The Performance Counters enum used to identify the index of each
performance counter and provide the total number of performance
counters (V3D_PERFCNT_NUM). But, this enum is only valid for V3D 4.2,
not for V3D 7.1.
As we implemented a new flexible structure to retrieve performance
counters information, we can deprecate this enum.
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240512222655.2792754-7-mcanal@igalia.com
V3D_PERFCNT_NUM represents the maximum number of performance counters
for V3D 4.2, but not for V3D 7.1. This means that, if we use
V3D_PERFCNT_NUM, we might go out-of-bounds on V3D 7.1.
Therefore, use the number of performance counters on V3D 7.1 as the
maximum number of counters. This will allow us to create arrays on the
stack with reasonable size. Note that userspace must use the value
provided by DRM_V3D_PARAM_MAX_PERF_COUNTERS.
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240512222655.2792754-6-mcanal@igalia.com
Userspace usually needs some information about the performance counters
available. Although we could replicate this information in the kernel
and user-space, let's use the kernel as the "single source of truth" to
avoid issues in the future (e.g. list of performance counters is updated
in user-space, but not in the kernel, generating invalid requests).
Therefore, create a new IOCTL to expose the performance counters
information, that is name, category, and description.
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240512222655.2792754-5-mcanal@igalia.com
Currently, even though V3D 7.1 has 93 performance counters, it is not
possible to create counters bigger than 87, as
`v3d_perfmon_create_ioctl()` understands that counters bigger than 87
are invalid.
Therefore, create a device variable to expose the maximum
number of counters for a given V3D version and make
`v3d_perfmon_create_ioctl()` check this variable.
This commit fixes CTS failures in the performance queries tests
`dEQP-VK.query_pool.performance_query.*` [1]
Link: ea1f09a5f2 [1]
Fixes: 6fd9487147 ("drm/v3d: add brcm,2712-v3d as a compatible V3D device")
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240512222655.2792754-3-mcanal@igalia.com
Add name, category and description for each one of the 93 performance
counters available on V3D.
Note that V3D 4.2 has 87 performance counters, while V3D 7.1 has 93.
Therefore, there are two performance counters arrays. The index of the
performance counter for each V3D version is represented by its position
on the array.
Signed-off-by: Maíra Canal <mcanal@igalia.com>
Reviewed-by: Iago Toral Quiroga <itoral@igalia.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240512222655.2792754-2-mcanal@igalia.com
The modeset lock protects the DDC code from concurrent modeset
operations, which use the same registers. Move that code from the
connector helpers into the DDC helpers .pre_xfer() and .post_xfer().
Both, .pre_xfer() and .post_xfer(), enclose the transfer of data blocks
over the I2C channel in the internal I2C function bit_xfer(). Both
calls are executed unconditionally if present. Invoking DDC transfers
from any where within the driver now takes the lock.
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Reviewed-by: Jocelyn Falempe <jfalempe@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240513125620.6337-10-tzimmermann@suse.de
Rename struct mga_i2c_chan to struct mgag200_ddc, define it in the
source file mgag200_i2c.c, and reorder its fields. Rename all related
variables from i2c to ddc. Also rename the i2c adapter accordingly.
Using the term 'ddc' documents the purpose of the code clearly. The
old term 'i2c' could refer to any functionality on an i2c bus. No
functional changes.
Signed-off-by: Thomas Zimmermann <tzimmermann@suse.de>
Reviewed-by: Jocelyn Falempe <jfalempe@redhat.com>
Link: https://patchwork.freedesktop.org/patch/msgid/20240513125620.6337-7-tzimmermann@suse.de
Consensus on the mailing lists is that panels shouldn't use a table of
init commands but should instead use init functions. We'll use the
same concepts as the recently introduced
mipi_dsi_generic_write_seq_multi() to make this clean/easy and also
not bloat the driver too much. Measuring before/after this change:
$ scripts/bloat-o-meter \
.../before/panel-innolux-p079zca.ko \
.../after/panel-innolux-p079zca.ko
add/remove: 3/2 grow/shrink: 0/1 up/down: 2356/-1944 (412)
Function old new delta
innolux_p097pfg_init - 1772 +1772
innolux_p097pfg_init.d - 480 +480
innolux_panel_write_multi - 104 +104
innolux_panel_prepare 412 308 -104
.compoundliteral 480 - -480
innolux_p097pfg_init_cmds 1360 - -1360
Total: Before=5802, After=6214, chg +7.10%
Note that, unlike some other drivers, we actually make this panel
driver _bigger_ by using the new functions. This is because the
innolux-p079zca panel driver didn't have as complex of a table and
thus the old table was more efficient than the code. The bloat is
still not giant (only 412 bytes).
Also note that we can't direclty use
mipi_dsi_generic_write_seq_multi() here because we need to deal with
the crazy "nop" that this driver sends after all commands. This means
that we have to write code that is "inspired" by the new macros.
Since we're touching all the tables, let's also convert hex numbers to
lower case as per kernel conventions.
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Link: https://lore.kernel.org/r/20240514102056.v5.9.I947e28c81f9ef7dcd3add6e193be72d6f8ea086f@changeid
Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20240514102056.v5.9.I947e28c81f9ef7dcd3add6e193be72d6f8ea086f@changeid
Consensus on the mailing lists is that panels shouldn't use a table of
init commands but should instead use init functions. With the recently
introduced mipi_dsi_dcs_write_seq_multi() this is not only clean/easy
but also saves space. Measuring before/after this change:
$ scripts/bloat-o-meter \
.../before/panel-ilitek-ili9882t.ko \
.../after/panel-ilitek-ili9882t.ko
add/remove: 3/2 grow/shrink: 0/2 up/down: 6834/-8177 (-1343)
Function old new delta
starry_ili9882t_init - 6152 +6152
starry_ili9882t_init.d - 678 +678
ili9882t_disable.d - 4 +4
ili9882t_disable 260 228 -32
ili9882t_prepare 540 396 -144
.compoundliteral 681 - -681
starry_ili9882t_init_cmd 7320 - -7320
Total: Before=11928, After=10585, chg -11.26%
Let's do the conversion.
Since we're touching all the tables, let's also convert hex numbers to
lower case as per kernel conventions.
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Link: https://lore.kernel.org/r/20240514102056.v5.8.I407b977f9508573103491df549dfaffc3688db13@changeid
Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20240514102056.v5.8.I407b977f9508573103491df549dfaffc3688db13@changeid
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%
Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Link: https://lore.kernel.org/r/20240514102056.v5.5.Ie94246c30fe95101e0e26dd5f96e976dbeb8f242@changeid
Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20240514102056.v5.5.Ie94246c30fe95101e0e26dd5f96e976dbeb8f242@changeid
Through a cooperative effort between Hsin-Yi Wang and Dmitry
Baryshkov, we have realized the dev_err() in the
mipi_dsi_*_write_seq() macros was causing quite a bit of bloat to the
kernel. Let's hoist this call into drm_mipi_dsi.c by adding a "chatty"
version of the functions that includes the print. While doing this,
add a bit more comments to these macros making it clear that they
print errors and also that they return out of _the caller's_ function.
Without any changes to clients this gives a nice savings. Specifically
the macro was inlined and thus the error report call was inlined into
every call to mipi_dsi_dcs_write_seq() and
mipi_dsi_generic_write_seq(). By using a call to a "chatty" function,
the usage is reduced to one call in the chatty function and a function
call at the invoking site.
Building with my build system shows one example:
$ scripts/bloat-o-meter \
.../before/panel-novatek-nt36672e.ko \
.../after/panel-novatek-nt36672e.ko
add/remove: 0/0 grow/shrink: 0/1 up/down: 0/-4404 (-4404)
Function old new delta
nt36672e_1080x2408_60hz_init 10640 6236 -4404
Total: Before=15055, After=10651, chg -29.25%
Note that given the change in location of the print it's harder to
include the "cmd" in the printout for mipi_dsi_dcs_write_seq() since,
theoretically, someone could call the new chatty function with a
zero-size array and it would be illegal to dereference data[0].
There's a printk format to print the whole buffer and this is probably
more useful for debugging anyway. Given that we're doing this for
mipi_dsi_dcs_write_seq(), let's also print the buffer for
mipi_dsi_generic_write_seq() in the error case.
It should be noted that the current consensus of DRM folks is that the
mipi_dsi_*_write_seq() should be deprecated due to the non-intuitive
return behavior. A future patch will formally mark them as deprecated
and provide an alternative.
Reviewed-by: Dmitry Baryshkov <dmitry.baryshkov@linaro.org>
Reviewed-by: Neil Armstrong <neil.armstrong@linaro.org>
Reviewed-by: Linus Walleij <linus.walleij@linaro.org>
Signed-off-by: Douglas Anderson <dianders@chromium.org>
Link: https://lore.kernel.org/r/20240514102056.v5.4.Id15fae80582bc74a0d4f1338987fa375738f45b9@changeid
Signed-off-by: Neil Armstrong <neil.armstrong@linaro.org>
Link: https://patchwork.freedesktop.org/patch/msgid/20240514102056.v5.4.Id15fae80582bc74a0d4f1338987fa375738f45b9@changeid
