[tvm] branch main updated: [Target][Minor] Add A6000 Target Tag (#14762)

[masahi]

This is an automated email from the ASF dual-hosted git repository.

masahi pushed a commit to branch main
in repository https://gitbox.apache.org/repos/asf/tvm.git


The following commit(s) were added to refs/heads/main by this push:
     new 706a5b2dd5 [Target][Minor] Add A6000 Target Tag (#14762)
706a5b2dd5 is described below

commit 706a5b2dd5ab8f22a88b84eeb4fd94f318f1e9a5
Author: Xiyou Zhou <xi...@octoml.ai>
AuthorDate: Thu May 4 19:44:34 2023 -0700

    [Target][Minor] Add A6000 Target Tag (#14762)
    
    Add a6000 tag and sources.
---
 src/target/tag.cc | 7 +++++++
 1 file changed, 7 insertions(+)

diff --git a/src/target/tag.cc b/src/target/tag.cc
index bd2b5958c5..037d2e5937 100644
--- a/src/target/tag.cc
+++ b/src/target/tag.cc
@@ -105,6 +105,12 @@ TVM_REGISTER_TARGET_TAG("nvidia/jetson-agx-xavier")
       {"registers_per_block", Integer(RegPerBlock)},              \
   });
 
+// Naming convention for CUDA tags see https://developer.nvidia.com/cuda-gpus
+// Parameters see Table 15. Technical Specifications per Compute Capability
+// https://docs.nvidia.com/cuda/cuda-c-programming-guide/index.html
+// Check `Maximum y- or z-dimension of a grid of thread blocks` for max 
threads per block
+// Check `Maximum amount of shared memory per thread block` for max shared 
memory per block
+// Note that above 48 KB requires dynamic shared memory
 TVM_REGISTER_CUDA_TAG("nvidia/tesla-k80", "sm_37", 49152, 65536);
 TVM_REGISTER_CUDA_TAG("nvidia/tesla-k40", "sm_35", 49152, 65536);
 TVM_REGISTER_CUDA_TAG("nvidia/tesla-k20", "sm_35", 49152, 65536);
@@ -129,6 +135,7 @@ TVM_REGISTER_CUDA_TAG("nvidia/tesla-k80", "sm_37", 49152, 
65536);
 TVM_REGISTER_CUDA_TAG("nvidia/tesla-k40", "sm_35", 49152, 65536);
 TVM_REGISTER_CUDA_TAG("nvidia/tesla-k20", "sm_35", 49152, 65536);
 TVM_REGISTER_CUDA_TAG("nvidia/tesla-k10", "sm_30", 49152, 65536);
+TVM_REGISTER_CUDA_TAG("nvidia/rtx-a6000", "sm_86", 49152, 65536);
 TVM_REGISTER_CUDA_TAG("nvidia/quadro-rtx-8000", "sm_75", 49152, 65536);
 TVM_REGISTER_CUDA_TAG("nvidia/quadro-rtx-6000", "sm_75", 49152, 65536);
 TVM_REGISTER_CUDA_TAG("nvidia/quadro-rtx-5000", "sm_75", 49152, 65536);