[AMDGPU] Disassembler: support for DPP

Review: http://reviews.llvm.org/D18642
llvm-svn: 265015

3 files changed, 112 insertions, 7 deletions

diff --git a/llvm/lib/Target/AMDGPU/Disassembler/AMDGPUDisassembler.cpp b/llvm/lib/Target/AMDGPU/Disassembler/AMDGPUDisassembler.cpp
index a3c58663ef8..2990b570f53 100644
--- a/llvm/lib/Target/AMDGPU/Disassembler/AMDGPUDisassembler.cpp
+++ b/llvm/lib/Target/AMDGPU/Disassembler/AMDGPUDisassembler.cpp
@@ -83,10 +83,10 @@ DECODE_OPERAND(SReg_512)
 //
 //===----------------------------------------------------------------------===//
 
-static inline uint32_t eatB32(ArrayRef<uint8_t>& Bytes) {
-  assert(Bytes.size() >= sizeof eatB32(Bytes));
-  const auto Res = support::endian::read32le(Bytes.data());
-  Bytes = Bytes.slice(sizeof Res);
+template <typename T> static inline T eatBytes(ArrayRef<uint8_t>& Bytes) {
+  assert(Bytes.size() >= sizeof(T));
+  const auto Res = support::endian::read<T, support::endianness::little>(Bytes.data());
+  Bytes = Bytes.slice(sizeof(T));
   return Res;
 }
 
@@ -123,8 +123,20 @@ DecodeStatus AMDGPUDisassembler::getInstruction(MCInst &MI, uint64_t &Size,
   do {
     // ToDo: better to switch encoding length using some bit predicate
     // but it is unknown yet, so try all we can
+    
+    // Try to decode DPP first to solve conflict with VOP1 and VOP2 encodings
+    if (Bytes.size() >= 8) {
+      const uint64_t QW = eatBytes<uint64_t>(Bytes);
+      Res = tryDecodeInst(DecoderTableDPP64, MI, QW, Address);
+      if (Res) break;
+    }
+
+    // Reinitialize Bytes as DPP64 could have eaten too much
+    Bytes = Bytes_.slice(0, MaxInstBytesNum);
+
+    // Try decode 32-bit instruction
     if (Bytes.size() < 4) break;
-    const uint32_t DW = eatB32(Bytes);
+    const uint32_t DW = eatBytes<uint32_t>(Bytes);
     Res = tryDecodeInst(DecoderTableVI32, MI, DW, Address);
     if (Res) break;
 
@@ -132,7 +144,7 @@ DecodeStatus AMDGPUDisassembler::getInstruction(MCInst &MI, uint64_t &Size,
     if (Res) break;
 
     if (Bytes.size() < 4) break;
-    const uint64_t QW = ((uint64_t)eatB32(Bytes) << 32) | DW;
+    const uint64_t QW = ((uint64_t)eatBytes<uint32_t>(Bytes) << 32) | DW;
     Res = tryDecodeInst(DecoderTableVI64, MI, QW, Address);
     if (Res) break;
 
@@ -261,7 +273,7 @@ MCOperand AMDGPUDisassembler::decodeLiteralConstant() const {
   if (Bytes.size() < 4)
     return errOperand(0, "cannot read literal, inst bytes left " +
                          Twine(Bytes.size()));
-  return MCOperand::createImm(eatB32(Bytes));
+  return MCOperand::createImm(eatBytes<uint32_t>(Bytes));
 }
 
 MCOperand AMDGPUDisassembler::decodeIntImmed(unsigned Imm) {
diff --git a/llvm/lib/Target/AMDGPU/SIInstrInfo.td b/llvm/lib/Target/AMDGPU/SIInstrInfo.td
index 37ee35c79de..669acd5c5ba 100644
--- a/llvm/lib/Target/AMDGPU/SIInstrInfo.td
+++ b/llvm/lib/Target/AMDGPU/SIInstrInfo.td
@@ -1705,6 +1705,8 @@ class VOP1_DPP <vop1 op, string opName, VOPProfile p> :
   VOP1_DPPe <op.VI>,
   VOP_DPP <p.OutsDPP, p.InsDPP, opName#p.AsmDPP, [], p.HasModifiers> {
   let AssemblerPredicates = [isVI];
+  let DecoderNamespace = "DPP";
+  let DisableDecoder = DisableVIDecoder;
   let src0_modifiers = !if(p.HasModifiers, ?, 0);
   let src1_modifiers = 0;
 }
@@ -1767,6 +1769,8 @@ class VOP2_DPP <vop2 op, string opName, VOPProfile p> :
   VOP2_DPPe <op.VI>,
   VOP_DPP <p.OutsDPP, p.InsDPP, opName#p.AsmDPP, [], p.HasModifiers> {
   let AssemblerPredicates = [isVI];
+  let DecoderNamespace = "DPP";
+  let DisableDecoder = DisableVIDecoder;
   let src0_modifiers = !if(p.HasModifiers, ?, 0);
   let src1_modifiers = !if(p.HasModifiers, ?, 0);
 }
diff --git a/llvm/test/MC/Disassembler/AMDGPU/dpp_vi.txt b/llvm/test/MC/Disassembler/AMDGPU/dpp_vi.txt
new file mode 100644
index 00000000000..117110290ed
--- /dev/null
+++ b/llvm/test/MC/Disassembler/AMDGPU/dpp_vi.txt
@@ -0,0 +1,89 @@
+# RUN: llvm-mc -arch=amdgcn -mcpu=tonga -disassemble -show-encoding < %s | FileCheck %s -check-prefix=VI
+
+# VI: v_mov_b32_dpp v0, v0 quad_perm:[0,2,1,1] row_mask:0xf bank_mask:0xf ; encoding: [0xfa,0x02,0x00,0x7e,0x00,0x58,0x00,0xff]
+0xfa 0x02 0x00 0x7e 0x00 0x58 0x00 0xff
+
+# VI: v_mov_b32_dpp v0, v0 row_shl:1 row_mask:0xf bank_mask:0xf ; encoding: [0xfa,0x02,0x00,0x7e,0x00,0x01,0x01,0xff]
+0xfa 0x02 0x00 0x7e 0x00 0x01 0x01 0xff
+
+# VI: v_mov_b32_dpp v0, v0 row_shr:15 row_mask:0xf bank_mask:0xf ; encoding: [0xfa,0x02,0x00,0x7e,0x00,0x1f,0x01,0xff]
+0xfa 0x02 0x00 0x7e 0x00 0x1f 0x01 0xff
+
+# VI: v_mov_b32_dpp v0, v0 row_ror:12 row_mask:0xf bank_mask:0xf ; encoding: [0xfa,0x02,0x00,0x7e,0x00,0x2c,0x01,0xff]
+0xfa 0x02 0x00 0x7e 0x00 0x2c 0x01 0xff
+
+# VI: v_mov_b32_dpp v0, v0 wave_shl:1 row_mask:0xf bank_mask:0xf ; encoding: [0xfa,0x02,0x00,0x7e,0x00,0x30,0x01,0xff]
+0xfa 0x02 0x00 0x7e 0x00 0x30 0x01 0xff
+
+# VI: v_mov_b32_dpp v0, v0 wave_rol:1 row_mask:0xf bank_mask:0xf ; encoding: [0xfa,0x02,0x00,0x7e,0x00,0x34,0x01,0xff]
+0xfa 0x02 0x00 0x7e 0x00 0x34 0x01 0xff
+
+# VI: v_mov_b32_dpp v0, v0 wave_shr:1 row_mask:0xf bank_mask:0xf ; encoding: [0xfa,0x02,0x00,0x7e,0x00,0x38,0x01,0xff]
+0xfa 0x02 0x00 0x7e 0x00 0x38 0x01 0xff
+
+# VI: v_mov_b32_dpp v0, v0 wave_ror:1 row_mask:0xf bank_mask:0xf ; encoding: [0xfa,0x02,0x00,0x7e,0x00,0x3c,0x01,0xff]
+0xfa 0x02 0x00 0x7e 0x00 0x3c 0x01 0xff
+
+# VI: v_mov_b32_dpp v0, v0 row_mirror row_mask:0xf bank_mask:0xf ; encoding: [0xfa,0x02,0x00,0x7e,0x00,0x40,0x01,0xff]
+0xfa 0x02 0x00 0x7e 0x00 0x40 0x01 0xff
+
+# VI: v_mov_b32_dpp v0, v0 row_half_mirror row_mask:0xf bank_mask:0xf ; encoding: [0xfa,0x02,0x00,0x7e,0x00,0x41,0x01,0xff]
+0xfa 0x02 0x00 0x7e 0x00 0x41 0x01 0xff
+
+# VI: v_mov_b32_dpp v0, v0 row_bcast:15 row_mask:0xf bank_mask:0xf ; encoding: [0xfa,0x02,0x00,0x7e,0x00,0x42,0x01,0xff]
+0xfa 0x02 0x00 0x7e 0x00 0x42 0x01 0xff
+
+# VI: v_mov_b32_dpp v0, v0 row_bcast:31 row_mask:0xf bank_mask:0xf ; encoding: [0xfa,0x02,0x00,0x7e,0x00,0x43,0x01,0xff]
+0xfa 0x02 0x00 0x7e 0x00 0x43 0x01 0xff
+
+# VI: v_mov_b32_dpp v0, v0 quad_perm:[1,3,0,1] row_mask:0xa bank_mask:0x1 bound_ctrl:0 ; encoding: [0xfa,0x02,0x00,0x7e,0x00,0x4d,0x08,0xa1]
+0xfa 0x02 0x00 0x7e 0x00 0x4d 0x08 0xa1
+
+# VI: v_mov_b32_dpp v0, v0 quad_perm:[1,3,0,1] row_mask:0xa bank_mask:0xf ; encoding: [0xfa,0x02,0x00,0x7e,0x00,0x4d,0x00,0xaf]
+0xfa 0x02 0x00 0x7e 0x00 0x4d 0x00 0xaf
+
+# VI: v_mov_b32_dpp v0, v0 quad_perm:[1,3,0,1] row_mask:0xf bank_mask:0x1 ; encoding: [0xfa,0x02,0x00,0x7e,0x00,0x4d,0x00,0xf1]
+0xfa 0x02 0x00 0x7e 0x00 0x4d 0x00 0xf1
+
+# VI: v_mov_b32_dpp v0, v0 quad_perm:[1,3,0,1] row_mask:0xf bank_mask:0xf bound_ctrl:0 ; encoding: [0xfa,0x02,0x00,0x7e,0x00,0x4d,0x08,0xff]
+0xfa 0x02 0x00 0x7e 0x00 0x4d 0x08 0xff
+
+# VI: v_mov_b32_dpp v0, v0 quad_perm:[1,3,0,1] row_mask:0xa bank_mask:0x1 ; encoding: [0xfa,0x02,0x00,0x7e,0x00,0x4d,0x00,0xa1]
+0xfa 0x02 0x00 0x7e 0x00 0x4d 0x00 0xa1
+
+# VI: v_mov_b32_dpp v0, v0 quad_perm:[1,3,0,1] row_mask:0xa bank_mask:0xf bound_ctrl:0 ; encoding: [0xfa,0x02,0x00,0x7e,0x00,0x4d,0x08,0xaf]
+0xfa 0x02 0x00 0x7e 0x00 0x4d 0x08 0xaf
+
+# VI: v_mov_b32_dpp v0, v0 quad_perm:[1,3,0,1] row_mask:0xf bank_mask:0x1 bound_ctrl:0 ; encoding: [0xfa,0x02,0x00,0x7e,0x00,0x4d,0x08,0xf1]
+0xfa 0x02 0x00 0x7e 0x00 0x4d 0x08 0xf1
+
+# VI: v_cvt_u32_f32_dpp v0, v0 row_shl:1 row_mask:0xa bank_mask:0x1 bound_ctrl:0 ; encoding: [0xfa,0x0e,0x00,0x7e,0x00,0x01,0x09,0xa1]
+0xfa 0x0e 0x00 0x7e 0x00 0x01 0x09 0xa1
+
+# VI: v_fract_f32_dpp v0, v0 row_shl:1 row_mask:0xa bank_mask:0x1 bound_ctrl:0 ; encoding: [0xfa,0x36,0x00,0x7e,0x00,0x01,0x09,0xa1]
+0xfa 0x36 0x00 0x7e 0x00 0x01 0x09 0xa1
+
+# VI: v_sin_f32_dpp v0, v0 row_shl:1 row_mask:0xa bank_mask:0x1 bound_ctrl:0 ; encoding: [0xfa,0x52,0x00,0x7e,0x00,0x01,0x09,0xa1]
+0xfa 0x52 0x00 0x7e 0x00 0x01 0x09 0xa1
+
+# VI: v_add_f32_dpp v0, v0, v0 row_shl:1 row_mask:0xa bank_mask:0x1 bound_ctrl:0 ; encoding: [0xfa,0x00,0x00,0x02,0x00,0x01,0x09,0xa1]
+0xfa 0x00 0x00 0x02 0x00 0x01 0x09 0xa1
+
+# VI: v_min_f32_dpp v0, v0, v0 row_shl:1 row_mask:0xa bank_mask:0x1 bound_ctrl:0 ; encoding: [0xfa,0x00,0x00,0x14,0x00,0x01,0x09,0xa1]
+0xfa 0x00 0x00 0x14 0x00 0x01 0x09 0xa1
+
+# VI: v_and_b32_dpp v0, v0, v0 row_shl:1 row_mask:0xa bank_mask:0x1 bound_ctrl:0 ; encoding: [0xfa,0x00,0x00,0x26,0x00,0x01,0x09,0xa1]
+0xfa 0x00 0x00 0x26 0x00 0x01 0x09 0xa1
+
+# VI: v_add_f32_dpp v0, -v0, v0 row_shl:1 row_mask:0xa bank_mask:0x1 bound_ctrl:0 ; encoding: [0xfa,0x00,0x00,0x02,0x00,0x01,0x19,0xa1]
+0xfa 0x00 0x00 0x02 0x00 0x01 0x19 0xa1
+
+# VI: v_add_f32_dpp v0, v0, |v0| row_shl:1 row_mask:0xa bank_mask:0x1 bound_ctrl:0 ; encoding: [0xfa,0x00,0x00,0x02,0x00,0x01,0x89,0xa1]
+0xfa 0x00 0x00 0x02 0x00 0x01 0x89 0xa1
+
+# VI: v_add_f32_dpp v0, -v0, |v0| row_shl:1 row_mask:0xa bank_mask:0x1 bound_ctrl:0 ; encoding: [0xfa,0x00,0x00,0x02,0x00,0x01,0x99,0xa1]
+0xfa 0x00 0x00 0x02 0x00 0x01 0x99 0xa1
+
+# VI: v_add_f32_dpp v0, |v0|, -v0 row_shl:1 row_mask:0xa bank_mask:0x1 bound_ctrl:0 ; encoding: [0xfa,0x00,0x00,0x02,0x00,0x01,0x69,0xa1]
+
+0xfa 0x00 0x00 0x02 0x00 0x01 0x69 0xa1
\ No newline at end of file