Skip to content

Commit 4e96a81

Browse files
[SYCL] Windows default build instructions without -DLLAMA_SYCL_F16 flag activated (ggml-org#6767)
* Fix FP32/FP16 build instructions * Fix typo * Recommended build instruction Co-authored-by: Neo Zhang Jianyu <jianyu.zhang@intel.com> * Recommended build instruction Co-authored-by: Neo Zhang Jianyu <jianyu.zhang@intel.com> * Recommended build instruction Co-authored-by: Neo Zhang Jianyu <jianyu.zhang@intel.com> * Add comments in Intel GPU linux --------- Co-authored-by: Anas Ahouzi <112881240+aahouzi-intel@users.noreply.github.com> Co-authored-by: Neo Zhang Jianyu <jianyu.zhang@intel.com>
1 parent 192090b commit 4e96a81

File tree

1 file changed

+12
-8
lines changed

1 file changed

+12
-8
lines changed

README-sycl.md

Lines changed: 12 additions & 8 deletions
Original file line numberDiff line numberDiff line change
@@ -229,12 +229,12 @@ source /opt/intel/oneapi/setvars.sh
229229
# Build LLAMA with MKL BLAS acceleration for intel GPU
230230
mkdir -p build && cd build
231231

232-
# Option 1: Use FP16 for better performance in long-prompt inference
233-
#cmake .. -DLLAMA_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx -DLLAMA_SYCL_F16=ON
234-
235-
# Option 2: Use FP32 by default
232+
# Option 1: Use FP32 (recommended for better performance in most cases)
236233
cmake .. -DLLAMA_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx
237234

235+
# Option 2: Use FP16
236+
cmake .. -DLLAMA_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx -DLLAMA_SYCL_F16=ON
237+
238238
#build all binary
239239
cmake --build . --config Release -j -v
240240
```
@@ -250,12 +250,12 @@ export CPLUS_INCLUDE_DIR=/path/to/oneMKL/include:$CPLUS_INCLUDE_DIR
250250
# Build LLAMA with Nvidia BLAS acceleration through SYCL
251251
mkdir -p build && cd build
252252

253-
# Option 1: Use FP16 for better performance in long-prompt inference
254-
cmake .. -DLLAMA_SYCL=ON -DLLAMA_SYCL_TARGET=NVIDIA -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx -DLLAMA_SYCL_F16=ON
255-
256-
# Option 2: Use FP32 by default
253+
# Option 1: Use FP32 (recommended for better performance in most cases)
257254
cmake .. -DLLAMA_SYCL=ON -DLLAMA_SYCL_TARGET=NVIDIA -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx
258255

256+
# Option 2: Use FP16
257+
cmake .. -DLLAMA_SYCL=ON -DLLAMA_SYCL_TARGET=NVIDIA -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icpx -DLLAMA_SYCL_F16=ON
258+
259259
#build all binary
260260
cmake --build . --config Release -j -v
261261

@@ -416,6 +416,10 @@ mkdir -p build
416416
cd build
417417
@call "C:\Program Files (x86)\Intel\oneAPI\setvars.bat" intel64 --force
418418
419+
# Option 1: Use FP32 (recommended for better performance in most cases)
420+
cmake -G "MinGW Makefiles" .. -DLLAMA_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icx -DCMAKE_BUILD_TYPE=Release
421+
422+
# Option 2: Or FP16
419423
cmake -G "MinGW Makefiles" .. -DLLAMA_SYCL=ON -DCMAKE_C_COMPILER=icx -DCMAKE_CXX_COMPILER=icx -DCMAKE_BUILD_TYPE=Release -DLLAMA_SYCL_F16=ON
420424
421425
make -j

0 commit comments

Comments
 (0)