[AMDGPU] Assembler: fix v_mac_f16 immediates

Reviewers: vpykhtin, artem.tamazov, tstellarAMD

Subscribers: arsenm, kzhuravl, wdng, nhaehnle, yaxunl, tony-tye

Differential Revision: https://reviews.llvm.org/D28802

git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@292224 91177308-0d34-0410-b5e6-96231b3b80d8

diff --git a/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp b/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp
index 3cf9a1d92469639b057ba2c8a52792a984a2094c..ec0a8e99206b5f75c97aeedeb398d5754e7f0806 100644 (file)
--- a/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp
+++ b/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp
@@ -1031,6 +1031,12 @@ bool AMDGPUOperand::isInlinableImm(MVT type) const {
     if (!canLosslesslyConvertToFPType(FPLiteral, type))
       return false;
 
+    if (type.getScalarSizeInBits() == 16) {
+      return AMDGPU::isInlinableLiteral16(
+        static_cast<int32_t>(FPLiteral.bitcastToAPInt().getZExtValue()),
+        AsmParser->hasInv2PiInlineImm());
+    }
+
     // Check if single precision literal is inlinable
     return AMDGPU::isInlinableLiteral32(
       static_cast<int32_t>(FPLiteral.bitcastToAPInt().getZExtValue()),
@@ -3547,6 +3553,7 @@ void AMDGPUAsmParser::cvtSdwaVOPC(MCInst &Inst, const OperandVector &Operands) {
 
 void AMDGPUAsmParser::cvtSDWA(MCInst &Inst, const OperandVector &Operands,
                               uint64_t BasicInstType) {
+  using namespace llvm::AMDGPU::SDWA;
   OptionalImmIndexMap OptionalIdx;
 
   unsigned I = 1;
@@ -3581,21 +3588,21 @@ void AMDGPUAsmParser::cvtSDWA(MCInst &Inst, const OperandVector &Operands,
     // V_NOP_sdwa_vi has no optional sdwa arguments
     switch (BasicInstType) {
     case SIInstrFlags::VOP1:
-      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaDstSel, 6);
-      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaDstUnused, 2);
-      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaSrc0Sel, 6);
+      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaDstSel, SdwaSel::DWORD);
+      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaDstUnused, DstUnused::UNUSED_PRESERVE);
+      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaSrc0Sel, SdwaSel::DWORD);
       break;
 
     case SIInstrFlags::VOP2:
-      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaDstSel, 6);
-      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaDstUnused, 2);
-      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaSrc0Sel, 6);
-      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaSrc1Sel, 6);
+      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaDstSel, SdwaSel::DWORD);
+      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaDstUnused, DstUnused::UNUSED_PRESERVE);
+      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaSrc0Sel, SdwaSel::DWORD);
+      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaSrc1Sel, SdwaSel::DWORD);
       break;
 
     case SIInstrFlags::VOPC:
-      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaSrc0Sel, 6);
-      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaSrc1Sel, 6);
+      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaSrc0Sel, SdwaSel::DWORD);
+      addOptionalImmOperand(Inst, Operands, OptionalIdx, AMDGPUOperand::ImmTySdwaSrc1Sel, SdwaSel::DWORD);
       break;
 
     default:

diff --git a/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp b/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp
index 5f651d4da5d22ef9dc1e5b6e8c599a6081bf483c..a24c44c8eb3f6d4fb0dabd80b15d301e898a721b 100644 (file)
--- a/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp
+++ b/lib/Target/AMDGPU/Utils/AMDGPUBaseInfo.cpp
@@ -440,7 +440,8 @@ bool isInlinableLiteral32(int32_t Literal, bool HasInv2Pi) {
 }
 
 bool isInlinableLiteral16(int16_t Literal, bool HasInv2Pi) {
-  assert(HasInv2Pi);
+  if (!HasInv2Pi)
+    return false;
 
   if (Literal >= -16 && Literal <= 64)
     return true;

diff --git a/test/MC/AMDGPU/vop3.s b/test/MC/AMDGPU/vop3.s
index 6a62df0d956140a69bd035bedb9e1862259277f8..908db41de408a16eb9fc9ac9c963ed2132ae8f5f 100644 (file)
--- a/test/MC/AMDGPU/vop3.s
+++ b/test/MC/AMDGPU/vop3.s
@@ -1,9 +1,10 @@
 // RUN: not llvm-mc -arch=amdgcn -show-encoding %s | FileCheck %s --check-prefix=SICI
-// RUN: llvm-mc -arch=amdgcn -mcpu=hawaii -show-encoding %s | FileCheck %s --check-prefix=CI
+// RUN: not llvm-mc -arch=amdgcn -mcpu=hawaii -show-encoding %s | FileCheck %s --check-prefix=CI
 // RUN: not llvm-mc -arch=amdgcn -mcpu=tonga -show-encoding %s | FileCheck %s --check-prefix=VI
 
-// RUN: not llvm-mc -arch=amdgcn -show-encoding %s 2>&1 | FileCheck %s --check-prefix=NOSI
-// RUN: not llvm-mc -arch=amdgcn -mcpu=tonga -show-encoding %s 2>&1 | FileCheck %s -check-prefix=NOVI
+// RUN: not llvm-mc -arch=amdgcn -show-encoding %s 2>&1 | FileCheck %s --check-prefix=NOSI --check-prefix=NOSICI
+// RUN: not llvm-mc -arch=amdgcn -mcpu=hawaii -show-encoding %s 2>&1 | FileCheck %s --check-prefix=NOSICI
+// RUN: not llvm-mc -arch=amdgcn -mcpu=tonga -show-encoding %s 2>&1 | FileCheck %s --check-prefix=NOVI
 
 
 //===----------------------------------------------------------------------===//
@@ -258,6 +259,18 @@ v_mac_f32_e64 v0, -v1, |v2|
 // SICI: v_mac_f32_e64 v0, -v1, |v2| ; encoding: [0x00,0x02,0x3e,0xd2,0x01,0x05,0x02,0x20]
 // VI:   v_mac_f32_e64 v0, -v1, |v2| ; encoding: [0x00,0x02,0x16,0xd1,0x01,0x05,0x02,0x20]
 
+v_mac_f16_e64 v0, 0.5, flat_scratch_lo
+// NOSICI: error:
+// VI: v_mac_f16_e64 v0, 0.5, flat_scratch_lo ; encoding: [0x00,0x00,0x23,0xd1,0xf0,0xcc,0x00,0x00]
+
+v_mac_f16_e64 v0, -4.0, flat_scratch_lo
+// NOSICI: error:
+// VI: v_mac_f16_e64 v0, -4.0, flat_scratch_lo ; encoding: [0x00,0x00,0x23,0xd1,0xf6,0xcc,0x00,0x20]
+
+v_mac_f16_e64 v0, flat_scratch_lo, -4.0
+// NOSICI: error:
+// VI: v_mac_f16_e64 v0, flat_scratch_lo, -4.0 ; encoding: [0x00,0x00,0x23,0xd1,0x66,0xec,0x01,0x40]
+
 ///===---------------------------------------------------------------------===//
 // VOP3 Instructions
 ///===---------------------------------------------------------------------===//