From b8b5ee4934c861fa26286e955c288a99ce3c2027 Mon Sep 17 00:00:00 2001
From: Raghuveer Devulapalli <raghuveer.devulapalli@intel.com>
Date: Fri, 22 Dec 2023 15:51:42 -0800
Subject: [PATCH 1/2] Add inlien keyword to avx512_qsort _Float16

---
 src/avx512fp16-16bit-qsort.hpp | 4 ++--
 1 file changed, 2 insertions(+), 2 deletions(-)

diff --git a/src/avx512fp16-16bit-qsort.hpp b/src/avx512fp16-16bit-qsort.hpp
index 8a9a49ed..1206f822 100644
--- a/src/avx512fp16-16bit-qsort.hpp
+++ b/src/avx512fp16-16bit-qsort.hpp
@@ -145,7 +145,7 @@ replace_inf_with_nan(_Float16 *arr, int64_t arrsize, int64_t nan_count)
 }
 
 template <>
-void avx512_qselect(_Float16 *arr, int64_t k, int64_t arrsize)
+inline void avx512_qselect(_Float16 *arr, int64_t k, int64_t arrsize)
 {
     if (arrsize > 1) {
         int64_t nan_count = replace_nan_with_inf(arr, arrsize);
@@ -156,7 +156,7 @@ void avx512_qselect(_Float16 *arr, int64_t k, int64_t arrsize)
 }
 
 template <>
-void avx512_qsort(_Float16 *arr, int64_t arrsize)
+inline void avx512_qsort(_Float16 *arr, int64_t arrsize)
 {
     if (arrsize > 1) {
         int64_t nan_count = replace_nan_with_inf(arr, arrsize);

From c20fb421455db62e764897f07c421b8b5b39fa7d Mon Sep 17 00:00:00 2001
From: Raghuveer Devulapalli <raghuveer.devulapalli@intel.com>
Date: Fri, 22 Dec 2023 15:59:04 -0800
Subject: [PATCH 2/2] More inline

---
 src/avx512-16bit-qsort.hpp | 14 +++++++-------
 1 file changed, 7 insertions(+), 7 deletions(-)

diff --git a/src/avx512-16bit-qsort.hpp b/src/avx512-16bit-qsort.hpp
index 606f8706..2eb44542 100644
--- a/src/avx512-16bit-qsort.hpp
+++ b/src/avx512-16bit-qsort.hpp
@@ -350,7 +350,7 @@ struct zmm_vector<uint16_t> {
 };
 
 template <>
-bool comparison_func<zmm_vector<float16>>(const uint16_t &a, const uint16_t &b)
+inline bool comparison_func<zmm_vector<float16>>(const uint16_t &a, const uint16_t &b)
 {
     uint16_t signa = a & 0x8000, signb = b & 0x8000;
     uint16_t expa = a & 0x7c00, expb = b & 0x7c00;
@@ -406,7 +406,7 @@ replace_inf_with_nan(uint16_t *arr, int64_t arrsize, int64_t nan_count)
 }
 
 template <>
-void avx512_qselect(int16_t *arr, int64_t k, int64_t arrsize)
+inline void avx512_qselect(int16_t *arr, int64_t k, int64_t arrsize)
 {
     if (arrsize > 1) {
         qselect_16bit_<zmm_vector<int16_t>, int16_t>(
@@ -415,7 +415,7 @@ void avx512_qselect(int16_t *arr, int64_t k, int64_t arrsize)
 }
 
 template <>
-void avx512_qselect(uint16_t *arr, int64_t k, int64_t arrsize)
+inline void avx512_qselect(uint16_t *arr, int64_t k, int64_t arrsize)
 {
     if (arrsize > 1) {
         qselect_16bit_<zmm_vector<uint16_t>, uint16_t>(
@@ -423,7 +423,7 @@ void avx512_qselect(uint16_t *arr, int64_t k, int64_t arrsize)
     }
 }
 
-void avx512_qselect_fp16(uint16_t *arr, int64_t k, int64_t arrsize)
+inline void avx512_qselect_fp16(uint16_t *arr, int64_t k, int64_t arrsize)
 {
     if (arrsize > 1) {
         int64_t nan_count = replace_nan_with_inf(arr, arrsize);
@@ -434,7 +434,7 @@ void avx512_qselect_fp16(uint16_t *arr, int64_t k, int64_t arrsize)
 }
 
 template <>
-void avx512_qsort(int16_t *arr, int64_t arrsize)
+inline void avx512_qsort(int16_t *arr, int64_t arrsize)
 {
     if (arrsize > 1) {
         qsort_16bit_<zmm_vector<int16_t>, int16_t>(
@@ -443,7 +443,7 @@ void avx512_qsort(int16_t *arr, int64_t arrsize)
 }
 
 template <>
-void avx512_qsort(uint16_t *arr, int64_t arrsize)
+inline void avx512_qsort(uint16_t *arr, int64_t arrsize)
 {
     if (arrsize > 1) {
         qsort_16bit_<zmm_vector<uint16_t>, uint16_t>(
@@ -451,7 +451,7 @@ void avx512_qsort(uint16_t *arr, int64_t arrsize)
     }
 }
 
-void avx512_qsort_fp16(uint16_t *arr, int64_t arrsize)
+inline void avx512_qsort_fp16(uint16_t *arr, int64_t arrsize)
 {
     if (arrsize > 1) {
         int64_t nan_count = replace_nan_with_inf(arr, arrsize);