reorg, refactor, recompile

squiggle.c

@@ -8,7 +8,7 @@
 #define NORMAL90CONFIDENCE 1.6448536269514727
 
 // Pseudo Random number generator
-uint64_t xorshift32(uint32_t* seed)
+static uint64_t xorshift32(uint32_t* seed)
 {
     // Algorithm "xor" from p. 4 of Marsaglia, "Xorshift RNGs"
     // See:
@@ -24,7 +24,7 @@ uint64_t xorshift32(uint32_t* seed)
     return *seed = x;
 }
 
-uint64_t xorshift64(uint64_t* seed)
+static uint64_t xorshift64(uint64_t* seed)
 {
     // same as above, but for generating doubles instead of floats
     uint64_t x = *seed;
@@ -196,16 +196,6 @@ double array_std(double* array, int length)
     return std;
 }
 
-void array_print(double xs[], int n)
-{
-    printf("[");
-    for (int i = 0; i < n - 1; i++) {
-        printf("%f, ", xs[i]);
-    }
-    printf("%f", xs[n - 1]);
-    printf("]\n");
-}
-
 // Mixture function
 double sample_mixture(double (*samplers[])(uint64_t*), double* weights, int n_dists, uint64_t* seed)
 {

squiggle_more.c

@@ -7,37 +7,56 @@
 #include <stdlib.h>
 #include "squiggle.h"
 
-/* Math constants */
-#define PI 3.14159265358979323846 // M_PI in gcc gnu99
-#define NORMAL90CONFIDENCE 1.6448536269514727
+/* Parallel sampler */
+void parallel_sampler(double (*sampler)(uint64_t* seed), double* results, int n_threads, int n_samples){
+    if((n_samples % n_threads) != 0){
+        fprintf(stderr, "Number of samples isn't divisible by number of threads, aborting\n");
+        exit(1);
+    }
+    uint64_t** seeds = malloc(n_threads * sizeof(uint64_t*));
+    for (uint64_t i = 0; i < n_threads; i++) {
+        seeds[i] = malloc(sizeof(uint64_t));
+        *seeds[i] = i + 1; // xorshift can't start with 0
+    }
 
-/* Some error niceties */
-// These won't be used until later 
-#define MAX_ERROR_LENGTH 500
-#define EXIT_ON_ERROR 0
-#define PROCESS_ERROR(error_msg) process_error(error_msg, EXIT_ON_ERROR, __FILE__, __LINE__)
+    int i;
+    #pragma omp parallel private(i)
+    {
+        #pragma omp for
+        for (i = 0; i < n_threads; i++) {
+            int lower_bound = i * (n_samples / n_threads);
+            int upper_bound = ((i+1) * (n_samples / n_threads)) - 1;
+            // printf("Lower bound: %d, upper bound: %d\n", lower_bound, upper_bound);
+            for (int j = lower_bound; j < upper_bound; j++) {
+                results[j] = sampler(seeds[i]);
+            }
+        }
+    }
+
+    for (uint64_t i = 0; i < n_threads; i++) {
+        free(seeds[i]);
+    }
+    free(seeds);
+}
 
 /* Get confidence intervals, given a sampler */
 // Not in core yet because I'm not sure how much I like the struct
 // and the built-in 100k samples
 // to do: add n to function parameters and document
+
 typedef struct ci_t {
     float low;
     float high;
 } ci;
-typedef struct ci_searcher_t {
-    double num;
-    int remaining;
-} ci_searcher;
 
-void swp(int i, int j, double xs[])
+static void swp(int i, int j, double xs[])
 {
     double tmp = xs[i];
     xs[i] = xs[j];
     xs[j] = tmp;
 }
 
-int partition(int low, int high, double xs[], int length)
+static int partition(int low, int high, double xs[], int length)
 {
     // To understand this function:
     // - see the note after gt variable definition
@@ -56,8 +75,9 @@ int partition(int low, int high, double xs[], int length)
     return gt;
 }
 
-double quickselect(int k, double xs[], int length)
+static double quickselect(int k, double xs[], int length)
 {
+    // https://en.wikipedia.org/wiki/Quickselect
     int low = 0;
     int high = length - 1;
     for (;;) {
@@ -75,37 +95,92 @@ double quickselect(int k, double xs[], int length)
     }
 }
 
-ci get_90_confidence_interval(double (*sampler)(uint64_t*), uint64_t* seed)
-{
-    int n = 100 * 1000;
-    double* samples_array = malloc(n * sizeof(double));
+ci sampler_get_ci(double (*sampler)(uint64_t*), ci interval, int n, uint64_t* seed){
+    double* xs = malloc(n * sizeof(double));
     for (int i = 0; i < n; i++) {
-        samples_array[i] = sampler(seed);
+        xs[i] = sampler(seed);
     }
-    // 10% confidence interval: n/20, n - n/20
-    ci_searcher low = {.x = samples_array[0], .remaining = n/20) };
-    ci_searcher high = {.x = samples_array[0], .remaining = n-(n/20) };
 
-    // test with finding the lowest
-    for(int j=1; i<n; j++){
-        if(low.x > samples_array[i]){
-            low.x = samples_array[i];
-        }
-    }
+    int low_k = floor(interval.low * n);
+    int high_k = ceil(interval.high * n);
 
     ci result = {
-        .low = samples_array[5000],
-        .high = samples_array[94999],
+        .low = quickselect(low_k, xs, n),
+        .high = quickselect(high_k, xs, n),
     };
-    free(samples_array);
+    free(xs);
+    return result;
 
+}
+
+ci get_90_confidence_interval(double (*sampler)(uint64_t*), uint64_t* seed)
+{
+    return sampler_get_ci(sampler, (ci) {.low = 0.05, .high = 0.95}, 1000000, seed);
+}
+
+/* Algebra manipulations */
+// here I discover named structs,
+// which mean that I don't have to be typing
+// struct blah all the time.
+
+#define NORMAL90CONFIDENCE 1.6448536269514727
+
+typedef struct normal_params_t {
+    double mean;
+    double std;
+} normal_params;
+
+normal_params algebra_sum_normals(normal_params a, normal_params b)
+{
+    normal_params result = {
+        .mean = a.mean + b.mean,
+        .std = sqrt((a.std * a.std) + (b.std * b.std)),
+    };
+    return result;
+}
+
+typedef struct lognormal_params_t {
+    double logmean;
+    double logstd;
+} lognormal_params;
+
+lognormal_params algebra_product_lognormals(lognormal_params a, lognormal_params b)
+{
+    lognormal_params result = {
+        .logmean = a.logmean + b.logmean,
+        .logstd = sqrt((a.logstd * a.logstd) + (b.logstd * b.logstd)),
+    };
+    return result;
+}
+
+lognormal_params convert_ci_to_lognormal_params(ci x)
+{
+    double loghigh = logf(x.high);
+    double loglow = logf(x.low);
+    double logmean = (loghigh + loglow) / 2.0;
+    double logstd = (loghigh - loglow) / (2.0 * NORMAL90CONFIDENCE);
+    lognormal_params result = { .logmean = logmean, .logstd = logstd };
+    return result;
+}
+
+ci convert_lognormal_params_to_ci(lognormal_params y)
+{
+    double h = y.logstd * NORMAL90CONFIDENCE;
+    double loghigh = y.logmean + h;
+    double loglow = y.logmean - h;
+    ci result = { .low = exp(loglow), .high = exp(loghigh) };
     return result;
 }
 
 /* Scaffolding to handle errors */
-// We are building towards sample from an arbitrary cdf
+// We will sample from an arbitrary cdf
 // and that operation might fail
 // so we build some scaffolding here
+
+#define MAX_ERROR_LENGTH 500
+#define EXIT_ON_ERROR 0
+#define PROCESS_ERROR(error_msg) process_error(error_msg, EXIT_ON_ERROR, __FILE__, __LINE__)
+
 struct box {
     int empty;
     double content;
@@ -297,85 +372,15 @@ double sampler_cdf_danger(struct box cdf(double), uint64_t* seed)
 	}
 }
 
-/* Algebra manipulations */
-// here I discover named structs,
-// which mean that I don't have to be typing
-// struct blah all the time.
-typedef struct normal_params_t {
-    double mean;
-    double std;
-} normal_params;
+/* array print: potentially useful for debugging */
 
-normal_params algebra_sum_normals(normal_params a, normal_params b)
+void array_print(double xs[], int n)
 {
-    normal_params result = {
-        .mean = a.mean + b.mean,
-        .std = sqrt((a.std * a.std) + (b.std * b.std)),
-    };
-    return result;
-}
-
-typedef struct lognormal_params_t {
-    double logmean;
-    double logstd;
-} lognormal_params;
-
-lognormal_params algebra_product_lognormals(lognormal_params a, lognormal_params b)
-{
-    lognormal_params result = {
-        .logmean = a.logmean + b.logmean,
-        .logstd = sqrt((a.logstd * a.logstd) + (b.logstd * b.logstd)),
-    };
-    return result;
-}
-
-lognormal_params convert_ci_to_lognormal_params(ci x)
-{
-    double loghigh = logf(x.high);
-    double loglow = logf(x.low);
-    double logmean = (loghigh + loglow) / 2.0;
-    double logstd = (loghigh - loglow) / (2.0 * NORMAL90CONFIDENCE);
-    lognormal_params result = { .logmean = logmean, .logstd = logstd };
-    return result;
-}
-
-ci convert_lognormal_params_to_ci(lognormal_params y)
-{
-    double h = y.logstd * NORMAL90CONFIDENCE;
-    double loghigh = y.logmean + h;
-    double loglow = y.logmean - h;
-    ci result = { .low = exp(loglow), .high = exp(loghigh) };
-    return result;
-}
-
-/* Parallel sampler */
-void parallel_sampler(double (*sampler)(uint64_t* seed), double* results, int n_threads, int n_samples){
-    if((n_samples % n_threads) != 0){
-        fprintf(stderr, "Number of samples isn't divisible by number of threads, aborting\n");
-        exit(1);
+    printf("[");
+    for (int i = 0; i < n - 1; i++) {
+        printf("%f, ", xs[i]);
     }
-    uint64_t** seeds = malloc(n_threads * sizeof(uint64_t*));
-    for (uint64_t i = 0; i < n_threads; i++) {
-        seeds[i] = malloc(sizeof(uint64_t));
-        *seeds[i] = i + 1; // xorshift can't start with 0
-    }
-
-    int i;
-    #pragma omp parallel private(i)
-    {
-        #pragma omp for
-        for (i = 0; i < n_threads; i++) {
-            int lower_bound = i * (n_samples / n_threads);
-            int upper_bound = ((i+1) * (n_samples / n_threads)) - 1;
-            // printf("Lower bound: %d, upper bound: %d\n", lower_bound, upper_bound);
-            for (int j = lower_bound; j < upper_bound; j++) {
-                results[j] = sampler(seeds[i]);
-            }
-        }
-    }
-
-    for (uint64_t i = 0; i < n_threads; i++) {
-        free(seeds[i]);
-    }
-    free(seeds);
+    printf("%f", xs[n - 1]);
+    printf("]\n");
 }
+

squiggle_more.h

@@ -1,35 +1,17 @@
 #ifndef SQUIGGLE_C_EXTRA
 #define SQUIGGLE_C_EXTRA 
 
-// Box
-struct box {
-    int empty;
-    double content;
-    char* error_msg;
-};
+/* Parallel sampling */
+void parallel_sampler(double (*sampler)(uint64_t* seed), double* results, int n_threads, int n_samples);
 
-// Macros to handle errors
-#define MAX_ERROR_LENGTH 500
-#define EXIT_ON_ERROR 0
-#define PROCESS_ERROR(error_msg) process_error(error_msg, EXIT_ON_ERROR, __FILE__, __LINE__)
-struct box process_error(const char* error_msg, int should_exit, char* file, int line);
-
-// Inverse cdf
-struct box inverse_cdf_double(double cdf(double), double p);
-struct box inverse_cdf_box(struct box cdf_box(double), double p);
-
-// Samplers from cdf
-struct box sampler_cdf_double(double cdf(double), uint64_t* seed);
-struct box sampler_cdf_box(struct box cdf(double), uint64_t* seed);
-
-// Get 90% confidence interval
+/* Get 90% confidence interval */
 typedef struct ci_t {
     float low;
     float high;
 } ci;
 ci get_90_confidence_interval(double (*sampler)(uint64_t*), uint64_t* seed);
 
-// small algebra manipulations
+/* Algebra manipulations */
 
 typedef struct normal_params_t {
     double mean;
@@ -46,6 +28,24 @@ lognormal_params algebra_product_lognormals(lognormal_params a, lognormal_params
 lognormal_params convert_ci_to_lognormal_params(ci x);
 ci               convert_lognormal_params_to_ci(lognormal_params y);
 
-void parallel_sampler(double (*sampler)(uint64_t* seed), double* results, int n_threads, int n_samples);
+/* Error handling */
+struct box {
+    int empty;
+    double content;
+    char* error_msg;
+};
+#define MAX_ERROR_LENGTH 500
+#define EXIT_ON_ERROR 0
+#define PROCESS_ERROR(error_msg) process_error(error_msg, EXIT_ON_ERROR, __FILE__, __LINE__)
+struct box process_error(const char* error_msg, int should_exit, char* file, int line);
+void array_print(double* array, int length);
+
+/* Inverse cdf */
+struct box inverse_cdf_double(double cdf(double), double p);
+struct box inverse_cdf_box(struct box cdf_box(double), double p);
+
+/* Samplers from cdf */
+struct box sampler_cdf_double(double cdf(double), uint64_t* seed);
+struct box sampler_cdf_box(struct box cdf(double), uint64_t* seed);
 
 #endif