[clang] 16b9410 - [X86] Cast to __v4hi instead of __m64 in the implementation of _mm_extract_pi16 and _mm_insert_pi16.

[Craig Topper via cfe-commits]

Author: Craig Topper
Date: 2020-01-22T16:00:23-06:00
New Revision: 16b9410caa35da976fa5f3cf6dd3d6f3776d51ca

URL: https://github.com/llvm/llvm-project/commit/16b9410caa35da976fa5f3cf6dd3d6f3776d51ca
DIFF: https://github.com/llvm/llvm-project/commit/16b9410caa35da976fa5f3cf6dd3d6f3776d51ca.diff

LOG: [X86] Cast to __v4hi instead of __m64 in the implementation of _mm_extract_pi16 and _mm_insert_pi16.

__m64 is a vector of 1 long long. But the builtins these intrinsics
are calling expect a vector of 4 shorts.

Fixes PR44589

Added: 
    

Modified: 
    clang/lib/Headers/xmmintrin.h

Removed: 
    


################################################################################
diff  --git a/clang/lib/Headers/xmmintrin.h b/clang/lib/Headers/xmmintrin.h
index 0e61eab44aeb..9b8de63f04d5 100644
--- a/clang/lib/Headers/xmmintrin.h
+++ b/clang/lib/Headers/xmmintrin.h
@@ -2181,7 +2181,7 @@ void _mm_sfence(void);
 ///    3: Bits [63:48] are copied to the destination.
 /// \returns A 16-bit integer containing the extracted 16 bits of packed data.
 #define _mm_extract_pi16(a, n) \
-  (int)__builtin_ia32_vec_ext_v4hi((__m64)a, (int)n)
+  (int)__builtin_ia32_vec_ext_v4hi((__v4hi)a, (int)n)
 
 /// Copies data from the 64-bit vector of [4 x i16] to the destination,
 ///    and inserts the lower 16-bits of an integer operand at the 16-bit offset
@@ -2212,7 +2212,7 @@ void _mm_sfence(void);
 /// \returns A 64-bit integer vector containing the copied packed data from the
 ///    operands.
 #define _mm_insert_pi16(a, d, n) \
-  (__m64)__builtin_ia32_vec_set_v4hi((__m64)a, (int)d, (int)n)
+  (__m64)__builtin_ia32_vec_set_v4hi((__v4hi)a, (int)d, (int)n)
 
 /// Compares each of the corresponding packed 16-bit integer values of
 ///    the 64-bit integer vectors, and writes the greater value to the