[Mesa-dev] [PATCH 2/3] radv: Use a lower max offchip buffer count.
Samuel Pitoiset
samuel.pitoiset at gmail.com
Mon Sep 3 07:10:17 UTC 2018
On 9/3/18 2:35 AM, Bas Nieuwenhuizen wrote:
> No clue what gets fixed by this but both radeonsi and amdvlk do it.
>
> CC: <mesa-stable at lists.freedesktop.org>
> ---
> src/amd/vulkan/radv_device.c | 24 ++++++++++++++++++++++--
> 1 file changed, 22 insertions(+), 2 deletions(-)
>
> diff --git a/src/amd/vulkan/radv_device.c b/src/amd/vulkan/radv_device.c
> index c300c88468a..3d208ab053d 100644
> --- a/src/amd/vulkan/radv_device.c
> +++ b/src/amd/vulkan/radv_device.c
> @@ -1901,10 +1901,30 @@ radv_get_hs_offchip_param(struct radv_device *device, uint32_t *max_offchip_buff
> device->physical_device->rad_info.family != CHIP_CARRIZO &&
> device->physical_device->rad_info.family != CHIP_STONEY;
> unsigned max_offchip_buffers_per_se = double_offchip_buffers ? 128 : 64;
> - unsigned max_offchip_buffers = max_offchip_buffers_per_se *
> - device->physical_device->rad_info.max_se;
> + unsigned max_offchip_buffers;
> unsigned offchip_granularity;
> unsigned hs_offchip_param;
> +
> + /*
> + * Per RadeonSI:
> + * This must be one less than the maximum number due to a hw limitation.
> + * Various hardware bugs in SI, CIK, and GFX9 need this.
> + *
> + * Per AMDVLK:
> + * Vega10 should limit max_offchip_buffers to 508 (4 * 127).
> + * Gfx7 should limit max_offchip_buffers to 508
> + * Gfx6 should limit max_offchip_buffers to 126 (2 * 63)
> + *
> + * Follow AMDVLK here.
> + */
Can you adjust the indentation here please?
Otherwise, series is:
Reviewed-by: Samuel Pitoiset <samuel.pitoiset at gmail.com>
> + if (device->physical_device->rad_info.family == CHIP_VEGA10 ||
> + device->physical_device->rad_info.chip_class == CIK ||
> + device->physical_device->rad_info.chip_class == SI)
> + --max_offchip_buffers_per_se;
> +
> + max_offchip_buffers = max_offchip_buffers_per_se *
> + device->physical_device->rad_info.max_se;
> +
> switch (device->tess_offchip_block_dw_size) {
> default:
> assert(0);
>
More information about the mesa-dev
mailing list