[clang] [llvm] Clang: convert `__m64` intrinsics to unconditionally use SSE2 instead of MMX. (PR #96540)
via cfe-commits
cfe-commits at lists.llvm.org
Mon Jun 24 12:26:59 PDT 2024
llvmbot wrote:
<!--LLVM PR SUMMARY COMMENT-->
@llvm/pr-subscribers-backend-x86
@llvm/pr-subscribers-clang
Author: James Y Knight (jyknight)
<details>
<summary>Changes</summary>
The MMX instruction set is legacy, and the SSE2 variants are in every way superior, when they are available -- and they have been available since the Pentium 4 was released, 20 years ago.
Therefore, we are switching the "MMX" intrinsics to depend on SSE2, unconditionally. This change entirely drops the ability to generate vectorized code using compiler intrinsics for chips with MMX but without SSE2: the Intel Pentium MMX, Pentium, II, and Pentium III (released 1997-1999), as well as AMD K6 and K7 series chips of around the same timeframe. (Note that targeting these older CPUs remains supported, simply without the ability to use MMX compiler intrinsics.)
Migrating away from the use of MMX also fixes a rather non-obvious requirement for users of the intrinsics API. The long-standing programming model for MMX requires that the programmer be aware of the x87/MMX mode-switching semantics, and manually call _mm_empty() between using any MMX instruction and any x87 FPU instruction. If you neglect to, then every future x87 operation will return a NaN result. This requirement is not at all obvious to users of these these intrinsics, and causes very difficult to detect bugs.
Additionally, in some circumstances, LLVM may reorder x87 and mmx operations around each-other, unaware of this mode switching issue. So, even inserting _mm_empty() calls appropriately will not always guarantee correct operation.
Eliminating the use of MMX fixes both these latter issues.
(Originally uploaded at https://reviews.llvm.org/D86855)
Works towards issue #<!-- -->41665.
---
Patch is 125.33 KiB, truncated to 20.00 KiB below, full version: https://github.com/llvm/llvm-project/pull/96540.diff
16 Files Affected:
- (modified) clang/include/clang/Basic/BuiltinsX86.def (+2-2)
- (modified) clang/lib/CodeGen/CGBuiltin.cpp (+2)
- (modified) clang/lib/Headers/emmintrin.h (+20-22)
- (modified) clang/lib/Headers/mmintrin.h (+177-128)
- (modified) clang/lib/Headers/tmmintrin.h (+57-40)
- (modified) clang/lib/Headers/xmmintrin.h (+89-99)
- (modified) clang/test/CodeGen/X86/mmx-builtins.c (+109-100)
- (modified) clang/test/CodeGen/X86/mmx-inline-asm.c (+1-1)
- (modified) clang/test/CodeGen/X86/mmx-shift-with-immediate.c (+9-9)
- (modified) clang/test/CodeGen/attr-target-x86-mmx.c (+3-4)
- (modified) clang/test/Headers/xmmintrin.c (+1-1)
- (modified) clang/test/Sema/x86-builtin-palignr.c (+1-1)
- (modified) llvm/include/llvm/IR/IntrinsicsX86.td (+2-2)
- (added) mmx-tests/Makefile (+29)
- (added) mmx-tests/mmx-tests.py (+301)
- (added) mmx-tests/test.c (+237)
``````````diff
diff --git a/clang/include/clang/Basic/BuiltinsX86.def b/clang/include/clang/Basic/BuiltinsX86.def
index 7074479786b97..612673127a376 100644
--- a/clang/include/clang/Basic/BuiltinsX86.def
+++ b/clang/include/clang/Basic/BuiltinsX86.def
@@ -150,8 +150,8 @@ TARGET_BUILTIN(__builtin_ia32_pmovmskb, "iV8c", "ncV:64:", "mmx,sse")
TARGET_BUILTIN(__builtin_ia32_pmulhuw, "V4sV4sV4s", "ncV:64:", "mmx,sse")
TARGET_BUILTIN(__builtin_ia32_psadbw, "V4sV8cV8c", "ncV:64:", "mmx,sse")
TARGET_BUILTIN(__builtin_ia32_pshufw, "V4sV4sIc", "ncV:64:", "mmx,sse")
-TARGET_BUILTIN(__builtin_ia32_vec_ext_v4hi, "iV4sIi", "ncV:64:", "mmx,sse")
-TARGET_BUILTIN(__builtin_ia32_vec_set_v4hi, "V4sV4siIi", "ncV:64:", "mmx,sse")
+TARGET_BUILTIN(__builtin_ia32_vec_ext_v4hi, "sV4sIi", "ncV:64:", "sse")
+TARGET_BUILTIN(__builtin_ia32_vec_set_v4hi, "V4sV4ssIi", "ncV:64:", "sse")
// MMX+SSE2
TARGET_BUILTIN(__builtin_ia32_cvtpd2pi, "V2iV2d", "ncV:64:", "mmx,sse2")
diff --git a/clang/lib/CodeGen/CGBuiltin.cpp b/clang/lib/CodeGen/CGBuiltin.cpp
index 931726a78dae9..4ccf0b1ac69b3 100644
--- a/clang/lib/CodeGen/CGBuiltin.cpp
+++ b/clang/lib/CodeGen/CGBuiltin.cpp
@@ -14355,6 +14355,7 @@ Value *CodeGenFunction::EmitX86BuiltinExpr(unsigned BuiltinID,
case X86::BI__builtin_ia32_vec_init_v2si:
return Builder.CreateBitCast(BuildVector(Ops),
llvm::Type::getX86_MMXTy(getLLVMContext()));
+ case X86::BI__builtin_ia32_vec_ext_v4hi:
case X86::BI__builtin_ia32_vec_ext_v2si:
case X86::BI__builtin_ia32_vec_ext_v16qi:
case X86::BI__builtin_ia32_vec_ext_v8hi:
@@ -14373,6 +14374,7 @@ Value *CodeGenFunction::EmitX86BuiltinExpr(unsigned BuiltinID,
// Otherwise we could just do this in the header file.
return Builder.CreateExtractElement(Ops[0], Index);
}
+ case X86::BI__builtin_ia32_vec_set_v4hi:
case X86::BI__builtin_ia32_vec_set_v16qi:
case X86::BI__builtin_ia32_vec_set_v8hi:
case X86::BI__builtin_ia32_vec_set_v4si:
diff --git a/clang/lib/Headers/emmintrin.h b/clang/lib/Headers/emmintrin.h
index e85bfc47aa5cc..02160285d58d6 100644
--- a/clang/lib/Headers/emmintrin.h
+++ b/clang/lib/Headers/emmintrin.h
@@ -49,12 +49,10 @@ typedef __bf16 __m128bh __attribute__((__vector_size__(16), __aligned__(16)));
#endif
/* Define the default attributes for the functions in this file. */
-#define __DEFAULT_FN_ATTRS \
- __attribute__((__always_inline__, __nodebug__, \
- __target__("sse2,no-evex512"), __min_vector_width__(128)))
-#define __DEFAULT_FN_ATTRS_MMX \
- __attribute__((__always_inline__, __nodebug__, \
- __target__("mmx,sse2,no-evex512"), __min_vector_width__(64)))
+#define __DEFAULT_FN_ATTRS __attribute__((__always_inline__, __nodebug__, __target__("sse2,no-evex512"), __min_vector_width__(128)))
+
+#define __trunc64(x) (__m64)__builtin_shufflevector((__v2di)(x), __extension__ (__v2di){}, 0)
+#define __anyext128(x) (__m128i)__builtin_shufflevector((__v2si)(x), __extension__ (__v2si){}, 0, 1, -1, -1)
/// Adds lower double-precision values in both operands and returns the
/// sum in the lower 64 bits of the result. The upper 64 bits of the result
@@ -1486,8 +1484,8 @@ static __inline__ int __DEFAULT_FN_ATTRS _mm_cvttsd_si32(__m128d __a) {
/// \param __a
/// A 128-bit vector of [2 x double].
/// \returns A 64-bit vector of [2 x i32] containing the converted values.
-static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX _mm_cvtpd_pi32(__m128d __a) {
- return (__m64)__builtin_ia32_cvtpd2pi((__v2df)__a);
+static __inline__ __m64 __DEFAULT_FN_ATTRS _mm_cvtpd_pi32(__m128d __a) {
+ return __trunc64(__builtin_ia32_cvtpd2dq((__v2df)__a));
}
/// Converts the two double-precision floating-point elements of a
@@ -1505,8 +1503,8 @@ static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX _mm_cvtpd_pi32(__m128d __a) {
/// \param __a
/// A 128-bit vector of [2 x double].
/// \returns A 64-bit vector of [2 x i32] containing the converted values.
-static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX _mm_cvttpd_pi32(__m128d __a) {
- return (__m64)__builtin_ia32_cvttpd2pi((__v2df)__a);
+static __inline__ __m64 __DEFAULT_FN_ATTRS _mm_cvttpd_pi32(__m128d __a) {
+ return __trunc64(__builtin_ia32_cvttpd2dq((__v2df)__a));
}
/// Converts the two signed 32-bit integer elements of a 64-bit vector of
@@ -1520,8 +1518,8 @@ static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX _mm_cvttpd_pi32(__m128d __a) {
/// \param __a
/// A 64-bit vector of [2 x i32].
/// \returns A 128-bit vector of [2 x double] containing the converted values.
-static __inline__ __m128d __DEFAULT_FN_ATTRS_MMX _mm_cvtpi32_pd(__m64 __a) {
- return __builtin_ia32_cvtpi2pd((__v2si)__a);
+static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_cvtpi32_pd(__m64 __a) {
+ return (__m128d) __builtin_convertvector((__v2si)__a, __v2df);
}
/// Returns the low-order element of a 128-bit vector of [2 x double] as
@@ -2108,9 +2106,8 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_add_epi32(__m128i __a,
/// \param __b
/// A 64-bit integer.
/// \returns A 64-bit integer containing the sum of both parameters.
-static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX _mm_add_si64(__m64 __a,
- __m64 __b) {
- return (__m64)__builtin_ia32_paddq((__v1di)__a, (__v1di)__b);
+static __inline__ __m64 __DEFAULT_FN_ATTRS _mm_add_si64(__m64 __a, __m64 __b) {
+ return (__m64)(((unsigned long long)__a) + ((unsigned long long)__b));
}
/// Adds the corresponding elements of two 128-bit vectors of [2 x i64],
@@ -2431,9 +2428,9 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_mullo_epi16(__m128i __a,
/// \param __b
/// A 64-bit integer containing one of the source operands.
/// \returns A 64-bit integer vector containing the product of both operands.
-static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX _mm_mul_su32(__m64 __a,
- __m64 __b) {
- return __builtin_ia32_pmuludq((__v2si)__a, (__v2si)__b);
+static __inline__ __m64 __DEFAULT_FN_ATTRS _mm_mul_su32(__m64 __a, __m64 __b) {
+ return __trunc64(__builtin_ia32_pmuludq128((__v4si)__anyext128(__a),
+ (__v4si)__anyext128(__b)));
}
/// Multiplies 32-bit unsigned integer values contained in the lower
@@ -2539,9 +2536,8 @@ static __inline__ __m128i __DEFAULT_FN_ATTRS _mm_sub_epi32(__m128i __a,
/// A 64-bit integer vector containing the subtrahend.
/// \returns A 64-bit integer vector containing the difference of the values in
/// the operands.
-static __inline__ __m64 __DEFAULT_FN_ATTRS_MMX _mm_sub_si64(__m64 __a,
- __m64 __b) {
- return (__m64)__builtin_ia32_psubq((__v1di)__a, (__v1di)__b);
+static __inline__ __m64 __DEFAULT_FN_ATTRS _mm_sub_si64(__m64 __a, __m64 __b) {
+ return (__m64)((unsigned long long)__a - (unsigned long long)__b);
}
/// Subtracts the corresponding elements of two [2 x i64] vectors.
@@ -4889,8 +4885,10 @@ void _mm_pause(void);
#if defined(__cplusplus)
} // extern "C"
#endif
+
+#undef __anyext128
+#undef __trunc64
#undef __DEFAULT_FN_ATTRS
-#undef __DEFAULT_FN_ATTRS_MMX
#define _MM_SHUFFLE2(x, y) (((x) << 1) | (y))
diff --git a/clang/lib/Headers/mmintrin.h b/clang/lib/Headers/mmintrin.h
index 4e154e2d85935..71d7487673c02 100644
--- a/clang/lib/Headers/mmintrin.h
+++ b/clang/lib/Headers/mmintrin.h
@@ -21,10 +21,29 @@ typedef int __v2si __attribute__((__vector_size__(8)));
typedef short __v4hi __attribute__((__vector_size__(8)));
typedef char __v8qi __attribute__((__vector_size__(8)));
+/* Unsigned types */
+typedef unsigned long long __v1du __attribute__ ((__vector_size__ (8)));
+typedef unsigned int __v2su __attribute__ ((__vector_size__ (8)));
+typedef unsigned short __v4hu __attribute__((__vector_size__(8)));
+typedef unsigned char __v8qu __attribute__((__vector_size__(8)));
+
+/* We need an explicitly signed variant for char. Note that this shouldn't
+ * appear in the interface though. */
+typedef signed char __v8qs __attribute__((__vector_size__(8)));
+
+/* SSE/SSE2 types */
+typedef long long __m128i __attribute__((__vector_size__(16), __aligned__(16)));
+typedef long long __v2di __attribute__ ((__vector_size__ (16)));
+typedef int __v4si __attribute__((__vector_size__(16)));
+typedef short __v8hi __attribute__((__vector_size__(16)));
+typedef char __v16qi __attribute__((__vector_size__(16)));
+
/* Define the default attributes for the functions in this file. */
-#define __DEFAULT_FN_ATTRS \
- __attribute__((__always_inline__, __nodebug__, __target__("mmx,no-evex512"), \
- __min_vector_width__(64)))
+#define __DEFAULT_FN_ATTRS_SSE2 __attribute__((__always_inline__, __nodebug__, __target__("sse2,no-evex512"), __min_vector_width__(64)))
+
+#define __trunc64(x) (__m64)__builtin_shufflevector((__v2di)(x), __extension__ (__v2di){}, 0)
+#define __anyext128(x) (__m128i)__builtin_shufflevector((__v2si)(x), __extension__ (__v2si){}, 0, 1, -1, -1)
+#define __extract2_32(a) (__m64)__builtin_shufflevector((__v4si)(a), __extension__ (__v4si){}, 0, 2);
/// Clears the MMX state by setting the state of the x87 stack registers
/// to empty.
@@ -50,10 +69,10 @@ _mm_empty(void) {
/// A 32-bit integer value.
/// \returns A 64-bit integer vector. The lower 32 bits contain the value of the
/// parameter. The upper 32 bits are set to 0.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
+static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
_mm_cvtsi32_si64(int __i)
{
- return (__m64)__builtin_ia32_vec_init_v2si(__i, 0);
+ return __extension__ (__m64)(__v2si){__i, 0};
}
/// Returns the lower 32 bits of a 64-bit integer vector as a 32-bit
@@ -67,10 +86,10 @@ _mm_cvtsi32_si64(int __i)
/// A 64-bit integer vector.
/// \returns A 32-bit signed integer value containing the lower 32 bits of the
/// parameter.
-static __inline__ int __DEFAULT_FN_ATTRS
+static __inline__ int __DEFAULT_FN_ATTRS_SSE2
_mm_cvtsi64_si32(__m64 __m)
{
- return __builtin_ia32_vec_ext_v2si((__v2si)__m, 0);
+ return ((__v2si)__m)[0];
}
/// Casts a 64-bit signed integer value into a 64-bit integer vector.
@@ -83,7 +102,7 @@ _mm_cvtsi64_si32(__m64 __m)
/// A 64-bit signed integer.
/// \returns A 64-bit integer vector containing the same bitwise pattern as the
/// parameter.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
+static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
_mm_cvtsi64_m64(long long __i)
{
return (__m64)__i;
@@ -99,7 +118,7 @@ _mm_cvtsi64_m64(long long __i)
/// A 64-bit integer vector.
/// \returns A 64-bit signed integer containing the same bitwise pattern as the
/// parameter.
-static __inline__ long long __DEFAULT_FN_ATTRS
+static __inline__ long long __DEFAULT_FN_ATTRS_SSE2
_mm_cvtm64_si64(__m64 __m)
{
return (long long)__m;
@@ -124,10 +143,11 @@ _mm_cvtm64_si64(__m64 __m)
/// written to the upper 32 bits of the result.
/// \returns A 64-bit integer vector of [8 x i8] containing the converted
/// values.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
+static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
_mm_packs_pi16(__m64 __m1, __m64 __m2)
{
- return (__m64)__builtin_ia32_packsswb((__v4hi)__m1, (__v4hi)__m2);
+ return __extract2_32(__builtin_ia32_packsswb128((__v8hi)__anyext128(__m1),
+ (__v8hi)__anyext128(__m2)));
}
/// Converts, with saturation, 32-bit signed integers from both 64-bit integer
@@ -149,10 +169,11 @@ _mm_packs_pi16(__m64 __m1, __m64 __m2)
/// written to the upper 32 bits of the result.
/// \returns A 64-bit integer vector of [4 x i16] containing the converted
/// values.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
+static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
_mm_packs_pi32(__m64 __m1, __m64 __m2)
{
- return (__m64)__builtin_ia32_packssdw((__v2si)__m1, (__v2si)__m2);
+ return __extract2_32(__builtin_ia32_packssdw128((__v4si)__anyext128(__m1),
+ (__v4si)__anyext128(__m2)));
}
/// Converts, with saturation, 16-bit signed integers from both 64-bit integer
@@ -174,10 +195,11 @@ _mm_packs_pi32(__m64 __m1, __m64 __m2)
/// written to the upper 32 bits of the result.
/// \returns A 64-bit integer vector of [8 x i8] containing the converted
/// values.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
+static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
_mm_packs_pu16(__m64 __m1, __m64 __m2)
{
- return (__m64)__builtin_ia32_packuswb((__v4hi)__m1, (__v4hi)__m2);
+ return __extract2_32(__builtin_ia32_packuswb128((__v8hi)__anyext128(__m1),
+ (__v8hi)__anyext128(__m2)));
}
/// Unpacks the upper 32 bits from two 64-bit integer vectors of [8 x i8]
@@ -201,10 +223,11 @@ _mm_packs_pu16(__m64 __m1, __m64 __m2)
/// Bits [63:56] are written to bits [63:56] of the result.
/// \returns A 64-bit integer vector of [8 x i8] containing the interleaved
/// values.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
+static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
_mm_unpackhi_pi8(__m64 __m1, __m64 __m2)
{
- return (__m64)__builtin_ia32_punpckhbw((__v8qi)__m1, (__v8qi)__m2);
+ return (__m64)__builtin_shufflevector((__v8qi)__m1, (__v8qi)__m2,
+ 4, 12, 5, 13, 6, 14, 7, 15);
}
/// Unpacks the upper 32 bits from two 64-bit integer vectors of
@@ -224,10 +247,11 @@ _mm_unpackhi_pi8(__m64 __m1, __m64 __m2)
/// Bits [63:48] are written to bits [63:48] of the result.
/// \returns A 64-bit integer vector of [4 x i16] containing the interleaved
/// values.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
+static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
_mm_unpackhi_pi16(__m64 __m1, __m64 __m2)
{
- return (__m64)__builtin_ia32_punpckhwd((__v4hi)__m1, (__v4hi)__m2);
+ return (__m64)__builtin_shufflevector((__v4hi)__m1, (__v4hi)__m2,
+ 2, 6, 3, 7);
}
/// Unpacks the upper 32 bits from two 64-bit integer vectors of
@@ -245,10 +269,10 @@ _mm_unpackhi_pi16(__m64 __m1, __m64 __m2)
/// the upper 32 bits of the result.
/// \returns A 64-bit integer vector of [2 x i32] containing the interleaved
/// values.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
+static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
_mm_unpackhi_pi32(__m64 __m1, __m64 __m2)
{
- return (__m64)__builtin_ia32_punpckhdq((__v2si)__m1, (__v2si)__m2);
+ return (__m64)__builtin_shufflevector((__v2si)__m1, (__v2si)__m2, 1, 3);
}
/// Unpacks the lower 32 bits from two 64-bit integer vectors of [8 x i8]
@@ -272,10 +296,11 @@ _mm_unpackhi_pi32(__m64 __m1, __m64 __m2)
/// Bits [31:24] are written to bits [63:56] of the result.
/// \returns A 64-bit integer vector of [8 x i8] containing the interleaved
/// values.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
+static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
_mm_unpacklo_pi8(__m64 __m1, __m64 __m2)
{
- return (__m64)__builtin_ia32_punpcklbw((__v8qi)__m1, (__v8qi)__m2);
+ return (__m64)__builtin_shufflevector((__v8qi)__m1, (__v8qi)__m2,
+ 0, 8, 1, 9, 2, 10, 3, 11);
}
/// Unpacks the lower 32 bits from two 64-bit integer vectors of
@@ -295,10 +320,11 @@ _mm_unpacklo_pi8(__m64 __m1, __m64 __m2)
/// Bits [31:16] are written to bits [63:48] of the result.
/// \returns A 64-bit integer vector of [4 x i16] containing the interleaved
/// values.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
+static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
_mm_unpacklo_pi16(__m64 __m1, __m64 __m2)
{
- return (__m64)__builtin_ia32_punpcklwd((__v4hi)__m1, (__v4hi)__m2);
+ return (__m64)__builtin_shufflevector((__v4hi)__m1, (__v4hi)__m2,
+ 0, 4, 1, 5);
}
/// Unpacks the lower 32 bits from two 64-bit integer vectors of
@@ -316,10 +342,10 @@ _mm_unpacklo_pi16(__m64 __m1, __m64 __m2)
/// the upper 32 bits of the result.
/// \returns A 64-bit integer vector of [2 x i32] containing the interleaved
/// values.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
+static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
_mm_unpacklo_pi32(__m64 __m1, __m64 __m2)
{
- return (__m64)__builtin_ia32_punpckldq((__v2si)__m1, (__v2si)__m2);
+ return (__m64)__builtin_shufflevector((__v2si)__m1, (__v2si)__m2, 0, 2);
}
/// Adds each 8-bit integer element of the first 64-bit integer vector
@@ -337,10 +363,10 @@ _mm_unpacklo_pi32(__m64 __m1, __m64 __m2)
/// A 64-bit integer vector of [8 x i8].
/// \returns A 64-bit integer vector of [8 x i8] containing the sums of both
/// parameters.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
+static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
_mm_add_pi8(__m64 __m1, __m64 __m2)
{
- return (__m64)__builtin_ia32_paddb((__v8qi)__m1, (__v8qi)__m2);
+ return (__m64)(((__v8qu)__m1) + ((__v8qu)__m2));
}
/// Adds each 16-bit integer element of the first 64-bit integer vector
@@ -358,10 +384,10 @@ _mm_add_pi8(__m64 __m1, __m64 __m2)
/// A 64-bit integer vector of [4 x i16].
/// \returns A 64-bit integer vector of [4 x i16] containing the sums of both
/// parameters.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
+static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
_mm_add_pi16(__m64 __m1, __m64 __m2)
{
- return (__m64)__builtin_ia32_paddw((__v4hi)__m1, (__v4hi)__m2);
+ return (__m64)(((__v4hu)__m1) + ((__v4hu)__m2));
}
/// Adds each 32-bit integer element of the first 64-bit integer vector
@@ -379,10 +405,10 @@ _mm_add_pi16(__m64 __m1, __m64 __m2)
/// A 64-bit integer vector of [2 x i32].
/// \returns A 64-bit integer vector of [2 x i32] containing the sums of both
/// parameters.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
+static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
_mm_add_pi32(__m64 __m1, __m64 __m2)
{
- return (__m64)__builtin_ia32_paddd((__v2si)__m1, (__v2si)__m2);
+ return (__m64)(((__v2su)__m1) + ((__v2su)__m2));
}
/// Adds, with saturation, each 8-bit signed integer element of the first
@@ -403,10 +429,10 @@ _mm_add_pi32(__m64 __m1, __m64 __m2)
/// A 64-bit integer vector of [8 x i8].
/// \returns A 64-bit integer vector of [8 x i8] containing the saturated sums
/// of both parameters.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
+static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
_mm_adds_pi8(__m64 __m1, __m64 __m2)
{
- return (__m64)__builtin_ia32_paddsb((__v8qi)__m1, (__v8qi)__m2);
+ return (__m64)__builtin_elementwise_add_sat((__v8qs)__m1, (__v8qs)__m2);
}
/// Adds, with saturation, each 16-bit signed integer element of the first
@@ -427,10 +453,10 @@ _mm_adds_pi8(__m64 __m1, __m64 __m2)
/// A 64-bit integer vector of [4 x i16].
/// \returns A 64-bit integer vector of [4 x i16] containing the saturated sums
/// of both parameters.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
+static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
_mm_adds_pi16(__m64 __m1, __m64 __m2)
{
- return (__m64)__builtin_ia32_paddsw((__v4hi)__m1, (__v4hi)__m2);
+ return (__m64)__builtin_elementwise_add_sat((__v4hi)__m1, (__v4hi)__m2);
}
/// Adds, with saturation, each 8-bit unsigned integer element of the first
@@ -450,10 +476,10 @@ _mm_adds_pi16(__m64 __m1, __m64 __m2)
/// A 64-bit integer vector of [8 x i8].
/// \returns A 64-bit integer vector of [8 x i8] containing the saturated
/// unsigned sums of both parameters.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
+static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
_mm_adds_pu8(__m64 __m1, __m64 __m2)
{
- return (__m64)__builtin_ia32_paddusb((__v8qi)__m1, (__v8qi)__m2);
+ return (__m64)__builtin_elementwise_add_sat((__v8qu)__m1, (__v8qu)__m2);
}
/// Adds, with saturation, each 16-bit unsigned integer element of the first
@@ -473,10 +499,10 @@ _mm_adds_pu8(__m64 __m1, __m64 __m2)
/// A 64-bit integer vector of [4 x i16].
/// \returns A 64-bit integer vector of [4 x i16] containing the saturated
/// unsigned sums of both parameters.
-static __inline__ __m64 __DEFAULT_FN_ATTRS
+static __inline__ __m64 __DEFAULT_FN_ATTRS_SSE2
_mm_adds_pu16(__m64 __m1, __m64 __m2)
{
- return (__m64)__builtin_ia32_paddusw(...
[truncated]
``````````
</details>
https://github.com/llvm/llvm-project/pull/96540
More information about the cfe-commits
mailing list