added memory barrier intrinsics to spirv-std (#769)

* added memory barrier intrinsics to spirv-std

* added function-level docs

* added unit-tests

constants could not be validated in tests because of the limited output

* clippy fixes

* improved documentation

* Added missing features to device_memory and all_memory barrier intrinsics.

crates/spirv-std/src/arch/barrier.rs

@@ -78,3 +78,114 @@ pub unsafe fn memory_barrier<
         semantics = const SEMANTICS,
     }
 }
+
+/// Blocks execution of all threads in a group until all group shared accesses have been completed.
+///
+/// This is an exact implementation of `GroupMemoryBarrier()`.
+///
+/// From <https://docs.microsoft.com/en-us/windows/win32/direct3dhlsl/groupmemorybarrier>
+#[spirv_std_macros::gpu_only]
+#[inline]
+pub unsafe fn workgroup_memory_barrier() {
+    memory_barrier::<
+        { crate::memory::Scope::Workgroup as u32 },
+        {
+            crate::memory::Semantics::WORKGROUP_MEMORY.bits()
+                | crate::memory::Semantics::ACQUIRE_RELEASE.bits()
+        },
+    >();
+}
+
+/// Blocks execution of all threads in a group until all group shared accesses have been completed and all threads in the group have reached this call.
+///
+/// This is an exact implementation of `GroupMemoryBarrierWithGroupSync()`.
+///
+/// From <https://docs.microsoft.com/en-us/windows/win32/direct3dhlsl/groupmemorybarrierwithgroupsync>
+#[spirv_std_macros::gpu_only]
+#[inline]
+pub unsafe fn workgroup_memory_barrier_with_group_sync() {
+    control_barrier::<
+        { crate::memory::Scope::Workgroup as u32 },
+        { crate::memory::Scope::Workgroup as u32 },
+        {
+            crate::memory::Semantics::WORKGROUP_MEMORY.bits()
+                | crate::memory::Semantics::ACQUIRE_RELEASE.bits()
+        },
+    >();
+}
+
+/// Blocks execution of all threads in a group until all device memory accesses have been completed.
+///
+/// This is an exact implementation of `DeviceMemoryBarrier()`.
+///
+/// From <https://docs.microsoft.com/en-us/windows/win32/direct3dhlsl/devicememorybarrier>
+#[spirv_std_macros::gpu_only]
+#[inline]
+pub unsafe fn device_memory_barrier() {
+    memory_barrier::<
+        { crate::memory::Scope::Device as u32 },
+        {
+            crate::memory::Semantics::IMAGE_MEMORY.bits()
+                | crate::memory::Semantics::UNIFORM_MEMORY.bits()
+                | crate::memory::Semantics::ACQUIRE_RELEASE.bits()
+        },
+    >();
+}
+
+/// Blocks execution of all threads in a group until all device memory accesses have been completed and all threads in the group have reached this call.
+///
+/// This is an exact implementation of `DeviceMemoryBarrierWithGroupSync()`.
+///
+/// From <https://docs.microsoft.com/en-us/windows/win32/direct3dhlsl/devicememorybarrierwithgroupsync>
+#[spirv_std_macros::gpu_only]
+#[inline]
+pub unsafe fn device_memory_barrier_with_group_sync() {
+    control_barrier::<
+        { crate::memory::Scope::Workgroup as u32 },
+        { crate::memory::Scope::Device as u32 },
+        {
+            crate::memory::Semantics::IMAGE_MEMORY.bits()
+                | crate::memory::Semantics::UNIFORM_MEMORY.bits()
+                | crate::memory::Semantics::ACQUIRE_RELEASE.bits()
+        },
+    >();
+}
+
+/// Blocks execution of all threads in a group until all memory accesses have been completed.
+///
+/// This is an exact implementation of `AllMemoryBarrier()`.
+///
+/// From <https://docs.microsoft.com/en-us/windows/win32/direct3dhlsl/allmemorybarrier>
+#[spirv_std_macros::gpu_only]
+#[inline]
+pub unsafe fn all_memory_barrier() {
+    memory_barrier::<
+        { crate::memory::Scope::Device as u32 },
+        {
+            crate::memory::Semantics::WORKGROUP_MEMORY.bits()
+                | crate::memory::Semantics::IMAGE_MEMORY.bits()
+                | crate::memory::Semantics::UNIFORM_MEMORY.bits()
+                | crate::memory::Semantics::ACQUIRE_RELEASE.bits()
+        },
+    >();
+}
+
+/// Blocks execution of all threads in a group until all memory accesses have been completed and all threads in the group have reached this call.
+///
+/// This is an exact implementation of `AllMemoryBarrierWithGroupSync()`.
+///
+/// From <https://docs.microsoft.com/en-us/windows/win32/direct3dhlsl/allmemorybarrierwithgroupsync>
+#[spirv_std_macros::gpu_only]
+#[inline]
+pub unsafe fn all_memory_barrier_with_group_sync() {
+    control_barrier::<
+        { crate::memory::Scope::Workgroup as u32 },
+        { crate::memory::Scope::Device as u32 },
+        {
+            crate::memory::Semantics::WORKGROUP_MEMORY.bits()
+                | crate::memory::Semantics::IMAGE_MEMORY.bits()
+                | crate::memory::Semantics::UNIFORM_MEMORY.bits()
+                | crate::memory::Semantics::ACQUIRE_RELEASE.bits()
+        },
+    >();
+}

tests/ui/arch/all_memory_barrier.rs

@@ -0,0 +1,16 @@
+// build-pass
+// compile-flags: -C target-feature=+VulkanMemoryModelDeviceScopeKHR,+ext:SPV_KHR_vulkan_memory_model
+// compile-flags: -C llvm-args=--disassemble-fn=all_memory_barrier::all_memory_barrier
+
+use spirv_std as _;
+
+unsafe fn all_memory_barrier() {
+    spirv_std::arch::all_memory_barrier();
+}
+
+#[spirv(compute(threads(1, 1, 1)))]
+pub fn main() {
+    unsafe {
+        all_memory_barrier();
+    }
+}

tests/ui/arch/all_memory_barrier.stderr

@@ -0,0 +1,7 @@
+%1 = OpFunction  %2  None %3
+%4 = OpLabel
+OpLine %5 72 4
+OpMemoryBarrier %6 %7
+OpLine %8 9 1
+OpReturn
+OpFunctionEnd

tests/ui/arch/all_memory_barrier_with_group_sync.rs

@@ -0,0 +1,16 @@
+// build-pass
+// compile-flags: -C target-feature=+VulkanMemoryModelDeviceScopeKHR,+ext:SPV_KHR_vulkan_memory_model
+// compile-flags: -C llvm-args=--disassemble-fn=all_memory_barrier_with_group_sync::all_memory_barrier_with_group_sync
+
+use spirv_std as _;
+
+unsafe fn all_memory_barrier_with_group_sync() {
+    spirv_std::arch::all_memory_barrier_with_group_sync();
+}
+
+#[spirv(compute(threads(1, 1, 1)))]
+pub fn main() {
+    unsafe {
+        all_memory_barrier_with_group_sync();
+    }
+}

tests/ui/arch/all_memory_barrier_with_group_sync.stderr

@@ -0,0 +1,7 @@
+%1 = OpFunction  %2  None %3
+%4 = OpLabel
+OpLine %5 38 4
+OpControlBarrier %6 %7 %8
+OpLine %9 9 1
+OpReturn
+OpFunctionEnd

tests/ui/arch/device_memory_barrier.rs

@@ -0,0 +1,16 @@
+// build-pass
+// compile-flags: -C target-feature=+VulkanMemoryModelDeviceScopeKHR,+ext:SPV_KHR_vulkan_memory_model
+// compile-flags: -C llvm-args=--disassemble-fn=device_memory_barrier::device_memory_barrier
+
+use spirv_std as _;
+
+unsafe fn device_memory_barrier() {
+    spirv_std::arch::device_memory_barrier();
+}
+
+#[spirv(compute(threads(1, 1, 1)))]
+pub fn main() {
+    unsafe {
+        device_memory_barrier();
+    }
+}

tests/ui/arch/device_memory_barrier.stderr

@@ -0,0 +1,7 @@
+%1 = OpFunction  %2  None %3
+%4 = OpLabel
+OpLine %5 72 4
+OpMemoryBarrier %6 %7
+OpLine %8 9 1
+OpReturn
+OpFunctionEnd

tests/ui/arch/device_memory_barrier_with_group_sync.rs

@@ -0,0 +1,16 @@
+// build-pass
+// compile-flags: -C target-feature=+VulkanMemoryModelDeviceScopeKHR,+ext:SPV_KHR_vulkan_memory_model
+// compile-flags: -C llvm-args=--disassemble-fn=device_memory_barrier_with_group_sync::device_memory_barrier_with_group_sync
+
+use spirv_std as _;
+
+unsafe fn device_memory_barrier_with_group_sync() {
+    spirv_std::arch::device_memory_barrier_with_group_sync();
+}
+
+#[spirv(compute(threads(1, 1, 1)))]
+pub fn main() {
+    unsafe {
+        device_memory_barrier_with_group_sync();
+    }
+}

tests/ui/arch/device_memory_barrier_with_group_sync.stderr

@@ -0,0 +1,7 @@
+%1 = OpFunction  %2  None %3
+%4 = OpLabel
+OpLine %5 38 4
+OpControlBarrier %6 %7 %8
+OpLine %9 9 1
+OpReturn
+OpFunctionEnd

tests/ui/arch/workgroup_memory_barrier.rs

@@ -0,0 +1,15 @@
+// build-pass
+// compile-flags: -C llvm-args=--disassemble-fn=workgroup_memory_barrier::workgroup_memory_barrier
+
+use spirv_std as _;
+
+unsafe fn workgroup_memory_barrier() {
+    spirv_std::arch::workgroup_memory_barrier();
+}
+
+#[spirv(compute(threads(1, 1, 1)))]
+pub fn main() {
+    unsafe {
+        workgroup_memory_barrier();
+    }
+}

tests/ui/arch/workgroup_memory_barrier.stderr

@@ -0,0 +1,7 @@
+%1 = OpFunction  %2  None %3
+%4 = OpLabel
+OpLine %5 72 4
+OpMemoryBarrier %6 %7
+OpLine %8 8 1
+OpReturn
+OpFunctionEnd

tests/ui/arch/workgroup_memory_barrier_with_group_sync.rs

@@ -0,0 +1,15 @@
+// build-pass
+// compile-flags: -C llvm-args=--disassemble-fn=workgroup_memory_barrier_with_group_sync::workgroup_memory_barrier_with_group_sync
+
+use spirv_std as _;
+
+unsafe fn workgroup_memory_barrier_with_group_sync() {
+    spirv_std::arch::workgroup_memory_barrier_with_group_sync();
+}
+
+#[spirv(compute(threads(1, 1, 1)))]
+pub fn main() {
+    unsafe {
+        workgroup_memory_barrier_with_group_sync();
+    }
+}

tests/ui/arch/workgroup_memory_barrier_with_group_sync.stderr

@@ -0,0 +1,7 @@
+%1 = OpFunction  %2  None %3
+%4 = OpLabel
+OpLine %5 38 4
+OpControlBarrier %6 %6 %7
+OpLine %8 8 1
+OpReturn
+OpFunctionEnd