anv: Implement VK_ANDROID_native_buffer (v9)
This implementation is correct (afaict), but takes two shortcuts
regarding the import/export of Android sync fds.
Shortcut 1. When Android calls vkAcquireImageANDROID to import a sync
fd into a VkSemaphore or VkFence, the driver instead simply blocks on
the sync fd, then puts the VkSemaphore or VkFence into the signalled
state. Thanks to implicit sync, this produces correct behavior (with
extra latency overhead, perhaps) despite its ugliness.
Shortcut 2. When Android calls vkQueueSignalReleaseImageANDROID to export
a collection of wait semaphores as a sync fd, the driver instead
submits the semaphores to the queue, then returns sync fd -1, which
informs the caller that no additional synchronization is needed.
Again, thanks to implicit sync, this produces correct behavior (with
extra batch submission overhead) despite its ugliness.
I chose to take the shortcuts instead of properly importing/exporting
the sync fds for two reasons:
Reason 1. I've already tested this patch with dEQP and with demos
apps. It works. I wanted to get the tested patches into the tree now,
and polish the implementation afterwards.
Reason 2. I want to run this on a 3.18 kernel (gasp!). In 3.18, i915
supports neither Android's sync_fence, nor upstream's sync_file, nor
drm_syncobj. Again, I tested these patches on Android with a 3.18
kernel and they work.
I plan to quickly follow-up with patches that remove the shortcuts and
properly import/export the sync fds.
Non-Testing
===========
I did not test at all using the Android.mk buildsystem. I may have broke
it. Please test and review that.
Testing
=======
I tested with 64-bit ARC++ on a Skylake Chromebook and a 3.18 kernel.
The following pass (as of patchset v9):
- a little spinning cube demo APK
- several Sascha demos
- dEQP-VK.info.*
- dEQP-VK.api.wsi.android.*
(except dEQP-VK.api.wsi.android.swapchain.*.image_usage, because
dEQP wants to create swapchains with VK_IMAGE_USAGE_STORAGE_BIT)
- dEQP-VK.api.smoke.*
- dEQP-VK.api.info.instance.*
- dEQP-VK.api.info.device.*
v2:
- Reject VkNativeBufferANDROID if the dma-buf's size is too small for
the VkImage.
- Stop abusing VkNativeBufferANDROID by passing it to vkAllocateMemory
during vkCreateImage. Instead, directly import its dma-buf during
vkCreateImage with anv_bo_cache_import(). [for jekstrand]
- Rebase onto Tapani's VK_EXT_debug_report changes.
- Drop `CPPFLAGS += $(top_srcdir)/include/android`. The dir does not
exist.
v3:
- Delete duplicate #include "anv_private.h". [per Tapani]
- Try to fix the Android-IA build in Android.vulkan.mk by following
Tapani's example.
v4:
- Unset EXEC_OBJECT_ASYNC and set EXEC_OBJECT_WRITE on the imported
gralloc buffer, just as we do for all other winsys buffers in
anv_wsi.c. [found by Tapani]
v5:
- Really fix the Android-IA build by ensuring that Android.vulkan.mk
uses Mesa' vulkan.h and not Android's. Insert -I$(MESA_TOP)/include
before -Iframeworks/native/vulkan/include. [for Tapani]
- In vkAcquireImageANDROID, submit signal operations to the
VkSemaphore and VkFence. [for zhou]
v6:
- Drop copy-paste duplication in vkGetSwapchainGrallocUsageANDROID().
[found by zhou]
- Improve comments in vkGetSwapchainGrallocUsageANDROID().
v7:
- Fix vkGetSwapchainGrallocUsageANDROID() to inspect its
VkImageUsageFlags parameter. [for tfiga]
- This fix regresses dEQP-VK.api.wsi.android.swapchain.*.image_usage
because dEQP wants to create swapchains with
VK_IMAGE_USAGE_STORAGE_BIT.
v8:
- Drop unneeded goto in vkAcquireImageANDROID. [for tfiga]
v8.1: (minor changes)
- Drop errant hunks added by rerere in anv_device.c.
- Drop explicit mention of VK_ANDROID_native_buffer in
anv_entrypoints_gen.py. [for jekstrand]
v9:
- Isolate as much Android code as possible, moving it from anv_image.c
to anv_android.c. Connect the files with anv_image_from_gralloc().
Remove VkNativeBufferANDROID params from all anv_image.c
funcs. [for krh]
- Replace some intel_loge() with vk_errorf() in anv_android.c.
- Use © in copyright line. [for krh]
Reviewed-by: Tapani Pälli <tapani.palli@intel.com> (v5)
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com> (v9)
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net> (v9)
Cc: zhoucm1 <david1.zhou@amd.com>
Cc: Tomasz Figa <tfiga@chromium.org>
2016-11-14 16:13:51 -08:00
|
|
|
/*
|
|
|
|
|
* Copyright © 2017, Google Inc.
|
|
|
|
|
*
|
|
|
|
|
* Permission is hereby granted, free of charge, to any person obtaining a
|
|
|
|
|
* copy of this software and associated documentation files (the "Software"),
|
|
|
|
|
* to deal in the Software without restriction, including without limitation
|
|
|
|
|
* the rights to use, copy, modify, merge, publish, distribute, sublicense,
|
|
|
|
|
* and/or sell copies of the Software, and to permit persons to whom the
|
|
|
|
|
* Software is furnished to do so, subject to the following conditions:
|
|
|
|
|
*
|
|
|
|
|
* The above copyright notice and this permission notice (including the next
|
|
|
|
|
* paragraph) shall be included in all copies or substantial portions of the
|
|
|
|
|
* Software.
|
|
|
|
|
*
|
|
|
|
|
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
|
|
|
|
|
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
|
|
|
|
|
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL
|
|
|
|
|
* THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
|
|
|
|
|
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING
|
|
|
|
|
* FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS
|
|
|
|
|
* IN THE SOFTWARE.
|
|
|
|
|
*/
|
|
|
|
|
|
|
|
|
|
#include "anv_private.h"
|
2019-03-18 19:06:05 -07:00
|
|
|
|
2025-09-16 14:39:40 -07:00
|
|
|
VkResult
|
|
|
|
|
anv_android_import_from_handle(struct anv_device *device,
|
|
|
|
|
const buffer_handle_t handle,
|
|
|
|
|
uint64_t modifier,
|
|
|
|
|
struct anv_bo **bo_out)
|
|
|
|
|
{
|
|
|
|
|
/* NOTE - We support buffers with only one handle but do not error on
|
|
|
|
|
* multiple handle case. Reason is that we want to support YUV formats
|
|
|
|
|
* where we have many logical planes but they all point to the same
|
|
|
|
|
* buffer, like is the case with VK_FORMAT_G8_B8R8_2PLANE_420_UNORM.
|
|
|
|
|
*
|
|
|
|
|
* Do not close the gralloc handle's dma_buf. The lifetime of the dma_buf
|
|
|
|
|
* must exceed that of the gralloc handle, and we do not own the gralloc
|
|
|
|
|
* handle.
|
|
|
|
|
*/
|
|
|
|
|
int dma_buf = (handle && handle->numFds) ? handle->data[0] : -1;
|
|
|
|
|
if (dma_buf < 0)
|
|
|
|
|
return VK_ERROR_INVALID_EXTERNAL_HANDLE;
|
|
|
|
|
|
|
|
|
|
enum anv_bo_alloc_flags alloc_flags = ANV_BO_ALLOC_EXTERNAL;
|
|
|
|
|
if (device->info->ver >= 20 && isl_drm_modifier_has_aux(modifier)) {
|
|
|
|
|
/* We always set scanout flag when importing buffers with Xe2
|
|
|
|
|
* modifiers, same as the rest in anv and iris drivers.
|
|
|
|
|
*/
|
|
|
|
|
alloc_flags |= ANV_BO_ALLOC_COMPRESSED | ANV_BO_ALLOC_SCANOUT;
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
return anv_device_import_bo(device,
|
|
|
|
|
dma_buf,
|
|
|
|
|
alloc_flags,
|
|
|
|
|
0 /* client_address */,
|
|
|
|
|
bo_out);
|
|
|
|
|
}
|
|
|
|
|
|
2018-11-08 10:20:35 +02:00
|
|
|
/*
|
|
|
|
|
* Called from anv_AllocateMemory when import AHardwareBuffer.
|
|
|
|
|
*/
|
|
|
|
|
VkResult
|
2025-10-23 14:14:25 -07:00
|
|
|
anv_import_ahb_memory(VkDevice device_h,
|
2023-03-20 18:00:38 -05:00
|
|
|
struct anv_device_memory *mem)
|
2018-11-08 10:20:35 +02:00
|
|
|
{
|
2019-03-18 19:06:05 -07:00
|
|
|
#if ANDROID_API_LEVEL >= 26
|
2018-11-08 10:20:35 +02:00
|
|
|
ANV_FROM_HANDLE(anv_device, device, device_h);
|
2025-10-24 13:46:58 -07:00
|
|
|
VkResult result;
|
|
|
|
|
VkImageDrmFormatModifierExplicitCreateInfoEXT mod_info;
|
|
|
|
|
VkSubresourceLayout layouts[ISL_MODIFIER_MAX_PLANES];
|
|
|
|
|
|
|
|
|
|
result = vk_android_get_ahb_layout(mem->vk.ahardware_buffer,
|
|
|
|
|
&mod_info, layouts,
|
|
|
|
|
ISL_MODIFIER_MAX_PLANES);
|
|
|
|
|
if (result != VK_SUCCESS)
|
|
|
|
|
return result;
|
2018-11-08 10:20:35 +02:00
|
|
|
|
|
|
|
|
/* Import from AHardwareBuffer to anv_device_memory. */
|
|
|
|
|
const native_handle_t *handle =
|
2023-03-20 18:00:38 -05:00
|
|
|
AHardwareBuffer_getNativeHandle(mem->vk.ahardware_buffer);
|
2018-11-08 10:20:35 +02:00
|
|
|
|
|
|
|
|
/* NOTE - We support buffers with only one handle but do not error on
|
|
|
|
|
* multiple handle case. Reason is that we want to support YUV formats
|
|
|
|
|
* where we have many logical planes but they all point to the same
|
|
|
|
|
* buffer, like is the case with VK_FORMAT_G8_B8R8_2PLANE_420_UNORM.
|
|
|
|
|
*/
|
2025-10-24 13:46:58 -07:00
|
|
|
return anv_android_import_from_handle(device, handle,
|
|
|
|
|
mod_info.drmFormatModifier,
|
|
|
|
|
&mem->bo);
|
2019-03-18 19:06:05 -07:00
|
|
|
#else
|
|
|
|
|
return VK_ERROR_EXTENSION_NOT_PRESENT;
|
|
|
|
|
#endif
|
2018-11-08 10:20:35 +02:00
|
|
|
}
|
|
|
|
|
|
2024-06-21 11:50:30 -07:00
|
|
|
VkResult
|
|
|
|
|
anv_android_get_tiling(struct anv_device *device,
|
|
|
|
|
struct u_gralloc_buffer_handle *gr_handle,
|
|
|
|
|
enum isl_tiling *tiling_out)
|
|
|
|
|
{
|
2025-06-16 00:54:47 -07:00
|
|
|
struct u_gralloc *gralloc = vk_android_get_ugralloc();
|
|
|
|
|
assert(gralloc);
|
2024-06-21 11:50:30 -07:00
|
|
|
|
|
|
|
|
struct u_gralloc_buffer_basic_info buf_info;
|
2025-06-16 00:54:47 -07:00
|
|
|
if (u_gralloc_get_buffer_basic_info(gralloc, gr_handle, &buf_info))
|
2024-06-21 11:50:30 -07:00
|
|
|
return vk_errorf(device, VK_ERROR_INVALID_EXTERNAL_HANDLE,
|
|
|
|
|
"failed to get tiling from gralloc buffer info");
|
|
|
|
|
|
|
|
|
|
const struct isl_drm_modifier_info *mod_info =
|
|
|
|
|
isl_drm_modifier_get_info(buf_info.modifier);
|
|
|
|
|
if (!mod_info) {
|
|
|
|
|
return vk_errorf(device, VK_ERROR_INVALID_EXTERNAL_HANDLE,
|
|
|
|
|
"invalid drm modifier from VkNativeBufferANDROID "
|
|
|
|
|
"gralloc buffer info 0x%"PRIx64"", buf_info.modifier);
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
*tiling_out = mod_info->tiling;
|
|
|
|
|
return VK_SUCCESS;
|
|
|
|
|
}
|
|
|
|
|
|
anv: Implement VK_ANDROID_native_buffer (v9)
This implementation is correct (afaict), but takes two shortcuts
regarding the import/export of Android sync fds.
Shortcut 1. When Android calls vkAcquireImageANDROID to import a sync
fd into a VkSemaphore or VkFence, the driver instead simply blocks on
the sync fd, then puts the VkSemaphore or VkFence into the signalled
state. Thanks to implicit sync, this produces correct behavior (with
extra latency overhead, perhaps) despite its ugliness.
Shortcut 2. When Android calls vkQueueSignalReleaseImageANDROID to export
a collection of wait semaphores as a sync fd, the driver instead
submits the semaphores to the queue, then returns sync fd -1, which
informs the caller that no additional synchronization is needed.
Again, thanks to implicit sync, this produces correct behavior (with
extra batch submission overhead) despite its ugliness.
I chose to take the shortcuts instead of properly importing/exporting
the sync fds for two reasons:
Reason 1. I've already tested this patch with dEQP and with demos
apps. It works. I wanted to get the tested patches into the tree now,
and polish the implementation afterwards.
Reason 2. I want to run this on a 3.18 kernel (gasp!). In 3.18, i915
supports neither Android's sync_fence, nor upstream's sync_file, nor
drm_syncobj. Again, I tested these patches on Android with a 3.18
kernel and they work.
I plan to quickly follow-up with patches that remove the shortcuts and
properly import/export the sync fds.
Non-Testing
===========
I did not test at all using the Android.mk buildsystem. I may have broke
it. Please test and review that.
Testing
=======
I tested with 64-bit ARC++ on a Skylake Chromebook and a 3.18 kernel.
The following pass (as of patchset v9):
- a little spinning cube demo APK
- several Sascha demos
- dEQP-VK.info.*
- dEQP-VK.api.wsi.android.*
(except dEQP-VK.api.wsi.android.swapchain.*.image_usage, because
dEQP wants to create swapchains with VK_IMAGE_USAGE_STORAGE_BIT)
- dEQP-VK.api.smoke.*
- dEQP-VK.api.info.instance.*
- dEQP-VK.api.info.device.*
v2:
- Reject VkNativeBufferANDROID if the dma-buf's size is too small for
the VkImage.
- Stop abusing VkNativeBufferANDROID by passing it to vkAllocateMemory
during vkCreateImage. Instead, directly import its dma-buf during
vkCreateImage with anv_bo_cache_import(). [for jekstrand]
- Rebase onto Tapani's VK_EXT_debug_report changes.
- Drop `CPPFLAGS += $(top_srcdir)/include/android`. The dir does not
exist.
v3:
- Delete duplicate #include "anv_private.h". [per Tapani]
- Try to fix the Android-IA build in Android.vulkan.mk by following
Tapani's example.
v4:
- Unset EXEC_OBJECT_ASYNC and set EXEC_OBJECT_WRITE on the imported
gralloc buffer, just as we do for all other winsys buffers in
anv_wsi.c. [found by Tapani]
v5:
- Really fix the Android-IA build by ensuring that Android.vulkan.mk
uses Mesa' vulkan.h and not Android's. Insert -I$(MESA_TOP)/include
before -Iframeworks/native/vulkan/include. [for Tapani]
- In vkAcquireImageANDROID, submit signal operations to the
VkSemaphore and VkFence. [for zhou]
v6:
- Drop copy-paste duplication in vkGetSwapchainGrallocUsageANDROID().
[found by zhou]
- Improve comments in vkGetSwapchainGrallocUsageANDROID().
v7:
- Fix vkGetSwapchainGrallocUsageANDROID() to inspect its
VkImageUsageFlags parameter. [for tfiga]
- This fix regresses dEQP-VK.api.wsi.android.swapchain.*.image_usage
because dEQP wants to create swapchains with
VK_IMAGE_USAGE_STORAGE_BIT.
v8:
- Drop unneeded goto in vkAcquireImageANDROID. [for tfiga]
v8.1: (minor changes)
- Drop errant hunks added by rerere in anv_device.c.
- Drop explicit mention of VK_ANDROID_native_buffer in
anv_entrypoints_gen.py. [for jekstrand]
v9:
- Isolate as much Android code as possible, moving it from anv_image.c
to anv_android.c. Connect the files with anv_image_from_gralloc().
Remove VkNativeBufferANDROID params from all anv_image.c
funcs. [for krh]
- Replace some intel_loge() with vk_errorf() in anv_android.c.
- Use © in copyright line. [for krh]
Reviewed-by: Tapani Pälli <tapani.palli@intel.com> (v5)
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com> (v9)
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net> (v9)
Cc: zhoucm1 <david1.zhou@amd.com>
Cc: Tomasz Figa <tfiga@chromium.org>
2016-11-14 16:13:51 -08:00
|
|
|
VkResult
|
2021-07-28 13:36:51 +03:00
|
|
|
anv_image_init_from_gralloc(struct anv_device *device,
|
|
|
|
|
struct anv_image *image,
|
|
|
|
|
const VkImageCreateInfo *base_info,
|
|
|
|
|
const VkNativeBufferANDROID *gralloc_info)
|
anv: Implement VK_ANDROID_native_buffer (v9)
This implementation is correct (afaict), but takes two shortcuts
regarding the import/export of Android sync fds.
Shortcut 1. When Android calls vkAcquireImageANDROID to import a sync
fd into a VkSemaphore or VkFence, the driver instead simply blocks on
the sync fd, then puts the VkSemaphore or VkFence into the signalled
state. Thanks to implicit sync, this produces correct behavior (with
extra latency overhead, perhaps) despite its ugliness.
Shortcut 2. When Android calls vkQueueSignalReleaseImageANDROID to export
a collection of wait semaphores as a sync fd, the driver instead
submits the semaphores to the queue, then returns sync fd -1, which
informs the caller that no additional synchronization is needed.
Again, thanks to implicit sync, this produces correct behavior (with
extra batch submission overhead) despite its ugliness.
I chose to take the shortcuts instead of properly importing/exporting
the sync fds for two reasons:
Reason 1. I've already tested this patch with dEQP and with demos
apps. It works. I wanted to get the tested patches into the tree now,
and polish the implementation afterwards.
Reason 2. I want to run this on a 3.18 kernel (gasp!). In 3.18, i915
supports neither Android's sync_fence, nor upstream's sync_file, nor
drm_syncobj. Again, I tested these patches on Android with a 3.18
kernel and they work.
I plan to quickly follow-up with patches that remove the shortcuts and
properly import/export the sync fds.
Non-Testing
===========
I did not test at all using the Android.mk buildsystem. I may have broke
it. Please test and review that.
Testing
=======
I tested with 64-bit ARC++ on a Skylake Chromebook and a 3.18 kernel.
The following pass (as of patchset v9):
- a little spinning cube demo APK
- several Sascha demos
- dEQP-VK.info.*
- dEQP-VK.api.wsi.android.*
(except dEQP-VK.api.wsi.android.swapchain.*.image_usage, because
dEQP wants to create swapchains with VK_IMAGE_USAGE_STORAGE_BIT)
- dEQP-VK.api.smoke.*
- dEQP-VK.api.info.instance.*
- dEQP-VK.api.info.device.*
v2:
- Reject VkNativeBufferANDROID if the dma-buf's size is too small for
the VkImage.
- Stop abusing VkNativeBufferANDROID by passing it to vkAllocateMemory
during vkCreateImage. Instead, directly import its dma-buf during
vkCreateImage with anv_bo_cache_import(). [for jekstrand]
- Rebase onto Tapani's VK_EXT_debug_report changes.
- Drop `CPPFLAGS += $(top_srcdir)/include/android`. The dir does not
exist.
v3:
- Delete duplicate #include "anv_private.h". [per Tapani]
- Try to fix the Android-IA build in Android.vulkan.mk by following
Tapani's example.
v4:
- Unset EXEC_OBJECT_ASYNC and set EXEC_OBJECT_WRITE on the imported
gralloc buffer, just as we do for all other winsys buffers in
anv_wsi.c. [found by Tapani]
v5:
- Really fix the Android-IA build by ensuring that Android.vulkan.mk
uses Mesa' vulkan.h and not Android's. Insert -I$(MESA_TOP)/include
before -Iframeworks/native/vulkan/include. [for Tapani]
- In vkAcquireImageANDROID, submit signal operations to the
VkSemaphore and VkFence. [for zhou]
v6:
- Drop copy-paste duplication in vkGetSwapchainGrallocUsageANDROID().
[found by zhou]
- Improve comments in vkGetSwapchainGrallocUsageANDROID().
v7:
- Fix vkGetSwapchainGrallocUsageANDROID() to inspect its
VkImageUsageFlags parameter. [for tfiga]
- This fix regresses dEQP-VK.api.wsi.android.swapchain.*.image_usage
because dEQP wants to create swapchains with
VK_IMAGE_USAGE_STORAGE_BIT.
v8:
- Drop unneeded goto in vkAcquireImageANDROID. [for tfiga]
v8.1: (minor changes)
- Drop errant hunks added by rerere in anv_device.c.
- Drop explicit mention of VK_ANDROID_native_buffer in
anv_entrypoints_gen.py. [for jekstrand]
v9:
- Isolate as much Android code as possible, moving it from anv_image.c
to anv_android.c. Connect the files with anv_image_from_gralloc().
Remove VkNativeBufferANDROID params from all anv_image.c
funcs. [for krh]
- Replace some intel_loge() with vk_errorf() in anv_android.c.
- Use © in copyright line. [for krh]
Reviewed-by: Tapani Pälli <tapani.palli@intel.com> (v5)
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com> (v9)
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net> (v9)
Cc: zhoucm1 <david1.zhou@amd.com>
Cc: Tomasz Figa <tfiga@chromium.org>
2016-11-14 16:13:51 -08:00
|
|
|
{
|
|
|
|
|
VkResult result;
|
|
|
|
|
|
|
|
|
|
struct anv_image_create_info anv_info = {
|
|
|
|
|
.vk_info = base_info,
|
|
|
|
|
.isl_extra_usage_flags = ISL_SURF_USAGE_DISABLE_AUX_BIT,
|
|
|
|
|
};
|
|
|
|
|
|
2025-10-24 13:46:58 -07:00
|
|
|
VkImageDrmFormatModifierExplicitCreateInfoEXT mod_info = {
|
|
|
|
|
.drmFormatModifier = DRM_FORMAT_MOD_INVALID,
|
|
|
|
|
};
|
anv: Implement VK_ANDROID_native_buffer (v9)
This implementation is correct (afaict), but takes two shortcuts
regarding the import/export of Android sync fds.
Shortcut 1. When Android calls vkAcquireImageANDROID to import a sync
fd into a VkSemaphore or VkFence, the driver instead simply blocks on
the sync fd, then puts the VkSemaphore or VkFence into the signalled
state. Thanks to implicit sync, this produces correct behavior (with
extra latency overhead, perhaps) despite its ugliness.
Shortcut 2. When Android calls vkQueueSignalReleaseImageANDROID to export
a collection of wait semaphores as a sync fd, the driver instead
submits the semaphores to the queue, then returns sync fd -1, which
informs the caller that no additional synchronization is needed.
Again, thanks to implicit sync, this produces correct behavior (with
extra batch submission overhead) despite its ugliness.
I chose to take the shortcuts instead of properly importing/exporting
the sync fds for two reasons:
Reason 1. I've already tested this patch with dEQP and with demos
apps. It works. I wanted to get the tested patches into the tree now,
and polish the implementation afterwards.
Reason 2. I want to run this on a 3.18 kernel (gasp!). In 3.18, i915
supports neither Android's sync_fence, nor upstream's sync_file, nor
drm_syncobj. Again, I tested these patches on Android with a 3.18
kernel and they work.
I plan to quickly follow-up with patches that remove the shortcuts and
properly import/export the sync fds.
Non-Testing
===========
I did not test at all using the Android.mk buildsystem. I may have broke
it. Please test and review that.
Testing
=======
I tested with 64-bit ARC++ on a Skylake Chromebook and a 3.18 kernel.
The following pass (as of patchset v9):
- a little spinning cube demo APK
- several Sascha demos
- dEQP-VK.info.*
- dEQP-VK.api.wsi.android.*
(except dEQP-VK.api.wsi.android.swapchain.*.image_usage, because
dEQP wants to create swapchains with VK_IMAGE_USAGE_STORAGE_BIT)
- dEQP-VK.api.smoke.*
- dEQP-VK.api.info.instance.*
- dEQP-VK.api.info.device.*
v2:
- Reject VkNativeBufferANDROID if the dma-buf's size is too small for
the VkImage.
- Stop abusing VkNativeBufferANDROID by passing it to vkAllocateMemory
during vkCreateImage. Instead, directly import its dma-buf during
vkCreateImage with anv_bo_cache_import(). [for jekstrand]
- Rebase onto Tapani's VK_EXT_debug_report changes.
- Drop `CPPFLAGS += $(top_srcdir)/include/android`. The dir does not
exist.
v3:
- Delete duplicate #include "anv_private.h". [per Tapani]
- Try to fix the Android-IA build in Android.vulkan.mk by following
Tapani's example.
v4:
- Unset EXEC_OBJECT_ASYNC and set EXEC_OBJECT_WRITE on the imported
gralloc buffer, just as we do for all other winsys buffers in
anv_wsi.c. [found by Tapani]
v5:
- Really fix the Android-IA build by ensuring that Android.vulkan.mk
uses Mesa' vulkan.h and not Android's. Insert -I$(MESA_TOP)/include
before -Iframeworks/native/vulkan/include. [for Tapani]
- In vkAcquireImageANDROID, submit signal operations to the
VkSemaphore and VkFence. [for zhou]
v6:
- Drop copy-paste duplication in vkGetSwapchainGrallocUsageANDROID().
[found by zhou]
- Improve comments in vkGetSwapchainGrallocUsageANDROID().
v7:
- Fix vkGetSwapchainGrallocUsageANDROID() to inspect its
VkImageUsageFlags parameter. [for tfiga]
- This fix regresses dEQP-VK.api.wsi.android.swapchain.*.image_usage
because dEQP wants to create swapchains with
VK_IMAGE_USAGE_STORAGE_BIT.
v8:
- Drop unneeded goto in vkAcquireImageANDROID. [for tfiga]
v8.1: (minor changes)
- Drop errant hunks added by rerere in anv_device.c.
- Drop explicit mention of VK_ANDROID_native_buffer in
anv_entrypoints_gen.py. [for jekstrand]
v9:
- Isolate as much Android code as possible, moving it from anv_image.c
to anv_android.c. Connect the files with anv_image_from_gralloc().
Remove VkNativeBufferANDROID params from all anv_image.c
funcs. [for krh]
- Replace some intel_loge() with vk_errorf() in anv_android.c.
- Use © in copyright line. [for krh]
Reviewed-by: Tapani Pälli <tapani.palli@intel.com> (v5)
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com> (v9)
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net> (v9)
Cc: zhoucm1 <david1.zhou@amd.com>
Cc: Tomasz Figa <tfiga@chromium.org>
2016-11-14 16:13:51 -08:00
|
|
|
|
2021-10-30 15:47:38 -05:00
|
|
|
enum isl_tiling tiling;
|
2025-06-16 00:54:47 -07:00
|
|
|
if (vk_android_get_ugralloc()) {
|
2025-10-24 12:47:11 -07:00
|
|
|
VkSubresourceLayout layouts[ISL_MODIFIER_MAX_PLANES];
|
|
|
|
|
result = vk_android_get_ahb_layout(gralloc_info->ahb,
|
|
|
|
|
&mod_info,
|
|
|
|
|
layouts,
|
|
|
|
|
ISL_MODIFIER_MAX_PLANES);
|
2024-06-21 11:50:30 -07:00
|
|
|
if (result != VK_SUCCESS)
|
|
|
|
|
return result;
|
2025-10-24 12:47:11 -07:00
|
|
|
const struct isl_drm_modifier_info *isl_mod_info =
|
|
|
|
|
isl_drm_modifier_get_info(mod_info.drmFormatModifier);
|
|
|
|
|
if (!isl_mod_info) {
|
|
|
|
|
return vk_errorf(device, VK_ERROR_INVALID_EXTERNAL_HANDLE,
|
|
|
|
|
"invalid modifier from gralloc info 0x%"PRIx64"",
|
|
|
|
|
mod_info.drmFormatModifier);
|
|
|
|
|
}
|
|
|
|
|
tiling = isl_mod_info->tiling;
|
2025-10-24 13:46:58 -07:00
|
|
|
}
|
|
|
|
|
|
|
|
|
|
struct anv_bo *bo = NULL;
|
|
|
|
|
/* If this function fails and if the imported bo was resident in the cache,
|
|
|
|
|
* we should avoid updating the bo's flags. Therefore, we defer updating
|
|
|
|
|
* the flags until success is certain.
|
|
|
|
|
*
|
|
|
|
|
*/
|
|
|
|
|
result = anv_android_import_from_handle(device,
|
|
|
|
|
gralloc_info->handle,
|
|
|
|
|
mod_info.drmFormatModifier,
|
|
|
|
|
&bo);
|
|
|
|
|
if (result != VK_SUCCESS) {
|
|
|
|
|
return vk_errorf(device, result,
|
|
|
|
|
"failed to import dma-buf from VkNativeBufferANDROID");
|
|
|
|
|
}
|
|
|
|
|
|
|
|
|
|
/* The bo has to be imported first to do this when the tiling hasn't been
|
|
|
|
|
* obtained.
|
|
|
|
|
*/
|
|
|
|
|
if (mod_info.drmFormatModifier == DRM_FORMAT_MOD_INVALID) {
|
2024-05-14 14:00:08 +00:00
|
|
|
/* Fallback to get_tiling API. */
|
|
|
|
|
result = anv_device_get_bo_tiling(device, bo, &tiling);
|
|
|
|
|
if (result != VK_SUCCESS) {
|
|
|
|
|
return vk_errorf(device, result,
|
|
|
|
|
"failed to get tiling from VkNativeBufferANDROID");
|
|
|
|
|
}
|
anv: Implement VK_ANDROID_native_buffer (v9)
This implementation is correct (afaict), but takes two shortcuts
regarding the import/export of Android sync fds.
Shortcut 1. When Android calls vkAcquireImageANDROID to import a sync
fd into a VkSemaphore or VkFence, the driver instead simply blocks on
the sync fd, then puts the VkSemaphore or VkFence into the signalled
state. Thanks to implicit sync, this produces correct behavior (with
extra latency overhead, perhaps) despite its ugliness.
Shortcut 2. When Android calls vkQueueSignalReleaseImageANDROID to export
a collection of wait semaphores as a sync fd, the driver instead
submits the semaphores to the queue, then returns sync fd -1, which
informs the caller that no additional synchronization is needed.
Again, thanks to implicit sync, this produces correct behavior (with
extra batch submission overhead) despite its ugliness.
I chose to take the shortcuts instead of properly importing/exporting
the sync fds for two reasons:
Reason 1. I've already tested this patch with dEQP and with demos
apps. It works. I wanted to get the tested patches into the tree now,
and polish the implementation afterwards.
Reason 2. I want to run this on a 3.18 kernel (gasp!). In 3.18, i915
supports neither Android's sync_fence, nor upstream's sync_file, nor
drm_syncobj. Again, I tested these patches on Android with a 3.18
kernel and they work.
I plan to quickly follow-up with patches that remove the shortcuts and
properly import/export the sync fds.
Non-Testing
===========
I did not test at all using the Android.mk buildsystem. I may have broke
it. Please test and review that.
Testing
=======
I tested with 64-bit ARC++ on a Skylake Chromebook and a 3.18 kernel.
The following pass (as of patchset v9):
- a little spinning cube demo APK
- several Sascha demos
- dEQP-VK.info.*
- dEQP-VK.api.wsi.android.*
(except dEQP-VK.api.wsi.android.swapchain.*.image_usage, because
dEQP wants to create swapchains with VK_IMAGE_USAGE_STORAGE_BIT)
- dEQP-VK.api.smoke.*
- dEQP-VK.api.info.instance.*
- dEQP-VK.api.info.device.*
v2:
- Reject VkNativeBufferANDROID if the dma-buf's size is too small for
the VkImage.
- Stop abusing VkNativeBufferANDROID by passing it to vkAllocateMemory
during vkCreateImage. Instead, directly import its dma-buf during
vkCreateImage with anv_bo_cache_import(). [for jekstrand]
- Rebase onto Tapani's VK_EXT_debug_report changes.
- Drop `CPPFLAGS += $(top_srcdir)/include/android`. The dir does not
exist.
v3:
- Delete duplicate #include "anv_private.h". [per Tapani]
- Try to fix the Android-IA build in Android.vulkan.mk by following
Tapani's example.
v4:
- Unset EXEC_OBJECT_ASYNC and set EXEC_OBJECT_WRITE on the imported
gralloc buffer, just as we do for all other winsys buffers in
anv_wsi.c. [found by Tapani]
v5:
- Really fix the Android-IA build by ensuring that Android.vulkan.mk
uses Mesa' vulkan.h and not Android's. Insert -I$(MESA_TOP)/include
before -Iframeworks/native/vulkan/include. [for Tapani]
- In vkAcquireImageANDROID, submit signal operations to the
VkSemaphore and VkFence. [for zhou]
v6:
- Drop copy-paste duplication in vkGetSwapchainGrallocUsageANDROID().
[found by zhou]
- Improve comments in vkGetSwapchainGrallocUsageANDROID().
v7:
- Fix vkGetSwapchainGrallocUsageANDROID() to inspect its
VkImageUsageFlags parameter. [for tfiga]
- This fix regresses dEQP-VK.api.wsi.android.swapchain.*.image_usage
because dEQP wants to create swapchains with
VK_IMAGE_USAGE_STORAGE_BIT.
v8:
- Drop unneeded goto in vkAcquireImageANDROID. [for tfiga]
v8.1: (minor changes)
- Drop errant hunks added by rerere in anv_device.c.
- Drop explicit mention of VK_ANDROID_native_buffer in
anv_entrypoints_gen.py. [for jekstrand]
v9:
- Isolate as much Android code as possible, moving it from anv_image.c
to anv_android.c. Connect the files with anv_image_from_gralloc().
Remove VkNativeBufferANDROID params from all anv_image.c
funcs. [for krh]
- Replace some intel_loge() with vk_errorf() in anv_android.c.
- Use © in copyright line. [for krh]
Reviewed-by: Tapani Pälli <tapani.palli@intel.com> (v5)
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com> (v9)
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net> (v9)
Cc: zhoucm1 <david1.zhou@amd.com>
Cc: Tomasz Figa <tfiga@chromium.org>
2016-11-14 16:13:51 -08:00
|
|
|
}
|
2021-10-30 15:47:38 -05:00
|
|
|
anv_info.isl_tiling_flags = 1u << tiling;
|
anv: Implement VK_ANDROID_native_buffer (v9)
This implementation is correct (afaict), but takes two shortcuts
regarding the import/export of Android sync fds.
Shortcut 1. When Android calls vkAcquireImageANDROID to import a sync
fd into a VkSemaphore or VkFence, the driver instead simply blocks on
the sync fd, then puts the VkSemaphore or VkFence into the signalled
state. Thanks to implicit sync, this produces correct behavior (with
extra latency overhead, perhaps) despite its ugliness.
Shortcut 2. When Android calls vkQueueSignalReleaseImageANDROID to export
a collection of wait semaphores as a sync fd, the driver instead
submits the semaphores to the queue, then returns sync fd -1, which
informs the caller that no additional synchronization is needed.
Again, thanks to implicit sync, this produces correct behavior (with
extra batch submission overhead) despite its ugliness.
I chose to take the shortcuts instead of properly importing/exporting
the sync fds for two reasons:
Reason 1. I've already tested this patch with dEQP and with demos
apps. It works. I wanted to get the tested patches into the tree now,
and polish the implementation afterwards.
Reason 2. I want to run this on a 3.18 kernel (gasp!). In 3.18, i915
supports neither Android's sync_fence, nor upstream's sync_file, nor
drm_syncobj. Again, I tested these patches on Android with a 3.18
kernel and they work.
I plan to quickly follow-up with patches that remove the shortcuts and
properly import/export the sync fds.
Non-Testing
===========
I did not test at all using the Android.mk buildsystem. I may have broke
it. Please test and review that.
Testing
=======
I tested with 64-bit ARC++ on a Skylake Chromebook and a 3.18 kernel.
The following pass (as of patchset v9):
- a little spinning cube demo APK
- several Sascha demos
- dEQP-VK.info.*
- dEQP-VK.api.wsi.android.*
(except dEQP-VK.api.wsi.android.swapchain.*.image_usage, because
dEQP wants to create swapchains with VK_IMAGE_USAGE_STORAGE_BIT)
- dEQP-VK.api.smoke.*
- dEQP-VK.api.info.instance.*
- dEQP-VK.api.info.device.*
v2:
- Reject VkNativeBufferANDROID if the dma-buf's size is too small for
the VkImage.
- Stop abusing VkNativeBufferANDROID by passing it to vkAllocateMemory
during vkCreateImage. Instead, directly import its dma-buf during
vkCreateImage with anv_bo_cache_import(). [for jekstrand]
- Rebase onto Tapani's VK_EXT_debug_report changes.
- Drop `CPPFLAGS += $(top_srcdir)/include/android`. The dir does not
exist.
v3:
- Delete duplicate #include "anv_private.h". [per Tapani]
- Try to fix the Android-IA build in Android.vulkan.mk by following
Tapani's example.
v4:
- Unset EXEC_OBJECT_ASYNC and set EXEC_OBJECT_WRITE on the imported
gralloc buffer, just as we do for all other winsys buffers in
anv_wsi.c. [found by Tapani]
v5:
- Really fix the Android-IA build by ensuring that Android.vulkan.mk
uses Mesa' vulkan.h and not Android's. Insert -I$(MESA_TOP)/include
before -Iframeworks/native/vulkan/include. [for Tapani]
- In vkAcquireImageANDROID, submit signal operations to the
VkSemaphore and VkFence. [for zhou]
v6:
- Drop copy-paste duplication in vkGetSwapchainGrallocUsageANDROID().
[found by zhou]
- Improve comments in vkGetSwapchainGrallocUsageANDROID().
v7:
- Fix vkGetSwapchainGrallocUsageANDROID() to inspect its
VkImageUsageFlags parameter. [for tfiga]
- This fix regresses dEQP-VK.api.wsi.android.swapchain.*.image_usage
because dEQP wants to create swapchains with
VK_IMAGE_USAGE_STORAGE_BIT.
v8:
- Drop unneeded goto in vkAcquireImageANDROID. [for tfiga]
v8.1: (minor changes)
- Drop errant hunks added by rerere in anv_device.c.
- Drop explicit mention of VK_ANDROID_native_buffer in
anv_entrypoints_gen.py. [for jekstrand]
v9:
- Isolate as much Android code as possible, moving it from anv_image.c
to anv_android.c. Connect the files with anv_image_from_gralloc().
Remove VkNativeBufferANDROID params from all anv_image.c
funcs. [for krh]
- Replace some intel_loge() with vk_errorf() in anv_android.c.
- Use © in copyright line. [for krh]
Reviewed-by: Tapani Pälli <tapani.palli@intel.com> (v5)
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com> (v9)
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net> (v9)
Cc: zhoucm1 <david1.zhou@amd.com>
Cc: Tomasz Figa <tfiga@chromium.org>
2016-11-14 16:13:51 -08:00
|
|
|
|
2023-07-30 09:20:52 +01:00
|
|
|
anv_info.stride = gralloc_info->stride;
|
2023-03-28 10:13:05 +00:00
|
|
|
|
2021-07-28 13:36:51 +03:00
|
|
|
result = anv_image_init(device, image, &anv_info);
|
anv: Implement VK_ANDROID_native_buffer (v9)
This implementation is correct (afaict), but takes two shortcuts
regarding the import/export of Android sync fds.
Shortcut 1. When Android calls vkAcquireImageANDROID to import a sync
fd into a VkSemaphore or VkFence, the driver instead simply blocks on
the sync fd, then puts the VkSemaphore or VkFence into the signalled
state. Thanks to implicit sync, this produces correct behavior (with
extra latency overhead, perhaps) despite its ugliness.
Shortcut 2. When Android calls vkQueueSignalReleaseImageANDROID to export
a collection of wait semaphores as a sync fd, the driver instead
submits the semaphores to the queue, then returns sync fd -1, which
informs the caller that no additional synchronization is needed.
Again, thanks to implicit sync, this produces correct behavior (with
extra batch submission overhead) despite its ugliness.
I chose to take the shortcuts instead of properly importing/exporting
the sync fds for two reasons:
Reason 1. I've already tested this patch with dEQP and with demos
apps. It works. I wanted to get the tested patches into the tree now,
and polish the implementation afterwards.
Reason 2. I want to run this on a 3.18 kernel (gasp!). In 3.18, i915
supports neither Android's sync_fence, nor upstream's sync_file, nor
drm_syncobj. Again, I tested these patches on Android with a 3.18
kernel and they work.
I plan to quickly follow-up with patches that remove the shortcuts and
properly import/export the sync fds.
Non-Testing
===========
I did not test at all using the Android.mk buildsystem. I may have broke
it. Please test and review that.
Testing
=======
I tested with 64-bit ARC++ on a Skylake Chromebook and a 3.18 kernel.
The following pass (as of patchset v9):
- a little spinning cube demo APK
- several Sascha demos
- dEQP-VK.info.*
- dEQP-VK.api.wsi.android.*
(except dEQP-VK.api.wsi.android.swapchain.*.image_usage, because
dEQP wants to create swapchains with VK_IMAGE_USAGE_STORAGE_BIT)
- dEQP-VK.api.smoke.*
- dEQP-VK.api.info.instance.*
- dEQP-VK.api.info.device.*
v2:
- Reject VkNativeBufferANDROID if the dma-buf's size is too small for
the VkImage.
- Stop abusing VkNativeBufferANDROID by passing it to vkAllocateMemory
during vkCreateImage. Instead, directly import its dma-buf during
vkCreateImage with anv_bo_cache_import(). [for jekstrand]
- Rebase onto Tapani's VK_EXT_debug_report changes.
- Drop `CPPFLAGS += $(top_srcdir)/include/android`. The dir does not
exist.
v3:
- Delete duplicate #include "anv_private.h". [per Tapani]
- Try to fix the Android-IA build in Android.vulkan.mk by following
Tapani's example.
v4:
- Unset EXEC_OBJECT_ASYNC and set EXEC_OBJECT_WRITE on the imported
gralloc buffer, just as we do for all other winsys buffers in
anv_wsi.c. [found by Tapani]
v5:
- Really fix the Android-IA build by ensuring that Android.vulkan.mk
uses Mesa' vulkan.h and not Android's. Insert -I$(MESA_TOP)/include
before -Iframeworks/native/vulkan/include. [for Tapani]
- In vkAcquireImageANDROID, submit signal operations to the
VkSemaphore and VkFence. [for zhou]
v6:
- Drop copy-paste duplication in vkGetSwapchainGrallocUsageANDROID().
[found by zhou]
- Improve comments in vkGetSwapchainGrallocUsageANDROID().
v7:
- Fix vkGetSwapchainGrallocUsageANDROID() to inspect its
VkImageUsageFlags parameter. [for tfiga]
- This fix regresses dEQP-VK.api.wsi.android.swapchain.*.image_usage
because dEQP wants to create swapchains with
VK_IMAGE_USAGE_STORAGE_BIT.
v8:
- Drop unneeded goto in vkAcquireImageANDROID. [for tfiga]
v8.1: (minor changes)
- Drop errant hunks added by rerere in anv_device.c.
- Drop explicit mention of VK_ANDROID_native_buffer in
anv_entrypoints_gen.py. [for jekstrand]
v9:
- Isolate as much Android code as possible, moving it from anv_image.c
to anv_android.c. Connect the files with anv_image_from_gralloc().
Remove VkNativeBufferANDROID params from all anv_image.c
funcs. [for krh]
- Replace some intel_loge() with vk_errorf() in anv_android.c.
- Use © in copyright line. [for krh]
Reviewed-by: Tapani Pälli <tapani.palli@intel.com> (v5)
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com> (v9)
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net> (v9)
Cc: zhoucm1 <david1.zhou@amd.com>
Cc: Tomasz Figa <tfiga@chromium.org>
2016-11-14 16:13:51 -08:00
|
|
|
if (result != VK_SUCCESS)
|
2021-07-28 13:36:51 +03:00
|
|
|
goto fail_init;
|
anv: Implement VK_ANDROID_native_buffer (v9)
This implementation is correct (afaict), but takes two shortcuts
regarding the import/export of Android sync fds.
Shortcut 1. When Android calls vkAcquireImageANDROID to import a sync
fd into a VkSemaphore or VkFence, the driver instead simply blocks on
the sync fd, then puts the VkSemaphore or VkFence into the signalled
state. Thanks to implicit sync, this produces correct behavior (with
extra latency overhead, perhaps) despite its ugliness.
Shortcut 2. When Android calls vkQueueSignalReleaseImageANDROID to export
a collection of wait semaphores as a sync fd, the driver instead
submits the semaphores to the queue, then returns sync fd -1, which
informs the caller that no additional synchronization is needed.
Again, thanks to implicit sync, this produces correct behavior (with
extra batch submission overhead) despite its ugliness.
I chose to take the shortcuts instead of properly importing/exporting
the sync fds for two reasons:
Reason 1. I've already tested this patch with dEQP and with demos
apps. It works. I wanted to get the tested patches into the tree now,
and polish the implementation afterwards.
Reason 2. I want to run this on a 3.18 kernel (gasp!). In 3.18, i915
supports neither Android's sync_fence, nor upstream's sync_file, nor
drm_syncobj. Again, I tested these patches on Android with a 3.18
kernel and they work.
I plan to quickly follow-up with patches that remove the shortcuts and
properly import/export the sync fds.
Non-Testing
===========
I did not test at all using the Android.mk buildsystem. I may have broke
it. Please test and review that.
Testing
=======
I tested with 64-bit ARC++ on a Skylake Chromebook and a 3.18 kernel.
The following pass (as of patchset v9):
- a little spinning cube demo APK
- several Sascha demos
- dEQP-VK.info.*
- dEQP-VK.api.wsi.android.*
(except dEQP-VK.api.wsi.android.swapchain.*.image_usage, because
dEQP wants to create swapchains with VK_IMAGE_USAGE_STORAGE_BIT)
- dEQP-VK.api.smoke.*
- dEQP-VK.api.info.instance.*
- dEQP-VK.api.info.device.*
v2:
- Reject VkNativeBufferANDROID if the dma-buf's size is too small for
the VkImage.
- Stop abusing VkNativeBufferANDROID by passing it to vkAllocateMemory
during vkCreateImage. Instead, directly import its dma-buf during
vkCreateImage with anv_bo_cache_import(). [for jekstrand]
- Rebase onto Tapani's VK_EXT_debug_report changes.
- Drop `CPPFLAGS += $(top_srcdir)/include/android`. The dir does not
exist.
v3:
- Delete duplicate #include "anv_private.h". [per Tapani]
- Try to fix the Android-IA build in Android.vulkan.mk by following
Tapani's example.
v4:
- Unset EXEC_OBJECT_ASYNC and set EXEC_OBJECT_WRITE on the imported
gralloc buffer, just as we do for all other winsys buffers in
anv_wsi.c. [found by Tapani]
v5:
- Really fix the Android-IA build by ensuring that Android.vulkan.mk
uses Mesa' vulkan.h and not Android's. Insert -I$(MESA_TOP)/include
before -Iframeworks/native/vulkan/include. [for Tapani]
- In vkAcquireImageANDROID, submit signal operations to the
VkSemaphore and VkFence. [for zhou]
v6:
- Drop copy-paste duplication in vkGetSwapchainGrallocUsageANDROID().
[found by zhou]
- Improve comments in vkGetSwapchainGrallocUsageANDROID().
v7:
- Fix vkGetSwapchainGrallocUsageANDROID() to inspect its
VkImageUsageFlags parameter. [for tfiga]
- This fix regresses dEQP-VK.api.wsi.android.swapchain.*.image_usage
because dEQP wants to create swapchains with
VK_IMAGE_USAGE_STORAGE_BIT.
v8:
- Drop unneeded goto in vkAcquireImageANDROID. [for tfiga]
v8.1: (minor changes)
- Drop errant hunks added by rerere in anv_device.c.
- Drop explicit mention of VK_ANDROID_native_buffer in
anv_entrypoints_gen.py. [for jekstrand]
v9:
- Isolate as much Android code as possible, moving it from anv_image.c
to anv_android.c. Connect the files with anv_image_from_gralloc().
Remove VkNativeBufferANDROID params from all anv_image.c
funcs. [for krh]
- Replace some intel_loge() with vk_errorf() in anv_android.c.
- Use © in copyright line. [for krh]
Reviewed-by: Tapani Pälli <tapani.palli@intel.com> (v5)
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com> (v9)
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net> (v9)
Cc: zhoucm1 <david1.zhou@amd.com>
Cc: Tomasz Figa <tfiga@chromium.org>
2016-11-14 16:13:51 -08:00
|
|
|
|
2021-02-05 12:07:08 -08:00
|
|
|
VkMemoryRequirements2 mem_reqs = {
|
|
|
|
|
.sType = VK_STRUCTURE_TYPE_MEMORY_REQUIREMENTS_2,
|
|
|
|
|
};
|
|
|
|
|
|
2021-10-05 10:44:42 -05:00
|
|
|
anv_image_get_memory_requirements(device, image, image->vk.aspects,
|
|
|
|
|
&mem_reqs);
|
2021-02-05 12:07:08 -08:00
|
|
|
|
|
|
|
|
VkDeviceSize aligned_image_size =
|
2022-12-02 14:37:31 +05:30
|
|
|
align64(mem_reqs.memoryRequirements.size,
|
|
|
|
|
mem_reqs.memoryRequirements.alignment);
|
2021-02-05 12:07:08 -08:00
|
|
|
|
|
|
|
|
if (bo->size < aligned_image_size) {
|
2021-09-24 12:06:32 -05:00
|
|
|
result = vk_errorf(device, VK_ERROR_INVALID_EXTERNAL_HANDLE,
|
anv: Implement VK_ANDROID_native_buffer (v9)
This implementation is correct (afaict), but takes two shortcuts
regarding the import/export of Android sync fds.
Shortcut 1. When Android calls vkAcquireImageANDROID to import a sync
fd into a VkSemaphore or VkFence, the driver instead simply blocks on
the sync fd, then puts the VkSemaphore or VkFence into the signalled
state. Thanks to implicit sync, this produces correct behavior (with
extra latency overhead, perhaps) despite its ugliness.
Shortcut 2. When Android calls vkQueueSignalReleaseImageANDROID to export
a collection of wait semaphores as a sync fd, the driver instead
submits the semaphores to the queue, then returns sync fd -1, which
informs the caller that no additional synchronization is needed.
Again, thanks to implicit sync, this produces correct behavior (with
extra batch submission overhead) despite its ugliness.
I chose to take the shortcuts instead of properly importing/exporting
the sync fds for two reasons:
Reason 1. I've already tested this patch with dEQP and with demos
apps. It works. I wanted to get the tested patches into the tree now,
and polish the implementation afterwards.
Reason 2. I want to run this on a 3.18 kernel (gasp!). In 3.18, i915
supports neither Android's sync_fence, nor upstream's sync_file, nor
drm_syncobj. Again, I tested these patches on Android with a 3.18
kernel and they work.
I plan to quickly follow-up with patches that remove the shortcuts and
properly import/export the sync fds.
Non-Testing
===========
I did not test at all using the Android.mk buildsystem. I may have broke
it. Please test and review that.
Testing
=======
I tested with 64-bit ARC++ on a Skylake Chromebook and a 3.18 kernel.
The following pass (as of patchset v9):
- a little spinning cube demo APK
- several Sascha demos
- dEQP-VK.info.*
- dEQP-VK.api.wsi.android.*
(except dEQP-VK.api.wsi.android.swapchain.*.image_usage, because
dEQP wants to create swapchains with VK_IMAGE_USAGE_STORAGE_BIT)
- dEQP-VK.api.smoke.*
- dEQP-VK.api.info.instance.*
- dEQP-VK.api.info.device.*
v2:
- Reject VkNativeBufferANDROID if the dma-buf's size is too small for
the VkImage.
- Stop abusing VkNativeBufferANDROID by passing it to vkAllocateMemory
during vkCreateImage. Instead, directly import its dma-buf during
vkCreateImage with anv_bo_cache_import(). [for jekstrand]
- Rebase onto Tapani's VK_EXT_debug_report changes.
- Drop `CPPFLAGS += $(top_srcdir)/include/android`. The dir does not
exist.
v3:
- Delete duplicate #include "anv_private.h". [per Tapani]
- Try to fix the Android-IA build in Android.vulkan.mk by following
Tapani's example.
v4:
- Unset EXEC_OBJECT_ASYNC and set EXEC_OBJECT_WRITE on the imported
gralloc buffer, just as we do for all other winsys buffers in
anv_wsi.c. [found by Tapani]
v5:
- Really fix the Android-IA build by ensuring that Android.vulkan.mk
uses Mesa' vulkan.h and not Android's. Insert -I$(MESA_TOP)/include
before -Iframeworks/native/vulkan/include. [for Tapani]
- In vkAcquireImageANDROID, submit signal operations to the
VkSemaphore and VkFence. [for zhou]
v6:
- Drop copy-paste duplication in vkGetSwapchainGrallocUsageANDROID().
[found by zhou]
- Improve comments in vkGetSwapchainGrallocUsageANDROID().
v7:
- Fix vkGetSwapchainGrallocUsageANDROID() to inspect its
VkImageUsageFlags parameter. [for tfiga]
- This fix regresses dEQP-VK.api.wsi.android.swapchain.*.image_usage
because dEQP wants to create swapchains with
VK_IMAGE_USAGE_STORAGE_BIT.
v8:
- Drop unneeded goto in vkAcquireImageANDROID. [for tfiga]
v8.1: (minor changes)
- Drop errant hunks added by rerere in anv_device.c.
- Drop explicit mention of VK_ANDROID_native_buffer in
anv_entrypoints_gen.py. [for jekstrand]
v9:
- Isolate as much Android code as possible, moving it from anv_image.c
to anv_android.c. Connect the files with anv_image_from_gralloc().
Remove VkNativeBufferANDROID params from all anv_image.c
funcs. [for krh]
- Replace some intel_loge() with vk_errorf() in anv_android.c.
- Use © in copyright line. [for krh]
Reviewed-by: Tapani Pälli <tapani.palli@intel.com> (v5)
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com> (v9)
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net> (v9)
Cc: zhoucm1 <david1.zhou@amd.com>
Cc: Tomasz Figa <tfiga@chromium.org>
2016-11-14 16:13:51 -08:00
|
|
|
"dma-buf from VkNativeBufferANDROID is too small for "
|
|
|
|
|
"VkImage: %"PRIu64"B < %"PRIu64"B",
|
2021-02-05 12:07:08 -08:00
|
|
|
bo->size, aligned_image_size);
|
anv: Implement VK_ANDROID_native_buffer (v9)
This implementation is correct (afaict), but takes two shortcuts
regarding the import/export of Android sync fds.
Shortcut 1. When Android calls vkAcquireImageANDROID to import a sync
fd into a VkSemaphore or VkFence, the driver instead simply blocks on
the sync fd, then puts the VkSemaphore or VkFence into the signalled
state. Thanks to implicit sync, this produces correct behavior (with
extra latency overhead, perhaps) despite its ugliness.
Shortcut 2. When Android calls vkQueueSignalReleaseImageANDROID to export
a collection of wait semaphores as a sync fd, the driver instead
submits the semaphores to the queue, then returns sync fd -1, which
informs the caller that no additional synchronization is needed.
Again, thanks to implicit sync, this produces correct behavior (with
extra batch submission overhead) despite its ugliness.
I chose to take the shortcuts instead of properly importing/exporting
the sync fds for two reasons:
Reason 1. I've already tested this patch with dEQP and with demos
apps. It works. I wanted to get the tested patches into the tree now,
and polish the implementation afterwards.
Reason 2. I want to run this on a 3.18 kernel (gasp!). In 3.18, i915
supports neither Android's sync_fence, nor upstream's sync_file, nor
drm_syncobj. Again, I tested these patches on Android with a 3.18
kernel and they work.
I plan to quickly follow-up with patches that remove the shortcuts and
properly import/export the sync fds.
Non-Testing
===========
I did not test at all using the Android.mk buildsystem. I may have broke
it. Please test and review that.
Testing
=======
I tested with 64-bit ARC++ on a Skylake Chromebook and a 3.18 kernel.
The following pass (as of patchset v9):
- a little spinning cube demo APK
- several Sascha demos
- dEQP-VK.info.*
- dEQP-VK.api.wsi.android.*
(except dEQP-VK.api.wsi.android.swapchain.*.image_usage, because
dEQP wants to create swapchains with VK_IMAGE_USAGE_STORAGE_BIT)
- dEQP-VK.api.smoke.*
- dEQP-VK.api.info.instance.*
- dEQP-VK.api.info.device.*
v2:
- Reject VkNativeBufferANDROID if the dma-buf's size is too small for
the VkImage.
- Stop abusing VkNativeBufferANDROID by passing it to vkAllocateMemory
during vkCreateImage. Instead, directly import its dma-buf during
vkCreateImage with anv_bo_cache_import(). [for jekstrand]
- Rebase onto Tapani's VK_EXT_debug_report changes.
- Drop `CPPFLAGS += $(top_srcdir)/include/android`. The dir does not
exist.
v3:
- Delete duplicate #include "anv_private.h". [per Tapani]
- Try to fix the Android-IA build in Android.vulkan.mk by following
Tapani's example.
v4:
- Unset EXEC_OBJECT_ASYNC and set EXEC_OBJECT_WRITE on the imported
gralloc buffer, just as we do for all other winsys buffers in
anv_wsi.c. [found by Tapani]
v5:
- Really fix the Android-IA build by ensuring that Android.vulkan.mk
uses Mesa' vulkan.h and not Android's. Insert -I$(MESA_TOP)/include
before -Iframeworks/native/vulkan/include. [for Tapani]
- In vkAcquireImageANDROID, submit signal operations to the
VkSemaphore and VkFence. [for zhou]
v6:
- Drop copy-paste duplication in vkGetSwapchainGrallocUsageANDROID().
[found by zhou]
- Improve comments in vkGetSwapchainGrallocUsageANDROID().
v7:
- Fix vkGetSwapchainGrallocUsageANDROID() to inspect its
VkImageUsageFlags parameter. [for tfiga]
- This fix regresses dEQP-VK.api.wsi.android.swapchain.*.image_usage
because dEQP wants to create swapchains with
VK_IMAGE_USAGE_STORAGE_BIT.
v8:
- Drop unneeded goto in vkAcquireImageANDROID. [for tfiga]
v8.1: (minor changes)
- Drop errant hunks added by rerere in anv_device.c.
- Drop explicit mention of VK_ANDROID_native_buffer in
anv_entrypoints_gen.py. [for jekstrand]
v9:
- Isolate as much Android code as possible, moving it from anv_image.c
to anv_android.c. Connect the files with anv_image_from_gralloc().
Remove VkNativeBufferANDROID params from all anv_image.c
funcs. [for krh]
- Replace some intel_loge() with vk_errorf() in anv_android.c.
- Use © in copyright line. [for krh]
Reviewed-by: Tapani Pälli <tapani.palli@intel.com> (v5)
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com> (v9)
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net> (v9)
Cc: zhoucm1 <david1.zhou@amd.com>
Cc: Tomasz Figa <tfiga@chromium.org>
2016-11-14 16:13:51 -08:00
|
|
|
goto fail_size;
|
|
|
|
|
}
|
|
|
|
|
|
2020-12-14 20:13:30 -08:00
|
|
|
assert(!image->disjoint);
|
anv: Implement VK_ANDROID_native_buffer (v9)
This implementation is correct (afaict), but takes two shortcuts
regarding the import/export of Android sync fds.
Shortcut 1. When Android calls vkAcquireImageANDROID to import a sync
fd into a VkSemaphore or VkFence, the driver instead simply blocks on
the sync fd, then puts the VkSemaphore or VkFence into the signalled
state. Thanks to implicit sync, this produces correct behavior (with
extra latency overhead, perhaps) despite its ugliness.
Shortcut 2. When Android calls vkQueueSignalReleaseImageANDROID to export
a collection of wait semaphores as a sync fd, the driver instead
submits the semaphores to the queue, then returns sync fd -1, which
informs the caller that no additional synchronization is needed.
Again, thanks to implicit sync, this produces correct behavior (with
extra batch submission overhead) despite its ugliness.
I chose to take the shortcuts instead of properly importing/exporting
the sync fds for two reasons:
Reason 1. I've already tested this patch with dEQP and with demos
apps. It works. I wanted to get the tested patches into the tree now,
and polish the implementation afterwards.
Reason 2. I want to run this on a 3.18 kernel (gasp!). In 3.18, i915
supports neither Android's sync_fence, nor upstream's sync_file, nor
drm_syncobj. Again, I tested these patches on Android with a 3.18
kernel and they work.
I plan to quickly follow-up with patches that remove the shortcuts and
properly import/export the sync fds.
Non-Testing
===========
I did not test at all using the Android.mk buildsystem. I may have broke
it. Please test and review that.
Testing
=======
I tested with 64-bit ARC++ on a Skylake Chromebook and a 3.18 kernel.
The following pass (as of patchset v9):
- a little spinning cube demo APK
- several Sascha demos
- dEQP-VK.info.*
- dEQP-VK.api.wsi.android.*
(except dEQP-VK.api.wsi.android.swapchain.*.image_usage, because
dEQP wants to create swapchains with VK_IMAGE_USAGE_STORAGE_BIT)
- dEQP-VK.api.smoke.*
- dEQP-VK.api.info.instance.*
- dEQP-VK.api.info.device.*
v2:
- Reject VkNativeBufferANDROID if the dma-buf's size is too small for
the VkImage.
- Stop abusing VkNativeBufferANDROID by passing it to vkAllocateMemory
during vkCreateImage. Instead, directly import its dma-buf during
vkCreateImage with anv_bo_cache_import(). [for jekstrand]
- Rebase onto Tapani's VK_EXT_debug_report changes.
- Drop `CPPFLAGS += $(top_srcdir)/include/android`. The dir does not
exist.
v3:
- Delete duplicate #include "anv_private.h". [per Tapani]
- Try to fix the Android-IA build in Android.vulkan.mk by following
Tapani's example.
v4:
- Unset EXEC_OBJECT_ASYNC and set EXEC_OBJECT_WRITE on the imported
gralloc buffer, just as we do for all other winsys buffers in
anv_wsi.c. [found by Tapani]
v5:
- Really fix the Android-IA build by ensuring that Android.vulkan.mk
uses Mesa' vulkan.h and not Android's. Insert -I$(MESA_TOP)/include
before -Iframeworks/native/vulkan/include. [for Tapani]
- In vkAcquireImageANDROID, submit signal operations to the
VkSemaphore and VkFence. [for zhou]
v6:
- Drop copy-paste duplication in vkGetSwapchainGrallocUsageANDROID().
[found by zhou]
- Improve comments in vkGetSwapchainGrallocUsageANDROID().
v7:
- Fix vkGetSwapchainGrallocUsageANDROID() to inspect its
VkImageUsageFlags parameter. [for tfiga]
- This fix regresses dEQP-VK.api.wsi.android.swapchain.*.image_usage
because dEQP wants to create swapchains with
VK_IMAGE_USAGE_STORAGE_BIT.
v8:
- Drop unneeded goto in vkAcquireImageANDROID. [for tfiga]
v8.1: (minor changes)
- Drop errant hunks added by rerere in anv_device.c.
- Drop explicit mention of VK_ANDROID_native_buffer in
anv_entrypoints_gen.py. [for jekstrand]
v9:
- Isolate as much Android code as possible, moving it from anv_image.c
to anv_android.c. Connect the files with anv_image_from_gralloc().
Remove VkNativeBufferANDROID params from all anv_image.c
funcs. [for krh]
- Replace some intel_loge() with vk_errorf() in anv_android.c.
- Use © in copyright line. [for krh]
Reviewed-by: Tapani Pälli <tapani.palli@intel.com> (v5)
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com> (v9)
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net> (v9)
Cc: zhoucm1 <david1.zhou@amd.com>
Cc: Tomasz Figa <tfiga@chromium.org>
2016-11-14 16:13:51 -08:00
|
|
|
assert(image->n_planes == 1);
|
2020-12-14 20:13:30 -08:00
|
|
|
assert(image->planes[0].primary_surface.memory_range.binding ==
|
|
|
|
|
ANV_IMAGE_MEMORY_BINDING_MAIN);
|
|
|
|
|
assert(image->bindings[ANV_IMAGE_MEMORY_BINDING_MAIN].address.bo == NULL);
|
|
|
|
|
assert(image->bindings[ANV_IMAGE_MEMORY_BINDING_MAIN].address.offset == 0);
|
|
|
|
|
image->bindings[ANV_IMAGE_MEMORY_BINDING_MAIN].address.bo = bo;
|
2021-02-05 12:07:07 -08:00
|
|
|
image->from_gralloc = true;
|
anv: Implement VK_ANDROID_native_buffer (v9)
This implementation is correct (afaict), but takes two shortcuts
regarding the import/export of Android sync fds.
Shortcut 1. When Android calls vkAcquireImageANDROID to import a sync
fd into a VkSemaphore or VkFence, the driver instead simply blocks on
the sync fd, then puts the VkSemaphore or VkFence into the signalled
state. Thanks to implicit sync, this produces correct behavior (with
extra latency overhead, perhaps) despite its ugliness.
Shortcut 2. When Android calls vkQueueSignalReleaseImageANDROID to export
a collection of wait semaphores as a sync fd, the driver instead
submits the semaphores to the queue, then returns sync fd -1, which
informs the caller that no additional synchronization is needed.
Again, thanks to implicit sync, this produces correct behavior (with
extra batch submission overhead) despite its ugliness.
I chose to take the shortcuts instead of properly importing/exporting
the sync fds for two reasons:
Reason 1. I've already tested this patch with dEQP and with demos
apps. It works. I wanted to get the tested patches into the tree now,
and polish the implementation afterwards.
Reason 2. I want to run this on a 3.18 kernel (gasp!). In 3.18, i915
supports neither Android's sync_fence, nor upstream's sync_file, nor
drm_syncobj. Again, I tested these patches on Android with a 3.18
kernel and they work.
I plan to quickly follow-up with patches that remove the shortcuts and
properly import/export the sync fds.
Non-Testing
===========
I did not test at all using the Android.mk buildsystem. I may have broke
it. Please test and review that.
Testing
=======
I tested with 64-bit ARC++ on a Skylake Chromebook and a 3.18 kernel.
The following pass (as of patchset v9):
- a little spinning cube demo APK
- several Sascha demos
- dEQP-VK.info.*
- dEQP-VK.api.wsi.android.*
(except dEQP-VK.api.wsi.android.swapchain.*.image_usage, because
dEQP wants to create swapchains with VK_IMAGE_USAGE_STORAGE_BIT)
- dEQP-VK.api.smoke.*
- dEQP-VK.api.info.instance.*
- dEQP-VK.api.info.device.*
v2:
- Reject VkNativeBufferANDROID if the dma-buf's size is too small for
the VkImage.
- Stop abusing VkNativeBufferANDROID by passing it to vkAllocateMemory
during vkCreateImage. Instead, directly import its dma-buf during
vkCreateImage with anv_bo_cache_import(). [for jekstrand]
- Rebase onto Tapani's VK_EXT_debug_report changes.
- Drop `CPPFLAGS += $(top_srcdir)/include/android`. The dir does not
exist.
v3:
- Delete duplicate #include "anv_private.h". [per Tapani]
- Try to fix the Android-IA build in Android.vulkan.mk by following
Tapani's example.
v4:
- Unset EXEC_OBJECT_ASYNC and set EXEC_OBJECT_WRITE on the imported
gralloc buffer, just as we do for all other winsys buffers in
anv_wsi.c. [found by Tapani]
v5:
- Really fix the Android-IA build by ensuring that Android.vulkan.mk
uses Mesa' vulkan.h and not Android's. Insert -I$(MESA_TOP)/include
before -Iframeworks/native/vulkan/include. [for Tapani]
- In vkAcquireImageANDROID, submit signal operations to the
VkSemaphore and VkFence. [for zhou]
v6:
- Drop copy-paste duplication in vkGetSwapchainGrallocUsageANDROID().
[found by zhou]
- Improve comments in vkGetSwapchainGrallocUsageANDROID().
v7:
- Fix vkGetSwapchainGrallocUsageANDROID() to inspect its
VkImageUsageFlags parameter. [for tfiga]
- This fix regresses dEQP-VK.api.wsi.android.swapchain.*.image_usage
because dEQP wants to create swapchains with
VK_IMAGE_USAGE_STORAGE_BIT.
v8:
- Drop unneeded goto in vkAcquireImageANDROID. [for tfiga]
v8.1: (minor changes)
- Drop errant hunks added by rerere in anv_device.c.
- Drop explicit mention of VK_ANDROID_native_buffer in
anv_entrypoints_gen.py. [for jekstrand]
v9:
- Isolate as much Android code as possible, moving it from anv_image.c
to anv_android.c. Connect the files with anv_image_from_gralloc().
Remove VkNativeBufferANDROID params from all anv_image.c
funcs. [for krh]
- Replace some intel_loge() with vk_errorf() in anv_android.c.
- Use © in copyright line. [for krh]
Reviewed-by: Tapani Pälli <tapani.palli@intel.com> (v5)
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com> (v9)
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net> (v9)
Cc: zhoucm1 <david1.zhou@amd.com>
Cc: Tomasz Figa <tfiga@chromium.org>
2016-11-14 16:13:51 -08:00
|
|
|
|
|
|
|
|
return VK_SUCCESS;
|
|
|
|
|
|
|
|
|
|
fail_size:
|
2021-07-28 13:36:51 +03:00
|
|
|
anv_image_finish(image);
|
|
|
|
|
fail_init:
|
2019-10-31 10:28:05 -05:00
|
|
|
anv_device_release_bo(device, bo);
|
anv: Implement VK_ANDROID_native_buffer (v9)
This implementation is correct (afaict), but takes two shortcuts
regarding the import/export of Android sync fds.
Shortcut 1. When Android calls vkAcquireImageANDROID to import a sync
fd into a VkSemaphore or VkFence, the driver instead simply blocks on
the sync fd, then puts the VkSemaphore or VkFence into the signalled
state. Thanks to implicit sync, this produces correct behavior (with
extra latency overhead, perhaps) despite its ugliness.
Shortcut 2. When Android calls vkQueueSignalReleaseImageANDROID to export
a collection of wait semaphores as a sync fd, the driver instead
submits the semaphores to the queue, then returns sync fd -1, which
informs the caller that no additional synchronization is needed.
Again, thanks to implicit sync, this produces correct behavior (with
extra batch submission overhead) despite its ugliness.
I chose to take the shortcuts instead of properly importing/exporting
the sync fds for two reasons:
Reason 1. I've already tested this patch with dEQP and with demos
apps. It works. I wanted to get the tested patches into the tree now,
and polish the implementation afterwards.
Reason 2. I want to run this on a 3.18 kernel (gasp!). In 3.18, i915
supports neither Android's sync_fence, nor upstream's sync_file, nor
drm_syncobj. Again, I tested these patches on Android with a 3.18
kernel and they work.
I plan to quickly follow-up with patches that remove the shortcuts and
properly import/export the sync fds.
Non-Testing
===========
I did not test at all using the Android.mk buildsystem. I may have broke
it. Please test and review that.
Testing
=======
I tested with 64-bit ARC++ on a Skylake Chromebook and a 3.18 kernel.
The following pass (as of patchset v9):
- a little spinning cube demo APK
- several Sascha demos
- dEQP-VK.info.*
- dEQP-VK.api.wsi.android.*
(except dEQP-VK.api.wsi.android.swapchain.*.image_usage, because
dEQP wants to create swapchains with VK_IMAGE_USAGE_STORAGE_BIT)
- dEQP-VK.api.smoke.*
- dEQP-VK.api.info.instance.*
- dEQP-VK.api.info.device.*
v2:
- Reject VkNativeBufferANDROID if the dma-buf's size is too small for
the VkImage.
- Stop abusing VkNativeBufferANDROID by passing it to vkAllocateMemory
during vkCreateImage. Instead, directly import its dma-buf during
vkCreateImage with anv_bo_cache_import(). [for jekstrand]
- Rebase onto Tapani's VK_EXT_debug_report changes.
- Drop `CPPFLAGS += $(top_srcdir)/include/android`. The dir does not
exist.
v3:
- Delete duplicate #include "anv_private.h". [per Tapani]
- Try to fix the Android-IA build in Android.vulkan.mk by following
Tapani's example.
v4:
- Unset EXEC_OBJECT_ASYNC and set EXEC_OBJECT_WRITE on the imported
gralloc buffer, just as we do for all other winsys buffers in
anv_wsi.c. [found by Tapani]
v5:
- Really fix the Android-IA build by ensuring that Android.vulkan.mk
uses Mesa' vulkan.h and not Android's. Insert -I$(MESA_TOP)/include
before -Iframeworks/native/vulkan/include. [for Tapani]
- In vkAcquireImageANDROID, submit signal operations to the
VkSemaphore and VkFence. [for zhou]
v6:
- Drop copy-paste duplication in vkGetSwapchainGrallocUsageANDROID().
[found by zhou]
- Improve comments in vkGetSwapchainGrallocUsageANDROID().
v7:
- Fix vkGetSwapchainGrallocUsageANDROID() to inspect its
VkImageUsageFlags parameter. [for tfiga]
- This fix regresses dEQP-VK.api.wsi.android.swapchain.*.image_usage
because dEQP wants to create swapchains with
VK_IMAGE_USAGE_STORAGE_BIT.
v8:
- Drop unneeded goto in vkAcquireImageANDROID. [for tfiga]
v8.1: (minor changes)
- Drop errant hunks added by rerere in anv_device.c.
- Drop explicit mention of VK_ANDROID_native_buffer in
anv_entrypoints_gen.py. [for jekstrand]
v9:
- Isolate as much Android code as possible, moving it from anv_image.c
to anv_android.c. Connect the files with anv_image_from_gralloc().
Remove VkNativeBufferANDROID params from all anv_image.c
funcs. [for krh]
- Replace some intel_loge() with vk_errorf() in anv_android.c.
- Use © in copyright line. [for krh]
Reviewed-by: Tapani Pälli <tapani.palli@intel.com> (v5)
Reviewed-by: Kristian H. Kristensen <hoegsberg@google.com> (v9)
Reviewed-by: Jason Ekstrand <jason@jlekstrand.net> (v9)
Cc: zhoucm1 <david1.zhou@amd.com>
Cc: Tomasz Figa <tfiga@chromium.org>
2016-11-14 16:13:51 -08:00
|
|
|
|
|
|
|
|
return result;
|
|
|
|
|
}
|