Skip to content

rv32imf_xpulpv2 ISA support for Siracusa platform #64

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
Victor-Jung merged 2 commits into from
May 6, 2025
Merged

rv32imf_xpulpv2 ISA support for Siracusa platform #64

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
888f0af
New ISA support for Siracusa platform
runwangdl May 6, 2025
ad34d96
Add CHANGELOG
runwangdl May 6, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 6 additions & 0 deletions CHANGELOG.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -277,3 +277,9 @@ Change main.c to use OUTPUTTYPE instead of float
### Removed
- Remove the link to the precompiled LLVM 12 in the `testRunner` for Snitch and in the CI.
- Remove the sourcing of the cursed PULP SDK script.

## rv32imf_xpulpv2 ISA support for Siracusa platform

### Changed
- The ISA for the Siracusa platform has been updated from rv32imc_zfinx_xpulpv2 to rv32imf_xpulpv2.
- All floating-point comparison tasks in deeploytest.c are now offloaded to Cluster 0 for execution.
288 changes: 174 additions & 114 deletions DeeployTest/Platforms/Siracusa/src/deeploytest.c
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -9,7 +9,8 @@
* Copyright (C) 2020 ETH Zurich and University of Bologna.
*
* Author: Moritz Scherer, ETH Zurich
*
* Author: Run Wang, ETH Zurich
*
* SPDX-License-Identifier: Apache-2.0
*
* Licensed under the Apache License, Version 2.0 (the License); you may
Expand All @@ -25,119 +26,178 @@
* limitations under the License.
*/

#include "CycleCounter.h"
#include "Network.h"
#include "dory_mem.h"
#include "pmsis.h"
#include "testinputs.h"
#include "testoutputs.h"

#define MAINSTACKSIZE 8000
#define SLAVESTACKSIZE 3800

struct pi_device cluster_dev;

void main(void) {
#ifndef CI
printf("HELLO WORLD:\r\n");
#endif
struct pi_cluster_conf conf;

pi_cluster_conf_init(&conf);
conf.id = 0;
pi_open_from_conf(&cluster_dev, &conf);
if (pi_cluster_open(&cluster_dev))
return;

mem_init();
#ifndef NOFLASH
open_fs();
#endif

printf("Intializing\r\n");

struct pi_cluster_task cluster_task;

pi_cluster_task(&cluster_task, InitNetwork, NULL);
cluster_task.stack_size = MAINSTACKSIZE;
cluster_task.slave_stack_size = SLAVESTACKSIZE;
pi_cluster_send_task_to_cl(&cluster_dev, &cluster_task);

#ifndef CI
printf("Initialized\r\n");
#endif
for (int buf = 0; buf < DeeployNetwork_num_inputs; buf++) {
if (DeeployNetwork_inputs[buf] >= 0x10000000) {
memcpy(DeeployNetwork_inputs[buf], testInputVector[buf],
DeeployNetwork_inputs_bytes[buf]);
}
}

#ifndef CI
printf("Input copied\r\n");
#endif
// RunNetwork(0, 1);
pi_cluster_task(&cluster_task, RunNetwork, NULL);
cluster_task.stack_size = MAINSTACKSIZE;
cluster_task.slave_stack_size = SLAVESTACKSIZE;
ResetTimer();
StartTimer();
pi_cluster_send_task_to_cl(&cluster_dev, &cluster_task);
StopTimer();

#ifndef CI
printf("Output:\r\n");
#endif

int32_t tot_err, tot_tested;
tot_err = 0;
tot_tested = 0;
OUTPUTTYPE *compbuf;
OUTPUTTYPE diff, expected, actual;
for (int buf = 0; buf < DeeployNetwork_num_outputs; buf++)
{
tot_tested += DeeployNetwork_outputs_bytes[buf] / sizeof(OUTPUTTYPE);
if (DeeployNetwork_outputs[buf] < 0x1000000)
{
compbuf = pi_l2_malloc(DeeployNetwork_outputs_bytes[buf]);
ram_read(compbuf, DeeployNetwork_outputs[buf],
DeeployNetwork_outputs_bytes[buf]);
}
else
{
compbuf = DeeployNetwork_outputs[buf];
}
for (int i = 0; i < DeeployNetwork_outputs_bytes[buf] / sizeof(OUTPUTTYPE); i++)
{
expected = ((OUTPUTTYPE *)testOutputVector[buf])[i];
actual = ((OUTPUTTYPE *)compbuf)[i];
diff = expected - actual;
if (ISOUTPUTFLOAT)
{
if ((diff < -1e-4) || (diff > 1e-4) || (isnan(diff)))
{
tot_err += 1;
printf("Expected: %10.6f ", expected);
printf("Actual: %10.6f ", actual);
printf("Diff: %10.6f at Index %12u in Output %u\r\n", diff, i, buf);
}
}
else
{
if (diff)
{
tot_err += 1;
printf("Expected: %4d ", expected);
printf("Actual: %4d ", actual);
printf("Diff: %4d at Index %12u in Output %u\r\n", diff, i, buf);
}
}
}
if (DeeployNetwork_outputs[buf] < 0x1000000)
{
pi_l2_free(compbuf, DeeployNetwork_outputs_bytes[buf]);
}
}
#include "CycleCounter.h"
#include "Network.h"
#include "dory_mem.h"
#include "pmsis.h"
#include "testinputs.h"
#include "testoutputs.h"

#define MAINSTACKSIZE 8000
#define SLAVESTACKSIZE 3800

struct pi_device cluster_dev;

typedef struct {
void *expected;
void *actual;
int num_elements;
int output_buf_index;
int *err_count;
} FloatCompareArgs;


void CompareFloatOnCluster(void *args)
{

if(pi_core_id() == 0)
{
FloatCompareArgs *compare_args = (FloatCompareArgs *)args;
float *expected = (float *)compare_args->expected;
float *actual = (float *)compare_args->actual;
int num_elements = compare_args->num_elements;
int output_buf_index = compare_args->output_buf_index;
int *err_count = compare_args->err_count;

int local_err_count = 0;

for (int i = 0; i < num_elements; i++)
{
float expected_val = expected[i];
float actual_val = actual[i];
float diff = expected_val - actual_val;


if ((diff < -1e-4) || (diff > 1e-4) || isnan(diff))
{
local_err_count += 1;

printf("Expected: %10.6f ", expected_val);
printf("Actual: %10.6f ", actual_val);
printf("Diff: %10.6f at Index %12u in Output %u\r\n", diff, i, output_buf_index);
}
}

*err_count = local_err_count;
}
}

void main(void)
{
#ifndef CI
printf("HELLO WORLD:\r\n");
#endif
struct pi_cluster_conf conf;

pi_cluster_conf_init(&conf);
conf.id = 0;
pi_open_from_conf(&cluster_dev, &conf);
if (pi_cluster_open(&cluster_dev))
return;

mem_init();
#ifndef NOFLASH
open_fs();
#endif

printf("Intializing\r\n");

struct pi_cluster_task cluster_task;

pi_cluster_task(&cluster_task, InitNetwork, NULL);
cluster_task.stack_size = MAINSTACKSIZE;
cluster_task.slave_stack_size = SLAVESTACKSIZE;
pi_cluster_send_task_to_cl(&cluster_dev, &cluster_task);

#ifndef CI
printf("Initialized\r\n");
#endif
for (int buf = 0; buf < DeeployNetwork_num_inputs; buf++)
{
if (DeeployNetwork_inputs[buf] >= 0x10000000)
{
memcpy(DeeployNetwork_inputs[buf], testInputVector[buf],
DeeployNetwork_inputs_bytes[buf]);
}
}

#ifndef CI
printf("Input copied\r\n");
#endif
// RunNetwork(0, 1);
pi_cluster_task(&cluster_task, RunNetwork, NULL);
cluster_task.stack_size = MAINSTACKSIZE;
cluster_task.slave_stack_size = SLAVESTACKSIZE;
ResetTimer();
StartTimer();
pi_cluster_send_task_to_cl(&cluster_dev, &cluster_task);
StopTimer();

#ifndef CI
printf("Output:\r\n");
#endif

uint32_t tot_err, tot_tested;
tot_err = 0;
tot_tested = 0;
void *compbuf;
FloatCompareArgs float_compare_args;
uint32_t float_error_count = 0;

for (int buf = 0; buf < DeeployNetwork_num_outputs; buf++)
{
tot_tested += DeeployNetwork_outputs_bytes[buf] / sizeof(OUTPUTTYPE);

if (DeeployNetwork_outputs[buf] < 0x1000000)
{
compbuf = pi_l2_malloc(DeeployNetwork_outputs_bytes[buf]);
ram_read(compbuf, DeeployNetwork_outputs[buf],
DeeployNetwork_outputs_bytes[buf]);
}
else
{
compbuf = DeeployNetwork_outputs[buf];
}

if (ISOUTPUTFLOAT)
{
float_error_count = 0;
float_compare_args.expected = testOutputVector[buf];
float_compare_args.actual = compbuf;
float_compare_args.num_elements = DeeployNetwork_outputs_bytes[buf] / sizeof(float);
float_compare_args.output_buf_index = buf;
float_compare_args.err_count = &float_error_count;

pi_cluster_task(&cluster_task, CompareFloatOnCluster, &float_compare_args);
cluster_task.stack_size = MAINSTACKSIZE;
cluster_task.slave_stack_size = SLAVESTACKSIZE;
pi_cluster_send_task_to_cl(&cluster_dev, &cluster_task);

tot_err += float_error_count;
}
else
{

for (int i = 0; i < DeeployNetwork_outputs_bytes[buf] / sizeof(OUTPUTTYPE); i++)
{
OUTPUTTYPE expected = ((OUTPUTTYPE *)testOutputVector[buf])[i];
OUTPUTTYPE actual = ((OUTPUTTYPE *)compbuf)[i];
OUTPUTTYPE diff = expected - actual;

if (diff)
{
tot_err += 1;
printf("Expected: %4d ", expected);
printf("Actual: %4d ", actual);
printf("Diff: %4d at Index %12u in Output %u\r\n", diff, i, buf);
}
}
}
if (DeeployNetwork_outputs[buf] < 0x1000000)
{
pi_l2_free(compbuf, DeeployNetwork_outputs_bytes[buf]);
}
}

printf("Runtime: %u cycles\r\n", getCycles());
printf("Errors: %u out of %u \r\n", tot_err, tot_tested);
Expand Down
2 changes: 1 addition & 1 deletion cmake/pulp/toolchain_llvm.cmake
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -10,7 +10,7 @@ set(CMAKE_ASM_COMPILER ${TOOLCHAIN_PREFIX}/clang)
set(CMAKE_OBJCOPY ${TOOLCHAIN_PREFIX}/${LLVM_TAG}-objcopy)
set(CMAKE_OBJDUMP ${TOOLCHAIN_PREFIX}/${LLVM_TAG}-objdump)

set(ISA rv32imc_zfinx_xpulpv2)
set(ISA rv32imf_xpulpv2)
set(PE 8)
set(FC 1)

Expand Down
Loading