Roy Spliet
2014-Nov-28 11:25 UTC
[Nouveau] [RESEND PATCH nouveau 3/3] volt: add support for GK20A
Hello Vince, One minor question inline. Op 28-11-14 om 12:13 schreef Vince Hsu:> The voltage value are calculated by the hardware characterized > result. > > Signed-off-by: Vince Hsu <vinceh at nvidia.com> > --- > > Resend this patch with the fuse change and proper patch prefix > per Thierry's request. > > drm/Kbuild | 1 + > drm/core/subdev/volt/gk20a.c | 1 + > nvkm/engine/device/nve0.c | 1 + > nvkm/include/subdev/volt.h | 1 + > nvkm/subdev/clock/gk20a.c | 15 ++++ > nvkm/subdev/volt/gk20a.c | 202 +++++++++++++++++++++++++++++++++++++++++++ > 6 files changed, 221 insertions(+) > create mode 120000 drm/core/subdev/volt/gk20a.c > create mode 100644 nvkm/subdev/volt/gk20a.c > > diff --git a/drm/Kbuild b/drm/Kbuild > index 728bc5b66b29..7c49e6655066 100644 > --- a/drm/Kbuild > +++ b/drm/Kbuild > @@ -225,6 +225,7 @@ nouveau-y += core/subdev/vm/nvc0.o > nouveau-y += core/subdev/volt/base.o > nouveau-y += core/subdev/volt/gpio.o > nouveau-y += core/subdev/volt/nv40.o > +nouveau-y += core/subdev/volt/gk20a.o > > nouveau-y += core/engine/falcon.o > nouveau-y += core/engine/xtensa.o > diff --git a/drm/core/subdev/volt/gk20a.c b/drm/core/subdev/volt/gk20a.c > new file mode 120000 > index 000000000000..2894eb1ede13 > --- /dev/null > +++ b/drm/core/subdev/volt/gk20a.c > @@ -0,0 +1 @@ > +../../../../nvkm/subdev/volt/gk20a.c > \ No newline at end of file > diff --git a/nvkm/engine/device/nve0.c b/nvkm/engine/device/nve0.c > index b1b2e484ecfa..674da1f095b2 100644 > --- a/nvkm/engine/device/nve0.c > +++ b/nvkm/engine/device/nve0.c > @@ -179,6 +179,7 @@ nve0_identify(struct nouveau_device *device) > device->oclass[NVDEV_ENGINE_GR ] = gk20a_graph_oclass; > device->oclass[NVDEV_ENGINE_COPY2 ] = &nve0_copy2_oclass; > device->oclass[NVDEV_ENGINE_PERFMON] = &nve0_perfmon_oclass; > + device->oclass[NVDEV_SUBDEV_VOLT ] = &gk20a_volt_oclass; > break; > case 0xf0: > device->cname = "GK110"; > diff --git a/nvkm/include/subdev/volt.h b/nvkm/include/subdev/volt.h > index 820b62ffd75b..67db5e58880d 100644 > --- a/nvkm/include/subdev/volt.h > +++ b/nvkm/include/subdev/volt.h > @@ -52,6 +52,7 @@ int _nouveau_volt_init(struct nouveau_object *); > #define _nouveau_volt_fini _nouveau_subdev_fini > > extern struct nouveau_oclass nv40_volt_oclass; > +extern struct nouveau_oclass gk20a_volt_oclass; > > int nouveau_voltgpio_init(struct nouveau_volt *); > int nouveau_voltgpio_get(struct nouveau_volt *); > diff --git a/nvkm/subdev/clock/gk20a.c b/nvkm/subdev/clock/gk20a.c > index 82abbea2be12..fb4fad374bdd 100644 > --- a/nvkm/subdev/clock/gk20a.c > +++ b/nvkm/subdev/clock/gk20a.c > @@ -470,76 +470,91 @@ gk20a_pstates[] = { > { > .base = { > .domain[nv_clk_src_gpc] = 72000, > + .voltage = 0, > }, > }, > { > .base = { > .domain[nv_clk_src_gpc] = 108000, > + .voltage = 1, > }, > }, > { > .base = { > .domain[nv_clk_src_gpc] = 180000, > + .voltage = 2, > }, > }, > { > .base = { > .domain[nv_clk_src_gpc] = 252000, > + .voltage = 3, > }, > }, > { > .base = { > .domain[nv_clk_src_gpc] = 324000, > + .voltage = 4, > }, > }, > { > .base = { > .domain[nv_clk_src_gpc] = 396000, > + .voltage = 5, > }, > }, > { > .base = { > .domain[nv_clk_src_gpc] = 468000, > + .voltage = 6, > }, > }, > { > .base = { > .domain[nv_clk_src_gpc] = 540000, > + .voltage = 7, > }, > }, > { > .base = { > .domain[nv_clk_src_gpc] = 612000, > + .voltage = 8, > }, > }, > { > .base = { > .domain[nv_clk_src_gpc] = 648000, > + .voltage = 9, > }, > }, > { > .base = { > .domain[nv_clk_src_gpc] = 684000, > + .voltage = 10, > }, > }, > { > .base = { > .domain[nv_clk_src_gpc] = 708000, > + .voltage = 11, > }, > }, > { > .base = { > .domain[nv_clk_src_gpc] = 756000, > + .voltage = 12, > }, > }, > { > .base = { > .domain[nv_clk_src_gpc] = 804000, > + .voltage = 13, > }, > }, > { > .base = { > .domain[nv_clk_src_gpc] = 852000, > + .voltage = 14, > }, > }, > };Is there a particular reason why this table is hard-coded rather than stored in the device tree? It doesn't seem to differ much between different gk20a's (but this might change with the denver-core version?), but I do anticipate a lot of "code" duplication when post-K1 cores are released and supported.> diff --git a/nvkm/subdev/volt/gk20a.c b/nvkm/subdev/volt/gk20a.c > new file mode 100644 > index 000000000000..87385966e67f > --- /dev/null > +++ b/nvkm/subdev/volt/gk20a.c > @@ -0,0 +1,202 @@ > +/* > + * Copyright (c) 2014, NVIDIA CORPORATION. All rights reserved. > + * > + * Permission is hereby granted, free of charge, to any person obtaining a > + * copy of this software and associated documentation files (the "Software"), > + * to deal in the Software without restriction, including without limitation > + * the rights to use, copy, modify, merge, publish, distribute, sublicense, > + * and/or sell copies of the Software, and to permit persons to whom the > + * Software is furnished to do so, subject to the following conditions: > + * > + * The above copyright notice and this permission notice shall be included in > + * all copies or substantial portions of the Software. > + * > + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR > + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, > + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL > + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER > + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING > + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER > + * DEALINGS IN THE SOFTWARE. > + */ > + > +#include <linux/kernel.h> > +#include <linux/regulator/consumer.h> > +#include <soc/tegra/fuse.h> > + > +#include <nouveau_platform.h> > +#include <subdev/volt.h> > + > +struct cvb_coef { > + int c0; > + int c1; > + int c2; > + int c3; > + int c4; > + int c5; > +}; > + > +struct gk20a_volt_priv { > + struct nouveau_volt base; > + struct regulator *vdd; > +}; > + > +const struct cvb_coef gk20a_cvb_coef[] = { > + /* MHz, c0, c1, c2, c3, c4, c5 */ > + /* 72 */ { 1209886, -36468, 515, 417, -13123, 203}, > + /* 108 */ { 1130804, -27659, 296, 298, -10834, 221}, > + /* 180 */ { 1162871, -27110, 247, 238, -10681, 268}, > + /* 252 */ { 1220458, -28654, 247, 179, -10376, 298}, > + /* 324 */ { 1280953, -30204, 247, 119, -9766, 304}, > + /* 396 */ { 1344547, -31777, 247, 119, -8545, 292}, > + /* 468 */ { 1420168, -34227, 269, 60, -7172, 256}, > + /* 540 */ { 1490757, -35955, 274, 60, -5188, 197}, > + /* 612 */ { 1599112, -42583, 398, 0, -1831, 119}, > + /* 648 */ { 1366986, -16459, -274, 0, -3204, 72}, > + /* 684 */ { 1391884, -17078, -274, -60, -1526, 30}, > + /* 708 */ { 1415522, -17497, -274, -60, -458, 0}, > + /* 756 */ { 1464061, -18331, -274, -119, 1831, -72}, > + /* 804 */ { 1524225, -20064, -254, -119, 4272, -155}, > + /* 852 */ { 1608418, -21643, -269, 0, 763, -48}, > +}; > + > +/** > + * cvb_mv = ((c2 * speedo / s_scale + c1) * speedo / s_scale + c0) > + */ > +static inline int > +gk20a_volt_get_cvb_voltage(int speedo, int s_scale, > + const struct cvb_coef *coef) > +{ > + int mv; > + > + mv = DIV_ROUND_CLOSEST(coef->c2 * speedo, s_scale); > + mv = DIV_ROUND_CLOSEST((mv + coef->c1) * speedo, s_scale) + coef->c0; > + return mv; > +} > + > +/** > + * cvb_t_mv > + * ((c2 * speedo / s_scale + c1) * speedo / s_scale + c0) + > + * ((c3 * speedo / s_scale + c4 + c5 * T / t_scale) * T / t_scale) > + */ > +static inline int > +gk20a_volt_get_cvb_t_voltage(int speedo, int temp, int s_scale, int t_scale, > + const struct cvb_coef *coef) > +{ > + int cvb_mv, mv; > + > + cvb_mv = gk20a_volt_get_cvb_voltage(speedo, s_scale, coef); > + > + mv = DIV_ROUND_CLOSEST(coef->c3 * speedo, s_scale) + coef->c4 + > + DIV_ROUND_CLOSEST(coef->c5 * temp, t_scale); > + mv = DIV_ROUND_CLOSEST(mv * temp, t_scale) + cvb_mv; > + return mv; > +} > + > +static int > +gk20a_volt_calc_voltage(const struct cvb_coef *coef) > +{ > + int speedo, mv; > + > + speedo = tegra_sku_info.gpu_speedo_value; > + > + mv = gk20a_volt_get_cvb_t_voltage(speedo, -10, 100, 10, coef); > + mv = DIV_ROUND_UP(mv, 1000); > + > + return mv * 1000; > +} > + > +static int > +gk20a_volt_vid_get(struct nouveau_volt *volt) > +{ > + struct gk20a_volt_priv *priv = (void *)volt; > + int i, uv; > + > + uv = regulator_get_voltage(priv->vdd); > + > + for (i = 0; i < volt->vid_nr; i++) > + if (volt->vid[i].uv >= uv) > + return i; > + > + return -EINVAL; > +} > + > +static int > +gk20a_volt_vid_set(struct nouveau_volt *volt, u8 vid) > +{ > + struct gk20a_volt_priv *priv = (void *)volt; > + > + nv_debug(volt, "set voltage as %duv\n", volt->vid[vid].uv); > + return regulator_set_voltage(priv->vdd, volt->vid[vid].uv, 1200000); > +} > + > +static int > +gk20a_volt_set_id(struct nouveau_volt *volt, u8 id, int condition) > +{ > + struct gk20a_volt_priv *priv = (void *)volt; > + int prev_uv = regulator_get_voltage(priv->vdd); > + int target_uv = volt->vid[id].uv; > + int ret; > + > + nv_debug(volt, "prev=%d, target=%d, condition=%d\n", > + prev_uv, target_uv, condition); > + if (!condition || > + (condition < 0 && target_uv < prev_uv) || > + (condition > 0 && target_uv > prev_uv)) { > + ret = gk20a_volt_vid_set(volt, volt->vid[id].vid); > + } else { > + ret = 0; > + } > + > + return ret; > +} > + > +static int > +gk20a_volt_ctor(struct nouveau_object *parent, struct nouveau_object *engine, > + struct nouveau_oclass *oclass, void *data, u32 size, > + struct nouveau_object **pobject) > +{ > + struct gk20a_volt_priv *priv; > + struct nouveau_volt *volt; > + struct nouveau_platform_device *plat; > + int i, ret, uv; > + > + ret = nouveau_volt_create(parent, engine, oclass, &priv); > + *pobject = nv_object(priv); > + if (ret) > + return ret; > + > + volt = &priv->base; > + > + plat = nv_device_to_platform(nv_device(parent)); > + > + uv = regulator_get_voltage(plat->gpu->vdd); > + nv_info(priv, "The default voltage is %duV\n", uv); > + > + priv->vdd = plat->gpu->vdd; > + priv->base.vid_get = gk20a_volt_vid_get; > + priv->base.vid_set = gk20a_volt_vid_set; > + priv->base.set_id = gk20a_volt_set_id; > + > + volt->vid_nr = ARRAY_SIZE(gk20a_cvb_coef); > + nv_debug(priv, "%s - vid_nr = %d\n", __func__, volt->vid_nr); > + for (i = 0; i < volt->vid_nr; i++) { > + volt->vid[i].vid = i; > + volt->vid[i].uv = gk20a_volt_calc_voltage(&gk20a_cvb_coef[i]); > + nv_debug(priv, "%2d: vid=%d, uv=%d\n", i, volt->vid[i].vid, > + volt->vid[i].uv); > + } > + > + return 0; > +} > + > +struct nouveau_oclass > +gk20a_volt_oclass = { > + .handle = NV_SUBDEV(VOLT, 0xea), > + .ofuncs = &(struct nouveau_ofuncs) { > + .ctor = gk20a_volt_ctor, > + .dtor = _nouveau_volt_dtor, > + .init = _nouveau_volt_init, > + .fini = _nouveau_volt_fini, > + }, > +};
Vince Hsu
2014-Nov-28 11:57 UTC
[Nouveau] [RESEND PATCH nouveau 3/3] volt: add support for GK20A
Hi Roy, On 11/28/2014 07:25 PM, Roy Spliet wrote:> Hello Vince, > > One minor question inline. > > Op 28-11-14 om 12:13 schreef Vince Hsu: >> The voltage value are calculated by the hardware characterized >> result. >> >> Signed-off-by: Vince Hsu <vinceh at nvidia.com> >> --- >> >> Resend this patch with the fuse change and proper patch prefix >> per Thierry's request. >> >> drm/Kbuild | 1 + >> drm/core/subdev/volt/gk20a.c | 1 + >> nvkm/engine/device/nve0.c | 1 + >> nvkm/include/subdev/volt.h | 1 + >> nvkm/subdev/clock/gk20a.c | 15 ++++ >> nvkm/subdev/volt/gk20a.c | 202 >> +++++++++++++++++++++++++++++++++++++++++++ >> 6 files changed, 221 insertions(+) >> create mode 120000 drm/core/subdev/volt/gk20a.c >> create mode 100644 nvkm/subdev/volt/gk20a.c >> >> diff --git a/drm/Kbuild b/drm/Kbuild >> index 728bc5b66b29..7c49e6655066 100644 >> --- a/drm/Kbuild >> +++ b/drm/Kbuild >> @@ -225,6 +225,7 @@ nouveau-y += core/subdev/vm/nvc0.o >> nouveau-y += core/subdev/volt/base.o >> nouveau-y += core/subdev/volt/gpio.o >> nouveau-y += core/subdev/volt/nv40.o >> +nouveau-y += core/subdev/volt/gk20a.o >> nouveau-y += core/engine/falcon.o >> nouveau-y += core/engine/xtensa.o >> diff --git a/drm/core/subdev/volt/gk20a.c b/drm/core/subdev/volt/gk20a.c >> new file mode 120000 >> index 000000000000..2894eb1ede13 >> --- /dev/null >> +++ b/drm/core/subdev/volt/gk20a.c >> @@ -0,0 +1 @@ >> +../../../../nvkm/subdev/volt/gk20a.c >> \ No newline at end of file >> diff --git a/nvkm/engine/device/nve0.c b/nvkm/engine/device/nve0.c >> index b1b2e484ecfa..674da1f095b2 100644 >> --- a/nvkm/engine/device/nve0.c >> +++ b/nvkm/engine/device/nve0.c >> @@ -179,6 +179,7 @@ nve0_identify(struct nouveau_device *device) >> device->oclass[NVDEV_ENGINE_GR ] = gk20a_graph_oclass; >> device->oclass[NVDEV_ENGINE_COPY2 ] = &nve0_copy2_oclass; >> device->oclass[NVDEV_ENGINE_PERFMON] = &nve0_perfmon_oclass; >> + device->oclass[NVDEV_SUBDEV_VOLT ] = &gk20a_volt_oclass; >> break; >> case 0xf0: >> device->cname = "GK110"; >> diff --git a/nvkm/include/subdev/volt.h b/nvkm/include/subdev/volt.h >> index 820b62ffd75b..67db5e58880d 100644 >> --- a/nvkm/include/subdev/volt.h >> +++ b/nvkm/include/subdev/volt.h >> @@ -52,6 +52,7 @@ int _nouveau_volt_init(struct nouveau_object *); >> #define _nouveau_volt_fini _nouveau_subdev_fini >> extern struct nouveau_oclass nv40_volt_oclass; >> +extern struct nouveau_oclass gk20a_volt_oclass; >> int nouveau_voltgpio_init(struct nouveau_volt *); >> int nouveau_voltgpio_get(struct nouveau_volt *); >> diff --git a/nvkm/subdev/clock/gk20a.c b/nvkm/subdev/clock/gk20a.c >> index 82abbea2be12..fb4fad374bdd 100644 >> --- a/nvkm/subdev/clock/gk20a.c >> +++ b/nvkm/subdev/clock/gk20a.c >> @@ -470,76 +470,91 @@ gk20a_pstates[] = { >> { >> .base = { >> .domain[nv_clk_src_gpc] = 72000, >> + .voltage = 0, >> }, >> }, >> { >> .base = { >> .domain[nv_clk_src_gpc] = 108000, >> + .voltage = 1, >> }, >> }, >> { >> .base = { >> .domain[nv_clk_src_gpc] = 180000, >> + .voltage = 2, >> }, >> }, >> { >> .base = { >> .domain[nv_clk_src_gpc] = 252000, >> + .voltage = 3, >> }, >> }, >> { >> .base = { >> .domain[nv_clk_src_gpc] = 324000, >> + .voltage = 4, >> }, >> }, >> { >> .base = { >> .domain[nv_clk_src_gpc] = 396000, >> + .voltage = 5, >> }, >> }, >> { >> .base = { >> .domain[nv_clk_src_gpc] = 468000, >> + .voltage = 6, >> }, >> }, >> { >> .base = { >> .domain[nv_clk_src_gpc] = 540000, >> + .voltage = 7, >> }, >> }, >> { >> .base = { >> .domain[nv_clk_src_gpc] = 612000, >> + .voltage = 8, >> }, >> }, >> { >> .base = { >> .domain[nv_clk_src_gpc] = 648000, >> + .voltage = 9, >> }, >> }, >> { >> .base = { >> .domain[nv_clk_src_gpc] = 684000, >> + .voltage = 10, >> }, >> }, >> { >> .base = { >> .domain[nv_clk_src_gpc] = 708000, >> + .voltage = 11, >> }, >> }, >> { >> .base = { >> .domain[nv_clk_src_gpc] = 756000, >> + .voltage = 12, >> }, >> }, >> { >> .base = { >> .domain[nv_clk_src_gpc] = 804000, >> + .voltage = 13, >> }, >> }, >> { >> .base = { >> .domain[nv_clk_src_gpc] = 852000, >> + .voltage = 14, >> }, >> }, >> }; > > Is there a particular reason why this table is hard-coded rather than > stored in the device tree? It doesn't seem to differ much between > different gk20a's (but this might change with the denver-core > version?), but I do anticipate a lot of "code" duplication when > post-K1 cores are released and supported.Hmmm.. That's probably because I just realized we have some other example like the bios subdev is using the device tress stuff. The table is chip specific, not board specific. Should we put it in device tree? And the table should be the same between all the gk20a's although there is no guarantee. If the post-K1 cores also integrate the gk20a, they should use the same driver including this file. If not, I believe we will have some clever way to handle that. That's unlikely to happen though. :) Thanks, Vince> >> diff --git a/nvkm/subdev/volt/gk20a.c b/nvkm/subdev/volt/gk20a.c >> new file mode 100644 >> index 000000000000..87385966e67f >> --- /dev/null >> +++ b/nvkm/subdev/volt/gk20a.c >> @@ -0,0 +1,202 @@ >> +/* >> + * Copyright (c) 2014, NVIDIA CORPORATION. All rights reserved. >> + * >> + * Permission is hereby granted, free of charge, to any person >> obtaining a >> + * copy of this software and associated documentation files (the >> "Software"), >> + * to deal in the Software without restriction, including without >> limitation >> + * the rights to use, copy, modify, merge, publish, distribute, >> sublicense, >> + * and/or sell copies of the Software, and to permit persons to whom >> the >> + * Software is furnished to do so, subject to the following conditions: >> + * >> + * The above copyright notice and this permission notice shall be >> included in >> + * all copies or substantial portions of the Software. >> + * >> + * THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, >> EXPRESS OR >> + * IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF >> MERCHANTABILITY, >> + * FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO >> EVENT SHALL >> + * THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES >> OR OTHER >> + * LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, >> ARISING >> + * FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER >> + * DEALINGS IN THE SOFTWARE. >> + */ >> + >> +#include <linux/kernel.h> >> +#include <linux/regulator/consumer.h> >> +#include <soc/tegra/fuse.h> >> + >> +#include <nouveau_platform.h> >> +#include <subdev/volt.h> >> + >> +struct cvb_coef { >> + int c0; >> + int c1; >> + int c2; >> + int c3; >> + int c4; >> + int c5; >> +}; >> + >> +struct gk20a_volt_priv { >> + struct nouveau_volt base; >> + struct regulator *vdd; >> +}; >> + >> +const struct cvb_coef gk20a_cvb_coef[] = { >> + /* MHz, c0, c1, c2, c3, c4, c5 */ >> + /* 72 */ { 1209886, -36468, 515, 417, -13123, 203}, >> + /* 108 */ { 1130804, -27659, 296, 298, -10834, 221}, >> + /* 180 */ { 1162871, -27110, 247, 238, -10681, 268}, >> + /* 252 */ { 1220458, -28654, 247, 179, -10376, 298}, >> + /* 324 */ { 1280953, -30204, 247, 119, -9766, 304}, >> + /* 396 */ { 1344547, -31777, 247, 119, -8545, 292}, >> + /* 468 */ { 1420168, -34227, 269, 60, -7172, 256}, >> + /* 540 */ { 1490757, -35955, 274, 60, -5188, 197}, >> + /* 612 */ { 1599112, -42583, 398, 0, -1831, 119}, >> + /* 648 */ { 1366986, -16459, -274, 0, -3204, 72}, >> + /* 684 */ { 1391884, -17078, -274, -60, -1526, 30}, >> + /* 708 */ { 1415522, -17497, -274, -60, -458, 0}, >> + /* 756 */ { 1464061, -18331, -274, -119, 1831, -72}, >> + /* 804 */ { 1524225, -20064, -254, -119, 4272, -155}, >> + /* 852 */ { 1608418, -21643, -269, 0, 763, -48}, >> +}; >> + >> +/** >> + * cvb_mv = ((c2 * speedo / s_scale + c1) * speedo / s_scale + c0) >> + */ >> +static inline int >> +gk20a_volt_get_cvb_voltage(int speedo, int s_scale, >> + const struct cvb_coef *coef) >> +{ >> + int mv; >> + >> + mv = DIV_ROUND_CLOSEST(coef->c2 * speedo, s_scale); >> + mv = DIV_ROUND_CLOSEST((mv + coef->c1) * speedo, s_scale) + >> coef->c0; >> + return mv; >> +} >> + >> +/** >> + * cvb_t_mv >> + * ((c2 * speedo / s_scale + c1) * speedo / s_scale + c0) + >> + * ((c3 * speedo / s_scale + c4 + c5 * T / t_scale) * T / t_scale) >> + */ >> +static inline int >> +gk20a_volt_get_cvb_t_voltage(int speedo, int temp, int s_scale, int >> t_scale, >> + const struct cvb_coef *coef) >> +{ >> + int cvb_mv, mv; >> + >> + cvb_mv = gk20a_volt_get_cvb_voltage(speedo, s_scale, coef); >> + >> + mv = DIV_ROUND_CLOSEST(coef->c3 * speedo, s_scale) + coef->c4 + >> + DIV_ROUND_CLOSEST(coef->c5 * temp, t_scale); >> + mv = DIV_ROUND_CLOSEST(mv * temp, t_scale) + cvb_mv; >> + return mv; >> +} >> + >> +static int >> +gk20a_volt_calc_voltage(const struct cvb_coef *coef) >> +{ >> + int speedo, mv; >> + >> + speedo = tegra_sku_info.gpu_speedo_value; >> + >> + mv = gk20a_volt_get_cvb_t_voltage(speedo, -10, 100, 10, coef); >> + mv = DIV_ROUND_UP(mv, 1000); >> + >> + return mv * 1000; >> +} >> + >> +static int >> +gk20a_volt_vid_get(struct nouveau_volt *volt) >> +{ >> + struct gk20a_volt_priv *priv = (void *)volt; >> + int i, uv; >> + >> + uv = regulator_get_voltage(priv->vdd); >> + >> + for (i = 0; i < volt->vid_nr; i++) >> + if (volt->vid[i].uv >= uv) >> + return i; >> + >> + return -EINVAL; >> +} >> + >> +static int >> +gk20a_volt_vid_set(struct nouveau_volt *volt, u8 vid) >> +{ >> + struct gk20a_volt_priv *priv = (void *)volt; >> + >> + nv_debug(volt, "set voltage as %duv\n", volt->vid[vid].uv); >> + return regulator_set_voltage(priv->vdd, volt->vid[vid].uv, >> 1200000); >> +} >> + >> +static int >> +gk20a_volt_set_id(struct nouveau_volt *volt, u8 id, int condition) >> +{ >> + struct gk20a_volt_priv *priv = (void *)volt; >> + int prev_uv = regulator_get_voltage(priv->vdd); >> + int target_uv = volt->vid[id].uv; >> + int ret; >> + >> + nv_debug(volt, "prev=%d, target=%d, condition=%d\n", >> + prev_uv, target_uv, condition); >> + if (!condition || >> + (condition < 0 && target_uv < prev_uv) || >> + (condition > 0 && target_uv > prev_uv)) { >> + ret = gk20a_volt_vid_set(volt, volt->vid[id].vid); >> + } else { >> + ret = 0; >> + } >> + >> + return ret; >> +} >> + >> +static int >> +gk20a_volt_ctor(struct nouveau_object *parent, struct nouveau_object >> *engine, >> + struct nouveau_oclass *oclass, void *data, u32 size, >> + struct nouveau_object **pobject) >> +{ >> + struct gk20a_volt_priv *priv; >> + struct nouveau_volt *volt; >> + struct nouveau_platform_device *plat; >> + int i, ret, uv; >> + >> + ret = nouveau_volt_create(parent, engine, oclass, &priv); >> + *pobject = nv_object(priv); >> + if (ret) >> + return ret; >> + >> + volt = &priv->base; >> + >> + plat = nv_device_to_platform(nv_device(parent)); >> + >> + uv = regulator_get_voltage(plat->gpu->vdd); >> + nv_info(priv, "The default voltage is %duV\n", uv); >> + >> + priv->vdd = plat->gpu->vdd; >> + priv->base.vid_get = gk20a_volt_vid_get; >> + priv->base.vid_set = gk20a_volt_vid_set; >> + priv->base.set_id = gk20a_volt_set_id; >> + >> + volt->vid_nr = ARRAY_SIZE(gk20a_cvb_coef); >> + nv_debug(priv, "%s - vid_nr = %d\n", __func__, volt->vid_nr); >> + for (i = 0; i < volt->vid_nr; i++) { >> + volt->vid[i].vid = i; >> + volt->vid[i].uv = gk20a_volt_calc_voltage(&gk20a_cvb_coef[i]); >> + nv_debug(priv, "%2d: vid=%d, uv=%d\n", i, volt->vid[i].vid, >> + volt->vid[i].uv); >> + } >> + >> + return 0; >> +} >> + >> +struct nouveau_oclass >> +gk20a_volt_oclass = { >> + .handle = NV_SUBDEV(VOLT, 0xea), >> + .ofuncs = &(struct nouveau_ofuncs) { >> + .ctor = gk20a_volt_ctor, >> + .dtor = _nouveau_volt_dtor, >> + .init = _nouveau_volt_init, >> + .fini = _nouveau_volt_fini, >> + }, >> +}; > > _______________________________________________ > Nouveau mailing list > Nouveau at lists.freedesktop.org > http://lists.freedesktop.org/mailman/listinfo/nouveau----------------------------------------------------------------------------------- This email message is for the sole use of the intended recipient(s) and may contain confidential information. Any unauthorized review, use, disclosure or distribution is prohibited. If you are not the intended recipient, please contact the sender by reply email and destroy all copies of the original message. -----------------------------------------------------------------------------------
Roy Spliet
2014-Nov-28 12:09 UTC
[Nouveau] [RESEND PATCH nouveau 3/3] volt: add support for GK20A
Hello Vince, Op 28-11-14 om 12:57 schreef Vince Hsu:> Hi Roy, > > On 11/28/2014 07:25 PM, Roy Spliet wrote: >> Hello Vince, >> >> One minor question inline. >> >> Op 28-11-14 om 12:13 schreef Vince Hsu: >>> The voltage value are calculated by the hardware characterized >>> result. >>> >>> Signed-off-by: Vince Hsu <vinceh at nvidia.com> >>> --- >>> >>> Resend this patch with the fuse change and proper patch prefix >>> per Thierry's request. >>> >>> drm/Kbuild | 1 + >>> drm/core/subdev/volt/gk20a.c | 1 + >>> nvkm/engine/device/nve0.c | 1 + >>> nvkm/include/subdev/volt.h | 1 + >>> nvkm/subdev/clock/gk20a.c | 15 ++++ >>> nvkm/subdev/volt/gk20a.c | 202 >>> +++++++++++++++++++++++++++++++++++++++++++ >>> 6 files changed, 221 insertions(+) >>> create mode 120000 drm/core/subdev/volt/gk20a.c >>> create mode 100644 nvkm/subdev/volt/gk20a.c >>> >>> diff --git a/drm/Kbuild b/drm/Kbuild >>> index 728bc5b66b29..7c49e6655066 100644 >>> --- a/drm/Kbuild >>> +++ b/drm/Kbuild >>> @@ -225,6 +225,7 @@ nouveau-y += core/subdev/vm/nvc0.o >>> nouveau-y += core/subdev/volt/base.o >>> nouveau-y += core/subdev/volt/gpio.o >>> nouveau-y += core/subdev/volt/nv40.o >>> +nouveau-y += core/subdev/volt/gk20a.o >>> nouveau-y += core/engine/falcon.o >>> nouveau-y += core/engine/xtensa.o >>> diff --git a/drm/core/subdev/volt/gk20a.c >>> b/drm/core/subdev/volt/gk20a.c >>> new file mode 120000 >>> index 000000000000..2894eb1ede13 >>> --- /dev/null >>> +++ b/drm/core/subdev/volt/gk20a.c >>> @@ -0,0 +1 @@ >>> +../../../../nvkm/subdev/volt/gk20a.c >>> \ No newline at end of file >>> diff --git a/nvkm/engine/device/nve0.c b/nvkm/engine/device/nve0.c >>> index b1b2e484ecfa..674da1f095b2 100644 >>> --- a/nvkm/engine/device/nve0.c >>> +++ b/nvkm/engine/device/nve0.c >>> @@ -179,6 +179,7 @@ nve0_identify(struct nouveau_device *device) >>> device->oclass[NVDEV_ENGINE_GR ] = gk20a_graph_oclass; >>> device->oclass[NVDEV_ENGINE_COPY2 ] = &nve0_copy2_oclass; >>> device->oclass[NVDEV_ENGINE_PERFMON] = &nve0_perfmon_oclass; >>> + device->oclass[NVDEV_SUBDEV_VOLT ] = &gk20a_volt_oclass; >>> break; >>> case 0xf0: >>> device->cname = "GK110"; >>> diff --git a/nvkm/include/subdev/volt.h b/nvkm/include/subdev/volt.h >>> index 820b62ffd75b..67db5e58880d 100644 >>> --- a/nvkm/include/subdev/volt.h >>> +++ b/nvkm/include/subdev/volt.h >>> @@ -52,6 +52,7 @@ int _nouveau_volt_init(struct nouveau_object *); >>> #define _nouveau_volt_fini _nouveau_subdev_fini >>> extern struct nouveau_oclass nv40_volt_oclass; >>> +extern struct nouveau_oclass gk20a_volt_oclass; >>> int nouveau_voltgpio_init(struct nouveau_volt *); >>> int nouveau_voltgpio_get(struct nouveau_volt *); >>> diff --git a/nvkm/subdev/clock/gk20a.c b/nvkm/subdev/clock/gk20a.c >>> index 82abbea2be12..fb4fad374bdd 100644 >>> --- a/nvkm/subdev/clock/gk20a.c >>> +++ b/nvkm/subdev/clock/gk20a.c >>> @@ -470,76 +470,91 @@ gk20a_pstates[] = { >>> { >>> .base = { >>> .domain[nv_clk_src_gpc] = 72000, >>> + .voltage = 0, >>> }, >>> }, >>> { >>> .base = { >>> .domain[nv_clk_src_gpc] = 108000, >>> + .voltage = 1, >>> }, >>> }, >>> { >>> .base = { >>> .domain[nv_clk_src_gpc] = 180000, >>> + .voltage = 2, >>> }, >>> }, >>> { >>> .base = { >>> .domain[nv_clk_src_gpc] = 252000, >>> + .voltage = 3, >>> }, >>> }, >>> { >>> .base = { >>> .domain[nv_clk_src_gpc] = 324000, >>> + .voltage = 4, >>> }, >>> }, >>> { >>> .base = { >>> .domain[nv_clk_src_gpc] = 396000, >>> + .voltage = 5, >>> }, >>> }, >>> { >>> .base = { >>> .domain[nv_clk_src_gpc] = 468000, >>> + .voltage = 6, >>> }, >>> }, >>> { >>> .base = { >>> .domain[nv_clk_src_gpc] = 540000, >>> + .voltage = 7, >>> }, >>> }, >>> { >>> .base = { >>> .domain[nv_clk_src_gpc] = 612000, >>> + .voltage = 8, >>> }, >>> }, >>> { >>> .base = { >>> .domain[nv_clk_src_gpc] = 648000, >>> + .voltage = 9, >>> }, >>> }, >>> { >>> .base = { >>> .domain[nv_clk_src_gpc] = 684000, >>> + .voltage = 10, >>> }, >>> }, >>> { >>> .base = { >>> .domain[nv_clk_src_gpc] = 708000, >>> + .voltage = 11, >>> }, >>> }, >>> { >>> .base = { >>> .domain[nv_clk_src_gpc] = 756000, >>> + .voltage = 12, >>> }, >>> }, >>> { >>> .base = { >>> .domain[nv_clk_src_gpc] = 804000, >>> + .voltage = 13, >>> }, >>> }, >>> { >>> .base = { >>> .domain[nv_clk_src_gpc] = 852000, >>> + .voltage = 14, >>> }, >>> }, >>> }; >> >> Is there a particular reason why this table is hard-coded rather than >> stored in the device tree? It doesn't seem to differ much between >> different gk20a's (but this might change with the denver-core >> version?), but I do anticipate a lot of "code" duplication when >> post-K1 cores are released and supported. > Hmmm.. That's probably because I just realized we have some other > example like the bios subdev is using the device tress stuff. The > table is chip specific, not board specific. Should we put it in device > tree? > And the table should be the same between all the gk20a's although > there is no guarantee. If the post-K1 cores also integrate the gk20a, > they should use the same driver including this file. If not, I believe > we will have some clever way to handle that. That's unlikely to happen > though. :) > > Thanks, > VinceI'm not sure if I completely understand your reply, so forgive me if I am stating some obvious things: The reason why I brought this up is because, the way I see it, DTS is the replacement for (V)BIOS on ARM platforms, giving a set of parameters that drivers (nouveau) can use for that particular instance (the Tegra K1 SoC) of some more generic IP (gk20a). All the other devices nouveau supports have a VBIOS to describe this kind of information to us, hence we haven't seen this before. For CPUs there are plenty of examples though of such params defined in DT: in arch/arm/boot/dts/ : imx6qdl.dtsi documents the min and max volt for regulators, while the CPUs have a little freq<->volt mapping in imx6q.dtsi. GPUs are new in a sense that NVIDIA is the first to actively support upstream development (thanks!) Secondly, we should keep in mind that DT is not tied to Linux; I believe Linaro's long term goal is to take the DT from the Linux tree and maintain it as a separate tree, to be used with U-boot, *BSD, maybe even Windows. These kind of parameters are not very platform-dependent and although they seem like a little detail that's easy to reproduce on every platform, looking at the sheer size of the VBIOS data that could mean a *lot* of duplication. I bring this up not because I think I know better, but rather because I believe it's a good discussion to have now that there still is little legacy on ARM SoCs in nouveau. DTS is still a work-in-progress, and at this moment we have the opportunity to consider what needs to be documented in there and what doesn't. I would actually love to hear other, more experienced developers chime in as well (Rob Clark? Ben Skeggs? Linaro folks?) and see how they feel. Cheers, Roy
Seemingly Similar Threads
- [RESEND PATCH nouveau 3/3] volt: add support for GK20A
- [PATCH 1/2] volt: allow non-bios voltage scaling
- [RESEND PATCH nouveau 3/3] volt: add support for GK20A
- [RESEND V2 PATCH 1/3] soc/tegra: fuse: export tegra_sku_info for module use
- [V3 PATCH 1/3] soc/tegra: fuse: export tegra_sku_info