EngineX-Ascend/enginex-ascend-910-llama.cpp
10
0
Fork 0
Code Issues Pull Requests Actions 4 Projects Releases Wiki Activity

metal : Add missing unary ops Metal support (#14660)

Browse Source
This commit is contained in:
Yavor Ivanov
2025-07-12 22:38:13 -07:00
committed by GitHub
parent 84b396e051
commit dcf7f2ea3c
2 changed files with 135 additions and 0 deletions
Download Patch File Download Diff File Expand all files Collapse all files

45
ggml/src/ggml-metal/ggml-metal.metal
View File

@@ -1199,6 +1199,51 @@ kernel void kernel_neg(
dst[tpig] = -src0[tpig];
}
kernel void kernel_abs(
device const float * src0,
device float * dst,
uint tpig[[thread_position_in_grid]]) {
dst[tpig] = fabs(src0[tpig]);
}
kernel void kernel_sgn(
device const float * src0,
device float * dst,
uint tpig[[thread_position_in_grid]]) {
device const float & x = src0[tpig];
dst[tpig] = (x > 0.0f) ? 1.0f : ((x < 0.0f) ? -1.0f : 0.0f);
}
kernel void kernel_step(
device const float * src0,
device float * dst,
uint tpig[[thread_position_in_grid]]) {
dst[tpig] = src0[tpig] > 0.0f ? 1.0f : 0.0f;
}
kernel void kernel_hardswish(
device const float * src0,
device float * dst,
uint tpig[[thread_position_in_grid]]) {
device const float & x = src0[tpig];
dst[tpig] = x * fmin(1.0f, fmax(0.0f, (x + 3.0f) / 6.0f));
}
kernel void kernel_hardsigmoid(
device const float * src0,
device float * dst,
uint tpig[[thread_position_in_grid]]) {
device const float & x = src0[tpig];
dst[tpig] = fmin(1.0f, fmax(0.0f, (x + 3.0f) / 6.0f));
}
kernel void kernel_exp(
device const float * src0,
device float * dst,
uint tpig[[thread_position_in_grid]]) {
dst[tpig] = exp(src0[tpig]);
}
kernel void kernel_reglu(
device const char * src0,
device const char * src1,