From 35c066a39b07d607a5270d66d274037aee06f103 Mon Sep 17 00:00:00 2001
From: hev <r@hev.cc>
Date: Wed, 29 May 2024 04:29:47 +0800
Subject: [PATCH] fix: f32 reduce_min for x86 (#2385)

Shift did not reach only 1 value left.

For f32x8:

```
// min(4 elements)
m1  = [v0, v1, v2, v3, v4, v5, v6, v7]
m2  = [v4, v5, v6, v7, xx, xx, xx, xx]
min = [v0, v5, v2, v3, xx, xx, xx, xx]

// min(2 elements)
m1  = [v0, v5, v2, v3, xx, xx, xx, xx]
m2  = [v2, v3, xx, xx, xx, xx, xx, xx]
min = [v0, v3, xx, xx, xx, xx, xx, xx]

// min(1 element) (This step is missing now)
m1  = [v0, v3, xx, xx, xx, xx, xx, xx]
m2  = [v3, xx, xx, xx, xx, xx, xx, xx]
min = [v0, xx, xx, xx, xx, xx, xx, xx]
```
---
 rust/lance-linalg/src/simd/f32.rs | 22 ++++++++++++++++------
 1 file changed, 16 insertions(+), 6 deletions(-)

diff --git a/rust/lance-linalg/src/simd/f32.rs b/rust/lance-linalg/src/simd/f32.rs
index f365185fa3..c204e116a9 100644
--- a/rust/lance-linalg/src/simd/f32.rs
+++ b/rust/lance-linalg/src/simd/f32.rs
@@ -177,9 +177,9 @@ impl SIMD<f32, 8> for f32x8 {
                 min = _mm256_min_ps(min, shift);
                 shift = _mm256_permute_ps(min, 14);
                 min = _mm256_min_ps(min, shift);
-                let mut results: [f32; 8] = [0f32; 8];
-                _mm256_storeu_ps(results.as_mut_ptr(), min);
-                results[0]
+                shift = _mm256_permute_ps(min, 1);
+                min = _mm256_min_ps(min, shift);
+                _mm256_cvtss_f32(min)
             }
         }
         #[cfg(target_arch = "aarch64")]
@@ -520,9 +520,9 @@ impl SIMD<f32, 16> for f32x16 {
             m1 = _mm256_min_ps(m1, m2);
             m2 = _mm256_permute_ps(m1, 14);
             m1 = _mm256_min_ps(m1, m2);
-            let mut results: [f32; 8] = [0f32; 8];
-            _mm256_storeu_ps(results.as_mut_ptr(), m1);
-            results[0]
+            m2 = _mm256_permute_ps(m1, 1);
+            m1 = _mm256_min_ps(m1, m2);
+            _mm256_cvtss_f32(m1)
         }
 
         #[cfg(target_arch = "aarch64")]
@@ -782,8 +782,10 @@ mod tests {
     fn test_f32x8_cmp_ops() {
         let a = [1.0_f32, 2.0, 5.0, 6.0, 7.0, 3.0, 2.0, 1.0];
         let b = [2.0_f32, 1.0, 4.0, 5.0, 9.0, 5.0, 6.0, 2.0];
+        let c = [2.0_f32, 1.0, 4.0, 5.0, 7.0, 3.0, 2.0, 1.0];
         let simd_a: f32x8 = (&a).into();
         let simd_b: f32x8 = (&b).into();
+        let simd_c: f32x8 = (&c).into();
 
         let min_simd = simd_a.min(&simd_b);
         assert_eq!(
@@ -792,6 +794,8 @@ mod tests {
         );
         let min_val = min_simd.reduce_min();
         assert_eq!(min_val, 1.0);
+        let min_val = simd_c.reduce_min();
+        assert_eq!(min_val, 1.0);
 
         assert_eq!(Some(2), simd_a.find(5.0));
         assert_eq!(Some(1), simd_a.find(2.0));
@@ -836,8 +840,12 @@ mod tests {
         let b = [
             2.0_f32, 1.0, 4.0, 5.0, 9.0, 5.0, 6.0, 2.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 2.0, 1.0,
         ];
+        let c = [
+            1.0_f32, 1.0, 4.0, 5.0, 7.0, 3.0, 2.0, 1.0, -0.5, 5.0, 6.0, 7.0, 8.0, 9.0, 1.0, -1.0,
+        ];
         let simd_a: f32x16 = (&a).into();
         let simd_b: f32x16 = (&b).into();
+        let simd_c: f32x16 = (&c).into();
 
         let min_simd = simd_a.min(&simd_b);
         assert_eq!(
@@ -846,6 +854,8 @@ mod tests {
         );
         let min_val = min_simd.reduce_min();
         assert_eq!(min_val, -0.5);
+        let min_val = simd_c.reduce_min();
+        assert_eq!(min_val, -1.0);
 
         assert_eq!(Some(2), simd_a.find(5.0));
         assert_eq!(Some(1), simd_a.find(2.0));