Skip to content

Commit

Permalink
[X86] Finally handle target of __builtin_ia32_cmp[p|s][s|d] from avx …
Browse files Browse the repository at this point in the history 
…into sse/sse2/avx (llvm#84136)

This patch relands llvm#67410 and fixes the cmpfail below:
#include <immintrin.h>
__attribute__((target("avx"))) void test(__m128 a, __m128 b) {
  _mm_cmp_ps(a, b, 14);
}

According to Intel SDM, SSE/SSE2 instructions cmp[p|s][s|d] are
supported when imm8 is in range of [0, 7]
  • Loading branch information
@FreddyLeaf
FreddyLeaf committed Mar 9, 2024
1 parent c58c827 commit fc0fc76
Show file tree
Hide file tree
Showing 11 changed files with 316 additions and 121 deletions.
8 changes: 4 additions & 4 deletions clang/include/clang/Basic/BuiltinsX86.def
View file
Original file line number Diff line number Diff line change
Expand Up @@ -226,6 +226,8 @@ TARGET_BUILTIN(__builtin_ia32_minps, "V4fV4fV4f", "ncV:128:", "sse")
TARGET_BUILTIN(__builtin_ia32_maxps, "V4fV4fV4f", "ncV:128:", "sse")
TARGET_BUILTIN(__builtin_ia32_minss, "V4fV4fV4f", "ncV:128:", "sse")
TARGET_BUILTIN(__builtin_ia32_maxss, "V4fV4fV4f", "ncV:128:", "sse")
TARGET_BUILTIN(__builtin_ia32_cmpps, "V4fV4fV4fIc", "ncV:128:", "sse")
TARGET_BUILTIN(__builtin_ia32_cmpss, "V4fV4fV4fIc", "ncV:128:", "sse")

TARGET_BUILTIN(__builtin_ia32_cmpeqpd, "V2dV2dV2d", "ncV:128:", "sse2")
TARGET_BUILTIN(__builtin_ia32_cmpltpd, "V2dV2dV2d", "ncV:128:", "sse2")
Expand All @@ -243,6 +245,8 @@ TARGET_BUILTIN(__builtin_ia32_cmpneqsd, "V2dV2dV2d", "ncV:128:", "sse2")
TARGET_BUILTIN(__builtin_ia32_cmpnltsd, "V2dV2dV2d", "ncV:128:", "sse2")
TARGET_BUILTIN(__builtin_ia32_cmpnlesd, "V2dV2dV2d", "ncV:128:", "sse2")
TARGET_BUILTIN(__builtin_ia32_cmpordsd, "V2dV2dV2d", "ncV:128:", "sse2")
TARGET_BUILTIN(__builtin_ia32_cmpsd, "V2dV2dV2dIc", "ncV:128:", "sse2")
TARGET_BUILTIN(__builtin_ia32_cmppd, "V2dV2dV2dIc", "ncV:128:", "sse2")
TARGET_BUILTIN(__builtin_ia32_minpd, "V2dV2dV2d", "ncV:128:", "sse2")
TARGET_BUILTIN(__builtin_ia32_maxpd, "V2dV2dV2d", "ncV:128:", "sse2")
TARGET_BUILTIN(__builtin_ia32_minsd, "V2dV2dV2d", "ncV:128:", "sse2")
Expand Down Expand Up @@ -462,12 +466,8 @@ TARGET_BUILTIN(__builtin_ia32_blendvps256, "V8fV8fV8fV8f", "ncV:256:", "avx")
TARGET_BUILTIN(__builtin_ia32_shufpd256, "V4dV4dV4dIi", "ncV:256:", "avx")
TARGET_BUILTIN(__builtin_ia32_shufps256, "V8fV8fV8fIi", "ncV:256:", "avx")
TARGET_BUILTIN(__builtin_ia32_dpps256, "V8fV8fV8fIc", "ncV:256:", "avx")
TARGET_BUILTIN(__builtin_ia32_cmppd, "V2dV2dV2dIc", "ncV:128:", "avx")
TARGET_BUILTIN(__builtin_ia32_cmppd256, "V4dV4dV4dIc", "ncV:256:", "avx")
TARGET_BUILTIN(__builtin_ia32_cmpps, "V4fV4fV4fIc", "ncV:128:", "avx")
TARGET_BUILTIN(__builtin_ia32_cmpps256, "V8fV8fV8fIc", "ncV:256:", "avx")
TARGET_BUILTIN(__builtin_ia32_cmpsd, "V2dV2dV2dIc", "ncV:128:", "avx")
TARGET_BUILTIN(__builtin_ia32_cmpss, "V4fV4fV4fIc", "ncV:128:", "avx")
TARGET_BUILTIN(__builtin_ia32_vextractf128_pd256, "V2dV4dIi", "ncV:256:", "avx")
TARGET_BUILTIN(__builtin_ia32_vextractf128_ps256, "V4fV8fIi", "ncV:256:", "avx")
TARGET_BUILTIN(__builtin_ia32_vextractf128_si256, "V4iV8iIi", "ncV:256:", "avx")
Expand Down
19 changes: 19 additions & 0 deletions clang/lib/CodeGen/CodeGenFunction.cpp
View file
Original file line number Diff line number Diff line change
Expand Up @@ -31,6 +31,7 @@
#include "clang/AST/StmtObjC.h"
#include "clang/Basic/Builtins.h"
#include "clang/Basic/CodeGenOptions.h"
#include "clang/Basic/TargetBuiltins.h"
#include "clang/Basic/TargetInfo.h"
#include "clang/CodeGen/CGFunctionInfo.h"
#include "clang/Frontend/FrontendDiagnostic.h"
Expand Down Expand Up @@ -2613,6 +2614,24 @@ void CGBuilderInserter::InsertHelper(
// called function.
void CodeGenFunction::checkTargetFeatures(const CallExpr *E,
const FunctionDecl *TargetDecl) {
// SemaChecking cannot handle below x86 builtins because they have different
// parameter ranges with different TargetAttribute of caller.
if (CGM.getContext().getTargetInfo().getTriple().isX86()) {
unsigned BuiltinID = TargetDecl->getBuiltinID();
if (BuiltinID == X86::BI__builtin_ia32_cmpps ||
BuiltinID == X86::BI__builtin_ia32_cmpss ||
BuiltinID == X86::BI__builtin_ia32_cmppd ||
BuiltinID == X86::BI__builtin_ia32_cmpsd) {
const FunctionDecl *FD = dyn_cast_or_null<FunctionDecl>(CurCodeDecl);
llvm::StringMap<bool> TargetFetureMap;
CGM.getContext().getFunctionFeatureMap(TargetFetureMap, FD);
llvm::APSInt Result =
*(E->getArg(2)->getIntegerConstantExpr(CGM.getContext()));
if (Result.getSExtValue() > 7 && !TargetFetureMap.lookup("avx"))
CGM.getDiags().Report(E->getBeginLoc(), diag::err_builtin_needs_feature)
<< TargetDecl->getDeclName() << "avx";
}
}
return checkTargetFeatures(E->getBeginLoc(), TargetDecl);
}

Expand Down
28 changes: 8 additions & 20 deletions clang/lib/Headers/avxintrin.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -1574,14 +1574,6 @@ _mm256_blendv_ps(__m256 __a, __m256 __b, __m256 __c)
(__v4df)(__m256d)(b), (int)(mask)))

/* Compare */
#define _CMP_EQ_OQ 0x00 /* Equal (ordered, non-signaling) */
#define _CMP_LT_OS 0x01 /* Less-than (ordered, signaling) */
#define _CMP_LE_OS 0x02 /* Less-than-or-equal (ordered, signaling) */
#define _CMP_UNORD_Q 0x03 /* Unordered (non-signaling) */
#define _CMP_NEQ_UQ 0x04 /* Not-equal (unordered, non-signaling) */
#define _CMP_NLT_US 0x05 /* Not-less-than (unordered, signaling) */
#define _CMP_NLE_US 0x06 /* Not-less-than-or-equal (unordered, signaling) */
#define _CMP_ORD_Q 0x07 /* Ordered (non-signaling) */
#define _CMP_EQ_UQ 0x08 /* Equal (unordered, non-signaling) */
#define _CMP_NGE_US 0x09 /* Not-greater-than-or-equal (unordered, signaling) */
#define _CMP_NGT_US 0x0a /* Not-greater-than (unordered, signaling) */
Expand All @@ -1607,6 +1599,7 @@ _mm256_blendv_ps(__m256 __a, __m256 __b, __m256 __c)
#define _CMP_GT_OQ 0x1e /* Greater-than (ordered, non-signaling) */
#define _CMP_TRUE_US 0x1f /* True (unordered, signaling) */

/* Below intrinsic defined in emmintrin.h can be used for AVX */
/// Compares each of the corresponding double-precision values of two
/// 128-bit vectors of [2 x double], using the operation specified by the
/// immediate integer operand.
Expand Down Expand Up @@ -1663,10 +1656,9 @@ _mm256_blendv_ps(__m256 __a, __m256 __b, __m256 __c)
/// 0x1E: Greater-than (ordered, non-signaling) \n
/// 0x1F: True (unordered, signaling)
/// \returns A 128-bit vector of [2 x double] containing the comparison results.
#define _mm_cmp_pd(a, b, c) \
((__m128d)__builtin_ia32_cmppd((__v2df)(__m128d)(a), \
(__v2df)(__m128d)(b), (c)))
/// \fn __m128d _mm_cmp_pd(__m128d a, __m128d b, const int c)

/* Below intrinsic defined in xmmintrin.h can be used for AVX */
/// Compares each of the corresponding values of two 128-bit vectors of
/// [4 x float], using the operation specified by the immediate integer
/// operand.
Expand Down Expand Up @@ -1723,9 +1715,7 @@ _mm256_blendv_ps(__m256 __a, __m256 __b, __m256 __c)
/// 0x1E: Greater-than (ordered, non-signaling) \n
/// 0x1F: True (unordered, signaling)
/// \returns A 128-bit vector of [4 x float] containing the comparison results.
#define _mm_cmp_ps(a, b, c) \
((__m128)__builtin_ia32_cmpps((__v4sf)(__m128)(a), \
(__v4sf)(__m128)(b), (c)))
/// \fn __m128 _mm_cmp_ps(__m128 a, __m128 b, const int c)

/// Compares each of the corresponding double-precision values of two
/// 256-bit vectors of [4 x double], using the operation specified by the
Expand Down Expand Up @@ -1847,6 +1837,7 @@ _mm256_blendv_ps(__m256 __a, __m256 __b, __m256 __c)
((__m256)__builtin_ia32_cmpps256((__v8sf)(__m256)(a), \
(__v8sf)(__m256)(b), (c)))

/* Below intrinsic defined in emmintrin.h can be used for AVX */
/// Compares each of the corresponding scalar double-precision values of
/// two 128-bit vectors of [2 x double], using the operation specified by the
/// immediate integer operand.
Expand Down Expand Up @@ -1902,10 +1893,9 @@ _mm256_blendv_ps(__m256 __a, __m256 __b, __m256 __c)
/// 0x1E: Greater-than (ordered, non-signaling) \n
/// 0x1F: True (unordered, signaling)
/// \returns A 128-bit vector of [2 x double] containing the comparison results.
#define _mm_cmp_sd(a, b, c) \
((__m128d)__builtin_ia32_cmpsd((__v2df)(__m128d)(a), \
(__v2df)(__m128d)(b), (c)))
/// \fn __m128d _mm_cmp_sd(__m128d a, __m128d b, const int c)

/* Below intrinsic defined in xmmintrin.h can be used for AVX */
/// Compares each of the corresponding scalar values of two 128-bit
/// vectors of [4 x float], using the operation specified by the immediate
/// integer operand.
Expand Down Expand Up @@ -1961,9 +1951,7 @@ _mm256_blendv_ps(__m256 __a, __m256 __b, __m256 __c)
/// 0x1E: Greater-than (ordered, non-signaling) \n
/// 0x1F: True (unordered, signaling)
/// \returns A 128-bit vector of [4 x float] containing the comparison results.
#define _mm_cmp_ss(a, b, c) \
((__m128)__builtin_ia32_cmpss((__v4sf)(__m128)(a), \
(__v4sf)(__m128)(b), (c)))
/// \fn __m128 _mm_cmp_ss(__m128 a, __m128 b, const int c)

/// Takes a [8 x i32] vector and returns the vector element value
/// indexed by the immediate constant operand.
Expand Down
68 changes: 68 additions & 0 deletions clang/lib/Headers/emmintrin.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -4745,6 +4745,74 @@ static __inline__ __m128d __DEFAULT_FN_ATTRS _mm_castsi128_pd(__m128i __a) {
return (__m128d)__a;
}

/// Compares each of the corresponding double-precision values of two
/// 128-bit vectors of [2 x double], using the operation specified by the
/// immediate integer operand.
///
/// Each comparison yields 0x0 for false, 0xFFFFFFFFFFFFFFFF for true.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128d _mm_cmp_pd(__m128d a, __m128d b, const int c);
/// \endcode
///
/// This intrinsic corresponds to the <c> (V)CMPPD </c> instruction.
///
/// \param a
/// A 128-bit vector of [2 x double].
/// \param b
/// A 128-bit vector of [2 x double].
/// \param c
/// An immediate integer operand, with bits [4:0] specifying which comparison
/// operation to use: \n
/// 0x00: Equal (ordered, non-signaling) \n
/// 0x01: Less-than (ordered, signaling) \n
/// 0x02: Less-than-or-equal (ordered, signaling) \n
/// 0x03: Unordered (non-signaling) \n
/// 0x04: Not-equal (unordered, non-signaling) \n
/// 0x05: Not-less-than (unordered, signaling) \n
/// 0x06: Not-less-than-or-equal (unordered, signaling) \n
/// 0x07: Ordered (non-signaling) \n
/// \returns A 128-bit vector of [2 x double] containing the comparison results.
#define _mm_cmp_pd(a, b, c) \
((__m128d)__builtin_ia32_cmppd((__v2df)(__m128d)(a), (__v2df)(__m128d)(b), \
(c)))

/// Compares each of the corresponding scalar double-precision values of
/// two 128-bit vectors of [2 x double], using the operation specified by the
/// immediate integer operand.
///
/// Each comparison yields 0x0 for false, 0xFFFFFFFFFFFFFFFF for true.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128d _mm_cmp_sd(__m128d a, __m128d b, const int c);
/// \endcode
///
/// This intrinsic corresponds to the <c> (V)CMPSD </c> instruction.
///
/// \param a
/// A 128-bit vector of [2 x double].
/// \param b
/// A 128-bit vector of [2 x double].
/// \param c
/// An immediate integer operand, with bits [4:0] specifying which comparison
/// operation to use: \n
/// 0x00: Equal (ordered, non-signaling) \n
/// 0x01: Less-than (ordered, signaling) \n
/// 0x02: Less-than-or-equal (ordered, signaling) \n
/// 0x03: Unordered (non-signaling) \n
/// 0x04: Not-equal (unordered, non-signaling) \n
/// 0x05: Not-less-than (unordered, signaling) \n
/// 0x06: Not-less-than-or-equal (unordered, signaling) \n
/// 0x07: Ordered (non-signaling) \n
/// \returns A 128-bit vector of [2 x double] containing the comparison results.
#define _mm_cmp_sd(a, b, c) \
((__m128d)__builtin_ia32_cmpsd((__v2df)(__m128d)(a), (__v2df)(__m128d)(b), \
(c)))

#if defined(__cplusplus)
extern "C" {
#endif
Expand Down
75 changes: 75 additions & 0 deletions clang/lib/Headers/xmmintrin.h
View file
Original file line number Diff line number Diff line change
Expand Up @@ -2940,6 +2940,81 @@ _mm_movemask_ps(__m128 __a)
return __builtin_ia32_movmskps((__v4sf)__a);
}

/* Compare */
#define _CMP_EQ_OQ 0x00 /* Equal (ordered, non-signaling) */
#define _CMP_LT_OS 0x01 /* Less-than (ordered, signaling) */
#define _CMP_LE_OS 0x02 /* Less-than-or-equal (ordered, signaling) */
#define _CMP_UNORD_Q 0x03 /* Unordered (non-signaling) */
#define _CMP_NEQ_UQ 0x04 /* Not-equal (unordered, non-signaling) */
#define _CMP_NLT_US 0x05 /* Not-less-than (unordered, signaling) */
#define _CMP_NLE_US 0x06 /* Not-less-than-or-equal (unordered, signaling) */
#define _CMP_ORD_Q 0x07 /* Ordered (non-signaling) */

/// Compares each of the corresponding values of two 128-bit vectors of
/// [4 x float], using the operation specified by the immediate integer
/// operand.
///
/// Each comparison yields 0x0 for false, 0xFFFFFFFF for true.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128 _mm_cmp_ps(__m128 a, __m128 b, const int c);
/// \endcode
///
/// This intrinsic corresponds to the <c> (V)CMPPS </c> instruction.
///
/// \param a
/// A 128-bit vector of [4 x float].
/// \param b
/// A 128-bit vector of [4 x float].
/// \param c
/// An immediate integer operand, with bits [4:0] specifying which comparison
/// operation to use: \n
/// 0x00: Equal (ordered, non-signaling) \n
/// 0x01: Less-than (ordered, signaling) \n
/// 0x02: Less-than-or-equal (ordered, signaling) \n
/// 0x03: Unordered (non-signaling) \n
/// 0x04: Not-equal (unordered, non-signaling) \n
/// 0x05: Not-less-than (unordered, signaling) \n
/// 0x06: Not-less-than-or-equal (unordered, signaling) \n
/// 0x07: Ordered (non-signaling) \n
/// \returns A 128-bit vector of [4 x float] containing the comparison results.
#define _mm_cmp_ps(a, b, c) \
((__m128)__builtin_ia32_cmpps((__v4sf)(__m128)(a), (__v4sf)(__m128)(b), (c)))

/// Compares each of the corresponding scalar values of two 128-bit
/// vectors of [4 x float], using the operation specified by the immediate
/// integer operand.
///
/// Each comparison yields 0x0 for false, 0xFFFFFFFF for true.
///
/// \headerfile <x86intrin.h>
///
/// \code
/// __m128 _mm_cmp_ss(__m128 a, __m128 b, const int c);
/// \endcode
///
/// This intrinsic corresponds to the <c> (V)CMPSS </c> instruction.
///
/// \param a
/// A 128-bit vector of [4 x float].
/// \param b
/// A 128-bit vector of [4 x float].
/// \param c
/// An immediate integer operand, with bits [4:0] specifying which comparison
/// operation to use: \n
/// 0x00: Equal (ordered, non-signaling) \n
/// 0x01: Less-than (ordered, signaling) \n
/// 0x02: Less-than-or-equal (ordered, signaling) \n
/// 0x03: Unordered (non-signaling) \n
/// 0x04: Not-equal (unordered, non-signaling) \n
/// 0x05: Not-less-than (unordered, signaling) \n
/// 0x06: Not-less-than-or-equal (unordered, signaling) \n
/// 0x07: Ordered (non-signaling) \n
/// \returns A 128-bit vector of [4 x float] containing the comparison results.
#define _mm_cmp_ss(a, b, c) \
((__m128)__builtin_ia32_cmpss((__v4sf)(__m128)(a), (__v4sf)(__m128)(b), (c)))

#define _MM_ALIGN16 __attribute__((aligned(16)))

Expand Down
11 changes: 11 additions & 0 deletions clang/test/CodeGen/X86/attribute-cmpsd-no-error.c
View file
Original file line number Diff line number Diff line change
@@ -0,0 +1,11 @@
// RUN: %clang_cc1 %s -ffreestanding -triple=x86_64-unknown-unknown-emit-llvm -o /dev/null -verify
// RUN: %clang_cc1 %s -ffreestanding -triple=i386-unknown-unknown-emit-llvm -o /dev/null -verify

// expected-no-diagnostics

#include <immintrin.h>

__attribute__((target("avx")))
__m128 test(__m128 a, __m128 b) {
return _mm_cmp_ps(a, b, 14);
}
Loading

0 comments on commit fc0fc76

Please sign in to comment.