Skip to content

[SPIRV] Add sign intrinsic part 1 - #101987

Merged
farzonl merged 1 commit into
llvm:mainfrom
tgymnich:hlsl-sign-1
Sep 9, 2024
Merged

[SPIRV] Add sign intrinsic part 1#101987
farzonl merged 1 commit into
llvm:mainfrom
tgymnich:hlsl-sign-1

Conversation

@tgymnich

@tgymnich tgymnich commented Aug 5, 2024 •
edited
Loading

Copy link
Copy Markdown
Member

partially fixes #70078

Changes

  • Added int_spv_sign intrinsic in IntrinsicsSPIRV.td
  • Added lowering and map to int_spv_sign in SPIRVInstructionSelector.cpp`
  • Added SPIR-V backend test case in llvm/test/CodeGen/SPIRV/hlsl-intrinsics/sign.ll

Related PRs

@github-actions

github-actions Bot commented Aug 5, 2024 •
edited
Loading

Copy link
Copy Markdown

✅ With the latest revision this PR passed the C/C++ code formatter.

This was referenced Aug 5, 2024
Merged
Merged
@tgymnich
tgymnich marked this pull request as ready for review August 5, 2024 14:26
@llvmbot

llvmbot commented Aug 5, 2024

Copy link
Copy Markdown
Member

@llvm/pr-subscribers-llvm-ir

Author: Tim Gymnich (tgymnich)

Changes

partially fixes #70078

Changes

  • Added int_spv_sign intrinsic in IntrinsicsSPIRV.td
  • Added lowering and map to int_spv_sign in SPIRVInstructionSelector.cpp`
  • Added SPIR-V backend test case in llvm/test/CodeGen/SPIRV/hlsl-intrinsics/sign.ll

Related PRs

Full diff: https://github.com/llvm/llvm-project/pull/101987.diff

3 Files Affected:

  • (modified) llvm/include/llvm/IR/IntrinsicsSPIRV.td (+1)
  • (modified) llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp (+52)
  • (added) llvm/test/CodeGen/SPIRV/hlsl-intrinsics/sign.ll (+143)
diff --git a/llvm/include/llvm/IR/IntrinsicsSPIRV.td b/llvm/include/llvm/IR/IntrinsicsSPIRV.td
index 3f77ef6bfcdbe..89eacf2ba70f9 100644
--- a/llvm/include/llvm/IR/IntrinsicsSPIRV.td
+++ b/llvm/include/llvm/IR/IntrinsicsSPIRV.td
@@ -65,4 +65,5 @@ let TargetPrefix = "spv" in {
     [IntrNoMem, IntrWillReturn] >;
   def int_spv_length : DefaultAttrsIntrinsic<[LLVMVectorElementType<0>], [llvm_anyfloat_ty]>;
   def int_spv_rsqrt : DefaultAttrsIntrinsic<[LLVMMatchType<0>], [llvm_anyfloat_ty]>;
+  def int_spv_sign : DefaultAttrsIntrinsic<[LLVMScalarOrSameVectorWidth<0, llvm_i32_ty>], [llvm_any_ty]>;
 }
diff --git a/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp b/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
index ed786bd33aa05..47e4b3dd40339 100644
--- a/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
+++ b/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
@@ -28,6 +28,7 @@
 #include "llvm/CodeGen/MachineInstrBuilder.h"
 #include "llvm/CodeGen/MachineModuleInfoImpls.h"
 #include "llvm/CodeGen/MachineRegisterInfo.h"
+#include "llvm/CodeGen/Register.h"
 #include "llvm/CodeGen/TargetOpcodes.h"
 #include "llvm/IR/IntrinsicsSPIRV.h"
 #include "llvm/Support/Debug.h"
@@ -178,6 +179,9 @@ class SPIRVInstructionSelector : public InstructionSelector {
   bool selectRsqrt(Register ResVReg, const SPIRVType *ResType,
                    MachineInstr &I) const;
 
+  bool selectSign(Register ResVReg, const SPIRVType *ResType,
+                  MachineInstr &I) const;
+
   void renderImm32(MachineInstrBuilder &MIB, const MachineInstr &I,
                    int OpIdx) const;
   void renderFImm32(MachineInstrBuilder &MIB, const MachineInstr &I,
@@ -1366,6 +1370,52 @@ bool SPIRVInstructionSelector::selectRsqrt(Register ResVReg,
       .constrainAllUses(TII, TRI, RBI);
 }
 
+bool SPIRVInstructionSelector::selectSign(Register ResVReg,
+                                          const SPIRVType *ResType,
+                                          MachineInstr &I) const {
+  assert(I.getNumOperands() == 3);
+  assert(I.getOperand(2).isReg());
+  MachineBasicBlock &BB = *I.getParent();
+  Register InputRegister = I.getOperand(2).getReg();
+  SPIRVType *InputType = GR.getSPIRVTypeForVReg(InputRegister);
+  auto &DL = I.getDebugLoc();
+
+  if (!InputType)
+    report_fatal_error("Input Type could not be determined.");
+
+  bool IsFloatTy = GR.isScalarOrVectorOfType(InputRegister, SPIRV::OpTypeFloat);
+
+  unsigned SignBitWidth = GR.getScalarOrVectorBitWidth(InputType);
+  unsigned ResBitWidth = GR.getScalarOrVectorBitWidth(ResType);
+
+  bool NeedsConversion = IsFloatTy || SignBitWidth != ResBitWidth;
+
+  auto SignOpcode = IsFloatTy ? GL::FSign : GL::SSign;
+  Register SignReg = NeedsConversion
+                         ? MRI->createVirtualRegister(&SPIRV::IDRegClass)
+                         : ResVReg;
+
+  bool Result =
+      BuildMI(BB, I, DL, TII.get(SPIRV::OpExtInst))
+          .addDef(SignReg)
+          .addUse(GR.getSPIRVTypeID(InputType))
+          .addImm(static_cast<uint32_t>(SPIRV::InstructionSet::GLSL_std_450))
+          .addImm(SignOpcode)
+          .addUse(InputRegister)
+          .constrainAllUses(TII, TRI, RBI);
+
+  if (NeedsConversion) {
+    auto ConvertOpcode = IsFloatTy ? SPIRV::OpConvertFToS : SPIRV::OpSConvert;
+    Result |= BuildMI(*I.getParent(), I, DL, TII.get(ConvertOpcode))
+                  .addDef(ResVReg)
+                  .addUse(GR.getSPIRVTypeID(ResType))
+                  .addUse(SignReg)
+                  .constrainAllUses(TII, TRI, RBI);
+  }
+
+  return Result;
+}
+
 bool SPIRVInstructionSelector::selectBitreverse(Register ResVReg,
                                                 const SPIRVType *ResType,
                                                 MachineInstr &I) const {
@@ -2082,6 +2132,8 @@ bool SPIRVInstructionSelector::selectIntrinsic(Register ResVReg,
     return selectFrac(ResVReg, ResType, I);
   case Intrinsic::spv_rsqrt:
     return selectRsqrt(ResVReg, ResType, I);
+  case Intrinsic::spv_sign:
+    return selectSign(ResVReg, ResType, I);
   case Intrinsic::spv_lifetime_start:
   case Intrinsic::spv_lifetime_end: {
     unsigned Op = IID == Intrinsic::spv_lifetime_start ? SPIRV::OpLifetimeStart
diff --git a/llvm/test/CodeGen/SPIRV/hlsl-intrinsics/sign.ll b/llvm/test/CodeGen/SPIRV/hlsl-intrinsics/sign.ll
new file mode 100644
index 0000000000000..52a41c3d3ad64
--- /dev/null
+++ b/llvm/test/CodeGen/SPIRV/hlsl-intrinsics/sign.ll
@@ -0,0 +1,143 @@
+; RUN: llc -O0 -mtriple=spirv-unknown-unknown %s -o - | FileCheck %s
+; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv-unknown-unknown %s -o - -filetype=obj | spirv-val %}
+
+; CHECK-DAG: %[[#op_ext_glsl:]] = OpExtInstImport "GLSL.std.450"
+
+; CHECK-DAG: %[[#float_16:]] = OpTypeFloat 16
+; CHECK-DAG: %[[#float_32:]] = OpTypeFloat 32
+; CHECK-DAG: %[[#float_64:]] = OpTypeFloat 64
+
+; CHECK-DAG: %[[#int_16:]] = OpTypeInt 16
+; CHECK-DAG: %[[#int_32:]] = OpTypeInt 32
+; CHECK-DAG: %[[#int_64:]] = OpTypeInt 64
+
+; CHECK-DAG: %[[#vec4_float_16:]] = OpTypeVector %[[#float_16]] 4
+; CHECK-DAG: %[[#vec4_float_32:]] = OpTypeVector %[[#float_32]] 4
+; CHECK-DAG: %[[#vec4_float_64:]] = OpTypeVector %[[#float_64]] 4
+
+; CHECK-DAG: %[[#vec4_int_16:]] = OpTypeVector %[[#int_16]] 4
+; CHECK-DAG: %[[#vec4_int_32:]] = OpTypeVector %[[#int_32]] 4
+; CHECK-DAG: %[[#vec4_int_64:]] = OpTypeVector %[[#int_64]] 4
+
+
+define noundef i32 @sign_half(half noundef %a) {
+entry:
+; CHECK: %[[#float_16_arg:]] = OpFunctionParameter %[[#float_16]]
+; CHECK: %[[#fsign:]] = OpExtInst %[[#float_16]] %[[#op_ext_glsl]] FSign %[[#float_16_arg]]
+; CHECK: %[[#]] = OpConvertFToS %[[#int_32]] %[[#fsign]]
+  %elt.sign = call i32 @llvm.spv.sign.f16(half %a)
+  ret i32 %elt.sign
+}
+
+define noundef i32 @sign_float(float noundef %a) {
+entry:
+; CHECK: %[[#float_32_arg:]] = OpFunctionParameter %[[#float_32]]
+; CHECK: %[[#fsign:]] = OpExtInst %[[#float_32]] %[[#op_ext_glsl]] FSign %[[#float_32_arg]]
+; CHECK: %[[#]] = OpConvertFToS %[[#int_32]] %[[#fsign]]
+  %elt.sign = call i32 @llvm.spv.sign.f32(float %a)
+  ret i32 %elt.sign
+}
+
+define noundef i32 @sign_double(double noundef %a) {
+entry:
+; CHECK: %[[#float_64_arg:]] = OpFunctionParameter %[[#float_64]]
+; CHECK: %[[#fsign:]] = OpExtInst %[[#float_64]] %[[#op_ext_glsl]] FSign %[[#float_64_arg]]
+; CHECK: %[[#]] = OpConvertFToS %[[#int_32]] %[[#fsign]]
+  %elt.sign = call i32 @llvm.spv.sign.f64(double %a)
+  ret i32 %elt.sign
+}
+
+define noundef i32 @sign_i16(i16 noundef %a) {
+entry:
+; CHECK: %[[#int_16_arg:]] = OpFunctionParameter %[[#int_16]]
+; CHECK: %[[#ssign:]] = OpExtInst %[[#int_16]] %[[#op_ext_glsl]] SSign %[[#int_16_arg]]
+; CHECK: %[[#]] = OpSConvert %[[#int_32]] %[[#ssign]]
+  %elt.sign = call i32 @llvm.spv.sign.i16(i16 %a)
+  ret i32 %elt.sign
+}
+
+define noundef i32 @sign_i32(i32 noundef %a) {
+entry:
+; CHECK: %[[#int_32_arg:]] = OpFunctionParameter %[[#int_32]]
+; CHECK: %[[#]] = OpExtInst %[[#int_32]] %[[#op_ext_glsl]] SSign %[[#int_32_arg]]
+  %elt.sign = call i32 @llvm.spv.sign.i32(i32 %a)
+  ret i32 %elt.sign
+}
+
+define noundef i32 @sign_i64(i64 noundef %a) {
+entry:
+; CHECK: %[[#int_64_arg:]] = OpFunctionParameter %[[#int_64]]
+; CHECK: %[[#ssign:]] = OpExtInst %[[#int_64]] %[[#op_ext_glsl]] SSign %[[#int_64_arg]]
+; CHECK: %[[#]] = OpSConvert %[[#int_32]] %[[#ssign]]
+  %elt.sign = call i32 @llvm.spv.sign.i64(i64 %a)
+  ret i32 %elt.sign
+}
+
+define noundef <4 x i32> @sign_half_vector(<4 x half> noundef %a) {
+entry:
+; CHECK: %[[#vec4_float_16_arg:]] = OpFunctionParameter %[[#vec4_float_16]]
+; CHECK: %[[#fsign:]] = OpExtInst %[[#vec4_float_16]] %[[#op_ext_glsl]] FSign %[[#vec4_float_16_arg]]
+; CHECK: %[[#]] = OpConvertFToS %[[#vec4_int_32]] %[[#fsign]]
+  %elt.sign = call <4 x i32> @llvm.spv.sign.v4f16(<4 x half> %a)
+  ret <4 x i32> %elt.sign
+}
+
+define noundef <4 x i32> @sign_float_vector(<4 x float> noundef %a) {
+entry:
+; CHECK: %[[#vec4_float_32_arg:]] = OpFunctionParameter %[[#vec4_float_32]]
+; CHECK: %[[#fsign:]] = OpExtInst %[[#vec4_float_32]] %[[#op_ext_glsl]] FSign %[[#vec4_float_32_arg]]
+; CHECK: %[[#]] = OpConvertFToS %[[#vec4_int_32]] %[[#fsign]]
+  %elt.sign = call <4 x i32> @llvm.spv.sign.v4f32(<4 x float> %a)
+  ret <4 x i32> %elt.sign
+}
+
+define noundef <4 x i32> @sign_double_vector(<4 x double> noundef %a) {
+entry:
+; CHECK: %[[#vec4_float_64_arg:]] = OpFunctionParameter %[[#vec4_float_64]]
+; CHECK: %[[#fsign:]] = OpExtInst %[[#vec4_float_64]] %[[#op_ext_glsl]] FSign %[[#vec4_float_64_arg]]
+; CHECK: %[[#]] = OpConvertFToS %[[#vec4_int_32]] %[[#fsign]]
+  %elt.sign = call <4 x i32> @llvm.spv.sign.v4f64(<4 x double> %a)
+  ret <4 x i32> %elt.sign
+}
+
+define noundef <4 x i32> @sign_i16_vector(<4 x i16> noundef %a) {
+entry:
+; CHECK: %[[#vec4_int_16_arg:]] = OpFunctionParameter %[[#vec4_int_16]]
+; CHECK: %[[#ssign:]] = OpExtInst %[[#vec4_int_16]] %[[#op_ext_glsl]] SSign %[[#vec4_int_16_arg]]
+; CHECK: %[[#]] = OpSConvert %[[#vec4_int_32]] %[[#ssign]]
+  %elt.sign = call <4 x i32> @llvm.spv.sign.v4i16(<4 x i16> %a)
+  ret <4 x i32> %elt.sign
+}
+
+define noundef <4 x i32> @sign_i32_vector(<4 x i32> noundef %a) {
+entry:
+; CHECK: %[[#vec4_int_32_arg:]] = OpFunctionParameter %[[#vec4_int_32]]
+; CHECK: %[[#]] = OpExtInst %[[#vec4_int_32]] %[[#op_ext_glsl]] SSign %[[#vec4_int_32_arg]]
+  %elt.sign = call <4 x i32> @llvm.spv.sign.v4i32(<4 x i32> %a)
+  ret <4 x i32> %elt.sign
+}
+
+define noundef <4 x i32> @sign_i64_vector(<4 x i64> noundef %a) {
+entry:
+; CHECK: %[[#vec4_int_64_arg:]] = OpFunctionParameter %[[#vec4_int_64]]
+; CHECK: %[[#ssign:]] = OpExtInst %[[#vec4_int_64]] %[[#op_ext_glsl]] SSign %[[#vec4_int_64_arg]]
+; CHECK: %[[#]] = OpSConvert %[[#vec4_int_32]] %[[#ssign]]
+  %elt.sign = call <4 x i32> @llvm.spv.sign.v4i64(<4 x i64> %a)
+  ret <4 x i32> %elt.sign
+}
+
+declare i32 @llvm.spv.sign.f16(half)
+declare i32 @llvm.spv.sign.f32(float)
+declare i32 @llvm.spv.sign.f64(double)
+
+declare i32 @llvm.spv.sign.i16(i16)
+declare i32 @llvm.spv.sign.i32(i32)
+declare i32 @llvm.spv.sign.i64(i64)
+
+declare <4 x i32> @llvm.spv.sign.v4f16(<4 x half>)
+declare <4 x i32> @llvm.spv.sign.v4f32(<4 x float>)
+declare <4 x i32> @llvm.spv.sign.v4f64(<4 x double>)
+
+declare <4 x i32> @llvm.spv.sign.v4i16(<4 x i16>)
+declare <4 x i32> @llvm.spv.sign.v4i32(<4 x i32>)
+declare <4 x i32> @llvm.spv.sign.v4i64(<4 x i64>)

@llvmbot

llvmbot commented Aug 5, 2024

Copy link
Copy Markdown
Member

@llvm/pr-subscribers-backend-spir-v

Author: Tim Gymnich (tgymnich)

Changes

partially fixes #70078

Changes

  • Added int_spv_sign intrinsic in IntrinsicsSPIRV.td
  • Added lowering and map to int_spv_sign in SPIRVInstructionSelector.cpp`
  • Added SPIR-V backend test case in llvm/test/CodeGen/SPIRV/hlsl-intrinsics/sign.ll

Related PRs

Full diff: https://github.com/llvm/llvm-project/pull/101987.diff

3 Files Affected:

  • (modified) llvm/include/llvm/IR/IntrinsicsSPIRV.td (+1)
  • (modified) llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp (+52)
  • (added) llvm/test/CodeGen/SPIRV/hlsl-intrinsics/sign.ll (+143)
diff --git a/llvm/include/llvm/IR/IntrinsicsSPIRV.td b/llvm/include/llvm/IR/IntrinsicsSPIRV.td
index 3f77ef6bfcdbe..89eacf2ba70f9 100644
--- a/llvm/include/llvm/IR/IntrinsicsSPIRV.td
+++ b/llvm/include/llvm/IR/IntrinsicsSPIRV.td
@@ -65,4 +65,5 @@ let TargetPrefix = "spv" in {
     [IntrNoMem, IntrWillReturn] >;
   def int_spv_length : DefaultAttrsIntrinsic<[LLVMVectorElementType<0>], [llvm_anyfloat_ty]>;
   def int_spv_rsqrt : DefaultAttrsIntrinsic<[LLVMMatchType<0>], [llvm_anyfloat_ty]>;
+  def int_spv_sign : DefaultAttrsIntrinsic<[LLVMScalarOrSameVectorWidth<0, llvm_i32_ty>], [llvm_any_ty]>;
 }
diff --git a/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp b/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
index ed786bd33aa05..47e4b3dd40339 100644
--- a/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
+++ b/llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
@@ -28,6 +28,7 @@
 #include "llvm/CodeGen/MachineInstrBuilder.h"
 #include "llvm/CodeGen/MachineModuleInfoImpls.h"
 #include "llvm/CodeGen/MachineRegisterInfo.h"
+#include "llvm/CodeGen/Register.h"
 #include "llvm/CodeGen/TargetOpcodes.h"
 #include "llvm/IR/IntrinsicsSPIRV.h"
 #include "llvm/Support/Debug.h"
@@ -178,6 +179,9 @@ class SPIRVInstructionSelector : public InstructionSelector {
   bool selectRsqrt(Register ResVReg, const SPIRVType *ResType,
                    MachineInstr &I) const;
 
+  bool selectSign(Register ResVReg, const SPIRVType *ResType,
+                  MachineInstr &I) const;
+
   void renderImm32(MachineInstrBuilder &MIB, const MachineInstr &I,
                    int OpIdx) const;
   void renderFImm32(MachineInstrBuilder &MIB, const MachineInstr &I,
@@ -1366,6 +1370,52 @@ bool SPIRVInstructionSelector::selectRsqrt(Register ResVReg,
       .constrainAllUses(TII, TRI, RBI);
 }
 
+bool SPIRVInstructionSelector::selectSign(Register ResVReg,
+                                          const SPIRVType *ResType,
+                                          MachineInstr &I) const {
+  assert(I.getNumOperands() == 3);
+  assert(I.getOperand(2).isReg());
+  MachineBasicBlock &BB = *I.getParent();
+  Register InputRegister = I.getOperand(2).getReg();
+  SPIRVType *InputType = GR.getSPIRVTypeForVReg(InputRegister);
+  auto &DL = I.getDebugLoc();
+
+  if (!InputType)
+    report_fatal_error("Input Type could not be determined.");
+
+  bool IsFloatTy = GR.isScalarOrVectorOfType(InputRegister, SPIRV::OpTypeFloat);
+
+  unsigned SignBitWidth = GR.getScalarOrVectorBitWidth(InputType);
+  unsigned ResBitWidth = GR.getScalarOrVectorBitWidth(ResType);
+
+  bool NeedsConversion = IsFloatTy || SignBitWidth != ResBitWidth;
+
+  auto SignOpcode = IsFloatTy ? GL::FSign : GL::SSign;
+  Register SignReg = NeedsConversion
+                         ? MRI->createVirtualRegister(&SPIRV::IDRegClass)
+                         : ResVReg;
+
+  bool Result =
+      BuildMI(BB, I, DL, TII.get(SPIRV::OpExtInst))
+          .addDef(SignReg)
+          .addUse(GR.getSPIRVTypeID(InputType))
+          .addImm(static_cast<uint32_t>(SPIRV::InstructionSet::GLSL_std_450))
+          .addImm(SignOpcode)
+          .addUse(InputRegister)
+          .constrainAllUses(TII, TRI, RBI);
+
+  if (NeedsConversion) {
+    auto ConvertOpcode = IsFloatTy ? SPIRV::OpConvertFToS : SPIRV::OpSConvert;
+    Result |= BuildMI(*I.getParent(), I, DL, TII.get(ConvertOpcode))
+                  .addDef(ResVReg)
+                  .addUse(GR.getSPIRVTypeID(ResType))
+                  .addUse(SignReg)
+                  .constrainAllUses(TII, TRI, RBI);
+  }
+
+  return Result;
+}
+
 bool SPIRVInstructionSelector::selectBitreverse(Register ResVReg,
                                                 const SPIRVType *ResType,
                                                 MachineInstr &I) const {
@@ -2082,6 +2132,8 @@ bool SPIRVInstructionSelector::selectIntrinsic(Register ResVReg,
     return selectFrac(ResVReg, ResType, I);
   case Intrinsic::spv_rsqrt:
     return selectRsqrt(ResVReg, ResType, I);
+  case Intrinsic::spv_sign:
+    return selectSign(ResVReg, ResType, I);
   case Intrinsic::spv_lifetime_start:
   case Intrinsic::spv_lifetime_end: {
     unsigned Op = IID == Intrinsic::spv_lifetime_start ? SPIRV::OpLifetimeStart
diff --git a/llvm/test/CodeGen/SPIRV/hlsl-intrinsics/sign.ll b/llvm/test/CodeGen/SPIRV/hlsl-intrinsics/sign.ll
new file mode 100644
index 0000000000000..52a41c3d3ad64
--- /dev/null
+++ b/llvm/test/CodeGen/SPIRV/hlsl-intrinsics/sign.ll
@@ -0,0 +1,143 @@
+; RUN: llc -O0 -mtriple=spirv-unknown-unknown %s -o - | FileCheck %s
+; RUN: %if spirv-tools %{ llc -O0 -mtriple=spirv-unknown-unknown %s -o - -filetype=obj | spirv-val %}
+
+; CHECK-DAG: %[[#op_ext_glsl:]] = OpExtInstImport "GLSL.std.450"
+
+; CHECK-DAG: %[[#float_16:]] = OpTypeFloat 16
+; CHECK-DAG: %[[#float_32:]] = OpTypeFloat 32
+; CHECK-DAG: %[[#float_64:]] = OpTypeFloat 64
+
+; CHECK-DAG: %[[#int_16:]] = OpTypeInt 16
+; CHECK-DAG: %[[#int_32:]] = OpTypeInt 32
+; CHECK-DAG: %[[#int_64:]] = OpTypeInt 64
+
+; CHECK-DAG: %[[#vec4_float_16:]] = OpTypeVector %[[#float_16]] 4
+; CHECK-DAG: %[[#vec4_float_32:]] = OpTypeVector %[[#float_32]] 4
+; CHECK-DAG: %[[#vec4_float_64:]] = OpTypeVector %[[#float_64]] 4
+
+; CHECK-DAG: %[[#vec4_int_16:]] = OpTypeVector %[[#int_16]] 4
+; CHECK-DAG: %[[#vec4_int_32:]] = OpTypeVector %[[#int_32]] 4
+; CHECK-DAG: %[[#vec4_int_64:]] = OpTypeVector %[[#int_64]] 4
+
+
+define noundef i32 @sign_half(half noundef %a) {
+entry:
+; CHECK: %[[#float_16_arg:]] = OpFunctionParameter %[[#float_16]]
+; CHECK: %[[#fsign:]] = OpExtInst %[[#float_16]] %[[#op_ext_glsl]] FSign %[[#float_16_arg]]
+; CHECK: %[[#]] = OpConvertFToS %[[#int_32]] %[[#fsign]]
+  %elt.sign = call i32 @llvm.spv.sign.f16(half %a)
+  ret i32 %elt.sign
+}
+
+define noundef i32 @sign_float(float noundef %a) {
+entry:
+; CHECK: %[[#float_32_arg:]] = OpFunctionParameter %[[#float_32]]
+; CHECK: %[[#fsign:]] = OpExtInst %[[#float_32]] %[[#op_ext_glsl]] FSign %[[#float_32_arg]]
+; CHECK: %[[#]] = OpConvertFToS %[[#int_32]] %[[#fsign]]
+  %elt.sign = call i32 @llvm.spv.sign.f32(float %a)
+  ret i32 %elt.sign
+}
+
+define noundef i32 @sign_double(double noundef %a) {
+entry:
+; CHECK: %[[#float_64_arg:]] = OpFunctionParameter %[[#float_64]]
+; CHECK: %[[#fsign:]] = OpExtInst %[[#float_64]] %[[#op_ext_glsl]] FSign %[[#float_64_arg]]
+; CHECK: %[[#]] = OpConvertFToS %[[#int_32]] %[[#fsign]]
+  %elt.sign = call i32 @llvm.spv.sign.f64(double %a)
+  ret i32 %elt.sign
+}
+
+define noundef i32 @sign_i16(i16 noundef %a) {
+entry:
+; CHECK: %[[#int_16_arg:]] = OpFunctionParameter %[[#int_16]]
+; CHECK: %[[#ssign:]] = OpExtInst %[[#int_16]] %[[#op_ext_glsl]] SSign %[[#int_16_arg]]
+; CHECK: %[[#]] = OpSConvert %[[#int_32]] %[[#ssign]]
+  %elt.sign = call i32 @llvm.spv.sign.i16(i16 %a)
+  ret i32 %elt.sign
+}
+
+define noundef i32 @sign_i32(i32 noundef %a) {
+entry:
+; CHECK: %[[#int_32_arg:]] = OpFunctionParameter %[[#int_32]]
+; CHECK: %[[#]] = OpExtInst %[[#int_32]] %[[#op_ext_glsl]] SSign %[[#int_32_arg]]
+  %elt.sign = call i32 @llvm.spv.sign.i32(i32 %a)
+  ret i32 %elt.sign
+}
+
+define noundef i32 @sign_i64(i64 noundef %a) {
+entry:
+; CHECK: %[[#int_64_arg:]] = OpFunctionParameter %[[#int_64]]
+; CHECK: %[[#ssign:]] = OpExtInst %[[#int_64]] %[[#op_ext_glsl]] SSign %[[#int_64_arg]]
+; CHECK: %[[#]] = OpSConvert %[[#int_32]] %[[#ssign]]
+  %elt.sign = call i32 @llvm.spv.sign.i64(i64 %a)
+  ret i32 %elt.sign
+}
+
+define noundef <4 x i32> @sign_half_vector(<4 x half> noundef %a) {
+entry:
+; CHECK: %[[#vec4_float_16_arg:]] = OpFunctionParameter %[[#vec4_float_16]]
+; CHECK: %[[#fsign:]] = OpExtInst %[[#vec4_float_16]] %[[#op_ext_glsl]] FSign %[[#vec4_float_16_arg]]
+; CHECK: %[[#]] = OpConvertFToS %[[#vec4_int_32]] %[[#fsign]]
+  %elt.sign = call <4 x i32> @llvm.spv.sign.v4f16(<4 x half> %a)
+  ret <4 x i32> %elt.sign
+}
+
+define noundef <4 x i32> @sign_float_vector(<4 x float> noundef %a) {
+entry:
+; CHECK: %[[#vec4_float_32_arg:]] = OpFunctionParameter %[[#vec4_float_32]]
+; CHECK: %[[#fsign:]] = OpExtInst %[[#vec4_float_32]] %[[#op_ext_glsl]] FSign %[[#vec4_float_32_arg]]
+; CHECK: %[[#]] = OpConvertFToS %[[#vec4_int_32]] %[[#fsign]]
+  %elt.sign = call <4 x i32> @llvm.spv.sign.v4f32(<4 x float> %a)
+  ret <4 x i32> %elt.sign
+}
+
+define noundef <4 x i32> @sign_double_vector(<4 x double> noundef %a) {
+entry:
+; CHECK: %[[#vec4_float_64_arg:]] = OpFunctionParameter %[[#vec4_float_64]]
+; CHECK: %[[#fsign:]] = OpExtInst %[[#vec4_float_64]] %[[#op_ext_glsl]] FSign %[[#vec4_float_64_arg]]
+; CHECK: %[[#]] = OpConvertFToS %[[#vec4_int_32]] %[[#fsign]]
+  %elt.sign = call <4 x i32> @llvm.spv.sign.v4f64(<4 x double> %a)
+  ret <4 x i32> %elt.sign
+}
+
+define noundef <4 x i32> @sign_i16_vector(<4 x i16> noundef %a) {
+entry:
+; CHECK: %[[#vec4_int_16_arg:]] = OpFunctionParameter %[[#vec4_int_16]]
+; CHECK: %[[#ssign:]] = OpExtInst %[[#vec4_int_16]] %[[#op_ext_glsl]] SSign %[[#vec4_int_16_arg]]
+; CHECK: %[[#]] = OpSConvert %[[#vec4_int_32]] %[[#ssign]]
+  %elt.sign = call <4 x i32> @llvm.spv.sign.v4i16(<4 x i16> %a)
+  ret <4 x i32> %elt.sign
+}
+
+define noundef <4 x i32> @sign_i32_vector(<4 x i32> noundef %a) {
+entry:
+; CHECK: %[[#vec4_int_32_arg:]] = OpFunctionParameter %[[#vec4_int_32]]
+; CHECK: %[[#]] = OpExtInst %[[#vec4_int_32]] %[[#op_ext_glsl]] SSign %[[#vec4_int_32_arg]]
+  %elt.sign = call <4 x i32> @llvm.spv.sign.v4i32(<4 x i32> %a)
+  ret <4 x i32> %elt.sign
+}
+
+define noundef <4 x i32> @sign_i64_vector(<4 x i64> noundef %a) {
+entry:
+; CHECK: %[[#vec4_int_64_arg:]] = OpFunctionParameter %[[#vec4_int_64]]
+; CHECK: %[[#ssign:]] = OpExtInst %[[#vec4_int_64]] %[[#op_ext_glsl]] SSign %[[#vec4_int_64_arg]]
+; CHECK: %[[#]] = OpSConvert %[[#vec4_int_32]] %[[#ssign]]
+  %elt.sign = call <4 x i32> @llvm.spv.sign.v4i64(<4 x i64> %a)
+  ret <4 x i32> %elt.sign
+}
+
+declare i32 @llvm.spv.sign.f16(half)
+declare i32 @llvm.spv.sign.f32(float)
+declare i32 @llvm.spv.sign.f64(double)
+
+declare i32 @llvm.spv.sign.i16(i16)
+declare i32 @llvm.spv.sign.i32(i32)
+declare i32 @llvm.spv.sign.i64(i64)
+
+declare <4 x i32> @llvm.spv.sign.v4f16(<4 x half>)
+declare <4 x i32> @llvm.spv.sign.v4f32(<4 x float>)
+declare <4 x i32> @llvm.spv.sign.v4f64(<4 x double>)
+
+declare <4 x i32> @llvm.spv.sign.v4i16(<4 x i16>)
+declare <4 x i32> @llvm.spv.sign.v4i32(<4 x i32>)
+declare <4 x i32> @llvm.spv.sign.v4i64(<4 x i64>)

farzonl
farzonl reviewed Aug 11, 2024
View reviewed changes
Comment thread llvm/lib/Target/SPIRV/SPIRVInstructionSelector.cpp
@tgymnich
tgymnich force-pushed the hlsl-sign-1 branch 2 times, most recently from cac6f19 to 87eb441 Compare August 29, 2024 18:18
farzonl pushed a commit that referenced this pull request Sep 5, 2024
@tgymnich
[DXIL] Add sign intrinsic part 2 (#101988)
5edede2 
makes progress on #70078

### Changes
- Added `int_dx_sign` intrinsic in `IntrinsicsDirectX.td`
- Added expansion for `int_dx_sign in `DXILIntrinsicExpansion.cpp`
- Added DXIL backend test case

### Related PRs
- #101987
- #101989
farzonl
farzonl approved these changes Sep 9, 2024
View reviewed changes

@farzonl farzonl left a comment

Copy link
Copy Markdown
Member

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

LGTM

pow2clk
pow2clk approved these changes Sep 9, 2024
View reviewed changes

@pow2clk pow2clk left a comment

Copy link
Copy Markdown
Contributor

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looks good!

@farzonl
farzonl merged commit a9a5a18 into llvm:main Sep 9, 2024
@tgymnich
tgymnich deleted the hlsl-sign-1 branch September 9, 2024 22:02
farzonl pushed a commit that referenced this pull request Sep 10, 2024
@tgymnich
[clang][HLSL] Add sign intrinsic part 3 (#101989)
dce5039 
partially fixes #70078

### Changes
- Implemented `sign` clang builtin
- Linked `sign` clang builtin with `hlsl_intrinsics.h`
- Added sema checks for `sign` to `CheckHLSLBuiltinFunctionCall` in
`SemaChecking.cpp`
- Add codegen for `sign` to `EmitHLSLBuiltinExpr` in `CGBuiltin.cpp`
- Add codegen tests to `clang/test/CodeGenHLSL/builtins/sign.hlsl`
- Add sema tests to `clang/test/SemaHLSL/BuiltIns/sign-errors.hlsl`

### Related PRs
- #101987
- #101988

### Discussion
- Should there be a `usign` intrinsic that handles the unsigned cases?
@tgymnich tgymnich mentioned this pull request Sep 12, 2024
Merged
farzonl pushed a commit that referenced this pull request Oct 10, 2024
@tgymnich
[clang][HLSL] Add sign intrinsic part 4 (#108396)
99608f1 
- Add handling for unsigned integers to hlsl_elementwise_sign
- Use `select` instead of adding dx and spirv intrinsics for unsigned
integers as [discussed previously
](#101988 (comment))

fixes #70078

### Related PRs
- #101987
- #101988
- #101989

cc @farzonl @pow2clk @bob80905 @bogner @llvm-beanz
DanielCChen pushed a commit to DanielCChen/llvm-project that referenced this pull request Oct 16, 2024
@tgymnich @DanielCChen
[clang][HLSL] Add sign intrinsic part 4 (llvm#108396)
6121292 
- Add handling for unsigned integers to hlsl_elementwise_sign
- Use `select` instead of adding dx and spirv intrinsics for unsigned
integers as [discussed previously
](llvm#101988 (comment))

fixes llvm#70078

### Related PRs
- llvm#101987
- llvm#101988
- llvm#101989

cc @farzonl @pow2clk @bob80905 @bogner @llvm-beanz
OpenHarmonySCM-noreply pushed a commit to openharmony/third_party_llvm-project that referenced this pull request Jul 19, 2026
@tgymnich
[DXIL] Add sign intrinsic part 2 (#101988)
a947aac 
makes progress on #70078

### Changes
- Added `int_dx_sign` intrinsic in `IntrinsicsDirectX.td`
- Added expansion for `int_dx_sign in `DXILIntrinsicExpansion.cpp`
- Added DXIL backend test case

### Related PRs
- llvm/llvm-project#101987
- llvm/llvm-project#101989
OpenHarmonySCM-noreply pushed a commit to openharmony/third_party_llvm-project that referenced this pull request Jul 19, 2026
@tgymnich
[clang][HLSL] Add sign intrinsic part 3 (#101989)
32963d9 
partially fixes #70078

### Changes
- Implemented `sign` clang builtin
- Linked `sign` clang builtin with `hlsl_intrinsics.h`
- Added sema checks for `sign` to `CheckHLSLBuiltinFunctionCall` in
`SemaChecking.cpp`
- Add codegen for `sign` to `EmitHLSLBuiltinExpr` in `CGBuiltin.cpp`
- Add codegen tests to `clang/test/CodeGenHLSL/builtins/sign.hlsl`
- Add sema tests to `clang/test/SemaHLSL/BuiltIns/sign-errors.hlsl`

### Related PRs
- llvm/llvm-project#101987
- llvm/llvm-project#101988

### Discussion
- Should there be a `usign` intrinsic that handles the unsigned cases?
OpenHarmonySCM-noreply pushed a commit to openharmony/third_party_llvm-project that referenced this pull request Jul 19, 2026
@tgymnich
[clang][HLSL] Add sign intrinsic part 4 (#108396)
8f6985f 
- Add handling for unsigned integers to hlsl_elementwise_sign
- Use `select` instead of adding dx and spirv intrinsics for unsigned
integers as [discussed previously
](llvm/llvm-project#101988 (comment))

fixes #70078

### Related PRs
- llvm/llvm-project#101987
- llvm/llvm-project#101988
- llvm/llvm-project#101989

cc @farzonl @pow2clk @bob80905 @bogner @llvm-beanz
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Reviewers

@farzonl farzonl farzonl approved these changes
+1 more reviewer
@pow2clk pow2clk pow2clk approved these changes
Reviewers whose approvals may not affect merge requirements

Assignees

No one assigned

Labels

backend:SPIR-V llvm:ir

Projects

None yet

Milestone

No milestone

Development

Successfully merging this pull request may close these issues.

Implement the sign HLSL Function

4 participants

@tgymnich @llvmbot @farzonl @pow2clk