[SYCL] Add tanh in BF16 math (#20883)

We received request for tanh in bfloat16 math functions. This PR adds it
for SPIRV targets using fp32 emulation and update extension doc.

---------

Signed-off-by: jinge90 <[email protected]>

Author: jinge90

sycl/doc/extensions/experimental/sycl_ext_oneapi_bfloat16_math_functions.asciidoc

@@ -57,14 +57,14 @@ specification.
 
 This extension adds `bfloat16` support to the `fma`, `fmin`, `fmax`, `fabs`,
 `isnan`, `ceil`, `floor`, `cos`, `sin`, `exp`, `exp2`, `exp10`, `log`, `log2`,
-`log10`, `rint`, `sqrt`, `rsqrt` and `trunc` SYCL floating point math functions.
-These functions can be used as element wise operations on matrices, supplementing
-the `bfloat16` support in the sycl_ext_oneapi_matrix extension.
+`log10`, `rint`, `sqrt`, `rsqrt`, `trunc` and `tanh` SYCL floating point math
+functions. These functions can be used as element wise operations on matrices,
+supplementing the `bfloat16` support in the sycl_ext_oneapi_matrix extension.
 
 The descriptions of the `fma`, `fmin`, `fmax`, `fabs`, `isnan`, `ceil`, `floor`,
 `cos`, `sin`, `exp`, `exp2`, `exp10`, `log`, `log2`, `log10`, `rint`, `sqrt`,
-`rsqrt` and `trunc` SYCL floating point math functions can be found in the SYCL
-specification:
+`rsqrt`, `trunc` and `tanh` SYCL floating point math functions can be found in
+the SYCL specification:
 https://www.khronos.org/registry/SYCL/specs/sycl-2020/html/sycl-2020.html#_math_functions.
 
 [NOTE]

sycl/include/sycl/ext/oneapi/experimental/bfloat16_math.hpp

@@ -476,6 +476,10 @@ BFLOAT16_MATH_FP32_WRAPPERS(trunc)
 BFLOAT16_MATH_FP32_WRAPPERS_MARRAY(trunc)
 BFLOAT16_MATH_FP32_WRAPPERS_VEC(trunc)
 
+BFLOAT16_MATH_FP32_WRAPPERS(tanh)
+BFLOAT16_MATH_FP32_WRAPPERS_MARRAY(tanh)
+BFLOAT16_MATH_FP32_WRAPPERS_VEC(tanh)
+
 #undef BFLOAT16_MATH_FP32_WRAPPERS
 #undef BFLOAT16_MATH_FP32_WRAPPERS_MARRAY
 #undef BFLOAT16_MATH_FP32_WRAPPERS_VEC

sycl/test-e2e/BFloat16/bfloat16_builtins.hpp

@@ -187,6 +187,7 @@ void test() {
   TEST(exp10, 1);
   TEST(exp2, 1);
   TEST(rint, 1);
+  TEST(tanh, 1);
 
   // log, log2, log10, sqrt, rsqrt testing, the input
   // must be positive.