Support REG_SEQUENCE in tablegen.

[oota-llvm.git] / lib / Target / R600 / SIInstructions.td

diff --git a/lib/Target/R600/SIInstructions.td b/lib/Target/R600/SIInstructions.td
index 92a53ddeb9fd53906f4629b87714950ec433e81c..36b52929668105edaea36178c144dee77d1690b5 100644 (file)
--- a/lib/Target/R600/SIInstructions.td
+++ b/lib/Target/R600/SIInstructions.td
@@ -22,14 +22,70 @@ def InterpSlot : Operand<i32> {
   let PrintMethod = "printInterpSlot";
 }
 
+def SendMsgImm : Operand<i32> {
+  let PrintMethod = "printSendMsg";
+}
+
 def isSI : Predicate<"Subtarget.getGeneration() "
-                      "== AMDGPUSubtarget::SOUTHERN_ISLANDS">;
+                      ">= AMDGPUSubtarget::SOUTHERN_ISLANDS">;
+
+def isCI : Predicate<"Subtarget.getGeneration() "
+                      ">= AMDGPUSubtarget::SEA_ISLANDS">;
+def HasFlatAddressSpace : Predicate<"Subtarget.hasFlatAddressSpace()">;
 
 def WAIT_FLAG : InstFlag<"printWaitFlag">;
 
-let Predicates = [isSI] in {
+let SubtargetPredicate = isSI in {
+
+//===----------------------------------------------------------------------===//
+// EXP Instructions
+//===----------------------------------------------------------------------===//
+
+defm EXP : EXP_m;
+
+//===----------------------------------------------------------------------===//
+// SMRD Instructions
+//===----------------------------------------------------------------------===//
+
+let mayLoad = 1 in {
+
+// We are using the SGPR_32 and not the SReg_32 register class for 32-bit
+// SMRD instructions, because the SGPR_32 register class does not include M0
+// and writing to M0 from an SMRD instruction will hang the GPU.
+defm S_LOAD_DWORD : SMRD_Helper <0x00, "S_LOAD_DWORD", SReg_64, SGPR_32>;
+defm S_LOAD_DWORDX2 : SMRD_Helper <0x01, "S_LOAD_DWORDX2", SReg_64, SReg_64>;
+defm S_LOAD_DWORDX4 : SMRD_Helper <0x02, "S_LOAD_DWORDX4", SReg_64, SReg_128>;
+defm S_LOAD_DWORDX8 : SMRD_Helper <0x03, "S_LOAD_DWORDX8", SReg_64, SReg_256>;
+defm S_LOAD_DWORDX16 : SMRD_Helper <0x04, "S_LOAD_DWORDX16", SReg_64, SReg_512>;
+
+defm S_BUFFER_LOAD_DWORD : SMRD_Helper <
+  0x08, "S_BUFFER_LOAD_DWORD", SReg_128, SGPR_32
+>;
 
-let neverHasSideEffects = 1 in {
+defm S_BUFFER_LOAD_DWORDX2 : SMRD_Helper <
+  0x09, "S_BUFFER_LOAD_DWORDX2", SReg_128, SReg_64
+>;
+
+defm S_BUFFER_LOAD_DWORDX4 : SMRD_Helper <
+  0x0a, "S_BUFFER_LOAD_DWORDX4", SReg_128, SReg_128
+>;
+
+defm S_BUFFER_LOAD_DWORDX8 : SMRD_Helper <
+  0x0b, "S_BUFFER_LOAD_DWORDX8", SReg_128, SReg_256
+>;
+
+defm S_BUFFER_LOAD_DWORDX16 : SMRD_Helper <
+  0x0c, "S_BUFFER_LOAD_DWORDX16", SReg_128, SReg_512
+>;
+
+} // mayLoad = 1
+
+//def S_MEMTIME : SMRD_ <0x0000001e, "S_MEMTIME", []>;
+//def S_DCACHE_INV : SMRD_ <0x0000001f, "S_DCACHE_INV", []>;
+
+//===----------------------------------------------------------------------===//
+// SOP1 Instructions
+//===----------------------------------------------------------------------===//
 
 let isMoveImm = 1 in {
 def S_MOV_B32 : SOP1_32 <0x00000003, "S_MOV_B32", []>;
@@ -38,33 +94,57 @@ def S_CMOV_B32 : SOP1_32 <0x00000005, "S_CMOV_B32", []>;
 def S_CMOV_B64 : SOP1_64 <0x00000006, "S_CMOV_B64", []>;
 } // End isMoveImm = 1
 
-def S_NOT_B32 : SOP1_32 <0x00000007, "S_NOT_B32", []>;
-def S_NOT_B64 : SOP1_64 <0x00000008, "S_NOT_B64", []>;
+def S_NOT_B32 : SOP1_32 <0x00000007, "S_NOT_B32",
+  [(set i32:$dst, (not i32:$src0))]
+>;
+
+def S_NOT_B64 : SOP1_64 <0x00000008, "S_NOT_B64",
+  [(set i64:$dst, (not i64:$src0))]
+>;
 def S_WQM_B32 : SOP1_32 <0x00000009, "S_WQM_B32", []>;
 def S_WQM_B64 : SOP1_64 <0x0000000a, "S_WQM_B64", []>;
-def S_BREV_B32 : SOP1_32 <0x0000000b, "S_BREV_B32", []>;
+def S_BREV_B32 : SOP1_32 <0x0000000b, "S_BREV_B32",
+  [(set i32:$dst, (AMDGPUbrev i32:$src0))]
+>;
 def S_BREV_B64 : SOP1_64 <0x0000000c, "S_BREV_B64", []>;
-} // End neverHasSideEffects = 1
 
 ////def S_BCNT0_I32_B32 : SOP1_BCNT0 <0x0000000d, "S_BCNT0_I32_B32", []>;
 ////def S_BCNT0_I32_B64 : SOP1_BCNT0 <0x0000000e, "S_BCNT0_I32_B64", []>;
-////def S_BCNT1_I32_B32 : SOP1_BCNT1 <0x0000000f, "S_BCNT1_I32_B32", []>;
-////def S_BCNT1_I32_B64 : SOP1_BCNT1 <0x00000010, "S_BCNT1_I32_B64", []>;
-////def S_FF0_I32_B32 : SOP1_FF0 <0x00000011, "S_FF0_I32_B32", []>;
+def S_BCNT1_I32_B32 : SOP1_32 <0x0000000f, "S_BCNT1_I32_B32",
+  [(set i32:$dst, (ctpop i32:$src0))]
+>;
+def S_BCNT1_I32_B64 : SOP1_32_64 <0x00000010, "S_BCNT1_I32_B64", []>;
+
+////def S_FF0_I32_B32 : SOP1_32 <0x00000011, "S_FF0_I32_B32", []>;
 ////def S_FF0_I32_B64 : SOP1_FF0 <0x00000012, "S_FF0_I32_B64", []>;
-////def S_FF1_I32_B32 : SOP1_FF1 <0x00000013, "S_FF1_I32_B32", []>;
+def S_FF1_I32_B32 : SOP1_32 <0x00000013, "S_FF1_I32_B32",
+  [(set i32:$dst, (cttz_zero_undef i32:$src0))]
+>;
 ////def S_FF1_I32_B64 : SOP1_FF1 <0x00000014, "S_FF1_I32_B64", []>;
-//def S_FLBIT_I32_B32 : SOP1_32 <0x00000015, "S_FLBIT_I32_B32", []>;
+
+def S_FLBIT_I32_B32 : SOP1_32 <0x00000015, "S_FLBIT_I32_B32",
+  [(set i32:$dst, (ctlz_zero_undef i32:$src0))]
+>;
+
 //def S_FLBIT_I32_B64 : SOP1_32 <0x00000016, "S_FLBIT_I32_B64", []>;
 def S_FLBIT_I32 : SOP1_32 <0x00000017, "S_FLBIT_I32", []>;
 //def S_FLBIT_I32_I64 : SOP1_32 <0x00000018, "S_FLBIT_I32_I64", []>;
-//def S_SEXT_I32_I8 : SOP1_32 <0x00000019, "S_SEXT_I32_I8", []>;
-//def S_SEXT_I32_I16 : SOP1_32 <0x0000001a, "S_SEXT_I32_I16", []>;
+def S_SEXT_I32_I8 : SOP1_32 <0x00000019, "S_SEXT_I32_I8",
+  [(set i32:$dst, (sext_inreg i32:$src0, i8))]
+>;
+def S_SEXT_I32_I16 : SOP1_32 <0x0000001a, "S_SEXT_I32_I16",
+  [(set i32:$dst, (sext_inreg i32:$src0, i16))]
+>;
+
 ////def S_BITSET0_B32 : SOP1_BITSET0 <0x0000001b, "S_BITSET0_B32", []>;
 ////def S_BITSET0_B64 : SOP1_BITSET0 <0x0000001c, "S_BITSET0_B64", []>;
 ////def S_BITSET1_B32 : SOP1_BITSET1 <0x0000001d, "S_BITSET1_B32", []>;
 ////def S_BITSET1_B64 : SOP1_BITSET1 <0x0000001e, "S_BITSET1_B64", []>;
-def S_GETPC_B64 : SOP1_64 <0x0000001f, "S_GETPC_B64", []>;
+def S_GETPC_B64 : SOP1 <
+  0x0000001f, (outs SReg_64:$dst), (ins), "S_GETPC_B64 $dst", []
+> {
+  let SSRC0 = 0;
+}
 def S_SETPC_B64 : SOP1_64 <0x00000020, "S_SETPC_B64", []>;
 def S_SWAPPC_B64 : SOP1_64 <0x00000021, "S_SWAPPC_B64", []>;
 def S_RFE_B64 : SOP1_64 <0x00000022, "S_RFE_B64", []>;
@@ -92,6 +172,154 @@ def S_MOVRELD_B64 : SOP1_64 <0x00000031, "S_MOVRELD_B64", []>;
 def S_MOV_REGRD_B32 : SOP1_32 <0x00000033, "S_MOV_REGRD_B32", []>;
 def S_ABS_I32 : SOP1_32 <0x00000034, "S_ABS_I32", []>;
 def S_MOV_FED_B32 : SOP1_32 <0x00000035, "S_MOV_FED_B32", []>;
+
+//===----------------------------------------------------------------------===//
+// SOP2 Instructions
+//===----------------------------------------------------------------------===//
+
+let Defs = [SCC] in { // Carry out goes to SCC
+let isCommutable = 1 in {
+def S_ADD_U32 : SOP2_32 <0x00000000, "S_ADD_U32", []>;
+def S_ADD_I32 : SOP2_32 <0x00000002, "S_ADD_I32",
+  [(set i32:$dst, (add SSrc_32:$src0, SSrc_32:$src1))]
+>;
+} // End isCommutable = 1
+
+def S_SUB_U32 : SOP2_32 <0x00000001, "S_SUB_U32", []>;
+def S_SUB_I32 : SOP2_32 <0x00000003, "S_SUB_I32",
+  [(set i32:$dst, (sub SSrc_32:$src0, SSrc_32:$src1))]
+>;
+
+let Uses = [SCC] in { // Carry in comes from SCC
+let isCommutable = 1 in {
+def S_ADDC_U32 : SOP2_32 <0x00000004, "S_ADDC_U32",
+  [(set i32:$dst, (adde (i32 SSrc_32:$src0), (i32 SSrc_32:$src1)))]>;
+} // End isCommutable = 1
+
+def S_SUBB_U32 : SOP2_32 <0x00000005, "S_SUBB_U32",
+  [(set i32:$dst, (sube (i32 SSrc_32:$src0), (i32 SSrc_32:$src1)))]>;
+} // End Uses = [SCC]
+} // End Defs = [SCC]
+
+def S_MIN_I32 : SOP2_32 <0x00000006, "S_MIN_I32",
+  [(set i32:$dst, (AMDGPUsmin i32:$src0, i32:$src1))]
+>;
+def S_MIN_U32 : SOP2_32 <0x00000007, "S_MIN_U32",
+  [(set i32:$dst, (AMDGPUumin i32:$src0, i32:$src1))]
+>;
+def S_MAX_I32 : SOP2_32 <0x00000008, "S_MAX_I32",
+  [(set i32:$dst, (AMDGPUsmax i32:$src0, i32:$src1))]
+>;
+def S_MAX_U32 : SOP2_32 <0x00000009, "S_MAX_U32",
+  [(set i32:$dst, (AMDGPUumax i32:$src0, i32:$src1))]
+>;
+
+def S_CSELECT_B32 : SOP2 <
+  0x0000000a, (outs SReg_32:$dst),
+  (ins SReg_32:$src0, SReg_32:$src1, SCCReg:$scc), "S_CSELECT_B32",
+  []
+>;
+
+def S_CSELECT_B64 : SOP2_64 <0x0000000b, "S_CSELECT_B64", []>;
+
+def S_AND_B32 : SOP2_32 <0x0000000e, "S_AND_B32",
+  [(set i32:$dst, (and i32:$src0, i32:$src1))]
+>;
+
+def S_AND_B64 : SOP2_64 <0x0000000f, "S_AND_B64",
+  [(set i64:$dst, (and i64:$src0, i64:$src1))]
+>;
+
+def S_OR_B32 : SOP2_32 <0x00000010, "S_OR_B32",
+  [(set i32:$dst, (or i32:$src0, i32:$src1))]
+>;
+
+def S_OR_B64 : SOP2_64 <0x00000011, "S_OR_B64",
+  [(set i64:$dst, (or i64:$src0, i64:$src1))]
+>;
+
+def S_XOR_B32 : SOP2_32 <0x00000012, "S_XOR_B32",
+  [(set i32:$dst, (xor i32:$src0, i32:$src1))]
+>;
+
+def S_XOR_B64 : SOP2_64 <0x00000013, "S_XOR_B64",
+  [(set i64:$dst, (xor i64:$src0, i64:$src1))]
+>;
+def S_ANDN2_B32 : SOP2_32 <0x00000014, "S_ANDN2_B32", []>;
+def S_ANDN2_B64 : SOP2_64 <0x00000015, "S_ANDN2_B64", []>;
+def S_ORN2_B32 : SOP2_32 <0x00000016, "S_ORN2_B32", []>;
+def S_ORN2_B64 : SOP2_64 <0x00000017, "S_ORN2_B64", []>;
+def S_NAND_B32 : SOP2_32 <0x00000018, "S_NAND_B32", []>;
+def S_NAND_B64 : SOP2_64 <0x00000019, "S_NAND_B64", []>;
+def S_NOR_B32 : SOP2_32 <0x0000001a, "S_NOR_B32", []>;
+def S_NOR_B64 : SOP2_64 <0x0000001b, "S_NOR_B64", []>;
+def S_XNOR_B32 : SOP2_32 <0x0000001c, "S_XNOR_B32", []>;
+def S_XNOR_B64 : SOP2_64 <0x0000001d, "S_XNOR_B64", []>;
+
+// Use added complexity so these patterns are preferred to the VALU patterns.
+let AddedComplexity = 1 in {
+
+def S_LSHL_B32 : SOP2_32 <0x0000001e, "S_LSHL_B32",
+  [(set i32:$dst, (shl i32:$src0, i32:$src1))]
+>;
+def S_LSHL_B64 : SOP2_SHIFT_64 <0x0000001f, "S_LSHL_B64",
+  [(set i64:$dst, (shl i64:$src0, i32:$src1))]
+>;
+def S_LSHR_B32 : SOP2_32 <0x00000020, "S_LSHR_B32",
+  [(set i32:$dst, (srl i32:$src0, i32:$src1))]
+>;
+def S_LSHR_B64 : SOP2_SHIFT_64 <0x00000021, "S_LSHR_B64",
+  [(set i64:$dst, (srl i64:$src0, i32:$src1))]
+>;
+def S_ASHR_I32 : SOP2_32 <0x00000022, "S_ASHR_I32",
+  [(set i32:$dst, (sra i32:$src0, i32:$src1))]
+>;
+def S_ASHR_I64 : SOP2_SHIFT_64 <0x00000023, "S_ASHR_I64",
+  [(set i64:$dst, (sra i64:$src0, i32:$src1))]
+>;
+
+
+def S_BFM_B32 : SOP2_32 <0x00000024, "S_BFM_B32", []>;
+def S_BFM_B64 : SOP2_64 <0x00000025, "S_BFM_B64", []>;
+def S_MUL_I32 : SOP2_32 <0x00000026, "S_MUL_I32",
+  [(set i32:$dst, (mul i32:$src0, i32:$src1))]
+>;
+
+} // End AddedComplexity = 1
+
+def S_BFE_U32 : SOP2_32 <0x00000027, "S_BFE_U32", []>;
+def S_BFE_I32 : SOP2_32 <0x00000028, "S_BFE_I32", []>;
+def S_BFE_U64 : SOP2_64 <0x00000029, "S_BFE_U64", []>;
+def S_BFE_I64 : SOP2_64 <0x0000002a, "S_BFE_I64", []>;
+//def S_CBRANCH_G_FORK : SOP2_ <0x0000002b, "S_CBRANCH_G_FORK", []>;
+def S_ABSDIFF_I32 : SOP2_32 <0x0000002c, "S_ABSDIFF_I32", []>;
+
+//===----------------------------------------------------------------------===//
+// SOPC Instructions
+//===----------------------------------------------------------------------===//
+
+def S_CMP_EQ_I32 : SOPC_32 <0x00000000, "S_CMP_EQ_I32">;
+def S_CMP_LG_I32 : SOPC_32 <0x00000001, "S_CMP_LG_I32">;
+def S_CMP_GT_I32 : SOPC_32 <0x00000002, "S_CMP_GT_I32">;
+def S_CMP_GE_I32 : SOPC_32 <0x00000003, "S_CMP_GE_I32">;
+def S_CMP_LT_I32 : SOPC_32 <0x00000004, "S_CMP_LT_I32">;
+def S_CMP_LE_I32 : SOPC_32 <0x00000005, "S_CMP_LE_I32">;
+def S_CMP_EQ_U32 : SOPC_32 <0x00000006, "S_CMP_EQ_U32">;
+def S_CMP_LG_U32 : SOPC_32 <0x00000007, "S_CMP_LG_U32">;
+def S_CMP_GT_U32 : SOPC_32 <0x00000008, "S_CMP_GT_U32">;
+def S_CMP_GE_U32 : SOPC_32 <0x00000009, "S_CMP_GE_U32">;
+def S_CMP_LT_U32 : SOPC_32 <0x0000000a, "S_CMP_LT_U32">;
+def S_CMP_LE_U32 : SOPC_32 <0x0000000b, "S_CMP_LE_U32">;
+////def S_BITCMP0_B32 : SOPC_BITCMP0 <0x0000000c, "S_BITCMP0_B32", []>;
+////def S_BITCMP1_B32 : SOPC_BITCMP1 <0x0000000d, "S_BITCMP1_B32", []>;
+////def S_BITCMP0_B64 : SOPC_BITCMP0 <0x0000000e, "S_BITCMP0_B64", []>;
+////def S_BITCMP1_B64 : SOPC_BITCMP1 <0x0000000f, "S_BITCMP1_B64", []>;
+//def S_SETVSKIP : SOPC_ <0x00000010, "S_SETVSKIP", []>;
+
+//===----------------------------------------------------------------------===//
+// SOPK Instructions
+//===----------------------------------------------------------------------===//
+
 def S_MOVK_I32 : SOPK_32 <0x00000000, "S_MOVK_I32", []>;
 def S_CMOVK_I32 : SOPK_32 <0x00000002, "S_CMOVK_I32", []>;
 
@@ -114,7 +342,7 @@ def S_CMPK_EQ_I32 : SOPK <
 >;
 */
 
-let isCompare = 1 in {
+let isCompare = 1, Defs = [SCC] in {
 def S_CMPK_LG_I32 : SOPK_32 <0x00000004, "S_CMPK_LG_I32", []>;
 def S_CMPK_GT_I32 : SOPK_32 <0x00000005, "S_CMPK_GT_I32", []>;
 def S_CMPK_GE_I32 : SOPK_32 <0x00000006, "S_CMPK_GE_I32", []>;
@@ -126,10 +354,13 @@ def S_CMPK_GT_U32 : SOPK_32 <0x0000000b, "S_CMPK_GT_U32", []>;
 def S_CMPK_GE_U32 : SOPK_32 <0x0000000c, "S_CMPK_GE_U32", []>;
 def S_CMPK_LT_U32 : SOPK_32 <0x0000000d, "S_CMPK_LT_U32", []>;
 def S_CMPK_LE_U32 : SOPK_32 <0x0000000e, "S_CMPK_LE_U32", []>;
-} // End isCompare = 1
+} // End isCompare = 1, Defs = [SCC]
+
+let Defs = [SCC], isCommutable = 1 in {
+  def S_ADDK_I32 : SOPK_32 <0x0000000f, "S_ADDK_I32", []>;
+  def S_MULK_I32 : SOPK_32 <0x00000010, "S_MULK_I32", []>;
+}
 
-def S_ADDK_I32 : SOPK_32 <0x0000000f, "S_ADDK_I32", []>;
-def S_MULK_I32 : SOPK_32 <0x00000010, "S_MULK_I32", []>;
 //def S_CBRANCH_I_FORK : SOPK_ <0x00000011, "S_CBRANCH_I_FORK", []>;
 def S_GETREG_B32 : SOPK_32 <0x00000012, "S_GETREG_B32", []>;
 def S_SETREG_B32 : SOPK_32 <0x00000013, "S_SETREG_B32", []>;
@@ -137,272 +368,483 @@ def S_GETREG_REGRD_B32 : SOPK_32 <0x00000014, "S_GETREG_REGRD_B32", []>;
 //def S_SETREG_IMM32_B32 : SOPK_32 <0x00000015, "S_SETREG_IMM32_B32", []>;
 //def EXP : EXP_ <0x00000000, "EXP", []>;
 
+//===----------------------------------------------------------------------===//
+// SOPP Instructions
+//===----------------------------------------------------------------------===//
+
+def S_NOP : SOPP <0x00000000, (ins i16imm:$simm16), "S_NOP $simm16", []>;
+
+let isTerminator = 1 in {
+
+def S_ENDPGM : SOPP <0x00000001, (ins), "S_ENDPGM",
+  [(IL_retflag)]> {
+  let simm16 = 0;
+  let isBarrier = 1;
+  let hasCtrlDep = 1;
+}
+
+let isBranch = 1 in {
+def S_BRANCH : SOPP <
+  0x00000002, (ins sopp_brtarget:$simm16), "S_BRANCH $simm16",
+  [(br bb:$simm16)]> {
+  let isBarrier = 1;
+}
+
+let DisableEncoding = "$scc" in {
+def S_CBRANCH_SCC0 : SOPP <
+  0x00000004, (ins sopp_brtarget:$simm16, SCCReg:$scc),
+  "S_CBRANCH_SCC0 $simm16", []
+>;
+def S_CBRANCH_SCC1 : SOPP <
+  0x00000005, (ins sopp_brtarget:$simm16, SCCReg:$scc),
+  "S_CBRANCH_SCC1 $simm16",
+  []
+>;
+} // End DisableEncoding = "$scc"
+
+def S_CBRANCH_VCCZ : SOPP <
+  0x00000006, (ins sopp_brtarget:$simm16, VCCReg:$vcc),
+  "S_CBRANCH_VCCZ $simm16",
+  []
+>;
+def S_CBRANCH_VCCNZ : SOPP <
+  0x00000007, (ins sopp_brtarget:$simm16, VCCReg:$vcc),
+  "S_CBRANCH_VCCNZ $simm16",
+  []
+>;
+
+let DisableEncoding = "$exec" in {
+def S_CBRANCH_EXECZ : SOPP <
+  0x00000008, (ins sopp_brtarget:$simm16, EXECReg:$exec),
+  "S_CBRANCH_EXECZ $simm16",
+  []
+>;
+def S_CBRANCH_EXECNZ : SOPP <
+  0x00000009, (ins sopp_brtarget:$simm16, EXECReg:$exec),
+  "S_CBRANCH_EXECNZ $simm16",
+  []
+>;
+} // End DisableEncoding = "$exec"
+
+
+} // End isBranch = 1
+} // End isTerminator = 1
+
+let hasSideEffects = 1 in {
+def S_BARRIER : SOPP <0x0000000a, (ins), "S_BARRIER",
+  [(int_AMDGPU_barrier_local)]
+> {
+  let simm16 = 0;
+  let isBarrier = 1;
+  let hasCtrlDep = 1;
+  let mayLoad = 1;
+  let mayStore = 1;
+}
+
+def S_WAITCNT : SOPP <0x0000000c, (ins WAIT_FLAG:$simm16), "S_WAITCNT $simm16",
+  []
+>;
+//def S_SETHALT : SOPP_ <0x0000000d, "S_SETHALT", []>;
+//def S_SLEEP : SOPP_ <0x0000000e, "S_SLEEP", []>;
+//def S_SETPRIO : SOPP_ <0x0000000f, "S_SETPRIO", []>;
+
+let Uses = [EXEC] in {
+  def S_SENDMSG : SOPP <0x00000010, (ins SendMsgImm:$simm16, M0Reg:$m0), "S_SENDMSG $simm16",
+      [(int_SI_sendmsg imm:$simm16, M0Reg:$m0)]
+  > {
+    let DisableEncoding = "$m0";
+  }
+} // End Uses = [EXEC]
+
+//def S_SENDMSGHALT : SOPP_ <0x00000011, "S_SENDMSGHALT", []>;
+//def S_TRAP : SOPP_ <0x00000012, "S_TRAP", []>;
+//def S_ICACHE_INV : SOPP_ <0x00000013, "S_ICACHE_INV", []>;
+//def S_INCPERFLEVEL : SOPP_ <0x00000014, "S_INCPERFLEVEL", []>;
+//def S_DECPERFLEVEL : SOPP_ <0x00000015, "S_DECPERFLEVEL", []>;
+//def S_TTRACEDATA : SOPP_ <0x00000016, "S_TTRACEDATA", []>;
+} // End hasSideEffects
+
+//===----------------------------------------------------------------------===//
+// VOPC Instructions
+//===----------------------------------------------------------------------===//
+
 let isCompare = 1 in {
 
-defm V_CMP_F_F32 : VOPC_32 <0x00000000, "V_CMP_F_F32">;
-defm V_CMP_LT_F32 : VOPC_32 <0x00000001, "V_CMP_LT_F32", f32, COND_LT>;
-defm V_CMP_EQ_F32 : VOPC_32 <0x00000002, "V_CMP_EQ_F32", f32, COND_EQ>;
-defm V_CMP_LE_F32 : VOPC_32 <0x00000003, "V_CMP_LE_F32", f32, COND_LE>;
-defm V_CMP_GT_F32 : VOPC_32 <0x00000004, "V_CMP_GT_F32", f32, COND_GT>;
-defm V_CMP_LG_F32 : VOPC_32 <0x00000005, "V_CMP_LG_F32", f32, COND_NE>;
-defm V_CMP_GE_F32 : VOPC_32 <0x00000006, "V_CMP_GE_F32", f32, COND_GE>;
-defm V_CMP_O_F32 : VOPC_32 <0x00000007, "V_CMP_O_F32">;
-defm V_CMP_U_F32 : VOPC_32 <0x00000008, "V_CMP_U_F32">;
-defm V_CMP_NGE_F32 : VOPC_32 <0x00000009, "V_CMP_NGE_F32">;
-defm V_CMP_NLG_F32 : VOPC_32 <0x0000000a, "V_CMP_NLG_F32">;
-defm V_CMP_NGT_F32 : VOPC_32 <0x0000000b, "V_CMP_NGT_F32">;
-defm V_CMP_NLE_F32 : VOPC_32 <0x0000000c, "V_CMP_NLE_F32">;
-defm V_CMP_NEQ_F32 : VOPC_32 <0x0000000d, "V_CMP_NEQ_F32", f32, COND_NE>;
-defm V_CMP_NLT_F32 : VOPC_32 <0x0000000e, "V_CMP_NLT_F32">;
-defm V_CMP_TRU_F32 : VOPC_32 <0x0000000f, "V_CMP_TRU_F32">;
+defm V_CMP_F_F32 : VOPC_F32 <vopc<0x0>, "V_CMP_F_F32">;
+defm V_CMP_LT_F32 : VOPC_F32 <vopc<0x1>, "V_CMP_LT_F32", COND_OLT>;
+defm V_CMP_EQ_F32 : VOPC_F32 <vopc<0x2>, "V_CMP_EQ_F32", COND_OEQ>;
+defm V_CMP_LE_F32 : VOPC_F32 <vopc<0x3>, "V_CMP_LE_F32", COND_OLE>;
+defm V_CMP_GT_F32 : VOPC_F32 <vopc<0x4>, "V_CMP_GT_F32", COND_OGT>;
+defm V_CMP_LG_F32 : VOPC_F32 <vopc<0x5>, "V_CMP_LG_F32">;
+defm V_CMP_GE_F32 : VOPC_F32 <vopc<0x6>, "V_CMP_GE_F32", COND_OGE>;
+defm V_CMP_O_F32 : VOPC_F32 <vopc<0x7>, "V_CMP_O_F32", COND_O>;
+defm V_CMP_U_F32 : VOPC_F32 <vopc<0x8>, "V_CMP_U_F32", COND_UO>;
+defm V_CMP_NGE_F32 : VOPC_F32 <vopc<0x9>, "V_CMP_NGE_F32">;
+defm V_CMP_NLG_F32 : VOPC_F32 <vopc<0xa>, "V_CMP_NLG_F32">;
+defm V_CMP_NGT_F32 : VOPC_F32 <vopc<0xb>, "V_CMP_NGT_F32">;
+defm V_CMP_NLE_F32 : VOPC_F32 <vopc<0xc>, "V_CMP_NLE_F32">;
+defm V_CMP_NEQ_F32 : VOPC_F32 <vopc<0xd>, "V_CMP_NEQ_F32", COND_UNE>;
+defm V_CMP_NLT_F32 : VOPC_F32 <vopc<0xe>, "V_CMP_NLT_F32">;
+defm V_CMP_TRU_F32 : VOPC_F32 <vopc<0xf>, "V_CMP_TRU_F32">;
 
-let hasSideEffects = 1, Defs = [EXEC] in {
+let hasSideEffects = 1 in {
 
-defm V_CMPX_F_F32 : VOPC_32 <0x00000010, "V_CMPX_F_F32">;
-defm V_CMPX_LT_F32 : VOPC_32 <0x00000011, "V_CMPX_LT_F32">;
-defm V_CMPX_EQ_F32 : VOPC_32 <0x00000012, "V_CMPX_EQ_F32">;
-defm V_CMPX_LE_F32 : VOPC_32 <0x00000013, "V_CMPX_LE_F32">;
-defm V_CMPX_GT_F32 : VOPC_32 <0x00000014, "V_CMPX_GT_F32">;
-defm V_CMPX_LG_F32 : VOPC_32 <0x00000015, "V_CMPX_LG_F32">;
-defm V_CMPX_GE_F32 : VOPC_32 <0x00000016, "V_CMPX_GE_F32">;
-defm V_CMPX_O_F32 : VOPC_32 <0x00000017, "V_CMPX_O_F32">;
-defm V_CMPX_U_F32 : VOPC_32 <0x00000018, "V_CMPX_U_F32">;
-defm V_CMPX_NGE_F32 : VOPC_32 <0x00000019, "V_CMPX_NGE_F32">;
-defm V_CMPX_NLG_F32 : VOPC_32 <0x0000001a, "V_CMPX_NLG_F32">;
-defm V_CMPX_NGT_F32 : VOPC_32 <0x0000001b, "V_CMPX_NGT_F32">;
-defm V_CMPX_NLE_F32 : VOPC_32 <0x0000001c, "V_CMPX_NLE_F32">;
-defm V_CMPX_NEQ_F32 : VOPC_32 <0x0000001d, "V_CMPX_NEQ_F32">;
-defm V_CMPX_NLT_F32 : VOPC_32 <0x0000001e, "V_CMPX_NLT_F32">;
-defm V_CMPX_TRU_F32 : VOPC_32 <0x0000001f, "V_CMPX_TRU_F32">;
+defm V_CMPX_F_F32 : VOPCX_F32 <vopc<0x10>, "V_CMPX_F_F32">;
+defm V_CMPX_LT_F32 : VOPCX_F32 <vopc<0x11>, "V_CMPX_LT_F32">;
+defm V_CMPX_EQ_F32 : VOPCX_F32 <vopc<0x12>, "V_CMPX_EQ_F32">;
+defm V_CMPX_LE_F32 : VOPCX_F32 <vopc<0x13>, "V_CMPX_LE_F32">;
+defm V_CMPX_GT_F32 : VOPCX_F32 <vopc<0x14>, "V_CMPX_GT_F32">;
+defm V_CMPX_LG_F32 : VOPCX_F32 <vopc<0x15>, "V_CMPX_LG_F32">;
+defm V_CMPX_GE_F32 : VOPCX_F32 <vopc<0x16>, "V_CMPX_GE_F32">;
+defm V_CMPX_O_F32 : VOPCX_F32 <vopc<0x17>, "V_CMPX_O_F32">;
+defm V_CMPX_U_F32 : VOPCX_F32 <vopc<0x18>, "V_CMPX_U_F32">;
+defm V_CMPX_NGE_F32 : VOPCX_F32 <vopc<0x19>, "V_CMPX_NGE_F32">;
+defm V_CMPX_NLG_F32 : VOPCX_F32 <vopc<0x1a>, "V_CMPX_NLG_F32">;
+defm V_CMPX_NGT_F32 : VOPCX_F32 <vopc<0x1b>, "V_CMPX_NGT_F32">;
+defm V_CMPX_NLE_F32 : VOPCX_F32 <vopc<0x1c>, "V_CMPX_NLE_F32">;
+defm V_CMPX_NEQ_F32 : VOPCX_F32 <vopc<0x1d>, "V_CMPX_NEQ_F32">;
+defm V_CMPX_NLT_F32 : VOPCX_F32 <vopc<0x1e>, "V_CMPX_NLT_F32">;
+defm V_CMPX_TRU_F32 : VOPCX_F32 <vopc<0x1f>, "V_CMPX_TRU_F32">;
 
-} // End hasSideEffects = 1, Defs = [EXEC]
+} // End hasSideEffects = 1
 
-defm V_CMP_F_F64 : VOPC_64 <0x00000020, "V_CMP_F_F64">;
-defm V_CMP_LT_F64 : VOPC_64 <0x00000021, "V_CMP_LT_F64", f64, COND_LT>;
-defm V_CMP_EQ_F64 : VOPC_64 <0x00000022, "V_CMP_EQ_F64", f64, COND_EQ>;
-defm V_CMP_LE_F64 : VOPC_64 <0x00000023, "V_CMP_LE_F64", f64, COND_LE>;
-defm V_CMP_GT_F64 : VOPC_64 <0x00000024, "V_CMP_GT_F64", f64, COND_GT>;
-defm V_CMP_LG_F64 : VOPC_64 <0x00000025, "V_CMP_LG_F64">;
-defm V_CMP_GE_F64 : VOPC_64 <0x00000026, "V_CMP_GE_F64", f64, COND_GE>;
-defm V_CMP_O_F64 : VOPC_64 <0x00000027, "V_CMP_O_F64">;
-defm V_CMP_U_F64 : VOPC_64 <0x00000028, "V_CMP_U_F64">;
-defm V_CMP_NGE_F64 : VOPC_64 <0x00000029, "V_CMP_NGE_F64">;
-defm V_CMP_NLG_F64 : VOPC_64 <0x0000002a, "V_CMP_NLG_F64">;
-defm V_CMP_NGT_F64 : VOPC_64 <0x0000002b, "V_CMP_NGT_F64">;
-defm V_CMP_NLE_F64 : VOPC_64 <0x0000002c, "V_CMP_NLE_F64">;
-defm V_CMP_NEQ_F64 : VOPC_64 <0x0000002d, "V_CMP_NEQ_F64", f64, COND_NE>;
-defm V_CMP_NLT_F64 : VOPC_64 <0x0000002e, "V_CMP_NLT_F64">;
-defm V_CMP_TRU_F64 : VOPC_64 <0x0000002f, "V_CMP_TRU_F64">;
+defm V_CMP_F_F64 : VOPC_F64 <vopc<0x20>, "V_CMP_F_F64">;
+defm V_CMP_LT_F64 : VOPC_F64 <vopc<0x21>, "V_CMP_LT_F64", COND_OLT>;
+defm V_CMP_EQ_F64 : VOPC_F64 <vopc<0x22>, "V_CMP_EQ_F64", COND_OEQ>;
+defm V_CMP_LE_F64 : VOPC_F64 <vopc<0x23>, "V_CMP_LE_F64", COND_OLE>;
+defm V_CMP_GT_F64 : VOPC_F64 <vopc<0x24>, "V_CMP_GT_F64", COND_OGT>;
+defm V_CMP_LG_F64 : VOPC_F64 <vopc<0x25>, "V_CMP_LG_F64">;
+defm V_CMP_GE_F64 : VOPC_F64 <vopc<0x26>, "V_CMP_GE_F64", COND_OGE>;
+defm V_CMP_O_F64 : VOPC_F64 <vopc<0x27>, "V_CMP_O_F64", COND_O>;
+defm V_CMP_U_F64 : VOPC_F64 <vopc<0x28>, "V_CMP_U_F64", COND_UO>;
+defm V_CMP_NGE_F64 : VOPC_F64 <vopc<0x29>, "V_CMP_NGE_F64">;
+defm V_CMP_NLG_F64 : VOPC_F64 <vopc<0x2a>, "V_CMP_NLG_F64">;
+defm V_CMP_NGT_F64 : VOPC_F64 <vopc<0x2b>, "V_CMP_NGT_F64">;
+defm V_CMP_NLE_F64 : VOPC_F64 <vopc<0x2c>, "V_CMP_NLE_F64">;
+defm V_CMP_NEQ_F64 : VOPC_F64 <vopc<0x2d>, "V_CMP_NEQ_F64", COND_UNE>;
+defm V_CMP_NLT_F64 : VOPC_F64 <vopc<0x2e>, "V_CMP_NLT_F64">;
+defm V_CMP_TRU_F64 : VOPC_F64 <vopc<0x2f>, "V_CMP_TRU_F64">;
 
-let hasSideEffects = 1, Defs = [EXEC] in {
+let hasSideEffects = 1 in {
 
-defm V_CMPX_F_F64 : VOPC_64 <0x00000030, "V_CMPX_F_F64">;
-defm V_CMPX_LT_F64 : VOPC_64 <0x00000031, "V_CMPX_LT_F64">;
-defm V_CMPX_EQ_F64 : VOPC_64 <0x00000032, "V_CMPX_EQ_F64">;
-defm V_CMPX_LE_F64 : VOPC_64 <0x00000033, "V_CMPX_LE_F64">;
-defm V_CMPX_GT_F64 : VOPC_64 <0x00000034, "V_CMPX_GT_F64">;
-defm V_CMPX_LG_F64 : VOPC_64 <0x00000035, "V_CMPX_LG_F64">;
-defm V_CMPX_GE_F64 : VOPC_64 <0x00000036, "V_CMPX_GE_F64">;
-defm V_CMPX_O_F64 : VOPC_64 <0x00000037, "V_CMPX_O_F64">;
-defm V_CMPX_U_F64 : VOPC_64 <0x00000038, "V_CMPX_U_F64">;
-defm V_CMPX_NGE_F64 : VOPC_64 <0x00000039, "V_CMPX_NGE_F64">;
-defm V_CMPX_NLG_F64 : VOPC_64 <0x0000003a, "V_CMPX_NLG_F64">;
-defm V_CMPX_NGT_F64 : VOPC_64 <0x0000003b, "V_CMPX_NGT_F64">;
-defm V_CMPX_NLE_F64 : VOPC_64 <0x0000003c, "V_CMPX_NLE_F64">;
-defm V_CMPX_NEQ_F64 : VOPC_64 <0x0000003d, "V_CMPX_NEQ_F64">;
-defm V_CMPX_NLT_F64 : VOPC_64 <0x0000003e, "V_CMPX_NLT_F64">;
-defm V_CMPX_TRU_F64 : VOPC_64 <0x0000003f, "V_CMPX_TRU_F64">;
+defm V_CMPX_F_F64 : VOPCX_F64 <vopc<0x30>, "V_CMPX_F_F64">;
+defm V_CMPX_LT_F64 : VOPCX_F64 <vopc<0x31>, "V_CMPX_LT_F64">;
+defm V_CMPX_EQ_F64 : VOPCX_F64 <vopc<0x32>, "V_CMPX_EQ_F64">;
+defm V_CMPX_LE_F64 : VOPCX_F64 <vopc<0x33>, "V_CMPX_LE_F64">;
+defm V_CMPX_GT_F64 : VOPCX_F64 <vopc<0x34>, "V_CMPX_GT_F64">;
+defm V_CMPX_LG_F64 : VOPCX_F64 <vopc<0x35>, "V_CMPX_LG_F64">;
+defm V_CMPX_GE_F64 : VOPCX_F64 <vopc<0x36>, "V_CMPX_GE_F64">;
+defm V_CMPX_O_F64 : VOPCX_F64 <vopc<0x37>, "V_CMPX_O_F64">;
+defm V_CMPX_U_F64 : VOPCX_F64 <vopc<0x38>, "V_CMPX_U_F64">;
+defm V_CMPX_NGE_F64 : VOPCX_F64 <vopc<0x39>, "V_CMPX_NGE_F64">;
+defm V_CMPX_NLG_F64 : VOPCX_F64 <vopc<0x3a>, "V_CMPX_NLG_F64">;
+defm V_CMPX_NGT_F64 : VOPCX_F64 <vopc<0x3b>, "V_CMPX_NGT_F64">;
+defm V_CMPX_NLE_F64 : VOPCX_F64 <vopc<0x3c>, "V_CMPX_NLE_F64">;
+defm V_CMPX_NEQ_F64 : VOPCX_F64 <vopc<0x3d>, "V_CMPX_NEQ_F64">;
+defm V_CMPX_NLT_F64 : VOPCX_F64 <vopc<0x3e>, "V_CMPX_NLT_F64">;
+defm V_CMPX_TRU_F64 : VOPCX_F64 <vopc<0x3f>, "V_CMPX_TRU_F64">;
 
-} // End hasSideEffects = 1, Defs = [EXEC]
+} // End hasSideEffects = 1
 
-defm V_CMPS_F_F32 : VOPC_32 <0x00000040, "V_CMPS_F_F32">;
-defm V_CMPS_LT_F32 : VOPC_32 <0x00000041, "V_CMPS_LT_F32">;
-defm V_CMPS_EQ_F32 : VOPC_32 <0x00000042, "V_CMPS_EQ_F32">;
-defm V_CMPS_LE_F32 : VOPC_32 <0x00000043, "V_CMPS_LE_F32">;
-defm V_CMPS_GT_F32 : VOPC_32 <0x00000044, "V_CMPS_GT_F32">;
-defm V_CMPS_LG_F32 : VOPC_32 <0x00000045, "V_CMPS_LG_F32">;
-defm V_CMPS_GE_F32 : VOPC_32 <0x00000046, "V_CMPS_GE_F32">;
-defm V_CMPS_O_F32 : VOPC_32 <0x00000047, "V_CMPS_O_F32">;
-defm V_CMPS_U_F32 : VOPC_32 <0x00000048, "V_CMPS_U_F32">;
-defm V_CMPS_NGE_F32 : VOPC_32 <0x00000049, "V_CMPS_NGE_F32">;
-defm V_CMPS_NLG_F32 : VOPC_32 <0x0000004a, "V_CMPS_NLG_F32">;
-defm V_CMPS_NGT_F32 : VOPC_32 <0x0000004b, "V_CMPS_NGT_F32">;
-defm V_CMPS_NLE_F32 : VOPC_32 <0x0000004c, "V_CMPS_NLE_F32">;
-defm V_CMPS_NEQ_F32 : VOPC_32 <0x0000004d, "V_CMPS_NEQ_F32">;
-defm V_CMPS_NLT_F32 : VOPC_32 <0x0000004e, "V_CMPS_NLT_F32">;
-defm V_CMPS_TRU_F32 : VOPC_32 <0x0000004f, "V_CMPS_TRU_F32">;
+defm V_CMPS_F_F32 : VOPC_F32 <vopc<0x40>, "V_CMPS_F_F32">;
+defm V_CMPS_LT_F32 : VOPC_F32 <vopc<0x41>, "V_CMPS_LT_F32">;
+defm V_CMPS_EQ_F32 : VOPC_F32 <vopc<0x42>, "V_CMPS_EQ_F32">;
+defm V_CMPS_LE_F32 : VOPC_F32 <vopc<0x43>, "V_CMPS_LE_F32">;
+defm V_CMPS_GT_F32 : VOPC_F32 <vopc<0x44>, "V_CMPS_GT_F32">;
+defm V_CMPS_LG_F32 : VOPC_F32 <vopc<0x45>, "V_CMPS_LG_F32">;
+defm V_CMPS_GE_F32 : VOPC_F32 <vopc<0x46>, "V_CMPS_GE_F32">;
+defm V_CMPS_O_F32 : VOPC_F32 <vopc<0x47>, "V_CMPS_O_F32">;
+defm V_CMPS_U_F32 : VOPC_F32 <vopc<0x48>, "V_CMPS_U_F32">;
+defm V_CMPS_NGE_F32 : VOPC_F32 <vopc<0x49>, "V_CMPS_NGE_F32">;
+defm V_CMPS_NLG_F32 : VOPC_F32 <vopc<0x4a>, "V_CMPS_NLG_F32">;
+defm V_CMPS_NGT_F32 : VOPC_F32 <vopc<0x4b>, "V_CMPS_NGT_F32">;
+defm V_CMPS_NLE_F32 : VOPC_F32 <vopc<0x4c>, "V_CMPS_NLE_F32">;
+defm V_CMPS_NEQ_F32 : VOPC_F32 <vopc<0x4d>, "V_CMPS_NEQ_F32">;
+defm V_CMPS_NLT_F32 : VOPC_F32 <vopc<0x4e>, "V_CMPS_NLT_F32">;
+defm V_CMPS_TRU_F32 : VOPC_F32 <vopc<0x4f>, "V_CMPS_TRU_F32">;
 
-let hasSideEffects = 1, Defs = [EXEC] in {
+let hasSideEffects = 1 in {
 
-defm V_CMPSX_F_F32 : VOPC_32 <0x00000050, "V_CMPSX_F_F32">;
-defm V_CMPSX_LT_F32 : VOPC_32 <0x00000051, "V_CMPSX_LT_F32">;
-defm V_CMPSX_EQ_F32 : VOPC_32 <0x00000052, "V_CMPSX_EQ_F32">;
-defm V_CMPSX_LE_F32 : VOPC_32 <0x00000053, "V_CMPSX_LE_F32">;
-defm V_CMPSX_GT_F32 : VOPC_32 <0x00000054, "V_CMPSX_GT_F32">;
-defm V_CMPSX_LG_F32 : VOPC_32 <0x00000055, "V_CMPSX_LG_F32">;
-defm V_CMPSX_GE_F32 : VOPC_32 <0x00000056, "V_CMPSX_GE_F32">;
-defm V_CMPSX_O_F32 : VOPC_32 <0x00000057, "V_CMPSX_O_F32">;
-defm V_CMPSX_U_F32 : VOPC_32 <0x00000058, "V_CMPSX_U_F32">;
-defm V_CMPSX_NGE_F32 : VOPC_32 <0x00000059, "V_CMPSX_NGE_F32">;
-defm V_CMPSX_NLG_F32 : VOPC_32 <0x0000005a, "V_CMPSX_NLG_F32">;
-defm V_CMPSX_NGT_F32 : VOPC_32 <0x0000005b, "V_CMPSX_NGT_F32">;
-defm V_CMPSX_NLE_F32 : VOPC_32 <0x0000005c, "V_CMPSX_NLE_F32">;
-defm V_CMPSX_NEQ_F32 : VOPC_32 <0x0000005d, "V_CMPSX_NEQ_F32">;
-defm V_CMPSX_NLT_F32 : VOPC_32 <0x0000005e, "V_CMPSX_NLT_F32">;
-defm V_CMPSX_TRU_F32 : VOPC_32 <0x0000005f, "V_CMPSX_TRU_F32">;
+defm V_CMPSX_F_F32 : VOPCX_F32 <vopc<0x50>, "V_CMPSX_F_F32">;
+defm V_CMPSX_LT_F32 : VOPCX_F32 <vopc<0x51>, "V_CMPSX_LT_F32">;
+defm V_CMPSX_EQ_F32 : VOPCX_F32 <vopc<0x52>, "V_CMPSX_EQ_F32">;
+defm V_CMPSX_LE_F32 : VOPCX_F32 <vopc<0x53>, "V_CMPSX_LE_F32">;
+defm V_CMPSX_GT_F32 : VOPCX_F32 <vopc<0x54>, "V_CMPSX_GT_F32">;
+defm V_CMPSX_LG_F32 : VOPCX_F32 <vopc<0x55>, "V_CMPSX_LG_F32">;
+defm V_CMPSX_GE_F32 : VOPCX_F32 <vopc<0x56>, "V_CMPSX_GE_F32">;
+defm V_CMPSX_O_F32 : VOPCX_F32 <vopc<0x57>, "V_CMPSX_O_F32">;
+defm V_CMPSX_U_F32 : VOPCX_F32 <vopc<0x58>, "V_CMPSX_U_F32">;
+defm V_CMPSX_NGE_F32 : VOPCX_F32 <vopc<0x59>, "V_CMPSX_NGE_F32">;
+defm V_CMPSX_NLG_F32 : VOPCX_F32 <vopc<0x5a>, "V_CMPSX_NLG_F32">;
+defm V_CMPSX_NGT_F32 : VOPCX_F32 <vopc<0x5b>, "V_CMPSX_NGT_F32">;
+defm V_CMPSX_NLE_F32 : VOPCX_F32 <vopc<0x5c>, "V_CMPSX_NLE_F32">;
+defm V_CMPSX_NEQ_F32 : VOPCX_F32 <vopc<0x5d>, "V_CMPSX_NEQ_F32">;
+defm V_CMPSX_NLT_F32 : VOPCX_F32 <vopc<0x5e>, "V_CMPSX_NLT_F32">;
+defm V_CMPSX_TRU_F32 : VOPCX_F32 <vopc<0x5f>, "V_CMPSX_TRU_F32">;
 
-} // End hasSideEffects = 1, Defs = [EXEC]
+} // End hasSideEffects = 1
 
-defm V_CMPS_F_F64 : VOPC_64 <0x00000060, "V_CMPS_F_F64">;
-defm V_CMPS_LT_F64 : VOPC_64 <0x00000061, "V_CMPS_LT_F64">;
-defm V_CMPS_EQ_F64 : VOPC_64 <0x00000062, "V_CMPS_EQ_F64">;
-defm V_CMPS_LE_F64 : VOPC_64 <0x00000063, "V_CMPS_LE_F64">;
-defm V_CMPS_GT_F64 : VOPC_64 <0x00000064, "V_CMPS_GT_F64">;
-defm V_CMPS_LG_F64 : VOPC_64 <0x00000065, "V_CMPS_LG_F64">;
-defm V_CMPS_GE_F64 : VOPC_64 <0x00000066, "V_CMPS_GE_F64">;
-defm V_CMPS_O_F64 : VOPC_64 <0x00000067, "V_CMPS_O_F64">;
-defm V_CMPS_U_F64 : VOPC_64 <0x00000068, "V_CMPS_U_F64">;
-defm V_CMPS_NGE_F64 : VOPC_64 <0x00000069, "V_CMPS_NGE_F64">;
-defm V_CMPS_NLG_F64 : VOPC_64 <0x0000006a, "V_CMPS_NLG_F64">;
-defm V_CMPS_NGT_F64 : VOPC_64 <0x0000006b, "V_CMPS_NGT_F64">;
-defm V_CMPS_NLE_F64 : VOPC_64 <0x0000006c, "V_CMPS_NLE_F64">;
-defm V_CMPS_NEQ_F64 : VOPC_64 <0x0000006d, "V_CMPS_NEQ_F64">;
-defm V_CMPS_NLT_F64 : VOPC_64 <0x0000006e, "V_CMPS_NLT_F64">;
-defm V_CMPS_TRU_F64 : VOPC_64 <0x0000006f, "V_CMPS_TRU_F64">;
+defm V_CMPS_F_F64 : VOPC_F64 <vopc<0x60>, "V_CMPS_F_F64">;
+defm V_CMPS_LT_F64 : VOPC_F64 <vopc<0x61>, "V_CMPS_LT_F64">;
+defm V_CMPS_EQ_F64 : VOPC_F64 <vopc<0x62>, "V_CMPS_EQ_F64">;
+defm V_CMPS_LE_F64 : VOPC_F64 <vopc<0x63>, "V_CMPS_LE_F64">;
+defm V_CMPS_GT_F64 : VOPC_F64 <vopc<0x64>, "V_CMPS_GT_F64">;
+defm V_CMPS_LG_F64 : VOPC_F64 <vopc<0x65>, "V_CMPS_LG_F64">;
+defm V_CMPS_GE_F64 : VOPC_F64 <vopc<0x66>, "V_CMPS_GE_F64">;
+defm V_CMPS_O_F64 : VOPC_F64 <vopc<0x67>, "V_CMPS_O_F64">;
+defm V_CMPS_U_F64 : VOPC_F64 <vopc<0x68>, "V_CMPS_U_F64">;
+defm V_CMPS_NGE_F64 : VOPC_F64 <vopc<0x69>, "V_CMPS_NGE_F64">;
+defm V_CMPS_NLG_F64 : VOPC_F64 <vopc<0x6a>, "V_CMPS_NLG_F64">;
+defm V_CMPS_NGT_F64 : VOPC_F64 <vopc<0x6b>, "V_CMPS_NGT_F64">;
+defm V_CMPS_NLE_F64 : VOPC_F64 <vopc<0x6c>, "V_CMPS_NLE_F64">;
+defm V_CMPS_NEQ_F64 : VOPC_F64 <vopc<0x6d>, "V_CMPS_NEQ_F64">;
+defm V_CMPS_NLT_F64 : VOPC_F64 <vopc<0x6e>, "V_CMPS_NLT_F64">;
+defm V_CMPS_TRU_F64 : VOPC_F64 <vopc<0x6f>, "V_CMPS_TRU_F64">;
 
 let hasSideEffects = 1, Defs = [EXEC] in {
 
-defm V_CMPSX_F_F64 : VOPC_64 <0x00000070, "V_CMPSX_F_F64">;
-defm V_CMPSX_LT_F64 : VOPC_64 <0x00000071, "V_CMPSX_LT_F64">;
-defm V_CMPSX_EQ_F64 : VOPC_64 <0x00000072, "V_CMPSX_EQ_F64">;
-defm V_CMPSX_LE_F64 : VOPC_64 <0x00000073, "V_CMPSX_LE_F64">;
-defm V_CMPSX_GT_F64 : VOPC_64 <0x00000074, "V_CMPSX_GT_F64">;
-defm V_CMPSX_LG_F64 : VOPC_64 <0x00000075, "V_CMPSX_LG_F64">;
-defm V_CMPSX_GE_F64 : VOPC_64 <0x00000076, "V_CMPSX_GE_F64">;
-defm V_CMPSX_O_F64 : VOPC_64 <0x00000077, "V_CMPSX_O_F64">;
-defm V_CMPSX_U_F64 : VOPC_64 <0x00000078, "V_CMPSX_U_F64">;
-defm V_CMPSX_NGE_F64 : VOPC_64 <0x00000079, "V_CMPSX_NGE_F64">;
-defm V_CMPSX_NLG_F64 : VOPC_64 <0x0000007a, "V_CMPSX_NLG_F64">;
-defm V_CMPSX_NGT_F64 : VOPC_64 <0x0000007b, "V_CMPSX_NGT_F64">;
-defm V_CMPSX_NLE_F64 : VOPC_64 <0x0000007c, "V_CMPSX_NLE_F64">;
-defm V_CMPSX_NEQ_F64 : VOPC_64 <0x0000007d, "V_CMPSX_NEQ_F64">;
-defm V_CMPSX_NLT_F64 : VOPC_64 <0x0000007e, "V_CMPSX_NLT_F64">;
-defm V_CMPSX_TRU_F64 : VOPC_64 <0x0000007f, "V_CMPSX_TRU_F64">;
+defm V_CMPSX_F_F64 : VOPC_F64 <vopc<0x70>, "V_CMPSX_F_F64">;
+defm V_CMPSX_LT_F64 : VOPC_F64 <vopc<0x71>, "V_CMPSX_LT_F64">;
+defm V_CMPSX_EQ_F64 : VOPC_F64 <vopc<0x72>, "V_CMPSX_EQ_F64">;
+defm V_CMPSX_LE_F64 : VOPC_F64 <vopc<0x73>, "V_CMPSX_LE_F64">;
+defm V_CMPSX_GT_F64 : VOPC_F64 <vopc<0x74>, "V_CMPSX_GT_F64">;
+defm V_CMPSX_LG_F64 : VOPC_F64 <vopc<0x75>, "V_CMPSX_LG_F64">;
+defm V_CMPSX_GE_F64 : VOPC_F64 <vopc<0x76>, "V_CMPSX_GE_F64">;
+defm V_CMPSX_O_F64 : VOPC_F64 <vopc<0x77>, "V_CMPSX_O_F64">;
+defm V_CMPSX_U_F64 : VOPC_F64 <vopc<0x78>, "V_CMPSX_U_F64">;
+defm V_CMPSX_NGE_F64 : VOPC_F64 <vopc<0x79>, "V_CMPSX_NGE_F64">;
+defm V_CMPSX_NLG_F64 : VOPC_F64 <vopc<0x7a>, "V_CMPSX_NLG_F64">;
+defm V_CMPSX_NGT_F64 : VOPC_F64 <vopc<0x7b>, "V_CMPSX_NGT_F64">;
+defm V_CMPSX_NLE_F64 : VOPC_F64 <vopc<0x7c>, "V_CMPSX_NLE_F64">;
+defm V_CMPSX_NEQ_F64 : VOPC_F64 <vopc<0x7d>, "V_CMPSX_NEQ_F64">;
+defm V_CMPSX_NLT_F64 : VOPC_F64 <vopc<0x7e>, "V_CMPSX_NLT_F64">;
+defm V_CMPSX_TRU_F64 : VOPC_F64 <vopc<0x7f>, "V_CMPSX_TRU_F64">;
 
 } // End hasSideEffects = 1, Defs = [EXEC]
 
-defm V_CMP_F_I32 : VOPC_32 <0x00000080, "V_CMP_F_I32">;
-defm V_CMP_LT_I32 : VOPC_32 <0x00000081, "V_CMP_LT_I32", i32, COND_LT>;
-defm V_CMP_EQ_I32 : VOPC_32 <0x00000082, "V_CMP_EQ_I32", i32, COND_EQ>;
-defm V_CMP_LE_I32 : VOPC_32 <0x00000083, "V_CMP_LE_I32", i32, COND_LE>;
-defm V_CMP_GT_I32 : VOPC_32 <0x00000084, "V_CMP_GT_I32", i32, COND_GT>;
-defm V_CMP_NE_I32 : VOPC_32 <0x00000085, "V_CMP_NE_I32", i32, COND_NE>;
-defm V_CMP_GE_I32 : VOPC_32 <0x00000086, "V_CMP_GE_I32", i32, COND_GE>;
-defm V_CMP_T_I32 : VOPC_32 <0x00000087, "V_CMP_T_I32">;
+defm V_CMP_F_I32 : VOPC_I32 <vopc<0x80>, "V_CMP_F_I32">;
+defm V_CMP_LT_I32 : VOPC_I32 <vopc<0x81>, "V_CMP_LT_I32", COND_SLT>;
+defm V_CMP_EQ_I32 : VOPC_I32 <vopc<0x82>, "V_CMP_EQ_I32", COND_EQ>;
+defm V_CMP_LE_I32 : VOPC_I32 <vopc<0x83>, "V_CMP_LE_I32", COND_SLE>;
+defm V_CMP_GT_I32 : VOPC_I32 <vopc<0x84>, "V_CMP_GT_I32", COND_SGT>;
+defm V_CMP_NE_I32 : VOPC_I32 <vopc<0x85>, "V_CMP_NE_I32", COND_NE>;
+defm V_CMP_GE_I32 : VOPC_I32 <vopc<0x86>, "V_CMP_GE_I32", COND_SGE>;
+defm V_CMP_T_I32 : VOPC_I32 <vopc<0x87>, "V_CMP_T_I32">;
 
-let hasSideEffects = 1, Defs = [EXEC] in {
+let hasSideEffects = 1 in {
 
-defm V_CMPX_F_I32 : VOPC_32 <0x00000090, "V_CMPX_F_I32">;
-defm V_CMPX_LT_I32 : VOPC_32 <0x00000091, "V_CMPX_LT_I32">;
-defm V_CMPX_EQ_I32 : VOPC_32 <0x00000092, "V_CMPX_EQ_I32">;
-defm V_CMPX_LE_I32 : VOPC_32 <0x00000093, "V_CMPX_LE_I32">;
-defm V_CMPX_GT_I32 : VOPC_32 <0x00000094, "V_CMPX_GT_I32">;
-defm V_CMPX_NE_I32 : VOPC_32 <0x00000095, "V_CMPX_NE_I32">;
-defm V_CMPX_GE_I32 : VOPC_32 <0x00000096, "V_CMPX_GE_I32">;
-defm V_CMPX_T_I32 : VOPC_32 <0x00000097, "V_CMPX_T_I32">;
+defm V_CMPX_F_I32 : VOPCX_I32 <vopc<0x90>, "V_CMPX_F_I32">;
+defm V_CMPX_LT_I32 : VOPCX_I32 <vopc<0x91>, "V_CMPX_LT_I32">;
+defm V_CMPX_EQ_I32 : VOPCX_I32 <vopc<0x92>, "V_CMPX_EQ_I32">;
+defm V_CMPX_LE_I32 : VOPCX_I32 <vopc<0x93>, "V_CMPX_LE_I32">;
+defm V_CMPX_GT_I32 : VOPCX_I32 <vopc<0x94>, "V_CMPX_GT_I32">;
+defm V_CMPX_NE_I32 : VOPCX_I32 <vopc<0x95>, "V_CMPX_NE_I32">;
+defm V_CMPX_GE_I32 : VOPCX_I32 <vopc<0x96>, "V_CMPX_GE_I32">;
+defm V_CMPX_T_I32 : VOPCX_I32 <vopc<0x97>, "V_CMPX_T_I32">;
 
-} // End hasSideEffects = 1, Defs = [EXEC]
+} // End hasSideEffects = 1
 
-defm V_CMP_F_I64 : VOPC_64 <0x000000a0, "V_CMP_F_I64">;
-defm V_CMP_LT_I64 : VOPC_64 <0x000000a1, "V_CMP_LT_I64">;
-defm V_CMP_EQ_I64 : VOPC_64 <0x000000a2, "V_CMP_EQ_I64">;
-defm V_CMP_LE_I64 : VOPC_64 <0x000000a3, "V_CMP_LE_I64">;
-defm V_CMP_GT_I64 : VOPC_64 <0x000000a4, "V_CMP_GT_I64">;
-defm V_CMP_NE_I64 : VOPC_64 <0x000000a5, "V_CMP_NE_I64">;
-defm V_CMP_GE_I64 : VOPC_64 <0x000000a6, "V_CMP_GE_I64">;
-defm V_CMP_T_I64 : VOPC_64 <0x000000a7, "V_CMP_T_I64">;
+defm V_CMP_F_I64 : VOPC_I64 <vopc<0xa0>, "V_CMP_F_I64">;
+defm V_CMP_LT_I64 : VOPC_I64 <vopc<0xa1>, "V_CMP_LT_I64", COND_SLT>;
+defm V_CMP_EQ_I64 : VOPC_I64 <vopc<0xa2>, "V_CMP_EQ_I64", COND_EQ>;
+defm V_CMP_LE_I64 : VOPC_I64 <vopc<0xa3>, "V_CMP_LE_I64", COND_SLE>;
+defm V_CMP_GT_I64 : VOPC_I64 <vopc<0xa4>, "V_CMP_GT_I64", COND_SGT>;
+defm V_CMP_NE_I64 : VOPC_I64 <vopc<0xa5>, "V_CMP_NE_I64", COND_NE>;
+defm V_CMP_GE_I64 : VOPC_I64 <vopc<0xa6>, "V_CMP_GE_I64", COND_SGE>;
+defm V_CMP_T_I64 : VOPC_I64 <vopc<0xa7>, "V_CMP_T_I64">;
 
-let hasSideEffects = 1, Defs = [EXEC] in {
+let hasSideEffects = 1 in {
 
-defm V_CMPX_F_I64 : VOPC_64 <0x000000b0, "V_CMPX_F_I64">;
-defm V_CMPX_LT_I64 : VOPC_64 <0x000000b1, "V_CMPX_LT_I64">;
-defm V_CMPX_EQ_I64 : VOPC_64 <0x000000b2, "V_CMPX_EQ_I64">;
-defm V_CMPX_LE_I64 : VOPC_64 <0x000000b3, "V_CMPX_LE_I64">;
-defm V_CMPX_GT_I64 : VOPC_64 <0x000000b4, "V_CMPX_GT_I64">;
-defm V_CMPX_NE_I64 : VOPC_64 <0x000000b5, "V_CMPX_NE_I64">;
-defm V_CMPX_GE_I64 : VOPC_64 <0x000000b6, "V_CMPX_GE_I64">;
-defm V_CMPX_T_I64 : VOPC_64 <0x000000b7, "V_CMPX_T_I64">;
+defm V_CMPX_F_I64 : VOPCX_I64 <vopc<0xb0>, "V_CMPX_F_I64">;
+defm V_CMPX_LT_I64 : VOPCX_I64 <vopc<0xb1>, "V_CMPX_LT_I64">;
+defm V_CMPX_EQ_I64 : VOPCX_I64 <vopc<0xb2>, "V_CMPX_EQ_I64">;
+defm V_CMPX_LE_I64 : VOPCX_I64 <vopc<0xb3>, "V_CMPX_LE_I64">;
+defm V_CMPX_GT_I64 : VOPCX_I64 <vopc<0xb4>, "V_CMPX_GT_I64">;
+defm V_CMPX_NE_I64 : VOPCX_I64 <vopc<0xb5>, "V_CMPX_NE_I64">;
+defm V_CMPX_GE_I64 : VOPCX_I64 <vopc<0xb6>, "V_CMPX_GE_I64">;
+defm V_CMPX_T_I64 : VOPCX_I64 <vopc<0xb7>, "V_CMPX_T_I64">;
 
-} // End hasSideEffects = 1, Defs = [EXEC]
+} // End hasSideEffects = 1
 
-defm V_CMP_F_U32 : VOPC_32 <0x000000c0, "V_CMP_F_U32">;
-defm V_CMP_LT_U32 : VOPC_32 <0x000000c1, "V_CMP_LT_U32">;
-defm V_CMP_EQ_U32 : VOPC_32 <0x000000c2, "V_CMP_EQ_U32">;
-defm V_CMP_LE_U32 : VOPC_32 <0x000000c3, "V_CMP_LE_U32">;
-defm V_CMP_GT_U32 : VOPC_32 <0x000000c4, "V_CMP_GT_U32">;
-defm V_CMP_NE_U32 : VOPC_32 <0x000000c5, "V_CMP_NE_U32">;
-defm V_CMP_GE_U32 : VOPC_32 <0x000000c6, "V_CMP_GE_U32">;
-defm V_CMP_T_U32 : VOPC_32 <0x000000c7, "V_CMP_T_U32">;
+defm V_CMP_F_U32 : VOPC_I32 <vopc<0xc0>, "V_CMP_F_U32">;
+defm V_CMP_LT_U32 : VOPC_I32 <vopc<0xc1>, "V_CMP_LT_U32", COND_ULT>;
+defm V_CMP_EQ_U32 : VOPC_I32 <vopc<0xc2>, "V_CMP_EQ_U32", COND_EQ>;
+defm V_CMP_LE_U32 : VOPC_I32 <vopc<0xc3>, "V_CMP_LE_U32", COND_ULE>;
+defm V_CMP_GT_U32 : VOPC_I32 <vopc<0xc4>, "V_CMP_GT_U32", COND_UGT>;
+defm V_CMP_NE_U32 : VOPC_I32 <vopc<0xc5>, "V_CMP_NE_U32", COND_NE>;
+defm V_CMP_GE_U32 : VOPC_I32 <vopc<0xc6>, "V_CMP_GE_U32", COND_UGE>;
+defm V_CMP_T_U32 : VOPC_I32 <vopc<0xc7>, "V_CMP_T_U32">;
 
-let hasSideEffects = 1, Defs = [EXEC] in {
+let hasSideEffects = 1 in {
 
-defm V_CMPX_F_U32 : VOPC_32 <0x000000d0, "V_CMPX_F_U32">;
-defm V_CMPX_LT_U32 : VOPC_32 <0x000000d1, "V_CMPX_LT_U32">;
-defm V_CMPX_EQ_U32 : VOPC_32 <0x000000d2, "V_CMPX_EQ_U32">;
-defm V_CMPX_LE_U32 : VOPC_32 <0x000000d3, "V_CMPX_LE_U32">;
-defm V_CMPX_GT_U32 : VOPC_32 <0x000000d4, "V_CMPX_GT_U32">;
-defm V_CMPX_NE_U32 : VOPC_32 <0x000000d5, "V_CMPX_NE_U32">;
-defm V_CMPX_GE_U32 : VOPC_32 <0x000000d6, "V_CMPX_GE_U32">;
-defm V_CMPX_T_U32 : VOPC_32 <0x000000d7, "V_CMPX_T_U32">;
+defm V_CMPX_F_U32 : VOPCX_I32 <vopc<0xd0>, "V_CMPX_F_U32">;
+defm V_CMPX_LT_U32 : VOPCX_I32 <vopc<0xd1>, "V_CMPX_LT_U32">;
+defm V_CMPX_EQ_U32 : VOPCX_I32 <vopc<0xd2>, "V_CMPX_EQ_U32">;
+defm V_CMPX_LE_U32 : VOPCX_I32 <vopc<0xd3>, "V_CMPX_LE_U32">;
+defm V_CMPX_GT_U32 : VOPCX_I32 <vopc<0xd4>, "V_CMPX_GT_U32">;
+defm V_CMPX_NE_U32 : VOPCX_I32 <vopc<0xd5>, "V_CMPX_NE_U32">;
+defm V_CMPX_GE_U32 : VOPCX_I32 <vopc<0xd6>, "V_CMPX_GE_U32">;
+defm V_CMPX_T_U32 : VOPCX_I32 <vopc<0xd7>, "V_CMPX_T_U32">;
 
-} // End hasSideEffects = 1, Defs = [EXEC]
+} // End hasSideEffects = 1
 
-defm V_CMP_F_U64 : VOPC_64 <0x000000e0, "V_CMP_F_U64">;
-defm V_CMP_LT_U64 : VOPC_64 <0x000000e1, "V_CMP_LT_U64">;
-defm V_CMP_EQ_U64 : VOPC_64 <0x000000e2, "V_CMP_EQ_U64">;
-defm V_CMP_LE_U64 : VOPC_64 <0x000000e3, "V_CMP_LE_U64">;
-defm V_CMP_GT_U64 : VOPC_64 <0x000000e4, "V_CMP_GT_U64">;
-defm V_CMP_NE_U64 : VOPC_64 <0x000000e5, "V_CMP_NE_U64">;
-defm V_CMP_GE_U64 : VOPC_64 <0x000000e6, "V_CMP_GE_U64">;
-defm V_CMP_T_U64 : VOPC_64 <0x000000e7, "V_CMP_T_U64">;
+defm V_CMP_F_U64 : VOPC_I64 <vopc<0xe0>, "V_CMP_F_U64">;
+defm V_CMP_LT_U64 : VOPC_I64 <vopc<0xe1>, "V_CMP_LT_U64", COND_ULT>;
+defm V_CMP_EQ_U64 : VOPC_I64 <vopc<0xe2>, "V_CMP_EQ_U64", COND_EQ>;
+defm V_CMP_LE_U64 : VOPC_I64 <vopc<0xe3>, "V_CMP_LE_U64", COND_ULE>;
+defm V_CMP_GT_U64 : VOPC_I64 <vopc<0xe4>, "V_CMP_GT_U64", COND_UGT>;
+defm V_CMP_NE_U64 : VOPC_I64 <vopc<0xe5>, "V_CMP_NE_U64", COND_NE>;
+defm V_CMP_GE_U64 : VOPC_I64 <vopc<0xe6>, "V_CMP_GE_U64", COND_UGE>;
+defm V_CMP_T_U64 : VOPC_I64 <vopc<0xe7>, "V_CMP_T_U64">;
 
-let hasSideEffects = 1, Defs = [EXEC] in {
+let hasSideEffects = 1 in {
 
-defm V_CMPX_F_U64 : VOPC_64 <0x000000f0, "V_CMPX_F_U64">;
-defm V_CMPX_LT_U64 : VOPC_64 <0x000000f1, "V_CMPX_LT_U64">;
-defm V_CMPX_EQ_U64 : VOPC_64 <0x000000f2, "V_CMPX_EQ_U64">;
-defm V_CMPX_LE_U64 : VOPC_64 <0x000000f3, "V_CMPX_LE_U64">;
-defm V_CMPX_GT_U64 : VOPC_64 <0x000000f4, "V_CMPX_GT_U64">;
-defm V_CMPX_NE_U64 : VOPC_64 <0x000000f5, "V_CMPX_NE_U64">;
-defm V_CMPX_GE_U64 : VOPC_64 <0x000000f6, "V_CMPX_GE_U64">;
-defm V_CMPX_T_U64 : VOPC_64 <0x000000f7, "V_CMPX_T_U64">;
+defm V_CMPX_F_U64 : VOPCX_I64 <vopc<0xf0>, "V_CMPX_F_U64">;
+defm V_CMPX_LT_U64 : VOPCX_I64 <vopc<0xf1>, "V_CMPX_LT_U64">;
+defm V_CMPX_EQ_U64 : VOPCX_I64 <vopc<0xf2>, "V_CMPX_EQ_U64">;
+defm V_CMPX_LE_U64 : VOPCX_I64 <vopc<0xf3>, "V_CMPX_LE_U64">;
+defm V_CMPX_GT_U64 : VOPCX_I64 <vopc<0xf4>, "V_CMPX_GT_U64">;
+defm V_CMPX_NE_U64 : VOPCX_I64 <vopc<0xf5>, "V_CMPX_NE_U64">;
+defm V_CMPX_GE_U64 : VOPCX_I64 <vopc<0xf6>, "V_CMPX_GE_U64">;
+defm V_CMPX_T_U64 : VOPCX_I64 <vopc<0xf7>, "V_CMPX_T_U64">;
 
-} // End hasSideEffects = 1, Defs = [EXEC]
+} // End hasSideEffects = 1
 
-defm V_CMP_CLASS_F32 : VOPC_32 <0x00000088, "V_CMP_CLASS_F32">;
+defm V_CMP_CLASS_F32 : VOPC_F32 <vopc<0x88>, "V_CMP_CLASS_F32">;
 
-let hasSideEffects = 1, Defs = [EXEC] in {
-defm V_CMPX_CLASS_F32 : VOPC_32 <0x00000098, "V_CMPX_CLASS_F32">;
-} // End hasSideEffects = 1, Defs = [EXEC]
+let hasSideEffects = 1 in {
+defm V_CMPX_CLASS_F32 : VOPCX_F32 <vopc<0x98>, "V_CMPX_CLASS_F32">;
+} // End hasSideEffects = 1
 
-defm V_CMP_CLASS_F64 : VOPC_64 <0x000000a8, "V_CMP_CLASS_F64">;
+defm V_CMP_CLASS_F64 : VOPC_F64 <vopc<0xa8>, "V_CMP_CLASS_F64">;
 
-let hasSideEffects = 1, Defs = [EXEC] in {
-defm V_CMPX_CLASS_F64 : VOPC_64 <0x000000b8, "V_CMPX_CLASS_F64">;
-} // End hasSideEffects = 1, Defs = [EXEC]
+let hasSideEffects = 1 in {
+defm V_CMPX_CLASS_F64 : VOPCX_F64 <vopc<0xb8>, "V_CMPX_CLASS_F64">;
+} // End hasSideEffects = 1
 
 } // End isCompare = 1
 
-def DS_ADD_U32_RTN : DS_1A1D_RET <0x20, "DS_ADD_U32_RTN", VReg_32>;
-def DS_SUB_U32_RTN : DS_1A1D_RET <0x21, "DS_SUB_U32_RTN", VReg_32>;
+//===----------------------------------------------------------------------===//
+// DS Instructions
+//===----------------------------------------------------------------------===//
+
+
+def DS_ADD_U32 : DS_1A1D_NORET <0x0, "DS_ADD_U32", VReg_32>;
+def DS_SUB_U32 : DS_1A1D_NORET <0x1, "DS_SUB_U32", VReg_32>;
+def DS_RSUB_U32 : DS_1A1D_NORET <0x2, "DS_RSUB_U32", VReg_32>;
+def DS_INC_U32 : DS_1A1D_NORET <0x3, "DS_INC_U32", VReg_32>;
+def DS_DEC_U32 : DS_1A1D_NORET <0x4, "DS_DEC_U32", VReg_32>;
+def DS_MIN_I32 : DS_1A1D_NORET <0x5, "DS_MIN_I32", VReg_32>;
+def DS_MAX_I32 : DS_1A1D_NORET <0x6, "DS_MAX_I32", VReg_32>;
+def DS_MIN_U32 : DS_1A1D_NORET <0x7, "DS_MIN_U32", VReg_32>;
+def DS_MAX_U32 : DS_1A1D_NORET <0x8, "DS_MAX_U32", VReg_32>;
+def DS_AND_B32 : DS_1A1D_NORET <0x9, "DS_AND_B32", VReg_32>;
+def DS_OR_B32 : DS_1A1D_NORET <0xa, "DS_OR_B32", VReg_32>;
+def DS_XOR_B32 : DS_1A1D_NORET <0xb, "DS_XOR_B32", VReg_32>;
+def DS_MSKOR_B32 : DS_1A1D_NORET <0xc, "DS_MSKOR_B32", VReg_32>;
+def DS_CMPST_B32 : DS_1A2D_NORET <0x10, "DS_CMPST_B32", VReg_32>;
+def DS_CMPST_F32 : DS_1A2D_NORET <0x11, "DS_CMPST_F32", VReg_32>;
+def DS_MIN_F32 : DS_1A1D_NORET <0x12, "DS_MIN_F32", VReg_32>;
+def DS_MAX_F32 : DS_1A1D_NORET <0x13, "DS_MAX_F32", VReg_32>;
+
+def DS_ADD_RTN_U32 : DS_1A1D_RET <0x20, "DS_ADD_RTN_U32", VReg_32, "DS_ADD_U32">;
+def DS_SUB_RTN_U32 : DS_1A1D_RET <0x21, "DS_SUB_RTN_U32", VReg_32, "DS_SUB_U32">;
+def DS_RSUB_RTN_U32 : DS_1A1D_RET <0x22, "DS_RSUB_RTN_U32", VReg_32, "DS_RSUB_U32">;
+def DS_INC_RTN_U32 : DS_1A1D_RET <0x23, "DS_INC_RTN_U32", VReg_32, "DS_INC_U32">;
+def DS_DEC_RTN_U32 : DS_1A1D_RET <0x24, "DS_DEC_RTN_U32", VReg_32, "DS_DEC_U32">;
+def DS_MIN_RTN_I32 : DS_1A1D_RET <0x25, "DS_MIN_RTN_I32", VReg_32, "DS_MIN_I32">;
+def DS_MAX_RTN_I32 : DS_1A1D_RET <0x26, "DS_MAX_RTN_I32", VReg_32, "DS_MAX_I32">;
+def DS_MIN_RTN_U32 : DS_1A1D_RET <0x27, "DS_MIN_RTN_U32", VReg_32, "DS_MIN_U32">;
+def DS_MAX_RTN_U32 : DS_1A1D_RET <0x28, "DS_MAX_RTN_U32", VReg_32, "DS_MAX_U32">;
+def DS_AND_RTN_B32 : DS_1A1D_RET <0x29, "DS_AND_RTN_B32", VReg_32, "DS_AND_B32">;
+def DS_OR_RTN_B32 : DS_1A1D_RET <0x2a, "DS_OR_RTN_B32", VReg_32, "DS_OR_B32">;
+def DS_XOR_RTN_B32 : DS_1A1D_RET <0x2b, "DS_XOR_RTN_B32", VReg_32, "DS_XOR_B32">;
+def DS_MSKOR_RTN_B32 : DS_1A1D_RET <0x2c, "DS_MSKOR_RTN_B32", VReg_32, "DS_MSKOR_B32">;
+def DS_WRXCHG_RTN_B32 : DS_1A1D_RET <0x2d, "DS_WRXCHG_RTN_B32", VReg_32>;
+//def DS_WRXCHG2_RTN_B32 : DS_2A0D_RET <0x2e, "DS_WRXCHG2_RTN_B32", VReg_32, "DS_WRXCHG2_B32">;
+//def DS_WRXCHG2ST64_RTN_B32 : DS_2A0D_RET <0x2f, "DS_WRXCHG2_RTN_B32", VReg_32, "DS_WRXCHG2ST64_B32">;
+def DS_CMPST_RTN_B32 : DS_1A2D_RET <0x30, "DS_CMPST_RTN_B32", VReg_32, "DS_CMPST_B32">;
+def DS_CMPST_RTN_F32 : DS_1A2D_RET <0x31, "DS_CMPST_RTN_F32", VReg_32, "DS_CMPST_F32">;
+def DS_MIN_RTN_F32 : DS_1A1D_RET <0x32, "DS_MIN_RTN_F32", VReg_32, "DS_MIN_F32">;
+def DS_MAX_RTN_F32 : DS_1A1D_RET <0x33, "DS_MAX_RTN_F32", VReg_32, "DS_MAX_F32">;
+
+let SubtargetPredicate = isCI in {
+def DS_WRAP_RTN_F32 : DS_1A1D_RET <0x34, "DS_WRAP_RTN_F32", VReg_32, "DS_WRAP_F32">;
+} // End isCI
+
+
+def DS_ADD_U64 : DS_1A1D_NORET <0x40, "DS_ADD_U64", VReg_64>;
+def DS_SUB_U64 : DS_1A1D_NORET <0x41, "DS_SUB_U64", VReg_64>;
+def DS_RSUB_U64 : DS_1A1D_NORET <0x42, "DS_RSUB_U64", VReg_64>;
+def DS_INC_U64 : DS_1A1D_NORET <0x43, "DS_INC_U64", VReg_64>;
+def DS_DEC_U64 : DS_1A1D_NORET <0x44, "DS_DEC_U64", VReg_64>;
+def DS_MIN_I64 : DS_1A1D_NORET <0x45, "DS_MIN_I64", VReg_64>;
+def DS_MAX_I64 : DS_1A1D_NORET <0x46, "DS_MAX_I64", VReg_64>;
+def DS_MIN_U64 : DS_1A1D_NORET <0x47, "DS_MIN_U64", VReg_64>;
+def DS_MAX_U64 : DS_1A1D_NORET <0x48, "DS_MAX_U64", VReg_64>;
+def DS_AND_B64 : DS_1A1D_NORET <0x49, "DS_AND_B64", VReg_64>;
+def DS_OR_B64 : DS_1A1D_NORET <0x4a, "DS_OR_B64", VReg_64>;
+def DS_XOR_B64 : DS_1A1D_NORET <0x4b, "DS_XOR_B64", VReg_64>;
+def DS_MSKOR_B64 : DS_1A1D_NORET <0x4c, "DS_MSKOR_B64", VReg_64>;
+def DS_CMPST_B64 : DS_1A2D_NORET <0x50, "DS_CMPST_B64", VReg_64>;
+def DS_CMPST_F64 : DS_1A2D_NORET <0x51, "DS_CMPST_F64", VReg_64>;
+def DS_MIN_F64 : DS_1A1D_NORET <0x52, "DS_MIN_F64", VReg_64>;
+def DS_MAX_F64 : DS_1A1D_NORET <0x53, "DS_MAX_F64", VReg_64>;
+
+def DS_ADD_RTN_U64 : DS_1A1D_RET <0x60, "DS_ADD_RTN_U64", VReg_64, "DS_ADD_U64">;
+def DS_SUB_RTN_U64 : DS_1A1D_RET <0x61, "DS_SUB_RTN_U64", VReg_64, "DS_SUB_U64">;
+def DS_RSUB_RTN_U64 : DS_1A1D_RET <0x62, "DS_RSUB_RTN_U64", VReg_64, "DS_RSUB_U64">;
+def DS_INC_RTN_U64 : DS_1A1D_RET <0x63, "DS_INC_RTN_U64", VReg_64, "DS_INC_U64">;
+def DS_DEC_RTN_U64 : DS_1A1D_RET <0x64, "DS_DEC_RTN_U64", VReg_64, "DS_DEC_U64">;
+def DS_MIN_RTN_I64 : DS_1A1D_RET <0x65, "DS_MIN_RTN_I64", VReg_64, "DS_MIN_I64">;
+def DS_MAX_RTN_I64 : DS_1A1D_RET <0x66, "DS_MAX_RTN_I64", VReg_64, "DS_MAX_I64">;
+def DS_MIN_RTN_U64 : DS_1A1D_RET <0x67, "DS_MIN_RTN_U64", VReg_64, "DS_MIN_U64">;
+def DS_MAX_RTN_U64 : DS_1A1D_RET <0x68, "DS_MAX_RTN_U64", VReg_64, "DS_MAX_U64">;
+def DS_AND_RTN_B64 : DS_1A1D_RET <0x69, "DS_AND_RTN_B64", VReg_64, "DS_AND_B64">;
+def DS_OR_RTN_B64 : DS_1A1D_RET <0x6a, "DS_OR_RTN_B64", VReg_64, "DS_OR_B64">;
+def DS_XOR_RTN_B64 : DS_1A1D_RET <0x6b, "DS_XOR_RTN_B64", VReg_64, "DS_XOR_B64">;
+def DS_MSKOR_RTN_B64 : DS_1A1D_RET <0x6c, "DS_MSKOR_RTN_B64", VReg_64, "DS_MSKOR_B64">;
+def DS_WRXCHG_RTN_B64 : DS_1A1D_RET <0x6d, "DS_WRXCHG_RTN_B64", VReg_64, "DS_WRXCHG_B64">;
+//def DS_WRXCHG2_RTN_B64 : DS_2A0D_RET <0x6e, "DS_WRXCHG2_RTN_B64", VReg_64, "DS_WRXCHG2_B64">;
+//def DS_WRXCHG2ST64_RTN_B64 : DS_2A0D_RET <0x6f, "DS_WRXCHG2_RTN_B64", VReg_64, "DS_WRXCHG2ST64_B64">;
+def DS_CMPST_RTN_B64 : DS_1A2D_RET <0x70, "DS_CMPST_RTN_B64", VReg_64, "DS_CMPST_B64">;
+def DS_CMPST_RTN_F64 : DS_1A2D_RET <0x71, "DS_CMPST_RTN_F64", VReg_64, "DS_CMPST_F64">;
+def DS_MIN_RTN_F64 : DS_1A1D_RET <0x72, "DS_MIN_F64", VReg_64, "DS_MIN_F64">;
+def DS_MAX_RTN_F64 : DS_1A1D_RET <0x73, "DS_MAX_F64", VReg_64, "DS_MAX_F64">;
+
+//let SubtargetPredicate = isCI in {
+// DS_CONDXCHG32_RTN_B64
+// DS_CONDXCHG32_RTN_B128
+//} // End isCI
+
+// TODO: _SRC2_* forms
+
 def DS_WRITE_B32 : DS_Store_Helper <0x0000000d, "DS_WRITE_B32", VReg_32>;
 def DS_WRITE_B8 : DS_Store_Helper <0x00000001e, "DS_WRITE_B8", VReg_32>;
 def DS_WRITE_B16 : DS_Store_Helper <0x00000001f, "DS_WRITE_B16", VReg_32>;
+def DS_WRITE_B64 : DS_Store_Helper <0x00000004d, "DS_WRITE_B64", VReg_64>;
+
 def DS_READ_B32 : DS_Load_Helper <0x00000036, "DS_READ_B32", VReg_32>;
 def DS_READ_I8 : DS_Load_Helper <0x00000039, "DS_READ_I8", VReg_32>;
 def DS_READ_U8 : DS_Load_Helper <0x0000003a, "DS_READ_U8", VReg_32>;
 def DS_READ_I16 : DS_Load_Helper <0x0000003b, "DS_READ_I16", VReg_32>;
 def DS_READ_U16 : DS_Load_Helper <0x0000003c, "DS_READ_U16", VReg_32>;
+def DS_READ_B64 : DS_Load_Helper <0x00000076, "DS_READ_B64", VReg_64>;
+
+// 2 forms.
+def DS_WRITE2_B32 : DS_Store2_Helper <0x0000000E, "DS_WRITE2_B32", VReg_32>;
+def DS_WRITE2ST64_B32 : DS_Store2_Helper <0x0000000F, "DS_WRITE2ST64_B32", VReg_32>;
+def DS_WRITE2_B64 : DS_Store2_Helper <0x0000004E, "DS_WRITE2_B64", VReg_64>;
+def DS_WRITE2ST64_B64 : DS_Store2_Helper <0x0000004F, "DS_WRITE2ST64_B64", VReg_64>;
+
+def DS_READ2_B32 : DS_Load2_Helper <0x00000037, "DS_READ2_B32", VReg_64>;
+def DS_READ2ST64_B32 : DS_Load2_Helper <0x00000038, "DS_READ2ST64_B32", VReg_64>;
+def DS_READ2_B64 : DS_Load2_Helper <0x00000075, "DS_READ2_B64", VReg_128>;
+def DS_READ2ST64_B64 : DS_Load2_Helper <0x00000076, "DS_READ2ST64_B64", VReg_128>;
+
+//===----------------------------------------------------------------------===//
+// MUBUF Instructions
+//===----------------------------------------------------------------------===//
 
 //def BUFFER_LOAD_FORMAT_X : MUBUF_ <0x00000000, "BUFFER_LOAD_FORMAT_X", []>;
 //def BUFFER_LOAD_FORMAT_XY : MUBUF_ <0x00000001, "BUFFER_LOAD_FORMAT_XY", []>;
@@ -412,45 +854,80 @@ defm BUFFER_LOAD_FORMAT_XYZW : MUBUF_Load_Helper <0x00000003, "BUFFER_LOAD_FORMA
 //def BUFFER_STORE_FORMAT_XY : MUBUF_ <0x00000005, "BUFFER_STORE_FORMAT_XY", []>;
 //def BUFFER_STORE_FORMAT_XYZ : MUBUF_ <0x00000006, "BUFFER_STORE_FORMAT_XYZ", []>;
 //def BUFFER_STORE_FORMAT_XYZW : MUBUF_ <0x00000007, "BUFFER_STORE_FORMAT_XYZW", []>;
-defm BUFFER_LOAD_UBYTE : MUBUF_Load_Helper <0x00000008, "BUFFER_LOAD_UBYTE", VReg_32>;
-defm BUFFER_LOAD_SBYTE : MUBUF_Load_Helper <0x00000009, "BUFFER_LOAD_SBYTE", VReg_32>;
-defm BUFFER_LOAD_USHORT : MUBUF_Load_Helper <0x0000000a, "BUFFER_LOAD_USHORT", VReg_32>;
-defm BUFFER_LOAD_SSHORT : MUBUF_Load_Helper <0x0000000b, "BUFFER_LOAD_SSHORT", VReg_32>;
-defm BUFFER_LOAD_DWORD : MUBUF_Load_Helper <0x0000000c, "BUFFER_LOAD_DWORD", VReg_32>;
-defm BUFFER_LOAD_DWORDX2 : MUBUF_Load_Helper <0x0000000d, "BUFFER_LOAD_DWORDX2", VReg_64>;
-defm BUFFER_LOAD_DWORDX4 : MUBUF_Load_Helper <0x0000000e, "BUFFER_LOAD_DWORDX4", VReg_128>;
+defm BUFFER_LOAD_UBYTE : MUBUF_Load_Helper <
+  0x00000008, "BUFFER_LOAD_UBYTE", VReg_32, i32, az_extloadi8_global
+>;
+defm BUFFER_LOAD_SBYTE : MUBUF_Load_Helper <
+  0x00000009, "BUFFER_LOAD_SBYTE", VReg_32, i32, sextloadi8_global
+>;
+defm BUFFER_LOAD_USHORT : MUBUF_Load_Helper <
+  0x0000000a, "BUFFER_LOAD_USHORT", VReg_32, i32, az_extloadi16_global
+>;
+defm BUFFER_LOAD_SSHORT : MUBUF_Load_Helper <
+  0x0000000b, "BUFFER_LOAD_SSHORT", VReg_32, i32, sextloadi16_global
+>;
+defm BUFFER_LOAD_DWORD : MUBUF_Load_Helper <
+  0x0000000c, "BUFFER_LOAD_DWORD", VReg_32, i32, global_load
+>;
+defm BUFFER_LOAD_DWORDX2 : MUBUF_Load_Helper <
+  0x0000000d, "BUFFER_LOAD_DWORDX2", VReg_64, v2i32, global_load
+>;
+defm BUFFER_LOAD_DWORDX4 : MUBUF_Load_Helper <
+  0x0000000e, "BUFFER_LOAD_DWORDX4", VReg_128, v4i32, global_load
+>;
 
-def BUFFER_STORE_BYTE : MUBUF_Store_Helper <
-  0x00000018, "BUFFER_STORE_BYTE", VReg_32
+defm BUFFER_STORE_BYTE : MUBUF_Store_Helper <
+  0x00000018, "BUFFER_STORE_BYTE", VReg_32, i32, truncstorei8_global
 >;
 
-def BUFFER_STORE_SHORT : MUBUF_Store_Helper <
-  0x0000001a, "BUFFER_STORE_SHORT", VReg_32
+defm BUFFER_STORE_SHORT : MUBUF_Store_Helper <
+  0x0000001a, "BUFFER_STORE_SHORT", VReg_32, i32, truncstorei16_global
 >;
 
-def BUFFER_STORE_DWORD : MUBUF_Store_Helper <
-  0x0000001c, "BUFFER_STORE_DWORD", VReg_32
+defm BUFFER_STORE_DWORD : MUBUF_Store_Helper <
+  0x0000001c, "BUFFER_STORE_DWORD", VReg_32, i32, global_store
 >;
 
-def BUFFER_STORE_DWORDX2 : MUBUF_Store_Helper <
-  0x0000001d, "BUFFER_STORE_DWORDX2", VReg_64
+defm BUFFER_STORE_DWORDX2 : MUBUF_Store_Helper <
+  0x0000001d, "BUFFER_STORE_DWORDX2", VReg_64, v2i32, global_store
 >;
 
-def BUFFER_STORE_DWORDX4 : MUBUF_Store_Helper <
-  0x0000001e, "BUFFER_STORE_DWORDX4", VReg_128
+defm BUFFER_STORE_DWORDX4 : MUBUF_Store_Helper <
+  0x0000001e, "BUFFER_STORE_DWORDX4", VReg_128, v4i32, global_store
 >;
 //def BUFFER_ATOMIC_SWAP : MUBUF_ <0x00000030, "BUFFER_ATOMIC_SWAP", []>;
+defm BUFFER_ATOMIC_SWAP : MUBUF_Atomic <
+  0x00000030, "BUFFER_ATOMIC_SWAP", VReg_32, i32, atomic_swap_global
+>;
 //def BUFFER_ATOMIC_CMPSWAP : MUBUF_ <0x00000031, "BUFFER_ATOMIC_CMPSWAP", []>;
-//def BUFFER_ATOMIC_ADD : MUBUF_ <0x00000032, "BUFFER_ATOMIC_ADD", []>;
-//def BUFFER_ATOMIC_SUB : MUBUF_ <0x00000033, "BUFFER_ATOMIC_SUB", []>;
+defm BUFFER_ATOMIC_ADD : MUBUF_Atomic <
+  0x00000032, "BUFFER_ATOMIC_ADD", VReg_32, i32, atomic_add_global
+>;
+defm BUFFER_ATOMIC_SUB : MUBUF_Atomic <
+  0x00000033, "BUFFER_ATOMIC_SUB", VReg_32, i32, atomic_sub_global
+>;
 //def BUFFER_ATOMIC_RSUB : MUBUF_ <0x00000034, "BUFFER_ATOMIC_RSUB", []>;
-//def BUFFER_ATOMIC_SMIN : MUBUF_ <0x00000035, "BUFFER_ATOMIC_SMIN", []>;
-//def BUFFER_ATOMIC_UMIN : MUBUF_ <0x00000036, "BUFFER_ATOMIC_UMIN", []>;
-//def BUFFER_ATOMIC_SMAX : MUBUF_ <0x00000037, "BUFFER_ATOMIC_SMAX", []>;
-//def BUFFER_ATOMIC_UMAX : MUBUF_ <0x00000038, "BUFFER_ATOMIC_UMAX", []>;
-//def BUFFER_ATOMIC_AND : MUBUF_ <0x00000039, "BUFFER_ATOMIC_AND", []>;
-//def BUFFER_ATOMIC_OR : MUBUF_ <0x0000003a, "BUFFER_ATOMIC_OR", []>;
-//def BUFFER_ATOMIC_XOR : MUBUF_ <0x0000003b, "BUFFER_ATOMIC_XOR", []>;
+defm BUFFER_ATOMIC_SMIN : MUBUF_Atomic <
+  0x00000035, "BUFFER_ATOMIC_SMIN", VReg_32, i32, atomic_min_global
+>;
+defm BUFFER_ATOMIC_UMIN : MUBUF_Atomic <
+  0x00000036, "BUFFER_ATOMIC_UMIN", VReg_32, i32, atomic_umin_global
+>;
+defm BUFFER_ATOMIC_SMAX : MUBUF_Atomic <
+  0x00000037, "BUFFER_ATOMIC_SMAX", VReg_32, i32, atomic_max_global
+>;
+defm BUFFER_ATOMIC_UMAX : MUBUF_Atomic <
+  0x00000038, "BUFFER_ATOMIC_UMAX", VReg_32, i32, atomic_umax_global
+>;
+defm BUFFER_ATOMIC_AND : MUBUF_Atomic <
+  0x00000039, "BUFFER_ATOMIC_AND", VReg_32, i32, atomic_and_global
+>;
+defm BUFFER_ATOMIC_OR : MUBUF_Atomic <
+  0x0000003a, "BUFFER_ATOMIC_OR", VReg_32, i32, atomic_or_global
+>;
+defm BUFFER_ATOMIC_XOR : MUBUF_Atomic <
+  0x0000003b, "BUFFER_ATOMIC_XOR", VReg_32, i32, atomic_xor_global
+>;
 //def BUFFER_ATOMIC_INC : MUBUF_ <0x0000003c, "BUFFER_ATOMIC_INC", []>;
 //def BUFFER_ATOMIC_DEC : MUBUF_ <0x0000003d, "BUFFER_ATOMIC_DEC", []>;
 //def BUFFER_ATOMIC_FCMPSWAP : MUBUF_ <0x0000003e, "BUFFER_ATOMIC_FCMPSWAP", []>;
@@ -475,47 +952,24 @@ def BUFFER_STORE_DWORDX4 : MUBUF_Store_Helper <
 //def BUFFER_ATOMIC_FMAX_X2 : MUBUF_X2 <0x00000060, "BUFFER_ATOMIC_FMAX_X2", []>;
 //def BUFFER_WBINVL1_SC : MUBUF_WBINVL1 <0x00000070, "BUFFER_WBINVL1_SC", []>;
 //def BUFFER_WBINVL1 : MUBUF_WBINVL1 <0x00000071, "BUFFER_WBINVL1", []>;
+
+//===----------------------------------------------------------------------===//
+// MTBUF Instructions
+//===----------------------------------------------------------------------===//
+
 //def TBUFFER_LOAD_FORMAT_X : MTBUF_ <0x00000000, "TBUFFER_LOAD_FORMAT_X", []>;
 //def TBUFFER_LOAD_FORMAT_XY : MTBUF_ <0x00000001, "TBUFFER_LOAD_FORMAT_XY", []>;
 //def TBUFFER_LOAD_FORMAT_XYZ : MTBUF_ <0x00000002, "TBUFFER_LOAD_FORMAT_XYZ", []>;
-def TBUFFER_LOAD_FORMAT_XYZW : MTBUF_Load_Helper <0x00000003, "TBUFFER_LOAD_FORMAT_XYZW", VReg_128>;
-def TBUFFER_STORE_FORMAT_X : MTBUF_Store_Helper <0x00000004, "TBUFFER_STORE_FORMAT_X", VReg_32>;
-def TBUFFER_STORE_FORMAT_XY : MTBUF_Store_Helper <0x00000005, "TBUFFER_STORE_FORMAT_XY", VReg_64>;
-def TBUFFER_STORE_FORMAT_XYZ : MTBUF_Store_Helper <0x00000006, "TBUFFER_STORE_FORMAT_XYZ", VReg_128>;
-def TBUFFER_STORE_FORMAT_XYZW : MTBUF_Store_Helper <0x00000007, "TBUFFER_STORE_FORMAT_XYZW", VReg_128>;
-
-let mayLoad = 1 in {
-
-defm S_LOAD_DWORD : SMRD_Helper <0x00, "S_LOAD_DWORD", SReg_64, SReg_32>;
-defm S_LOAD_DWORDX2 : SMRD_Helper <0x01, "S_LOAD_DWORDX2", SReg_64, SReg_64>;
-defm S_LOAD_DWORDX4 : SMRD_Helper <0x02, "S_LOAD_DWORDX4", SReg_64, SReg_128>;
-defm S_LOAD_DWORDX8 : SMRD_Helper <0x03, "S_LOAD_DWORDX8", SReg_64, SReg_256>;
-defm S_LOAD_DWORDX16 : SMRD_Helper <0x04, "S_LOAD_DWORDX16", SReg_64, SReg_512>;
-
-defm S_BUFFER_LOAD_DWORD : SMRD_Helper <
-  0x08, "S_BUFFER_LOAD_DWORD", SReg_128, SReg_32
->;
-
-defm S_BUFFER_LOAD_DWORDX2 : SMRD_Helper <
-  0x09, "S_BUFFER_LOAD_DWORDX2", SReg_128, SReg_64
->;
-
-defm S_BUFFER_LOAD_DWORDX4 : SMRD_Helper <
-  0x0a, "S_BUFFER_LOAD_DWORDX4", SReg_128, SReg_128
->;
-
-defm S_BUFFER_LOAD_DWORDX8 : SMRD_Helper <
-  0x0b, "S_BUFFER_LOAD_DWORDX8", SReg_128, SReg_256
->;
-
-defm S_BUFFER_LOAD_DWORDX16 : SMRD_Helper <
-  0x0c, "S_BUFFER_LOAD_DWORDX16", SReg_128, SReg_512
->;
+defm TBUFFER_LOAD_FORMAT_XYZW : MTBUF_Load_Helper <0x00000003, "TBUFFER_LOAD_FORMAT_XYZW", VReg_128>;
+defm TBUFFER_STORE_FORMAT_X : MTBUF_Store_Helper <0x00000004, "TBUFFER_STORE_FORMAT_X", VReg_32>;
+defm TBUFFER_STORE_FORMAT_XY : MTBUF_Store_Helper <0x00000005, "TBUFFER_STORE_FORMAT_XY", VReg_64>;
+defm TBUFFER_STORE_FORMAT_XYZ : MTBUF_Store_Helper <0x00000006, "TBUFFER_STORE_FORMAT_XYZ", VReg_128>;
+defm TBUFFER_STORE_FORMAT_XYZW : MTBUF_Store_Helper <0x00000007, "TBUFFER_STORE_FORMAT_XYZW", VReg_128>;
 
-} // mayLoad = 1
+//===----------------------------------------------------------------------===//
+// MIMG Instructions
+//===----------------------------------------------------------------------===//
 
-//def S_MEMTIME : SMRD_ <0x0000001e, "S_MEMTIME", []>;
-//def S_DCACHE_INV : SMRD_ <0x0000001f, "S_DCACHE_INV", []>;
 defm IMAGE_LOAD : MIMG_NoSampler <0x00000000, "IMAGE_LOAD">;
 defm IMAGE_LOAD_MIP : MIMG_NoSampler <0x00000001, "IMAGE_LOAD_MIP">;
 //def IMAGE_LOAD_PCK : MIMG_NoPattern_ <"IMAGE_LOAD_PCK", 0x00000002>;
@@ -544,179 +998,302 @@ defm IMAGE_GET_RESINFO : MIMG_NoSampler <0x0000000e, "IMAGE_GET_RESINFO">;
 //def IMAGE_ATOMIC_FCMPSWAP : MIMG_NoPattern_ <"IMAGE_ATOMIC_FCMPSWAP", 0x0000001d>;
 //def IMAGE_ATOMIC_FMIN : MIMG_NoPattern_ <"IMAGE_ATOMIC_FMIN", 0x0000001e>;
 //def IMAGE_ATOMIC_FMAX : MIMG_NoPattern_ <"IMAGE_ATOMIC_FMAX", 0x0000001f>;
-defm IMAGE_SAMPLE : MIMG_Sampler <0x00000020, "IMAGE_SAMPLE">;
-//def IMAGE_SAMPLE_CL : MIMG_NoPattern_ <"IMAGE_SAMPLE_CL", 0x00000021>;
-defm IMAGE_SAMPLE_D : MIMG_Sampler <0x00000022, "IMAGE_SAMPLE_D">;
-//def IMAGE_SAMPLE_D_CL : MIMG_NoPattern_ <"IMAGE_SAMPLE_D_CL", 0x00000023>;
-defm IMAGE_SAMPLE_L : MIMG_Sampler <0x00000024, "IMAGE_SAMPLE_L">;
-defm IMAGE_SAMPLE_B : MIMG_Sampler <0x00000025, "IMAGE_SAMPLE_B">;
-//def IMAGE_SAMPLE_B_CL : MIMG_NoPattern_ <"IMAGE_SAMPLE_B_CL", 0x00000026>;
-//def IMAGE_SAMPLE_LZ : MIMG_NoPattern_ <"IMAGE_SAMPLE_LZ", 0x00000027>;
-defm IMAGE_SAMPLE_C : MIMG_Sampler <0x00000028, "IMAGE_SAMPLE_C">;
-//def IMAGE_SAMPLE_C_CL : MIMG_NoPattern_ <"IMAGE_SAMPLE_C_CL", 0x00000029>;
-defm IMAGE_SAMPLE_C_D : MIMG_Sampler <0x0000002a, "IMAGE_SAMPLE_C_D">;
-//def IMAGE_SAMPLE_C_D_CL : MIMG_NoPattern_ <"IMAGE_SAMPLE_C_D_CL", 0x0000002b>;
-defm IMAGE_SAMPLE_C_L : MIMG_Sampler <0x0000002c, "IMAGE_SAMPLE_C_L">;
-defm IMAGE_SAMPLE_C_B : MIMG_Sampler <0x0000002d, "IMAGE_SAMPLE_C_B">;
-//def IMAGE_SAMPLE_C_B_CL : MIMG_NoPattern_ <"IMAGE_SAMPLE_C_B_CL", 0x0000002e>;
-//def IMAGE_SAMPLE_C_LZ : MIMG_NoPattern_ <"IMAGE_SAMPLE_C_LZ", 0x0000002f>;
-//def IMAGE_SAMPLE_O : MIMG_NoPattern_ <"IMAGE_SAMPLE_O", 0x00000030>;
-//def IMAGE_SAMPLE_CL_O : MIMG_NoPattern_ <"IMAGE_SAMPLE_CL_O", 0x00000031>;
-//def IMAGE_SAMPLE_D_O : MIMG_NoPattern_ <"IMAGE_SAMPLE_D_O", 0x00000032>;
-//def IMAGE_SAMPLE_D_CL_O : MIMG_NoPattern_ <"IMAGE_SAMPLE_D_CL_O", 0x00000033>;
-//def IMAGE_SAMPLE_L_O : MIMG_NoPattern_ <"IMAGE_SAMPLE_L_O", 0x00000034>;
-//def IMAGE_SAMPLE_B_O : MIMG_NoPattern_ <"IMAGE_SAMPLE_B_O", 0x00000035>;
-//def IMAGE_SAMPLE_B_CL_O : MIMG_NoPattern_ <"IMAGE_SAMPLE_B_CL_O", 0x00000036>;
-//def IMAGE_SAMPLE_LZ_O : MIMG_NoPattern_ <"IMAGE_SAMPLE_LZ_O", 0x00000037>;
-//def IMAGE_SAMPLE_C_O : MIMG_NoPattern_ <"IMAGE_SAMPLE_C_O", 0x00000038>;
-//def IMAGE_SAMPLE_C_CL_O : MIMG_NoPattern_ <"IMAGE_SAMPLE_C_CL_O", 0x00000039>;
-//def IMAGE_SAMPLE_C_D_O : MIMG_NoPattern_ <"IMAGE_SAMPLE_C_D_O", 0x0000003a>;
-//def IMAGE_SAMPLE_C_D_CL_O : MIMG_NoPattern_ <"IMAGE_SAMPLE_C_D_CL_O", 0x0000003b>;
-//def IMAGE_SAMPLE_C_L_O : MIMG_NoPattern_ <"IMAGE_SAMPLE_C_L_O", 0x0000003c>;
-//def IMAGE_SAMPLE_C_B_O : MIMG_NoPattern_ <"IMAGE_SAMPLE_C_B_O", 0x0000003d>;
-//def IMAGE_SAMPLE_C_B_CL_O : MIMG_NoPattern_ <"IMAGE_SAMPLE_C_B_CL_O", 0x0000003e>;
-//def IMAGE_SAMPLE_C_LZ_O : MIMG_NoPattern_ <"IMAGE_SAMPLE_C_LZ_O", 0x0000003f>;
-//def IMAGE_GATHER4 : MIMG_NoPattern_GATHER4 <"IMAGE_GATHER4", 0x00000040>;
-//def IMAGE_GATHER4_CL : MIMG_NoPattern_GATHER4 <"IMAGE_GATHER4_CL", 0x00000041>;
-//def IMAGE_GATHER4_L : MIMG_NoPattern_GATHER4 <"IMAGE_GATHER4_L", 0x00000044>;
-//def IMAGE_GATHER4_B : MIMG_NoPattern_GATHER4 <"IMAGE_GATHER4_B", 0x00000045>;
-//def IMAGE_GATHER4_B_CL : MIMG_NoPattern_GATHER4 <"IMAGE_GATHER4_B_CL", 0x00000046>;
-//def IMAGE_GATHER4_LZ : MIMG_NoPattern_GATHER4 <"IMAGE_GATHER4_LZ", 0x00000047>;
-//def IMAGE_GATHER4_C : MIMG_NoPattern_GATHER4 <"IMAGE_GATHER4_C", 0x00000048>;
-//def IMAGE_GATHER4_C_CL : MIMG_NoPattern_GATHER4 <"IMAGE_GATHER4_C_CL", 0x00000049>;
-//def IMAGE_GATHER4_C_L : MIMG_NoPattern_GATHER4 <"IMAGE_GATHER4_C_L", 0x0000004c>;
-//def IMAGE_GATHER4_C_B : MIMG_NoPattern_GATHER4 <"IMAGE_GATHER4_C_B", 0x0000004d>;
-//def IMAGE_GATHER4_C_B_CL : MIMG_NoPattern_GATHER4 <"IMAGE_GATHER4_C_B_CL", 0x0000004e>;
-//def IMAGE_GATHER4_C_LZ : MIMG_NoPattern_GATHER4 <"IMAGE_GATHER4_C_LZ", 0x0000004f>;
-//def IMAGE_GATHER4_O : MIMG_NoPattern_GATHER4 <"IMAGE_GATHER4_O", 0x00000050>;
-//def IMAGE_GATHER4_CL_O : MIMG_NoPattern_GATHER4 <"IMAGE_GATHER4_CL_O", 0x00000051>;
-//def IMAGE_GATHER4_L_O : MIMG_NoPattern_GATHER4 <"IMAGE_GATHER4_L_O", 0x00000054>;
-//def IMAGE_GATHER4_B_O : MIMG_NoPattern_GATHER4 <"IMAGE_GATHER4_B_O", 0x00000055>;
-//def IMAGE_GATHER4_B_CL_O : MIMG_NoPattern_GATHER4 <"IMAGE_GATHER4_B_CL_O", 0x00000056>;
-//def IMAGE_GATHER4_LZ_O : MIMG_NoPattern_GATHER4 <"IMAGE_GATHER4_LZ_O", 0x00000057>;
-//def IMAGE_GATHER4_C_O : MIMG_NoPattern_GATHER4 <"IMAGE_GATHER4_C_O", 0x00000058>;
-//def IMAGE_GATHER4_C_CL_O : MIMG_NoPattern_GATHER4 <"IMAGE_GATHER4_C_CL_O", 0x00000059>;
-//def IMAGE_GATHER4_C_L_O : MIMG_NoPattern_GATHER4 <"IMAGE_GATHER4_C_L_O", 0x0000005c>;
-//def IMAGE_GATHER4_C_B_O : MIMG_NoPattern_GATHER4 <"IMAGE_GATHER4_C_B_O", 0x0000005d>;
-//def IMAGE_GATHER4_C_B_CL_O : MIMG_NoPattern_GATHER4 <"IMAGE_GATHER4_C_B_CL_O", 0x0000005e>;
-//def IMAGE_GATHER4_C_LZ_O : MIMG_NoPattern_GATHER4 <"IMAGE_GATHER4_C_LZ_O", 0x0000005f>;
-//def IMAGE_GET_LOD : MIMG_NoPattern_ <"IMAGE_GET_LOD", 0x00000060>;
-//def IMAGE_SAMPLE_CD : MIMG_NoPattern_ <"IMAGE_SAMPLE_CD", 0x00000068>;
-//def IMAGE_SAMPLE_CD_CL : MIMG_NoPattern_ <"IMAGE_SAMPLE_CD_CL", 0x00000069>;
-//def IMAGE_SAMPLE_C_CD : MIMG_NoPattern_ <"IMAGE_SAMPLE_C_CD", 0x0000006a>;
-//def IMAGE_SAMPLE_C_CD_CL : MIMG_NoPattern_ <"IMAGE_SAMPLE_C_CD_CL", 0x0000006b>;
-//def IMAGE_SAMPLE_CD_O : MIMG_NoPattern_ <"IMAGE_SAMPLE_CD_O", 0x0000006c>;
-//def IMAGE_SAMPLE_CD_CL_O : MIMG_NoPattern_ <"IMAGE_SAMPLE_CD_CL_O", 0x0000006d>;
-//def IMAGE_SAMPLE_C_CD_O : MIMG_NoPattern_ <"IMAGE_SAMPLE_C_CD_O", 0x0000006e>;
-//def IMAGE_SAMPLE_C_CD_CL_O : MIMG_NoPattern_ <"IMAGE_SAMPLE_C_CD_CL_O", 0x0000006f>;
+defm IMAGE_SAMPLE           : MIMG_Sampler <0x00000020, "IMAGE_SAMPLE">;
+defm IMAGE_SAMPLE_CL        : MIMG_Sampler <0x00000021, "IMAGE_SAMPLE_CL">;
+defm IMAGE_SAMPLE_D         : MIMG_Sampler <0x00000022, "IMAGE_SAMPLE_D">;
+defm IMAGE_SAMPLE_D_CL      : MIMG_Sampler <0x00000023, "IMAGE_SAMPLE_D_CL">;
+defm IMAGE_SAMPLE_L         : MIMG_Sampler <0x00000024, "IMAGE_SAMPLE_L">;
+defm IMAGE_SAMPLE_B         : MIMG_Sampler <0x00000025, "IMAGE_SAMPLE_B">;
+defm IMAGE_SAMPLE_B_CL      : MIMG_Sampler <0x00000026, "IMAGE_SAMPLE_B_CL">;
+defm IMAGE_SAMPLE_LZ        : MIMG_Sampler <0x00000027, "IMAGE_SAMPLE_LZ">;
+defm IMAGE_SAMPLE_C         : MIMG_Sampler <0x00000028, "IMAGE_SAMPLE_C">;
+defm IMAGE_SAMPLE_C_CL      : MIMG_Sampler <0x00000029, "IMAGE_SAMPLE_C_CL">;
+defm IMAGE_SAMPLE_C_D       : MIMG_Sampler <0x0000002a, "IMAGE_SAMPLE_C_D">;
+defm IMAGE_SAMPLE_C_D_CL    : MIMG_Sampler <0x0000002b, "IMAGE_SAMPLE_C_D_CL">;
+defm IMAGE_SAMPLE_C_L       : MIMG_Sampler <0x0000002c, "IMAGE_SAMPLE_C_L">;
+defm IMAGE_SAMPLE_C_B       : MIMG_Sampler <0x0000002d, "IMAGE_SAMPLE_C_B">;
+defm IMAGE_SAMPLE_C_B_CL    : MIMG_Sampler <0x0000002e, "IMAGE_SAMPLE_C_B_CL">;
+defm IMAGE_SAMPLE_C_LZ      : MIMG_Sampler <0x0000002f, "IMAGE_SAMPLE_C_LZ">;
+defm IMAGE_SAMPLE_O         : MIMG_Sampler <0x00000030, "IMAGE_SAMPLE_O">;
+defm IMAGE_SAMPLE_CL_O      : MIMG_Sampler <0x00000031, "IMAGE_SAMPLE_CL_O">;
+defm IMAGE_SAMPLE_D_O       : MIMG_Sampler <0x00000032, "IMAGE_SAMPLE_D_O">;
+defm IMAGE_SAMPLE_D_CL_O    : MIMG_Sampler <0x00000033, "IMAGE_SAMPLE_D_CL_O">;
+defm IMAGE_SAMPLE_L_O       : MIMG_Sampler <0x00000034, "IMAGE_SAMPLE_L_O">;
+defm IMAGE_SAMPLE_B_O       : MIMG_Sampler <0x00000035, "IMAGE_SAMPLE_B_O">;
+defm IMAGE_SAMPLE_B_CL_O    : MIMG_Sampler <0x00000036, "IMAGE_SAMPLE_B_CL_O">;
+defm IMAGE_SAMPLE_LZ_O      : MIMG_Sampler <0x00000037, "IMAGE_SAMPLE_LZ_O">;
+defm IMAGE_SAMPLE_C_O       : MIMG_Sampler <0x00000038, "IMAGE_SAMPLE_C_O">;
+defm IMAGE_SAMPLE_C_CL_O    : MIMG_Sampler <0x00000039, "IMAGE_SAMPLE_C_CL_O">;
+defm IMAGE_SAMPLE_C_D_O     : MIMG_Sampler <0x0000003a, "IMAGE_SAMPLE_C_D_O">;
+defm IMAGE_SAMPLE_C_D_CL_O  : MIMG_Sampler <0x0000003b, "IMAGE_SAMPLE_C_D_CL_O">;
+defm IMAGE_SAMPLE_C_L_O     : MIMG_Sampler <0x0000003c, "IMAGE_SAMPLE_C_L_O">;
+defm IMAGE_SAMPLE_C_B_O     : MIMG_Sampler <0x0000003d, "IMAGE_SAMPLE_C_B_O">;
+defm IMAGE_SAMPLE_C_B_CL_O  : MIMG_Sampler <0x0000003e, "IMAGE_SAMPLE_C_B_CL_O">;
+defm IMAGE_SAMPLE_C_LZ_O    : MIMG_Sampler <0x0000003f, "IMAGE_SAMPLE_C_LZ_O">;
+defm IMAGE_GATHER4          : MIMG_Gather <0x00000040, "IMAGE_GATHER4">;
+defm IMAGE_GATHER4_CL       : MIMG_Gather <0x00000041, "IMAGE_GATHER4_CL">;
+defm IMAGE_GATHER4_L        : MIMG_Gather <0x00000044, "IMAGE_GATHER4_L">;
+defm IMAGE_GATHER4_B        : MIMG_Gather <0x00000045, "IMAGE_GATHER4_B">;
+defm IMAGE_GATHER4_B_CL     : MIMG_Gather <0x00000046, "IMAGE_GATHER4_B_CL">;
+defm IMAGE_GATHER4_LZ       : MIMG_Gather <0x00000047, "IMAGE_GATHER4_LZ">;
+defm IMAGE_GATHER4_C        : MIMG_Gather <0x00000048, "IMAGE_GATHER4_C">;
+defm IMAGE_GATHER4_C_CL     : MIMG_Gather <0x00000049, "IMAGE_GATHER4_C_CL">;
+defm IMAGE_GATHER4_C_L      : MIMG_Gather <0x0000004c, "IMAGE_GATHER4_C_L">;
+defm IMAGE_GATHER4_C_B      : MIMG_Gather <0x0000004d, "IMAGE_GATHER4_C_B">;
+defm IMAGE_GATHER4_C_B_CL   : MIMG_Gather <0x0000004e, "IMAGE_GATHER4_C_B_CL">;
+defm IMAGE_GATHER4_C_LZ     : MIMG_Gather <0x0000004f, "IMAGE_GATHER4_C_LZ">;
+defm IMAGE_GATHER4_O        : MIMG_Gather <0x00000050, "IMAGE_GATHER4_O">;
+defm IMAGE_GATHER4_CL_O     : MIMG_Gather <0x00000051, "IMAGE_GATHER4_CL_O">;
+defm IMAGE_GATHER4_L_O      : MIMG_Gather <0x00000054, "IMAGE_GATHER4_L_O">;
+defm IMAGE_GATHER4_B_O      : MIMG_Gather <0x00000055, "IMAGE_GATHER4_B_O">;
+defm IMAGE_GATHER4_B_CL_O   : MIMG_Gather <0x00000056, "IMAGE_GATHER4_B_CL_O">;
+defm IMAGE_GATHER4_LZ_O     : MIMG_Gather <0x00000057, "IMAGE_GATHER4_LZ_O">;
+defm IMAGE_GATHER4_C_O      : MIMG_Gather <0x00000058, "IMAGE_GATHER4_C_O">;
+defm IMAGE_GATHER4_C_CL_O   : MIMG_Gather <0x00000059, "IMAGE_GATHER4_C_CL_O">;
+defm IMAGE_GATHER4_C_L_O    : MIMG_Gather <0x0000005c, "IMAGE_GATHER4_C_L_O">;
+defm IMAGE_GATHER4_C_B_O    : MIMG_Gather <0x0000005d, "IMAGE_GATHER4_C_B_O">;
+defm IMAGE_GATHER4_C_B_CL_O : MIMG_Gather <0x0000005e, "IMAGE_GATHER4_C_B_CL_O">;
+defm IMAGE_GATHER4_C_LZ_O   : MIMG_Gather <0x0000005f, "IMAGE_GATHER4_C_LZ_O">;
+defm IMAGE_GET_LOD          : MIMG_Sampler <0x00000060, "IMAGE_GET_LOD">;
+defm IMAGE_SAMPLE_CD        : MIMG_Sampler <0x00000068, "IMAGE_SAMPLE_CD">;
+defm IMAGE_SAMPLE_CD_CL     : MIMG_Sampler <0x00000069, "IMAGE_SAMPLE_CD_CL">;
+defm IMAGE_SAMPLE_C_CD      : MIMG_Sampler <0x0000006a, "IMAGE_SAMPLE_C_CD">;
+defm IMAGE_SAMPLE_C_CD_CL   : MIMG_Sampler <0x0000006b, "IMAGE_SAMPLE_C_CD_CL">;
+defm IMAGE_SAMPLE_CD_O      : MIMG_Sampler <0x0000006c, "IMAGE_SAMPLE_CD_O">;
+defm IMAGE_SAMPLE_CD_CL_O   : MIMG_Sampler <0x0000006d, "IMAGE_SAMPLE_CD_CL_O">;
+defm IMAGE_SAMPLE_C_CD_O    : MIMG_Sampler <0x0000006e, "IMAGE_SAMPLE_C_CD_O">;
+defm IMAGE_SAMPLE_C_CD_CL_O : MIMG_Sampler <0x0000006f, "IMAGE_SAMPLE_C_CD_CL_O">;
 //def IMAGE_RSRC256 : MIMG_NoPattern_RSRC256 <"IMAGE_RSRC256", 0x0000007e>;
 //def IMAGE_SAMPLER : MIMG_NoPattern_ <"IMAGE_SAMPLER", 0x0000007f>;
+
+//===----------------------------------------------------------------------===//
+// Flat Instructions
+//===----------------------------------------------------------------------===//
+
+let Predicates = [HasFlatAddressSpace] in {
+def FLAT_LOAD_UBYTE : FLAT_Load_Helper <0x00000008, "FLAT_LOAD_UBYTE", VReg_32>;
+def FLAT_LOAD_SBYTE : FLAT_Load_Helper <0x00000009, "FLAT_LOAD_SBYTE", VReg_32>;
+def FLAT_LOAD_USHORT : FLAT_Load_Helper <0x0000000a, "FLAT_LOAD_USHORT", VReg_32>;
+def FLAT_LOAD_SSHORT : FLAT_Load_Helper <0x0000000b, "FLAT_LOAD_SSHORT", VReg_32>;
+def FLAT_LOAD_DWORD : FLAT_Load_Helper <0x0000000c, "FLAT_LOAD_DWORD", VReg_32>;
+def FLAT_LOAD_DWORDX2 : FLAT_Load_Helper <0x0000000d, "FLAT_LOAD_DWORDX2", VReg_64>;
+def FLAT_LOAD_DWORDX4 : FLAT_Load_Helper <0x0000000e, "FLAT_LOAD_DWORDX4", VReg_128>;
+def FLAT_LOAD_DWORDX3 : FLAT_Load_Helper <0x00000010, "FLAT_LOAD_DWORDX3", VReg_96>;
+
+def FLAT_STORE_BYTE : FLAT_Store_Helper <
+  0x00000018, "FLAT_STORE_BYTE", VReg_32
+>;
+
+def FLAT_STORE_SHORT : FLAT_Store_Helper <
+  0x0000001a, "FLAT_STORE_SHORT", VReg_32
+>;
+
+def FLAT_STORE_DWORD : FLAT_Store_Helper <
+  0x0000001c, "FLAT_STORE_DWORD", VReg_32
+>;
+
+def FLAT_STORE_DWORDX2 : FLAT_Store_Helper <
+  0x0000001d, "FLAT_STORE_DWORDX2", VReg_64
+>;
+
+def FLAT_STORE_DWORDX4 : FLAT_Store_Helper <
+  0x0000001e, "FLAT_STORE_DWORDX4", VReg_128
+>;
+
+def FLAT_STORE_DWORDX3 : FLAT_Store_Helper <
+  0x0000001e, "FLAT_STORE_DWORDX3", VReg_96
+>;
+
+//def FLAT_ATOMIC_SWAP : FLAT_ <0x00000030, "FLAT_ATOMIC_SWAP", []>;
+//def FLAT_ATOMIC_CMPSWAP : FLAT_ <0x00000031, "FLAT_ATOMIC_CMPSWAP", []>;
+//def FLAT_ATOMIC_ADD : FLAT_ <0x00000032, "FLAT_ATOMIC_ADD", []>;
+//def FLAT_ATOMIC_SUB : FLAT_ <0x00000033, "FLAT_ATOMIC_SUB", []>;
+//def FLAT_ATOMIC_RSUB : FLAT_ <0x00000034, "FLAT_ATOMIC_RSUB", []>;
+//def FLAT_ATOMIC_SMIN : FLAT_ <0x00000035, "FLAT_ATOMIC_SMIN", []>;
+//def FLAT_ATOMIC_UMIN : FLAT_ <0x00000036, "FLAT_ATOMIC_UMIN", []>;
+//def FLAT_ATOMIC_SMAX : FLAT_ <0x00000037, "FLAT_ATOMIC_SMAX", []>;
+//def FLAT_ATOMIC_UMAX : FLAT_ <0x00000038, "FLAT_ATOMIC_UMAX", []>;
+//def FLAT_ATOMIC_AND : FLAT_ <0x00000039, "FLAT_ATOMIC_AND", []>;
+//def FLAT_ATOMIC_OR : FLAT_ <0x0000003a, "FLAT_ATOMIC_OR", []>;
+//def FLAT_ATOMIC_XOR : FLAT_ <0x0000003b, "FLAT_ATOMIC_XOR", []>;
+//def FLAT_ATOMIC_INC : FLAT_ <0x0000003c, "FLAT_ATOMIC_INC", []>;
+//def FLAT_ATOMIC_DEC : FLAT_ <0x0000003d, "FLAT_ATOMIC_DEC", []>;
+//def FLAT_ATOMIC_FCMPSWAP : FLAT_ <0x0000003e, "FLAT_ATOMIC_FCMPSWAP", []>;
+//def FLAT_ATOMIC_FMIN : FLAT_ <0x0000003f, "FLAT_ATOMIC_FMIN", []>;
+//def FLAT_ATOMIC_FMAX : FLAT_ <0x00000040, "FLAT_ATOMIC_FMAX", []>;
+//def FLAT_ATOMIC_SWAP_X2 : FLAT_X2 <0x00000050, "FLAT_ATOMIC_SWAP_X2", []>;
+//def FLAT_ATOMIC_CMPSWAP_X2 : FLAT_X2 <0x00000051, "FLAT_ATOMIC_CMPSWAP_X2", []>;
+//def FLAT_ATOMIC_ADD_X2 : FLAT_X2 <0x00000052, "FLAT_ATOMIC_ADD_X2", []>;
+//def FLAT_ATOMIC_SUB_X2 : FLAT_X2 <0x00000053, "FLAT_ATOMIC_SUB_X2", []>;
+//def FLAT_ATOMIC_RSUB_X2 : FLAT_X2 <0x00000054, "FLAT_ATOMIC_RSUB_X2", []>;
+//def FLAT_ATOMIC_SMIN_X2 : FLAT_X2 <0x00000055, "FLAT_ATOMIC_SMIN_X2", []>;
+//def FLAT_ATOMIC_UMIN_X2 : FLAT_X2 <0x00000056, "FLAT_ATOMIC_UMIN_X2", []>;
+//def FLAT_ATOMIC_SMAX_X2 : FLAT_X2 <0x00000057, "FLAT_ATOMIC_SMAX_X2", []>;
+//def FLAT_ATOMIC_UMAX_X2 : FLAT_X2 <0x00000058, "FLAT_ATOMIC_UMAX_X2", []>;
+//def FLAT_ATOMIC_AND_X2 : FLAT_X2 <0x00000059, "FLAT_ATOMIC_AND_X2", []>;
+//def FLAT_ATOMIC_OR_X2 : FLAT_X2 <0x0000005a, "FLAT_ATOMIC_OR_X2", []>;
+//def FLAT_ATOMIC_XOR_X2 : FLAT_X2 <0x0000005b, "FLAT_ATOMIC_XOR_X2", []>;
+//def FLAT_ATOMIC_INC_X2 : FLAT_X2 <0x0000005c, "FLAT_ATOMIC_INC_X2", []>;
+//def FLAT_ATOMIC_DEC_X2 : FLAT_X2 <0x0000005d, "FLAT_ATOMIC_DEC_X2", []>;
+//def FLAT_ATOMIC_FCMPSWAP_X2 : FLAT_X2 <0x0000005e, "FLAT_ATOMIC_FCMPSWAP_X2", []>;
+//def FLAT_ATOMIC_FMIN_X2 : FLAT_X2 <0x0000005f, "FLAT_ATOMIC_FMIN_X2", []>;
+//def FLAT_ATOMIC_FMAX_X2 : FLAT_X2 <0x00000060, "FLAT_ATOMIC_FMAX_X2", []>;
+
+} // End HasFlatAddressSpace predicate
+//===----------------------------------------------------------------------===//
+// VOP1 Instructions
+//===----------------------------------------------------------------------===//
+
 //def V_NOP : VOP1_ <0x00000000, "V_NOP", []>;
 
+let isMoveImm = 1 in {
+defm V_MOV_B32 : VOP1Inst <vop1<0x1>, "V_MOV_B32", VOP_I32_I32>;
+} // End isMoveImm = 1
+
+let Uses = [EXEC] in {
+
+def V_READFIRSTLANE_B32 : VOP1 <
+  0x00000002,
+  (outs SReg_32:$vdst),
+  (ins VReg_32:$src0),
+  "V_READFIRSTLANE_B32 $vdst, $src0",
+  []
+>;
 
-let neverHasSideEffects = 1, isMoveImm = 1 in {
-defm V_MOV_B32 : VOP1_32 <0x00000001, "V_MOV_B32", []>;
-} // End neverHasSideEffects = 1, isMoveImm = 1
+}
 
-defm V_READFIRSTLANE_B32 : VOP1_32 <0x00000002, "V_READFIRSTLANE_B32", []>;
-defm V_CVT_I32_F64 : VOP1_32_64 <0x00000003, "V_CVT_I32_F64",
-  [(set i32:$dst, (fp_to_sint f64:$src0))]
+defm V_CVT_I32_F64 : VOP1Inst <vop1<0x3>, "V_CVT_I32_F64",
+  VOP_I32_F64, fp_to_sint
+>;
+defm V_CVT_F64_I32 : VOP1Inst <vop1<0x4>, "V_CVT_F64_I32",
+  VOP_F64_I32, sint_to_fp
+>;
+defm V_CVT_F32_I32 : VOP1Inst <vop1<0x5>, "V_CVT_F32_I32",
+  VOP_F32_I32, sint_to_fp
 >;
-defm V_CVT_F64_I32 : VOP1_64_32 <0x00000004, "V_CVT_F64_I32",
-  [(set f64:$dst, (sint_to_fp i32:$src0))]
+defm V_CVT_F32_U32 : VOP1Inst <vop1<0x6>, "V_CVT_F32_U32",
+  VOP_F32_I32, uint_to_fp
 >;
-defm V_CVT_F32_I32 : VOP1_32 <0x00000005, "V_CVT_F32_I32",
-  [(set f32:$dst, (sint_to_fp i32:$src0))]
+defm V_CVT_U32_F32 : VOP1Inst <vop1<0x7>, "V_CVT_U32_F32",
+  VOP_I32_F32, fp_to_uint
 >;
-defm V_CVT_F32_U32 : VOP1_32 <0x00000006, "V_CVT_F32_U32",
-  [(set f32:$dst, (uint_to_fp i32:$src0))]
+defm V_CVT_I32_F32 : VOP1Inst <vop1<0x8>, "V_CVT_I32_F32",
+  VOP_I32_F32, fp_to_sint
 >;
-defm V_CVT_U32_F32 : VOP1_32 <0x00000007, "V_CVT_U32_F32",
-  [(set i32:$dst, (fp_to_uint f32:$src0))]
+defm V_MOV_FED_B32 : VOP1Inst <vop1<0x9>, "V_MOV_FED_B32", VOP_I32_I32>;
+defm V_CVT_F16_F32 : VOP1Inst <vop1<0xa>, "V_CVT_F16_F32",
+  VOP_I32_F32, fp_to_f16
 >;
-defm V_CVT_I32_F32 : VOP1_32 <0x00000008, "V_CVT_I32_F32",
-  [(set i32:$dst, (fp_to_sint f32:$src0))]
+defm V_CVT_F32_F16 : VOP1Inst <vop1<0xb>, "V_CVT_F32_F16",
+  VOP_F32_I32, f16_to_fp
 >;
-defm V_MOV_FED_B32 : VOP1_32 <0x00000009, "V_MOV_FED_B32", []>;
-////def V_CVT_F16_F32 : VOP1_F16 <0x0000000a, "V_CVT_F16_F32", []>;
-//defm V_CVT_F32_F16 : VOP1_32 <0x0000000b, "V_CVT_F32_F16", []>;
 //defm V_CVT_RPI_I32_F32 : VOP1_32 <0x0000000c, "V_CVT_RPI_I32_F32", []>;
 //defm V_CVT_FLR_I32_F32 : VOP1_32 <0x0000000d, "V_CVT_FLR_I32_F32", []>;
 //defm V_CVT_OFF_F32_I4 : VOP1_32 <0x0000000e, "V_CVT_OFF_F32_I4", []>;
-defm V_CVT_F32_F64 : VOP1_32_64 <0x0000000f, "V_CVT_F32_F64",
-  [(set f32:$dst, (fround f64:$src0))]
->;
-defm V_CVT_F64_F32 : VOP1_64_32 <0x00000010, "V_CVT_F64_F32",
-  [(set f64:$dst, (fextend f32:$src0))]
->;
-//defm V_CVT_F32_UBYTE0 : VOP1_32 <0x00000011, "V_CVT_F32_UBYTE0", []>;
-//defm V_CVT_F32_UBYTE1 : VOP1_32 <0x00000012, "V_CVT_F32_UBYTE1", []>;
-//defm V_CVT_F32_UBYTE2 : VOP1_32 <0x00000013, "V_CVT_F32_UBYTE2", []>;
-//defm V_CVT_F32_UBYTE3 : VOP1_32 <0x00000014, "V_CVT_F32_UBYTE3", []>;
-//defm V_CVT_U32_F64 : VOP1_32 <0x00000015, "V_CVT_U32_F64", []>;
-//defm V_CVT_F64_U32 : VOP1_64 <0x00000016, "V_CVT_F64_U32", []>;
-defm V_FRACT_F32 : VOP1_32 <0x00000020, "V_FRACT_F32",
-  [(set f32:$dst, (AMDGPUfract f32:$src0))]
->;
-defm V_TRUNC_F32 : VOP1_32 <0x00000021, "V_TRUNC_F32",
-  [(set f32:$dst, (int_AMDGPU_trunc f32:$src0))]
->;
-defm V_CEIL_F32 : VOP1_32 <0x00000022, "V_CEIL_F32",
-  [(set f32:$dst, (fceil f32:$src0))]
->;
-defm V_RNDNE_F32 : VOP1_32 <0x00000023, "V_RNDNE_F32",
-  [(set f32:$dst, (frint f32:$src0))]
->;
-defm V_FLOOR_F32 : VOP1_32 <0x00000024, "V_FLOOR_F32",
-  [(set f32:$dst, (ffloor f32:$src0))]
->;
-defm V_EXP_F32 : VOP1_32 <0x00000025, "V_EXP_F32",
-  [(set f32:$dst, (fexp2 f32:$src0))]
->;
-defm V_LOG_CLAMP_F32 : VOP1_32 <0x00000026, "V_LOG_CLAMP_F32", []>;
-defm V_LOG_F32 : VOP1_32 <0x00000027, "V_LOG_F32",
-  [(set f32:$dst, (flog2 f32:$src0))]
->;
-defm V_RCP_CLAMP_F32 : VOP1_32 <0x00000028, "V_RCP_CLAMP_F32", []>;
-defm V_RCP_LEGACY_F32 : VOP1_32 <0x00000029, "V_RCP_LEGACY_F32", []>;
-defm V_RCP_F32 : VOP1_32 <0x0000002a, "V_RCP_F32",
-  [(set f32:$dst, (fdiv FP_ONE, f32:$src0))]
->;
-defm V_RCP_IFLAG_F32 : VOP1_32 <0x0000002b, "V_RCP_IFLAG_F32", []>;
-defm V_RSQ_CLAMP_F32 : VOP1_32 <0x0000002c, "V_RSQ_CLAMP_F32", []>;
-defm V_RSQ_LEGACY_F32 : VOP1_32 <
-  0x0000002d, "V_RSQ_LEGACY_F32",
-  [(set f32:$dst, (int_AMDGPU_rsq f32:$src0))]
->;
-defm V_RSQ_F32 : VOP1_32 <0x0000002e, "V_RSQ_F32", []>;
-defm V_RCP_F64 : VOP1_64 <0x0000002f, "V_RCP_F64",
-  [(set f64:$dst, (fdiv FP_ONE, f64:$src0))]
->;
-defm V_RCP_CLAMP_F64 : VOP1_64 <0x00000030, "V_RCP_CLAMP_F64", []>;
-defm V_RSQ_F64 : VOP1_64 <0x00000031, "V_RSQ_F64", []>;
-defm V_RSQ_CLAMP_F64 : VOP1_64 <0x00000032, "V_RSQ_CLAMP_F64", []>;
-defm V_SQRT_F32 : VOP1_32 <0x00000033, "V_SQRT_F32",
-  [(set f32:$dst, (fsqrt f32:$src0))]
->;
-defm V_SQRT_F64 : VOP1_64 <0x00000034, "V_SQRT_F64",
-  [(set f64:$dst, (fsqrt f64:$src0))]
->;
-defm V_SIN_F32 : VOP1_32 <0x00000035, "V_SIN_F32", []>;
-defm V_COS_F32 : VOP1_32 <0x00000036, "V_COS_F32", []>;
-defm V_NOT_B32 : VOP1_32 <0x00000037, "V_NOT_B32", []>;
-defm V_BFREV_B32 : VOP1_32 <0x00000038, "V_BFREV_B32", []>;
-defm V_FFBH_U32 : VOP1_32 <0x00000039, "V_FFBH_U32", []>;
-defm V_FFBL_B32 : VOP1_32 <0x0000003a, "V_FFBL_B32", []>;
-defm V_FFBH_I32 : VOP1_32 <0x0000003b, "V_FFBH_I32", []>;
-//defm V_FREXP_EXP_I32_F64 : VOP1_32 <0x0000003c, "V_FREXP_EXP_I32_F64", []>;
-defm V_FREXP_MANT_F64 : VOP1_64 <0x0000003d, "V_FREXP_MANT_F64", []>;
-defm V_FRACT_F64 : VOP1_64 <0x0000003e, "V_FRACT_F64", []>;
-//defm V_FREXP_EXP_I32_F32 : VOP1_32 <0x0000003f, "V_FREXP_EXP_I32_F32", []>;
-defm V_FREXP_MANT_F32 : VOP1_32 <0x00000040, "V_FREXP_MANT_F32", []>;
+defm V_CVT_F32_F64 : VOP1Inst <vop1<0xf>, "V_CVT_F32_F64",
+  VOP_F32_F64, fround
+>;
+defm V_CVT_F64_F32 : VOP1Inst <vop1<0x10>, "V_CVT_F64_F32",
+  VOP_F64_F32, fextend
+>;
+defm V_CVT_F32_UBYTE0 : VOP1Inst <vop1<0x11>, "V_CVT_F32_UBYTE0",
+  VOP_F32_I32, AMDGPUcvt_f32_ubyte0
+>;
+defm V_CVT_F32_UBYTE1 : VOP1Inst <vop1<0x12>, "V_CVT_F32_UBYTE1",
+  VOP_F32_I32, AMDGPUcvt_f32_ubyte1
+>;
+defm V_CVT_F32_UBYTE2 : VOP1Inst <vop1<0x13>, "V_CVT_F32_UBYTE2",
+  VOP_F32_I32, AMDGPUcvt_f32_ubyte2
+>;
+defm V_CVT_F32_UBYTE3 : VOP1Inst <vop1<0x14>, "V_CVT_F32_UBYTE3",
+  VOP_F32_I32, AMDGPUcvt_f32_ubyte3
+>;
+defm V_CVT_U32_F64 : VOP1Inst <vop1<0x15>, "V_CVT_U32_F64",
+  VOP_I32_F64, fp_to_uint
+>;
+defm V_CVT_F64_U32 : VOP1Inst <vop1<0x16>, "V_CVT_F64_U32",
+  VOP_F64_I32, uint_to_fp
+>;
+
+defm V_FRACT_F32 : VOP1Inst <vop1<0x20>, "V_FRACT_F32",
+  VOP_F32_F32, AMDGPUfract
+>;
+defm V_TRUNC_F32 : VOP1Inst <vop1<0x21>, "V_TRUNC_F32",
+  VOP_F32_F32, ftrunc
+>;
+defm V_CEIL_F32 : VOP1Inst <vop1<0x22>, "V_CEIL_F32",
+  VOP_F32_F32, fceil
+>;
+defm V_RNDNE_F32 : VOP1Inst <vop1<0x23>, "V_RNDNE_F32",
+  VOP_F32_F32, frint
+>;
+defm V_FLOOR_F32 : VOP1Inst <vop1<0x24>, "V_FLOOR_F32",
+  VOP_F32_F32, ffloor
+>;
+defm V_EXP_F32 : VOP1Inst <vop1<0x25>, "V_EXP_F32",
+  VOP_F32_F32, fexp2
+>;
+defm V_LOG_CLAMP_F32 : VOP1Inst <vop1<0x26>, "V_LOG_CLAMP_F32", VOP_F32_F32>;
+defm V_LOG_F32 : VOP1Inst <vop1<0x27>, "V_LOG_F32",
+  VOP_F32_F32, flog2
+>;
+
+defm V_RCP_CLAMP_F32 : VOP1Inst <vop1<0x28>, "V_RCP_CLAMP_F32", VOP_F32_F32>;
+defm V_RCP_LEGACY_F32 : VOP1Inst <vop1<0x29>, "V_RCP_LEGACY_F32", VOP_F32_F32>;
+defm V_RCP_F32 : VOP1Inst <vop1<0x2a>, "V_RCP_F32",
+  VOP_F32_F32, AMDGPUrcp
+>;
+defm V_RCP_IFLAG_F32 : VOP1Inst <vop1<0x2b>, "V_RCP_IFLAG_F32", VOP_F32_F32>;
+defm V_RSQ_CLAMP_F32 : VOP1Inst <vop1<0x2c>, "V_RSQ_CLAMP_F32",
+  VOP_F32_F32, AMDGPUrsq_clamped
+>;
+defm V_RSQ_LEGACY_F32 : VOP1Inst <vop1<0x2d>, "V_RSQ_LEGACY_F32",
+  VOP_F32_F32, AMDGPUrsq_legacy
+>;
+defm V_RSQ_F32 : VOP1Inst <vop1<0x2e>, "V_RSQ_F32",
+  VOP_F32_F32, AMDGPUrsq
+>;
+defm V_RCP_F64 : VOP1Inst <vop1<0x2f>, "V_RCP_F64",
+  VOP_F64_F64, AMDGPUrcp
+>;
+defm V_RCP_CLAMP_F64 : VOP1Inst <vop1<0x30>, "V_RCP_CLAMP_F64", VOP_F64_F64>;
+defm V_RSQ_F64 : VOP1Inst <vop1<0x31>, "V_RSQ_F64",
+  VOP_F64_F64, AMDGPUrsq
+>;
+defm V_RSQ_CLAMP_F64 : VOP1Inst <vop1<0x32>, "V_RSQ_CLAMP_F64",
+  VOP_F64_F64, AMDGPUrsq_clamped
+>;
+defm V_SQRT_F32 : VOP1Inst <vop1<0x33>, "V_SQRT_F32",
+  VOP_F32_F32, fsqrt
+>;
+defm V_SQRT_F64 : VOP1Inst <vop1<0x34>, "V_SQRT_F64",
+  VOP_F64_F64, fsqrt
+>;
+defm V_SIN_F32 : VOP1Inst <vop1<0x35>, "V_SIN_F32",
+  VOP_F32_F32, AMDGPUsin
+>;
+defm V_COS_F32 : VOP1Inst <vop1<0x36>, "V_COS_F32",
+  VOP_F32_F32, AMDGPUcos
+>;
+defm V_NOT_B32 : VOP1Inst <vop1<0x37>, "V_NOT_B32", VOP_I32_I32>;
+defm V_BFREV_B32 : VOP1Inst <vop1<0x38>, "V_BFREV_B32", VOP_I32_I32>;
+defm V_FFBH_U32 : VOP1Inst <vop1<0x39>, "V_FFBH_U32", VOP_I32_I32>;
+defm V_FFBL_B32 : VOP1Inst <vop1<0x3a>, "V_FFBL_B32", VOP_I32_I32>;
+defm V_FFBH_I32 : VOP1Inst <vop1<0x3b>, "V_FFBH_I32", VOP_I32_I32>;
+//defm V_FREXP_EXP_I32_F64 : VOPInst <0x0000003c, "V_FREXP_EXP_I32_F64", VOP_I32_F32>;
+defm V_FREXP_MANT_F64 : VOP1Inst <vop1<0x3d>, "V_FREXP_MANT_F64", VOP_F64_F64>;
+defm V_FRACT_F64 : VOP1Inst <vop1<0x3e>, "V_FRACT_F64", VOP_F64_F64>;
+//defm V_FREXP_EXP_I32_F32 : VOPInst <0x0000003f, "V_FREXP_EXP_I32_F32", VOP_I32_F32>;
+defm V_FREXP_MANT_F32 : VOP1Inst <vop1<0x40>, "V_FREXP_MANT_F32", VOP_F32_F32>;
 //def V_CLREXCP : VOP1_ <0x00000041, "V_CLREXCP", []>;
-defm V_MOVRELD_B32 : VOP1_32 <0x00000042, "V_MOVRELD_B32", []>;
-defm V_MOVRELS_B32 : VOP1_32 <0x00000043, "V_MOVRELS_B32", []>;
-defm V_MOVRELSD_B32 : VOP1_32 <0x00000044, "V_MOVRELSD_B32", []>;
+defm V_MOVRELD_B32 : VOP1Inst <vop1<0x42>, "V_MOVRELD_B32", VOP_I32_I32>;
+defm V_MOVRELS_B32 : VOP1Inst <vop1<0x43>, "V_MOVRELS_B32", VOP_I32_I32>;
+defm V_MOVRELSD_B32 : VOP1Inst <vop1<0x44>, "V_MOVRELSD_B32", VOP_I32_I32>;
+
+
+//===----------------------------------------------------------------------===//
+// VINTRP Instructions
+//===----------------------------------------------------------------------===//
 
 def V_INTERP_P1_F32 : VINTRP <
   0x00000000,
@@ -734,103 +1311,23 @@ def V_INTERP_P2_F32 : VINTRP <
   "V_INTERP_P2_F32 $dst, [$src0], $j, $attr_chan, $attr, [$m0]",
   []> {
 
-  let Constraints = "$src0 = $dst";
-  let DisableEncoding = "$src0,$m0";
-
-}
-
-def V_INTERP_MOV_F32 : VINTRP <
-  0x00000002,
-  (outs VReg_32:$dst),
-  (ins InterpSlot:$src0, i32imm:$attr_chan, i32imm:$attr, M0Reg:$m0),
-  "V_INTERP_MOV_F32 $dst, $src0, $attr_chan, $attr, [$m0]",
-  []> {
-  let DisableEncoding = "$m0";
-}
-
-//def S_NOP : SOPP_ <0x00000000, "S_NOP", []>;
-
-let isTerminator = 1 in {
-
-def S_ENDPGM : SOPP <0x00000001, (ins), "S_ENDPGM",
-  [(IL_retflag)]> {
-  let SIMM16 = 0;
-  let isBarrier = 1;
-  let hasCtrlDep = 1;
-}
-
-let isBranch = 1 in {
-def S_BRANCH : SOPP <
-  0x00000002, (ins brtarget:$target), "S_BRANCH $target",
-  [(br bb:$target)]> {
-  let isBarrier = 1;
-}
-
-let DisableEncoding = "$scc" in {
-def S_CBRANCH_SCC0 : SOPP <
-  0x00000004, (ins brtarget:$target, SCCReg:$scc),
-  "S_CBRANCH_SCC0 $target", []
->;
-def S_CBRANCH_SCC1 : SOPP <
-  0x00000005, (ins brtarget:$target, SCCReg:$scc),
-  "S_CBRANCH_SCC1 $target",
-  []
->;
-} // End DisableEncoding = "$scc"
-
-def S_CBRANCH_VCCZ : SOPP <
-  0x00000006, (ins brtarget:$target, VCCReg:$vcc),
-  "S_CBRANCH_VCCZ $target",
-  []
->;
-def S_CBRANCH_VCCNZ : SOPP <
-  0x00000007, (ins brtarget:$target, VCCReg:$vcc),
-  "S_CBRANCH_VCCNZ $target",
-  []
->;
-
-let DisableEncoding = "$exec" in {
-def S_CBRANCH_EXECZ : SOPP <
-  0x00000008, (ins brtarget:$target, EXECReg:$exec),
-  "S_CBRANCH_EXECZ $target",
-  []
->;
-def S_CBRANCH_EXECNZ : SOPP <
-  0x00000009, (ins brtarget:$target, EXECReg:$exec),
-  "S_CBRANCH_EXECNZ $target",
-  []
->;
-} // End DisableEncoding = "$exec"
-
+  let Constraints = "$src0 = $dst";
+  let DisableEncoding = "$src0,$m0";
 
-} // End isBranch = 1
-} // End isTerminator = 1
+}
 
-let hasSideEffects = 1 in {
-def S_BARRIER : SOPP <0x0000000a, (ins), "S_BARRIER",
-  [(int_AMDGPU_barrier_local)]
-> {
-  let SIMM16 = 0;
-  let isBarrier = 1;
-  let hasCtrlDep = 1;
-  let mayLoad = 1;
-  let mayStore = 1;
+def V_INTERP_MOV_F32 : VINTRP <
+  0x00000002,
+  (outs VReg_32:$dst),
+  (ins InterpSlot:$src0, i32imm:$attr_chan, i32imm:$attr, M0Reg:$m0),
+  "V_INTERP_MOV_F32 $dst, $src0, $attr_chan, $attr, [$m0]",
+  []> {
+  let DisableEncoding = "$m0";
 }
 
-def S_WAITCNT : SOPP <0x0000000c, (ins WAIT_FLAG:$simm16), "S_WAITCNT $simm16",
-  []
->;
-} // End hasSideEffects
-//def S_SETHALT : SOPP_ <0x0000000d, "S_SETHALT", []>;
-//def S_SLEEP : SOPP_ <0x0000000e, "S_SLEEP", []>;
-//def S_SETPRIO : SOPP_ <0x0000000f, "S_SETPRIO", []>;
-//def S_SENDMSG : SOPP_ <0x00000010, "S_SENDMSG", []>;
-//def S_SENDMSGHALT : SOPP_ <0x00000011, "S_SENDMSGHALT", []>;
-//def S_TRAP : SOPP_ <0x00000012, "S_TRAP", []>;
-//def S_ICACHE_INV : SOPP_ <0x00000013, "S_ICACHE_INV", []>;
-//def S_INCPERFLEVEL : SOPP_ <0x00000014, "S_INCPERFLEVEL", []>;
-//def S_DECPERFLEVEL : SOPP_ <0x00000015, "S_DECPERFLEVEL", []>;
-//def S_TTRACEDATA : SOPP_ <0x00000016, "S_TTRACEDATA", []>;
+//===----------------------------------------------------------------------===//
+// VOP2 Instructions
+//===----------------------------------------------------------------------===//
 
 def V_CNDMASK_B32_e32 : VOP2 <0x00000000, (outs VReg_32:$dst),
   (ins VSrc_32:$src0, VReg_32:$src1, VCCReg:$vcc),
@@ -841,207 +1338,220 @@ def V_CNDMASK_B32_e32 : VOP2 <0x00000000, (outs VReg_32:$dst),
 }
 
 def V_CNDMASK_B32_e64 : VOP3 <0x00000100, (outs VReg_32:$dst),
-  (ins VSrc_32:$src0, VSrc_32:$src1, SSrc_64:$src2,
-   InstFlag:$abs, InstFlag:$clamp, InstFlag:$omod, InstFlag:$neg),
-  "V_CNDMASK_B32_e64 $dst, $src0, $src1, $src2, $abs, $clamp, $omod, $neg",
+  (ins VSrc_32:$src0, VSrc_32:$src1, SSrc_64:$src2),
+  "V_CNDMASK_B32_e64 $dst, $src0, $src1, $src2",
   [(set i32:$dst, (select i1:$src2, i32:$src1, i32:$src0))]
->;
-
-//f32 pattern for V_CNDMASK_B32_e64
-def : Pat <
-  (f32 (select i1:$src2, f32:$src1, f32:$src0)),
-  (V_CNDMASK_B32_e64 $src0, $src1, $src2)
->;
+> {
+  let src0_modifiers = 0;
+  let src1_modifiers = 0;
+  let src2_modifiers = 0;
+}
 
-def : Pat <
-  (i32 (trunc i64:$val)),
-  (EXTRACT_SUBREG $val, sub0)
+def V_READLANE_B32 : VOP2 <
+  0x00000001,
+  (outs SReg_32:$vdst),
+  (ins VReg_32:$src0, SSrc_32:$vsrc1),
+  "V_READLANE_B32 $vdst, $src0, $vsrc1",
+  []
 >;
 
-//use two V_CNDMASK_B32_e64 instructions for f64
-def : Pat <
-  (f64 (select i1:$src2, f64:$src1, f64:$src0)),
-  (INSERT_SUBREG (INSERT_SUBREG (f64 (IMPLICIT_DEF)),
-  (V_CNDMASK_B32_e64 (EXTRACT_SUBREG $src0, sub0),
-                     (EXTRACT_SUBREG $src1, sub0),
-                     $src2), sub0),
-  (V_CNDMASK_B32_e64 (EXTRACT_SUBREG $src0, sub1),
-                     (EXTRACT_SUBREG $src1, sub1),
-                     $src2), sub1)
+def V_WRITELANE_B32 : VOP2 <
+  0x00000002,
+  (outs VReg_32:$vdst),
+  (ins SReg_32:$src0, SSrc_32:$vsrc1),
+  "V_WRITELANE_B32 $vdst, $src0, $vsrc1",
+  []
 >;
 
-defm V_READLANE_B32 : VOP2_32 <0x00000001, "V_READLANE_B32", []>;
-defm V_WRITELANE_B32 : VOP2_32 <0x00000002, "V_WRITELANE_B32", []>;
-
 let isCommutable = 1 in {
-defm V_ADD_F32 : VOP2_32 <0x00000003, "V_ADD_F32",
-  [(set f32:$dst, (fadd f32:$src0, f32:$src1))]
+defm V_ADD_F32 : VOP2Inst <vop2<0x3>, "V_ADD_F32",
+  VOP_F32_F32_F32, fadd
 >;
 
-defm V_SUB_F32 : VOP2_32 <0x00000004, "V_SUB_F32",
-  [(set f32:$dst, (fsub f32:$src0, f32:$src1))]
+defm V_SUB_F32 : VOP2Inst <vop2<0x4>, "V_SUB_F32", VOP_F32_F32_F32, fsub>;
+defm V_SUBREV_F32 : VOP2Inst <vop2<0x5>, "V_SUBREV_F32",
+  VOP_F32_F32_F32, null_frag, "V_SUB_F32"
 >;
-defm V_SUBREV_F32 : VOP2_32 <0x00000005, "V_SUBREV_F32", [], "V_SUB_F32">;
 } // End isCommutable = 1
 
-defm V_MAC_LEGACY_F32 : VOP2_32 <0x00000006, "V_MAC_LEGACY_F32", []>;
+defm V_MAC_LEGACY_F32 : VOP2Inst <vop2<0x6>, "V_MAC_LEGACY_F32",
+  VOP_F32_F32_F32
+>;
 
 let isCommutable = 1 in {
 
-defm V_MUL_LEGACY_F32 : VOP2_32 <
-  0x00000007, "V_MUL_LEGACY_F32",
-  [(set f32:$dst, (int_AMDGPU_mul f32:$src0, f32:$src1))]
+defm V_MUL_LEGACY_F32 : VOP2Inst <vop2<0x7>, "V_MUL_LEGACY_F32",
+  VOP_F32_F32_F32, int_AMDGPU_mul
 >;
 
-defm V_MUL_F32 : VOP2_32 <0x00000008, "V_MUL_F32",
-  [(set f32:$dst, (fmul f32:$src0, f32:$src1))]
+defm V_MUL_F32 : VOP2Inst <vop2<0x8>, "V_MUL_F32",
+  VOP_F32_F32_F32, fmul
 >;
 
 
-defm V_MUL_I32_I24 : VOP2_32 <0x00000009, "V_MUL_I32_I24",
-  [(set i32:$dst, (mul I24:$src0, I24:$src1))]
+defm V_MUL_I32_I24 : VOP2Inst <vop2<0x9>, "V_MUL_I32_I24",
+  VOP_I32_I32_I32, AMDGPUmul_i24
 >;
 //defm V_MUL_HI_I32_I24 : VOP2_32 <0x0000000a, "V_MUL_HI_I32_I24", []>;
-defm V_MUL_U32_U24 : VOP2_32 <0x0000000b, "V_MUL_U32_U24",
-  [(set i32:$dst, (mul U24:$src0, U24:$src1))]
+defm V_MUL_U32_U24 : VOP2Inst <vop2<0xb>, "V_MUL_U32_U24",
+  VOP_I32_I32_I32, AMDGPUmul_u24
 >;
 //defm V_MUL_HI_U32_U24 : VOP2_32 <0x0000000c, "V_MUL_HI_U32_U24", []>;
 
 
-defm V_MIN_LEGACY_F32 : VOP2_32 <0x0000000d, "V_MIN_LEGACY_F32",
-  [(set f32:$dst, (AMDGPUfmin f32:$src0, f32:$src1))]
+defm V_MIN_LEGACY_F32 : VOP2Inst <vop2<0xd>, "V_MIN_LEGACY_F32",
+  VOP_F32_F32_F32, AMDGPUfmin
 >;
 
-defm V_MAX_LEGACY_F32 : VOP2_32 <0x0000000e, "V_MAX_LEGACY_F32",
-  [(set f32:$dst, (AMDGPUfmax f32:$src0, f32:$src1))]
+defm V_MAX_LEGACY_F32 : VOP2Inst <vop2<0xe>, "V_MAX_LEGACY_F32",
+  VOP_F32_F32_F32, AMDGPUfmax
 >;
 
-defm V_MIN_F32 : VOP2_32 <0x0000000f, "V_MIN_F32", []>;
-defm V_MAX_F32 : VOP2_32 <0x00000010, "V_MAX_F32", []>;
-defm V_MIN_I32 : VOP2_32 <0x00000011, "V_MIN_I32",
-  [(set i32:$dst, (AMDGPUsmin i32:$src0, i32:$src1))]
->;
-defm V_MAX_I32 : VOP2_32 <0x00000012, "V_MAX_I32",
-  [(set i32:$dst, (AMDGPUsmax i32:$src0, i32:$src1))]
+defm V_MIN_F32 : VOP2Inst <vop2<0xf>, "V_MIN_F32", VOP_F32_F32_F32, fminnum>;
+defm V_MAX_F32 : VOP2Inst <vop2<0x10>, "V_MAX_F32", VOP_F32_F32_F32, fmaxnum>;
+defm V_MIN_I32 : VOP2Inst <vop2<0x11>, "V_MIN_I32", VOP_I32_I32_I32, AMDGPUsmin>;
+defm V_MAX_I32 : VOP2Inst <vop2<0x12>, "V_MAX_I32", VOP_I32_I32_I32, AMDGPUsmax>;
+defm V_MIN_U32 : VOP2Inst <vop2<0x13>, "V_MIN_U32", VOP_I32_I32_I32, AMDGPUumin>;
+defm V_MAX_U32 : VOP2Inst <vop2<0x14>, "V_MAX_U32", VOP_I32_I32_I32, AMDGPUumax>;
+
+defm V_LSHR_B32 : VOP2Inst <vop2<0x15>, "V_LSHR_B32", VOP_I32_I32_I32, srl>;
+
+defm V_LSHRREV_B32 : VOP2Inst <
+  vop2<0x16>, "V_LSHRREV_B32", VOP_I32_I32_I32, null_frag, "V_LSHR_B32"
 >;
-defm V_MIN_U32 : VOP2_32 <0x00000013, "V_MIN_U32",
-  [(set i32:$dst, (AMDGPUumin i32:$src0, i32:$src1))]
+
+defm V_ASHR_I32 : VOP2Inst <vop2<0x17>, "V_ASHR_I32",
+  VOP_I32_I32_I32, sra
 >;
-defm V_MAX_U32 : VOP2_32 <0x00000014, "V_MAX_U32",
-  [(set i32:$dst, (AMDGPUumax i32:$src0, i32:$src1))]
+defm V_ASHRREV_I32 : VOP2Inst <
+  vop2<0x18>, "V_ASHRREV_I32", VOP_I32_I32_I32, null_frag, "V_ASHR_I32"
 >;
 
-defm V_LSHR_B32 : VOP2_32 <0x00000015, "V_LSHR_B32",
-  [(set i32:$dst, (srl i32:$src0, i32:$src1))]
->;
-defm V_LSHRREV_B32 : VOP2_32 <0x00000016, "V_LSHRREV_B32", [], "V_LSHR_B32">;
+let hasPostISelHook = 1 in {
 
-defm V_ASHR_I32 : VOP2_32 <0x00000017, "V_ASHR_I32",
-  [(set i32:$dst, (sra i32:$src0, i32:$src1))]
->;
-defm V_ASHRREV_I32 : VOP2_32 <0x00000018, "V_ASHRREV_I32", [], "V_ASHR_I32">;
+defm V_LSHL_B32 : VOP2Inst <vop2<0x19>, "V_LSHL_B32", VOP_I32_I32_I32, shl>;
 
-defm V_LSHL_B32 : VOP2_32 <0x00000019, "V_LSHL_B32",
-  [(set i32:$dst, (shl i32:$src0, i32:$src1))]
+}
+defm V_LSHLREV_B32 : VOP2Inst <
+  vop2<0x1a>, "V_LSHLREV_B32", VOP_I32_I32_I32, null_frag, "V_LSHL_B32"
 >;
-defm V_LSHLREV_B32 : VOP2_32 <0x0000001a, "V_LSHLREV_B32", [], "V_LSHL_B32">;
 
-defm V_AND_B32 : VOP2_32 <0x0000001b, "V_AND_B32",
-  [(set i32:$dst, (and i32:$src0, i32:$src1))]
+defm V_AND_B32 : VOP2Inst <vop2<0x1b>, "V_AND_B32",
+  VOP_I32_I32_I32, and>;
+defm V_OR_B32 : VOP2Inst <vop2<0x1c>, "V_OR_B32",
+  VOP_I32_I32_I32, or
 >;
-defm V_OR_B32 : VOP2_32 <0x0000001c, "V_OR_B32",
-  [(set i32:$dst, (or i32:$src0, i32:$src1))]
->;
-defm V_XOR_B32 : VOP2_32 <0x0000001d, "V_XOR_B32",
-  [(set i32:$dst, (xor i32:$src0, i32:$src1))]
+defm V_XOR_B32 : VOP2Inst <vop2<0x1d>, "V_XOR_B32",
+  VOP_I32_I32_I32, xor
 >;
 
 } // End isCommutable = 1
 
-defm V_BFM_B32 : VOP2_32 <0x0000001e, "V_BFM_B32", []>;
-defm V_MAC_F32 : VOP2_32 <0x0000001f, "V_MAC_F32", []>;
-defm V_MADMK_F32 : VOP2_32 <0x00000020, "V_MADMK_F32", []>;
-defm V_MADAK_F32 : VOP2_32 <0x00000021, "V_MADAK_F32", []>;
-//defm V_BCNT_U32_B32 : VOP2_32 <0x00000022, "V_BCNT_U32_B32", []>;
-defm V_MBCNT_LO_U32_B32 : VOP2_32 <0x00000023, "V_MBCNT_LO_U32_B32", []>;
-defm V_MBCNT_HI_U32_B32 : VOP2_32 <0x00000024, "V_MBCNT_HI_U32_B32", []>;
+defm V_BFM_B32 : VOP2Inst <vop2<0x1e>, "V_BFM_B32",
+  VOP_I32_I32_I32, AMDGPUbfm>;
+defm V_MAC_F32 : VOP2Inst <vop2<0x1f>, "V_MAC_F32", VOP_F32_F32_F32>;
+defm V_MADMK_F32 : VOP2Inst <vop2<0x20>, "V_MADMK_F32", VOP_F32_F32_F32>;
+defm V_MADAK_F32 : VOP2Inst <vop2<0x21>, "V_MADAK_F32", VOP_F32_F32_F32>;
+defm V_BCNT_U32_B32 : VOP2Inst <vop2<0x22>, "V_BCNT_U32_B32", VOP_I32_I32_I32>;
+defm V_MBCNT_LO_U32_B32 : VOP2Inst <vop2<0x23>, "V_MBCNT_LO_U32_B32",
+  VOP_I32_I32_I32
+>;
+defm V_MBCNT_HI_U32_B32 : VOP2Inst <vop2<0x24>, "V_MBCNT_HI_U32_B32",
+  VOP_I32_I32_I32
+>;
 
 let isCommutable = 1, Defs = [VCC] in { // Carry-out goes to VCC
-defm V_ADD_I32 : VOP2b_32 <0x00000025, "V_ADD_I32",
-  [(set i32:$dst, (add (i32 VSrc_32:$src0), (i32 VReg_32:$src1)))]
+// No patterns so that the scalar instructions are always selected.
+// The scalar versions will be replaced with vector when needed later.
+defm V_ADD_I32 : VOP2bInst <vop2<0x25>, "V_ADD_I32",
+  VOP_I32_I32_I32, add
+>;
+defm V_SUB_I32 : VOP2bInst <vop2<0x26>, "V_SUB_I32",
+  VOP_I32_I32_I32, sub
+>;
+defm V_SUBREV_I32 : VOP2bInst <vop2<0x27>, "V_SUBREV_I32",
+  VOP_I32_I32_I32, null_frag, "V_SUB_I32"
 >;
 
-defm V_SUB_I32 : VOP2b_32 <0x00000026, "V_SUB_I32",
-  [(set i32:$dst, (sub i32:$src0, i32:$src1))]
+let Uses = [VCC] in { // Carry-in comes from VCC
+defm V_ADDC_U32 : VOP2bInst <vop2<0x28>, "V_ADDC_U32",
+  VOP_I32_I32_I32_VCC, adde
+>;
+defm V_SUBB_U32 : VOP2bInst <vop2<0x29>, "V_SUBB_U32",
+  VOP_I32_I32_I32_VCC, sube
+>;
+defm V_SUBBREV_U32 : VOP2bInst <vop2<0x2a>, "V_SUBBREV_U32",
+  VOP_I32_I32_I32_VCC, null_frag, "V_SUBB_U32"
 >;
-defm V_SUBREV_I32 : VOP2b_32 <0x00000027, "V_SUBREV_I32", [], "V_SUB_I32">;
 
-let Uses = [VCC] in { // Carry-out comes from VCC
-defm V_ADDC_U32 : VOP2b_32 <0x00000028, "V_ADDC_U32", []>;
-defm V_SUBB_U32 : VOP2b_32 <0x00000029, "V_SUBB_U32", []>;
-defm V_SUBBREV_U32 : VOP2b_32 <0x0000002a, "V_SUBBREV_U32", [], "V_SUBB_U32">;
 } // End Uses = [VCC]
 } // End isCommutable = 1, Defs = [VCC]
 
-defm V_LDEXP_F32 : VOP2_32 <0x0000002b, "V_LDEXP_F32", []>;
+defm V_LDEXP_F32 : VOP2Inst <vop2<0x2b>, "V_LDEXP_F32",
+  VOP_F32_F32_I32, AMDGPUldexp
+>;
 ////def V_CVT_PKACCUM_U8_F32 : VOP2_U8 <0x0000002c, "V_CVT_PKACCUM_U8_F32", []>;
 ////def V_CVT_PKNORM_I16_F32 : VOP2_I16 <0x0000002d, "V_CVT_PKNORM_I16_F32", []>;
 ////def V_CVT_PKNORM_U16_F32 : VOP2_U16 <0x0000002e, "V_CVT_PKNORM_U16_F32", []>;
-defm V_CVT_PKRTZ_F16_F32 : VOP2_32 <0x0000002f, "V_CVT_PKRTZ_F16_F32",
- [(set i32:$dst, (int_SI_packf16 f32:$src0, f32:$src1))]
+defm V_CVT_PKRTZ_F16_F32 : VOP2Inst <vop2<0x2f>, "V_CVT_PKRTZ_F16_F32",
+ VOP_I32_F32_F32, int_SI_packf16
 >;
 ////def V_CVT_PK_U16_U32 : VOP2_U16 <0x00000030, "V_CVT_PK_U16_U32", []>;
 ////def V_CVT_PK_I16_I32 : VOP2_I16 <0x00000031, "V_CVT_PK_I16_I32", []>;
-def S_CMP_EQ_I32 : SOPC_32 <0x00000000, "S_CMP_EQ_I32", []>;
-def S_CMP_LG_I32 : SOPC_32 <0x00000001, "S_CMP_LG_I32", []>;
-def S_CMP_GT_I32 : SOPC_32 <0x00000002, "S_CMP_GT_I32", []>;
-def S_CMP_GE_I32 : SOPC_32 <0x00000003, "S_CMP_GE_I32", []>;
-def S_CMP_LT_I32 : SOPC_32 <0x00000004, "S_CMP_LT_I32", []>;
-def S_CMP_LE_I32 : SOPC_32 <0x00000005, "S_CMP_LE_I32", []>;
-def S_CMP_EQ_U32 : SOPC_32 <0x00000006, "S_CMP_EQ_U32", []>;
-def S_CMP_LG_U32 : SOPC_32 <0x00000007, "S_CMP_LG_U32", []>;
-def S_CMP_GT_U32 : SOPC_32 <0x00000008, "S_CMP_GT_U32", []>;
-def S_CMP_GE_U32 : SOPC_32 <0x00000009, "S_CMP_GE_U32", []>;
-def S_CMP_LT_U32 : SOPC_32 <0x0000000a, "S_CMP_LT_U32", []>;
-def S_CMP_LE_U32 : SOPC_32 <0x0000000b, "S_CMP_LE_U32", []>;
-////def S_BITCMP0_B32 : SOPC_BITCMP0 <0x0000000c, "S_BITCMP0_B32", []>;
-////def S_BITCMP1_B32 : SOPC_BITCMP1 <0x0000000d, "S_BITCMP1_B32", []>;
-////def S_BITCMP0_B64 : SOPC_BITCMP0 <0x0000000e, "S_BITCMP0_B64", []>;
-////def S_BITCMP1_B64 : SOPC_BITCMP1 <0x0000000f, "S_BITCMP1_B64", []>;
-//def S_SETVSKIP : SOPC_ <0x00000010, "S_SETVSKIP", []>;
 
-let neverHasSideEffects = 1 in {
+//===----------------------------------------------------------------------===//
+// VOP3 Instructions
+//===----------------------------------------------------------------------===//
 
-def V_MAD_LEGACY_F32 : VOP3_32 <0x00000140, "V_MAD_LEGACY_F32", []>;
-def V_MAD_F32 : VOP3_32 <0x00000141, "V_MAD_F32", []>;
-def V_MAD_I32_I24 : VOP3_32 <0x00000142, "V_MAD_I32_I24",
-  [(set i32:$dst, (add (mul I24:$src0, I24:$src1), i32:$src2))]
+defm V_MAD_LEGACY_F32 : VOP3Inst <vop3<0x140>, "V_MAD_LEGACY_F32",
+  VOP_F32_F32_F32_F32
+>;
+defm V_MAD_F32 : VOP3Inst <vop3<0x141>, "V_MAD_F32",
+  VOP_F32_F32_F32_F32, fmad
 >;
-def V_MAD_U32_U24 : VOP3_32 <0x00000143, "V_MAD_U32_U24",
-  [(set i32:$dst, (add (mul U24:$src0, U24:$src1), i32:$src2))]
+defm V_MAD_I32_I24 : VOP3Inst <vop3<0x142>, "V_MAD_I32_I24",
+  VOP_I32_I32_I32_I32, AMDGPUmad_i24
+>;
+defm V_MAD_U32_U24 : VOP3Inst <vop3<0x143>, "V_MAD_U32_U24",
+  VOP_I32_I32_I32_I32, AMDGPUmad_u24
 >;
 
-} // End neverHasSideEffects
-def V_CUBEID_F32 : VOP3_32 <0x00000144, "V_CUBEID_F32", []>;
-def V_CUBESC_F32 : VOP3_32 <0x00000145, "V_CUBESC_F32", []>;
-def V_CUBETC_F32 : VOP3_32 <0x00000146, "V_CUBETC_F32", []>;
-def V_CUBEMA_F32 : VOP3_32 <0x00000147, "V_CUBEMA_F32", []>;
-def V_BFE_U32 : VOP3_32 <0x00000148, "V_BFE_U32", []>;
-def V_BFE_I32 : VOP3_32 <0x00000149, "V_BFE_I32", []>;
-def V_BFI_B32 : VOP3_32 <0x0000014a, "V_BFI_B32", []>;
-defm : BFIPatterns <V_BFI_B32>;
-def V_FMA_F32 : VOP3_32 <0x0000014b, "V_FMA_F32",
-  [(set f32:$dst, (fma f32:$src0, f32:$src1, f32:$src2))]
+defm V_CUBEID_F32 : VOP3Inst <vop3<0x144>, "V_CUBEID_F32",
+  VOP_F32_F32_F32_F32
+>;
+defm V_CUBESC_F32 : VOP3Inst <vop3<0x145>, "V_CUBESC_F32",
+  VOP_F32_F32_F32_F32
+>;
+defm V_CUBETC_F32 : VOP3Inst <vop3<0x146>, "V_CUBETC_F32",
+  VOP_F32_F32_F32_F32
+>;
+defm V_CUBEMA_F32 : VOP3Inst <vop3<0x147>, "V_CUBEMA_F32",
+  VOP_F32_F32_F32_F32
+>;
+defm V_BFE_U32 : VOP3Inst <vop3<0x148>, "V_BFE_U32",
+  VOP_I32_I32_I32_I32, AMDGPUbfe_u32
+>;
+defm V_BFE_I32 : VOP3Inst <vop3<0x149>, "V_BFE_I32",
+  VOP_I32_I32_I32_I32, AMDGPUbfe_i32
 >;
-def V_FMA_F64 : VOP3_64 <0x0000014c, "V_FMA_F64",
-  [(set f64:$dst, (fma f64:$src0, f64:$src1, f64:$src2))]
+defm V_BFI_B32 : VOP3Inst <vop3<0x14a>, "V_BFI_B32",
+  VOP_I32_I32_I32_I32, AMDGPUbfi
+>;
+defm V_FMA_F32 : VOP3Inst <vop3<0x14b>, "V_FMA_F32",
+  VOP_F32_F32_F32_F32, fma
+>;
+defm V_FMA_F64 : VOP3Inst <vop3<0x14c>, "V_FMA_F64",
+  VOP_F64_F64_F64_F64, fma
 >;
 //def V_LERP_U8 : VOP3_U8 <0x0000014d, "V_LERP_U8", []>;
-def V_ALIGNBIT_B32 : VOP3_32 <0x0000014e, "V_ALIGNBIT_B32", []>;
-def : ROTRPattern <V_ALIGNBIT_B32>;
-
-def V_ALIGNBYTE_B32 : VOP3_32 <0x0000014f, "V_ALIGNBYTE_B32", []>;
-def V_MULLIT_F32 : VOP3_32 <0x00000150, "V_MULLIT_F32", []>;
+defm V_ALIGNBIT_B32 : VOP3Inst <vop3<0x14e>, "V_ALIGNBIT_B32",
+  VOP_I32_I32_I32_I32
+>;
+defm V_ALIGNBYTE_B32 : VOP3Inst <vop3<0x14f>, "V_ALIGNBYTE_B32",
+  VOP_I32_I32_I32_I32
+>;
+defm V_MULLIT_F32 : VOP3Inst <vop3<0x150>, "V_MULLIT_F32",
+  VOP_F32_F32_F32_F32>;
 ////def V_MIN3_F32 : VOP3_MIN3 <0x00000151, "V_MIN3_F32", []>;
 ////def V_MIN3_I32 : VOP3_MIN3 <0x00000152, "V_MIN3_I32", []>;
 ////def V_MIN3_U32 : VOP3_MIN3 <0x00000153, "V_MIN3_U32", []>;
@@ -1054,148 +1564,114 @@ def V_MULLIT_F32 : VOP3_32 <0x00000150, "V_MULLIT_F32", []>;
 //def V_SAD_U8 : VOP3_U8 <0x0000015a, "V_SAD_U8", []>;
 //def V_SAD_HI_U8 : VOP3_U8 <0x0000015b, "V_SAD_HI_U8", []>;
 //def V_SAD_U16 : VOP3_U16 <0x0000015c, "V_SAD_U16", []>;
-def V_SAD_U32 : VOP3_32 <0x0000015d, "V_SAD_U32", []>;
+defm V_SAD_U32 : VOP3Inst <vop3<0x15d>, "V_SAD_U32",
+  VOP_I32_I32_I32_I32
+>;
 ////def V_CVT_PK_U8_F32 : VOP3_U8 <0x0000015e, "V_CVT_PK_U8_F32", []>;
-def V_DIV_FIXUP_F32 : VOP3_32 <0x0000015f, "V_DIV_FIXUP_F32", []>;
-def V_DIV_FIXUP_F64 : VOP3_64 <0x00000160, "V_DIV_FIXUP_F64", []>;
+defm V_DIV_FIXUP_F32 : VOP3Inst <
+  vop3<0x15f>, "V_DIV_FIXUP_F32", VOP_F32_F32_F32_F32, AMDGPUdiv_fixup
+>;
+defm V_DIV_FIXUP_F64 : VOP3Inst <
+  vop3<0x160>, "V_DIV_FIXUP_F64", VOP_F64_F64_F64_F64, AMDGPUdiv_fixup
+>;
 
-def V_LSHL_B64 : VOP3_64_Shift <0x00000161, "V_LSHL_B64",
-  [(set i64:$dst, (shl i64:$src0, i32:$src1))]
+defm V_LSHL_B64 : VOP3Inst <vop3<0x161>, "V_LSHL_B64",
+  VOP_I64_I64_I32, shl
 >;
-def V_LSHR_B64 : VOP3_64_Shift <0x00000162, "V_LSHR_B64",
-  [(set i64:$dst, (srl i64:$src0, i32:$src1))]
+defm V_LSHR_B64 : VOP3Inst <vop3<0x162>, "V_LSHR_B64",
+  VOP_I64_I64_I32, srl
 >;
-def V_ASHR_I64 : VOP3_64_Shift <0x00000163, "V_ASHR_I64",
-  [(set i64:$dst, (sra i64:$src0, i32:$src1))]
+defm V_ASHR_I64 : VOP3Inst <vop3<0x163>, "V_ASHR_I64",
+  VOP_I64_I64_I32, sra
 >;
 
 let isCommutable = 1 in {
 
-def V_ADD_F64 : VOP3_64 <0x00000164, "V_ADD_F64", []>;
-def V_MUL_F64 : VOP3_64 <0x00000165, "V_MUL_F64", []>;
-def V_MIN_F64 : VOP3_64 <0x00000166, "V_MIN_F64", []>;
-def V_MAX_F64 : VOP3_64 <0x00000167, "V_MAX_F64", []>;
-
-} // isCommutable = 1
-
-def : Pat <
-  (fadd f64:$src0, f64:$src1),
-  (V_ADD_F64 $src0, $src1, (i64 0))
+defm V_ADD_F64 : VOP3Inst <vop3<0x164>, "V_ADD_F64",
+  VOP_F64_F64_F64, fadd
+>;
+defm V_MUL_F64 : VOP3Inst <vop3<0x165>, "V_MUL_F64",
+  VOP_F64_F64_F64, fmul
 >;
 
-def : Pat <
-  (fmul f64:$src0, f64:$src1),
-  (V_MUL_F64 $src0, $src1, (i64 0))
+defm V_MIN_F64 : VOP3Inst <vop3<0x166>, "V_MIN_F64",
+  VOP_F64_F64_F64, fminnum
 >;
+defm V_MAX_F64 : VOP3Inst <vop3<0x167>, "V_MAX_F64",
+  VOP_F64_F64_F64, fmaxnum
+>;
+
+} // isCommutable = 1
 
-def V_LDEXP_F64 : VOP3_64 <0x00000168, "V_LDEXP_F64", []>;
+defm V_LDEXP_F64 : VOP3Inst <vop3<0x168>, "V_LDEXP_F64",
+  VOP_F64_F64_I32, AMDGPUldexp
+>;
 
 let isCommutable = 1 in {
 
-def V_MUL_LO_U32 : VOP3_32 <0x00000169, "V_MUL_LO_U32", []>;
-def V_MUL_HI_U32 : VOP3_32 <0x0000016a, "V_MUL_HI_U32", []>;
-def V_MUL_LO_I32 : VOP3_32 <0x0000016b, "V_MUL_LO_I32", []>;
-def V_MUL_HI_I32 : VOP3_32 <0x0000016c, "V_MUL_HI_I32", []>;
+defm V_MUL_LO_U32 : VOP3Inst <vop3<0x169>, "V_MUL_LO_U32",
+  VOP_I32_I32_I32
+>;
+defm V_MUL_HI_U32 : VOP3Inst <vop3<0x16a>, "V_MUL_HI_U32",
+  VOP_I32_I32_I32
+>;
+defm V_MUL_LO_I32 : VOP3Inst <vop3<0x16b>, "V_MUL_LO_I32",
+  VOP_I32_I32_I32
+>;
+defm V_MUL_HI_I32 : VOP3Inst <vop3<0x16c>, "V_MUL_HI_I32",
+  VOP_I32_I32_I32
+>;
 
 } // isCommutable = 1
 
-def : Pat <
-  (mul i32:$src0, i32:$src1),
-  (V_MUL_LO_I32 $src0, $src1, (i32 0))
->;
+defm V_DIV_SCALE_F32 : VOP3b_32 <vop3<0x16d>, "V_DIV_SCALE_F32", []>;
 
-def : Pat <
-  (mulhu i32:$src0, i32:$src1),
-  (V_MUL_HI_U32 $src0, $src1, (i32 0))
->;
+// Double precision division pre-scale.
+defm V_DIV_SCALE_F64 : VOP3b_64 <vop3<0x16e>, "V_DIV_SCALE_F64", []>;
 
-def : Pat <
-  (mulhs i32:$src0, i32:$src1),
-  (V_MUL_HI_I32 $src0, $src1, (i32 0))
+defm V_DIV_FMAS_F32 : VOP3Inst <vop3<0x16f>, "V_DIV_FMAS_F32",
+  VOP_F32_F32_F32_F32, AMDGPUdiv_fmas
+>;
+defm V_DIV_FMAS_F64 : VOP3Inst <vop3<0x170>, "V_DIV_FMAS_F64",
+  VOP_F64_F64_F64_F64, AMDGPUdiv_fmas
 >;
-
-def V_DIV_SCALE_F32 : VOP3_32 <0x0000016d, "V_DIV_SCALE_F32", []>;
-def V_DIV_SCALE_F64 : VOP3_64 <0x0000016e, "V_DIV_SCALE_F64", []>;
-def V_DIV_FMAS_F32 : VOP3_32 <0x0000016f, "V_DIV_FMAS_F32", []>;
-def V_DIV_FMAS_F64 : VOP3_64 <0x00000170, "V_DIV_FMAS_F64", []>;
 //def V_MSAD_U8 : VOP3_U8 <0x00000171, "V_MSAD_U8", []>;
 //def V_QSAD_U8 : VOP3_U8 <0x00000172, "V_QSAD_U8", []>;
 //def V_MQSAD_U8 : VOP3_U8 <0x00000173, "V_MQSAD_U8", []>;
-def V_TRIG_PREOP_F64 : VOP3_64 <0x00000174, "V_TRIG_PREOP_F64", []>;
-def S_ADD_U32 : SOP2_32 <0x00000000, "S_ADD_U32", []>;
-def S_SUB_U32 : SOP2_32 <0x00000001, "S_SUB_U32", []>;
-def S_ADD_I32 : SOP2_32 <0x00000002, "S_ADD_I32", []>;
-def S_SUB_I32 : SOP2_32 <0x00000003, "S_SUB_I32", []>;
-def S_ADDC_U32 : SOP2_32 <0x00000004, "S_ADDC_U32", []>;
-def S_SUBB_U32 : SOP2_32 <0x00000005, "S_SUBB_U32", []>;
-def S_MIN_I32 : SOP2_32 <0x00000006, "S_MIN_I32", []>;
-def S_MIN_U32 : SOP2_32 <0x00000007, "S_MIN_U32", []>;
-def S_MAX_I32 : SOP2_32 <0x00000008, "S_MAX_I32", []>;
-def S_MAX_U32 : SOP2_32 <0x00000009, "S_MAX_U32", []>;
-
-def S_CSELECT_B32 : SOP2 <
-  0x0000000a, (outs SReg_32:$dst),
-  (ins SReg_32:$src0, SReg_32:$src1, SCCReg:$scc), "S_CSELECT_B32",
-  []
+defm V_TRIG_PREOP_F64 : VOP3Inst <
+  vop3<0x174>, "V_TRIG_PREOP_F64", VOP_F64_F64_I32, AMDGPUtrig_preop
 >;
 
-def S_CSELECT_B64 : SOP2_64 <0x0000000b, "S_CSELECT_B64", []>;
+//===----------------------------------------------------------------------===//
+// Pseudo Instructions
+//===----------------------------------------------------------------------===//
 
-def S_AND_B32 : SOP2_32 <0x0000000e, "S_AND_B32", []>;
+let isCodeGenOnly = 1, isPseudo = 1 in {
 
-def S_AND_B64 : SOP2_64 <0x0000000f, "S_AND_B64",
-  [(set i64:$dst, (and i64:$src0, i64:$src1))]
+def V_MOV_I1 : InstSI <
+  (outs VReg_1:$dst),
+  (ins i1imm:$src),
+  "", [(set i1:$dst, (imm:$src))]
 >;
 
-def : Pat <
-  (i1 (and i1:$src0, i1:$src1)),
-  (S_AND_B64 $src0, $src1)
+def V_AND_I1 : InstSI <
+   (outs VReg_1:$dst), (ins VReg_1:$src0, VReg_1:$src1), "",
+   [(set i1:$dst, (and i1:$src0, i1:$src1))]
 >;
 
-def S_OR_B32 : SOP2_32 <0x00000010, "S_OR_B32", []>;
-def S_OR_B64 : SOP2_64 <0x00000011, "S_OR_B64", []>;
-def : Pat <
-  (i1 (or i1:$src0, i1:$src1)),
-  (S_OR_B64 $src0, $src1)
+def V_OR_I1 : InstSI <
+   (outs VReg_1:$dst), (ins VReg_1:$src0, VReg_1:$src1), "",
+   [(set i1:$dst, (or i1:$src0, i1:$src1))]
 >;
-def S_XOR_B32 : SOP2_32 <0x00000012, "S_XOR_B32", []>;
-def S_XOR_B64 : SOP2_64 <0x00000013, "S_XOR_B64",
+
+def V_XOR_I1 : InstSI <
+  (outs VReg_1:$dst), (ins VReg_1:$src0, VReg_1:$src1), "",
   [(set i1:$dst, (xor i1:$src0, i1:$src1))]
 >;
-def S_ANDN2_B32 : SOP2_32 <0x00000014, "S_ANDN2_B32", []>;
-def S_ANDN2_B64 : SOP2_64 <0x00000015, "S_ANDN2_B64", []>;
-def S_ORN2_B32 : SOP2_32 <0x00000016, "S_ORN2_B32", []>;
-def S_ORN2_B64 : SOP2_64 <0x00000017, "S_ORN2_B64", []>;
-def S_NAND_B32 : SOP2_32 <0x00000018, "S_NAND_B32", []>;
-def S_NAND_B64 : SOP2_64 <0x00000019, "S_NAND_B64", []>;
-def S_NOR_B32 : SOP2_32 <0x0000001a, "S_NOR_B32", []>;
-def S_NOR_B64 : SOP2_64 <0x0000001b, "S_NOR_B64", []>;
-def S_XNOR_B32 : SOP2_32 <0x0000001c, "S_XNOR_B32", []>;
-def S_XNOR_B64 : SOP2_64 <0x0000001d, "S_XNOR_B64", []>;
-def S_LSHL_B32 : SOP2_32 <0x0000001e, "S_LSHL_B32", []>;
-def S_LSHL_B64 : SOP2_64 <0x0000001f, "S_LSHL_B64", []>;
-def S_LSHR_B32 : SOP2_32 <0x00000020, "S_LSHR_B32", []>;
-def S_LSHR_B64 : SOP2_64 <0x00000021, "S_LSHR_B64", []>;
-def S_ASHR_I32 : SOP2_32 <0x00000022, "S_ASHR_I32", []>;
-def S_ASHR_I64 : SOP2_64 <0x00000023, "S_ASHR_I64", []>;
-def S_BFM_B32 : SOP2_32 <0x00000024, "S_BFM_B32", []>;
-def S_BFM_B64 : SOP2_64 <0x00000025, "S_BFM_B64", []>;
-def S_MUL_I32 : SOP2_32 <0x00000026, "S_MUL_I32", []>;
-def S_BFE_U32 : SOP2_32 <0x00000027, "S_BFE_U32", []>;
-def S_BFE_I32 : SOP2_32 <0x00000028, "S_BFE_I32", []>;
-def S_BFE_U64 : SOP2_64 <0x00000029, "S_BFE_U64", []>;
-def S_BFE_I64 : SOP2_64 <0x0000002a, "S_BFE_I64", []>;
-//def S_CBRANCH_G_FORK : SOP2_ <0x0000002b, "S_CBRANCH_G_FORK", []>;
-def S_ABSDIFF_I32 : SOP2_32 <0x0000002c, "S_ABSDIFF_I32", []>;
-
-let isCodeGenOnly = 1, isPseudo = 1 in {
 
-def LOAD_CONST : AMDGPUShaderInst <
-  (outs GPRF32:$dst),
-  (ins i32imm:$src),
-  "LOAD_CONST $dst, $src",
-  [(set GPRF32:$dst, (int_AMDGPU_load_const imm:$src))]
->;
+let hasSideEffects = 1 in {
+def SGPR_USE : InstSI <(outs),(ins), "", []>;
+}
 
 // SI pseudo instructions. These are used by the CFG structurizer pass
 // and should be lowered to ISA instructions prior to codegen.
@@ -1205,19 +1681,19 @@ let mayLoad = 1, mayStore = 1, hasSideEffects = 1,
 
 let isBranch = 1, isTerminator = 1 in {
 
-def SI_IF : InstSI <
+def SI_IF: InstSI <
   (outs SReg_64:$dst),
   (ins SReg_64:$vcc, brtarget:$target),
-  "SI_IF $dst, $vcc, $target",
+  "",
   [(set i64:$dst, (int_SI_if i1:$vcc, bb:$target))]
 >;
 
 def SI_ELSE : InstSI <
   (outs SReg_64:$dst),
   (ins SReg_64:$src, brtarget:$target),
-  "SI_ELSE $dst, $src, $target",
-  [(set i64:$dst, (int_SI_else i64:$src, bb:$target))]> {
-
+  "",
+  [(set i64:$dst, (int_SI_else i64:$src, bb:$target))]
+> {
   let Constraints = "$src = $dst";
 }
 
@@ -1260,8 +1736,8 @@ def SI_END_CF : InstSI <
 
 def SI_KILL : InstSI <
   (outs),
-  (ins VReg_32:$src),
-  "SI_KIL $src",
+  (ins VSrc_32:$src),
+  "SI_KILL $src",
   [(int_AMDGPU_kill f32:$src)]
 >;
 
@@ -1270,6 +1746,36 @@ def SI_KILL : InstSI <
 
 let Uses = [EXEC], Defs = [EXEC,VCC,M0] in {
 
+//defm SI_ : RegisterLoadStore <VReg_32, FRAMEri, ADDRIndirect>;
+
+let UseNamedOperandTable = 1 in {
+
+def SI_RegisterLoad : InstSI <
+  (outs VReg_32:$dst, SReg_64:$temp),
+  (ins FRAMEri32:$addr, i32imm:$chan),
+  "", []
+> {
+  let isRegisterLoad = 1;
+  let mayLoad = 1;
+}
+
+class SIRegStore<dag outs> : InstSI <
+  outs,
+  (ins VReg_32:$val, FRAMEri32:$addr, i32imm:$chan),
+  "", []
+> {
+  let isRegisterStore = 1;
+  let mayStore = 1;
+}
+
+let usesCustomInserter = 1 in {
+def SI_RegisterStorePseudo : SIRegStore<(outs)>;
+} // End usesCustomInserter = 1
+def SI_RegisterStore : SIRegStore<(outs SReg_64:$temp)>;
+
+
+} // End UseNamedOperandTable = 1
+
 def SI_INDIRECT_SRC : InstSI <
   (outs VReg_32:$dst, SReg_64:$temp),
   (ins unknown:$src, VSrc_32:$idx, i32imm:$off),
@@ -1286,97 +1792,380 @@ class SI_INDIRECT_DST<RegisterClass rc> : InstSI <
   let Constraints = "$src = $dst";
 }
 
+def SI_INDIRECT_DST_V1 : SI_INDIRECT_DST<VReg_32>;
 def SI_INDIRECT_DST_V2 : SI_INDIRECT_DST<VReg_64>;
 def SI_INDIRECT_DST_V4 : SI_INDIRECT_DST<VReg_128>;
 def SI_INDIRECT_DST_V8 : SI_INDIRECT_DST<VReg_256>;
 def SI_INDIRECT_DST_V16 : SI_INDIRECT_DST<VReg_512>;
 
-} // Uses = [EXEC,VCC,M0], Defs = [EXEC,VCC,M0]
+} // Uses = [EXEC,VCC,M0], Defs = [EXEC,VCC,M0]
+
+let usesCustomInserter = 1 in {
+
+// This pseudo instruction takes a pointer as input and outputs a resource
+// constant that can be used with the ADDR64 MUBUF instructions.
+def SI_ADDR64_RSRC : InstSI <
+  (outs SReg_128:$srsrc),
+  (ins SSrc_64:$ptr),
+  "", []
+>;
+
+def V_SUB_F64 : InstSI <
+  (outs VReg_64:$dst),
+  (ins VReg_64:$src0, VReg_64:$src1),
+  "V_SUB_F64 $dst, $src0, $src1",
+  [(set f64:$dst, (fsub f64:$src0, f64:$src1))]
+>;
+
+} // end usesCustomInserter
+
+multiclass SI_SPILL_SGPR <RegisterClass sgpr_class> {
+
+  def _SAVE : InstSI <
+    (outs),
+    (ins sgpr_class:$src, i32imm:$frame_idx),
+    "", []
+  >;
+
+  def _RESTORE : InstSI <
+    (outs sgpr_class:$dst),
+    (ins i32imm:$frame_idx),
+    "", []
+  >;
+
+}
+
+defm SI_SPILL_S32  : SI_SPILL_SGPR <SReg_32>;
+defm SI_SPILL_S64  : SI_SPILL_SGPR <SReg_64>;
+defm SI_SPILL_S128 : SI_SPILL_SGPR <SReg_128>;
+defm SI_SPILL_S256 : SI_SPILL_SGPR <SReg_256>;
+defm SI_SPILL_S512 : SI_SPILL_SGPR <SReg_512>;
+
+multiclass SI_SPILL_VGPR <RegisterClass vgpr_class> {
+  def _SAVE : InstSI <
+    (outs),
+    (ins vgpr_class:$src, i32imm:$frame_idx),
+    "", []
+  >;
+
+  def _RESTORE : InstSI <
+    (outs vgpr_class:$dst),
+    (ins i32imm:$frame_idx),
+    "", []
+  >;
+}
+
+defm SI_SPILL_V32  : SI_SPILL_VGPR <VReg_32>;
+defm SI_SPILL_V64  : SI_SPILL_VGPR <VReg_64>;
+defm SI_SPILL_V96  : SI_SPILL_VGPR <VReg_96>;
+defm SI_SPILL_V128 : SI_SPILL_VGPR <VReg_128>;
+defm SI_SPILL_V256 : SI_SPILL_VGPR <VReg_256>;
+defm SI_SPILL_V512 : SI_SPILL_VGPR <VReg_512>;
+
+let Defs = [SCC] in {
+
+def SI_CONSTDATA_PTR : InstSI <
+  (outs SReg_64:$dst),
+  (ins),
+  "", [(set SReg_64:$dst, (i64 SIconstdata_ptr))]
+>;
+
+} // End Defs = [SCC]
+
+} // end IsCodeGenOnly, isPseudo
+
+} // end SubtargetPredicate = SI
+
+let Predicates = [isSI] in {
+
+def : Pat<
+  (int_AMDGPU_cndlt f32:$src0, f32:$src1, f32:$src2),
+  (V_CNDMASK_B32_e64 $src2, $src1,
+                     (V_CMP_GT_F32_e64 SRCMODS.NONE, 0, SRCMODS.NONE, $src0,
+                                       DSTCLAMP.NONE, DSTOMOD.NONE))
+>;
+
+def : Pat <
+  (int_AMDGPU_kilp),
+  (SI_KILL 0xbf800000)
+>;
+
+/* int_SI_vs_load_input */
+def : Pat<
+  (SIload_input v4i32:$tlst, imm:$attr_offset, i32:$buf_idx_vgpr),
+  (BUFFER_LOAD_FORMAT_XYZW_IDXEN $tlst, $buf_idx_vgpr, imm:$attr_offset, 0, 0, 0, 0)
+>;
+
+/* int_SI_export */
+def : Pat <
+  (int_SI_export imm:$en, imm:$vm, imm:$done, imm:$tgt, imm:$compr,
+                 f32:$src0, f32:$src1, f32:$src2, f32:$src3),
+  (EXP imm:$en, imm:$tgt, imm:$compr, imm:$done, imm:$vm,
+       $src0, $src1, $src2, $src3)
+>;
+
+//===----------------------------------------------------------------------===//
+// SMRD Patterns
+//===----------------------------------------------------------------------===//
+
+multiclass SMRD_Pattern <SMRD Instr_IMM, SMRD Instr_SGPR, ValueType vt> {
+
+  // 1. Offset as 8bit DWORD immediate
+  def : Pat <
+    (constant_load (add i64:$sbase, (i64 IMM8bitDWORD:$offset))),
+    (vt (Instr_IMM $sbase, (as_dword_i32imm $offset)))
+  >;
+
+  // 2. Offset loaded in an 32bit SGPR
+  def : Pat <
+    (constant_load (add i64:$sbase, (i64 IMM32bit:$offset))),
+    (vt (Instr_SGPR $sbase, (S_MOV_B32 (i32 (as_i32imm $offset)))))
+  >;
+
+  // 3. No offset at all
+  def : Pat <
+    (constant_load i64:$sbase),
+    (vt (Instr_IMM $sbase, 0))
+  >;
+}
 
-let usesCustomInserter = 1 in {
+defm : SMRD_Pattern <S_LOAD_DWORD_IMM, S_LOAD_DWORD_SGPR, f32>;
+defm : SMRD_Pattern <S_LOAD_DWORD_IMM, S_LOAD_DWORD_SGPR, i32>;
+defm : SMRD_Pattern <S_LOAD_DWORDX2_IMM, S_LOAD_DWORDX2_SGPR, v2i32>;
+defm : SMRD_Pattern <S_LOAD_DWORDX4_IMM, S_LOAD_DWORDX4_SGPR, v4i32>;
+defm : SMRD_Pattern <S_LOAD_DWORDX8_IMM, S_LOAD_DWORDX8_SGPR, v32i8>;
+defm : SMRD_Pattern <S_LOAD_DWORDX8_IMM, S_LOAD_DWORDX8_SGPR, v8i32>;
+defm : SMRD_Pattern <S_LOAD_DWORDX16_IMM, S_LOAD_DWORDX16_SGPR, v16i32>;
 
-// This pseudo instruction takes a pointer as input and outputs a resource
-// constant that can be used with the ADDR64 MUBUF instructions.
-def SI_ADDR64_RSRC : InstSI <
-  (outs SReg_128:$srsrc),
-  (ins SReg_64:$ptr),
-  "", []
+// 1. Offset as 8bit DWORD immediate
+def : Pat <
+  (SIload_constant v4i32:$sbase, IMM8bitDWORD:$offset),
+  (S_BUFFER_LOAD_DWORD_IMM $sbase, (as_dword_i32imm $offset))
 >;
 
-def V_SUB_F64 : InstSI <
-  (outs VReg_64:$dst),
-  (ins VReg_64:$src0, VReg_64:$src1),
-  "V_SUB_F64 $dst, $src0, $src1",
-  []
+// 2. Offset loaded in an 32bit SGPR
+def : Pat <
+  (SIload_constant v4i32:$sbase, imm:$offset),
+  (S_BUFFER_LOAD_DWORD_SGPR $sbase, (S_MOV_B32 imm:$offset))
 >;
 
-} // end usesCustomInserter
+} // Predicates = [isSI] in {
 
-} // end IsCodeGenOnly, isPseudo
+//===----------------------------------------------------------------------===//
+// SOP1 Patterns
+//===----------------------------------------------------------------------===//
 
-def : Pat<
-  (int_AMDGPU_cndlt f32:$src0, f32:$src1, f32:$src2),
-  (V_CNDMASK_B32_e64 $src2, $src1, (V_CMP_GT_F32_e64 0, $src0))
+def : Pat <
+  (i64 (ctpop i64:$src)),
+    (i64 (REG_SEQUENCE SReg_64,
+     (S_BCNT1_I32_B64 $src), sub0,
+     (S_MOV_B32 0), sub1))
 >;
 
+//===----------------------------------------------------------------------===//
+// SOP2 Patterns
+//===----------------------------------------------------------------------===//
+
+// V_ADD_I32_e32/S_ADD_U32 produces carry in VCC/SCC. For the vector
+// case, the sgpr-copies pass will fix this to use the vector version.
 def : Pat <
-  (int_AMDGPU_kilp),
-  (SI_KILL (V_MOV_B32_e32 0xbf800000))
+  (i32 (addc i32:$src0, i32:$src1)),
+  (S_ADD_U32 $src0, $src1)
 >;
 
-/* int_SI_vs_load_input */
-def : Pat<
-  (SIload_input i128:$tlst, IMM12bit:$attr_offset, i32:$buf_idx_vgpr),
-  (BUFFER_LOAD_FORMAT_XYZW_IDXEN $tlst, $buf_idx_vgpr, imm:$attr_offset)
->;
+let  Predicates = [isSI] in {
+
+//===----------------------------------------------------------------------===//
+// SOPP Patterns
+//===----------------------------------------------------------------------===//
 
-/* int_SI_export */
 def : Pat <
-  (int_SI_export imm:$en, imm:$vm, imm:$done, imm:$tgt, imm:$compr,
-                 f32:$src0, f32:$src1, f32:$src2, f32:$src3),
-  (EXP imm:$en, imm:$tgt, imm:$compr, imm:$done, imm:$vm,
-       $src0, $src1, $src2, $src3)
+  (int_AMDGPU_barrier_global),
+  (S_BARRIER)
 >;
 
+//===----------------------------------------------------------------------===//
+// VOP1 Patterns
+//===----------------------------------------------------------------------===//
+
+let Predicates = [UnsafeFPMath] in {
+def : RcpPat<V_RCP_F64_e32, f64>;
+defm : RsqPat<V_RSQ_F64_e32, f64>;
+defm : RsqPat<V_RSQ_F32_e32, f32>;
+}
+
+//===----------------------------------------------------------------------===//
+// VOP2 Patterns
+//===----------------------------------------------------------------------===//
+
 def : Pat <
-  (f64 (fsub f64:$src0, f64:$src1)),
-  (V_SUB_F64 $src0, $src1)
+  (i32 (add (i32 (ctpop i32:$popcnt)), i32:$val)),
+  (V_BCNT_U32_B32_e64 $popcnt, $val)
 >;
 
 /********** ======================= **********/
 /********** Image sampling patterns **********/
 /********** ======================= **********/
 
+// Image + sampler
+class SampleRawPattern<SDPatternOperator name, MIMG opcode, ValueType vt> : Pat <
+  (name vt:$addr, v8i32:$rsrc, v4i32:$sampler, i32:$dmask, i32:$unorm,
+        i32:$r128, i32:$da, i32:$glc, i32:$slc, i32:$tfe, i32:$lwe),
+  (opcode (as_i32imm $dmask), (as_i1imm $unorm), (as_i1imm $glc), (as_i1imm $da),
+          (as_i1imm $r128), (as_i1imm $tfe), (as_i1imm $lwe), (as_i1imm $slc),
+          $addr, $rsrc, $sampler)
+>;
+
+multiclass SampleRawPatterns<SDPatternOperator name, string opcode> {
+  def : SampleRawPattern<name, !cast<MIMG>(opcode # _V4_V1), i32>;
+  def : SampleRawPattern<name, !cast<MIMG>(opcode # _V4_V2), v2i32>;
+  def : SampleRawPattern<name, !cast<MIMG>(opcode # _V4_V4), v4i32>;
+  def : SampleRawPattern<name, !cast<MIMG>(opcode # _V4_V8), v8i32>;
+  def : SampleRawPattern<name, !cast<MIMG>(opcode # _V4_V16), v16i32>;
+}
+
+// Image only
+class ImagePattern<SDPatternOperator name, MIMG opcode, ValueType vt> : Pat <
+  (name vt:$addr, v8i32:$rsrc, i32:$dmask, i32:$unorm,
+        i32:$r128, i32:$da, i32:$glc, i32:$slc, i32:$tfe, i32:$lwe),
+  (opcode (as_i32imm $dmask), (as_i1imm $unorm), (as_i1imm $glc), (as_i1imm $da),
+          (as_i1imm $r128), (as_i1imm $tfe), (as_i1imm $lwe), (as_i1imm $slc),
+          $addr, $rsrc)
+>;
+
+multiclass ImagePatterns<SDPatternOperator name, string opcode> {
+  def : ImagePattern<name, !cast<MIMG>(opcode # _V4_V1), i32>;
+  def : ImagePattern<name, !cast<MIMG>(opcode # _V4_V2), v2i32>;
+  def : ImagePattern<name, !cast<MIMG>(opcode # _V4_V4), v4i32>;
+}
+
+// Basic sample
+defm : SampleRawPatterns<int_SI_image_sample,           "IMAGE_SAMPLE">;
+defm : SampleRawPatterns<int_SI_image_sample_cl,        "IMAGE_SAMPLE_CL">;
+defm : SampleRawPatterns<int_SI_image_sample_d,         "IMAGE_SAMPLE_D">;
+defm : SampleRawPatterns<int_SI_image_sample_d_cl,      "IMAGE_SAMPLE_D_CL">;
+defm : SampleRawPatterns<int_SI_image_sample_l,         "IMAGE_SAMPLE_L">;
+defm : SampleRawPatterns<int_SI_image_sample_b,         "IMAGE_SAMPLE_B">;
+defm : SampleRawPatterns<int_SI_image_sample_b_cl,      "IMAGE_SAMPLE_B_CL">;
+defm : SampleRawPatterns<int_SI_image_sample_lz,        "IMAGE_SAMPLE_LZ">;
+defm : SampleRawPatterns<int_SI_image_sample_cd,        "IMAGE_SAMPLE_CD">;
+defm : SampleRawPatterns<int_SI_image_sample_cd_cl,     "IMAGE_SAMPLE_CD_CL">;
+
+// Sample with comparison
+defm : SampleRawPatterns<int_SI_image_sample_c,         "IMAGE_SAMPLE_C">;
+defm : SampleRawPatterns<int_SI_image_sample_c_cl,      "IMAGE_SAMPLE_C_CL">;
+defm : SampleRawPatterns<int_SI_image_sample_c_d,       "IMAGE_SAMPLE_C_D">;
+defm : SampleRawPatterns<int_SI_image_sample_c_d_cl,    "IMAGE_SAMPLE_C_D_CL">;
+defm : SampleRawPatterns<int_SI_image_sample_c_l,       "IMAGE_SAMPLE_C_L">;
+defm : SampleRawPatterns<int_SI_image_sample_c_b,       "IMAGE_SAMPLE_C_B">;
+defm : SampleRawPatterns<int_SI_image_sample_c_b_cl,    "IMAGE_SAMPLE_C_B_CL">;
+defm : SampleRawPatterns<int_SI_image_sample_c_lz,      "IMAGE_SAMPLE_C_LZ">;
+defm : SampleRawPatterns<int_SI_image_sample_c_cd,      "IMAGE_SAMPLE_C_CD">;
+defm : SampleRawPatterns<int_SI_image_sample_c_cd_cl,   "IMAGE_SAMPLE_C_CD_CL">;
+
+// Sample with offsets
+defm : SampleRawPatterns<int_SI_image_sample_o,         "IMAGE_SAMPLE_O">;
+defm : SampleRawPatterns<int_SI_image_sample_cl_o,      "IMAGE_SAMPLE_CL_O">;
+defm : SampleRawPatterns<int_SI_image_sample_d_o,       "IMAGE_SAMPLE_D_O">;
+defm : SampleRawPatterns<int_SI_image_sample_d_cl_o,    "IMAGE_SAMPLE_D_CL_O">;
+defm : SampleRawPatterns<int_SI_image_sample_l_o,       "IMAGE_SAMPLE_L_O">;
+defm : SampleRawPatterns<int_SI_image_sample_b_o,       "IMAGE_SAMPLE_B_O">;
+defm : SampleRawPatterns<int_SI_image_sample_b_cl_o,    "IMAGE_SAMPLE_B_CL_O">;
+defm : SampleRawPatterns<int_SI_image_sample_lz_o,      "IMAGE_SAMPLE_LZ_O">;
+defm : SampleRawPatterns<int_SI_image_sample_cd_o,      "IMAGE_SAMPLE_CD_O">;
+defm : SampleRawPatterns<int_SI_image_sample_cd_cl_o,   "IMAGE_SAMPLE_CD_CL_O">;
+
+// Sample with comparison and offsets
+defm : SampleRawPatterns<int_SI_image_sample_c_o,       "IMAGE_SAMPLE_C_O">;
+defm : SampleRawPatterns<int_SI_image_sample_c_cl_o,    "IMAGE_SAMPLE_C_CL_O">;
+defm : SampleRawPatterns<int_SI_image_sample_c_d_o,     "IMAGE_SAMPLE_C_D_O">;
+defm : SampleRawPatterns<int_SI_image_sample_c_d_cl_o,  "IMAGE_SAMPLE_C_D_CL_O">;
+defm : SampleRawPatterns<int_SI_image_sample_c_l_o,     "IMAGE_SAMPLE_C_L_O">;
+defm : SampleRawPatterns<int_SI_image_sample_c_b_o,     "IMAGE_SAMPLE_C_B_O">;
+defm : SampleRawPatterns<int_SI_image_sample_c_b_cl_o,  "IMAGE_SAMPLE_C_B_CL_O">;
+defm : SampleRawPatterns<int_SI_image_sample_c_lz_o,    "IMAGE_SAMPLE_C_LZ_O">;
+defm : SampleRawPatterns<int_SI_image_sample_c_cd_o,    "IMAGE_SAMPLE_C_CD_O">;
+defm : SampleRawPatterns<int_SI_image_sample_c_cd_cl_o, "IMAGE_SAMPLE_C_CD_CL_O">;
+
+// Gather opcodes
+// Only the variants which make sense are defined.
+def : SampleRawPattern<int_SI_gather4,           IMAGE_GATHER4_V4_V2,        v2i32>;
+def : SampleRawPattern<int_SI_gather4,           IMAGE_GATHER4_V4_V4,        v4i32>;
+def : SampleRawPattern<int_SI_gather4_cl,        IMAGE_GATHER4_CL_V4_V4,     v4i32>;
+def : SampleRawPattern<int_SI_gather4_l,         IMAGE_GATHER4_L_V4_V4,      v4i32>;
+def : SampleRawPattern<int_SI_gather4_b,         IMAGE_GATHER4_B_V4_V4,      v4i32>;
+def : SampleRawPattern<int_SI_gather4_b_cl,      IMAGE_GATHER4_B_CL_V4_V4,   v4i32>;
+def : SampleRawPattern<int_SI_gather4_b_cl,      IMAGE_GATHER4_B_CL_V4_V8,   v8i32>;
+def : SampleRawPattern<int_SI_gather4_lz,        IMAGE_GATHER4_LZ_V4_V2,     v2i32>;
+def : SampleRawPattern<int_SI_gather4_lz,        IMAGE_GATHER4_LZ_V4_V4,     v4i32>;
+
+def : SampleRawPattern<int_SI_gather4_c,         IMAGE_GATHER4_C_V4_V4,      v4i32>;
+def : SampleRawPattern<int_SI_gather4_c_cl,      IMAGE_GATHER4_C_CL_V4_V4,   v4i32>;
+def : SampleRawPattern<int_SI_gather4_c_cl,      IMAGE_GATHER4_C_CL_V4_V8,   v8i32>;
+def : SampleRawPattern<int_SI_gather4_c_l,       IMAGE_GATHER4_C_L_V4_V4,    v4i32>;
+def : SampleRawPattern<int_SI_gather4_c_l,       IMAGE_GATHER4_C_L_V4_V8,    v8i32>;
+def : SampleRawPattern<int_SI_gather4_c_b,       IMAGE_GATHER4_C_B_V4_V4,    v4i32>;
+def : SampleRawPattern<int_SI_gather4_c_b,       IMAGE_GATHER4_C_B_V4_V8,    v8i32>;
+def : SampleRawPattern<int_SI_gather4_c_b_cl,    IMAGE_GATHER4_C_B_CL_V4_V8, v8i32>;
+def : SampleRawPattern<int_SI_gather4_c_lz,      IMAGE_GATHER4_C_LZ_V4_V4,   v4i32>;
+
+def : SampleRawPattern<int_SI_gather4_o,         IMAGE_GATHER4_O_V4_V4,      v4i32>;
+def : SampleRawPattern<int_SI_gather4_cl_o,      IMAGE_GATHER4_CL_O_V4_V4,   v4i32>;
+def : SampleRawPattern<int_SI_gather4_cl_o,      IMAGE_GATHER4_CL_O_V4_V8,   v8i32>;
+def : SampleRawPattern<int_SI_gather4_l_o,       IMAGE_GATHER4_L_O_V4_V4,    v4i32>;
+def : SampleRawPattern<int_SI_gather4_l_o,       IMAGE_GATHER4_L_O_V4_V8,    v8i32>;
+def : SampleRawPattern<int_SI_gather4_b_o,       IMAGE_GATHER4_B_O_V4_V4,    v4i32>;
+def : SampleRawPattern<int_SI_gather4_b_o,       IMAGE_GATHER4_B_O_V4_V8,    v8i32>;
+def : SampleRawPattern<int_SI_gather4_b_cl_o,    IMAGE_GATHER4_B_CL_O_V4_V8, v8i32>;
+def : SampleRawPattern<int_SI_gather4_lz_o,      IMAGE_GATHER4_LZ_O_V4_V4,   v4i32>;
+
+def : SampleRawPattern<int_SI_gather4_c_o,       IMAGE_GATHER4_C_O_V4_V4,    v4i32>;
+def : SampleRawPattern<int_SI_gather4_c_o,       IMAGE_GATHER4_C_O_V4_V8,    v8i32>;
+def : SampleRawPattern<int_SI_gather4_c_cl_o,    IMAGE_GATHER4_C_CL_O_V4_V8, v8i32>;
+def : SampleRawPattern<int_SI_gather4_c_l_o,     IMAGE_GATHER4_C_L_O_V4_V8,  v8i32>;
+def : SampleRawPattern<int_SI_gather4_c_b_o,     IMAGE_GATHER4_C_B_O_V4_V8,  v8i32>;
+def : SampleRawPattern<int_SI_gather4_c_b_cl_o,  IMAGE_GATHER4_C_B_CL_O_V4_V8, v8i32>;
+def : SampleRawPattern<int_SI_gather4_c_lz_o,    IMAGE_GATHER4_C_LZ_O_V4_V4, v4i32>;
+def : SampleRawPattern<int_SI_gather4_c_lz_o,    IMAGE_GATHER4_C_LZ_O_V4_V8, v8i32>;
+
+def : SampleRawPattern<int_SI_getlod, IMAGE_GET_LOD_V4_V1, i32>;
+def : SampleRawPattern<int_SI_getlod, IMAGE_GET_LOD_V4_V2, v2i32>;
+def : SampleRawPattern<int_SI_getlod, IMAGE_GET_LOD_V4_V4, v4i32>;
+
+def : ImagePattern<int_SI_getresinfo, IMAGE_GET_RESINFO_V4_V1, i32>;
+defm : ImagePatterns<int_SI_image_load, "IMAGE_LOAD">;
+defm : ImagePatterns<int_SI_image_load_mip, "IMAGE_LOAD_MIP">;
+
 /* SIsample for simple 1D texture lookup */
 def : Pat <
-  (SIsample i32:$addr, v32i8:$rsrc, i128:$sampler, imm),
+  (SIsample i32:$addr, v32i8:$rsrc, v4i32:$sampler, imm),
   (IMAGE_SAMPLE_V4_V1 0xf, 0, 0, 0, 0, 0, 0, 0, $addr, $rsrc, $sampler)
 >;
 
 class SamplePattern<SDNode name, MIMG opcode, ValueType vt> : Pat <
-    (name vt:$addr, v32i8:$rsrc, i128:$sampler, imm),
+    (name vt:$addr, v32i8:$rsrc, v4i32:$sampler, imm),
     (opcode 0xf, 0, 0, 0, 0, 0, 0, 0, $addr, $rsrc, $sampler)
 >;
 
 class SampleRectPattern<SDNode name, MIMG opcode, ValueType vt> : Pat <
-    (name vt:$addr, v32i8:$rsrc, i128:$sampler, TEX_RECT),
+    (name vt:$addr, v32i8:$rsrc, v4i32:$sampler, TEX_RECT),
     (opcode 0xf, 1, 0, 0, 0, 0, 0, 0, $addr, $rsrc, $sampler)
 >;
 
 class SampleArrayPattern<SDNode name, MIMG opcode, ValueType vt> : Pat <
-    (name vt:$addr, v32i8:$rsrc, i128:$sampler, TEX_ARRAY),
+    (name vt:$addr, v32i8:$rsrc, v4i32:$sampler, TEX_ARRAY),
     (opcode 0xf, 0, 0, 1, 0, 0, 0, 0, $addr, $rsrc, $sampler)
 >;
 
 class SampleShadowPattern<SDNode name, MIMG opcode,
                           ValueType vt> : Pat <
-    (name vt:$addr, v32i8:$rsrc, i128:$sampler, TEX_SHADOW),
+    (name vt:$addr, v32i8:$rsrc, v4i32:$sampler, TEX_SHADOW),
     (opcode 0xf, 0, 0, 0, 0, 0, 0, 0, $addr, $rsrc, $sampler)
 >;
 
 class SampleShadowArrayPattern<SDNode name, MIMG opcode,
                                ValueType vt> : Pat <
-    (name vt:$addr, v32i8:$rsrc, i128:$sampler, TEX_SHADOW_ARRAY),
+    (name vt:$addr, v32i8:$rsrc, v4i32:$sampler, TEX_SHADOW_ARRAY),
     (opcode 0xf, 0, 0, 1, 0, 0, 0, 0, $addr, $rsrc, $sampler)
 >;
 
@@ -1561,37 +2350,92 @@ def : BitConvert <f64, i64, VReg_64>;
 def : BitConvert <v2f32, v2i32, VReg_64>;
 def : BitConvert <v2i32, v2f32, VReg_64>;
 def : BitConvert <v2i32, i64, VReg_64>;
-
+def : BitConvert <i64, v2i32, VReg_64>;
+def : BitConvert <v2f32, i64, VReg_64>;
+def : BitConvert <i64, v2f32, VReg_64>;
+def : BitConvert <v2i32, f64, VReg_64>;
+def : BitConvert <f64, v2i32, VReg_64>;
 def : BitConvert <v4f32, v4i32, VReg_128>;
 def : BitConvert <v4i32, v4f32, VReg_128>;
-def : BitConvert <v4i32, i128,  VReg_128>;
-def : BitConvert <i128, v4i32,  VReg_128>;
 
+def : BitConvert <v8f32, v8i32, SReg_256>;
+def : BitConvert <v8i32, v8f32, SReg_256>;
 def : BitConvert <v8i32, v32i8, SReg_256>;
 def : BitConvert <v32i8, v8i32, SReg_256>;
 def : BitConvert <v8i32, v32i8, VReg_256>;
+def : BitConvert <v8i32, v8f32, VReg_256>;
+def : BitConvert <v8f32, v8i32, VReg_256>;
 def : BitConvert <v32i8, v8i32, VReg_256>;
 
+def : BitConvert <v16i32, v16f32, VReg_512>;
+def : BitConvert <v16f32, v16i32, VReg_512>;
+
 /********** =================== **********/
 /********** Src & Dst modifiers **********/
 /********** =================== **********/
 
+def FCLAMP_SI : AMDGPUShaderInst <
+  (outs VReg_32:$dst),
+  (ins VSrc_32:$src0),
+  "FCLAMP_SI $dst, $src0",
+  []
+> {
+  let usesCustomInserter = 1;
+}
+
+def : Pat <
+  (AMDGPUclamp f32:$src, (f32 FP_ZERO), (f32 FP_ONE)),
+  (FCLAMP_SI f32:$src)
+>;
+
+/********** ================================ **********/
+/********** Floating point absolute/negative **********/
+/********** ================================ **********/
+
+// Prevent expanding both fneg and fabs.
+
+// FIXME: Should use S_OR_B32
 def : Pat <
-  (int_AMDIL_clamp f32:$src, (f32 FP_ZERO), (f32 FP_ONE)),
-  (V_ADD_F32_e64 $src, (i32 0 /* SRC1 */),
-   0 /* ABS */, 1 /* CLAMP */, 0 /* OMOD */, 0 /* NEG */)
+  (fneg (fabs f32:$src)),
+  (V_OR_B32_e32 $src, (V_MOV_B32_e32 0x80000000)) /* Set sign bit */
+>;
+
+// FIXME: Should use S_OR_B32
+def : Pat <
+  (fneg (fabs f64:$src)),
+    (f64 (INSERT_SUBREG
+      (INSERT_SUBREG (f64 (IMPLICIT_DEF)),
+       (i32 (EXTRACT_SUBREG f64:$src, sub0)), sub0),
+    (V_OR_B32_e32 (EXTRACT_SUBREG f64:$src, sub1),
+                  (V_MOV_B32_e32 0x80000000)), sub1)) // Set sign bit.
 >;
 
 def : Pat <
   (fabs f32:$src),
-  (V_ADD_F32_e64 $src, (i32 0 /* SRC1 */),
-   1 /* ABS */, 0 /* CLAMP */, 0 /* OMOD */, 0 /* NEG */)
+  (V_AND_B32_e32 $src, (V_MOV_B32_e32 0x7fffffff))
 >;
 
 def : Pat <
   (fneg f32:$src),
-  (V_ADD_F32_e64 $src, (i32 0 /* SRC1 */),
-   0 /* ABS */, 0 /* CLAMP */, 0 /* OMOD */, 1 /* NEG */)
+  (V_XOR_B32_e32 $src, (V_MOV_B32_e32 0x80000000))
+>;
+
+def : Pat <
+  (fabs f64:$src),
+    (f64 (INSERT_SUBREG
+      (INSERT_SUBREG (f64 (IMPLICIT_DEF)),
+       (i32 (EXTRACT_SUBREG f64:$src, sub0)), sub0),
+    (V_AND_B32_e32 (EXTRACT_SUBREG f64:$src, sub1),
+                    (V_MOV_B32_e32 0x7fffffff)), sub1)) // Set sign bit.
+>;
+
+def : Pat <
+  (fneg f64:$src),
+    (f64 (INSERT_SUBREG
+      (INSERT_SUBREG (f64 (IMPLICIT_DEF)),
+       (i32 (EXTRACT_SUBREG f64:$src, sub0)), sub0),
+    (V_XOR_B32_e32 (EXTRACT_SUBREG f64:$src, sub1),
+                    (V_MOV_B32_e32 0x80000000)), sub1))
 >;
 
 /********** ================== **********/
@@ -1618,31 +2462,11 @@ def : Pat <
   (V_MOV_B32_e32 fpimm:$imm)
 >;
 
-def : Pat <
-  (i1 imm:$imm),
-  (S_MOV_B64 imm:$imm)
->;
-
 def : Pat <
   (i64 InlineImm<i64>:$imm),
   (S_MOV_B64 InlineImm<i64>:$imm)
 >;
 
-// i64 immediates aren't supported in hardware, split it into two 32bit values
-def : Pat <
-  (i64 imm:$imm),
-  (INSERT_SUBREG (INSERT_SUBREG (i64 (IMPLICIT_DEF)),
-    (S_MOV_B32 (i32 (LO32 imm:$imm))), sub0),
-    (S_MOV_B32 (i32 (HI32 imm:$imm))), sub1)
->;
-
-def : Pat <
-  (f64 fpimm:$imm),
-  (INSERT_SUBREG (INSERT_SUBREG (f64 (IMPLICIT_DEF)),
-    (V_MOV_B32_e32 (f32 (LO32f fpimm:$imm))), sub0),
-    (V_MOV_B32_e32 (f32 (HI32f fpimm:$imm))), sub1)
->;
-
 /********** ===================== **********/
 /********** Interpolation Paterns **********/
 /********** ===================== **********/
@@ -1672,45 +2496,36 @@ def : Pat <
   (V_MUL_LEGACY_F32_e32 $src0, (V_RCP_LEGACY_F32_e32 $src1))
 >;
 
-def : Pat<
-  (fdiv f32:$src0, f32:$src1),
-  (V_MUL_F32_e32 $src0, (V_RCP_F32_e32 $src1))
->;
-
 def : Pat<
   (fdiv f64:$src0, f64:$src1),
-  (V_MUL_F64 $src0, (V_RCP_F64_e32 $src1), (i64 0))
->;
-
-def : Pat <
-  (fcos f32:$src0),
-  (V_COS_F32_e32 (V_MUL_F32_e32 $src0, (V_MOV_B32_e32 CONST.TWO_PI_INV)))
->;
-
-def : Pat <
-  (fsin f32:$src0),
-  (V_SIN_F32_e32 (V_MUL_F32_e32 $src0, (V_MOV_B32_e32 CONST.TWO_PI_INV)))
+  (V_MUL_F64 0 /* src0_modifiers */, $src0,
+             0 /* src1_modifiers */, (V_RCP_F64_e32 $src1),
+             0 /* clamp */, 0 /* omod */)
 >;
 
 def : Pat <
   (int_AMDGPU_cube v4f32:$src),
   (INSERT_SUBREG (INSERT_SUBREG (INSERT_SUBREG (INSERT_SUBREG (v4f32 (IMPLICIT_DEF)),
-    (V_CUBETC_F32 (EXTRACT_SUBREG $src, sub0),
-                  (EXTRACT_SUBREG $src, sub1),
-                  (EXTRACT_SUBREG $src, sub2)),
-                   sub0),
-    (V_CUBESC_F32 (EXTRACT_SUBREG $src, sub0),
-                  (EXTRACT_SUBREG $src, sub1),
-                  (EXTRACT_SUBREG $src, sub2)),
-                   sub1),
-    (V_CUBEMA_F32 (EXTRACT_SUBREG $src, sub0),
-                  (EXTRACT_SUBREG $src, sub1),
-                  (EXTRACT_SUBREG $src, sub2)),
-                   sub2),
-    (V_CUBEID_F32 (EXTRACT_SUBREG $src, sub0),
-                  (EXTRACT_SUBREG $src, sub1),
-                  (EXTRACT_SUBREG $src, sub2)),
-                   sub3)
+    (V_CUBETC_F32 0 /* src0_modifiers */, (EXTRACT_SUBREG $src, sub0),
+                  0 /* src1_modifiers */, (EXTRACT_SUBREG $src, sub1),
+                  0 /* src2_modifiers */, (EXTRACT_SUBREG $src, sub2),
+                  0 /* clamp */, 0 /* omod */),
+                  sub0),
+    (V_CUBESC_F32 0 /* src0_modifiers */, (EXTRACT_SUBREG $src, sub0),
+                  0 /* src1_modifiers */,(EXTRACT_SUBREG $src, sub1),
+                  0 /* src2_modifiers */,(EXTRACT_SUBREG $src, sub2),
+                  0 /* clamp */, 0 /* omod */),
+                  sub1),
+    (V_CUBEMA_F32 0 /* src1_modifiers */,(EXTRACT_SUBREG $src, sub0),
+                  0 /* src1_modifiers */,(EXTRACT_SUBREG $src, sub1),
+                  0 /* src1_modifiers */,(EXTRACT_SUBREG $src, sub2),
+                  0 /* clamp */, 0 /* omod */),
+                  sub2),
+    (V_CUBEID_F32 0 /* src1_modifiers */,(EXTRACT_SUBREG $src, sub0),
+                  0 /* src1_modifiers */,(EXTRACT_SUBREG $src, sub1),
+                  0 /* src1_modifiers */,(EXTRACT_SUBREG $src, sub2),
+                  0 /* clamp */, 0 /* omod */),
+                  sub3)
 >;
 
 def : Pat <
@@ -1718,22 +2533,18 @@ def : Pat <
   (V_CNDMASK_B32_e64 (i32 0), (i32 -1), $src0)
 >;
 
-// 1. Offset as 8bit DWORD immediate
-def : Pat <
-  (SIload_constant i128:$sbase, IMM8bitDWORD:$offset),
-  (S_BUFFER_LOAD_DWORD_IMM $sbase, IMM8bitDWORD:$offset)
+class Ext32Pat <SDNode ext> : Pat <
+  (i32 (ext i1:$src0)),
+  (V_CNDMASK_B32_e64 (i32 0), (i32 1), $src0)
 >;
 
-// 2. Offset loaded in an 32bit SGPR
-def : Pat <
-  (SIload_constant i128:$sbase, imm:$offset),
-  (S_BUFFER_LOAD_DWORD_SGPR $sbase, (S_MOV_B32 imm:$offset))
->;
+def : Ext32Pat <zext>;
+def : Ext32Pat <anyext>;
 
-// 3. Offset in an 32Bit VGPR
+// Offset in an 32Bit VGPR
 def : Pat <
-  (SIload_constant i128:$sbase, i32:$voff),
-  (BUFFER_LOAD_DWORD_OFFEN $sbase, $voff)
+  (SIload_constant v4i32:$sbase, i32:$voff),
+  (BUFFER_LOAD_DWORD_OFFEN $sbase, $voff, 0, 0, 0, 0, 0)
 >;
 
 // The multiplication scales from [0,1] to the unsigned integer range
@@ -1747,25 +2558,39 @@ def : Pat <
 def : Pat <
   (int_SI_tid),
   (V_MBCNT_HI_U32_B32_e32 0xffffffff,
-                          (V_MBCNT_LO_U32_B32_e64 0xffffffff, 0, 0, 0, 0, 0))
+                          (V_MBCNT_LO_U32_B32_e64 0xffffffff, 0))
 >;
 
-/********** ================== **********/
-/**********   VOP3 Patterns    **********/
-/********** ================== **********/
+//===----------------------------------------------------------------------===//
+// VOP3 Patterns
+//===----------------------------------------------------------------------===//
+
+def : IMad24Pat<V_MAD_I32_I24>;
+def : UMad24Pat<V_MAD_U32_U24>;
+
+def : Pat <
+  (mulhu i32:$src0, i32:$src1),
+  (V_MUL_HI_U32 $src0, $src1)
+>;
 
 def : Pat <
-  (f32 (fadd (fmul f32:$src0, f32:$src1), f32:$src2)),
-  (V_MAD_F32 $src0, $src1, $src2)
+  (mulhs i32:$src0, i32:$src1),
+  (V_MUL_HI_I32 $src0, $src1)
 >;
 
+def : Vop3ModPat<V_MAD_F32, VOP_F32_F32_F32_F32, AMDGPUmad>;
+
+
+defm : BFIPatterns <V_BFI_B32, S_MOV_B32>;
+def : ROTRPattern <V_ALIGNBIT_B32>;
+
 /********** ======================= **********/
 /**********   Load/Store Patterns   **********/
 /********** ======================= **********/
 
 class DSReadPat <DS inst, ValueType vt, PatFrag frag> : Pat <
-  (frag i32:$src0),
-  (vt (inst 0, $src0, $src0, $src0, 0, 0))
+  (vt (frag (DS1Addr1Offset i32:$ptr, i32:$offset))),
+  (inst (i1 0), $ptr, (as_i16imm $offset))
 >;
 
 def : DSReadPat <DS_READ_I8,  i32, sextloadi8_local>;
@@ -1773,126 +2598,210 @@ def : DSReadPat <DS_READ_U8,  i32, az_extloadi8_local>;
 def : DSReadPat <DS_READ_I16, i32, sextloadi16_local>;
 def : DSReadPat <DS_READ_U16, i32, az_extloadi16_local>;
 def : DSReadPat <DS_READ_B32, i32, local_load>;
+
+let AddedComplexity = 100 in {
+
+def : DSReadPat <DS_READ_B64, v2i32, local_load_aligned8bytes>;
+
+} // End AddedComplexity = 100
+
 def : Pat <
-    (local_load i32:$src0),
-    (i32 (DS_READ_B32 0, $src0, $src0, $src0, 0, 0))
+  (v2i32 (local_load (DS64Bit4ByteAligned i32:$ptr, i8:$offset0,
+                                                    i8:$offset1))),
+  (DS_READ2_B32 (i1 0), $ptr, $offset0, $offset1)
 >;
 
 class DSWritePat <DS inst, ValueType vt, PatFrag frag> : Pat <
-  (frag i32:$src1, i32:$src0),
-  (inst 0, $src0, $src1, $src1, 0, 0)
+  (frag vt:$value, (DS1Addr1Offset i32:$ptr, i32:$offset)),
+  (inst (i1 0), $ptr, $value, (as_i16imm $offset))
 >;
 
 def : DSWritePat <DS_WRITE_B8, i32, truncstorei8_local>;
 def : DSWritePat <DS_WRITE_B16, i32, truncstorei16_local>;
 def : DSWritePat <DS_WRITE_B32, i32, local_store>;
 
-def : Pat <(atomic_load_add_local i32:$ptr, i32:$val),
-           (DS_ADD_U32_RTN 0, $ptr, $val, 0, 0)>;
+let AddedComplexity = 100 in {
 
-def : Pat <(atomic_load_sub_local i32:$ptr, i32:$val),
-           (DS_SUB_U32_RTN 0, $ptr, $val, 0, 0)>;
+def : DSWritePat <DS_WRITE_B64, v2i32, local_store_aligned8bytes>;
+} // End AddedComplexity = 100
 
-/********** ================== **********/
-/**********   SMRD Patterns    **********/
-/********** ================== **********/
+def : Pat <
+  (local_store v2i32:$value, (DS64Bit4ByteAligned i32:$ptr, i8:$offset0,
+                                                            i8:$offset1)),
+  (DS_WRITE2_B32 (i1 0), $ptr, (EXTRACT_SUBREG $value, sub0),
+                        (EXTRACT_SUBREG $value, sub1), $offset0, $offset1)
+>;
 
-multiclass SMRD_Pattern <SMRD Instr_IMM, SMRD Instr_SGPR, ValueType vt> {
+class DSAtomicRetPat<DS inst, ValueType vt, PatFrag frag> : Pat <
+  (frag (DS1Addr1Offset i32:$ptr, i32:$offset), vt:$value),
+  (inst (i1 0), $ptr, $value, (as_i16imm $offset))
+>;
 
-  // 1. Offset as 8bit DWORD immediate
-  def : Pat <
-    (constant_load (SIadd64bit32bit i64:$sbase, IMM8bitDWORD:$offset)),
-    (vt (Instr_IMM $sbase, IMM8bitDWORD:$offset))
-  >;
+// Special case of DSAtomicRetPat for add / sub 1 -> inc / dec
+//
+// We need to use something for the data0, so we set a register to
+// -1. For the non-rtn variants, the manual says it does
+// DS[A] = (DS[A] >= D0) ? 0 : DS[A] + 1, and setting D0 to uint_max
+// will always do the increment so I'm assuming it's the same.
+//
+// We also load this -1 with s_mov_b32 / s_mov_b64 even though this
+// needs to be a VGPR. The SGPR copy pass will fix this, and it's
+// easier since there is no v_mov_b64.
+class DSAtomicIncRetPat<DS inst, ValueType vt,
+                        Instruction LoadImm, PatFrag frag> : Pat <
+  (frag (DS1Addr1Offset i32:$ptr, i32:$offset), (vt 1)),
+  (inst (i1 0), $ptr, (LoadImm (vt -1)), (as_i16imm $offset))
+>;
 
-  // 2. Offset loaded in an 32bit SGPR
-  def : Pat <
-    (constant_load (SIadd64bit32bit i64:$sbase, imm:$offset)),
-    (vt (Instr_SGPR $sbase, (S_MOV_B32 imm:$offset)))
-  >;
 
-  // 3. No offset at all
-  def : Pat <
-    (constant_load i64:$sbase),
-    (vt (Instr_IMM $sbase, 0))
-  >;
-}
+class DSAtomicCmpXChg <DS inst, ValueType vt, PatFrag frag> : Pat <
+  (frag (DS1Addr1Offset i32:$ptr, i32:$offset), vt:$cmp, vt:$swap),
+  (inst (i1 0), $ptr, $cmp, $swap, (as_i16imm $offset))
+>;
+
+
+// 32-bit atomics.
+def : DSAtomicIncRetPat<DS_INC_RTN_U32, i32,
+                        S_MOV_B32, atomic_load_add_local>;
+def : DSAtomicIncRetPat<DS_DEC_RTN_U32, i32,
+                        S_MOV_B32, atomic_load_sub_local>;
+
+def : DSAtomicRetPat<DS_WRXCHG_RTN_B32, i32, atomic_swap_local>;
+def : DSAtomicRetPat<DS_ADD_RTN_U32, i32, atomic_load_add_local>;
+def : DSAtomicRetPat<DS_SUB_RTN_U32, i32, atomic_load_sub_local>;
+def : DSAtomicRetPat<DS_AND_RTN_B32, i32, atomic_load_and_local>;
+def : DSAtomicRetPat<DS_OR_RTN_B32, i32, atomic_load_or_local>;
+def : DSAtomicRetPat<DS_XOR_RTN_B32, i32, atomic_load_xor_local>;
+def : DSAtomicRetPat<DS_MIN_RTN_I32, i32, atomic_load_min_local>;
+def : DSAtomicRetPat<DS_MAX_RTN_I32, i32, atomic_load_max_local>;
+def : DSAtomicRetPat<DS_MIN_RTN_U32, i32, atomic_load_umin_local>;
+def : DSAtomicRetPat<DS_MAX_RTN_U32, i32, atomic_load_umax_local>;
+
+def : DSAtomicCmpXChg<DS_CMPST_RTN_B32, i32, atomic_cmp_swap_32_local>;
+
+// 64-bit atomics.
+def : DSAtomicIncRetPat<DS_INC_RTN_U64, i64,
+                        S_MOV_B64, atomic_load_add_local>;
+def : DSAtomicIncRetPat<DS_DEC_RTN_U64, i64,
+                        S_MOV_B64, atomic_load_sub_local>;
+
+def : DSAtomicRetPat<DS_WRXCHG_RTN_B64, i64, atomic_swap_local>;
+def : DSAtomicRetPat<DS_ADD_RTN_U64, i64, atomic_load_add_local>;
+def : DSAtomicRetPat<DS_SUB_RTN_U64, i64, atomic_load_sub_local>;
+def : DSAtomicRetPat<DS_AND_RTN_B64, i64, atomic_load_and_local>;
+def : DSAtomicRetPat<DS_OR_RTN_B64, i64, atomic_load_or_local>;
+def : DSAtomicRetPat<DS_XOR_RTN_B64, i64, atomic_load_xor_local>;
+def : DSAtomicRetPat<DS_MIN_RTN_I64, i64, atomic_load_min_local>;
+def : DSAtomicRetPat<DS_MAX_RTN_I64, i64, atomic_load_max_local>;
+def : DSAtomicRetPat<DS_MIN_RTN_U64, i64, atomic_load_umin_local>;
+def : DSAtomicRetPat<DS_MAX_RTN_U64, i64, atomic_load_umax_local>;
+
+def : DSAtomicCmpXChg<DS_CMPST_RTN_B64, i64, atomic_cmp_swap_64_local>;
 
-defm : SMRD_Pattern <S_LOAD_DWORD_IMM, S_LOAD_DWORD_SGPR, f32>;
-defm : SMRD_Pattern <S_LOAD_DWORD_IMM, S_LOAD_DWORD_SGPR, i32>;
-defm : SMRD_Pattern <S_LOAD_DWORDX2_IMM, S_LOAD_DWORDX2_SGPR, i64>;
-defm : SMRD_Pattern <S_LOAD_DWORDX2_IMM, S_LOAD_DWORDX2_SGPR, v2i32>;
-defm : SMRD_Pattern <S_LOAD_DWORDX4_IMM, S_LOAD_DWORDX4_SGPR, i128>;
-defm : SMRD_Pattern <S_LOAD_DWORDX4_IMM, S_LOAD_DWORDX4_SGPR, v4i32>;
-defm : SMRD_Pattern <S_LOAD_DWORDX8_IMM, S_LOAD_DWORDX8_SGPR, v32i8>;
-defm : SMRD_Pattern <S_LOAD_DWORDX8_IMM, S_LOAD_DWORDX8_SGPR, v8i32>;
-defm : SMRD_Pattern <S_LOAD_DWORDX16_IMM, S_LOAD_DWORDX16_SGPR, v16i32>;
 
 //===----------------------------------------------------------------------===//
 // MUBUF Patterns
 //===----------------------------------------------------------------------===//
 
 multiclass MUBUFLoad_Pattern <MUBUF Instr_ADDR64, ValueType vt,
-                              PatFrag global_ld, PatFrag constant_ld> {
+                              PatFrag constant_ld> {
   def : Pat <
-    (vt (global_ld (add i64:$ptr, (i64 IMM12bit:$offset)))),
-    (Instr_ADDR64 (SI_ADDR64_RSRC (i64 0)), $ptr, (as_i16imm $offset))
+     (vt (constant_ld (MUBUFAddr64 v4i32:$srsrc, i64:$vaddr, i16:$offset))),
+     (Instr_ADDR64 $srsrc, $vaddr, $offset)
   >;
+}
 
-  def : Pat <
-    (vt (global_ld i64:$ptr)),
-    (Instr_ADDR64 (SI_ADDR64_RSRC (i64 0)), $ptr, 0)
-  >;
+defm : MUBUFLoad_Pattern <BUFFER_LOAD_SBYTE_ADDR64, i32, sextloadi8_constant>;
+defm : MUBUFLoad_Pattern <BUFFER_LOAD_UBYTE_ADDR64, i32, az_extloadi8_constant>;
+defm : MUBUFLoad_Pattern <BUFFER_LOAD_SSHORT_ADDR64, i32, sextloadi16_constant>;
+defm : MUBUFLoad_Pattern <BUFFER_LOAD_USHORT_ADDR64, i32, az_extloadi16_constant>;
+defm : MUBUFLoad_Pattern <BUFFER_LOAD_DWORD_ADDR64, i32, constant_load>;
+defm : MUBUFLoad_Pattern <BUFFER_LOAD_DWORDX2_ADDR64, v2i32, constant_load>;
+defm : MUBUFLoad_Pattern <BUFFER_LOAD_DWORDX4_ADDR64, v4i32, constant_load>;
+
+class MUBUFScratchLoadPat <MUBUF Instr, ValueType vt, PatFrag ld> : Pat <
+  (vt (ld (MUBUFScratch v4i32:$srsrc, i32:$vaddr,
+                        i32:$soffset, u16imm:$offset))),
+  (Instr $srsrc, $vaddr, $soffset, $offset, 0, 0, 0)
+>;
+
+def : MUBUFScratchLoadPat <BUFFER_LOAD_SBYTE_OFFEN, i32, sextloadi8_private>;
+def : MUBUFScratchLoadPat <BUFFER_LOAD_UBYTE_OFFEN, i32, extloadi8_private>;
+def : MUBUFScratchLoadPat <BUFFER_LOAD_SSHORT_OFFEN, i32, sextloadi16_private>;
+def : MUBUFScratchLoadPat <BUFFER_LOAD_USHORT_OFFEN, i32, extloadi16_private>;
+def : MUBUFScratchLoadPat <BUFFER_LOAD_DWORD_OFFEN, i32, load_private>;
+def : MUBUFScratchLoadPat <BUFFER_LOAD_DWORDX2_OFFEN, v2i32, load_private>;
+def : MUBUFScratchLoadPat <BUFFER_LOAD_DWORDX4_OFFEN, v4i32, load_private>;
+
+// BUFFER_LOAD_DWORD*, addr64=0
+multiclass MUBUF_Load_Dword <ValueType vt, MUBUF offset, MUBUF offen, MUBUF idxen,
+                             MUBUF bothen> {
 
   def : Pat <
-     (vt (global_ld (add i64:$ptr, i64:$offset))),
-     (Instr_ADDR64 (SI_ADDR64_RSRC $ptr), $offset, 0)
+    (vt (int_SI_buffer_load_dword v4i32:$rsrc, (i32 imm), i32:$soffset,
+                                  imm:$offset, 0, 0, imm:$glc, imm:$slc,
+                                  imm:$tfe)),
+    (offset $rsrc, (as_i16imm $offset), $soffset, (as_i1imm $glc),
+            (as_i1imm $slc), (as_i1imm $tfe))
   >;
 
   def : Pat <
-     (vt (constant_ld (add i64:$ptr, i64:$offset))),
-     (Instr_ADDR64 (SI_ADDR64_RSRC $ptr), $offset, 0)
+    (vt (int_SI_buffer_load_dword v4i32:$rsrc, i32:$vaddr, i32:$soffset,
+                                  imm:$offset, 1, 0, imm:$glc, imm:$slc,
+                                  imm:$tfe)),
+    (offen $rsrc, $vaddr, $soffset, (as_i16imm $offset), (as_i1imm $glc), (as_i1imm $slc),
+           (as_i1imm $tfe))
   >;
-}
-
-defm : MUBUFLoad_Pattern <BUFFER_LOAD_SBYTE_ADDR64, i32,
-                          sextloadi8_global, sextloadi8_constant>;
-defm : MUBUFLoad_Pattern <BUFFER_LOAD_UBYTE_ADDR64, i32,
-                          az_extloadi8_global, az_extloadi8_constant>;
-defm : MUBUFLoad_Pattern <BUFFER_LOAD_SSHORT_ADDR64, i32,
-                          sextloadi16_global, sextloadi16_constant>;
-defm : MUBUFLoad_Pattern <BUFFER_LOAD_USHORT_ADDR64, i32,
-                          az_extloadi16_global, az_extloadi16_constant>;
-defm : MUBUFLoad_Pattern <BUFFER_LOAD_DWORD_ADDR64, i32,
-                          global_load, constant_load>;
-defm : MUBUFLoad_Pattern <BUFFER_LOAD_DWORDX2_ADDR64, i64,
-                          global_load, constant_load>;
-defm : MUBUFLoad_Pattern <BUFFER_LOAD_DWORDX2_ADDR64, i64,
-                          az_extloadi32_global, az_extloadi32_constant>;
-defm : MUBUFLoad_Pattern <BUFFER_LOAD_DWORDX2_ADDR64, v2i32,
-                          global_load, constant_load>;
-defm : MUBUFLoad_Pattern <BUFFER_LOAD_DWORDX4_ADDR64, v4i32,
-                          global_load, constant_load>;
-
-multiclass MUBUFStore_Pattern <MUBUF Instr, ValueType vt, PatFrag st> {
 
   def : Pat <
-    (st vt:$value, i64:$ptr),
-    (Instr $value, (SI_ADDR64_RSRC (i64 0)), $ptr, 0)
+    (vt (int_SI_buffer_load_dword v4i32:$rsrc, i32:$vaddr, i32:$soffset,
+                                  imm:$offset, 0, 1, imm:$glc, imm:$slc,
+                                  imm:$tfe)),
+    (idxen $rsrc, $vaddr, (as_i16imm $offset), $soffset, (as_i1imm $glc),
+           (as_i1imm $slc), (as_i1imm $tfe))
   >;
 
   def : Pat <
-    (st vt:$value, (add i64:$ptr, i64:$offset)),
-    (Instr $value, (SI_ADDR64_RSRC $ptr), $offset, 0)
-   >;
+    (vt (int_SI_buffer_load_dword v4i32:$rsrc, v2i32:$vaddr, i32:$soffset,
+                                  imm, 1, 1, imm:$glc, imm:$slc,
+                                  imm:$tfe)),
+    (bothen $rsrc, $vaddr, $soffset, (as_i1imm $glc), (as_i1imm $slc),
+            (as_i1imm $tfe))
+  >;
 }
 
-defm : MUBUFStore_Pattern <BUFFER_STORE_BYTE, i32, truncstorei8_global>;
-defm : MUBUFStore_Pattern <BUFFER_STORE_SHORT, i32, truncstorei16_global>;
-defm : MUBUFStore_Pattern <BUFFER_STORE_DWORD, i32, global_store>;
-defm : MUBUFStore_Pattern <BUFFER_STORE_DWORDX2, i64, global_store>;
-defm : MUBUFStore_Pattern <BUFFER_STORE_DWORDX2, v2i32, global_store>;
-defm : MUBUFStore_Pattern <BUFFER_STORE_DWORDX4, v4i32, global_store>;
+defm : MUBUF_Load_Dword <i32, BUFFER_LOAD_DWORD_OFFSET, BUFFER_LOAD_DWORD_OFFEN,
+                         BUFFER_LOAD_DWORD_IDXEN, BUFFER_LOAD_DWORD_BOTHEN>;
+defm : MUBUF_Load_Dword <v2i32, BUFFER_LOAD_DWORDX2_OFFSET, BUFFER_LOAD_DWORDX2_OFFEN,
+                         BUFFER_LOAD_DWORDX2_IDXEN, BUFFER_LOAD_DWORDX2_BOTHEN>;
+defm : MUBUF_Load_Dword <v4i32, BUFFER_LOAD_DWORDX4_OFFSET, BUFFER_LOAD_DWORDX4_OFFEN,
+                         BUFFER_LOAD_DWORDX4_IDXEN, BUFFER_LOAD_DWORDX4_BOTHEN>;
+
+class MUBUFScratchStorePat <MUBUF Instr, ValueType vt, PatFrag st> : Pat <
+  (st vt:$value, (MUBUFScratch v4i32:$srsrc, i32:$vaddr, i32:$soffset,
+                               u16imm:$offset)),
+  (Instr $value, $srsrc, $vaddr, $soffset, $offset, 0, 0, 0)
+>;
+
+def : MUBUFScratchStorePat <BUFFER_STORE_BYTE_OFFEN, i32, truncstorei8_private>;
+def : MUBUFScratchStorePat <BUFFER_STORE_SHORT_OFFEN, i32, truncstorei16_private>;
+def : MUBUFScratchStorePat <BUFFER_STORE_DWORD_OFFEN, i32, store_private>;
+def : MUBUFScratchStorePat <BUFFER_STORE_DWORDX2_OFFEN, v2i32, store_private>;
+def : MUBUFScratchStorePat <BUFFER_STORE_DWORDX4_OFFEN, v4i32, store_private>;
+
+/*
+class MUBUFStore_Pattern <MUBUF Instr, ValueType vt, PatFrag st> : Pat <
+  (st vt:$value, (MUBUFScratch v4i32:$srsrc, i64:$vaddr, u16imm:$offset)),
+  (Instr $value, $srsrc, $vaddr, $offset)
+>;
+
+def : MUBUFStore_Pattern <BUFFER_STORE_BYTE_ADDR64, i32, truncstorei8_private>;
+def : MUBUFStore_Pattern <BUFFER_STORE_SHORT_ADDR64, i32, truncstorei16_private>;
+def : MUBUFStore_Pattern <BUFFER_STORE_DWORD_ADDR64, i32, store_private>;
+def : MUBUFStore_Pattern <BUFFER_STORE_DWORDX2_ADDR64, v2i32, store_private>;
+def : MUBUFStore_Pattern <BUFFER_STORE_DWORDX4_ADDR64, v4i32, store_private>;
+
+*/
 
 //===----------------------------------------------------------------------===//
 // MTBUF Patterns
@@ -1900,7 +2809,7 @@ defm : MUBUFStore_Pattern <BUFFER_STORE_DWORDX4, v4i32, global_store>;
 
 // TBUFFER_STORE_FORMAT_*, addr64=0
 class MTBUF_StoreResource <ValueType vt, int num_channels, MTBUF opcode> : Pat<
-  (SItbuffer_store i128:$rsrc, vt:$vdata, num_channels, i32:$vaddr,
+  (SItbuffer_store v4i32:$rsrc, vt:$vdata, num_channels, i32:$vaddr,
                    i32:$soffset, imm:$inst_offset, imm:$dfmt,
                    imm:$nfmt, imm:$offen, imm:$idxen,
                    imm:$glc, imm:$slc, imm:$tfe),
@@ -1915,79 +2824,250 @@ def : MTBUF_StoreResource <v2i32, 2, TBUFFER_STORE_FORMAT_XY>;
 def : MTBUF_StoreResource <v4i32, 3, TBUFFER_STORE_FORMAT_XYZ>;
 def : MTBUF_StoreResource <v4i32, 4, TBUFFER_STORE_FORMAT_XYZW>;
 
+let SubtargetPredicate = isCI in {
+
+// Sea island new arithmetic instructinos
+defm V_TRUNC_F64 : VOP1Inst <vop1<0x17>, "V_TRUNC_F64",
+  VOP_F64_F64, ftrunc
+>;
+defm V_CEIL_F64 : VOP1Inst <vop1<0x18>, "V_CEIL_F64",
+  VOP_F64_F64, fceil
+>;
+defm V_FLOOR_F64 : VOP1Inst <vop1<0x1A>, "V_FLOOR_F64",
+  VOP_F64_F64, ffloor
+>;
+defm V_RNDNE_F64 : VOP1Inst <vop1<0x19>, "V_RNDNE_F64",
+  VOP_F64_F64, frint
+>;
+
+defm V_QSAD_PK_U16_U8 : VOP3Inst <vop3<0x173>, "V_QSAD_PK_U16_U8",
+  VOP_I32_I32_I32
+>;
+defm V_MQSAD_U16_U8 : VOP3Inst <vop3<0x172>, "V_MQSAD_U16_U8",
+  VOP_I32_I32_I32
+>;
+defm V_MQSAD_U32_U8 : VOP3Inst <vop3<0x175>, "V_MQSAD_U32_U8",
+  VOP_I32_I32_I32
+>;
+defm V_MAD_U64_U32 : VOP3Inst <vop3<0x176>, "V_MAD_U64_U32",
+  VOP_I64_I32_I32_I64
+>;
+
+// XXX - Does this set VCC?
+defm V_MAD_I64_I32 : VOP3Inst <vop3<0x177>, "V_MAD_I64_I32",
+  VOP_I64_I32_I32_I64
+>;
+
+// Remaining instructions:
+// FLAT_*
+// S_CBRANCH_CDBGUSER
+// S_CBRANCH_CDBGSYS
+// S_CBRANCH_CDBGSYS_OR_USER
+// S_CBRANCH_CDBGSYS_AND_USER
+// S_DCACHE_INV_VOL
+// V_EXP_LEGACY_F32
+// V_LOG_LEGACY_F32
+// DS_NOP
+// DS_GWS_SEMA_RELEASE_ALL
+// DS_WRAP_RTN_B32
+// DS_CNDXCHG32_RTN_B64
+// DS_WRITE_B96
+// DS_WRITE_B128
+// DS_CONDXCHG32_RTN_B128
+// DS_READ_B96
+// DS_READ_B128
+// BUFFER_LOAD_DWORDX3
+// BUFFER_STORE_DWORDX3
+
+} // End iSCI
+
+//===----------------------------------------------------------------------===//
+// Flat Patterns
+//===----------------------------------------------------------------------===//
+
+class FLATLoad_Pattern <FLAT Instr_ADDR64, ValueType vt,
+                             PatFrag flat_ld> :
+  Pat <(vt (flat_ld i64:$ptr)),
+       (Instr_ADDR64 $ptr)
+>;
+
+def : FLATLoad_Pattern <FLAT_LOAD_SBYTE, i32, sextloadi8_flat>;
+def : FLATLoad_Pattern <FLAT_LOAD_UBYTE, i32, az_extloadi8_flat>;
+def : FLATLoad_Pattern <FLAT_LOAD_SSHORT, i32, sextloadi16_flat>;
+def : FLATLoad_Pattern <FLAT_LOAD_USHORT, i32, az_extloadi16_flat>;
+def : FLATLoad_Pattern <FLAT_LOAD_DWORD, i32, flat_load>;
+def : FLATLoad_Pattern <FLAT_LOAD_DWORDX2, i64, flat_load>;
+def : FLATLoad_Pattern <FLAT_LOAD_DWORDX2, i64, az_extloadi32_flat>;
+def : FLATLoad_Pattern <FLAT_LOAD_DWORDX2, v2i32, flat_load>;
+def : FLATLoad_Pattern <FLAT_LOAD_DWORDX4, v4i32, flat_load>;
+
+class FLATStore_Pattern <FLAT Instr, ValueType vt, PatFrag st> :
+  Pat <(st vt:$value, i64:$ptr),
+        (Instr $value, $ptr)
+  >;
+
+def : FLATStore_Pattern <FLAT_STORE_BYTE, i32, truncstorei8_flat>;
+def : FLATStore_Pattern <FLAT_STORE_SHORT, i32, truncstorei16_flat>;
+def : FLATStore_Pattern <FLAT_STORE_DWORD, i32, flat_store>;
+def : FLATStore_Pattern <FLAT_STORE_DWORDX2, i64, flat_store>;
+def : FLATStore_Pattern <FLAT_STORE_DWORDX2, v2i32, flat_store>;
+def : FLATStore_Pattern <FLAT_STORE_DWORDX4, v4i32, flat_store>;
+
 /********** ====================== **********/
 /**********   Indirect adressing   **********/
 /********** ====================== **********/
 
-multiclass SI_INDIRECT_Pattern <ValueType vt, SI_INDIRECT_DST IndDst> {
+multiclass SI_INDIRECT_Pattern <ValueType vt, ValueType eltvt, SI_INDIRECT_DST IndDst> {
 
   // 1. Extract with offset
   def : Pat<
     (vector_extract vt:$vec, (add i32:$idx, imm:$off)),
-    (f32 (SI_INDIRECT_SRC (IMPLICIT_DEF), $vec, $idx, imm:$off))
+    (eltvt (SI_INDIRECT_SRC (IMPLICIT_DEF), $vec, $idx, imm:$off))
   >;
 
   // 2. Extract without offset
   def : Pat<
     (vector_extract vt:$vec, i32:$idx),
-    (f32 (SI_INDIRECT_SRC (IMPLICIT_DEF), $vec, $idx, 0))
+    (eltvt (SI_INDIRECT_SRC (IMPLICIT_DEF), $vec, $idx, 0))
   >;
 
   // 3. Insert with offset
   def : Pat<
-    (vector_insert vt:$vec, f32:$val, (add i32:$idx, imm:$off)),
+    (vector_insert vt:$vec, eltvt:$val, (add i32:$idx, imm:$off)),
     (IndDst (IMPLICIT_DEF), $vec, $idx, imm:$off, $val)
   >;
 
   // 4. Insert without offset
   def : Pat<
-    (vector_insert vt:$vec, f32:$val, i32:$idx),
+    (vector_insert vt:$vec, eltvt:$val, i32:$idx),
     (IndDst (IMPLICIT_DEF), $vec, $idx, 0, $val)
   >;
 }
 
-defm : SI_INDIRECT_Pattern <v2f32, SI_INDIRECT_DST_V2>;
-defm : SI_INDIRECT_Pattern <v4f32, SI_INDIRECT_DST_V4>;
-defm : SI_INDIRECT_Pattern <v8f32, SI_INDIRECT_DST_V8>;
-defm : SI_INDIRECT_Pattern <v16f32, SI_INDIRECT_DST_V16>;
+defm : SI_INDIRECT_Pattern <v2f32, f32, SI_INDIRECT_DST_V2>;
+defm : SI_INDIRECT_Pattern <v4f32, f32, SI_INDIRECT_DST_V4>;
+defm : SI_INDIRECT_Pattern <v8f32, f32, SI_INDIRECT_DST_V8>;
+defm : SI_INDIRECT_Pattern <v16f32, f32, SI_INDIRECT_DST_V16>;
 
-/********** =============== **********/
-/**********   Conditions    **********/
-/********** =============== **********/
+defm : SI_INDIRECT_Pattern <v2i32, i32, SI_INDIRECT_DST_V2>;
+defm : SI_INDIRECT_Pattern <v4i32, i32, SI_INDIRECT_DST_V4>;
+defm : SI_INDIRECT_Pattern <v8i32, i32, SI_INDIRECT_DST_V8>;
+defm : SI_INDIRECT_Pattern <v16i32, i32, SI_INDIRECT_DST_V16>;
 
-def : Pat<
-  (i1 (setcc f32:$src0, f32:$src1, SETO)),
-  (V_CMP_O_F32_e64 $src0, $src1)
+//===----------------------------------------------------------------------===//
+// Conversion Patterns
+//===----------------------------------------------------------------------===//
+
+def : Pat<(i32 (sext_inreg i32:$src, i1)),
+  (S_BFE_I32 i32:$src, 65536)>; // 0 | 1 << 16
+
+// TODO: Match 64-bit BFE. SI has a 64-bit BFE, but it's scalar only so it
+// might not be worth the effort, and will need to expand to shifts when
+// fixing SGPR copies.
+
+// Handle sext_inreg in i64
+def : Pat <
+  (i64 (sext_inreg i64:$src, i1)),
+  (INSERT_SUBREG (INSERT_SUBREG (i64 (IMPLICIT_DEF)),
+    (S_BFE_I32 (EXTRACT_SUBREG i64:$src, sub0), 65536), sub0), // 0 | 1 << 16
+    (S_MOV_B32 -1), sub1)
 >;
 
-def : Pat<
-  (i1 (setcc f32:$src0, f32:$src1, SETUO)),
-  (V_CMP_U_F32_e64 $src0, $src1)
+def : Pat <
+  (i64 (sext_inreg i64:$src, i8)),
+  (INSERT_SUBREG (INSERT_SUBREG (i64 (IMPLICIT_DEF)),
+    (S_SEXT_I32_I8 (EXTRACT_SUBREG i64:$src, sub0)), sub0),
+    (S_MOV_B32 -1), sub1)
 >;
 
-//============================================================================//
+def : Pat <
+  (i64 (sext_inreg i64:$src, i16)),
+  (INSERT_SUBREG (INSERT_SUBREG (i64 (IMPLICIT_DEF)),
+    (S_SEXT_I32_I16 (EXTRACT_SUBREG i64:$src, sub0)), sub0),
+    (S_MOV_B32 -1), sub1)
+>;
+
+class ZExt_i64_i32_Pat <SDNode ext> : Pat <
+  (i64 (ext i32:$src)),
+  (INSERT_SUBREG (INSERT_SUBREG (i64 (IMPLICIT_DEF)), $src, sub0),
+    (S_MOV_B32 0), sub1)
+>;
+
+class ZExt_i64_i1_Pat <SDNode ext> : Pat <
+  (i64 (ext i1:$src)),
+  (INSERT_SUBREG
+    (INSERT_SUBREG (i64 (IMPLICIT_DEF)),
+      (V_CNDMASK_B32_e64 (i32 0), (i32 1), $src), sub0),
+    (S_MOV_B32 0), sub1)
+>;
+
+
+def : ZExt_i64_i32_Pat<zext>;
+def : ZExt_i64_i32_Pat<anyext>;
+def : ZExt_i64_i1_Pat<zext>;
+def : ZExt_i64_i1_Pat<anyext>;
+
+def : Pat <
+  (i64 (sext i32:$src)),
+    (INSERT_SUBREG
+      (INSERT_SUBREG (i64 (IMPLICIT_DEF)), $src, sub0),
+      (S_ASHR_I32 $src, 31), sub1)
+>;
+
+def : Pat <
+  (i64 (sext i1:$src)),
+  (INSERT_SUBREG
+    (INSERT_SUBREG
+      (i64 (IMPLICIT_DEF)),
+      (V_CNDMASK_B32_e64 0, -1, $src), sub0),
+    (V_CNDMASK_B32_e64 0, -1, $src), sub1)
+>;
+
+def : Pat <
+  (f32 (sint_to_fp i1:$src)),
+  (V_CNDMASK_B32_e64 (i32 0), CONST.FP32_NEG_ONE, $src)
+>;
+
+def : Pat <
+  (f32 (uint_to_fp i1:$src)),
+  (V_CNDMASK_B32_e64 (i32 0), CONST.FP32_ONE, $src)
+>;
+
+def : Pat <
+  (f64 (sint_to_fp i1:$src)),
+    (V_CVT_F64_I32_e32 (V_CNDMASK_B32_e64 (i32 0), (i32 -1), $src))
+>;
+
+def : Pat <
+  (f64 (uint_to_fp i1:$src)),
+  (V_CVT_F64_U32_e32 (V_CNDMASK_B32_e64 (i32 0), (i32 1), $src))
+>;
+
+//===----------------------------------------------------------------------===//
 // Miscellaneous Patterns
 //===----------------------------------------------------------------------===//
 
 def : Pat <
-  (i64 (trunc i128:$x)),
-  (INSERT_SUBREG (INSERT_SUBREG (i64 (IMPLICIT_DEF)),
-    (i32 (EXTRACT_SUBREG $x, sub0)), sub0),
-    (i32 (EXTRACT_SUBREG $x, sub1)), sub1)
+  (i32 (trunc i64:$a)),
+  (EXTRACT_SUBREG $a, sub0)
 >;
 
 def : Pat <
-  (or i64:$a, i64:$b),
-  (INSERT_SUBREG
-    (INSERT_SUBREG (IMPLICIT_DEF),
-      (V_OR_B32_e32 (EXTRACT_SUBREG $a, sub0), (EXTRACT_SUBREG $b, sub0)), sub0),
-    (V_OR_B32_e32 (EXTRACT_SUBREG $a, sub1), (EXTRACT_SUBREG $b, sub1)), sub1)
+  (i1 (trunc i32:$a)),
+  (V_CMP_EQ_I32_e64 (V_AND_B32_e64 (i32 1), $a), 1)
+>;
+
+def : Pat <
+  (i32 (bswap i32:$a)),
+  (V_BFI_B32 (S_MOV_B32 0x00ff00ff),
+             (V_ALIGNBIT_B32 $a, $a, 24),
+             (V_ALIGNBIT_B32 $a, $a, 8))
 >;
 
 //============================================================================//
 // Miscellaneous Optimization Patterns
 //============================================================================//
 
-def : SHA256MaPattern <V_BFI_B32, V_XOR_B32_e32>;
+def : SHA256MaPattern <V_BFI_B32, V_XOR_B32_e64>;
 
 } // End isSI predicate