https://github.com/ssahasra created https://github.com/llvm/llvm-project/pull/135803
The documentation for the `convergent` attribute claims that OpenCL and CUDA require "all threads" in a group to call the same convergent operation. This is true only for OpenCL, and in general, the `convergent` attribute is used in LLVM IR on operations that have no such constraint. >From 5024a4e5a9dc92744296cf59db7363602bca1807 Mon Sep 17 00:00:00 2001 From: Sameer Sahasrabuddhe <sameer.sahasrabud...@amd.com> Date: Tue, 15 Apr 2025 21:43:35 +0530 Subject: [PATCH] [clang] convergent attribute does not require "all threads" The documentation for the `convergent` attribute claims that OpenCL and CUDA require "all threads" in a group to call the same convergent operation. This is true only for OpenCL, and in general, the `convergent` attribute is used in LLVM IR on operations that have no such constraint. --- clang/include/clang/Basic/AttrDocs.td | 4 ---- 1 file changed, 4 deletions(-) diff --git a/clang/include/clang/Basic/AttrDocs.td b/clang/include/clang/Basic/AttrDocs.td index 97a5f24d35d7d..41e1918b4336f 100644 --- a/clang/include/clang/Basic/AttrDocs.td +++ b/clang/include/clang/Basic/AttrDocs.td @@ -1680,10 +1680,6 @@ translated into the LLVM ``convergent`` attribute, which indicates that the call instructions of a function with this attribute cannot be made control-dependent on any additional values. -In languages designed for SPMD/SIMT programming model, e.g. OpenCL or CUDA, -the call instructions of a function with this attribute must be executed by -all work items or threads in a work group or sub group. - This attribute is different from ``noduplicate`` because it allows duplicating function calls if it can be proved that the duplicated function calls are not made control-dependent on any additional values, e.g., unrolling a loop _______________________________________________ cfe-commits mailing list cfe-commits@lists.llvm.org https://lists.llvm.org/cgi-bin/mailman/listinfo/cfe-commits
