llvm · MrSidims · May 26, 2025 · May 20, 2025 · May 22, 2025 · May 23, 2025
diff --git a/llvm/docs/SPIRVUsage.rst b/llvm/docs/SPIRVUsage.rst
@@ -215,6 +215,8 @@ list of supported SPIR-V extensions, sorted alphabetically by their extension na
      - Adds a bitwise instruction on three operands and a look-up table index for specifying the bitwise operation to perform. 
    * - ``SPV_INTEL_subgroup_matrix_multiply_accumulate``
      - Adds an instruction to compute the matrix product of an M x K matrix with a K x N matrix and then add an M x N matrix. 
+   * - ``SPV_INTEL_int4``
+     - Adds support for 4-bit integer type, and allow this type to be used in cooperative matrices.
 
 To enable multiple extensions, list them separated by comma. For example, to enable support for atomic operations on floating-point numbers and arbitrary precision integers, use:
 

diff --git a/llvm/lib/Target/SPIRV/SPIRVCommandLine.cpp b/llvm/lib/Target/SPIRV/SPIRVCommandLine.cpp
@@ -99,7 +99,8 @@ static const std::map<std::string, SPIRV::Extension::Extension, std::less<>>
         {"SPV_INTEL_ternary_bitwise_function",
          SPIRV::Extension::Extension::SPV_INTEL_ternary_bitwise_function},
         {"SPV_INTEL_2d_block_io",
-         SPIRV::Extension::Extension::SPV_INTEL_2d_block_io}};
+         SPIRV::Extension::Extension::SPV_INTEL_2d_block_io},
+        {"SPV_INTEL_int4", SPIRV::Extension::Extension::SPV_INTEL_int4}};
 
 bool SPIRVExtensionsParser::parse(cl::Option &O, StringRef ArgName,
                                   StringRef ArgValue,

diff --git a/llvm/lib/Target/SPIRV/SPIRVGlobalRegistry.cpp b/llvm/lib/Target/SPIRV/SPIRVGlobalRegistry.cpp
@@ -154,7 +154,8 @@ unsigned SPIRVGlobalRegistry::adjustOpTypeIntWidth(unsigned Width) const {
     report_fatal_error("Unsupported integer width!");
   const SPIRVSubtarget &ST = cast<SPIRVSubtarget>(CurMF->getSubtarget());
   if (ST.canUseExtension(
-          SPIRV::Extension::SPV_INTEL_arbitrary_precision_integers))
+          SPIRV::Extension::SPV_INTEL_arbitrary_precision_integers) ||
+      ST.canUseExtension(SPIRV::Extension::SPV_INTEL_int4))
     return Width;
   if (Width <= 8)
     Width = 8;
@@ -174,9 +175,14 @@ SPIRVType *SPIRVGlobalRegistry::getOpTypeInt(unsigned Width,
   const SPIRVSubtarget &ST =
       cast<SPIRVSubtarget>(MIRBuilder.getMF().getSubtarget());
   return createOpType(MIRBuilder, [&](MachineIRBuilder &MIRBuilder) {
-    if ((!isPowerOf2_32(Width) || Width < 8) &&
-        ST.canUseExtension(
-            SPIRV::Extension::SPV_INTEL_arbitrary_precision_integers)) {
+    if (Width == 4 && ST.canUseExtension(SPIRV::Extension::SPV_INTEL_int4)) {
+      MIRBuilder.buildInstr(SPIRV::OpExtension)
+          .addImm(SPIRV::Extension::SPV_INTEL_int4);
+      MIRBuilder.buildInstr(SPIRV::OpCapability)
+          .addImm(SPIRV::Capability::Int4TypeINTEL);
+    } else if ((!isPowerOf2_32(Width) || Width < 8) &&
+               ST.canUseExtension(
+                   SPIRV::Extension::SPV_INTEL_arbitrary_precision_integers)) {
       MIRBuilder.buildInstr(SPIRV::OpExtension)
           .addImm(SPIRV::Extension::SPV_INTEL_arbitrary_precision_integers);
       MIRBuilder.buildInstr(SPIRV::OpCapability)
@@ -1563,6 +1569,13 @@ SPIRVType *SPIRVGlobalRegistry::getOrCreateOpTypeCoopMatr(
   const MachineInstr *NewMI =
       createOpType(MIRBuilder, [&](MachineIRBuilder &MIRBuilder) {
         SPIRVType *SpvTypeInt32 = getOrCreateSPIRVIntegerType(32, MIRBuilder);
+        const Type *ET = getTypeForSPIRVType(ElemType);
+        if (ET->isIntegerTy() && ET->getIntegerBitWidth() == 4 &&
+            cast<SPIRVSubtarget>(MIRBuilder.getMF().getSubtarget())
+                .canUseExtension(SPIRV::Extension::SPV_INTEL_int4)) {
+          MIRBuilder.buildInstr(SPIRV::OpCapability)
+              .addImm(SPIRV::Capability::Int4CooperativeMatrixINTEL);
+        }
         return MIRBuilder.buildInstr(SPIRV::OpTypeCooperativeMatrixKHR)
             .addDef(createTypeVReg(MIRBuilder))
             .addUse(getSPIRVTypeID(ElemType))

diff --git a/llvm/lib/Target/SPIRV/SPIRVLegalizerInfo.cpp b/llvm/lib/Target/SPIRV/SPIRVLegalizerInfo.cpp
@@ -128,7 +128,8 @@ SPIRVLegalizerInfo::SPIRVLegalizerInfo(const SPIRVSubtarget &ST) {
   bool IsExtendedInts =
       ST.canUseExtension(
           SPIRV::Extension::SPV_INTEL_arbitrary_precision_integers) ||
-      ST.canUseExtension(SPIRV::Extension::SPV_KHR_bit_instructions);
+      ST.canUseExtension(SPIRV::Extension::SPV_KHR_bit_instructions) ||
+      ST.canUseExtension(SPIRV::Extension::SPV_INTEL_int4);
   auto extendedScalarsAndVectors =
       [IsExtendedInts](const LegalityQuery &Query) {
         const LLT Ty = Query.Types[0];

diff --git a/llvm/lib/Target/SPIRV/SPIRVPreLegalizer.cpp b/llvm/lib/Target/SPIRV/SPIRVPreLegalizer.cpp
@@ -380,16 +380,31 @@ static SPIRVType *propagateSPIRVType(MachineInstr *MI, SPIRVGlobalRegistry *GR,
 // To support current approach and limitations wrt. bit width here we widen a
 // scalar register with a bit width greater than 1 to valid sizes and cap it to
 // 64 width.
-static void widenScalarLLTNextPow2(Register Reg, MachineRegisterInfo &MRI) {
+static unsigned widenBitWidthToNextPow2(unsigned BitWidth) {
+  if (BitWidth == 1)
+    return 1; // No need to widen 1-bit values
+  return std::min(std::max(1u << Log2_32_Ceil(BitWidth), 8u), 64u);
+}
+
+static void widenScalarType(Register Reg, MachineRegisterInfo &MRI) {
   LLT RegType = MRI.getType(Reg);
   if (!RegType.isScalar())
     return;
-  unsigned Sz = RegType.getScalarSizeInBits();
-  if (Sz == 1)
-    return;
-  unsigned NewSz = std::min(std::max(1u << Log2_32_Ceil(Sz), 8u), 64u);
-  if (NewSz != Sz)
-    MRI.setType(Reg, LLT::scalar(NewSz));
+  unsigned CurrentWidth = RegType.getScalarSizeInBits();
+  unsigned NewWidth = widenBitWidthToNextPow2(CurrentWidth);
+  if (NewWidth != CurrentWidth)
+    MRI.setType(Reg, LLT::scalar(NewWidth));
+}
+
+static void widenCImmType(MachineOperand &MOP) {
+  const ConstantInt *CImmVal = MOP.getCImm();
+  unsigned CurrentWidth = CImmVal->getBitWidth();
+  unsigned NewWidth = widenBitWidthToNextPow2(CurrentWidth);
+  if (NewWidth != CurrentWidth) {
+    // Replace the immediate value with the widened version
+    MOP.setCImm(ConstantInt::get(CImmVal->getType()->getContext(),
+                                 CImmVal->getValue().zextOrTrunc(NewWidth)));
+  }
 }
 
 static void setInsertPtAfterDef(MachineIRBuilder &MIB, MachineInstr *Def) {
@@ -492,7 +507,8 @@ generateAssignInstrs(MachineFunction &MF, SPIRVGlobalRegistry *GR,
   bool IsExtendedInts =
       ST->canUseExtension(
           SPIRV::Extension::SPV_INTEL_arbitrary_precision_integers) ||
-      ST->canUseExtension(SPIRV::Extension::SPV_KHR_bit_instructions);
+      ST->canUseExtension(SPIRV::Extension::SPV_KHR_bit_instructions) ||
+      ST->canUseExtension(SPIRV::Extension::SPV_INTEL_int4);
 
   for (MachineBasicBlock *MBB : post_order(&MF)) {
     if (MBB->empty())
@@ -505,10 +521,13 @@ generateAssignInstrs(MachineFunction &MF, SPIRVGlobalRegistry *GR,
       unsigned MIOp = MI.getOpcode();
 
       if (!IsExtendedInts) {
-        // validate bit width of scalar registers
-        for (const auto &MOP : MI.operands())
+        // validate bit width of scalar registers and constant immediates
+        for (auto &MOP : MI.operands()) {
           if (MOP.isReg())
-            widenScalarLLTNextPow2(MOP.getReg(), MRI);
+            widenScalarType(MOP.getReg(), MRI);
+          else if (MOP.isCImm())
+            widenCImmType(MOP);
+        }
       }
 
       if (isSpvIntrinsic(MI, Intrinsic::spv_assign_ptr_type)) {

diff --git a/llvm/lib/Target/SPIRV/SPIRVSymbolicOperands.td b/llvm/lib/Target/SPIRV/SPIRVSymbolicOperands.td
@@ -316,6 +316,7 @@ defm SPV_INTEL_fp_max_error : ExtensionOperand<119>;
 defm SPV_INTEL_ternary_bitwise_function : ExtensionOperand<120>;
 defm SPV_INTEL_subgroup_matrix_multiply_accumulate : ExtensionOperand<121>;
 defm SPV_INTEL_2d_block_io : ExtensionOperand<122>;
+defm SPV_INTEL_int4 : ExtensionOperand<123>;
 
 //===----------------------------------------------------------------------===//
 // Multiclass used to define Capabilities enum values and at the same time
@@ -521,6 +522,8 @@ defm SubgroupMatrixMultiplyAccumulateINTEL : CapabilityOperand<6236, 0, 0, [SPV_
 defm Subgroup2DBlockIOINTEL : CapabilityOperand<6228, 0, 0, [SPV_INTEL_2d_block_io], []>;
 defm Subgroup2DBlockTransformINTEL : CapabilityOperand<6229, 0, 0, [SPV_INTEL_2d_block_io], [Subgroup2DBlockIOINTEL]>;
 defm Subgroup2DBlockTransposeINTEL : CapabilityOperand<6230, 0, 0, [SPV_INTEL_2d_block_io], [Subgroup2DBlockIOINTEL]>;
+defm Int4TypeINTEL : CapabilityOperand<5112, 0, 0, [SPV_INTEL_int4], []>;
+defm Int4CooperativeMatrixINTEL : CapabilityOperand<5114, 0, 0, [SPV_INTEL_int4], [Int4TypeINTEL, CooperativeMatrixKHR]>;
 
 //===----------------------------------------------------------------------===//
 // Multiclass used to define SourceLanguage enum values and at the same time

diff --git a/llvm/test/CodeGen/SPIRV/extensions/SPV_INTEL_int4/cooperative_matrix.ll b/llvm/test/CodeGen/SPIRV/extensions/SPV_INTEL_int4/cooperative_matrix.ll
@@ -0,0 +1,20 @@
+; RUN: llc -verify-machineinstrs -O0 -mtriple=spirv32-unknown-unknown --spirv-ext=+SPV_INTEL_int4,+SPV_KHR_cooperative_matrix %s -o - | FileCheck %s
+; RUNx: %if spirv-tools %{ llc -O0 -mtriple=spirv32-unknown-unknown --spirv-ext=+SPV_INTEL_int4,+SPV_KHR_cooperative_matrix %s -o - -filetype=obj | spirv-val %}
+
+; CHECK-DAG: Capability Int4TypeINTEL
+; CHECK-DAG: Capability CooperativeMatrixKHR
+; CHECK-DAG: Extension "SPV_INTEL_int4"
+; CHECK-DAG: Capability Int4CooperativeMatrixINTEL
+; CHECK-DAG: Extension "SPV_KHR_cooperative_matrix"
+
+; CHECK: %[[#Int4Ty:]] = OpTypeInt 4 0
+; CHECK: %[[#CoopMatTy:]] = OpTypeCooperativeMatrixKHR %[[#Int4Ty]]
+; CHECK: CompositeConstruct %[[#CoopMatTy]]
+
+define spir_kernel void @foo() {
+entry:
+  %call.i.i = tail call spir_func noundef target("spirv.CooperativeMatrixKHR", i4, 3, 12, 12, 2) @_Z26__spirv_CompositeConstruct(i4 noundef 0)
+  ret void
+}
+
+declare dso_local spir_func noundef target("spirv.CooperativeMatrixKHR", i4, 3, 12, 12, 2) @_Z26__spirv_CompositeConstruct(i4 noundef)
diff --git a/llvm/test/CodeGen/SPIRV/extensions/SPV_INTEL_int4/negative.ll b/llvm/test/CodeGen/SPIRV/extensions/SPV_INTEL_int4/negative.ll
@@ -0,0 +1,29 @@
+; RUN: llc -verify-machineinstrs -O0 -mtriple=spirv32-unknown-unknown --spirv-ext=+SPV_INTEL_arbitrary_precision_integers %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-INT-4
+
+; RUN: llc -verify-machineinstrs -O0 -mtriple=spirv32-unknown-unknown %s -o - | FileCheck %s --check-prefixes=CHECK,CHECK-INT-8
+; No error would be reported in comparison to Khronos llvm-spirv, because type adjustments to integer size are made 
+; in case no appropriate extension is enabled. Here we expect that the type is adjusted to 8 bits.
+
+; CHECK-SPIRV: Capability ArbitraryPrecisionIntegersINTEL
+; CHECK-SPIRV: Extension "SPV_INTEL_arbitrary_precision_integers"
+; CHECK-INT-4: %[[#Int4:]] = OpTypeInt 4 0
+; CHECK-INT-8: %[[#Int4:]] = OpTypeInt 8 0
+; CHECK: OpTypeFunction %[[#]] %[[#Int4]]
+; CHECK: %[[#Int4PtrTy:]] = OpTypePointer Function %[[#Int4]]
+; CHECK: %[[#Const:]] = OpConstant %[[#Int4]]  1
+
+; CHECK: %[[#Int4Ptr:]] = OpVariable %[[#Int4PtrTy]] Function
+; CHECK: OpStore %[[#Int4Ptr]] %[[#Const]]
+; CHECK: %[[#Load:]] = OpLoad %[[#Int4]] %[[#Int4Ptr]]
+; CHECK: OpFunctionCall %[[#]] %[[#]] %[[#Load]]
+
+define spir_kernel void @foo() {
+entry:
+  %0 = alloca i4
+  store i4 1, ptr %0
+  %1 = load i4, ptr %0
+  call spir_func void @boo(i4 %1)
+  ret void
+}
+
+declare spir_func void @boo(i4)
diff --git a/llvm/test/CodeGen/SPIRV/extensions/SPV_INTEL_int4/trivial.ll b/llvm/test/CodeGen/SPIRV/extensions/SPV_INTEL_int4/trivial.ll
@@ -0,0 +1,25 @@
+; RUN: llc -verify-machineinstrs -O0 -mtriple=spirv32-unknown-unknown --spirv-ext=+SPV_INTEL_int4 %s -o - | FileCheck %s
+; RUNx: %if spirv-tools %{ llc -O0 -mtriple=spirv32-unknown-unknown --spirv-ext=+SPV_INTEL_int4 %s -o - -filetype=obj | spirv-val %}
+
+; CHECK: Capability Int4TypeINTEL
+; CHECK: Extension "SPV_INTEL_int4"
+; CHECK: %[[#Int4:]] = OpTypeInt  4 0
+; CHECK: OpTypeFunction %[[#]] %[[#Int4]]
+; CHECK: %[[#Int4PtrTy:]] = OpTypePointer Function %[[#Int4]]
+; CHECK: %[[#Const:]] = OpConstant %[[#Int4]]  1
+
+; CHECK: %[[#Int4Ptr:]] = OpVariable %[[#Int4PtrTy]] Function
+; CHECK: OpStore %[[#Int4Ptr]] %[[#Const]]
+; CHECK: %[[#Load:]] = OpLoad %[[#Int4]] %[[#Int4Ptr]]
+; CHECK: OpFunctionCall %[[#]] %[[#]] %[[#Load]]
+
+define spir_kernel void @foo() {
+entry:
+  %0 = alloca i4
+  store i4 1, ptr %0
+  %1 = load i4, ptr %0
+  call spir_func void @boo(i4 %1)
+  ret void
+}
+
+declare spir_func void @boo(i4)