AMDGPU: Add runtime metadata KeyArgSExt and KeyArgZExt

[yaxunl]

Summary

Runtime sign ext or zero ext char/short types to 32 bit. We need a metadata to tell runtime whether to sign ext or zero ext.

[yaxunl]

We will follow HSA kernel function ABI so that runtime will not sign/zero extend kernel arguments.

lib/Target/AMDGPU/AMDGPUAsmPrinter.cpp

   }
 }
 
+// \param SExt Argument is sign extended by runtime.
+// \param ZExt Argument is zero extended by runtime.
 static void emitRuntimeMetadataForKernelArg(const DataLayout &DL,
     MCStreamer &OutStreamer, Type *T,
     RuntimeMD::KernelArg::Kind Kind,
+    bool SExt = false, bool ZExt = false,
     StringRef BaseTypeName = "", StringRef TypeName = "",
     StringRef ArgName = "", StringRef TypeQual = "", StringRef AccQual = "") {
   // Emit KeyArgBegin.
   OutStreamer.EmitIntValue(RuntimeMD::KeyArgBegin, 1);
 
   // Emit KeyArgSize and KeyArgAlign.
+  // For scalar integer argument smaller than 32 bit, runtime extends them to
+  // 32 bit, therefore their size is 4 bytes.
+  bool Ext = SExt || ZExt;
   emitRuntimeMDIntValue(OutStreamer, RuntimeMD::KeyArgSize,
-                        DL.getTypeAllocSize(T), 4);
+                        Ext ? 4 : DL.getTypeAllocSize(T), 4);
   emitRuntimeMDIntValue(OutStreamer, RuntimeMD::KeyArgAlign,
-                        DL.getABITypeAlignment(T), 4);
+                        Ext ? 4 : DL.getABITypeAlignment(T), 4);
   if (auto PT = dyn_cast<PointerType>(T)) {
     auto ET = PT->getElementType();
     if (PT->getAddressSpace() == AMDGPUAS::LOCAL_ADDRESS && ET->isSized())

                         DL.getABITypeAlignment(ET), 4);
   }
 
+  // Emit KeyArgSExt and KeyArgZExt.
+  if (SExt)
+    OutStreamer.EmitIntValue(RuntimeMD::KeyArgSExt, 1);
+  if (ZExt)
+    OutStreamer.EmitIntValue(RuntimeMD::KeyArgZExt, 1);
+
   // Emit KeyArgTypeName.
   if (!TypeName.empty())
     emitRuntimeMDStringValue(OutStreamer, RuntimeMD::KeyArgTypeName, TypeName);

   for (auto &Arg : F.args()) {
     unsigned I = Arg.getArgNo();
     Type *T = Arg.getType();
+    // For scalar integer type kernel argument smaller than 32 bit, runtime
+    // extends them to 32 bit.
+    bool Ext = T->isIntegerTy() && T->getIntegerBitWidth() < 32;
     auto TypeName = dyn_cast<MDString>(F.getMetadata(
         "kernel_arg_type")->getOperand(I))->getString();
     auto BaseTypeName = cast<MDString>(F.getMetadata(

                    RuntimeMD::KernelArg::GlobalBuffer) :
                    RuntimeMD::KernelArg::ByValue);
     emitRuntimeMetadataForKernelArg(DL, *OutStreamer, T,
-        Kind, BaseTypeName, TypeName, ArgName, TypeQual, AccQual);
+                                    Kind, Ext ? Arg.hasSExtAttr() : false,
+                                    Ext ? Arg.hasZExtAttr() : false,
+                                    BaseTypeName, TypeName, ArgName, TypeQual,
+                                    AccQual);
   }
 
   // Emit hidden kernel arguments for OpenCL kernels.

lib/Target/AMDGPU/AMDGPURuntimeMetadata.h

     KeyPrintfInfo               = 30, // Prinf function call information
     KeyArgActualAcc             = 31, // The actual kernel argument access qualifier
     KeyArgPointeeAlign          = 32, // Alignment of pointee type
+    KeyArgSExt                  = 33, // Argument needs sign extension
+    KeyArgZExt                  = 34, // Argument needs zero extension
   };
 
   enum Language : uint8_t {

test/CodeGen/AMDGPU/runtime-metadata.ll

 ; CHECK-NEXT: .ascii  "test_char"
 ; CHECK-NEXT: .byte 7
 ; CHECK-NEXT: .byte 9
-; CHECK-NEXT: .long 1
+; CHECK-NEXT: .long 4
 ; CHECK-NEXT: .byte 10
-; CHECK-NEXT: .long 1
+; CHECK-NEXT: .long 4
+; CHECK-NEXT: .byte 33
 ; CHECK-NEXT: .byte 11
 ; CHECK-NEXT: .long 4
 ; CHECK-NEXT: .ascii  "char"

 ; CHECK-NEXT: .byte 8
 ; CHECK-NEXT: .byte 5
 
-define amdgpu_kernel void @test_char(i8 %a) !kernel_arg_addr_space !1 !kernel_arg_access_qual !2 !kernel_arg_type !9 !kernel_arg_base_type !9 !kernel_arg_type_qual !4 {
+define amdgpu_kernel void @test_char(i8 signext %a) !kernel_arg_addr_space !1 !kernel_arg_access_qual !2 !kernel_arg_type !9 !kernel_arg_base_type !9 !kernel_arg_type_qual !4 {
   ret void
 }
 

   ret void
 }
 
+; CHECK-LABEL:{{^}}test_sext:
+; CHECK: .section        .AMDGPU.runtime_metadata
+; CHECK-NEXT: .byte 4
+; CHECK-NEXT: .byte 6
+; CHECK-NEXT: .long 9
+; CHECK-NEXT: .ascii  "test_sext"
+; CHECK-NEXT: .byte 7
+; CHECK-NEXT: .byte 9
+; CHECK-NEXT: .long 4
+; CHECK-NEXT: .byte 10
+; CHECK-NEXT: .long 4
+; CHECK-NEXT: .byte 33
+; CHECK-NEXT: .byte 11
+; CHECK-NEXT: .long 4
+; CHECK-NEXT: .ascii  "char"
+; CHECK-NEXT: .byte 13
+; CHECK-NEXT: .byte 0
+; CHECK-NEXT: .byte 14
+; CHECK-NEXT: .short  1
+; CHECK-NEXT: .byte 16
+; CHECK-NEXT: .byte 0
+; CHECK-NEXT: .byte 8
+; CHECK-NEXT: .byte 7
+; CHECK-NEXT: .byte 9
+; CHECK-NEXT: .long 4
+; CHECK-NEXT: .byte 10
+; CHECK-NEXT: .long 4
+; CHECK-NEXT: .byte 33
+; CHECK-NEXT: .byte 11
+; CHECK-NEXT: .long 5
+; CHECK-NEXT: .ascii  "short"
+; CHECK-NEXT: .byte 13
+; CHECK-NEXT: .byte 0
+; CHECK-NEXT: .byte 14
+; CHECK-NEXT: .short  3
+; CHECK-NEXT: .byte 16
+; CHECK-NEXT: .byte 0
+; CHECK-NEXT: .byte 8
+; CHECK-NEXT: .byte 7
+; CHECK-NEXT: .byte 9
+; CHECK-NEXT: .long 2
+; CHECK-NEXT: .byte 10
+; CHECK-NEXT: .long 2
+; CHECK-NEXT: .byte 11
+; CHECK-NEXT: .long 5
+; CHECK-NEXT: .ascii  "char2"
+; CHECK-NEXT: .byte 13
+; CHECK-NEXT: .byte 0
+; CHECK-NEXT: .byte 14
+; CHECK-NEXT: .short  1
+; CHECK-NEXT: .byte 16
+; CHECK-NEXT: .byte 0
+; CHECK-NEXT: .byte 8
+; CHECK-NEXT: .byte 7
+; CHECK-NEXT: .byte 9
+; CHECK-NEXT: .long 4
+; CHECK-NEXT: .byte 10
+; CHECK-NEXT: .long 4
+; CHECK-NEXT: .byte 11
+; CHECK-NEXT: .long 6
+; CHECK-NEXT: .ascii  "short2"
+; CHECK-NEXT: .byte 13
+; CHECK-NEXT: .byte 0
+; CHECK-NEXT: .byte 14
+; CHECK-NEXT: .short  3
+; CHECK-NEXT: .byte 16
+; CHECK-NEXT: .byte 0
+; CHECK-NEXT: .byte 8
+; CHECK-NEXT: .byte 7
+; CHECK-NEXT: .byte 9
+; CHECK-NEXT: .long 4
+; CHECK-NEXT: .byte 10
+; CHECK-NEXT: .long 4
+; CHECK-NEXT: .byte 11
+; CHECK-NEXT: .long 3
+; CHECK-NEXT: .ascii  "int"
+; CHECK-NEXT: .byte 13
+; CHECK-NEXT: .byte 0
+; CHECK-NEXT: .byte 14
+; CHECK-NEXT: .short  6
+; CHECK-NEXT: .byte 16
+; CHECK-NEXT: .byte 0
+; CHECK-NEXT: .byte 8
+; CHECK-NEXT: .byte 7
+; CHECK-NEXT: .byte 9
+; CHECK-NEXT: .long 8
+; CHECK-NEXT: .byte 10
+; CHECK-NEXT: .long 8
+; CHECK-NEXT: .byte 13
+; CHECK-NEXT: .byte 7
+; CHECK-NEXT: .byte 14
+; CHECK-NEXT: .short  9
+; CHECK-NEXT: .byte 8
+; CHECK-NEXT: .byte 7
+; CHECK-NEXT: .byte 9
+; CHECK-NEXT: .long 8
+; CHECK-NEXT: .byte 10
+; CHECK-NEXT: .long 8
+; CHECK-NEXT: .byte 13
+; CHECK-NEXT: .byte 8
+; CHECK-NEXT: .byte 14
+; CHECK-NEXT: .short  9
+; CHECK-NEXT: .byte 8
+; CHECK-NEXT: .byte 7
+; CHECK-NEXT: .byte 9
+; CHECK-NEXT: .long 8
+; CHECK-NEXT: .byte 10
+; CHECK-NEXT: .long 8
+; CHECK-NEXT: .byte 13
+; CHECK-NEXT: .byte 9
+; CHECK-NEXT: .byte 14
+; CHECK-NEXT: .short  9
+; CHECK-NEXT: .byte 8
+; CHECK-NEXT: .byte 7
+; CHECK-NEXT: .byte 9
+; CHECK-NEXT: .long 8
+; CHECK-NEXT: .byte 10
+; CHECK-NEXT: .long 8
+; CHECK-NEXT: .byte 13
+; CHECK-NEXT: .byte 11
+; CHECK-NEXT: .byte 14
+; CHECK-NEXT: .short  1
+; CHECK-NEXT: .byte 15
+; CHECK-NEXT: .byte 1
+; CHECK-NEXT: .byte 8
+; CHECK-NEXT: .byte 5
+
+define amdgpu_kernel void @test_sext(i8 signext %a, i16 signext %b, <2 x i8> %c, <2 x i16> %d, i32 signext %e) !kernel_arg_addr_space !111 !kernel_arg_access_qual !112 !kernel_arg_type !113 !kernel_arg_base_type !113 !kernel_arg_type_qual !114 {
+  ret void
+}
+
+; CHECK-LABEL:{{^}}test_zext:
+; CHECK: .section        .AMDGPU.runtime_metadata
+; CHECK-NEXT: .byte 4
+; CHECK-NEXT: .byte 6
+; CHECK-NEXT: .long 9
+; CHECK-NEXT: .ascii  "test_zext"
+; CHECK-NEXT: .byte 7
+; CHECK-NEXT: .byte 9
+; CHECK-NEXT: .long 4
+; CHECK-NEXT: .byte 10
+; CHECK-NEXT: .long 4
+; CHECK-NEXT: .byte 34
+; CHECK-NEXT: .byte 11
+; CHECK-NEXT: .long 5
+; CHECK-NEXT: .ascii  "uchar"
+; CHECK-NEXT: .byte 13
+; CHECK-NEXT: .byte 0
+; CHECK-NEXT: .byte 14
+; CHECK-NEXT: .short  2
+; CHECK-NEXT: .byte 16
+; CHECK-NEXT: .byte 0
+; CHECK-NEXT: .byte 8
+; CHECK-NEXT: .byte 7
+; CHECK-NEXT: .byte 9
+; CHECK-NEXT: .long 4
+; CHECK-NEXT: .byte 10
+; CHECK-NEXT: .long 4
+; CHECK-NEXT: .byte 34
+; CHECK-NEXT: .byte 11
+; CHECK-NEXT: .long 6
+; CHECK-NEXT: .ascii  "ushort"
+; CHECK-NEXT: .byte 13
+; CHECK-NEXT: .byte 0
+; CHECK-NEXT: .byte 14
+; CHECK-NEXT: .short  4
+; CHECK-NEXT: .byte 16
+; CHECK-NEXT: .byte 0
+; CHECK-NEXT: .byte 8
+; CHECK-NEXT: .byte 7
+; CHECK-NEXT: .byte 9
+; CHECK-NEXT: .long 2
+; CHECK-NEXT: .byte 10
+; CHECK-NEXT: .long 2
+; CHECK-NEXT: .byte 11
+; CHECK-NEXT: .long 6
+; CHECK-NEXT: .ascii  "uchar2"
+; CHECK-NEXT: .byte 13
+; CHECK-NEXT: .byte 0
+; CHECK-NEXT: .byte 14
+; CHECK-NEXT: .short  2
+; CHECK-NEXT: .byte 16
+; CHECK-NEXT: .byte 0
+; CHECK-NEXT: .byte 8
+; CHECK-NEXT: .byte 7
+; CHECK-NEXT: .byte 9
+; CHECK-NEXT: .long 4
+; CHECK-NEXT: .byte 10
+; CHECK-NEXT: .long 4
+; CHECK-NEXT: .byte 11
+; CHECK-NEXT: .long 7
+; CHECK-NEXT: .ascii  "ushort2"
+; CHECK-NEXT: .byte 13
+; CHECK-NEXT: .byte 0
+; CHECK-NEXT: .byte 14
+; CHECK-NEXT: .short  4
+; CHECK-NEXT: .byte 16
+; CHECK-NEXT: .byte 0
+; CHECK-NEXT: .byte 8
+; CHECK-NEXT: .byte 7
+; CHECK-NEXT: .byte 9
+; CHECK-NEXT: .long 4
+; CHECK-NEXT: .byte 10
+; CHECK-NEXT: .long 4
+; CHECK-NEXT: .byte 11
+; CHECK-NEXT: .long 4
+; CHECK-NEXT: .ascii  "uint"
+; CHECK-NEXT: .byte 13
+; CHECK-NEXT: .byte 0
+; CHECK-NEXT: .byte 14
+; CHECK-NEXT: .short  7
+; CHECK-NEXT: .byte 16
+; CHECK-NEXT: .byte 0
+; CHECK-NEXT: .byte 8
+; CHECK-NEXT: .byte 7
+; CHECK-NEXT: .byte 9
+; CHECK-NEXT: .long 8
+; CHECK-NEXT: .byte 10
+; CHECK-NEXT: .long 8
+; CHECK-NEXT: .byte 13
+; CHECK-NEXT: .byte 7
+; CHECK-NEXT: .byte 14
+; CHECK-NEXT: .short  9
+; CHECK-NEXT: .byte 8
+; CHECK-NEXT: .byte 7
+; CHECK-NEXT: .byte 9
+; CHECK-NEXT: .long 8
+; CHECK-NEXT: .byte 10
+; CHECK-NEXT: .long 8
+; CHECK-NEXT: .byte 13
+; CHECK-NEXT: .byte 8
+; CHECK-NEXT: .byte 14
+; CHECK-NEXT: .short  9
+; CHECK-NEXT: .byte 8
+; CHECK-NEXT: .byte 7
+; CHECK-NEXT: .byte 9
+; CHECK-NEXT: .long 8
+; CHECK-NEXT: .byte 10
+; CHECK-NEXT: .long 8
+; CHECK-NEXT: .byte 13
+; CHECK-NEXT: .byte 9
+; CHECK-NEXT: .byte 14
+; CHECK-NEXT: .short  9
+; CHECK-NEXT: .byte 8
+; CHECK-NEXT: .byte 7
+; CHECK-NEXT: .byte 9
+; CHECK-NEXT: .long 8
+; CHECK-NEXT: .byte 10
+; CHECK-NEXT: .long 8
+; CHECK-NEXT: .byte 13
+; CHECK-NEXT: .byte 11
+; CHECK-NEXT: .byte 14
+; CHECK-NEXT: .short  1
+; CHECK-NEXT: .byte 15
+; CHECK-NEXT: .byte 1
+; CHECK-NEXT: .byte 8
+; CHECK-NEXT: .byte 5
+
+
+define amdgpu_kernel void @test_zext(i8 zeroext %a, i16 zeroext %b, <2 x i8> %c, <2 x i16> %d, i32 zeroext %e) !kernel_arg_addr_space !111 !kernel_arg_access_qual !112 !kernel_arg_type !121 !kernel_arg_base_type !121 !kernel_arg_type_qual !114 {
+  ret void
+}
+
 !llvm.printf.fmts = !{!100, !101}
 
 !1 = !{i32 0}

 !94 = !{!"", !"", !"", !"", !"", !"", !""}
 !100 = !{!"1:1:4:%d\5Cn"}
 !101 = !{!"2:1:8:%g\5Cn"}
+!111 = !{i32 0, i32 0, i32 0, i32 0, i32 0}
+!112 = !{!"none", !"none", !"none", !"none", !"none"}
+!113 = !{!"char", !"short", !"char2", !"short2", !"int"}
+!114 = !{!"", !"", !"", !"", !""}
+!121 = !{!"uchar", !"ushort", !"uchar2", !"ushort2", !"uint"}