Problem 205

P(Peter 9d4 beats Colin 6d6).

View problem on Project Euler

Performance comparison

Flow source

# Project Euler 205
# P(Peter 9d4 beats Colin 6d6).

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

function roll_dp(dices: i32, sides: i32, count: ptr<i64>) -> void {
    count[0] = 1
    let mut d: i32 = 0
    while d < dices {
        let mut next: ptr<i64> = calloc(40, 8)
        let mut t: i32 = 0
        while t <= 36 {
            if count[t] != 0 {
                let mut s: i32 = 1
                while s <= sides {
                    next[t + s] = next[t + s] + count[t]
                    s = s + 1
                }
            }
            t = t + 1
        }
        t = 0
        while t <= 36 {
            count[t] = next[t]
            t = t + 1
        }
        free(next)
        d = d + 1
    }
}

function main() -> i32 {
    let peter: ptr<i64> = calloc(40, 8)
    let colin: ptr<i64> = calloc(40, 8)
    if peter == null || colin == null { return 1 }
    roll_dp(9, 4, peter)
    roll_dp(6, 6, colin)

    let mut sum_p: f64 = 0.0
    let mut sum_c: f64 = 0.0
    let mut t: i32 = 0
    while t <= 36 {
        sum_p = sum_p + (peter[t] as f64)
        sum_c = sum_c + (colin[t] as f64)
        t = t + 1
    }

    let mut win: f64 = 0.0
    t = 1
    while t <= 36 {
        let mut num_wins: f64 = 0.0
        let mut u: i32 = 1
        while u < t {
            num_wins = num_wins + (colin[u] as f64)
            u = u + 1
        }
        win = win + (num_wins / sum_c) * ((peter[t] as f64) / sum_p)
        t = t + 1
    }
    printf("%.7f\n", win)
    free(peter); free(colin)
    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; }

void roll_dp_i32_i32_ptr_i64(int32_t dices, int32_t sides, int64_t* count);
int32_t main(void);



void roll_dp_i32_i32_ptr_i64(int32_t dices, int32_t sides, int64_t* count) {
    count[0] = 1;
    int32_t d = 0;
    while (d < dices) {
        int64_t* next = (int64_t*)(calloc(40, 8));
        int32_t t = 0;
        while (t <= 36) {
            if (count[t] != 0) {
                int32_t s = 1;
                while (s <= sides) {
                    next[(t + s)] = (next[(t + s)] + count[t]);
                    s = (s + 1);
                }
            }
            t = (t + 1);
        }
        t = 0;
        while (t <= 36) {
            count[t] = next[t];
            t = (t + 1);
        }
        free(next);
        d = (d + 1);
    }
}

int32_t main(void) {
    int64_t* peter = (int64_t*)(calloc(40, 8));
    int64_t* colin = (int64_t*)(calloc(40, 8));
    if ((peter == NULL || colin == NULL)) {
        return 1;
    }
    roll_dp_i32_i32_ptr_i64(9, 4, peter);
    roll_dp_i32_i32_ptr_i64(6, 6, colin);
    double sum_p = 0.0;
    double sum_c = 0.0;
    int32_t t = 0;
    while (t <= 36) {
        sum_p = (sum_p + ((double)(peter[t])));
        sum_c = (sum_c + ((double)(colin[t])));
        t = (t + 1);
    }
    double win = 0.0;
    t = 1;
    while (t <= 36) {
        double num_wins = 0.0;
        int32_t u = 1;
        while (u < t) {
            num_wins = (num_wins + ((double)(colin[u])));
            u = (u + 1);
        }
        win = (win + ((num_wins / sum_c) * (((double)(peter[t])) / sum_p)));
        t = (t + 1);
    }
    printf("%.7f\n", win);
    free(peter);
    free(colin);
    return 0;
}

Generated MLIR

module {
  llvm.func @printf(!llvm.ptr, ...) -> i32
  llvm.mlir.global internal constant @str_0("%.7f\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 @roll_dp(%arg0: i32, %arg1: i32, %arg2: !llvm.ptr) -> () {
    %0 = arith.constant 1 : i32
    %1 = arith.constant 0 : i32
    %2 = arith.extsi %0 : i32 to i64
    %3 = arith.extsi %1 : i32 to i64
    %4 = llvm.getelementptr %arg2[%3] : (!llvm.ptr, i64) -> !llvm.ptr, i64
    llvm.store %2, %4 : i64, !llvm.ptr
    %5 = arith.constant 0 : i32
    %6 = llvm.mlir.constant(1 : i64) : i64
    %7 = llvm.alloca %6 x i32 : (i64) -> !llvm.ptr
    llvm.store %5, %7 : i32, !llvm.ptr
    cf.br ^bb0
    ^bb0:
    %8 = llvm.load %7 : !llvm.ptr -> i32
    %9 = arith.cmpi slt, %8, %arg0 : i32
    cf.cond_br %9, ^bb1, ^bb2
    ^bb1:
      %11 = arith.constant 40 : i32
      %12 = arith.constant 8 : i32
      %13 = arith.extsi %11 : i32 to i64
      %14 = arith.extsi %12 : i32 to i64
      %10 = func.call @calloc(%13, %14) : (i64, i64) -> !llvm.ptr
      %15 = llvm.mlir.constant(1 : i64) : i64
      %16 = llvm.alloca %15 x !llvm.ptr : (i64) -> !llvm.ptr
      llvm.store %10, %16 : !llvm.ptr, !llvm.ptr
      %17 = arith.constant 0 : i32
      %18 = llvm.mlir.constant(1 : i64) : i64
      %19 = llvm.alloca %18 x i32 : (i64) -> !llvm.ptr
      llvm.store %17, %19 : i32, !llvm.ptr
      cf.br ^bb3
      ^bb3:
      %20 = llvm.load %19 : !llvm.ptr -> i32
      %21 = arith.constant 36 : i32
      %22 = arith.cmpi sle, %20, %21 : i32
      cf.cond_br %22, ^bb4, ^bb5
      ^bb4:
        %24 = llvm.load %19 : !llvm.ptr -> i32
        %25 = arith.extsi %24 : i32 to i64
        %26 = llvm.getelementptr %arg2[%25] : (!llvm.ptr, i64) -> !llvm.ptr, i64
        %23 = llvm.load %26 : !llvm.ptr -> i64
        %27 = arith.constant 0 : i32
        %29 = arith.extsi %27 : i32 to i64
        %28 = arith.cmpi ne, %23, %29 : i64
        cf.cond_br %28, ^bb6, ^bb7
        ^bb6:
          %30 = arith.constant 1 : i32
          %31 = llvm.mlir.constant(1 : i64) : i64
          %32 = llvm.alloca %31 x i32 : (i64) -> !llvm.ptr
          llvm.store %30, %32 : i32, !llvm.ptr
          cf.br ^bb9
          ^bb9:
          %33 = llvm.load %32 : !llvm.ptr -> i32
          %34 = arith.cmpi sle, %33, %arg1 : i32
          cf.cond_br %34, ^bb10, ^bb11
          ^bb10:
            %36 = llvm.load %16 : !llvm.ptr -> !llvm.ptr
            %37 = llvm.load %19 : !llvm.ptr -> i32
            %38 = llvm.load %32 : !llvm.ptr -> i32
            %39 = arith.addi %37, %38 : i32
            %40 = arith.extsi %39 : i32 to i64
            %41 = llvm.getelementptr %36[%40] : (!llvm.ptr, i64) -> !llvm.ptr, i64
            %35 = llvm.load %41 : !llvm.ptr -> i64
            %43 = llvm.load %19 : !llvm.ptr -> i32
            %44 = arith.extsi %43 : i32 to i64
            %45 = llvm.getelementptr %arg2[%44] : (!llvm.ptr, i64) -> !llvm.ptr, i64
            %42 = llvm.load %45 : !llvm.ptr -> i64
            %46 = arith.addi %35, %42 : i64
            %47 = llvm.load %16 : !llvm.ptr -> !llvm.ptr
            %48 = llvm.load %19 : !llvm.ptr -> i32
            %49 = llvm.load %32 : !llvm.ptr -> i32
            %50 = arith.addi %48, %49 : i32
            %51 = arith.extsi %50 : i32 to i64
            %52 = llvm.getelementptr %47[%51] : (!llvm.ptr, i64) -> !llvm.ptr, i64
            llvm.store %46, %52 : i64, !llvm.ptr
            %53 = llvm.load %32 : !llvm.ptr -> i32
            %54 = arith.constant 1 : i32
            %55 = arith.addi %53, %54 : i32
            llvm.store %55, %32 : i32, !llvm.ptr
            cf.br ^bb9
          ^bb11:
          cf.br ^bb8
        ^bb7:
          cf.br ^bb8
        ^bb8:
        %56 = llvm.load %19 : !llvm.ptr -> i32
        %57 = arith.constant 1 : i32
        %58 = arith.addi %56, %57 : i32
        llvm.store %58, %19 : i32, !llvm.ptr
        cf.br ^bb3
      ^bb5:
      %59 = arith.constant 0 : i32
      llvm.store %59, %19 : i32, !llvm.ptr
      cf.br ^bb12
      ^bb12:
      %60 = llvm.load %19 : !llvm.ptr -> i32
      %61 = arith.constant 36 : i32
      %62 = arith.cmpi sle, %60, %61 : i32
      cf.cond_br %62, ^bb13, ^bb14
      ^bb13:
        %64 = llvm.load %16 : !llvm.ptr -> !llvm.ptr
        %65 = llvm.load %19 : !llvm.ptr -> i32
        %66 = arith.extsi %65 : i32 to i64
        %67 = llvm.getelementptr %64[%66] : (!llvm.ptr, i64) -> !llvm.ptr, i64
        %63 = llvm.load %67 : !llvm.ptr -> i64
        %68 = llvm.load %19 : !llvm.ptr -> i32
        %69 = arith.extsi %68 : i32 to i64
        %70 = llvm.getelementptr %arg2[%69] : (!llvm.ptr, i64) -> !llvm.ptr, i64
        llvm.store %63, %70 : i64, !llvm.ptr
        %71 = llvm.load %19 : !llvm.ptr -> i32
        %72 = arith.constant 1 : i32
        %73 = arith.addi %71, %72 : i32
        llvm.store %73, %19 : i32, !llvm.ptr
        cf.br ^bb12
      ^bb14:
      %75 = llvm.load %16 : !llvm.ptr -> !llvm.ptr
      func.call @free(%75) : (!llvm.ptr) -> ()
      %76 = llvm.load %7 : !llvm.ptr -> i32
      %77 = arith.constant 1 : i32
      %78 = arith.addi %76, %77 : i32
      llvm.store %78, %7 : i32, !llvm.ptr
      cf.br ^bb0
    ^bb2:
    func.return
  }
  func.func @main() -> i32 {
    %80 = arith.constant 40 : i32
    %81 = arith.constant 8 : i32
    %82 = arith.extsi %80 : i32 to i64
    %83 = arith.extsi %81 : i32 to i64
    %79 = func.call @calloc(%82, %83) : (i64, i64) -> !llvm.ptr
    %85 = arith.constant 40 : i32
    %86 = arith.constant 8 : i32
    %87 = arith.extsi %85 : i32 to i64
    %88 = arith.extsi %86 : i32 to i64
    %84 = func.call @calloc(%87, %88) : (i64, i64) -> !llvm.ptr
    %89 = llvm.mlir.zero : !llvm.ptr
    %90 = llvm.icmp "eq" %79, %89 : !llvm.ptr
    %91 = scf.if %90 -> (i1) {
      %92 = arith.constant true
      scf.yield %92 : i1
    } else {
      %93 = llvm.mlir.zero : !llvm.ptr
      %94 = llvm.icmp "eq" %84, %93 : !llvm.ptr
      scf.yield %94 : i1
    }
    cf.cond_br %91, ^bb15, ^bb16
    ^bb15:
      %95 = arith.constant 1 : i32
      func.return %95 : i32
    ^bb16:
      cf.br ^bb17
    ^bb17:
    %97 = arith.constant 9 : i32
    %98 = arith.constant 4 : i32
    func.call @roll_dp(%97, %98, %79) : (i32, i32, !llvm.ptr) -> ()
    %100 = arith.constant 6 : i32
    %101 = arith.constant 6 : i32
    func.call @roll_dp(%100, %101, %84) : (i32, i32, !llvm.ptr) -> ()
    %102 = arith.constant 0.0 : f32
    %103 = arith.extf %102 : f32 to f64
    %104 = llvm.mlir.constant(1 : i64) : i64
    %105 = llvm.alloca %104 x f64 : (i64) -> !llvm.ptr
    llvm.store %103, %105 : f64, !llvm.ptr
    %106 = arith.constant 0.0 : f32
    %107 = arith.extf %106 : f32 to f64
    %108 = llvm.mlir.constant(1 : i64) : i64
    %109 = llvm.alloca %108 x f64 : (i64) -> !llvm.ptr
    llvm.store %107, %109 : f64, !llvm.ptr
    %110 = arith.constant 0 : i32
    %111 = llvm.mlir.constant(1 : i64) : i64
    %112 = llvm.alloca %111 x i32 : (i64) -> !llvm.ptr
    llvm.store %110, %112 : i32, !llvm.ptr
    cf.br ^bb18
    ^bb18:
    %113 = llvm.load %112 : !llvm.ptr -> i32
    %114 = arith.constant 36 : i32
    %115 = arith.cmpi sle, %113, %114 : i32
    cf.cond_br %115, ^bb19, ^bb20
    ^bb19:
      %116 = llvm.load %105 : !llvm.ptr -> f64
      %118 = llvm.load %112 : !llvm.ptr -> i32
      %119 = arith.extsi %118 : i32 to i64
      %120 = llvm.getelementptr %79[%119] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      %117 = llvm.load %120 : !llvm.ptr -> i64
      %121 = arith.sitofp %117 : i64 to f64
      %122 = arith.addf %116, %121 : f64
      llvm.store %122, %105 : f64, !llvm.ptr
      %123 = llvm.load %109 : !llvm.ptr -> f64
      %125 = llvm.load %112 : !llvm.ptr -> i32
      %126 = arith.extsi %125 : i32 to i64
      %127 = llvm.getelementptr %84[%126] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      %124 = llvm.load %127 : !llvm.ptr -> i64
      %128 = arith.sitofp %124 : i64 to f64
      %129 = arith.addf %123, %128 : f64
      llvm.store %129, %109 : f64, !llvm.ptr
      %130 = llvm.load %112 : !llvm.ptr -> i32
      %131 = arith.constant 1 : i32
      %132 = arith.addi %130, %131 : i32
      llvm.store %132, %112 : i32, !llvm.ptr
      cf.br ^bb18
    ^bb20:
    %133 = arith.constant 0.0 : f32
    %134 = arith.extf %133 : f32 to f64
    %135 = llvm.mlir.constant(1 : i64) : i64
    %136 = llvm.alloca %135 x f64 : (i64) -> !llvm.ptr
    llvm.store %134, %136 : f64, !llvm.ptr
    %137 = arith.constant 1 : i32
    llvm.store %137, %112 : i32, !llvm.ptr
    cf.br ^bb21
    ^bb21:
    %138 = llvm.load %112 : !llvm.ptr -> i32
    %139 = arith.constant 36 : i32
    %140 = arith.cmpi sle, %138, %139 : i32
    cf.cond_br %140, ^bb22, ^bb23
    ^bb22:
      %141 = arith.constant 0.0 : f32
      %142 = arith.extf %141 : f32 to f64
      %143 = llvm.mlir.constant(1 : i64) : i64
      %144 = llvm.alloca %143 x f64 : (i64) -> !llvm.ptr
      llvm.store %142, %144 : f64, !llvm.ptr
      %145 = arith.constant 1 : i32
      %146 = llvm.mlir.constant(1 : i64) : i64
      %147 = llvm.alloca %146 x i32 : (i64) -> !llvm.ptr
      llvm.store %145, %147 : i32, !llvm.ptr
      cf.br ^bb24
      ^bb24:
      %148 = llvm.load %147 : !llvm.ptr -> i32
      %149 = llvm.load %112 : !llvm.ptr -> i32
      %150 = arith.cmpi slt, %148, %149 : i32
      cf.cond_br %150, ^bb25, ^bb26
      ^bb25:
        %151 = llvm.load %144 : !llvm.ptr -> f64
        %153 = llvm.load %147 : !llvm.ptr -> i32
        %154 = arith.extsi %153 : i32 to i64
        %155 = llvm.getelementptr %84[%154] : (!llvm.ptr, i64) -> !llvm.ptr, i64
        %152 = llvm.load %155 : !llvm.ptr -> i64
        %156 = arith.sitofp %152 : i64 to f64
        %157 = arith.addf %151, %156 : f64
        llvm.store %157, %144 : f64, !llvm.ptr
        %158 = llvm.load %147 : !llvm.ptr -> i32
        %159 = arith.constant 1 : i32
        %160 = arith.addi %158, %159 : i32
        llvm.store %160, %147 : i32, !llvm.ptr
        cf.br ^bb24
      ^bb26:
      %161 = llvm.load %136 : !llvm.ptr -> f64
      %162 = llvm.load %144 : !llvm.ptr -> f64
      %163 = llvm.load %109 : !llvm.ptr -> f64
      %164 = arith.divf %162, %163 : f64
      %166 = llvm.load %112 : !llvm.ptr -> i32
      %167 = arith.extsi %166 : i32 to i64
      %168 = llvm.getelementptr %79[%167] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      %165 = llvm.load %168 : !llvm.ptr -> i64
      %169 = arith.sitofp %165 : i64 to f64
      %170 = llvm.load %105 : !llvm.ptr -> f64
      %171 = arith.divf %169, %170 : f64
      %172 = arith.mulf %164, %171 : f64
      %173 = arith.addf %161, %172 : f64
      llvm.store %173, %136 : f64, !llvm.ptr
      %174 = llvm.load %112 : !llvm.ptr -> i32
      %175 = arith.constant 1 : i32
      %176 = arith.addi %174, %175 : i32
      llvm.store %176, %112 : i32, !llvm.ptr
      cf.br ^bb21
    ^bb23:
    %177 = llvm.mlir.addressof @str_0 : !llvm.ptr
    %178 = llvm.load %136 : !llvm.ptr -> f64
    %179 = llvm.call @printf(%177, %178) vararg(!llvm.func<i32 (ptr, ...)>) : (!llvm.ptr, f64) -> i32
    func.call @free(%79) : (!llvm.ptr) -> ()
    func.call @free(%84) : (!llvm.ptr) -> ()
    %182 = arith.constant 0 : i32
    func.return %182 : i32
  }
}