Enhancement and bug fixes for 1.9.3 release

Misc infrastructure updates, update sampler support for bindless images, update new format support for images

scripts/Doxyfile

@@ -1226,7 +1226,7 @@ HTML_COLORSTYLE_GAMMA  = 80
 # The default value is: NO.
 # This tag requires that the tag GENERATE_HTML is set to YES.
 
-HTML_TIMESTAMP         = YES
+# HTML_TIMESTAMP         = YES
 
 # If the HTML_DYNAMIC_MENUS tag is set to YES then the generated HTML
 # documentation will contain a main index with vertical navigation menus that
@@ -1662,7 +1662,7 @@ EXTRA_SEARCH_MAPPINGS  =
 # If the GENERATE_LATEX tag is set to YES, doxygen will generate LaTeX output.
 # The default value is: YES.
 
-GENERATE_LATEX         = YES
+GENERATE_LATEX         = NO
 
 # The LATEX_OUTPUT tag is used to specify where the LaTeX docs will be put. If a
 # relative path is entered the value of OUTPUT_DIRECTORY will be put in front of
@@ -1854,7 +1854,7 @@ LATEX_BIB_STYLE        = plain
 # The default value is: NO.
 # This tag requires that the tag GENERATE_LATEX is set to YES.
 
-LATEX_TIMESTAMP        = YES
+# LATEX_TIMESTAMP        = YES
 
 # The LATEX_EMOJI_DIRECTORY tag is used to specify the (relative or absolute)
 # path from which the emoji images will be read. If a relative path is entered,
@@ -2142,7 +2142,8 @@ INCLUDE_FILE_PATTERNS  =
 # recursively expanded use the := operator instead of the = operator.
 # This tag requires that the tag ENABLE_PREPROCESSING is set to YES.
 
-PREDEFINED             = __cplusplus
+PREDEFINED             = __cplusplus \
+                         "module=modul3"
 
 # If the MACRO_EXPANSION and EXPAND_ONLY_PREDEF tags are set to YES then this
 # tag can be used to specify a list of macro names that should be expanded. The

scripts/core/EXT_EXP_BindlessImages.rst

@@ -47,10 +47,17 @@ In this extension, we propose the following additions:
  * Provide a new image descriptor and flags for Bindless images.
  * Support for creation of images on linearly allocated memory backed by USM.
  * Extension API to create an image handle from pitched memory
+ * Create Bindless sampled images
 
 A "Bindless image" can be created by passing ${x}_image_bindless_exp_desc_t to pNext member of
 ${x}_image_desc_t and set the flags value as ${X}_IMAGE_BINDLESS_EXP_FLAG_BINDLESS
 
+A "Bindless sampled image" can be created by passing ${x}_image_bindless_exp_desc_t to pNext member of
+${x}_image_desc_t and setting the flags to a combination of ${X}_IMAGE_BINDLESS_EXP_FLAG_BINDLESS and ${X}_IMAGE_BINDLESS_EXP_FLAG_SAMPLED_IMAGE
+When image view is created from bindless sampled image, sampling modes can be redefined by passing sampler descriptor in pNext field of ${x}_image_bindless_exp_desc_t struct.
+Image view created from bindless sampled image without setting ${X}_IMAGE_BINDLESS_EXP_FLAG_SAMPLED_IMAGE is an unsampled image.
+Sampled image view can be created from bindless unsampled image by setting ${X}_IMAGE_BINDLESS_EXP_FLAG_SAMPLED_IMAGE and passing sampler descriptor in pNext field of ${x}_image_bindless_exp_desc_t struct.
+
 This extension is complimentary to and may be used in conjunction with the `ZE_extension_image_view <https://spec.oneapi.io/level-zero/latest/core/EXT_ImageView.html#image-view-extension>`_ extension
 
 Programming example with Bindless images
@@ -178,4 +185,101 @@ Programming example with pitched memory usage
 
     // Once all operations on the image are complete we need destroy image handle and free memory
     ${x}ImageDestroy(hImage);
-    ${x}MemFree(hContext, pitchedPtr);
+    ${x}MemFree(hContext, pitchedPtr);
+
+Programming example with Bindless sampled images
+~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~
+
+.. parsed-literal::
+
+    // 2D image dimensions
+    size_t imageWidth = 1024;
+    size_t imageHeight = 1024;
+
+    // Single-precision float image format with one channel
+    ${x}_image_format_t imageFormat = {
+        ZE_IMAGE_FORMAT_LAYOUT_32, ZE_IMAGE_FORMAT_TYPE_FLOAT,
+        ZE_IMAGE_FORMAT_SWIZZLE_R, ZE_IMAGE_FORMAT_SWIZZLE_X,
+        ZE_IMAGE_FORMAT_SWIZZLE_R, ZE_IMAGE_FORMAT_SWIZZLE_X
+    }
+
+    // Define sampler descriptor
+    ${x}_sampler_desc_t samplerDesc = {
+        ZE_STRUCTURE_TYPE_SAMPLER_DESC,
+        nullptr,
+        ZE_SAMPLER_ADDRESS_MODE_CLAMP,
+        ZE_SAMPLER_FILTER_MODE_LINEAR,
+        true
+    };
+
+    // Create an image descriptor for bindless image
+    ${x}_image_desc_t imageDesc = {
+        ZE_STRUCTURE_TYPE_IMAGE_DESC,
+        nullptr,
+        0,
+        ZE_IMAGE_TYPE_2D,
+        imageFormat,
+        imageWidth, imageHeight, 0, 0, 0
+    };
+
+    ${x}_image_bindless_exp_desc_t bindlessImageDesc = {ZE_STRUCTURE_TYPE_BINDLESS_IMAGE_EXP_DESC};
+    bindlessImageDesc.flags = ZE_IMAGE_BINDLESS_EXP_FLAG_BINDLESS | ZE_IMAGE_BINDLESS_EXP_FLAG_SAMPLED_IMAGE;
+    imageDesc.pNext = &bindlessImageDesc;
+
+    bindlessImageDesc.pNext = &samplerDesc;
+
+    // Create bindless sampled image
+    // pass ZE_IMAGE_BINDLESS_EXP_FLAG_BINDLESS and ZE_IMAGE_BINDLESS_EXP_FLAG_SAMPLED_IMAGE to zeImageCreate(),
+    ${x}_image_handle_t hImage;
+    ${x}ImageCreate(hContext, hDevice, &imageDesc, &hImage);
+
+    // Create an image view from bindless sampled image
+    // define sampler descriptor for view
+    ${x}_sampler_desc_t samplerDescForView = {
+        ZE_STRUCTURE_TYPE_SAMPLER_DESC,
+        nullptr,
+        ZE_SAMPLER_ADDRESS_MODE_CLAMP,
+        ZE_SAMPLER_FILTER_MODE_NEAREST,
+        true
+    };
+
+    ${x}_image_format_t imageViewFormat = {
+        ZE_IMAGE_FORMAT_LAYOUT_32, ZE_IMAGE_FORMAT_TYPE_UINT,
+        ZE_IMAGE_FORMAT_SWIZZLE_R, ZE_IMAGE_FORMAT_SWIZZLE_X,
+        ZE_IMAGE_FORMAT_SWIZZLE_R, ZE_IMAGE_FORMAT_SWIZZLE_X
+    }
+
+    // image descriptor for bindless image view
+    ${x}_image_desc_t imageViewDesc = {
+        ZE_STRUCTURE_TYPE_IMAGE_DESC,
+        nullptr,
+        0,
+        ZE_IMAGE_TYPE_2D,
+        imageViewFormat,
+        128, 128, 0, 0, 0
+    };
+    imageViewDesc.pNext = &bindlessImageDesc;
+    bindlessImageDesc.pNext = &samplerDescForView;
+    ${x}_image_handle_t hImageView;
+
+    ${x}ImageViewCreateExt(hContext, hDevice, &imageViewDesc, hImage, &hImageView);
+
+    // If ZE_IMAGE_BINDLESS_EXP_FLAG_SAMPLED_IMAGE is not set, unsampled image is created
+    ${x}_image_handle_t hUnsampledImageView;
+    bindlessImageDesc.flags = ZE_IMAGE_BINDLESS_EXP_FLAG_BINDLESS;
+    bindlessImageDesc.pNext = nullptr;
+    ${x}ImageViewCreateExt(hContext, hDevice, &imageViewDesc, hImage, &hUnsampledImageView);
+
+    // Create an image view from bindless unsampled image
+    ${x}_image_handle_t hUnsampledImage;
+    ${x}_image_handle_t hSampledImageView;
+    bindlessImageDesc.flags = ZE_IMAGE_BINDLESS_EXP_FLAG_BINDLESS;
+    bindlessImageDesc.pNext = nullptr;
+    imageDesc.pNext = &bindlessImageDesc;
+
+    // create unsampled image
+    ${x}ImageCreate(hContext, hDevice, &imageDesc, &hUnsampledImage);
+
+    bindlessImageDesc.flags = ZE_IMAGE_BINDLESS_EXP_FLAG_BINDLESS | ZE_IMAGE_BINDLESS_EXP_FLAG_SAMPLED_IMAGE;
+    bindlessImageDesc.pNext = &samplerDescForView;
+    ${x}ImageViewCreateExt(hContext, hDevice, &imageDesc, hUnsampledImage, &hSampledImageView);

scripts/core/EXT_Exp_ImageView.rst

@@ -6,6 +6,11 @@ from templates import helper as th
     x=tags['$x']
     X=x.upper()
 %>
+
+<%!
+    from parse_specs import _version_compare_gequal
+%>
+
 :orphan:
 
 .. _ZE_experimental_image_view:
@@ -14,7 +19,7 @@ from templates import helper as th
  Image View Extension
 =========================
 
-%if ver >= 1.5:
+%if _version_compare_gequal(ver, "1.5"):
 This experimental extension is deprecated and replaced by the :ref:`${th.subt(namespace, tags, X)}_extension_image_view <${th.subt(namespace, tags, X)}_extension_image_view>` standard extension.
 %endif
 

scripts/core/EXT_Exp_ImageViewPlanar.rst

@@ -6,6 +6,11 @@ from templates import helper as th
     x=tags['$x']
     X=x.upper()
 %>
+
+<%!
+    from parse_specs import _version_compare_gequal
+%>
+
 :orphan:
 
 .. _ZE_experimental_image_view_planar:
@@ -14,7 +19,7 @@ from templates import helper as th
  Image View Planar Extension
 =============================
 
-%if ver >= 1.5:
+%if _version_compare_gequal(ver, "1.5"):
 This experimental extension is deprecated and replaced by the :ref:`${th.subt(namespace, tags, X)}_extension_image_view_planar <${th.subt(namespace, tags, X)}_extension_image_view_planar>` standard extension.
 %endif
 

scripts/core/PROG.rst

@@ -4,6 +4,11 @@
     x=tags['$x']
     X=x.upper()
 %>
+
+<%!
+    from parse_specs import _version_compare_less, _version_compare_gequal
+%>
+
 .. _core-programming-guide:
 
 ========================
@@ -47,7 +52,7 @@ The following diagram illustrates the relationship between the driver, device an
 
 .. image:: ../images/core_device.png
 
-%if ver >= 1.7:
+%if _version_compare_gequal(ver, "1.7"):
 Level Zero device model hierarchy is composed of **Root Devices** and **Sub-Devices**: A root-device may contain two or more sub-devices and a sub-device shall belong to a single root-device.
 A root-device may not contain a single sub-device, as that would be the same root-device. A root device may also be a device with no sub-devices.
 
@@ -620,10 +625,10 @@ External memory handles may be imported from other APIs, or exported for use in
 Importing and exporting external memory is an optional feature.
 Devices may describe the types of external memory handles they support using ${x}DeviceGetExternalMemoryProperties.
 
-%if ver >= 1.5:
+%if _version_compare_gequal(ver, "1.5"):
 Importing and exporting external memory is supported for device and host memory allocations and images.
 %endif
-%if ver < 1.5:
+%if _version_compare_less(ver, "1.5"):
 Importing and exporting external memory is supported for device memory allocations and images.
 %endif
 
@@ -1104,10 +1109,10 @@ A kernel timestamp event is a special type of event that records device timestam
 .. parsed-literal::
 
        // Get timestamp frequency
-%if ver >= 1.1:
+%if _version_compare_gequal(ver, "1.1"):
        const double timestampFreq = NS_IN_SEC / device_properties.timerResolution;
 %endif
-%if ver < 1.1:
+%if _version_compare_less(ver, "1.1"):
        const uint64_t timestampFreq = device_properties.timerResolution;
 %endif
        const uint64_t timestampMaxValue = ~(-1L << device_properties.kernelTimestampValidBits);
@@ -1712,7 +1717,7 @@ Environment Variables
 
 The following table documents the supported knobs for overriding default functional behavior.
 
-%if ver < 1.7:
+%if _version_compare_less(ver, "1.7"):
 
 +-----------------+-------------------------------------+------------+-----------------------------------------------------------------------------------+
 | Category        | Name                                | Values     | Description                                                                       |
@@ -1726,7 +1731,7 @@ The following table documents the supported knobs for overriding default functio
 
 %endif
 
-%if ver >= 1.7:
+%if _version_compare_gequal(ver, "1.7"):
 
 +-----------------+-------------------------------------+-----------------------------------+-----------------------------------------------------------------------------------+
 | Category        | Name                                | Values                            | Description                                                                       |
@@ -1766,7 +1771,7 @@ The values are specific to system configuration; e.g., the number of devices and
 The values are specific to the order in which devices are reported by the driver; i.e., the first device maps to ordinal 0, the second device to ordinal 1, and so forth.
 If the affinity mask is not set, then all devices and sub-devices are reported; as is the default behavior.
 
-%if ver >= 1.7:
+%if _version_compare_gequal(ver, "1.7"):
 The affinity mask masks the devices as defined by value set in the ${X}_FLAT_DEVICE_HIERARCHY environment variable, i.e., a Level Zero driver shall read
 first ${X}_FLAT_DEVICE_HIERARCHY to determine the device handles to be used by the application and then interpret the values passed in ${X}_AFFINITY_MASK
 based on the device model selected.
@@ -1776,7 +1781,7 @@ The order of the devices reported by the ${x}DeviceGet is implementation-specifi
 
 The order of the devices reported by the ${x}DeviceGet can be forced to be consistent by setting the ${X}_ENABLE_PCI_ID_DEVICE_ORDER environment variable.
 
-%if ver < 1.7:
+%if _version_compare_less(ver, "1.7"):
 The following examples demonstrate proper usage for a system configuration of two devices, each with four sub-devices:
 
 - `0, 1`: all devices and sub-devices are reported (same as default)
@@ -1788,7 +1793,7 @@ The following examples demonstrate proper usage for a system configuration of tw
 
 %endif
 
-%if ver >= 1.7:
+%if _version_compare_gequal(ver, "1.7"):
 The following examples demonstrate proper usage for a system configuration composed of two physical devices, each of which can be further
 sub-divided into four smaller devices. For the purpose of these examples, we will refer to the two physical devices as `parent devices`
 and to the smaller sub-devices as `tiles`.
@@ -2125,10 +2130,10 @@ such as multiple levels of indirection, there are two methods available:
 
     + If the driver is unable to make all allocations resident, then the call to ${x}CommandQueueExecuteCommandLists will return ${X}_RESULT_ERROR_OUT_OF_DEVICE_MEMORY
 
-%if ver >= 1.6:
+%if _version_compare_gequal(ver, "1.6"):
 2. Explicit ${x}ContextMakeMemoryResident APIs are included for the application to dynamically change residency as needed.
 %endif
-%if ver < 1.6:
+%if _version_compare_less(ver, "1.6"):
 2. Explicit ${x}ContextMakeMemoryResident APIs are included for the application to dynamically change residency as needed. (Windows-only)
 %endif
 
@@ -2283,18 +2288,18 @@ The following code examples demonstrate how to use the memory IPC APIs:
        ${x}MemCloseIpcHandle(hContext, dptr);
        
 
-%if ver >= 1.6:
+%if _version_compare_gequal(ver, "1.6"):
 5. Finally, return the IPC handle to the driver with  ${x}MemPutIpcHandle and
    free the device pointer in the sending process. If ${x}MemPutIpcHandle is not called,
    any actions performed by that call are eventually done by ${x}MemFree.
 %endif
-%if ver < 1.6:
+%if _version_compare_less(ver, "1.6"):
 5. Finally, free the device pointer in the sending process:
 %endif
 
 .. parsed-literal::
 
-%if ver >= 1.6:
+%if _version_compare_gequal(ver, "1.6"):
        ${x}MemPutIpcHandle(hContext, hIpc);
 %endif
        ${x}MemFree(hContext, dptr);
@@ -2384,19 +2389,19 @@ Note, there is no guaranteed address equivalence for the values of ``hEvent`` in
        ${x}EventDestroy(hEvent);
        ${x}EventPoolCloseIpcHandle(&hEventPool);
 
-%if ver >= 1.6:
+%if _version_compare_gequal(ver, "1.6"):
 5. Finally, return the IPC handle to the driver with ${x}EventPoolPutIpcHandle and
    free the event pool in the sending process. If ${x}EventPoolPutIpcHandle is not called,
    any actions performed by that call are eventually done by ${x}EventPoolDestroy.
 %endif
-%if ver < 1.6:
+%if _version_compare_less(ver, "1.6"):
 5. Finally, free the event pool handle in the sending process:
 %endif
 
 .. parsed-literal::
 
        ${x}EventDestroy(hEvent);
-%if ver >= 1.6:
+%if _version_compare_gequal(ver, "1.6"):
        ${x}EventPoolPutIpcHandle(hContext, hIpcEventPool);
 %endif
        ${x}EventPoolDestroy(hEventPool);

scripts/core/SPIRV.rst

@@ -7,6 +7,10 @@ from templates import helper as th
     x=tags['$x']
     X=x.upper()
 %>
+<%!
+    from parse_specs import _version_compare_gequal
+%>
+
 ==========================
  SPIR-V Programming Guide
 ==========================
@@ -412,7 +416,7 @@ The following restrictions apply to the
    words, the write must begin at a 32-bit boundary. There is no
    restriction on the y-component of the coordinate.
 
-%if ver >= 1.1:
+%if _version_compare_gequal(ver, "1.1"):
 Floating-Point Atomics
 ----------------------
 
@@ -459,7 +463,7 @@ Additionally:
 
 %endif
 
-%if ver >= 1.2:
+%if _version_compare_gequal(ver, "1.2"):
 Extended Subgroups
 ------------------
 
@@ -651,7 +655,7 @@ optional *ClusterSize* operand.
 
 %endif
 
-%if ver >= 1.2:
+%if _version_compare_gequal(ver, "1.2"):
 Linkonce ODR
 ------------
 
@@ -664,7 +668,7 @@ include the **LinkOnceODR** linkage type.
 
 %endif
 
-%if ver >= 1.5:
+%if _version_compare_gequal(ver, "1.5"):
 Bfloat16 Conversions
 --------------------