Merge pull request #500 from howjmay/vsqrt

feat: Add vsqrt[q]_[f32|f64]
howjmay · Aug 1, 2024 · c235c2a · c235c2a
2 parents 4e39149 + 3f8d547
commit c235c2a
Show file tree

Hide file tree

Showing 3 changed files with 72 additions and 12 deletions.
diff --git a/neon2rvv.h b/neon2rvv.h
@@ -4800,13 +4800,13 @@ FORCE_INLINE float32_t vrecpss_f32(float32_t a, float32_t b) { return 2.0 - a *
 
 FORCE_INLINE float64_t vrecpsd_f64(float64_t a, float64_t b) { return 2.0 - a * b; }
 
-// FORCE_INLINE float32x2_t vsqrt_f32(float32x2_t a);
+FORCE_INLINE float32x2_t vsqrt_f32(float32x2_t a) { return __riscv_vfsqrt_v_f32m1(a, 2); }
 
-// FORCE_INLINE float32x4_t vsqrtq_f32(float32x4_t a);
+FORCE_INLINE float32x4_t vsqrtq_f32(float32x4_t a) { return __riscv_vfsqrt_v_f32m1(a, 4); }
 
-// FORCE_INLINE float64x1_t vsqrt_f64(float64x1_t a);
+FORCE_INLINE float64x1_t vsqrt_f64(float64x1_t a) { return __riscv_vfsqrt_v_f64m1(a, 1); }
 
-// FORCE_INLINE float64x2_t vsqrtq_f64(float64x2_t a);
+FORCE_INLINE float64x2_t vsqrtq_f64(float64x2_t a) { return __riscv_vfsqrt_v_f64m1(a, 2); }
 
 FORCE_INLINE float32x2_t vrsqrts_f32(float32x2_t a, float32x2_t b) {
   return __riscv_vfdiv_vf_f32m1(__riscv_vfnmsac_vv_f32m1(vdup_n_f32(3.0), a, b, 2), 2.0, 2);

diff --git a/tests/impl.cpp b/tests/impl.cpp
@@ -17010,13 +17010,73 @@ result_t test_vrecpsd_f64(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) {
 #endif  // ENABLE_TEST_ALL
 }
 
-result_t test_vsqrt_f32(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) { return TEST_UNIMPL; }
+result_t test_vsqrt_f32(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) {
+#ifdef ENABLE_TEST_ALL
+  const float *_a = (const float *)impl.test_cases_float_pointer1;
+  float _c[2];
+  for (int i = 0; i < 2; i++) {
+    _c[i] = sqrtf(_a[i]);
+  }
+
+  float32x2_t a = vld1_f32(_a);
+  float32x2_t c = vsqrt_f32(a);
+
+  return validate_float_error(c, _c[0], _c[1], 0.0001f);
+#else
+  return TEST_UNIMPL;
+#endif  // ENABLE_TEST_ALL
+}
+
+result_t test_vsqrtq_f32(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) {
+#ifdef ENABLE_TEST_ALL
+  const float *_a = (const float *)impl.test_cases_float_pointer1;
+  float _c[4];
+  for (int i = 0; i < 4; i++) {
+    _c[i] = sqrtf(_a[i]);
+  }
+
+  float32x4_t a = vld1q_f32(_a);
+  float32x4_t c = vsqrtq_f32(a);
+
+  return validate_float_error(c, _c[0], _c[1], _c[2], _c[3], 0.0001f);
+#else
+  return TEST_UNIMPL;
+#endif  // ENABLE_TEST_ALL
+}
+
+result_t test_vsqrt_f64(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) {
+#ifdef ENABLE_TEST_ALL
+  const double *_a = (const double *)impl.test_cases_float_pointer1;
+  double _c[2];
+  for (int i = 0; i < 1; i++) {
+    _c[i] = sqrt(_a[i]);
+  }
+
+  float64x1_t a = vld1_f64(_a);
+  float64x1_t c = vsqrt_f64(a);
+
+  return validate_double_error(c, _c[0], 0.0001f);
+#else
+  return TEST_UNIMPL;
+#endif  // ENABLE_TEST_ALL
+}
 
-result_t test_vsqrtq_f32(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) { return TEST_UNIMPL; }
+result_t test_vsqrtq_f64(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) {
+#ifdef ENABLE_TEST_ALL
+  const double *_a = (const double *)impl.test_cases_float_pointer1;
+  double _c[4];
+  for (int i = 0; i < 4; i++) {
+    _c[i] = sqrt(_a[i]);
+  }
 
-result_t test_vsqrt_f64(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) { return TEST_UNIMPL; }
+  float64x2_t a = vld1q_f64(_a);
+  float64x2_t c = vsqrtq_f64(a);
 
-result_t test_vsqrtq_f64(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) { return TEST_UNIMPL; }
+  return validate_double_error(c, _c[0], _c[1], 0.0001f);
+#else
+  return TEST_UNIMPL;
+#endif  // ENABLE_TEST_ALL
+}
 
 result_t test_vrsqrts_f32(const NEON2RVV_TEST_IMPL &impl, uint32_t iter) {
 #ifdef ENABLE_TEST_ALL

diff --git a/tests/impl.h b/tests/impl.h
@@ -961,10 +961,10 @@
   _(vrecpsq_f64)                                                                 \
   _(vrecpss_f32)                                                                 \
   _(vrecpsd_f64)                                                                 \
-  /*_(vsqrt_f32)                                                              */ \
-  /*_(vsqrtq_f32)                                                             */ \
-  /*_(vsqrt_f64)                                                              */ \
-  /*_(vsqrtq_f64)                                                             */ \
+  _(vsqrt_f32)                                                                   \
+  _(vsqrtq_f32)                                                                  \
+  _(vsqrt_f64)                                                                   \
+  _(vsqrtq_f64)                                                                  \
   _(vrsqrts_f32)                                                                 \
   _(vrsqrtsq_f32)                                                                \
   _(vrsqrts_f64)                                                                 \