Problem 018

Maximum path sum through the triangle (bottom-up DP).

View problem on Project Euler

Performance comparison

Flow source

# Project Euler 018
# Maximum path sum through the triangle (bottom-up DP).

extern {
    function fopen(path: string, mode: string) -> ptr<void>
    function fgetc(f: ptr<void>) -> i32
    function fclose(f: ptr<void>) -> i32
    function calloc(n: i64, size: i64) -> ptr<void>
    function free(p: ptr<void>) -> void
}

function read_int(f: ptr<void>) -> i32 {
    let mut c: i32 = fgetc(f)
    while c >= 0 && (c < 48 || c > 57) {
        c = fgetc(f)
    }
    if c < 0 {
        return -1
    }
    let mut v: i32 = 0
    while c >= 48 && c <= 57 {
        v = v * 10 + (c - 48)
        c = fgetc(f)
    }
    return v
}

function max2(a: i64, b: i64) -> i64 {
    if a > b {
        return a
    }
    return b
}

function main() -> i32 {
    let rows: i32 = 15
    let cells: i32 = rows * (rows + 1) / 2
    let t: ptr<i64> = calloc(cells as i64, 8)
    if t == null {
        return 1
    }

    let f: ptr<void> = fopen("data/p018.txt", "r")
    if f == null {
        printf("failed to read data/p018.txt\n")
        free(t)
        return 1
    }

    let mut idx: i32 = 0
    while idx < cells {
        let v: i32 = read_int(f)
        if v < 0 {
            break
        }
        t[idx] = v as i64
        idx = idx + 1
    }
    fclose(f)

    let mut r: i32 = rows - 2
    while r >= 0 {
        let base: i32 = r * (r + 1) / 2
        let below: i32 = (r + 1) * (r + 2) / 2
        let mut c: i32 = 0
        while c <= r {
            t[base + c] = t[base + c] + max2(t[below + c], t[below + c + 1])
            c = c + 1
        }
        r = r - 1
    }

    printf("%lld\n", t[0])
    free(t)
    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 read_int_ptr_void(void* f);
int64_t max2_i64_i64(int64_t a, int64_t b);
int32_t main(void);






int32_t read_int_ptr_void(void* f) {
    int32_t c = fgetc(f);
    while ((c >= 0 && (c < 48 || c > 57))) {
        c = fgetc(f);
    }
    if (c < 0) {
        return (-1);
    }
    int32_t v = 0;
    while ((c >= 48 && c <= 57)) {
        v = ((v * 10) + (c - 48));
        c = fgetc(f);
    }
    return v;
}

int64_t max2_i64_i64(int64_t a, int64_t b) {
    if (a > b) {
        return a;
    }
    return b;
}

int32_t main(void) {
    int32_t rows = 15;
    int32_t cells = FLOW_CHECKED_DIV(((rows * (rows + 1))), (2));
    int64_t* t = (int64_t*)(calloc(((int64_t)(cells)), 8));
    if (t == NULL) {
        return 1;
    }
    void* f = (void*)(fopen("data/p018.txt", "r"));
    if (f == NULL) {
        printf("failed to read data/p018.txt\n");
        free(t);
        return 1;
    }
    int32_t idx = 0;
    while (idx < cells) {
        int32_t v = read_int_ptr_void(f);
        if (v < 0) {
            break;
        }
        t[idx] = ((int64_t)(v));
        idx = (idx + 1);
    }
    fclose(f);
    int32_t r = (rows - 2);
    while (r >= 0) {
        int32_t base = FLOW_CHECKED_DIV(((r * (r + 1))), (2));
        int32_t below = FLOW_CHECKED_DIV((((r + 1) * (r + 2))), (2));
        int32_t c = 0;
        while (c <= r) {
            t[(base + c)] = (t[(base + c)] + max2_i64_i64(t[(below + c)], t[((below + c) + 1)]));
            c = (c + 1);
        }
        r = (r - 1);
    }
    printf("%lld\n", t[0]);
    free(t);
    return 0;
}

Generated MLIR

module {
  llvm.func @printf(!llvm.ptr, ...) -> i32
  llvm.mlir.global internal constant @str_0("data/p018.txt\00") {addr_space = 0 : i32} : !llvm.array<14 x i8>
  llvm.mlir.global internal constant @str_1("r\00") {addr_space = 0 : i32} : !llvm.array<2 x i8>
  llvm.mlir.global internal constant @str_2("failed to read data/p018.txt\n\00") {addr_space = 0 : i32} : !llvm.array<30 x i8>
  llvm.mlir.global internal constant @str_3("%lld\n\00") {addr_space = 0 : i32} : !llvm.array<6 x i8>
  func.func private @fopen(!llvm.ptr, !llvm.ptr) -> !llvm.ptr
  func.func private @fgetc(!llvm.ptr) -> i32
  func.func private @fclose(!llvm.ptr) -> i32
  func.func private @calloc(i64, i64) -> !llvm.ptr
  func.func private @free(!llvm.ptr) -> ()
  func.func @read_int(%arg0: !llvm.ptr) -> i32 {
    %0 = func.call @fgetc(%arg0) : (!llvm.ptr) -> i32
    %1 = llvm.mlir.constant(1 : i64) : i64
    %2 = llvm.alloca %1 x i32 : (i64) -> !llvm.ptr
    llvm.store %0, %2 : i32, !llvm.ptr
    cf.br ^bb0
    ^bb0:
    %3 = llvm.load %2 : !llvm.ptr -> i32
    %4 = arith.constant 0 : i32
    %5 = arith.cmpi sge, %3, %4 : i32
    %6 = scf.if %5 -> (i1) {
      %7 = llvm.load %2 : !llvm.ptr -> i32
      %8 = arith.constant 48 : i32
      %9 = arith.cmpi slt, %7, %8 : i32
      %10 = scf.if %9 -> (i1) {
        %11 = arith.constant true
        scf.yield %11 : i1
      } else {
        %12 = llvm.load %2 : !llvm.ptr -> i32
        %13 = arith.constant 57 : i32
        %14 = arith.cmpi sgt, %12, %13 : i32
        scf.yield %14 : i1
      }
      scf.yield %10 : i1
    } else {
      %15 = arith.constant false
      scf.yield %15 : i1
    }
    cf.cond_br %6, ^bb1, ^bb2
    ^bb1:
      %16 = func.call @fgetc(%arg0) : (!llvm.ptr) -> i32
      llvm.store %16, %2 : i32, !llvm.ptr
      cf.br ^bb0
    ^bb2:
    %17 = llvm.load %2 : !llvm.ptr -> i32
    %18 = arith.constant 0 : i32
    %19 = arith.cmpi slt, %17, %18 : i32
    cf.cond_br %19, ^bb3, ^bb4
    ^bb3:
      %20 = arith.constant 1 : i32
      %22 = arith.constant 0 : i32
      %21 = arith.subi %22, %20 : i32
      func.return %21 : i32
    ^bb4:
      cf.br ^bb5
    ^bb5:
    %23 = arith.constant 0 : i32
    %24 = llvm.mlir.constant(1 : i64) : i64
    %25 = llvm.alloca %24 x i32 : (i64) -> !llvm.ptr
    llvm.store %23, %25 : i32, !llvm.ptr
    cf.br ^bb6
    ^bb6:
    %26 = llvm.load %2 : !llvm.ptr -> i32
    %27 = arith.constant 48 : i32
    %28 = arith.cmpi sge, %26, %27 : i32
    %29 = scf.if %28 -> (i1) {
      %30 = llvm.load %2 : !llvm.ptr -> i32
      %31 = arith.constant 57 : i32
      %32 = arith.cmpi sle, %30, %31 : i32
      scf.yield %32 : i1
    } else {
      %33 = arith.constant false
      scf.yield %33 : i1
    }
    cf.cond_br %29, ^bb7, ^bb8
    ^bb7:
      %34 = llvm.load %25 : !llvm.ptr -> i32
      %35 = arith.constant 10 : i32
      %36 = arith.muli %34, %35 : i32
      %37 = llvm.load %2 : !llvm.ptr -> i32
      %38 = arith.constant 48 : i32
      %39 = arith.subi %37, %38 : i32
      %40 = arith.addi %36, %39 : i32
      llvm.store %40, %25 : i32, !llvm.ptr
      %41 = func.call @fgetc(%arg0) : (!llvm.ptr) -> i32
      llvm.store %41, %2 : i32, !llvm.ptr
      cf.br ^bb6
    ^bb8:
    %42 = llvm.load %25 : !llvm.ptr -> i32
    func.return %42 : i32
  }
  func.func @max2(%arg0: i64, %arg1: i64) -> i64 {
    %43 = arith.cmpi sgt, %arg0, %arg1 : i64
    cf.cond_br %43, ^bb9, ^bb10
    ^bb9:
      func.return %arg0 : i64
    ^bb10:
      cf.br ^bb11
    ^bb11:
    func.return %arg1 : i64
  }
  func.func @main() -> i32 {
    %44 = arith.constant 15 : i32
    %45 = arith.constant 1 : i32
    %46 = arith.addi %44, %45 : i32
    %47 = arith.muli %44, %46 : i32
    %48 = arith.constant 2 : i32
    %49 = arith.divsi %47, %48 : i32
    %51 = arith.extsi %49 : i32 to i64
    %52 = arith.constant 8 : i32
    %53 = arith.extsi %52 : i32 to i64
    %50 = func.call @calloc(%51, %53) : (i64, i64) -> !llvm.ptr
    %54 = llvm.mlir.zero : !llvm.ptr
    %55 = llvm.icmp "eq" %50, %54 : !llvm.ptr
    cf.cond_br %55, ^bb12, ^bb13
    ^bb12:
      %56 = arith.constant 1 : i32
      func.return %56 : i32
    ^bb13:
      cf.br ^bb14
    ^bb14:
    %58 = llvm.mlir.addressof @str_0 : !llvm.ptr
    %59 = llvm.mlir.addressof @str_1 : !llvm.ptr
    %57 = func.call @fopen(%58, %59) : (!llvm.ptr, !llvm.ptr) -> !llvm.ptr
    %60 = llvm.mlir.zero : !llvm.ptr
    %61 = llvm.icmp "eq" %57, %60 : !llvm.ptr
    cf.cond_br %61, ^bb15, ^bb16
    ^bb15:
      %62 = llvm.mlir.addressof @str_2 : !llvm.ptr
      %63 = llvm.call @printf(%62) vararg(!llvm.func<i32 (ptr, ...)>) : (!llvm.ptr) -> i32
      func.call @free(%50) : (!llvm.ptr) -> ()
      %65 = arith.constant 1 : i32
      func.return %65 : i32
    ^bb16:
      cf.br ^bb17
    ^bb17:
    %66 = arith.constant 0 : i32
    %67 = llvm.mlir.constant(1 : i64) : i64
    %68 = llvm.alloca %67 x i32 : (i64) -> !llvm.ptr
    llvm.store %66, %68 : i32, !llvm.ptr
    cf.br ^bb18
    ^bb18:
    %69 = llvm.load %68 : !llvm.ptr -> i32
    %70 = arith.cmpi slt, %69, %49 : i32
    cf.cond_br %70, ^bb19, ^bb20
    ^bb19:
      %71 = func.call @read_int(%57) : (!llvm.ptr) -> i32
      %72 = arith.constant 0 : i32
      %73 = arith.cmpi slt, %71, %72 : i32
      cf.cond_br %73, ^bb21, ^bb22
      ^bb21:
        cf.br ^bb20
      ^bb22:
        cf.br ^bb23
      ^bb23:
      %74 = arith.extsi %71 : i32 to i64
      %75 = llvm.load %68 : !llvm.ptr -> i32
      %76 = arith.extsi %75 : i32 to i64
      %77 = llvm.getelementptr %50[%76] : (!llvm.ptr, i64) -> !llvm.ptr, i64
      llvm.store %74, %77 : i64, !llvm.ptr
      %78 = llvm.load %68 : !llvm.ptr -> i32
      %79 = arith.constant 1 : i32
      %80 = arith.addi %78, %79 : i32
      llvm.store %80, %68 : i32, !llvm.ptr
      cf.br ^bb18
    ^bb20:
    %81 = func.call @fclose(%57) : (!llvm.ptr) -> i32
    %82 = arith.constant 2 : i32
    %83 = arith.subi %44, %82 : i32
    %84 = llvm.mlir.constant(1 : i64) : i64
    %85 = llvm.alloca %84 x i32 : (i64) -> !llvm.ptr
    llvm.store %83, %85 : i32, !llvm.ptr
    cf.br ^bb24
    ^bb24:
    %86 = llvm.load %85 : !llvm.ptr -> i32
    %87 = arith.constant 0 : i32
    %88 = arith.cmpi sge, %86, %87 : i32
    cf.cond_br %88, ^bb25, ^bb26
    ^bb25:
      %89 = llvm.load %85 : !llvm.ptr -> i32
      %90 = llvm.load %85 : !llvm.ptr -> i32
      %91 = arith.constant 1 : i32
      %92 = arith.addi %90, %91 : i32
      %93 = arith.muli %89, %92 : i32
      %94 = arith.constant 2 : i32
      %95 = arith.divsi %93, %94 : i32
      %96 = llvm.load %85 : !llvm.ptr -> i32
      %97 = arith.constant 1 : i32
      %98 = arith.addi %96, %97 : i32
      %99 = llvm.load %85 : !llvm.ptr -> i32
      %100 = arith.constant 2 : i32
      %101 = arith.addi %99, %100 : i32
      %102 = arith.muli %98, %101 : i32
      %103 = arith.constant 2 : i32
      %104 = arith.divsi %102, %103 : i32
      %105 = arith.constant 0 : i32
      %106 = llvm.mlir.constant(1 : i64) : i64
      %107 = llvm.alloca %106 x i32 : (i64) -> !llvm.ptr
      llvm.store %105, %107 : i32, !llvm.ptr
      cf.br ^bb27
      ^bb27:
      %108 = llvm.load %107 : !llvm.ptr -> i32
      %109 = llvm.load %85 : !llvm.ptr -> i32
      %110 = arith.cmpi sle, %108, %109 : i32
      cf.cond_br %110, ^bb28, ^bb29
      ^bb28:
        %112 = llvm.load %107 : !llvm.ptr -> i32
        %113 = arith.addi %95, %112 : i32
        %114 = arith.extsi %113 : i32 to i64
        %115 = llvm.getelementptr %50[%114] : (!llvm.ptr, i64) -> !llvm.ptr, i64
        %111 = llvm.load %115 : !llvm.ptr -> i64
        %118 = llvm.load %107 : !llvm.ptr -> i32
        %119 = arith.addi %104, %118 : i32
        %120 = arith.extsi %119 : i32 to i64
        %121 = llvm.getelementptr %50[%120] : (!llvm.ptr, i64) -> !llvm.ptr, i64
        %117 = llvm.load %121 : !llvm.ptr -> i64
        %123 = llvm.load %107 : !llvm.ptr -> i32
        %124 = arith.addi %104, %123 : i32
        %125 = arith.constant 1 : i32
        %126 = arith.addi %124, %125 : i32
        %127 = arith.extsi %126 : i32 to i64
        %128 = llvm.getelementptr %50[%127] : (!llvm.ptr, i64) -> !llvm.ptr, i64
        %122 = llvm.load %128 : !llvm.ptr -> i64
        %116 = func.call @max2(%117, %122) : (i64, i64) -> i64
        %129 = arith.addi %111, %116 : i64
        %130 = llvm.load %107 : !llvm.ptr -> i32
        %131 = arith.addi %95, %130 : i32
        %132 = arith.extsi %131 : i32 to i64
        %133 = llvm.getelementptr %50[%132] : (!llvm.ptr, i64) -> !llvm.ptr, i64
        llvm.store %129, %133 : i64, !llvm.ptr
        %134 = llvm.load %107 : !llvm.ptr -> i32
        %135 = arith.constant 1 : i32
        %136 = arith.addi %134, %135 : i32
        llvm.store %136, %107 : i32, !llvm.ptr
        cf.br ^bb27
      ^bb29:
      %137 = llvm.load %85 : !llvm.ptr -> i32
      %138 = arith.constant 1 : i32
      %139 = arith.subi %137, %138 : i32
      llvm.store %139, %85 : i32, !llvm.ptr
      cf.br ^bb24
    ^bb26:
    %140 = llvm.mlir.addressof @str_3 : !llvm.ptr
    %142 = arith.constant 0 : i32
    %143 = arith.extsi %142 : i32 to i64
    %144 = llvm.getelementptr %50[%143] : (!llvm.ptr, i64) -> !llvm.ptr, i64
    %141 = llvm.load %144 : !llvm.ptr -> i64
    %145 = llvm.call @printf(%140, %141) vararg(!llvm.func<i32 (ptr, ...)>) : (!llvm.ptr, i64) -> i32
    func.call @free(%50) : (!llvm.ptr) -> ()
    %147 = arith.constant 0 : i32
    func.return %147 : i32
  }
}