bsc 3.3.4

CHANGES

@@ -1,3 +1,6 @@
+Changes in 3.3.4 (January, 24 2024)
+- Implemented GPU acceleration of reverse Burrows�Wheeler transform.
+
 Changes in 3.3.3 (November, 26 2023)
 - Fixed out-of-bound memory access issue for large inputs.
 - Slightly improved compression performance.

README

@@ -10,7 +10,7 @@ block-sorting data compression algorithms.
 libbsc is a library based on bsc, it uses the same algorithms
 as bsc and enables you to compress memory blocks.
 
-Copyright (c) 2009-2023 Ilya Grebnov <[email protected]>
+Copyright (c) 2009-2024 Ilya Grebnov <[email protected]>
 
 See file AUTHORS for a full list of contributors.
 
@@ -21,7 +21,7 @@ See the bsc and libbsc web site:
 Software License:
 -----------------
 
-Copyright (c) 2009-2023 Ilya Grebnov <[email protected]>
+Copyright (c) 2009-2024 Ilya Grebnov <[email protected]>
 
 Licensed under the Apache License, Version 2.0 (the "License");
 you may not use this file except in compliance with the License.
@@ -54,7 +54,8 @@ Compression and decompression requirements are the same and in bytes, can
 be estimated as 16Mb + 5 x block size x number of blocks processed in parallel.
 
 GPU memory usage for NVIDIA CUDA technology is different from CPU memory usage
-and can be estimated as 20 x block size for ST and 21 x block size for BWT.
+and can be estimated as 20 x block size for ST, 21 x block size for forward BWT
+and 7 x block size for inverse BWT.
 
 
 NVIDIA GPU acceleration:

VERSION

@@ -1 +1 @@
-3.3.3
+3.3.4

bsc.cpp

@@ -8,7 +8,7 @@
 This file is a part of bsc and/or libbsc, a program and a library for
 lossless, block-sorting data compression.
 
-   Copyright (c) 2009-2021 Ilya Grebnov <[email protected]>
+   Copyright (c) 2009-2024 Ilya Grebnov <[email protected]>
 
    Licensed under the Apache License, Version 2.0 (the "License");
    you may not use this file except in compliance with the License.
@@ -869,8 +869,8 @@ void ProcessCommandline(int argc, char * argv[])
 
 int main(int argc, char * argv[])
 {
-    fprintf(stdout, "This is bsc, Block Sorting Compressor. Version 3.3.3. 26 November 2023.\n");
-    fprintf(stdout, "Copyright (c) 2009-2023 Ilya Grebnov <[email protected]>.\n\n");
+    fprintf(stdout, "This is bsc, Block Sorting Compressor. Version 3.3.4. 24 January 2024.\n");
+    fprintf(stdout, "Copyright (c) 2009-2024 Ilya Grebnov <[email protected]>.\n\n");
 
 #if defined(_OPENMP) && defined(__INTEL_COMPILER)
 

libbsc/adler32/adler32.cpp

@@ -8,7 +8,7 @@
 This file is a part of bsc and/or libbsc, a program and a library for
 lossless, block-sorting data compression.
 
-   Copyright (c) 2009-2021 Ilya Grebnov <[email protected]>
+   Copyright (c) 2009-2024 Ilya Grebnov <[email protected]>
 
    Licensed under the Apache License, Version 2.0 (the "License");
    you may not use this file except in compliance with the License.

libbsc/adler32/adler32.h

@@ -8,7 +8,7 @@
 This file is a part of bsc and/or libbsc, a program and a library for
 lossless, block-sorting data compression.
 
-   Copyright (c) 2009-2021 Ilya Grebnov <[email protected]>
+   Copyright (c) 2009-2024 Ilya Grebnov <[email protected]>
 
    Licensed under the Apache License, Version 2.0 (the "License");
    you may not use this file except in compliance with the License.

libbsc/bwt/bwt.cpp

@@ -8,7 +8,7 @@
 This file is a part of bsc and/or libbsc, a program and a library for
 lossless, block-sorting data compression.
 
-   Copyright (c) 2009-2021 Ilya Grebnov <[email protected]>
+   Copyright (c) 2009-2024 Ilya Grebnov <[email protected]>
 
    Licensed under the Apache License, Version 2.0 (the "License");
    you may not use this file except in compliance with the License.
@@ -226,13 +226,63 @@ int bsc_bwt_encode(unsigned char * T, int n, unsigned char * num_indexes, int *
     return LIBBSC_NOT_ENOUGH_MEMORY;
 }
 
+int bsc_bwt_gpu_decode(unsigned char * T, int n, int index, int features)
+{
+    int result = -1;
+
+#ifdef LIBBSC_CUDA_SUPPORT
+    if (features & LIBBSC_FEATURE_CUDA)
+    {
+        int storage_approx_length = (n / 3) | 0x1fffff;
+
+#ifdef LIBBSC_OPENMP
+        omp_set_lock(&bwt_cuda_lock);
+
+        if (bwt_cuda_device_storage_size < storage_approx_length)
+        {
+            if (bwt_cuda_device_storage != NULL)
+            {
+                libcubwt_free_device_storage(bwt_cuda_device_storage);
+
+                bwt_cuda_device_storage = NULL;
+                bwt_cuda_device_storage_size = 0;
+            }
+
+            if (libcubwt_allocate_device_storage(&bwt_cuda_device_storage, storage_approx_length + (storage_approx_length / 32)) == LIBCUBWT_NO_ERROR)
+            {
+                bwt_cuda_device_storage_size = storage_approx_length + (storage_approx_length / 32);
+            }
+        }
+
+        if (bwt_cuda_device_storage_size >= storage_approx_length)
+        {
+            result = (int)libcubwt_unbwt(bwt_cuda_device_storage, T, T, n, NULL, index);
+        } 
+
+        omp_unset_lock(&bwt_cuda_lock);
+#else
+        void * bwt_cuda_device_storage = NULL;
+
+        if (libcubwt_allocate_device_storage(&bwt_cuda_device_storage, storage_approx_length) == LIBCUBWT_NO_ERROR)
+        {
+            result = (int)libcubwt_unbwt(bwt_cuda_device_storage, T, T, n, NULL, index);
+
+            libcubwt_free_device_storage(bwt_cuda_device_storage);
+        }
+#endif
+    }
+#endif
+
+    return result;
+}
+
 int bsc_bwt_decode(unsigned char * T, int n, int index, unsigned char num_indexes, int * indexes, int features)
 {
     if ((T == NULL) || (n < 0) || (index <= 0) || (index > n))
     {
         return LIBBSC_BAD_PARAMETER;
     }
-    if (n <= 1)
+    if (n <= 1 || bsc_bwt_gpu_decode(T, n, index, features) == 0)
     {
         return LIBBSC_NO_ERROR;
     }

libbsc/bwt/bwt.h

@@ -8,7 +8,7 @@
 This file is a part of bsc and/or libbsc, a program and a library for
 lossless, block-sorting data compression.
 
-   Copyright (c) 2009-2021 Ilya Grebnov <[email protected]>
+   Copyright (c) 2009-2024 Ilya Grebnov <[email protected]>
 
    Licensed under the Apache License, Version 2.0 (the "License");
    you may not use this file except in compliance with the License.

libbsc/bwt/libcubwt/AUTHORS

@@ -7,4 +7,5 @@
   Leyuan Wang, Sean Baxter, John D. Owens, Yury Shukhrov,
   Rory Mitchell, Jacopo Pantaleoni, Duane Merrill,
   Georgy Evtushenko, Allison Vacanti, Robert Crovella,
-  Mark Harris.
+  Mark Harris, Vitaly Osipov, Andy Adinets, Elias Stehle,
+  Michael Maniscalco.

libbsc/bwt/libcubwt/CHANGES

@@ -1,3 +1,6 @@
+Changes in 1.6.0 (January 24, 2024)
+- Inverse Burrows-Wheeler transform.
+
 Changes in 1.5.0 (March 24, 2023)
 - Reduced memory usage and improved performance.
 

libbsc/bwt/libcubwt/VERSION

@@ -1 +1 @@
-1.5.0
+1.6.0