Index: include/llvm/IR/IntrinsicsX86.td =================================================================== --- include/llvm/IR/IntrinsicsX86.td +++ include/llvm/IR/IntrinsicsX86.td @@ -3753,51 +3753,480 @@ // Vector convert let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.". - def int_x86_avx512_mask_cvttps2dq_512: GCCBuiltin<"__builtin_ia32_cvttps2dq512_mask">, - Intrinsic<[llvm_v16i32_ty], [llvm_v16f32_ty, llvm_v16i32_ty, - llvm_i16_ty, llvm_i32_ty], [IntrNoMem]>; - def int_x86_avx512_mask_cvttps2udq_512: GCCBuiltin<"__builtin_ia32_cvttps2udq512_mask">, - Intrinsic<[llvm_v16i32_ty], [llvm_v16f32_ty, llvm_v16i32_ty, - llvm_i16_ty, llvm_i32_ty], [IntrNoMem]>; - def int_x86_avx512_mask_cvttpd2dq_512: GCCBuiltin<"__builtin_ia32_cvttpd2dq512_mask">, - Intrinsic<[llvm_v8i32_ty], [llvm_v8f64_ty, llvm_v8i32_ty, - llvm_i8_ty, llvm_i32_ty], [IntrNoMem]>; - def int_x86_avx512_mask_cvttpd2udq_512: GCCBuiltin<"__builtin_ia32_cvttpd2udq512_mask">, - Intrinsic<[llvm_v8i32_ty], [llvm_v8f64_ty, llvm_v8i32_ty, - llvm_i8_ty, llvm_i32_ty], [IntrNoMem]>; + def int_x86_avx512_mask_cvtdq2pd_128 : + GCCBuiltin<"__builtin_ia32_cvtdq2pd128_mask">, + Intrinsic<[llvm_v2f64_ty], + [llvm_v4i32_ty, llvm_v2f64_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtdq2pd_256 : + GCCBuiltin<"__builtin_ia32_cvtdq2pd256_mask">, + Intrinsic<[llvm_v4f64_ty], + [llvm_v4i32_ty, llvm_v4f64_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtdq2pd_512 : + GCCBuiltin<"__builtin_ia32_cvtdq2pd512_mask">, + Intrinsic<[llvm_v8f64_ty], + [llvm_v8i32_ty, llvm_v8f64_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtdq2ps_128 : + GCCBuiltin<"__builtin_ia32_cvtdq2ps128_mask">, + Intrinsic<[llvm_v4f32_ty], + [llvm_v4i32_ty, llvm_v4f32_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtdq2ps_256 : + GCCBuiltin<"__builtin_ia32_cvtdq2ps256_mask">, + Intrinsic<[llvm_v8f32_ty], + [llvm_v8i32_ty, llvm_v8f32_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtdq2ps_512 : + GCCBuiltin<"__builtin_ia32_cvtdq2ps512_mask">, + Intrinsic<[llvm_v16f32_ty], + [llvm_v16i32_ty, llvm_v16f32_ty, llvm_i16_ty, llvm_i32_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtpd2dq_128 : + GCCBuiltin<"__builtin_ia32_cvtpd2dq128_mask">, + Intrinsic<[llvm_v4i32_ty], + [llvm_v2f64_ty, llvm_v4i32_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtpd2dq_256 : + GCCBuiltin<"__builtin_ia32_cvtpd2dq256_mask">, + Intrinsic<[llvm_v4i32_ty], + [llvm_v4f64_ty, llvm_v4i32_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtpd2dq_512 : + GCCBuiltin<"__builtin_ia32_cvtpd2dq512_mask">, + Intrinsic<[llvm_v8i32_ty], + [llvm_v8f64_ty, llvm_v8i32_ty, llvm_i8_ty, llvm_i32_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtpd2ps_256 : + GCCBuiltin<"__builtin_ia32_cvtpd2ps256_mask">, + Intrinsic<[llvm_v4f32_ty], + [llvm_v4f64_ty, llvm_v4f32_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtpd2ps_512 : + GCCBuiltin<"__builtin_ia32_cvtpd2ps512_mask">, + Intrinsic<[llvm_v8f32_ty], + [llvm_v8f64_ty, llvm_v8f32_ty, llvm_i8_ty, llvm_i32_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtpd2ps : + GCCBuiltin<"__builtin_ia32_cvtpd2ps_mask">, + Intrinsic<[llvm_v4f32_ty], + [llvm_v2f64_ty, llvm_v4f32_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtpd2qq_128 : + GCCBuiltin<"__builtin_ia32_cvtpd2qq128_mask">, + Intrinsic<[llvm_v2i64_ty], + [llvm_v2f64_ty, llvm_v2i64_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtpd2qq_256 : + GCCBuiltin<"__builtin_ia32_cvtpd2qq256_mask">, + Intrinsic<[llvm_v4i64_ty], + [llvm_v4f64_ty, llvm_v4i64_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtpd2qq_512 : + GCCBuiltin<"__builtin_ia32_cvtpd2qq512_mask">, + Intrinsic<[llvm_v8i64_ty], + [llvm_v8f64_ty, llvm_v8i64_ty, llvm_i8_ty, llvm_i32_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtpd2udq_128 : + GCCBuiltin<"__builtin_ia32_cvtpd2udq128_mask">, + Intrinsic<[llvm_v4i32_ty], + [llvm_v2f64_ty, llvm_v4i32_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtpd2udq_256 : + GCCBuiltin<"__builtin_ia32_cvtpd2udq256_mask">, + Intrinsic<[llvm_v4i32_ty], + [llvm_v4f64_ty, llvm_v4i32_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtpd2udq_512 : + GCCBuiltin<"__builtin_ia32_cvtpd2udq512_mask">, + Intrinsic<[llvm_v8i32_ty], + [llvm_v8f64_ty, llvm_v8i32_ty, llvm_i8_ty, llvm_i32_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtpd2uqq_128 : + GCCBuiltin<"__builtin_ia32_cvtpd2uqq128_mask">, + Intrinsic<[llvm_v2i64_ty], + [llvm_v2f64_ty, llvm_v2i64_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtpd2uqq_256 : + GCCBuiltin<"__builtin_ia32_cvtpd2uqq256_mask">, + Intrinsic<[llvm_v4i64_ty], + [llvm_v4f64_ty, llvm_v4i64_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtpd2uqq_512 : + GCCBuiltin<"__builtin_ia32_cvtpd2uqq512_mask">, + Intrinsic<[llvm_v8i64_ty], + [llvm_v8f64_ty, llvm_v8i64_ty, llvm_i8_ty, llvm_i32_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtps2dq_128 : + GCCBuiltin<"__builtin_ia32_cvtps2dq128_mask">, + Intrinsic<[llvm_v4i32_ty], + [llvm_v4f32_ty, llvm_v4i32_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtps2dq_256 : + GCCBuiltin<"__builtin_ia32_cvtps2dq256_mask">, + Intrinsic<[llvm_v8i32_ty], + [llvm_v8f32_ty, llvm_v8i32_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtps2dq_512 : + GCCBuiltin<"__builtin_ia32_cvtps2dq512_mask">, + Intrinsic<[llvm_v16i32_ty], + [llvm_v16f32_ty, llvm_v16i32_ty, llvm_i16_ty, llvm_i32_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtps2pd_128 : + GCCBuiltin<"__builtin_ia32_cvtps2pd128_mask">, + Intrinsic<[llvm_v2f64_ty], + [llvm_v4f32_ty, llvm_v2f64_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtps2pd_256 : + GCCBuiltin<"__builtin_ia32_cvtps2pd256_mask">, + Intrinsic<[llvm_v4f64_ty], + [llvm_v4f32_ty, llvm_v4f64_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtps2pd_512 : + GCCBuiltin<"__builtin_ia32_cvtps2pd512_mask">, + Intrinsic<[llvm_v8f64_ty], + [llvm_v8f32_ty, llvm_v8f64_ty, llvm_i8_ty, llvm_i32_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtps2qq_128 : + GCCBuiltin<"__builtin_ia32_cvtps2qq128_mask">, + Intrinsic<[llvm_v2i64_ty], + [llvm_v4f32_ty, llvm_v2i64_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtps2qq_256 : + GCCBuiltin<"__builtin_ia32_cvtps2qq256_mask">, + Intrinsic<[llvm_v4i64_ty], + [llvm_v4f32_ty, llvm_v4i64_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtps2qq_512 : + GCCBuiltin<"__builtin_ia32_cvtps2qq512_mask">, + Intrinsic<[llvm_v8i64_ty], + [llvm_v8f32_ty, llvm_v8i64_ty, llvm_i8_ty, llvm_i32_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtps2udq_128 : + GCCBuiltin<"__builtin_ia32_cvtps2udq128_mask">, + Intrinsic<[llvm_v4i32_ty], + [llvm_v4f32_ty, llvm_v4i32_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtps2udq_256 : + GCCBuiltin<"__builtin_ia32_cvtps2udq256_mask">, + Intrinsic<[llvm_v8i32_ty], + [llvm_v8f32_ty, llvm_v8i32_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtps2udq_512 : + GCCBuiltin<"__builtin_ia32_cvtps2udq512_mask">, + Intrinsic<[llvm_v16i32_ty], + [llvm_v16f32_ty, llvm_v16i32_ty, llvm_i16_ty, llvm_i32_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtps2uqq_128 : + GCCBuiltin<"__builtin_ia32_cvtps2uqq128_mask">, + Intrinsic<[llvm_v2i64_ty], + [llvm_v4f32_ty, llvm_v2i64_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtps2uqq_256 : + GCCBuiltin<"__builtin_ia32_cvtps2uqq256_mask">, + Intrinsic<[llvm_v4i64_ty], + [llvm_v4f32_ty, llvm_v4i64_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtps2uqq_512 : + GCCBuiltin<"__builtin_ia32_cvtps2uqq512_mask">, + Intrinsic<[llvm_v8i64_ty], + [llvm_v8f32_ty, llvm_v8i64_ty, llvm_i8_ty, llvm_i32_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtqq2pd_128 : + GCCBuiltin<"__builtin_ia32_cvtqq2pd128_mask">, + Intrinsic<[llvm_v2f64_ty], + [llvm_v2i64_ty, llvm_v2f64_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtqq2pd_256 : + GCCBuiltin<"__builtin_ia32_cvtqq2pd256_mask">, + Intrinsic<[llvm_v4f64_ty], + [llvm_v4i64_ty, llvm_v4f64_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtqq2pd_512 : + GCCBuiltin<"__builtin_ia32_cvtqq2pd512_mask">, + Intrinsic<[llvm_v8f64_ty], + [llvm_v8i64_ty, llvm_v8f64_ty, llvm_i8_ty, llvm_i32_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtqq2ps_128 : + GCCBuiltin<"__builtin_ia32_cvtqq2ps128_mask">, + Intrinsic<[llvm_v4f32_ty], + [llvm_v2i64_ty, llvm_v4f32_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtqq2ps_256 : + GCCBuiltin<"__builtin_ia32_cvtqq2ps256_mask">, + Intrinsic<[llvm_v4f32_ty], + [llvm_v4i64_ty, llvm_v4f32_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtqq2ps_512 : + GCCBuiltin<"__builtin_ia32_cvtqq2ps512_mask">, + Intrinsic<[llvm_v8f32_ty], + [llvm_v8i64_ty, llvm_v8f32_ty, llvm_i8_ty, llvm_i32_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvttpd2dq_128 : + GCCBuiltin<"__builtin_ia32_cvttpd2dq128_mask">, + Intrinsic<[llvm_v4i32_ty], + [llvm_v2f64_ty, llvm_v4i32_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvttpd2dq_256 : + GCCBuiltin<"__builtin_ia32_cvttpd2dq256_mask">, + Intrinsic<[llvm_v4i32_ty], + [llvm_v4f64_ty, llvm_v4i32_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvttpd2dq_512 : + GCCBuiltin<"__builtin_ia32_cvttpd2dq512_mask">, + Intrinsic<[llvm_v8i32_ty], + [llvm_v8f64_ty, llvm_v8i32_ty, llvm_i8_ty, llvm_i32_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvttpd2qq_128 : + GCCBuiltin<"__builtin_ia32_cvttpd2qq128_mask">, + Intrinsic<[llvm_v2i64_ty], + [llvm_v2f64_ty, llvm_v2i64_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvttpd2qq_256 : + GCCBuiltin<"__builtin_ia32_cvttpd2qq256_mask">, + Intrinsic<[llvm_v4i64_ty], + [llvm_v4f64_ty, llvm_v4i64_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvttpd2qq_512 : + GCCBuiltin<"__builtin_ia32_cvttpd2qq512_mask">, + Intrinsic<[llvm_v8i64_ty], + [llvm_v8f64_ty, llvm_v8i64_ty, llvm_i8_ty, llvm_i32_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvttpd2udq_128 : + GCCBuiltin<"__builtin_ia32_cvttpd2udq128_mask">, + Intrinsic<[llvm_v4i32_ty], + [llvm_v2f64_ty, llvm_v4i32_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvttpd2udq_256 : + GCCBuiltin<"__builtin_ia32_cvttpd2udq256_mask">, + Intrinsic<[llvm_v4i32_ty], + [llvm_v4f64_ty, llvm_v4i32_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvttpd2udq_512 : + GCCBuiltin<"__builtin_ia32_cvttpd2udq512_mask">, + Intrinsic<[llvm_v8i32_ty], + [llvm_v8f64_ty, llvm_v8i32_ty, llvm_i8_ty, llvm_i32_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvttpd2uqq_128 : + GCCBuiltin<"__builtin_ia32_cvttpd2uqq128_mask">, + Intrinsic<[llvm_v2i64_ty], + [llvm_v2f64_ty, llvm_v2i64_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvttpd2uqq_256 : + GCCBuiltin<"__builtin_ia32_cvttpd2uqq256_mask">, + Intrinsic<[llvm_v4i64_ty], + [llvm_v4f64_ty, llvm_v4i64_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvttpd2uqq_512 : + GCCBuiltin<"__builtin_ia32_cvttpd2uqq512_mask">, + Intrinsic<[llvm_v8i64_ty], + [llvm_v8f64_ty, llvm_v8i64_ty, llvm_i8_ty, llvm_i32_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvttps2dq_128 : + GCCBuiltin<"__builtin_ia32_cvttps2dq128_mask">, + Intrinsic<[llvm_v4i32_ty], + [llvm_v4f32_ty, llvm_v4i32_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvttps2dq_256 : + GCCBuiltin<"__builtin_ia32_cvttps2dq256_mask">, + Intrinsic<[llvm_v8i32_ty], + [llvm_v8f32_ty, llvm_v8i32_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvttps2dq_512 : + GCCBuiltin<"__builtin_ia32_cvttps2dq512_mask">, + Intrinsic<[llvm_v16i32_ty], + [llvm_v16f32_ty, llvm_v16i32_ty, llvm_i16_ty, llvm_i32_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvttps2qq_128 : + GCCBuiltin<"__builtin_ia32_cvttps2qq128_mask">, + Intrinsic<[llvm_v2i64_ty], + [llvm_v4f32_ty, llvm_v2i64_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvttps2qq_256 : + GCCBuiltin<"__builtin_ia32_cvttps2qq256_mask">, + Intrinsic<[llvm_v4i64_ty], + [llvm_v4f32_ty, llvm_v4i64_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvttps2qq_512 : + GCCBuiltin<"__builtin_ia32_cvttps2qq512_mask">, + Intrinsic<[llvm_v8i64_ty], + [llvm_v8f32_ty, llvm_v8i64_ty, llvm_i8_ty, llvm_i32_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvttps2udq_128 : + GCCBuiltin<"__builtin_ia32_cvttps2udq128_mask">, + Intrinsic<[llvm_v4i32_ty], + [llvm_v4f32_ty, llvm_v4i32_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvttps2udq_256 : + GCCBuiltin<"__builtin_ia32_cvttps2udq256_mask">, + Intrinsic<[llvm_v8i32_ty], + [llvm_v8f32_ty, llvm_v8i32_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvttps2udq_512 : + GCCBuiltin<"__builtin_ia32_cvttps2udq512_mask">, + Intrinsic<[llvm_v16i32_ty], + [llvm_v16f32_ty, llvm_v16i32_ty, llvm_i16_ty, llvm_i32_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvttps2uqq_128 : + GCCBuiltin<"__builtin_ia32_cvttps2uqq128_mask">, + Intrinsic<[llvm_v2i64_ty], + [llvm_v4f32_ty, llvm_v2i64_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvttps2uqq_256 : + GCCBuiltin<"__builtin_ia32_cvttps2uqq256_mask">, + Intrinsic<[llvm_v4i64_ty], + [llvm_v4f32_ty, llvm_v4i64_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvttps2uqq_512 : + GCCBuiltin<"__builtin_ia32_cvttps2uqq512_mask">, + Intrinsic<[llvm_v8i64_ty], + [llvm_v8f32_ty, llvm_v8i64_ty, llvm_i8_ty, llvm_i32_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtudq2pd_128 : + GCCBuiltin<"__builtin_ia32_cvtudq2pd128_mask">, + Intrinsic<[llvm_v2f64_ty], + [llvm_v4i32_ty, llvm_v2f64_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtudq2pd_256 : + GCCBuiltin<"__builtin_ia32_cvtudq2pd256_mask">, + Intrinsic<[llvm_v4f64_ty], + [llvm_v4i32_ty, llvm_v4f64_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtudq2pd_512 : + GCCBuiltin<"__builtin_ia32_cvtudq2pd512_mask">, + Intrinsic<[llvm_v8f64_ty], + [llvm_v8i32_ty, llvm_v8f64_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtudq2ps_128 : + GCCBuiltin<"__builtin_ia32_cvtudq2ps128_mask">, + Intrinsic<[llvm_v4f32_ty], + [llvm_v4i32_ty, llvm_v4f32_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtudq2ps_256 : + GCCBuiltin<"__builtin_ia32_cvtudq2ps256_mask">, + Intrinsic<[llvm_v8f32_ty], + [llvm_v8i32_ty, llvm_v8f32_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtudq2ps_512 : + GCCBuiltin<"__builtin_ia32_cvtudq2ps512_mask">, + Intrinsic<[llvm_v16f32_ty], + [llvm_v16i32_ty, llvm_v16f32_ty, llvm_i16_ty, llvm_i32_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtuqq2pd_128 : + GCCBuiltin<"__builtin_ia32_cvtuqq2pd128_mask">, + Intrinsic<[llvm_v2f64_ty], + [llvm_v2i64_ty, llvm_v2f64_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtuqq2pd_256 : + GCCBuiltin<"__builtin_ia32_cvtuqq2pd256_mask">, + Intrinsic<[llvm_v4f64_ty], + [llvm_v4i64_ty, llvm_v4f64_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtuqq2pd_512 : + GCCBuiltin<"__builtin_ia32_cvtuqq2pd512_mask">, + Intrinsic<[llvm_v8f64_ty], + [llvm_v8i64_ty, llvm_v8f64_ty, llvm_i8_ty, llvm_i32_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtuqq2ps_128 : + GCCBuiltin<"__builtin_ia32_cvtuqq2ps128_mask">, + Intrinsic<[llvm_v4f32_ty], + [llvm_v2i64_ty, llvm_v4f32_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtuqq2ps_256 : + GCCBuiltin<"__builtin_ia32_cvtuqq2ps256_mask">, + Intrinsic<[llvm_v4f32_ty], + [llvm_v4i64_ty, llvm_v4f32_ty, llvm_i8_ty], + [IntrNoMem]>; + + def int_x86_avx512_mask_cvtuqq2ps_512 : + GCCBuiltin<"__builtin_ia32_cvtuqq2ps512_mask">, + Intrinsic<[llvm_v8f32_ty], + [llvm_v8i64_ty, llvm_v8f32_ty, llvm_i8_ty, llvm_i32_ty], + [IntrNoMem]>; + def int_x86_avx512_mask_rndscale_ps_512: GCCBuiltin<"__builtin_ia32_rndscaleps_mask">, Intrinsic<[llvm_v16f32_ty], [llvm_v16f32_ty, llvm_i32_ty, llvm_v16f32_ty, llvm_i16_ty, llvm_i32_ty], [IntrNoMem]>; def int_x86_avx512_mask_rndscale_pd_512: GCCBuiltin<"__builtin_ia32_rndscalepd_mask">, Intrinsic<[llvm_v8f64_ty], [llvm_v8f64_ty, llvm_i32_ty, llvm_v8f64_ty, llvm_i8_ty, llvm_i32_ty], [IntrNoMem]>; - def int_x86_avx512_mask_cvtps2dq_512: GCCBuiltin<"__builtin_ia32_cvtps2dq512_mask">, - Intrinsic<[llvm_v16i32_ty], [llvm_v16f32_ty, llvm_v16i32_ty, - llvm_i16_ty, llvm_i32_ty], [IntrNoMem]>; - def int_x86_avx512_mask_cvtpd2dq_512: GCCBuiltin<"__builtin_ia32_cvtpd2dq512_mask">, - Intrinsic<[llvm_v8i32_ty], [llvm_v8f64_ty, llvm_v8i32_ty, - llvm_i8_ty, llvm_i32_ty], [IntrNoMem]>; - def int_x86_avx512_mask_cvtps2udq_512: GCCBuiltin<"__builtin_ia32_cvtps2udq512_mask">, - Intrinsic<[llvm_v16i32_ty], [llvm_v16f32_ty, llvm_v16i32_ty, - llvm_i16_ty, llvm_i32_ty], [IntrNoMem]>; - def int_x86_avx512_mask_cvtpd2udq_512: GCCBuiltin<"__builtin_ia32_cvtpd2udq512_mask">, - Intrinsic<[llvm_v8i32_ty], [llvm_v8f64_ty, llvm_v8i32_ty, - llvm_i8_ty, llvm_i32_ty], [IntrNoMem]>; - def int_x86_avx512_mask_cvtdq2ps_512 : GCCBuiltin<"__builtin_ia32_cvtdq2ps512_mask">, - Intrinsic<[llvm_v16f32_ty], [llvm_v16i32_ty, llvm_v16f32_ty, - llvm_i16_ty, llvm_i32_ty], [IntrNoMem]>; - def int_x86_avx512_mask_cvtdq2pd_512 : GCCBuiltin<"__builtin_ia32_cvtdq2pd512_mask">, - Intrinsic<[llvm_v8f64_ty], [llvm_v8i32_ty, llvm_v8f64_ty, - llvm_i8_ty], [IntrNoMem]>; - def int_x86_avx512_mask_cvtudq2ps_512 : GCCBuiltin<"__builtin_ia32_cvtudq2ps512_mask">, - Intrinsic<[llvm_v16f32_ty], [llvm_v16i32_ty, llvm_v16f32_ty, - llvm_i16_ty, llvm_i32_ty], [IntrNoMem]>; - def int_x86_avx512_mask_cvtudq2pd_512 : GCCBuiltin<"__builtin_ia32_cvtudq2pd512_mask">, - Intrinsic<[llvm_v8f64_ty], [llvm_v8i32_ty, llvm_v8f64_ty, - llvm_i8_ty], [IntrNoMem]>; - def int_x86_avx512_mask_cvtpd2ps_512 : GCCBuiltin<"__builtin_ia32_cvtpd2ps512_mask">, - Intrinsic<[llvm_v8f32_ty], [llvm_v8f64_ty, llvm_v8f32_ty, - llvm_i8_ty, llvm_i32_ty], [IntrNoMem]>; } // Vector load with broadcast Index: lib/Target/X86/X86ISelLowering.cpp =================================================================== --- lib/Target/X86/X86ISelLowering.cpp +++ lib/Target/X86/X86ISelLowering.cpp @@ -15358,28 +15358,45 @@ SDValue PassThru = Op.getOperand(2); SDValue Mask = Op.getOperand(3); SDValue RoundingMode; + // We allways add rounding mode to the Node. + // If the rounding mode is not specified, we add the + // "current direction" mode. if (Op.getNumOperands() == 4) - RoundingMode = DAG.getConstant(X86::STATIC_ROUNDING::CUR_DIRECTION, dl, MVT::i32); + RoundingMode = + DAG.getConstant(X86::STATIC_ROUNDING::CUR_DIRECTION, dl, MVT::i32); else RoundingMode = Op.getOperand(4); unsigned IntrWithRoundingModeOpcode = IntrData->Opc1; - if (IntrWithRoundingModeOpcode != 0) { - unsigned Round = cast(RoundingMode)->getZExtValue(); - if (Round != X86::STATIC_ROUNDING::CUR_DIRECTION) + if (IntrWithRoundingModeOpcode != 0) + if (cast(RoundingMode)->getZExtValue() != + X86::STATIC_ROUNDING::CUR_DIRECTION) return getVectorMaskingNode(DAG.getNode(IntrWithRoundingModeOpcode, dl, Op.getValueType(), Src, RoundingMode), Mask, PassThru, Subtarget, DAG); - } return getVectorMaskingNode(DAG.getNode(IntrData->Opc0, dl, VT, Src, RoundingMode), Mask, PassThru, Subtarget, DAG); } case INTR_TYPE_1OP_MASK: { SDValue Src = Op.getOperand(1); - SDValue Passthru = Op.getOperand(2); + SDValue PassThru = Op.getOperand(2); SDValue Mask = Op.getOperand(3); + // We add rounding mode to the Node when + // - RM Opcode is specified and + // - RM is not "current direction". + unsigned IntrWithRoundingModeOpcode = IntrData->Opc1; + if (IntrWithRoundingModeOpcode != 0) { + SDValue Rnd = Op.getOperand(4); + unsigned Round = cast(Rnd)->getZExtValue(); + if (Round != X86::STATIC_ROUNDING::CUR_DIRECTION) { + return getVectorMaskingNode(DAG.getNode(IntrWithRoundingModeOpcode, + dl, Op.getValueType(), + Src, Rnd), + Mask, PassThru, Subtarget, DAG); + } + } return getVectorMaskingNode(DAG.getNode(IntrData->Opc0, dl, VT, Src), - Mask, Passthru, Subtarget, DAG); + Mask, PassThru, Subtarget, DAG); } case INTR_TYPE_SCALAR_MASK_RM: { SDValue Src1 = Op.getOperand(1); Index: lib/Target/X86/X86IntrinsicsInfo.h =================================================================== --- lib/Target/X86/X86IntrinsicsInfo.h +++ lib/Target/X86/X86IntrinsicsInfo.h @@ -416,6 +416,163 @@ X86_INTRINSIC_DATA(avx512_mask_compress_q_512, COMPRESS_EXPAND_IN_REG, X86ISD::COMPRESS, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtdq2pd_128, INTR_TYPE_1OP_MASK, + X86ISD::CVTDQ2PD, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtdq2pd_256, INTR_TYPE_1OP_MASK, + ISD::SINT_TO_FP, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtdq2pd_512, INTR_TYPE_1OP_MASK, + ISD::SINT_TO_FP, 0), // no rm + X86_INTRINSIC_DATA(avx512_mask_cvtdq2ps_128, INTR_TYPE_1OP_MASK, + ISD::SINT_TO_FP, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtdq2ps_256, INTR_TYPE_1OP_MASK, + ISD::SINT_TO_FP, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtdq2ps_512, INTR_TYPE_1OP_MASK, + ISD::SINT_TO_FP, ISD::SINT_TO_FP), //er + X86_INTRINSIC_DATA(avx512_mask_cvtpd2dq_128, INTR_TYPE_1OP_MASK, + X86ISD::FP_TO_SINT_RND, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtpd2dq_256, INTR_TYPE_1OP_MASK, + X86ISD::FP_TO_SINT_RND, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtpd2dq_512, INTR_TYPE_1OP_MASK, + X86ISD::FP_TO_SINT_RND, X86ISD::FP_TO_SINT_RND), + X86_INTRINSIC_DATA(avx512_mask_cvtpd2ps, INTR_TYPE_1OP_MASK, + X86ISD::VFPROUND, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtpd2ps_256, INTR_TYPE_1OP_MASK_RM, + ISD::FP_ROUND, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtpd2ps_512, INTR_TYPE_1OP_MASK_RM, + ISD::FP_ROUND, X86ISD::VFPROUND), + X86_INTRINSIC_DATA(avx512_mask_cvtpd2qq_128, INTR_TYPE_1OP_MASK, + X86ISD::FP_TO_SINT_RND, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtpd2qq_256, INTR_TYPE_1OP_MASK, + X86ISD::FP_TO_SINT_RND, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtpd2qq_512, INTR_TYPE_1OP_MASK, + X86ISD::FP_TO_SINT_RND, X86ISD::FP_TO_SINT_RND), + X86_INTRINSIC_DATA(avx512_mask_cvtpd2udq_128, INTR_TYPE_1OP_MASK, + X86ISD::FP_TO_UINT_RND, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtpd2udq_256, INTR_TYPE_1OP_MASK, + X86ISD::FP_TO_UINT_RND, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtpd2udq_512, INTR_TYPE_1OP_MASK, + X86ISD::FP_TO_UINT_RND, X86ISD::FP_TO_UINT_RND), + X86_INTRINSIC_DATA(avx512_mask_cvtpd2uqq_128, INTR_TYPE_1OP_MASK, + X86ISD::FP_TO_UINT_RND, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtpd2uqq_256, INTR_TYPE_1OP_MASK, + X86ISD::FP_TO_UINT_RND, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtpd2uqq_512, INTR_TYPE_1OP_MASK, + X86ISD::FP_TO_UINT_RND, X86ISD::FP_TO_UINT_RND), + X86_INTRINSIC_DATA(avx512_mask_cvtps2dq_128, INTR_TYPE_1OP_MASK, + X86ISD::FP_TO_SINT_RND, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtps2dq_256, INTR_TYPE_1OP_MASK, + X86ISD::FP_TO_SINT_RND, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtps2dq_512, INTR_TYPE_1OP_MASK, + X86ISD::FP_TO_SINT_RND, X86ISD::FP_TO_SINT_RND), + X86_INTRINSIC_DATA(avx512_mask_cvtps2pd_128, INTR_TYPE_1OP_MASK, + X86ISD::VFPEXT, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtps2pd_256, INTR_TYPE_1OP_MASK, + ISD::FP_EXTEND, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtps2pd_512, INTR_TYPE_1OP_MASK, + ISD::FP_EXTEND, X86ISD::VFPEXT), + X86_INTRINSIC_DATA(avx512_mask_cvtps2qq_128, INTR_TYPE_1OP_MASK, + X86ISD::FP_TO_SINT_RND, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtps2qq_256, INTR_TYPE_1OP_MASK, + X86ISD::FP_TO_SINT_RND, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtps2qq_512, INTR_TYPE_1OP_MASK, + X86ISD::FP_TO_SINT_RND, X86ISD::FP_TO_SINT_RND), + X86_INTRINSIC_DATA(avx512_mask_cvtps2udq_128, INTR_TYPE_1OP_MASK, + X86ISD::FP_TO_UINT_RND, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtps2udq_256, INTR_TYPE_1OP_MASK, + X86ISD::FP_TO_UINT_RND, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtps2udq_512, INTR_TYPE_1OP_MASK, + X86ISD::FP_TO_UINT_RND, X86ISD::FP_TO_UINT_RND), + X86_INTRINSIC_DATA(avx512_mask_cvtps2uqq_128, INTR_TYPE_1OP_MASK, + X86ISD::FP_TO_UINT_RND, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtps2uqq_256, INTR_TYPE_1OP_MASK, + X86ISD::FP_TO_UINT_RND, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtps2uqq_512, INTR_TYPE_1OP_MASK, + X86ISD::FP_TO_UINT_RND, X86ISD::FP_TO_UINT_RND), + X86_INTRINSIC_DATA(avx512_mask_cvtqq2pd_128, INTR_TYPE_1OP_MASK, + ISD::SINT_TO_FP, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtqq2pd_256, INTR_TYPE_1OP_MASK, + ISD::SINT_TO_FP, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtqq2pd_512, INTR_TYPE_1OP_MASK, + ISD::SINT_TO_FP, ISD::SINT_TO_FP), + X86_INTRINSIC_DATA(avx512_mask_cvtqq2ps_128, INTR_TYPE_1OP_MASK, + ISD::SINT_TO_FP, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtqq2ps_256, INTR_TYPE_1OP_MASK, + ISD::SINT_TO_FP, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtqq2ps_512, INTR_TYPE_1OP_MASK, + ISD::SINT_TO_FP, ISD::SINT_TO_FP), + X86_INTRINSIC_DATA(avx512_mask_cvttpd2dq_128, INTR_TYPE_1OP_MASK, + ISD::FP_TO_SINT, 0), + X86_INTRINSIC_DATA(avx512_mask_cvttpd2dq_256, INTR_TYPE_1OP_MASK, + ISD::FP_TO_SINT, 0), + X86_INTRINSIC_DATA(avx512_mask_cvttpd2dq_512, INTR_TYPE_1OP_MASK, + ISD::FP_TO_SINT, ISD::FP_TO_SINT), + X86_INTRINSIC_DATA(avx512_mask_cvttpd2qq_128, INTR_TYPE_1OP_MASK, + ISD::FP_TO_SINT, 0), + X86_INTRINSIC_DATA(avx512_mask_cvttpd2qq_256, INTR_TYPE_1OP_MASK, + ISD::FP_TO_SINT, 0), + X86_INTRINSIC_DATA(avx512_mask_cvttpd2qq_512, INTR_TYPE_1OP_MASK, + ISD::FP_TO_SINT, ISD::FP_TO_SINT), + X86_INTRINSIC_DATA(avx512_mask_cvttpd2udq_128, INTR_TYPE_1OP_MASK, + ISD::FP_TO_UINT, 0), + X86_INTRINSIC_DATA(avx512_mask_cvttpd2udq_256, INTR_TYPE_1OP_MASK, + ISD::FP_TO_UINT, 0), + X86_INTRINSIC_DATA(avx512_mask_cvttpd2udq_512, INTR_TYPE_1OP_MASK, + ISD::FP_TO_UINT, ISD::FP_TO_UINT), + X86_INTRINSIC_DATA(avx512_mask_cvttpd2uqq_128, INTR_TYPE_1OP_MASK, + ISD::FP_TO_UINT, 0), + X86_INTRINSIC_DATA(avx512_mask_cvttpd2uqq_256, INTR_TYPE_1OP_MASK, + ISD::FP_TO_UINT, 0), + X86_INTRINSIC_DATA(avx512_mask_cvttpd2uqq_512, INTR_TYPE_1OP_MASK, + ISD::FP_TO_UINT, ISD::FP_TO_UINT), + X86_INTRINSIC_DATA(avx512_mask_cvttps2dq_128, INTR_TYPE_1OP_MASK, + ISD::FP_TO_SINT, 0), + X86_INTRINSIC_DATA(avx512_mask_cvttps2dq_256, INTR_TYPE_1OP_MASK, + ISD::FP_TO_SINT, 0), + X86_INTRINSIC_DATA(avx512_mask_cvttps2dq_512, INTR_TYPE_1OP_MASK, + ISD::FP_TO_SINT, ISD::FP_TO_SINT), + X86_INTRINSIC_DATA(avx512_mask_cvttps2qq_128, INTR_TYPE_1OP_MASK, + ISD::FP_TO_SINT, 0), + X86_INTRINSIC_DATA(avx512_mask_cvttps2qq_256, INTR_TYPE_1OP_MASK, + ISD::FP_TO_SINT, 0), + X86_INTRINSIC_DATA(avx512_mask_cvttps2qq_512, INTR_TYPE_1OP_MASK, + ISD::FP_TO_SINT, ISD::FP_TO_SINT), + X86_INTRINSIC_DATA(avx512_mask_cvttps2udq_128, INTR_TYPE_1OP_MASK, + ISD::FP_TO_UINT, 0), + X86_INTRINSIC_DATA(avx512_mask_cvttps2udq_256, INTR_TYPE_1OP_MASK, + ISD::FP_TO_UINT, 0), + X86_INTRINSIC_DATA(avx512_mask_cvttps2udq_512, INTR_TYPE_1OP_MASK, + ISD::FP_TO_UINT, ISD::FP_TO_UINT), + X86_INTRINSIC_DATA(avx512_mask_cvttps2uqq_128, INTR_TYPE_1OP_MASK, + ISD::FP_TO_UINT, 0), + X86_INTRINSIC_DATA(avx512_mask_cvttps2uqq_256, INTR_TYPE_1OP_MASK, + ISD::FP_TO_UINT, 0), + X86_INTRINSIC_DATA(avx512_mask_cvttps2uqq_512, INTR_TYPE_1OP_MASK, + ISD::FP_TO_UINT, ISD::FP_TO_UINT), + X86_INTRINSIC_DATA(avx512_mask_cvtudq2pd_128, INTR_TYPE_1OP_MASK, + X86ISD::CVTUDQ2PD, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtudq2pd_256, INTR_TYPE_1OP_MASK, + ISD::UINT_TO_FP, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtudq2pd_512, INTR_TYPE_1OP_MASK, + ISD::UINT_TO_FP, 0), // no rm + X86_INTRINSIC_DATA(avx512_mask_cvtudq2ps_128, INTR_TYPE_1OP_MASK, + ISD::UINT_TO_FP, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtudq2ps_256, INTR_TYPE_1OP_MASK, + ISD::UINT_TO_FP, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtudq2ps_512, INTR_TYPE_1OP_MASK, + ISD::UINT_TO_FP, ISD::UINT_TO_FP), + X86_INTRINSIC_DATA(avx512_mask_cvtuqq2pd_128, INTR_TYPE_1OP_MASK, + ISD::UINT_TO_FP, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtuqq2pd_256, INTR_TYPE_1OP_MASK, + ISD::UINT_TO_FP, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtuqq2pd_512, INTR_TYPE_1OP_MASK, + ISD::UINT_TO_FP, ISD::UINT_TO_FP), + X86_INTRINSIC_DATA(avx512_mask_cvtuqq2ps_128, INTR_TYPE_1OP_MASK, + ISD::UINT_TO_FP, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtuqq2ps_256, INTR_TYPE_1OP_MASK, + ISD::UINT_TO_FP, 0), + X86_INTRINSIC_DATA(avx512_mask_cvtuqq2ps_512, INTR_TYPE_1OP_MASK, + ISD::UINT_TO_FP, ISD::UINT_TO_FP), + X86_INTRINSIC_DATA(avx512_mask_div_pd_128, INTR_TYPE_2OP_MASK, ISD::FDIV, 0), X86_INTRINSIC_DATA(avx512_mask_div_pd_256, INTR_TYPE_2OP_MASK, ISD::FDIV, 0), X86_INTRINSIC_DATA(avx512_mask_div_pd_512, INTR_TYPE_2OP_MASK, ISD::FDIV, Index: test/CodeGen/X86/avx512-intrinsics.ll =================================================================== --- test/CodeGen/X86/avx512-intrinsics.ll +++ test/CodeGen/X86/avx512-intrinsics.ll @@ -3118,3 +3118,237 @@ %res2 = fadd <16 x float> %res, %res1 ret <16 x float> %res2 } + +declare <8 x double> @llvm.x86.avx512.mask.cvtdq2pd.512(<8 x i32>, <8 x double>, i8) + +define <8 x double>@test_int_x86_avx512_mask_cvt_dq2pd_512(<8 x i32> %x0, <8 x double> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_dq2pd_512: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvtdq2pd %ymm0, %zmm1 {%k1} +; CHECK-NEXT: vcvtdq2pd %ymm0, %zmm0 +; CHECK-NEXT: vaddpd %zmm0, %zmm1, %zmm0 +; CHECK-NEXT: retq + %res = call <8 x double> @llvm.x86.avx512.mask.cvtdq2pd.512(<8 x i32> %x0, <8 x double> %x1, i8 %x2) + %res1 = call <8 x double> @llvm.x86.avx512.mask.cvtdq2pd.512(<8 x i32> %x0, <8 x double> %x1, i8 -1) + %res2 = fadd <8 x double> %res, %res1 + ret <8 x double> %res2 +} + +declare <16 x float> @llvm.x86.avx512.mask.cvtdq2ps.512(<16 x i32>, <16 x float>, i16, i32) + +define <16 x float>@test_int_x86_avx512_mask_cvt_dq2ps_512(<16 x i32> %x0, <16 x float> %x1, i16 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_dq2ps_512: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovw %edi, %k1 +; CHECK-NEXT: vcvtdq2ps %zmm0, %zmm1 {%k1} +; CHECK-NEXT: vcvtdq2ps {rn-sae}, %zmm0, %zmm0 +; CHECK-NEXT: vaddps %zmm0, %zmm1, %zmm0 +; CHECK-NEXT: retq + %res = call <16 x float> @llvm.x86.avx512.mask.cvtdq2ps.512(<16 x i32> %x0, <16 x float> %x1, i16 %x2, i32 4) + %res1 = call <16 x float> @llvm.x86.avx512.mask.cvtdq2ps.512(<16 x i32> %x0, <16 x float> %x1, i16 -1, i32 0) + %res2 = fadd <16 x float> %res, %res1 + ret <16 x float> %res2 +} + +declare <8 x i32> @llvm.x86.avx512.mask.cvtpd2dq.512(<8 x double>, <8 x i32>, i8, i32) + +define <8 x i32>@test_int_x86_avx512_mask_cvt_pd2dq_512(<8 x double> %x0, <8 x i32> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_pd2dq_512: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvtpd2dq %zmm0, %ymm1 {%k1} +; CHECK-NEXT: vcvtpd2dq {rn-sae}, %zmm0, %ymm0 +; CHECK-NEXT: vpaddd %ymm0, %ymm1, %ymm0 +; CHECK-NEXT: retq + %res = call <8 x i32> @llvm.x86.avx512.mask.cvtpd2dq.512(<8 x double> %x0, <8 x i32> %x1, i8 %x2, i32 4) + %res1 = call <8 x i32> @llvm.x86.avx512.mask.cvtpd2dq.512(<8 x double> %x0, <8 x i32> %x1, i8 -1, i32 0) + %res2 = add <8 x i32> %res, %res1 + ret <8 x i32> %res2 +} + +declare <8 x float> @llvm.x86.avx512.mask.cvtpd2ps.512(<8 x double>, <8 x float>, i8, i32) + +define <8 x float>@test_int_x86_avx512_mask_cvt_pd2ps_512(<8 x double> %x0, <8 x float> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_pd2ps_512: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvtpd2ps %zmm0, %ymm1 {%k1} +; CHECK-NEXT: vcvtpd2ps {ru-sae}, %zmm0, %ymm0 +; CHECK-NEXT: vaddps %ymm0, %ymm1, %ymm0 +; CHECK-NEXT: retq + %res = call <8 x float> @llvm.x86.avx512.mask.cvtpd2ps.512(<8 x double> %x0, <8 x float> %x1, i8 %x2, i32 4) + %res1 = call <8 x float> @llvm.x86.avx512.mask.cvtpd2ps.512(<8 x double> %x0, <8 x float> %x1, i8 -1, i32 2) + %res2 = fadd <8 x float> %res, %res1 + ret <8 x float> %res2 +} + +declare <8 x i32> @llvm.x86.avx512.mask.cvtpd2udq.512(<8 x double>, <8 x i32>, i8, i32) + +define <8 x i32>@test_int_x86_avx512_mask_cvt_pd2udq_512(<8 x double> %x0, <8 x i32> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_pd2udq_512: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvtpd2udq {ru-sae}, %zmm0, %ymm1 {%k1} +; CHECK-NEXT: vcvtpd2udq {rn-sae}, %zmm0, %ymm0 +; CHECK-NEXT: vpaddd %ymm0, %ymm1, %ymm0 +; CHECK-NEXT: retq + %res = call <8 x i32> @llvm.x86.avx512.mask.cvtpd2udq.512(<8 x double> %x0, <8 x i32> %x1, i8 %x2, i32 2) + %res1 = call <8 x i32> @llvm.x86.avx512.mask.cvtpd2udq.512(<8 x double> %x0, <8 x i32> %x1, i8 -1, i32 0) + %res2 = add <8 x i32> %res, %res1 + ret <8 x i32> %res2 +} + +declare <16 x i32> @llvm.x86.avx512.mask.cvtps2dq.512(<16 x float>, <16 x i32>, i16, i32) + +define <16 x i32>@test_int_x86_avx512_mask_cvt_ps2dq_512(<16 x float> %x0, <16 x i32> %x1, i16 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_ps2dq_512: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovw %edi, %k1 +; CHECK-NEXT: vcvtps2dq {ru-sae}, %zmm0, %zmm1 {%k1} +; CHECK-NEXT: vcvtps2dq {rn-sae}, %zmm0, %zmm0 +; CHECK-NEXT: vpaddd %zmm0, %zmm1, %zmm0 +; CHECK-NEXT: retq + %res = call <16 x i32> @llvm.x86.avx512.mask.cvtps2dq.512(<16 x float> %x0, <16 x i32> %x1, i16 %x2, i32 2) + %res1 = call <16 x i32> @llvm.x86.avx512.mask.cvtps2dq.512(<16 x float> %x0, <16 x i32> %x1, i16 -1, i32 0) + %res2 = add <16 x i32> %res, %res1 + ret <16 x i32> %res2 +} + +declare <8 x double> @llvm.x86.avx512.mask.cvtps2pd.512(<8 x float>, <8 x double>, i8, i32) + +define <8 x double>@test_int_x86_avx512_mask_cvt_ps2pd_512(<8 x float> %x0, <8 x double> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_ps2pd_512: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvtps2pd %ymm0, %zmm1 {%k1} +; CHECK-NEXT: vcvtps2pd {sae}, %ymm0, %zmm0 +; CHECK-NEXT: vaddpd %zmm0, %zmm1, %zmm0 +; CHECK-NEXT: retq + %res = call <8 x double> @llvm.x86.avx512.mask.cvtps2pd.512(<8 x float> %x0, <8 x double> %x1, i8 %x2, i32 4) + %res1 = call <8 x double> @llvm.x86.avx512.mask.cvtps2pd.512(<8 x float> %x0, <8 x double> %x1, i8 -1, i32 8) + %res2 = fadd <8 x double> %res, %res1 + ret <8 x double> %res2 +} + +declare <16 x i32> @llvm.x86.avx512.mask.cvtps2udq.512(<16 x float>, <16 x i32>, i16, i32) + +define <16 x i32>@test_int_x86_avx512_mask_cvt_ps2udq_512(<16 x float> %x0, <16 x i32> %x1, i16 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_ps2udq_512: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovw %edi, %k1 +; CHECK-NEXT: vcvtps2udq {ru-sae}, %zmm0, %zmm1 {%k1} +; CHECK-NEXT: vcvtps2udq {rn-sae}, %zmm0, %zmm0 +; CHECK-NEXT: vpaddd %zmm0, %zmm1, %zmm0 +; CHECK-NEXT: retq + %res = call <16 x i32> @llvm.x86.avx512.mask.cvtps2udq.512(<16 x float> %x0, <16 x i32> %x1, i16 %x2, i32 2) + %res1 = call <16 x i32> @llvm.x86.avx512.mask.cvtps2udq.512(<16 x float> %x0, <16 x i32> %x1, i16 -1, i32 0) + %res2 = add <16 x i32> %res, %res1 + ret <16 x i32> %res2 +} + +declare <8 x i32> @llvm.x86.avx512.mask.cvttpd2dq.512(<8 x double>, <8 x i32>, i8, i32) + +define <8 x i32>@test_int_x86_avx512_mask_cvtt_pd2dq_512(<8 x double> %x0, <8 x i32> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvtt_pd2dq_512: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvttpd2dq %zmm0, %ymm1 {%k1} +; CHECK-NEXT: vcvttpd2dq {sae}, %zmm0, %ymm0 +; CHECK-NEXT: vpaddd %ymm0, %ymm1, %ymm0 +; CHECK-NEXT: retq + %res = call <8 x i32> @llvm.x86.avx512.mask.cvttpd2dq.512(<8 x double> %x0, <8 x i32> %x1, i8 %x2, i32 4) + %res1 = call <8 x i32> @llvm.x86.avx512.mask.cvttpd2dq.512(<8 x double> %x0, <8 x i32> %x1, i8 -1, i32 8) + %res2 = add <8 x i32> %res, %res1 + ret <8 x i32> %res2 +} + +declare <8 x double> @llvm.x86.avx512.mask.cvtudq2pd.512(<8 x i32>, <8 x double>, i8) + +define <8 x double>@test_int_x86_avx512_mask_cvt_udq2pd_512(<8 x i32> %x0, <8 x double> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_udq2pd_512: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvtudq2pd %ymm0, %zmm1 {%k1} +; CHECK-NEXT: vcvtudq2pd %ymm0, %zmm0 +; CHECK-NEXT: vaddpd %zmm0, %zmm1, %zmm0 +; CHECK-NEXT: retq + %res = call <8 x double> @llvm.x86.avx512.mask.cvtudq2pd.512(<8 x i32> %x0, <8 x double> %x1, i8 %x2) + %res1 = call <8 x double> @llvm.x86.avx512.mask.cvtudq2pd.512(<8 x i32> %x0, <8 x double> %x1, i8 -1) + %res2 = fadd <8 x double> %res, %res1 + ret <8 x double> %res2 +} + + +declare <16 x float> @llvm.x86.avx512.mask.cvtudq2ps.512(<16 x i32>, <16 x float>, i16, i32) + +define <16 x float>@test_int_x86_avx512_mask_cvt_udq2ps_512(<16 x i32> %x0, <16 x float> %x1, i16 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_udq2ps_512: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovw %edi, %k1 +; CHECK-NEXT: vcvtudq2ps %zmm0, %zmm1 {%k1} +; CHECK-NEXT: vcvtudq2ps {rn-sae}, %zmm0, %zmm0 +; CHECK-NEXT: vaddps %zmm0, %zmm1, %zmm0 +; CHECK-NEXT: retq + %res = call <16 x float> @llvm.x86.avx512.mask.cvtudq2ps.512(<16 x i32> %x0, <16 x float> %x1, i16 %x2, i32 4) + %res1 = call <16 x float> @llvm.x86.avx512.mask.cvtudq2ps.512(<16 x i32> %x0, <16 x float> %x1, i16 -1, i32 0) + %res2 = fadd <16 x float> %res, %res1 + ret <16 x float> %res2 +} + +declare <8 x i32> @llvm.x86.avx512.mask.cvttpd2udq.512(<8 x double>, <8 x i32>, i8, i32) + +define <8 x i32>@test_int_x86_avx512_mask_cvtt_pd2udq_512(<8 x double> %x0, <8 x i32> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvtt_pd2udq_512: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvttpd2udq %zmm0, %ymm1 {%k1} +; CHECK-NEXT: vcvttpd2udq {sae}, %zmm0, %ymm0 +; CHECK-NEXT: vpaddd %ymm0, %ymm1, %ymm0 +; CHECK-NEXT: retq + %res = call <8 x i32> @llvm.x86.avx512.mask.cvttpd2udq.512(<8 x double> %x0, <8 x i32> %x1, i8 %x2, i32 4) + %res1 = call <8 x i32> @llvm.x86.avx512.mask.cvttpd2udq.512(<8 x double> %x0, <8 x i32> %x1, i8 -1, i32 8) + %res2 = add <8 x i32> %res, %res1 + ret <8 x i32> %res2 +} + +declare <16 x i32> @llvm.x86.avx512.mask.cvttps2dq.512(<16 x float>, <16 x i32>, i16, i32) + +define <16 x i32>@test_int_x86_avx512_mask_cvtt_ps2dq_512(<16 x float> %x0, <16 x i32> %x1, i16 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvtt_ps2dq_512: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovw %edi, %k1 +; CHECK-NEXT: vcvttps2dq %zmm0, %zmm1 {%k1} +; CHECK-NEXT: vcvttps2dq {sae}, %zmm0, %zmm0 +; CHECK-NEXT: vpaddd %zmm0, %zmm1, %zmm0 +; CHECK-NEXT: retq + %res = call <16 x i32> @llvm.x86.avx512.mask.cvttps2dq.512(<16 x float> %x0, <16 x i32> %x1, i16 %x2, i32 4) + %res1 = call <16 x i32> @llvm.x86.avx512.mask.cvttps2dq.512(<16 x float> %x0, <16 x i32> %x1, i16 -1, i32 8) + %res2 = add <16 x i32> %res, %res1 + ret <16 x i32> %res2 +} + +declare <16 x i32> @llvm.x86.avx512.mask.cvttps2udq.512(<16 x float>, <16 x i32>, i16, i32) + +define <16 x i32>@test_int_x86_avx512_mask_cvtt_ps2udq_512(<16 x float> %x0, <16 x i32> %x1, i16 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvtt_ps2udq_512: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovw %edi, %k1 +; CHECK-NEXT: vcvttps2udq %zmm0, %zmm1 {%k1} +; CHECK-NEXT: vcvttps2udq {sae}, %zmm0, %zmm0 +; CHECK-NEXT: vpaddd %zmm0, %zmm1, %zmm0 +; CHECK-NEXT: retq + %res = call <16 x i32> @llvm.x86.avx512.mask.cvttps2udq.512(<16 x float> %x0, <16 x i32> %x1, i16 %x2, i32 4) + %res1 = call <16 x i32> @llvm.x86.avx512.mask.cvttps2udq.512(<16 x float> %x0, <16 x i32> %x1, i16 -1, i32 8) + %res2 = add <16 x i32> %res, %res1 + ret <16 x i32> %res2 +} + Index: test/CodeGen/X86/avx512bw-intrinsics.ll =================================================================== --- test/CodeGen/X86/avx512bw-intrinsics.ll +++ test/CodeGen/X86/avx512bw-intrinsics.ll @@ -1,4 +1,4 @@ -; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=skx --show-mc-encoding| FileCheck %s +; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=knl -mattr=+avx512bw --show-mc-encoding| FileCheck %s define i64 @test_pcmpeq_b(<64 x i8> %a, <64 x i8> %b) { ; CHECK-LABEL: test_pcmpeq_b @@ -304,28 +304,6 @@ declare i32 @llvm.x86.avx512.mask.ucmp.w.512(<32 x i16>, <32 x i16>, i32, i32) nounwind readnone -; CHECK-LABEL: test_x86_mask_blend_b_256 -; CHECK: vpblendmb -define <32 x i8> @test_x86_mask_blend_b_256(i32 %a0, <32 x i8> %a1, <32 x i8> %a2) { - %res = call <32 x i8> @llvm.x86.avx512.mask.blend.b.256(<32 x i8> %a1, <32 x i8> %a2, i32 %a0) ; <<32 x i8>> [#uses=1] - ret <32 x i8> %res -} -declare <32 x i8> @llvm.x86.avx512.mask.blend.b.256(<32 x i8>, <32 x i8>, i32) nounwind readonly - -; CHECK-LABEL: test_x86_mask_blend_w_256 -define <16 x i16> @test_x86_mask_blend_w_256(i16 %mask, <16 x i16> %a1, <16 x i16> %a2) { - ; CHECK: vpblendmw - %res = call <16 x i16> @llvm.x86.avx512.mask.blend.w.256(<16 x i16> %a1, <16 x i16> %a2, i16 %mask) ; <<16 x i16>> [#uses=1] - ret <16 x i16> %res -} -declare <16 x i16> @llvm.x86.avx512.mask.blend.w.256(<16 x i16>, <16 x i16>, i16) nounwind readonly - -; CHECK-LABEL: test_x86_mask_blend_b_512 -; CHECK: vpblendmb -define <64 x i8> @test_x86_mask_blend_b_512(i64 %a0, <64 x i8> %a1, <64 x i8> %a2) { - %res = call <64 x i8> @llvm.x86.avx512.mask.blend.b.512(<64 x i8> %a1, <64 x i8> %a2, i64 %a0) ; <<64 x i8>> [#uses=1] - ret <64 x i8> %res -} declare <64 x i8> @llvm.x86.avx512.mask.blend.b.512(<64 x i8>, <64 x i8>, i64) nounwind readonly ; CHECK-LABEL: test_x86_mask_blend_w_512 @@ -336,21 +314,12 @@ } declare <32 x i16> @llvm.x86.avx512.mask.blend.w.512(<32 x i16>, <32 x i16>, i32) nounwind readonly -; CHECK-LABEL: test_x86_mask_blend_b_128 +; CHECK-LABEL: test_x86_mask_blend_b_512 ; CHECK: vpblendmb -define <16 x i8> @test_x86_mask_blend_b_128(i16 %a0, <16 x i8> %a1, <16 x i8> %a2) { - %res = call <16 x i8> @llvm.x86.avx512.mask.blend.b.128(<16 x i8> %a1, <16 x i8> %a2, i16 %a0) ; <<16 x i8>> [#uses=1] - ret <16 x i8> %res -} -declare <16 x i8> @llvm.x86.avx512.mask.blend.b.128(<16 x i8>, <16 x i8>, i16) nounwind readonly - -; CHECK-LABEL: test_x86_mask_blend_w_128 -define <8 x i16> @test_x86_mask_blend_w_128(i8 %mask, <8 x i16> %a1, <8 x i16> %a2) { - ; CHECK: vpblendmw - %res = call <8 x i16> @llvm.x86.avx512.mask.blend.w.128(<8 x i16> %a1, <8 x i16> %a2, i8 %mask) ; <<8 x i16>> [#uses=1] - ret <8 x i16> %res +define <64 x i8> @test_x86_mask_blend_b_512(i64 %a0, <64 x i8> %a1, <64 x i8> %a2) { + %res = call <64 x i8> @llvm.x86.avx512.mask.blend.b.512(<64 x i8> %a1, <64 x i8> %a2, i64 %a0) ; <<64 x i8>> [#uses=1] + ret <64 x i8> %res } -declare <8 x i16> @llvm.x86.avx512.mask.blend.w.128(<8 x i16>, <8 x i16>, i8) nounwind readonly define <32 x i16> @test_mask_packs_epi32_rr_512(<16 x i32> %a, <16 x i32> %b) { ;CHECK-LABEL: test_mask_packs_epi32_rr_512 Index: test/CodeGen/X86/avx512bwvl-intrinsics.ll =================================================================== --- test/CodeGen/X86/avx512bwvl-intrinsics.ll +++ test/CodeGen/X86/avx512bwvl-intrinsics.ll @@ -3763,6 +3763,38 @@ ret <16 x i16> %res2 } +; CHECK-LABEL: test_x86_mask_blend_b_256 +; CHECK: vpblendmb +define <32 x i8> @test_x86_mask_blend_b_256(i32 %a0, <32 x i8> %a1, <32 x i8> %a2) { + %res = call <32 x i8> @llvm.x86.avx512.mask.blend.b.256(<32 x i8> %a1, <32 x i8> %a2, i32 %a0) ; <<32 x i8>> [#uses=1] + ret <32 x i8> %res +} +declare <32 x i8> @llvm.x86.avx512.mask.blend.b.256(<32 x i8>, <32 x i8>, i32) nounwind readonly + +; CHECK-LABEL: test_x86_mask_blend_w_256 +define <16 x i16> @test_x86_mask_blend_w_256(i16 %mask, <16 x i16> %a1, <16 x i16> %a2) { + ; CHECK: vpblendmw + %res = call <16 x i16> @llvm.x86.avx512.mask.blend.w.256(<16 x i16> %a1, <16 x i16> %a2, i16 %mask) ; <<16 x i16>> [#uses=1] + ret <16 x i16> %res +} +declare <16 x i16> @llvm.x86.avx512.mask.blend.w.256(<16 x i16>, <16 x i16>, i16) nounwind readonly + +; CHECK-LABEL: test_x86_mask_blend_b_128 +; CHECK: vpblendmb +define <16 x i8> @test_x86_mask_blend_b_128(i16 %a0, <16 x i8> %a1, <16 x i8> %a2) { + %res = call <16 x i8> @llvm.x86.avx512.mask.blend.b.128(<16 x i8> %a1, <16 x i8> %a2, i16 %a0) ; <<16 x i8>> [#uses=1] + ret <16 x i8> %res +} +declare <16 x i8> @llvm.x86.avx512.mask.blend.b.128(<16 x i8>, <16 x i8>, i16) nounwind readonly + +; CHECK-LABEL: test_x86_mask_blend_w_128 +define <8 x i16> @test_x86_mask_blend_w_128(i8 %mask, <8 x i16> %a1, <8 x i16> %a2) { + ; CHECK: vpblendmw + %res = call <8 x i16> @llvm.x86.avx512.mask.blend.w.128(<8 x i16> %a1, <8 x i16> %a2, i8 %mask) ; <<8 x i16>> [#uses=1] + ret <8 x i16> %res +} +declare <8 x i16> @llvm.x86.avx512.mask.blend.w.128(<8 x i16>, <8 x i16>, i8) nounwind readonly + declare <8 x i16> @llvm.x86.avx512.mask.pmulhu.w.128(<8 x i16>, <8 x i16>, <8 x i16>, i8) ; CHECK-LABEL: @test_int_x86_avx512_mask_pmulhu_w_128 Index: test/CodeGen/X86/avx512dq-intrinsics.ll =================================================================== --- test/CodeGen/X86/avx512dq-intrinsics.ll +++ test/CodeGen/X86/avx512dq-intrinsics.ll @@ -0,0 +1,194 @@ +; RUN: llc < %s -mtriple=x86_64-apple-darwin -mcpu=knl -mattr=+avx512dq | FileCheck %s + +declare <8 x i64> @llvm.x86.avx512.mask.cvtpd2qq.512(<8 x double>, <8 x i64>, i8, i32) + +define <8 x i64>@test_int_x86_avx512_mask_cvt_pd2qq_512(<8 x double> %x0, <8 x i64> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_pd2qq_512: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvtpd2qq {ru-sae}, %zmm0, %zmm1 {%k1} +; CHECK-NEXT: vcvtpd2qq {rn-sae}, %zmm0, %zmm0 +; CHECK-NEXT: vpaddq %zmm0, %zmm1, %zmm0 +; CHECK-NEXT: retq + %res = call <8 x i64> @llvm.x86.avx512.mask.cvtpd2qq.512(<8 x double> %x0, <8 x i64> %x1, i8 %x2, i32 2) + %res1 = call <8 x i64> @llvm.x86.avx512.mask.cvtpd2qq.512(<8 x double> %x0, <8 x i64> %x1, i8 -1, i32 0) + %res2 = add <8 x i64> %res, %res1 + ret <8 x i64> %res2 +} + +declare <8 x i64> @llvm.x86.avx512.mask.cvtpd2uqq.512(<8 x double>, <8 x i64>, i8, i32) + +define <8 x i64>@test_int_x86_avx512_mask_cvt_pd2uqq_512(<8 x double> %x0, <8 x i64> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_pd2uqq_512: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvtpd2uqq {ru-sae}, %zmm0, %zmm1 {%k1} +; CHECK-NEXT: vcvtpd2uqq {rn-sae}, %zmm0, %zmm0 +; CHECK-NEXT: vpaddq %zmm0, %zmm1, %zmm0 +; CHECK-NEXT: retq + %res = call <8 x i64> @llvm.x86.avx512.mask.cvtpd2uqq.512(<8 x double> %x0, <8 x i64> %x1, i8 %x2, i32 2) + %res1 = call <8 x i64> @llvm.x86.avx512.mask.cvtpd2uqq.512(<8 x double> %x0, <8 x i64> %x1, i8 -1, i32 0) + %res2 = add <8 x i64> %res, %res1 + ret <8 x i64> %res2 +} + +declare <8 x i64> @llvm.x86.avx512.mask.cvtps2qq.512(<8 x float>, <8 x i64>, i8, i32) + +define <8 x i64>@test_int_x86_avx512_mask_cvt_ps2qq_512(<8 x float> %x0, <8 x i64> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_ps2qq_512: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvtps2qq {ru-sae}, %ymm0, %zmm1 {%k1} +; CHECK-NEXT: vcvtps2qq {rn-sae}, %ymm0, %zmm0 +; CHECK-NEXT: vpaddq %zmm0, %zmm1, %zmm0 +; CHECK-NEXT: retq + %res = call <8 x i64> @llvm.x86.avx512.mask.cvtps2qq.512(<8 x float> %x0, <8 x i64> %x1, i8 %x2, i32 2) + %res1 = call <8 x i64> @llvm.x86.avx512.mask.cvtps2qq.512(<8 x float> %x0, <8 x i64> %x1, i8 -1, i32 0) + %res2 = add <8 x i64> %res, %res1 + ret <8 x i64> %res2 +} + +declare <8 x i64> @llvm.x86.avx512.mask.cvtps2uqq.512(<8 x float>, <8 x i64>, i8, i32) + +define <8 x i64>@test_int_x86_avx512_mask_cvt_ps2uqq_512(<8 x float> %x0, <8 x i64> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_ps2uqq_512: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvtps2uqq {ru-sae}, %ymm0, %zmm1 {%k1} +; CHECK-NEXT: vcvtps2uqq {rn-sae}, %ymm0, %zmm0 +; CHECK-NEXT: vpaddq %zmm0, %zmm1, %zmm0 +; CHECK-NEXT: retq + %res = call <8 x i64> @llvm.x86.avx512.mask.cvtps2uqq.512(<8 x float> %x0, <8 x i64> %x1, i8 %x2, i32 2) + %res1 = call <8 x i64> @llvm.x86.avx512.mask.cvtps2uqq.512(<8 x float> %x0, <8 x i64> %x1, i8 -1, i32 0) + %res2 = add <8 x i64> %res, %res1 + ret <8 x i64> %res2 +} + +declare <8 x double> @llvm.x86.avx512.mask.cvtqq2pd.512(<8 x i64>, <8 x double>, i8, i32) + +define <8 x double>@test_int_x86_avx512_mask_cvt_qq2pd_512(<8 x i64> %x0, <8 x double> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_qq2pd_512: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvtqq2pd %zmm0, %zmm1 {%k1} +; CHECK-NEXT: vcvtqq2pd {rn-sae}, %zmm0, %zmm0 +; CHECK-NEXT: vaddpd %zmm0, %zmm1, %zmm0 +; CHECK-NEXT: retq + %res = call <8 x double> @llvm.x86.avx512.mask.cvtqq2pd.512(<8 x i64> %x0, <8 x double> %x1, i8 %x2, i32 4) + %res1 = call <8 x double> @llvm.x86.avx512.mask.cvtqq2pd.512(<8 x i64> %x0, <8 x double> %x1, i8 -1, i32 0) + %res2 = fadd <8 x double> %res, %res1 + ret <8 x double> %res2 +} + +declare <8 x float> @llvm.x86.avx512.mask.cvtqq2ps.512(<8 x i64>, <8 x float>, i8, i32) + +define <8 x float>@test_int_x86_avx512_mask_cvt_qq2ps_512(<8 x i64> %x0, <8 x float> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_qq2ps_512: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvtqq2ps %zmm0, %ymm1 {%k1} +; CHECK-NEXT: vcvtqq2ps {rn-sae}, %zmm0, %ymm0 +; CHECK-NEXT: vaddps %ymm0, %ymm1, %ymm0 +; CHECK-NEXT: retq + %res = call <8 x float> @llvm.x86.avx512.mask.cvtqq2ps.512(<8 x i64> %x0, <8 x float> %x1, i8 %x2, i32 4) + %res1 = call <8 x float> @llvm.x86.avx512.mask.cvtqq2ps.512(<8 x i64> %x0, <8 x float> %x1, i8 -1, i32 0) + %res2 = fadd <8 x float> %res, %res1 + ret <8 x float> %res2 +} + +declare <8 x i64> @llvm.x86.avx512.mask.cvttpd2qq.512(<8 x double>, <8 x i64>, i8, i32) + +define <8 x i64>@test_int_x86_avx512_mask_cvtt_pd2qq_512(<8 x double> %x0, <8 x i64> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvtt_pd2qq_512: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvttpd2qq %zmm0, %zmm1 {%k1} +; CHECK-NEXT: vcvttpd2qq {sae}, %zmm0, %zmm0 +; CHECK-NEXT: vpaddq %zmm0, %zmm1, %zmm0 +; CHECK-NEXT: retq + %res = call <8 x i64> @llvm.x86.avx512.mask.cvttpd2qq.512(<8 x double> %x0, <8 x i64> %x1, i8 %x2, i32 4) + %res1 = call <8 x i64> @llvm.x86.avx512.mask.cvttpd2qq.512(<8 x double> %x0, <8 x i64> %x1, i8 -1, i32 8) + %res2 = add <8 x i64> %res, %res1 + ret <8 x i64> %res2 +} + +declare <8 x i64> @llvm.x86.avx512.mask.cvttpd2uqq.512(<8 x double>, <8 x i64>, i8, i32) + +define <8 x i64>@test_int_x86_avx512_mask_cvtt_pd2uqq_512(<8 x double> %x0, <8 x i64> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvtt_pd2uqq_512: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvttpd2uqq %zmm0, %zmm1 {%k1} +; CHECK-NEXT: vcvttpd2uqq {sae}, %zmm0, %zmm0 +; CHECK-NEXT: vpaddq %zmm0, %zmm1, %zmm0 +; CHECK-NEXT: retq + %res = call <8 x i64> @llvm.x86.avx512.mask.cvttpd2uqq.512(<8 x double> %x0, <8 x i64> %x1, i8 %x2, i32 4) + %res1 = call <8 x i64> @llvm.x86.avx512.mask.cvttpd2uqq.512(<8 x double> %x0, <8 x i64> %x1, i8 -1, i32 8) + %res2 = add <8 x i64> %res, %res1 + ret <8 x i64> %res2 +} + +declare <8 x i64> @llvm.x86.avx512.mask.cvttps2qq.512(<8 x float>, <8 x i64>, i8, i32) + +define <8 x i64>@test_int_x86_avx512_mask_cvtt_ps2qq_512(<8 x float> %x0, <8 x i64> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvtt_ps2qq_512: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvttps2qq %ymm0, %zmm1 {%k1} +; CHECK-NEXT: vcvttps2qq {sae}, %ymm0, %zmm0 +; CHECK-NEXT: vpaddq %zmm0, %zmm1, %zmm0 +; CHECK-NEXT: retq + %res = call <8 x i64> @llvm.x86.avx512.mask.cvttps2qq.512(<8 x float> %x0, <8 x i64> %x1, i8 %x2, i32 4) + %res1 = call <8 x i64> @llvm.x86.avx512.mask.cvttps2qq.512(<8 x float> %x0, <8 x i64> %x1, i8 -1, i32 8) + %res2 = add <8 x i64> %res, %res1 + ret <8 x i64> %res2 +} + +declare <8 x i64> @llvm.x86.avx512.mask.cvttps2uqq.512(<8 x float>, <8 x i64>, i8, i32) + +define <8 x i64>@test_int_x86_avx512_mask_cvtt_ps2uqq_512(<8 x float> %x0, <8 x i64> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvtt_ps2uqq_512: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvttps2uqq %ymm0, %zmm1 {%k1} +; CHECK-NEXT: vcvttps2uqq {sae}, %ymm0, %zmm0 +; CHECK-NEXT: vpaddq %zmm0, %zmm1, %zmm0 +; CHECK-NEXT: retq + %res = call <8 x i64> @llvm.x86.avx512.mask.cvttps2uqq.512(<8 x float> %x0, <8 x i64> %x1, i8 %x2, i32 4) + %res1 = call <8 x i64> @llvm.x86.avx512.mask.cvttps2uqq.512(<8 x float> %x0, <8 x i64> %x1, i8 -1, i32 8) + %res2 = add <8 x i64> %res, %res1 + ret <8 x i64> %res2 +} + +declare <8 x double> @llvm.x86.avx512.mask.cvtuqq2pd.512(<8 x i64>, <8 x double>, i8, i32) + +define <8 x double>@test_int_x86_avx512_mask_cvt_uqq2pd_512(<8 x i64> %x0, <8 x double> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_uqq2pd_512: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvtuqq2pd %zmm0, %zmm1 {%k1} +; CHECK-NEXT: vcvtuqq2pd {rn-sae}, %zmm0, %zmm0 +; CHECK-NEXT: vaddpd %zmm0, %zmm1, %zmm0 +; CHECK-NEXT: retq + %res = call <8 x double> @llvm.x86.avx512.mask.cvtuqq2pd.512(<8 x i64> %x0, <8 x double> %x1, i8 %x2, i32 4) + %res1 = call <8 x double> @llvm.x86.avx512.mask.cvtuqq2pd.512(<8 x i64> %x0, <8 x double> %x1, i8 -1, i32 0) + %res2 = fadd <8 x double> %res, %res1 + ret <8 x double> %res2 +} + +declare <8 x float> @llvm.x86.avx512.mask.cvtuqq2ps.512(<8 x i64>, <8 x float>, i8, i32) + +define <8 x float>@test_int_x86_avx512_mask_cvt_uqq2ps_512(<8 x i64> %x0, <8 x float> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_uqq2ps_512: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvtuqq2ps %zmm0, %ymm1 {%k1} +; CHECK-NEXT: vcvtuqq2ps {rn-sae}, %zmm0, %ymm0 +; CHECK-NEXT: vaddps %ymm0, %ymm1, %ymm0 +; CHECK-NEXT: retq + %res = call <8 x float> @llvm.x86.avx512.mask.cvtuqq2ps.512(<8 x i64> %x0, <8 x float> %x1, i8 %x2, i32 4) + %res1 = call <8 x float> @llvm.x86.avx512.mask.cvtuqq2ps.512(<8 x i64> %x0, <8 x float> %x1, i8 -1, i32 0) + %res2 = fadd <8 x float> %res, %res1 + ret <8 x float> %res2 +} + Index: test/CodeGen/X86/avx512dqvl-intrinsics.ll =================================================================== --- test/CodeGen/X86/avx512dqvl-intrinsics.ll +++ test/CodeGen/X86/avx512dqvl-intrinsics.ll @@ -1134,7 +1134,7 @@ define <16 x float> @test_mask_xor_ps_rmbk_512(<16 x float> %a, float* %ptr_b, <16 x float> %passThru, i16 %mask) { ;CHECK-LABEL: test_mask_xor_ps_rmbk_512 - ;CHECK: vxorps (%rdi){1to16}, %zmm0, %zmm1 {%k1} ## encoding: [0x62,0xf1,0x7c,0x59,0x57,0x0f] + ;CHECK: vxorps (%rdi){1to16}, %zmm0, %zmm1 {%k1} %q = load float, float* %ptr_b %vecinit.i = insertelement <16 x float> undef, float %q, i32 0 %b = shufflevector <16 x float> %vecinit.i, <16 x float> undef, <16 x i32> zeroinitializer @@ -1144,7 +1144,7 @@ define <16 x float> @test_mask_xor_ps_rmbkz_512(<16 x float> %a, float* %ptr_b, i16 %mask) { ;CHECK-LABEL: test_mask_xor_ps_rmbkz_512 - ;CHECK: vxorps (%rdi){1to16}, %zmm0, %zmm0 {%k1} {z} ## encoding: [0x62,0xf1,0x7c,0xd9,0x57,0x07] + ;CHECK: vxorps (%rdi){1to16}, %zmm0, %zmm0 {%k1} {z} %q = load float, float* %ptr_b %vecinit.i = insertelement <16 x float> undef, float %q, i32 0 %b = shufflevector <16 x float> %vecinit.i, <16 x float> undef, <16 x i32> zeroinitializer @@ -1152,4 +1152,388 @@ ret <16 x float> %res } -declare <16 x float> @llvm.x86.avx512.mask.xor.ps.512(<16 x float>, <16 x float>, <16 x float>, i16) \ No newline at end of file +declare <16 x float> @llvm.x86.avx512.mask.xor.ps.512(<16 x float>, <16 x float>, <16 x float>, i16) + +declare <2 x i64> @llvm.x86.avx512.mask.cvtpd2qq.128(<2 x double>, <2 x i64>, i8) + +define <2 x i64>@test_int_x86_avx512_mask_cvt_pd2qq_128(<2 x double> %x0, <2 x i64> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_pd2qq_128: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvtpd2qq %xmm0, %xmm1 {%k1} +; CHECK-NEXT: vcvtpd2qq %xmm0, %xmm0 +; CHECK-NEXT: vpaddq %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <2 x i64> @llvm.x86.avx512.mask.cvtpd2qq.128(<2 x double> %x0, <2 x i64> %x1, i8 %x2) + %res1 = call <2 x i64> @llvm.x86.avx512.mask.cvtpd2qq.128(<2 x double> %x0, <2 x i64> %x1, i8 -1) + %res2 = add <2 x i64> %res, %res1 + ret <2 x i64> %res2 +} + +declare <4 x i64> @llvm.x86.avx512.mask.cvtpd2qq.256(<4 x double>, <4 x i64>, i8) + +define <4 x i64>@test_int_x86_avx512_mask_cvt_pd2qq_256(<4 x double> %x0, <4 x i64> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_pd2qq_256: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvtpd2qq %ymm0, %ymm1 {%k1} +; CHECK-NEXT: vcvtpd2qq %ymm0, %ymm0 +; CHECK-NEXT: vpaddq %ymm0, %ymm1, %ymm0 +; CHECK-NEXT: retq + %res = call <4 x i64> @llvm.x86.avx512.mask.cvtpd2qq.256(<4 x double> %x0, <4 x i64> %x1, i8 %x2) + %res1 = call <4 x i64> @llvm.x86.avx512.mask.cvtpd2qq.256(<4 x double> %x0, <4 x i64> %x1, i8 -1) + %res2 = add <4 x i64> %res, %res1 + ret <4 x i64> %res2 +} + +declare <2 x i64> @llvm.x86.avx512.mask.cvtpd2uqq.128(<2 x double>, <2 x i64>, i8) + +define <2 x i64>@test_int_x86_avx512_mask_cvt_pd2uqq_128(<2 x double> %x0, <2 x i64> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_pd2uqq_128: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvtpd2uqq %xmm0, %xmm1 {%k1} +; CHECK-NEXT: vcvtpd2uqq %xmm0, %xmm0 +; CHECK-NEXT: vpaddq %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <2 x i64> @llvm.x86.avx512.mask.cvtpd2uqq.128(<2 x double> %x0, <2 x i64> %x1, i8 %x2) + %res1 = call <2 x i64> @llvm.x86.avx512.mask.cvtpd2uqq.128(<2 x double> %x0, <2 x i64> %x1, i8 -1) + %res2 = add <2 x i64> %res, %res1 + ret <2 x i64> %res2 +} + +declare <4 x i64> @llvm.x86.avx512.mask.cvtpd2uqq.256(<4 x double>, <4 x i64>, i8) + +define <4 x i64>@test_int_x86_avx512_mask_cvt_pd2uqq_256(<4 x double> %x0, <4 x i64> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_pd2uqq_256: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvtpd2uqq %ymm0, %ymm1 {%k1} +; CHECK-NEXT: vcvtpd2uqq %ymm0, %ymm0 +; CHECK-NEXT: vpaddq %ymm0, %ymm1, %ymm0 +; CHECK-NEXT: retq + %res = call <4 x i64> @llvm.x86.avx512.mask.cvtpd2uqq.256(<4 x double> %x0, <4 x i64> %x1, i8 %x2) + %res1 = call <4 x i64> @llvm.x86.avx512.mask.cvtpd2uqq.256(<4 x double> %x0, <4 x i64> %x1, i8 -1) + %res2 = add <4 x i64> %res, %res1 + ret <4 x i64> %res2 +} + +declare <2 x i64> @llvm.x86.avx512.mask.cvtps2qq.128(<4 x float>, <2 x i64>, i8) + +define <2 x i64>@test_int_x86_avx512_mask_cvt_ps2qq_128(<4 x float> %x0, <2 x i64> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_ps2qq_128: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvtps2qq %xmm0, %xmm1 {%k1} +; CHECK-NEXT: vcvtps2qq %xmm0, %xmm0 +; CHECK-NEXT: vpaddq %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <2 x i64> @llvm.x86.avx512.mask.cvtps2qq.128(<4 x float> %x0, <2 x i64> %x1, i8 %x2) + %res1 = call <2 x i64> @llvm.x86.avx512.mask.cvtps2qq.128(<4 x float> %x0, <2 x i64> %x1, i8 -1) + %res2 = add <2 x i64> %res, %res1 + ret <2 x i64> %res2 +} + +declare <4 x i64> @llvm.x86.avx512.mask.cvtps2qq.256(<4 x float>, <4 x i64>, i8) + +define <4 x i64>@test_int_x86_avx512_mask_cvt_ps2qq_256(<4 x float> %x0, <4 x i64> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_ps2qq_256: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvtps2qq %xmm0, %ymm1 {%k1} +; CHECK-NEXT: vcvtps2qq %xmm0, %ymm0 +; CHECK-NEXT: vpaddq %ymm0, %ymm1, %ymm0 +; CHECK-NEXT: retq + %res = call <4 x i64> @llvm.x86.avx512.mask.cvtps2qq.256(<4 x float> %x0, <4 x i64> %x1, i8 %x2) + %res1 = call <4 x i64> @llvm.x86.avx512.mask.cvtps2qq.256(<4 x float> %x0, <4 x i64> %x1, i8 -1) + %res2 = add <4 x i64> %res, %res1 + ret <4 x i64> %res2 +} + +declare <2 x i64> @llvm.x86.avx512.mask.cvtps2uqq.128(<4 x float>, <2 x i64>, i8) + +define <2 x i64>@test_int_x86_avx512_mask_cvt_ps2uqq_128(<4 x float> %x0, <2 x i64> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_ps2uqq_128: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvtps2uqq %xmm0, %xmm1 {%k1} +; CHECK-NEXT: vcvtps2uqq %xmm0, %xmm0 +; CHECK-NEXT: vpaddq %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <2 x i64> @llvm.x86.avx512.mask.cvtps2uqq.128(<4 x float> %x0, <2 x i64> %x1, i8 %x2) + %res1 = call <2 x i64> @llvm.x86.avx512.mask.cvtps2uqq.128(<4 x float> %x0, <2 x i64> %x1, i8 -1) + %res2 = add <2 x i64> %res, %res1 + ret <2 x i64> %res2 +} + +declare <4 x i64> @llvm.x86.avx512.mask.cvtps2uqq.256(<4 x float>, <4 x i64>, i8) + +define <4 x i64>@test_int_x86_avx512_mask_cvt_ps2uqq_256(<4 x float> %x0, <4 x i64> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_ps2uqq_256: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvtps2uqq %xmm0, %ymm1 {%k1} +; CHECK-NEXT: vcvtps2uqq %xmm0, %ymm0 +; CHECK-NEXT: vpaddq %ymm0, %ymm1, %ymm0 +; CHECK-NEXT: retq + %res = call <4 x i64> @llvm.x86.avx512.mask.cvtps2uqq.256(<4 x float> %x0, <4 x i64> %x1, i8 %x2) + %res1 = call <4 x i64> @llvm.x86.avx512.mask.cvtps2uqq.256(<4 x float> %x0, <4 x i64> %x1, i8 -1) + %res2 = add <4 x i64> %res, %res1 + ret <4 x i64> %res2 +} + +declare <2 x double> @llvm.x86.avx512.mask.cvtqq2pd.128(<2 x i64>, <2 x double>, i8) + +define <2 x double>@test_int_x86_avx512_mask_cvt_qq2pd_128(<2 x i64> %x0, <2 x double> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_qq2pd_128: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvtqq2pd %xmm0, %xmm1 {%k1} +; CHECK-NEXT: vcvtqq2pd %xmm0, %xmm0 +; CHECK-NEXT: vaddpd %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <2 x double> @llvm.x86.avx512.mask.cvtqq2pd.128(<2 x i64> %x0, <2 x double> %x1, i8 %x2) + %res1 = call <2 x double> @llvm.x86.avx512.mask.cvtqq2pd.128(<2 x i64> %x0, <2 x double> %x1, i8 -1) + %res2 = fadd <2 x double> %res, %res1 + ret <2 x double> %res2 +} + +declare <4 x double> @llvm.x86.avx512.mask.cvtqq2pd.256(<4 x i64>, <4 x double>, i8) + +define <4 x double>@test_int_x86_avx512_mask_cvt_qq2pd_256(<4 x i64> %x0, <4 x double> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_qq2pd_256: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvtqq2pd %ymm0, %ymm1 {%k1} +; CHECK-NEXT: vcvtqq2pd %ymm0, %ymm0 +; CHECK-NEXT: vaddpd %ymm0, %ymm1, %ymm0 +; CHECK-NEXT: retq + %res = call <4 x double> @llvm.x86.avx512.mask.cvtqq2pd.256(<4 x i64> %x0, <4 x double> %x1, i8 %x2) + %res1 = call <4 x double> @llvm.x86.avx512.mask.cvtqq2pd.256(<4 x i64> %x0, <4 x double> %x1, i8 -1) + %res2 = fadd <4 x double> %res, %res1 + ret <4 x double> %res2 +} + +declare <4 x float> @llvm.x86.avx512.mask.cvtqq2ps.128(<2 x i64>, <4 x float>, i8) + +define <4 x float>@test_int_x86_avx512_mask_cvt_qq2ps_128(<2 x i64> %x0, <4 x float> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_qq2ps_128: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvtqq2ps %xmm0, %xmm1 {%k1} +; CHECK-NEXT: vcvtqq2ps %xmm0, %xmm0 +; CHECK-NEXT: vaddps %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <4 x float> @llvm.x86.avx512.mask.cvtqq2ps.128(<2 x i64> %x0, <4 x float> %x1, i8 %x2) + %res1 = call <4 x float> @llvm.x86.avx512.mask.cvtqq2ps.128(<2 x i64> %x0, <4 x float> %x1, i8 -1) + %res2 = fadd <4 x float> %res, %res1 + ret <4 x float> %res2 +} + +declare <4 x float> @llvm.x86.avx512.mask.cvtqq2ps.256(<4 x i64>, <4 x float>, i8) + +define <4 x float>@test_int_x86_avx512_mask_cvt_qq2ps_256(<4 x i64> %x0, <4 x float> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_qq2ps_256: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvtqq2ps %ymm0, %xmm1 {%k1} +; CHECK-NEXT: vcvtqq2ps %ymm0, %xmm0 +; CHECK-NEXT: vaddps %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <4 x float> @llvm.x86.avx512.mask.cvtqq2ps.256(<4 x i64> %x0, <4 x float> %x1, i8 %x2) + %res1 = call <4 x float> @llvm.x86.avx512.mask.cvtqq2ps.256(<4 x i64> %x0, <4 x float> %x1, i8 -1) + %res2 = fadd <4 x float> %res, %res1 + ret <4 x float> %res2 +} + +declare <2 x i64> @llvm.x86.avx512.mask.cvttpd2qq.128(<2 x double>, <2 x i64>, i8) + +define <2 x i64>@test_int_x86_avx512_mask_cvtt_pd2qq_128(<2 x double> %x0, <2 x i64> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvtt_pd2qq_128: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvttpd2qq %xmm0, %xmm1 {%k1} +; CHECK-NEXT: vcvttpd2qq %xmm0, %xmm0 +; CHECK-NEXT: vpaddq %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <2 x i64> @llvm.x86.avx512.mask.cvttpd2qq.128(<2 x double> %x0, <2 x i64> %x1, i8 %x2) + %res1 = call <2 x i64> @llvm.x86.avx512.mask.cvttpd2qq.128(<2 x double> %x0, <2 x i64> %x1, i8 -1) + %res2 = add <2 x i64> %res, %res1 + ret <2 x i64> %res2 +} + +declare <4 x i64> @llvm.x86.avx512.mask.cvttpd2qq.256(<4 x double>, <4 x i64>, i8) + +define <4 x i64>@test_int_x86_avx512_mask_cvtt_pd2qq_256(<4 x double> %x0, <4 x i64> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvtt_pd2qq_256: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvttpd2qq %ymm0, %ymm1 {%k1} +; CHECK-NEXT: vcvttpd2qq %ymm0, %ymm0 +; CHECK-NEXT: vpaddq %ymm0, %ymm1, %ymm0 +; CHECK-NEXT: retq + %res = call <4 x i64> @llvm.x86.avx512.mask.cvttpd2qq.256(<4 x double> %x0, <4 x i64> %x1, i8 %x2) + %res1 = call <4 x i64> @llvm.x86.avx512.mask.cvttpd2qq.256(<4 x double> %x0, <4 x i64> %x1, i8 -1) + %res2 = add <4 x i64> %res, %res1 + ret <4 x i64> %res2 +} + +declare <2 x i64> @llvm.x86.avx512.mask.cvttpd2uqq.128(<2 x double>, <2 x i64>, i8) + +define <2 x i64>@test_int_x86_avx512_mask_cvtt_pd2uqq_128(<2 x double> %x0, <2 x i64> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvtt_pd2uqq_128: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvttpd2uqq %xmm0, %xmm1 {%k1} +; CHECK-NEXT: vcvttpd2uqq %xmm0, %xmm0 +; CHECK-NEXT: vpaddq %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <2 x i64> @llvm.x86.avx512.mask.cvttpd2uqq.128(<2 x double> %x0, <2 x i64> %x1, i8 %x2) + %res1 = call <2 x i64> @llvm.x86.avx512.mask.cvttpd2uqq.128(<2 x double> %x0, <2 x i64> %x1, i8 -1) + %res2 = add <2 x i64> %res, %res1 + ret <2 x i64> %res2 +} + +declare <4 x i64> @llvm.x86.avx512.mask.cvttpd2uqq.256(<4 x double>, <4 x i64>, i8) + +define <4 x i64>@test_int_x86_avx512_mask_cvtt_pd2uqq_256(<4 x double> %x0, <4 x i64> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvtt_pd2uqq_256: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvttpd2uqq %ymm0, %ymm1 {%k1} +; CHECK-NEXT: vcvttpd2uqq %ymm0, %ymm0 +; CHECK-NEXT: vpaddq %ymm0, %ymm1, %ymm0 +; CHECK-NEXT: retq + %res = call <4 x i64> @llvm.x86.avx512.mask.cvttpd2uqq.256(<4 x double> %x0, <4 x i64> %x1, i8 %x2) + %res1 = call <4 x i64> @llvm.x86.avx512.mask.cvttpd2uqq.256(<4 x double> %x0, <4 x i64> %x1, i8 -1) + %res2 = add <4 x i64> %res, %res1 + ret <4 x i64> %res2 +} + +declare <2 x i64> @llvm.x86.avx512.mask.cvttps2qq.128(<4 x float>, <2 x i64>, i8) + +define <2 x i64>@test_int_x86_avx512_mask_cvtt_ps2qq_128(<4 x float> %x0, <2 x i64> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvtt_ps2qq_128: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvttps2qq %xmm0, %xmm1 {%k1} +; CHECK-NEXT: vcvttps2qq %xmm0, %xmm0 +; CHECK-NEXT: vpaddq %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <2 x i64> @llvm.x86.avx512.mask.cvttps2qq.128(<4 x float> %x0, <2 x i64> %x1, i8 %x2) + %res1 = call <2 x i64> @llvm.x86.avx512.mask.cvttps2qq.128(<4 x float> %x0, <2 x i64> %x1, i8 -1) + %res2 = add <2 x i64> %res, %res1 + ret <2 x i64> %res2 +} + +declare <4 x i64> @llvm.x86.avx512.mask.cvttps2qq.256(<4 x float>, <4 x i64>, i8) + +define <4 x i64>@test_int_x86_avx512_mask_cvtt_ps2qq_256(<4 x float> %x0, <4 x i64> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvtt_ps2qq_256: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvttps2qq %xmm0, %ymm1 {%k1} +; CHECK-NEXT: vcvttps2qq %xmm0, %ymm0 +; CHECK-NEXT: vpaddq %ymm0, %ymm1, %ymm0 +; CHECK-NEXT: retq + %res = call <4 x i64> @llvm.x86.avx512.mask.cvttps2qq.256(<4 x float> %x0, <4 x i64> %x1, i8 %x2) + %res1 = call <4 x i64> @llvm.x86.avx512.mask.cvttps2qq.256(<4 x float> %x0, <4 x i64> %x1, i8 -1) + %res2 = add <4 x i64> %res, %res1 + ret <4 x i64> %res2 +} + +declare <2 x double> @llvm.x86.avx512.mask.cvtuqq2pd.128(<2 x i64>, <2 x double>, i8) + +define <2 x double>@test_int_x86_avx512_mask_cvt_uqq2pd_128(<2 x i64> %x0, <2 x double> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_uqq2pd_128: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvtuqq2pd %xmm0, %xmm1 {%k1} +; CHECK-NEXT: vcvtuqq2pd %xmm0, %xmm0 +; CHECK-NEXT: vaddpd %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <2 x double> @llvm.x86.avx512.mask.cvtuqq2pd.128(<2 x i64> %x0, <2 x double> %x1, i8 %x2) + %res1 = call <2 x double> @llvm.x86.avx512.mask.cvtuqq2pd.128(<2 x i64> %x0, <2 x double> %x1, i8 -1) + %res2 = fadd <2 x double> %res, %res1 + ret <2 x double> %res2 +} + +declare <4 x double> @llvm.x86.avx512.mask.cvtuqq2pd.256(<4 x i64>, <4 x double>, i8) + +define <4 x double>@test_int_x86_avx512_mask_cvt_uqq2pd_256(<4 x i64> %x0, <4 x double> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_uqq2pd_256: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvtuqq2pd %ymm0, %ymm1 {%k1} +; CHECK-NEXT: vcvtuqq2pd %ymm0, %ymm0 +; CHECK-NEXT: vaddpd %ymm0, %ymm1, %ymm0 +; CHECK-NEXT: retq + %res = call <4 x double> @llvm.x86.avx512.mask.cvtuqq2pd.256(<4 x i64> %x0, <4 x double> %x1, i8 %x2) + %res1 = call <4 x double> @llvm.x86.avx512.mask.cvtuqq2pd.256(<4 x i64> %x0, <4 x double> %x1, i8 -1) + %res2 = fadd <4 x double> %res, %res1 + ret <4 x double> %res2 +} + +declare <4 x float> @llvm.x86.avx512.mask.cvtuqq2ps.128(<2 x i64>, <4 x float>, i8) + +define <4 x float>@test_int_x86_avx512_mask_cvt_uqq2ps_128(<2 x i64> %x0, <4 x float> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_uqq2ps_128: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvtuqq2ps %xmm0, %xmm1 {%k1} +; CHECK-NEXT: vcvtuqq2ps %xmm0, %xmm0 +; CHECK-NEXT: vaddps %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <4 x float> @llvm.x86.avx512.mask.cvtuqq2ps.128(<2 x i64> %x0, <4 x float> %x1, i8 %x2) + %res1 = call <4 x float> @llvm.x86.avx512.mask.cvtuqq2ps.128(<2 x i64> %x0, <4 x float> %x1, i8 -1) + %res2 = fadd <4 x float> %res, %res1 + ret <4 x float> %res2 +} + +declare <4 x float> @llvm.x86.avx512.mask.cvtuqq2ps.256(<4 x i64>, <4 x float>, i8) + +define <4 x float>@test_int_x86_avx512_mask_cvt_uqq2ps_256(<4 x i64> %x0, <4 x float> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_uqq2ps_256: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvtuqq2ps %ymm0, %xmm1 {%k1} +; CHECK-NEXT: vcvtuqq2ps %ymm0, %xmm0 +; CHECK-NEXT: vaddps %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <4 x float> @llvm.x86.avx512.mask.cvtuqq2ps.256(<4 x i64> %x0, <4 x float> %x1, i8 %x2) + %res1 = call <4 x float> @llvm.x86.avx512.mask.cvtuqq2ps.256(<4 x i64> %x0, <4 x float> %x1, i8 -1) + %res2 = fadd <4 x float> %res, %res1 + ret <4 x float> %res2 +} + +declare <2 x i64> @llvm.x86.avx512.mask.cvttps2uqq.128(<4 x float>, <2 x i64>, i8) + +define <2 x i64>@test_int_x86_avx512_mask_cvtt_ps2uqq_128(<4 x float> %x0, <2 x i64> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvtt_ps2uqq_128: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvttps2uqq %xmm0, %xmm1 {%k1} +; CHECK-NEXT: vcvttps2uqq %xmm0, %xmm0 +; CHECK-NEXT: vpaddq %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <2 x i64> @llvm.x86.avx512.mask.cvttps2uqq.128(<4 x float> %x0, <2 x i64> %x1, i8 %x2) + %res1 = call <2 x i64> @llvm.x86.avx512.mask.cvttps2uqq.128(<4 x float> %x0, <2 x i64> %x1, i8 -1) + %res2 = add <2 x i64> %res, %res1 + ret <2 x i64> %res2 +} + +declare <4 x i64> @llvm.x86.avx512.mask.cvttps2uqq.256(<4 x float>, <4 x i64>, i8) + +define <4 x i64>@test_int_x86_avx512_mask_cvtt_ps2uqq_256(<4 x float> %x0, <4 x i64> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvtt_ps2uqq_256: +; CHECK: ## BB#0: +; CHECK-NEXT: kmovb %edi, %k1 +; CHECK-NEXT: vcvttps2uqq %xmm0, %ymm1 {%k1} +; CHECK-NEXT: vcvttps2uqq %xmm0, %ymm0 +; CHECK-NEXT: vpaddq %ymm0, %ymm1, %ymm0 +; CHECK-NEXT: retq + %res = call <4 x i64> @llvm.x86.avx512.mask.cvttps2uqq.256(<4 x float> %x0, <4 x i64> %x1, i8 %x2) + %res1 = call <4 x i64> @llvm.x86.avx512.mask.cvttps2uqq.256(<4 x float> %x0, <4 x i64> %x1, i8 -1) + %res2 = add <4 x i64> %res, %res1 + ret <4 x i64> %res2 +} Index: test/CodeGen/X86/avx512vl-intrinsics.ll =================================================================== --- test/CodeGen/X86/avx512vl-intrinsics.ll +++ test/CodeGen/X86/avx512vl-intrinsics.ll @@ -3003,4 +3003,481 @@ %res1 = call <8 x float> @llvm.x86.avx512.mask.scalef.ps.256(<8 x float> %x0, <8 x float> %x1, <8 x float> %x2, i8 -1) %res2 = fadd <8 x float> %res, %res1 ret <8 x float> %res2 -} \ No newline at end of file +} + +declare <2 x double> @llvm.x86.avx512.mask.cvtdq2pd.128(<4 x i32>, <2 x double>, i8) + +define <2 x double>@test_int_x86_avx512_mask_cvt_dq2pd_128(<4 x i32> %x0, <2 x double> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_dq2pd_128: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvtdq2pd %xmm0, %xmm1 {%k1} +; CHECK-NEXT: vcvtdq2pd %xmm0, %xmm0 +; CHECK-NEXT: vaddpd %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <2 x double> @llvm.x86.avx512.mask.cvtdq2pd.128(<4 x i32> %x0, <2 x double> %x1, i8 %x2) + %res1 = call <2 x double> @llvm.x86.avx512.mask.cvtdq2pd.128(<4 x i32> %x0, <2 x double> %x1, i8 -1) + %res2 = fadd <2 x double> %res, %res1 + ret <2 x double> %res2 +} + +declare <4 x double> @llvm.x86.avx512.mask.cvtdq2pd.256(<4 x i32>, <4 x double>, i8) + +define <4 x double>@test_int_x86_avx512_mask_cvt_dq2pd_256(<4 x i32> %x0, <4 x double> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_dq2pd_256: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvtdq2pd %xmm0, %ymm1 {%k1} +; CHECK-NEXT: vcvtdq2pd %xmm0, %ymm0 +; CHECK-NEXT: vaddpd %ymm0, %ymm1, %ymm0 +; CHECK-NEXT: retq + %res = call <4 x double> @llvm.x86.avx512.mask.cvtdq2pd.256(<4 x i32> %x0, <4 x double> %x1, i8 %x2) + %res1 = call <4 x double> @llvm.x86.avx512.mask.cvtdq2pd.256(<4 x i32> %x0, <4 x double> %x1, i8 -1) + %res2 = fadd <4 x double> %res, %res1 + ret <4 x double> %res2 +} + +declare <4 x float> @llvm.x86.avx512.mask.cvtdq2ps.128(<4 x i32>, <4 x float>, i8) + +define <4 x float>@test_int_x86_avx512_mask_cvt_dq2ps_128(<4 x i32> %x0, <4 x float> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_dq2ps_128: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvtdq2ps %xmm0, %xmm1 {%k1} +; CHECK-NEXT: vcvtdq2ps %xmm0, %xmm0 +; CHECK-NEXT: vaddps %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <4 x float> @llvm.x86.avx512.mask.cvtdq2ps.128(<4 x i32> %x0, <4 x float> %x1, i8 %x2) + %res1 = call <4 x float> @llvm.x86.avx512.mask.cvtdq2ps.128(<4 x i32> %x0, <4 x float> %x1, i8 -1) + %res2 = fadd <4 x float> %res, %res1 + ret <4 x float> %res2 +} + +declare <8 x float> @llvm.x86.avx512.mask.cvtdq2ps.256(<8 x i32>, <8 x float>, i8) + +define <8 x float>@test_int_x86_avx512_mask_cvt_dq2ps_256(<8 x i32> %x0, <8 x float> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_dq2ps_256: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvtdq2ps %ymm0, %ymm1 {%k1} +; CHECK-NEXT: vcvtdq2ps %ymm0, %ymm0 +; CHECK-NEXT: vaddps %ymm0, %ymm1, %ymm0 +; CHECK-NEXT: retq + %res = call <8 x float> @llvm.x86.avx512.mask.cvtdq2ps.256(<8 x i32> %x0, <8 x float> %x1, i8 %x2) + %res1 = call <8 x float> @llvm.x86.avx512.mask.cvtdq2ps.256(<8 x i32> %x0, <8 x float> %x1, i8 -1) + %res2 = fadd <8 x float> %res, %res1 + ret <8 x float> %res2 +} + +declare <4 x i32> @llvm.x86.avx512.mask.cvtpd2dq.128(<2 x double>, <4 x i32>, i8) + +define <4 x i32>@test_int_x86_avx512_mask_cvt_pd2dq_128(<2 x double> %x0, <4 x i32> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_pd2dq_128: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvtpd2dq %xmm0, %xmm1 {%k1} +; CHECK-NEXT: vcvtpd2dq %xmm0, %xmm0 +; CHECK-NEXT: vpaddd %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <4 x i32> @llvm.x86.avx512.mask.cvtpd2dq.128(<2 x double> %x0, <4 x i32> %x1, i8 %x2) + %res1 = call <4 x i32> @llvm.x86.avx512.mask.cvtpd2dq.128(<2 x double> %x0, <4 x i32> %x1, i8 -1) + %res2 = add <4 x i32> %res, %res1 + ret <4 x i32> %res2 +} + +declare <4 x i32> @llvm.x86.avx512.mask.cvtpd2dq.256(<4 x double>, <4 x i32>, i8) + +define <4 x i32>@test_int_x86_avx512_mask_cvt_pd2dq_256(<4 x double> %x0, <4 x i32> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_pd2dq_256: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvtpd2dq %ymm0, %xmm1 {%k1} +; CHECK-NEXT: vcvtpd2dq %ymm0, %xmm0 +; CHECK-NEXT: vpaddd %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <4 x i32> @llvm.x86.avx512.mask.cvtpd2dq.256(<4 x double> %x0, <4 x i32> %x1, i8 %x2) + %res1 = call <4 x i32> @llvm.x86.avx512.mask.cvtpd2dq.256(<4 x double> %x0, <4 x i32> %x1, i8 -1) + %res2 = add <4 x i32> %res, %res1 + ret <4 x i32> %res2 +} + +declare <4 x float> @llvm.x86.avx512.mask.cvtpd2ps.256(<4 x double>, <4 x float>, i8) + +define <4 x float>@test_int_x86_avx512_mask_cvt_pd2ps_256(<4 x double> %x0, <4 x float> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_pd2ps_256: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvtpd2ps %ymm0, %xmm1 {%k1} +; CHECK-NEXT: vcvtpd2ps %ymm0, %xmm0 +; CHECK-NEXT: vaddps %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <4 x float> @llvm.x86.avx512.mask.cvtpd2ps.256(<4 x double> %x0, <4 x float> %x1, i8 %x2) + %res1 = call <4 x float> @llvm.x86.avx512.mask.cvtpd2ps.256(<4 x double> %x0, <4 x float> %x1, i8 -1) + %res2 = fadd <4 x float> %res, %res1 + ret <4 x float> %res2 +} + +declare <4 x float> @llvm.x86.avx512.mask.cvtpd2ps(<2 x double>, <4 x float>, i8) + +define <4 x float>@test_int_x86_avx512_mask_cvt_pd2ps(<2 x double> %x0, <4 x float> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_pd2ps: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvtpd2ps %xmm0, %xmm1 {%k1} +; CHECK-NEXT: vcvtpd2ps %xmm0, %xmm0 +; CHECK-NEXT: vaddps %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <4 x float> @llvm.x86.avx512.mask.cvtpd2ps(<2 x double> %x0, <4 x float> %x1, i8 %x2) + %res1 = call <4 x float> @llvm.x86.avx512.mask.cvtpd2ps(<2 x double> %x0, <4 x float> %x1, i8 -1) + %res2 = fadd <4 x float> %res, %res1 + ret <4 x float> %res2 +} + +declare <4 x i32> @llvm.x86.avx512.mask.cvtpd2udq.128(<2 x double>, <4 x i32>, i8) + +define <4 x i32>@test_int_x86_avx512_mask_cvt_pd2udq_128(<2 x double> %x0, <4 x i32> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_pd2udq_128: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvtpd2udq %xmm0, %xmm1 {%k1} +; CHECK-NEXT: vcvtpd2udq %xmm0, %xmm0 +; CHECK-NEXT: vpaddd %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <4 x i32> @llvm.x86.avx512.mask.cvtpd2udq.128(<2 x double> %x0, <4 x i32> %x1, i8 %x2) + %res1 = call <4 x i32> @llvm.x86.avx512.mask.cvtpd2udq.128(<2 x double> %x0, <4 x i32> %x1, i8 -1) + %res2 = add <4 x i32> %res, %res1 + ret <4 x i32> %res2 +} + +declare <4 x i32> @llvm.x86.avx512.mask.cvtpd2udq.256(<4 x double>, <4 x i32>, i8) + +define <4 x i32>@test_int_x86_avx512_mask_cvt_pd2udq_256(<4 x double> %x0, <4 x i32> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_pd2udq_256: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvtpd2udq %ymm0, %xmm1 {%k1} +; CHECK-NEXT: vcvtpd2udq %ymm0, %xmm0 +; CHECK-NEXT: vpaddd %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <4 x i32> @llvm.x86.avx512.mask.cvtpd2udq.256(<4 x double> %x0, <4 x i32> %x1, i8 %x2) + %res1 = call <4 x i32> @llvm.x86.avx512.mask.cvtpd2udq.256(<4 x double> %x0, <4 x i32> %x1, i8 -1) + %res2 = add <4 x i32> %res, %res1 + ret <4 x i32> %res2 +} + +declare <4 x i32> @llvm.x86.avx512.mask.cvtps2dq.128(<4 x float>, <4 x i32>, i8) + +define <4 x i32>@test_int_x86_avx512_mask_cvt_ps2dq_128(<4 x float> %x0, <4 x i32> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_ps2dq_128: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvtps2dq %xmm0, %xmm1 {%k1} +; CHECK-NEXT: vcvtps2dq %xmm0, %xmm0 +; CHECK-NEXT: vpaddd %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <4 x i32> @llvm.x86.avx512.mask.cvtps2dq.128(<4 x float> %x0, <4 x i32> %x1, i8 %x2) + %res1 = call <4 x i32> @llvm.x86.avx512.mask.cvtps2dq.128(<4 x float> %x0, <4 x i32> %x1, i8 -1) + %res2 = add <4 x i32> %res, %res1 + ret <4 x i32> %res2 +} + +declare <8 x i32> @llvm.x86.avx512.mask.cvtps2dq.256(<8 x float>, <8 x i32>, i8) + +define <8 x i32>@test_int_x86_avx512_mask_cvt_ps2dq_256(<8 x float> %x0, <8 x i32> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_ps2dq_256: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvtps2dq %ymm0, %ymm1 {%k1} +; CHECK-NEXT: vcvtps2dq %ymm0, %ymm0 +; CHECK-NEXT: vpaddd %ymm0, %ymm1, %ymm0 +; CHECK-NEXT: retq + %res = call <8 x i32> @llvm.x86.avx512.mask.cvtps2dq.256(<8 x float> %x0, <8 x i32> %x1, i8 %x2) + %res1 = call <8 x i32> @llvm.x86.avx512.mask.cvtps2dq.256(<8 x float> %x0, <8 x i32> %x1, i8 -1) + %res2 = add <8 x i32> %res, %res1 + ret <8 x i32> %res2 +} + +declare <2 x double> @llvm.x86.avx512.mask.cvtps2pd.128(<4 x float>, <2 x double>, i8) + +define <2 x double>@test_int_x86_avx512_mask_cvt_ps2pd_128(<4 x float> %x0, <2 x double> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_ps2pd_128: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvtps2pd %xmm0, %xmm1 {%k1} +; CHECK-NEXT: vcvtps2pd %xmm0, %xmm0 +; CHECK-NEXT: vaddpd %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <2 x double> @llvm.x86.avx512.mask.cvtps2pd.128(<4 x float> %x0, <2 x double> %x1, i8 %x2) + %res1 = call <2 x double> @llvm.x86.avx512.mask.cvtps2pd.128(<4 x float> %x0, <2 x double> %x1, i8 -1) + %res2 = fadd <2 x double> %res, %res1 + ret <2 x double> %res2 +} + +declare <4 x double> @llvm.x86.avx512.mask.cvtps2pd.256(<4 x float>, <4 x double>, i8) + +define <4 x double>@test_int_x86_avx512_mask_cvt_ps2pd_256(<4 x float> %x0, <4 x double> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_ps2pd_256: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvtps2pd %xmm0, %ymm1 {%k1} +; CHECK-NEXT: vcvtps2pd %xmm0, %ymm0 +; CHECK-NEXT: vaddpd %ymm0, %ymm1, %ymm0 +; CHECK-NEXT: retq + %res = call <4 x double> @llvm.x86.avx512.mask.cvtps2pd.256(<4 x float> %x0, <4 x double> %x1, i8 %x2) + %res1 = call <4 x double> @llvm.x86.avx512.mask.cvtps2pd.256(<4 x float> %x0, <4 x double> %x1, i8 -1) + %res2 = fadd <4 x double> %res, %res1 + ret <4 x double> %res2 +} + +declare <4 x i32> @llvm.x86.avx512.mask.cvtps2udq.128(<4 x float>, <4 x i32>, i8) + +define <4 x i32>@test_int_x86_avx512_mask_cvt_ps2udq_128(<4 x float> %x0, <4 x i32> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_ps2udq_128: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvtps2udq %xmm0, %xmm1 {%k1} +; CHECK-NEXT: vcvtps2udq %xmm0, %xmm0 +; CHECK-NEXT: vpaddd %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <4 x i32> @llvm.x86.avx512.mask.cvtps2udq.128(<4 x float> %x0, <4 x i32> %x1, i8 %x2) + %res1 = call <4 x i32> @llvm.x86.avx512.mask.cvtps2udq.128(<4 x float> %x0, <4 x i32> %x1, i8 -1) + %res2 = add <4 x i32> %res, %res1 + ret <4 x i32> %res2 +} + +declare <8 x i32> @llvm.x86.avx512.mask.cvtps2udq.256(<8 x float>, <8 x i32>, i8) + +define <8 x i32>@test_int_x86_avx512_mask_cvt_ps2udq_256(<8 x float> %x0, <8 x i32> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_ps2udq_256: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvtps2udq %ymm0, %ymm1 {%k1} +; CHECK-NEXT: vcvtps2udq %ymm0, %ymm0 +; CHECK-NEXT: vpaddd %ymm0, %ymm1, %ymm0 +; CHECK-NEXT: retq + %res = call <8 x i32> @llvm.x86.avx512.mask.cvtps2udq.256(<8 x float> %x0, <8 x i32> %x1, i8 %x2) + %res1 = call <8 x i32> @llvm.x86.avx512.mask.cvtps2udq.256(<8 x float> %x0, <8 x i32> %x1, i8 -1) + %res2 = add <8 x i32> %res, %res1 + ret <8 x i32> %res2 +} + +declare <4 x i32> @llvm.x86.avx512.mask.cvttpd2dq.128(<2 x double>, <4 x i32>, i8) + +define <4 x i32>@test_int_x86_avx512_mask_cvtt_pd2dq_128(<2 x double> %x0, <4 x i32> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvtt_pd2dq_128: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvttpd2dq %xmm0, %xmm1 {%k1} +; CHECK-NEXT: vcvttpd2dq %xmm0, %xmm0 +; CHECK-NEXT: vpaddd %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <4 x i32> @llvm.x86.avx512.mask.cvttpd2dq.128(<2 x double> %x0, <4 x i32> %x1, i8 %x2) + %res1 = call <4 x i32> @llvm.x86.avx512.mask.cvttpd2dq.128(<2 x double> %x0, <4 x i32> %x1, i8 -1) + %res2 = add <4 x i32> %res, %res1 + ret <4 x i32> %res2 +} + +declare <4 x i32> @llvm.x86.avx512.mask.cvttpd2dq.256(<4 x double>, <4 x i32>, i8) + +define <4 x i32>@test_int_x86_avx512_mask_cvtt_pd2dq_256(<4 x double> %x0, <4 x i32> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvtt_pd2dq_256: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvttpd2dq %ymm0, %xmm1 {%k1} +; CHECK-NEXT: vcvttpd2dq %ymm0, %xmm0 +; CHECK-NEXT: vpaddd %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <4 x i32> @llvm.x86.avx512.mask.cvttpd2dq.256(<4 x double> %x0, <4 x i32> %x1, i8 %x2) + %res1 = call <4 x i32> @llvm.x86.avx512.mask.cvttpd2dq.256(<4 x double> %x0, <4 x i32> %x1, i8 -1) + %res2 = add <4 x i32> %res, %res1 + ret <4 x i32> %res2 +} + +declare <4 x i32> @llvm.x86.avx512.mask.cvttpd2udq.128(<2 x double>, <4 x i32>, i8) + +define <4 x i32>@test_int_x86_avx512_mask_cvtt_pd2udq_128(<2 x double> %x0, <4 x i32> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvtt_pd2udq_128: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvttpd2udq %xmm0, %xmm1 {%k1} +; CHECK-NEXT: vcvttpd2udq %xmm0, %xmm0 +; CHECK-NEXT: vpaddd %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <4 x i32> @llvm.x86.avx512.mask.cvttpd2udq.128(<2 x double> %x0, <4 x i32> %x1, i8 %x2) + %res1 = call <4 x i32> @llvm.x86.avx512.mask.cvttpd2udq.128(<2 x double> %x0, <4 x i32> %x1, i8 -1) + %res2 = add <4 x i32> %res, %res1 + ret <4 x i32> %res2 +} + +declare <4 x i32> @llvm.x86.avx512.mask.cvttpd2udq.256(<4 x double>, <4 x i32>, i8) + +define <4 x i32>@test_int_x86_avx512_mask_cvtt_pd2udq_256(<4 x double> %x0, <4 x i32> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvtt_pd2udq_256: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvttpd2udq %ymm0, %xmm1 {%k1} +; CHECK-NEXT: vcvttpd2udq %ymm0, %xmm0 +; CHECK-NEXT: vpaddd %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <4 x i32> @llvm.x86.avx512.mask.cvttpd2udq.256(<4 x double> %x0, <4 x i32> %x1, i8 %x2) + %res1 = call <4 x i32> @llvm.x86.avx512.mask.cvttpd2udq.256(<4 x double> %x0, <4 x i32> %x1, i8 -1) + %res2 = add <4 x i32> %res, %res1 + ret <4 x i32> %res2 +} + +declare <4 x i32> @llvm.x86.avx512.mask.cvttps2dq.128(<4 x float>, <4 x i32>, i8) + +define <4 x i32>@test_int_x86_avx512_mask_cvtt_ps2dq_128(<4 x float> %x0, <4 x i32> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvtt_ps2dq_128: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvttps2dq %xmm0, %xmm1 {%k1} +; CHECK-NEXT: vcvttps2dq %xmm0, %xmm0 +; CHECK-NEXT: vpaddd %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <4 x i32> @llvm.x86.avx512.mask.cvttps2dq.128(<4 x float> %x0, <4 x i32> %x1, i8 %x2) + %res1 = call <4 x i32> @llvm.x86.avx512.mask.cvttps2dq.128(<4 x float> %x0, <4 x i32> %x1, i8 -1) + %res2 = add <4 x i32> %res, %res1 + ret <4 x i32> %res2 +} + +declare <8 x i32> @llvm.x86.avx512.mask.cvttps2dq.256(<8 x float>, <8 x i32>, i8) + +define <8 x i32>@test_int_x86_avx512_mask_cvtt_ps2dq_256(<8 x float> %x0, <8 x i32> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvtt_ps2dq_256: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvttps2dq %ymm0, %ymm1 {%k1} +; CHECK-NEXT: vcvttps2dq %ymm0, %ymm0 +; CHECK-NEXT: vpaddd %ymm0, %ymm1, %ymm0 +; CHECK-NEXT: retq + %res = call <8 x i32> @llvm.x86.avx512.mask.cvttps2dq.256(<8 x float> %x0, <8 x i32> %x1, i8 %x2) + %res1 = call <8 x i32> @llvm.x86.avx512.mask.cvttps2dq.256(<8 x float> %x0, <8 x i32> %x1, i8 -1) + %res2 = add <8 x i32> %res, %res1 + ret <8 x i32> %res2 +} + +declare <4 x i32> @llvm.x86.avx512.mask.cvttps2udq.128(<4 x float>, <4 x i32>, i8) + +define <4 x i32>@test_int_x86_avx512_mask_cvtt_ps2udq_128(<4 x float> %x0, <4 x i32> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvtt_ps2udq_128: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvttps2udq %xmm0, %xmm1 {%k1} +; CHECK-NEXT: vcvttps2udq %xmm0, %xmm0 +; CHECK-NEXT: vpaddd %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <4 x i32> @llvm.x86.avx512.mask.cvttps2udq.128(<4 x float> %x0, <4 x i32> %x1, i8 %x2) + %res1 = call <4 x i32> @llvm.x86.avx512.mask.cvttps2udq.128(<4 x float> %x0, <4 x i32> %x1, i8 -1) + %res2 = add <4 x i32> %res, %res1 + ret <4 x i32> %res2 +} + +declare <8 x i32> @llvm.x86.avx512.mask.cvttps2udq.256(<8 x float>, <8 x i32>, i8) + +define <8 x i32>@test_int_x86_avx512_mask_cvtt_ps2udq_256(<8 x float> %x0, <8 x i32> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvtt_ps2udq_256: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvttps2udq %ymm0, %ymm1 {%k1} +; CHECK-NEXT: vcvttps2udq %ymm0, %ymm0 +; CHECK-NEXT: vpaddd %ymm0, %ymm1, %ymm0 +; CHECK-NEXT: retq + %res = call <8 x i32> @llvm.x86.avx512.mask.cvttps2udq.256(<8 x float> %x0, <8 x i32> %x1, i8 %x2) + %res1 = call <8 x i32> @llvm.x86.avx512.mask.cvttps2udq.256(<8 x float> %x0, <8 x i32> %x1, i8 -1) + %res2 = add <8 x i32> %res, %res1 + ret <8 x i32> %res2 +} + +declare <2 x double> @llvm.x86.avx512.mask.cvtudq2pd.128(<4 x i32>, <2 x double>, i8) + +define <2 x double>@test_int_x86_avx512_mask_cvt_udq2pd_128(<4 x i32> %x0, <2 x double> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_udq2pd_128: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvtudq2pd %xmm0, %xmm1 {%k1} +; CHECK-NEXT: vcvtudq2pd %xmm0, %xmm0 +; CHECK-NEXT: vaddpd %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <2 x double> @llvm.x86.avx512.mask.cvtudq2pd.128(<4 x i32> %x0, <2 x double> %x1, i8 %x2) + %res1 = call <2 x double> @llvm.x86.avx512.mask.cvtudq2pd.128(<4 x i32> %x0, <2 x double> %x1, i8 -1) + %res2 = fadd <2 x double> %res, %res1 + ret <2 x double> %res2 +} + +declare <4 x double> @llvm.x86.avx512.mask.cvtudq2pd.256(<4 x i32>, <4 x double>, i8) + +define <4 x double>@test_int_x86_avx512_mask_cvt_udq2pd_256(<4 x i32> %x0, <4 x double> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_udq2pd_256: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvtudq2pd %xmm0, %ymm1 {%k1} +; CHECK-NEXT: vcvtudq2pd %xmm0, %ymm0 +; CHECK-NEXT: vaddpd %ymm0, %ymm1, %ymm0 +; CHECK-NEXT: retq + %res = call <4 x double> @llvm.x86.avx512.mask.cvtudq2pd.256(<4 x i32> %x0, <4 x double> %x1, i8 %x2) + %res1 = call <4 x double> @llvm.x86.avx512.mask.cvtudq2pd.256(<4 x i32> %x0, <4 x double> %x1, i8 -1) + %res2 = fadd <4 x double> %res, %res1 + ret <4 x double> %res2 +} + +declare <4 x float> @llvm.x86.avx512.mask.cvtudq2ps.128(<4 x i32>, <4 x float>, i8) + +define <4 x float>@test_int_x86_avx512_mask_cvt_udq2ps_128(<4 x i32> %x0, <4 x float> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_udq2ps_128: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvtudq2ps %xmm0, %xmm1 {%k1} +; CHECK-NEXT: vcvtudq2ps %xmm0, %xmm0 +; CHECK-NEXT: vaddps %xmm0, %xmm1, %xmm0 +; CHECK-NEXT: retq + %res = call <4 x float> @llvm.x86.avx512.mask.cvtudq2ps.128(<4 x i32> %x0, <4 x float> %x1, i8 %x2) + %res1 = call <4 x float> @llvm.x86.avx512.mask.cvtudq2ps.128(<4 x i32> %x0, <4 x float> %x1, i8 -1) + %res2 = fadd <4 x float> %res, %res1 + ret <4 x float> %res2 +} + +declare <8 x float> @llvm.x86.avx512.mask.cvtudq2ps.256(<8 x i32>, <8 x float>, i8) + +define <8 x float>@test_int_x86_avx512_mask_cvt_udq2ps_256(<8 x i32> %x0, <8 x float> %x1, i8 %x2) { +; CHECK-LABEL: test_int_x86_avx512_mask_cvt_udq2ps_256: +; CHECK: ## BB#0: +; CHECK-NEXT: movzbl %dil, %eax +; CHECK-NEXT: kmovw %eax, %k1 +; CHECK-NEXT: vcvtudq2ps %ymm0, %ymm1 {%k1} +; CHECK-NEXT: vcvtudq2ps %ymm0, %ymm0 +; CHECK-NEXT: vaddps %ymm0, %ymm1, %ymm0 +; CHECK-NEXT: retq + %res = call <8 x float> @llvm.x86.avx512.mask.cvtudq2ps.256(<8 x i32> %x0, <8 x float> %x1, i8 %x2) + %res1 = call <8 x float> @llvm.x86.avx512.mask.cvtudq2ps.256(<8 x i32> %x0, <8 x float> %x1, i8 -1) + %res2 = fadd <8 x float> %res, %res1 + ret <8 x float> %res2 +} +