Massive rewrite of MMX:

The x86_mmx type is used for MMX intrinsics, parameters and
return values where these use MMX registers, and is also
supported in load, store, and bitcast.

Only the above operations generate MMX instructions, and optimizations
do not operate on or produce MMX intrinsics. 

MMX-sized vectors <2 x i32> etc. are lowered to XMM or split into
smaller pieces.  Optimizations may occur on these forms and the
result casted back to x86_mmx, provided the result feeds into a
previous existing x86_mmx operation.

The point of all this is prevent optimizations from introducing
MMX operations, which is unsafe due to the EMMS problem.



git-svn-id: https://llvm.org/svn/llvm-project/llvm/trunk@115243 91177308-0d34-0410-b5e6-96231b3b80d8

1 files changed, 143 insertions, 167 deletions

diff --git a/include/llvm/IntrinsicsX86.td b/include/llvm/IntrinsicsX86.td
index 58ad8d40af..8650bbc8c7 100644
--- a/include/llvm/IntrinsicsX86.td
+++ b/include/llvm/IntrinsicsX86.td
@@ -130,12 +130,12 @@ let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
               Intrinsic<[llvm_v4f32_ty], [llvm_v4f32_ty,
                          llvm_i64_ty], [IntrNoMem]>;
   def int_x86_sse_cvtps2pi : GCCBuiltin<"__builtin_ia32_cvtps2pi">,
-              Intrinsic<[llvm_v2i32_ty], [llvm_v4f32_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_v4f32_ty], [IntrNoMem]>;
   def int_x86_sse_cvttps2pi: GCCBuiltin<"__builtin_ia32_cvttps2pi">,
-              Intrinsic<[llvm_v2i32_ty], [llvm_v4f32_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_v4f32_ty], [IntrNoMem]>;
   def int_x86_sse_cvtpi2ps : GCCBuiltin<"__builtin_ia32_cvtpi2ps">,
               Intrinsic<[llvm_v4f32_ty], [llvm_v4f32_ty,
-                         llvm_v2i32_ty], [IntrNoMem]>;
+                         llvm_x86mmx_ty], [IntrNoMem]>;
 }
 
 // SIMD load ops
@@ -445,11 +445,11 @@ let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
               Intrinsic<[llvm_v2f64_ty], [llvm_v2f64_ty,
                          llvm_v4f32_ty], [IntrNoMem]>;
   def int_x86_sse_cvtpd2pi : GCCBuiltin<"__builtin_ia32_cvtpd2pi">,
-              Intrinsic<[llvm_v2i32_ty], [llvm_v2f64_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_v2f64_ty], [IntrNoMem]>;
   def int_x86_sse_cvttpd2pi: GCCBuiltin<"__builtin_ia32_cvttpd2pi">,
-              Intrinsic<[llvm_v2i32_ty], [llvm_v2f64_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_v2f64_ty], [IntrNoMem]>;
   def int_x86_sse_cvtpi2pd : GCCBuiltin<"__builtin_ia32_cvtpi2pd">,
-              Intrinsic<[llvm_v2f64_ty], [llvm_v2i32_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_v2f64_ty], [llvm_x86mmx_ty], [IntrNoMem]>;
 }
 
 // SIMD load ops
@@ -563,50 +563,50 @@ let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
 // Horizontal arithmetic ops
 let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
   def int_x86_ssse3_phadd_w         : GCCBuiltin<"__builtin_ia32_phaddw">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty,
-                         llvm_v4i16_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
   def int_x86_ssse3_phadd_w_128     : GCCBuiltin<"__builtin_ia32_phaddw128">,
               Intrinsic<[llvm_v8i16_ty], [llvm_v8i16_ty,
                          llvm_v8i16_ty], [IntrNoMem]>;
 
   def int_x86_ssse3_phadd_d         : GCCBuiltin<"__builtin_ia32_phaddd">,
-              Intrinsic<[llvm_v2i32_ty], [llvm_v2i32_ty,
-                         llvm_v2i32_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
   def int_x86_ssse3_phadd_d_128     : GCCBuiltin<"__builtin_ia32_phaddd128">,
               Intrinsic<[llvm_v4i32_ty], [llvm_v4i32_ty,
                          llvm_v4i32_ty], [IntrNoMem]>;
 
   def int_x86_ssse3_phadd_sw        : GCCBuiltin<"__builtin_ia32_phaddsw">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty,
-                         llvm_v4i16_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
   def int_x86_ssse3_phadd_sw_128    : GCCBuiltin<"__builtin_ia32_phaddsw128">,
               Intrinsic<[llvm_v4i32_ty], [llvm_v4i32_ty,
                          llvm_v4i32_ty], [IntrNoMem]>;
 
   def int_x86_ssse3_phsub_w         : GCCBuiltin<"__builtin_ia32_phsubw">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty,
-                         llvm_v4i16_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
   def int_x86_ssse3_phsub_w_128     : GCCBuiltin<"__builtin_ia32_phsubw128">,
               Intrinsic<[llvm_v8i16_ty], [llvm_v8i16_ty,
                          llvm_v8i16_ty], [IntrNoMem]>;
 
   def int_x86_ssse3_phsub_d         : GCCBuiltin<"__builtin_ia32_phsubd">,
-              Intrinsic<[llvm_v2i32_ty], [llvm_v2i32_ty,
-                         llvm_v2i32_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
   def int_x86_ssse3_phsub_d_128     : GCCBuiltin<"__builtin_ia32_phsubd128">,
               Intrinsic<[llvm_v4i32_ty], [llvm_v4i32_ty,
                          llvm_v4i32_ty], [IntrNoMem]>;
 
   def int_x86_ssse3_phsub_sw        : GCCBuiltin<"__builtin_ia32_phsubsw">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty,
-                         llvm_v4i16_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
   def int_x86_ssse3_phsub_sw_128    : GCCBuiltin<"__builtin_ia32_phsubsw128">,
               Intrinsic<[llvm_v8i16_ty], [llvm_v8i16_ty,
                          llvm_v8i16_ty], [IntrNoMem]>;
 
   def int_x86_ssse3_pmadd_ub_sw     : GCCBuiltin<"__builtin_ia32_pmaddubsw">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty,
-                         llvm_v4i16_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
   def int_x86_ssse3_pmadd_ub_sw_128 : GCCBuiltin<"__builtin_ia32_pmaddubsw128">,
               Intrinsic<[llvm_v8i16_ty], [llvm_v8i16_ty,
                          llvm_v8i16_ty], [IntrNoMem]>;
@@ -615,8 +615,8 @@ let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
 // Packed multiply high with round and scale
 let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
   def int_x86_ssse3_pmul_hr_sw      : GCCBuiltin<"__builtin_ia32_pmulhrsw">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty,
-                         llvm_v4i16_ty], [IntrNoMem, Commutative]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem, Commutative]>;
   def int_x86_ssse3_pmul_hr_sw_128  : GCCBuiltin<"__builtin_ia32_pmulhrsw128">,
               Intrinsic<[llvm_v8i16_ty], [llvm_v8i16_ty,
                          llvm_v8i16_ty], [IntrNoMem, Commutative]>;
@@ -625,35 +625,35 @@ let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
 // Shuffle ops
 let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
   def int_x86_ssse3_pshuf_b         : GCCBuiltin<"__builtin_ia32_pshufb">,
-              Intrinsic<[llvm_v8i8_ty], [llvm_v8i8_ty,
-                         llvm_v8i8_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
   def int_x86_ssse3_pshuf_b_128     : GCCBuiltin<"__builtin_ia32_pshufb128">,
               Intrinsic<[llvm_v16i8_ty], [llvm_v16i8_ty,
                          llvm_v16i8_ty], [IntrNoMem]>;
   def int_x86_ssse3_pshuf_w         : GCCBuiltin<"__builtin_ia32_pshufw">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty, llvm_i32_ty],
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty, llvm_i32_ty],
                          [IntrNoMem]>;
 }
 
 // Sign ops
 let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
   def int_x86_ssse3_psign_b         : GCCBuiltin<"__builtin_ia32_psignb">,
-              Intrinsic<[llvm_v8i8_ty], [llvm_v8i8_ty,
-                         llvm_v8i8_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
   def int_x86_ssse3_psign_b_128     : GCCBuiltin<"__builtin_ia32_psignb128">,
               Intrinsic<[llvm_v16i8_ty], [llvm_v16i8_ty,
                          llvm_v16i8_ty], [IntrNoMem]>;
 
   def int_x86_ssse3_psign_w         : GCCBuiltin<"__builtin_ia32_psignw">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty,
-                         llvm_v4i16_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
   def int_x86_ssse3_psign_w_128     : GCCBuiltin<"__builtin_ia32_psignw128">,
               Intrinsic<[llvm_v8i16_ty], [llvm_v8i16_ty,
                          llvm_v8i16_ty], [IntrNoMem]>;
 
   def int_x86_ssse3_psign_d         : GCCBuiltin<"__builtin_ia32_psignd">,
-              Intrinsic<[llvm_v2i32_ty], [llvm_v2i32_ty,
-                         llvm_v2i32_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
   def int_x86_ssse3_psign_d_128     : GCCBuiltin<"__builtin_ia32_psignd128">,
               Intrinsic<[llvm_v4i32_ty], [llvm_v4i32_ty,
                          llvm_v4i32_ty], [IntrNoMem]>;
@@ -662,17 +662,17 @@ let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
 // Absolute value ops
 let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
   def int_x86_ssse3_pabs_b     : GCCBuiltin<"__builtin_ia32_pabsb">,
-              Intrinsic<[llvm_v8i8_ty], [llvm_v8i8_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty], [IntrNoMem]>;
   def int_x86_ssse3_pabs_b_128 : GCCBuiltin<"__builtin_ia32_pabsb128">,
               Intrinsic<[llvm_v16i8_ty], [llvm_v16i8_ty], [IntrNoMem]>;
 
   def int_x86_ssse3_pabs_w     : GCCBuiltin<"__builtin_ia32_pabsw">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty], [IntrNoMem]>;
   def int_x86_ssse3_pabs_w_128 : GCCBuiltin<"__builtin_ia32_pabsw128">,
               Intrinsic<[llvm_v8i16_ty], [llvm_v8i16_ty], [IntrNoMem]>;
 
   def int_x86_ssse3_pabs_d     : GCCBuiltin<"__builtin_ia32_pabsd">,
-              Intrinsic<[llvm_v2i32_ty], [llvm_v2i32_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty], [IntrNoMem]>;
   def int_x86_ssse3_pabs_d_128 : GCCBuiltin<"__builtin_ia32_pabsd128">,
               Intrinsic<[llvm_v4i32_ty], [llvm_v4i32_ty], [IntrNoMem]>;
 }
@@ -1328,281 +1328,257 @@ let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
 let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
   // Addition
   def int_x86_mmx_padd_b : GCCBuiltin<"__builtin_ia32_paddb">,
-              Intrinsic<[llvm_v8i8_ty], [llvm_v8i8_ty, llvm_v8i8_ty],
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty, llvm_x86mmx_ty],
                         [IntrNoMem]>;
   def int_x86_mmx_padd_w : GCCBuiltin<"__builtin_ia32_paddw">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty, llvm_v4i16_ty],
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty, llvm_x86mmx_ty],
                         [IntrNoMem]>;
   def int_x86_mmx_padd_d : GCCBuiltin<"__builtin_ia32_paddd">,
-              Intrinsic<[llvm_v2i32_ty], [llvm_v2i32_ty, llvm_v2i32_ty],
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty, llvm_x86mmx_ty],
                         [IntrNoMem]>;
   def int_x86_mmx_padd_q : GCCBuiltin<"__builtin_ia32_paddq">,
-              Intrinsic<[llvm_v1i64_ty], [llvm_v1i64_ty, llvm_v1i64_ty],
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty, llvm_x86mmx_ty],
                         [IntrNoMem]>;
 
   def int_x86_mmx_padds_b : GCCBuiltin<"__builtin_ia32_paddsb">,
-              Intrinsic<[llvm_v8i8_ty], [llvm_v8i8_ty,
-                         llvm_v8i8_ty], [IntrNoMem, Commutative]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem, Commutative]>;
   def int_x86_mmx_padds_w : GCCBuiltin<"__builtin_ia32_paddsw">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty,
-                         llvm_v4i16_ty], [IntrNoMem, Commutative]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem, Commutative]>;
 
   def int_x86_mmx_paddus_b : GCCBuiltin<"__builtin_ia32_paddusb">,
-              Intrinsic<[llvm_v8i8_ty], [llvm_v8i8_ty,
-                         llvm_v8i8_ty], [IntrNoMem, Commutative]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem, Commutative]>;
   def int_x86_mmx_paddus_w : GCCBuiltin<"__builtin_ia32_paddusw">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty,
-                         llvm_v4i16_ty], [IntrNoMem, Commutative]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem, Commutative]>;
 
   // Subtraction
   def int_x86_mmx_psub_b : GCCBuiltin<"__builtin_ia32_psubb">,
-              Intrinsic<[llvm_v8i8_ty], [llvm_v8i8_ty, llvm_v8i8_ty],
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty, llvm_x86mmx_ty],
                         [IntrNoMem]>;
   def int_x86_mmx_psub_w : GCCBuiltin<"__builtin_ia32_psubw">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty, llvm_v4i16_ty],
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty, llvm_x86mmx_ty],
                         [IntrNoMem]>;
   def int_x86_mmx_psub_d : GCCBuiltin<"__builtin_ia32_psubd">,
-              Intrinsic<[llvm_v2i32_ty], [llvm_v2i32_ty, llvm_v2i32_ty],
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty, llvm_x86mmx_ty],
                         [IntrNoMem]>;
   def int_x86_mmx_psub_q : GCCBuiltin<"__builtin_ia32_psubq">,
-              Intrinsic<[llvm_v1i64_ty], [llvm_v1i64_ty, llvm_v1i64_ty],
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty, llvm_x86mmx_ty],
                         [IntrNoMem]>;
 
   def int_x86_mmx_psubs_b : GCCBuiltin<"__builtin_ia32_psubsb">,
-              Intrinsic<[llvm_v8i8_ty], [llvm_v8i8_ty,
-                         llvm_v8i8_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
   def int_x86_mmx_psubs_w : GCCBuiltin<"__builtin_ia32_psubsw">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty,
-                         llvm_v4i16_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
 
   def int_x86_mmx_psubus_b : GCCBuiltin<"__builtin_ia32_psubusb">,
-              Intrinsic<[llvm_v8i8_ty], [llvm_v8i8_ty,
-                         llvm_v8i8_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
   def int_x86_mmx_psubus_w : GCCBuiltin<"__builtin_ia32_psubusw">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty,
-                         llvm_v4i16_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
 
   // Multiplication
   def int_x86_mmx_pmulh_w : GCCBuiltin<"__builtin_ia32_pmulhw">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty,
-                         llvm_v4i16_ty], [IntrNoMem, Commutative]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem, Commutative]>;
   def int_x86_mmx_pmull_w : GCCBuiltin<"__builtin_ia32_pmullw">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty,
-                         llvm_v4i16_ty], [IntrNoMem, Commutative]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem, Commutative]>;
   def int_x86_mmx_pmulhu_w : GCCBuiltin<"__builtin_ia32_pmulhuw">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty,
-                         llvm_v4i16_ty], [IntrNoMem, Commutative]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem, Commutative]>;
   def int_x86_mmx_pmulu_dq : GCCBuiltin<"__builtin_ia32_pmuludq">,
-              Intrinsic<[llvm_v2i32_ty], [llvm_v2i32_ty,
-                         llvm_v2i32_ty], [IntrNoMem, Commutative]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem, Commutative]>;
   def int_x86_mmx_pmadd_wd : GCCBuiltin<"__builtin_ia32_pmaddwd">,
-              Intrinsic<[llvm_v2i32_ty], [llvm_v4i16_ty,
-                         llvm_v4i16_ty], [IntrNoMem, Commutative]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem, Commutative]>;
 
   // Bitwise operations
   def int_x86_mmx_pand : GCCBuiltin<"__builtin_ia32_pand">,
-              Intrinsic<[llvm_v1i64_ty], [llvm_v1i64_ty, llvm_v1i64_ty],
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty, llvm_x86mmx_ty],
                         [IntrNoMem]>;
   def int_x86_mmx_pandn : GCCBuiltin<"__builtin_ia32_pandn">,
-              Intrinsic<[llvm_v1i64_ty], [llvm_v1i64_ty, llvm_v1i64_ty],
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty, llvm_x86mmx_ty],
                         [IntrNoMem]>;
   def int_x86_mmx_por : GCCBuiltin<"__builtin_ia32_por">,
-              Intrinsic<[llvm_v1i64_ty], [llvm_v1i64_ty, llvm_v1i64_ty],
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty, llvm_x86mmx_ty],
                         [IntrNoMem]>;
   def int_x86_mmx_pxor : GCCBuiltin<"__builtin_ia32_pxor">,
-              Intrinsic<[llvm_v1i64_ty], [llvm_v1i64_ty, llvm_v1i64_ty],
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty, llvm_x86mmx_ty],
                         [IntrNoMem]>;
 
   // Averages
   def int_x86_mmx_pavg_b : GCCBuiltin<"__builtin_ia32_pavgb">,
-              Intrinsic<[llvm_v8i8_ty], [llvm_v8i8_ty,
-                         llvm_v8i8_ty], [IntrNoMem, Commutative]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem, Commutative]>;
   def int_x86_mmx_pavg_w : GCCBuiltin<"__builtin_ia32_pavgw">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty,
-                         llvm_v4i16_ty], [IntrNoMem, Commutative]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem, Commutative]>;
 
   // Maximum
   def int_x86_mmx_pmaxu_b : GCCBuiltin<"__builtin_ia32_pmaxub">,
-              Intrinsic<[llvm_v8i8_ty], [llvm_v8i8_ty,
-                         llvm_v8i8_ty], [IntrNoMem, Commutative]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem, Commutative]>;
   def int_x86_mmx_pmaxs_w : GCCBuiltin<"__builtin_ia32_pmaxsw">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty,
-                         llvm_v4i16_ty], [IntrNoMem, Commutative]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem, Commutative]>;
 
   // Minimum
   def int_x86_mmx_pminu_b : GCCBuiltin<"__builtin_ia32_pminub">,
-              Intrinsic<[llvm_v8i8_ty], [llvm_v8i8_ty,
-                         llvm_v8i8_ty], [IntrNoMem, Commutative]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem, Commutative]>;
   def int_x86_mmx_pmins_w : GCCBuiltin<"__builtin_ia32_pminsw">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty,
-                         llvm_v4i16_ty], [IntrNoMem, Commutative]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem, Commutative]>;
 
   // Packed sum of absolute differences
   def int_x86_mmx_psad_bw : GCCBuiltin<"__builtin_ia32_psadbw">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v8i8_ty,
-                         llvm_v8i8_ty], [IntrNoMem, Commutative]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem, Commutative]>;
 }
 
 // Integer shift ops.
 let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
   // Shift left logical
   def int_x86_mmx_psll_w : GCCBuiltin<"__builtin_ia32_psllw">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty,
-                         llvm_v1i64_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
   def int_x86_mmx_psll_d : GCCBuiltin<"__builtin_ia32_pslld">,
-              Intrinsic<[llvm_v2i32_ty], [llvm_v2i32_ty,
-                         llvm_v1i64_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
   def int_x86_mmx_psll_q : GCCBuiltin<"__builtin_ia32_psllq">,
-              Intrinsic<[llvm_v1i64_ty], [llvm_v1i64_ty,
-                         llvm_v1i64_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
 
   def int_x86_mmx_psrl_w : GCCBuiltin<"__builtin_ia32_psrlw">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty,
-                         llvm_v1i64_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
   def int_x86_mmx_psrl_d : GCCBuiltin<"__builtin_ia32_psrld">,
-              Intrinsic<[llvm_v2i32_ty], [llvm_v2i32_ty,
-                         llvm_v1i64_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
   def int_x86_mmx_psrl_q : GCCBuiltin<"__builtin_ia32_psrlq">,
-              Intrinsic<[llvm_v1i64_ty], [llvm_v1i64_ty,
-                         llvm_v1i64_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
 
   def int_x86_mmx_psra_w : GCCBuiltin<"__builtin_ia32_psraw">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty,
-                         llvm_v1i64_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
   def int_x86_mmx_psra_d : GCCBuiltin<"__builtin_ia32_psrad">,
-              Intrinsic<[llvm_v2i32_ty], [llvm_v2i32_ty,
-                         llvm_v1i64_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
 
   def int_x86_mmx_pslli_w : GCCBuiltin<"__builtin_ia32_psllwi">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty,
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
                          llvm_i32_ty], [IntrNoMem]>;
   def int_x86_mmx_pslli_d : GCCBuiltin<"__builtin_ia32_pslldi">,
-              Intrinsic<[llvm_v2i32_ty], [llvm_v2i32_ty,
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
                          llvm_i32_ty], [IntrNoMem]>;
   def int_x86_mmx_pslli_q : GCCBuiltin<"__builtin_ia32_psllqi">,
-              Intrinsic<[llvm_v1i64_ty], [llvm_v1i64_ty,
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
                          llvm_i32_ty], [IntrNoMem]>;
 
   def int_x86_mmx_psrli_w : GCCBuiltin<"__builtin_ia32_psrlwi">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty,
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
                          llvm_i32_ty], [IntrNoMem]>;
   def int_x86_mmx_psrli_d : GCCBuiltin<"__builtin_ia32_psrldi">,
-              Intrinsic<[llvm_v2i32_ty], [llvm_v2i32_ty,
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
                          llvm_i32_ty], [IntrNoMem]>;
   def int_x86_mmx_psrli_q : GCCBuiltin<"__builtin_ia32_psrlqi">,
-              Intrinsic<[llvm_v1i64_ty], [llvm_v1i64_ty,
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
                          llvm_i32_ty], [IntrNoMem]>;
 
   def int_x86_mmx_psrai_w : GCCBuiltin<"__builtin_ia32_psrawi">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty,
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
                          llvm_i32_ty], [IntrNoMem]>;
   def int_x86_mmx_psrai_d : GCCBuiltin<"__builtin_ia32_psradi">,
-              Intrinsic<[llvm_v2i32_ty], [llvm_v2i32_ty,
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
                          llvm_i32_ty], [IntrNoMem]>;
 }
 
 // Pack ops.
 let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
   def int_x86_mmx_packsswb : GCCBuiltin<"__builtin_ia32_packsswb">,
-              Intrinsic<[llvm_v8i8_ty], [llvm_v4i16_ty,
-                         llvm_v4i16_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
   def int_x86_mmx_packssdw : GCCBuiltin<"__builtin_ia32_packssdw">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v2i32_ty,
-                         llvm_v2i32_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
   def int_x86_mmx_packuswb : GCCBuiltin<"__builtin_ia32_packuswb">,
-              Intrinsic<[llvm_v8i8_ty], [llvm_v4i16_ty,
-                         llvm_v4i16_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
 }
 
 // Unpacking ops.
 let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
   def int_x86_mmx_punpckhbw : GCCBuiltin<"__builtin_ia32_punpckhbw">,
-              Intrinsic<[llvm_v8i8_ty], [llvm_v8i8_ty, llvm_v8i8_ty],
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty, llvm_x86mmx_ty],
                         [IntrNoMem]>;
   def int_x86_mmx_punpckhwd : GCCBuiltin<"__builtin_ia32_punpckhwd">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty, llvm_v4i16_ty],
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty, llvm_x86mmx_ty],
                         [IntrNoMem]>;
   def int_x86_mmx_punpckhdq : GCCBuiltin<"__builtin_ia32_punpckhdq">,
-              Intrinsic<[llvm_v2i32_ty], [llvm_v2i32_ty, llvm_v2i32_ty],
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty, llvm_x86mmx_ty],
                         [IntrNoMem]>;
   def int_x86_mmx_punpcklbw : GCCBuiltin<"__builtin_ia32_punpcklbw">,
-              Intrinsic<[llvm_v8i8_ty], [llvm_v8i8_ty, llvm_v8i8_ty],
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty, llvm_x86mmx_ty],
                         [IntrNoMem]>;
   def int_x86_mmx_punpcklwd : GCCBuiltin<"__builtin_ia32_punpcklwd">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty, llvm_v4i16_ty],
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty, llvm_x86mmx_ty],
                         [IntrNoMem]>;
   def int_x86_mmx_punpckldq : GCCBuiltin<"__builtin_ia32_punpckldq">,
-              Intrinsic<[llvm_v2i32_ty], [llvm_v2i32_ty, llvm_v2i32_ty],
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty, llvm_x86mmx_ty],
                         [IntrNoMem]>;
 }
 
 // Integer comparison ops
 let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
   def int_x86_mmx_pcmpeq_b : GCCBuiltin<"__builtin_ia32_pcmpeqb">,
-              Intrinsic<[llvm_v8i8_ty], [llvm_v8i8_ty,
-                         llvm_v8i8_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
   def int_x86_mmx_pcmpeq_w : GCCBuiltin<"__builtin_ia32_pcmpeqw">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty,
-                         llvm_v4i16_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
   def int_x86_mmx_pcmpeq_d : GCCBuiltin<"__builtin_ia32_pcmpeqd">,
-              Intrinsic<[llvm_v2i32_ty], [llvm_v2i32_ty,
-                         llvm_v2i32_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
 
   def int_x86_mmx_pcmpgt_b : GCCBuiltin<"__builtin_ia32_pcmpgtb">,
-              Intrinsic<[llvm_v8i8_ty], [llvm_v8i8_ty,
-                         llvm_v8i8_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
   def int_x86_mmx_pcmpgt_w : GCCBuiltin<"__builtin_ia32_pcmpgtw">,
-              Intrinsic<[llvm_v4i16_ty], [llvm_v4i16_ty,
-                         llvm_v4i16_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
   def int_x86_mmx_pcmpgt_d : GCCBuiltin<"__builtin_ia32_pcmpgtd">,
-              Intrinsic<[llvm_v2i32_ty], [llvm_v2i32_ty,
-                         llvm_v2i32_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty,
+                         llvm_x86mmx_ty], [IntrNoMem]>;
 }
 
 // Misc.
 let TargetPrefix = "x86" in {  // All intrinsics start with "llvm.x86.".
   def int_x86_mmx_maskmovq : GCCBuiltin<"__builtin_ia32_maskmovq">,
-              Intrinsic<[], [llvm_v8i8_ty, llvm_v8i8_ty, llvm_ptr_ty], []>;
+              Intrinsic<[], [llvm_x86mmx_ty, llvm_x86mmx_ty, llvm_ptr_ty], []>;
 
   def int_x86_mmx_pmovmskb : GCCBuiltin<"__builtin_ia32_pmovmskb">,
-              Intrinsic<[llvm_i32_ty], [llvm_v8i8_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_i32_ty], [llvm_x86mmx_ty], [IntrNoMem]>;
 
   def int_x86_mmx_movnt_dq : GCCBuiltin<"__builtin_ia32_movntq">,
-              Intrinsic<[], [llvm_ptr_ty, llvm_v1i64_ty], []>;
+              Intrinsic<[], [llvm_ptrx86mmx_ty, llvm_x86mmx_ty], []>;
 
   def int_x86_mmx_palignr_b : GCCBuiltin<"__builtin_ia32_palignr">,
-              Intrinsic<[llvm_v1i64_ty], [llvm_v1i64_ty, 
-                        llvm_v1i64_ty, llvm_i8_ty], [IntrNoMem]>;
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty, 
+                        llvm_x86mmx_ty, llvm_i8_ty], [IntrNoMem]>;
 
-  def int_x86_mmx_pextr_w :
-              Intrinsic<[llvm_i32_ty], [llvm_v1i64_ty, llvm_i32_ty], 
+  def int_x86_mmx_pextr_w : GCCBuiltin<"__builtin_ia32_vec_ext_v4hi">,
+              Intrinsic<[llvm_i32_ty], [llvm_x86mmx_ty, llvm_i32_ty], 
                         [IntrNoMem]>;
 
-  def int_x86_mmx_pinsr_w :
-              Intrinsic<[llvm_v1i64_ty], [llvm_v1i64_ty, 
+  def int_x86_mmx_pinsr_w : GCCBuiltin<"__builtin_ia32_vec_set_v4hi">,
+              Intrinsic<[llvm_x86mmx_ty], [llvm_x86mmx_ty, 
                         llvm_i32_ty, llvm_i32_ty], [IntrNoMem]>;
-
-  def int_x86_mmx_cvtsi32_si64 : 
-              Intrinsic<[llvm_v1i64_ty], [llvm_i32_ty], [IntrNoMem]>;
-  def int_x86_mmx_cvtsi64_si32 : 
-              Intrinsic<[llvm_i32_ty], [llvm_v1i64_ty], [IntrNoMem]>;
-
-  def int_x86_mmx_vec_init_b : GCCBuiltin<"__builtin_ia32_vec_init_v8qi">,
-              Intrinsic<[llvm_v8i8_ty],
-                        [llvm_i8_ty, llvm_i8_ty, llvm_i8_ty, llvm_i8_ty,
-                         llvm_i8_ty, llvm_i8_ty, llvm_i8_ty, llvm_i8_ty],
-                        [IntrNoMem]>;
-  def int_x86_mmx_vec_init_w : GCCBuiltin<"__builtin_ia32_vec_init_v4hi">,
-              Intrinsic<[llvm_v4i16_ty],
-                        [llvm_i16_ty, llvm_i16_ty, llvm_i16_ty, llvm_i16_ty],
-                        [IntrNoMem]>;
-  def int_x86_mmx_vec_init_d : GCCBuiltin<"__builtin_ia32_vec_init_v2si">,
-              Intrinsic<[llvm_v2i32_ty],
-                        [llvm_i32_ty, llvm_i32_ty],
-                        [IntrNoMem]>;
-
-  def int_x86_mmx_vec_ext_d : GCCBuiltin<"__builtin_ia32_vec_ext_v2si">,
-              Intrinsic<[llvm_v2i32_ty],
-                        [llvm_v2i32_ty, llvm_i32_ty],
-                        [IntrNoMem]>;
 }