This is the mail archive of the
gcc-patches@gcc.gnu.org
mailing list for the GCC project.
[PATCH][AArch64] Implement vfma_f64, vmla_f64, vfms_f64, vmls_f64 intrinsics
- From: Kyrill Tkachov <kyrylo dot tkachov at arm dot com>
- To: GCC Patches <gcc-patches at gcc dot gnu dot org>
- Cc: Marcus Shawcroft <Marcus dot Shawcroft at arm dot com>
- Date: Fri, 20 Jun 2014 15:17:22 +0100
- Subject: [PATCH][AArch64] Implement vfma_f64, vmla_f64, vfms_f64, vmls_f64 intrinsics
- Authentication-results: sourceware.org; auth=none
Hi all,
Now that Alan fixed the float64x1_t machinery, this patch implements
some low-hanging intrinsics
in arm_neon.h.
Tested aarch64-none-elf and bootstrapped on aarch64-linux.
Ok for trunk?
Thanks,
Kyrill
2014-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
* config/aarch64/arm_neon.h (vfma_f64): New intrinsic.
(vmla_f64): Likewise.
(vfms_f64): Likewise.
(vmls_f64): Likewise.
2014-06-20 Kyrylo Tkachov <kyrylo.tkachov@arm.com>
* gcc.target/aarch64/simd/vfma_f64.c: New test.
* gcc.target/aarch64/simd/vmla_f64.c: Likewise.
* gcc.target/aarch64/simd/vfms_f64.c: Likewise.
* gcc.target/aarch64/simd/vmls_f64.c: Likewise.
commit ffb5a3efe38e50c0d410b5517e030aa37cad88b7
Author: Kyrylo Tkachov <kyrylo.tkachov@arm.com>
Date: Wed Jun 18 14:25:17 2014 +0100
[AArch64] Implement vfma_f64, vmla_f64, vfms_f64, vmls_f64 in arm_neon.h
diff --git a/gcc/config/aarch64/arm_neon.h b/gcc/config/aarch64/arm_neon.h
index fd520f5..2809b3e 100644
--- a/gcc/config/aarch64/arm_neon.h
+++ b/gcc/config/aarch64/arm_neon.h
@@ -16701,6 +16701,14 @@ vextq_u64 (uint64x2_t __a, uint64x2_t __b, __const int __c)
#endif
}
+/* vfma */
+
+__extension__ static __inline float64x1_t __attribute__ ((__always_inline__))
+vfma_f64 (float64x1_t __a, float64x1_t __b, float64x1_t __c)
+{
+ return (float64x1_t) {__builtin_fma (__b[0], __c[0], __a[0])};
+}
+
/* vfma_lane */
__extension__ static __inline float32x2_t __attribute__ ((__always_inline__))
@@ -16804,6 +16812,14 @@ vfmaq_laneq_f64 (float64x2_t __a, float64x2_t __b,
__a);
}
+/* vfms */
+
+__extension__ static __inline float64x1_t __attribute__ ((__always_inline__))
+vfms_f64 (float64x1_t __a, float64x1_t __b, float64x1_t __c)
+{
+ return (float64x1_t) {__builtin_fma (-__b[0], __c[0], __a[0])};
+}
+
/* vfms_lane */
__extension__ static __inline float32x2_t __attribute__ ((__always_inline__))
@@ -18432,6 +18448,12 @@ vmla_f32 (float32x2_t a, float32x2_t b, float32x2_t c)
return a + b * c;
}
+__extension__ static __inline float64x1_t __attribute__ ((__always_inline__))
+vmla_f64 (float64x1_t __a, float64x1_t __b, float64x1_t __c)
+{
+ return __a + __b * __c;
+}
+
__extension__ static __inline float32x4_t __attribute__ ((__always_inline__))
vmlaq_f32 (float32x4_t a, float32x4_t b, float32x4_t c)
{
@@ -18600,6 +18622,12 @@ vmls_f32 (float32x2_t a, float32x2_t b, float32x2_t c)
return a - b * c;
}
+__extension__ static __inline float64x1_t __attribute__ ((__always_inline__))
+vmls_f64 (float64x1_t __a, float64x1_t __b, float64x1_t __c)
+{
+ return __a - __b * __c;
+}
+
__extension__ static __inline float32x4_t __attribute__ ((__always_inline__))
vmlsq_f32 (float32x4_t a, float32x4_t b, float32x4_t c)
{
diff --git a/gcc/testsuite/gcc.target/aarch64/simd/vfma_f64.c b/gcc/testsuite/gcc.target/aarch64/simd/vfma_f64.c
new file mode 100644
index 0000000..d6bcf1c
--- /dev/null
+++ b/gcc/testsuite/gcc.target/aarch64/simd/vfma_f64.c
@@ -0,0 +1,41 @@
+/* Test the vfma_f64 AArch64 SIMD intrinsic. */
+
+/* { dg-do run } */
+/* { dg-options "-save-temps -O3" } */
+
+#include "arm_neon.h"
+#include <stdio.h>
+
+#define EPS 1.0e-15
+
+
+extern void abort (void);
+
+int
+main (void)
+{
+ float64x1_t arg1;
+ float64x1_t arg2;
+ float64x1_t arg3;
+
+ float64_t expected;
+ float64_t actual;
+
+ arg1 = vcreate_f64 (0x3fe3955382d35b0eULL);
+ arg2 = vcreate_f64 (0x3fa88480812d6670ULL);
+ arg3 = vcreate_f64 (0x3fd5791ae2a92572ULL);
+
+ expected = 0.6280448184360076;
+ actual = vget_lane_f64 (vfma_f64 (arg1, arg2, arg3), 0);
+
+ if (__builtin_fabs (expected - actual) > EPS)
+ {
+ fprintf (stderr, "Expected: %lf, got %lf\n", expected, actual);
+ abort ();
+ }
+
+ return 0;
+}
+
+/* { dg-final { scan-assembler-times "fmadd\[ \t\]+\[dD\]\[0-9\]+, ?\[dD\]\[0-9\]+, ?\[dD\]\[0-9\]+, ?\[dD\]\[0-9\]+\n" 1 } } */
+/* { dg-final { cleanup-saved-temps } } */
diff --git a/gcc/testsuite/gcc.target/aarch64/simd/vfms_f64.c b/gcc/testsuite/gcc.target/aarch64/simd/vfms_f64.c
new file mode 100644
index 0000000..3f34758
--- /dev/null
+++ b/gcc/testsuite/gcc.target/aarch64/simd/vfms_f64.c
@@ -0,0 +1,41 @@
+/* Test the vfms_f64 AArch64 SIMD intrinsic. */
+
+/* { dg-do run } */
+/* { dg-options "-save-temps -O3" } */
+
+#include "arm_neon.h"
+#include <stdio.h>
+
+#define EPS 1.0e-15
+
+
+extern void abort (void);
+
+int
+main (void)
+{
+ float64x1_t arg1;
+ float64x1_t arg2;
+ float64x1_t arg3;
+
+ float64_t expected;
+ float64_t actual;
+
+ arg1 = vcreate_f64 (0x3fe730af8db9e6f7ULL);
+ arg2 = vcreate_f64 (0x3fe6b78680fa29ceULL);
+ arg3 = vcreate_f64 (0x3feea3cbf921fbe0ULL);
+
+ expected = 4.4964705746355915e-2;
+ actual = vget_lane_f64 (vfms_f64 (arg1, arg2, arg3), 0);
+
+ if (__builtin_fabs (expected - actual) > EPS)
+ {
+ fprintf (stderr, "Expected: %lf, got %lf\n", expected, actual);
+ abort ();
+ }
+
+ return 0;
+}
+
+/* { dg-final { scan-assembler-times "fmsub\[ \t\]+\[dD\]\[0-9\]+, ?\[dD\]\[0-9\]+, ?\[dD\]\[0-9\]+, ?\[dD\]\[0-9\]+\n" 1 } } */
+/* { dg-final { cleanup-saved-temps } } */
diff --git a/gcc/testsuite/gcc.target/aarch64/simd/vmla_f64.c b/gcc/testsuite/gcc.target/aarch64/simd/vmla_f64.c
new file mode 100644
index 0000000..59ce326
--- /dev/null
+++ b/gcc/testsuite/gcc.target/aarch64/simd/vmla_f64.c
@@ -0,0 +1,37 @@
+/* Test the vmla_f64 AArch64 SIMD intrinsic. */
+
+/* { dg-do run } */
+/* { dg-options "-O3" } */
+
+#include "arm_neon.h"
+#include <stdio.h>
+
+#define EPS 1.0e-15
+
+extern void abort (void);
+
+int
+main (void)
+{
+ float64x1_t arg1;
+ float64x1_t arg2;
+ float64x1_t arg3;
+
+ float64_t expected;
+ float64_t actual;
+
+ arg1 = vcreate_f64 (0x3fc4de626b6bbe9cULL);
+ arg2 = vcreate_f64 (0x3fb7e454dbe84408ULL);
+ arg3 = vcreate_f64 (0x3fdd359b94201a3aULL);
+
+ expected = 0.20563116414665633;
+ actual = vget_lane_f64 (vmla_f64 (arg1, arg2, arg3), 0);
+
+ if (__builtin_fabs (expected - actual) > EPS)
+ {
+ fprintf (stderr, "Expected: %lf, got %lf\n", expected, actual);
+ abort ();
+ }
+
+ return 0;
+}
diff --git a/gcc/testsuite/gcc.target/aarch64/simd/vmls_f64.c b/gcc/testsuite/gcc.target/aarch64/simd/vmls_f64.c
new file mode 100644
index 0000000..9163609
--- /dev/null
+++ b/gcc/testsuite/gcc.target/aarch64/simd/vmls_f64.c
@@ -0,0 +1,37 @@
+/* Test the vmls_f64 AArch64 SIMD intrinsic. */
+
+/* { dg-do run } */
+/* { dg-options "-O3" } */
+
+#include "arm_neon.h"
+#include <stdio.h>
+
+#define EPS 1.0e-15
+
+extern void abort (void);
+
+int
+main (void)
+{
+ float64x1_t arg1;
+ float64x1_t arg2;
+ float64x1_t arg3;
+
+ float64_t expected;
+ float64_t actual;
+
+ arg1 = vcreate_f64 (0x3fea7ec860271ad9ULL);
+ arg2 = vcreate_f64 (0x3fca04faa09302e8ULL);
+ arg3 = vcreate_f64 (0x3fecfec8c67415a0ULL);
+
+ expected = 0.6437868393361155;
+ actual = vget_lane_f64 (vmls_f64 (arg1, arg2, arg3), 0);
+
+ if (__builtin_fabs (expected - actual) > EPS)
+ {
+ fprintf (stderr, "Expected: %lf, got %lf\n", expected, actual);
+ abort ();
+ }
+
+ return 0;
+}