svcount") %predcnt) {
32entry:
33; CHECK:  %0:ppr = COPY $p0
34; CHECK:  STR_PXI %0, %stack.0.predcnt.addr, 0 :: (store (<vscale x 1 x s16>) into %ir.predcnt.addr)
35; CHECK:  %1:pnr_3b = COPY %0
36; CHECK:  INLINEASM &"fadd z0.h, $0/m, z0.h, #0.5", 1 /* sideeffect attdialect */, {{[0-9]+}} /* reguse:PNR_3b */, %1
37; CHECK:  RET_ReallyLR
38  %predcnt.addr = alloca target("aarch64.svcount"), align 2
39  store target("aarch64.svcount") %predcnt, ptr %predcnt.addr, align 2
40  %0 = load target("aarch64.svcount"), ptr %predcnt.addr, align 2
41  call void asm sideeffect "fadd z0.h, $0/m, z0.h, #0.5", "@3Upl"(target("aarch64.svcount") %0)
42  ret void
43}
44