1*67e74705SXin Li// RUN: %clang_cc1 -x cl -O1 -emit-llvm %s -o - -triple x86_64-linux-gnu | FileCheck %s -check-prefix=OPT 2*67e74705SXin Li// RUN: %clang_cc1 -x cl -O0 -emit-llvm %s -o - -triple x86_64-linux-gnu | FileCheck %s -check-prefix=NOOPT 3*67e74705SXin Li 4*67e74705SXin Li// OpenCL essentially reduces all shift amounts to the last word-size 5*67e74705SXin Li// bits before evaluating. Test this both for variables and constants 6*67e74705SXin Li// evaluated in the front-end. 7*67e74705SXin Li 8*67e74705SXin Li// OPT: @gtest1 = local_unnamed_addr constant i64 2147483648 9*67e74705SXin Li__constant const unsigned long gtest1 = 1UL << 31; 10*67e74705SXin Li 11*67e74705SXin Li// NOOPT: @negativeShift32 12*67e74705SXin Liint negativeShift32(int a,int b) { 13*67e74705SXin Li // NOOPT: %array0 = alloca [256 x i8] 14*67e74705SXin Li char array0[((int)1)<<40]; 15*67e74705SXin Li // NOOPT: %array1 = alloca [256 x i8] 16*67e74705SXin Li char array1[((int)1)<<(-24)]; 17*67e74705SXin Li 18*67e74705SXin Li // NOOPT: ret i32 65536 19*67e74705SXin Li return ((int)1)<<(-16); 20*67e74705SXin Li} 21*67e74705SXin Li 22*67e74705SXin Li//OPT: @positiveShift32 23*67e74705SXin Liint positiveShift32(int a,int b) { 24*67e74705SXin Li //OPT: [[M32:%.+]] = and i32 %b, 31 25*67e74705SXin Li //OPT-NEXT: [[C32:%.+]] = shl i32 %a, [[M32]] 26*67e74705SXin Li int c = a<<b; 27*67e74705SXin Li int d = ((int)1)<<33; 28*67e74705SXin Li //OPT-NEXT: [[E32:%.+]] = add nsw i32 [[C32]], 2 29*67e74705SXin Li int e = c + d; 30*67e74705SXin Li //OPT-NEXT: ret i32 [[E32]] 31*67e74705SXin Li return e; 32*67e74705SXin Li} 33*67e74705SXin Li 34*67e74705SXin Li//OPT: @positiveShift64 35*67e74705SXin Lilong positiveShift64(long a,long b) { 36*67e74705SXin Li //OPT: [[M64:%.+]] = and i64 %b, 63 37*67e74705SXin Li //OPT-NEXT: [[C64:%.+]] = ashr i64 %a, [[M64]] 38*67e74705SXin Li long c = a>>b; 39*67e74705SXin Li long d = ((long)8)>>65; 40*67e74705SXin Li //OPT-NEXT: [[E64:%.+]] = add nsw i64 [[C64]], 4 41*67e74705SXin Li long e = c + d; 42*67e74705SXin Li //OPT-NEXT: ret i64 [[E64]] 43*67e74705SXin Li return e; 44*67e74705SXin Li} 45*67e74705SXin Li 46*67e74705SXin Litypedef __attribute__((ext_vector_type(4))) int int4; 47*67e74705SXin Li 48*67e74705SXin Li//OPT: @vectorVectorTest 49*67e74705SXin Liint4 vectorVectorTest(int4 a,int4 b) { 50*67e74705SXin Li //OPT: [[VM:%.+]] = and <4 x i32> %b, <i32 31, i32 31, i32 31, i32 31> 51*67e74705SXin Li //OPT-NEXT: [[VC:%.+]] = shl <4 x i32> %a, [[VM]] 52*67e74705SXin Li int4 c = a << b; 53*67e74705SXin Li //OPT-NEXT: [[VF:%.+]] = add <4 x i32> [[VC]], <i32 2, i32 4, i32 16, i32 8> 54*67e74705SXin Li int4 d = {1, 1, 1, 1}; 55*67e74705SXin Li int4 e = {33, 34, -28, -29}; 56*67e74705SXin Li int4 f = c + (d << e); 57*67e74705SXin Li //OPT-NEXT: ret <4 x i32> [[VF]] 58*67e74705SXin Li return f; 59*67e74705SXin Li} 60*67e74705SXin Li 61*67e74705SXin Li//OPT: @vectorScalarTest 62*67e74705SXin Liint4 vectorScalarTest(int4 a,int b) { 63*67e74705SXin Li //OPT: [[SP0:%.+]] = insertelement <4 x i32> undef, i32 %b, i32 0 64*67e74705SXin Li //OPT: [[SP1:%.+]] = shufflevector <4 x i32> [[SP0]], <4 x i32> undef, <4 x i32> zeroinitializer 65*67e74705SXin Li //OPT: [[VSM:%.+]] = and <4 x i32> [[SP1]], <i32 31, i32 31, i32 31, i32 31> 66*67e74705SXin Li //OPT-NEXT: [[VSC:%.+]] = shl <4 x i32> %a, [[VSM]] 67*67e74705SXin Li int4 c = a << b; 68*67e74705SXin Li //OPT-NEXT: [[VSF:%.+]] = add <4 x i32> [[VSC]], <i32 4, i32 4, i32 4, i32 4> 69*67e74705SXin Li int4 d = {1, 1, 1, 1}; 70*67e74705SXin Li int4 f = c + (d << 34); 71*67e74705SXin Li //OPT-NEXT: ret <4 x i32> [[VSF]] 72*67e74705SXin Li return f; 73*67e74705SXin Li} 74