Committed on behalf of Matthew Wahab

gcc/
2017-05-16  Matthew Wahab  <matthew.wahab@arm.com>

	* config/arm/arm_neon.h (vadd_f16): Use standard arithmetic
	operations in fast-math mode.
	(vaddq_f16): Likewise.
	(vmul_f16): Likewise.
	(vmulq_f16): Likewise.
	(vsub_f16): Likewise.
	(vsubq_f16): Likewise.
	* config/arm/neon.md (add<mode>3): New.
	(sub<mode>3): New.
	(fma:<VH:mode>3): New.  Also remove outdated comment.
	(mul<mode>3): New.

testsuite/
2017-05-16  Matthew Wahab  <matthew.wahab@arm.com>

	* gcc.target/arm/armv8_2-fp16-arith-1.c: Expand comment.  Update
	expected output of vadd, vsub and vmul instructions.
	* gcc.target/arm/armv8_2-fp16-arith-2.c: New.
	* gcc.target/arm/armv8_2-fp16-neon-2.c: New.
	* gcc.target/arm/armv8_2-fp16-neon-3.c: New.

From-SVN: r248090

gcc/ChangeLog

+2017-05-16  Matthew Wahab  <matthew.wahab@arm.com>
+
+	* config/arm/arm_neon.h (vadd_f16): Use standard arithmetic
+	operations in fast-math mode.
+	(vaddq_f16): Likewise.
+	(vmul_f16): Likewise.
+	(vmulq_f16): Likewise.
+	(vsub_f16): Likewise.
+	(vsubq_f16): Likewise.
+	* config/arm/neon.md (add<mode>3): New.
+	(sub<mode>3): New.
+	(fma:<VH:mode>3): New.  Also remove outdated comment.
+	(mul<mode>3): New.
+
 2017-05-16  Martin Liska  <mliska@suse.cz>
 
 	PR ipa/79849.

gcc/config/arm/arm_neon.h

@@ -17069,14 +17069,22 @@ __extension__ extern __inline float16x4_t
 __attribute__  ((__always_inline__, __gnu_inline__, __artificial__))
 vadd_f16 (float16x4_t __a, float16x4_t __b)
 {
+#ifdef __FAST_MATH__
+  return __a + __b;
+#else
   return __builtin_neon_vaddv4hf (__a, __b);
+#endif
 }
 
 __extension__ extern __inline float16x8_t
 __attribute__  ((__always_inline__, __gnu_inline__, __artificial__))
 vaddq_f16 (float16x8_t __a, float16x8_t __b)
 {
+#ifdef __FAST_MATH__
+  return __a + __b;
+#else
   return __builtin_neon_vaddv8hf (__a, __b);
+#endif
 }
 
 __extension__ extern __inline uint16x4_t
@@ -17587,7 +17595,11 @@ __extension__ extern __inline float16x4_t
 __attribute__  ((__always_inline__, __gnu_inline__, __artificial__))
 vmul_f16 (float16x4_t __a, float16x4_t __b)
 {
+#ifdef __FAST_MATH__
+  return __a * __b;
+#else
   return __builtin_neon_vmulfv4hf (__a, __b);
+#endif
 }
 
 __extension__ extern __inline float16x4_t
@@ -17608,7 +17620,11 @@ __extension__ extern __inline float16x8_t
 __attribute__  ((__always_inline__, __gnu_inline__, __artificial__))
 vmulq_f16 (float16x8_t __a, float16x8_t __b)
 {
+#ifdef __FAST_MATH__
+  return __a * __b;
+#else
   return __builtin_neon_vmulfv8hf (__a, __b);
+#endif
 }
 
 __extension__ extern __inline float16x8_t
@@ -17804,14 +17820,22 @@ __extension__ extern __inline float16x4_t
 __attribute__  ((__always_inline__, __gnu_inline__, __artificial__))
 vsub_f16 (float16x4_t __a, float16x4_t __b)
 {
+#ifdef __FAST_MATH__
+  return __a - __b;
+#else
   return __builtin_neon_vsubv4hf (__a, __b);
+#endif
 }
 
 __extension__ extern __inline float16x8_t
 __attribute__  ((__always_inline__, __gnu_inline__, __artificial__))
 vsubq_f16 (float16x8_t __a, float16x8_t __b)
 {
+#ifdef __FAST_MATH__
+  return __a - __b;
+#else
   return __builtin_neon_vsubv8hf (__a, __b);
+#endif
 }
 
 #endif /* __ARM_FEATURE_VECTOR_FP16_ARITHMETIC.  */

gcc/config/arm/neon.md

@@ -505,6 +505,23 @@
                     (const_string "neon_add<q>")))]
 )
 
+;; As with SFmode, full support for HFmode vector arithmetic is only available
+;; when flag-unsafe-math-optimizations is enabled.
+
+(define_insn "add<mode>3"
+  [(set
+    (match_operand:VH 0 "s_register_operand" "=w")
+    (plus:VH
+     (match_operand:VH 1 "s_register_operand" "w")
+     (match_operand:VH 2 "s_register_operand" "w")))]
+ "TARGET_NEON_FP16INST && flag_unsafe_math_optimizations"
+ "vadd.<V_if_elem>\t%<V_reg>0, %<V_reg>1, %<V_reg>2"
+ [(set (attr "type")
+   (if_then_else (match_test "<Is_float_mode>")
+    (const_string "neon_fp_addsub_s<q>")
+    (const_string "neon_add<q>")))]
+)
+
 (define_insn "add<mode>3_fp16"
   [(set
     (match_operand:VH 0 "s_register_operand" "=w")
@@ -557,6 +574,17 @@
                     (const_string "neon_sub<q>")))]
 )
 
+(define_insn "sub<mode>3"
+ [(set
+   (match_operand:VH 0 "s_register_operand" "=w")
+   (minus:VH
+    (match_operand:VH 1 "s_register_operand" "w")
+    (match_operand:VH 2 "s_register_operand" "w")))]
+ "TARGET_NEON_FP16INST && flag_unsafe_math_optimizations"
+ "vsub.<V_if_elem>\t%<V_reg>0, %<V_reg>1, %<V_reg>2"
+ [(set_attr "type" "neon_sub<q>")]
+)
+
 (define_insn "sub<mode>3_fp16"
  [(set
    (match_operand:VH 0 "s_register_operand" "=w")
@@ -664,8 +692,17 @@
   [(set_attr "type" "neon_fp_mla_s<q>")]
 )
 
-;; There is limited support for unsafe-math optimizations using the NEON FP16
-;; arithmetic instructions, so only the intrinsic is currently supported.
+(define_insn "fma<VH:mode>4"
+ [(set (match_operand:VH 0 "register_operand" "=w")
+   (fma:VH
+    (match_operand:VH 1 "register_operand" "w")
+    (match_operand:VH 2 "register_operand" "w")
+    (match_operand:VH 3 "register_operand" "0")))]
+ "TARGET_NEON_FP16INST && flag_unsafe_math_optimizations"
+ "vfma.<V_if_elem>\\t%<V_reg>0, %<V_reg>1, %<V_reg>2"
+ [(set_attr "type" "neon_fp_mla_s<q>")]
+)
+
 (define_insn "fma<VH:mode>4_intrinsic"
  [(set (match_operand:VH 0 "register_operand" "=w")
    (fma:VH
@@ -2175,6 +2212,17 @@
                     (const_string "neon_mul_<V_elem_ch><q>")))]
 )
 
+(define_insn "mul<mode>3"
+ [(set
+   (match_operand:VH 0 "s_register_operand" "=w")
+   (mult:VH
+    (match_operand:VH 1 "s_register_operand" "w")
+    (match_operand:VH 2 "s_register_operand" "w")))]
+  "TARGET_NEON_FP16INST && flag_unsafe_math_optimizations"
+  "vmul.f16\t%<V_reg>0, %<V_reg>1, %<V_reg>2"
+ [(set_attr "type" "neon_mul_<VH_elem_ch><q>")]
+)
+
 (define_insn "neon_vmulf<mode>"
  [(set
    (match_operand:VH 0 "s_register_operand" "=w")

gcc/testsuite/ChangeLog

+2016-05-16  Matthew Wahab  <matthew.wahab@arm.com>
+
+	* gcc.target/arm/armv8_2-fp16-arith-1.c: Expand comment.  Update
+	expected output of vadd, vsub and vmul instructions.
+	* gcc.target/arm/armv8_2-fp16-arith-2.c: New.
+	* gcc.target/arm/armv8_2-fp16-neon-2.c: New.
+	* gcc.target/arm/armv8_2-fp16-neon-3.c: New.
+
 2017-05-15  Jerry DeLisle  <jvdelisle@gcc.gnu.org>
 
 	PR libgfortran/80727

gcc/testsuite/gcc.target/arm/armv8_2-fp16-arith-1.c

@@ -3,7 +3,8 @@
 /* { dg-options "-O2 -ffast-math" }  */
 /* { dg-add-options arm_v8_2a_fp16_neon }  */
 
-/* Test instructions generated for half-precision arithmetic.  */
+/* Test instructions generated for half-precision arithmetic with
+   unsafe-math-optimizations enabled.  */
 
 typedef __fp16 float16_t;
 typedef __simd64_float16_t float16x4_t;
@@ -90,9 +91,18 @@ TEST_CMP (greaterthanqual, >=, int16x8_t, float16x8_t)
 /* { dg-final { scan-assembler-times {vneg\.f16\tq[0-9]+, q[0-9]+} 1 } }  */
 /* { dg-final { scan-assembler-times {vabs\.f16\ts[0-9]+, s[0-9]+} 2 } }  */
 
-/* { dg-final { scan-assembler-times {vadd\.f16\ts[0-9]+, s[0-9]+, s[0-9]+} 13 } }  */
-/* { dg-final { scan-assembler-times {vsub\.f16\ts[0-9]+, s[0-9]+, s[0-9]+} 13 } }  */
-/* { dg-final { scan-assembler-times {vmul\.f16\ts[0-9]+, s[0-9]+, s[0-9]+} 13 } }  */
+/* { dg-final { scan-assembler-times {vadd\.f16\ts[0-9]+, s[0-9]+, s[0-9]+} 1 } }  */
+/* { dg-final { scan-assembler-times {vadd\.f16\td[0-9]+, d[0-9]+, d[0-9]+} 1 } }  */
+/* { dg-final { scan-assembler-times {vadd\.f16\tq[0-9]+, q[0-9]+, q[0-9]+} 1 } }  */
+
+/* { dg-final { scan-assembler-times {vsub\.f16\ts[0-9]+, s[0-9]+, s[0-9]+} 1 } }  */
+/* { dg-final { scan-assembler-times {vsub\.f16\td[0-9]+, d[0-9]+, d[0-9]+} 1 } }  */
+/* { dg-final { scan-assembler-times {vsub\.f16\tq[0-9]+, q[0-9]+, q[0-9]+} 1 } }  */
+
+/* { dg-final { scan-assembler-times {vmul\.f16\ts[0-9]+, s[0-9]+, s[0-9]+} 1 } }  */
+/* { dg-final { scan-assembler-times {vmul\.f16\td[0-9]+, d[0-9]+, d[0-9]+} 1 } }  */
+/* { dg-final { scan-assembler-times {vmul\.f16\tq[0-9]+, q[0-9]+, q[0-9]+} 1 } }  */
+
 /* { dg-final { scan-assembler-times {vdiv\.f16\ts[0-9]+, s[0-9]+, s[0-9]+} 13 } }  */
 /* { dg-final { scan-assembler-times {vcmp\.f32\ts[0-9]+, s[0-9]+} 26 } }  */
 /* { dg-final { scan-assembler-times {vcmpe\.f32\ts[0-9]+, s[0-9]+} 52 } }  */

gcc/testsuite/gcc.target/arm/armv8_2-fp16-arith-2.c

+/* { dg-do compile }  */
+/* { dg-require-effective-target arm_v8_2a_fp16_neon_ok }  */
+/* { dg-options "-O2 -fno-fast-math" }  */
+/* { dg-add-options arm_v8_2a_fp16_neon }  */
+
+/* Test instructions generated for half-precision arithmetic without
+   unsafe-math-optimizations.  */
+
+typedef __fp16 float16_t;
+typedef __simd64_float16_t float16x4_t;
+typedef __simd128_float16_t float16x8_t;
+
+typedef short int16x4_t __attribute__ ((vector_size (8)));
+typedef short int int16x8_t  __attribute__ ((vector_size (16)));
+
+float16_t
+fp16_abs (float16_t a)
+{
+  return (a < 0) ? -a : a;
+}
+
+#define TEST_UNOP(NAME, OPERATOR, TY)		\
+  TY test_##NAME##_##TY (TY a)			\
+  {						\
+    return OPERATOR (a);			\
+  }
+
+#define TEST_BINOP(NAME, OPERATOR, TY)		\
+  TY test_##NAME##_##TY (TY a, TY b)		\
+  {						\
+    return a OPERATOR b;			\
+  }
+
+#define TEST_CMP(NAME, OPERATOR, RTY, TY)	\
+  RTY test_##NAME##_##TY (TY a, TY b)		\
+  {						\
+    return a OPERATOR b;			\
+  }
+
+/* Scalars.  */
+
+TEST_UNOP (neg, -, float16_t)
+TEST_UNOP (abs, fp16_abs, float16_t)
+
+TEST_BINOP (add, +, float16_t)
+TEST_BINOP (sub, -, float16_t)
+TEST_BINOP (mult, *, float16_t)
+TEST_BINOP (div, /, float16_t)
+
+TEST_CMP (equal, ==, int, float16_t)
+TEST_CMP (unequal, !=, int, float16_t)
+TEST_CMP (lessthan, <, int, float16_t)
+TEST_CMP (greaterthan, >, int, float16_t)
+TEST_CMP (lessthanequal, <=, int, float16_t)
+TEST_CMP (greaterthanqual, >=, int, float16_t)
+
+/* Vectors of size 4.  */
+
+TEST_UNOP (neg, -, float16x4_t)
+
+TEST_BINOP (add, +, float16x4_t)
+TEST_BINOP (sub, -, float16x4_t)
+TEST_BINOP (mult, *, float16x4_t)
+TEST_BINOP (div, /, float16x4_t)
+
+TEST_CMP (equal, ==, int16x4_t, float16x4_t)
+TEST_CMP (unequal, !=, int16x4_t, float16x4_t)
+TEST_CMP (lessthan, <, int16x4_t, float16x4_t)
+TEST_CMP (greaterthan, >, int16x4_t, float16x4_t)
+TEST_CMP (lessthanequal, <=, int16x4_t, float16x4_t)
+TEST_CMP (greaterthanqual, >=, int16x4_t, float16x4_t)
+
+/* Vectors of size 8.  */
+
+TEST_UNOP (neg, -, float16x8_t)
+
+TEST_BINOP (add, +, float16x8_t)
+TEST_BINOP (sub, -, float16x8_t)
+TEST_BINOP (mult, *, float16x8_t)
+TEST_BINOP (div, /, float16x8_t)
+
+TEST_CMP (equal, ==, int16x8_t, float16x8_t)
+TEST_CMP (unequal, !=, int16x8_t, float16x8_t)
+TEST_CMP (lessthan, <, int16x8_t, float16x8_t)
+TEST_CMP (greaterthan, >, int16x8_t, float16x8_t)
+TEST_CMP (lessthanequal, <=, int16x8_t, float16x8_t)
+TEST_CMP (greaterthanqual, >=, int16x8_t, float16x8_t)
+
+/* { dg-final { scan-assembler-times {vneg\.f16\ts[0-9]+, s[0-9]+} 1 } }  */
+/* { dg-final { scan-assembler-times {vneg\.f16\td[0-9]+, d[0-9]+} 1 } }  */
+/* { dg-final { scan-assembler-times {vneg\.f16\tq[0-9]+, q[0-9]+} 1 } }  */
+
+/* { dg-final { scan-assembler-times {vadd\.f16\ts[0-9]+, s[0-9]+, s[0-9]+} 13 } }  */
+/* { dg-final { scan-assembler-times {vsub\.f16\ts[0-9]+, s[0-9]+, s[0-9]+} 13 } }  */
+/* { dg-final { scan-assembler-times {vmul\.f16\ts[0-9]+, s[0-9]+, s[0-9]+} 13 } }  */
+/* { dg-final { scan-assembler-times {vdiv\.f16\ts[0-9]+, s[0-9]+, s[0-9]+} 13 } }  */
+/* { dg-final { scan-assembler-times {vcmp\.f32\ts[0-9]+, s[0-9]+} 26 } }  */
+
+/* { dg-final { scan-assembler-times {vcmpe\.f32\ts[0-9]+, s[0-9]+} 52 } }  */
+/* { dg-final { scan-assembler-times {vcmpe\.f32\ts[0-9]+, #0} 2 } }  */
+
+/* { dg-final { scan-assembler-not {vabs\.f16} } }  */
+
+/* { dg-final { scan-assembler-not {vadd\.f32} } }  */
+/* { dg-final { scan-assembler-not {vsub\.f32} } }  */
+/* { dg-final { scan-assembler-not {vmul\.f32} } }  */
+/* { dg-final { scan-assembler-not {vdiv\.f32} } }  */
+/* { dg-final { scan-assembler-not {vcmp\.f16} } }  */
+/* { dg-final { scan-assembler-not {vcmpe\.f16} } }  */

gcc/testsuite/gcc.target/arm/armv8_2-fp16-neon-3.c

+/* { dg-do compile }  */
+/* { dg-require-effective-target arm_v8_2a_fp16_neon_ok }  */
+/* { dg-options "-O2 -ffast-math" }  */
+/* { dg-add-options arm_v8_2a_fp16_neon }  */
+
+/* Test compiler use of FP16 FMA/FMS instructions with -ffast-math.  */
+
+#include <arm_neon.h>
+
+float16x4_t
+test_vfma_1 (float16x4_t a, float16x4_t b, float16x4_t c)
+{
+  return vadd_f16 (vmul_f16 (a, b), c);
+}
+
+float16x4_t
+test_vfma_2 (float16x4_t a, float16x4_t b, float16x4_t c)
+{
+  return vsub_f16 (vmul_f16 (a, b), vneg_f16 (c));
+}
+
+float16x4_t
+test_vfma_3 (float16x4_t a, float16x4_t b, float16x4_t c)
+{
+  return vsub_f16 (vmul_f16 (vneg_f16 (a), vneg_f16 (b)), vneg_f16 (c));
+}
+
+float16x4_t
+test_vfma_4 (float16x4_t a, float16x4_t b, float16x4_t c)
+{
+  return vsub_f16 (vmul_f16 (a, b), vneg_f16 (c));
+}
+/* { dg-final { scan-assembler-times {vfma\.f16\td[0-9]+, d[0-9]+, d[0-9]+} 4 } }  */
+
+float16x8_t
+test_vfmaq_1 (float16x8_t a, float16x8_t b, float16x8_t c)
+{
+  return vaddq_f16 (vmulq_f16 (a, b), c);
+}
+
+float16x8_t
+test_vfmaq_2 (float16x8_t a, float16x8_t b, float16x8_t c)
+{
+  return vsubq_f16 (vmulq_f16 (a, b), vnegq_f16 (c));
+}
+
+float16x8_t
+test_vfmaq_3 (float16x8_t a, float16x8_t b, float16x8_t c)
+{
+  return vsubq_f16 (vmulq_f16 (vnegq_f16 (a), vnegq_f16 (b)), vnegq_f16 (c));
+}
+
+float16x8_t
+test_vfmaq_4 (float16x8_t a, float16x8_t b, float16x8_t c)
+{
+  return vsubq_f16 (vmulq_f16 (a, b), vnegq_f16 (c));
+}
+/* { dg-final { scan-assembler-times {vfma\.f16\tq[0-9]+, q[0-9]+, q[0-9]+} 4 } }  */
+
+float16x4_t
+test_vfms_1 (float16x4_t a, float16x4_t b, float16x4_t c)
+{
+  return vsub_f16 (c, vmul_f16 (a, b));
+}
+
+float16x4_t
+test_vfms_2 (float16x4_t a, float16x4_t b, float16x4_t c)
+{
+  return vsub_f16 (a, vmul_f16 (b, c));
+}
+
+float16x4_t
+test_vfms_3 (float16x4_t a, float16x4_t b, float16x4_t c)
+{
+  return vadd_f16 (vmul_f16 (vneg_f16 (a), b), c);
+}
+
+float16x4_t
+test_vfms_4 (float16x4_t a, float16x4_t b, float16x4_t c)
+{
+  return vadd_f16 (vmul_f16 (a, vneg_f16 (b)), c);
+}
+/* { dg-final { scan-assembler-times {vfms\.f16\td[0-9]+, d[0-9]+, d[0-9]+} 4 } } */
+
+float16x8_t
+test_vfmsq_1 (float16x8_t a, float16x8_t b, float16x8_t c)
+{
+  return vsubq_f16 (c, vmulq_f16 (a, b));
+}
+
+float16x8_t
+test_vfmsq_2 (float16x8_t a, float16x8_t b, float16x8_t c)
+{
+  return vsubq_f16 (a, vmulq_f16 (b, c));
+}
+
+float16x8_t
+test_vfmsq_3 (float16x8_t a, float16x8_t b, float16x8_t c)
+{
+  return vaddq_f16 (vmulq_f16 (vnegq_f16 (a), b), c);
+}
+
+float16x8_t
+test_vfmsq_4 (float16x8_t a, float16x8_t b, float16x8_t c)
+{
+  return vaddq_f16 (vmulq_f16 (a, vnegq_f16 (b)), c);
+}
+/* { dg-final { scan-assembler-times {vfms\.f16\tq[0-9]+, q[0-9]+, q[0-9]+} 4 } } */