Problem 219

Minimal cost for 10^9 binary codes with costs 1 and 4.

View problem on Project Euler

Performance comparison

Flow source

# Project Euler 219
# Minimal cost for 10^9 binary codes with costs 1 and 4.

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

function main() -> i32 {
    let LIMIT: i64 = 1000000000
    let costs: ptr<i64> = calloc(80, 8)
    if costs == null { return 1 }
    costs[1] = 1
    costs[4] = 1
    let mut total: i64 = 5
    let mut current: i64 = 1
    let mut remaining: i64 = LIMIT - 2
    while remaining > 0 {
        while costs[current] == 0 { current = current + 1 }
        let mut block: i64 = costs[current]
        if block > remaining { block = remaining }
        remaining = remaining - block
        costs[current] = costs[current] - block
        costs[current + 1] = costs[current + 1] + block
        costs[current + 4] = costs[current + 4] + block
        total = total + block * (current + 5)
    }
    printf("%lld\n", total)
    free(costs)
    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; }

int32_t main(void);



int32_t main(void) {
    int64_t LIMIT = 1000000000;
    int64_t* costs = (int64_t*)(calloc(80, 8));
    if (costs == NULL) {
        return 1;
    }
    costs[1] = 1;
    costs[4] = 1;
    int64_t total = 5;
    int64_t current = 1;
    int64_t remaining = (LIMIT - 2);
    while (remaining > 0) {
        while (costs[current] == 0) {
            current = (current + 1);
        }
        int64_t block = costs[current];
        if (block > remaining) {
            block = remaining;
        }
        remaining = (remaining - block);
        costs[current] = (costs[current] - block);
        costs[(current + 1)] = (costs[(current + 1)] + block);
        costs[(current + 4)] = (costs[(current + 4)] + block);
        total = (total + (block * (current + 5)));
    }
    printf("%lld\n", total);
    free(costs);
    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 @main() -> i32 {
    %0 = arith.constant 1000000000 : i32
    %1 = arith.extsi %0 : i32 to i64
    %3 = arith.constant 80 : i32
    %4 = arith.constant 8 : i32
    %5 = arith.extsi %3 : i32 to i64
    %6 = arith.extsi %4 : i32 to i64
    %2 = func.call @calloc(%5, %6) : (i64, i64) -> !llvm.ptr
    %7 = llvm.mlir.zero : !llvm.ptr
    %8 = llvm.icmp "eq" %2, %7 : !llvm.ptr
    cf.cond_br %8, ^bb0, ^bb1
    ^bb0:
      %9 = arith.constant 1 : i32
      func.return %9 : i32
    ^bb1:
      cf.br ^bb2
    ^bb2:
    %10 = arith.constant 1 : i32
    %11 = arith.constant 1 : i32
    %12 = arith.extsi %10 : i32 to i64
    %13 = arith.extsi %11 : i32 to i64
    %14 = llvm.getelementptr %2[%13] : (!llvm.ptr, i64) -> !llvm.ptr, i64
    llvm.store %12, %14 : i64, !llvm.ptr
    %15 = arith.constant 1 : i32
    %16 = arith.constant 4 : i32
    %17 = arith.extsi %15 : i32 to i64
    %18 = arith.extsi %16 : i32 to i64
    %19 = llvm.getelementptr %2[%18] : (!llvm.ptr, i64) -> !llvm.ptr, i64
    llvm.store %17, %19 : i64, !llvm.ptr
    %20 = arith.constant 5 : i32
    %21 = arith.extsi %20 : i32 to i64
    %22 = llvm.mlir.constant(1 : i64) : i64
    %23 = llvm.alloca %22 x i64 : (i64) -> !llvm.ptr
    llvm.store %21, %23 : i64, !llvm.ptr
    %24 = arith.constant 1 : i32
    %25 = arith.extsi %24 : i32 to i64
    %26 = llvm.mlir.constant(1 : i64) : i64
    %27 = llvm.alloca %26 x i64 : (i64) -> !llvm.ptr
    llvm.store %25, %27 : i64, !llvm.ptr
    %28 = arith.constant 2 : i32
    %30 = arith.extsi %28 : i32 to i64
    %29 = arith.subi %1, %30 : i64
    %31 = llvm.mlir.constant(1 : i64) : i64
    %32 = llvm.alloca %31 x i64 : (i64) -> !llvm.ptr
    llvm.store %29, %32 : i64, !llvm.ptr
    cf.br ^bb3
    ^bb3:
    %33 = llvm.load %32 : !llvm.ptr -> i64
    %34 = arith.constant 0 : i32
    %36 = arith.extsi %34 : i32 to i64
    %35 = arith.cmpi sgt, %33, %36 : i64
    cf.cond_br %35, ^bb4, ^bb5
    ^bb4:
      cf.br ^bb6
      ^bb6:
      %38 = llvm.load %27 : !llvm.ptr -> i64
      %39 = llvm.getelementptr %2[%38] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      %37 = llvm.load %39 : !llvm.ptr -> i64
      %40 = arith.constant 0 : i32
      %42 = arith.extsi %40 : i32 to i64
      %41 = arith.cmpi eq, %37, %42 : i64
      cf.cond_br %41, ^bb7, ^bb8
      ^bb7:
        %43 = llvm.load %27 : !llvm.ptr -> i64
        %44 = arith.constant 1 : i32
        %46 = arith.extsi %44 : i32 to i64
        %45 = arith.addi %43, %46 : i64
        llvm.store %45, %27 : i64, !llvm.ptr
        cf.br ^bb6
      ^bb8:
      %48 = llvm.load %27 : !llvm.ptr -> i64
      %49 = llvm.getelementptr %2[%48] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      %47 = llvm.load %49 : !llvm.ptr -> i64
      %50 = llvm.mlir.constant(1 : i64) : i64
      %51 = llvm.alloca %50 x i64 : (i64) -> !llvm.ptr
      llvm.store %47, %51 : i64, !llvm.ptr
      %52 = llvm.load %51 : !llvm.ptr -> i64
      %53 = llvm.load %32 : !llvm.ptr -> i64
      %54 = arith.cmpi sgt, %52, %53 : i64
      cf.cond_br %54, ^bb9, ^bb10
      ^bb9:
        %55 = llvm.load %32 : !llvm.ptr -> i64
        llvm.store %55, %51 : i64, !llvm.ptr
        cf.br ^bb11
      ^bb10:
        cf.br ^bb11
      ^bb11:
      %56 = llvm.load %32 : !llvm.ptr -> i64
      %57 = llvm.load %51 : !llvm.ptr -> i64
      %58 = arith.subi %56, %57 : i64
      llvm.store %58, %32 : i64, !llvm.ptr
      %60 = llvm.load %27 : !llvm.ptr -> i64
      %61 = llvm.getelementptr %2[%60] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      %59 = llvm.load %61 : !llvm.ptr -> i64
      %62 = llvm.load %51 : !llvm.ptr -> i64
      %63 = arith.subi %59, %62 : i64
      %64 = llvm.load %27 : !llvm.ptr -> i64
      %65 = llvm.getelementptr %2[%64] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      llvm.store %63, %65 : i64, !llvm.ptr
      %67 = llvm.load %27 : !llvm.ptr -> i64
      %68 = arith.constant 1 : i32
      %70 = arith.extsi %68 : i32 to i64
      %69 = arith.addi %67, %70 : i64
      %71 = llvm.getelementptr %2[%69] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      %66 = llvm.load %71 : !llvm.ptr -> i64
      %72 = llvm.load %51 : !llvm.ptr -> i64
      %73 = arith.addi %66, %72 : i64
      %74 = llvm.load %27 : !llvm.ptr -> i64
      %75 = arith.constant 1 : i32
      %77 = arith.extsi %75 : i32 to i64
      %76 = arith.addi %74, %77 : i64
      %78 = llvm.getelementptr %2[%76] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      llvm.store %73, %78 : i64, !llvm.ptr
      %80 = llvm.load %27 : !llvm.ptr -> i64
      %81 = arith.constant 4 : i32
      %83 = arith.extsi %81 : i32 to i64
      %82 = arith.addi %80, %83 : i64
      %84 = llvm.getelementptr %2[%82] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      %79 = llvm.load %84 : !llvm.ptr -> i64
      %85 = llvm.load %51 : !llvm.ptr -> i64
      %86 = arith.addi %79, %85 : i64
      %87 = llvm.load %27 : !llvm.ptr -> i64
      %88 = arith.constant 4 : i32
      %90 = arith.extsi %88 : i32 to i64
      %89 = arith.addi %87, %90 : i64
      %91 = llvm.getelementptr %2[%89] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      llvm.store %86, %91 : i64, !llvm.ptr
      %92 = llvm.load %23 : !llvm.ptr -> i64
      %93 = llvm.load %51 : !llvm.ptr -> i64
      %94 = llvm.load %27 : !llvm.ptr -> i64
      %95 = arith.constant 5 : i32
      %97 = arith.extsi %95 : i32 to i64
      %96 = arith.addi %94, %97 : i64
      %98 = arith.muli %93, %96 : i64
      %99 = arith.addi %92, %98 : i64
      llvm.store %99, %23 : i64, !llvm.ptr
      cf.br ^bb3
    ^bb5:
    %100 = llvm.mlir.addressof @str_0 : !llvm.ptr
    %101 = llvm.load %23 : !llvm.ptr -> i64
    %102 = llvm.call @printf(%100, %101) vararg(!llvm.func<i32 (ptr, ...)>) : (!llvm.ptr, i64) -> i32
    func.call @free(%2) : (!llvm.ptr) -> ()
    %104 = arith.constant 0 : i32
    func.return %104 : i32
  }
}