video_core: Enable ImageGather with subpixel offset on Intel
This commit is contained in:
parent
780240e697
commit
fe91066f46
|
@ -52,9 +52,9 @@ struct Profile {
|
||||||
bool need_declared_frag_colors{};
|
bool need_declared_frag_colors{};
|
||||||
/// Prevents fast math optimizations that may cause inaccuracies
|
/// Prevents fast math optimizations that may cause inaccuracies
|
||||||
bool need_fastmath_off{};
|
bool need_fastmath_off{};
|
||||||
/// Some GPU vendors use a lower fixed point format of 16.8 when calculating pixel coordinates
|
/// Some GPU vendors use a different rounding precision when calculating texture pixel
|
||||||
/// in the ImageGather instruction than the Maxwell architecture does. Applying an offset does
|
/// coordinates with the 16.8 format in the ImageGather instruction than the Maxwell
|
||||||
/// fix this mismatching rounding behaviour.
|
/// architecture. Applying an offset does fix this mismatching rounding behaviour.
|
||||||
bool need_gather_subpixel_offset{};
|
bool need_gather_subpixel_offset{};
|
||||||
|
|
||||||
/// OpFClamp is broken and OpFMax + OpFMin should be used instead
|
/// OpFClamp is broken and OpFMax + OpFMin should be used instead
|
||||||
|
|
|
@ -169,7 +169,6 @@ Device::Device(Core::Frontend::EmuWindow& emu_window) {
|
||||||
has_draw_texture = GLAD_GL_NV_draw_texture;
|
has_draw_texture = GLAD_GL_NV_draw_texture;
|
||||||
warp_size_potentially_larger_than_guest = !is_nvidia && !is_intel;
|
warp_size_potentially_larger_than_guest = !is_nvidia && !is_intel;
|
||||||
need_fastmath_off = is_nvidia;
|
need_fastmath_off = is_nvidia;
|
||||||
need_gather_subpixel_offset = is_amd;
|
|
||||||
can_report_memory = GLAD_GL_NVX_gpu_memory_info;
|
can_report_memory = GLAD_GL_NVX_gpu_memory_info;
|
||||||
|
|
||||||
// At the moment of writing this, only Nvidia's driver optimizes BufferSubData on exclusive
|
// At the moment of writing this, only Nvidia's driver optimizes BufferSubData on exclusive
|
||||||
|
|
|
@ -160,10 +160,6 @@ public:
|
||||||
return need_fastmath_off;
|
return need_fastmath_off;
|
||||||
}
|
}
|
||||||
|
|
||||||
bool NeedsGatherSubpixelOffset() const {
|
|
||||||
return need_gather_subpixel_offset;
|
|
||||||
}
|
|
||||||
|
|
||||||
bool HasCbufFtouBug() const {
|
bool HasCbufFtouBug() const {
|
||||||
return has_cbuf_ftou_bug;
|
return has_cbuf_ftou_bug;
|
||||||
}
|
}
|
||||||
|
@ -180,6 +176,10 @@ public:
|
||||||
return vendor_name == "ATI Technologies Inc.";
|
return vendor_name == "ATI Technologies Inc.";
|
||||||
}
|
}
|
||||||
|
|
||||||
|
bool IsIntel() const {
|
||||||
|
return vendor_name == "Intel";
|
||||||
|
}
|
||||||
|
|
||||||
bool CanReportMemoryUsage() const {
|
bool CanReportMemoryUsage() const {
|
||||||
return can_report_memory;
|
return can_report_memory;
|
||||||
}
|
}
|
||||||
|
@ -229,7 +229,6 @@ private:
|
||||||
bool has_draw_texture{};
|
bool has_draw_texture{};
|
||||||
bool warp_size_potentially_larger_than_guest{};
|
bool warp_size_potentially_larger_than_guest{};
|
||||||
bool need_fastmath_off{};
|
bool need_fastmath_off{};
|
||||||
bool need_gather_subpixel_offset{};
|
|
||||||
bool has_cbuf_ftou_bug{};
|
bool has_cbuf_ftou_bug{};
|
||||||
bool has_bool_ref_bug{};
|
bool has_bool_ref_bug{};
|
||||||
bool can_report_memory{};
|
bool can_report_memory{};
|
||||||
|
|
|
@ -218,7 +218,7 @@ ShaderCache::ShaderCache(RasterizerOpenGL& rasterizer_, Core::Frontend::EmuWindo
|
||||||
.lower_left_origin_mode = true,
|
.lower_left_origin_mode = true,
|
||||||
.need_declared_frag_colors = true,
|
.need_declared_frag_colors = true,
|
||||||
.need_fastmath_off = device.NeedsFastmathOff(),
|
.need_fastmath_off = device.NeedsFastmathOff(),
|
||||||
.need_gather_subpixel_offset = device.NeedsGatherSubpixelOffset(),
|
.need_gather_subpixel_offset = device.IsAmd() || device.IsIntel(),
|
||||||
|
|
||||||
.has_broken_spirv_clamp = true,
|
.has_broken_spirv_clamp = true,
|
||||||
.has_broken_unsigned_image_offsets = true,
|
.has_broken_unsigned_image_offsets = true,
|
||||||
|
|
|
@ -329,7 +329,9 @@ PipelineCache::PipelineCache(RasterizerVulkan& rasterizer_, const Device& device
|
||||||
|
|
||||||
.lower_left_origin_mode = false,
|
.lower_left_origin_mode = false,
|
||||||
.need_declared_frag_colors = false,
|
.need_declared_frag_colors = false,
|
||||||
.need_gather_subpixel_offset = device.NeedsGatherSubpixelOffset(),
|
.need_gather_subpixel_offset = driver_id == VK_DRIVER_ID_AMD_PROPRIETARY ||
|
||||||
|
driver_id == VK_DRIVER_ID_INTEL_PROPRIETARY_WINDOWS ||
|
||||||
|
driver_id == VK_DRIVER_ID_INTEL_OPEN_SOURCE_MESA,
|
||||||
|
|
||||||
.has_broken_spirv_clamp = driver_id == VK_DRIVER_ID_INTEL_PROPRIETARY_WINDOWS,
|
.has_broken_spirv_clamp = driver_id == VK_DRIVER_ID_INTEL_PROPRIETARY_WINDOWS,
|
||||||
.has_broken_spirv_position_input = driver_id == VK_DRIVER_ID_QUALCOMM_PROPRIETARY,
|
.has_broken_spirv_position_input = driver_id == VK_DRIVER_ID_QUALCOMM_PROPRIETARY,
|
||||||
|
|
|
@ -431,7 +431,6 @@ Device::Device(VkInstance instance_, vk::PhysicalDevice physical_, VkSurfaceKHR
|
||||||
"AMD GCN4 and earlier have broken VK_IMAGE_CREATE_CUBE_COMPATIBLE_BIT");
|
"AMD GCN4 and earlier have broken VK_IMAGE_CREATE_CUBE_COMPATIBLE_BIT");
|
||||||
has_broken_cube_compatibility = true;
|
has_broken_cube_compatibility = true;
|
||||||
}
|
}
|
||||||
need_gather_subpixel_offset = true;
|
|
||||||
}
|
}
|
||||||
if (extensions.sampler_filter_minmax && is_amd) {
|
if (extensions.sampler_filter_minmax && is_amd) {
|
||||||
// Disable ext_sampler_filter_minmax on AMD GCN4 and lower as it is broken.
|
// Disable ext_sampler_filter_minmax on AMD GCN4 and lower as it is broken.
|
||||||
|
|
|
@ -554,10 +554,6 @@ public:
|
||||||
return features.robustness2.nullDescriptor;
|
return features.robustness2.nullDescriptor;
|
||||||
}
|
}
|
||||||
|
|
||||||
bool NeedsGatherSubpixelOffset() const {
|
|
||||||
return need_gather_subpixel_offset;
|
|
||||||
}
|
|
||||||
|
|
||||||
u32 GetMaxVertexInputAttributes() const {
|
u32 GetMaxVertexInputAttributes() const {
|
||||||
return properties.properties.limits.maxVertexInputAttributes;
|
return properties.properties.limits.maxVertexInputAttributes;
|
||||||
}
|
}
|
||||||
|
@ -668,7 +664,6 @@ private:
|
||||||
bool must_emulate_bgr565{}; ///< Emulates BGR565 by swizzling RGB565 format.
|
bool must_emulate_bgr565{}; ///< Emulates BGR565 by swizzling RGB565 format.
|
||||||
bool dynamic_state3_blending{}; ///< Has all blending features of dynamic_state3.
|
bool dynamic_state3_blending{}; ///< Has all blending features of dynamic_state3.
|
||||||
bool dynamic_state3_enables{}; ///< Has all enables features of dynamic_state3.
|
bool dynamic_state3_enables{}; ///< Has all enables features of dynamic_state3.
|
||||||
bool need_gather_subpixel_offset{}; ///< Needs offset at ImageGather for correct rounding.
|
|
||||||
u64 device_access_memory{}; ///< Total size of device local memory in bytes.
|
u64 device_access_memory{}; ///< Total size of device local memory in bytes.
|
||||||
u32 sets_per_pool{}; ///< Sets per Description Pool
|
u32 sets_per_pool{}; ///< Sets per Description Pool
|
||||||
|
|
||||||
|
|
Reference in New Issue