Merge pull request #990 from InfiniTensor/demo131

Demo-131 Cuda graph with optimized paged attention

Merge pull request #990 from InfiniTensor/demo131
Demo-131 Cuda graph with optimized paged attention
784139b9 · thatPepe · GitHub · 3c8fb3c0 · 1d6527cb · 784139b9
Unverified Commit 784139b9 authored Feb 13, 2026 by thatPepe Committed by GitHub Feb 13, 2026
20 changed files
--- a/src/infiniop/devices/moore/moore_kernel_common.h
+++ b/src/infiniop/devices/moore/moore_kernel_common.h
--- a/src/infiniop/devices/nvidia/nvidia_common.cu
+++ b/src/infiniop/devices/nvidia/nvidia_common.cu
--- a/src/infiniop/devices/nvidia/nvidia_handle.h
+++ b/src/infiniop/devices/nvidia/nvidia_handle.h
--- a/src/infiniop/devices/nvidia/nvidia_kernel_common.cuh
+++ b/src/infiniop/devices/nvidia/nvidia_kernel_common.cuh
--- a/src/infiniop/elementwise/bang/elementwise_bang.h
+++ b/src/infiniop/elementwise/bang/elementwise_bang.h
--- a/src/infiniop/elementwise/bang/elementwise_bang_kernel.h
+++ b/src/infiniop/elementwise/bang/elementwise_bang_kernel.h
--- a/src/infiniop/ninetoothed/build.py
+++ b/src/infiniop/ninetoothed/build.py
--- a/src/infiniop/ninetoothed/utils.h
+++ b/src/infiniop/ninetoothed/utils.h
--- a/src/infiniop/ops/add/bang/add_bang.mlu
+++ b/src/infiniop/ops/add/bang/add_bang.mlu
--- a/src/infiniop/ops/add/bang/add_bang_internal.mlu
+++ b/src/infiniop/ops/add/bang/add_bang_internal.mlu
--- a/src/infiniop/ops/add/operator.cc
+++ b/src/infiniop/ops/add/operator.cc
--- a/src/infiniop/ops/add_rms_norm/add_rms_norm.h
+++ b/src/infiniop/ops/add_rms_norm/add_rms_norm.h
--- a/src/infiniop/ops/add_rms_norm/cpu/add_rms_norm_cpu.cc
+++ b/src/infiniop/ops/add_rms_norm/cpu/add_rms_norm_cpu.cc
--- a/src/infiniop/ops/add_rms_norm/info.h
+++ b/src/infiniop/ops/add_rms_norm/info.h
--- a/src/infiniop/ops/add_rms_norm/metax/add_rms_norm_metax.cuh
+++ b/src/infiniop/ops/add_rms_norm/metax/add_rms_norm_metax.cuh
--- a/src/infiniop/ops/add_rms_norm/metax/add_rms_norm_metax.maca
+++ b/src/infiniop/ops/add_rms_norm/metax/add_rms_norm_metax.maca
--- a/src/infiniop/ops/add_rms_norm/moore/add_rms_norm_moore.h
+++ b/src/infiniop/ops/add_rms_norm/moore/add_rms_norm_moore.h
--- a/src/infiniop/ops/add_rms_norm/moore/add_rms_norm_moore.mu
+++ b/src/infiniop/ops/add_rms_norm/moore/add_rms_norm_moore.mu
--- a/src/infiniop/ops/add_rms_norm/nvidia/add_rms_norm_nvidia.cu
+++ b/src/infiniop/ops/add_rms_norm/nvidia/add_rms_norm_nvidia.cu
--- a/src/infiniop/ops/add_rms_norm/operator.cc
+++ b/src/infiniop/ops/add_rms_norm/operator.cc