[Mesa-dev] [PATCH 3/3] anv: Disable constant buffer 0 being relative.

Jason Ekstrand jason at jlekstrand.net
Fri Jun 15 20:21:17 UTC 2018


On Fri, Jun 15, 2018 at 1:12 PM, Rafael Antognolli <
rafael.antognolli at intel.com> wrote:

> If we are on gen8+ and have context isolation support, just make that
> constant buffer address be absolute, so we can use it for push UBOs too.
> ---
>  src/intel/vulkan/anv_device.c  |  5 ++++-
>  src/intel/vulkan/anv_private.h |  1 +
>  src/intel/vulkan/genX_state.c  | 27 +++++++++++++++++++++++++++
>  3 files changed, 32 insertions(+), 1 deletion(-)
>
> diff --git a/src/intel/vulkan/anv_device.c b/src/intel/vulkan/anv_device.c
> index d1637f097e8..002b05f15f8 100644
> --- a/src/intel/vulkan/anv_device.c
> +++ b/src/intel/vulkan/anv_device.c
> @@ -430,7 +430,8 @@ anv_physical_device_init(struct anv_physical_device
> *device,
>     device->compiler->shader_debug_log = compiler_debug_log;
>     device->compiler->shader_perf_log = compiler_perf_log;
>     device->compiler->supports_pull_constants = false;
> -   device->compiler->constant_buffer_0_is_relative = true;
> +   device->compiler->constant_buffer_0_is_relative =
> +      device->info.gen < 8 || !device->has_context_isolation;
>
>     isl_device_init(&device->isl_dev, &device->info, swizzled);
>
> @@ -1519,6 +1520,8 @@ VkResult anv_CreateDevice(
>     device->chipset_id = physical_device->chipset_id;
>     device->no_hw = physical_device->no_hw;
>     device->lost = false;
> +   device->constant_buffer_0_is_relative =
> +      physical_device->compiler->constant_buffer_0_is_relative;
>
>     if (pAllocator)
>        device->alloc = *pAllocator;
> diff --git a/src/intel/vulkan/anv_private.h b/src/intel/vulkan/anv_
> private.h
> index 6af2a0bd3f3..d7297da6f57 100644
> --- a/src/intel/vulkan/anv_private.h
> +++ b/src/intel/vulkan/anv_private.h
> @@ -983,6 +983,7 @@ struct anv_device {
>      pthread_mutex_t                             mutex;
>      pthread_cond_t                              queue_submit;
>      bool                                        lost;
> +    bool
> constant_buffer_0_is_relative;
>

I don't think we need a device boolean for this.  Let's just dig it out of
the compiler or the physical device.  I think the physical device probably
makes a bit more sense.  Other than that, the series looks good to me.

--Jason


>  };
>
>  static inline struct anv_state_pool *
> diff --git a/src/intel/vulkan/genX_state.c b/src/intel/vulkan/genX_state.c
> index c6e54046910..46e020c8134 100644
> --- a/src/intel/vulkan/genX_state.c
> +++ b/src/intel/vulkan/genX_state.c
> @@ -169,6 +169,33 @@ genX(init_device_state)(struct anv_device *device)
>     gen10_emit_wa_lri_to_cache_mode_zero(&batch);
>  #endif
>
> +   /* Set the "CONSTANT_BUFFER Address Offset Disable" bit, so
> +    * 3DSTATE_CONSTANT_XS buffer 0 is an absolute address.
> +    *
> +    * This is only safe on kernels with context isolation support.
> +    */
> +   if (GEN_GEN >= 8 &&
> +       !device->constant_buffer_0_is_relative) {
> +      UNUSED uint32_t tmp_reg;
> +#if GEN_GEN >= 9
> +      anv_pack_struct(&tmp_reg, GENX(CS_DEBUG_MODE2),
> +                      .CONSTANT_BUFFERAddressOffsetDisable = true,
> +                      .CONSTANT_BUFFERAddressOffsetDisableMask = true);
> +      anv_batch_emit(&batch, GENX(MI_LOAD_REGISTER_IMM), lri) {
> +         lri.RegisterOffset = GENX(CS_DEBUG_MODE2_num);
> +         lri.DataDWord      = tmp_reg;
> +      }
> +#elif GEN_GEN == 8
> +      anv_pack_struct(&tmp_reg, GENX(INSTPM),
> +                      .CONSTANT_BUFFERAddressOffsetDisable = true,
> +                      .CONSTANT_BUFFERAddressOffsetDisableMask = true);
> +      anv_batch_emit(&batch, GENX(MI_LOAD_REGISTER_IMM), lri) {
> +         lri.RegisterOffset = GENX(INSTPM_num);
> +         lri.DataDWord      = tmp_reg;
> +      }
> +#endif
> +   }
> +
>     anv_batch_emit(&batch, GENX(MI_BATCH_BUFFER_END), bbe);
>
>     assert(batch.next <= batch.end);
> --
> 2.14.3
>
> _______________________________________________
> mesa-dev mailing list
> mesa-dev at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/mesa-dev
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.freedesktop.org/archives/mesa-dev/attachments/20180615/40b8f2cc/attachment-0001.html>


More information about the mesa-dev mailing list