[CIR][CIRGen][Builtin][X86] Masked compress Intrinsics

Bhuvan1527 · Bhuvan1527 · commit 2944b9e6bb6f · 2025-11-28T14:45:51.000+05:30
This pr is related to the issue #167765 Added the support Masked compress builtin in CIR codeGen
diff --git a/clang/lib/CIR/CodeGen/CIRGenBuiltinX86.cpp b/clang/lib/CIR/CodeGen/CIRGenBuiltinX86.cpp
@@ -84,13 +84,10 @@ static mlir::Value getMaskVecValue(CIRGenBuilderTy &builder, mlir::Location loc,
   }
   return maskVec;
 }
-static mlir::Value emitX86CompressExpand(CIRGenFunction &cgf, const CallExpr *expr,ArrayRef<mlir::Value> ops, bool IsCompress, const std::string &ID){
-  auto ResultTy = cast<cir::VectorType>(ops[1].getType());
-  mlir::Value MaskValue = getMaskVecValue(cgf, expr, ops[2],  cast<cir::VectorType>(ResultTy).getSize());
-  llvm::SmallVector<mlir::Value, 4> op{ops[0], ops[1], MaskValue};
-  
-  return emitIntrinsicCallOp(cgf,expr, ID, ResultTy, op);
-
+static mlir::Value emitX86CompressExpand(CIRGenBuilderTy &builder, mlir::Location loc, mlir::Value source, mlir::Value mask, mlir::Value inputVector, const std::string &id){
+  auto ResultTy = cast<cir::VectorType>(mask.getType());
+  mlir::Value MaskValue = getMaskVecValue(builder, loc, inputVector,  cast<cir::VectorType>(ResultTy).getSize());
+  return emitIntrinsicCallOp(builder, loc, id, ResultTy, source, mask, MaskValue);
 }
 
 mlir::Value CIRGenFunction::emitX86BuiltinExpr(unsigned builtinID,
@@ -429,6 +426,10 @@ mlir::Value CIRGenFunction::emitX86BuiltinExpr(unsigned builtinID,
   case X86::BI__builtin_ia32_compressstoreqi128_mask:
   case X86::BI__builtin_ia32_compressstoreqi256_mask:
   case X86::BI__builtin_ia32_compressstoreqi512_mask:
+    cgm.errorNYI(expr->getSourceRange(),
+                 std::string("unimplemented X86 builtin call: ") +
+                     getContext().BuiltinInfo.getName(builtinID));
+    return {};
   case X86::BI__builtin_ia32_expanddf128_mask:
   case X86::BI__builtin_ia32_expanddf256_mask:
   case X86::BI__builtin_ia32_expanddf512_mask:
@@ -446,7 +447,10 @@ mlir::Value CIRGenFunction::emitX86BuiltinExpr(unsigned builtinID,
   case X86::BI__builtin_ia32_expandhi512_mask:
   case X86::BI__builtin_ia32_expandqi128_mask:
   case X86::BI__builtin_ia32_expandqi256_mask:
-  case X86::BI__builtin_ia32_expandqi512_mask:
+  case X86::BI__builtin_ia32_expandqi512_mask:{
+    mlir::Location loc = getLoc(expr->getExprLoc());
+    return emitX86CompressExpand(builder, loc, ops[0], ops[1], ops[2], "x86_avx512_mask_expand");
+  }
   case X86::BI__builtin_ia32_compressdf128_mask:
   case X86::BI__builtin_ia32_compressdf256_mask:
   case X86::BI__builtin_ia32_compressdf512_mask:
@@ -465,7 +469,8 @@ mlir::Value CIRGenFunction::emitX86BuiltinExpr(unsigned builtinID,
   case X86::BI__builtin_ia32_compressqi128_mask:
   case X86::BI__builtin_ia32_compressqi256_mask:
   case X86::BI__builtin_ia32_compressqi512_mask:{
-    return emitX86CompressExpand(*this, expr, ops, true, "x86_avx512_mask_compress");
+    mlir::Location loc = getLoc(expr->getExprLoc());
+    return emitX86CompressExpand(builder, loc, ops[0], ops[1], ops[2], "x86_avx512_mask_compress");
   }
   case X86::BI__builtin_ia32_gather3div2df:
   case X86::BI__builtin_ia32_gather3div2di:
@@ -791,16 +796,7 @@ mlir::Value CIRGenFunction::emitX86BuiltinExpr(unsigned builtinID,
   case X86::BI__builtin_ia32_sqrtsh_round_mask:
   case X86::BI__builtin_ia32_sqrtsd_round_mask:
   case X86::BI__builtin_ia32_sqrtss_round_mask:
-  case X86::BI__builtin_ia32_sqrtpd256:
-  case X86::BI__builtin_ia32_sqrtpd:
-  case X86::BI__builtin_ia32_sqrtps256:
-  case X86::BI__builtin_ia32_sqrtps:
-  case X86::BI__builtin_ia32_sqrtph256:
-  case X86::BI__builtin_ia32_sqrtph:
   case X86::BI__builtin_ia32_sqrtph512:
-  case X86::BI__builtin_ia32_vsqrtbf16256:
-  case X86::BI__builtin_ia32_vsqrtbf16:
-  case X86::BI__builtin_ia32_vsqrtbf16512:
   case X86::BI__builtin_ia32_sqrtps512:
   case X86::BI__builtin_ia32_sqrtpd512:
   case X86::BI__builtin_ia32_pmuludq128:
@@ -953,7 +949,6 @@ mlir::Value CIRGenFunction::emitX86BuiltinExpr(unsigned builtinID,
   case X86::BI__builtin_ia32_vcvtph2ps256_mask:
   case X86::BI__builtin_ia32_vcvtph2ps512_mask:
   case X86::BI__builtin_ia32_cvtneps2bf16_128_mask:
-  case X86::BI__builtin_ia32_cvtsbf162ss_32:
   case X86::BI__builtin_ia32_cvtneps2bf16_256_mask:
   case X86::BI__builtin_ia32_cvtneps2bf16_512_mask:
   case X86::BI__cpuid:
diff --git a/clang/test/CIR/CodeGenBuiltins/X86/avx512vl-builtins.c b/clang/test/CIR/CodeGenBuiltins/X86/avx512vl-builtins.c
@@ -0,0 +1,158 @@
+// RUN: %clang_cc1 -x c -flax-vector-conversions=none -ffreestanding %s -triple=x86_64-unknown-linux -target-feature +avx512vlvbmi2 -fclangir -emit-cir -o %t.cir -Wall -Werror -Wsign-conversion
+// RUN: FileCheck --check-prefix=CIR --input-file=%t.cir %s
+// RUN: %clang_cc1 -x c -flax-vector-conversions=none -ffreestanding %s -triple=x86_64-unknown-linux -target-feature +avx512vlvbmi2 -fclangir -emit-llvm -o %t.ll -Wall -Werror -Wsign-conversion
+// RUN: FileCheck --check-prefixes=LLVM --input-file=%t.ll %s
+
+// RUN: %clang_cc1 -x c++ -flax-vector-conversions=none -ffreestanding %s -triple=x86_64-unknown-linux -target-feature +avx512vlvbmi2 -fclangir -emit-cir -o %t.cir -Wall -Werror -Wsign-conversion
+// RUN: FileCheck --check-prefix=CIR --input-file=%t.cir %s
+// RUN: %clang_cc1 -x c++ -flax-vector-conversions=none -ffreestanding %s -triple=x86_64-unknown-linux -target-feature +avx512vlvbmi2 -fclangir -emit-llvm -o %t.ll -Wall -Werror -Wsign-conversion
+// RUN: FileCheck --check-prefixes=LLVM --input-file=%t.ll %s
+
+#include <immintrin.h>
+
+
+__m128d test_mm_mask_expand_pd(__m128d __W, __mmask8 __U, __m128d __A) {
+
+  return _mm_mask_expand_pd(__W,__U,__A); 
+}
+__m128d test_mm_maskz_expand_pd(__mmask8 __U, __m128d __A) {
+
+  return _mm_maskz_expand_pd(__U,__A); 
+}
+__m256d test_mm256_mask_expand_pd(__m256d __W, __mmask8 __U, __m256d __A) {
+
+  return _mm256_mask_expand_pd(__W,__U,__A); 
+}
+__m256d test_mm256_maskz_expand_pd(__mmask8 __U, __m256d __A) {
+
+  return _mm256_maskz_expand_pd(__U,__A); 
+}
+__m128i test_mm_mask_expand_epi64(__m128i __W, __mmask8 __U, __m128i __A) {
+
+  return _mm_mask_expand_epi64(__W,__U,__A); 
+}
+__m128i test_mm_maskz_expand_epi64(__mmask8 __U, __m128i __A) {
+
+  return _mm_maskz_expand_epi64(__U,__A); 
+}
+__m256i test_mm256_mask_expand_epi64(__m256i __W, __mmask8 __U, __m256i __A) {
+
+  return _mm256_mask_expand_epi64(__W,__U,__A); 
+}
+__m256i test_mm256_maskz_expand_epi64(__mmask8 __U, __m256i __A) {
+
+  return _mm256_maskz_expand_epi64(__U,__A); 
+}
+
+__m128 test_mm_mask_expand_ps(__m128 __W, __mmask8 __U, __m128 __A) {
+
+  return _mm_mask_expand_ps(__W,__U,__A); 
+}
+__m128 test_mm_maskz_expand_ps(__mmask8 __U, __m128 __A) {
+
+  return _mm_maskz_expand_ps(__U,__A); 
+}
+__m256 test_mm256_mask_expand_ps(__m256 __W, __mmask8 __U, __m256 __A) {
+
+  return _mm256_mask_expand_ps(__W,__U,__A); 
+}
+__m256 test_mm256_maskz_expand_ps(__mmask8 __U, __m256 __A) {
+
+  return _mm256_maskz_expand_ps(__U,__A); 
+}
+__m128i test_mm_mask_expand_epi32(__m128i __W, __mmask8 __U, __m128i __A) {
+
+  return _mm_mask_expand_epi32(__W,__U,__A); 
+}
+__m128i test_mm_maskz_expand_epi32(__mmask8 __U, __m128i __A) {
+
+  return _mm_maskz_expand_epi32(__U,__A); 
+}
+__m256i test_mm256_mask_expand_epi32(__m256i __W, __mmask8 __U, __m256i __A) {
+
+  return _mm256_mask_expand_epi32(__W,__U,__A); 
+}
+__m256i test_mm256_maskz_expand_epi32(__mmask8 __U, __m256i __A) {
+
+  return _mm256_maskz_expand_epi32(__U,__A); 
+}
+
+__m128d test_mm_mask_compress_pd(__m128d __W, __mmask8 __U, __m128d __A) {
+
+  return _mm_mask_compress_pd(__W,__U,__A); 
+}
+
+__m128d test_mm_maskz_compress_pd(__mmask8 __U, __m128d __A) {
+
+  return _mm_maskz_compress_pd(__U,__A); 
+}
+
+__m256d test_mm256_mask_compress_pd(__m256d __W, __mmask8 __U, __m256d __A) {
+
+  return _mm256_mask_compress_pd(__W,__U,__A); 
+}
+
+__m256d test_mm256_maskz_compress_pd(__mmask8 __U, __m256d __A) {
+
+  return _mm256_maskz_compress_pd(__U,__A); 
+}
+
+__m128i test_mm_mask_compress_epi64(__m128i __W, __mmask8 __U, __m128i __A) {
+
+  return _mm_mask_compress_epi64(__W,__U,__A); 
+}
+
+__m128i test_mm_maskz_compress_epi64(__mmask8 __U, __m128i __A) {
+
+  return _mm_maskz_compress_epi64(__U,__A); 
+}
+
+__m256i test_mm256_mask_compress_epi64(__m256i __W, __mmask8 __U, __m256i __A) {
+
+  return _mm256_mask_compress_epi64(__W,__U,__A); 
+}
+
+__m256i test_mm256_maskz_compress_epi64(__mmask8 __U, __m256i __A) {
+
+  return _mm256_maskz_compress_epi64(__U,__A); 
+}
+
+__m128 test_mm_mask_compress_ps(__m128 __W, __mmask8 __U, __m128 __A) {
+
+  return _mm_mask_compress_ps(__W,__U,__A); 
+}
+
+__m128 test_mm_maskz_compress_ps(__mmask8 __U, __m128 __A) {
+
+  return _mm_maskz_compress_ps(__U,__A); 
+}
+
+__m256 test_mm256_mask_compress_ps(__m256 __W, __mmask8 __U, __m256 __A) {
+
+  return _mm256_mask_compress_ps(__W,__U,__A); 
+}
+
+__m256 test_mm256_maskz_compress_ps(__mmask8 __U, __m256 __A) {
+
+  return _mm256_maskz_compress_ps(__U,__A); 
+}
+
+__m128i test_mm_mask_compress_epi32(__m128i __W, __mmask8 __U, __m128i __A) {
+
+  return _mm_mask_compress_epi32(__W,__U,__A); 
+}
+
+__m128i test_mm_maskz_compress_epi32(__mmask8 __U, __m128i __A) {
+
+  return _mm_maskz_compress_epi32(__U,__A); 
+}
+
+__m256i test_mm256_mask_compress_epi32(__m256i __W, __mmask8 __U, __m256i __A) {
+
+  return _mm256_mask_compress_epi32(__W,__U,__A); 
+}
+
+__m256i test_mm256_maskz_compress_epi32(__mmask8 __U, __m256i __A) {
+
+  return _mm256_maskz_compress_epi32(__U,__A); 
+}
diff --git a/clang/test/CIR/CodeGenBuiltins/X86/avx512vlvbmi2-builtins.c b/clang/test/CIR/CodeGenBuiltins/X86/avx512vlvbmi2-builtins.c
@@ -0,0 +1,53 @@
+
+// RUN: %clang_cc1 -x c -flax-vector-conversions=none -ffreestanding %s -triple=x86_64-unknown-linux -target-feature +avx512vlvbmi2 -fclangir -emit-cir -o %t.cir -Wall -Werror -Wsign-conversion
+// RUN: FileCheck --check-prefix=CIR --input-file=%t.cir %s
+// RUN: %clang_cc1 -x c -flax-vector-conversions=none -ffreestanding %s -triple=x86_64-unknown-linux -target-feature +avx512vlvbmi2 -fclangir -emit-llvm -o %t.ll -Wall -Werror -Wsign-conversion
+// RUN: FileCheck --check-prefixes=LLVM --input-file=%t.ll %s
+
+// RUN: %clang_cc1 -x c++ -flax-vector-conversions=none -ffreestanding %s -triple=x86_64-unknown-linux -target-feature +avx512vlvbmi2 -fclangir -emit-cir -o %t.cir -Wall -Werror -Wsign-conversion
+// RUN: FileCheck --check-prefix=CIR --input-file=%t.cir %s
+// RUN: %clang_cc1 -x c++ -flax-vector-conversions=none -ffreestanding %s -triple=x86_64-unknown-linux -target-feature +avx512vlvbmi2 -fclangir -emit-llvm -o %t.ll -Wall -Werror -Wsign-conversion
+// RUN: FileCheck --check-prefixes=LLVM --input-file=%t.ll %s
+
+#include <immintrin.h>
+
+
+__m128i test_mm_mask_compress_epi16(__m128i __S, __mmask8 __U, __m128i __D) {
+
+  return _mm_mask_compress_epi16(__S, __U, __D);
+}
+
+__m128i test_mm_maskz_compress_epi16(__mmask8 __U, __m128i __D) {
+
+  return _mm_maskz_compress_epi16(__U, __D);
+}
+
+__m128i test_mm_mask_compress_epi8(__m128i __S, __mmask16 __U, __m128i __D) {
+
+  return _mm_mask_compress_epi8(__S, __U, __D);
+}
+
+__m128i test_mm_maskz_compress_epi8(__mmask16 __U, __m128i __D) {
+
+  return _mm_maskz_compress_epi8(__U, __D);
+}
+
+__m128i test_mm_mask_expand_epi16(__m128i __S, __mmask8 __U, __m128i __D) {
+
+  return _mm_mask_expand_epi16(__S, __U, __D);
+}
+
+__m128i test_mm_maskz_expand_epi16(__mmask8 __U, __m128i __D) {
+
+  return _mm_maskz_expand_epi16(__U, __D);
+}
+
+__m128i test_mm_mask_expand_epi8(__m128i __S, __mmask16 __U, __m128i __D) {
+
+  return _mm_mask_expand_epi8(__S, __U, __D);
+}
+
+__m128i test_mm_maskz_expand_epi8(__mmask16 __U, __m128i __D) {
+
+  return _mm_maskz_expand_epi8(__U, __D);
+}
