76 lines · plain
1; RUN: opt < %s -mtriple=x86_64-unknown-unknown -mattr=+amx-int8 -mattr=+avx512f -x86-lower-amx-type -S | FileCheck %s2; RUN: opt < %s -mtriple=x86_64-unknown-unknown -mattr=+amx-int8 -mattr=+avx512f -passes=x86-lower-amx-type -S | FileCheck %s3 4@buf = dso_local global [1024 x i8] zeroinitializer, align 165@buf2 = dso_local global [1024 x i8] zeroinitializer, align 166 7; Function Attrs: nounwind uwtable8define dso_local void @test_api(i32 %cond, i16 signext %row, i16 signext %col) local_unnamed_addr {9 10; CHECK-LABEL: entry:11; CHECK: %{{[0-9]+}} = alloca <256 x i32>, align 102412; CHECK-NEXT: %{{[0-9]+}} = alloca <256 x i32>, align 102413; CHECK-NEXT: %{{[0-9]+}} = alloca <256 x i32>, align 102414; CHECK-NEXT: %{{[0-9]+}} = alloca <256 x i32>, align 102415; CHECK-NEXT: %tobool.not = icmp eq i32 %cond, 016; CHECK-NEXT: br i1 %tobool.not, label %if.else, label %if.then17; CHECK: if.then:18; CHECK-NEXT: %{{[0-9]+}} = tail call x86_amx @llvm.x86.tileloadd64.internal(i16 %row, i16 8, ptr @buf, i64 32)19; CHECK-NEXT: call void @llvm.x86.tilestored64.internal(i16 %row, i16 8, ptr %{{[0-9]+}}, i64 64, x86_amx %{{[0-9]+}})20; CHECK-NEXT: %{{[0-9]+}} = tail call x86_amx @llvm.x86.tileloadd64.internal(i16 8, i16 %col, ptr @buf, i64 32)21; CHECK-NEXT: call void @llvm.x86.tilestored64.internal(i16 8, i16 %col, ptr %{{[0-9]+}}, i64 64, x86_amx %{{[0-9]+}})22; CHECK-NEXT: %{{[0-9]+}} = tail call x86_amx @llvm.x86.tileloadd64.internal(i16 %row, i16 %col, ptr @buf, i64 32)23; CHECK-NEXT: call void @llvm.x86.tilestored64.internal(i16 %row, i16 %col, ptr %{{[0-9]+}}, i64 64, x86_amx %{{[0-9]+}})24; CHECK-NEXT: br label %if.end25; CHECK: if.else:26; CHECK-NEXT: %{{[0-9]+}} = tail call x86_amx @llvm.x86.tileloadd64.internal(i16 %row, i16 8, ptr @buf2, i64 32)27; CHECK-NEXT: call void @llvm.x86.tilestored64.internal(i16 %row, i16 8, ptr %{{[0-9]+}}, i64 64, x86_amx %{{[0-9]+}})28; CHECK-NEXT: %{{[0-9]+}} = tail call x86_amx @llvm.x86.tileloadd64.internal(i16 8, i16 %col, ptr @buf2, i64 32)29; CHECK-NEXT: call void @llvm.x86.tilestored64.internal(i16 8, i16 %col, ptr %{{[0-9]+}}, i64 64, x86_amx %{{[0-9]+}})30; CHECK-NEXT: %{{[0-9]+}} = tail call x86_amx @llvm.x86.tileloadd64.internal(i16 %row, i16 %col, ptr @buf2, i64 32)31; CHECK-NEXT: call void @llvm.x86.tilestored64.internal(i16 %row, i16 %col, ptr %{{[0-9]+}}, i64 64, x86_amx %{{[0-9]+}})32; CHECK-NEXT: br label %if.end33; CHECK: if.end:34; CHECK-NEXT: %{{[0-9]+}} = call x86_amx @llvm.x86.tileloadd64.internal(i16 %row, i16 8, ptr %{{[0-9]+}}, i64 64)35; CHECK-NEXT: %{{[0-9]+}} = call x86_amx @llvm.x86.tileloadd64.internal(i16 8, i16 %col, ptr %{{[0-9]+}}, i64 64)36; CHECK-NEXT: %{{[0-9]+}} = call x86_amx @llvm.x86.tileloadd64.internal(i16 %row, i16 %col, ptr %{{[0-9]+}}, i64 64)37; CHECK-NEXT: %{{[0-9]+}} = tail call x86_amx @llvm.x86.tdpbssd.internal(i16 %row, i16 %col, i16 8, x86_amx %{{[0-9]+}}, x86_amx %{{[0-9]+}}, x86_amx %{{[0-9]+}})38; CHECK-NEXT: call void @llvm.x86.tilestored64.internal(i16 %row, i16 %col, ptr %{{[0-9]+}}, i64 64, x86_amx %{{[0-9]+}})39; CHECK-NEXT: %{{[0-9]+}} = call x86_amx @llvm.x86.tileloadd64.internal(i16 %row, i16 %col, ptr %{{[0-9]+}}, i64 64)40; CHECK-NEXT: tail call void @llvm.x86.tilestored64.internal(i16 %row, i16 %col, ptr @buf, i64 32, x86_amx %{{[0-9]+}})41; CHECK-NEXT: ret void42 43entry:44 %tobool.not = icmp eq i32 %cond, 045 br i1 %tobool.not, label %if.else, label %if.then46 47if.then: ; preds = %entry48 %0 = tail call x86_amx @llvm.x86.tileloadd64.internal(i16 %row, i16 8, ptr @buf, i64 32)49 %1 = tail call x86_amx @llvm.x86.tileloadd64.internal(i16 8, i16 %col, ptr @buf, i64 32)50 %2 = tail call x86_amx @llvm.x86.tileloadd64.internal(i16 %row, i16 %col, ptr @buf, i64 32)51 br label %if.end52 53if.else: ; preds = %entry54 %3 = tail call x86_amx @llvm.x86.tileloadd64.internal(i16 %row, i16 8, ptr @buf2, i64 32)55 %4 = tail call x86_amx @llvm.x86.tileloadd64.internal(i16 8, i16 %col, ptr @buf2, i64 32)56 %5 = tail call x86_amx @llvm.x86.tileloadd64.internal(i16 %row, i16 %col, ptr @buf2, i64 32)57 br label %if.end58 59if.end: ; preds = %if.else, %if.then60 %a.sroa.1094.0.in = phi x86_amx [ %3, %if.else ], [ %0, %if.then ]61 %b.sroa.1069.0.in = phi x86_amx [ %4, %if.else ], [ %1, %if.then ]62 %c.sroa.1044.0.in = phi x86_amx [ %5, %if.else ], [ %2, %if.then ]63 %6 = tail call x86_amx @llvm.x86.tdpbssd.internal(i16 %row, i16 %col, i16 8, x86_amx %c.sroa.1044.0.in, x86_amx %a.sroa.1094.0.in, x86_amx %b.sroa.1069.0.in)64 tail call void @llvm.x86.tilestored64.internal(i16 %row, i16 %col, ptr @buf, i64 32, x86_amx %6)65 ret void66}67 68; Function Attrs: nounwind69declare x86_amx @llvm.x86.tileloadd64.internal(i16, i16, ptr, i64)70 71; Function Attrs: nounwind72declare x86_amx @llvm.x86.tdpbssd.internal(i16, i16, i16, x86_amx, x86_amx, x86_amx)73 74; Function Attrs: nounwind75declare void @llvm.x86.tilestored64.internal(i16, i16, ptr, i64, x86_amx)76