From 87298cb451a7650f8d2d24f34f1ddd662c2aeecc Mon Sep 17 00:00:00 2001
From: Geolm <moyaguillaume@gmail.com>
Date: Mon, 29 Jan 2024 17:36:55 -0500
Subject: [PATCH] added benchmark on neon

---
 tests/benchmark.c | 23 ++++++++++++++++++++++-
 1 file changed, 22 insertions(+), 1 deletion(-)

diff --git a/tests/benchmark.c b/tests/benchmark.c
index b400604..7d2c21e 100644
--- a/tests/benchmark.c
+++ b/tests/benchmark.c
@@ -40,7 +40,19 @@ int benchmark(approximation_function function, const char* name)
 
     output = _mm256_cvtss_f32(result);
 #else
+    float32x4_t step = vdupq_n_f32(FLT_EPSILON);
+    float32x4_t input = vld1q_f32(init_array);
+    float32x4_t result = vdupq_n_f32(0.f);
 
+     start = stm_now();
+
+    for(uint32_t i=0; i<NUM_ITERATIONS; ++i)
+    {
+        result = vaddq_f32(result, function(input));
+        input = vaddq_f32(input, step);
+    }
+
+    output = vgetq_lane_f32(result, 0);
 #endif
 
     printf(".%s: %3.3f ms \n", name, stm_ms(stm_since(start)));
@@ -73,7 +85,16 @@ int main(int argc, char * argv[])
     output += benchmark(mm256_log_ps, "mm256_log_ps");
     output += benchmark(mm256_log2_ps, "mm256_log2_ps");
 #else
-
+    output += benchmark(vacosq_f32, "vacosq_f32");
+    output += benchmark(vasinq_f32, "vasinq_f32");
+    output += benchmark(vatanq_f32, "vatanq_f32");
+    output += benchmark(vcbrtq_f32, "vcbrtq_f32");
+    output += benchmark(vcosq_f32, "vcosq_f32");
+    output += benchmark(vsinq_f32, "vsinq_f32");
+    output += benchmark(vexpq_f32, "vexpq_f32");
+    output += benchmark(vexp2q_f32, "vexp2q_f32");
+    output += benchmark(vlogq_f32, "vlogq_f32");
+    output += benchmark(vlog2q_f32, "vlog2q_f32");
 #endif
 
     printf("\n%d\n", output);