Ryujinx/moltenvk
1
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

Work around problems with explicit LoD with arrayed depth images on Apple Silicon.

Browse Source 
Use an explicit gradient to make it sample the correct level.

Update SPIRV-Cross to pull in the change needed for this.
This commit is contained in:
Chip Davis 2023-02-03 14:46:37 -08:00
parent e8885a24e8
commit 2c01c8b7e0
5 changed files with 12 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

2
Docs/Whats_New.md
View File

@ -20,6 +20,8 @@ Released TBA
- Fix issue where extension `VK_KHR_fragment_shader_barycentric` - Fix issue where extension `VK_KHR_fragment_shader_barycentric`
was sometimes incorrectly disabled due to a Metal driver bug. was sometimes incorrectly disabled due to a Metal driver bug.
- Work around problems with using explicit LoD with arrayed depth images
on Apple Silicon.

2
ExternalRevisions/SPIRV-Cross_repo_revision
View File

@ -1 +1 @@
eb9b2732982b8eea3395216be825a74dd84fe4be 4e2fdb25671c742a9fbe93a6034eb1542244c7e1

1
MoltenVK/MoltenVK/API/vk_mvk_moltenvk.h
View File

@ -979,6 +979,7 @@ typedef struct {
VkBool32 programmableSamplePositions; /**< If true, programmable MSAA sample positions are supported. */ VkBool32 programmableSamplePositions; /**< If true, programmable MSAA sample positions are supported. */
VkBool32 shaderBarycentricCoordinates; /**< If true, fragment shader barycentric coordinates are supported. */ VkBool32 shaderBarycentricCoordinates; /**< If true, fragment shader barycentric coordinates are supported. */
MTLArgumentBuffersTier argumentBuffersTier; /**< The argument buffer tier available on this device, as a Metal enumeration. */ MTLArgumentBuffersTier argumentBuffersTier; /**< The argument buffer tier available on this device, as a Metal enumeration. */
VkBool32 needsSampleDrefLodArrayWorkaround; /**< If true, sampling from arrayed depth images with explicit LoD is broken and needs a workaround. */
} MVKPhysicalDeviceMetalFeatures; } MVKPhysicalDeviceMetalFeatures;
/** MoltenVK performance of a particular type of activity. */ /** MoltenVK performance of a particular type of activity. */

3
MoltenVK/MoltenVK/GPUObjects/MVKDevice.mm
View File

@ -1493,6 +1493,9 @@ void MVKPhysicalDevice::initMetalFeatures() {
_metalFeatures.clearColorFloatRounding = MVK_FLOAT_ROUNDING_DOWN; _metalFeatures.clearColorFloatRounding = MVK_FLOAT_ROUNDING_DOWN;
break; break;
case kAppleVendorId: case kAppleVendorId:
// TODO: Other GPUs?
_metalFeatures.needsSampleDrefLodArrayWorkaround = true;
// fallthrough
case kIntelVendorId: case kIntelVendorId:
case kNVVendorId: case kNVVendorId:
default: default:

6
MoltenVK/MoltenVK/GPUObjects/MVKPipeline.mm
View File

@ -1152,6 +1152,9 @@ bool MVKGraphicsPipeline::addFragmentShaderToPipeline(MTLRenderPipelineDescripto
shaderConfig.options.mslOptions.capture_output_to_buffer = false; shaderConfig.options.mslOptions.capture_output_to_buffer = false;
shaderConfig.options.mslOptions.fixed_subgroup_size = mvkIsAnyFlagEnabled(_pFragmentSS->flags, VK_PIPELINE_SHADER_STAGE_CREATE_ALLOW_VARYING_SUBGROUP_SIZE_BIT_EXT) ? 0 : _device->_pMetalFeatures->maxSubgroupSize; shaderConfig.options.mslOptions.fixed_subgroup_size = mvkIsAnyFlagEnabled(_pFragmentSS->flags, VK_PIPELINE_SHADER_STAGE_CREATE_ALLOW_VARYING_SUBGROUP_SIZE_BIT_EXT) ? 0 : _device->_pMetalFeatures->maxSubgroupSize;
shaderConfig.options.mslOptions.check_discarded_frag_stores = true; shaderConfig.options.mslOptions.check_discarded_frag_stores = true;
if (_device->_pMetalFeatures->needsSampleDrefLodArrayWorkaround) {
shaderConfig.options.mslOptions.sample_dref_lod_array_as_grad = true;
}
if (_isRasterizing && pCreateInfo->pMultisampleState) { // Must ignore allowed bad pMultisampleState pointer if rasterization disabled if (_isRasterizing && pCreateInfo->pMultisampleState) { // Must ignore allowed bad pMultisampleState pointer if rasterization disabled
if (pCreateInfo->pMultisampleState->pSampleMask && pCreateInfo->pMultisampleState->pSampleMask[0] != 0xffffffff) { if (pCreateInfo->pMultisampleState->pSampleMask && pCreateInfo->pMultisampleState->pSampleMask[0] != 0xffffffff) {
shaderConfig.options.mslOptions.additional_fixed_sample_mask = pCreateInfo->pMultisampleState->pSampleMask[0]; shaderConfig.options.mslOptions.additional_fixed_sample_mask = pCreateInfo->pMultisampleState->pSampleMask[0];
@ -2253,7 +2256,8 @@ namespace SPIRV_CROSS_NAMESPACE {
opt.vertex_index_type, opt.vertex_index_type,
opt.force_sample_rate_shading, opt.force_sample_rate_shading,
opt.manual_helper_invocation_updates, opt.manual_helper_invocation_updates,
opt.check_discarded_frag_stores); opt.check_discarded_frag_stores,
opt.sample_dref_lod_array_as_grad);
} }
template<class Archive> template<class Archive>