Home | History | Annotate | Download | only in AMDGPU
      1 ; RUN: llc -march=amdgcn -mcpu=verde -enable-misched=0 -post-RA-scheduler=0 < %s | FileCheck %s
      2 ; RUN: llc -regalloc=basic -march=amdgcn -mcpu=tonga -enable-misched=0 -post-RA-scheduler=0 < %s | FileCheck %s
      3  ;
      4 ; There is something about Tonga that causes this test to spend a lot of time
      5 ; in the default register allocator.
      6 
      7 
      8 ; When the offset of VGPR spills into scratch space gets too large, an additional SGPR
      9 ; is used to calculate the scratch load/store address. Make sure that this
     10 ; mechanism works even when many spills happen.
     11 
     12 ; Just test that it compiles successfully.
     13 ; CHECK-LABEL: test
     14 define amdgpu_kernel void @test(<1280 x i32> addrspace(1)* %out, <1280 x i32> addrspace(1)* %in) {
     15 entry:
     16   %lo = call i32 @llvm.amdgcn.mbcnt.lo(i32 -1, i32 0)
     17   %tid = call i32 @llvm.amdgcn.mbcnt.hi(i32 -1, i32 %lo)
     18 
     19   %aptr = getelementptr <1280 x i32>, <1280 x i32> addrspace(1)* %in, i32 %tid
     20   %a = load <1280 x i32>, <1280 x i32> addrspace(1)* %aptr
     21 
     22 ; mark most VGPR registers as used to increase register pressure
     23   call void asm sideeffect "", "~{v4},~{v8},~{v12},~{v16},~{v20},~{v24},~{v28},~{v32}" ()
     24   call void asm sideeffect "", "~{v36},~{v40},~{v44},~{v48},~{v52},~{v56},~{v60},~{v64}" ()
     25   call void asm sideeffect "", "~{v68},~{v72},~{v76},~{v80},~{v84},~{v88},~{v92},~{v96}" ()
     26   call void asm sideeffect "", "~{v100},~{v104},~{v108},~{v112},~{v116},~{v120},~{v124},~{v128}" ()
     27   call void asm sideeffect "", "~{v132},~{v136},~{v140},~{v144},~{v148},~{v152},~{v156},~{v160}" ()
     28   call void asm sideeffect "", "~{v164},~{v168},~{v172},~{v176},~{v180},~{v184},~{v188},~{v192}" ()
     29   call void asm sideeffect "", "~{v196},~{v200},~{v204},~{v208},~{v212},~{v216},~{v220},~{v224}" ()
     30 
     31   %outptr = getelementptr <1280 x i32>, <1280 x i32> addrspace(1)* %out, i32 %tid
     32   store <1280 x i32> %a, <1280 x i32> addrspace(1)* %outptr
     33 
     34   ret void
     35 }
     36 
     37 declare i32 @llvm.amdgcn.mbcnt.lo(i32, i32) #1
     38 declare i32 @llvm.amdgcn.mbcnt.hi(i32, i32) #1
     39 
     40 attributes #1 = { nounwind readnone }
     41