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

HIP: Cleanup hipification header (#15285)

Browse Source 
add expicit conversion operator to support older versions of rocm
Switch over to hip_bf16 from legacy hip_bfloat16
Simplify RDNA3 define
Reduce swap over of new hipblas api to rocm 6.5 as this version is used for rocm 7.0 previews

---------

Co-authored-by: Johannes Gäßler <johannesg@5d6.de>
This commit is contained in:
uvos
2025-08-14 16:23:56 +02:00
committed by GitHub
parent b204a5a234
commit 5ba36f6103
7 changed files with 32 additions and 33 deletions
Download Patch File Download Diff File Expand all files Collapse all files

7
ggml/src/ggml-cuda/getrows.cu
View File

@@ -1,5 +1,6 @@
#include "getrows.cuh"
#include "dequantize.cuh"
#include "convert.cuh"
template<int qk, int qr, dequantize_kernel_t dequantize_kernel, typename dst_t>
static __global__ void k_get_rows(
@@ -34,8 +35,8 @@ static __global__ void k_get_rows(
dfloat2 v;
dequantize_kernel(src0_row, ib, iqs, v);
dst_row[iybs + iqs + 0] = float(v.x);
dst_row[iybs + iqs + y_offset] = float(v.y);
dst_row[iybs + iqs + 0] = ggml_cuda_cast<dst_t>(v.x);
dst_row[iybs + iqs + y_offset] = ggml_cuda_cast<dst_t>(v.y);
}
template<typename src0_t, typename dst_t>
@@ -62,7 +63,7 @@ static __global__ void k_get_rows_float(
dst_t * dst_row = dst + i10*s1 + i11*s2 + i12*s3;
const src0_t * src0_row = (const src0_t *)((const char *) src0 + i01*nb01 + i11*nb02 + i12*nb03);
dst_row[i00] = float(src0_row[i00]);
dst_row[i00] = ggml_cuda_cast<dst_t>(src0_row[i00]);
}
template<typename grad_t, typename dst_t>