[llvm] 409c443 - [AMDGPU] Validate dst and src2 non-overlapping restriction in asm

Stanislav Mekhanoshin via llvm-commits llvm-commits at lists.llvm.org
Wed Jan 26 15:14:14 PST 2022


Author: Stanislav Mekhanoshin
Date: 2022-01-26T15:14:06-08:00
New Revision: 409c4436f9564883cac324b4371ec2b0919f721b

URL: https://github.com/llvm/llvm-project/commit/409c4436f9564883cac324b4371ec2b0919f721b
DIFF: https://github.com/llvm/llvm-project/commit/409c4436f9564883cac324b4371ec2b0919f721b.diff

LOG: [AMDGPU] Validate dst and src2 non-overlapping restriction in asm

Differential Revision: https://reviews.llvm.org/D118089

Added: 
    

Modified: 
    llvm/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp
    llvm/test/MC/AMDGPU/accvgpr-altnames.s
    llvm/test/MC/AMDGPU/gfx90a_err_pos.s
    llvm/test/MC/AMDGPU/mai-err.s
    llvm/test/MC/AMDGPU/mai-gfx90a.s
    llvm/test/MC/AMDGPU/mai.s
    llvm/test/MC/AMDGPU/misaligned-vgpr-tuples-err.s

Removed: 
    


################################################################################
diff  --git a/llvm/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp b/llvm/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp
index f6c54fe3e7bf9..c1c88d9a74626 100644
--- a/llvm/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp
+++ b/llvm/lib/Target/AMDGPU/AsmParser/AMDGPUAsmParser.cpp
@@ -1548,6 +1548,7 @@ class AMDGPUAsmParser : public MCTargetAsmParser {
   bool validateVccOperand(unsigned Reg) const;
   bool validateVOPLiteral(const MCInst &Inst, const OperandVector &Operands);
   bool validateMAIAccWrite(const MCInst &Inst, const OperandVector &Operands);
+  bool validateMFMA(const MCInst &Inst, const OperandVector &Operands);
   bool validateAGPRLdSt(const MCInst &Inst) const;
   bool validateVGPRAlign(const MCInst &Inst) const;
   bool validateGWS(const MCInst &Inst, const OperandVector &Operands);
@@ -3613,6 +3614,40 @@ bool AMDGPUAsmParser::validateMAIAccWrite(const MCInst &Inst,
   return true;
 }
 
+bool AMDGPUAsmParser::validateMFMA(const MCInst &Inst,
+                                   const OperandVector &Operands) {
+  const unsigned Opc = Inst.getOpcode();
+  const MCInstrDesc &Desc = MII.get(Opc);
+
+  if ((Desc.TSFlags & SIInstrFlags::IsMAI) == 0)
+    return true;
+
+  const int Src2Idx = AMDGPU::getNamedOperandIdx(Opc, AMDGPU::OpName::src2);
+  if (Src2Idx == -1)
+    return true;
+
+  const MCOperand &Src2 = Inst.getOperand(Src2Idx);
+  if (!Src2.isReg())
+    return true;
+
+  MCRegister Src2Reg = Src2.getReg();
+  MCRegister DstReg = Inst.getOperand(0).getReg();
+  if (Src2Reg == DstReg)
+    return true;
+
+  const MCRegisterInfo *TRI = getContext().getRegisterInfo();
+  if (TRI->getRegClass(Desc.OpInfo[0].RegClass).getSizeInBits() <= 128)
+    return true;
+
+  if (isRegIntersect(Src2Reg, DstReg, TRI)) {
+    Error(getRegLoc(mc2PseudoReg(Src2Reg), Operands),
+          "source 2 operand must not partially overlap with dst");
+    return false;
+  }
+
+  return true;
+}
+
 bool AMDGPUAsmParser::validateDivScale(const MCInst &Inst) {
   switch (Inst.getOpcode()) {
   default:
@@ -4297,6 +4332,9 @@ bool AMDGPUAsmParser::validateInstruction(const MCInst &Inst,
   if (!validateMAIAccWrite(Inst, Operands)) {
     return false;
   }
+  if (!validateMFMA(Inst, Operands)) {
+    return false;
+  }
   if (!validateCoherencyBits(Inst, Operands, IDLoc)) {
     return false;
   }

diff  --git a/llvm/test/MC/AMDGPU/accvgpr-altnames.s b/llvm/test/MC/AMDGPU/accvgpr-altnames.s
index cc52162480a5d..b0445f0192860 100644
--- a/llvm/test/MC/AMDGPU/accvgpr-altnames.s
+++ b/llvm/test/MC/AMDGPU/accvgpr-altnames.s
@@ -6,5 +6,5 @@ v_accvgpr_read_b32 v2, acc0
 v_accvgpr_write_b32 acc2, -2.0
 // GFX908: v_accvgpr_write_b32 a2, -2.0    ; encoding: [0x02,0x40,0xd9,0xd3,0xf5,0x00,0x00,0x18]
 
-v_mfma_f32_32x32x1f32 acc[0:31], acc0, acc1, acc[1:32]
-// GFX908: v_mfma_f32_32x32x1f32 a[0:31], a0, a1, a[1:32] ; encoding: [0x00,0x00,0xc0,0xd3,0x00,0x03,0x06,0x1c]
+v_mfma_f32_32x32x1f32 acc[0:31], acc0, acc1, acc[32:63]
+// GFX908: v_mfma_f32_32x32x1f32 a[0:31], a0, a1, a[32:63] ; encoding: [0x00,0x00,0xc0,0xd3,0x00,0x03,0x82,0x1c]

diff  --git a/llvm/test/MC/AMDGPU/gfx90a_err_pos.s b/llvm/test/MC/AMDGPU/gfx90a_err_pos.s
index e7377c51df13e..69f4f1ad5d5de 100644
--- a/llvm/test/MC/AMDGPU/gfx90a_err_pos.s
+++ b/llvm/test/MC/AMDGPU/gfx90a_err_pos.s
@@ -7,3 +7,8 @@ ds_gws_init a1 offset:65535 gds
 // CHECK: error: vgpr must be even aligned
 // CHECK-NEXT:{{^}}ds_gws_init a1 offset:65535 gds
 // CHECK-NEXT:{{^}}            ^
+
+v_mfma_f64_16x16x4f64 v[0:7], v[0:1], v[2:3], v[2:9]
+// CHECK: error: source 2 operand must not partially overlap with dst
+// CHECK-NEXT:{{^}}v_mfma_f64_16x16x4f64 v[0:7], v[0:1], v[2:3], v[2:9]
+// CHECK-NEXT:{{^}}                                              ^

diff  --git a/llvm/test/MC/AMDGPU/mai-err.s b/llvm/test/MC/AMDGPU/mai-err.s
index 5a1c6e7045a52..9048583569193 100644
--- a/llvm/test/MC/AMDGPU/mai-err.s
+++ b/llvm/test/MC/AMDGPU/mai-err.s
@@ -39,19 +39,19 @@ v_accvgpr_write_b32 a0, 65
 v_accvgpr_write_b32 a0, v0
 // GFX900: error: instruction not supported on this GPU
 
-v_mfma_f32_32x32x1f32 v[0:31], v0, v1, a[1:32]
+v_mfma_f32_32x32x1f32 v[0:31], v0, v1, a[0:31]
 // GFX908: error: invalid operand for instruction
 // GFX900: error: instruction not supported on this GPU
 
-v_mfma_f32_32x32x1f32 a[0:31], v0, v1, v[1:32]
+v_mfma_f32_32x32x1f32 a[0:31], v0, v1, v[0:31]
 // GFX908: error: invalid operand for instruction
 // GFX900: error: instruction not supported on this GPU
 
-v_mfma_f32_32x32x1f32 a[0:31], s0, v1, a[1:32]
+v_mfma_f32_32x32x1f32 a[0:31], s0, v1, a[0:31]
 // GFX908: error: invalid operand for instruction
 // GFX900: error: instruction not supported on this GPU
 
-v_mfma_f32_32x32x1f32 a[0:31], 1, v1, a[1:32]
+v_mfma_f32_32x32x1f32 a[0:31], 1, v1, a[0:31]
 // GFX908: error: invalid operand for instruction
 // GFX900: error: instruction not supported on this GPU
 
@@ -698,3 +698,11 @@ v_mfma_f32_16x16x8bf16 a[0:3], a0, a1, -2.0
 v_mfma_f32_16x16x8bf16 a[0:3], a0, a1, -2.0 cbsz:3 abid:2 blgp:7
 // GFX908: error: inline constants are not allowed for this operand
 // GFX900: error: instruction not supported on this GPU
+
+v_mfma_f32_32x32x1f32 a[0:31], v0, v1, a[2:33]
+// GFX908: error: source 2 operand must not partially overlap with dst
+// GFX900: error: instruction not supported on this GPU
+
+v_mfma_f32_16x16x1f32 a[0:15], v0, v1, a[2:17]
+// GFX908: error: source 2 operand must not partially overlap with dst
+// GFX900: error: instruction not supported on this GPU

diff  --git a/llvm/test/MC/AMDGPU/mai-gfx90a.s b/llvm/test/MC/AMDGPU/mai-gfx90a.s
index 14cec6a05730e..db116f8a20cd2 100644
--- a/llvm/test/MC/AMDGPU/mai-gfx90a.s
+++ b/llvm/test/MC/AMDGPU/mai-gfx90a.s
@@ -27,53 +27,53 @@ v_accvgpr_write a2, v255
 v_accvgpr_mov_b32 a1, a2
 // GFX90A: v_accvgpr_mov_b32 a1, a2        ; encoding: [0x02,0xa5,0x02,0x7e]
 
-v_mfma_f32_32x32x1f32 a[0:31], v0, v1, a[2:33]
-// GFX90A: v_mfma_f32_32x32x1f32 a[0:31], v0, v1, a[2:33] ; encoding: [0x00,0x80,0xc0,0xd3,0x00,0x03,0x0a,0x04]
+v_mfma_f32_32x32x1f32 a[0:31], v0, v1, a[34:65]
+// GFX90A: v_mfma_f32_32x32x1f32 a[0:31], v0, v1, a[34:65] ; encoding: [0x00,0x80,0xc0,0xd3,0x00,0x03,0x8a,0x04]
 
-v_mfma_f32_32x32x1f32 a[0:31], v0, v1, a[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x1f32 a[0:31], v0, v1, a[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc0,0xd3,0x00,0x03,0x0a,0xe4]
+v_mfma_f32_32x32x1f32 a[0:31], v0, v1, a[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x1f32 a[0:31], v0, v1, a[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc0,0xd3,0x00,0x03,0x8a,0xe4]
 
-v_mfma_f32_32x32x1f32 a[0:31], v0, a1, a[2:33]
-// GFX90A: v_mfma_f32_32x32x1f32 a[0:31], v0, a1, a[2:33] ; encoding: [0x00,0x80,0xc0,0xd3,0x00,0x03,0x0a,0x14]
+v_mfma_f32_32x32x1f32 a[0:31], v0, a1, a[34:65]
+// GFX90A: v_mfma_f32_32x32x1f32 a[0:31], v0, a1, a[34:65] ; encoding: [0x00,0x80,0xc0,0xd3,0x00,0x03,0x8a,0x14]
 
-v_mfma_f32_32x32x1f32 a[0:31], v0, a1, a[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x1f32 a[0:31], v0, a1, a[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc0,0xd3,0x00,0x03,0x0a,0xf4]
+v_mfma_f32_32x32x1f32 a[0:31], v0, a1, a[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x1f32 a[0:31], v0, a1, a[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc0,0xd3,0x00,0x03,0x8a,0xf4]
 
-v_mfma_f32_32x32x1f32 a[0:31], a0, v1, a[2:33]
-// GFX90A: v_mfma_f32_32x32x1f32 a[0:31], a0, v1, a[2:33] ; encoding: [0x00,0x80,0xc0,0xd3,0x00,0x03,0x0a,0x0c]
+v_mfma_f32_32x32x1f32 a[0:31], a0, v1, a[34:65]
+// GFX90A: v_mfma_f32_32x32x1f32 a[0:31], a0, v1, a[34:65] ; encoding: [0x00,0x80,0xc0,0xd3,0x00,0x03,0x8a,0x0c]
 
-v_mfma_f32_32x32x1f32 a[0:31], a0, v1, a[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x1f32 a[0:31], a0, v1, a[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc0,0xd3,0x00,0x03,0x0a,0xec]
+v_mfma_f32_32x32x1f32 a[0:31], a0, v1, a[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x1f32 a[0:31], a0, v1, a[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc0,0xd3,0x00,0x03,0x8a,0xec]
 
-v_mfma_f32_32x32x1f32 a[0:31], a0, a1, a[2:33]
-// GFX90A: v_mfma_f32_32x32x1f32 a[0:31], a0, a1, a[2:33] ; encoding: [0x00,0x80,0xc0,0xd3,0x00,0x03,0x0a,0x1c]
+v_mfma_f32_32x32x1f32 a[0:31], a0, a1, a[34:65]
+// GFX90A: v_mfma_f32_32x32x1f32 a[0:31], a0, a1, a[34:65] ; encoding: [0x00,0x80,0xc0,0xd3,0x00,0x03,0x8a,0x1c]
 
-v_mfma_f32_32x32x1f32 a[0:31], a0, a1, a[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x1f32 a[0:31], a0, a1, a[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc0,0xd3,0x00,0x03,0x0a,0xfc]
+v_mfma_f32_32x32x1f32 a[0:31], a0, a1, a[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x1f32 a[0:31], a0, a1, a[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc0,0xd3,0x00,0x03,0x8a,0xfc]
 
-v_mfma_f32_32x32x1f32 v[0:31], v0, v1, v[2:33]
-// GFX90A: v_mfma_f32_32x32x1f32 v[0:31], v0, v1, v[2:33] ; encoding: [0x00,0x00,0xc0,0xd3,0x00,0x03,0x0a,0x04]
+v_mfma_f32_32x32x1f32 v[0:31], v0, v1, v[34:65]
+// GFX90A: v_mfma_f32_32x32x1f32 v[0:31], v0, v1, v[34:65] ; encoding: [0x00,0x00,0xc0,0xd3,0x00,0x03,0x8a,0x04]
 
-v_mfma_f32_32x32x1f32 v[0:31], v0, v1, v[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x1f32 v[0:31], v0, v1, v[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc0,0xd3,0x00,0x03,0x0a,0xe4]
+v_mfma_f32_32x32x1f32 v[0:31], v0, v1, v[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x1f32 v[0:31], v0, v1, v[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc0,0xd3,0x00,0x03,0x8a,0xe4]
 
-v_mfma_f32_32x32x1f32 v[0:31], v0, a1, v[2:33]
-// GFX90A: v_mfma_f32_32x32x1f32 v[0:31], v0, a1, v[2:33] ; encoding: [0x00,0x00,0xc0,0xd3,0x00,0x03,0x0a,0x14]
+v_mfma_f32_32x32x1f32 v[0:31], v0, a1, v[34:65]
+// GFX90A: v_mfma_f32_32x32x1f32 v[0:31], v0, a1, v[34:65] ; encoding: [0x00,0x00,0xc0,0xd3,0x00,0x03,0x8a,0x14]
 
-v_mfma_f32_32x32x1f32 v[0:31], v0, a1, v[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x1f32 v[0:31], v0, a1, v[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc0,0xd3,0x00,0x03,0x0a,0xf4]
+v_mfma_f32_32x32x1f32 v[0:31], v0, a1, v[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x1f32 v[0:31], v0, a1, v[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc0,0xd3,0x00,0x03,0x8a,0xf4]
 
-v_mfma_f32_32x32x1f32 v[0:31], a0, v1, v[2:33]
-// GFX90A: v_mfma_f32_32x32x1f32 v[0:31], a0, v1, v[2:33] ; encoding: [0x00,0x00,0xc0,0xd3,0x00,0x03,0x0a,0x0c]
+v_mfma_f32_32x32x1f32 v[0:31], a0, v1, v[34:65]
+// GFX90A: v_mfma_f32_32x32x1f32 v[0:31], a0, v1, v[34:65] ; encoding: [0x00,0x00,0xc0,0xd3,0x00,0x03,0x8a,0x0c]
 
-v_mfma_f32_32x32x1f32 v[0:31], a0, v1, v[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x1f32 v[0:31], a0, v1, v[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc0,0xd3,0x00,0x03,0x0a,0xec]
+v_mfma_f32_32x32x1f32 v[0:31], a0, v1, v[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x1f32 v[0:31], a0, v1, v[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc0,0xd3,0x00,0x03,0x8a,0xec]
 
-v_mfma_f32_32x32x1f32 v[0:31], a0, a1, v[2:33]
-// GFX90A: v_mfma_f32_32x32x1f32 v[0:31], a0, a1, v[2:33] ; encoding: [0x00,0x00,0xc0,0xd3,0x00,0x03,0x0a,0x1c]
+v_mfma_f32_32x32x1f32 v[0:31], a0, a1, v[34:65]
+// GFX90A: v_mfma_f32_32x32x1f32 v[0:31], a0, a1, v[34:65] ; encoding: [0x00,0x00,0xc0,0xd3,0x00,0x03,0x8a,0x1c]
 
-v_mfma_f32_32x32x1f32 v[0:31], a0, a1, v[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x1f32 v[0:31], a0, a1, v[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc0,0xd3,0x00,0x03,0x0a,0xfc]
+v_mfma_f32_32x32x1f32 v[0:31], a0, a1, v[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x1f32 v[0:31], a0, a1, v[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc0,0xd3,0x00,0x03,0x8a,0xfc]
 
 v_mfma_f32_32x32x1f32 a[0:31], v0, v1, -2.0
 // GFX90A: v_mfma_f32_32x32x1f32 a[0:31], v0, v1, -2.0 ; encoding: [0x00,0x80,0xc0,0xd3,0x00,0x03,0xd6,0x03]
@@ -123,53 +123,53 @@ v_mfma_f32_32x32x1f32 a[0:31], a0, a1, -2.0 cbsz:3 abid:2 blgp:7
 v_mfma_f32_32x32x1f32 v[0:31], a0, a1, -2.0 cbsz:3 abid:2 blgp:7
 // GFX90A: v_mfma_f32_32x32x1f32 v[0:31], a0, a1, -2.0 cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc0,0xd3,0x00,0x03,0xd6,0xfb]
 
-v_mfma_f32_16x16x1f32 a[0:15], v0, v1, a[2:17]
-// GFX90A: v_mfma_f32_16x16x1f32 a[0:15], v0, v1, a[2:17] ; encoding: [0x00,0x80,0xc1,0xd3,0x00,0x03,0x0a,0x04]
+v_mfma_f32_16x16x1f32 a[0:15], v0, v1, a[18:33]
+// GFX90A: v_mfma_f32_16x16x1f32 a[0:15], v0, v1, a[18:33] ; encoding: [0x00,0x80,0xc1,0xd3,0x00,0x03,0x4a,0x04]
 
-v_mfma_f32_16x16x1f32 a[0:15], v0, v1, a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x1f32 a[0:15], v0, v1, a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc1,0xd3,0x00,0x03,0x0a,0xe4]
+v_mfma_f32_16x16x1f32 a[0:15], v0, v1, a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x1f32 a[0:15], v0, v1, a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc1,0xd3,0x00,0x03,0x4a,0xe4]
 
-v_mfma_f32_16x16x1f32 a[0:15], v0, a1, a[2:17]
-// GFX90A: v_mfma_f32_16x16x1f32 a[0:15], v0, a1, a[2:17] ; encoding: [0x00,0x80,0xc1,0xd3,0x00,0x03,0x0a,0x14]
+v_mfma_f32_16x16x1f32 a[0:15], v0, a1, a[18:33]
+// GFX90A: v_mfma_f32_16x16x1f32 a[0:15], v0, a1, a[18:33] ; encoding: [0x00,0x80,0xc1,0xd3,0x00,0x03,0x4a,0x14]
 
-v_mfma_f32_16x16x1f32 a[0:15], v0, a1, a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x1f32 a[0:15], v0, a1, a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc1,0xd3,0x00,0x03,0x0a,0xf4]
+v_mfma_f32_16x16x1f32 a[0:15], v0, a1, a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x1f32 a[0:15], v0, a1, a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc1,0xd3,0x00,0x03,0x4a,0xf4]
 
-v_mfma_f32_16x16x1f32 a[0:15], a0, v1, a[2:17]
-// GFX90A: v_mfma_f32_16x16x1f32 a[0:15], a0, v1, a[2:17] ; encoding: [0x00,0x80,0xc1,0xd3,0x00,0x03,0x0a,0x0c]
+v_mfma_f32_16x16x1f32 a[0:15], a0, v1, a[18:33]
+// GFX90A: v_mfma_f32_16x16x1f32 a[0:15], a0, v1, a[18:33] ; encoding: [0x00,0x80,0xc1,0xd3,0x00,0x03,0x4a,0x0c]
 
-v_mfma_f32_16x16x1f32 a[0:15], a0, v1, a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x1f32 a[0:15], a0, v1, a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc1,0xd3,0x00,0x03,0x0a,0xec]
+v_mfma_f32_16x16x1f32 a[0:15], a0, v1, a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x1f32 a[0:15], a0, v1, a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc1,0xd3,0x00,0x03,0x4a,0xec]
 
-v_mfma_f32_16x16x1f32 a[0:15], a0, a1, a[2:17]
-// GFX90A: v_mfma_f32_16x16x1f32 a[0:15], a0, a1, a[2:17] ; encoding: [0x00,0x80,0xc1,0xd3,0x00,0x03,0x0a,0x1c]
+v_mfma_f32_16x16x1f32 a[0:15], a0, a1, a[18:33]
+// GFX90A: v_mfma_f32_16x16x1f32 a[0:15], a0, a1, a[18:33] ; encoding: [0x00,0x80,0xc1,0xd3,0x00,0x03,0x4a,0x1c]
 
-v_mfma_f32_16x16x1f32 a[0:15], a0, a1, a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x1f32 a[0:15], a0, a1, a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc1,0xd3,0x00,0x03,0x0a,0xfc]
+v_mfma_f32_16x16x1f32 a[0:15], a0, a1, a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x1f32 a[0:15], a0, a1, a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc1,0xd3,0x00,0x03,0x4a,0xfc]
 
-v_mfma_f32_16x16x1f32 v[0:15], v0, v1, v[2:17]
-// GFX90A: v_mfma_f32_16x16x1f32 v[0:15], v0, v1, v[2:17] ; encoding: [0x00,0x00,0xc1,0xd3,0x00,0x03,0x0a,0x04]
+v_mfma_f32_16x16x1f32 v[0:15], v0, v1, v[18:33]
+// GFX90A: v_mfma_f32_16x16x1f32 v[0:15], v0, v1, v[18:33] ; encoding: [0x00,0x00,0xc1,0xd3,0x00,0x03,0x4a,0x04]
 
-v_mfma_f32_16x16x1f32 v[0:15], v0, v1, v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x1f32 v[0:15], v0, v1, v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc1,0xd3,0x00,0x03,0x0a,0xe4]
+v_mfma_f32_16x16x1f32 v[0:15], v0, v1, v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x1f32 v[0:15], v0, v1, v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc1,0xd3,0x00,0x03,0x4a,0xe4]
 
-v_mfma_f32_16x16x1f32 v[0:15], v0, a1, v[2:17]
-// GFX90A: v_mfma_f32_16x16x1f32 v[0:15], v0, a1, v[2:17] ; encoding: [0x00,0x00,0xc1,0xd3,0x00,0x03,0x0a,0x14]
+v_mfma_f32_16x16x1f32 v[0:15], v0, a1, v[18:33]
+// GFX90A: v_mfma_f32_16x16x1f32 v[0:15], v0, a1, v[18:33] ; encoding: [0x00,0x00,0xc1,0xd3,0x00,0x03,0x4a,0x14]
 
-v_mfma_f32_16x16x1f32 v[0:15], v0, a1, v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x1f32 v[0:15], v0, a1, v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc1,0xd3,0x00,0x03,0x0a,0xf4]
+v_mfma_f32_16x16x1f32 v[0:15], v0, a1, v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x1f32 v[0:15], v0, a1, v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc1,0xd3,0x00,0x03,0x4a,0xf4]
 
-v_mfma_f32_16x16x1f32 v[0:15], a0, v1, v[2:17]
-// GFX90A: v_mfma_f32_16x16x1f32 v[0:15], a0, v1, v[2:17] ; encoding: [0x00,0x00,0xc1,0xd3,0x00,0x03,0x0a,0x0c]
+v_mfma_f32_16x16x1f32 v[0:15], a0, v1, v[18:33]
+// GFX90A: v_mfma_f32_16x16x1f32 v[0:15], a0, v1, v[18:33] ; encoding: [0x00,0x00,0xc1,0xd3,0x00,0x03,0x4a,0x0c]
 
-v_mfma_f32_16x16x1f32 v[0:15], a0, v1, v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x1f32 v[0:15], a0, v1, v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc1,0xd3,0x00,0x03,0x0a,0xec]
+v_mfma_f32_16x16x1f32 v[0:15], a0, v1, v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x1f32 v[0:15], a0, v1, v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc1,0xd3,0x00,0x03,0x4a,0xec]
 
-v_mfma_f32_16x16x1f32 v[0:15], a0, a1, v[2:17]
-// GFX90A: v_mfma_f32_16x16x1f32 v[0:15], a0, a1, v[2:17] ; encoding: [0x00,0x00,0xc1,0xd3,0x00,0x03,0x0a,0x1c]
+v_mfma_f32_16x16x1f32 v[0:15], a0, a1, v[18:33]
+// GFX90A: v_mfma_f32_16x16x1f32 v[0:15], a0, a1, v[18:33] ; encoding: [0x00,0x00,0xc1,0xd3,0x00,0x03,0x4a,0x1c]
 
-v_mfma_f32_16x16x1f32 v[0:15], a0, a1, v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x1f32 v[0:15], a0, a1, v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc1,0xd3,0x00,0x03,0x0a,0xfc]
+v_mfma_f32_16x16x1f32 v[0:15], a0, a1, v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x1f32 v[0:15], a0, a1, v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc1,0xd3,0x00,0x03,0x4a,0xfc]
 
 v_mfma_f32_16x16x1f32 a[0:15], v0, v1, -2.0
 // GFX90A: v_mfma_f32_16x16x1f32 a[0:15], v0, v1, -2.0 ; encoding: [0x00,0x80,0xc1,0xd3,0x00,0x03,0xd6,0x03]
@@ -315,53 +315,53 @@ v_mfma_f32_4x4x1f32 a[0:3], a0, a1, -2.0 cbsz:3 abid:2 blgp:7
 v_mfma_f32_4x4x1f32 v[0:3], a0, a1, -2.0 cbsz:3 abid:2 blgp:7
 // GFX90A: v_mfma_f32_4x4x1f32 v[0:3], a0, a1, -2.0 cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc2,0xd3,0x00,0x03,0xd6,0xfb]
 
-v_mfma_f32_32x32x2f32 a[0:15], v0, v1, a[2:17]
-// GFX90A: v_mfma_f32_32x32x2f32 a[0:15], v0, v1, a[2:17] ; encoding: [0x00,0x80,0xc4,0xd3,0x00,0x03,0x0a,0x04]
+v_mfma_f32_32x32x2f32 a[0:15], v0, v1, a[18:33]
+// GFX90A: v_mfma_f32_32x32x2f32 a[0:15], v0, v1, a[18:33] ; encoding: [0x00,0x80,0xc4,0xd3,0x00,0x03,0x4a,0x04]
 
-v_mfma_f32_32x32x2f32 a[0:15], v0, v1, a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x2f32 a[0:15], v0, v1, a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc4,0xd3,0x00,0x03,0x0a,0xe4]
+v_mfma_f32_32x32x2f32 a[0:15], v0, v1, a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x2f32 a[0:15], v0, v1, a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc4,0xd3,0x00,0x03,0x4a,0xe4]
 
-v_mfma_f32_32x32x2f32 a[0:15], v0, a1, a[2:17]
-// GFX90A: v_mfma_f32_32x32x2f32 a[0:15], v0, a1, a[2:17] ; encoding: [0x00,0x80,0xc4,0xd3,0x00,0x03,0x0a,0x14]
+v_mfma_f32_32x32x2f32 a[0:15], v0, a1, a[18:33]
+// GFX90A: v_mfma_f32_32x32x2f32 a[0:15], v0, a1, a[18:33] ; encoding: [0x00,0x80,0xc4,0xd3,0x00,0x03,0x4a,0x14]
 
-v_mfma_f32_32x32x2f32 a[0:15], v0, a1, a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x2f32 a[0:15], v0, a1, a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc4,0xd3,0x00,0x03,0x0a,0xf4]
+v_mfma_f32_32x32x2f32 a[0:15], v0, a1, a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x2f32 a[0:15], v0, a1, a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc4,0xd3,0x00,0x03,0x4a,0xf4]
 
-v_mfma_f32_32x32x2f32 a[0:15], a0, v1, a[2:17]
-// GFX90A: v_mfma_f32_32x32x2f32 a[0:15], a0, v1, a[2:17] ; encoding: [0x00,0x80,0xc4,0xd3,0x00,0x03,0x0a,0x0c]
+v_mfma_f32_32x32x2f32 a[0:15], a0, v1, a[18:33]
+// GFX90A: v_mfma_f32_32x32x2f32 a[0:15], a0, v1, a[18:33] ; encoding: [0x00,0x80,0xc4,0xd3,0x00,0x03,0x4a,0x0c]
 
-v_mfma_f32_32x32x2f32 a[0:15], a0, v1, a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x2f32 a[0:15], a0, v1, a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc4,0xd3,0x00,0x03,0x0a,0xec]
+v_mfma_f32_32x32x2f32 a[0:15], a0, v1, a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x2f32 a[0:15], a0, v1, a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc4,0xd3,0x00,0x03,0x4a,0xec]
 
-v_mfma_f32_32x32x2f32 a[0:15], a0, a1, a[2:17]
-// GFX90A: v_mfma_f32_32x32x2f32 a[0:15], a0, a1, a[2:17] ; encoding: [0x00,0x80,0xc4,0xd3,0x00,0x03,0x0a,0x1c]
+v_mfma_f32_32x32x2f32 a[0:15], a0, a1, a[18:33]
+// GFX90A: v_mfma_f32_32x32x2f32 a[0:15], a0, a1, a[18:33] ; encoding: [0x00,0x80,0xc4,0xd3,0x00,0x03,0x4a,0x1c]
 
-v_mfma_f32_32x32x2f32 a[0:15], a0, a1, a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x2f32 a[0:15], a0, a1, a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc4,0xd3,0x00,0x03,0x0a,0xfc]
+v_mfma_f32_32x32x2f32 a[0:15], a0, a1, a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x2f32 a[0:15], a0, a1, a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc4,0xd3,0x00,0x03,0x4a,0xfc]
 
-v_mfma_f32_32x32x2f32 v[0:15], v0, v1, v[2:17]
-// GFX90A: v_mfma_f32_32x32x2f32 v[0:15], v0, v1, v[2:17] ; encoding: [0x00,0x00,0xc4,0xd3,0x00,0x03,0x0a,0x04]
+v_mfma_f32_32x32x2f32 v[0:15], v0, v1, v[18:33]
+// GFX90A: v_mfma_f32_32x32x2f32 v[0:15], v0, v1, v[18:33] ; encoding: [0x00,0x00,0xc4,0xd3,0x00,0x03,0x4a,0x04]
 
-v_mfma_f32_32x32x2f32 v[0:15], v0, v1, v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x2f32 v[0:15], v0, v1, v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc4,0xd3,0x00,0x03,0x0a,0xe4]
+v_mfma_f32_32x32x2f32 v[0:15], v0, v1, v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x2f32 v[0:15], v0, v1, v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc4,0xd3,0x00,0x03,0x4a,0xe4]
 
-v_mfma_f32_32x32x2f32 v[0:15], v0, a1, v[2:17]
-// GFX90A: v_mfma_f32_32x32x2f32 v[0:15], v0, a1, v[2:17] ; encoding: [0x00,0x00,0xc4,0xd3,0x00,0x03,0x0a,0x14]
+v_mfma_f32_32x32x2f32 v[0:15], v0, a1, v[18:33]
+// GFX90A: v_mfma_f32_32x32x2f32 v[0:15], v0, a1, v[18:33] ; encoding: [0x00,0x00,0xc4,0xd3,0x00,0x03,0x4a,0x14]
 
-v_mfma_f32_32x32x2f32 v[0:15], v0, a1, v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x2f32 v[0:15], v0, a1, v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc4,0xd3,0x00,0x03,0x0a,0xf4]
+v_mfma_f32_32x32x2f32 v[0:15], v0, a1, v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x2f32 v[0:15], v0, a1, v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc4,0xd3,0x00,0x03,0x4a,0xf4]
 
-v_mfma_f32_32x32x2f32 v[0:15], a0, v1, v[2:17]
-// GFX90A: v_mfma_f32_32x32x2f32 v[0:15], a0, v1, v[2:17] ; encoding: [0x00,0x00,0xc4,0xd3,0x00,0x03,0x0a,0x0c]
+v_mfma_f32_32x32x2f32 v[0:15], a0, v1, v[18:33]
+// GFX90A: v_mfma_f32_32x32x2f32 v[0:15], a0, v1, v[18:33] ; encoding: [0x00,0x00,0xc4,0xd3,0x00,0x03,0x4a,0x0c]
 
-v_mfma_f32_32x32x2f32 v[0:15], a0, v1, v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x2f32 v[0:15], a0, v1, v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc4,0xd3,0x00,0x03,0x0a,0xec]
+v_mfma_f32_32x32x2f32 v[0:15], a0, v1, v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x2f32 v[0:15], a0, v1, v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc4,0xd3,0x00,0x03,0x4a,0xec]
 
-v_mfma_f32_32x32x2f32 v[0:15], a0, a1, v[2:17]
-// GFX90A: v_mfma_f32_32x32x2f32 v[0:15], a0, a1, v[2:17] ; encoding: [0x00,0x00,0xc4,0xd3,0x00,0x03,0x0a,0x1c]
+v_mfma_f32_32x32x2f32 v[0:15], a0, a1, v[18:33]
+// GFX90A: v_mfma_f32_32x32x2f32 v[0:15], a0, a1, v[18:33] ; encoding: [0x00,0x00,0xc4,0xd3,0x00,0x03,0x4a,0x1c]
 
-v_mfma_f32_32x32x2f32 v[0:15], a0, a1, v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x2f32 v[0:15], a0, a1, v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc4,0xd3,0x00,0x03,0x0a,0xfc]
+v_mfma_f32_32x32x2f32 v[0:15], a0, a1, v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x2f32 v[0:15], a0, a1, v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc4,0xd3,0x00,0x03,0x4a,0xfc]
 
 v_mfma_f32_32x32x2f32 a[0:15], v0, v1, -2.0
 // GFX90A: v_mfma_f32_32x32x2f32 a[0:15], v0, v1, -2.0 ; encoding: [0x00,0x80,0xc4,0xd3,0x00,0x03,0xd6,0x03]
@@ -507,53 +507,53 @@ v_mfma_f32_16x16x4f32 a[0:3], a0, a1, -2.0 cbsz:3 abid:2 blgp:7
 v_mfma_f32_16x16x4f32 v[0:3], a0, a1, -2.0 cbsz:3 abid:2 blgp:7
 // GFX90A: v_mfma_f32_16x16x4f32 v[0:3], a0, a1, -2.0 cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc5,0xd3,0x00,0x03,0xd6,0xfb]
 
-v_mfma_f32_32x32x4f16 a[0:31], v[0:1], v[2:3], a[2:33]
-// GFX90A: v_mfma_f32_32x32x4f16 a[0:31], v[0:1], v[2:3], a[2:33] ; encoding: [0x00,0x80,0xc8,0xd3,0x00,0x05,0x0a,0x04]
+v_mfma_f32_32x32x4f16 a[0:31], v[0:1], v[2:3], a[34:65]
+// GFX90A: v_mfma_f32_32x32x4f16 a[0:31], v[0:1], v[2:3], a[34:65] ; encoding: [0x00,0x80,0xc8,0xd3,0x00,0x05,0x8a,0x04]
 
-v_mfma_f32_32x32x4f16 a[0:31], v[0:1], v[2:3], a[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x4f16 a[0:31], v[0:1], v[2:3], a[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc8,0xd3,0x00,0x05,0x0a,0xe4]
+v_mfma_f32_32x32x4f16 a[0:31], v[0:1], v[2:3], a[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x4f16 a[0:31], v[0:1], v[2:3], a[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc8,0xd3,0x00,0x05,0x8a,0xe4]
 
-v_mfma_f32_32x32x4f16 a[0:31], v[0:1], a[2:3], a[2:33]
-// GFX90A: v_mfma_f32_32x32x4f16 a[0:31], v[0:1], a[2:3], a[2:33] ; encoding: [0x00,0x80,0xc8,0xd3,0x00,0x05,0x0a,0x14]
+v_mfma_f32_32x32x4f16 a[0:31], v[0:1], a[2:3], a[34:65]
+// GFX90A: v_mfma_f32_32x32x4f16 a[0:31], v[0:1], a[2:3], a[34:65] ; encoding: [0x00,0x80,0xc8,0xd3,0x00,0x05,0x8a,0x14]
 
-v_mfma_f32_32x32x4f16 a[0:31], v[0:1], a[2:3], a[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x4f16 a[0:31], v[0:1], a[2:3], a[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc8,0xd3,0x00,0x05,0x0a,0xf4]
+v_mfma_f32_32x32x4f16 a[0:31], v[0:1], a[2:3], a[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x4f16 a[0:31], v[0:1], a[2:3], a[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc8,0xd3,0x00,0x05,0x8a,0xf4]
 
-v_mfma_f32_32x32x4f16 a[0:31], a[0:1], v[2:3], a[2:33]
-// GFX90A: v_mfma_f32_32x32x4f16 a[0:31], a[0:1], v[2:3], a[2:33] ; encoding: [0x00,0x80,0xc8,0xd3,0x00,0x05,0x0a,0x0c]
+v_mfma_f32_32x32x4f16 a[0:31], a[0:1], v[2:3], a[34:65]
+// GFX90A: v_mfma_f32_32x32x4f16 a[0:31], a[0:1], v[2:3], a[34:65] ; encoding: [0x00,0x80,0xc8,0xd3,0x00,0x05,0x8a,0x0c]
 
-v_mfma_f32_32x32x4f16 a[0:31], a[0:1], v[2:3], a[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x4f16 a[0:31], a[0:1], v[2:3], a[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc8,0xd3,0x00,0x05,0x0a,0xec]
+v_mfma_f32_32x32x4f16 a[0:31], a[0:1], v[2:3], a[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x4f16 a[0:31], a[0:1], v[2:3], a[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc8,0xd3,0x00,0x05,0x8a,0xec]
 
-v_mfma_f32_32x32x4f16 a[0:31], a[0:1], a[2:3], a[2:33]
-// GFX90A: v_mfma_f32_32x32x4f16 a[0:31], a[0:1], a[2:3], a[2:33] ; encoding: [0x00,0x80,0xc8,0xd3,0x00,0x05,0x0a,0x1c]
+v_mfma_f32_32x32x4f16 a[0:31], a[0:1], a[2:3], a[34:65]
+// GFX90A: v_mfma_f32_32x32x4f16 a[0:31], a[0:1], a[2:3], a[34:65] ; encoding: [0x00,0x80,0xc8,0xd3,0x00,0x05,0x8a,0x1c]
 
-v_mfma_f32_32x32x4f16 a[0:31], a[0:1], a[2:3], a[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x4f16 a[0:31], a[0:1], a[2:3], a[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc8,0xd3,0x00,0x05,0x0a,0xfc]
+v_mfma_f32_32x32x4f16 a[0:31], a[0:1], a[2:3], a[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x4f16 a[0:31], a[0:1], a[2:3], a[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc8,0xd3,0x00,0x05,0x8a,0xfc]
 
-v_mfma_f32_32x32x4f16 v[0:31], v[0:1], v[2:3], v[2:33]
-// GFX90A: v_mfma_f32_32x32x4f16 v[0:31], v[0:1], v[2:3], v[2:33] ; encoding: [0x00,0x00,0xc8,0xd3,0x00,0x05,0x0a,0x04]
+v_mfma_f32_32x32x4f16 v[0:31], v[0:1], v[2:3], v[34:65]
+// GFX90A: v_mfma_f32_32x32x4f16 v[0:31], v[0:1], v[2:3], v[34:65] ; encoding: [0x00,0x00,0xc8,0xd3,0x00,0x05,0x8a,0x04]
 
-v_mfma_f32_32x32x4f16 v[0:31], v[0:1], v[2:3], v[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x4f16 v[0:31], v[0:1], v[2:3], v[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc8,0xd3,0x00,0x05,0x0a,0xe4]
+v_mfma_f32_32x32x4f16 v[0:31], v[0:1], v[2:3], v[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x4f16 v[0:31], v[0:1], v[2:3], v[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc8,0xd3,0x00,0x05,0x8a,0xe4]
 
-v_mfma_f32_32x32x4f16 v[0:31], v[0:1], a[2:3], v[2:33]
-// GFX90A: v_mfma_f32_32x32x4f16 v[0:31], v[0:1], a[2:3], v[2:33] ; encoding: [0x00,0x00,0xc8,0xd3,0x00,0x05,0x0a,0x14]
+v_mfma_f32_32x32x4f16 v[0:31], v[0:1], a[2:3], v[34:65]
+// GFX90A: v_mfma_f32_32x32x4f16 v[0:31], v[0:1], a[2:3], v[34:65] ; encoding: [0x00,0x00,0xc8,0xd3,0x00,0x05,0x8a,0x14]
 
-v_mfma_f32_32x32x4f16 v[0:31], v[0:1], a[2:3], v[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x4f16 v[0:31], v[0:1], a[2:3], v[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc8,0xd3,0x00,0x05,0x0a,0xf4]
+v_mfma_f32_32x32x4f16 v[0:31], v[0:1], a[2:3], v[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x4f16 v[0:31], v[0:1], a[2:3], v[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc8,0xd3,0x00,0x05,0x8a,0xf4]
 
-v_mfma_f32_32x32x4f16 v[0:31], a[0:1], v[2:3], v[2:33]
-// GFX90A: v_mfma_f32_32x32x4f16 v[0:31], a[0:1], v[2:3], v[2:33] ; encoding: [0x00,0x00,0xc8,0xd3,0x00,0x05,0x0a,0x0c]
+v_mfma_f32_32x32x4f16 v[0:31], a[0:1], v[2:3], v[34:65]
+// GFX90A: v_mfma_f32_32x32x4f16 v[0:31], a[0:1], v[2:3], v[34:65] ; encoding: [0x00,0x00,0xc8,0xd3,0x00,0x05,0x8a,0x0c]
 
-v_mfma_f32_32x32x4f16 v[0:31], a[0:1], v[2:3], v[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x4f16 v[0:31], a[0:1], v[2:3], v[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc8,0xd3,0x00,0x05,0x0a,0xec]
+v_mfma_f32_32x32x4f16 v[0:31], a[0:1], v[2:3], v[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x4f16 v[0:31], a[0:1], v[2:3], v[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc8,0xd3,0x00,0x05,0x8a,0xec]
 
-v_mfma_f32_32x32x4f16 v[0:31], a[0:1], a[2:3], v[2:33]
-// GFX90A: v_mfma_f32_32x32x4f16 v[0:31], a[0:1], a[2:3], v[2:33] ; encoding: [0x00,0x00,0xc8,0xd3,0x00,0x05,0x0a,0x1c]
+v_mfma_f32_32x32x4f16 v[0:31], a[0:1], a[2:3], v[34:65]
+// GFX90A: v_mfma_f32_32x32x4f16 v[0:31], a[0:1], a[2:3], v[34:65] ; encoding: [0x00,0x00,0xc8,0xd3,0x00,0x05,0x8a,0x1c]
 
-v_mfma_f32_32x32x4f16 v[0:31], a[0:1], a[2:3], v[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x4f16 v[0:31], a[0:1], a[2:3], v[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc8,0xd3,0x00,0x05,0x0a,0xfc]
+v_mfma_f32_32x32x4f16 v[0:31], a[0:1], a[2:3], v[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x4f16 v[0:31], a[0:1], a[2:3], v[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc8,0xd3,0x00,0x05,0x8a,0xfc]
 
 v_mfma_f32_32x32x4f16 a[0:31], v[0:1], v[2:3], -2.0
 // GFX90A: v_mfma_f32_32x32x4f16 a[0:31], v[0:1], v[2:3], -2.0 ; encoding: [0x00,0x80,0xc8,0xd3,0x00,0x05,0xd6,0x03]
@@ -603,53 +603,53 @@ v_mfma_f32_32x32x4f16 a[0:31], a[0:1], a[2:3], -2.0 cbsz:3 abid:2 blgp:7
 v_mfma_f32_32x32x4f16 v[0:31], a[0:1], a[2:3], -2.0 cbsz:3 abid:2 blgp:7
 // GFX90A: v_mfma_f32_32x32x4f16 v[0:31], a[0:1], a[2:3], -2.0 cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc8,0xd3,0x00,0x05,0xd6,0xfb]
 
-v_mfma_f32_16x16x4f16 a[0:15], v[0:1], v[2:3], a[2:17]
-// GFX90A: v_mfma_f32_16x16x4f16 a[0:15], v[0:1], v[2:3], a[2:17] ; encoding: [0x00,0x80,0xc9,0xd3,0x00,0x05,0x0a,0x04]
+v_mfma_f32_16x16x4f16 a[0:15], v[0:1], v[2:3], a[18:33]
+// GFX90A: v_mfma_f32_16x16x4f16 a[0:15], v[0:1], v[2:3], a[18:33] ; encoding: [0x00,0x80,0xc9,0xd3,0x00,0x05,0x4a,0x04]
 
-v_mfma_f32_16x16x4f16 a[0:15], v[0:1], v[2:3], a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x4f16 a[0:15], v[0:1], v[2:3], a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc9,0xd3,0x00,0x05,0x0a,0xe4]
+v_mfma_f32_16x16x4f16 a[0:15], v[0:1], v[2:3], a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x4f16 a[0:15], v[0:1], v[2:3], a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc9,0xd3,0x00,0x05,0x4a,0xe4]
 
-v_mfma_f32_16x16x4f16 a[0:15], v[0:1], a[2:3], a[2:17]
-// GFX90A: v_mfma_f32_16x16x4f16 a[0:15], v[0:1], a[2:3], a[2:17] ; encoding: [0x00,0x80,0xc9,0xd3,0x00,0x05,0x0a,0x14]
+v_mfma_f32_16x16x4f16 a[0:15], v[0:1], a[2:3], a[18:33]
+// GFX90A: v_mfma_f32_16x16x4f16 a[0:15], v[0:1], a[2:3], a[18:33] ; encoding: [0x00,0x80,0xc9,0xd3,0x00,0x05,0x4a,0x14]
 
-v_mfma_f32_16x16x4f16 a[0:15], v[0:1], a[2:3], a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x4f16 a[0:15], v[0:1], a[2:3], a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc9,0xd3,0x00,0x05,0x0a,0xf4]
+v_mfma_f32_16x16x4f16 a[0:15], v[0:1], a[2:3], a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x4f16 a[0:15], v[0:1], a[2:3], a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc9,0xd3,0x00,0x05,0x4a,0xf4]
 
-v_mfma_f32_16x16x4f16 a[0:15], a[0:1], v[2:3], a[2:17]
-// GFX90A: v_mfma_f32_16x16x4f16 a[0:15], a[0:1], v[2:3], a[2:17] ; encoding: [0x00,0x80,0xc9,0xd3,0x00,0x05,0x0a,0x0c]
+v_mfma_f32_16x16x4f16 a[0:15], a[0:1], v[2:3], a[18:33]
+// GFX90A: v_mfma_f32_16x16x4f16 a[0:15], a[0:1], v[2:3], a[18:33] ; encoding: [0x00,0x80,0xc9,0xd3,0x00,0x05,0x4a,0x0c]
 
-v_mfma_f32_16x16x4f16 a[0:15], a[0:1], v[2:3], a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x4f16 a[0:15], a[0:1], v[2:3], a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc9,0xd3,0x00,0x05,0x0a,0xec]
+v_mfma_f32_16x16x4f16 a[0:15], a[0:1], v[2:3], a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x4f16 a[0:15], a[0:1], v[2:3], a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc9,0xd3,0x00,0x05,0x4a,0xec]
 
-v_mfma_f32_16x16x4f16 a[0:15], a[0:1], a[2:3], a[2:17]
-// GFX90A: v_mfma_f32_16x16x4f16 a[0:15], a[0:1], a[2:3], a[2:17] ; encoding: [0x00,0x80,0xc9,0xd3,0x00,0x05,0x0a,0x1c]
+v_mfma_f32_16x16x4f16 a[0:15], a[0:1], a[2:3], a[18:33]
+// GFX90A: v_mfma_f32_16x16x4f16 a[0:15], a[0:1], a[2:3], a[18:33] ; encoding: [0x00,0x80,0xc9,0xd3,0x00,0x05,0x4a,0x1c]
 
-v_mfma_f32_16x16x4f16 a[0:15], a[0:1], a[2:3], a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x4f16 a[0:15], a[0:1], a[2:3], a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc9,0xd3,0x00,0x05,0x0a,0xfc]
+v_mfma_f32_16x16x4f16 a[0:15], a[0:1], a[2:3], a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x4f16 a[0:15], a[0:1], a[2:3], a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xc9,0xd3,0x00,0x05,0x4a,0xfc]
 
-v_mfma_f32_16x16x4f16 v[0:15], v[0:1], v[2:3], v[2:17]
-// GFX90A: v_mfma_f32_16x16x4f16 v[0:15], v[0:1], v[2:3], v[2:17] ; encoding: [0x00,0x00,0xc9,0xd3,0x00,0x05,0x0a,0x04]
+v_mfma_f32_16x16x4f16 v[0:15], v[0:1], v[2:3], v[18:33]
+// GFX90A: v_mfma_f32_16x16x4f16 v[0:15], v[0:1], v[2:3], v[18:33] ; encoding: [0x00,0x00,0xc9,0xd3,0x00,0x05,0x4a,0x04]
 
-v_mfma_f32_16x16x4f16 v[0:15], v[0:1], v[2:3], v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x4f16 v[0:15], v[0:1], v[2:3], v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc9,0xd3,0x00,0x05,0x0a,0xe4]
+v_mfma_f32_16x16x4f16 v[0:15], v[0:1], v[2:3], v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x4f16 v[0:15], v[0:1], v[2:3], v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc9,0xd3,0x00,0x05,0x4a,0xe4]
 
-v_mfma_f32_16x16x4f16 v[0:15], v[0:1], a[2:3], v[2:17]
-// GFX90A: v_mfma_f32_16x16x4f16 v[0:15], v[0:1], a[2:3], v[2:17] ; encoding: [0x00,0x00,0xc9,0xd3,0x00,0x05,0x0a,0x14]
+v_mfma_f32_16x16x4f16 v[0:15], v[0:1], a[2:3], v[18:33]
+// GFX90A: v_mfma_f32_16x16x4f16 v[0:15], v[0:1], a[2:3], v[18:33] ; encoding: [0x00,0x00,0xc9,0xd3,0x00,0x05,0x4a,0x14]
 
-v_mfma_f32_16x16x4f16 v[0:15], v[0:1], a[2:3], v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x4f16 v[0:15], v[0:1], a[2:3], v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc9,0xd3,0x00,0x05,0x0a,0xf4]
+v_mfma_f32_16x16x4f16 v[0:15], v[0:1], a[2:3], v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x4f16 v[0:15], v[0:1], a[2:3], v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc9,0xd3,0x00,0x05,0x4a,0xf4]
 
-v_mfma_f32_16x16x4f16 v[0:15], a[0:1], v[2:3], v[2:17]
-// GFX90A: v_mfma_f32_16x16x4f16 v[0:15], a[0:1], v[2:3], v[2:17] ; encoding: [0x00,0x00,0xc9,0xd3,0x00,0x05,0x0a,0x0c]
+v_mfma_f32_16x16x4f16 v[0:15], a[0:1], v[2:3], v[18:33]
+// GFX90A: v_mfma_f32_16x16x4f16 v[0:15], a[0:1], v[2:3], v[18:33] ; encoding: [0x00,0x00,0xc9,0xd3,0x00,0x05,0x4a,0x0c]
 
-v_mfma_f32_16x16x4f16 v[0:15], a[0:1], v[2:3], v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x4f16 v[0:15], a[0:1], v[2:3], v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc9,0xd3,0x00,0x05,0x0a,0xec]
+v_mfma_f32_16x16x4f16 v[0:15], a[0:1], v[2:3], v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x4f16 v[0:15], a[0:1], v[2:3], v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc9,0xd3,0x00,0x05,0x4a,0xec]
 
-v_mfma_f32_16x16x4f16 v[0:15], a[0:1], a[2:3], v[2:17]
-// GFX90A: v_mfma_f32_16x16x4f16 v[0:15], a[0:1], a[2:3], v[2:17] ; encoding: [0x00,0x00,0xc9,0xd3,0x00,0x05,0x0a,0x1c]
+v_mfma_f32_16x16x4f16 v[0:15], a[0:1], a[2:3], v[18:33]
+// GFX90A: v_mfma_f32_16x16x4f16 v[0:15], a[0:1], a[2:3], v[18:33] ; encoding: [0x00,0x00,0xc9,0xd3,0x00,0x05,0x4a,0x1c]
 
-v_mfma_f32_16x16x4f16 v[0:15], a[0:1], a[2:3], v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x4f16 v[0:15], a[0:1], a[2:3], v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc9,0xd3,0x00,0x05,0x0a,0xfc]
+v_mfma_f32_16x16x4f16 v[0:15], a[0:1], a[2:3], v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x4f16 v[0:15], a[0:1], a[2:3], v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc9,0xd3,0x00,0x05,0x4a,0xfc]
 
 v_mfma_f32_16x16x4f16 a[0:15], v[0:1], v[2:3], -2.0
 // GFX90A: v_mfma_f32_16x16x4f16 a[0:15], v[0:1], v[2:3], -2.0 ; encoding: [0x00,0x80,0xc9,0xd3,0x00,0x05,0xd6,0x03]
@@ -795,53 +795,53 @@ v_mfma_f32_4x4x4f16 a[0:3], a[0:1], a[2:3], -2.0 cbsz:3 abid:2 blgp:7
 v_mfma_f32_4x4x4f16 v[0:3], a[0:1], a[2:3], -2.0 cbsz:3 abid:2 blgp:7
 // GFX90A: v_mfma_f32_4x4x4f16 v[0:3], a[0:1], a[2:3], -2.0 cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xca,0xd3,0x00,0x05,0xd6,0xfb]
 
-v_mfma_f32_32x32x8f16 a[0:15], v[0:1], v[2:3], a[2:17]
-// GFX90A: v_mfma_f32_32x32x8f16 a[0:15], v[0:1], v[2:3], a[2:17] ; encoding: [0x00,0x80,0xcc,0xd3,0x00,0x05,0x0a,0x04]
+v_mfma_f32_32x32x8f16 a[0:15], v[0:1], v[2:3], a[18:33]
+// GFX90A: v_mfma_f32_32x32x8f16 a[0:15], v[0:1], v[2:3], a[18:33] ; encoding: [0x00,0x80,0xcc,0xd3,0x00,0x05,0x4a,0x04]
 
-v_mfma_f32_32x32x8f16 a[0:15], v[0:1], v[2:3], a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x8f16 a[0:15], v[0:1], v[2:3], a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xcc,0xd3,0x00,0x05,0x0a,0xe4]
+v_mfma_f32_32x32x8f16 a[0:15], v[0:1], v[2:3], a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x8f16 a[0:15], v[0:1], v[2:3], a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xcc,0xd3,0x00,0x05,0x4a,0xe4]
 
-v_mfma_f32_32x32x8f16 a[0:15], v[0:1], a[2:3], a[2:17]
-// GFX90A: v_mfma_f32_32x32x8f16 a[0:15], v[0:1], a[2:3], a[2:17] ; encoding: [0x00,0x80,0xcc,0xd3,0x00,0x05,0x0a,0x14]
+v_mfma_f32_32x32x8f16 a[0:15], v[0:1], a[2:3], a[18:33]
+// GFX90A: v_mfma_f32_32x32x8f16 a[0:15], v[0:1], a[2:3], a[18:33] ; encoding: [0x00,0x80,0xcc,0xd3,0x00,0x05,0x4a,0x14]
 
-v_mfma_f32_32x32x8f16 a[0:15], v[0:1], a[2:3], a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x8f16 a[0:15], v[0:1], a[2:3], a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xcc,0xd3,0x00,0x05,0x0a,0xf4]
+v_mfma_f32_32x32x8f16 a[0:15], v[0:1], a[2:3], a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x8f16 a[0:15], v[0:1], a[2:3], a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xcc,0xd3,0x00,0x05,0x4a,0xf4]
 
-v_mfma_f32_32x32x8f16 a[0:15], a[0:1], v[2:3], a[2:17]
-// GFX90A: v_mfma_f32_32x32x8f16 a[0:15], a[0:1], v[2:3], a[2:17] ; encoding: [0x00,0x80,0xcc,0xd3,0x00,0x05,0x0a,0x0c]
+v_mfma_f32_32x32x8f16 a[0:15], a[0:1], v[2:3], a[18:33]
+// GFX90A: v_mfma_f32_32x32x8f16 a[0:15], a[0:1], v[2:3], a[18:33] ; encoding: [0x00,0x80,0xcc,0xd3,0x00,0x05,0x4a,0x0c]
 
-v_mfma_f32_32x32x8f16 a[0:15], a[0:1], v[2:3], a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x8f16 a[0:15], a[0:1], v[2:3], a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xcc,0xd3,0x00,0x05,0x0a,0xec]
+v_mfma_f32_32x32x8f16 a[0:15], a[0:1], v[2:3], a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x8f16 a[0:15], a[0:1], v[2:3], a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xcc,0xd3,0x00,0x05,0x4a,0xec]
 
-v_mfma_f32_32x32x8f16 a[0:15], a[0:1], a[2:3], a[2:17]
-// GFX90A: v_mfma_f32_32x32x8f16 a[0:15], a[0:1], a[2:3], a[2:17] ; encoding: [0x00,0x80,0xcc,0xd3,0x00,0x05,0x0a,0x1c]
+v_mfma_f32_32x32x8f16 a[0:15], a[0:1], a[2:3], a[18:33]
+// GFX90A: v_mfma_f32_32x32x8f16 a[0:15], a[0:1], a[2:3], a[18:33] ; encoding: [0x00,0x80,0xcc,0xd3,0x00,0x05,0x4a,0x1c]
 
-v_mfma_f32_32x32x8f16 a[0:15], a[0:1], a[2:3], a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x8f16 a[0:15], a[0:1], a[2:3], a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xcc,0xd3,0x00,0x05,0x0a,0xfc]
+v_mfma_f32_32x32x8f16 a[0:15], a[0:1], a[2:3], a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x8f16 a[0:15], a[0:1], a[2:3], a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xcc,0xd3,0x00,0x05,0x4a,0xfc]
 
-v_mfma_f32_32x32x8f16 v[0:15], v[0:1], v[2:3], v[2:17]
-// GFX90A: v_mfma_f32_32x32x8f16 v[0:15], v[0:1], v[2:3], v[2:17] ; encoding: [0x00,0x00,0xcc,0xd3,0x00,0x05,0x0a,0x04]
+v_mfma_f32_32x32x8f16 v[0:15], v[0:1], v[2:3], v[18:33]
+// GFX90A: v_mfma_f32_32x32x8f16 v[0:15], v[0:1], v[2:3], v[18:33] ; encoding: [0x00,0x00,0xcc,0xd3,0x00,0x05,0x4a,0x04]
 
-v_mfma_f32_32x32x8f16 v[0:15], v[0:1], v[2:3], v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x8f16 v[0:15], v[0:1], v[2:3], v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xcc,0xd3,0x00,0x05,0x0a,0xe4]
+v_mfma_f32_32x32x8f16 v[0:15], v[0:1], v[2:3], v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x8f16 v[0:15], v[0:1], v[2:3], v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xcc,0xd3,0x00,0x05,0x4a,0xe4]
 
-v_mfma_f32_32x32x8f16 v[0:15], v[0:1], a[2:3], v[2:17]
-// GFX90A: v_mfma_f32_32x32x8f16 v[0:15], v[0:1], a[2:3], v[2:17] ; encoding: [0x00,0x00,0xcc,0xd3,0x00,0x05,0x0a,0x14]
+v_mfma_f32_32x32x8f16 v[0:15], v[0:1], a[2:3], v[18:33]
+// GFX90A: v_mfma_f32_32x32x8f16 v[0:15], v[0:1], a[2:3], v[18:33] ; encoding: [0x00,0x00,0xcc,0xd3,0x00,0x05,0x4a,0x14]
 
-v_mfma_f32_32x32x8f16 v[0:15], v[0:1], a[2:3], v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x8f16 v[0:15], v[0:1], a[2:3], v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xcc,0xd3,0x00,0x05,0x0a,0xf4]
+v_mfma_f32_32x32x8f16 v[0:15], v[0:1], a[2:3], v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x8f16 v[0:15], v[0:1], a[2:3], v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xcc,0xd3,0x00,0x05,0x4a,0xf4]
 
-v_mfma_f32_32x32x8f16 v[0:15], a[0:1], v[2:3], v[2:17]
-// GFX90A: v_mfma_f32_32x32x8f16 v[0:15], a[0:1], v[2:3], v[2:17] ; encoding: [0x00,0x00,0xcc,0xd3,0x00,0x05,0x0a,0x0c]
+v_mfma_f32_32x32x8f16 v[0:15], a[0:1], v[2:3], v[18:33]
+// GFX90A: v_mfma_f32_32x32x8f16 v[0:15], a[0:1], v[2:3], v[18:33] ; encoding: [0x00,0x00,0xcc,0xd3,0x00,0x05,0x4a,0x0c]
 
-v_mfma_f32_32x32x8f16 v[0:15], a[0:1], v[2:3], v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x8f16 v[0:15], a[0:1], v[2:3], v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xcc,0xd3,0x00,0x05,0x0a,0xec]
+v_mfma_f32_32x32x8f16 v[0:15], a[0:1], v[2:3], v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x8f16 v[0:15], a[0:1], v[2:3], v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xcc,0xd3,0x00,0x05,0x4a,0xec]
 
-v_mfma_f32_32x32x8f16 v[0:15], a[0:1], a[2:3], v[2:17]
-// GFX90A: v_mfma_f32_32x32x8f16 v[0:15], a[0:1], a[2:3], v[2:17] ; encoding: [0x00,0x00,0xcc,0xd3,0x00,0x05,0x0a,0x1c]
+v_mfma_f32_32x32x8f16 v[0:15], a[0:1], a[2:3], v[18:33]
+// GFX90A: v_mfma_f32_32x32x8f16 v[0:15], a[0:1], a[2:3], v[18:33] ; encoding: [0x00,0x00,0xcc,0xd3,0x00,0x05,0x4a,0x1c]
 
-v_mfma_f32_32x32x8f16 v[0:15], a[0:1], a[2:3], v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x8f16 v[0:15], a[0:1], a[2:3], v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xcc,0xd3,0x00,0x05,0x0a,0xfc]
+v_mfma_f32_32x32x8f16 v[0:15], a[0:1], a[2:3], v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x8f16 v[0:15], a[0:1], a[2:3], v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xcc,0xd3,0x00,0x05,0x4a,0xfc]
 
 v_mfma_f32_32x32x8f16 a[0:15], v[0:1], v[2:3], -2.0
 // GFX90A: v_mfma_f32_32x32x8f16 a[0:15], v[0:1], v[2:3], -2.0 ; encoding: [0x00,0x80,0xcc,0xd3,0x00,0x05,0xd6,0x03]
@@ -987,53 +987,53 @@ v_mfma_f32_16x16x16f16 a[0:3], a[0:1], a[2:3], -2.0 cbsz:3 abid:2 blgp:7
 v_mfma_f32_16x16x16f16 v[0:3], a[0:1], a[2:3], -2.0 cbsz:3 abid:2 blgp:7
 // GFX90A: v_mfma_f32_16x16x16f16 v[0:3], a[0:1], a[2:3], -2.0 cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xcd,0xd3,0x00,0x05,0xd6,0xfb]
 
-v_mfma_i32_32x32x4i8 a[0:31], v0, v1, a[2:33]
-// GFX90A: v_mfma_i32_32x32x4i8 a[0:31], v0, v1, a[2:33] ; encoding: [0x00,0x80,0xd0,0xd3,0x00,0x03,0x0a,0x04]
+v_mfma_i32_32x32x4i8 a[0:31], v0, v1, a[34:65]
+// GFX90A: v_mfma_i32_32x32x4i8 a[0:31], v0, v1, a[34:65] ; encoding: [0x00,0x80,0xd0,0xd3,0x00,0x03,0x8a,0x04]
 
-v_mfma_i32_32x32x4i8 a[0:31], v0, v1, a[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_i32_32x32x4i8 a[0:31], v0, v1, a[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xd0,0xd3,0x00,0x03,0x0a,0xe4]
+v_mfma_i32_32x32x4i8 a[0:31], v0, v1, a[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_i32_32x32x4i8 a[0:31], v0, v1, a[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xd0,0xd3,0x00,0x03,0x8a,0xe4]
 
-v_mfma_i32_32x32x4i8 a[0:31], v0, a1, a[2:33]
-// GFX90A: v_mfma_i32_32x32x4i8 a[0:31], v0, a1, a[2:33] ; encoding: [0x00,0x80,0xd0,0xd3,0x00,0x03,0x0a,0x14]
+v_mfma_i32_32x32x4i8 a[0:31], v0, a1, a[34:65]
+// GFX90A: v_mfma_i32_32x32x4i8 a[0:31], v0, a1, a[34:65] ; encoding: [0x00,0x80,0xd0,0xd3,0x00,0x03,0x8a,0x14]
 
-v_mfma_i32_32x32x4i8 a[0:31], v0, a1, a[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_i32_32x32x4i8 a[0:31], v0, a1, a[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xd0,0xd3,0x00,0x03,0x0a,0xf4]
+v_mfma_i32_32x32x4i8 a[0:31], v0, a1, a[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_i32_32x32x4i8 a[0:31], v0, a1, a[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xd0,0xd3,0x00,0x03,0x8a,0xf4]
 
-v_mfma_i32_32x32x4i8 a[0:31], a0, v1, a[2:33]
-// GFX90A: v_mfma_i32_32x32x4i8 a[0:31], a0, v1, a[2:33] ; encoding: [0x00,0x80,0xd0,0xd3,0x00,0x03,0x0a,0x0c]
+v_mfma_i32_32x32x4i8 a[0:31], a0, v1, a[34:65]
+// GFX90A: v_mfma_i32_32x32x4i8 a[0:31], a0, v1, a[34:65] ; encoding: [0x00,0x80,0xd0,0xd3,0x00,0x03,0x8a,0x0c]
 
-v_mfma_i32_32x32x4i8 a[0:31], a0, v1, a[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_i32_32x32x4i8 a[0:31], a0, v1, a[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xd0,0xd3,0x00,0x03,0x0a,0xec]
+v_mfma_i32_32x32x4i8 a[0:31], a0, v1, a[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_i32_32x32x4i8 a[0:31], a0, v1, a[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xd0,0xd3,0x00,0x03,0x8a,0xec]
 
-v_mfma_i32_32x32x4i8 a[0:31], a0, a1, a[2:33]
-// GFX90A: v_mfma_i32_32x32x4i8 a[0:31], a0, a1, a[2:33] ; encoding: [0x00,0x80,0xd0,0xd3,0x00,0x03,0x0a,0x1c]
+v_mfma_i32_32x32x4i8 a[0:31], a0, a1, a[34:65]
+// GFX90A: v_mfma_i32_32x32x4i8 a[0:31], a0, a1, a[34:65] ; encoding: [0x00,0x80,0xd0,0xd3,0x00,0x03,0x8a,0x1c]
 
-v_mfma_i32_32x32x4i8 a[0:31], a0, a1, a[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_i32_32x32x4i8 a[0:31], a0, a1, a[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xd0,0xd3,0x00,0x03,0x0a,0xfc]
+v_mfma_i32_32x32x4i8 a[0:31], a0, a1, a[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_i32_32x32x4i8 a[0:31], a0, a1, a[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xd0,0xd3,0x00,0x03,0x8a,0xfc]
 
-v_mfma_i32_32x32x4i8 v[0:31], v0, v1, v[2:33]
-// GFX90A: v_mfma_i32_32x32x4i8 v[0:31], v0, v1, v[2:33] ; encoding: [0x00,0x00,0xd0,0xd3,0x00,0x03,0x0a,0x04]
+v_mfma_i32_32x32x4i8 v[0:31], v0, v1, v[34:65]
+// GFX90A: v_mfma_i32_32x32x4i8 v[0:31], v0, v1, v[34:65] ; encoding: [0x00,0x00,0xd0,0xd3,0x00,0x03,0x8a,0x04]
 
-v_mfma_i32_32x32x4i8 v[0:31], v0, v1, v[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_i32_32x32x4i8 v[0:31], v0, v1, v[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd0,0xd3,0x00,0x03,0x0a,0xe4]
+v_mfma_i32_32x32x4i8 v[0:31], v0, v1, v[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_i32_32x32x4i8 v[0:31], v0, v1, v[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd0,0xd3,0x00,0x03,0x8a,0xe4]
 
-v_mfma_i32_32x32x4i8 v[0:31], v0, a1, v[2:33]
-// GFX90A: v_mfma_i32_32x32x4i8 v[0:31], v0, a1, v[2:33] ; encoding: [0x00,0x00,0xd0,0xd3,0x00,0x03,0x0a,0x14]
+v_mfma_i32_32x32x4i8 v[0:31], v0, a1, v[34:65]
+// GFX90A: v_mfma_i32_32x32x4i8 v[0:31], v0, a1, v[34:65] ; encoding: [0x00,0x00,0xd0,0xd3,0x00,0x03,0x8a,0x14]
 
-v_mfma_i32_32x32x4i8 v[0:31], v0, a1, v[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_i32_32x32x4i8 v[0:31], v0, a1, v[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd0,0xd3,0x00,0x03,0x0a,0xf4]
+v_mfma_i32_32x32x4i8 v[0:31], v0, a1, v[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_i32_32x32x4i8 v[0:31], v0, a1, v[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd0,0xd3,0x00,0x03,0x8a,0xf4]
 
-v_mfma_i32_32x32x4i8 v[0:31], a0, v1, v[2:33]
-// GFX90A: v_mfma_i32_32x32x4i8 v[0:31], a0, v1, v[2:33] ; encoding: [0x00,0x00,0xd0,0xd3,0x00,0x03,0x0a,0x0c]
+v_mfma_i32_32x32x4i8 v[0:31], a0, v1, v[34:65]
+// GFX90A: v_mfma_i32_32x32x4i8 v[0:31], a0, v1, v[34:65] ; encoding: [0x00,0x00,0xd0,0xd3,0x00,0x03,0x8a,0x0c]
 
-v_mfma_i32_32x32x4i8 v[0:31], a0, v1, v[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_i32_32x32x4i8 v[0:31], a0, v1, v[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd0,0xd3,0x00,0x03,0x0a,0xec]
+v_mfma_i32_32x32x4i8 v[0:31], a0, v1, v[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_i32_32x32x4i8 v[0:31], a0, v1, v[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd0,0xd3,0x00,0x03,0x8a,0xec]
 
-v_mfma_i32_32x32x4i8 v[0:31], a0, a1, v[2:33]
-// GFX90A: v_mfma_i32_32x32x4i8 v[0:31], a0, a1, v[2:33] ; encoding: [0x00,0x00,0xd0,0xd3,0x00,0x03,0x0a,0x1c]
+v_mfma_i32_32x32x4i8 v[0:31], a0, a1, v[34:65]
+// GFX90A: v_mfma_i32_32x32x4i8 v[0:31], a0, a1, v[34:65] ; encoding: [0x00,0x00,0xd0,0xd3,0x00,0x03,0x8a,0x1c]
 
-v_mfma_i32_32x32x4i8 v[0:31], a0, a1, v[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_i32_32x32x4i8 v[0:31], a0, a1, v[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd0,0xd3,0x00,0x03,0x0a,0xfc]
+v_mfma_i32_32x32x4i8 v[0:31], a0, a1, v[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_i32_32x32x4i8 v[0:31], a0, a1, v[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd0,0xd3,0x00,0x03,0x8a,0xfc]
 
 v_mfma_i32_32x32x4i8 a[0:31], v0, v1, 2
 // GFX90A: v_mfma_i32_32x32x4i8 a[0:31], v0, v1, 2 ; encoding: [0x00,0x80,0xd0,0xd3,0x00,0x03,0x0a,0x02]
@@ -1083,53 +1083,53 @@ v_mfma_i32_32x32x4i8 a[0:31], a0, a1, 2 cbsz:3 abid:2 blgp:7
 v_mfma_i32_32x32x4i8 v[0:31], a0, a1, 2 cbsz:3 abid:2 blgp:7
 // GFX90A: v_mfma_i32_32x32x4i8 v[0:31], a0, a1, 2 cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd0,0xd3,0x00,0x03,0x0a,0xfa]
 
-v_mfma_i32_16x16x4i8 a[0:15], v0, v1, a[2:17]
-// GFX90A: v_mfma_i32_16x16x4i8 a[0:15], v0, v1, a[2:17] ; encoding: [0x00,0x80,0xd1,0xd3,0x00,0x03,0x0a,0x04]
+v_mfma_i32_16x16x4i8 a[0:15], v0, v1, a[18:33]
+// GFX90A: v_mfma_i32_16x16x4i8 a[0:15], v0, v1, a[18:33] ; encoding: [0x00,0x80,0xd1,0xd3,0x00,0x03,0x4a,0x04]
 
-v_mfma_i32_16x16x4i8 a[0:15], v0, v1, a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_i32_16x16x4i8 a[0:15], v0, v1, a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xd1,0xd3,0x00,0x03,0x0a,0xe4]
+v_mfma_i32_16x16x4i8 a[0:15], v0, v1, a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_i32_16x16x4i8 a[0:15], v0, v1, a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xd1,0xd3,0x00,0x03,0x4a,0xe4]
 
-v_mfma_i32_16x16x4i8 a[0:15], v0, a1, a[2:17]
-// GFX90A: v_mfma_i32_16x16x4i8 a[0:15], v0, a1, a[2:17] ; encoding: [0x00,0x80,0xd1,0xd3,0x00,0x03,0x0a,0x14]
+v_mfma_i32_16x16x4i8 a[0:15], v0, a1, a[18:33]
+// GFX90A: v_mfma_i32_16x16x4i8 a[0:15], v0, a1, a[18:33] ; encoding: [0x00,0x80,0xd1,0xd3,0x00,0x03,0x4a,0x14]
 
-v_mfma_i32_16x16x4i8 a[0:15], v0, a1, a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_i32_16x16x4i8 a[0:15], v0, a1, a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xd1,0xd3,0x00,0x03,0x0a,0xf4]
+v_mfma_i32_16x16x4i8 a[0:15], v0, a1, a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_i32_16x16x4i8 a[0:15], v0, a1, a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xd1,0xd3,0x00,0x03,0x4a,0xf4]
 
-v_mfma_i32_16x16x4i8 a[0:15], a0, v1, a[2:17]
-// GFX90A: v_mfma_i32_16x16x4i8 a[0:15], a0, v1, a[2:17] ; encoding: [0x00,0x80,0xd1,0xd3,0x00,0x03,0x0a,0x0c]
+v_mfma_i32_16x16x4i8 a[0:15], a0, v1, a[18:33]
+// GFX90A: v_mfma_i32_16x16x4i8 a[0:15], a0, v1, a[18:33] ; encoding: [0x00,0x80,0xd1,0xd3,0x00,0x03,0x4a,0x0c]
 
-v_mfma_i32_16x16x4i8 a[0:15], a0, v1, a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_i32_16x16x4i8 a[0:15], a0, v1, a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xd1,0xd3,0x00,0x03,0x0a,0xec]
+v_mfma_i32_16x16x4i8 a[0:15], a0, v1, a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_i32_16x16x4i8 a[0:15], a0, v1, a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xd1,0xd3,0x00,0x03,0x4a,0xec]
 
-v_mfma_i32_16x16x4i8 a[0:15], a0, a1, a[2:17]
-// GFX90A: v_mfma_i32_16x16x4i8 a[0:15], a0, a1, a[2:17] ; encoding: [0x00,0x80,0xd1,0xd3,0x00,0x03,0x0a,0x1c]
+v_mfma_i32_16x16x4i8 a[0:15], a0, a1, a[18:33]
+// GFX90A: v_mfma_i32_16x16x4i8 a[0:15], a0, a1, a[18:33] ; encoding: [0x00,0x80,0xd1,0xd3,0x00,0x03,0x4a,0x1c]
 
-v_mfma_i32_16x16x4i8 a[0:15], a0, a1, a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_i32_16x16x4i8 a[0:15], a0, a1, a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xd1,0xd3,0x00,0x03,0x0a,0xfc]
+v_mfma_i32_16x16x4i8 a[0:15], a0, a1, a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_i32_16x16x4i8 a[0:15], a0, a1, a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xd1,0xd3,0x00,0x03,0x4a,0xfc]
 
-v_mfma_i32_16x16x4i8 v[0:15], v0, v1, v[2:17]
-// GFX90A: v_mfma_i32_16x16x4i8 v[0:15], v0, v1, v[2:17] ; encoding: [0x00,0x00,0xd1,0xd3,0x00,0x03,0x0a,0x04]
+v_mfma_i32_16x16x4i8 v[0:15], v0, v1, v[18:33]
+// GFX90A: v_mfma_i32_16x16x4i8 v[0:15], v0, v1, v[18:33] ; encoding: [0x00,0x00,0xd1,0xd3,0x00,0x03,0x4a,0x04]
 
-v_mfma_i32_16x16x4i8 v[0:15], v0, v1, v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_i32_16x16x4i8 v[0:15], v0, v1, v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd1,0xd3,0x00,0x03,0x0a,0xe4]
+v_mfma_i32_16x16x4i8 v[0:15], v0, v1, v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_i32_16x16x4i8 v[0:15], v0, v1, v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd1,0xd3,0x00,0x03,0x4a,0xe4]
 
-v_mfma_i32_16x16x4i8 v[0:15], v0, a1, v[2:17]
-// GFX90A: v_mfma_i32_16x16x4i8 v[0:15], v0, a1, v[2:17] ; encoding: [0x00,0x00,0xd1,0xd3,0x00,0x03,0x0a,0x14]
+v_mfma_i32_16x16x4i8 v[0:15], v0, a1, v[18:33]
+// GFX90A: v_mfma_i32_16x16x4i8 v[0:15], v0, a1, v[18:33] ; encoding: [0x00,0x00,0xd1,0xd3,0x00,0x03,0x4a,0x14]
 
-v_mfma_i32_16x16x4i8 v[0:15], v0, a1, v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_i32_16x16x4i8 v[0:15], v0, a1, v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd1,0xd3,0x00,0x03,0x0a,0xf4]
+v_mfma_i32_16x16x4i8 v[0:15], v0, a1, v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_i32_16x16x4i8 v[0:15], v0, a1, v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd1,0xd3,0x00,0x03,0x4a,0xf4]
 
-v_mfma_i32_16x16x4i8 v[0:15], a0, v1, v[2:17]
-// GFX90A: v_mfma_i32_16x16x4i8 v[0:15], a0, v1, v[2:17] ; encoding: [0x00,0x00,0xd1,0xd3,0x00,0x03,0x0a,0x0c]
+v_mfma_i32_16x16x4i8 v[0:15], a0, v1, v[18:33]
+// GFX90A: v_mfma_i32_16x16x4i8 v[0:15], a0, v1, v[18:33] ; encoding: [0x00,0x00,0xd1,0xd3,0x00,0x03,0x4a,0x0c]
 
-v_mfma_i32_16x16x4i8 v[0:15], a0, v1, v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_i32_16x16x4i8 v[0:15], a0, v1, v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd1,0xd3,0x00,0x03,0x0a,0xec]
+v_mfma_i32_16x16x4i8 v[0:15], a0, v1, v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_i32_16x16x4i8 v[0:15], a0, v1, v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd1,0xd3,0x00,0x03,0x4a,0xec]
 
-v_mfma_i32_16x16x4i8 v[0:15], a0, a1, v[2:17]
-// GFX90A: v_mfma_i32_16x16x4i8 v[0:15], a0, a1, v[2:17] ; encoding: [0x00,0x00,0xd1,0xd3,0x00,0x03,0x0a,0x1c]
+v_mfma_i32_16x16x4i8 v[0:15], a0, a1, v[18:33]
+// GFX90A: v_mfma_i32_16x16x4i8 v[0:15], a0, a1, v[18:33] ; encoding: [0x00,0x00,0xd1,0xd3,0x00,0x03,0x4a,0x1c]
 
-v_mfma_i32_16x16x4i8 v[0:15], a0, a1, v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_i32_16x16x4i8 v[0:15], a0, a1, v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd1,0xd3,0x00,0x03,0x0a,0xfc]
+v_mfma_i32_16x16x4i8 v[0:15], a0, a1, v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_i32_16x16x4i8 v[0:15], a0, a1, v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd1,0xd3,0x00,0x03,0x4a,0xfc]
 
 v_mfma_i32_16x16x4i8 a[0:15], v0, v1, 2
 // GFX90A: v_mfma_i32_16x16x4i8 a[0:15], v0, v1, 2 ; encoding: [0x00,0x80,0xd1,0xd3,0x00,0x03,0x0a,0x02]
@@ -1275,53 +1275,53 @@ v_mfma_i32_4x4x4i8 a[0:3], a0, a1, 2 cbsz:3 abid:2 blgp:7
 v_mfma_i32_4x4x4i8 v[0:3], a0, a1, 2 cbsz:3 abid:2 blgp:7
 // GFX90A: v_mfma_i32_4x4x4i8 v[0:3], a0, a1, 2 cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd2,0xd3,0x00,0x03,0x0a,0xfa]
 
-v_mfma_i32_32x32x8i8 a[0:15], v0, v1, a[2:17]
-// GFX90A: v_mfma_i32_32x32x8i8 a[0:15], v0, v1, a[2:17] ; encoding: [0x00,0x80,0xd4,0xd3,0x00,0x03,0x0a,0x04]
+v_mfma_i32_32x32x8i8 a[0:15], v0, v1, a[18:33]
+// GFX90A: v_mfma_i32_32x32x8i8 a[0:15], v0, v1, a[18:33] ; encoding: [0x00,0x80,0xd4,0xd3,0x00,0x03,0x4a,0x04]
 
-v_mfma_i32_32x32x8i8 a[0:15], v0, v1, a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_i32_32x32x8i8 a[0:15], v0, v1, a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xd4,0xd3,0x00,0x03,0x0a,0xe4]
+v_mfma_i32_32x32x8i8 a[0:15], v0, v1, a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_i32_32x32x8i8 a[0:15], v0, v1, a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xd4,0xd3,0x00,0x03,0x4a,0xe4]
 
-v_mfma_i32_32x32x8i8 a[0:15], v0, a1, a[2:17]
-// GFX90A: v_mfma_i32_32x32x8i8 a[0:15], v0, a1, a[2:17] ; encoding: [0x00,0x80,0xd4,0xd3,0x00,0x03,0x0a,0x14]
+v_mfma_i32_32x32x8i8 a[0:15], v0, a1, a[18:33]
+// GFX90A: v_mfma_i32_32x32x8i8 a[0:15], v0, a1, a[18:33] ; encoding: [0x00,0x80,0xd4,0xd3,0x00,0x03,0x4a,0x14]
 
-v_mfma_i32_32x32x8i8 a[0:15], v0, a1, a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_i32_32x32x8i8 a[0:15], v0, a1, a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xd4,0xd3,0x00,0x03,0x0a,0xf4]
+v_mfma_i32_32x32x8i8 a[0:15], v0, a1, a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_i32_32x32x8i8 a[0:15], v0, a1, a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xd4,0xd3,0x00,0x03,0x4a,0xf4]
 
-v_mfma_i32_32x32x8i8 a[0:15], a0, v1, a[2:17]
-// GFX90A: v_mfma_i32_32x32x8i8 a[0:15], a0, v1, a[2:17] ; encoding: [0x00,0x80,0xd4,0xd3,0x00,0x03,0x0a,0x0c]
+v_mfma_i32_32x32x8i8 a[0:15], a0, v1, a[18:33]
+// GFX90A: v_mfma_i32_32x32x8i8 a[0:15], a0, v1, a[18:33] ; encoding: [0x00,0x80,0xd4,0xd3,0x00,0x03,0x4a,0x0c]
 
-v_mfma_i32_32x32x8i8 a[0:15], a0, v1, a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_i32_32x32x8i8 a[0:15], a0, v1, a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xd4,0xd3,0x00,0x03,0x0a,0xec]
+v_mfma_i32_32x32x8i8 a[0:15], a0, v1, a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_i32_32x32x8i8 a[0:15], a0, v1, a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xd4,0xd3,0x00,0x03,0x4a,0xec]
 
-v_mfma_i32_32x32x8i8 a[0:15], a0, a1, a[2:17]
-// GFX90A: v_mfma_i32_32x32x8i8 a[0:15], a0, a1, a[2:17] ; encoding: [0x00,0x80,0xd4,0xd3,0x00,0x03,0x0a,0x1c]
+v_mfma_i32_32x32x8i8 a[0:15], a0, a1, a[18:33]
+// GFX90A: v_mfma_i32_32x32x8i8 a[0:15], a0, a1, a[18:33] ; encoding: [0x00,0x80,0xd4,0xd3,0x00,0x03,0x4a,0x1c]
 
-v_mfma_i32_32x32x8i8 a[0:15], a0, a1, a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_i32_32x32x8i8 a[0:15], a0, a1, a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xd4,0xd3,0x00,0x03,0x0a,0xfc]
+v_mfma_i32_32x32x8i8 a[0:15], a0, a1, a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_i32_32x32x8i8 a[0:15], a0, a1, a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xd4,0xd3,0x00,0x03,0x4a,0xfc]
 
-v_mfma_i32_32x32x8i8 v[0:15], v0, v1, v[2:17]
-// GFX90A: v_mfma_i32_32x32x8i8 v[0:15], v0, v1, v[2:17] ; encoding: [0x00,0x00,0xd4,0xd3,0x00,0x03,0x0a,0x04]
+v_mfma_i32_32x32x8i8 v[0:15], v0, v1, v[18:33]
+// GFX90A: v_mfma_i32_32x32x8i8 v[0:15], v0, v1, v[18:33] ; encoding: [0x00,0x00,0xd4,0xd3,0x00,0x03,0x4a,0x04]
 
-v_mfma_i32_32x32x8i8 v[0:15], v0, v1, v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_i32_32x32x8i8 v[0:15], v0, v1, v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd4,0xd3,0x00,0x03,0x0a,0xe4]
+v_mfma_i32_32x32x8i8 v[0:15], v0, v1, v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_i32_32x32x8i8 v[0:15], v0, v1, v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd4,0xd3,0x00,0x03,0x4a,0xe4]
 
-v_mfma_i32_32x32x8i8 v[0:15], v0, a1, v[2:17]
-// GFX90A: v_mfma_i32_32x32x8i8 v[0:15], v0, a1, v[2:17] ; encoding: [0x00,0x00,0xd4,0xd3,0x00,0x03,0x0a,0x14]
+v_mfma_i32_32x32x8i8 v[0:15], v0, a1, v[18:33]
+// GFX90A: v_mfma_i32_32x32x8i8 v[0:15], v0, a1, v[18:33] ; encoding: [0x00,0x00,0xd4,0xd3,0x00,0x03,0x4a,0x14]
 
-v_mfma_i32_32x32x8i8 v[0:15], v0, a1, v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_i32_32x32x8i8 v[0:15], v0, a1, v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd4,0xd3,0x00,0x03,0x0a,0xf4]
+v_mfma_i32_32x32x8i8 v[0:15], v0, a1, v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_i32_32x32x8i8 v[0:15], v0, a1, v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd4,0xd3,0x00,0x03,0x4a,0xf4]
 
-v_mfma_i32_32x32x8i8 v[0:15], a0, v1, v[2:17]
-// GFX90A: v_mfma_i32_32x32x8i8 v[0:15], a0, v1, v[2:17] ; encoding: [0x00,0x00,0xd4,0xd3,0x00,0x03,0x0a,0x0c]
+v_mfma_i32_32x32x8i8 v[0:15], a0, v1, v[18:33]
+// GFX90A: v_mfma_i32_32x32x8i8 v[0:15], a0, v1, v[18:33] ; encoding: [0x00,0x00,0xd4,0xd3,0x00,0x03,0x4a,0x0c]
 
-v_mfma_i32_32x32x8i8 v[0:15], a0, v1, v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_i32_32x32x8i8 v[0:15], a0, v1, v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd4,0xd3,0x00,0x03,0x0a,0xec]
+v_mfma_i32_32x32x8i8 v[0:15], a0, v1, v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_i32_32x32x8i8 v[0:15], a0, v1, v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd4,0xd3,0x00,0x03,0x4a,0xec]
 
-v_mfma_i32_32x32x8i8 v[0:15], a0, a1, v[2:17]
-// GFX90A: v_mfma_i32_32x32x8i8 v[0:15], a0, a1, v[2:17] ; encoding: [0x00,0x00,0xd4,0xd3,0x00,0x03,0x0a,0x1c]
+v_mfma_i32_32x32x8i8 v[0:15], a0, a1, v[18:33]
+// GFX90A: v_mfma_i32_32x32x8i8 v[0:15], a0, a1, v[18:33] ; encoding: [0x00,0x00,0xd4,0xd3,0x00,0x03,0x4a,0x1c]
 
-v_mfma_i32_32x32x8i8 v[0:15], a0, a1, v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_i32_32x32x8i8 v[0:15], a0, a1, v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd4,0xd3,0x00,0x03,0x0a,0xfc]
+v_mfma_i32_32x32x8i8 v[0:15], a0, a1, v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_i32_32x32x8i8 v[0:15], a0, a1, v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd4,0xd3,0x00,0x03,0x4a,0xfc]
 
 v_mfma_i32_32x32x8i8 a[0:15], v0, v1, 2
 // GFX90A: v_mfma_i32_32x32x8i8 a[0:15], v0, v1, 2 ; encoding: [0x00,0x80,0xd4,0xd3,0x00,0x03,0x0a,0x02]
@@ -1467,53 +1467,53 @@ v_mfma_i32_16x16x16i8 a[0:3], a0, a1, 2 cbsz:3 abid:2 blgp:7
 v_mfma_i32_16x16x16i8 v[0:3], a0, a1, 2 cbsz:3 abid:2 blgp:7
 // GFX90A: v_mfma_i32_16x16x16i8 v[0:3], a0, a1, 2 cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd5,0xd3,0x00,0x03,0x0a,0xfa]
 
-v_mfma_f32_32x32x2bf16 a[0:31], v0, v1, a[2:33]
-// GFX90A: v_mfma_f32_32x32x2bf16 a[0:31], v0, v1, a[2:33] ; encoding: [0x00,0x80,0xe8,0xd3,0x00,0x03,0x0a,0x04]
+v_mfma_f32_32x32x2bf16 a[0:31], v0, v1, a[34:65]
+// GFX90A: v_mfma_f32_32x32x2bf16 a[0:31], v0, v1, a[34:65] ; encoding: [0x00,0x80,0xe8,0xd3,0x00,0x03,0x8a,0x04]
 
-v_mfma_f32_32x32x2bf16 a[0:31], v0, v1, a[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x2bf16 a[0:31], v0, v1, a[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe8,0xd3,0x00,0x03,0x0a,0xe4]
+v_mfma_f32_32x32x2bf16 a[0:31], v0, v1, a[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x2bf16 a[0:31], v0, v1, a[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe8,0xd3,0x00,0x03,0x8a,0xe4]
 
-v_mfma_f32_32x32x2bf16 a[0:31], v0, a1, a[2:33]
-// GFX90A: v_mfma_f32_32x32x2bf16 a[0:31], v0, a1, a[2:33] ; encoding: [0x00,0x80,0xe8,0xd3,0x00,0x03,0x0a,0x14]
+v_mfma_f32_32x32x2bf16 a[0:31], v0, a1, a[34:65]
+// GFX90A: v_mfma_f32_32x32x2bf16 a[0:31], v0, a1, a[34:65] ; encoding: [0x00,0x80,0xe8,0xd3,0x00,0x03,0x8a,0x14]
 
-v_mfma_f32_32x32x2bf16 a[0:31], v0, a1, a[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x2bf16 a[0:31], v0, a1, a[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe8,0xd3,0x00,0x03,0x0a,0xf4]
+v_mfma_f32_32x32x2bf16 a[0:31], v0, a1, a[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x2bf16 a[0:31], v0, a1, a[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe8,0xd3,0x00,0x03,0x8a,0xf4]
 
-v_mfma_f32_32x32x2bf16 a[0:31], a0, v1, a[2:33]
-// GFX90A: v_mfma_f32_32x32x2bf16 a[0:31], a0, v1, a[2:33] ; encoding: [0x00,0x80,0xe8,0xd3,0x00,0x03,0x0a,0x0c]
+v_mfma_f32_32x32x2bf16 a[0:31], a0, v1, a[34:65]
+// GFX90A: v_mfma_f32_32x32x2bf16 a[0:31], a0, v1, a[34:65] ; encoding: [0x00,0x80,0xe8,0xd3,0x00,0x03,0x8a,0x0c]
 
-v_mfma_f32_32x32x2bf16 a[0:31], a0, v1, a[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x2bf16 a[0:31], a0, v1, a[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe8,0xd3,0x00,0x03,0x0a,0xec]
+v_mfma_f32_32x32x2bf16 a[0:31], a0, v1, a[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x2bf16 a[0:31], a0, v1, a[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe8,0xd3,0x00,0x03,0x8a,0xec]
 
-v_mfma_f32_32x32x2bf16 a[0:31], a0, a1, a[2:33]
-// GFX90A: v_mfma_f32_32x32x2bf16 a[0:31], a0, a1, a[2:33] ; encoding: [0x00,0x80,0xe8,0xd3,0x00,0x03,0x0a,0x1c]
+v_mfma_f32_32x32x2bf16 a[0:31], a0, a1, a[34:65]
+// GFX90A: v_mfma_f32_32x32x2bf16 a[0:31], a0, a1, a[34:65] ; encoding: [0x00,0x80,0xe8,0xd3,0x00,0x03,0x8a,0x1c]
 
-v_mfma_f32_32x32x2bf16 a[0:31], a0, a1, a[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x2bf16 a[0:31], a0, a1, a[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe8,0xd3,0x00,0x03,0x0a,0xfc]
+v_mfma_f32_32x32x2bf16 a[0:31], a0, a1, a[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x2bf16 a[0:31], a0, a1, a[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe8,0xd3,0x00,0x03,0x8a,0xfc]
 
-v_mfma_f32_32x32x2bf16 v[0:31], v0, v1, v[2:33]
-// GFX90A: v_mfma_f32_32x32x2bf16 v[0:31], v0, v1, v[2:33] ; encoding: [0x00,0x00,0xe8,0xd3,0x00,0x03,0x0a,0x04]
+v_mfma_f32_32x32x2bf16 v[0:31], v0, v1, v[34:65]
+// GFX90A: v_mfma_f32_32x32x2bf16 v[0:31], v0, v1, v[34:65] ; encoding: [0x00,0x00,0xe8,0xd3,0x00,0x03,0x8a,0x04]
 
-v_mfma_f32_32x32x2bf16 v[0:31], v0, v1, v[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x2bf16 v[0:31], v0, v1, v[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe8,0xd3,0x00,0x03,0x0a,0xe4]
+v_mfma_f32_32x32x2bf16 v[0:31], v0, v1, v[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x2bf16 v[0:31], v0, v1, v[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe8,0xd3,0x00,0x03,0x8a,0xe4]
 
-v_mfma_f32_32x32x2bf16 v[0:31], v0, a1, v[2:33]
-// GFX90A: v_mfma_f32_32x32x2bf16 v[0:31], v0, a1, v[2:33] ; encoding: [0x00,0x00,0xe8,0xd3,0x00,0x03,0x0a,0x14]
+v_mfma_f32_32x32x2bf16 v[0:31], v0, a1, v[34:65]
+// GFX90A: v_mfma_f32_32x32x2bf16 v[0:31], v0, a1, v[34:65] ; encoding: [0x00,0x00,0xe8,0xd3,0x00,0x03,0x8a,0x14]
 
-v_mfma_f32_32x32x2bf16 v[0:31], v0, a1, v[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x2bf16 v[0:31], v0, a1, v[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe8,0xd3,0x00,0x03,0x0a,0xf4]
+v_mfma_f32_32x32x2bf16 v[0:31], v0, a1, v[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x2bf16 v[0:31], v0, a1, v[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe8,0xd3,0x00,0x03,0x8a,0xf4]
 
-v_mfma_f32_32x32x2bf16 v[0:31], a0, v1, v[2:33]
-// GFX90A: v_mfma_f32_32x32x2bf16 v[0:31], a0, v1, v[2:33] ; encoding: [0x00,0x00,0xe8,0xd3,0x00,0x03,0x0a,0x0c]
+v_mfma_f32_32x32x2bf16 v[0:31], a0, v1, v[34:65]
+// GFX90A: v_mfma_f32_32x32x2bf16 v[0:31], a0, v1, v[34:65] ; encoding: [0x00,0x00,0xe8,0xd3,0x00,0x03,0x8a,0x0c]
 
-v_mfma_f32_32x32x2bf16 v[0:31], a0, v1, v[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x2bf16 v[0:31], a0, v1, v[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe8,0xd3,0x00,0x03,0x0a,0xec]
+v_mfma_f32_32x32x2bf16 v[0:31], a0, v1, v[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x2bf16 v[0:31], a0, v1, v[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe8,0xd3,0x00,0x03,0x8a,0xec]
 
-v_mfma_f32_32x32x2bf16 v[0:31], a0, a1, v[2:33]
-// GFX90A: v_mfma_f32_32x32x2bf16 v[0:31], a0, a1, v[2:33] ; encoding: [0x00,0x00,0xe8,0xd3,0x00,0x03,0x0a,0x1c]
+v_mfma_f32_32x32x2bf16 v[0:31], a0, a1, v[34:65]
+// GFX90A: v_mfma_f32_32x32x2bf16 v[0:31], a0, a1, v[34:65] ; encoding: [0x00,0x00,0xe8,0xd3,0x00,0x03,0x8a,0x1c]
 
-v_mfma_f32_32x32x2bf16 v[0:31], a0, a1, v[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x2bf16 v[0:31], a0, a1, v[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe8,0xd3,0x00,0x03,0x0a,0xfc]
+v_mfma_f32_32x32x2bf16 v[0:31], a0, a1, v[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x2bf16 v[0:31], a0, a1, v[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe8,0xd3,0x00,0x03,0x8a,0xfc]
 
 v_mfma_f32_32x32x2bf16 a[0:31], v0, v1, -2.0
 // GFX90A: v_mfma_f32_32x32x2bf16 a[0:31], v0, v1, -2.0 ; encoding: [0x00,0x80,0xe8,0xd3,0x00,0x03,0xd6,0x03]
@@ -1563,53 +1563,53 @@ v_mfma_f32_32x32x2bf16 a[0:31], a0, a1, -2.0 cbsz:3 abid:2 blgp:7
 v_mfma_f32_32x32x2bf16 v[0:31], a0, a1, -2.0 cbsz:3 abid:2 blgp:7
 // GFX90A: v_mfma_f32_32x32x2bf16 v[0:31], a0, a1, -2.0 cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe8,0xd3,0x00,0x03,0xd6,0xfb]
 
-v_mfma_f32_16x16x2bf16 a[0:15], v0, v1, a[2:17]
-// GFX90A: v_mfma_f32_16x16x2bf16 a[0:15], v0, v1, a[2:17] ; encoding: [0x00,0x80,0xe9,0xd3,0x00,0x03,0x0a,0x04]
+v_mfma_f32_16x16x2bf16 a[0:15], v0, v1, a[18:33]
+// GFX90A: v_mfma_f32_16x16x2bf16 a[0:15], v0, v1, a[18:33] ; encoding: [0x00,0x80,0xe9,0xd3,0x00,0x03,0x4a,0x04]
 
-v_mfma_f32_16x16x2bf16 a[0:15], v0, v1, a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x2bf16 a[0:15], v0, v1, a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe9,0xd3,0x00,0x03,0x0a,0xe4]
+v_mfma_f32_16x16x2bf16 a[0:15], v0, v1, a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x2bf16 a[0:15], v0, v1, a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe9,0xd3,0x00,0x03,0x4a,0xe4]
 
-v_mfma_f32_16x16x2bf16 a[0:15], v0, a1, a[2:17]
-// GFX90A: v_mfma_f32_16x16x2bf16 a[0:15], v0, a1, a[2:17] ; encoding: [0x00,0x80,0xe9,0xd3,0x00,0x03,0x0a,0x14]
+v_mfma_f32_16x16x2bf16 a[0:15], v0, a1, a[18:33]
+// GFX90A: v_mfma_f32_16x16x2bf16 a[0:15], v0, a1, a[18:33] ; encoding: [0x00,0x80,0xe9,0xd3,0x00,0x03,0x4a,0x14]
 
-v_mfma_f32_16x16x2bf16 a[0:15], v0, a1, a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x2bf16 a[0:15], v0, a1, a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe9,0xd3,0x00,0x03,0x0a,0xf4]
+v_mfma_f32_16x16x2bf16 a[0:15], v0, a1, a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x2bf16 a[0:15], v0, a1, a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe9,0xd3,0x00,0x03,0x4a,0xf4]
 
-v_mfma_f32_16x16x2bf16 a[0:15], a0, v1, a[2:17]
-// GFX90A: v_mfma_f32_16x16x2bf16 a[0:15], a0, v1, a[2:17] ; encoding: [0x00,0x80,0xe9,0xd3,0x00,0x03,0x0a,0x0c]
+v_mfma_f32_16x16x2bf16 a[0:15], a0, v1, a[18:33]
+// GFX90A: v_mfma_f32_16x16x2bf16 a[0:15], a0, v1, a[18:33] ; encoding: [0x00,0x80,0xe9,0xd3,0x00,0x03,0x4a,0x0c]
 
-v_mfma_f32_16x16x2bf16 a[0:15], a0, v1, a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x2bf16 a[0:15], a0, v1, a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe9,0xd3,0x00,0x03,0x0a,0xec]
+v_mfma_f32_16x16x2bf16 a[0:15], a0, v1, a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x2bf16 a[0:15], a0, v1, a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe9,0xd3,0x00,0x03,0x4a,0xec]
 
-v_mfma_f32_16x16x2bf16 a[0:15], a0, a1, a[2:17]
-// GFX90A: v_mfma_f32_16x16x2bf16 a[0:15], a0, a1, a[2:17] ; encoding: [0x00,0x80,0xe9,0xd3,0x00,0x03,0x0a,0x1c]
+v_mfma_f32_16x16x2bf16 a[0:15], a0, a1, a[18:33]
+// GFX90A: v_mfma_f32_16x16x2bf16 a[0:15], a0, a1, a[18:33] ; encoding: [0x00,0x80,0xe9,0xd3,0x00,0x03,0x4a,0x1c]
 
-v_mfma_f32_16x16x2bf16 a[0:15], a0, a1, a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x2bf16 a[0:15], a0, a1, a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe9,0xd3,0x00,0x03,0x0a,0xfc]
+v_mfma_f32_16x16x2bf16 a[0:15], a0, a1, a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x2bf16 a[0:15], a0, a1, a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe9,0xd3,0x00,0x03,0x4a,0xfc]
 
-v_mfma_f32_16x16x2bf16 v[0:15], v0, v1, v[2:17]
-// GFX90A: v_mfma_f32_16x16x2bf16 v[0:15], v0, v1, v[2:17] ; encoding: [0x00,0x00,0xe9,0xd3,0x00,0x03,0x0a,0x04]
+v_mfma_f32_16x16x2bf16 v[0:15], v0, v1, v[18:33]
+// GFX90A: v_mfma_f32_16x16x2bf16 v[0:15], v0, v1, v[18:33] ; encoding: [0x00,0x00,0xe9,0xd3,0x00,0x03,0x4a,0x04]
 
-v_mfma_f32_16x16x2bf16 v[0:15], v0, v1, v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x2bf16 v[0:15], v0, v1, v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe9,0xd3,0x00,0x03,0x0a,0xe4]
+v_mfma_f32_16x16x2bf16 v[0:15], v0, v1, v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x2bf16 v[0:15], v0, v1, v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe9,0xd3,0x00,0x03,0x4a,0xe4]
 
-v_mfma_f32_16x16x2bf16 v[0:15], v0, a1, v[2:17]
-// GFX90A: v_mfma_f32_16x16x2bf16 v[0:15], v0, a1, v[2:17] ; encoding: [0x00,0x00,0xe9,0xd3,0x00,0x03,0x0a,0x14]
+v_mfma_f32_16x16x2bf16 v[0:15], v0, a1, v[18:33]
+// GFX90A: v_mfma_f32_16x16x2bf16 v[0:15], v0, a1, v[18:33] ; encoding: [0x00,0x00,0xe9,0xd3,0x00,0x03,0x4a,0x14]
 
-v_mfma_f32_16x16x2bf16 v[0:15], v0, a1, v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x2bf16 v[0:15], v0, a1, v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe9,0xd3,0x00,0x03,0x0a,0xf4]
+v_mfma_f32_16x16x2bf16 v[0:15], v0, a1, v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x2bf16 v[0:15], v0, a1, v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe9,0xd3,0x00,0x03,0x4a,0xf4]
 
-v_mfma_f32_16x16x2bf16 v[0:15], a0, v1, v[2:17]
-// GFX90A: v_mfma_f32_16x16x2bf16 v[0:15], a0, v1, v[2:17] ; encoding: [0x00,0x00,0xe9,0xd3,0x00,0x03,0x0a,0x0c]
+v_mfma_f32_16x16x2bf16 v[0:15], a0, v1, v[18:33]
+// GFX90A: v_mfma_f32_16x16x2bf16 v[0:15], a0, v1, v[18:33] ; encoding: [0x00,0x00,0xe9,0xd3,0x00,0x03,0x4a,0x0c]
 
-v_mfma_f32_16x16x2bf16 v[0:15], a0, v1, v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x2bf16 v[0:15], a0, v1, v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe9,0xd3,0x00,0x03,0x0a,0xec]
+v_mfma_f32_16x16x2bf16 v[0:15], a0, v1, v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x2bf16 v[0:15], a0, v1, v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe9,0xd3,0x00,0x03,0x4a,0xec]
 
-v_mfma_f32_16x16x2bf16 v[0:15], a0, a1, v[2:17]
-// GFX90A: v_mfma_f32_16x16x2bf16 v[0:15], a0, a1, v[2:17] ; encoding: [0x00,0x00,0xe9,0xd3,0x00,0x03,0x0a,0x1c]
+v_mfma_f32_16x16x2bf16 v[0:15], a0, a1, v[18:33]
+// GFX90A: v_mfma_f32_16x16x2bf16 v[0:15], a0, a1, v[18:33] ; encoding: [0x00,0x00,0xe9,0xd3,0x00,0x03,0x4a,0x1c]
 
-v_mfma_f32_16x16x2bf16 v[0:15], a0, a1, v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x2bf16 v[0:15], a0, a1, v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe9,0xd3,0x00,0x03,0x0a,0xfc]
+v_mfma_f32_16x16x2bf16 v[0:15], a0, a1, v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x2bf16 v[0:15], a0, a1, v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe9,0xd3,0x00,0x03,0x4a,0xfc]
 
 v_mfma_f32_16x16x2bf16 a[0:15], v0, v1, -2.0
 // GFX90A: v_mfma_f32_16x16x2bf16 a[0:15], v0, v1, -2.0 ; encoding: [0x00,0x80,0xe9,0xd3,0x00,0x03,0xd6,0x03]
@@ -1755,53 +1755,53 @@ v_mfma_f32_4x4x2bf16 a[0:3], a0, a1, -2.0 cbsz:3 abid:2 blgp:7
 v_mfma_f32_4x4x2bf16 v[0:3], a0, a1, -2.0 cbsz:3 abid:2 blgp:7
 // GFX90A: v_mfma_f32_4x4x2bf16 v[0:3], a0, a1, -2.0 cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xeb,0xd3,0x00,0x03,0xd6,0xfb]
 
-v_mfma_f32_32x32x4bf16 a[0:15], v0, v1, a[2:17]
-// GFX90A: v_mfma_f32_32x32x4bf16 a[0:15], v0, v1, a[2:17] ; encoding: [0x00,0x80,0xec,0xd3,0x00,0x03,0x0a,0x04]
+v_mfma_f32_32x32x4bf16 a[0:15], v0, v1, a[18:33]
+// GFX90A: v_mfma_f32_32x32x4bf16 a[0:15], v0, v1, a[18:33] ; encoding: [0x00,0x80,0xec,0xd3,0x00,0x03,0x4a,0x04]
 
-v_mfma_f32_32x32x4bf16 a[0:15], v0, v1, a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x4bf16 a[0:15], v0, v1, a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xec,0xd3,0x00,0x03,0x0a,0xe4]
+v_mfma_f32_32x32x4bf16 a[0:15], v0, v1, a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x4bf16 a[0:15], v0, v1, a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xec,0xd3,0x00,0x03,0x4a,0xe4]
 
-v_mfma_f32_32x32x4bf16 a[0:15], v0, a1, a[2:17]
-// GFX90A: v_mfma_f32_32x32x4bf16 a[0:15], v0, a1, a[2:17] ; encoding: [0x00,0x80,0xec,0xd3,0x00,0x03,0x0a,0x14]
+v_mfma_f32_32x32x4bf16 a[0:15], v0, a1, a[18:33]
+// GFX90A: v_mfma_f32_32x32x4bf16 a[0:15], v0, a1, a[18:33] ; encoding: [0x00,0x80,0xec,0xd3,0x00,0x03,0x4a,0x14]
 
-v_mfma_f32_32x32x4bf16 a[0:15], v0, a1, a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x4bf16 a[0:15], v0, a1, a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xec,0xd3,0x00,0x03,0x0a,0xf4]
+v_mfma_f32_32x32x4bf16 a[0:15], v0, a1, a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x4bf16 a[0:15], v0, a1, a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xec,0xd3,0x00,0x03,0x4a,0xf4]
 
-v_mfma_f32_32x32x4bf16 a[0:15], a0, v1, a[2:17]
-// GFX90A: v_mfma_f32_32x32x4bf16 a[0:15], a0, v1, a[2:17] ; encoding: [0x00,0x80,0xec,0xd3,0x00,0x03,0x0a,0x0c]
+v_mfma_f32_32x32x4bf16 a[0:15], a0, v1, a[18:33]
+// GFX90A: v_mfma_f32_32x32x4bf16 a[0:15], a0, v1, a[18:33] ; encoding: [0x00,0x80,0xec,0xd3,0x00,0x03,0x4a,0x0c]
 
-v_mfma_f32_32x32x4bf16 a[0:15], a0, v1, a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x4bf16 a[0:15], a0, v1, a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xec,0xd3,0x00,0x03,0x0a,0xec]
+v_mfma_f32_32x32x4bf16 a[0:15], a0, v1, a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x4bf16 a[0:15], a0, v1, a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xec,0xd3,0x00,0x03,0x4a,0xec]
 
-v_mfma_f32_32x32x4bf16 a[0:15], a0, a1, a[2:17]
-// GFX90A: v_mfma_f32_32x32x4bf16 a[0:15], a0, a1, a[2:17] ; encoding: [0x00,0x80,0xec,0xd3,0x00,0x03,0x0a,0x1c]
+v_mfma_f32_32x32x4bf16 a[0:15], a0, a1, a[18:33]
+// GFX90A: v_mfma_f32_32x32x4bf16 a[0:15], a0, a1, a[18:33] ; encoding: [0x00,0x80,0xec,0xd3,0x00,0x03,0x4a,0x1c]
 
-v_mfma_f32_32x32x4bf16 a[0:15], a0, a1, a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x4bf16 a[0:15], a0, a1, a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xec,0xd3,0x00,0x03,0x0a,0xfc]
+v_mfma_f32_32x32x4bf16 a[0:15], a0, a1, a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x4bf16 a[0:15], a0, a1, a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xec,0xd3,0x00,0x03,0x4a,0xfc]
 
-v_mfma_f32_32x32x4bf16 v[0:15], v0, v1, v[2:17]
-// GFX90A: v_mfma_f32_32x32x4bf16 v[0:15], v0, v1, v[2:17] ; encoding: [0x00,0x00,0xec,0xd3,0x00,0x03,0x0a,0x04]
+v_mfma_f32_32x32x4bf16 v[0:15], v0, v1, v[18:33]
+// GFX90A: v_mfma_f32_32x32x4bf16 v[0:15], v0, v1, v[18:33] ; encoding: [0x00,0x00,0xec,0xd3,0x00,0x03,0x4a,0x04]
 
-v_mfma_f32_32x32x4bf16 v[0:15], v0, v1, v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x4bf16 v[0:15], v0, v1, v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xec,0xd3,0x00,0x03,0x0a,0xe4]
+v_mfma_f32_32x32x4bf16 v[0:15], v0, v1, v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x4bf16 v[0:15], v0, v1, v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xec,0xd3,0x00,0x03,0x4a,0xe4]
 
-v_mfma_f32_32x32x4bf16 v[0:15], v0, a1, v[2:17]
-// GFX90A: v_mfma_f32_32x32x4bf16 v[0:15], v0, a1, v[2:17] ; encoding: [0x00,0x00,0xec,0xd3,0x00,0x03,0x0a,0x14]
+v_mfma_f32_32x32x4bf16 v[0:15], v0, a1, v[18:33]
+// GFX90A: v_mfma_f32_32x32x4bf16 v[0:15], v0, a1, v[18:33] ; encoding: [0x00,0x00,0xec,0xd3,0x00,0x03,0x4a,0x14]
 
-v_mfma_f32_32x32x4bf16 v[0:15], v0, a1, v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x4bf16 v[0:15], v0, a1, v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xec,0xd3,0x00,0x03,0x0a,0xf4]
+v_mfma_f32_32x32x4bf16 v[0:15], v0, a1, v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x4bf16 v[0:15], v0, a1, v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xec,0xd3,0x00,0x03,0x4a,0xf4]
 
-v_mfma_f32_32x32x4bf16 v[0:15], a0, v1, v[2:17]
-// GFX90A: v_mfma_f32_32x32x4bf16 v[0:15], a0, v1, v[2:17] ; encoding: [0x00,0x00,0xec,0xd3,0x00,0x03,0x0a,0x0c]
+v_mfma_f32_32x32x4bf16 v[0:15], a0, v1, v[18:33]
+// GFX90A: v_mfma_f32_32x32x4bf16 v[0:15], a0, v1, v[18:33] ; encoding: [0x00,0x00,0xec,0xd3,0x00,0x03,0x4a,0x0c]
 
-v_mfma_f32_32x32x4bf16 v[0:15], a0, v1, v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x4bf16 v[0:15], a0, v1, v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xec,0xd3,0x00,0x03,0x0a,0xec]
+v_mfma_f32_32x32x4bf16 v[0:15], a0, v1, v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x4bf16 v[0:15], a0, v1, v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xec,0xd3,0x00,0x03,0x4a,0xec]
 
-v_mfma_f32_32x32x4bf16 v[0:15], a0, a1, v[2:17]
-// GFX90A: v_mfma_f32_32x32x4bf16 v[0:15], a0, a1, v[2:17] ; encoding: [0x00,0x00,0xec,0xd3,0x00,0x03,0x0a,0x1c]
+v_mfma_f32_32x32x4bf16 v[0:15], a0, a1, v[18:33]
+// GFX90A: v_mfma_f32_32x32x4bf16 v[0:15], a0, a1, v[18:33] ; encoding: [0x00,0x00,0xec,0xd3,0x00,0x03,0x4a,0x1c]
 
-v_mfma_f32_32x32x4bf16 v[0:15], a0, a1, v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x4bf16 v[0:15], a0, a1, v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xec,0xd3,0x00,0x03,0x0a,0xfc]
+v_mfma_f32_32x32x4bf16 v[0:15], a0, a1, v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x4bf16 v[0:15], a0, a1, v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xec,0xd3,0x00,0x03,0x4a,0xfc]
 
 v_mfma_f32_32x32x4bf16 a[0:15], v0, v1, -2.0
 // GFX90A: v_mfma_f32_32x32x4bf16 a[0:15], v0, v1, -2.0 ; encoding: [0x00,0x80,0xec,0xd3,0x00,0x03,0xd6,0x03]
@@ -1947,53 +1947,53 @@ v_mfma_f32_16x16x8bf16 a[0:3], a0, a1, -2.0 cbsz:3 abid:2 blgp:7
 v_mfma_f32_16x16x8bf16 v[0:3], a0, a1, -2.0 cbsz:3 abid:2 blgp:7
 // GFX90A: v_mfma_f32_16x16x8bf16 v[0:3], a0, a1, -2.0 cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xed,0xd3,0x00,0x03,0xd6,0xfb]
 
-v_mfma_f32_32x32x4bf16_1k a[0:31], v[0:1], v[2:3], a[2:33]
-// GFX90A: v_mfma_f32_32x32x4bf16_1k a[0:31], v[0:1], v[2:3], a[2:33] ; encoding: [0x00,0x80,0xe3,0xd3,0x00,0x05,0x0a,0x04]
+v_mfma_f32_32x32x4bf16_1k a[0:31], v[0:1], v[2:3], a[34:65]
+// GFX90A: v_mfma_f32_32x32x4bf16_1k a[0:31], v[0:1], v[2:3], a[34:65] ; encoding: [0x00,0x80,0xe3,0xd3,0x00,0x05,0x8a,0x04]
 
-v_mfma_f32_32x32x4bf16_1k a[0:31], v[0:1], v[2:3], a[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x4bf16_1k a[0:31], v[0:1], v[2:3], a[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe3,0xd3,0x00,0x05,0x0a,0xe4]
+v_mfma_f32_32x32x4bf16_1k a[0:31], v[0:1], v[2:3], a[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x4bf16_1k a[0:31], v[0:1], v[2:3], a[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe3,0xd3,0x00,0x05,0x8a,0xe4]
 
-v_mfma_f32_32x32x4bf16_1k a[0:31], v[0:1], a[2:3], a[2:33]
-// GFX90A: v_mfma_f32_32x32x4bf16_1k a[0:31], v[0:1], a[2:3], a[2:33] ; encoding: [0x00,0x80,0xe3,0xd3,0x00,0x05,0x0a,0x14]
+v_mfma_f32_32x32x4bf16_1k a[0:31], v[0:1], a[2:3], a[34:65]
+// GFX90A: v_mfma_f32_32x32x4bf16_1k a[0:31], v[0:1], a[2:3], a[34:65] ; encoding: [0x00,0x80,0xe3,0xd3,0x00,0x05,0x8a,0x14]
 
-v_mfma_f32_32x32x4bf16_1k a[0:31], v[0:1], a[2:3], a[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x4bf16_1k a[0:31], v[0:1], a[2:3], a[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe3,0xd3,0x00,0x05,0x0a,0xf4]
+v_mfma_f32_32x32x4bf16_1k a[0:31], v[0:1], a[2:3], a[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x4bf16_1k a[0:31], v[0:1], a[2:3], a[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe3,0xd3,0x00,0x05,0x8a,0xf4]
 
-v_mfma_f32_32x32x4bf16_1k a[0:31], a[0:1], v[2:3], a[2:33]
-// GFX90A: v_mfma_f32_32x32x4bf16_1k a[0:31], a[0:1], v[2:3], a[2:33] ; encoding: [0x00,0x80,0xe3,0xd3,0x00,0x05,0x0a,0x0c]
+v_mfma_f32_32x32x4bf16_1k a[0:31], a[0:1], v[2:3], a[34:65]
+// GFX90A: v_mfma_f32_32x32x4bf16_1k a[0:31], a[0:1], v[2:3], a[34:65] ; encoding: [0x00,0x80,0xe3,0xd3,0x00,0x05,0x8a,0x0c]
 
-v_mfma_f32_32x32x4bf16_1k a[0:31], a[0:1], v[2:3], a[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x4bf16_1k a[0:31], a[0:1], v[2:3], a[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe3,0xd3,0x00,0x05,0x0a,0xec]
+v_mfma_f32_32x32x4bf16_1k a[0:31], a[0:1], v[2:3], a[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x4bf16_1k a[0:31], a[0:1], v[2:3], a[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe3,0xd3,0x00,0x05,0x8a,0xec]
 
-v_mfma_f32_32x32x4bf16_1k a[0:31], a[0:1], a[2:3], a[2:33]
-// GFX90A: v_mfma_f32_32x32x4bf16_1k a[0:31], a[0:1], a[2:3], a[2:33] ; encoding: [0x00,0x80,0xe3,0xd3,0x00,0x05,0x0a,0x1c]
+v_mfma_f32_32x32x4bf16_1k a[0:31], a[0:1], a[2:3], a[34:65]
+// GFX90A: v_mfma_f32_32x32x4bf16_1k a[0:31], a[0:1], a[2:3], a[34:65] ; encoding: [0x00,0x80,0xe3,0xd3,0x00,0x05,0x8a,0x1c]
 
-v_mfma_f32_32x32x4bf16_1k a[0:31], a[0:1], a[2:3], a[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x4bf16_1k a[0:31], a[0:1], a[2:3], a[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe3,0xd3,0x00,0x05,0x0a,0xfc]
+v_mfma_f32_32x32x4bf16_1k a[0:31], a[0:1], a[2:3], a[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x4bf16_1k a[0:31], a[0:1], a[2:3], a[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe3,0xd3,0x00,0x05,0x8a,0xfc]
 
-v_mfma_f32_32x32x4bf16_1k v[0:31], v[0:1], v[2:3], v[2:33]
-// GFX90A: v_mfma_f32_32x32x4bf16_1k v[0:31], v[0:1], v[2:3], v[2:33] ; encoding: [0x00,0x00,0xe3,0xd3,0x00,0x05,0x0a,0x04]
+v_mfma_f32_32x32x4bf16_1k v[0:31], v[0:1], v[2:3], v[34:65]
+// GFX90A: v_mfma_f32_32x32x4bf16_1k v[0:31], v[0:1], v[2:3], v[34:65] ; encoding: [0x00,0x00,0xe3,0xd3,0x00,0x05,0x8a,0x04]
 
-v_mfma_f32_32x32x4bf16_1k v[0:31], v[0:1], v[2:3], v[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x4bf16_1k v[0:31], v[0:1], v[2:3], v[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe3,0xd3,0x00,0x05,0x0a,0xe4]
+v_mfma_f32_32x32x4bf16_1k v[0:31], v[0:1], v[2:3], v[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x4bf16_1k v[0:31], v[0:1], v[2:3], v[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe3,0xd3,0x00,0x05,0x8a,0xe4]
 
-v_mfma_f32_32x32x4bf16_1k v[0:31], v[0:1], a[2:3], v[2:33]
-// GFX90A: v_mfma_f32_32x32x4bf16_1k v[0:31], v[0:1], a[2:3], v[2:33] ; encoding: [0x00,0x00,0xe3,0xd3,0x00,0x05,0x0a,0x14]
+v_mfma_f32_32x32x4bf16_1k v[0:31], v[0:1], a[2:3], v[34:65]
+// GFX90A: v_mfma_f32_32x32x4bf16_1k v[0:31], v[0:1], a[2:3], v[34:65] ; encoding: [0x00,0x00,0xe3,0xd3,0x00,0x05,0x8a,0x14]
 
-v_mfma_f32_32x32x4bf16_1k v[0:31], v[0:1], a[2:3], v[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x4bf16_1k v[0:31], v[0:1], a[2:3], v[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe3,0xd3,0x00,0x05,0x0a,0xf4]
+v_mfma_f32_32x32x4bf16_1k v[0:31], v[0:1], a[2:3], v[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x4bf16_1k v[0:31], v[0:1], a[2:3], v[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe3,0xd3,0x00,0x05,0x8a,0xf4]
 
-v_mfma_f32_32x32x4bf16_1k v[0:31], a[0:1], v[2:3], v[2:33]
-// GFX90A: v_mfma_f32_32x32x4bf16_1k v[0:31], a[0:1], v[2:3], v[2:33] ; encoding: [0x00,0x00,0xe3,0xd3,0x00,0x05,0x0a,0x0c]
+v_mfma_f32_32x32x4bf16_1k v[0:31], a[0:1], v[2:3], v[34:65]
+// GFX90A: v_mfma_f32_32x32x4bf16_1k v[0:31], a[0:1], v[2:3], v[34:65] ; encoding: [0x00,0x00,0xe3,0xd3,0x00,0x05,0x8a,0x0c]
 
-v_mfma_f32_32x32x4bf16_1k v[0:31], a[0:1], v[2:3], v[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x4bf16_1k v[0:31], a[0:1], v[2:3], v[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe3,0xd3,0x00,0x05,0x0a,0xec]
+v_mfma_f32_32x32x4bf16_1k v[0:31], a[0:1], v[2:3], v[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x4bf16_1k v[0:31], a[0:1], v[2:3], v[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe3,0xd3,0x00,0x05,0x8a,0xec]
 
-v_mfma_f32_32x32x4bf16_1k v[0:31], a[0:1], a[2:3], v[2:33]
-// GFX90A: v_mfma_f32_32x32x4bf16_1k v[0:31], a[0:1], a[2:3], v[2:33] ; encoding: [0x00,0x00,0xe3,0xd3,0x00,0x05,0x0a,0x1c]
+v_mfma_f32_32x32x4bf16_1k v[0:31], a[0:1], a[2:3], v[34:65]
+// GFX90A: v_mfma_f32_32x32x4bf16_1k v[0:31], a[0:1], a[2:3], v[34:65] ; encoding: [0x00,0x00,0xe3,0xd3,0x00,0x05,0x8a,0x1c]
 
-v_mfma_f32_32x32x4bf16_1k v[0:31], a[0:1], a[2:3], v[2:33] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x4bf16_1k v[0:31], a[0:1], a[2:3], v[2:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe3,0xd3,0x00,0x05,0x0a,0xfc]
+v_mfma_f32_32x32x4bf16_1k v[0:31], a[0:1], a[2:3], v[34:65] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x4bf16_1k v[0:31], a[0:1], a[2:3], v[34:65] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe3,0xd3,0x00,0x05,0x8a,0xfc]
 
 v_mfma_f32_32x32x4bf16_1k a[0:31], v[0:1], v[2:3], -2.0
 // GFX90A: v_mfma_f32_32x32x4bf16_1k a[0:31], v[0:1], v[2:3], -2.0 ; encoding: [0x00,0x80,0xe3,0xd3,0x00,0x05,0xd6,0x03]
@@ -2043,53 +2043,53 @@ v_mfma_f32_32x32x4bf16_1k a[0:31], a[0:1], a[2:3], -2.0 cbsz:3 abid:2 blgp:7
 v_mfma_f32_32x32x4bf16_1k v[0:31], a[0:1], a[2:3], -2.0 cbsz:3 abid:2 blgp:7
 // GFX90A: v_mfma_f32_32x32x4bf16_1k v[0:31], a[0:1], a[2:3], -2.0 cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe3,0xd3,0x00,0x05,0xd6,0xfb]
 
-v_mfma_f32_16x16x4bf16_1k a[0:15], v[0:1], v[2:3], a[2:17]
-// GFX90A: v_mfma_f32_16x16x4bf16_1k a[0:15], v[0:1], v[2:3], a[2:17] ; encoding: [0x00,0x80,0xe4,0xd3,0x00,0x05,0x0a,0x04]
+v_mfma_f32_16x16x4bf16_1k a[0:15], v[0:1], v[2:3], a[18:33]
+// GFX90A: v_mfma_f32_16x16x4bf16_1k a[0:15], v[0:1], v[2:3], a[18:33] ; encoding: [0x00,0x80,0xe4,0xd3,0x00,0x05,0x4a,0x04]
 
-v_mfma_f32_16x16x4bf16_1k a[0:15], v[0:1], v[2:3], a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x4bf16_1k a[0:15], v[0:1], v[2:3], a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe4,0xd3,0x00,0x05,0x0a,0xe4]
+v_mfma_f32_16x16x4bf16_1k a[0:15], v[0:1], v[2:3], a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x4bf16_1k a[0:15], v[0:1], v[2:3], a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe4,0xd3,0x00,0x05,0x4a,0xe4]
 
-v_mfma_f32_16x16x4bf16_1k a[0:15], v[0:1], a[2:3], a[2:17]
-// GFX90A: v_mfma_f32_16x16x4bf16_1k a[0:15], v[0:1], a[2:3], a[2:17] ; encoding: [0x00,0x80,0xe4,0xd3,0x00,0x05,0x0a,0x14]
+v_mfma_f32_16x16x4bf16_1k a[0:15], v[0:1], a[2:3], a[18:33]
+// GFX90A: v_mfma_f32_16x16x4bf16_1k a[0:15], v[0:1], a[2:3], a[18:33] ; encoding: [0x00,0x80,0xe4,0xd3,0x00,0x05,0x4a,0x14]
 
-v_mfma_f32_16x16x4bf16_1k a[0:15], v[0:1], a[2:3], a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x4bf16_1k a[0:15], v[0:1], a[2:3], a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe4,0xd3,0x00,0x05,0x0a,0xf4]
+v_mfma_f32_16x16x4bf16_1k a[0:15], v[0:1], a[2:3], a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x4bf16_1k a[0:15], v[0:1], a[2:3], a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe4,0xd3,0x00,0x05,0x4a,0xf4]
 
-v_mfma_f32_16x16x4bf16_1k a[0:15], a[0:1], v[2:3], a[2:17]
-// GFX90A: v_mfma_f32_16x16x4bf16_1k a[0:15], a[0:1], v[2:3], a[2:17] ; encoding: [0x00,0x80,0xe4,0xd3,0x00,0x05,0x0a,0x0c]
+v_mfma_f32_16x16x4bf16_1k a[0:15], a[0:1], v[2:3], a[18:33]
+// GFX90A: v_mfma_f32_16x16x4bf16_1k a[0:15], a[0:1], v[2:3], a[18:33] ; encoding: [0x00,0x80,0xe4,0xd3,0x00,0x05,0x4a,0x0c]
 
-v_mfma_f32_16x16x4bf16_1k a[0:15], a[0:1], v[2:3], a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x4bf16_1k a[0:15], a[0:1], v[2:3], a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe4,0xd3,0x00,0x05,0x0a,0xec]
+v_mfma_f32_16x16x4bf16_1k a[0:15], a[0:1], v[2:3], a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x4bf16_1k a[0:15], a[0:1], v[2:3], a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe4,0xd3,0x00,0x05,0x4a,0xec]
 
-v_mfma_f32_16x16x4bf16_1k a[0:15], a[0:1], a[2:3], a[2:17]
-// GFX90A: v_mfma_f32_16x16x4bf16_1k a[0:15], a[0:1], a[2:3], a[2:17] ; encoding: [0x00,0x80,0xe4,0xd3,0x00,0x05,0x0a,0x1c]
+v_mfma_f32_16x16x4bf16_1k a[0:15], a[0:1], a[2:3], a[18:33]
+// GFX90A: v_mfma_f32_16x16x4bf16_1k a[0:15], a[0:1], a[2:3], a[18:33] ; encoding: [0x00,0x80,0xe4,0xd3,0x00,0x05,0x4a,0x1c]
 
-v_mfma_f32_16x16x4bf16_1k a[0:15], a[0:1], a[2:3], a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x4bf16_1k a[0:15], a[0:1], a[2:3], a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe4,0xd3,0x00,0x05,0x0a,0xfc]
+v_mfma_f32_16x16x4bf16_1k a[0:15], a[0:1], a[2:3], a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x4bf16_1k a[0:15], a[0:1], a[2:3], a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe4,0xd3,0x00,0x05,0x4a,0xfc]
 
-v_mfma_f32_16x16x4bf16_1k v[0:15], v[0:1], v[2:3], v[2:17]
-// GFX90A: v_mfma_f32_16x16x4bf16_1k v[0:15], v[0:1], v[2:3], v[2:17] ; encoding: [0x00,0x00,0xe4,0xd3,0x00,0x05,0x0a,0x04]
+v_mfma_f32_16x16x4bf16_1k v[0:15], v[0:1], v[2:3], v[18:33]
+// GFX90A: v_mfma_f32_16x16x4bf16_1k v[0:15], v[0:1], v[2:3], v[18:33] ; encoding: [0x00,0x00,0xe4,0xd3,0x00,0x05,0x4a,0x04]
 
-v_mfma_f32_16x16x4bf16_1k v[0:15], v[0:1], v[2:3], v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x4bf16_1k v[0:15], v[0:1], v[2:3], v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe4,0xd3,0x00,0x05,0x0a,0xe4]
+v_mfma_f32_16x16x4bf16_1k v[0:15], v[0:1], v[2:3], v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x4bf16_1k v[0:15], v[0:1], v[2:3], v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe4,0xd3,0x00,0x05,0x4a,0xe4]
 
-v_mfma_f32_16x16x4bf16_1k v[0:15], v[0:1], a[2:3], v[2:17]
-// GFX90A: v_mfma_f32_16x16x4bf16_1k v[0:15], v[0:1], a[2:3], v[2:17] ; encoding: [0x00,0x00,0xe4,0xd3,0x00,0x05,0x0a,0x14]
+v_mfma_f32_16x16x4bf16_1k v[0:15], v[0:1], a[2:3], v[18:33]
+// GFX90A: v_mfma_f32_16x16x4bf16_1k v[0:15], v[0:1], a[2:3], v[18:33] ; encoding: [0x00,0x00,0xe4,0xd3,0x00,0x05,0x4a,0x14]
 
-v_mfma_f32_16x16x4bf16_1k v[0:15], v[0:1], a[2:3], v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x4bf16_1k v[0:15], v[0:1], a[2:3], v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe4,0xd3,0x00,0x05,0x0a,0xf4]
+v_mfma_f32_16x16x4bf16_1k v[0:15], v[0:1], a[2:3], v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x4bf16_1k v[0:15], v[0:1], a[2:3], v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe4,0xd3,0x00,0x05,0x4a,0xf4]
 
-v_mfma_f32_16x16x4bf16_1k v[0:15], a[0:1], v[2:3], v[2:17]
-// GFX90A: v_mfma_f32_16x16x4bf16_1k v[0:15], a[0:1], v[2:3], v[2:17] ; encoding: [0x00,0x00,0xe4,0xd3,0x00,0x05,0x0a,0x0c]
+v_mfma_f32_16x16x4bf16_1k v[0:15], a[0:1], v[2:3], v[18:33]
+// GFX90A: v_mfma_f32_16x16x4bf16_1k v[0:15], a[0:1], v[2:3], v[18:33] ; encoding: [0x00,0x00,0xe4,0xd3,0x00,0x05,0x4a,0x0c]
 
-v_mfma_f32_16x16x4bf16_1k v[0:15], a[0:1], v[2:3], v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x4bf16_1k v[0:15], a[0:1], v[2:3], v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe4,0xd3,0x00,0x05,0x0a,0xec]
+v_mfma_f32_16x16x4bf16_1k v[0:15], a[0:1], v[2:3], v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x4bf16_1k v[0:15], a[0:1], v[2:3], v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe4,0xd3,0x00,0x05,0x4a,0xec]
 
-v_mfma_f32_16x16x4bf16_1k v[0:15], a[0:1], a[2:3], v[2:17]
-// GFX90A: v_mfma_f32_16x16x4bf16_1k v[0:15], a[0:1], a[2:3], v[2:17] ; encoding: [0x00,0x00,0xe4,0xd3,0x00,0x05,0x0a,0x1c]
+v_mfma_f32_16x16x4bf16_1k v[0:15], a[0:1], a[2:3], v[18:33]
+// GFX90A: v_mfma_f32_16x16x4bf16_1k v[0:15], a[0:1], a[2:3], v[18:33] ; encoding: [0x00,0x00,0xe4,0xd3,0x00,0x05,0x4a,0x1c]
 
-v_mfma_f32_16x16x4bf16_1k v[0:15], a[0:1], a[2:3], v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_16x16x4bf16_1k v[0:15], a[0:1], a[2:3], v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe4,0xd3,0x00,0x05,0x0a,0xfc]
+v_mfma_f32_16x16x4bf16_1k v[0:15], a[0:1], a[2:3], v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_16x16x4bf16_1k v[0:15], a[0:1], a[2:3], v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe4,0xd3,0x00,0x05,0x4a,0xfc]
 
 v_mfma_f32_16x16x4bf16_1k a[0:15], v[0:1], v[2:3], -2.0
 // GFX90A: v_mfma_f32_16x16x4bf16_1k a[0:15], v[0:1], v[2:3], -2.0 ; encoding: [0x00,0x80,0xe4,0xd3,0x00,0x05,0xd6,0x03]
@@ -2235,53 +2235,53 @@ v_mfma_f32_4x4x4bf16_1k a[0:3], a[0:1], a[2:3], -2.0 cbsz:3 abid:2 blgp:7
 v_mfma_f32_4x4x4bf16_1k v[0:3], a[0:1], a[2:3], -2.0 cbsz:3 abid:2 blgp:7
 // GFX90A: v_mfma_f32_4x4x4bf16_1k v[0:3], a[0:1], a[2:3], -2.0 cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe5,0xd3,0x00,0x05,0xd6,0xfb]
 
-v_mfma_f32_32x32x8bf16_1k a[0:15], v[0:1], v[2:3], a[2:17]
-// GFX90A: v_mfma_f32_32x32x8bf16_1k a[0:15], v[0:1], v[2:3], a[2:17] ; encoding: [0x00,0x80,0xe6,0xd3,0x00,0x05,0x0a,0x04]
+v_mfma_f32_32x32x8bf16_1k a[0:15], v[0:1], v[2:3], a[18:33]
+// GFX90A: v_mfma_f32_32x32x8bf16_1k a[0:15], v[0:1], v[2:3], a[18:33] ; encoding: [0x00,0x80,0xe6,0xd3,0x00,0x05,0x4a,0x04]
 
-v_mfma_f32_32x32x8bf16_1k a[0:15], v[0:1], v[2:3], a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x8bf16_1k a[0:15], v[0:1], v[2:3], a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe6,0xd3,0x00,0x05,0x0a,0xe4]
+v_mfma_f32_32x32x8bf16_1k a[0:15], v[0:1], v[2:3], a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x8bf16_1k a[0:15], v[0:1], v[2:3], a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe6,0xd3,0x00,0x05,0x4a,0xe4]
 
-v_mfma_f32_32x32x8bf16_1k a[0:15], v[0:1], a[2:3], a[2:17]
-// GFX90A: v_mfma_f32_32x32x8bf16_1k a[0:15], v[0:1], a[2:3], a[2:17] ; encoding: [0x00,0x80,0xe6,0xd3,0x00,0x05,0x0a,0x14]
+v_mfma_f32_32x32x8bf16_1k a[0:15], v[0:1], a[2:3], a[18:33]
+// GFX90A: v_mfma_f32_32x32x8bf16_1k a[0:15], v[0:1], a[2:3], a[18:33] ; encoding: [0x00,0x80,0xe6,0xd3,0x00,0x05,0x4a,0x14]
 
-v_mfma_f32_32x32x8bf16_1k a[0:15], v[0:1], a[2:3], a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x8bf16_1k a[0:15], v[0:1], a[2:3], a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe6,0xd3,0x00,0x05,0x0a,0xf4]
+v_mfma_f32_32x32x8bf16_1k a[0:15], v[0:1], a[2:3], a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x8bf16_1k a[0:15], v[0:1], a[2:3], a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe6,0xd3,0x00,0x05,0x4a,0xf4]
 
-v_mfma_f32_32x32x8bf16_1k a[0:15], a[0:1], v[2:3], a[2:17]
-// GFX90A: v_mfma_f32_32x32x8bf16_1k a[0:15], a[0:1], v[2:3], a[2:17] ; encoding: [0x00,0x80,0xe6,0xd3,0x00,0x05,0x0a,0x0c]
+v_mfma_f32_32x32x8bf16_1k a[0:15], a[0:1], v[2:3], a[18:33]
+// GFX90A: v_mfma_f32_32x32x8bf16_1k a[0:15], a[0:1], v[2:3], a[18:33] ; encoding: [0x00,0x80,0xe6,0xd3,0x00,0x05,0x4a,0x0c]
 
-v_mfma_f32_32x32x8bf16_1k a[0:15], a[0:1], v[2:3], a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x8bf16_1k a[0:15], a[0:1], v[2:3], a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe6,0xd3,0x00,0x05,0x0a,0xec]
+v_mfma_f32_32x32x8bf16_1k a[0:15], a[0:1], v[2:3], a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x8bf16_1k a[0:15], a[0:1], v[2:3], a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe6,0xd3,0x00,0x05,0x4a,0xec]
 
-v_mfma_f32_32x32x8bf16_1k a[0:15], a[0:1], a[2:3], a[2:17]
-// GFX90A: v_mfma_f32_32x32x8bf16_1k a[0:15], a[0:1], a[2:3], a[2:17] ; encoding: [0x00,0x80,0xe6,0xd3,0x00,0x05,0x0a,0x1c]
+v_mfma_f32_32x32x8bf16_1k a[0:15], a[0:1], a[2:3], a[18:33]
+// GFX90A: v_mfma_f32_32x32x8bf16_1k a[0:15], a[0:1], a[2:3], a[18:33] ; encoding: [0x00,0x80,0xe6,0xd3,0x00,0x05,0x4a,0x1c]
 
-v_mfma_f32_32x32x8bf16_1k a[0:15], a[0:1], a[2:3], a[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x8bf16_1k a[0:15], a[0:1], a[2:3], a[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe6,0xd3,0x00,0x05,0x0a,0xfc]
+v_mfma_f32_32x32x8bf16_1k a[0:15], a[0:1], a[2:3], a[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x8bf16_1k a[0:15], a[0:1], a[2:3], a[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xe6,0xd3,0x00,0x05,0x4a,0xfc]
 
-v_mfma_f32_32x32x8bf16_1k v[0:15], v[0:1], v[2:3], v[2:17]
-// GFX90A: v_mfma_f32_32x32x8bf16_1k v[0:15], v[0:1], v[2:3], v[2:17] ; encoding: [0x00,0x00,0xe6,0xd3,0x00,0x05,0x0a,0x04]
+v_mfma_f32_32x32x8bf16_1k v[0:15], v[0:1], v[2:3], v[18:33]
+// GFX90A: v_mfma_f32_32x32x8bf16_1k v[0:15], v[0:1], v[2:3], v[18:33] ; encoding: [0x00,0x00,0xe6,0xd3,0x00,0x05,0x4a,0x04]
 
-v_mfma_f32_32x32x8bf16_1k v[0:15], v[0:1], v[2:3], v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x8bf16_1k v[0:15], v[0:1], v[2:3], v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe6,0xd3,0x00,0x05,0x0a,0xe4]
+v_mfma_f32_32x32x8bf16_1k v[0:15], v[0:1], v[2:3], v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x8bf16_1k v[0:15], v[0:1], v[2:3], v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe6,0xd3,0x00,0x05,0x4a,0xe4]
 
-v_mfma_f32_32x32x8bf16_1k v[0:15], v[0:1], a[2:3], v[2:17]
-// GFX90A: v_mfma_f32_32x32x8bf16_1k v[0:15], v[0:1], a[2:3], v[2:17] ; encoding: [0x00,0x00,0xe6,0xd3,0x00,0x05,0x0a,0x14]
+v_mfma_f32_32x32x8bf16_1k v[0:15], v[0:1], a[2:3], v[18:33]
+// GFX90A: v_mfma_f32_32x32x8bf16_1k v[0:15], v[0:1], a[2:3], v[18:33] ; encoding: [0x00,0x00,0xe6,0xd3,0x00,0x05,0x4a,0x14]
 
-v_mfma_f32_32x32x8bf16_1k v[0:15], v[0:1], a[2:3], v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x8bf16_1k v[0:15], v[0:1], a[2:3], v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe6,0xd3,0x00,0x05,0x0a,0xf4]
+v_mfma_f32_32x32x8bf16_1k v[0:15], v[0:1], a[2:3], v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x8bf16_1k v[0:15], v[0:1], a[2:3], v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe6,0xd3,0x00,0x05,0x4a,0xf4]
 
-v_mfma_f32_32x32x8bf16_1k v[0:15], a[0:1], v[2:3], v[2:17]
-// GFX90A: v_mfma_f32_32x32x8bf16_1k v[0:15], a[0:1], v[2:3], v[2:17] ; encoding: [0x00,0x00,0xe6,0xd3,0x00,0x05,0x0a,0x0c]
+v_mfma_f32_32x32x8bf16_1k v[0:15], a[0:1], v[2:3], v[18:33]
+// GFX90A: v_mfma_f32_32x32x8bf16_1k v[0:15], a[0:1], v[2:3], v[18:33] ; encoding: [0x00,0x00,0xe6,0xd3,0x00,0x05,0x4a,0x0c]
 
-v_mfma_f32_32x32x8bf16_1k v[0:15], a[0:1], v[2:3], v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x8bf16_1k v[0:15], a[0:1], v[2:3], v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe6,0xd3,0x00,0x05,0x0a,0xec]
+v_mfma_f32_32x32x8bf16_1k v[0:15], a[0:1], v[2:3], v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x8bf16_1k v[0:15], a[0:1], v[2:3], v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe6,0xd3,0x00,0x05,0x4a,0xec]
 
-v_mfma_f32_32x32x8bf16_1k v[0:15], a[0:1], a[2:3], v[2:17]
-// GFX90A: v_mfma_f32_32x32x8bf16_1k v[0:15], a[0:1], a[2:3], v[2:17] ; encoding: [0x00,0x00,0xe6,0xd3,0x00,0x05,0x0a,0x1c]
+v_mfma_f32_32x32x8bf16_1k v[0:15], a[0:1], a[2:3], v[18:33]
+// GFX90A: v_mfma_f32_32x32x8bf16_1k v[0:15], a[0:1], a[2:3], v[18:33] ; encoding: [0x00,0x00,0xe6,0xd3,0x00,0x05,0x4a,0x1c]
 
-v_mfma_f32_32x32x8bf16_1k v[0:15], a[0:1], a[2:3], v[2:17] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f32_32x32x8bf16_1k v[0:15], a[0:1], a[2:3], v[2:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe6,0xd3,0x00,0x05,0x0a,0xfc]
+v_mfma_f32_32x32x8bf16_1k v[0:15], a[0:1], a[2:3], v[18:33] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f32_32x32x8bf16_1k v[0:15], a[0:1], a[2:3], v[18:33] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe6,0xd3,0x00,0x05,0x4a,0xfc]
 
 v_mfma_f32_32x32x8bf16_1k a[0:15], v[0:1], v[2:3], -2.0
 // GFX90A: v_mfma_f32_32x32x8bf16_1k a[0:15], v[0:1], v[2:3], -2.0 ; encoding: [0x00,0x80,0xe6,0xd3,0x00,0x05,0xd6,0x03]
@@ -2427,11 +2427,11 @@ v_mfma_f32_16x16x16bf16_1k a[0:3], a[0:1], a[2:3], -2.0 cbsz:3 abid:2 blgp:7
 v_mfma_f32_16x16x16bf16_1k v[0:3], a[0:1], a[2:3], -2.0 cbsz:3 abid:2 blgp:7
 // GFX90A: v_mfma_f32_16x16x16bf16_1k v[0:3], a[0:1], a[2:3], -2.0 cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe7,0xd3,0x00,0x05,0xd6,0xfb]
 
-v_mfma_f64_16x16x4f64 v[0:7], v[0:1], v[2:3], v[2:9]
-// GFX90A: v_mfma_f64_16x16x4f64 v[0:7], v[0:1], v[2:3], v[2:9] ; encoding: [0x00,0x00,0xee,0xd3,0x00,0x05,0x0a,0x04]
+v_mfma_f64_16x16x4f64 v[0:7], v[0:1], v[2:3], v[10:17]
+// GFX90A: v_mfma_f64_16x16x4f64 v[0:7], v[0:1], v[2:3], v[10:17] ; encoding: [0x00,0x00,0xee,0xd3,0x00,0x05,0x2a,0x04]
 
-v_mfma_f64_16x16x4f64 v[0:7], v[0:1], v[2:3], v[2:9] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f64_16x16x4f64 v[0:7], v[0:1], v[2:3], v[2:9] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xee,0xd3,0x00,0x05,0x0a,0xe4]
+v_mfma_f64_16x16x4f64 v[0:7], v[0:1], v[2:3], v[10:17] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f64_16x16x4f64 v[0:7], v[0:1], v[2:3], v[10:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xee,0xd3,0x00,0x05,0x2a,0xe4]
 
 v_mfma_f64_16x16x4f64 v[0:7], v[0:1], v[2:3], -2.0
 // GFX90A: v_mfma_f64_16x16x4f64 v[0:7], v[0:1], v[2:3], -2.0 ; encoding: [0x00,0x00,0xee,0xd3,0x00,0x05,0xd6,0x03]
@@ -2451,11 +2451,11 @@ v_mfma_f64_4x4x4f64 v[0:1], v[0:1], v[2:3], -2.0
 v_mfma_f64_4x4x4f64 v[0:1], v[0:1], v[2:3], -2.0 cbsz:3 abid:2 blgp:7
 // GFX90A: v_mfma_f64_4x4x4f64 v[0:1], v[0:1], v[2:3], -2.0 cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xef,0xd3,0x00,0x05,0xd6,0xe3]
 
-v_mfma_f64_16x16x4f64 v[0:7], a[0:1], v[2:3], v[2:9]
-// GFX90A: v_mfma_f64_16x16x4f64 v[0:7], a[0:1], v[2:3], v[2:9] ; encoding: [0x00,0x00,0xee,0xd3,0x00,0x05,0x0a,0x0c]
+v_mfma_f64_16x16x4f64 v[0:7], a[0:1], v[2:3], v[10:17]
+// GFX90A: v_mfma_f64_16x16x4f64 v[0:7], a[0:1], v[2:3], v[10:17] ; encoding: [0x00,0x00,0xee,0xd3,0x00,0x05,0x2a,0x0c]
 
-v_mfma_f64_16x16x4f64 v[0:7], v[0:1], a[2:3], v[2:9] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f64_16x16x4f64 v[0:7], v[0:1], a[2:3], v[2:9] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xee,0xd3,0x00,0x05,0x0a,0xf4]
+v_mfma_f64_16x16x4f64 v[0:7], v[0:1], a[2:3], v[10:17] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f64_16x16x4f64 v[0:7], v[0:1], a[2:3], v[10:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xee,0xd3,0x00,0x05,0x2a,0xf4]
 
 v_mfma_f64_16x16x4f64 v[0:7], a[0:1], a[2:3], -2.0
 // GFX90A: v_mfma_f64_16x16x4f64 v[0:7], a[0:1], a[2:3], -2.0 ; encoding: [0x00,0x00,0xee,0xd3,0x00,0x05,0xd6,0x1b]
@@ -2469,11 +2469,11 @@ v_mfma_f64_4x4x4f64 v[0:1], v[0:1], a[2:3], v[2:3] cbsz:3 abid:2 blgp:7
 v_mfma_f64_4x4x4f64 v[0:1], a[0:1], a[2:3], -2.0
 // GFX90A: v_mfma_f64_4x4x4f64 v[0:1], a[0:1], a[2:3], -2.0 ; encoding: [0x00,0x00,0xef,0xd3,0x00,0x05,0xd6,0x1b]
 
-v_mfma_f64_16x16x4f64 a[0:7], v[0:1], v[2:3], a[2:9]
-// GFX90A: v_mfma_f64_16x16x4f64 a[0:7], v[0:1], v[2:3], a[2:9] ; encoding: [0x00,0x80,0xee,0xd3,0x00,0x05,0x0a,0x04]
+v_mfma_f64_16x16x4f64 a[0:7], v[0:1], v[2:3], a[10:17]
+// GFX90A: v_mfma_f64_16x16x4f64 a[0:7], v[0:1], v[2:3], a[10:17] ; encoding: [0x00,0x80,0xee,0xd3,0x00,0x05,0x2a,0x04]
 
-v_mfma_f64_16x16x4f64 a[0:7], v[0:1], v[2:3], a[2:9] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f64_16x16x4f64 a[0:7], v[0:1], v[2:3], a[2:9] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xee,0xd3,0x00,0x05,0x0a,0xe4]
+v_mfma_f64_16x16x4f64 a[0:7], v[0:1], v[2:3], a[10:17] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f64_16x16x4f64 a[0:7], v[0:1], v[2:3], a[10:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xee,0xd3,0x00,0x05,0x2a,0xe4]
 
 v_mfma_f64_16x16x4f64 a[0:7], v[0:1], v[2:3], -2.0
 // GFX90A: v_mfma_f64_16x16x4f64 a[0:7], v[0:1], v[2:3], -2.0 ; encoding: [0x00,0x80,0xee,0xd3,0x00,0x05,0xd6,0x03]
@@ -2496,11 +2496,11 @@ v_mfma_f64_4x4x4f64 a[0:1], v[0:1], v[2:3], 0
 v_mfma_f64_4x4x4f64 a[0:1], v[0:1], v[2:3], -2.0 cbsz:3 abid:2 blgp:7
 // GFX90A: v_mfma_f64_4x4x4f64 a[0:1], v[0:1], v[2:3], -2.0 cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xef,0xd3,0x00,0x05,0xd6,0xe3]
 
-v_mfma_f64_16x16x4f64 a[0:7], a[0:1], v[2:3], a[2:9]
-// GFX90A: v_mfma_f64_16x16x4f64 a[0:7], a[0:1], v[2:3], a[2:9] ; encoding: [0x00,0x80,0xee,0xd3,0x00,0x05,0x0a,0x0c]
+v_mfma_f64_16x16x4f64 a[0:7], a[0:1], v[2:3], a[10:17]
+// GFX90A: v_mfma_f64_16x16x4f64 a[0:7], a[0:1], v[2:3], a[10:17] ; encoding: [0x00,0x80,0xee,0xd3,0x00,0x05,0x2a,0x0c]
 
-v_mfma_f64_16x16x4f64 a[0:7], v[0:1], a[2:3], a[2:9] cbsz:3 abid:2 blgp:7
-// GFX90A: v_mfma_f64_16x16x4f64 a[0:7], v[0:1], a[2:3], a[2:9] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xee,0xd3,0x00,0x05,0x0a,0xf4]
+v_mfma_f64_16x16x4f64 a[0:7], v[0:1], a[2:3], a[10:17] cbsz:3 abid:2 blgp:7
+// GFX90A: v_mfma_f64_16x16x4f64 a[0:7], v[0:1], a[2:3], a[10:17] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x93,0xee,0xd3,0x00,0x05,0x2a,0xf4]
 
 v_mfma_f64_16x16x4f64 a[0:7], a[0:1], a[2:3], -2.0
 // GFX90A: v_mfma_f64_16x16x4f64 a[0:7], a[0:1], a[2:3], -2.0 ; encoding: [0x00,0x80,0xee,0xd3,0x00,0x05,0xd6,0x1b]

diff  --git a/llvm/test/MC/AMDGPU/mai.s b/llvm/test/MC/AMDGPU/mai.s
index 5629df067590b..221d3b3f029da 100644
--- a/llvm/test/MC/AMDGPU/mai.s
+++ b/llvm/test/MC/AMDGPU/mai.s
@@ -43,53 +43,59 @@ v_accvgpr_write a2, shared_base
 v_accvgpr_write a2, pops_exiting_wave_id
 // NOGFX908: error: source operand must be either a VGPR or an inline constant
 
-v_mfma_f32_32x32x1f32 a[0:31], v0, v1, a[1:32]
-// GFX908: v_mfma_f32_32x32x1f32 a[0:31], v0, v1, a[1:32] ; encoding: [0x00,0x00,0xc0,0xd3,0x00,0x03,0x06,0x04]
+v_mfma_f32_32x32x1f32 a[0:31], v0, v1, a[0:31]
+// GFX908: v_mfma_f32_32x32x1f32 a[0:31], v0, v1, a[0:31] ; encoding: [0x00,0x00,0xc0,0xd3,0x00,0x03,0x02,0x04]
 
-v_mfma_f32_32x32x1f32 a[0:31], v0, v1, a[1:32] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_32x32x1f32 a[0:31], v0, v1, a[1:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc0,0xd3,0x00,0x03,0x06,0xe4]
+v_mfma_f32_32x32x1f32 a[0:31], v0, v1, a[33:64]
+// GFX908: v_mfma_f32_32x32x1f32 a[0:31], v0, v1, a[33:64] ; encoding: [0x00,0x00,0xc0,0xd3,0x00,0x03,0x86,0x04]
 
-v_mfma_f32_32x32x1f32 a[0:31], v0, a1, a[1:32]
-// GFX908: v_mfma_f32_32x32x1f32 a[0:31], v0, a1, a[1:32] ; encoding: [0x00,0x00,0xc0,0xd3,0x00,0x03,0x06,0x14]
+v_mfma_f32_32x32x1f32 a[0:31], v0, v1, a[33:64] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_32x32x1f32 a[0:31], v0, v1, a[33:64] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc0,0xd3,0x00,0x03,0x86,0xe4]
 
-v_mfma_f32_32x32x1f32 a[0:31], v0, a1, a[1:32] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_32x32x1f32 a[0:31], v0, a1, a[1:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc0,0xd3,0x00,0x03,0x06,0xf4]
+v_mfma_f32_32x32x1f32 a[0:31], v0, a1, a[33:64]
+// GFX908: v_mfma_f32_32x32x1f32 a[0:31], v0, a1, a[33:64] ; encoding: [0x00,0x00,0xc0,0xd3,0x00,0x03,0x86,0x14]
 
-v_mfma_f32_32x32x1f32 a[0:31], a0, v1, a[1:32]
-// GFX908: v_mfma_f32_32x32x1f32 a[0:31], a0, v1, a[1:32] ; encoding: [0x00,0x00,0xc0,0xd3,0x00,0x03,0x06,0x0c]
+v_mfma_f32_32x32x1f32 a[0:31], v0, a1, a[33:64] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_32x32x1f32 a[0:31], v0, a1, a[33:64] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc0,0xd3,0x00,0x03,0x86,0xf4]
 
-v_mfma_f32_32x32x1f32 a[0:31], a0, v1, a[1:32] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_32x32x1f32 a[0:31], a0, v1, a[1:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc0,0xd3,0x00,0x03,0x06,0xec]
+v_mfma_f32_32x32x1f32 a[0:31], a0, v1, a[33:64]
+// GFX908: v_mfma_f32_32x32x1f32 a[0:31], a0, v1, a[33:64] ; encoding: [0x00,0x00,0xc0,0xd3,0x00,0x03,0x86,0x0c]
 
-v_mfma_f32_32x32x1f32 a[0:31], a0, a1, a[1:32]
-// GFX908: v_mfma_f32_32x32x1f32 a[0:31], a0, a1, a[1:32] ; encoding: [0x00,0x00,0xc0,0xd3,0x00,0x03,0x06,0x1c]
+v_mfma_f32_32x32x1f32 a[0:31], a0, v1, a[33:64] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_32x32x1f32 a[0:31], a0, v1, a[33:64] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc0,0xd3,0x00,0x03,0x86,0xec]
 
-v_mfma_f32_32x32x1f32 a[0:31], a0, a1, a[1:32] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_32x32x1f32 a[0:31], a0, a1, a[1:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc0,0xd3,0x00,0x03,0x06,0xfc]
+v_mfma_f32_32x32x1f32 a[0:31], a0, a1, a[33:64]
+// GFX908: v_mfma_f32_32x32x1f32 a[0:31], a0, a1, a[33:64] ; encoding: [0x00,0x00,0xc0,0xd3,0x00,0x03,0x86,0x1c]
 
-v_mfma_f32_16x16x1f32 a[0:15], v0, v1, a[1:16]
-// GFX908: v_mfma_f32_16x16x1f32 a[0:15], v0, v1, a[1:16] ; encoding: [0x00,0x00,0xc1,0xd3,0x00,0x03,0x06,0x04]
+v_mfma_f32_32x32x1f32 a[0:31], a0, a1, a[33:64] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_32x32x1f32 a[0:31], a0, a1, a[33:64] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc0,0xd3,0x00,0x03,0x86,0xfc]
 
-v_mfma_f32_16x16x1f32 a[0:15], v0, v1, a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_16x16x1f32 a[0:15], v0, v1, a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc1,0xd3,0x00,0x03,0x06,0xe4]
+v_mfma_f32_16x16x1f32 a[0:15], v0, v1, a[17:32]
+// GFX908: v_mfma_f32_16x16x1f32 a[0:15], v0, v1, a[17:32] ; encoding: [0x00,0x00,0xc1,0xd3,0x00,0x03,0x46,0x04]
 
-v_mfma_f32_16x16x1f32 a[0:15], v0, a1, a[1:16]
-// GFX908: v_mfma_f32_16x16x1f32 a[0:15], v0, a1, a[1:16] ; encoding: [0x00,0x00,0xc1,0xd3,0x00,0x03,0x06,0x14]
+v_mfma_f32_16x16x1f32 a[0:15], v0, v1, a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_16x16x1f32 a[0:15], v0, v1, a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc1,0xd3,0x00,0x03,0x46,0xe4]
 
-v_mfma_f32_16x16x1f32 a[0:15], v0, a1, a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_16x16x1f32 a[0:15], v0, a1, a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc1,0xd3,0x00,0x03,0x06,0xf4]
+v_mfma_f32_16x16x1f32 a[0:15], v0, a1, a[17:32]
+// GFX908: v_mfma_f32_16x16x1f32 a[0:15], v0, a1, a[17:32] ; encoding: [0x00,0x00,0xc1,0xd3,0x00,0x03,0x46,0x14]
 
-v_mfma_f32_16x16x1f32 a[0:15], a0, v1, a[1:16]
-// GFX908: v_mfma_f32_16x16x1f32 a[0:15], a0, v1, a[1:16] ; encoding: [0x00,0x00,0xc1,0xd3,0x00,0x03,0x06,0x0c]
+v_mfma_f32_16x16x1f32 a[0:15], v0, a1, a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_16x16x1f32 a[0:15], v0, a1, a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc1,0xd3,0x00,0x03,0x46,0xf4]
 
-v_mfma_f32_16x16x1f32 a[0:15], a0, v1, a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_16x16x1f32 a[0:15], a0, v1, a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc1,0xd3,0x00,0x03,0x06,0xec]
+v_mfma_f32_16x16x1f32 a[0:15], a0, v1, a[17:32]
+// GFX908: v_mfma_f32_16x16x1f32 a[0:15], a0, v1, a[17:32] ; encoding: [0x00,0x00,0xc1,0xd3,0x00,0x03,0x46,0x0c]
 
-v_mfma_f32_16x16x1f32 a[0:15], a0, a1, a[1:16]
-// GFX908: v_mfma_f32_16x16x1f32 a[0:15], a0, a1, a[1:16] ; encoding: [0x00,0x00,0xc1,0xd3,0x00,0x03,0x06,0x1c]
+v_mfma_f32_16x16x1f32 a[0:15], a0, v1, a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_16x16x1f32 a[0:15], a0, v1, a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc1,0xd3,0x00,0x03,0x46,0xec]
 
-v_mfma_f32_16x16x1f32 a[0:15], a0, a1, a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_16x16x1f32 a[0:15], a0, a1, a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc1,0xd3,0x00,0x03,0x06,0xfc]
+v_mfma_f32_16x16x1f32 a[0:15], a0, a1, a[17:32]
+// GFX908: v_mfma_f32_16x16x1f32 a[0:15], a0, a1, a[17:32] ; encoding: [0x00,0x00,0xc1,0xd3,0x00,0x03,0x46,0x1c]
+
+v_mfma_f32_16x16x1f32 a[0:15], a0, a1, a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_16x16x1f32 a[0:15], a0, a1, a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc1,0xd3,0x00,0x03,0x46,0xfc]
+
+v_mfma_f32_4x4x1f32 a[0:3], v0, v1, a[1:4]
+// GFX908: v_mfma_f32_4x4x1f32 a[0:3], v0, v1, a[1:4] ; encoding: [0x00,0x00,0xc2,0xd3,0x00,0x03,0x06,0x04]
 
 v_mfma_f32_4x4x1f32 a[0:3], v0, v1, a[1:4]
 // GFX908: v_mfma_f32_4x4x1f32 a[0:3], v0, v1, a[1:4] ; encoding: [0x00,0x00,0xc2,0xd3,0x00,0x03,0x06,0x04]
@@ -115,29 +121,29 @@ v_mfma_f32_4x4x1f32 a[0:3], a0, a1, a[1:4]
 v_mfma_f32_4x4x1f32 a[0:3], a0, a1, a[1:4] cbsz:3 abid:2 blgp:7
 // GFX908: v_mfma_f32_4x4x1f32 a[0:3], a0, a1, a[1:4] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc2,0xd3,0x00,0x03,0x06,0xfc]
 
-v_mfma_f32_32x32x2f32 a[0:15], v0, v1, a[1:16]
-// GFX908: v_mfma_f32_32x32x2f32 a[0:15], v0, v1, a[1:16] ; encoding: [0x00,0x00,0xc4,0xd3,0x00,0x03,0x06,0x04]
+v_mfma_f32_32x32x2f32 a[0:15], v0, v1, a[17:32]
+// GFX908: v_mfma_f32_32x32x2f32 a[0:15], v0, v1, a[17:32] ; encoding: [0x00,0x00,0xc4,0xd3,0x00,0x03,0x46,0x04]
 
-v_mfma_f32_32x32x2f32 a[0:15], v0, v1, a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_32x32x2f32 a[0:15], v0, v1, a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc4,0xd3,0x00,0x03,0x06,0xe4]
+v_mfma_f32_32x32x2f32 a[0:15], v0, v1, a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_32x32x2f32 a[0:15], v0, v1, a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc4,0xd3,0x00,0x03,0x46,0xe4]
 
-v_mfma_f32_32x32x2f32 a[0:15], v0, a1, a[1:16]
-// GFX908: v_mfma_f32_32x32x2f32 a[0:15], v0, a1, a[1:16] ; encoding: [0x00,0x00,0xc4,0xd3,0x00,0x03,0x06,0x14]
+v_mfma_f32_32x32x2f32 a[0:15], v0, a1, a[17:32]
+// GFX908: v_mfma_f32_32x32x2f32 a[0:15], v0, a1, a[17:32] ; encoding: [0x00,0x00,0xc4,0xd3,0x00,0x03,0x46,0x14]
 
-v_mfma_f32_32x32x2f32 a[0:15], v0, a1, a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_32x32x2f32 a[0:15], v0, a1, a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc4,0xd3,0x00,0x03,0x06,0xf4]
+v_mfma_f32_32x32x2f32 a[0:15], v0, a1, a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_32x32x2f32 a[0:15], v0, a1, a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc4,0xd3,0x00,0x03,0x46,0xf4]
 
-v_mfma_f32_32x32x2f32 a[0:15], a0, v1, a[1:16]
-// GFX908: v_mfma_f32_32x32x2f32 a[0:15], a0, v1, a[1:16] ; encoding: [0x00,0x00,0xc4,0xd3,0x00,0x03,0x06,0x0c]
+v_mfma_f32_32x32x2f32 a[0:15], a0, v1, a[17:32]
+// GFX908: v_mfma_f32_32x32x2f32 a[0:15], a0, v1, a[17:32] ; encoding: [0x00,0x00,0xc4,0xd3,0x00,0x03,0x46,0x0c]
 
-v_mfma_f32_32x32x2f32 a[0:15], a0, v1, a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_32x32x2f32 a[0:15], a0, v1, a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc4,0xd3,0x00,0x03,0x06,0xec]
+v_mfma_f32_32x32x2f32 a[0:15], a0, v1, a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_32x32x2f32 a[0:15], a0, v1, a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc4,0xd3,0x00,0x03,0x46,0xec]
 
-v_mfma_f32_32x32x2f32 a[0:15], a0, a1, a[1:16]
-// GFX908: v_mfma_f32_32x32x2f32 a[0:15], a0, a1, a[1:16] ; encoding: [0x00,0x00,0xc4,0xd3,0x00,0x03,0x06,0x1c]
+v_mfma_f32_32x32x2f32 a[0:15], a0, a1, a[17:32]
+// GFX908: v_mfma_f32_32x32x2f32 a[0:15], a0, a1, a[17:32] ; encoding: [0x00,0x00,0xc4,0xd3,0x00,0x03,0x46,0x1c]
 
-v_mfma_f32_32x32x2f32 a[0:15], a0, a1, a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_32x32x2f32 a[0:15], a0, a1, a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc4,0xd3,0x00,0x03,0x06,0xfc]
+v_mfma_f32_32x32x2f32 a[0:15], a0, a1, a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_32x32x2f32 a[0:15], a0, a1, a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc4,0xd3,0x00,0x03,0x46,0xfc]
 
 v_mfma_f32_16x16x4f32 a[0:3], v0, v1, a[1:4]
 // GFX908: v_mfma_f32_16x16x4f32 a[0:3], v0, v1, a[1:4] ; encoding: [0x00,0x00,0xc5,0xd3,0x00,0x03,0x06,0x04]
@@ -163,53 +169,53 @@ v_mfma_f32_16x16x4f32 a[0:3], a0, a1, a[1:4]
 v_mfma_f32_16x16x4f32 a[0:3], a0, a1, a[1:4] cbsz:3 abid:2 blgp:7
 // GFX908: v_mfma_f32_16x16x4f32 a[0:3], a0, a1, a[1:4] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc5,0xd3,0x00,0x03,0x06,0xfc]
 
-v_mfma_f32_32x32x4f16 a[0:31], v[0:1], v[1:2], a[1:32]
-// GFX908: v_mfma_f32_32x32x4f16 a[0:31], v[0:1], v[1:2], a[1:32] ; encoding: [0x00,0x00,0xc8,0xd3,0x00,0x03,0x06,0x04]
+v_mfma_f32_32x32x4f16 a[0:31], v[0:1], v[1:2], a[33:64]
+// GFX908: v_mfma_f32_32x32x4f16 a[0:31], v[0:1], v[1:2], a[33:64] ; encoding: [0x00,0x00,0xc8,0xd3,0x00,0x03,0x86,0x04]
 
-v_mfma_f32_32x32x4f16 a[0:31], v[0:1], v[1:2], a[1:32] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_32x32x4f16 a[0:31], v[0:1], v[1:2], a[1:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc8,0xd3,0x00,0x03,0x06,0xe4]
+v_mfma_f32_32x32x4f16 a[0:31], v[0:1], v[1:2], a[33:64] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_32x32x4f16 a[0:31], v[0:1], v[1:2], a[33:64] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc8,0xd3,0x00,0x03,0x86,0xe4]
 
-v_mfma_f32_32x32x4f16 a[0:31], v[0:1], a[1:2], a[1:32]
-// GFX908: v_mfma_f32_32x32x4f16 a[0:31], v[0:1], a[1:2], a[1:32] ; encoding: [0x00,0x00,0xc8,0xd3,0x00,0x03,0x06,0x14]
+v_mfma_f32_32x32x4f16 a[0:31], v[0:1], a[1:2], a[33:64]
+// GFX908: v_mfma_f32_32x32x4f16 a[0:31], v[0:1], a[1:2], a[33:64] ; encoding: [0x00,0x00,0xc8,0xd3,0x00,0x03,0x86,0x14]
 
-v_mfma_f32_32x32x4f16 a[0:31], v[0:1], a[1:2], a[1:32] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_32x32x4f16 a[0:31], v[0:1], a[1:2], a[1:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc8,0xd3,0x00,0x03,0x06,0xf4]
+v_mfma_f32_32x32x4f16 a[0:31], v[0:1], a[1:2], a[33:64] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_32x32x4f16 a[0:31], v[0:1], a[1:2], a[33:64] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc8,0xd3,0x00,0x03,0x86,0xf4]
 
-v_mfma_f32_32x32x4f16 a[0:31], a[0:1], v[1:2], a[1:32]
-// GFX908: v_mfma_f32_32x32x4f16 a[0:31], a[0:1], v[1:2], a[1:32] ; encoding: [0x00,0x00,0xc8,0xd3,0x00,0x03,0x06,0x0c]
+v_mfma_f32_32x32x4f16 a[0:31], a[0:1], v[1:2], a[33:64]
+// GFX908: v_mfma_f32_32x32x4f16 a[0:31], a[0:1], v[1:2], a[33:64] ; encoding: [0x00,0x00,0xc8,0xd3,0x00,0x03,0x86,0x0c]
 
-v_mfma_f32_32x32x4f16 a[0:31], a[0:1], v[1:2], a[1:32] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_32x32x4f16 a[0:31], a[0:1], v[1:2], a[1:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc8,0xd3,0x00,0x03,0x06,0xec]
+v_mfma_f32_32x32x4f16 a[0:31], a[0:1], v[1:2], a[33:64] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_32x32x4f16 a[0:31], a[0:1], v[1:2], a[33:64] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc8,0xd3,0x00,0x03,0x86,0xec]
 
-v_mfma_f32_32x32x4f16 a[0:31], a[0:1], a[1:2], a[1:32]
-// GFX908: v_mfma_f32_32x32x4f16 a[0:31], a[0:1], a[1:2], a[1:32] ; encoding: [0x00,0x00,0xc8,0xd3,0x00,0x03,0x06,0x1c]
+v_mfma_f32_32x32x4f16 a[0:31], a[0:1], a[1:2], a[33:64]
+// GFX908: v_mfma_f32_32x32x4f16 a[0:31], a[0:1], a[1:2], a[33:64] ; encoding: [0x00,0x00,0xc8,0xd3,0x00,0x03,0x86,0x1c]
 
-v_mfma_f32_32x32x4f16 a[0:31], a[0:1], a[1:2], a[1:32] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_32x32x4f16 a[0:31], a[0:1], a[1:2], a[1:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc8,0xd3,0x00,0x03,0x06,0xfc]
+v_mfma_f32_32x32x4f16 a[0:31], a[0:1], a[1:2], a[33:64] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_32x32x4f16 a[0:31], a[0:1], a[1:2], a[33:64] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc8,0xd3,0x00,0x03,0x86,0xfc]
 
-v_mfma_f32_16x16x4f16 a[0:15], v[0:1], v[1:2], a[1:16]
-// GFX908: v_mfma_f32_16x16x4f16 a[0:15], v[0:1], v[1:2], a[1:16] ; encoding: [0x00,0x00,0xc9,0xd3,0x00,0x03,0x06,0x04]
+v_mfma_f32_16x16x4f16 a[0:15], v[0:1], v[1:2], a[17:32]
+// GFX908: v_mfma_f32_16x16x4f16 a[0:15], v[0:1], v[1:2], a[17:32] ; encoding: [0x00,0x00,0xc9,0xd3,0x00,0x03,0x46,0x04]
 
-v_mfma_f32_16x16x4f16 a[0:15], v[0:1], v[1:2], a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_16x16x4f16 a[0:15], v[0:1], v[1:2], a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc9,0xd3,0x00,0x03,0x06,0xe4]
+v_mfma_f32_16x16x4f16 a[0:15], v[0:1], v[1:2], a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_16x16x4f16 a[0:15], v[0:1], v[1:2], a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc9,0xd3,0x00,0x03,0x46,0xe4]
 
-v_mfma_f32_16x16x4f16 a[0:15], v[0:1], a[1:2], a[1:16]
-// GFX908: v_mfma_f32_16x16x4f16 a[0:15], v[0:1], a[1:2], a[1:16] ; encoding: [0x00,0x00,0xc9,0xd3,0x00,0x03,0x06,0x14]
+v_mfma_f32_16x16x4f16 a[0:15], v[0:1], a[1:2], a[17:32]
+// GFX908: v_mfma_f32_16x16x4f16 a[0:15], v[0:1], a[1:2], a[17:32] ; encoding: [0x00,0x00,0xc9,0xd3,0x00,0x03,0x46,0x14]
 
-v_mfma_f32_16x16x4f16 a[0:15], v[0:1], a[1:2], a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_16x16x4f16 a[0:15], v[0:1], a[1:2], a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc9,0xd3,0x00,0x03,0x06,0xf4]
+v_mfma_f32_16x16x4f16 a[0:15], v[0:1], a[1:2], a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_16x16x4f16 a[0:15], v[0:1], a[1:2], a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc9,0xd3,0x00,0x03,0x46,0xf4]
 
-v_mfma_f32_16x16x4f16 a[0:15], a[0:1], v[1:2], a[1:16]
-// GFX908: v_mfma_f32_16x16x4f16 a[0:15], a[0:1], v[1:2], a[1:16] ; encoding: [0x00,0x00,0xc9,0xd3,0x00,0x03,0x06,0x0c]
+v_mfma_f32_16x16x4f16 a[0:15], a[0:1], v[1:2], a[17:32]
+// GFX908: v_mfma_f32_16x16x4f16 a[0:15], a[0:1], v[1:2], a[17:32] ; encoding: [0x00,0x00,0xc9,0xd3,0x00,0x03,0x46,0x0c]
 
-v_mfma_f32_16x16x4f16 a[0:15], a[0:1], v[1:2], a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_16x16x4f16 a[0:15], a[0:1], v[1:2], a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc9,0xd3,0x00,0x03,0x06,0xec]
+v_mfma_f32_16x16x4f16 a[0:15], a[0:1], v[1:2], a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_16x16x4f16 a[0:15], a[0:1], v[1:2], a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc9,0xd3,0x00,0x03,0x46,0xec]
 
-v_mfma_f32_16x16x4f16 a[0:15], a[0:1], a[1:2], a[1:16]
-// GFX908: v_mfma_f32_16x16x4f16 a[0:15], a[0:1], a[1:2], a[1:16] ; encoding: [0x00,0x00,0xc9,0xd3,0x00,0x03,0x06,0x1c]
+v_mfma_f32_16x16x4f16 a[0:15], a[0:1], a[1:2], a[17:32]
+// GFX908: v_mfma_f32_16x16x4f16 a[0:15], a[0:1], a[1:2], a[17:32] ; encoding: [0x00,0x00,0xc9,0xd3,0x00,0x03,0x46,0x1c]
 
-v_mfma_f32_16x16x4f16 a[0:15], a[0:1], a[1:2], a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_16x16x4f16 a[0:15], a[0:1], a[1:2], a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc9,0xd3,0x00,0x03,0x06,0xfc]
+v_mfma_f32_16x16x4f16 a[0:15], a[0:1], a[1:2], a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_16x16x4f16 a[0:15], a[0:1], a[1:2], a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xc9,0xd3,0x00,0x03,0x46,0xfc]
 
 v_mfma_f32_4x4x4f16 a[0:3], v[0:1], v[1:2], a[1:4]
 // GFX908: v_mfma_f32_4x4x4f16 a[0:3], v[0:1], v[1:2], a[1:4] ; encoding: [0x00,0x00,0xca,0xd3,0x00,0x03,0x06,0x04]
@@ -235,29 +241,29 @@ v_mfma_f32_4x4x4f16 a[0:3], a[0:1], a[1:2], a[1:4]
 v_mfma_f32_4x4x4f16 a[0:3], a[0:1], a[1:2], a[1:4] cbsz:3 abid:2 blgp:7
 // GFX908: v_mfma_f32_4x4x4f16 a[0:3], a[0:1], a[1:2], a[1:4] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xca,0xd3,0x00,0x03,0x06,0xfc]
 
-v_mfma_f32_32x32x8f16 a[0:15], v[0:1], v[1:2], a[1:16]
-// GFX908: v_mfma_f32_32x32x8f16 a[0:15], v[0:1], v[1:2], a[1:16] ; encoding: [0x00,0x00,0xcc,0xd3,0x00,0x03,0x06,0x04]
+v_mfma_f32_32x32x8f16 a[0:15], v[0:1], v[1:2], a[17:32]
+// GFX908: v_mfma_f32_32x32x8f16 a[0:15], v[0:1], v[1:2], a[17:32] ; encoding: [0x00,0x00,0xcc,0xd3,0x00,0x03,0x46,0x04]
 
-v_mfma_f32_32x32x8f16 a[0:15], v[0:1], v[1:2], a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_32x32x8f16 a[0:15], v[0:1], v[1:2], a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xcc,0xd3,0x00,0x03,0x06,0xe4]
+v_mfma_f32_32x32x8f16 a[0:15], v[0:1], v[1:2], a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_32x32x8f16 a[0:15], v[0:1], v[1:2], a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xcc,0xd3,0x00,0x03,0x46,0xe4]
 
-v_mfma_f32_32x32x8f16 a[0:15], v[0:1], a[1:2], a[1:16]
-// GFX908: v_mfma_f32_32x32x8f16 a[0:15], v[0:1], a[1:2], a[1:16] ; encoding: [0x00,0x00,0xcc,0xd3,0x00,0x03,0x06,0x14]
+v_mfma_f32_32x32x8f16 a[0:15], v[0:1], a[1:2], a[17:32]
+// GFX908: v_mfma_f32_32x32x8f16 a[0:15], v[0:1], a[1:2], a[17:32] ; encoding: [0x00,0x00,0xcc,0xd3,0x00,0x03,0x46,0x14]
 
-v_mfma_f32_32x32x8f16 a[0:15], v[0:1], a[1:2], a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_32x32x8f16 a[0:15], v[0:1], a[1:2], a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xcc,0xd3,0x00,0x03,0x06,0xf4]
+v_mfma_f32_32x32x8f16 a[0:15], v[0:1], a[1:2], a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_32x32x8f16 a[0:15], v[0:1], a[1:2], a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xcc,0xd3,0x00,0x03,0x46,0xf4]
 
-v_mfma_f32_32x32x8f16 a[0:15], a[0:1], v[1:2], a[1:16]
-// GFX908: v_mfma_f32_32x32x8f16 a[0:15], a[0:1], v[1:2], a[1:16] ; encoding: [0x00,0x00,0xcc,0xd3,0x00,0x03,0x06,0x0c]
+v_mfma_f32_32x32x8f16 a[0:15], a[0:1], v[1:2], a[17:32]
+// GFX908: v_mfma_f32_32x32x8f16 a[0:15], a[0:1], v[1:2], a[17:32] ; encoding: [0x00,0x00,0xcc,0xd3,0x00,0x03,0x46,0x0c]
 
-v_mfma_f32_32x32x8f16 a[0:15], a[0:1], v[1:2], a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_32x32x8f16 a[0:15], a[0:1], v[1:2], a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xcc,0xd3,0x00,0x03,0x06,0xec]
+v_mfma_f32_32x32x8f16 a[0:15], a[0:1], v[1:2], a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_32x32x8f16 a[0:15], a[0:1], v[1:2], a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xcc,0xd3,0x00,0x03,0x46,0xec]
 
-v_mfma_f32_32x32x8f16 a[0:15], a[0:1], a[1:2], a[1:16]
-// GFX908: v_mfma_f32_32x32x8f16 a[0:15], a[0:1], a[1:2], a[1:16] ; encoding: [0x00,0x00,0xcc,0xd3,0x00,0x03,0x06,0x1c]
+v_mfma_f32_32x32x8f16 a[0:15], a[0:1], a[1:2], a[17:32]
+// GFX908: v_mfma_f32_32x32x8f16 a[0:15], a[0:1], a[1:2], a[17:32] ; encoding: [0x00,0x00,0xcc,0xd3,0x00,0x03,0x46,0x1c]
 
-v_mfma_f32_32x32x8f16 a[0:15], a[0:1], a[1:2], a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_32x32x8f16 a[0:15], a[0:1], a[1:2], a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xcc,0xd3,0x00,0x03,0x06,0xfc]
+v_mfma_f32_32x32x8f16 a[0:15], a[0:1], a[1:2], a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_32x32x8f16 a[0:15], a[0:1], a[1:2], a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xcc,0xd3,0x00,0x03,0x46,0xfc]
 
 v_mfma_f32_16x16x16f16 a[0:3], v[0:1], v[1:2], a[1:4]
 // GFX908: v_mfma_f32_16x16x16f16 a[0:3], v[0:1], v[1:2], a[1:4] ; encoding: [0x00,0x00,0xcd,0xd3,0x00,0x03,0x06,0x04]
@@ -283,53 +289,53 @@ v_mfma_f32_16x16x16f16 a[0:3], a[0:1], a[1:2], a[1:4]
 v_mfma_f32_16x16x16f16 a[0:3], a[0:1], a[1:2], a[1:4] cbsz:3 abid:2 blgp:7
 // GFX908: v_mfma_f32_16x16x16f16 a[0:3], a[0:1], a[1:2], a[1:4] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xcd,0xd3,0x00,0x03,0x06,0xfc]
 
-v_mfma_i32_32x32x4i8 a[0:31], v0, v1, a[1:32]
-// GFX908: v_mfma_i32_32x32x4i8 a[0:31], v0, v1, a[1:32] ; encoding: [0x00,0x00,0xd0,0xd3,0x00,0x03,0x06,0x04]
+v_mfma_i32_32x32x4i8 a[0:31], v0, v1, a[33:64]
+// GFX908: v_mfma_i32_32x32x4i8 a[0:31], v0, v1, a[33:64] ; encoding: [0x00,0x00,0xd0,0xd3,0x00,0x03,0x86,0x04]
 
-v_mfma_i32_32x32x4i8 a[0:31], v0, v1, a[1:32] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_i32_32x32x4i8 a[0:31], v0, v1, a[1:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd0,0xd3,0x00,0x03,0x06,0xe4]
+v_mfma_i32_32x32x4i8 a[0:31], v0, v1, a[33:64] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_i32_32x32x4i8 a[0:31], v0, v1, a[33:64] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd0,0xd3,0x00,0x03,0x86,0xe4]
 
-v_mfma_i32_32x32x4i8 a[0:31], v0, a1, a[1:32]
-// GFX908: v_mfma_i32_32x32x4i8 a[0:31], v0, a1, a[1:32] ; encoding: [0x00,0x00,0xd0,0xd3,0x00,0x03,0x06,0x14]
+v_mfma_i32_32x32x4i8 a[0:31], v0, a1, a[33:64]
+// GFX908: v_mfma_i32_32x32x4i8 a[0:31], v0, a1, a[33:64] ; encoding: [0x00,0x00,0xd0,0xd3,0x00,0x03,0x86,0x14]
 
-v_mfma_i32_32x32x4i8 a[0:31], v0, a1, a[1:32] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_i32_32x32x4i8 a[0:31], v0, a1, a[1:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd0,0xd3,0x00,0x03,0x06,0xf4]
+v_mfma_i32_32x32x4i8 a[0:31], v0, a1, a[33:64] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_i32_32x32x4i8 a[0:31], v0, a1, a[33:64] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd0,0xd3,0x00,0x03,0x86,0xf4]
 
-v_mfma_i32_32x32x4i8 a[0:31], a0, v1, a[1:32]
-// GFX908: v_mfma_i32_32x32x4i8 a[0:31], a0, v1, a[1:32] ; encoding: [0x00,0x00,0xd0,0xd3,0x00,0x03,0x06,0x0c]
+v_mfma_i32_32x32x4i8 a[0:31], a0, v1, a[33:64]
+// GFX908: v_mfma_i32_32x32x4i8 a[0:31], a0, v1, a[33:64] ; encoding: [0x00,0x00,0xd0,0xd3,0x00,0x03,0x86,0x0c]
 
-v_mfma_i32_32x32x4i8 a[0:31], a0, v1, a[1:32] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_i32_32x32x4i8 a[0:31], a0, v1, a[1:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd0,0xd3,0x00,0x03,0x06,0xec]
+v_mfma_i32_32x32x4i8 a[0:31], a0, v1, a[33:64] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_i32_32x32x4i8 a[0:31], a0, v1, a[33:64] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd0,0xd3,0x00,0x03,0x86,0xec]
 
-v_mfma_i32_32x32x4i8 a[0:31], a0, a1, a[1:32]
-// GFX908: v_mfma_i32_32x32x4i8 a[0:31], a0, a1, a[1:32] ; encoding: [0x00,0x00,0xd0,0xd3,0x00,0x03,0x06,0x1c]
+v_mfma_i32_32x32x4i8 a[0:31], a0, a1, a[33:64]
+// GFX908: v_mfma_i32_32x32x4i8 a[0:31], a0, a1, a[33:64] ; encoding: [0x00,0x00,0xd0,0xd3,0x00,0x03,0x86,0x1c]
 
-v_mfma_i32_32x32x4i8 a[0:31], a0, a1, a[1:32] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_i32_32x32x4i8 a[0:31], a0, a1, a[1:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd0,0xd3,0x00,0x03,0x06,0xfc]
+v_mfma_i32_32x32x4i8 a[0:31], a0, a1, a[33:64] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_i32_32x32x4i8 a[0:31], a0, a1, a[33:64] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd0,0xd3,0x00,0x03,0x86,0xfc]
 
-v_mfma_i32_16x16x4i8 a[0:15], v0, v1, a[1:16]
-// GFX908: v_mfma_i32_16x16x4i8 a[0:15], v0, v1, a[1:16] ; encoding: [0x00,0x00,0xd1,0xd3,0x00,0x03,0x06,0x04]
+v_mfma_i32_16x16x4i8 a[0:15], v0, v1, a[17:32]
+// GFX908: v_mfma_i32_16x16x4i8 a[0:15], v0, v1, a[17:32] ; encoding: [0x00,0x00,0xd1,0xd3,0x00,0x03,0x46,0x04]
 
-v_mfma_i32_16x16x4i8 a[0:15], v0, v1, a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_i32_16x16x4i8 a[0:15], v0, v1, a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd1,0xd3,0x00,0x03,0x06,0xe4]
+v_mfma_i32_16x16x4i8 a[0:15], v0, v1, a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_i32_16x16x4i8 a[0:15], v0, v1, a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd1,0xd3,0x00,0x03,0x46,0xe4]
 
-v_mfma_i32_16x16x4i8 a[0:15], v0, a1, a[1:16]
-// GFX908: v_mfma_i32_16x16x4i8 a[0:15], v0, a1, a[1:16] ; encoding: [0x00,0x00,0xd1,0xd3,0x00,0x03,0x06,0x14]
+v_mfma_i32_16x16x4i8 a[0:15], v0, a1, a[17:32]
+// GFX908: v_mfma_i32_16x16x4i8 a[0:15], v0, a1, a[17:32] ; encoding: [0x00,0x00,0xd1,0xd3,0x00,0x03,0x46,0x14]
 
-v_mfma_i32_16x16x4i8 a[0:15], v0, a1, a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_i32_16x16x4i8 a[0:15], v0, a1, a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd1,0xd3,0x00,0x03,0x06,0xf4]
+v_mfma_i32_16x16x4i8 a[0:15], v0, a1, a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_i32_16x16x4i8 a[0:15], v0, a1, a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd1,0xd3,0x00,0x03,0x46,0xf4]
 
-v_mfma_i32_16x16x4i8 a[0:15], a0, v1, a[1:16]
-// GFX908: v_mfma_i32_16x16x4i8 a[0:15], a0, v1, a[1:16] ; encoding: [0x00,0x00,0xd1,0xd3,0x00,0x03,0x06,0x0c]
+v_mfma_i32_16x16x4i8 a[0:15], a0, v1, a[17:32]
+// GFX908: v_mfma_i32_16x16x4i8 a[0:15], a0, v1, a[17:32] ; encoding: [0x00,0x00,0xd1,0xd3,0x00,0x03,0x46,0x0c]
 
-v_mfma_i32_16x16x4i8 a[0:15], a0, v1, a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_i32_16x16x4i8 a[0:15], a0, v1, a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd1,0xd3,0x00,0x03,0x06,0xec]
+v_mfma_i32_16x16x4i8 a[0:15], a0, v1, a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_i32_16x16x4i8 a[0:15], a0, v1, a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd1,0xd3,0x00,0x03,0x46,0xec]
 
-v_mfma_i32_16x16x4i8 a[0:15], a0, a1, a[1:16]
-// GFX908: v_mfma_i32_16x16x4i8 a[0:15], a0, a1, a[1:16] ; encoding: [0x00,0x00,0xd1,0xd3,0x00,0x03,0x06,0x1c]
+v_mfma_i32_16x16x4i8 a[0:15], a0, a1, a[17:32]
+// GFX908: v_mfma_i32_16x16x4i8 a[0:15], a0, a1, a[17:32] ; encoding: [0x00,0x00,0xd1,0xd3,0x00,0x03,0x46,0x1c]
 
-v_mfma_i32_16x16x4i8 a[0:15], a0, a1, a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_i32_16x16x4i8 a[0:15], a0, a1, a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd1,0xd3,0x00,0x03,0x06,0xfc]
+v_mfma_i32_16x16x4i8 a[0:15], a0, a1, a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_i32_16x16x4i8 a[0:15], a0, a1, a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd1,0xd3,0x00,0x03,0x46,0xfc]
 
 v_mfma_i32_4x4x4i8 a[0:3], v0, v1, a[1:4]
 // GFX908: v_mfma_i32_4x4x4i8 a[0:3], v0, v1, a[1:4] ; encoding: [0x00,0x00,0xd2,0xd3,0x00,0x03,0x06,0x04]
@@ -355,29 +361,29 @@ v_mfma_i32_4x4x4i8 a[0:3], a0, a1, a[1:4]
 v_mfma_i32_4x4x4i8 a[0:3], a0, a1, a[1:4] cbsz:3 abid:2 blgp:7
 // GFX908: v_mfma_i32_4x4x4i8 a[0:3], a0, a1, a[1:4] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd2,0xd3,0x00,0x03,0x06,0xfc]
 
-v_mfma_i32_32x32x8i8 a[0:15], v0, v1, a[1:16]
-// GFX908: v_mfma_i32_32x32x8i8 a[0:15], v0, v1, a[1:16] ; encoding: [0x00,0x00,0xd4,0xd3,0x00,0x03,0x06,0x04]
+v_mfma_i32_32x32x8i8 a[0:15], v0, v1, a[17:32]
+// GFX908: v_mfma_i32_32x32x8i8 a[0:15], v0, v1, a[17:32] ; encoding: [0x00,0x00,0xd4,0xd3,0x00,0x03,0x46,0x04]
 
-v_mfma_i32_32x32x8i8 a[0:15], v0, v1, a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_i32_32x32x8i8 a[0:15], v0, v1, a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd4,0xd3,0x00,0x03,0x06,0xe4]
+v_mfma_i32_32x32x8i8 a[0:15], v0, v1, a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_i32_32x32x8i8 a[0:15], v0, v1, a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd4,0xd3,0x00,0x03,0x46,0xe4]
 
-v_mfma_i32_32x32x8i8 a[0:15], v0, a1, a[1:16]
-// GFX908: v_mfma_i32_32x32x8i8 a[0:15], v0, a1, a[1:16] ; encoding: [0x00,0x00,0xd4,0xd3,0x00,0x03,0x06,0x14]
+v_mfma_i32_32x32x8i8 a[0:15], v0, a1, a[17:32]
+// GFX908: v_mfma_i32_32x32x8i8 a[0:15], v0, a1, a[17:32] ; encoding: [0x00,0x00,0xd4,0xd3,0x00,0x03,0x46,0x14]
 
-v_mfma_i32_32x32x8i8 a[0:15], v0, a1, a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_i32_32x32x8i8 a[0:15], v0, a1, a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd4,0xd3,0x00,0x03,0x06,0xf4]
+v_mfma_i32_32x32x8i8 a[0:15], v0, a1, a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_i32_32x32x8i8 a[0:15], v0, a1, a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd4,0xd3,0x00,0x03,0x46,0xf4]
 
-v_mfma_i32_32x32x8i8 a[0:15], a0, v1, a[1:16]
-// GFX908: v_mfma_i32_32x32x8i8 a[0:15], a0, v1, a[1:16] ; encoding: [0x00,0x00,0xd4,0xd3,0x00,0x03,0x06,0x0c]
+v_mfma_i32_32x32x8i8 a[0:15], a0, v1, a[17:32]
+// GFX908: v_mfma_i32_32x32x8i8 a[0:15], a0, v1, a[17:32] ; encoding: [0x00,0x00,0xd4,0xd3,0x00,0x03,0x46,0x0c]
 
-v_mfma_i32_32x32x8i8 a[0:15], a0, v1, a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_i32_32x32x8i8 a[0:15], a0, v1, a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd4,0xd3,0x00,0x03,0x06,0xec]
+v_mfma_i32_32x32x8i8 a[0:15], a0, v1, a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_i32_32x32x8i8 a[0:15], a0, v1, a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd4,0xd3,0x00,0x03,0x46,0xec]
 
-v_mfma_i32_32x32x8i8 a[0:15], a0, a1, a[1:16]
-// GFX908: v_mfma_i32_32x32x8i8 a[0:15], a0, a1, a[1:16] ; encoding: [0x00,0x00,0xd4,0xd3,0x00,0x03,0x06,0x1c]
+v_mfma_i32_32x32x8i8 a[0:15], a0, a1, a[17:32]
+// GFX908: v_mfma_i32_32x32x8i8 a[0:15], a0, a1, a[17:32] ; encoding: [0x00,0x00,0xd4,0xd3,0x00,0x03,0x46,0x1c]
 
-v_mfma_i32_32x32x8i8 a[0:15], a0, a1, a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_i32_32x32x8i8 a[0:15], a0, a1, a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd4,0xd3,0x00,0x03,0x06,0xfc]
+v_mfma_i32_32x32x8i8 a[0:15], a0, a1, a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_i32_32x32x8i8 a[0:15], a0, a1, a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd4,0xd3,0x00,0x03,0x46,0xfc]
 
 v_mfma_i32_16x16x16i8 a[0:3], v0, v1, a[1:4]
 // GFX908: v_mfma_i32_16x16x16i8 a[0:3], v0, v1, a[1:4] ; encoding: [0x00,0x00,0xd5,0xd3,0x00,0x03,0x06,0x04]
@@ -403,53 +409,53 @@ v_mfma_i32_16x16x16i8 a[0:3], a0, a1, a[1:4]
 v_mfma_i32_16x16x16i8 a[0:3], a0, a1, a[1:4] cbsz:3 abid:2 blgp:7
 // GFX908: v_mfma_i32_16x16x16i8 a[0:3], a0, a1, a[1:4] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xd5,0xd3,0x00,0x03,0x06,0xfc]
 
-v_mfma_f32_32x32x2bf16 a[0:31], v0, v1, a[1:32]
-// GFX908: v_mfma_f32_32x32x2bf16 a[0:31], v0, v1, a[1:32] ; encoding: [0x00,0x00,0xe8,0xd3,0x00,0x03,0x06,0x04]
+v_mfma_f32_32x32x2bf16 a[0:31], v0, v1, a[33:64]
+// GFX908: v_mfma_f32_32x32x2bf16 a[0:31], v0, v1, a[33:64] ; encoding: [0x00,0x00,0xe8,0xd3,0x00,0x03,0x86,0x04]
 
-v_mfma_f32_32x32x2bf16 a[0:31], v0, v1, a[1:32] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_32x32x2bf16 a[0:31], v0, v1, a[1:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe8,0xd3,0x00,0x03,0x06,0xe4]
+v_mfma_f32_32x32x2bf16 a[0:31], v0, v1, a[33:64] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_32x32x2bf16 a[0:31], v0, v1, a[33:64] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe8,0xd3,0x00,0x03,0x86,0xe4]
 
-v_mfma_f32_32x32x2bf16 a[0:31], v0, a1, a[1:32]
-// GFX908: v_mfma_f32_32x32x2bf16 a[0:31], v0, a1, a[1:32] ; encoding: [0x00,0x00,0xe8,0xd3,0x00,0x03,0x06,0x14]
+v_mfma_f32_32x32x2bf16 a[0:31], v0, a1, a[33:64]
+// GFX908: v_mfma_f32_32x32x2bf16 a[0:31], v0, a1, a[33:64] ; encoding: [0x00,0x00,0xe8,0xd3,0x00,0x03,0x86,0x14]
 
-v_mfma_f32_32x32x2bf16 a[0:31], v0, a1, a[1:32] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_32x32x2bf16 a[0:31], v0, a1, a[1:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe8,0xd3,0x00,0x03,0x06,0xf4]
+v_mfma_f32_32x32x2bf16 a[0:31], v0, a1, a[33:64] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_32x32x2bf16 a[0:31], v0, a1, a[33:64] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe8,0xd3,0x00,0x03,0x86,0xf4]
 
-v_mfma_f32_32x32x2bf16 a[0:31], a0, v1, a[1:32]
-// GFX908: v_mfma_f32_32x32x2bf16 a[0:31], a0, v1, a[1:32] ; encoding: [0x00,0x00,0xe8,0xd3,0x00,0x03,0x06,0x0c]
+v_mfma_f32_32x32x2bf16 a[0:31], a0, v1, a[33:64]
+// GFX908: v_mfma_f32_32x32x2bf16 a[0:31], a0, v1, a[33:64] ; encoding: [0x00,0x00,0xe8,0xd3,0x00,0x03,0x86,0x0c]
 
-v_mfma_f32_32x32x2bf16 a[0:31], a0, v1, a[1:32] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_32x32x2bf16 a[0:31], a0, v1, a[1:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe8,0xd3,0x00,0x03,0x06,0xec]
+v_mfma_f32_32x32x2bf16 a[0:31], a0, v1, a[33:64] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_32x32x2bf16 a[0:31], a0, v1, a[33:64] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe8,0xd3,0x00,0x03,0x86,0xec]
 
-v_mfma_f32_32x32x2bf16 a[0:31], a0, a1, a[1:32]
-// GFX908: v_mfma_f32_32x32x2bf16 a[0:31], a0, a1, a[1:32] ; encoding: [0x00,0x00,0xe8,0xd3,0x00,0x03,0x06,0x1c]
+v_mfma_f32_32x32x2bf16 a[0:31], a0, a1, a[33:64]
+// GFX908: v_mfma_f32_32x32x2bf16 a[0:31], a0, a1, a[33:64] ; encoding: [0x00,0x00,0xe8,0xd3,0x00,0x03,0x86,0x1c]
 
-v_mfma_f32_32x32x2bf16 a[0:31], a0, a1, a[1:32] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_32x32x2bf16 a[0:31], a0, a1, a[1:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe8,0xd3,0x00,0x03,0x06,0xfc]
+v_mfma_f32_32x32x2bf16 a[0:31], a0, a1, a[33:64] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_32x32x2bf16 a[0:31], a0, a1, a[33:64] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe8,0xd3,0x00,0x03,0x86,0xfc]
 
-v_mfma_f32_16x16x2bf16 a[0:15], v0, v1, a[1:16]
-// GFX908: v_mfma_f32_16x16x2bf16 a[0:15], v0, v1, a[1:16] ; encoding: [0x00,0x00,0xe9,0xd3,0x00,0x03,0x06,0x04]
+v_mfma_f32_16x16x2bf16 a[0:15], v0, v1, a[17:32]
+// GFX908: v_mfma_f32_16x16x2bf16 a[0:15], v0, v1, a[17:32] ; encoding: [0x00,0x00,0xe9,0xd3,0x00,0x03,0x46,0x04]
 
-v_mfma_f32_16x16x2bf16 a[0:15], v0, v1, a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_16x16x2bf16 a[0:15], v0, v1, a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe9,0xd3,0x00,0x03,0x06,0xe4]
+v_mfma_f32_16x16x2bf16 a[0:15], v0, v1, a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_16x16x2bf16 a[0:15], v0, v1, a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe9,0xd3,0x00,0x03,0x46,0xe4]
 
-v_mfma_f32_16x16x2bf16 a[0:15], v0, a1, a[1:16]
-// GFX908: v_mfma_f32_16x16x2bf16 a[0:15], v0, a1, a[1:16] ; encoding: [0x00,0x00,0xe9,0xd3,0x00,0x03,0x06,0x14]
+v_mfma_f32_16x16x2bf16 a[0:15], v0, a1, a[17:32]
+// GFX908: v_mfma_f32_16x16x2bf16 a[0:15], v0, a1, a[17:32] ; encoding: [0x00,0x00,0xe9,0xd3,0x00,0x03,0x46,0x14]
 
-v_mfma_f32_16x16x2bf16 a[0:15], v0, a1, a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_16x16x2bf16 a[0:15], v0, a1, a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe9,0xd3,0x00,0x03,0x06,0xf4]
+v_mfma_f32_16x16x2bf16 a[0:15], v0, a1, a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_16x16x2bf16 a[0:15], v0, a1, a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe9,0xd3,0x00,0x03,0x46,0xf4]
 
-v_mfma_f32_16x16x2bf16 a[0:15], a0, v1, a[1:16]
-// GFX908: v_mfma_f32_16x16x2bf16 a[0:15], a0, v1, a[1:16] ; encoding: [0x00,0x00,0xe9,0xd3,0x00,0x03,0x06,0x0c]
+v_mfma_f32_16x16x2bf16 a[0:15], a0, v1, a[17:32]
+// GFX908: v_mfma_f32_16x16x2bf16 a[0:15], a0, v1, a[17:32] ; encoding: [0x00,0x00,0xe9,0xd3,0x00,0x03,0x46,0x0c]
 
-v_mfma_f32_16x16x2bf16 a[0:15], a0, v1, a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_16x16x2bf16 a[0:15], a0, v1, a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe9,0xd3,0x00,0x03,0x06,0xec]
+v_mfma_f32_16x16x2bf16 a[0:15], a0, v1, a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_16x16x2bf16 a[0:15], a0, v1, a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe9,0xd3,0x00,0x03,0x46,0xec]
 
-v_mfma_f32_16x16x2bf16 a[0:15], a0, a1, a[1:16]
-// GFX908: v_mfma_f32_16x16x2bf16 a[0:15], a0, a1, a[1:16] ; encoding: [0x00,0x00,0xe9,0xd3,0x00,0x03,0x06,0x1c]
+v_mfma_f32_16x16x2bf16 a[0:15], a0, a1, a[17:32]
+// GFX908: v_mfma_f32_16x16x2bf16 a[0:15], a0, a1, a[17:32] ; encoding: [0x00,0x00,0xe9,0xd3,0x00,0x03,0x46,0x1c]
 
-v_mfma_f32_16x16x2bf16 a[0:15], a0, a1, a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_16x16x2bf16 a[0:15], a0, a1, a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe9,0xd3,0x00,0x03,0x06,0xfc]
+v_mfma_f32_16x16x2bf16 a[0:15], a0, a1, a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_16x16x2bf16 a[0:15], a0, a1, a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xe9,0xd3,0x00,0x03,0x46,0xfc]
 
 v_mfma_f32_4x4x2bf16 a[0:3], v0, v1, a[1:4]
 // GFX908: v_mfma_f32_4x4x2bf16 a[0:3], v0, v1, a[1:4] ; encoding: [0x00,0x00,0xeb,0xd3,0x00,0x03,0x06,0x04]
@@ -475,29 +481,29 @@ v_mfma_f32_4x4x2bf16 a[0:3], a0, a1, a[1:4]
 v_mfma_f32_4x4x2bf16 a[0:3], a0, a1, a[1:4] cbsz:3 abid:2 blgp:7
 // GFX908: v_mfma_f32_4x4x2bf16 a[0:3], a0, a1, a[1:4] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xeb,0xd3,0x00,0x03,0x06,0xfc]
 
-v_mfma_f32_32x32x4bf16 a[0:15], v0, v1, a[1:16]
-// GFX908: v_mfma_f32_32x32x4bf16 a[0:15], v0, v1, a[1:16] ; encoding: [0x00,0x00,0xec,0xd3,0x00,0x03,0x06,0x04]
+v_mfma_f32_32x32x4bf16 a[0:15], v0, v1, a[17:32]
+// GFX908: v_mfma_f32_32x32x4bf16 a[0:15], v0, v1, a[17:32] ; encoding: [0x00,0x00,0xec,0xd3,0x00,0x03,0x46,0x04]
 
-v_mfma_f32_32x32x4bf16 a[0:15], v0, v1, a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_32x32x4bf16 a[0:15], v0, v1, a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xec,0xd3,0x00,0x03,0x06,0xe4]
+v_mfma_f32_32x32x4bf16 a[0:15], v0, v1, a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_32x32x4bf16 a[0:15], v0, v1, a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xec,0xd3,0x00,0x03,0x46,0xe4]
 
-v_mfma_f32_32x32x4bf16 a[0:15], v0, a1, a[1:16]
-// GFX908: v_mfma_f32_32x32x4bf16 a[0:15], v0, a1, a[1:16] ; encoding: [0x00,0x00,0xec,0xd3,0x00,0x03,0x06,0x14]
+v_mfma_f32_32x32x4bf16 a[0:15], v0, a1, a[17:32]
+// GFX908: v_mfma_f32_32x32x4bf16 a[0:15], v0, a1, a[17:32] ; encoding: [0x00,0x00,0xec,0xd3,0x00,0x03,0x46,0x14]
 
-v_mfma_f32_32x32x4bf16 a[0:15], v0, a1, a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_32x32x4bf16 a[0:15], v0, a1, a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xec,0xd3,0x00,0x03,0x06,0xf4]
+v_mfma_f32_32x32x4bf16 a[0:15], v0, a1, a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_32x32x4bf16 a[0:15], v0, a1, a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xec,0xd3,0x00,0x03,0x46,0xf4]
 
-v_mfma_f32_32x32x4bf16 a[0:15], a0, v1, a[1:16]
-// GFX908: v_mfma_f32_32x32x4bf16 a[0:15], a0, v1, a[1:16] ; encoding: [0x00,0x00,0xec,0xd3,0x00,0x03,0x06,0x0c]
+v_mfma_f32_32x32x4bf16 a[0:15], a0, v1, a[17:32]
+// GFX908: v_mfma_f32_32x32x4bf16 a[0:15], a0, v1, a[17:32] ; encoding: [0x00,0x00,0xec,0xd3,0x00,0x03,0x46,0x0c]
 
-v_mfma_f32_32x32x4bf16 a[0:15], a0, v1, a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_32x32x4bf16 a[0:15], a0, v1, a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xec,0xd3,0x00,0x03,0x06,0xec]
+v_mfma_f32_32x32x4bf16 a[0:15], a0, v1, a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_32x32x4bf16 a[0:15], a0, v1, a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xec,0xd3,0x00,0x03,0x46,0xec]
 
-v_mfma_f32_32x32x4bf16 a[0:15], a0, a1, a[1:16]
-// GFX908: v_mfma_f32_32x32x4bf16 a[0:15], a0, a1, a[1:16] ; encoding: [0x00,0x00,0xec,0xd3,0x00,0x03,0x06,0x1c]
+v_mfma_f32_32x32x4bf16 a[0:15], a0, a1, a[17:32]
+// GFX908: v_mfma_f32_32x32x4bf16 a[0:15], a0, a1, a[17:32] ; encoding: [0x00,0x00,0xec,0xd3,0x00,0x03,0x46,0x1c]
 
-v_mfma_f32_32x32x4bf16 a[0:15], a0, a1, a[1:16] cbsz:3 abid:2 blgp:7
-// GFX908: v_mfma_f32_32x32x4bf16 a[0:15], a0, a1, a[1:16] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xec,0xd3,0x00,0x03,0x06,0xfc]
+v_mfma_f32_32x32x4bf16 a[0:15], a0, a1, a[17:32] cbsz:3 abid:2 blgp:7
+// GFX908: v_mfma_f32_32x32x4bf16 a[0:15], a0, a1, a[17:32] cbsz:3 abid:2 blgp:7 ; encoding: [0x00,0x13,0xec,0xd3,0x00,0x03,0x46,0xfc]
 
 v_mfma_f32_16x16x8bf16 a[0:3], v0, v1, a[1:4]
 // GFX908: v_mfma_f32_16x16x8bf16 a[0:3], v0, v1, a[1:4] ; encoding: [0x00,0x00,0xed,0xd3,0x00,0x03,0x06,0x04]

diff  --git a/llvm/test/MC/AMDGPU/misaligned-vgpr-tuples-err.s b/llvm/test/MC/AMDGPU/misaligned-vgpr-tuples-err.s
index d2cd4383bb417..4056a8ac90333 100644
--- a/llvm/test/MC/AMDGPU/misaligned-vgpr-tuples-err.s
+++ b/llvm/test/MC/AMDGPU/misaligned-vgpr-tuples-err.s
@@ -18,8 +18,8 @@ v_mfma_f32_32x32x8f16 a[0:15], a[1:2], v[0:1], a[0:15]
 v_mfma_i32_4x4x4i8 a[1:4], a0, v1, 2
 // GFX90A: error: invalid register class: vgpr tuples must be 64 bit aligned
 
-v_mfma_f32_16x16x1f32 a[0:15], a0, v1, a[1:16]
+v_mfma_f32_16x16x1f32 a[0:15], a0, v1, a[17:32]
 // GFX90A: error: invalid register class: vgpr tuples must be 64 bit aligned
 
-v_mfma_f32_32x32x1f32 a[0:31], v0, v1, a[1:32]
+v_mfma_f32_32x32x1f32 a[0:31], v0, v1, a[33:64]
 // GFX90A: error: invalid register class: vgpr tuples must be 64 bit aligned


        


More information about the llvm-commits mailing list