Merge pull request #1298 from billhollings/auto-gpu-capture-fixes
Fixes to auto GPU capture
diff --git a/MoltenVK/MoltenVK/API/vk_mvk_moltenvk.h b/MoltenVK/MoltenVK/API/vk_mvk_moltenvk.h
index ed74893..3043525 100644
--- a/MoltenVK/MoltenVK/API/vk_mvk_moltenvk.h
+++ b/MoltenVK/MoltenVK/API/vk_mvk_moltenvk.h
@@ -58,6 +58,15 @@
#define VK_MVK_MOLTENVK_SPEC_VERSION 31
#define VK_MVK_MOLTENVK_EXTENSION_NAME "VK_MVK_moltenvk"
+/** Identifies the scope for Metal to run an automatic GPU capture for diagnostic debugging purposes. */
+typedef enum MVKConfigAutoGPUCaptureScopeBits {
+ MVK_CONFIG_AUTO_GPU_CAPTURE_SCOPE_NONE = 0, /**< No automatic GPU capture. */
+ MVK_CONFIG_AUTO_GPU_CAPTURE_SCOPE_DEVICE = 1, /**< Automatically capture all GPU activity during the lifetime of a VkDevice. */
+ MVK_CONFIG_AUTO_GPU_CAPTURE_SCOPE_FRAME = 2, /**< Automatically capture all GPU activity during the rendering and presentation of the first frame. */
+ MVK_CONFIG_AUTO_GPU_CAPTURE_SCOPE_MAX_ENUM = 0x7FFFFFFF
+} MVKConfigAutoGPUCaptureScopeBits;
+typedef VkFlags MVKConfigAutoGPUCaptureScope;
+
/** Identifies extensions to advertise as part of MoltenVK configuration. */
typedef enum MVKConfigAdvertiseExtensionBits {
MVK_CONFIG_ADVERTISE_EXTENSIONS_ALL = 0x00000001, /**< All supported extensions. */
@@ -390,18 +399,21 @@
VkBool32 switchSystemGPU;
/**
- * If enabled, arbitrary VkImageView component swizzles are supported, as defined
- * in VkImageViewCreateInfo::components when creating a VkImageView.
+ * Older versions of Metal do not natively support per-texture swizzling. When running on
+ * such a system, and this parameter is enabled, arbitrary VkImageView component swizzles
+ * are supported, as defined in VkImageViewCreateInfo::components when creating a VkImageView.
*
- * If disabled, a very limited set of VkImageView component swizzles are supported
- * via format substitutions.
+ * If disabled, and native Metal per-texture swizzling is not available on the platform,
+ * a very limited set of VkImageView component swizzles are supported via format substitutions.
*
- * Metal does not natively support per-texture swizzling. If this parameter is enabled
- * both when a VkImageView is created, and when any pipeline that uses that VkImageView
- * is compiled, VkImageView swizzling is automatically performed in the converted Metal
- * shader code during all texture sampling and reading operations, regardless of whether
- * a swizzle is required for the VkImageView associated with the Metal texture.
- * This may result in reduced performance.
+ * If Metal supports native per-texture swizzling, this parameter is ignored.
+
+ * When running on an older version of Metal that does not support native per-texture
+ * swizzling, if this parameter is enabled, both when a VkImageView is created, and
+ * when any pipeline that uses that VkImageView is compiled, VkImageView swizzling is
+ * automatically performed in the converted Metal shader code during all texture sampling
+ * and reading operations, regardless of whether a swizzle is required for the VkImageView
+ * associated with the Metal texture. This may result in reduced performance.
*
* The value of this parameter may be changed at any time during application runtime,
* and the changed value will immediately effect subsequent MoltenVK behaviour.
@@ -420,8 +432,9 @@
* in a call to vkGetPhysicalDeviceImageFormatProperties2KHR() to query for an VkImageView
* format that will require full swizzling to be enabled, and this feature is not enabled.
*
- * If this parameter is disabled, the following limited set of VkImageView swizzles are
- * supported by MoltenVK, via automatic format substitution:
+ * If this parameter is disabled, and native Metal per-texture swizzling is not available
+ * on the platform, the following limited set of VkImageView swizzles are supported by
+ * MoltenVK, via automatic format substitution:
*
* Texture format Swizzle
* -------------- -------
@@ -585,15 +598,14 @@
* Controls whether Metal should run an automatic GPU capture without the user having to
* trigger it manually via the Xcode user interface, and controls the scope under which
* that GPU capture will occur. This is useful when trying to capture a one-shot GPU trace,
- * such as when running a Vulkan CTS test case. For the automatic GPU capture to occur,
- * the Xcode scheme under which the app is run must have the Metal GPU capture option
- * enabled. MVK_CONFIG_AUTO_GPU_CAPTURE_SCOPE should not be set to manually trigger a
- * GPU capture via the Xcode user interface.
+ * such as when running a Vulkan CTS test case. For the automatic GPU capture to occur, the
+ * Xcode scheme under which the app is run must have the Metal GPU capture option enabled.
+ * This parameter should not be set to manually trigger a GPU capture via the Xcode user interface.
*
- * To automatically trigger a GPU capture, set this value as follows:
- * 0: No automatic GPU capture.
- * 1: Capture all GPU commands issued during the lifetime of the VkDevice.
- * 2: Capture all GPU commands issued during the first rendered frame.
+ * When the value of this parameter is MVK_CONFIG_AUTO_GPU_CAPTURE_SCOPE_FRAME,
+ * the queue for which the GPU activity is captured is identifed by the values of
+ * the defaultGPUCaptureScopeQueueFamilyIndex and defaultGPUCaptureScopeQueueIndex
+ * configuration parameters.
*
* The value of this parameter must be changed before creating a VkDevice,
* for the change to take effect.
@@ -603,7 +615,7 @@
* runtime environment variable or MoltenVK compile-time build setting.
* If neither is set, no automatic GPU capture will occur.
*/
- uint32_t autoGPUCaptureScope;
+ MVKConfigAutoGPUCaptureScope autoGPUCaptureScope;
/**
* The path to a file where the automatic GPU capture should be saved, if autoGPUCaptureScope
diff --git a/MoltenVK/MoltenVK/GPUObjects/MVKDevice.h b/MoltenVK/MoltenVK/GPUObjects/MVKDevice.h
index 2a821d0..c65ffae 100644
--- a/MoltenVK/MoltenVK/GPUObjects/MVKDevice.h
+++ b/MoltenVK/MoltenVK/GPUObjects/MVKDevice.h
@@ -698,27 +698,17 @@
* capture scope, and if so, starts capturing from the specified Metal capture object.
* The capture will be made either to Xcode, or to a file if one has been configured.
*
- * The autoGPUCaptureScope parameter must be one of:
- * - MVK_CONFIG_AUTO_GPU_CAPTURE_SCOPE_NONE
- * - MVK_CONFIG_AUTO_GPU_CAPTURE_SCOPE_DEVICE
- * - MVK_CONFIG_AUTO_GPU_CAPTURE_SCOPE_FRAME
- *
* The mtlCaptureObject must be one of:
* - MTLDevice for scope MVK_CONFIG_AUTO_GPU_CAPTURE_SCOPE_DEVICE
* - MTLCommandQueue for scope MVK_CONFIG_AUTO_GPU_CAPTURE_SCOPE_FRAME.
*/
- void startAutoGPUCapture(int32_t autoGPUCaptureScope, id mtlCaptureObject);
+ void startAutoGPUCapture(MVKConfigAutoGPUCaptureScope autoGPUCaptureScope, id mtlCaptureObject);
/**
* Checks if automatic GPU capture is enabled for the specified
* auto capture scope, and if so, stops capturing.
- *
- * The autoGPUCaptureScope parameter must be one of:
- * - MVK_CONFIG_AUTO_GPU_CAPTURE_SCOPE_NONE
- * - MVK_CONFIG_AUTO_GPU_CAPTURE_SCOPE_DEVICE
- * - MVK_CONFIG_AUTO_GPU_CAPTURE_SCOPE_FRAME
*/
- void stopAutoGPUCapture(int32_t autoGPUCaptureScope);
+ void stopAutoGPUCapture(MVKConfigAutoGPUCaptureScope autoGPUCaptureScope);
/** Returns whether this instance is currently automatically capturing a GPU trace. */
inline bool isCurrentlyAutoGPUCapturing() { return _isCurrentlyAutoGPUCapturing; }
diff --git a/MoltenVK/MoltenVK/GPUObjects/MVKDevice.mm b/MoltenVK/MoltenVK/GPUObjects/MVKDevice.mm
index a023c20..9b9c923 100644
--- a/MoltenVK/MoltenVK/GPUObjects/MVKDevice.mm
+++ b/MoltenVK/MoltenVK/GPUObjects/MVKDevice.mm
@@ -3680,7 +3680,7 @@
_enabledInlineUniformBlockFeatures.descriptorBindingInlineUniformBlockUpdateAfterBind));
}
-void MVKDevice::startAutoGPUCapture(int32_t autoGPUCaptureScope, id mtlCaptureObject) {
+void MVKDevice::startAutoGPUCapture(MVKConfigAutoGPUCaptureScope autoGPUCaptureScope, id mtlCaptureObject) {
if (_isCurrentlyAutoGPUCapturing || (mvkGetMVKConfiguration()->autoGPUCaptureScope != autoGPUCaptureScope)) { return; }
@@ -3729,7 +3729,7 @@
}
}
-void MVKDevice::stopAutoGPUCapture(int32_t autoGPUCaptureScope) {
+void MVKDevice::stopAutoGPUCapture(MVKConfigAutoGPUCaptureScope autoGPUCaptureScope) {
if (_isCurrentlyAutoGPUCapturing && mvkGetMVKConfiguration()->autoGPUCaptureScope == autoGPUCaptureScope) {
[[MTLCaptureManager sharedCaptureManager] stopCapture];
_isCurrentlyAutoGPUCapturing = false;
diff --git a/MoltenVK/MoltenVK/GPUObjects/MVKQueue.mm b/MoltenVK/MoltenVK/GPUObjects/MVKQueue.mm
index c502387..7e549b3 100644
--- a/MoltenVK/MoltenVK/GPUObjects/MVKQueue.mm
+++ b/MoltenVK/MoltenVK/GPUObjects/MVKQueue.mm
@@ -214,13 +214,15 @@
void MVKQueue::initGPUCaptureScopes() {
_submissionCaptureScope = new MVKGPUCaptureScope(this);
- if (_queueFamily->getIndex() == mvkGetMVKConfiguration()->defaultGPUCaptureScopeQueueFamilyIndex &&
- _index == mvkGetMVKConfiguration()->defaultGPUCaptureScopeQueueIndex) {
+ const MVKConfiguration* pMVKConfig = mvkGetMVKConfiguration();
+ if (_queueFamily->getIndex() == pMVKConfig->defaultGPUCaptureScopeQueueFamilyIndex &&
+ _index == pMVKConfig->defaultGPUCaptureScopeQueueIndex) {
+
+ getDevice()->startAutoGPUCapture(MVK_CONFIG_AUTO_GPU_CAPTURE_SCOPE_FRAME, _mtlQueue);
_submissionCaptureScope->makeDefault();
+
}
_submissionCaptureScope->beginScope(); // Allow Xcode to capture the first frame if desired.
-
- getDevice()->startAutoGPUCapture(MVK_CONFIG_AUTO_GPU_CAPTURE_SCOPE_FRAME, _mtlQueue);
}
MVKQueue::~MVKQueue() {
diff --git a/MoltenVK/MoltenVK/Utility/MVKEnvironment.h b/MoltenVK/MoltenVK/Utility/MVKEnvironment.h
index 7aa09f2..e93ad4a 100644
--- a/MoltenVK/MoltenVK/Utility/MVKEnvironment.h
+++ b/MoltenVK/MoltenVK/Utility/MVKEnvironment.h
@@ -222,9 +222,6 @@
* developer having to trigger it manually via the Xcode UI. This is useful when trying
* to capture a one-shot trace, such as when running a Vulkan CTS test case.
*/
-#define MVK_CONFIG_AUTO_GPU_CAPTURE_SCOPE_NONE 0
-#define MVK_CONFIG_AUTO_GPU_CAPTURE_SCOPE_DEVICE 1
-#define MVK_CONFIG_AUTO_GPU_CAPTURE_SCOPE_FRAME 2
#ifndef MVK_CONFIG_AUTO_GPU_CAPTURE_SCOPE
# define MVK_CONFIG_AUTO_GPU_CAPTURE_SCOPE MVK_CONFIG_AUTO_GPU_CAPTURE_SCOPE_NONE
#endif
