| Message ID | 20240426165839.v2.4.Ie94246c30fe95101e0e26dd5f96e976dbeb8f242@changeid (mailing list archive) |
|---|---|
| State | New, archived |
| Headers | show |
| Series | drm/mipi-dsi: Reduce bloat and add funcs for cleaner init seqs | expand |
Hi Douglas. On Fri, Apr 26, 2024 at 04:58:37PM -0700, Douglas Anderson wrote: > The current mipi_dsi_*_write_seq() macros are non-intutitive because > they contain a hidden "return" statement that will return out of the > _caller_ of the macro. Let's mark them as deprecated and instead > introduce some new macros that are more intuitive. > > These new macros are less optimal when an error occurs but should > behave more optimally when there is no error. Specifically these new > macros cause smaller code to get generated and the code size savings > (less to fetch from RAM, less cache space used, less RAM used) are > important. Since the error case isn't something we need to optimize > for and these new macros are easier to understand and more flexible, > they should be used. The whole ignore-and-return-early-if-an-error-occured concept is nice and make this easy to understand and use. I have a few nits in the following, but the overall concept is nice. Sam > > After converting to use these new functions, one example shows some > nice savings while also being easier to understand. > > $ scripts/bloat-o-meter \ > ...after/panel-novatek-nt36672e.ko \ > ...ctx/panel-novatek-nt36672e.ko > add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-988 (-988) > Function old new delta > nt36672e_1080x2408_60hz_init 6236 5248 -988 > Total: Before=10651, After=9663, chg -9.28% > > Signed-off-by: Douglas Anderson <dianders@chromium.org> > --- > Right now this patch introduces two new functions in > drm_mipi_dsi.c. Alternatively we could have changed the prototype of > the "chatty" functions and made the deprecated macros adapt to the new > prototype. While this sounds nice, it bloated callers of the > deprecated functioin a bit because it caused the compiler to emit less > optimal code. It doesn't seem terrible to add two more functions, so I > went that way. The concern here is when it will be cleaned up. As a minimum please consider adding an item to todo.rst that details what should be done to get rid of the now obsolete chatty functions so someone can pick it up. > > Changes in v2: > - New > > drivers/gpu/drm/drm_mipi_dsi.c | 56 +++++++++++++++++++++++++++++++++ > include/drm/drm_mipi_dsi.h | 57 ++++++++++++++++++++++++++++++++++ > 2 files changed, 113 insertions(+) > > diff --git a/drivers/gpu/drm/drm_mipi_dsi.c b/drivers/gpu/drm/drm_mipi_dsi.c > index 1e062eda3b88..b7c75a4396e6 100644 > --- a/drivers/gpu/drm/drm_mipi_dsi.c > +++ b/drivers/gpu/drm/drm_mipi_dsi.c > @@ -792,6 +792,34 @@ int mipi_dsi_generic_write_chatty(struct mipi_dsi_device *dsi, > } > EXPORT_SYMBOL(mipi_dsi_generic_write_chatty); > > +/** > + * mipi_dsi_generic_write_multi() - mipi_dsi_generic_write_chatty() w/ accum_err > + * @ctx: Context for multiple DSI transactions > + * @payload: buffer containing the payload > + * @size: size of payload buffer > + * > + * Like mipi_dsi_generic_write_chatty() but deals with errors in a way that > + * makes it convenient to make several calls in a row. > + */ > +void mipi_dsi_generic_write_multi(struct mipi_dsi_multi_context *ctx, > + const void *payload, size_t size) > +{ > + struct mipi_dsi_device *dsi = ctx->dsi; > + struct device *dev = &dsi->dev; > + ssize_t ret; > + > + if (ctx->accum_err) > + return; > + > + ret = mipi_dsi_generic_write(dsi, payload, size); > + if (ret < 0) { > + ctx->accum_err = ret; > + dev_err_ratelimited(dev, "sending generic data %*ph failed: %d\n", > + (int)size, payload, ctx->accum_err); > + } I see no point in using ratelimited and then change it in the next patch. > +} > +EXPORT_SYMBOL(mipi_dsi_generic_write_multi); > + > /** > * mipi_dsi_generic_read() - receive data using a generic read packet > * @dsi: DSI peripheral device > @@ -908,6 +936,34 @@ int mipi_dsi_dcs_write_buffer_chatty(struct mipi_dsi_device *dsi, > } > EXPORT_SYMBOL(mipi_dsi_dcs_write_buffer_chatty); > > +/** > + * mipi_dsi_dcs_write_buffer_multi - mipi_dsi_dcs_write_buffer_chatty() w/ accum_err > + * @ctx: Context for multiple DSI transactions > + * @data: buffer containing data to be transmitted > + * @len: size of transmission buffer > + * > + * Like mipi_dsi_dcs_write_buffer_chatty() but deals with errors in a way that > + * makes it convenient to make several calls in a row. > + */ > +void mipi_dsi_dcs_write_buffer_multi(struct mipi_dsi_multi_context *ctx, > + const void *data, size_t len) > +{ > + struct mipi_dsi_device *dsi = ctx->dsi; > + struct device *dev = &dsi->dev; > + ssize_t ret; > + > + if (ctx->accum_err) > + return; > + > + ret = mipi_dsi_dcs_write_buffer(dsi, data, len); > + if (ret < 0) { > + ctx->accum_err = ret; > + dev_err_ratelimited(dev, "sending dcs data %*ph failed: %d\n", > + (int)len, data, ctx->accum_err); > + } > +} > +EXPORT_SYMBOL(mipi_dsi_dcs_write_buffer_multi); > + > /** > * mipi_dsi_dcs_write() - send DCS write command > * @dsi: DSI peripheral device > diff --git a/include/drm/drm_mipi_dsi.h b/include/drm/drm_mipi_dsi.h > index 6d68d9927f46..379594649a42 100644 > --- a/include/drm/drm_mipi_dsi.h > +++ b/include/drm/drm_mipi_dsi.h > @@ -197,6 +197,22 @@ struct mipi_dsi_device { > struct drm_dsc_config *dsc; > }; > > +/** > + * struct mipi_dsi_multi_context - Context to call multiple MIPI DSI funcs in a row > + * @dsi: Pointer to the MIPI DSI device > + * @accum_err: Storage for the accumulated error over the multiple calls. Init > + * to 0. If a function encounters an error then the error code will be > + * stored here. If you call a function and this points to a non-zero > + * value then the function will be a noop. This allows calling a function > + * many times in a row and just checking the error at the end to see if > + * any of them failed. > + */ > + > +struct mipi_dsi_multi_context { > + struct mipi_dsi_device *dsi; > + int accum_err; > +}; Inline comments is - I think - preferred. > + > #define MIPI_DSI_MODULE_PREFIX "mipi-dsi:" > > #define to_mipi_dsi_device(__dev) container_of_const(__dev, struct mipi_dsi_device, dev) > @@ -258,6 +274,8 @@ ssize_t mipi_dsi_generic_write(struct mipi_dsi_device *dsi, const void *payload, > size_t size); > int mipi_dsi_generic_write_chatty(struct mipi_dsi_device *dsi, > const void *payload, size_t size); > +void mipi_dsi_generic_write_multi(struct mipi_dsi_multi_context *ctx, > + const void *payload, size_t size); > ssize_t mipi_dsi_generic_read(struct mipi_dsi_device *dsi, const void *params, > size_t num_params, void *data, size_t size); > > @@ -283,6 +301,8 @@ ssize_t mipi_dsi_dcs_write_buffer(struct mipi_dsi_device *dsi, > const void *data, size_t len); > int mipi_dsi_dcs_write_buffer_chatty(struct mipi_dsi_device *dsi, > const void *data, size_t len); > +void mipi_dsi_dcs_write_buffer_multi(struct mipi_dsi_multi_context *ctx, > + const void *data, size_t len); > ssize_t mipi_dsi_dcs_write(struct mipi_dsi_device *dsi, u8 cmd, > const void *data, size_t len); > ssize_t mipi_dsi_dcs_read(struct mipi_dsi_device *dsi, u8 cmd, void *data, > @@ -319,6 +339,9 @@ int mipi_dsi_dcs_get_display_brightness_large(struct mipi_dsi_device *dsi, > * This macro will print errors for you and will RETURN FROM THE CALLING > * FUNCTION (yes this is non-intuitive) upon error. > * > + * Because of the non-intuitive return behavior, THIS MACRO IS DEPRECATED. > + * Please replace calls of it with mipi_dsi_generic_write_seq_multi(). > + * > * @dsi: DSI peripheral device > * @seq: buffer containing the payload > */ > @@ -331,12 +354,30 @@ int mipi_dsi_dcs_get_display_brightness_large(struct mipi_dsi_device *dsi, > return ret; \ > } while (0) > > +/** > + * mipi_dsi_generic_write_seq_multi - transmit data using a generic write packet > + * > + * This macro will print errors for you and error handling is optimized for > + * callers that call this multiple times in a row. > + * > + * @ctx: Context for multiple DSI transactions > + * @seq: buffer containing the payload > + */ > +#define mipi_dsi_generic_write_seq_multi(ctx, seq...) \ > + do { \ > + static const u8 d[] = { seq }; \ > + mipi_dsi_generic_write_multi(ctx, d, ARRAY_SIZE(d)); \ > + } while (0) > + > /** > * mipi_dsi_dcs_write_seq - transmit a DCS command with payload > * > * This macro will print errors for you and will RETURN FROM THE CALLING > * FUNCTION (yes this is non-intuitive) upon error. > * > + * Because of the non-intuitive return behavior, THIS MACRO IS DEPRECATED. > + * Please replace calls of it with mipi_dsi_dcs_write_seq_multi(). > + * > * @dsi: DSI peripheral device > * @cmd: Command > * @seq: buffer containing data to be transmitted > @@ -350,6 +391,22 @@ int mipi_dsi_dcs_get_display_brightness_large(struct mipi_dsi_device *dsi, > return ret; \ > } while (0) > > +/** > + * mipi_dsi_dcs_write_seq_multi - transmit a DCS command with payload > + * > + * This macro will print errors for you and error handling is optimized for > + * callers that call this multiple times in a row. > + * > + * @ctx: Context for multiple DSI transactions > + * @cmd: Command > + * @seq: buffer containing data to be transmitted > + */ > +#define mipi_dsi_dcs_write_seq_multi(ctx, cmd, seq...) \ > + do { \ > + static const u8 d[] = { cmd, seq }; \ > + mipi_dsi_dcs_write_buffer_multi(ctx, d, ARRAY_SIZE(d)); \ > + } while (0) > + > /** > * struct mipi_dsi_driver - DSI driver > * @driver: device driver model driver > -- > 2.44.0.769.g3c40516874-goog
Hello Mister Anderson, On 27/04/2024 01:58, Douglas Anderson wrote: > The current mipi_dsi_*_write_seq() macros are non-intutitive because > they contain a hidden "return" statement that will return out of the > _caller_ of the macro. Let's mark them as deprecated and instead > introduce some new macros that are more intuitive. > > These new macros are less optimal when an error occurs but should > behave more optimally when there is no error. Specifically these new > macros cause smaller code to get generated and the code size savings > (less to fetch from RAM, less cache space used, less RAM used) are > important. Since the error case isn't something we need to optimize > for and these new macros are easier to understand and more flexible, > they should be used. > > After converting to use these new functions, one example shows some > nice savings while also being easier to understand. > > $ scripts/bloat-o-meter \ > ...after/panel-novatek-nt36672e.ko \ > ...ctx/panel-novatek-nt36672e.ko > add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-988 (-988) > Function old new delta > nt36672e_1080x2408_60hz_init 6236 5248 -988 > Total: Before=10651, After=9663, chg -9.28% > > Signed-off-by: Douglas Anderson <dianders@chromium.org> > --- > Right now this patch introduces two new functions in > drm_mipi_dsi.c. Alternatively we could have changed the prototype of > the "chatty" functions and made the deprecated macros adapt to the new > prototype. While this sounds nice, it bloated callers of the > deprecated functioin a bit because it caused the compiler to emit less > optimal code. It doesn't seem terrible to add two more functions, so I > went that way. > > Changes in v2: > - New > > drivers/gpu/drm/drm_mipi_dsi.c | 56 +++++++++++++++++++++++++++++++++ > include/drm/drm_mipi_dsi.h | 57 ++++++++++++++++++++++++++++++++++ > 2 files changed, 113 insertions(+) > > diff --git a/drivers/gpu/drm/drm_mipi_dsi.c b/drivers/gpu/drm/drm_mipi_dsi.c > index 1e062eda3b88..b7c75a4396e6 100644 > --- a/drivers/gpu/drm/drm_mipi_dsi.c > +++ b/drivers/gpu/drm/drm_mipi_dsi.c > @@ -792,6 +792,34 @@ int mipi_dsi_generic_write_chatty(struct mipi_dsi_device *dsi, > } > EXPORT_SYMBOL(mipi_dsi_generic_write_chatty); > <snip> > }; > > +/** > + * struct mipi_dsi_multi_context - Context to call multiple MIPI DSI funcs in a row > + * @dsi: Pointer to the MIPI DSI device > + * @accum_err: Storage for the accumulated error over the multiple calls. Init > + * to 0. If a function encounters an error then the error code will be > + * stored here. If you call a function and this points to a non-zero > + * value then the function will be a noop. This allows calling a function > + * many times in a row and just checking the error at the end to see if > + * any of them failed. > + */ > + > +struct mipi_dsi_multi_context { > + struct mipi_dsi_device *dsi; > + int accum_err; > +}; I like the design, but having a context struct seems over-engineered while we could pass a single int over without encapsulating it with mipi_dsi_multi_context. void mipi_dsi_dcs_write_buffer_multi(struct mipi_dsi_multi_context *ctx, const void *data, size_t len); vs void mipi_dsi_dcs_write_buffer_multi(struct mipi_dsi_device *dsi, int *accum_err, const void *data, size_t len); is the same, and it avoids having to declare a mipi_dsi_multi_context and set ctx->dsi, and I'll find it much easier to migrate, just add a &ret and make sure ret is initialized to 0. <snip>
Hi, On Fri, Apr 26, 2024 at 11:33 PM Sam Ravnborg <sam@ravnborg.org> wrote: > > > --- > > Right now this patch introduces two new functions in > > drm_mipi_dsi.c. Alternatively we could have changed the prototype of > > the "chatty" functions and made the deprecated macros adapt to the new > > prototype. While this sounds nice, it bloated callers of the > > deprecated functioin a bit because it caused the compiler to emit less > > optimal code. It doesn't seem terrible to add two more functions, so I > > went that way. > The concern here is when it will be cleaned up. As a minimum please > consider adding an item to todo.rst that details what should be done > to get rid of the now obsolete chatty functions so someone can pick it > up. Sure, I can add something to do TODO. Do folks agree that's the preferred way to do things? A few thoughts I had: 1. In theory it could be useful to keep _both_ the "chatty" and "multi" variants of the functions. In at least a few places the "multi" variant was awkward. The resulting auo_kd101n80_45na_init() [1] looked awkward. If I was writing just this function I would have chosen an API more like the "chatty" one, but since I was trying to make all the init functions similar I kept them all using the "multi" API. Does it make sense to keep both long term? [1] https://lore.kernel.org/all/20240426165839.v2.7.Ib5030ab5cd41b4e08b1958bd7e51571725723008@changeid/ 2. Going completely the opposite direction, we could also not bother saving as much space today and _force_ everyone to move to the new "multi" API to get the full space savings. So I guess three options: a) leave my patches the way they are and add a TODO, b) Keep the "chatty" variants long term and remove my "after-the-cut", or c) Don't get as much space savings today but don't introduce a new exported function. Opinions? > > @@ -792,6 +792,34 @@ int mipi_dsi_generic_write_chatty(struct mipi_dsi_device *dsi, > > } > > EXPORT_SYMBOL(mipi_dsi_generic_write_chatty); > > > > +/** > > + * mipi_dsi_generic_write_multi() - mipi_dsi_generic_write_chatty() w/ accum_err > > + * @ctx: Context for multiple DSI transactions > > + * @payload: buffer containing the payload > > + * @size: size of payload buffer > > + * > > + * Like mipi_dsi_generic_write_chatty() but deals with errors in a way that > > + * makes it convenient to make several calls in a row. > > + */ > > +void mipi_dsi_generic_write_multi(struct mipi_dsi_multi_context *ctx, > > + const void *payload, size_t size) > > +{ > > + struct mipi_dsi_device *dsi = ctx->dsi; > > + struct device *dev = &dsi->dev; > > + ssize_t ret; > > + > > + if (ctx->accum_err) > > + return; > > + > > + ret = mipi_dsi_generic_write(dsi, payload, size); > > + if (ret < 0) { > > + ctx->accum_err = ret; > > + dev_err_ratelimited(dev, "sending generic data %*ph failed: %d\n", > > + (int)size, payload, ctx->accum_err); > > + } > I see no point in using ratelimited and then change it in the next > patch. Sure, I'll re-order patches. > > @@ -197,6 +197,22 @@ struct mipi_dsi_device { > > struct drm_dsc_config *dsc; > > }; > > > > +/** > > + * struct mipi_dsi_multi_context - Context to call multiple MIPI DSI funcs in a row > > + * @dsi: Pointer to the MIPI DSI device > > + * @accum_err: Storage for the accumulated error over the multiple calls. Init > > + * to 0. If a function encounters an error then the error code will be > > + * stored here. If you call a function and this points to a non-zero > > + * value then the function will be a noop. This allows calling a function > > + * many times in a row and just checking the error at the end to see if > > + * any of them failed. > > + */ > > + > > +struct mipi_dsi_multi_context { > > + struct mipi_dsi_device *dsi; > > + int accum_err; > > +}; > Inline comments is - I think - preferred. Sure, I'll update in the next version.
Hi, On Mon, Apr 29, 2024 at 2:38 AM Neil Armstrong <neil.armstrong@linaro.org> wrote: > > > +/** > > + * struct mipi_dsi_multi_context - Context to call multiple MIPI DSI funcs in a row > > + * @dsi: Pointer to the MIPI DSI device > > + * @accum_err: Storage for the accumulated error over the multiple calls. Init > > + * to 0. If a function encounters an error then the error code will be > > + * stored here. If you call a function and this points to a non-zero > > + * value then the function will be a noop. This allows calling a function > > + * many times in a row and just checking the error at the end to see if > > + * any of them failed. > > + */ > > + > > +struct mipi_dsi_multi_context { > > + struct mipi_dsi_device *dsi; > > + int accum_err; > > +}; > > I like the design, but having a context struct seems over-engineered while we could pass > a single int over without encapsulating it with mipi_dsi_multi_context. > > void mipi_dsi_dcs_write_buffer_multi(struct mipi_dsi_multi_context *ctx, > const void *data, size_t len); > vs > void mipi_dsi_dcs_write_buffer_multi(struct mipi_dsi_device *dsi, int *accum_err, > const void *data, size_t len); > > is the same, and it avoids having to declare a mipi_dsi_multi_context and set ctx->dsi, > and I'll find it much easier to migrate, just add a &ret and make sure ret is initialized to 0. Yeah, I had the same reaction when Jani suggested the context style [1] and I actually coded it up exactly as you suggest above. I then changed my mind and went with the context. My motivation was that when I tested it I found that using the context produced smaller code. Specifically, from the description of this patch we see we end up with: Total: Before=10651, After=9663, chg -9.28% ...when I didn't have the context and I had the accum_err then instead of getting ~9% smaller I believe it actually got ~0.5% bigger. This makes some sense as the caller has to pass 4 parameters to each call instead of 3. It's not a giant size difference but it was at least some motivation that helped push me in this direction. I'd also say that when I looked at the code in the end the style grew on me. It's really not too terrible to have one line in your functions that looks like: struct mipi_dsi_multi_context ctx = { .dsi = boe->dsi }; ...and if that becomes the canonical way to do it then it's really hard to accidentally forget to initialize the error value. With the other API it's _very_ easy to forget to initialize the error value and the compiler won't yell at you. It also makes it very obvious to the caller that this function is doing something a little different than most Linux APIs with that error return. So I guess I'd say that I ended up being pretty happy with the "context" even if it does feel a little over engineered and I'd argue to keep it that way. That being said, if you feel strongly about it then we can perhaps get others to chime in to see which style they prefer? Let me know what you think. [1] https://lore.kernel.org/r/8734r85tcf.fsf@intel.com
On Mon, 29 Apr 2024, Doug Anderson <dianders@chromium.org> wrote: > Hi, > > On Mon, Apr 29, 2024 at 2:38 AM Neil Armstrong > <neil.armstrong@linaro.org> wrote: >> >> > +/** >> > + * struct mipi_dsi_multi_context - Context to call multiple MIPI DSI funcs in a row >> > + * @dsi: Pointer to the MIPI DSI device >> > + * @accum_err: Storage for the accumulated error over the multiple calls. Init >> > + * to 0. If a function encounters an error then the error code will be >> > + * stored here. If you call a function and this points to a non-zero >> > + * value then the function will be a noop. This allows calling a function >> > + * many times in a row and just checking the error at the end to see if >> > + * any of them failed. >> > + */ >> > + >> > +struct mipi_dsi_multi_context { >> > + struct mipi_dsi_device *dsi; >> > + int accum_err; >> > +}; >> >> I like the design, but having a context struct seems over-engineered while we could pass >> a single int over without encapsulating it with mipi_dsi_multi_context. >> >> void mipi_dsi_dcs_write_buffer_multi(struct mipi_dsi_multi_context *ctx, >> const void *data, size_t len); >> vs >> void mipi_dsi_dcs_write_buffer_multi(struct mipi_dsi_device *dsi, int *accum_err, >> const void *data, size_t len); >> >> is the same, and it avoids having to declare a mipi_dsi_multi_context and set ctx->dsi, >> and I'll find it much easier to migrate, just add a &ret and make sure ret is initialized to 0. > > Yeah, I had the same reaction when Jani suggested the context style > [1] and I actually coded it up exactly as you suggest above. I then > changed my mind and went with the context. My motivation was that when > I tested it I found that using the context produced smaller code. > Specifically, from the description of this patch we see we end up > with: > > Total: Before=10651, After=9663, chg -9.28% > > ...when I didn't have the context and I had the accum_err then instead > of getting ~9% smaller I believe it actually got ~0.5% bigger. This > makes some sense as the caller has to pass 4 parameters to each call > instead of 3. > > It's not a giant size difference but it was at least some motivation > that helped push me in this direction. I'd also say that when I looked > at the code in the end the style grew on me. It's really not too > terrible to have one line in your functions that looks like: > > struct mipi_dsi_multi_context ctx = { .dsi = boe->dsi }; > > ...and if that becomes the canonical way to do it then it's really > hard to accidentally forget to initialize the error value. With the > other API it's _very_ easy to forget to initialize the error value and > the compiler won't yell at you. It also makes it very obvious to the > caller that this function is doing something a little different than > most Linux APIs with that error return. > > So I guess I'd say that I ended up being pretty happy with the > "context" even if it does feel a little over engineered and I'd argue > to keep it that way. That being said, if you feel strongly about it > then we can perhaps get others to chime in to see which style they > prefer? Let me know what you think. > > > [1] https://lore.kernel.org/r/8734r85tcf.fsf@intel.com FWIW, I don't feel strongly about this, and I could be persuaded either way, but I've got this gut feeling that an extensible context parameter might be benefitial future proofing in this case. BR, Jani.
Hi, On Mon, Apr 29, 2024 at 8:39 AM Jani Nikula <jani.nikula@linux.intel.com> wrote: > > On Mon, 29 Apr 2024, Doug Anderson <dianders@chromium.org> wrote: > > Hi, > > > > On Mon, Apr 29, 2024 at 2:38 AM Neil Armstrong > > <neil.armstrong@linaro.org> wrote: > >> > >> > +/** > >> > + * struct mipi_dsi_multi_context - Context to call multiple MIPI DSI funcs in a row > >> > + * @dsi: Pointer to the MIPI DSI device > >> > + * @accum_err: Storage for the accumulated error over the multiple calls. Init > >> > + * to 0. If a function encounters an error then the error code will be > >> > + * stored here. If you call a function and this points to a non-zero > >> > + * value then the function will be a noop. This allows calling a function > >> > + * many times in a row and just checking the error at the end to see if > >> > + * any of them failed. > >> > + */ > >> > + > >> > +struct mipi_dsi_multi_context { > >> > + struct mipi_dsi_device *dsi; > >> > + int accum_err; > >> > +}; > >> > >> I like the design, but having a context struct seems over-engineered while we could pass > >> a single int over without encapsulating it with mipi_dsi_multi_context. > >> > >> void mipi_dsi_dcs_write_buffer_multi(struct mipi_dsi_multi_context *ctx, > >> const void *data, size_t len); > >> vs > >> void mipi_dsi_dcs_write_buffer_multi(struct mipi_dsi_device *dsi, int *accum_err, > >> const void *data, size_t len); > >> > >> is the same, and it avoids having to declare a mipi_dsi_multi_context and set ctx->dsi, > >> and I'll find it much easier to migrate, just add a &ret and make sure ret is initialized to 0. > > > > Yeah, I had the same reaction when Jani suggested the context style > > [1] and I actually coded it up exactly as you suggest above. I then > > changed my mind and went with the context. My motivation was that when > > I tested it I found that using the context produced smaller code. > > Specifically, from the description of this patch we see we end up > > with: > > > > Total: Before=10651, After=9663, chg -9.28% > > > > ...when I didn't have the context and I had the accum_err then instead > > of getting ~9% smaller I believe it actually got ~0.5% bigger. This > > makes some sense as the caller has to pass 4 parameters to each call > > instead of 3. > > > > It's not a giant size difference but it was at least some motivation > > that helped push me in this direction. I'd also say that when I looked > > at the code in the end the style grew on me. It's really not too > > terrible to have one line in your functions that looks like: > > > > struct mipi_dsi_multi_context ctx = { .dsi = boe->dsi }; > > > > ...and if that becomes the canonical way to do it then it's really > > hard to accidentally forget to initialize the error value. With the > > other API it's _very_ easy to forget to initialize the error value and > > the compiler won't yell at you. It also makes it very obvious to the > > caller that this function is doing something a little different than > > most Linux APIs with that error return. > > > > So I guess I'd say that I ended up being pretty happy with the > > "context" even if it does feel a little over engineered and I'd argue > > to keep it that way. That being said, if you feel strongly about it > > then we can perhaps get others to chime in to see which style they > > prefer? Let me know what you think. > > > > > > [1] https://lore.kernel.org/r/8734r85tcf.fsf@intel.com > > FWIW, I don't feel strongly about this, and I could be persuaded either > way, but I've got this gut feeling that an extensible context parameter > might be benefitial future proofing in this case. I've gone ahead and sent out a v3 where I still leave it as taking `struct mipi_dsi_multi_context`. Neil: if you feel strongly about it, I have no problem sending out a v4. I still think that the size savings of using the context are a good "tiebreaking" factor in choosing between the two styles... ;-) -Doug
On Wed, 1 May 2024 at 18:49, Doug Anderson <dianders@chromium.org> wrote: > > Hi, > > On Mon, Apr 29, 2024 at 8:39 AM Jani Nikula <jani.nikula@linux.intel.com> wrote: > > > > On Mon, 29 Apr 2024, Doug Anderson <dianders@chromium.org> wrote: > > > Hi, > > > > > > On Mon, Apr 29, 2024 at 2:38 AM Neil Armstrong > > > <neil.armstrong@linaro.org> wrote: > > >> > > >> > +/** > > >> > + * struct mipi_dsi_multi_context - Context to call multiple MIPI DSI funcs in a row > > >> > + * @dsi: Pointer to the MIPI DSI device > > >> > + * @accum_err: Storage for the accumulated error over the multiple calls. Init > > >> > + * to 0. If a function encounters an error then the error code will be > > >> > + * stored here. If you call a function and this points to a non-zero > > >> > + * value then the function will be a noop. This allows calling a function > > >> > + * many times in a row and just checking the error at the end to see if > > >> > + * any of them failed. > > >> > + */ > > >> > + > > >> > +struct mipi_dsi_multi_context { > > >> > + struct mipi_dsi_device *dsi; > > >> > + int accum_err; > > >> > +}; > > >> > > >> I like the design, but having a context struct seems over-engineered while we could pass > > >> a single int over without encapsulating it with mipi_dsi_multi_context. > > >> > > >> void mipi_dsi_dcs_write_buffer_multi(struct mipi_dsi_multi_context *ctx, > > >> const void *data, size_t len); > > >> vs > > >> void mipi_dsi_dcs_write_buffer_multi(struct mipi_dsi_device *dsi, int *accum_err, > > >> const void *data, size_t len); > > >> > > >> is the same, and it avoids having to declare a mipi_dsi_multi_context and set ctx->dsi, > > >> and I'll find it much easier to migrate, just add a &ret and make sure ret is initialized to 0. > > > > > > Yeah, I had the same reaction when Jani suggested the context style > > > [1] and I actually coded it up exactly as you suggest above. I then > > > changed my mind and went with the context. My motivation was that when > > > I tested it I found that using the context produced smaller code. > > > Specifically, from the description of this patch we see we end up > > > with: > > > > > > Total: Before=10651, After=9663, chg -9.28% > > > > > > ...when I didn't have the context and I had the accum_err then instead > > > of getting ~9% smaller I believe it actually got ~0.5% bigger. This > > > makes some sense as the caller has to pass 4 parameters to each call > > > instead of 3. > > > > > > It's not a giant size difference but it was at least some motivation > > > that helped push me in this direction. I'd also say that when I looked > > > at the code in the end the style grew on me. It's really not too > > > terrible to have one line in your functions that looks like: > > > > > > struct mipi_dsi_multi_context ctx = { .dsi = boe->dsi }; > > > > > > ...and if that becomes the canonical way to do it then it's really > > > hard to accidentally forget to initialize the error value. With the > > > other API it's _very_ easy to forget to initialize the error value and > > > the compiler won't yell at you. It also makes it very obvious to the > > > caller that this function is doing something a little different than > > > most Linux APIs with that error return. > > > > > > So I guess I'd say that I ended up being pretty happy with the > > > "context" even if it does feel a little over engineered and I'd argue > > > to keep it that way. That being said, if you feel strongly about it > > > then we can perhaps get others to chime in to see which style they > > > prefer? Let me know what you think. > > > > > > > > > [1] https://lore.kernel.org/r/8734r85tcf.fsf@intel.com > > > > FWIW, I don't feel strongly about this, and I could be persuaded either > > way, but I've got this gut feeling that an extensible context parameter > > might be benefitial future proofing in this case. > > I've gone ahead and sent out a v3 where I still leave it as taking > `struct mipi_dsi_multi_context`. Neil: if you feel strongly about it, > I have no problem sending out a v4. I still think that the size > savings of using the context are a good "tiebreaking" factor in > choosing between the two styles... ;-) I like the idea of context. If later on we need to add additional data (like DSC PPS or drm_mode), we can simply extend the context structure. > > -Doug
On Mon, Apr 29, 2024 at 4:26 PM Doug Anderson <dianders@chromium.org> wrote: > So I guess I'd say that I ended up being pretty happy with the > "context" even if it does feel a little over engineered and I'd argue > to keep it that way. That being said, if you feel strongly about it > then we can perhaps get others to chime in to see which style they > prefer? Let me know what you think. I'm in favor of this design. Reviewed-by: Linus Walleij <linus.walleij@linaro.org> Yours, Linus Walleij
diff --git a/drivers/gpu/drm/drm_mipi_dsi.c b/drivers/gpu/drm/drm_mipi_dsi.c index 1e062eda3b88..b7c75a4396e6 100644 --- a/drivers/gpu/drm/drm_mipi_dsi.c +++ b/drivers/gpu/drm/drm_mipi_dsi.c @@ -792,6 +792,34 @@ int mipi_dsi_generic_write_chatty(struct mipi_dsi_device *dsi, } EXPORT_SYMBOL(mipi_dsi_generic_write_chatty); +/** + * mipi_dsi_generic_write_multi() - mipi_dsi_generic_write_chatty() w/ accum_err + * @ctx: Context for multiple DSI transactions + * @payload: buffer containing the payload + * @size: size of payload buffer + * + * Like mipi_dsi_generic_write_chatty() but deals with errors in a way that + * makes it convenient to make several calls in a row. + */ +void mipi_dsi_generic_write_multi(struct mipi_dsi_multi_context *ctx, + const void *payload, size_t size) +{ + struct mipi_dsi_device *dsi = ctx->dsi; + struct device *dev = &dsi->dev; + ssize_t ret; + + if (ctx->accum_err) + return; + + ret = mipi_dsi_generic_write(dsi, payload, size); + if (ret < 0) { + ctx->accum_err = ret; + dev_err_ratelimited(dev, "sending generic data %*ph failed: %d\n", + (int)size, payload, ctx->accum_err); + } +} +EXPORT_SYMBOL(mipi_dsi_generic_write_multi); + /** * mipi_dsi_generic_read() - receive data using a generic read packet * @dsi: DSI peripheral device @@ -908,6 +936,34 @@ int mipi_dsi_dcs_write_buffer_chatty(struct mipi_dsi_device *dsi, } EXPORT_SYMBOL(mipi_dsi_dcs_write_buffer_chatty); +/** + * mipi_dsi_dcs_write_buffer_multi - mipi_dsi_dcs_write_buffer_chatty() w/ accum_err + * @ctx: Context for multiple DSI transactions + * @data: buffer containing data to be transmitted + * @len: size of transmission buffer + * + * Like mipi_dsi_dcs_write_buffer_chatty() but deals with errors in a way that + * makes it convenient to make several calls in a row. + */ +void mipi_dsi_dcs_write_buffer_multi(struct mipi_dsi_multi_context *ctx, + const void *data, size_t len) +{ + struct mipi_dsi_device *dsi = ctx->dsi; + struct device *dev = &dsi->dev; + ssize_t ret; + + if (ctx->accum_err) + return; + + ret = mipi_dsi_dcs_write_buffer(dsi, data, len); + if (ret < 0) { + ctx->accum_err = ret; + dev_err_ratelimited(dev, "sending dcs data %*ph failed: %d\n", + (int)len, data, ctx->accum_err); + } +} +EXPORT_SYMBOL(mipi_dsi_dcs_write_buffer_multi); + /** * mipi_dsi_dcs_write() - send DCS write command * @dsi: DSI peripheral device diff --git a/include/drm/drm_mipi_dsi.h b/include/drm/drm_mipi_dsi.h index 6d68d9927f46..379594649a42 100644 --- a/include/drm/drm_mipi_dsi.h +++ b/include/drm/drm_mipi_dsi.h @@ -197,6 +197,22 @@ struct mipi_dsi_device { struct drm_dsc_config *dsc; }; +/** + * struct mipi_dsi_multi_context - Context to call multiple MIPI DSI funcs in a row + * @dsi: Pointer to the MIPI DSI device + * @accum_err: Storage for the accumulated error over the multiple calls. Init + * to 0. If a function encounters an error then the error code will be + * stored here. If you call a function and this points to a non-zero + * value then the function will be a noop. This allows calling a function + * many times in a row and just checking the error at the end to see if + * any of them failed. + */ + +struct mipi_dsi_multi_context { + struct mipi_dsi_device *dsi; + int accum_err; +}; + #define MIPI_DSI_MODULE_PREFIX "mipi-dsi:" #define to_mipi_dsi_device(__dev) container_of_const(__dev, struct mipi_dsi_device, dev) @@ -258,6 +274,8 @@ ssize_t mipi_dsi_generic_write(struct mipi_dsi_device *dsi, const void *payload, size_t size); int mipi_dsi_generic_write_chatty(struct mipi_dsi_device *dsi, const void *payload, size_t size); +void mipi_dsi_generic_write_multi(struct mipi_dsi_multi_context *ctx, + const void *payload, size_t size); ssize_t mipi_dsi_generic_read(struct mipi_dsi_device *dsi, const void *params, size_t num_params, void *data, size_t size); @@ -283,6 +301,8 @@ ssize_t mipi_dsi_dcs_write_buffer(struct mipi_dsi_device *dsi, const void *data, size_t len); int mipi_dsi_dcs_write_buffer_chatty(struct mipi_dsi_device *dsi, const void *data, size_t len); +void mipi_dsi_dcs_write_buffer_multi(struct mipi_dsi_multi_context *ctx, + const void *data, size_t len); ssize_t mipi_dsi_dcs_write(struct mipi_dsi_device *dsi, u8 cmd, const void *data, size_t len); ssize_t mipi_dsi_dcs_read(struct mipi_dsi_device *dsi, u8 cmd, void *data, @@ -319,6 +339,9 @@ int mipi_dsi_dcs_get_display_brightness_large(struct mipi_dsi_device *dsi, * This macro will print errors for you and will RETURN FROM THE CALLING * FUNCTION (yes this is non-intuitive) upon error. * + * Because of the non-intuitive return behavior, THIS MACRO IS DEPRECATED. + * Please replace calls of it with mipi_dsi_generic_write_seq_multi(). + * * @dsi: DSI peripheral device * @seq: buffer containing the payload */ @@ -331,12 +354,30 @@ int mipi_dsi_dcs_get_display_brightness_large(struct mipi_dsi_device *dsi, return ret; \ } while (0) +/** + * mipi_dsi_generic_write_seq_multi - transmit data using a generic write packet + * + * This macro will print errors for you and error handling is optimized for + * callers that call this multiple times in a row. + * + * @ctx: Context for multiple DSI transactions + * @seq: buffer containing the payload + */ +#define mipi_dsi_generic_write_seq_multi(ctx, seq...) \ + do { \ + static const u8 d[] = { seq }; \ + mipi_dsi_generic_write_multi(ctx, d, ARRAY_SIZE(d)); \ + } while (0) + /** * mipi_dsi_dcs_write_seq - transmit a DCS command with payload * * This macro will print errors for you and will RETURN FROM THE CALLING * FUNCTION (yes this is non-intuitive) upon error. * + * Because of the non-intuitive return behavior, THIS MACRO IS DEPRECATED. + * Please replace calls of it with mipi_dsi_dcs_write_seq_multi(). + * * @dsi: DSI peripheral device * @cmd: Command * @seq: buffer containing data to be transmitted @@ -350,6 +391,22 @@ int mipi_dsi_dcs_get_display_brightness_large(struct mipi_dsi_device *dsi, return ret; \ } while (0) +/** + * mipi_dsi_dcs_write_seq_multi - transmit a DCS command with payload + * + * This macro will print errors for you and error handling is optimized for + * callers that call this multiple times in a row. + * + * @ctx: Context for multiple DSI transactions + * @cmd: Command + * @seq: buffer containing data to be transmitted + */ +#define mipi_dsi_dcs_write_seq_multi(ctx, cmd, seq...) \ + do { \ + static const u8 d[] = { cmd, seq }; \ + mipi_dsi_dcs_write_buffer_multi(ctx, d, ARRAY_SIZE(d)); \ + } while (0) + /** * struct mipi_dsi_driver - DSI driver * @driver: device driver model driver
The current mipi_dsi_*_write_seq() macros are non-intutitive because they contain a hidden "return" statement that will return out of the _caller_ of the macro. Let's mark them as deprecated and instead introduce some new macros that are more intuitive. These new macros are less optimal when an error occurs but should behave more optimally when there is no error. Specifically these new macros cause smaller code to get generated and the code size savings (less to fetch from RAM, less cache space used, less RAM used) are important. Since the error case isn't something we need to optimize for and these new macros are easier to understand and more flexible, they should be used. After converting to use these new functions, one example shows some nice savings while also being easier to understand. $ scripts/bloat-o-meter \ ...after/panel-novatek-nt36672e.ko \ ...ctx/panel-novatek-nt36672e.ko add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-988 (-988) Function old new delta nt36672e_1080x2408_60hz_init 6236 5248 -988 Total: Before=10651, After=9663, chg -9.28% Signed-off-by: Douglas Anderson <dianders@chromium.org> --- Right now this patch introduces two new functions in drm_mipi_dsi.c. Alternatively we could have changed the prototype of the "chatty" functions and made the deprecated macros adapt to the new prototype. While this sounds nice, it bloated callers of the deprecated functioin a bit because it caused the compiler to emit less optimal code. It doesn't seem terrible to add two more functions, so I went that way. Changes in v2: - New drivers/gpu/drm/drm_mipi_dsi.c | 56 +++++++++++++++++++++++++++++++++ include/drm/drm_mipi_dsi.h | 57 ++++++++++++++++++++++++++++++++++ 2 files changed, 113 insertions(+)