Commit dbbd41a
committed
[GPU] Fix ambiguous fmin/fmax calls for unsigned integer types in eltwise kernel
Fixes #33618
When MIN/MAX eltwise operations are used with UINT8/UINT16/UINT32 inputs,
the generated OpenCL kernel incorrectly used fmin/fmax functions which
are only defined for floating-point types, causing compilation errors:
'call to fmin/fmax is ambiguous'.
This fix adds UINT8, UINT16, UINT32, and INT16 to the integer type
checks, ensuring that OpenCL's min/max functions (which support integer
types) are used instead of fmin/fmax.
This bug affects ONNX Clip operations (opset 11+) which are lowered to
Maximum + Minimum operations in OpenVINO.1 parent 608e396 commit dbbd41a
File tree
1 file changed
+13
-9
lines changed- src/plugins/intel_gpu/src/kernel_selector/kernels/eltwise
1 file changed
+13
-9
lines changedLines changed: 13 additions & 9 deletions
| Original file line number | Diff line number | Diff line change | |
|---|---|---|---|
| |||
241 | 241 | | |
242 | 242 | | |
243 | 243 | | |
| 244 | + | |
| 245 | + | |
| 246 | + | |
| 247 | + | |
| 248 | + | |
| 249 | + | |
| 250 | + | |
| 251 | + | |
| 252 | + | |
| 253 | + | |
244 | 254 | | |
245 | | - | |
246 | | - | |
247 | | - | |
| 255 | + | |
248 | 256 | | |
249 | | - | |
250 | | - | |
251 | | - | |
| 257 | + | |
252 | 258 | | |
253 | 259 | | |
254 | 260 | | |
| |||
257 | 263 | | |
258 | 264 | | |
259 | 265 | | |
260 | | - | |
261 | | - | |
262 | | - | |
| 266 | + | |
263 | 267 | | |
264 | 268 | | |
265 | 269 | | |
| |||
0 commit comments