[NVPTX] Add missing patterns for div.approx with immediate denominator

[oota-llvm.git] / test / CodeGen / R600 / local-memory.ll

diff --git a/test/CodeGen/R600/local-memory.ll b/test/CodeGen/R600/local-memory.ll
index 9ebb7692a680442e396db0d6a46a1c4955eeb7e1..8db78ad46a6146918f1437f724a03c8a2847676e 100644 (file)
--- a/test/CodeGen/R600/local-memory.ll
+++ b/test/CodeGen/R600/local-memory.ll
@@ -1,18 +1,23 @@
 ; RUN: llc < %s -march=r600 -mcpu=redwood | FileCheck --check-prefix=EG-CHECK %s
-; RUN: llc < %s -march=r600 -mcpu=verde | FileCheck --check-prefix=SI-CHECK %s
+; RUN: llc < %s -march=r600 -mcpu=verde -verify-machineinstrs | FileCheck --check-prefix=SI-CHECK %s
+; RUN: llc < %s -march=r600 -mcpu=bonaire -verify-machineinstrs | FileCheck --check-prefix=CI-CHECK %s
 
-@local_memory.local_mem = internal addrspace(3) unnamed_addr global [16 x i32] zeroinitializer, align 4
+@local_memory.local_mem = internal addrspace(3) unnamed_addr global [128 x i32] zeroinitializer, align 4
 
-; EG-CHECK: @local_memory
-; SI-CHECK: @local_memory
+; EG-CHECK-LABEL: @local_memory
+; SI-CHECK-LABEL: @local_memory
+; CI-CHECK-LABEL: @local_memory
 
 ; Check that the LDS size emitted correctly
 ; EG-CHECK: .long 166120
-; EG-CHECK-NEXT: .long 16
+; EG-CHECK-NEXT: .long 128
 ; SI-CHECK: .long 47180
-; SI-CHECK-NEXT: .long 32768
+; SI-CHECK-NEXT: .long 65536
+; CI-CHECK: .long 47180
+; CI-CHECK-NEXT: .long 32768
 
 ; EG-CHECK: LDS_WRITE
+; SI-CHECK-NOT: S_WQM_B64
 ; SI-CHECK: DS_WRITE_B32 0
 
 ; GROUP_BARRIER must be the last instruction in a clause
@@ -21,18 +26,18 @@
 ; SI-CHECK: S_BARRIER
 
 ; EG-CHECK: LDS_READ_RET
-; SI-CHECK: DS_READ_B32 {{VGPR[0-9]+}}, 0
+; SI-CHECK: DS_READ_B32 {{v[0-9]+}}, 0
 
 define void @local_memory(i32 addrspace(1)* %out) {
 entry:
   %y.i = call i32 @llvm.r600.read.tidig.x() #0
-  %arrayidx = getelementptr inbounds [16 x i32] addrspace(3)* @local_memory.local_mem, i32 0, i32 %y.i
+  %arrayidx = getelementptr inbounds [128 x i32] addrspace(3)* @local_memory.local_mem, i32 0, i32 %y.i
   store i32 %y.i, i32 addrspace(3)* %arrayidx, align 4
   %add = add nsw i32 %y.i, 1
   %cmp = icmp eq i32 %add, 16
   %.add = select i1 %cmp, i32 0, i32 %add
   call void @llvm.AMDGPU.barrier.local()
-  %arrayidx1 = getelementptr inbounds [16 x i32] addrspace(3)* @local_memory.local_mem, i32 0, i32 %.add
+  %arrayidx1 = getelementptr inbounds [128 x i32] addrspace(3)* @local_memory.local_mem, i32 0, i32 %.add
   %0 = load i32 addrspace(3)* %arrayidx1, align 4
   %arrayidx2 = getelementptr inbounds i32 addrspace(1)* %out, i32 %y.i
   store i32 %0, i32 addrspace(1)* %arrayidx2, align 4