[openmp][amdgpu] Add DynamicLdsSize to AMDGPUImplicitArgsTy

[skc7]

#65273 "hidden_dynamic_lds_size" argument will be added in the reserved section at offset 120 of the implicit argument layout
Add DynamicLdsSize to AMDGPUImplicitArgsTy struct at offset 120 and fill the dynamic LDS size before kernel launch.

[saiislam]

Let's hold till @arsenm approves #65273 .
Otherwise, the patch looks fine to me.

[llvmbot]

@llvm/pr-subscribers-backend-amdgpu

Changes

Add DynamicLdsSize to AMDGPUImplicitArgsTy struct and fill the dynamic LDS size before kernel launch.

---

Full diff: https://github.com/llvm/llvm-project/pull/65325.diff

2 Files Affected:

• (modified) openmp/libomptarget/plugins-nextgen/amdgpu/src/rtl.cpp (+1)
• (modified) openmp/libomptarget/plugins-nextgen/amdgpu/utils/UtilitiesRTL.h (+2-1)

diff --git a/openmp/libomptarget/plugins-nextgen/amdgpu/src/rtl.cpp b/openmp/libomptarget/plugins-nextgen/amdgpu/src/rtl.cpp
index 66a25e29d016276..da2b3021728f9bb 100644
--- a/openmp/libomptarget/plugins-nextgen/amdgpu/src/rtl.cpp
+++ b/openmp/libomptarget/plugins-nextgen/amdgpu/src/rtl.cpp
@@ -3031,6 +3031,7 @@ Error AMDGPUKernelTy::launchImpl(GenericDeviceTy &GenericDevice,
     ImplArgs->GroupSizeY = 1;
     ImplArgs->GroupSizeZ = 1;
     ImplArgs->GridDims = 1;
+    ImplArgs->DynamicLdsSize = KernelArgs.DynCGroupMem;
   }
 
   // Push the kernel launch into the stream.
diff --git a/openmp/libomptarget/plugins-nextgen/amdgpu/utils/UtilitiesRTL.h b/openmp/libomptarget/plugins-nextgen/amdgpu/utils/UtilitiesRTL.h
index b39545ab7d02ba2..d9828f0aa37a912 100644
--- a/openmp/libomptarget/plugins-nextgen/amdgpu/utils/UtilitiesRTL.h
+++ b/openmp/libomptarget/plugins-nextgen/amdgpu/utils/UtilitiesRTL.h
@@ -43,7 +43,8 @@ struct AMDGPUImplicitArgsTy {
   uint16_t GroupSizeZ;
   uint8_t Unused0[46]; // 46 byte offset.
   uint16_t GridDims;
-  uint8_t Unused1[190]; // 190 byte offset.
+  uint32_t DynamicLdsSize;
+  uint8_t Unused1[186]; // 186 byte offset.
 };
 
 // Dummy struct for COV4 implicitargs.

[saiislam]

LGTM!