Skip to content

Commit 2e992dc

Browse files
authored
[SYCL][DOC] Document -foffload-fp32-prec-div/sqrt options (#17033)
Signed-off-by: Sidorov, Dmitry <dmitry.sidorov@intel.com>
1 parent a4f9764 commit 2e992dc

File tree

1 file changed

+13
-2
lines changed

1 file changed

+13
-2
lines changed

sycl/doc/UsersManual.md

Lines changed: 13 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -322,6 +322,18 @@ and not recommended to use in production environment.
322322

323323
## Other options
324324

325+
**`-f[no-]offload-fp32-prec-sqrt`**
326+
327+
Enable use of correctly rounded `sycl::sqrt` function as defined by IEE754.
328+
Without this flag, the default precision requirement for `sycl::sqrt` is 3
329+
ULP.
330+
331+
**`-f[no-]offload-fp32-prec-div`**
332+
333+
Enable use of correctly rounded divide operation as defined by IEE754.
334+
Without this flag, the default precision requirement for divide in SYCL is
335+
2.5 ULP.
336+
325337
**`-fsycl-device-only`**
326338

327339
Compile only device part of the code and ignore host part.
@@ -365,15 +377,14 @@ and not recommended to use in production environment.
365377
options (e.g. -c, -E, -S) may interfere with the expected output set during
366378
the host compilation. Doing so is considered undefined behavior.
367379

368-
**`-fsycl-fp32-prec-sqrt`**
380+
**`-fsycl-fp32-prec-sqrt`** [DEPRECATED]
369381

370382
Enable use of correctly rounded `sycl::sqrt` function as defined by IEE754.
371383
Without this flag, the default precision requirement for `sycl::sqrt` is 3
372384
ULP.
373385

374386
NOTE: This flag is currently only supported with the CUDA and HIP targets.
375387

376-
377388
**`-f[no-]sycl-esimd-force-stateless-mem`** [EXPERIMENTAL]
378389

379390
Enforces stateless memory access and enables the automatic conversion of

0 commit comments

Comments
 (0)