Speed things up by ignoring return values.

The remote bitbang implementation is really slow when reading back data.
During many operations, like writing a block of memory, we don't need to
see the return data. By communicating this to the lower layers, we get
huge speedups. Downloads to spike now are 295KB/s.

That means the gdbserver.py download test now runs to completion,
unfortunately it fails. Everything else is still passing, though, so I'm
committing this change.

Change-Id: I44cc9db1ade0908c1a12d09b23fc8e529a802d88
This commit is contained in:
Tim Newsome 2017-02-22 16:04:36 -08:00
parent f9168b09d0
commit 3bd661b679
1 changed files with 45 additions and 13 deletions

View File

@ -218,17 +218,24 @@ static void dump_field(const struct scan_field *field)
unsigned int out_op = get_field(out, DTM_DMI_OP);
unsigned int out_data = get_field(out, DTM_DMI_DATA);
unsigned int out_address = out >> DTM_DMI_ADDRESS_OFFSET;
uint64_t in = buf_get_u64(field->in_value, 0, field->num_bits);
unsigned int in_op = get_field(in, DTM_DMI_OP);
unsigned int in_data = get_field(in, DTM_DMI_DATA);
unsigned int in_address = in >> DTM_DMI_ADDRESS_OFFSET;
log_printf_lf(LOG_LVL_DEBUG,
__FILE__, __LINE__, "scan",
"%db %s %08x @%02x -> %s %08x @%02x",
field->num_bits,
op_string[out_op], out_data, out_address,
status_string[in_op], in_data, in_address);
if (field->in_value) {
uint64_t in = buf_get_u64(field->in_value, 0, field->num_bits);
unsigned int in_op = get_field(in, DTM_DMI_OP);
unsigned int in_data = get_field(in, DTM_DMI_DATA);
unsigned int in_address = in >> DTM_DMI_ADDRESS_OFFSET;
log_printf_lf(LOG_LVL_DEBUG,
__FILE__, __LINE__, "scan",
"%db %s %08x @%02x -> %s %08x @%02x",
field->num_bits,
op_string[out_op], out_data, out_address,
status_string[in_op], in_data, in_address);
} else {
log_printf_lf(LOG_LVL_DEBUG,
__FILE__, __LINE__, "scan", "%db %s %08x @%02x -> ?",
field->num_bits, op_string[out_op], out_data, out_address);
}
}
static riscv013_info_t *get_info(const struct target *target)
@ -309,7 +316,9 @@ static void scans_add_dmi_write(scans_t *scans, unsigned address,
uint8_t *out = scans->out + data_offset;
field->num_bits = info->abits + DMI_OP_SIZE + DMI_DATA_SIZE;
field->in_value = scans->in + data_offset;
// We gain a lot of speed in remote bitbang by not looking at the return
// value.
field->in_value = NULL;
field->out_value = out;
buf_set_u64(out, DMI_OP_START, DMI_OP_SIZE, DMI_OP_WRITE);
@ -466,9 +475,12 @@ static dmi_status_t dmi_scan(struct target *target, uint16_t *address_in,
struct scan_field field = {
.num_bits = info->abits + DMI_OP_SIZE + DMI_DATA_SIZE,
.out_value = out,
.in_value = in
};
if (address_in || data_in) {
field.in_value = in;
}
assert(info->abits != 0);
buf_set_u64(out, DMI_OP_START, DMI_OP_SIZE, op);
@ -2098,6 +2110,23 @@ static int read_memory(struct target *target, uint32_t address,
return ERROR_OK;
}
/**
* If there was a DMI error, clear that error and return 1.
* Otherwise return 0.
*/
static int check_dmi_error(struct target *target)
{
dmi_status_t status = dmi_scan(target, NULL, NULL, DMI_OP_NOP, 0, 0,
false);
if (status != DMI_STATUS_SUCCESS) {
// Clear errors.
dtmcontrol_scan(target, DTM_DTMCONTROL_DMIRESET);
increase_dmi_busy_delay(target);
return 1;
}
return 0;
}
static int write_memory(struct target *target, uint32_t address,
uint32_t size, uint32_t count, const uint8_t *buffer)
{
@ -2163,6 +2192,9 @@ static int write_memory(struct target *target, uint32_t address,
if (result != ERROR_OK)
return result;
int dmi_error = check_dmi_error(target);
// Clear autoexec.
dmi_write(target, DMI_ABSTRACTCS, DMI_ABSTRACTCS_CMDERR);
uint32_t abstractcs = dmi_read(target, DMI_ABSTRACTCS);
unsigned cmderr = get_field(abstractcs, DMI_ABSTRACTCS_CMDERR);
@ -2172,7 +2204,7 @@ static int write_memory(struct target *target, uint32_t address,
} else if (cmderr) {
LOG_ERROR("cmderr=%d", get_field(abstractcs, DMI_ABSTRACTCS_CMDERR));
return ERROR_FAIL;
} else {
} else if (!dmi_error) {
break;
}
}