ported memaccess

This commit is contained in:
2026-04-09 17:03:39 +01:00
parent 746dce577f
commit 365a35e61b
6 changed files with 684 additions and 0 deletions

View File

@@ -0,0 +1,45 @@
scp ../../encoding.h home-1:/home/akilan/
scp ../../entry.S home-1:/home/akilan/
scp ../../link.ld home-1:/home/akilan/
scp ../../riscv_test_p.h home-1:/home/akilan/
scp ../../riscv_test.h home-1:/home/akilan/
scp ../../string.c home-1:/home/akilan/
scp ../../vm.c home-1:/home/akilan/
scp ../../test.S home-1:/home/akilan/
scp ../../test_macros.h home-1:/home/akilan/
scp ../../malloc.c home-1:/home/akilan/
# scp ../../test.c home-1:/home/akilan/
scp glibc.c home-1:/home/akilan/
# GCC compile
# ssh home-1 'cd /home/akilan/ && export PATH=/opt/riscv/bin:$PATH && riscv64-unknown-elf-gcc -DENTROPY=0xf21e02b -mcmodel=medany -nostdlib -nostartfiles -ffreestanding -fno-builtin -mabi=lp64 -T link.ld vm.c string.c entry.S test.S -o testC'
# Clang compile (architecture Regular RISCV)
ssh home-1 'clang --target=riscv64-unknown-elf -DENTROPY=0xf21e02b -DDEFINE_MALLOC -DDEFINE_FREE --gcc-toolchain=/opt/riscv -mcmodel=medany -nostdlib -nostartfiles -ffreestanding -fno-builtin -fno-builtin-malloc -mabi=lp64 -T link.ld vm.c string.c malloc.c entry.S test.S glibc.c -o testC'
# Clang using the cheri clang compiler
# ssh home-1 'cd cheri/output/sdk/bin/ && ./clang --target=riscv64-unknown-freebsd -DENTROPY=0xf21e02b --sysroot=cheribsd-riscv64-for-purecap-rootfs.cfg -mcmodel=medany -nostdlib -nostartfiles -ffreestanding -fno-builtin -march=rv64imafdcxcheri -mabi=lp64d -T link.ld vm.c string.c entry.S test.S -o testC'
scp home-1:/home/akilan/testC ../../../
# scp home-1:/home/akilan/cheri/output/sdk/bin/testC .
# riscv64-unknown-elf-gcc \
# -march=rv64imac \
# -mabi=lp64 \
# -nostartfiles \
# -fno-builtin \
# -T link.ld \
# entry.S vm.c string.c \
# clang \
# --target=riscv64-unknown-elf \
# -march=rv64imac \
# -mabi=lp64 \
# -nostdlib \
# -nostartfiles \
# -fuse-ld=lld \
# -T link.ld \
# entry.S vm.c string.c \
# -I. \
# -o output.elf%

View File

@@ -0,0 +1,229 @@
/* Benchmark malloc and free functions.
Copyright (C) 2019-2021 Free Software Foundation, Inc.
This file is part of the GNU C Library.
The GNU C Library is free software; you can redistribute it and/or
modify it under the terms of the GNU Lesser General Public
License as published by the Free Software Foundation; either
version 2.1 of the License, or (at your option) any later version.
The GNU C Library is distributed in the hope that it will be useful,
but WITHOUT ANY WARRANTY; without even the implied warranty of
MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU
Lesser General Public License for more details.
You should have received a copy of the GNU Lesser General Public
License along with the GNU C Library; if not, see
<https://www.gnu.org/licenses/>. */
// modified by Daan Leijen to fit the bench suite and add lifo/fifo free order.
// #include <pthread.h>
// #include <stdio.h>
// #include <stdlib.h>
// #include <sys/resource.h>
// #include "malloc.h"
// #include <cheriintrin.h>
#include <stdint.h>
#include <stddef.h>
// #define malloc MALLOCCHERI
// #define free FREECHERI
// #include "bench-timing.h"
// #include "json-lib.h"
// void free(void * __capability ptr);
// void * __capability malloc(size_t size);
void* malloc(size_t size);
void free(void *ptr);
/* Benchmark the malloc/free performance of a varying number of blocks of a
given size. This enables performance tracking of the t-cache and fastbins.
It tests 3 different scenarios: single-threaded using main arena,
multi-threaded using thread-arena, and main arena with SINGLE_THREAD_P
false. */
// source: https://github.com/daanx/mimalloc-bench/blob/master/bench/glibc-bench/bench-malloc-thread.c
#define NUM_ITERS 20
#define NUM_ALLOCS 4
#define MAX_ALLOCS 1600
// Daan: disable timing
typedef long timing_t;
#define TIMING_NOW(s)
#define TIMING_DIFF(e,start,stop)
typedef struct
{
size_t iters;
size_t size;
int n;
timing_t elapsed;
} malloc_args;
static void
do_benchmark(malloc_args *args, char **arr)
{
// timing_t start, stop;
size_t iters = args->iters;
size_t size = args->size;
int n = args->n;
// TIMING_NOW (start);
for (int j = 0; j < iters; j++)
{
for (int i = 0; i < n; i++) {
arr[i] = malloc (size);
for(int g = 0; g < size; g++) { arr[i][g] =(char)g; }
}
// free half in fifo order
for (int i = 0; i < n/2; i++) {
free (arr[i]);
arr[i] = NULL;
}
// and the other half in lifo order
for(int i = n-1; i >= n/2; i--) {
free(arr[i]);
arr[i] = NULL;
}
}
// TIMING_NOW (stop);
// TIMING_DIFF (args->elapsed, start, stop);
}
static malloc_args tests[3][NUM_ALLOCS];
static int allocs[NUM_ALLOCS] = { 25, 100, 400, MAX_ALLOCS };
// static void *
// thread_test (void *p)
// {
// char **arr = (char**)p;
// /* Run benchmark multi-threaded. */
// for (int i = 0; i < NUM_ALLOCS; i++)
// do_benchmark (&tests[2][i], arr);
// return p;
// }
void
bench (unsigned long size)
{
size_t iters = NUM_ITERS;
char**arr = (char**)malloc (MAX_ALLOCS * sizeof (void*));
//char * __capability * __capability arr = malloc(MAX_ALLOCS * sizeof(char * __capability));
for (int t = 0; t < 3; t++)
for (int i = 0; i < NUM_ALLOCS; i++)
{
tests[t][i].n = allocs[i];
tests[t][i].size = size;
tests[t][i].iters = iters / allocs[i];
/* Do a quick warmup run. */
if (t == 0)
do_benchmark (&tests[0][i], arr);
}
/* Run benchmark single threaded in main_arena. */
for (int i = 0; i < NUM_ALLOCS; i++)
do_benchmark (&tests[0][i], arr);
/* Run benchmark in a thread_arena. */
// pthread_t t;
// pthread_create (&t, NULL, thread_test, (void*)arr);
// pthread_join (t, NULL);
/* Repeat benchmark in main_arena with SINGLE_THREAD_P == false. */
// for (int i = 0; i < NUM_ALLOCS; i++)
// do_benchmark (&tests[1][i], arr);
free (arr);
arr = NULL;
/*
json_ctx_t json_ctx;
json_init (&json_ctx, 0, stdout);
json_document_begin (&json_ctx);
json_attr_string (&json_ctx, "timing_type", TIMING_TYPE);
json_attr_object_begin (&json_ctx, "functions");
json_attr_object_begin (&json_ctx, "malloc");
char s[100];
double iters2 = iters;
json_attr_object_begin (&json_ctx, "");
json_attr_double (&json_ctx, "malloc_block_size", size);
struct rusage usage;
getrusage (RUSAGE_SELF, &usage);
json_attr_double (&json_ctx, "max_rss", usage.ru_maxrss);
for (int i = 0; i < NUM_ALLOCS; i++)
{
sprintf (s, "main_arena_st_allocs_%04d_time", allocs[i]);
json_attr_double (&json_ctx, s, tests[0][i].elapsed / iters2);
}
for (int i = 0; i < NUM_ALLOCS; i++)
{
sprintf (s, "main_arena_mt_allocs_%04d_time", allocs[i]);
json_attr_double (&json_ctx, s, tests[1][i].elapsed / iters2);
}
for (int i = 0; i < NUM_ALLOCS; i++)
{
sprintf (s, "thread_arena__allocs_%04d_time", allocs[i]);
json_attr_double (&json_ctx, s, tests[2][i].elapsed / iters2);
}
json_attr_object_end (&json_ctx);
json_attr_object_end (&json_ctx);
json_attr_object_end (&json_ctx);
json_document_end (&json_ctx);
*/
}
// static void usage (const char *name)
// {
// fprintf (stderr, "%s: <alloc_size>\n", name);
// exit (1);
// }
int main (void)
{
// INITREGULARALLOC();
long size = 3;
// if (argc == 2)
// size = strtol (argv[1], NULL, 0);
// if (argc > 2 || size <= 0)
// usage (argv[0]);
// bench (size);
// bench (2*size);
// bench (4*size);
// bench (8*size);
// bench (16*size);
// bench (32*size);
return 0;
}

View File

@@ -0,0 +1,45 @@
scp ../../encoding.h home-1:/home/akilan/
scp ../../entry.S home-1:/home/akilan/
scp ../../link.ld home-1:/home/akilan/
scp ../../riscv_test_p.h home-1:/home/akilan/
scp ../../riscv_test.h home-1:/home/akilan/
scp ../../string.c home-1:/home/akilan/
scp ../../vm.c home-1:/home/akilan/
scp ../../test.S home-1:/home/akilan/
scp ../../test_macros.h home-1:/home/akilan/
scp ../../malloc.c home-1:/home/akilan/
# scp ../../test.c home-1:/home/akilan/
scp memaccess.c home-1:/home/akilan/
# GCC compile
# ssh home-1 'cd /home/akilan/ && export PATH=/opt/riscv/bin:$PATH && riscv64-unknown-elf-gcc -DENTROPY=0xf21e02b -mcmodel=medany -nostdlib -nostartfiles -ffreestanding -fno-builtin -mabi=lp64 -T link.ld vm.c string.c entry.S test.S -o testC'
# Clang compile (architecture Regular RISCV)
ssh home-1 'clang --target=riscv64-unknown-elf -DENTROPY=0xf21e02b -DDEFINE_MALLOC -DDEFINE_FREE --gcc-toolchain=/opt/riscv -mcmodel=medany -nostdlib -nostartfiles -ffreestanding -fno-builtin -fno-builtin-malloc -mabi=lp64 -T link.ld vm.c string.c malloc.c entry.S test.S memaccess.c -o testC'
# Clang using the cheri clang compiler
# ssh home-1 'cd cheri/output/sdk/bin/ && ./clang --target=riscv64-unknown-freebsd -DENTROPY=0xf21e02b --sysroot=cheribsd-riscv64-for-purecap-rootfs.cfg -mcmodel=medany -nostdlib -nostartfiles -ffreestanding -fno-builtin -march=rv64imafdcxcheri -mabi=lp64d -T link.ld vm.c string.c entry.S test.S -o testC'
scp home-1:/home/akilan/testC ../../../
# scp home-1:/home/akilan/cheri/output/sdk/bin/testC .
# riscv64-unknown-elf-gcc \
# -march=rv64imac \
# -mabi=lp64 \
# -nostartfiles \
# -fno-builtin \
# -T link.ld \
# entry.S vm.c string.c \
# clang \
# --target=riscv64-unknown-elf \
# -march=rv64imac \
# -mabi=lp64 \
# -nostdlib \
# -nostartfiles \
# -fuse-ld=lld \
# -T link.ld \
# entry.S vm.c string.c \
# -I. \
# -o output.elf%

View File

@@ -0,0 +1,364 @@
/*
memaccesstest.c
Author Alex Bordei at bigstep
All rights reserved.
*/
// source: https://github.com/bigstepinc/memaccesstest/tree/master
// #include <stdio.h>
// #include <cheriintrin.h>
#include <stddef.h>
#include <stdint.h>
// #include <stdlib.h>
// #include <pthread.h>
// #include "config.h"
// #ifndef __MACH__
// #include <time.h>
// #else
// #include <mach/mach_time.h>
// #endif
// #include "malloc.h"
// #define malloc MALLOCCHERI
// #define free FREECHERI
#define NPAD 7
#define MIN_WSS sizeof(struct l)
#define MAX_WSS 10
void free(void *ptr);
void *malloc(size_t size);
// struct l{
// struct l *n;
// struct l *p;
// long int pad[NPAD-1];
// };
struct l{
struct l *n;
struct l *p;
long int pad[NPAD-1];
};
// uint64_t rand64()
// {
// // Combin 4 parts of low 16-bit of each rand()
// uint64_t R0 = (uint64_t)random() << 48;
// uint64_t R1 = (uint64_t)random() << 48 >> 16;
// uint64_t R2 = (uint64_t)random() << 48 >> 32;
// uint64_t R3 = (uint64_t)random() << 48 >> 48;
// return R0 | R1 | R2 | R3;
// };
#ifdef DEBUG
// void print_ll_lin(struct l *root,unsigned long working_set_size)
// {
// struct l *current=root->n;
// unsigned long count;
// printf("0");
// count=0;
// while(current!=root)
// {
// // printf("->%ld",current-root);
// current=current->n;
// count++;
// if(count>working_set_size/sizeof(struct l))
// {
// // printf("Error:cycle!");
// break;
// }
// }
// // printf("\n");
// }
// void print_ll_lin_back(struct l *root,unsigned long working_set_size)
// {
// struct l *current=root->p;
// unsigned long count;
// printf("0");
// count=0;
// while(current!=root)
// {
// // printf("->%ld",current-root);
// current=current->p;
// count++;
// if(count>working_set_size/sizeof(struct l))
// {
// // printf("Error:cycle!");
// break;
// }
// }
// printf("\n");
// }
// void print_ll(struct l *root,unsigned long working_set_size)
// {
// unsigned long i;
// for(i=0;i<(working_set_size/sizeof(struct l));i++)
// printf("DEBUG: %ld<-%ld->%ld\n",(root+i)->p-root,i,((root+i)->n)-root);
// printf("DEBUG ");
// print_ll_lin(root, working_set_size);
// printf("DEBUG ");
// print_ll_lin_back(root, working_set_size);
// }
#endif
// void build_sequencial_ll(struct l * root, long working_set_size)
void build_sequencial_ll(struct l *root, long working_set_size)
{
unsigned long i;
struct l *current;
for(i=0;i<(working_set_size/sizeof(struct l));i++)
{
current=root+i;
if(i==(working_set_size/sizeof(struct l)-1))
current->n=root;
else
current->n=current+1;
if(i==0)
current->p=root+working_set_size/sizeof(struct l)-1;
else
current->p=current-1;
}
}
// void build_random_ll(struct l *root, long working_set_size)
// {
// unsigned long i;
// struct l *a;
// struct l *b;
// struct l *oldan;
// struct l *oldbn;
// struct l *oldap;
// struct l *oldbp;
// unsigned long rnd;
// unsigned long tries;
// build_sequencial_ll(root, working_set_size);
// //construct the array of elements
// //link the list to the next element in the array
// for(i=1;i<(working_set_size/sizeof(struct l)-1);i++)
// {
// a=(root+i);
// b=a;
// tries=5;
// while((b==a || b==root) && tries)
// {
// rnd=rand64() % (working_set_size/sizeof(struct l)-1);
// b=root+rnd;
// tries--;
// }
// if(!tries)
// continue;
// oldan=a->n; //a->n is lost along the way
// oldbn=b->n; //b->n is lost along the way
// oldap=a->p; //a->p is lost along the way
// oldbp=b->p; //b->p is lost along the way
// a->p->n=b;
// b->p->n=a;
// if(b!=oldan)
// {
// b->n=oldan;
// oldan->p=b;
// }
// else
// {
// b->n=a;
// a->p=b;
// }
// if(a!=oldbn)
// {
// a->n=oldbn;
// oldbn->p=a;
// }
// else
// {
// a->n=b;
// b->p=a;
// }
// if(b!=oldap)
// b->p=oldap;
// else
// b->p=a;
// if(a!=oldbp)
// a->p=oldbp;
// else
// a->p=b;
// #ifdef DEBUG
// printf("DEBUG: Round i=%ld\n",i);
// printf("DEBUG: A=%ld B=%ld\n",a-root,b-root);
// print_ll(root, working_set_size);
// #endif
// }
// }
struct thread_start_data
{
struct l *root;
unsigned long working_set_size;
int thread_index;
};
void * walk(void *param)
{
struct thread_start_data *data=param;
unsigned long start,end;
struct l *root=data->root;
// struct l* current=root->n;
struct l *current = root->n;
// #ifdef __MACH__
// start=mach_absolute_time();
// #else
// struct timespec ts;
// clock_gettime(CLOCK_MONOTONIC_RAW,&ts);
// start=ts.tv_nsec;
// #endif
/////////////////////////////////////
//traverse the linked list
while(current!=root)
{
// #ifdef ADDTEST
// current->pad[0]+=current->n->pad[0];
// #endif
// #ifdef INCTEST
// current->pad[0]++;
// #endif
current=current->n;
}
/////////////////////////////////////
// #ifdef __MACH__
// end=mach_absolute_time();
// #else
// clock_gettime(CLOCK_MONOTONIC_RAW,&ts);
// end=ts.tv_nsec;
// #endif
// printf("%d,%ld,%ld,%ld\n",data->thread_index,data->working_set_size, end-start, data->working_set_size/sizeof(struct l));
// free(data);
return NULL;
}
int main(void)
{
// clock_t tic = clock();
// INITREGULARALLOC();
//#ifdef DEBUG
// printf("DEBUG:sizeof(struct l)=%ld\n",sizeof(struct l));
// #endif
struct l *root;
unsigned long working_set_size;
int i;
// struct thread_start_data *tdata;
struct thread_start_data *tdata;
// pthread_t threads[NTHREADS];
root=malloc(MAX_WSS/sizeof(struct l) * sizeof(struct l));
for(working_set_size=MIN_WSS;working_set_size<MAX_WSS;working_set_size=working_set_size<<1)
{
// #ifdef DEBUG
// printf("DEBUG:Working set size=%ld elements=%ld\n",working_set_size,working_set_size/sizeof(struct l));
// #endif
// clock_t toc = clock();
// printf("Elapsed build get here before ll: %f seconds\n", (double)(toc - tic) / CLOCKS_PER_SEC);
// #ifdef RANDOM_LIST
// build_random_ll(root, working_set_size);
// #else
// clock_t tic1 = clock();
build_sequencial_ll(root, working_set_size);
// build_sequencial_ll(root, working_set_size);
// clock_t toc1 = clock();
// printf("Elapsed build ll: %f seconds\n", (double)(toc1 - tic1) / CLOCKS_PER_SEC);
// #endif
// #if NTHREADS>0
// for(i=0;i<NTHREADS;i++)
// {
// tdata= malloc(sizeof(struct thread_start_data));
// tdata->thread_index=i;
// tdata->root=root;
// tdata->working_set_size=working_set_size;
// pthread_create(&threads[i], NULL, walk, (void*) tdata);
// }
// for(i=0;i<NTHREADS;i++)
// pthread_join(threads[i],NULL);
// #else
tdata = malloc(sizeof(struct thread_start_data));
tdata->thread_index=0;
tdata->root=root;
tdata->working_set_size=working_set_size;
// clock_t tic2 = clock();
for (int i = 0; i < 64; i++) {
walk(tdata);
}
// tdata = malloc(sizeof(struct thread_start_data));
// tdata->thread_index=0;
// tdata->root=root;
// tdata->working_set_size=working_set_size;
// // clock_t tic2 = clock();
// for (int i = 0; i < 64; i++) {
// walk(tdata);
// }
// clock_t toc2 = clock();
// printf("Elapsed build walk: %f seconds\n", (double)(toc2 - tic2) / CLOCKS_PER_SEC);
// #endif
}
// printf("-----------------------------------------------");
free(root);
root = NULL;
return 0;
}

Binary file not shown.

View File

@@ -0,0 +1 @@
*.txt