llvm_i32_ty], [IntrNoMem]>;
def int_x86_sse2_psrl_w : GCCBuiltin<"__builtin_ia32_psrlw128">,
Intrinsic<[llvm_v8i16_ty, llvm_v8i16_ty,
- llvm_v4i32_ty], [IntrNoMem]>;
+ llvm_v8i16_ty], [IntrNoMem]>;
def int_x86_sse2_psrl_d : GCCBuiltin<"__builtin_ia32_psrld128">,
Intrinsic<[llvm_v4i32_ty, llvm_v4i32_ty,
llvm_v4i32_ty], [IntrNoMem]>;
def int_x86_sse2_psrl_q : GCCBuiltin<"__builtin_ia32_psrlq128">,
Intrinsic<[llvm_v2i64_ty, llvm_v2i64_ty,
- llvm_v4i32_ty], [IntrNoMem]>;
+ llvm_v2i64_ty], [IntrNoMem]>;
def int_x86_sse2_psrl_dq : GCCBuiltin<"__builtin_ia32_psrldqi128">,
Intrinsic<[llvm_v2i64_ty, llvm_v2i64_ty,
llvm_i32_ty], [IntrNoMem]>;
llvm_v2i64_ty, llvm_i32_ty], [IntrNoMem]>;
}
+//===----------------------------------------------------------------------===//
+// SSE4.1
+
+// FP rounding ops
+let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
+ def int_x86_sse41_round_ss : GCCBuiltin<"__builtin_ia32_roundss">,
+ Intrinsic<[llvm_v4f32_ty, llvm_v4f32_ty,
+ llvm_i32_ty], [IntrNoMem]>;
+ def int_x86_sse41_round_ps : GCCBuiltin<"__builtin_ia32_roundps">,
+ Intrinsic<[llvm_v4f32_ty, llvm_v4f32_ty,
+ llvm_i32_ty], [IntrNoMem]>;
+ def int_x86_sse41_round_sd : GCCBuiltin<"__builtin_ia32_roundsd">,
+ Intrinsic<[llvm_v2f64_ty, llvm_v2f64_ty,
+ llvm_i32_ty], [IntrNoMem]>;
+ def int_x86_sse41_round_pd : GCCBuiltin<"__builtin_ia32_roundpd">,
+ Intrinsic<[llvm_v2f64_ty, llvm_v2f64_ty,
+ llvm_i32_ty], [IntrNoMem]>;
+}
+
+// Vector sign and zero extend
+let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
+ def int_x86_sse41_pmovsxbd : GCCBuiltin<"__builtin_ia32_pmovsxbd128">,
+ Intrinsic<[llvm_v4i32_ty, llvm_v16i8_ty]>;
+ def int_x86_sse41_pmovsxbq : GCCBuiltin<"__builtin_ia32_pmovsxbq128">,
+ Intrinsic<[llvm_v2i64_ty, llvm_v16i8_ty]>;
+ def int_x86_sse41_pmovsxbw : GCCBuiltin<"__builtin_ia32_pmovsxbw128">,
+ Intrinsic<[llvm_v8i16_ty, llvm_v16i8_ty]>;
+ def int_x86_sse41_pmovsxdq : GCCBuiltin<"__builtin_ia32_pmovsxdq128">,
+ Intrinsic<[llvm_v2i64_ty, llvm_v4i32_ty]>;
+ def int_x86_sse41_pmovsxwd : GCCBuiltin<"__builtin_ia32_pmovsxwd128">,
+ Intrinsic<[llvm_v4i32_ty, llvm_v8i16_ty]>;
+ def int_x86_sse41_pmovsxwq : GCCBuiltin<"__builtin_ia32_pmovsxwq128">,
+ Intrinsic<[llvm_v2i64_ty, llvm_v8i16_ty]>;
+ def int_x86_sse41_pmovzxbd : GCCBuiltin<"__builtin_ia32_pmovzxbd128">,
+ Intrinsic<[llvm_v4i32_ty, llvm_v16i8_ty]>;
+ def int_x86_sse41_pmovzxbq : GCCBuiltin<"__builtin_ia32_pmovzxbq128">,
+ Intrinsic<[llvm_v2i64_ty, llvm_v16i8_ty]>;
+ def int_x86_sse41_pmovzxbw : GCCBuiltin<"__builtin_ia32_pmovzxbw128">,
+ Intrinsic<[llvm_v8i16_ty, llvm_v16i8_ty]>;
+ def int_x86_sse41_pmovzxdq : GCCBuiltin<"__builtin_ia32_pmovzxdq128">,
+ Intrinsic<[llvm_v2i64_ty, llvm_v4i32_ty]>;
+ def int_x86_sse41_pmovzxwd : GCCBuiltin<"__builtin_ia32_pmovzxwd128">,
+ Intrinsic<[llvm_v4i32_ty, llvm_v8i16_ty]>;
+ def int_x86_sse41_pmovzxwq : GCCBuiltin<"__builtin_ia32_pmovzxwq128">,
+ Intrinsic<[llvm_v2i64_ty, llvm_v8i16_ty]>;
+}
+
+// Vector min element
+let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
+ def int_x86_sse41_phminposuw : GCCBuiltin<"__builtin_ia32_phminposuw128">,
+ Intrinsic<[llvm_v8i16_ty, llvm_v8i16_ty]>;
+}
+
+// Vector compare, min, max
+let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
+ def int_x86_sse41_pcmpeqq : GCCBuiltin<"__builtin_ia32_pcmpeqq">,
+ Intrinsic<[llvm_v2i64_ty, llvm_v2i64_ty, llvm_v2i64_ty]>;
+ def int_x86_sse41_pmaxsb : GCCBuiltin<"__builtin_ia32_pmaxsb128">,
+ Intrinsic<[llvm_v16i8_ty, llvm_v16i8_ty, llvm_v16i8_ty]>;
+ def int_x86_sse41_pmaxsd : GCCBuiltin<"__builtin_ia32_pmaxsd128">,
+ Intrinsic<[llvm_v4i32_ty, llvm_v4i32_ty, llvm_v4i32_ty]>;
+ def int_x86_sse41_pmaxud : GCCBuiltin<"__builtin_ia32_pmaxud128">,
+ Intrinsic<[llvm_v4i32_ty, llvm_v4i32_ty, llvm_v4i32_ty]>;
+ def int_x86_sse41_pmaxuw : GCCBuiltin<"__builtin_ia32_pmaxuw128">,
+ Intrinsic<[llvm_v8i16_ty, llvm_v8i16_ty, llvm_v8i16_ty]>;
+ def int_x86_sse41_pminsb : GCCBuiltin<"__builtin_ia32_pminsb128">,
+ Intrinsic<[llvm_v16i8_ty, llvm_v16i8_ty, llvm_v16i8_ty]>;
+ def int_x86_sse41_pminsd : GCCBuiltin<"__builtin_ia32_pminsd128">,
+ Intrinsic<[llvm_v4i32_ty, llvm_v4i32_ty, llvm_v4i32_ty]>;
+ def int_x86_sse41_pminud : GCCBuiltin<"__builtin_ia32_pminud128">,
+ Intrinsic<[llvm_v4i32_ty, llvm_v4i32_ty, llvm_v4i32_ty]>;
+ def int_x86_sse41_pminuw : GCCBuiltin<"__builtin_ia32_pminuw128">,
+ Intrinsic<[llvm_v8i16_ty, llvm_v8i16_ty, llvm_v8i16_ty]>;
+}
+
+// Vector pack
+let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
+ def int_x86_sse41_packusdw : GCCBuiltin<"__builtin_ia32_packusdw128">,
+ Intrinsic<[llvm_v8i16_ty, llvm_v4i32_ty, llvm_v4i32_ty]>;
+}
+
+// Vector multiply
+let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
+ def int_x86_sse41_pmuldq : GCCBuiltin<"__builtin_ia32_pmuldq128">,
+ Intrinsic<[llvm_v2i64_ty, llvm_v4i32_ty, llvm_v4i32_ty]>;
+ def int_x86_sse41_pmulld : GCCBuiltin<"__builtin_ia32_pmulld128">,
+ Intrinsic<[llvm_v4i32_ty, llvm_v4i32_ty, llvm_v4i32_ty]>;
+}
+
+// Vector extract
+let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
+ def int_x86_sse41_pextrb :
+ Intrinsic<[llvm_i32_ty, llvm_v16i8_ty, llvm_i32_ty]>;
+ def int_x86_sse41_pextrd :
+ Intrinsic<[llvm_i32_ty, llvm_v4i32_ty, llvm_i32_ty]>;
+ def int_x86_sse41_pextrq :
+ Intrinsic<[llvm_i64_ty, llvm_v2i64_ty, llvm_i32_ty]>;
+ def int_x86_sse41_extractps : GCCBuiltin<"__builtin_ia32_extractps128">,
+ Intrinsic<[llvm_i32_ty, llvm_v4f32_ty, llvm_i32_ty]>;
+}
+
+// Vector insert
+let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
+ def int_x86_sse41_pinsrb : GCCBuiltin<"__builtin_ia32_vec_set_v16qi">,
+ Intrinsic<[llvm_v16i8_ty, llvm_v16i8_ty, llvm_i32_ty, llvm_i32_ty]>;
+ def int_x86_sse41_pinsrd : GCCBuiltin<"__builtin_ia32_vec_set_v4si">,
+ Intrinsic<[llvm_v4i32_ty, llvm_v4i32_ty, llvm_i32_ty, llvm_i32_ty]>;
+ def int_x86_sse41_pinsrq : GCCBuiltin<"__builtin_ia32_vec_set_v2di">,
+ Intrinsic<[llvm_v2i64_ty, llvm_v2i64_ty, llvm_i64_ty, llvm_i32_ty]>;
+ def int_x86_sse41_insertps : GCCBuiltin<"__builtin_ia32_insertps128">,
+ Intrinsic<[llvm_v4f32_ty, llvm_v4f32_ty, llvm_v4f32_ty, llvm_i32_ty]>;
+}
+
+// Vector blend
+let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
+ def int_x86_sse41_pblendvb : GCCBuiltin<"__builtin_ia32_pblendvb128">,
+ Intrinsic<[llvm_v16i8_ty, llvm_v16i8_ty, llvm_v16i8_ty, llvm_v16i8_ty]>;
+ def int_x86_sse41_pblendw : GCCBuiltin<"__builtin_ia32_pblendw128">,
+ Intrinsic<[llvm_v8i16_ty, llvm_v8i16_ty, llvm_v8i16_ty, llvm_i32_ty]>;
+ def int_x86_sse41_blendpd : GCCBuiltin<"__builtin_ia32_blendpd">,
+ Intrinsic<[llvm_v2f64_ty, llvm_v2f64_ty, llvm_v2f64_ty, llvm_i32_ty]>;
+ def int_x86_sse41_blendps : GCCBuiltin<"__builtin_ia32_blendps">,
+ Intrinsic<[llvm_v4f32_ty, llvm_v4f32_ty, llvm_v4f32_ty, llvm_i32_ty]>;
+ def int_x86_sse41_blendvpd : GCCBuiltin<"__builtin_ia32_blendvpd">,
+ Intrinsic<[llvm_v2f64_ty, llvm_v2f64_ty, llvm_v2f64_ty, llvm_v2f64_ty]>;
+ def int_x86_sse41_blendvps : GCCBuiltin<"__builtin_ia32_blendvps">,
+ Intrinsic<[llvm_v4f32_ty, llvm_v4f32_ty, llvm_v4f32_ty, llvm_v4f32_ty]>;
+}
+
+// Vector dot product
+let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
+ def int_x86_sse41_dppd : GCCBuiltin<"__builtin_ia32_dppd">,
+ Intrinsic<[llvm_v2f64_ty, llvm_v2f64_ty, llvm_v2f64_ty, llvm_i32_ty]>;
+ def int_x86_sse41_dpps : GCCBuiltin<"__builtin_ia32_dpps">,
+ Intrinsic<[llvm_v4f32_ty, llvm_v4f32_ty, llvm_v4f32_ty, llvm_i32_ty]>;
+}
+
+// Vector sum of absolute differences
+let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
+ def int_x86_sse41_mpsadbw : GCCBuiltin<"__builtin_ia32_mpsadbw128">,
+ Intrinsic<[llvm_v16i8_ty, llvm_v16i8_ty, llvm_v16i8_ty, llvm_i32_ty]>;
+}
+
+// Vector sum of absolute differences
+let TargetPrefix = "x86" in { // All intrinsics start with "llvm.x86.".
+ def int_x86_sse41_movntdqa : GCCBuiltin<"__builtin_ia32_movntdqa">,
+ Intrinsic<[llvm_v2i64_ty, llvm_ptr_ty]>;
+}
+
+
//===----------------------------------------------------------------------===//
// MMX
projects / oota-llvm.git / blobdiff