+2017-04-07 Jakub Jelinek <jakub@redhat.com>
+
+ PR target/80322
+ PR target/80323
+ PR target/80325
+ PR target/80326
+ * config/i386/avxintrin.h (_mm256_cvtsd_f64, _mm256_cvtss_f32): New
+ intrinsics.
+ * config/i386/avx512fintrin.h (_mm512_int2mask, _mm512_mask2int,
+ _mm512_abs_ps, _mm512_mask_abs_ps, _mm512_abs_pd, _mm512_mask_abs_pd,
+ _mm512_cvtsd_f64, _mm512_cvtss_f32): Likewise.
+
2017-04-07 Andreas Tobler <andreast@gcc.gnu.org>
* config/aarch64/aarch64-freebsd.h: Define WCHAR_TYPE.
typedef unsigned char __mmask8;
typedef unsigned short __mmask16;
+extern __inline __mmask16
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm512_int2mask (int __M)
+{
+ return (__mmask16) __M;
+}
+
+extern __inline int
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm512_mask2int (__mmask16 __M)
+{
+ return (int) __M;
+}
+
extern __inline __m512i
__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
_mm512_set_epi64 (long long __A, long long __B, long long __C,
return __Y;
}
+#define _mm512_undefined _mm512_undefined_ps
+
extern __inline __m512d
__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
_mm512_undefined_pd (void)
(__v8di) __B, __U);
}
+extern __inline __m512
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm512_abs_ps (__m512 __A)
+{
+ return (__m512) _mm512_and_epi32 ((__m512i) __A,
+ _mm512_set1_epi32 (0x7fffffff));
+}
+
+extern __inline __m512
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm512_mask_abs_ps (__m512 __W, __mmask16 __U, __m512 __A)
+{
+ return (__m512) _mm512_mask_and_epi32 ((__m512i) __W, __U, (__m512i) __A,
+ _mm512_set1_epi32 (0x7fffffff));
+}
+
+extern __inline __m512d
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm512_abs_pd (__m512 __A)
+{
+ return (__m512d) _mm512_and_epi64 ((__m512i) __A,
+ _mm512_set1_epi64 (0x7fffffffffffffffLL));
+}
+
+extern __inline __m512d
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm512_mask_abs_pd (__m512d __W, __mmask8 __U, __m512 __A)
+{
+ return (__m512d)
+ _mm512_mask_and_epi64 ((__m512i) __W, __U, (__m512i) __A,
+ _mm512_set1_epi64 (0x7fffffffffffffffLL));
+}
+
extern __inline __m512i
__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
_mm512_unpackhi_epi32 (__m512i __A, __m512i __B)
_MM_FROUND_CUR_DIRECTION);
}
+extern __inline double
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm512_cvtsd_f64 (__m512d __A)
+{
+ return __A[0];
+}
+
+extern __inline float
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm512_cvtss_f32 (__m512 __A)
+{
+ return __A[0];
+}
+
#ifdef __x86_64__
extern __inline __m128
__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
return (__m256i)__builtin_ia32_cvttps2dq256 ((__v8sf) __A);
}
+extern __inline double
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_cvtsd_f64 (__m256d __A)
+{
+ return __A[0];
+}
+
+extern __inline float
+__attribute__ ((__gnu_inline__, __always_inline__, __artificial__))
+_mm256_cvtss_f32 (__m256 __A)
+{
+ return __A[0];
+}
+
#ifdef __OPTIMIZE__
extern __inline __m128d __attribute__((__gnu_inline__, __always_inline__, __artificial__))
_mm256_extractf128_pd (__m256d __X, const int __N)
+2017-04-07 Jakub Jelinek <jakub@redhat.com>
+
+ PR target/80322
+ PR target/80323
+ PR target/80325
+ PR target/80326
+ * gcc.target/i386/avx512f-undefined-1.c: New test.
+ * gcc.target/i386/avx512f-cvtsd-1.c: New test.
+ * gcc.target/i386/avx-cvtsd-1.c: New test.
+ * gcc.target/i386/avx512f-cvtss-1.c: New test.
+ * gcc.target/i386/avx512f-abspd-1.c: New test.
+ * gcc.target/i386/avx-cvtss-1.c: New test.
+ * gcc.target/i386/avx512f-absps-1.c: New test.
+ * gcc.target/i386/avx512f-int2mask-1.c: New test.
+ * gcc.target/i386/avx512f-mask2int-1.c: New test.
+
2017-04-07 Marek Polacek <polacek@redhat.com>
PR sanitizer/80348
--- /dev/null
+/* { dg-do run } */
+/* { dg-require-effective-target avx } */
+/* { dg-options "-O2 -mavx" } */
+
+#include "avx-check.h"
+
+__attribute__((noinline, noclone)) double
+foo (__m256d x)
+{
+ return _mm256_cvtsd_f64 (x);
+}
+
+static void
+avx_test (void)
+{
+ if (_mm256_cvtsd_f64 (_mm256_set_pd (13.5, 24.5, 23.0, 22.5)) != 22.5)
+ __builtin_abort ();
+
+ if (foo (_mm256_set_pd (24.25, 23.75, 22.0, 12.25)) != 12.25)
+ __builtin_abort ();
+}
--- /dev/null
+/* { dg-do run } */
+/* { dg-require-effective-target avx } */
+/* { dg-options "-O2 -mavx" } */
+
+#include "avx-check.h"
+
+__attribute__((noinline, noclone)) double
+foo (__m256 x)
+{
+ return _mm256_cvtss_f32 (x);
+}
+
+static void
+avx_test (void)
+{
+ if (_mm256_cvtss_f32 (_mm256_set_ps (5.5f, 24.5f, 23.0f, 22.5f,
+ 2.0f, 3.0f, 4.0f, 13.5f)) != 13.5f)
+ __builtin_abort ();
+
+ if (foo (_mm256_set_ps (5.25f, 24.25f, 23.75f, 22.0f,
+ 2.0f, 3.0f, 4.0f, 12.25f)) != 12.25f)
+ __builtin_abort ();
+}
--- /dev/null
+/* { dg-do run } */
+/* { dg-options "-O2 -mavx512f" } */
+/* { dg-require-effective-target avx512f } */
+
+#define AVX512F
+
+#include "avx512f-helper.h"
+
+#define SIZE (AVX512F_LEN / 32)
+#include "avx512f-mask-type.h"
+
+static void
+CALC (float *i1, float *r)
+{
+ int i;
+
+ for (i = 0; i < SIZE; i++)
+ if (i1[i] < 0)
+ r[i] = -i1[i];
+ else
+ r[i] = i1[i];
+}
+
+void
+TEST (void)
+{
+ float ck[SIZE];
+ int i;
+ UNION_TYPE (AVX512F_LEN, ) s, d, dm;
+ MASK_TYPE mask = MASK_VALUE;
+
+ for (i = 0; i < SIZE; i++)
+ {
+ s.a[i] = i * ((i & 1) ? 3.5f : -7.5f);
+ d.a[i] = DEFAULT_VALUE;
+ dm.a[i] = DEFAULT_VALUE;
+ }
+
+ CALC (s.a, ck);
+
+ d.x = INTRINSIC (_abs_ps) (s.x);
+ dm.x = INTRINSIC (_mask_abs_ps) (dm.x, mask, s.x);
+
+ if (UNION_CHECK (AVX512F_LEN, ) (d, ck))
+ abort ();
+
+ MASK_MERGE () (ck, mask, SIZE);
+ if (UNION_CHECK (AVX512F_LEN, ) (dm, ck))
+ abort ();
+}
--- /dev/null
+/* { dg-do run } */
+/* { dg-options "-O2 -mavx512f" } */
+/* { dg-require-effective-target avx512f } */
+
+#define AVX512F
+
+#include "avx512f-helper.h"
+
+#define SIZE (AVX512F_LEN / 32)
+#include "avx512f-mask-type.h"
+
+static void
+CALC (float *i1, float *r)
+{
+ int i;
+
+ for (i = 0; i < SIZE; i++)
+ if (i1[i] < 0)
+ r[i] = -i1[i];
+ else
+ r[i] = i1[i];
+}
+
+void
+TEST (void)
+{
+ float ck[SIZE];
+ int i;
+ UNION_TYPE (AVX512F_LEN, ) s, d, dm;
+ MASK_TYPE mask = MASK_VALUE;
+
+ for (i = 0; i < SIZE; i++)
+ {
+ s.a[i] = i * ((i & 1) ? 3.5f : -7.5f);
+ d.a[i] = DEFAULT_VALUE;
+ dm.a[i] = DEFAULT_VALUE;
+ }
+
+ CALC (s.a, ck);
+
+ d.x = INTRINSIC (_abs_ps) (s.x);
+ dm.x = INTRINSIC (_mask_abs_ps) (dm.x, mask, s.x);
+
+ if (UNION_CHECK (AVX512F_LEN, ) (d, ck))
+ abort ();
+
+ MASK_MERGE () (ck, mask, SIZE);
+ if (UNION_CHECK (AVX512F_LEN, ) (dm, ck))
+ abort ();
+}
--- /dev/null
+/* { dg-do run } */
+/* { dg-require-effective-target avx512f } */
+/* { dg-options "-O2 -mavx512f" } */
+
+#include "avx512f-check.h"
+
+__attribute__((noinline, noclone)) double
+foo (__m512d x)
+{
+ return _mm512_cvtsd_f64 (x);
+}
+
+static void
+avx512f_test (void)
+{
+ if (_mm512_cvtsd_f64 (_mm512_set_pd (5.5, 24.5, 23.0, 22.5,
+ 2.0, 3.0, 4.0, 13.5)) != 13.5)
+ __builtin_abort ();
+
+ if (foo (_mm512_set_pd (5.25, 24.25, 23.75, 22.0,
+ 2.0, 3.0, 4.0, 12.25)) != 12.25)
+ __builtin_abort ();
+}
--- /dev/null
+/* { dg-do run } */
+/* { dg-require-effective-target avx512f } */
+/* { dg-options "-O2 -mavx512f" } */
+
+#include "avx512f-check.h"
+
+__attribute__((noinline, noclone)) double
+foo (__m512 x)
+{
+ return _mm512_cvtss_f32 (x);
+}
+
+static void
+avx512f_test (void)
+{
+ if (_mm512_cvtss_f32 (_mm512_set_ps (13.0f, 24.5f, 23.0f, 22.5f,
+ 2.0f, 3.0f, 4.0f, 5.0f,
+ 6.0f, 7.0f, 8.0f, 9.0f,
+ 10.0f, 11.0f, 12.0f, 13.5f)) != 13.5f)
+ __builtin_abort ();
+
+ if (foo (_mm512_set_ps (13.25f, 24.25f, 23.75f, 22.0f,
+ 2.0f, 3.0f, 4.0f, 5.0f, 6.0f, 7.0f, 8.0f, 9.0f,
+ 10.0f, 11.0f, 12.0f, 12.25f)) != 12.25f)
+ __builtin_abort ();
+}
--- /dev/null
+/* { dg-do run } */
+/* { dg-options "-mavx512f -O2" } */
+/* { dg-require-effective-target avx512f } */
+
+#include "avx512f-check.h"
+
+static void
+avx512f_test (void)
+{
+ __m512i a = _mm512_set_epi32 (1, 17, 2, 12, 4, 14, 6, 16,
+ 8, 11, 10, 20, 12, 22, 14, 24);
+ __m512i b = _mm512_set_epi32 (0, 1, 11, 3, 13, 5, 15, 7,
+ 17, 9, 19, 11, 21, 13, 23, 16);
+ __mmask16 c = _mm512_kmov (_mm512_int2mask (2 | 8));
+ __m512i d = _mm512_mask_mov_epi32 (a, c, b);
+ __m512i e = _mm512_set_epi32 (1, 17, 2, 12, 4, 14, 6, 16,
+ 8, 11, 10, 20, 21, 22, 23, 24);
+ if (_mm512_mask2int (_mm512_cmpeq_epi32_mask (d, e)) != 0xffff)
+ __builtin_abort ();
+}
--- /dev/null
+/* { dg-do run } */
+/* { dg-options "-mavx512f -O2" } */
+/* { dg-require-effective-target avx512f } */
+
+#include "avx512f-check.h"
+
+static void
+avx512f_test (void)
+{
+ __m512i a
+ = _mm512_set_epi32 (1, 17, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 16);
+ __m512i b
+ = _mm512_set_epi32 (0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 16);
+ __mmask16 c = _mm512_cmpgt_epu32_mask (a, b);
+ if (_mm512_mask2int (c) != 0xc000)
+ __builtin_abort ();
+}
--- /dev/null
+/* { dg-do run } */
+/* { dg-options "-mavx512f -O2" } */
+/* { dg-require-effective-target avx512f } */
+
+#include "avx512f-check.h"
+
+static void
+avx512f_test (void)
+{
+ __m512 a = _mm512_undefined_ps ();
+ __m512 b = _mm512_undefined ();
+ __m512d c = _mm512_undefined_pd ();
+ __m512i d = _mm512_undefined_epi32 ();
+ __m512i e = _mm512_set1_epi32 (0);
+ __m512i f = _mm512_and_epi32 ((__m512i) a, e);
+ __m512i g = _mm512_and_epi32 ((__m512i) b, e);
+ __m512i h = _mm512_and_epi32 ((__m512i) c, e);
+ __m512i i = _mm512_and_epi32 (d, e);
+ if (_mm512_cmpeq_epi32_mask (f, e) != 0xffff
+ || _mm512_cmpeq_epi32_mask (g, e) != 0xffff
+ || _mm512_cmpeq_epi32_mask (h, e) != 0xffff
+ || _mm512_cmpeq_epi32_mask (i, e) != 0xffff)
+ __builtin_abort ();
+}
projects / gcc.git / commitdiff