1 ; RUN: llc < %s -march=nvptx -mcpu=sm_20 | FileCheck %s 2 3 target datalayout = "e-p:32:32:32-i1:8:8-i8:8:8-i16:16:16-i32:32:32-i64:64:64-f32:32:32-f64:64:64-v16:16:16-v32:32:32-v64:64:64-v128:128:128-n16:32:64" 4 5 6 define void @reg_plus_offset(i32* %a) { 7 ; CHECK: ldu.global.u32 %r{{[0-9]+}}, [%r{{[0-9]+}}+32]; 8 ; CHECK: ldu.global.u32 %r{{[0-9]+}}, [%r{{[0-9]+}}+36]; 9 %p2 = getelementptr i32, i32* %a, i32 8 10 %t1 = call i32 @llvm.nvvm.ldu.global.i.i32.p0i32(i32* %p2, i32 4) 11 %p3 = getelementptr i32, i32* %a, i32 9 12 %t2 = call i32 @llvm.nvvm.ldu.global.i.i32.p0i32(i32* %p3, i32 4) 13 %t3 = mul i32 %t1, %t2 14 store i32 %t3, i32* %a 15 ret void 16 } 17 18 declare i32 @llvm.nvvm.ldu.global.i.i32.p0i32(i32*, i32) 19 declare i32 @llvm.nvvm.read.ptx.sreg.tid.x() 20
