[NVPTX] Add Intrinsics for discard.* (#128404)

[NVPTX] Add  Intrinsics for discard.*
This PR adds intrinsics for all variations of discard.*

* These intrinsics supports generic or global for all variations.
* The lowering is handled from nvvm to nvptx tablegen directly.
* Lit tests are added as part of discard.ll
* The generated PTX is verified with a 12.3 ptxas executable.
* Added docs for these intrinsics in NVPTXUsage.rst.

For more information, refer to the PTX ISA

<https://docs.nvidia.com/cuda/parallel-thread-execution/#data-movement-and-conversion-instructions-discard>_.

---------

Co-authored-by: abmajumder <abmajumder@nvidia.com>
Co-authored-by: gonzalobg <65027571+gonzalobg@users.noreply.github.com>

Author: abhilash1910

llvm/docs/NVPTXUsage.rst

@@ -671,6 +671,45 @@ level on which the priority is to be applied. The only supported value for the s
 For more information, refer to the PTX ISA
 `<https://docs.nvidia.com/cuda/parallel-thread-execution/#data-movement-and-conversion-instructions-applypriority>`_.
 
+``llvm.nvvm.discard.*``'
+^^^^^^^^^^^^^^^^^^^^^^^^
+
+Syntax:
+"""""""
+
+.. code-block:: llvm
+
+  declare void  @llvm.nvvm.discard.global.L2(ptr addrspace(1) %global_ptr, i64 immarg)
+  declare void  @llvm.nvvm.discard.L2(ptr %ptr, i64 immarg)
+
+Overview:
+"""""""""
+
+The *effects* of the ``@llvm.nvvm.discard.L2*`` intrinsics are those of a non-atomic 
+non-volatile ``llvm.memset`` that writes ``undef`` to the destination 
+address range ``[%ptr, %ptr + immarg)``. The ``%ptr`` must be aligned by 128 bytes.
+Subsequent reads from the address range may read ``undef`` until the memory is overwritten 
+with a different value.
+These operations *hint* the implementation that data in the L2 cache can be destructively 
+discarded without writing it back to memory. 
+The operand ``immarg`` is an integer constant that specifies the length in bytes of the 
+address range ``[%ptr, %ptr + immarg)`` to write ``undef`` into. 
+The only supported value for the ``immarg`` operand is ``128``. 
+If generic addressing is used and the specified address does not fall within the 
+address window of global memory (``addrspace(1)``) the behavior is undefined.
+
+.. code-block:: llvm
+ 
+   call void @llvm.nvvm.discard.L2(ptr %p, i64 128)  ;; writes `undef` to [p, p+128)
+   %a = load i64, ptr %p. ;; loads 8 bytes containing undef
+   %b = load i64, ptr %p  ;; loads 8 bytes containing undef
+   ;; comparing %a and %b compares `undef` values!
+   %fa = freeze i64 %a  ;; freezes undef to stable bit-pattern
+   %fb = freeze i64 %b  ;; freezes undef to stable bit-pattern
+   ;; %fa may compare different to %fb!
+   
+For more information, refer to the  `CUDA C++ discard documentation <https://nvidia.github.io/cccl/libcudacxx/extended_api/memory_access_properties/discard_memory.html>`__ and to the `PTX ISA discard documentation <https://docs.nvidia.com/cuda/parallel-thread-execution/#data-movement-and-conversion-instructions-discard>`__ .
+
 '``llvm.nvvm.cp.async.bulk.tensor.g2s.tile.[1-5]d``'
 ^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^^
 

llvm/include/llvm/IR/IntrinsicsNVVM.td

@@ -5082,6 +5082,14 @@ def int_nvvm_applypriority_L2_evict_normal
     [IntrArgMemOnly, ReadOnly<ArgIndex<0>>, NoCapture<ArgIndex<0>>,
     ImmArg<ArgIndex<1>>]>;
 
+// Intrinsics for discard
+def int_nvvm_discard_global_L2 : DefaultAttrsIntrinsic<[], 
+    [llvm_global_ptr_ty, llvm_i64_ty], [NoCapture<ArgIndex<0>>,
+    ImmArg<ArgIndex<1>>, IntrHasSideEffects]>;
+
+def int_nvvm_discard_L2 : DefaultAttrsIntrinsic<[], 
+    [llvm_ptr_ty, llvm_i64_ty], [NoCapture<ArgIndex<0>>,
+    ImmArg<ArgIndex<1>>, IntrHasSideEffects]>;
 
 // Intrinsics for Bulk Copy using TMA (non-tensor)
 // From Global to Shared Cluster

llvm/lib/Target/NVPTX/NVPTXIntrinsics.td

@@ -800,6 +800,17 @@ class APPLYPRIORITY_L2_INTRS<string addr> :
 def APPLYPRIORITY_L2_EVICT_NORMAL        : APPLYPRIORITY_L2_INTRS<"">;
 def APPLYPRIORITY_GLOBAL_L2_EVICT_NORMAL : APPLYPRIORITY_L2_INTRS<"global">;
 
+//Discard Intrinsics
+class DISCARD_L2_INTRS<string Addr> :
+          NVPTXInst<(outs), (ins Int64Regs:$addr),
+          StrJoin<".", ["discard", Addr , "L2"]>.ret # " [$addr], 128;",
+          [(!cast<Intrinsic>(StrJoin<"_", ["int_nvvm_discard", Addr , "L2"]>.ret)
+          i64:$addr, (i64 128))]>,
+          Requires<[hasPTX<74>, hasSM<80>]>;
+
+def DISCARD_L2        : DISCARD_L2_INTRS<"">;
+def DISCARD_GLOBAL_L2 : DISCARD_L2_INTRS<"global">;
+
 //-----------------------------------
 // MBarrier Functions
 //-----------------------------------

llvm/test/CodeGen/NVPTX/discard.ll

@@ -0,0 +1,35 @@
+; NOTE: Assertions have been autogenerated by utils/update_llc_test_checks.py UTC_ARGS: --version 5
+; RUN: llc < %s -mtriple=nvptx64 -mcpu=sm_80 -mattr=+ptx74| FileCheck --check-prefixes=CHECK-PTX64 %s
+; RUN: %if ptxas-11.4 %{ llc < %s -mtriple=nvptx64 -mcpu=sm_80 -mattr=+ptx74| %ptxas-verify -arch=sm_80 %}
+
+target triple = "nvptx64-nvidia-cuda"
+
+declare void  @llvm.nvvm.discard.global.L2(ptr addrspace(1) %global_ptr, i64 immarg %size)
+declare void  @llvm.nvvm.discard.L2(ptr %ptr, i64 immarg %size)
+
+define void @discard_global_L2(ptr addrspace(1) %global_ptr) {
+; CHECK-PTX64-LABEL: discard_global_L2(
+; CHECK-PTX64:       {
+; CHECK-PTX64-NEXT:    .reg .b64 %rd<2>;
+; CHECK-PTX64-EMPTY:
+; CHECK-PTX64-NEXT:  // %bb.0:
+; CHECK-PTX64-NEXT:    ld.param.u64 %rd1, [discard_global_L2_param_0];
+; CHECK-PTX64-NEXT:    discard.global.L2 [%rd1], 128;
+; CHECK-PTX64-NEXT:    ret;
+  tail call void @llvm.nvvm.discard.global.L2(ptr addrspace(1) %global_ptr, i64 128)
+  ret void
+}
+
+define void @discard_L2(ptr %ptr) {
+; CHECK-PTX64-LABEL: discard_L2(
+; CHECK-PTX64:       {
+; CHECK-PTX64-NEXT:    .reg .b64 %rd<2>;
+; CHECK-PTX64-EMPTY:
+; CHECK-PTX64-NEXT:  // %bb.0:
+; CHECK-PTX64-NEXT:    ld.param.u64 %rd1, [discard_L2_param_0];
+; CHECK-PTX64-NEXT:    discard.L2 [%rd1], 128;
+; CHECK-PTX64-NEXT:    ret;
+  tail call void @llvm.nvvm.discard.L2(ptr %ptr, i64 128)
+  ret void
+}
+