]> git.djapps.eu Git - pkg/ggml/sources/whisper.cpp/commit
CUDA: fuse rms_norm + mul + rope (+ view + set_rows) (llama/26767)
authorRafail Giavrimis <redacted>
Sat, 8 Aug 2026 16:32:37 +0000 (17:32 +0100)
committerGeorgi Gerganov <redacted>
Fri, 14 Aug 2026 19:16:06 +0000 (22:16 +0300)
commit10791af67498011a99d361f02d3684e5d3b5c0c9
treea6360d3477e3eb9972535c64d42e9f54c12841d1
parent068d3b0b4530506a884fe88aa6e219e870697992
CUDA: fuse rms_norm + mul + rope (+ view + set_rows) (llama/26767)

* CUDA: fuse rms_norm + mul + rope (+ view + set_rows)

* tests: add broadcast weight case to rms_norm_mul_rope

* CUDA: check memory ranges before rms_norm rope fusion

* CUDA: check memory ranges in rope set_rows fusion
ggml/src/ggml-cuda/ggml-cuda.cu
ggml/src/ggml-cuda/rope.cu
ggml/src/ggml-cuda/rope.cuh