[NFC] Fix cvt_f32_ubyte.ll test.

[dfukalov]

Summary

Remove (unintended) infinite loop in the test.

[vangthao]

LGTM

llvm/test/CodeGen/AMDGPU/cvt_f32_ubyte.ll

 ; SI-NEXT:    v_cvt_u32_f32_e32 v0, v0
 ; SI-NEXT:    s_waitcnt vmcnt(2)
 ; SI-NEXT:    buffer_store_byte v1, off, s[0:3], 0
 ; SI-NEXT:    s_waitcnt vmcnt(2)
 ; SI-NEXT:    buffer_store_byte v2, off, s[0:3], 0
 ; SI-NEXT:    s_waitcnt vmcnt(2)
 ; SI-NEXT:    buffer_store_byte v3, off, s[0:3], 0
 ; SI-NEXT:    buffer_store_byte v0, off, s[0:3], 0
-; SI-NEXT:  .LBB40_1: ; %for.body.i
-; SI-NEXT:    ; =>This Inner Loop Header: Depth=1
-; SI-NEXT:    s_branch .LBB40_1
+; SI-NEXT:    s_endpgm
 ;
 ; VI-LABEL: cvt_f32_ubyte0_vector:
 ; VI:       ; %bb.0: ; %entry
 ; VI-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x0
 ; VI-NEXT:    s_mov_b32 s3, 0xf000
 ; VI-NEXT:    s_mov_b32 s2, -1
 ; VI-NEXT:    s_waitcnt lgkmcnt(0)
 ; VI-NEXT:    buffer_load_ubyte v0, off, s[0:3], 0 offset:3

 ; VI-NEXT:    v_cvt_i32_f32_e32 v0, v0
 ; VI-NEXT:    s_waitcnt vmcnt(2)
 ; VI-NEXT:    buffer_store_byte v1, off, s[0:3], 0
 ; VI-NEXT:    s_waitcnt vmcnt(2)
 ; VI-NEXT:    buffer_store_byte v2, off, s[0:3], 0
 ; VI-NEXT:    s_waitcnt vmcnt(2)
 ; VI-NEXT:    buffer_store_byte v3, off, s[0:3], 0
 ; VI-NEXT:    buffer_store_byte v0, off, s[0:3], 0
-; VI-NEXT:  .LBB40_1: ; %for.body.i
-; VI-NEXT:    ; =>This Inner Loop Header: Depth=1
-; VI-NEXT:    s_branch .LBB40_1
+; VI-NEXT:    s_endpgm
 ;
 ; GFX10-LABEL: cvt_f32_ubyte0_vector:
 ; GFX10:       ; %bb.0: ; %entry
 ; GFX10-NEXT:    s_load_dwordx2 s[0:1], s[0:1], 0x0
 ; GFX10-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX10-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX10-NEXT:    s_clause 0x3
 ; GFX10-NEXT:    global_load_ubyte v1, v0, s[0:1] offset:3

 ; GFX10-NEXT:    v_cvt_i32_f32_e32 v0, v0
 ; GFX10-NEXT:    s_waitcnt vmcnt(2)
 ; GFX10-NEXT:    global_store_byte v[0:1], v2, off
 ; GFX10-NEXT:    s_waitcnt vmcnt(1)
 ; GFX10-NEXT:    global_store_byte v[0:1], v3, off
 ; GFX10-NEXT:    s_waitcnt vmcnt(0)
 ; GFX10-NEXT:    global_store_byte v[0:1], v4, off
 ; GFX10-NEXT:    global_store_byte v[0:1], v0, off
-; GFX10-NEXT:  .LBB40_1: ; %for.body.i
-; GFX10-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX10-NEXT:    s_branch .LBB40_1
+; GFX10-NEXT:    s_endpgm
 ;
 ; GFX9-LABEL: cvt_f32_ubyte0_vector:
 ; GFX9:       ; %bb.0: ; %entry
-; GFX9-NEXT:  .LBB40_1: ; %for.body.i
-; GFX9-NEXT:    ; =>This Inner Loop Header: Depth=1
 ; GFX9-NEXT:    global_load_dwordx2 v[0:1], v[0:1], off
 ; GFX9-NEXT:    s_waitcnt vmcnt(0)
 ; GFX9-NEXT:    global_load_ubyte v2, v[0:1], off offset:3
 ; GFX9-NEXT:    global_load_dword v3, v[0:1], off
 ; GFX9-NEXT:    global_load_ubyte v4, v[0:1], off offset:2
 ; GFX9-NEXT:    global_load_ubyte v5, v[0:1], off offset:1
 ; GFX9-NEXT:    global_load_ubyte v6, v[0:1], off
 ; GFX9-NEXT:    s_waitcnt vmcnt(4)
 ; GFX9-NEXT:    v_cvt_f32_ubyte0_e32 v0, v2
 ; GFX9-NEXT:    s_waitcnt vmcnt(3)
 ; GFX9-NEXT:    v_fma_f32 v0, v3, v0, 0.5
 ; GFX9-NEXT:    v_cvt_i32_f32_e32 v0, v0
 ; GFX9-NEXT:    s_waitcnt vmcnt(2)
 ; GFX9-NEXT:    global_store_byte v[0:1], v4, off
 ; GFX9-NEXT:    s_waitcnt vmcnt(2)
 ; GFX9-NEXT:    global_store_byte v[0:1], v5, off
 ; GFX9-NEXT:    s_waitcnt vmcnt(2)
 ; GFX9-NEXT:    global_store_byte v[0:1], v6, off
 ; GFX9-NEXT:    global_store_byte v[0:1], v0, off
-; GFX9-NEXT:    s_branch .LBB40_1
+; GFX9-NEXT:    s_endpgm
 ;
 ; GFX11-LABEL: cvt_f32_ubyte0_vector:
 ; GFX11:       ; %bb.0: ; %entry
 ; GFX11-NEXT:    s_load_b64 s[0:1], s[0:1], 0x0
 ; GFX11-NEXT:    v_mov_b32_e32 v0, 0
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-NEXT:    s_clause 0x3
 ; GFX11-NEXT:    global_load_u8 v1, v0, s[0:1] offset:3
 ; GFX11-NEXT:    global_load_u8 v2, v0, s[0:1] offset:2
 ; GFX11-NEXT:    global_load_u8 v3, v0, s[0:1] offset:1
 ; GFX11-NEXT:    global_load_u8 v0, v0, s[0:1]
 ; GFX11-NEXT:    s_load_b32 s0, s[0:1], 0x0
 ; GFX11-NEXT:    s_waitcnt vmcnt(3)
 ; GFX11-NEXT:    v_cvt_f32_ubyte0_e32 v1, v1
 ; GFX11-NEXT:    s_waitcnt lgkmcnt(0)
 ; GFX11-NEXT:    s_delay_alu instid0(VALU_DEP_1) | instskip(NEXT) | instid1(VALU_DEP_1)
 ; GFX11-NEXT:    v_fma_f32 v1, s0, v1, 0.5
 ; GFX11-NEXT:    v_cvt_i32_f32_e32 v1, v1
 ; GFX11-NEXT:    s_waitcnt vmcnt(0)
 ; GFX11-NEXT:    s_clause 0x3
 ; GFX11-NEXT:    global_store_b8 v[0:1], v2, off
 ; GFX11-NEXT:    global_store_b8 v[0:1], v3, off
 ; GFX11-NEXT:    global_store_b8 v[0:1], v0, off
 ; GFX11-NEXT:    global_store_b8 v[0:1], v1, off
-; GFX11-NEXT:  .LBB40_1: ; %for.body.i
-; GFX11-NEXT:    ; =>This Inner Loop Header: Depth=1
-; GFX11-NEXT:    s_branch .LBB40_1
+; GFX11-NEXT:    s_sendmsg sendmsg(MSG_DEALLOC_VGPRS)
+; GFX11-NEXT:    s_endpgm
 entry:
   br label %for.body.i
 
 for.body.i:                                       ; preds = %for.body.i, %entry
   %retval.sroa.0.0.copyload = load %Vec*, %Vec* addrspace(1)* undef, align 8
   %add.ptr = getelementptr inbounds %Vec, %Vec* %retval.sroa.0.0.copyload, i64 undef
   %retval.sroa.0.0..sroa_cast = bitcast %Vec* %add.ptr to i32*
   %retval.sroa.0.0..sroa_cast_adr = addrspacecast i32* %retval.sroa.0.0..sroa_cast to i32 addrspace(1)*
for.body.i:                                       ; preds = %for.body.i, %entry
   %retval.sroa.4.0.insert.shift = shl nuw i32 %retval.sroa.4.0.insert.ext, 24
   %retval.sroa.3.0.insert.ext = and i32 %retval.sroa.0.0.copyload.i, 16711680
   %retval.sroa.3.0.insert.insert = or i32 %retval.sroa.4.0.insert.shift, %retval.sroa.3.0.insert.ext
   %retval.sroa.2.0.insert.ext = and i32 %retval.sroa.0.0.copyload.i, 65280
   %retval.sroa.2.0.insert.insert = or i32 %retval.sroa.3.0.insert.insert, %retval.sroa.2.0.insert.ext
   %retval.sroa.0.0.insert.ext = and i32 %retval.sroa.0.0.copyload.i, 255
   %retval.sroa.0.0.insert.insert = or i32 %retval.sroa.2.0.insert.insert, %retval.sroa.0.0.insert.ext
   store i32 %retval.sroa.0.0.insert.insert, i32 addrspace(1)* undef, align 1
-  br label %for.body.i
+  ret void
 }