ggml-cpu: extend support for RVV floating-point kernels (#17318)

* cmake: add BF16 RVV flag for ggml-cpu

* ggml-cpu: add floating-point conversion kernels

* ggml: add floating-point kernels

Co-authored-by: Rehan Qasim <rehan.qasim@10xengineers.ai>

* ggml-cpu: fix lmul in vec_dot_bf16

* ggml-cpu: change redsum to lmul 4, fix leftover

---------

Co-authored-by: Rehan Qasim <rehan.qasim@10xengineers.ai>
This commit is contained in:
Taimur Ahmad
2025-12-18 16:02:09 +02:00
committed by GitHub
co-authored by Rehan Qasim
parent 4d1316c440
commit f716588e63
4 changed files with 221 additions and 29 deletions
+4
View File
@@ -458,6 +458,7 @@ function(ggml_add_cpu_backend_variant_impl tag_name)
if (GGML_RV_ZFH)
string(APPEND MARCH_STR "_zfh")
endif()
if (GGML_XTHEADVECTOR)
string(APPEND MARCH_STR "_xtheadvector")
elseif (GGML_RVV)
@@ -465,6 +466,9 @@ function(ggml_add_cpu_backend_variant_impl tag_name)
if (GGML_RV_ZVFH)
string(APPEND MARCH_STR "_zvfh")
endif()
if (GGML_RV_ZVFBFWMA)
string(APPEND MARCH_STR "_zvfbfwma")
endif()
endif()
if (GGML_RV_ZICBOP)
string(APPEND MARCH_STR "_zicbop")