v1.1.0 update (#7)

* updates for v1.1.0

* minor revision

* Makefile revision

Author: cheng-zhao

CHANGELOG.md

@@ -1,7 +1,15 @@
 # FCFC Change Log
 
+## [1.1.0] – 2025-06-11
+
+-   Fixed an early-return bug affecting tiny catalogs.
+-   Corrected separation bins of the saved projected correlation data ([PR #4](pull/4)).
+-   Revised the normalization factor for weighted pair counts ([PR #3](pull/3))
+-   Added a `Makefile` for benchmark code compilation.
+-   Fixed a macro typo reported in [Issue #2](issues/2).
+-   Updated the FCFC paper reference in the README.
+
 ## [1.0.1] – 2022-06-26
 
 -   Refactorized the implementation of MPI and OpenMP parallelisms.
 -   Optimized MPI communications with custom structs, and fixed the deadlock due to multiple broadcasts.
-

Makefile

@@ -73,4 +73,4 @@ FCFC_2PT_BOX: src/fcfc/2pt_box
 	$(C_CMPLR) $(CFLAGS) -o $@ $(SRCS) $(wildcard $</*.c) $(LIBS) $(INCL) -I$<
 
 clean:
-	rm $(TARGETS)
+	rm -f $(TARGETS)

README.md

@@ -32,7 +32,7 @@ This program is compliant with the ISO C99 and IEEE POSIX.1-2008 standards, and
 
 FCFC is written by Cheng Zhao (赵成), and is distributed under the [MIT license](LICENSE.txt). If you use this program in research work that results in publications, please cite the following paper:
 
-> Zhao et al. 2020, [arXiv:2007.08997](https://ui.adsabs.harvard.edu/abs/2020arXiv200708997Z/abstract)
+> Zhao, C., 2023, [“Fast correlation function calculator: A high-performance pair-counting toolkit”](https://doi.org/10.1051/0004-6361/202346015), *Astronomy and Astrophysics*, 672, A83. \[[arXiv:2301.12557](https://arxiv.org/abs/2301.12557)\] \[[ADS Abstract](https://ui.adsabs.harvard.edu/abs/2023A%26A...672A..83Z)\]
 
 <sub>[\[TOC\]](#table-of-contents)</sub>
 
@@ -81,12 +81,12 @@ The list of available command line options can be consulted using the `-h` or `-
 
 An introduction of the components and the corresponding configuration parameters are listed below:
 
-| Component    | Description                                                     | Configuration parameters               |
-|:------------:|-----------------------------------------------------------------|:--------------------------------------:|
-| FCFC_2PT     | Compute 2PCF for survey-like data                               | [FCFC_2PT.md](doc/FCFC_2PT.md)         |
-| FCFC_2PT_BOX | Compute 2PCF for periodic simulation boxes<sup>[*](#tab1)</sup> | [FCFC_2PT_BOX.md](doc/FCFC_2PT_BOX.md) |
+| Component      | Description                                                      | Configuration parameters                 |
+|:--------------:|------------------------------------------------------------------|:----------------------------------------:|
+| FCFC\_2PT      | Compute 2PCF for survey-like data                                | [FCFC\_2PT.md](doc/FCFC_2PT.md)          |
+| FCFC\_2PT\_BOX | Compute 2PCF for periodic simulation boxes<sup>[\*](#tab1)</sup> | [FCFC\_2PT\_BOX.md](doc/FCFC_2PT_BOX.md) |
 
-<span id="tab1">*: treat the 3<sup>rd</sup> dimension (*z*-direction) as the line of sight</span>
+<span id="tab1">\*: treat the 3<sup>rd</sup> dimension (*z*-direction) as the line of sight</span>
 
 <sub>[\[TOC\]](#table-of-contents)</sub>
 
@@ -98,4 +98,3 @@ This program benefits from the following open-source projects:
 -   [https://github.com/swenson/sort](https://github.com/swenson/sort)
 
 <sub>[\[TOC\]](#table-of-contents)</sub>
-

benchmark/Makefile

@@ -0,0 +1,73 @@
+CC = gcc
+CFLAGS = -std=c99 -D_POSIX_C_SOURCE=200809 -Wall -O3 -march=native
+LIBS = -lm
+INCL = -Icommon
+SRCS = $(wildcard common/*.c)
+CFLAGS += -DBENCHMARK_TIMING
+
+TARGETS = BENCHMARK_struct_serial BENCHMARK_struct_avx BENCHMARK_struct_single_serial BENCHMARK_struct_single_avx BENCHMARK_hist_serial BENCHMARK_hist_avx BENCHMARK_hist_avx2 BENCHMARK_hist_single_serial BENCHMARK_hist_single_avx BENCHMARK_hist_single_avx2
+
+# Benchmarks with AVX512
+#TARGETS += BENCHMARK_struct_avx512 BENCHMARK_struct_single_avx512 BENCHMARK_hist_avx512 BENCHMARK_hist_single_avx512
+
+# Benchmarks for OpenMP strategy
+#TARGETS += BENCHMARK_struct_omp_dynamic BENCHMARK_struct_omp_static
+
+
+.PHONY: all
+all: $(TARGETS)
+
+BENCHMARK_struct_serial: struct
+	$(CC) $(CFLAGS) -DBENCHMARK_TREE_PREALLOC -o $@ $(SRCS) $(wildcard $</*.c) $(LIBS) $(INCL) -I$<
+
+BENCHMARK_struct_avx: struct
+	$(CC) $(CFLAGS) -DBENCHMARK_TREE_PREALLOC -DWITH_AVX -o $@ $(SRCS) $(wildcard $</*.c) $(LIBS) $(INCL) -I$<
+
+BENCHMARK_struct_avx512: struct
+	$(CC) $(CFLAGS) -DBENCHMARK_TREE_PREALLOC -DWITH_AVX512 -o $@ $(SRCS) $(wildcard $</*.c) $(LIBS) $(INCL) -I$<
+
+BENCHMARK_struct_single_serial: struct
+	$(CC) $(CFLAGS) -DBENCHMARK_TREE_PREALLOC -DSINGLE_PREC -o $@ $(SRCS) $(wildcard $</*.c) $(LIBS) $(INCL) -I$<
+
+BENCHMARK_struct_single_avx: struct
+	$(CC) $(CFLAGS) -DBENCHMARK_TREE_PREALLOC -DSINGLE_PREC -DWITH_AVX -o $@ $(SRCS) $(wildcard $</*.c) $(LIBS) $(INCL) -I$<
+
+BENCHMARK_struct_single_avx512: struct
+	$(CC) $(CFLAGS) -DBENCHMARK_TREE_PREALLOC -DSINGLE_PREC -DWITH_AVX512 -o $@ $(SRCS) $(wildcard $</*.c) $(LIBS) $(INCL) -I$<
+
+
+BENCHMARK_hist_serial: histogram
+	$(CC) $(CFLAGS) -o $@ $(SRCS) $(wildcard $</*.c) $(LIBS) $(INCL) -I$<
+
+BENCHMARK_hist_avx: histogram
+	$(CC) $(CFLAGS) -DWITH_AVX -o $@ $(SRCS) $(wildcard $</*.c) $(LIBS) $(INCL) -I$<
+
+BENCHMARK_hist_avx2: histogram
+	$(CC) $(CFLAGS) -DWITH_AVX2 -o $@ $(SRCS) $(wildcard $</*.c) $(LIBS) $(INCL) -I$<
+
+BENCHMARK_hist_avx512: histogram
+	$(CC) $(CFLAGS) -DWITH_AVX512 -o $@ $(SRCS) $(wildcard $</*.c) $(LIBS) $(INCL) -I$<
+
+BENCHMARK_hist_single_serial: histogram
+	$(CC) $(CFLAGS) -DSINGLE_PREC -o $@ $(SRCS) $(wildcard $</*.c) $(LIBS) $(INCL) -I$<
+
+BENCHMARK_hist_single_avx: histogram
+	$(CC) $(CFLAGS) -DSINGLE_PREC -DWITH_AVX -o $@ $(SRCS) $(wildcard $</*.c) $(LIBS) $(INCL) -I$<
+
+BENCHMARK_hist_single_avx2: histogram
+	$(CC) $(CFLAGS) -DSINGLE_PREC -DWITH_AVX2 -o $@ $(SRCS) $(wildcard $</*.c) $(LIBS) $(INCL) -I$<
+
+BENCHMARK_hist_single_avx512: histogram
+	$(CC) $(CFLAGS) -DSINGLE_PREC -DWITH_AVX512 -o $@ $(SRCS) $(wildcard $</*.c) $(LIBS) $(INCL) -I$<
+
+
+BENCHMARK_struct_omp_dynamic: struct
+	$(CC) $(CFLAGS) -DBENCHMARK_TIMING -DBENCHMARK_OMP_DYNAMIC -fopenmp -o $@ $(SRCS) $(wildcard $</*.c) $(LIBS) $(INCL) -I$<
+
+BENCHMARK_struct_omp_static: struct
+	$(CC) $(CFLAGS) -DBENCHMARK_TIMING -DBENCHMARK_OMP_STATIC -fopenmp -o $@ $(SRCS) $(wildcard $</*.c) $(LIBS) $(INCL) -I$<
+
+
+.PHONY: clean
+clean:
+	rm -f $(TARGETS)

benchmark/common/define_comm.h

@@ -49,7 +49,7 @@
   typedef float real;
   #define REAL_TOL      1e-6    /* tolerance for float number comparison */
   #ifdef FLT_EPSILON
-    #define REAL_EPS    REAL_EPS
+    #define REAL_EPS    FLT_EPSILON
   #else
     #define REAL_EPS    1e-7
   #endif

benchmark/histogram/benchmark.c

@@ -226,7 +226,16 @@ int main(int argc, char *argv[]) {
     printf(FMT_KEY(METHOD) ": FastLog2 (order %d)\n\n",
         method - BENCHMARK_HIST_FASTLOG2_START);
   }
-  else printf(FMT_KEY(METHOD) ": %s\n\n",
+  else printf(FMT_KEY(METHOD)
+#if BENCHMARK_SIMD  ==  BENCHMARK_SIMD_NONE
+      ": %s (no SIMD)\n\n",
+#elif BENCHMARK_SIMD  ==  BENCHMARK_SIMD_AVX
+      ": %s (AVX)\n\n",
+#elif BENCHMARK_SIMD  ==  BENCHMARK_SIMD_AVX2
+      ": %s (AVX2)\n\n",
+#elif BENCHMARK_SIMD  ==  BENCHMARK_SIMD_AVX512
+      ": %s (AVX512)\n\n",
+#endif
       method >= 0 ? benchmark_method_name[method] : "hybrid lookup table");
 
   /* Evaluate the distance histogram. */

doc/logo/plot_fcfc_logo.py

@@ -159,7 +159,7 @@ def export(self, filename, dpi=300):
     import matplotlib.patches as mpatches
 
     # Create the figure.
-    fig = plt.figure(figsize=(float(self.width)/dpi,float(self.height)/dpi), \
+    plt.figure(figsize=(float(self.width)/dpi,float(self.height)/dpi), \
           dpi=dpi, frameon=False)
     ax = plt.axes([0,0,1,1])
     ax.set_axis_off()

options.mk

@@ -3,7 +3,11 @@
 # * MPICC is used when WITH_MPI == T.
 CC := gcc
 MPICC := mpicc
-CFLAGS = -O3 -march=native -flto
+CFLAGS = -O3 -march=native
+
+# LTO option. Recommended flags for different compilers:
+#   gcc: -flto=auto, intel: -ipo, clang: -flto
+CFLAGS += -flto=auto
 
 # Set `WITH_MPI := T` to enable MPI.
 WITH_MPI := 

src/fcfc/2pt/build_tree.c

@@ -264,12 +264,17 @@ void *tree_create(const CONF *conf, CF *cf, const int idx
 
     /* Process weights */
     if (conf->has_wt[idx]) {
-      double sumw = 0;
+      double sumw, sumw2;
+      sumw = sumw2 = 0;
   #ifdef OMP
-    #pragma omp parallel for reduction(+:sumw) default(none) shared(data)
+    #pragma omp parallel for reduction(+:sumw,sumw2) default(none) shared(data)
   #endif
-      for (size_t i = 0; i < data->n; i++) sumw += data->w[i];
+      for (size_t i = 0; i < data->n; i++) {
+        sumw += data->w[i];
+        sumw2 += data->w[i] * data->w[i];
+      }
       data->wt = sumw;
+      data->w2 = sumw2;
     }
     else if (cf->cat_wt[idx]) {
   #if FCFC_SIMD  ==  FCFC_SIMD_NONE
@@ -285,7 +290,7 @@ void *tree_create(const CONF *conf, CF *cf, const int idx
     #pragma omp parallel for default(none) shared(data)
   #endif
       for (size_t i = 0; i < data->n; i++) data->w[i] = 1;
-      data->wt = (double) data->n;
+      data->wt = data->w2 = (double) data->n;
     }
 
     /* Construct the tree. */

src/fcfc/2pt/eval_cf.c

@@ -12,13 +12,11 @@
 
 #include "eval_cf.h"
 #include "count_func.h"
-#include "read_file.h"
 #include "read_res.h"
 #include "save_res.h"
 #include "legpoly.h"
 #include "build_tree.h"
 #include <stdlib.h>
-#include <math.h>
 
 #ifdef OMP
 #include <omp.h>
@@ -125,7 +123,8 @@ static int eval_pairs(const CONF *conf, CF *cf
         /* Double auto pairs. */
         if (cf->wt[i]) {
           for (size_t k = 0; k < cf->ntot; k++) cf->cnt[i][k].d *= 2;
-          cf->norm[i] = cf->data[cat[0]].wt * (cf->data[cat[0]].wt - 1);
+          cf->norm[i] = cf->data[cat[0]].wt * cf->data[cat[0]].wt
+              - cf->data[cat[0]].w2;
         }
         else {
           for (size_t k = 0; k < cf->ntot; k++) cf->cnt[i][k].i *= 2;