Problem 297

Sum of Zeckendorf lengths for 0 < n < 10^17.

View problem on Project Euler

Performance comparison

Flow source

# Project Euler 297
# Sum of Zeckendorf lengths for 0 < n < 10^17.

extern {
    function calloc(n: i64, size: i64) -> ptr<void>
    function free(p: ptr<void>) -> void
}

function search(x: i64, fib: ptr<i64>, fib_sum: ptr<i64>) -> i64 {
    let mut pos: i64 = 0
    while fib[pos + 1] <= x {
        pos = pos + 1
    }
    let reduced: i64 = x - fib[pos]
    if reduced == 0 {
        return fib_sum[pos]
    }
    return fib_sum[pos] + reduced + search(reduced, fib, fib_sum)
}

function main() -> i32 {
    let limit: i64 = 100000000000000000
    let fib: ptr<i64> = calloc(100, 8)
    let fib_sum: ptr<i64> = calloc(100, 8)
    if fib == null || fib_sum == null { return 1 }
    fib[0] = 1
    fib[1] = 2
    fib_sum[0] = 1
    fib_sum[1] = 2
    let mut size: i64 = 2
    while fib[size - 1] < limit {
        fib[size] = fib[size - 1] + fib[size - 2]
        fib_sum[size] = fib_sum[size - 1] + fib_sum[size - 2] + fib[size - 2] - 1
        size = size + 1
    }
    printf("%lld\n", search(limit - 1, fib, fib_sum))
    free(fib)
    free(fib_sum)
    return 0
}

Generated C

#include <stdint.h>
#include <stdbool.h>
#include <stdio.h>
#include <stdlib.h>
#include <string.h>

/* Flow runtime helpers */
typedef struct flow_temp_node { struct flow_temp_node* next; } flow_temp_node;
static flow_temp_node* flow_temp_head = NULL;
static int flow_temp_atexit_set = 0;
__attribute__((unused)) static void flow_temp_free_all(void) {
    while (flow_temp_head) {
        flow_temp_node* n = flow_temp_head;
        flow_temp_head = n->next;
        free(n);
    }
}
__attribute__((unused)) static void* flow_temp_alloc(size_t nbytes) {
    flow_temp_node* node = (flow_temp_node*)malloc(sizeof(flow_temp_node) + nbytes);
    if (!node) return NULL;
    node->next = flow_temp_head;
    flow_temp_head = node;
    if (!flow_temp_atexit_set) {
        flow_temp_atexit_set = 1;
        atexit(flow_temp_free_all);
    }
    return (void*)(node + 1);
}
#ifndef FLOW_DIAG
#define FLOW_DIAG(msg) fprintf(stderr, "%s", (msg))
#endif
#ifndef FLOW_LOG
#define FLOW_LOG(fmt, ...) printf(fmt, __VA_ARGS__)
#endif
#ifndef FLOW_LOG_EMPTY
#define FLOW_LOG_EMPTY(fmt) printf(fmt)
#endif
static char* flow_strcat(const char* a, const char* b) {
    size_t la = strlen(a ? a : ""), lb = strlen(b ? b : "");
    char* r = (char*)flow_temp_alloc(la + lb + 1);
    if (!r) return NULL;
    if (la) memcpy(r, a, la);
    if (lb) memcpy(r + la, b, lb);
    r[la + lb] = '\0';
    return r;
}

#define __flow_in_arr(arr, val) __extension__ ({ \
    int _found = 0; \
    size_t _n = sizeof(arr)/sizeof((arr)[0]); \
    for (size_t _i = 0; _i < _n; _i++) { \
        if ((arr)[_i] == (val)) { _found = 1; break; } \
    } _found; })

/* Unified fault handler (MISRA #279) — override with -DFLOW_FAULT_HANDLER=fn */
#ifndef FLOW_FAULT_HANDLER
__attribute__((unused)) static inline void flow_fault_handler(const char* msg) {
    fprintf(stderr, "flow: %s\n", msg ? msg : "fault");
    abort();
#if defined(__GNUC__) || defined(__clang__)
    __builtin_unreachable();
#endif
}
#else
#define flow_fault_handler FLOW_FAULT_HANDLER
#endif
#define flow_div_by_zero_handler() flow_fault_handler("division by zero")
#define flow_shift_ub_handler() flow_fault_handler("invalid shift (amount out of range or left-shift of negative)")

#ifndef FLOW_CHECKED_DIV
#define FLOW_CHECKED_DIV(L, R) (((R) != 0) ? ((L) / (R)) : (flow_div_by_zero_handler(), (L) * 0))
#endif
#ifndef FLOW_CHECKED_MOD
#define FLOW_CHECKED_MOD(L, R) (((R) != 0) ? ((L) % (R)) : (flow_div_by_zero_handler(), (L) * 0))
#endif
#ifndef FLOW_CHECKED_SHL
#define FLOW_CHECKED_SHL(L, R) ((((R) >= 0) && ((unsigned long long)(R) < (sizeof(L) * 8ull)) && ((L) >= 0)) ? ((L) << (R)) : (flow_shift_ub_handler(), (L) * 0))
#endif
#ifndef FLOW_CHECKED_SHR
#define FLOW_CHECKED_SHR(L, R) ((((R) >= 0) && ((unsigned long long)(R) < (sizeof(L) * 8ull))) ? ((L) >> (R)) : (flow_shift_ub_handler(), (L) * 0))
#endif

#include <math.h>

void* _ui_state = NULL;

static inline float i32_to_f32(int32_t v) { return (float)v; }

/* Host stub for @gpu kernels (device codegen replaces this). */
static inline int32_t gpu_thread_id(void) { return 0; }

int64_t search_i64_ptr_i64_ptr_i64(int64_t x, int64_t* fib, int64_t* fib_sum);
int32_t main(void);



int64_t search_i64_ptr_i64_ptr_i64(int64_t x, int64_t* fib, int64_t* fib_sum) {
    int64_t pos = 0;
    while (fib[(pos + 1)] <= x) {
        pos = (pos + 1);
    }
    int64_t reduced = (x - fib[pos]);
    if (reduced == 0) {
        return fib_sum[pos];
    }
    return ((fib_sum[pos] + reduced) + search_i64_ptr_i64_ptr_i64(reduced, fib, fib_sum));
}

int32_t main(void) {
    int64_t limit = 100000000000000000;
    int64_t* fib = (int64_t*)(calloc(100, 8));
    int64_t* fib_sum = (int64_t*)(calloc(100, 8));
    if ((fib == NULL || fib_sum == NULL)) {
        return 1;
    }
    fib[0] = 1;
    fib[1] = 2;
    fib_sum[0] = 1;
    fib_sum[1] = 2;
    int64_t size = 2;
    while (fib[(size - 1)] < limit) {
        fib[size] = (fib[(size - 1)] + fib[(size - 2)]);
        fib_sum[size] = (((fib_sum[(size - 1)] + fib_sum[(size - 2)]) + fib[(size - 2)]) - 1);
        size = (size + 1);
    }
    printf("%lld\n", search_i64_ptr_i64_ptr_i64((limit - 1), fib, fib_sum));
    free(fib);
    free(fib_sum);
    return 0;
}

Generated MLIR

module {
  llvm.func @printf(!llvm.ptr, ...) -> i32
  llvm.mlir.global internal constant @str_0("%lld\n\00") {addr_space = 0 : i32} : !llvm.array<6 x i8>
  func.func private @calloc(i64, i64) -> !llvm.ptr
  func.func private @free(!llvm.ptr) -> ()
  func.func @search(%arg0: i64, %arg1: !llvm.ptr, %arg2: !llvm.ptr) -> i64 {
    %0 = arith.constant 0 : i32
    %1 = arith.extsi %0 : i32 to i64
    %2 = llvm.mlir.constant(1 : i64) : i64
    %3 = llvm.alloca %2 x i64 : (i64) -> !llvm.ptr
    llvm.store %1, %3 : i64, !llvm.ptr
    cf.br ^bb0
    ^bb0:
    %5 = llvm.load %3 : !llvm.ptr -> i64
    %6 = arith.constant 1 : i32
    %8 = arith.extsi %6 : i32 to i64
    %7 = arith.addi %5, %8 : i64
    %9 = llvm.getelementptr %arg1[%7] : (!llvm.ptr, i64) -> !llvm.ptr, i64
    %4 = llvm.load %9 : !llvm.ptr -> i64
    %10 = arith.cmpi sle, %4, %arg0 : i64
    cf.cond_br %10, ^bb1, ^bb2
    ^bb1:
      %11 = llvm.load %3 : !llvm.ptr -> i64
      %12 = arith.constant 1 : i32
      %14 = arith.extsi %12 : i32 to i64
      %13 = arith.addi %11, %14 : i64
      llvm.store %13, %3 : i64, !llvm.ptr
      cf.br ^bb0
    ^bb2:
    %16 = llvm.load %3 : !llvm.ptr -> i64
    %17 = llvm.getelementptr %arg1[%16] : (!llvm.ptr, i64) -> !llvm.ptr, i64
    %15 = llvm.load %17 : !llvm.ptr -> i64
    %18 = arith.subi %arg0, %15 : i64
    %19 = arith.constant 0 : i32
    %21 = arith.extsi %19 : i32 to i64
    %20 = arith.cmpi eq, %18, %21 : i64
    cf.cond_br %20, ^bb3, ^bb4
    ^bb3:
      %23 = llvm.load %3 : !llvm.ptr -> i64
      %24 = llvm.getelementptr %arg2[%23] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      %22 = llvm.load %24 : !llvm.ptr -> i64
      func.return %22 : i64
    ^bb4:
      cf.br ^bb5
    ^bb5:
    %26 = llvm.load %3 : !llvm.ptr -> i64
    %27 = llvm.getelementptr %arg2[%26] : (!llvm.ptr, i64) -> !llvm.ptr, i64
    %25 = llvm.load %27 : !llvm.ptr -> i64
    %28 = arith.addi %25, %18 : i64
    %29 = func.call @search(%18, %arg1, %arg2) : (i64, !llvm.ptr, !llvm.ptr) -> i64
    %30 = arith.addi %28, %29 : i64
    func.return %30 : i64
  }
  func.func @main() -> i32 {
    %31 = arith.constant 99999995705032704 : i32
    %32 = arith.extsi %31 : i32 to i64
    %34 = arith.constant 100 : i32
    %35 = arith.constant 8 : i32
    %36 = arith.extsi %34 : i32 to i64
    %37 = arith.extsi %35 : i32 to i64
    %33 = func.call @calloc(%36, %37) : (i64, i64) -> !llvm.ptr
    %39 = arith.constant 100 : i32
    %40 = arith.constant 8 : i32
    %41 = arith.extsi %39 : i32 to i64
    %42 = arith.extsi %40 : i32 to i64
    %38 = func.call @calloc(%41, %42) : (i64, i64) -> !llvm.ptr
    %43 = llvm.mlir.zero : !llvm.ptr
    %44 = llvm.icmp "eq" %33, %43 : !llvm.ptr
    %45 = scf.if %44 -> (i1) {
      %46 = arith.constant true
      scf.yield %46 : i1
    } else {
      %47 = llvm.mlir.zero : !llvm.ptr
      %48 = llvm.icmp "eq" %38, %47 : !llvm.ptr
      scf.yield %48 : i1
    }
    cf.cond_br %45, ^bb6, ^bb7
    ^bb6:
      %49 = arith.constant 1 : i32
      func.return %49 : i32
    ^bb7:
      cf.br ^bb8
    ^bb8:
    %50 = arith.constant 1 : i32
    %51 = arith.constant 0 : i32
    %52 = arith.extsi %50 : i32 to i64
    %53 = arith.extsi %51 : i32 to i64
    %54 = llvm.getelementptr %33[%53] : (!llvm.ptr, i64) -> !llvm.ptr, i64
    llvm.store %52, %54 : i64, !llvm.ptr
    %55 = arith.constant 2 : i32
    %56 = arith.constant 1 : i32
    %57 = arith.extsi %55 : i32 to i64
    %58 = arith.extsi %56 : i32 to i64
    %59 = llvm.getelementptr %33[%58] : (!llvm.ptr, i64) -> !llvm.ptr, i64
    llvm.store %57, %59 : i64, !llvm.ptr
    %60 = arith.constant 1 : i32
    %61 = arith.constant 0 : i32
    %62 = arith.extsi %60 : i32 to i64
    %63 = arith.extsi %61 : i32 to i64
    %64 = llvm.getelementptr %38[%63] : (!llvm.ptr, i64) -> !llvm.ptr, i64
    llvm.store %62, %64 : i64, !llvm.ptr
    %65 = arith.constant 2 : i32
    %66 = arith.constant 1 : i32
    %67 = arith.extsi %65 : i32 to i64
    %68 = arith.extsi %66 : i32 to i64
    %69 = llvm.getelementptr %38[%68] : (!llvm.ptr, i64) -> !llvm.ptr, i64
    llvm.store %67, %69 : i64, !llvm.ptr
    %70 = arith.constant 2 : i32
    %71 = arith.extsi %70 : i32 to i64
    %72 = llvm.mlir.constant(1 : i64) : i64
    %73 = llvm.alloca %72 x i64 : (i64) -> !llvm.ptr
    llvm.store %71, %73 : i64, !llvm.ptr
    cf.br ^bb9
    ^bb9:
    %75 = llvm.load %73 : !llvm.ptr -> i64
    %76 = arith.constant 1 : i32
    %78 = arith.extsi %76 : i32 to i64
    %77 = arith.subi %75, %78 : i64
    %79 = llvm.getelementptr %33[%77] : (!llvm.ptr, i64) -> !llvm.ptr, i64
    %74 = llvm.load %79 : !llvm.ptr -> i64
    %80 = arith.cmpi slt, %74, %32 : i64
    cf.cond_br %80, ^bb10, ^bb11
    ^bb10:
      %82 = llvm.load %73 : !llvm.ptr -> i64
      %83 = arith.constant 1 : i32
      %85 = arith.extsi %83 : i32 to i64
      %84 = arith.subi %82, %85 : i64
      %86 = llvm.getelementptr %33[%84] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      %81 = llvm.load %86 : !llvm.ptr -> i64
      %88 = llvm.load %73 : !llvm.ptr -> i64
      %89 = arith.constant 2 : i32
      %91 = arith.extsi %89 : i32 to i64
      %90 = arith.subi %88, %91 : i64
      %92 = llvm.getelementptr %33[%90] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      %87 = llvm.load %92 : !llvm.ptr -> i64
      %93 = arith.addi %81, %87 : i64
      %94 = llvm.load %73 : !llvm.ptr -> i64
      %95 = llvm.getelementptr %33[%94] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      llvm.store %93, %95 : i64, !llvm.ptr
      %97 = llvm.load %73 : !llvm.ptr -> i64
      %98 = arith.constant 1 : i32
      %100 = arith.extsi %98 : i32 to i64
      %99 = arith.subi %97, %100 : i64
      %101 = llvm.getelementptr %38[%99] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      %96 = llvm.load %101 : !llvm.ptr -> i64
      %103 = llvm.load %73 : !llvm.ptr -> i64
      %104 = arith.constant 2 : i32
      %106 = arith.extsi %104 : i32 to i64
      %105 = arith.subi %103, %106 : i64
      %107 = llvm.getelementptr %38[%105] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      %102 = llvm.load %107 : !llvm.ptr -> i64
      %108 = arith.addi %96, %102 : i64
      %110 = llvm.load %73 : !llvm.ptr -> i64
      %111 = arith.constant 2 : i32
      %113 = arith.extsi %111 : i32 to i64
      %112 = arith.subi %110, %113 : i64
      %114 = llvm.getelementptr %33[%112] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      %109 = llvm.load %114 : !llvm.ptr -> i64
      %115 = arith.addi %108, %109 : i64
      %116 = arith.constant 1 : i32
      %118 = arith.extsi %116 : i32 to i64
      %117 = arith.subi %115, %118 : i64
      %119 = llvm.load %73 : !llvm.ptr -> i64
      %120 = llvm.getelementptr %38[%119] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      llvm.store %117, %120 : i64, !llvm.ptr
      %121 = llvm.load %73 : !llvm.ptr -> i64
      %122 = arith.constant 1 : i32
      %124 = arith.extsi %122 : i32 to i64
      %123 = arith.addi %121, %124 : i64
      llvm.store %123, %73 : i64, !llvm.ptr
      cf.br ^bb9
    ^bb11:
    %125 = llvm.mlir.addressof @str_0 : !llvm.ptr
    %127 = arith.constant 1 : i32
    %129 = arith.extsi %127 : i32 to i64
    %128 = arith.subi %32, %129 : i64
    %126 = func.call @search(%128, %33, %38) : (i64, !llvm.ptr, !llvm.ptr) -> i64
    %130 = llvm.call @printf(%125, %126) vararg(!llvm.func<i32 (ptr, ...)>) : (!llvm.ptr, i64) -> i32
    func.call @free(%33) : (!llvm.ptr) -> ()
    func.call @free(%38) : (!llvm.ptr) -> ()
    %133 = arith.constant 0 : i32
    func.return %133 : i32
  }
}