AMDGPU/SI: Add amdgpu_kernel calling convention. Part 1.

Summary:
This will be used for AMDGPU_HSA_KERNEL symbol type in output ELF.

Also, in the future unused non-kernels may be optimized.

For now, also accept SPIR_KERNEL for HCC frontend.

Also, add bitcode compatibility tests for missing calling conventions
except AVR_BUILTIN which doesn't have parse code.

Reviewers: tstellarAMD, arsenm

Subscribers: arsenm, joker.eph, llvm-commits

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@268717 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/include/llvm/IR/CallingConv.h b/include/llvm/IR/CallingConv.h
index 362a96811c1bf6c63fff9d75f6eb176cce55d0e2..4987b7e943f2311d395a9cb964c2cf45017d197d 100644 (file)
--- a/include/llvm/IR/CallingConv.h
+++ b/include/llvm/IR/CallingConv.h
@@ -190,6 +190,9 @@ namespace CallingConv {
     /// Calling convention used for Mesa compute shaders.
     AMDGPU_CS = 90,
 
+    /// Calling convention for AMDGPU code object kernels.
+    AMDGPU_KERNEL = 91,
+
     /// The highest possible calling convention ID. Must be some 2^k - 1.
     MaxID = 1023
   };

diff --git a/lib/AsmParser/LLLexer.cpp b/lib/AsmParser/LLLexer.cpp
index d9b90bc92c9fcd6a71e5e45812fa86bc56b63f23..fd0081f6f972091284dd9910dd4b687decec6c49 100644 (file)
--- a/lib/AsmParser/LLLexer.cpp
+++ b/lib/AsmParser/LLLexer.cpp
@@ -604,6 +604,7 @@ lltok::Kind LLLexer::LexIdentifier() {
   KEYWORD(amdgpu_gs);
   KEYWORD(amdgpu_ps);
   KEYWORD(amdgpu_cs);
+  KEYWORD(amdgpu_kernel);
 
   KEYWORD(cc);
   KEYWORD(c);

diff --git a/lib/AsmParser/LLParser.cpp b/lib/AsmParser/LLParser.cpp
index d9b7f3efe94c391c175b72174eaef71ff2d5ce96..a8f44e982a279d9b6c625d4447754a62481fed3c 100644 (file)
--- a/lib/AsmParser/LLParser.cpp
+++ b/lib/AsmParser/LLParser.cpp
@@ -1636,6 +1636,7 @@ bool LLParser::ParseOptionalDLLStorageClass(unsigned &Res) {
 ///   ::= 'amdgpu_gs'
 ///   ::= 'amdgpu_ps'
 ///   ::= 'amdgpu_cs'
+///   ::= 'amdgpu_kernel'
 ///   ::= 'cc' UINT
 ///
 bool LLParser::ParseOptionalCallingConv(unsigned &CC) {
@@ -1675,6 +1676,7 @@ bool LLParser::ParseOptionalCallingConv(unsigned &CC) {
   case lltok::kw_amdgpu_gs:      CC = CallingConv::AMDGPU_GS; break;
   case lltok::kw_amdgpu_ps:      CC = CallingConv::AMDGPU_PS; break;
   case lltok::kw_amdgpu_cs:      CC = CallingConv::AMDGPU_CS; break;
+  case lltok::kw_amdgpu_kernel:  CC = CallingConv::AMDGPU_KERNEL; break;
   case lltok::kw_cc: {
       Lex.Lex();
       return ParseUInt32(CC);

diff --git a/lib/AsmParser/LLToken.h b/lib/AsmParser/LLToken.h
index 5d0e98dd62383995b7aae9d00a5c98708d51aa4a..d755e694cde4b6bd725501740c9b5dc92c20024e 100644 (file)
--- a/lib/AsmParser/LLToken.h
+++ b/lib/AsmParser/LLToken.h
@@ -111,6 +111,7 @@ namespace lltok {
     kw_amdgpu_gs,
     kw_amdgpu_ps,
     kw_amdgpu_cs,
+    kw_amdgpu_kernel,
 
     // Attributes:
     kw_attributes,

diff --git a/lib/IR/AsmWriter.cpp b/lib/IR/AsmWriter.cpp
index 43165bf65ed0cf3ee037ade7a458878cf13d7b74..c709416dc82163562ab85a2c8200cc1c0976af26 100644 (file)
--- a/lib/IR/AsmWriter.cpp
+++ b/lib/IR/AsmWriter.cpp
@@ -332,6 +332,7 @@ static void PrintCallingConv(unsigned cc, raw_ostream &Out) {
   case CallingConv::AMDGPU_GS:     Out << "amdgpu_gs"; break;
   case CallingConv::AMDGPU_PS:     Out << "amdgpu_ps"; break;
   case CallingConv::AMDGPU_CS:     Out << "amdgpu_cs"; break;
+  case CallingConv::AMDGPU_KERNEL: Out << "amdgpu_kernel"; break;
   }
 }
 

diff --git a/test/Bitcode/compatibility.ll b/test/Bitcode/compatibility.ll
index d1f249114d5ea94c04b0a873ad28a6c3d25d4302..4f0f29cfc8731d8b414441ee08783d6267794329 100644 (file)
--- a/test/Bitcode/compatibility.ll
+++ b/test/Bitcode/compatibility.ll
@@ -428,6 +428,46 @@ declare cc80 void @f.cc80()
 ; CHECK: declare x86_vectorcallcc void @f.cc80()
 declare x86_vectorcallcc void @f.x86_vectorcallcc()
 ; CHECK: declare x86_vectorcallcc void @f.x86_vectorcallcc()
+declare cc81 void @f.cc81()
+; CHECK: declare hhvmcc void @f.cc81()
+declare hhvmcc void @f.hhvmcc()
+; CHECK: declare hhvmcc void @f.hhvmcc()
+declare cc82 void @f.cc82()
+; CHECK: declare hhvm_ccc void @f.cc82()
+declare hhvm_ccc void @f.hhvm_ccc()
+; CHECK: declare hhvm_ccc void @f.hhvm_ccc()
+declare cc83 void @f.cc83()
+; CHECK: declare x86_intrcc void @f.cc83()
+declare x86_intrcc void @f.x86_intrcc()
+; CHECK: declare x86_intrcc void @f.x86_intrcc()
+declare cc84 void @f.cc84()
+; CHECK: declare avr_intrcc void @f.cc84()
+declare avr_intrcc void @f.avr_intrcc()
+; CHECK: declare avr_intrcc void @f.avr_intrcc()
+declare cc85 void @f.cc85()
+; CHECK: declare avr_signalcc void @f.cc85()
+declare avr_signalcc void @f.avr_signalcc()
+; CHECK: declare avr_signalcc void @f.avr_signalcc()
+declare cc87 void @f.cc87()
+; CHECK: declare amdgpu_vs void @f.cc87()
+declare amdgpu_vs void @f.amdgpu_vs()
+; CHECK: declare amdgpu_vs void @f.amdgpu_vs()
+declare cc88 void @f.cc88()
+; CHECK: declare amdgpu_gs void @f.cc88()
+declare amdgpu_gs void @f.amdgpu_gs()
+; CHECK: declare amdgpu_gs void @f.amdgpu_gs()
+declare cc89 void @f.cc89()
+; CHECK: declare amdgpu_ps void @f.cc89()
+declare amdgpu_ps void @f.amdgpu_ps()
+; CHECK: declare amdgpu_ps void @f.amdgpu_ps()
+declare cc90 void @f.cc90()
+; CHECK: declare amdgpu_cs void @f.cc90()
+declare amdgpu_cs void @f.amdgpu_cs()
+; CHECK: declare amdgpu_cs void @f.amdgpu_cs()
+declare cc91 void @f.cc91()
+; CHECK: declare amdgpu_kernel void @f.cc91()
+declare amdgpu_kernel void @f.amdgpu_kernel()
+; CHECK: declare amdgpu_kernel void @f.amdgpu_kernel()
 declare cc1023 void @f.cc1023()
 ; CHECK: declare cc1023 void @f.cc1023()
 

diff --git a/test/CodeGen/AMDGPU/hsa.ll b/test/CodeGen/AMDGPU/hsa.ll
index 9017b6253d27a6c950a81fc327720e96a46ed571..1f07242d29a10e82eeb6106bb930de2e27646db6 100644 (file)
--- a/test/CodeGen/AMDGPU/hsa.ll
+++ b/test/CodeGen/AMDGPU/hsa.ll
@@ -53,7 +53,7 @@
 ; HSA: .Lfunc_end0:
 ; HSA: .size   simple, .Lfunc_end0-simple
 
-define void @simple(i32 addrspace(1)* %out) {
+define amdgpu_kernel void @simple(i32 addrspace(1)* %out) {
 entry:
   store i32 0, i32 addrspace(1)* %out
   ret void