Patchwork [AArch64] Convert NEON frint implementations to use builtins.

login
register
mail settings
Submitter James Greenhalgh
Date April 26, 2013, 12:45 p.m.
Message ID <1366980340-2426-1-git-send-email-james.greenhalgh@arm.com>
Download mbox | patch
Permalink /patch/239857/
State New
Headers show

Comments

James Greenhalgh - April 26, 2013, 12:45 p.m.
Hi,

This patch renames the vrnd<a, i, m, n, p, x><q> intrinsics,
which previously were vrnd<q><a, i, m, n, p, x>

At the same time, we move these intrinsics to an RTL-based intrinsic.

Regression tested on aarch64-none-elf with no issues.

Thanks,
James

---
gcc/

2013-04-26  James Greenhalgh  <james.greenhalgh@arm.com>

	* config/aarch64/arm_neon.h (vrndq<a,m,n,p>_f<32, 64>): Rename to...
	(vrnd<a,m,n,p>q_f<32, 64>): ...This, implement using builtin.
	(vrnd<a,m,n,p>_f32): Implement using builtins.
	(vrnd<i,x><q>_f<32, 64>): New.

gcc/testsuite/

2013-04-26  James Greenhalgh  <james.greenhalgh@arm.com>

	* gcc.target/aarch64/vect-vrnd.c: New.
Marcus Shawcroft - April 26, 2013, 1:35 p.m.
On 26 April 2013 13:45, James Greenhalgh <james.greenhalgh@arm.com> wrote:
>
> Hi,
>
> This patch renames the vrnd<a, i, m, n, p, x><q> intrinsics,
> which previously were vrnd<q><a, i, m, n, p, x>
>
> At the same time, we move these intrinsics to an RTL-based intrinsic.
>
> Regression tested on aarch64-none-elf with no issues.
>
> Thanks,
> James
>
> ---
> gcc/
>
> 2013-04-26  James Greenhalgh  <james.greenhalgh@arm.com>
>
>         * config/aarch64/arm_neon.h (vrndq<a,m,n,p>_f<32, 64>): Rename to...
>         (vrnd<a,m,n,p>q_f<32, 64>): ...This, implement using builtin.
>         (vrnd<a,m,n,p>_f32): Implement using builtins.
>         (vrnd<i,x><q>_f<32, 64>): New.
>
> gcc/testsuite/
>
> 2013-04-26  James Greenhalgh  <james.greenhalgh@arm.com>
>
>         * gcc.target/aarch64/vect-vrnd.c: New.

OK
/Marcus

Patch

diff --git a/gcc/config/aarch64/arm_neon.h b/gcc/config/aarch64/arm_neon.h
index 6f5ca8e..c868a46 100644
--- a/gcc/config/aarch64/arm_neon.h
+++ b/gcc/config/aarch64/arm_neon.h
@@ -14941,171 +14941,6 @@  vrev64q_u32 (uint32x4_t a)
   return result;
 }
 
-__extension__ static __inline float32x2_t __attribute__ ((__always_inline__))
-vrnd_f32 (float32x2_t a)
-{
-  float32x2_t result;
-  __asm__ ("frintz %0.2s,%1.2s"
-           : "=w"(result)
-           : "w"(a)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline float32x2_t __attribute__ ((__always_inline__))
-vrnda_f32 (float32x2_t a)
-{
-  float32x2_t result;
-  __asm__ ("frinta %0.2s,%1.2s"
-           : "=w"(result)
-           : "w"(a)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline float32x2_t __attribute__ ((__always_inline__))
-vrndm_f32 (float32x2_t a)
-{
-  float32x2_t result;
-  __asm__ ("frintm %0.2s,%1.2s"
-           : "=w"(result)
-           : "w"(a)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline float32x2_t __attribute__ ((__always_inline__))
-vrndn_f32 (float32x2_t a)
-{
-  float32x2_t result;
-  __asm__ ("frintn %0.2s,%1.2s"
-           : "=w"(result)
-           : "w"(a)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline float32x2_t __attribute__ ((__always_inline__))
-vrndp_f32 (float32x2_t a)
-{
-  float32x2_t result;
-  __asm__ ("frintp %0.2s,%1.2s"
-           : "=w"(result)
-           : "w"(a)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline float32x4_t __attribute__ ((__always_inline__))
-vrndq_f32 (float32x4_t a)
-{
-  float32x4_t result;
-  __asm__ ("frintz %0.4s,%1.4s"
-           : "=w"(result)
-           : "w"(a)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline float64x2_t __attribute__ ((__always_inline__))
-vrndq_f64 (float64x2_t a)
-{
-  float64x2_t result;
-  __asm__ ("frintz %0.2d,%1.2d"
-           : "=w"(result)
-           : "w"(a)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline float32x4_t __attribute__ ((__always_inline__))
-vrndqa_f32 (float32x4_t a)
-{
-  float32x4_t result;
-  __asm__ ("frinta %0.4s,%1.4s"
-           : "=w"(result)
-           : "w"(a)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline float64x2_t __attribute__ ((__always_inline__))
-vrndqa_f64 (float64x2_t a)
-{
-  float64x2_t result;
-  __asm__ ("frinta %0.2d,%1.2d"
-           : "=w"(result)
-           : "w"(a)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline float32x4_t __attribute__ ((__always_inline__))
-vrndqm_f32 (float32x4_t a)
-{
-  float32x4_t result;
-  __asm__ ("frintm %0.4s,%1.4s"
-           : "=w"(result)
-           : "w"(a)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline float64x2_t __attribute__ ((__always_inline__))
-vrndqm_f64 (float64x2_t a)
-{
-  float64x2_t result;
-  __asm__ ("frintm %0.2d,%1.2d"
-           : "=w"(result)
-           : "w"(a)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline float32x4_t __attribute__ ((__always_inline__))
-vrndqn_f32 (float32x4_t a)
-{
-  float32x4_t result;
-  __asm__ ("frintn %0.4s,%1.4s"
-           : "=w"(result)
-           : "w"(a)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline float64x2_t __attribute__ ((__always_inline__))
-vrndqn_f64 (float64x2_t a)
-{
-  float64x2_t result;
-  __asm__ ("frintn %0.2d,%1.2d"
-           : "=w"(result)
-           : "w"(a)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline float32x4_t __attribute__ ((__always_inline__))
-vrndqp_f32 (float32x4_t a)
-{
-  float32x4_t result;
-  __asm__ ("frintp %0.4s,%1.4s"
-           : "=w"(result)
-           : "w"(a)
-           : /* No clobbers */);
-  return result;
-}
-
-__extension__ static __inline float64x2_t __attribute__ ((__always_inline__))
-vrndqp_f64 (float64x2_t a)
-{
-  float64x2_t result;
-  __asm__ ("frintp %0.2d,%1.2d"
-           : "=w"(result)
-           : "w"(a)
-           : /* No clobbers */);
-  return result;
-}
-
 #define vrshrn_high_n_s16(a, b, c)                                      \
   __extension__                                                         \
     ({                                                                  \
@@ -23069,6 +22904,145 @@  vrecpxd_f64 (float64_t __a)
   return __builtin_aarch64_frecpxdf (__a);
 }
 
+/* vrnd  */
+
+__extension__ static __inline float32x2_t __attribute__ ((__always_inline__))
+vrnd_f32 (float32x2_t __a)
+{
+  return __builtin_aarch64_btruncv2sf (__a);
+}
+
+__extension__ static __inline float32x4_t __attribute__ ((__always_inline__))
+vrndq_f32 (float32x4_t __a)
+{
+  return __builtin_aarch64_btruncv4sf (__a);
+}
+
+__extension__ static __inline float64x2_t __attribute__ ((__always_inline__))
+vrndq_f64 (float64x2_t __a)
+{
+  return __builtin_aarch64_btruncv2df (__a);
+}
+
+/* vrnda  */
+
+__extension__ static __inline float32x2_t __attribute__ ((__always_inline__))
+vrnda_f32 (float32x2_t __a)
+{
+  return __builtin_aarch64_roundv2sf (__a);
+}
+
+__extension__ static __inline float32x4_t __attribute__ ((__always_inline__))
+vrndaq_f32 (float32x4_t __a)
+{
+  return __builtin_aarch64_roundv4sf (__a);
+}
+
+__extension__ static __inline float64x2_t __attribute__ ((__always_inline__))
+vrndaq_f64 (float64x2_t __a)
+{
+  return __builtin_aarch64_roundv2df (__a);
+}
+
+/* vrndi  */
+
+__extension__ static __inline float32x2_t __attribute__ ((__always_inline__))
+vrndi_f32 (float32x2_t __a)
+{
+  return __builtin_aarch64_nearbyintv2sf (__a);
+}
+
+__extension__ static __inline float32x4_t __attribute__ ((__always_inline__))
+vrndiq_f32 (float32x4_t __a)
+{
+  return __builtin_aarch64_nearbyintv4sf (__a);
+}
+
+__extension__ static __inline float64x2_t __attribute__ ((__always_inline__))
+vrndiq_f64 (float64x2_t __a)
+{
+  return __builtin_aarch64_nearbyintv2df (__a);
+}
+
+/* vrndm  */
+
+__extension__ static __inline float32x2_t __attribute__ ((__always_inline__))
+vrndm_f32 (float32x2_t __a)
+{
+  return __builtin_aarch64_floorv2sf (__a);
+}
+
+__extension__ static __inline float32x4_t __attribute__ ((__always_inline__))
+vrndmq_f32 (float32x4_t __a)
+{
+  return __builtin_aarch64_floorv4sf (__a);
+}
+
+__extension__ static __inline float64x2_t __attribute__ ((__always_inline__))
+vrndmq_f64 (float64x2_t __a)
+{
+  return __builtin_aarch64_floorv2df (__a);
+}
+
+/* vrndn  */
+
+__extension__ static __inline float32x2_t __attribute__ ((__always_inline__))
+vrndn_f32 (float32x2_t __a)
+{
+  return __builtin_aarch64_frintnv2sf (__a);
+}
+__extension__ static __inline float32x4_t __attribute__ ((__always_inline__))
+vrndnq_f32 (float32x4_t __a)
+{
+  return __builtin_aarch64_frintnv4sf (__a);
+}
+
+__extension__ static __inline float64x2_t __attribute__ ((__always_inline__))
+vrndnq_f64 (float64x2_t __a)
+{
+  return __builtin_aarch64_frintnv2df (__a);
+}
+
+/* vrndp  */
+
+__extension__ static __inline float32x2_t __attribute__ ((__always_inline__))
+vrndp_f32 (float32x2_t __a)
+{
+  return __builtin_aarch64_ceilv2sf (__a);
+}
+
+__extension__ static __inline float32x4_t __attribute__ ((__always_inline__))
+vrndpq_f32 (float32x4_t __a)
+{
+  return __builtin_aarch64_ceilv4sf (__a);
+}
+
+__extension__ static __inline float64x2_t __attribute__ ((__always_inline__))
+vrndpq_f64 (float64x2_t __a)
+{
+  return __builtin_aarch64_ceilv2df (__a);
+}
+
+/* vrndx  */
+
+__extension__ static __inline float32x2_t __attribute__ ((__always_inline__))
+vrndx_f32 (float32x2_t __a)
+{
+  return __builtin_aarch64_rintv2sf (__a);
+}
+
+__extension__ static __inline float32x4_t __attribute__ ((__always_inline__))
+vrndxq_f32 (float32x4_t __a)
+{
+  return __builtin_aarch64_rintv4sf (__a);
+}
+
+__extension__ static __inline float64x2_t __attribute__ ((__always_inline__))
+vrndxq_f64 (float64x2_t __a)
+{
+  return __builtin_aarch64_rintv2df (__a);
+}
+
 /* vrshl */
 
 __extension__ static __inline int8x8_t __attribute__ ((__always_inline__))
diff --git a/gcc/testsuite/gcc.target/aarch64/vect-vrnd.c b/gcc/testsuite/gcc.target/aarch64/vect-vrnd.c
new file mode 100644
index 0000000..aa3fd9b
--- /dev/null
+++ b/gcc/testsuite/gcc.target/aarch64/vect-vrnd.c
@@ -0,0 +1,117 @@ 
+/* { dg-do run } */
+/* { dg-options "-O3 --save-temps" } */
+
+#include <arm_neon.h>
+
+extern void abort (void);
+extern float fabsf (float);
+extern double fabs (double);
+
+extern double trunc (double);
+extern double round (double);
+extern double nearbyint (double);
+extern double floor (double);
+extern double ceil (double);
+extern double rint (double);
+
+extern float truncf (float);
+extern float roundf (float);
+extern float nearbyintf (float);
+extern float floorf (float);
+extern float ceilf (float);
+extern float rintf (float);
+
+#define NUM_TESTS 8
+#define DELTA 0.000001
+
+float input_f32[] = {0.1f, -0.1f, 0.4f, 10.3f,
+		     200.0f, -800.0f, -13.0f, -0.5f};
+double input_f64[] = {0.1, -0.1, 0.4, 10.3,
+		      200.0, -800.0, -13.0, -0.5};
+
+#define TEST(SUFFIX, Q, WIDTH, LANES, C_FN, F)		     		\
+int									\
+test_vrnd##SUFFIX##_float##WIDTH##x##LANES##_t (void)			\
+{									\
+  int ret = 1;								\
+  int i = 0;								\
+  int nlanes = LANES;							\
+  float##WIDTH##_t expected_out[NUM_TESTS];				\
+  float##WIDTH##_t actual_out[NUM_TESTS];				\
+									\
+  for (i = 0; i < NUM_TESTS; i++)					\
+    {									\
+      expected_out[i] = C_FN##F (input_f##WIDTH[i]);			\
+      /* Don't vectorize this.  */					\
+      asm volatile ("" : : : "memory");					\
+    }									\
+									\
+  /* Prevent the compiler from noticing these two loops do the same	\
+     thing and optimizing away the comparison.  */			\
+  asm volatile ("" : : : "memory");					\
+									\
+  for (i = 0; i < NUM_TESTS; i+=nlanes)					\
+    {									\
+      float##WIDTH##x##LANES##_t out =					\
+	vrnd##SUFFIX##Q##_f##WIDTH					\
+		(vld1##Q##_f##WIDTH (input_f##WIDTH + i));		\
+      vst1##Q##_f##WIDTH (actual_out + i, out);				\
+    }									\
+									\
+  for (i = 0; i < NUM_TESTS; i++)					\
+    ret &= fabs##F (expected_out[i] - actual_out[i]) < DELTA;		\
+									\
+  return ret;								\
+}									\
+
+
+#define BUILD_VARIANTS(SUFFIX, C_FN)	\
+TEST (SUFFIX,  , 32, 2, C_FN, f)	\
+TEST (SUFFIX, q, 32, 4, C_FN, f)	\
+TEST (SUFFIX, q, 64, 2, C_FN,  )	\
+
+BUILD_VARIANTS ( , trunc)
+/* { dg-final { scan-assembler "frintz\\tv\[0-9\]+\.2s, v\[0-9\]+\.2s" } } */
+/* { dg-final { scan-assembler "frintz\\tv\[0-9\]+\.4s, v\[0-9\]+\.4s" } } */
+/* { dg-final { scan-assembler "frintz\\tv\[0-9\]+\.2d, v\[0-9\]+\.2d" } } */
+BUILD_VARIANTS (a, round)
+/* { dg-final { scan-assembler "frinta\\tv\[0-9\]+\.2s, v\[0-9\]+\.2s" } } */
+/* { dg-final { scan-assembler "frinta\\tv\[0-9\]+\.4s, v\[0-9\]+\.4s" } } */
+/* { dg-final { scan-assembler "frinta\\tv\[0-9\]+\.2d, v\[0-9\]+\.2d" } } */
+BUILD_VARIANTS (i, nearbyint)
+/* { dg-final { scan-assembler "frinti\\tv\[0-9\]+\.2s, v\[0-9\]+\.2s" } } */
+/* { dg-final { scan-assembler "frinti\\tv\[0-9\]+\.4s, v\[0-9\]+\.4s" } } */
+/* { dg-final { scan-assembler "frinti\\tv\[0-9\]+\.2d, v\[0-9\]+\.2d" } } */
+BUILD_VARIANTS (m, floor)
+/* { dg-final { scan-assembler "frintm\\tv\[0-9\]+\.2s, v\[0-9\]+\.2s" } } */
+/* { dg-final { scan-assembler "frintm\\tv\[0-9\]+\.4s, v\[0-9\]+\.4s" } } */
+/* { dg-final { scan-assembler "frintm\\tv\[0-9\]+\.2d, v\[0-9\]+\.2d" } } */
+BUILD_VARIANTS (p, ceil)
+/* { dg-final { scan-assembler "frintp\\tv\[0-9\]+\.2s, v\[0-9\]+\.2s" } } */
+/* { dg-final { scan-assembler "frintp\\tv\[0-9\]+\.4s, v\[0-9\]+\.4s" } } */
+/* { dg-final { scan-assembler "frintp\\tv\[0-9\]+\.2d, v\[0-9\]+\.2d" } } */
+BUILD_VARIANTS (x, rint)
+/* { dg-final { scan-assembler "frintx\\tv\[0-9\]+\.2s, v\[0-9\]+\.2s" } } */
+/* { dg-final { scan-assembler "frintx\\tv\[0-9\]+\.4s, v\[0-9\]+\.4s" } } */
+/* { dg-final { scan-assembler "frintx\\tv\[0-9\]+\.2d, v\[0-9\]+\.2d" } } */
+
+#undef TEST
+#define TEST(SUFFIX, Q, WIDTH, LANES, C_FN, F)			\
+{								\
+  if (!test_vrnd##SUFFIX##_float##WIDTH##x##LANES##_t ())	\
+    abort ();							\
+}
+
+int
+main (int argc, char **argv)
+{
+  BUILD_VARIANTS ( , trunc)
+  BUILD_VARIANTS (a, round)
+  BUILD_VARIANTS (i, nearbyint)
+  BUILD_VARIANTS (m, floor)
+  BUILD_VARIANTS (p, ceil)
+  BUILD_VARIANTS (x, rint)
+  return 0;
+}
+
+/* { dg-final { cleanup-saved-temps } } */