[PATCH RFC 6/8] drm/msm: adreno: enable GMU bandwidth for A740 and A750
Neil Armstrong
neil.armstrong at linaro.org
Fri Nov 15 09:20:01 UTC 2024
On 15/11/2024 08:33, Dmitry Baryshkov wrote:
> On Wed, Nov 13, 2024 at 04:48:32PM +0100, Neil Armstrong wrote:
>> Now all the DDR bandwidth voting via the GPU Management Unit (GMU)
>> is in place, let's declare the Bus Control Modules (BCMs) and
>
> s/let's //g
>
>> it's parameters in the GPU info struct and add the GMU_BW_VOTE
>> quirk to enable it.
>
> Can we define a function that checks for info.bcm[0].name isntead of
> adding a quirk?
Probably, I'll need ideas to how design this better, perhaps a simple
capability bitfield in a6xx_info ?
There's other feature that are lacking, like ACD or BCL which are not supported
on all a6xx/a7xx gpus.
>
>>
>> Signed-off-by: Neil Armstrong <neil.armstrong at linaro.org>
>> ---
>> drivers/gpu/drm/msm/adreno/a6xx_catalog.c | 26 ++++++++++++++++++++++++--
>> 1 file changed, 24 insertions(+), 2 deletions(-)
>>
>> diff --git a/drivers/gpu/drm/msm/adreno/a6xx_catalog.c b/drivers/gpu/drm/msm/adreno/a6xx_catalog.c
>> index 0c560e84ad5a53bb4e8a49ba4e153ce9cf33f7ae..014a24256b832d8e03fe06a6516b5348a5c0474a 100644
>> --- a/drivers/gpu/drm/msm/adreno/a6xx_catalog.c
>> +++ b/drivers/gpu/drm/msm/adreno/a6xx_catalog.c
>> @@ -1379,7 +1379,8 @@ static const struct adreno_info a7xx_gpus[] = {
>> .inactive_period = DRM_MSM_INACTIVE_PERIOD,
>> .quirks = ADRENO_QUIRK_HAS_CACHED_COHERENT |
>> ADRENO_QUIRK_HAS_HW_APRIV |
>> - ADRENO_QUIRK_PREEMPTION,
>> + ADRENO_QUIRK_PREEMPTION |
>> + ADRENO_QUIRK_GMU_BW_VOTE,
>> .init = a6xx_gpu_init,
>> .zapfw = "a740_zap.mdt",
>> .a6xx = &(const struct a6xx_info) {
>> @@ -1388,6 +1389,16 @@ static const struct adreno_info a7xx_gpus[] = {
>> .pwrup_reglist = &a7xx_pwrup_reglist,
>> .gmu_chipid = 0x7020100,
>> .gmu_cgc_mode = 0x00020202,
>> + .bcm = {
>> + [0] = { .name = "SH0", .buswidth = 16 },
>> + [1] = { .name = "MC0", .buswidth = 4 },
>> + [2] = {
>> + .name = "ACV",
>> + .fixed = true,
>> + .perfmode = BIT(3),
>> + .perfmode_bw = 16500000,
>
> Is it a platform property or GPU / GMU property? Can expect that there
> might be several SoCs having the same GPU, but different perfmode_bw
> entry?
I presume this is SoC specific ? But today the XXX_build_bw_table() are
already SoC specific, so where should this go ?
Downstream specifies this in the adreno-gpulist.h, which is the equivalent
here.
Neil
>
>> + },
>> + },
>> },
>> .address_space_size = SZ_16G,
>> .preempt_record_size = 4192 * SZ_1K,
>> @@ -1424,7 +1435,8 @@ static const struct adreno_info a7xx_gpus[] = {
>> .inactive_period = DRM_MSM_INACTIVE_PERIOD,
>> .quirks = ADRENO_QUIRK_HAS_CACHED_COHERENT |
>> ADRENO_QUIRK_HAS_HW_APRIV |
>> - ADRENO_QUIRK_PREEMPTION,
>> + ADRENO_QUIRK_PREEMPTION |
>> + ADRENO_QUIRK_GMU_BW_VOTE,
>> .init = a6xx_gpu_init,
>> .zapfw = "gen70900_zap.mbn",
>> .a6xx = &(const struct a6xx_info) {
>> @@ -1432,6 +1444,16 @@ static const struct adreno_info a7xx_gpus[] = {
>> .pwrup_reglist = &a7xx_pwrup_reglist,
>> .gmu_chipid = 0x7090100,
>> .gmu_cgc_mode = 0x00020202,
>> + .bcm = {
>> + [0] = { .name = "SH0", .buswidth = 16 },
>> + [1] = { .name = "MC0", .buswidth = 4 },
>> + [2] = {
>> + .name = "ACV",
>> + .fixed = true,
>> + .perfmode = BIT(2),
>> + .perfmode_bw = 10687500,
>> + },
>> + },
>> },
>> .address_space_size = SZ_16G,
>> .preempt_record_size = 3572 * SZ_1K,
>>
>> --
>> 2.34.1
>>
>
More information about the dri-devel
mailing list