Getting rid of extra CR characters at ends of lines.

git-svn-id: svn://svn.icms.temple.edu/lammps-ro/trunk@5285 f3b2605a-c512-4ea7-a41b-209d697bcdaa

lib/gpu/atomic_gpu_memory.cpp

@@ -10,11 +10,11 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
 ------------------------------------------------------------------------- */
-
+ 
 #include "atomic_gpu_memory.h"
 #define AtomicGPUMemoryT AtomicGPUMemory<numtyp, acctyp>
 

lib/gpu/atomic_gpu_memory.h

@@ -10,7 +10,7 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
 ------------------------------------------------------------------------- */

lib/gpu/charge_gpu_memory.cpp

@@ -10,7 +10,7 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
 ------------------------------------------------------------------------- */

lib/gpu/charge_gpu_memory.h

@@ -10,7 +10,7 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
 ------------------------------------------------------------------------- */

lib/gpu/cmm_cut_gpu.cpp

@@ -10,7 +10,7 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
 ------------------------------------------------------------------------- */

lib/gpu/cmm_cut_gpu_memory.cpp

@@ -10,7 +10,7 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
 ------------------------------------------------------------------------- */

lib/gpu/cmm_cut_gpu_memory.h

@@ -10,7 +10,7 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
 ------------------------------------------------------------------------- */

lib/gpu/cmmc_long_gpu.cpp

@@ -10,7 +10,7 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
 ------------------------------------------------------------------------- */

lib/gpu/cmmc_long_gpu_memory.cpp

@@ -10,7 +10,7 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
 ------------------------------------------------------------------------- */

lib/gpu/cmmc_long_gpu_memory.h

@@ -10,7 +10,7 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
 ------------------------------------------------------------------------- */

lib/gpu/cudpp_mini/cta/radixsort_cta.cu

@@ -1,337 +1,337 @@
-// -------------------------------------------------------------
+// -------------------------------------------------------------
-// CUDPP -- CUDA Data Parallel Primitives library
+// CUDPP -- CUDA Data Parallel Primitives library
-// -------------------------------------------------------------
+// -------------------------------------------------------------
-// $Revision$
+// $Revision$
-// $Date$
+// $Date$
-// ------------------------------------------------------------- 
+// ------------------------------------------------------------- 
-// This source code is distributed under the terms of license.txt 
+// This source code is distributed under the terms of license.txt 
-// in the root directory of this source distribution.
+// in the root directory of this source distribution.
-// ------------------------------------------------------------- 
+// ------------------------------------------------------------- 
-#include <cudpp_globals.h>
+#include <cudpp_globals.h>
-#include "cudpp_radixsort.h"
+#include "cudpp_radixsort.h"
-#include "cta/scan_cta.cu"
+#include "cta/scan_cta.cu"
-#include <cudpp.h>
+#include <cudpp.h>
-#include <stdio.h>
+#include <stdio.h>
-
+ 
-#include <cudpp_util.h>
+#include <cudpp_util.h>
-#include <math.h>
+#include <math.h>
-#include "sharedmem.h"
+#include "sharedmem.h"
-
+
-
+
-#ifdef __DEVICE_EMULATION__
+#ifdef __DEVICE_EMULATION__
-#define __EMUSYNC __syncthreads()
+#define __EMUSYNC __syncthreads()
-#else
+#else
-#define __EMUSYNC
+#define __EMUSYNC
-#endif
+#endif
-
+
-/**
+/**
- * @file
+ * @file
- * sort_cta.cu
+ * sort_cta.cu
- * 
+ * 
- * @brief CUDPP CTA-level sort routines
+ * @brief CUDPP CTA-level sort routines
- */
+ */
-
+
-/** \addtogroup cudpp_cta 
+/** \addtogroup cudpp_cta 
-* @{
+* @{
-*/
+*/
-
+
-/** @name Radix Sort Functions
+/** @name Radix Sort Functions
-* @{
+* @{
-*/
+*/
-
+
-
+
-typedef unsigned int uint;
+typedef unsigned int uint;
-
+
-/**
+/**
- * @brief Flips bits of single-precision floating-point number (parameterized by doFlip)
+ * @brief Flips bits of single-precision floating-point number (parameterized by doFlip)
- * 
+ * 
- *  flip a float for sorting
+ *  flip a float for sorting
- *  finds SIGN of fp number.
+ *  finds SIGN of fp number.
- *  if it's 1 (negative float), it flips all bits
+ *  if it's 1 (negative float), it flips all bits
- *  if it's 0 (positive float), it flips the sign only
+ *  if it's 0 (positive float), it flips the sign only
- * @param[in] f floating-point input (passed as unsigned int)
+ * @param[in] f floating-point input (passed as unsigned int)
- * @see floatUnflip
+ * @see floatUnflip
-**/
+**/
-
+
-template <bool doFlip>
+template <bool doFlip>
-__device__ uint floatFlip(uint f)
+__device__ uint floatFlip(uint f)
-{
+{
-    if (doFlip)
+    if (doFlip)
-    {
+    {
-        uint mask = -int(f >> 31) | 0x80000000;
+        uint mask = -int(f >> 31) | 0x80000000;
-        return f ^ mask;
+        return f ^ mask;
-    }
+    }
-    else
+    else
-        return f;
+        return f;
-}
+}
-
+
-/**
+/**
- * @brief Reverses bit-flip of single-precision floating-point number (parameterized by doFlip)
+ * @brief Reverses bit-flip of single-precision floating-point number (parameterized by doFlip)
- * 
+ * 
- * flip a float back (invert FloatFlip)
+ * flip a float back (invert FloatFlip)
- *  signed was flipped from above, so:
+ *  signed was flipped from above, so:
- *  if sign is 1 (negative), it flips the sign bit back
+ *  if sign is 1 (negative), it flips the sign bit back
- *  if sign is 0 (positive), it flips all bits back
+ *  if sign is 0 (positive), it flips all bits back
- * @param[in] f floating-point input (passed as unsigned int)
+ * @param[in] f floating-point input (passed as unsigned int)
- * @see floatFlip
+ * @see floatFlip
-**/
+**/
-template <bool doFlip>
+template <bool doFlip>
-__device__ uint floatUnflip(uint f)
+__device__ uint floatUnflip(uint f)
-{
+{
-    if (doFlip)
+    if (doFlip)
-    {
+    {
-        uint mask = ((f >> 31) - 1) | 0x80000000;
+        uint mask = ((f >> 31) - 1) | 0x80000000;
-        return f ^ mask;
+        return f ^ mask;
-    }
+    }
-    else
+    else
-        return f;
+        return f;
-}
+}
-
+
-/**
+/**
- * @brief Scans one warp quickly, optimized for 32-element warps, using shared memory
+ * @brief Scans one warp quickly, optimized for 32-element warps, using shared memory
- * 
+ * 
- * Scans each warp in parallel ("warp-scan"), one element per thread.
+ * Scans each warp in parallel ("warp-scan"), one element per thread.
- * uses 2 numElements of shared memory per thread (64 numElements per warp)
+ * uses 2 numElements of shared memory per thread (64 numElements per warp)
- * 
+ * 
- * @param[in] val Elements per thread to scan
+ * @param[in] val Elements per thread to scan
- * @param[in,out] sData
+ * @param[in,out] sData
-**/
+**/
-template<class T, int maxlevel>
+template<class T, int maxlevel>
-__device__ T scanwarp(T val, volatile T* sData)
+__device__ T scanwarp(T val, volatile T* sData)
-{
+{
-    // The following is the same as 2 * WARP_SIZE * warpId + threadInWarp = 
+    // The following is the same as 2 * WARP_SIZE * warpId + threadInWarp = 
-    // 64*(threadIdx.x >> 5) + (threadIdx.x & (WARP_SIZE - 1))
+    // 64*(threadIdx.x >> 5) + (threadIdx.x & (WARP_SIZE - 1))
-    int idx = 2 * threadIdx.x - (threadIdx.x & (WARP_SIZE - 1));
+    int idx = 2 * threadIdx.x - (threadIdx.x & (WARP_SIZE - 1));
-    sData[idx] = 0;
+    sData[idx] = 0;
-    idx += WARP_SIZE;
+    idx += WARP_SIZE;
-    T t = sData[idx] = val;          __EMUSYNC;
+    T t = sData[idx] = val;          __EMUSYNC;
-
+
-#ifdef __DEVICE_EMULATION__             
+#ifdef __DEVICE_EMULATION__             
-        t = sData[idx -  1]; __EMUSYNC; 
+        t = sData[idx -  1]; __EMUSYNC; 
-        sData[idx] += t;       __EMUSYNC;
+        sData[idx] += t;       __EMUSYNC;
-        t = sData[idx -  2];   __EMUSYNC; 
+        t = sData[idx -  2];   __EMUSYNC; 
-        sData[idx] += t;       __EMUSYNC;
+        sData[idx] += t;       __EMUSYNC;
-        t = sData[idx -  4];   __EMUSYNC; 
+        t = sData[idx -  4];   __EMUSYNC; 
-        sData[idx] += t;       __EMUSYNC;
+        sData[idx] += t;       __EMUSYNC;
-        t = sData[idx -  8];   __EMUSYNC; 
+        t = sData[idx -  8];   __EMUSYNC; 
-        sData[idx] += t;       __EMUSYNC;
+        sData[idx] += t;       __EMUSYNC;
-        t = sData[idx - 16];   __EMUSYNC; 
+        t = sData[idx - 16];   __EMUSYNC; 
-        sData[idx] += t;       __EMUSYNC;
+        sData[idx] += t;       __EMUSYNC;
-#else
+#else
-        if (0 <= maxlevel) { sData[idx] = t = t + sData[idx - 1]; } __EMUSYNC;
+        if (0 <= maxlevel) { sData[idx] = t = t + sData[idx - 1]; } __EMUSYNC;
-        if (1 <= maxlevel) { sData[idx] = t = t + sData[idx - 2]; } __EMUSYNC;
+        if (1 <= maxlevel) { sData[idx] = t = t + sData[idx - 2]; } __EMUSYNC;
-        if (2 <= maxlevel) { sData[idx] = t = t + sData[idx - 4]; } __EMUSYNC;
+        if (2 <= maxlevel) { sData[idx] = t = t + sData[idx - 4]; } __EMUSYNC;
-        if (3 <= maxlevel) { sData[idx] = t = t + sData[idx - 8]; } __EMUSYNC;
+        if (3 <= maxlevel) { sData[idx] = t = t + sData[idx - 8]; } __EMUSYNC;
-        if (4 <= maxlevel) { sData[idx] = t = t + sData[idx -16]; } __EMUSYNC;
+        if (4 <= maxlevel) { sData[idx] = t = t + sData[idx -16]; } __EMUSYNC;
-#endif          
+#endif          
-        return sData[idx] - val;  // convert inclusive -> exclusive
+        return sData[idx] - val;  // convert inclusive -> exclusive
-}
+}
-
+
-/**
+/**
- * @brief Scans 4*CTA_SIZE unsigned ints in a block
+ * @brief Scans 4*CTA_SIZE unsigned ints in a block
- *
+ *
- * scan4 scans 4*CTA_SIZE numElements in a block (4 per
+ * scan4 scans 4*CTA_SIZE numElements in a block (4 per
- * thread), using a warp-scan algorithm
+ * thread), using a warp-scan algorithm
- * 
+ * 
- * @param[in] idata 4-vector of integers to scan
+ * @param[in] idata 4-vector of integers to scan
-**/
+**/
-__device__ uint4 scan4(uint4 idata)
+__device__ uint4 scan4(uint4 idata)
-{    
+{    
-    extern  __shared__  uint ptr[];
+    extern  __shared__  uint ptr[];
-    
+    
-    uint idx = threadIdx.x;
+    uint idx = threadIdx.x;
-
+
-    uint4 val4 = idata;
+    uint4 val4 = idata;
-    uint sum[3];
+    uint sum[3];
-    sum[0] = val4.x;
+    sum[0] = val4.x;
-    sum[1] = val4.y + sum[0];
+    sum[1] = val4.y + sum[0];
-    sum[2] = val4.z + sum[1];
+    sum[2] = val4.z + sum[1];
-    
+    
-    uint val = val4.w + sum[2];
+    uint val = val4.w + sum[2];
-    
+    
-    val = scanwarp<uint, 4>(val, ptr);
+    val = scanwarp<uint, 4>(val, ptr);
-    __syncthreads();
+    __syncthreads();
-
+
-    if ((idx & (WARP_SIZE - 1)) == WARP_SIZE - 1)
+    if ((idx & (WARP_SIZE - 1)) == WARP_SIZE - 1)
-    {
+    {
-        ptr[idx >> 5] = val + val4.w + sum[2];
+        ptr[idx >> 5] = val + val4.w + sum[2];
-    }
+    }
-    __syncthreads();
+    __syncthreads();
-
+
-#ifndef __DEVICE_EMULATION__
+#ifndef __DEVICE_EMULATION__
-    if (idx < WARP_SIZE)
+    if (idx < WARP_SIZE)
-#endif
+#endif
-    {
+    {
-        ptr[idx] = scanwarp<uint, 2>(ptr[idx], ptr);
+        ptr[idx] = scanwarp<uint, 2>(ptr[idx], ptr);
-    }
+    }
-    __syncthreads();
+    __syncthreads();
-
+
-    val += ptr[idx >> 5];
+    val += ptr[idx >> 5];
-
+
-    val4.x = val;
+    val4.x = val;
-    val4.y = val + sum[0];
+    val4.y = val + sum[0];
-    val4.z = val + sum[1];
+    val4.z = val + sum[1];
-    val4.w = val + sum[2];      
+    val4.w = val + sum[2];      
-        
+        
-    return val4;
+    return val4;
-}
+}
-
+
-/**
+/**
- * @brief Computes output position for each thread given predicate; trues come first then falses
+ * @brief Computes output position for each thread given predicate; trues come first then falses
- * 
+ * 
- * Rank is the core of the radix sort loop.  Given a predicate, it
+ * Rank is the core of the radix sort loop.  Given a predicate, it
- * computes the output position for each thread in an ordering where all
+ * computes the output position for each thread in an ordering where all
- * True threads come first, followed by all False threads. 
+ * True threads come first, followed by all False threads. 
- * This version handles 4 predicates per thread; hence, "rank4".
+ * This version handles 4 predicates per thread; hence, "rank4".
- *
+ *
- * @param[in] preds true/false values for each of the 4 elements in this thread
+ * @param[in] preds true/false values for each of the 4 elements in this thread
- *
+ *
- * @todo is the description of "preds" correct?
+ * @todo is the description of "preds" correct?
-**/
+**/
-template <int ctasize>
+template <int ctasize>
-__device__ uint4 rank4(uint4 preds)
+__device__ uint4 rank4(uint4 preds)
-{
+{
-    uint4 address = scan4(preds);  
+    uint4 address = scan4(preds);  
-
+
-    __shared__ uint numtrue;
+    __shared__ uint numtrue;
-    if (threadIdx.x == ctasize-1)
+    if (threadIdx.x == ctasize-1)
-    {
+    {
-        numtrue = address.w + preds.w;
+        numtrue = address.w + preds.w;
-    }
+    }
-    __syncthreads();
+    __syncthreads();
-
+
-    uint4 rank;
+    uint4 rank;
-    uint idx = threadIdx.x << 2;
+    uint idx = threadIdx.x << 2;
-    rank.x = (preds.x) ? address.x : numtrue + idx   - address.x;
+    rank.x = (preds.x) ? address.x : numtrue + idx   - address.x;
-    rank.y = (preds.y) ? address.y : numtrue + idx + 1 - address.y;
+    rank.y = (preds.y) ? address.y : numtrue + idx + 1 - address.y;
-    rank.z = (preds.z) ? address.z : numtrue + idx + 2 - address.z;
+    rank.z = (preds.z) ? address.z : numtrue + idx + 2 - address.z;
-    rank.w = (preds.w) ? address.w : numtrue + idx + 3 - address.w;     
+    rank.w = (preds.w) ? address.w : numtrue + idx + 3 - address.w;     
-                
+                
-    return rank;
+    return rank;
-}
+}
-
+
-/**
+/**
- * @brief Sorts one block
+ * @brief Sorts one block
- *
+ *
- * Uses rank to sort one bit at a time: Sorts a block according
+ * Uses rank to sort one bit at a time: Sorts a block according
- * to bits startbit -> nbits + startbit
+ * to bits startbit -> nbits + startbit
- * @param[in,out] key
+ * @param[in,out] key
- * @param[in,out] value
+ * @param[in,out] value
-**/
+**/
-template<uint nbits, uint startbit>
+template<uint nbits, uint startbit>
-__device__ void radixSortBlock(uint4 &key, uint4 &value)
+__device__ void radixSortBlock(uint4 &key, uint4 &value)
-{
+{
-    extern __shared__ uint sMem1[];
+    extern __shared__ uint sMem1[];
-    for(uint shift = startbit; shift < (startbit + nbits); ++shift)
+    for(uint shift = startbit; shift < (startbit + nbits); ++shift)
-    {        
+    {        
-        uint4 lsb;
+        uint4 lsb;
-        lsb.x = !((key.x >> shift) & 0x1);
+        lsb.x = !((key.x >> shift) & 0x1);
-        lsb.y = !((key.y >> shift) & 0x1);
+        lsb.y = !((key.y >> shift) & 0x1);
-        lsb.z = !((key.z >> shift) & 0x1);
+        lsb.z = !((key.z >> shift) & 0x1);
-        lsb.w = !((key.w >> shift) & 0x1); 
+        lsb.w = !((key.w >> shift) & 0x1); 
-
+
-        uint4 r = rank4<256>(lsb);
+        uint4 r = rank4<256>(lsb);
-
+
-#if 1
+#if 1
-        // This arithmetic strides the ranks across 4 SORT_CTA_SIZE regions
+        // This arithmetic strides the ranks across 4 SORT_CTA_SIZE regions
-        sMem1[(r.x & 3) * SORT_CTA_SIZE + (r.x >> 2)] = key.x;
+        sMem1[(r.x & 3) * SORT_CTA_SIZE + (r.x >> 2)] = key.x;
-        sMem1[(r.y & 3) * SORT_CTA_SIZE + (r.y >> 2)] = key.y;
+        sMem1[(r.y & 3) * SORT_CTA_SIZE + (r.y >> 2)] = key.y;
-        sMem1[(r.z & 3) * SORT_CTA_SIZE + (r.z >> 2)] = key.z;
+        sMem1[(r.z & 3) * SORT_CTA_SIZE + (r.z >> 2)] = key.z;
-        sMem1[(r.w & 3) * SORT_CTA_SIZE + (r.w >> 2)] = key.w; 
+        sMem1[(r.w & 3) * SORT_CTA_SIZE + (r.w >> 2)] = key.w; 
-        __syncthreads();
+        __syncthreads();
-
+
-        // The above allows us to read without 4-way bank conflicts:
+        // The above allows us to read without 4-way bank conflicts:
-        key.x = sMem1[threadIdx.x];
+        key.x = sMem1[threadIdx.x];
-        key.y = sMem1[threadIdx.x +     SORT_CTA_SIZE];
+        key.y = sMem1[threadIdx.x +     SORT_CTA_SIZE];
-        key.z = sMem1[threadIdx.x + 2 * SORT_CTA_SIZE];
+        key.z = sMem1[threadIdx.x + 2 * SORT_CTA_SIZE];
-        key.w = sMem1[threadIdx.x + 3 * SORT_CTA_SIZE];
+        key.w = sMem1[threadIdx.x + 3 * SORT_CTA_SIZE];
-
+
-        __syncthreads();
+        __syncthreads();
-
+
-        sMem1[(r.x & 3) * SORT_CTA_SIZE + (r.x >> 2)] = value.x;
+        sMem1[(r.x & 3) * SORT_CTA_SIZE + (r.x >> 2)] = value.x;
-        sMem1[(r.y & 3) * SORT_CTA_SIZE + (r.y >> 2)] = value.y;
+        sMem1[(r.y & 3) * SORT_CTA_SIZE + (r.y >> 2)] = value.y;
-        sMem1[(r.z & 3) * SORT_CTA_SIZE + (r.z >> 2)] = value.z;
+        sMem1[(r.z & 3) * SORT_CTA_SIZE + (r.z >> 2)] = value.z;
-        sMem1[(r.w & 3) * SORT_CTA_SIZE + (r.w >> 2)] = value.w;
+        sMem1[(r.w & 3) * SORT_CTA_SIZE + (r.w >> 2)] = value.w;
-        __syncthreads();
+        __syncthreads();
-
+
-        value.x = sMem1[threadIdx.x];
+        value.x = sMem1[threadIdx.x];
-        value.y = sMem1[threadIdx.x +     SORT_CTA_SIZE];
+        value.y = sMem1[threadIdx.x +     SORT_CTA_SIZE];
-        value.z = sMem1[threadIdx.x + 2 * SORT_CTA_SIZE];
+        value.z = sMem1[threadIdx.x + 2 * SORT_CTA_SIZE];
-        value.w = sMem1[threadIdx.x + 3 * SORT_CTA_SIZE];
+        value.w = sMem1[threadIdx.x + 3 * SORT_CTA_SIZE];
-#else
+#else
-        sMem1[r.x] = key.x;
+        sMem1[r.x] = key.x;
-        sMem1[r.y] = key.y;
+        sMem1[r.y] = key.y;
-        sMem1[r.z] = key.z;
+        sMem1[r.z] = key.z;
-        sMem1[r.w] = key.w;
+        sMem1[r.w] = key.w;
-        __syncthreads();
+        __syncthreads();
-
+
-        // This access has 4-way bank conflicts
+        // This access has 4-way bank conflicts
-        key = sMem[threadIdx.x];
+        key = sMem[threadIdx.x];
-
+
-        __syncthreads();
+        __syncthreads();
-
+
-        sMem1[r.x] = value.x;
+        sMem1[r.x] = value.x;
-        sMem1[r.y] = value.y;
+        sMem1[r.y] = value.y;
-        sMem1[r.z] = value.z;
+        sMem1[r.z] = value.z;
-        sMem1[r.w] = value.w;
+        sMem1[r.w] = value.w;
-        __syncthreads();
+        __syncthreads();
-
+
-        value = sMem[threadIdx.x];
+        value = sMem[threadIdx.x];
-#endif
+#endif
-
+
-        __syncthreads();
+        __syncthreads();
-    }
+    }
-}
+}
-
+
-/**
+/**
- * @brief Sorts one block. Key-only version.
+ * @brief Sorts one block. Key-only version.
- *
+ *
- * Uses rank to sort one bit at a time: Sorts a block according
+ * Uses rank to sort one bit at a time: Sorts a block according
- * to bits startbit -> nbits + startbit
+ * to bits startbit -> nbits + startbit
- * @param[in,out] key
+ * @param[in,out] key
-**/
+**/
-
+
-template<uint nbits, uint startbit>
+template<uint nbits, uint startbit>
-__device__ void radixSortBlockKeysOnly(uint4 &key)
+__device__ void radixSortBlockKeysOnly(uint4 &key)
-{
+{
-    extern __shared__ uint sMem1[];
+    extern __shared__ uint sMem1[];
-    for(uint shift = startbit; shift < (startbit + nbits); ++shift)
+    for(uint shift = startbit; shift < (startbit + nbits); ++shift)
-    {                   
+    {                   
-        uint4 lsb;
+        uint4 lsb;
-        lsb.x = !((key.x >> shift) & 0x1);
+        lsb.x = !((key.x >> shift) & 0x1);
-        lsb.y = !((key.y >> shift) & 0x1);
+        lsb.y = !((key.y >> shift) & 0x1);
-        lsb.z = !((key.z >> shift) & 0x1);
+        lsb.z = !((key.z >> shift) & 0x1);
-        lsb.w = !((key.w >> shift) & 0x1);
+        lsb.w = !((key.w >> shift) & 0x1);
-
+
-        uint4 r = rank4<256>(lsb);
+        uint4 r = rank4<256>(lsb);
-
+
-#if 1
+#if 1
-        // This arithmetic strides the ranks across 4 CTA_SIZE regions
+        // This arithmetic strides the ranks across 4 CTA_SIZE regions
-        sMem1[(r.x & 3) * SORT_CTA_SIZE + (r.x >> 2)] = key.x;
+        sMem1[(r.x & 3) * SORT_CTA_SIZE + (r.x >> 2)] = key.x;
-        sMem1[(r.y & 3) * SORT_CTA_SIZE + (r.y >> 2)] = key.y;
+        sMem1[(r.y & 3) * SORT_CTA_SIZE + (r.y >> 2)] = key.y;
-        sMem1[(r.z & 3) * SORT_CTA_SIZE + (r.z >> 2)] = key.z;
+        sMem1[(r.z & 3) * SORT_CTA_SIZE + (r.z >> 2)] = key.z;
-        sMem1[(r.w & 3) * SORT_CTA_SIZE + (r.w >> 2)] = key.w;
+        sMem1[(r.w & 3) * SORT_CTA_SIZE + (r.w >> 2)] = key.w;
-        __syncthreads();
+        __syncthreads();
-
+
-        // The above allows us to read without 4-way bank conflicts:
+        // The above allows us to read without 4-way bank conflicts:
-        key.x = sMem1[threadIdx.x];
+        key.x = sMem1[threadIdx.x];
-        key.y = sMem1[threadIdx.x +     SORT_CTA_SIZE];
+        key.y = sMem1[threadIdx.x +     SORT_CTA_SIZE];
-        key.z = sMem1[threadIdx.x + 2 * SORT_CTA_SIZE];
+        key.z = sMem1[threadIdx.x + 2 * SORT_CTA_SIZE];
-        key.w = sMem1[threadIdx.x + 3 * SORT_CTA_SIZE];
+        key.w = sMem1[threadIdx.x + 3 * SORT_CTA_SIZE];
-#else
+#else
-        sMem1[r.x] = key.x;
+        sMem1[r.x] = key.x;
-        sMem1[r.y] = key.y;
+        sMem1[r.y] = key.y;
-        sMem1[r.z] = key.z;
+        sMem1[r.z] = key.z;
-        sMem1[r.w] = key.w;
+        sMem1[r.w] = key.w;
-        __syncthreads();
+        __syncthreads();
-
+
-        // This access has 4-way bank conflicts
+        // This access has 4-way bank conflicts
-        key = sMem[threadIdx.x];
+        key = sMem[threadIdx.x];
-#endif
+#endif
-
+
-        __syncthreads();
+        __syncthreads();
-    }
+    }
-}
+}
-
+
-/** @} */ // end radix sort functions
+/** @} */ // end radix sort functions
-/** @} */ // end cudpp_cta
+/** @} */ // end cudpp_cta

lib/gpu/cudpp_mini/cudpp.cpp

@@ -7,7 +7,7 @@
 // This source code is distributed under the terms of license.txt in
 // the root directory of this source distribution.
 // ------------------------------------------------------------- 
-
+ 
 /**
  * @file
  * cudpp.cpp

lib/gpu/cudpp_mini/cudpp.h

@@ -7,7 +7,7 @@
 // This source code is distributed under the terms of license.txt in
 // the root directory of this source distribution.
 // ------------------------------------------------------------- 
-
+ 
 /**
  * @file
  * cudpp.h

lib/gpu/cudpp_mini/cudpp_globals.h

@@ -7,7 +7,7 @@
 // This source code is distributed under the terms of license.txt in
 // the root directory of this source distribution.
 // ------------------------------------------------------------- 
-
+ 
 /**
  * @file
  * cudpp_globals.h

lib/gpu/cudpp_mini/cudpp_maximal_launch.cpp

@@ -8,7 +8,7 @@
 // in the root directory of this source distribution.
 // ------------------------------------------------------------- 
 #include "cudpp_maximal_launch.h"
-
+ 
 inline size_t min(size_t x, size_t y)
 {
     return (x <= y) ? x : y;

lib/gpu/cudpp_mini/cudpp_maximal_launch.h

@@ -9,7 +9,7 @@
 // ------------------------------------------------------------- 
 #ifndef _MAXIMAL_LAUNCH_H_
 #define _MAXIMAL_LAUNCH_H_
-
+ 
 #include "cuda_runtime.h"
 
 extern "C"

lib/gpu/cudpp_mini/cudpp_plan.cpp

@@ -7,7 +7,7 @@
 // This source code is distributed under the terms of license.txt
 // in the root directory of this source distribution.
 // ------------------------------------------------------------- 
-
+ 
 #include "cudpp.h"
 #include "cudpp_plan_manager.h"
 #include "cudpp_scan.h"

lib/gpu/cudpp_mini/cudpp_plan.h

@@ -9,7 +9,7 @@
 // ------------------------------------------------------------- 
 #ifndef __CUDPP_PLAN_H__
 #define __CUDPP_PLAN_H__
-
+ 
 typedef void* KernelPointer;
 
 extern "C" size_t getNumCTAs(KernelPointer kernel);

lib/gpu/cudpp_mini/cudpp_plan_manager.cpp

@@ -11,7 +11,7 @@
 #include "cudpp_plan.h"
 #include "cudpp_plan_manager.h"
 #include "cudpp_maximal_launch.h"
-
+ 
 typedef void* KernelPointer;
 
 extern "C" size_t getNumCTAs(KernelPointer kernel)

lib/gpu/cudpp_mini/cudpp_plan_manager.h

@@ -9,7 +9,7 @@
 // ------------------------------------------------------------- 
 #ifndef __CUDPP_PLAN_MANAGER_H__
 #define __CUDPP_PLAN_MANAGER_H__
-
+ 
 #include <map>
 
 class CUDPPPlan;

lib/gpu/cudpp_mini/cudpp_radixsort.h

@@ -9,7 +9,7 @@
 // ------------------------------------------------------------- 
 #ifndef   __RADIXSORT_H__
 #define   __RADIXSORT_H__
-
+ 
 #define SORT_CTA_SIZE 256 //This CTA_SIZE must equal 16 * number of radices 
 
 #include "cudpp_globals.h"

lib/gpu/cudpp_mini/cudpp_scan.h

@@ -7,7 +7,7 @@
 // This source code is distributed under the terms of license.txt 
 // in the root directory of this source distribution.
 // ------------------------------------------------------------- 
-
+ 
 /**
 * @file
 * cudpp_scan.h

lib/gpu/cudpp_mini/cudpp_util.h

@@ -7,7 +7,7 @@
 // This source code is distributed under the terms of license.txt in
 // the root directory of this source distribution.
 // ------------------------------------------------------------- 
-
+ 
 /**
  * @file
  * cudpp_util.h

lib/gpu/cudpp_mini/cutil.h

@@ -26,7 +26,7 @@
 * 227.7202-4 (JUNE 1995), all U.S. Government End Users acquire the 
 * source code with only those rights set forth herein.
 */
-
+ 
 
 /* CUda UTility Library */
 

lib/gpu/cudpp_mini/license.txt

@@ -23,3 +23,4 @@ DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF
 LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE 
 OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF 
 ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
+ 

lib/gpu/cudpp_mini/sharedmem.h

@@ -7,7 +7,7 @@
 // This source code is distributed under the terms of license.txt 
 // in the root directory of this source distribution.
 // ------------------------------------------------------------- 
-
+ 
 /**
  * @file
  * sharedmem.h

lib/gpu/gb_gpu.cpp

@@ -10,7 +10,7 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
 ------------------------------------------------------------------------- */

lib/gpu/gb_gpu_memory.cpp

@@ -10,7 +10,7 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
 ------------------------------------------------------------------------- */

lib/gpu/geryon/VERSION.txt

@@ -1 +1,2 @@
 Geryon Version 10.280
+ 

lib/gpu/geryon/file_to_cstr.sh

@@ -6,7 +6,7 @@
 # requires: sed, rm, mv
 #
 # Author: Axel Kohlmeyer, Temple University
-
+ 
 num_args=$#
 
 # we write to a scratch file, since

lib/gpu/geryon/nvc_device.h

@@ -13,7 +13,7 @@
     copyright            : (C) 2009 by W. Michael Brown
     email                : brownw@ornl.gov
  ***************************************************************************/
-
+ 
 /* -----------------------------------------------------------------------
    Copyright (2009) Sandia Corporation.  Under the terms of Contract
    DE-AC04-94AL85000 with Sandia Corporation, the U.S. Government retains

lib/gpu/geryon/nvc_macros.h

@@ -6,7 +6,7 @@
 #undef _GLIBCXX_ATOMIC_BUILTINS
 #endif // _GLIBCXX_ATOMIC_BUILTINS
 #endif // __APPLE__
-
+ 
 #include <stdio.h>
 #include <cassert>
 #include <cuda_runtime.h>

lib/gpu/geryon/nvc_texture.h

@@ -20,7 +20,7 @@
    certain rights in this software.  This software is distributed under 
    the Simplified BSD License.
    ----------------------------------------------------------------------- */
-
+ 
 #ifndef NVC_TEXTURE
 #define NVC_TEXTURE
 

lib/gpu/geryon/nvd_device.h

@@ -20,7 +20,7 @@
    certain rights in this software.  This software is distributed under 
    the Simplified BSD License.
    ----------------------------------------------------------------------- */
-
+ 
 #ifndef NVD_DEVICE
 #define NVD_DEVICE
 

lib/gpu/geryon/nvd_kernel.h

@@ -20,7 +20,7 @@
    certain rights in this software.  This software is distributed under 
    the Simplified BSD License.
    ----------------------------------------------------------------------- */
-
+ 
 #ifndef NVD_KERNEL
 #define NVD_KERNEL
 

lib/gpu/geryon/nvd_macros.h

@@ -1,6 +1,6 @@
 #ifndef NVD_MACROS_H
 #define NVD_MACROS_H
-
+ 
 #include <stdio.h>
 #include <cassert>
 #include <cuda.h>

lib/gpu/geryon/nvd_mat.h

@@ -20,7 +20,7 @@
    certain rights in this software.  This software is distributed under 
    the Simplified BSD License.
    ----------------------------------------------------------------------- */
-
+ 
 /*! \file */
    
 #ifndef NVD_MAT_H

lib/gpu/geryon/nvd_memory.h

@@ -20,7 +20,7 @@
    certain rights in this software.  This software is distributed under 
    the Simplified BSD License.
    ----------------------------------------------------------------------- */
-
+ 
 #ifndef NVD_MEMORY_H
 #define NVD_MEMORY_H
 

lib/gpu/geryon/nvd_texture.h

@@ -20,7 +20,7 @@
    certain rights in this software.  This software is distributed under 
    the Simplified BSD License.
    ----------------------------------------------------------------------- */
-
+ 
 #ifndef NVD_TEXTURE
 #define NVD_TEXTURE
 

lib/gpu/geryon/nvd_timer.h

@@ -20,7 +20,7 @@
    certain rights in this software.  This software is distributed under 
    the Simplified BSD License.
    ----------------------------------------------------------------------- */
-
+ 
 #ifndef NVD_TIMER_H
 #define NVD_TIMER_H
 

lib/gpu/geryon/ocl_device.h

@@ -20,7 +20,7 @@
    certain rights in this software.  This software is distributed under 
    the Simplified BSD License.
    ----------------------------------------------------------------------- */
-
+ 
 #ifndef OCL_DEVICE
 #define OCL_DEVICE
 

lib/gpu/geryon/ocl_kernel.h

@@ -13,7 +13,7 @@
     copyright            : (C) 2010 by W. Michael Brown
     email                : brownw@ornl.gov
  ***************************************************************************/
-
+ 
 /* -----------------------------------------------------------------------
    Copyright (2010) Sandia Corporation.  Under the terms of Contract
    DE-AC04-94AL85000 with Sandia Corporation, the U.S. Government retains

lib/gpu/geryon/ocl_mat.h

@@ -20,7 +20,7 @@
    certain rights in this software.  This software is distributed under 
    the Simplified BSD License.
    ----------------------------------------------------------------------- */
-
+ 
 /*! \file */
    
 #ifndef OCL_MAT_H

lib/gpu/geryon/ocl_texture.h

@@ -20,7 +20,7 @@
    certain rights in this software.  This software is distributed under 
    the Simplified BSD License.
    ----------------------------------------------------------------------- */
-
+ 
 #ifndef OCL_TEXTURE
 #define OCL_TEXTURE
 

lib/gpu/geryon/ocl_timer.h

@@ -20,7 +20,7 @@
    certain rights in this software.  This software is distributed under 
    the Simplified BSD License.
    ----------------------------------------------------------------------- */
-
+ 
 #ifndef OCL_TIMER_H
 #define OCL_TIMER_H
 

lib/gpu/geryon/ucl_arg_kludge.h

@@ -13,7 +13,7 @@
     copyright            : (C) 2010 by W. Michael Brown
     email                : brownw@ornl.gov
  ***************************************************************************/
-
+ 
 /* -----------------------------------------------------------------------
    Copyright (2010) Sandia Corporation.  Under the terms of Contract
    DE-AC04-94AL85000 with Sandia Corporation, the U.S. Government retains

lib/gpu/geryon/ucl_basemat.h

@@ -13,7 +13,7 @@
     copyright            : (C) 2009 by W. Michael Brown
     email                : brownw@ornl.gov
  ***************************************************************************/
-
+ 
 /* -----------------------------------------------------------------------
    Copyright (2009) Sandia Corporation.  Under the terms of Contract
    DE-AC04-94AL85000 with Sandia Corporation, the U.S. Government retains

lib/gpu/geryon/ucl_copy.h

@@ -13,7 +13,7 @@
     copyright            : (C) 2010 by W. Michael Brown
     email                : brownw@ornl.gov
  ***************************************************************************/
-
+ 
 /* -----------------------------------------------------------------------
    Copyright (2010) Sandia Corporation.  Under the terms of Contract
    DE-AC04-94AL85000 with Sandia Corporation, the U.S. Government retains

lib/gpu/geryon/ucl_d_mat.h

@@ -13,7 +13,7 @@
     copyright            : (C) 2009 by W. Michael Brown
     email                : brownw@ornl.gov
  ***************************************************************************/
-
+ 
 /* -----------------------------------------------------------------------
    Copyright (2009) Sandia Corporation.  Under the terms of Contract
    DE-AC04-94AL85000 with Sandia Corporation, the U.S. Government retains

lib/gpu/geryon/ucl_d_vec.h

@@ -13,7 +13,7 @@
     copyright            : (C) 2009 by W. Michael Brown
     email                : brownw@ornl.gov
  ***************************************************************************/
-
+ 
 /* -----------------------------------------------------------------------
    Copyright (2009) Sandia Corporation.  Under the terms of Contract
    DE-AC04-94AL85000 with Sandia Corporation, the U.S. Government retains

lib/gpu/geryon/ucl_get_devices.cpp

@@ -13,7 +13,7 @@
     copyright            : (C) 2009 by W. Michael Brown
     email                : brownw@ornl.gov
  ***************************************************************************/
-
+ 
 /* -----------------------------------------------------------------------
    Copyright (2009) Sandia Corporation.  Under the terms of Contract
    DE-AC04-94AL85000 with Sandia Corporation, the U.S. Government retains

lib/gpu/geryon/ucl_h_mat.h

@@ -13,7 +13,7 @@
     copyright            : (C) 2009 by W. Michael Brown
     email                : brownw@ornl.gov
  ***************************************************************************/
-
+ 
 /* -----------------------------------------------------------------------
    Copyright (2009) Sandia Corporation.  Under the terms of Contract
    DE-AC04-94AL85000 with Sandia Corporation, the U.S. Government retains

lib/gpu/geryon/ucl_h_vec.h

@@ -13,7 +13,7 @@
     copyright            : (C) 2009 by W. Michael Brown
     email                : brownw@ornl.gov
  ***************************************************************************/
-
+ 
 /* -----------------------------------------------------------------------
    Copyright (2009) Sandia Corporation.  Under the terms of Contract
    DE-AC04-94AL85000 with Sandia Corporation, the U.S. Government retains

lib/gpu/geryon/ucl_nv_kernel.h

@@ -13,7 +13,7 @@
     copyright            : (C) 2010 by W. Michael Brown
     email                : brownw@ornl.gov
  ***************************************************************************/
-
+ 
 /* -----------------------------------------------------------------------
    Copyright (2010) Sandia Corporation.  Under the terms of Contract
    DE-AC04-94AL85000 with Sandia Corporation, the U.S. Government retains

lib/gpu/geryon/ucl_print.h

@@ -13,7 +13,7 @@
     copyright            : (C) 2010 by W. Michael Brown
     email                : brownw@ornl.gov
  ***************************************************************************/
-
+ 
 /* -----------------------------------------------------------------------
    Copyright (2010) Sandia Corporation.  Under the terms of Contract
    DE-AC04-94AL85000 with Sandia Corporation, the U.S. Government retains

lib/gpu/geryon/ucl_types.h

@@ -13,7 +13,7 @@
     copyright            : (C) 2010 by W. Michael Brown
     email                : brownw@ornl.gov
  ***************************************************************************/
-
+ 
 /* -----------------------------------------------------------------------
    Copyright (2010) Sandia Corporation.  Under the terms of Contract
    DE-AC04-94AL85000 with Sandia Corporation, the U.S. Government retains

lib/gpu/lj96_cut_gpu.cpp

@@ -10,7 +10,7 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
 ------------------------------------------------------------------------- */

lib/gpu/lj96_cut_gpu_memory.cpp

@@ -10,7 +10,7 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
 ------------------------------------------------------------------------- */

lib/gpu/lj96_cut_gpu_memory.h

@@ -10,7 +10,7 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
 ------------------------------------------------------------------------- */

lib/gpu/lj_cut_gpu.cpp

@@ -10,7 +10,7 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
 ------------------------------------------------------------------------- */

lib/gpu/lj_cut_gpu_memory.cpp

@@ -10,7 +10,7 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
 ------------------------------------------------------------------------- */

lib/gpu/lj_cut_gpu_memory.h

@@ -10,7 +10,7 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
 ------------------------------------------------------------------------- */

lib/gpu/ljc_cut_gpu.cpp

@@ -10,7 +10,7 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
 ------------------------------------------------------------------------- */

lib/gpu/ljc_cut_gpu_memory.cpp

@@ -10,7 +10,7 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
 ------------------------------------------------------------------------- */

lib/gpu/ljc_cut_gpu_memory.h

@@ -10,7 +10,7 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
 ------------------------------------------------------------------------- */

lib/gpu/ljcl_cut_gpu.cpp

@@ -10,7 +10,7 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
 ------------------------------------------------------------------------- */

lib/gpu/ljcl_cut_gpu_memory.cpp

@@ -10,7 +10,7 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
 ------------------------------------------------------------------------- */

lib/gpu/ljcl_cut_gpu_memory.h

@@ -10,7 +10,7 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
 ------------------------------------------------------------------------- */

lib/gpu/pair_gpu_atom.cpp

@@ -10,7 +10,7 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
 ------------------------------------------------------------------------- */

lib/gpu/pair_gpu_balance.h

@@ -10,7 +10,7 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
 ------------------------------------------------------------------------- */

lib/gpu/pair_gpu_device.cpp

@@ -10,7 +10,7 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
 ------------------------------------------------------------------------- */

lib/gpu/pair_gpu_device.h

@@ -10,7 +10,7 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
 ------------------------------------------------------------------------- */

lib/gpu/pair_gpu_nbor.cpp

@@ -10,7 +10,7 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
                          Peng Wang (Nvidia), penwang@nvidia.com

lib/gpu/pair_gpu_precision.h

@@ -10,7 +10,7 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
 ------------------------------------------------------------------------- */

lib/gpu/pair_win_sort.cpp

@@ -10,7 +10,7 @@
 
    See the README file in the top-level LAMMPS directory.
 ------------------------------------------------------------------------- */
-
+ 
 /* ----------------------------------------------------------------------
    Contributing authors: Mike Brown (ORNL), brownw@ornl.gov
 ------------------------------------------------------------------------- */