mirror of https://gitee.com/openkylin/linux.git
drm/nouveau/pll/gk104: fix PLL instability due to bad configuration with gddr5
This patch uses an approach closer to the nvidia driver to configure both PLLs for high gddr5 memory clocks (usually above 2400MHz) Previously nouveau used the one PLL as it was used for the lower clocks and just adjusted the second PLL to get as close as possible to the requested clock. This means for my card, that I got a 4050 MHz clock although 4008 MHz was requested. Now the driver iterates over a list of PLL configuration also used by the nvidia driver and then adjust the second PLL to get near the requested clock. Also it hold to some restriction I found while analyzing the PLL configurations This won't fix all gddr5 high clock issues itself, but it should be fine on hybrid gpu systems as found on many laptops these days. Also switching while normal desktop usage should be a lot more stable than before. v2: move the pll code into ramgk104 Signed-off-by: Karol Herbst <nouveau@karolherbst.de> Signed-off-by: Ben Skeggs <bskeggs@redhat.com>
This commit is contained in:
parent
0d42743dfa
commit
78eaf335e4
|
@ -970,6 +970,67 @@ gk104_ram_calc_data(struct gk104_ram *ram, u32 khz, struct nvkm_ram_data *data)
|
||||||
return -EINVAL;
|
return -EINVAL;
|
||||||
}
|
}
|
||||||
|
|
||||||
|
static int
|
||||||
|
gk104_calc_pll_output(int fN, int M, int N, int P, int clk)
|
||||||
|
{
|
||||||
|
return ((clk * N) + (((u16)(fN + 4096) * clk) >> 13)) / (M * P);
|
||||||
|
}
|
||||||
|
|
||||||
|
static int
|
||||||
|
gk104_pll_calc_hiclk(int target_khz, int crystal,
|
||||||
|
int *N1, int *fN1, int *M1, int *P1,
|
||||||
|
int *N2, int *M2, int *P2)
|
||||||
|
{
|
||||||
|
int best_clk = 0, best_err = target_khz, p_ref, n_ref;
|
||||||
|
bool upper = false;
|
||||||
|
|
||||||
|
*M1 = 1;
|
||||||
|
/* M has to be 1, otherwise it gets unstable */
|
||||||
|
*M2 = 1;
|
||||||
|
/* can be 1 or 2, sticking with 1 for simplicity */
|
||||||
|
*P2 = 1;
|
||||||
|
|
||||||
|
for (p_ref = 0x7; p_ref >= 0x5; --p_ref) {
|
||||||
|
for (n_ref = 0x25; n_ref <= 0x2b; ++n_ref) {
|
||||||
|
int cur_N, cur_clk, cur_err;
|
||||||
|
|
||||||
|
cur_clk = gk104_calc_pll_output(0, 1, n_ref, p_ref, crystal);
|
||||||
|
cur_N = target_khz / cur_clk;
|
||||||
|
cur_err = target_khz
|
||||||
|
- gk104_calc_pll_output(0xf000, 1, cur_N, 1, cur_clk);
|
||||||
|
|
||||||
|
/* we found a better combination */
|
||||||
|
if (cur_err < best_err) {
|
||||||
|
best_err = cur_err;
|
||||||
|
best_clk = cur_clk;
|
||||||
|
*N2 = cur_N;
|
||||||
|
*N1 = n_ref;
|
||||||
|
*P1 = p_ref;
|
||||||
|
upper = false;
|
||||||
|
}
|
||||||
|
|
||||||
|
cur_N += 1;
|
||||||
|
cur_err = gk104_calc_pll_output(0xf000, 1, cur_N, 1, cur_clk)
|
||||||
|
- target_khz;
|
||||||
|
if (cur_err < best_err) {
|
||||||
|
best_err = cur_err;
|
||||||
|
best_clk = cur_clk;
|
||||||
|
*N2 = cur_N;
|
||||||
|
*N1 = n_ref;
|
||||||
|
*P1 = p_ref;
|
||||||
|
upper = true;
|
||||||
|
}
|
||||||
|
}
|
||||||
|
}
|
||||||
|
|
||||||
|
/* adjust fN to get closer to the target clock */
|
||||||
|
*fN1 = (u16)((((best_err / *N2 * *P2) * (*P1 * *M1)) << 13) / crystal);
|
||||||
|
if (upper)
|
||||||
|
*fN1 = (u16)(1 - *fN1);
|
||||||
|
|
||||||
|
return gk104_calc_pll_output(*fN1, 1, *N1, *P1, crystal);
|
||||||
|
}
|
||||||
|
|
||||||
static int
|
static int
|
||||||
gk104_ram_calc_xits(struct gk104_ram *ram, struct nvkm_ram_data *next)
|
gk104_ram_calc_xits(struct gk104_ram *ram, struct nvkm_ram_data *next)
|
||||||
{
|
{
|
||||||
|
@ -994,31 +1055,24 @@ gk104_ram_calc_xits(struct gk104_ram *ram, struct nvkm_ram_data *next)
|
||||||
* kepler boards, no idea how/why they're chosen.
|
* kepler boards, no idea how/why they're chosen.
|
||||||
*/
|
*/
|
||||||
refclk = next->freq;
|
refclk = next->freq;
|
||||||
if (ram->mode == 2)
|
|
||||||
refclk = fuc->mempll.refclk;
|
|
||||||
|
|
||||||
/* calculate refpll coefficients */
|
|
||||||
ret = gt215_pll_calc(subdev, &fuc->refpll, refclk, &ram->N1,
|
|
||||||
&ram->fN1, &ram->M1, &ram->P1);
|
|
||||||
fuc->mempll.refclk = ret;
|
|
||||||
if (ret <= 0) {
|
|
||||||
nvkm_error(subdev, "unable to calc refpll\n");
|
|
||||||
return -EINVAL;
|
|
||||||
}
|
|
||||||
|
|
||||||
/* calculate mempll coefficients, if we're using it */
|
|
||||||
if (ram->mode == 2) {
|
if (ram->mode == 2) {
|
||||||
/* post-divider doesn't work... the reg takes the values but
|
ret = gk104_pll_calc_hiclk(next->freq, subdev->device->crystal,
|
||||||
* appears to completely ignore it. there *is* a bit at
|
&ram->N1, &ram->fN1, &ram->M1, &ram->P1,
|
||||||
* bit 28 that appears to divide the clock by 2 if set.
|
&ram->N2, &ram->M2, &ram->P2);
|
||||||
*/
|
fuc->mempll.refclk = ret;
|
||||||
fuc->mempll.min_p = 1;
|
|
||||||
fuc->mempll.max_p = 2;
|
|
||||||
|
|
||||||
ret = gt215_pll_calc(subdev, &fuc->mempll, next->freq,
|
|
||||||
&ram->N2, NULL, &ram->M2, &ram->P2);
|
|
||||||
if (ret <= 0) {
|
if (ret <= 0) {
|
||||||
nvkm_error(subdev, "unable to calc mempll\n");
|
nvkm_error(subdev, "unable to calc plls\n");
|
||||||
|
return -EINVAL;
|
||||||
|
}
|
||||||
|
nvkm_debug(subdev, "sucessfully calced PLLs for clock %i kHz"
|
||||||
|
" (refclock: %i kHz)\n", next->freq, ret);
|
||||||
|
} else {
|
||||||
|
/* calculate refpll coefficients */
|
||||||
|
ret = gt215_pll_calc(subdev, &fuc->refpll, refclk, &ram->N1,
|
||||||
|
&ram->fN1, &ram->M1, &ram->P1);
|
||||||
|
fuc->mempll.refclk = ret;
|
||||||
|
if (ret <= 0) {
|
||||||
|
nvkm_error(subdev, "unable to calc refpll\n");
|
||||||
return -EINVAL;
|
return -EINVAL;
|
||||||
}
|
}
|
||||||
}
|
}
|
||||||
|
|
Loading…
Reference in New Issue