954 lines · plain
1;; #pragma OPENCL EXTENSION cl_khr_subgroup_ballot : enable2;; #pragma OPENCL EXTENSION cl_khr_fp16 : enable3;; #pragma OPENCL EXTENSION cl_khr_fp64 : enable4;;5;; kernel void testNonUniformBroadcastChars()6;; {7;; char16 v = 0;8;; v.s0 = sub_group_non_uniform_broadcast(v.s0, 0);9;; v.s01 = sub_group_non_uniform_broadcast(v.s01, 0);10;; v.s012 = sub_group_non_uniform_broadcast(v.s012, 0);11;; v.s0123 = sub_group_non_uniform_broadcast(v.s0123, 0);12;; v.s01234567 = sub_group_non_uniform_broadcast(v.s01234567, 0);13;; v = sub_group_non_uniform_broadcast(v, 0);14;; v.s0 = sub_group_broadcast_first(v.s0);15;; }16;;17;; kernel void testNonUniformBroadcastUChars()18;; {19;; uchar16 v = 0;20;; v.s0 = sub_group_non_uniform_broadcast(v.s0, 0);21;; v.s01 = sub_group_non_uniform_broadcast(v.s01, 0);22;; v.s012 = sub_group_non_uniform_broadcast(v.s012, 0);23;; v.s0123 = sub_group_non_uniform_broadcast(v.s0123, 0);24;; v.s01234567 = sub_group_non_uniform_broadcast(v.s01234567, 0);25;; v = sub_group_non_uniform_broadcast(v, 0);26;; v.s0 = sub_group_broadcast_first(v.s0);27;; }28;;29;; kernel void testNonUniformBroadcastShorts()30;; {31;; short16 v = 0;32;; v.s0 = sub_group_non_uniform_broadcast(v.s0, 0);33;; v.s01 = sub_group_non_uniform_broadcast(v.s01, 0);34;; v.s012 = sub_group_non_uniform_broadcast(v.s012, 0);35;; v.s0123 = sub_group_non_uniform_broadcast(v.s0123, 0);36;; v.s01234567 = sub_group_non_uniform_broadcast(v.s01234567, 0);37;; v = sub_group_non_uniform_broadcast(v, 0);38;; v.s0 = sub_group_broadcast_first(v.s0);39;; }40;;41;; kernel void testNonUniformBroadcastUShorts()42;; {43;; ushort16 v = 0;44;; v.s0 = sub_group_non_uniform_broadcast(v.s0, 0);45;; v.s01 = sub_group_non_uniform_broadcast(v.s01, 0);46;; v.s012 = sub_group_non_uniform_broadcast(v.s012, 0);47;; v.s0123 = sub_group_non_uniform_broadcast(v.s0123, 0);48;; v.s01234567 = sub_group_non_uniform_broadcast(v.s01234567, 0);49;; v = sub_group_non_uniform_broadcast(v, 0);50;; v.s0 = sub_group_broadcast_first(v.s0);51;; }52;;53;; kernel void testNonUniformBroadcastInts()54;; {55;; int16 v = 0;56;; v.s0 = sub_group_non_uniform_broadcast(v.s0, 0);57;; v.s01 = sub_group_non_uniform_broadcast(v.s01, 0);58;; v.s012 = sub_group_non_uniform_broadcast(v.s012, 0);59;; v.s0123 = sub_group_non_uniform_broadcast(v.s0123, 0);60;; v.s01234567 = sub_group_non_uniform_broadcast(v.s01234567, 0);61;; v = sub_group_non_uniform_broadcast(v, 0);62;; v.s0 = sub_group_broadcast_first(v.s0);63;; }64;;65;; kernel void testNonUniformBroadcastUInts()66;; {67;; uint16 v = 0;68;; v.s0 = sub_group_non_uniform_broadcast(v.s0, 0);69;; v.s01 = sub_group_non_uniform_broadcast(v.s01, 0);70;; v.s012 = sub_group_non_uniform_broadcast(v.s012, 0);71;; v.s0123 = sub_group_non_uniform_broadcast(v.s0123, 0);72;; v.s01234567 = sub_group_non_uniform_broadcast(v.s01234567, 0);73;; v = sub_group_non_uniform_broadcast(v, 0);74;; v.s0 = sub_group_broadcast_first(v.s0);75;; }76;;77;; kernel void testNonUniformBroadcastLongs()78;; {79;; long16 v = 0;80;; v.s0 = sub_group_non_uniform_broadcast(v.s0, 0);81;; v.s01 = sub_group_non_uniform_broadcast(v.s01, 0);82;; v.s012 = sub_group_non_uniform_broadcast(v.s012, 0);83;; v.s0123 = sub_group_non_uniform_broadcast(v.s0123, 0);84;; v.s01234567 = sub_group_non_uniform_broadcast(v.s01234567, 0);85;; v = sub_group_non_uniform_broadcast(v, 0);86;; v.s0 = sub_group_broadcast_first(v.s0);87;; }88;;89;; kernel void testNonUniformBroadcastULongs()90;; {91;; ulong16 v = 0;92;; v.s0 = sub_group_non_uniform_broadcast(v.s0, 0);93;; v.s01 = sub_group_non_uniform_broadcast(v.s01, 0);94;; v.s012 = sub_group_non_uniform_broadcast(v.s012, 0);95;; v.s0123 = sub_group_non_uniform_broadcast(v.s0123, 0);96;; v.s01234567 = sub_group_non_uniform_broadcast(v.s01234567, 0);97;; v = sub_group_non_uniform_broadcast(v, 0);98;; v.s0 = sub_group_broadcast_first(v.s0);99;; }100;;101;; kernel void testNonUniformBroadcastFloats()102;; {103;; float16 v = 0;104;; v.s0 = sub_group_non_uniform_broadcast(v.s0, 0);105;; v.s01 = sub_group_non_uniform_broadcast(v.s01, 0);106;; v.s012 = sub_group_non_uniform_broadcast(v.s012, 0);107;; v.s0123 = sub_group_non_uniform_broadcast(v.s0123, 0);108;; v.s01234567 = sub_group_non_uniform_broadcast(v.s01234567, 0);109;; v = sub_group_non_uniform_broadcast(v, 0);110;; v.s0 = sub_group_broadcast_first(v.s0);111;; }112;;113;; kernel void testNonUniformBroadcastHalfs()114;; {115;; half16 v = 0;116;; v.s0 = sub_group_non_uniform_broadcast(v.s0, 0);117;; v.s01 = sub_group_non_uniform_broadcast(v.s01, 0);118;; v.s012 = sub_group_non_uniform_broadcast(v.s012, 0);119;; v.s0123 = sub_group_non_uniform_broadcast(v.s0123, 0);120;; v.s01234567 = sub_group_non_uniform_broadcast(v.s01234567, 0);121;; v = sub_group_non_uniform_broadcast(v, 0);122;; v.s0 = sub_group_broadcast_first(v.s0);123;; }124;;125;; kernel void testNonUniformBroadcastDoubles()126;; {127;; double16 v = 0;128;; v.s0 = sub_group_non_uniform_broadcast(v.s0, 0);129;; v.s01 = sub_group_non_uniform_broadcast(v.s01, 0);130;; v.s012 = sub_group_non_uniform_broadcast(v.s012, 0);131;; v.s0123 = sub_group_non_uniform_broadcast(v.s0123, 0);132;; v.s01234567 = sub_group_non_uniform_broadcast(v.s01234567, 0);133;; v = sub_group_non_uniform_broadcast(v, 0);134;; v.s0 = sub_group_broadcast_first(v.s0);135;; }136;;137;; kernel void testBallotOperations(global uint* dst)138;; {139;; uint4 v = sub_group_ballot(0);140;; dst[0] = sub_group_inverse_ballot(v);141;; dst[1] = sub_group_ballot_bit_extract(v, 0);142;; dst[2] = sub_group_ballot_bit_count(v);143;; dst[3] = sub_group_ballot_inclusive_scan(v);144;; dst[4] = sub_group_ballot_exclusive_scan(v);145;; dst[5] = sub_group_ballot_find_lsb(v);146;; dst[6] = sub_group_ballot_find_msb(v);147;; }148;;149;; kernel void testSubgroupMasks(global uint4* dst)150;; {151;; dst[0] = get_sub_group_eq_mask();152;; dst[1] = get_sub_group_ge_mask();153;; dst[2] = get_sub_group_gt_mask();154;; dst[3] = get_sub_group_le_mask();155;; dst[4] = get_sub_group_lt_mask();156;; }157 158; RUN: llc -verify-machineinstrs -O0 -mtriple=spirv64-unknown-unknown %s -o - | FileCheck %s --check-prefix=CHECK-SPIRV159 160; CHECK-SPIRV-DAG: OpCapability GroupNonUniformBallot161 162; CHECK-SPIRV-DAG: OpDecorate %[[#eqMask:]] BuiltIn SubgroupEqMask163; CHECK-SPIRV-DAG: OpDecorate %[[#geMask:]] BuiltIn SubgroupGeMask164; CHECK-SPIRV-DAG: OpDecorate %[[#gtMask:]] BuiltIn SubgroupGtMask165; CHECK-SPIRV-DAG: OpDecorate %[[#leMask:]] BuiltIn SubgroupLeMask166; CHECK-SPIRV-DAG: OpDecorate %[[#ltMask:]] BuiltIn SubgroupLtMask167 168; CHECK-SPIRV-DAG: %[[#bool:]] = OpTypeBool169; CHECK-SPIRV-DAG: %[[#char:]] = OpTypeInt 8 0170; CHECK-SPIRV-DAG: %[[#short:]] = OpTypeInt 16 0171; CHECK-SPIRV-DAG: %[[#int:]] = OpTypeInt 32 0172; CHECK-SPIRV-DAG: %[[#long:]] = OpTypeInt 64 0173; CHECK-SPIRV-DAG: %[[#half:]] = OpTypeFloat 16174; CHECK-SPIRV-DAG: %[[#float:]] = OpTypeFloat 32175; CHECK-SPIRV-DAG: %[[#double:]] = OpTypeFloat 64176 177; CHECK-SPIRV-DAG: %[[#char2:]] = OpTypeVector %[[#char]] 2178; CHECK-SPIRV-DAG: %[[#char3:]] = OpTypeVector %[[#char]] 3179; CHECK-SPIRV-DAG: %[[#char4:]] = OpTypeVector %[[#char]] 4180; CHECK-SPIRV-DAG: %[[#char8:]] = OpTypeVector %[[#char]] 8181; CHECK-SPIRV-DAG: %[[#char16:]] = OpTypeVector %[[#char]] 16182 183; CHECK-SPIRV-DAG: %[[#short2:]] = OpTypeVector %[[#short]] 2184; CHECK-SPIRV-DAG: %[[#short3:]] = OpTypeVector %[[#short]] 3185; CHECK-SPIRV-DAG: %[[#short4:]] = OpTypeVector %[[#short]] 4186; CHECK-SPIRV-DAG: %[[#short8:]] = OpTypeVector %[[#short]] 8187; CHECK-SPIRV-DAG: %[[#short16:]] = OpTypeVector %[[#short]] 16188 189; CHECK-SPIRV-DAG: %[[#int2:]] = OpTypeVector %[[#int]] 2190; CHECK-SPIRV-DAG: %[[#int3:]] = OpTypeVector %[[#int]] 3191; CHECK-SPIRV-DAG: %[[#int4:]] = OpTypeVector %[[#int]] 4192; CHECK-SPIRV-DAG: %[[#int8:]] = OpTypeVector %[[#int]] 8193; CHECK-SPIRV-DAG: %[[#int16:]] = OpTypeVector %[[#int]] 16194 195; CHECK-SPIRV-DAG: %[[#long2:]] = OpTypeVector %[[#long]] 2196; CHECK-SPIRV-DAG: %[[#long3:]] = OpTypeVector %[[#long]] 3197; CHECK-SPIRV-DAG: %[[#long4:]] = OpTypeVector %[[#long]] 4198; CHECK-SPIRV-DAG: %[[#long8:]] = OpTypeVector %[[#long]] 8199; CHECK-SPIRV-DAG: %[[#long16:]] = OpTypeVector %[[#long]] 16200 201; CHECK-SPIRV-DAG: %[[#float2:]] = OpTypeVector %[[#float]] 2202; CHECK-SPIRV-DAG: %[[#float3:]] = OpTypeVector %[[#float]] 3203; CHECK-SPIRV-DAG: %[[#float4:]] = OpTypeVector %[[#float]] 4204; CHECK-SPIRV-DAG: %[[#float8:]] = OpTypeVector %[[#float]] 8205; CHECK-SPIRV-DAG: %[[#float16:]] = OpTypeVector %[[#float]] 16206 207; CHECK-SPIRV-DAG: %[[#half2:]] = OpTypeVector %[[#half]] 2208; CHECK-SPIRV-DAG: %[[#half3:]] = OpTypeVector %[[#half]] 3209; CHECK-SPIRV-DAG: %[[#half4:]] = OpTypeVector %[[#half]] 4210; CHECK-SPIRV-DAG: %[[#half8:]] = OpTypeVector %[[#half]] 8211; CHECK-SPIRV-DAG: %[[#half16:]] = OpTypeVector %[[#half]] 16212 213; CHECK-SPIRV-DAG: %[[#double2:]] = OpTypeVector %[[#double]] 2214; CHECK-SPIRV-DAG: %[[#double3:]] = OpTypeVector %[[#double]] 3215; CHECK-SPIRV-DAG: %[[#double4:]] = OpTypeVector %[[#double]] 4216; CHECK-SPIRV-DAG: %[[#double8:]] = OpTypeVector %[[#double]] 8217; CHECK-SPIRV-DAG: %[[#double16:]] = OpTypeVector %[[#double]] 16218 219; CHECK-SPIRV-DAG: %[[#false:]] = OpConstantFalse %[[#bool]]220; CHECK-SPIRV-DAG: %[[#ScopeSubgroup:]] = OpConstant %[[#int]] 3{{$}}221; CHECK-SPIRV-DAG: %[[#char_0:]] = OpConstantNull %[[#char]]222; CHECK-SPIRV-DAG: %[[#short_0:]] = OpConstantNull %[[#short]]223; CHECK-SPIRV-DAG: %[[#int_0:]] = OpConstantNull %[[#int]]224; CHECK-SPIRV-DAG: %[[#long_0:]] = OpConstantNull %[[#long]]225; CHECK-SPIRV-DAG: %[[#half_0:]] = OpConstantNull %[[#half]]226; CHECK-SPIRV-DAG: %[[#float_0:]] = OpConstantNull %[[#float]]227; CHECK-SPIRV-DAG: %[[#double_0:]] = OpConstantNull %[[#double]]228 229; CHECK-SPIRV: OpFunction230; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#char]] %[[#ScopeSubgroup]] %[[#char_0]] %[[#int_0]]231; CHECK-SPIRV: %[[#char2_0:]] = OpVectorShuffle %[[#char2]]232; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#char2]] %[[#ScopeSubgroup]] %[[#char2_0]] %[[#int_0]]233; CHECK-SPIRV: %[[#char3_0:]] = OpVectorShuffle %[[#char3]]234; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#char3]] %[[#ScopeSubgroup]] %[[#char3_0]] %[[#int_0]]235; CHECK-SPIRV: %[[#char4_0:]] = OpVectorShuffle %[[#char4]]236; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#char4]] %[[#ScopeSubgroup]] %[[#char4_0]] %[[#int_0]]237; CHECK-SPIRV: %[[#char8_0:]] = OpVectorShuffle %[[#char8]]238; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#char8]] %[[#ScopeSubgroup]] %[[#char8_0]] %[[#int_0]]239; CHECK-SPIRV: %[[#]] = OpVectorShuffle %[[#char16]]240; CHECK-SPIRV: %[[#char16_0:]] = OpVectorShuffle %[[#char16]]241; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#char16]] %[[#ScopeSubgroup]] %[[#char16_0]] %[[#int_0]]242; CHECK-SPIRV: %[[#char_value:]] = OpCompositeExtract %[[#char]]243; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcastFirst %[[#char]] %[[#ScopeSubgroup]] %[[#char_value]]244; CHECK-SPIRV: OpFunctionEnd245 246define dso_local spir_kernel void @testNonUniformBroadcastChars() local_unnamed_addr {247 %1 = tail call spir_func signext i8 @_Z31sub_group_non_uniform_broadcastcj(i8 signext 0, i32 0)248 %2 = insertelement <16 x i8> <i8 undef, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>, i8 %1, i64 0249 %3 = shufflevector <16 x i8> %2, <16 x i8> undef, <2 x i32> <i32 0, i32 1>250 %4 = tail call spir_func <2 x i8> @_Z31sub_group_non_uniform_broadcastDv2_cj(<2 x i8> %3, i32 0)251 %5 = shufflevector <2 x i8> %4, <2 x i8> undef, <16 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>252 %6 = shufflevector <16 x i8> %5, <16 x i8> %2, <16 x i32> <i32 0, i32 1, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>253 %7 = shufflevector <16 x i8> %6, <16 x i8> undef, <3 x i32> <i32 0, i32 1, i32 2>254 %8 = tail call spir_func <3 x i8> @_Z31sub_group_non_uniform_broadcastDv3_cj(<3 x i8> %7, i32 0)255 %9 = shufflevector <3 x i8> %8, <3 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>256 %10 = shufflevector <16 x i8> %9, <16 x i8> %6, <16 x i32> <i32 0, i32 1, i32 2, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>257 %11 = shufflevector <16 x i8> %10, <16 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>258 %12 = tail call spir_func <4 x i8> @_Z31sub_group_non_uniform_broadcastDv4_cj(<4 x i8> %11, i32 0)259 %13 = shufflevector <4 x i8> %12, <4 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>260 %14 = shufflevector <16 x i8> %13, <16 x i8> %10, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>261 %15 = shufflevector <16 x i8> %14, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>262 %16 = tail call spir_func <8 x i8> @_Z31sub_group_non_uniform_broadcastDv8_cj(<8 x i8> %15, i32 0)263 %17 = shufflevector <8 x i8> %16, <8 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>264 %18 = shufflevector <16 x i8> %17, <16 x i8> %14, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>265 %19 = tail call spir_func <16 x i8> @_Z31sub_group_non_uniform_broadcastDv16_cj(<16 x i8> %18, i32 0)266 %20 = extractelement <16 x i8> %19, i64 0267 %21 = tail call spir_func signext i8 @_Z25sub_group_broadcast_firstc(i8 signext %20)268 ret void269}270 271declare dso_local spir_func signext i8 @_Z31sub_group_non_uniform_broadcastcj(i8 signext, i32) local_unnamed_addr272 273declare dso_local spir_func <2 x i8> @_Z31sub_group_non_uniform_broadcastDv2_cj(<2 x i8>, i32) local_unnamed_addr274 275declare dso_local spir_func <3 x i8> @_Z31sub_group_non_uniform_broadcastDv3_cj(<3 x i8>, i32) local_unnamed_addr276 277declare dso_local spir_func <4 x i8> @_Z31sub_group_non_uniform_broadcastDv4_cj(<4 x i8>, i32) local_unnamed_addr278 279declare dso_local spir_func <8 x i8> @_Z31sub_group_non_uniform_broadcastDv8_cj(<8 x i8>, i32) local_unnamed_addr280 281declare dso_local spir_func <16 x i8> @_Z31sub_group_non_uniform_broadcastDv16_cj(<16 x i8>, i32) local_unnamed_addr282 283declare dso_local spir_func signext i8 @_Z25sub_group_broadcast_firstc(i8 signext) local_unnamed_addr284 285; CHECK-SPIRV: OpFunction286; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#char]] %[[#ScopeSubgroup]] %[[#char_0]] %[[#int_0]]287; CHECK-SPIRV: %[[#char2_0:]] = OpVectorShuffle %[[#char2]]288; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#char2]] %[[#ScopeSubgroup]] %[[#char2_0]] %[[#int_0]]289; CHECK-SPIRV: %[[#char3_0:]] = OpVectorShuffle %[[#char3]]290; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#char3]] %[[#ScopeSubgroup]] %[[#char3_0]] %[[#int_0]]291; CHECK-SPIRV: %[[#char4_0:]] = OpVectorShuffle %[[#char4]]292; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#char4]] %[[#ScopeSubgroup]] %[[#char4_0]] %[[#int_0]]293; CHECK-SPIRV: %[[#char8_0:]] = OpVectorShuffle %[[#char8]]294; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#char8]] %[[#ScopeSubgroup]] %[[#char8_0]] %[[#int_0]]295; CHECK-SPIRV: %[[#]] = OpVectorShuffle %[[#char16]]296; CHECK-SPIRV: %[[#char16_0:]] = OpVectorShuffle %[[#char16]]297; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#char16]] %[[#ScopeSubgroup]] %[[#char16_0]] %[[#int_0]]298; CHECK-SPIRV: %[[#char_value:]] = OpCompositeExtract %[[#char]]299; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcastFirst %[[#char]] %[[#ScopeSubgroup]] %[[#char_value]]300; CHECK-SPIRV: OpFunctionEnd301 302define dso_local spir_kernel void @testNonUniformBroadcastUChars() local_unnamed_addr {303 %1 = tail call spir_func zeroext i8 @_Z31sub_group_non_uniform_broadcasthj(i8 zeroext 0, i32 0)304 %2 = insertelement <16 x i8> <i8 undef, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0, i8 0>, i8 %1, i64 0305 %3 = shufflevector <16 x i8> %2, <16 x i8> undef, <2 x i32> <i32 0, i32 1>306 %4 = tail call spir_func <2 x i8> @_Z31sub_group_non_uniform_broadcastDv2_hj(<2 x i8> %3, i32 0)307 %5 = shufflevector <2 x i8> %4, <2 x i8> undef, <16 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>308 %6 = shufflevector <16 x i8> %5, <16 x i8> %2, <16 x i32> <i32 0, i32 1, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>309 %7 = shufflevector <16 x i8> %6, <16 x i8> undef, <3 x i32> <i32 0, i32 1, i32 2>310 %8 = tail call spir_func <3 x i8> @_Z31sub_group_non_uniform_broadcastDv3_hj(<3 x i8> %7, i32 0)311 %9 = shufflevector <3 x i8> %8, <3 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>312 %10 = shufflevector <16 x i8> %9, <16 x i8> %6, <16 x i32> <i32 0, i32 1, i32 2, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>313 %11 = shufflevector <16 x i8> %10, <16 x i8> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>314 %12 = tail call spir_func <4 x i8> @_Z31sub_group_non_uniform_broadcastDv4_hj(<4 x i8> %11, i32 0)315 %13 = shufflevector <4 x i8> %12, <4 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>316 %14 = shufflevector <16 x i8> %13, <16 x i8> %10, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>317 %15 = shufflevector <16 x i8> %14, <16 x i8> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>318 %16 = tail call spir_func <8 x i8> @_Z31sub_group_non_uniform_broadcastDv8_hj(<8 x i8> %15, i32 0)319 %17 = shufflevector <8 x i8> %16, <8 x i8> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>320 %18 = shufflevector <16 x i8> %17, <16 x i8> %14, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>321 %19 = tail call spir_func <16 x i8> @_Z31sub_group_non_uniform_broadcastDv16_hj(<16 x i8> %18, i32 0)322 %20 = extractelement <16 x i8> %19, i64 0323 %21 = tail call spir_func zeroext i8 @_Z25sub_group_broadcast_firsth(i8 zeroext %20)324 ret void325}326 327declare dso_local spir_func zeroext i8 @_Z31sub_group_non_uniform_broadcasthj(i8 zeroext, i32) local_unnamed_addr328 329declare dso_local spir_func <2 x i8> @_Z31sub_group_non_uniform_broadcastDv2_hj(<2 x i8>, i32) local_unnamed_addr330 331declare dso_local spir_func <3 x i8> @_Z31sub_group_non_uniform_broadcastDv3_hj(<3 x i8>, i32) local_unnamed_addr332 333declare dso_local spir_func <4 x i8> @_Z31sub_group_non_uniform_broadcastDv4_hj(<4 x i8>, i32) local_unnamed_addr334 335declare dso_local spir_func <8 x i8> @_Z31sub_group_non_uniform_broadcastDv8_hj(<8 x i8>, i32) local_unnamed_addr336 337declare dso_local spir_func <16 x i8> @_Z31sub_group_non_uniform_broadcastDv16_hj(<16 x i8>, i32) local_unnamed_addr338 339declare dso_local spir_func zeroext i8 @_Z25sub_group_broadcast_firsth(i8 zeroext) local_unnamed_addr340 341; CHECK-SPIRV: OpFunction342; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#short]] %[[#ScopeSubgroup]] %[[#short_0]] %[[#int_0]]343; CHECK-SPIRV: %[[#short2_0:]] = OpVectorShuffle %[[#short2]]344; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#short2]] %[[#ScopeSubgroup]] %[[#short2_0]] %[[#int_0]]345; CHECK-SPIRV: %[[#short3_0:]] = OpVectorShuffle %[[#short3]]346; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#short3]] %[[#ScopeSubgroup]] %[[#short3_0]] %[[#int_0]]347; CHECK-SPIRV: %[[#short4_0:]] = OpVectorShuffle %[[#short4]]348; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#short4]] %[[#ScopeSubgroup]] %[[#short4_0]] %[[#int_0]]349; CHECK-SPIRV: %[[#short8_0:]] = OpVectorShuffle %[[#short8]]350; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#short8]] %[[#ScopeSubgroup]] %[[#short8_0]] %[[#int_0]]351; CHECK-SPIRV: %[[#]] = OpVectorShuffle %[[#short16]]352; CHECK-SPIRV: %[[#short16_0:]] = OpVectorShuffle %[[#short16]]353; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#short16]] %[[#ScopeSubgroup]] %[[#short16_0]] %[[#int_0]]354; CHECK-SPIRV: %[[#short_value:]] = OpCompositeExtract %[[#short]]355; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcastFirst %[[#short]] %[[#ScopeSubgroup]] %[[#short_value]]356; CHECK-SPIRV: OpFunctionEnd357 358define dso_local spir_kernel void @testNonUniformBroadcastShorts() local_unnamed_addr {359 %1 = tail call spir_func signext i16 @_Z31sub_group_non_uniform_broadcastsj(i16 signext 0, i32 0)360 %2 = insertelement <16 x i16> <i16 undef, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0>, i16 %1, i64 0361 %3 = shufflevector <16 x i16> %2, <16 x i16> undef, <2 x i32> <i32 0, i32 1>362 %4 = tail call spir_func <2 x i16> @_Z31sub_group_non_uniform_broadcastDv2_sj(<2 x i16> %3, i32 0)363 %5 = shufflevector <2 x i16> %4, <2 x i16> undef, <16 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>364 %6 = shufflevector <16 x i16> %5, <16 x i16> %2, <16 x i32> <i32 0, i32 1, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>365 %7 = shufflevector <16 x i16> %6, <16 x i16> undef, <3 x i32> <i32 0, i32 1, i32 2>366 %8 = tail call spir_func <3 x i16> @_Z31sub_group_non_uniform_broadcastDv3_sj(<3 x i16> %7, i32 0)367 %9 = shufflevector <3 x i16> %8, <3 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>368 %10 = shufflevector <16 x i16> %9, <16 x i16> %6, <16 x i32> <i32 0, i32 1, i32 2, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>369 %11 = shufflevector <16 x i16> %10, <16 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>370 %12 = tail call spir_func <4 x i16> @_Z31sub_group_non_uniform_broadcastDv4_sj(<4 x i16> %11, i32 0)371 %13 = shufflevector <4 x i16> %12, <4 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>372 %14 = shufflevector <16 x i16> %13, <16 x i16> %10, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>373 %15 = shufflevector <16 x i16> %14, <16 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>374 %16 = tail call spir_func <8 x i16> @_Z31sub_group_non_uniform_broadcastDv8_sj(<8 x i16> %15, i32 0)375 %17 = shufflevector <8 x i16> %16, <8 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>376 %18 = shufflevector <16 x i16> %17, <16 x i16> %14, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>377 %19 = tail call spir_func <16 x i16> @_Z31sub_group_non_uniform_broadcastDv16_sj(<16 x i16> %18, i32 0)378 %20 = extractelement <16 x i16> %19, i64 0379 %21 = tail call spir_func signext i16 @_Z25sub_group_broadcast_firsts(i16 signext %20)380 ret void381}382 383declare dso_local spir_func signext i16 @_Z31sub_group_non_uniform_broadcastsj(i16 signext, i32) local_unnamed_addr384 385declare dso_local spir_func <2 x i16> @_Z31sub_group_non_uniform_broadcastDv2_sj(<2 x i16>, i32) local_unnamed_addr386 387declare dso_local spir_func <3 x i16> @_Z31sub_group_non_uniform_broadcastDv3_sj(<3 x i16>, i32) local_unnamed_addr388 389declare dso_local spir_func <4 x i16> @_Z31sub_group_non_uniform_broadcastDv4_sj(<4 x i16>, i32) local_unnamed_addr390 391declare dso_local spir_func <8 x i16> @_Z31sub_group_non_uniform_broadcastDv8_sj(<8 x i16>, i32) local_unnamed_addr392 393declare dso_local spir_func <16 x i16> @_Z31sub_group_non_uniform_broadcastDv16_sj(<16 x i16>, i32) local_unnamed_addr394 395declare dso_local spir_func signext i16 @_Z25sub_group_broadcast_firsts(i16 signext) local_unnamed_addr396 397; CHECK-SPIRV: OpFunction398; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#short]] %[[#ScopeSubgroup]] %[[#short_0]] %[[#int_0]]399; CHECK-SPIRV: %[[#short2_0:]] = OpVectorShuffle %[[#short2]]400; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#short2]] %[[#ScopeSubgroup]] %[[#short2_0]] %[[#int_0]]401; CHECK-SPIRV: %[[#short3_0:]] = OpVectorShuffle %[[#short3]]402; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#short3]] %[[#ScopeSubgroup]] %[[#short3_0]] %[[#int_0]]403; CHECK-SPIRV: %[[#short4_0:]] = OpVectorShuffle %[[#short4]]404; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#short4]] %[[#ScopeSubgroup]] %[[#short4_0]] %[[#int_0]]405; CHECK-SPIRV: %[[#short8_0:]] = OpVectorShuffle %[[#short8]]406; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#short8]] %[[#ScopeSubgroup]] %[[#short8_0]] %[[#int_0]]407; CHECK-SPIRV: %[[#]] = OpVectorShuffle %[[#short16]]408; CHECK-SPIRV: %[[#short16_0:]] = OpVectorShuffle %[[#short16]]409; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#short16]] %[[#ScopeSubgroup]] %[[#short16_0]] %[[#int_0]]410; CHECK-SPIRV: %[[#short_value:]] = OpCompositeExtract %[[#short]]411; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcastFirst %[[#short]] %[[#ScopeSubgroup]] %[[#short_value]]412; CHECK-SPIRV: OpFunctionEnd413 414define dso_local spir_kernel void @testNonUniformBroadcastUShorts() local_unnamed_addr {415 %1 = tail call spir_func zeroext i16 @_Z31sub_group_non_uniform_broadcasttj(i16 zeroext 0, i32 0)416 %2 = insertelement <16 x i16> <i16 undef, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0, i16 0>, i16 %1, i64 0417 %3 = shufflevector <16 x i16> %2, <16 x i16> undef, <2 x i32> <i32 0, i32 1>418 %4 = tail call spir_func <2 x i16> @_Z31sub_group_non_uniform_broadcastDv2_tj(<2 x i16> %3, i32 0)419 %5 = shufflevector <2 x i16> %4, <2 x i16> undef, <16 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>420 %6 = shufflevector <16 x i16> %5, <16 x i16> %2, <16 x i32> <i32 0, i32 1, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>421 %7 = shufflevector <16 x i16> %6, <16 x i16> undef, <3 x i32> <i32 0, i32 1, i32 2>422 %8 = tail call spir_func <3 x i16> @_Z31sub_group_non_uniform_broadcastDv3_tj(<3 x i16> %7, i32 0)423 %9 = shufflevector <3 x i16> %8, <3 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>424 %10 = shufflevector <16 x i16> %9, <16 x i16> %6, <16 x i32> <i32 0, i32 1, i32 2, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>425 %11 = shufflevector <16 x i16> %10, <16 x i16> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>426 %12 = tail call spir_func <4 x i16> @_Z31sub_group_non_uniform_broadcastDv4_tj(<4 x i16> %11, i32 0)427 %13 = shufflevector <4 x i16> %12, <4 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>428 %14 = shufflevector <16 x i16> %13, <16 x i16> %10, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>429 %15 = shufflevector <16 x i16> %14, <16 x i16> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>430 %16 = tail call spir_func <8 x i16> @_Z31sub_group_non_uniform_broadcastDv8_tj(<8 x i16> %15, i32 0)431 %17 = shufflevector <8 x i16> %16, <8 x i16> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>432 %18 = shufflevector <16 x i16> %17, <16 x i16> %14, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>433 %19 = tail call spir_func <16 x i16> @_Z31sub_group_non_uniform_broadcastDv16_tj(<16 x i16> %18, i32 0)434 %20 = extractelement <16 x i16> %19, i64 0435 %21 = tail call spir_func zeroext i16 @_Z25sub_group_broadcast_firstt(i16 zeroext %20)436 ret void437}438 439declare dso_local spir_func zeroext i16 @_Z31sub_group_non_uniform_broadcasttj(i16 zeroext, i32) local_unnamed_addr440 441declare dso_local spir_func <2 x i16> @_Z31sub_group_non_uniform_broadcastDv2_tj(<2 x i16>, i32) local_unnamed_addr442 443declare dso_local spir_func <3 x i16> @_Z31sub_group_non_uniform_broadcastDv3_tj(<3 x i16>, i32) local_unnamed_addr444 445declare dso_local spir_func <4 x i16> @_Z31sub_group_non_uniform_broadcastDv4_tj(<4 x i16>, i32) local_unnamed_addr446 447declare dso_local spir_func <8 x i16> @_Z31sub_group_non_uniform_broadcastDv8_tj(<8 x i16>, i32) local_unnamed_addr448 449declare dso_local spir_func <16 x i16> @_Z31sub_group_non_uniform_broadcastDv16_tj(<16 x i16>, i32) local_unnamed_addr450 451declare dso_local spir_func zeroext i16 @_Z25sub_group_broadcast_firstt(i16 zeroext) local_unnamed_addr452 453; CHECK-SPIRV: OpFunction454; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#int]] %[[#ScopeSubgroup]] %[[#int_0]] %[[#int_0]]455; CHECK-SPIRV: %[[#int2_0:]] = OpVectorShuffle %[[#int2]]456; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#int2]] %[[#ScopeSubgroup]] %[[#int2_0]] %[[#int_0]]457; CHECK-SPIRV: %[[#int3_0:]] = OpVectorShuffle %[[#int3]]458; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#int3]] %[[#ScopeSubgroup]] %[[#int3_0]] %[[#int_0]]459; CHECK-SPIRV: %[[#int4_0:]] = OpVectorShuffle %[[#int4]]460; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#int4]] %[[#ScopeSubgroup]] %[[#int4_0]] %[[#int_0]]461; CHECK-SPIRV: %[[#int8_0:]] = OpVectorShuffle %[[#int8]]462; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#int8]] %[[#ScopeSubgroup]] %[[#int8_0]] %[[#int_0]]463; CHECK-SPIRV: %[[#]] = OpVectorShuffle %[[#int16]]464; CHECK-SPIRV: %[[#int16_0:]] = OpVectorShuffle %[[#int16]]465; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#int16]] %[[#ScopeSubgroup]] %[[#int16_0]] %[[#int_0]]466; CHECK-SPIRV: %[[#int_value:]] = OpCompositeExtract %[[#int]]467; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcastFirst %[[#int]] %[[#ScopeSubgroup]] %[[#int_value]]468; CHECK-SPIRV: OpFunctionEnd469 470define dso_local spir_kernel void @testNonUniformBroadcastInts() local_unnamed_addr {471 %1 = tail call spir_func i32 @_Z31sub_group_non_uniform_broadcastij(i32 0, i32 0)472 %2 = insertelement <16 x i32> <i32 undef, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>, i32 %1, i64 0473 %3 = shufflevector <16 x i32> %2, <16 x i32> undef, <2 x i32> <i32 0, i32 1>474 %4 = tail call spir_func <2 x i32> @_Z31sub_group_non_uniform_broadcastDv2_ij(<2 x i32> %3, i32 0)475 %5 = shufflevector <2 x i32> %4, <2 x i32> undef, <16 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>476 %6 = shufflevector <16 x i32> %5, <16 x i32> %2, <16 x i32> <i32 0, i32 1, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>477 %7 = shufflevector <16 x i32> %6, <16 x i32> undef, <3 x i32> <i32 0, i32 1, i32 2>478 %8 = tail call spir_func <3 x i32> @_Z31sub_group_non_uniform_broadcastDv3_ij(<3 x i32> %7, i32 0)479 %9 = shufflevector <3 x i32> %8, <3 x i32> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>480 %10 = shufflevector <16 x i32> %9, <16 x i32> %6, <16 x i32> <i32 0, i32 1, i32 2, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>481 %11 = shufflevector <16 x i32> %10, <16 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>482 %12 = tail call spir_func <4 x i32> @_Z31sub_group_non_uniform_broadcastDv4_ij(<4 x i32> %11, i32 0)483 %13 = shufflevector <4 x i32> %12, <4 x i32> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>484 %14 = shufflevector <16 x i32> %13, <16 x i32> %10, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>485 %15 = shufflevector <16 x i32> %14, <16 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>486 %16 = tail call spir_func <8 x i32> @_Z31sub_group_non_uniform_broadcastDv8_ij(<8 x i32> %15, i32 0)487 %17 = shufflevector <8 x i32> %16, <8 x i32> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>488 %18 = shufflevector <16 x i32> %17, <16 x i32> %14, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>489 %19 = tail call spir_func <16 x i32> @_Z31sub_group_non_uniform_broadcastDv16_ij(<16 x i32> %18, i32 0)490 %20 = extractelement <16 x i32> %19, i64 0491 %21 = tail call spir_func i32 @_Z25sub_group_broadcast_firsti(i32 %20)492 ret void493}494 495declare dso_local spir_func i32 @_Z31sub_group_non_uniform_broadcastij(i32, i32) local_unnamed_addr496 497declare dso_local spir_func <2 x i32> @_Z31sub_group_non_uniform_broadcastDv2_ij(<2 x i32>, i32) local_unnamed_addr498 499declare dso_local spir_func <3 x i32> @_Z31sub_group_non_uniform_broadcastDv3_ij(<3 x i32>, i32) local_unnamed_addr500 501declare dso_local spir_func <4 x i32> @_Z31sub_group_non_uniform_broadcastDv4_ij(<4 x i32>, i32) local_unnamed_addr502 503declare dso_local spir_func <8 x i32> @_Z31sub_group_non_uniform_broadcastDv8_ij(<8 x i32>, i32) local_unnamed_addr504 505declare dso_local spir_func <16 x i32> @_Z31sub_group_non_uniform_broadcastDv16_ij(<16 x i32>, i32) local_unnamed_addr506 507declare dso_local spir_func i32 @_Z25sub_group_broadcast_firsti(i32) local_unnamed_addr508 509; CHECK-SPIRV: OpFunction510; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#int]] %[[#ScopeSubgroup]] %[[#int_0]] %[[#int_0]]511; CHECK-SPIRV: %[[#int2_0:]] = OpVectorShuffle %[[#int2]]512; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#int2]] %[[#ScopeSubgroup]] %[[#int2_0]] %[[#int_0]]513; CHECK-SPIRV: %[[#int3_0:]] = OpVectorShuffle %[[#int3]]514; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#int3]] %[[#ScopeSubgroup]] %[[#int3_0]] %[[#int_0]]515; CHECK-SPIRV: %[[#int4_0:]] = OpVectorShuffle %[[#int4]]516; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#int4]] %[[#ScopeSubgroup]] %[[#int4_0]] %[[#int_0]]517; CHECK-SPIRV: %[[#int8_0:]] = OpVectorShuffle %[[#int8]]518; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#int8]] %[[#ScopeSubgroup]] %[[#int8_0]] %[[#int_0]]519; CHECK-SPIRV: %[[#]] = OpVectorShuffle %[[#int16]]520; CHECK-SPIRV: %[[#int16_0:]] = OpVectorShuffle %[[#int16]]521; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#int16]] %[[#ScopeSubgroup]] %[[#int16_0]] %[[#int_0]]522; CHECK-SPIRV: %[[#int_value:]] = OpCompositeExtract %[[#int]]523; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcastFirst %[[#int]] %[[#ScopeSubgroup]] %[[#int_value]]524; CHECK-SPIRV: OpFunctionEnd525 526define dso_local spir_kernel void @testNonUniformBroadcastUInts() local_unnamed_addr {527 %1 = tail call spir_func i32 @_Z31sub_group_non_uniform_broadcastjj(i32 0, i32 0)528 %2 = insertelement <16 x i32> <i32 undef, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0, i32 0>, i32 %1, i64 0529 %3 = shufflevector <16 x i32> %2, <16 x i32> undef, <2 x i32> <i32 0, i32 1>530 %4 = tail call spir_func <2 x i32> @_Z31sub_group_non_uniform_broadcastDv2_jj(<2 x i32> %3, i32 0)531 %5 = shufflevector <2 x i32> %4, <2 x i32> undef, <16 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>532 %6 = shufflevector <16 x i32> %5, <16 x i32> %2, <16 x i32> <i32 0, i32 1, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>533 %7 = shufflevector <16 x i32> %6, <16 x i32> undef, <3 x i32> <i32 0, i32 1, i32 2>534 %8 = tail call spir_func <3 x i32> @_Z31sub_group_non_uniform_broadcastDv3_jj(<3 x i32> %7, i32 0)535 %9 = shufflevector <3 x i32> %8, <3 x i32> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>536 %10 = shufflevector <16 x i32> %9, <16 x i32> %6, <16 x i32> <i32 0, i32 1, i32 2, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>537 %11 = shufflevector <16 x i32> %10, <16 x i32> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>538 %12 = tail call spir_func <4 x i32> @_Z31sub_group_non_uniform_broadcastDv4_jj(<4 x i32> %11, i32 0)539 %13 = shufflevector <4 x i32> %12, <4 x i32> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>540 %14 = shufflevector <16 x i32> %13, <16 x i32> %10, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>541 %15 = shufflevector <16 x i32> %14, <16 x i32> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>542 %16 = tail call spir_func <8 x i32> @_Z31sub_group_non_uniform_broadcastDv8_jj(<8 x i32> %15, i32 0)543 %17 = shufflevector <8 x i32> %16, <8 x i32> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>544 %18 = shufflevector <16 x i32> %17, <16 x i32> %14, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>545 %19 = tail call spir_func <16 x i32> @_Z31sub_group_non_uniform_broadcastDv16_jj(<16 x i32> %18, i32 0)546 %20 = extractelement <16 x i32> %19, i64 0547 %21 = tail call spir_func i32 @_Z25sub_group_broadcast_firstj(i32 %20)548 ret void549}550 551declare dso_local spir_func i32 @_Z31sub_group_non_uniform_broadcastjj(i32, i32) local_unnamed_addr552 553declare dso_local spir_func <2 x i32> @_Z31sub_group_non_uniform_broadcastDv2_jj(<2 x i32>, i32) local_unnamed_addr554 555declare dso_local spir_func <3 x i32> @_Z31sub_group_non_uniform_broadcastDv3_jj(<3 x i32>, i32) local_unnamed_addr556 557declare dso_local spir_func <4 x i32> @_Z31sub_group_non_uniform_broadcastDv4_jj(<4 x i32>, i32) local_unnamed_addr558 559declare dso_local spir_func <8 x i32> @_Z31sub_group_non_uniform_broadcastDv8_jj(<8 x i32>, i32) local_unnamed_addr560 561declare dso_local spir_func <16 x i32> @_Z31sub_group_non_uniform_broadcastDv16_jj(<16 x i32>, i32) local_unnamed_addr562 563declare dso_local spir_func i32 @_Z25sub_group_broadcast_firstj(i32) local_unnamed_addr564 565; CHECK-SPIRV: OpFunction566; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#long]] %[[#ScopeSubgroup]] %[[#long_0]] %[[#int_0]]567; CHECK-SPIRV: %[[#long2_0:]] = OpVectorShuffle %[[#long2]]568; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#long2]] %[[#ScopeSubgroup]] %[[#long2_0]] %[[#int_0]]569; CHECK-SPIRV: %[[#long3_0:]] = OpVectorShuffle %[[#long3]]570; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#long3]] %[[#ScopeSubgroup]] %[[#long3_0]] %[[#int_0]]571; CHECK-SPIRV: %[[#long4_0:]] = OpVectorShuffle %[[#long4]]572; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#long4]] %[[#ScopeSubgroup]] %[[#long4_0]] %[[#int_0]]573; CHECK-SPIRV: %[[#long8_0:]] = OpVectorShuffle %[[#long8]]574; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#long8]] %[[#ScopeSubgroup]] %[[#long8_0]] %[[#int_0]]575; CHECK-SPIRV: %[[#]] = OpVectorShuffle %[[#long16]]576; CHECK-SPIRV: %[[#long16_0:]] = OpVectorShuffle %[[#long16]]577; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#long16]] %[[#ScopeSubgroup]] %[[#long16_0]] %[[#int_0]]578; CHECK-SPIRV: %[[#long_value:]] = OpCompositeExtract %[[#long]]579; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcastFirst %[[#long]] %[[#ScopeSubgroup]] %[[#long_value]]580; CHECK-SPIRV: OpFunctionEnd581 582define dso_local spir_kernel void @testNonUniformBroadcastLongs() local_unnamed_addr {583 %1 = tail call spir_func i64 @_Z31sub_group_non_uniform_broadcastlj(i64 0, i32 0)584 %2 = insertelement <16 x i64> <i64 undef, i64 0, i64 0, i64 0, i64 0, i64 0, i64 0, i64 0, i64 0, i64 0, i64 0, i64 0, i64 0, i64 0, i64 0, i64 0>, i64 %1, i64 0585 %3 = shufflevector <16 x i64> %2, <16 x i64> undef, <2 x i32> <i32 0, i32 1>586 %4 = tail call spir_func <2 x i64> @_Z31sub_group_non_uniform_broadcastDv2_lj(<2 x i64> %3, i32 0)587 %5 = shufflevector <2 x i64> %4, <2 x i64> undef, <16 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>588 %6 = shufflevector <16 x i64> %5, <16 x i64> %2, <16 x i32> <i32 0, i32 1, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>589 %7 = shufflevector <16 x i64> %6, <16 x i64> undef, <3 x i32> <i32 0, i32 1, i32 2>590 %8 = tail call spir_func <3 x i64> @_Z31sub_group_non_uniform_broadcastDv3_lj(<3 x i64> %7, i32 0)591 %9 = shufflevector <3 x i64> %8, <3 x i64> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>592 %10 = shufflevector <16 x i64> %9, <16 x i64> %6, <16 x i32> <i32 0, i32 1, i32 2, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>593 %11 = shufflevector <16 x i64> %10, <16 x i64> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>594 %12 = tail call spir_func <4 x i64> @_Z31sub_group_non_uniform_broadcastDv4_lj(<4 x i64> %11, i32 0)595 %13 = shufflevector <4 x i64> %12, <4 x i64> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>596 %14 = shufflevector <16 x i64> %13, <16 x i64> %10, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>597 %15 = shufflevector <16 x i64> %14, <16 x i64> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>598 %16 = tail call spir_func <8 x i64> @_Z31sub_group_non_uniform_broadcastDv8_lj(<8 x i64> %15, i32 0)599 %17 = shufflevector <8 x i64> %16, <8 x i64> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>600 %18 = shufflevector <16 x i64> %17, <16 x i64> %14, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>601 %19 = tail call spir_func <16 x i64> @_Z31sub_group_non_uniform_broadcastDv16_lj(<16 x i64> %18, i32 0)602 %20 = extractelement <16 x i64> %19, i64 0603 %21 = tail call spir_func i64 @_Z25sub_group_broadcast_firstl(i64 %20)604 ret void605}606 607declare dso_local spir_func i64 @_Z31sub_group_non_uniform_broadcastlj(i64, i32) local_unnamed_addr608 609declare dso_local spir_func <2 x i64> @_Z31sub_group_non_uniform_broadcastDv2_lj(<2 x i64>, i32) local_unnamed_addr610 611declare dso_local spir_func <3 x i64> @_Z31sub_group_non_uniform_broadcastDv3_lj(<3 x i64>, i32) local_unnamed_addr612 613declare dso_local spir_func <4 x i64> @_Z31sub_group_non_uniform_broadcastDv4_lj(<4 x i64>, i32) local_unnamed_addr614 615declare dso_local spir_func <8 x i64> @_Z31sub_group_non_uniform_broadcastDv8_lj(<8 x i64>, i32) local_unnamed_addr616 617declare dso_local spir_func <16 x i64> @_Z31sub_group_non_uniform_broadcastDv16_lj(<16 x i64>, i32) local_unnamed_addr618 619declare dso_local spir_func i64 @_Z25sub_group_broadcast_firstl(i64) local_unnamed_addr620 621; CHECK-SPIRV: OpFunction622; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#long]] %[[#ScopeSubgroup]] %[[#long_0]] %[[#int_0]]623; CHECK-SPIRV: %[[#long2_0:]] = OpVectorShuffle %[[#long2]]624; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#long2]] %[[#ScopeSubgroup]] %[[#long2_0]] %[[#int_0]]625; CHECK-SPIRV: %[[#long3_0:]] = OpVectorShuffle %[[#long3]]626; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#long3]] %[[#ScopeSubgroup]] %[[#long3_0]] %[[#int_0]]627; CHECK-SPIRV: %[[#long4_0:]] = OpVectorShuffle %[[#long4]]628; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#long4]] %[[#ScopeSubgroup]] %[[#long4_0]] %[[#int_0]]629; CHECK-SPIRV: %[[#long8_0:]] = OpVectorShuffle %[[#long8]]630; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#long8]] %[[#ScopeSubgroup]] %[[#long8_0]] %[[#int_0]]631; CHECK-SPIRV: %[[#]] = OpVectorShuffle %[[#long16]]632; CHECK-SPIRV: %[[#long16_0:]] = OpVectorShuffle %[[#long16]]633; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#long16]] %[[#ScopeSubgroup]] %[[#long16_0]] %[[#int_0]]634; CHECK-SPIRV: %[[#long_value:]] = OpCompositeExtract %[[#long]]635; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcastFirst %[[#long]] %[[#ScopeSubgroup]] %[[#long_value]]636; CHECK-SPIRV: OpFunctionEnd637 638define dso_local spir_kernel void @testNonUniformBroadcastULongs() local_unnamed_addr {639 %1 = tail call spir_func i64 @_Z31sub_group_non_uniform_broadcastmj(i64 0, i32 0)640 %2 = insertelement <16 x i64> <i64 undef, i64 0, i64 0, i64 0, i64 0, i64 0, i64 0, i64 0, i64 0, i64 0, i64 0, i64 0, i64 0, i64 0, i64 0, i64 0>, i64 %1, i64 0641 %3 = shufflevector <16 x i64> %2, <16 x i64> undef, <2 x i32> <i32 0, i32 1>642 %4 = tail call spir_func <2 x i64> @_Z31sub_group_non_uniform_broadcastDv2_mj(<2 x i64> %3, i32 0)643 %5 = shufflevector <2 x i64> %4, <2 x i64> undef, <16 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>644 %6 = shufflevector <16 x i64> %5, <16 x i64> %2, <16 x i32> <i32 0, i32 1, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>645 %7 = shufflevector <16 x i64> %6, <16 x i64> undef, <3 x i32> <i32 0, i32 1, i32 2>646 %8 = tail call spir_func <3 x i64> @_Z31sub_group_non_uniform_broadcastDv3_mj(<3 x i64> %7, i32 0)647 %9 = shufflevector <3 x i64> %8, <3 x i64> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>648 %10 = shufflevector <16 x i64> %9, <16 x i64> %6, <16 x i32> <i32 0, i32 1, i32 2, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>649 %11 = shufflevector <16 x i64> %10, <16 x i64> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>650 %12 = tail call spir_func <4 x i64> @_Z31sub_group_non_uniform_broadcastDv4_mj(<4 x i64> %11, i32 0)651 %13 = shufflevector <4 x i64> %12, <4 x i64> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>652 %14 = shufflevector <16 x i64> %13, <16 x i64> %10, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>653 %15 = shufflevector <16 x i64> %14, <16 x i64> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>654 %16 = tail call spir_func <8 x i64> @_Z31sub_group_non_uniform_broadcastDv8_mj(<8 x i64> %15, i32 0)655 %17 = shufflevector <8 x i64> %16, <8 x i64> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>656 %18 = shufflevector <16 x i64> %17, <16 x i64> %14, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>657 %19 = tail call spir_func <16 x i64> @_Z31sub_group_non_uniform_broadcastDv16_mj(<16 x i64> %18, i32 0)658 %20 = extractelement <16 x i64> %19, i64 0659 %21 = tail call spir_func i64 @_Z25sub_group_broadcast_firstm(i64 %20)660 ret void661}662 663declare dso_local spir_func i64 @_Z31sub_group_non_uniform_broadcastmj(i64, i32) local_unnamed_addr664 665declare dso_local spir_func <2 x i64> @_Z31sub_group_non_uniform_broadcastDv2_mj(<2 x i64>, i32) local_unnamed_addr666 667declare dso_local spir_func <3 x i64> @_Z31sub_group_non_uniform_broadcastDv3_mj(<3 x i64>, i32) local_unnamed_addr668 669declare dso_local spir_func <4 x i64> @_Z31sub_group_non_uniform_broadcastDv4_mj(<4 x i64>, i32) local_unnamed_addr670 671declare dso_local spir_func <8 x i64> @_Z31sub_group_non_uniform_broadcastDv8_mj(<8 x i64>, i32) local_unnamed_addr672 673declare dso_local spir_func <16 x i64> @_Z31sub_group_non_uniform_broadcastDv16_mj(<16 x i64>, i32) local_unnamed_addr674 675declare dso_local spir_func i64 @_Z25sub_group_broadcast_firstm(i64) local_unnamed_addr676 677; CHECK-SPIRV: OpFunction678; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#float]] %[[#ScopeSubgroup]] %[[#float_0]] %[[#int_0]]679; CHECK-SPIRV: %[[#float2_0:]] = OpVectorShuffle %[[#float2]]680; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#float2]] %[[#ScopeSubgroup]] %[[#float2_0]] %[[#int_0]]681; CHECK-SPIRV: %[[#float3_0:]] = OpVectorShuffle %[[#float3]]682; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#float3]] %[[#ScopeSubgroup]] %[[#float3_0]] %[[#int_0]]683; CHECK-SPIRV: %[[#float4_0:]] = OpVectorShuffle %[[#float4]]684; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#float4]] %[[#ScopeSubgroup]] %[[#float4_0]] %[[#int_0]]685; CHECK-SPIRV: %[[#float8_0:]] = OpVectorShuffle %[[#float8]]686; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#float8]] %[[#ScopeSubgroup]] %[[#float8_0]] %[[#int_0]]687; CHECK-SPIRV: %[[#]] = OpVectorShuffle %[[#float16]]688; CHECK-SPIRV: %[[#float16_0:]] = OpVectorShuffle %[[#float16]]689; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#float16]] %[[#ScopeSubgroup]] %[[#float16_0]] %[[#int_0]]690; CHECK-SPIRV: %[[#float_value:]] = OpCompositeExtract %[[#float]]691; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcastFirst %[[#float]] %[[#ScopeSubgroup]] %[[#float_value]]692; CHECK-SPIRV: OpFunctionEnd693 694define dso_local spir_kernel void @testNonUniformBroadcastFloats() local_unnamed_addr {695 %1 = tail call spir_func float @_Z31sub_group_non_uniform_broadcastfj(float 0.000000e+00, i32 0)696 %2 = insertelement <16 x float> <float undef, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00, float 0.000000e+00>, float %1, i64 0697 %3 = shufflevector <16 x float> %2, <16 x float> undef, <2 x i32> <i32 0, i32 1>698 %4 = tail call spir_func <2 x float> @_Z31sub_group_non_uniform_broadcastDv2_fj(<2 x float> %3, i32 0)699 %5 = shufflevector <2 x float> %4, <2 x float> undef, <16 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>700 %6 = shufflevector <16 x float> %5, <16 x float> %2, <16 x i32> <i32 0, i32 1, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>701 %7 = shufflevector <16 x float> %6, <16 x float> undef, <3 x i32> <i32 0, i32 1, i32 2>702 %8 = tail call spir_func <3 x float> @_Z31sub_group_non_uniform_broadcastDv3_fj(<3 x float> %7, i32 0)703 %9 = shufflevector <3 x float> %8, <3 x float> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>704 %10 = shufflevector <16 x float> %9, <16 x float> %6, <16 x i32> <i32 0, i32 1, i32 2, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>705 %11 = shufflevector <16 x float> %10, <16 x float> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>706 %12 = tail call spir_func <4 x float> @_Z31sub_group_non_uniform_broadcastDv4_fj(<4 x float> %11, i32 0)707 %13 = shufflevector <4 x float> %12, <4 x float> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>708 %14 = shufflevector <16 x float> %13, <16 x float> %10, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>709 %15 = shufflevector <16 x float> %14, <16 x float> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>710 %16 = tail call spir_func <8 x float> @_Z31sub_group_non_uniform_broadcastDv8_fj(<8 x float> %15, i32 0)711 %17 = shufflevector <8 x float> %16, <8 x float> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>712 %18 = shufflevector <16 x float> %17, <16 x float> %14, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>713 %19 = tail call spir_func <16 x float> @_Z31sub_group_non_uniform_broadcastDv16_fj(<16 x float> %18, i32 0)714 %20 = extractelement <16 x float> %19, i64 0715 %21 = tail call spir_func float @_Z25sub_group_broadcast_firstf(float %20)716 ret void717}718 719declare dso_local spir_func float @_Z31sub_group_non_uniform_broadcastfj(float, i32) local_unnamed_addr720 721declare dso_local spir_func <2 x float> @_Z31sub_group_non_uniform_broadcastDv2_fj(<2 x float>, i32) local_unnamed_addr722 723declare dso_local spir_func <3 x float> @_Z31sub_group_non_uniform_broadcastDv3_fj(<3 x float>, i32) local_unnamed_addr724 725declare dso_local spir_func <4 x float> @_Z31sub_group_non_uniform_broadcastDv4_fj(<4 x float>, i32) local_unnamed_addr726 727declare dso_local spir_func <8 x float> @_Z31sub_group_non_uniform_broadcastDv8_fj(<8 x float>, i32) local_unnamed_addr728 729declare dso_local spir_func <16 x float> @_Z31sub_group_non_uniform_broadcastDv16_fj(<16 x float>, i32) local_unnamed_addr730 731declare dso_local spir_func float @_Z25sub_group_broadcast_firstf(float) local_unnamed_addr732 733; CHECK-SPIRV: OpFunction734; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#half]] %[[#ScopeSubgroup]] %[[#half_0]] %[[#int_0]]735; CHECK-SPIRV: %[[#half2_0:]] = OpVectorShuffle %[[#half2]]736; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#half2]] %[[#ScopeSubgroup]] %[[#half2_0]] %[[#int_0]]737; CHECK-SPIRV: %[[#half3_0:]] = OpVectorShuffle %[[#half3]]738; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#half3]] %[[#ScopeSubgroup]] %[[#half3_0]] %[[#int_0]]739; CHECK-SPIRV: %[[#half4_0:]] = OpVectorShuffle %[[#half4]]740; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#half4]] %[[#ScopeSubgroup]] %[[#half4_0]] %[[#int_0]]741; CHECK-SPIRV: %[[#half8_0:]] = OpVectorShuffle %[[#half8]]742; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#half8]] %[[#ScopeSubgroup]] %[[#half8_0]] %[[#int_0]]743; CHECK-SPIRV: %[[#]] = OpVectorShuffle %[[#half16]]744; CHECK-SPIRV: %[[#half16_0:]] = OpVectorShuffle %[[#half16]]745; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#half16]] %[[#ScopeSubgroup]] %[[#half16_0]] %[[#int_0]]746; CHECK-SPIRV: %[[#half_value:]] = OpCompositeExtract %[[#half]]747; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcastFirst %[[#half]] %[[#ScopeSubgroup]] %[[#half_value]]748; CHECK-SPIRV: OpFunctionEnd749 750define dso_local spir_kernel void @testNonUniformBroadcastHalfs() local_unnamed_addr {751 %1 = tail call spir_func half @_Z31sub_group_non_uniform_broadcastDhj(half 0xH0000, i32 0)752 %2 = insertelement <16 x half> <half undef, half 0xH0000, half 0xH0000, half 0xH0000, half 0xH0000, half 0xH0000, half 0xH0000, half 0xH0000, half 0xH0000, half 0xH0000, half 0xH0000, half 0xH0000, half 0xH0000, half 0xH0000, half 0xH0000, half 0xH0000>, half %1, i64 0753 %3 = shufflevector <16 x half> %2, <16 x half> undef, <2 x i32> <i32 0, i32 1>754 %4 = tail call spir_func <2 x half> @_Z31sub_group_non_uniform_broadcastDv2_Dhj(<2 x half> %3, i32 0)755 %5 = shufflevector <2 x half> %4, <2 x half> undef, <16 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>756 %6 = shufflevector <16 x half> %5, <16 x half> %2, <16 x i32> <i32 0, i32 1, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>757 %7 = shufflevector <16 x half> %6, <16 x half> undef, <3 x i32> <i32 0, i32 1, i32 2>758 %8 = tail call spir_func <3 x half> @_Z31sub_group_non_uniform_broadcastDv3_Dhj(<3 x half> %7, i32 0)759 %9 = shufflevector <3 x half> %8, <3 x half> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>760 %10 = shufflevector <16 x half> %9, <16 x half> %6, <16 x i32> <i32 0, i32 1, i32 2, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>761 %11 = shufflevector <16 x half> %10, <16 x half> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>762 %12 = tail call spir_func <4 x half> @_Z31sub_group_non_uniform_broadcastDv4_Dhj(<4 x half> %11, i32 0)763 %13 = shufflevector <4 x half> %12, <4 x half> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>764 %14 = shufflevector <16 x half> %13, <16 x half> %10, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>765 %15 = shufflevector <16 x half> %14, <16 x half> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>766 %16 = tail call spir_func <8 x half> @_Z31sub_group_non_uniform_broadcastDv8_Dhj(<8 x half> %15, i32 0)767 %17 = shufflevector <8 x half> %16, <8 x half> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>768 %18 = shufflevector <16 x half> %17, <16 x half> %14, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>769 %19 = tail call spir_func <16 x half> @_Z31sub_group_non_uniform_broadcastDv16_Dhj(<16 x half> %18, i32 0)770 %20 = extractelement <16 x half> %19, i64 0771 %21 = tail call spir_func half @_Z25sub_group_broadcast_firstDh(half %20)772 ret void773}774 775declare dso_local spir_func half @_Z31sub_group_non_uniform_broadcastDhj(half, i32) local_unnamed_addr776 777declare dso_local spir_func <2 x half> @_Z31sub_group_non_uniform_broadcastDv2_Dhj(<2 x half>, i32) local_unnamed_addr778 779declare dso_local spir_func <3 x half> @_Z31sub_group_non_uniform_broadcastDv3_Dhj(<3 x half>, i32) local_unnamed_addr780 781declare dso_local spir_func <4 x half> @_Z31sub_group_non_uniform_broadcastDv4_Dhj(<4 x half>, i32) local_unnamed_addr782 783declare dso_local spir_func <8 x half> @_Z31sub_group_non_uniform_broadcastDv8_Dhj(<8 x half>, i32) local_unnamed_addr784 785declare dso_local spir_func <16 x half> @_Z31sub_group_non_uniform_broadcastDv16_Dhj(<16 x half>, i32) local_unnamed_addr786 787declare dso_local spir_func half @_Z25sub_group_broadcast_firstDh(half) local_unnamed_addr788 789; CHECK-SPIRV: OpFunction790; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#double]] %[[#ScopeSubgroup]] %[[#double_0]] %[[#int_0]]791; CHECK-SPIRV: %[[#double2_0:]] = OpVectorShuffle %[[#double2]]792; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#double2]] %[[#ScopeSubgroup]] %[[#double2_0]] %[[#int_0]]793; CHECK-SPIRV: %[[#double3_0:]] = OpVectorShuffle %[[#double3]]794; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#double3]] %[[#ScopeSubgroup]] %[[#double3_0]] %[[#int_0]]795; CHECK-SPIRV: %[[#double4_0:]] = OpVectorShuffle %[[#double4]]796; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#double4]] %[[#ScopeSubgroup]] %[[#double4_0]] %[[#int_0]]797; CHECK-SPIRV: %[[#double8_0:]] = OpVectorShuffle %[[#double8]]798; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#double8]] %[[#ScopeSubgroup]] %[[#double8_0]] %[[#int_0]]799; CHECK-SPIRV: %[[#]] = OpVectorShuffle %[[#double16]]800; CHECK-SPIRV: %[[#double16_0:]] = OpVectorShuffle %[[#double16]]801; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcast %[[#double16]] %[[#ScopeSubgroup]] %[[#double16_0]] %[[#int_0]]802; CHECK-SPIRV: %[[#double_value:]] = OpCompositeExtract %[[#double]]803; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBroadcastFirst %[[#double]] %[[#ScopeSubgroup]] %[[#double_value]]804; CHECK-SPIRV: OpFunctionEnd805 806define dso_local spir_kernel void @testNonUniformBroadcastDoubles() local_unnamed_addr {807 %1 = tail call spir_func double @_Z31sub_group_non_uniform_broadcastdj(double 0.000000e+00, i32 0)808 %2 = insertelement <16 x double> <double undef, double 0.000000e+00, double 0.000000e+00, double 0.000000e+00, double 0.000000e+00, double 0.000000e+00, double 0.000000e+00, double 0.000000e+00, double 0.000000e+00, double 0.000000e+00, double 0.000000e+00, double 0.000000e+00, double 0.000000e+00, double 0.000000e+00, double 0.000000e+00, double 0.000000e+00>, double %1, i64 0809 %3 = shufflevector <16 x double> %2, <16 x double> undef, <2 x i32> <i32 0, i32 1>810 %4 = tail call spir_func <2 x double> @_Z31sub_group_non_uniform_broadcastDv2_dj(<2 x double> %3, i32 0)811 %5 = shufflevector <2 x double> %4, <2 x double> undef, <16 x i32> <i32 0, i32 1, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>812 %6 = shufflevector <16 x double> %5, <16 x double> %2, <16 x i32> <i32 0, i32 1, i32 18, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>813 %7 = shufflevector <16 x double> %6, <16 x double> undef, <3 x i32> <i32 0, i32 1, i32 2>814 %8 = tail call spir_func <3 x double> @_Z31sub_group_non_uniform_broadcastDv3_dj(<3 x double> %7, i32 0)815 %9 = shufflevector <3 x double> %8, <3 x double> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>816 %10 = shufflevector <16 x double> %9, <16 x double> %6, <16 x i32> <i32 0, i32 1, i32 2, i32 19, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>817 %11 = shufflevector <16 x double> %10, <16 x double> undef, <4 x i32> <i32 0, i32 1, i32 2, i32 3>818 %12 = tail call spir_func <4 x double> @_Z31sub_group_non_uniform_broadcastDv4_dj(<4 x double> %11, i32 0)819 %13 = shufflevector <4 x double> %12, <4 x double> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>820 %14 = shufflevector <16 x double> %13, <16 x double> %10, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 20, i32 21, i32 22, i32 23, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>821 %15 = shufflevector <16 x double> %14, <16 x double> undef, <8 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7>822 %16 = tail call spir_func <8 x double> @_Z31sub_group_non_uniform_broadcastDv8_dj(<8 x double> %15, i32 0)823 %17 = shufflevector <8 x double> %16, <8 x double> undef, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef, i32 undef>824 %18 = shufflevector <16 x double> %17, <16 x double> %14, <16 x i32> <i32 0, i32 1, i32 2, i32 3, i32 4, i32 5, i32 6, i32 7, i32 24, i32 25, i32 26, i32 27, i32 28, i32 29, i32 30, i32 31>825 %19 = tail call spir_func <16 x double> @_Z31sub_group_non_uniform_broadcastDv16_dj(<16 x double> %18, i32 0)826 %20 = extractelement <16 x double> %19, i64 0827 %21 = tail call spir_func double @_Z25sub_group_broadcast_firstd(double %20)828 ret void829}830 831declare dso_local spir_func double @_Z31sub_group_non_uniform_broadcastdj(double, i32) local_unnamed_addr832 833declare dso_local spir_func <2 x double> @_Z31sub_group_non_uniform_broadcastDv2_dj(<2 x double>, i32) local_unnamed_addr834 835declare dso_local spir_func <3 x double> @_Z31sub_group_non_uniform_broadcastDv3_dj(<3 x double>, i32) local_unnamed_addr836 837declare dso_local spir_func <4 x double> @_Z31sub_group_non_uniform_broadcastDv4_dj(<4 x double>, i32) local_unnamed_addr838 839declare dso_local spir_func <8 x double> @_Z31sub_group_non_uniform_broadcastDv8_dj(<8 x double>, i32) local_unnamed_addr840 841declare dso_local spir_func <16 x double> @_Z31sub_group_non_uniform_broadcastDv16_dj(<16 x double>, i32) local_unnamed_addr842 843declare dso_local spir_func double @_Z25sub_group_broadcast_firstd(double) local_unnamed_addr844 845; CHECK-SPIRV: OpFunction846; CHECK-SPIRV: %[[#ballot:]] = OpGroupNonUniformBallot %[[#int4]] %[[#ScopeSubgroup]] %[[#false]]847; CHECK-SPIRV: %[[#ballot2:]] = OpGroupNonUniformBallot %[[#int4]] %[[#ScopeSubgroup]] %[[#false]]848; CHECK-SPIRV: %[[#]] = OpGroupNonUniformInverseBallot %[[#bool]] %[[#ScopeSubgroup]] %[[#ballot]]849; CHECK-SPIRV: %[[#]] = OpGroupNonUniformInverseBallot %[[#bool]] %[[#ScopeSubgroup]] %[[#ballot2]]850; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBallotBitExtract %[[#bool]] %[[#ScopeSubgroup]] %[[#ballot]] %[[#int_0]]851; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBallotBitExtract %[[#bool]] %[[#ScopeSubgroup]] %[[#ballot2]] %[[#int_0]]852; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBallotBitCount %[[#int]] %[[#ScopeSubgroup]] Reduce %[[#ballot]]853; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBallotBitCount %[[#int]] %[[#ScopeSubgroup]] Reduce %[[#ballot2]]854; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBallotBitCount %[[#int]] %[[#ScopeSubgroup]] InclusiveScan %[[#ballot]]855; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBallotBitCount %[[#int]] %[[#ScopeSubgroup]] InclusiveScan %[[#ballot2]]856; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBallotBitCount %[[#int]] %[[#ScopeSubgroup]] ExclusiveScan %[[#ballot]]857; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBallotBitCount %[[#int]] %[[#ScopeSubgroup]] ExclusiveScan %[[#ballot2]]858; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBallotFindLSB %[[#int]] %[[#ScopeSubgroup]] %[[#ballot]]859; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBallotFindLSB %[[#int]] %[[#ScopeSubgroup]] %[[#ballot2]]860; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBallotFindMSB %[[#int]] %[[#ScopeSubgroup]] %[[#ballot]]861; CHECK-SPIRV: %[[#]] = OpGroupNonUniformBallotFindMSB %[[#int]] %[[#ScopeSubgroup]] %[[#ballot2]]862; CHECK-SPIRV: OpFunctionEnd863 864define dso_local spir_kernel void @testBallotOperations(i32 addrspace(1)* nocapture) local_unnamed_addr {865 %2 = tail call spir_func <4 x i32> @_Z16sub_group_balloti(i32 0)866 %r2 = tail call spir_func <4 x i32> @__spirv_GroupNonUniformBallot(i32 3, i1 false)867 %3 = tail call spir_func i32 @_Z24sub_group_inverse_ballotDv4_j(<4 x i32> %2)868 %r3 = tail call spir_func i1 @__spirv_GroupNonUniformInverseBallot(i32 3, <4 x i32> %r2)869 store i32 %3, i32 addrspace(1)* %0, align 4870 %4 = tail call spir_func i32 @_Z28sub_group_ballot_bit_extractDv4_jj(<4 x i32> %2, i32 0)871 %r4 = tail call spir_func i32 @__spirv_GroupNonUniformBallotBitExtract(i32 3, <4 x i32> %r2, i32 0)872 %5 = getelementptr inbounds i32, i32 addrspace(1)* %0, i64 1873 store i32 %4, i32 addrspace(1)* %5, align 4874 %6 = tail call spir_func i32 @_Z26sub_group_ballot_bit_countDv4_j(<4 x i32> %2)875 %r6 = tail call spir_func i32 @__spirv_GroupNonUniformBallotBitCount(i32 3, i32 0, <4 x i32> %r2)876 %7 = getelementptr inbounds i32, i32 addrspace(1)* %0, i64 2877 store i32 %6, i32 addrspace(1)* %7, align 4878 %8 = tail call spir_func i32 @_Z31sub_group_ballot_inclusive_scanDv4_j(<4 x i32> %2)879 %r8 = tail call spir_func i32 @__spirv_GroupNonUniformBallotBitCount(i32 3, i32 1, <4 x i32> %r2)880 %9 = getelementptr inbounds i32, i32 addrspace(1)* %0, i64 3881 store i32 %8, i32 addrspace(1)* %9, align 4882 %10 = tail call spir_func i32 @_Z31sub_group_ballot_exclusive_scanDv4_j(<4 x i32> %2)883 %r10 = tail call spir_func i32 @__spirv_GroupNonUniformBallotBitCount(i32 3, i32 2, <4 x i32> %r2)884 %11 = getelementptr inbounds i32, i32 addrspace(1)* %0, i64 4885 store i32 %10, i32 addrspace(1)* %11, align 4886 %12 = tail call spir_func i32 @_Z25sub_group_ballot_find_lsbDv4_j(<4 x i32> %2)887 %r12 = tail call spir_func i32 @__spirv_GroupNonUniformBallotFindLSB(i32 3, <4 x i32> %r2)888 %13 = getelementptr inbounds i32, i32 addrspace(1)* %0, i64 5889 store i32 %12, i32 addrspace(1)* %13, align 4890 %14 = tail call spir_func i32 @_Z25sub_group_ballot_find_msbDv4_j(<4 x i32> %2)891 %r14 = tail call spir_func i32 @__spirv_GroupNonUniformBallotFindMSB(i32 3, <4 x i32> %r2)892 %15 = getelementptr inbounds i32, i32 addrspace(1)* %0, i64 6893 store i32 %14, i32 addrspace(1)* %15, align 4894 ret void895}896 897declare dso_local spir_func <4 x i32> @_Z16sub_group_balloti(i32) local_unnamed_addr898declare dso_local spir_func <4 x i32> @__spirv_GroupNonUniformBallot(i32, i1)899 900declare dso_local spir_func i32 @_Z24sub_group_inverse_ballotDv4_j(<4 x i32>) local_unnamed_addr901declare dso_local spir_func i1 @__spirv_GroupNonUniformInverseBallot(i32, <4 x i32>)902 903declare dso_local spir_func i32 @_Z28sub_group_ballot_bit_extractDv4_jj(<4 x i32>, i32) local_unnamed_addr904declare dso_local spir_func i1 @__spirv_GroupNonUniformBallotBitExtract(i32, <4 x i32>, i32) local_unnamed_addr905 906declare dso_local spir_func i32 @_Z26sub_group_ballot_bit_countDv4_j(<4 x i32>) local_unnamed_addr907declare dso_local spir_func i32 @__spirv_GroupNonUniformBallotBitCount(i32, i32, <4 x i32>)908 909declare dso_local spir_func i32 @_Z31sub_group_ballot_inclusive_scanDv4_j(<4 x i32>) local_unnamed_addr910 911declare dso_local spir_func i32 @_Z31sub_group_ballot_exclusive_scanDv4_j(<4 x i32>) local_unnamed_addr912 913declare dso_local spir_func i32 @_Z25sub_group_ballot_find_lsbDv4_j(<4 x i32>) local_unnamed_addr914declare dso_local spir_func i32 @__spirv_GroupNonUniformBallotFindLSB(i32, <4 x i32>)915 916declare dso_local spir_func i32 @_Z25sub_group_ballot_find_msbDv4_j(<4 x i32>) local_unnamed_addr917declare dso_local spir_func i32 @__spirv_GroupNonUniformBallotFindMSB(i32, <4 x i32>)918 919; CHECK-SPIRV: OpFunction920; CHECK-SPIRV: %[[#]] = OpLoad %[[#int4]] %[[#eqMask]]921; CHECK-SPIRV: %[[#]] = OpLoad %[[#int4]] %[[#geMask]]922; CHECK-SPIRV: %[[#]] = OpLoad %[[#int4]] %[[#gtMask]]923; CHECK-SPIRV: %[[#]] = OpLoad %[[#int4]] %[[#leMask]]924; CHECK-SPIRV: %[[#]] = OpLoad %[[#int4]] %[[#ltMask]]925; CHECK-SPIRV: OpFunctionEnd926 927define dso_local spir_kernel void @testSubgroupMasks(<4 x i32> addrspace(1)* nocapture) local_unnamed_addr {928 %2 = tail call spir_func <4 x i32> @_Z21get_sub_group_eq_maskv()929 store <4 x i32> %2, <4 x i32> addrspace(1)* %0, align 16930 %3 = tail call spir_func <4 x i32> @_Z21get_sub_group_ge_maskv()931 %4 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %0, i64 1932 store <4 x i32> %3, <4 x i32> addrspace(1)* %4, align 16933 %5 = tail call spir_func <4 x i32> @_Z21get_sub_group_gt_maskv()934 %6 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %0, i64 2935 store <4 x i32> %5, <4 x i32> addrspace(1)* %6, align 16936 %7 = tail call spir_func <4 x i32> @_Z21get_sub_group_le_maskv()937 %8 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %0, i64 3938 store <4 x i32> %7, <4 x i32> addrspace(1)* %8, align 16939 %9 = tail call spir_func <4 x i32> @_Z21get_sub_group_lt_maskv()940 %10 = getelementptr inbounds <4 x i32>, <4 x i32> addrspace(1)* %0, i64 4941 store <4 x i32> %9, <4 x i32> addrspace(1)* %10, align 16942 ret void943}944 945declare dso_local spir_func <4 x i32> @_Z21get_sub_group_eq_maskv() local_unnamed_addr946 947declare dso_local spir_func <4 x i32> @_Z21get_sub_group_ge_maskv() local_unnamed_addr948 949declare dso_local spir_func <4 x i32> @_Z21get_sub_group_gt_maskv() local_unnamed_addr950 951declare dso_local spir_func <4 x i32> @_Z21get_sub_group_le_maskv() local_unnamed_addr952 953declare dso_local spir_func <4 x i32> @_Z21get_sub_group_lt_maskv() local_unnamed_addr954