vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

sycl : Fixes to broken builds and test-backend-ops (#10257)

Browse source 
* Fixes broken build for the SYCL CUDA backend caused by non-explicit gemm call in outprod (merged in with RWKV6 in
Optimize RWKV6 Operator Naming and Implement Multi-core CPU/ SYCL Acceleration #10133)

* Marks permuted MUL_MAT as unsupported to be able to run test-backend-ops

* Fixes asserts in norm to fix debug builds.
This commit is contained in:
Alberto Cabrera Pérez 2024-11-13 09:40:57 +00:00 • committed by GitHub
parent 80dd7ff22f
commit 2e82ffa4af
No known key found for this signature in database
GPG key ID: B5690EEEBB952194
3 changed files with 9 additions and 4 deletions
Download patch file Download diff file Expand all files Collapse all files

6
ggml/src/ggml-sycl/norm.cpp
View file

@ -8,7 +8,6 @@ static void norm_f32(const float* x, float* dst, const int ncols, const float ep
const int nthreads = item_ct1.get_local_range(2);
const int nwarps = nthreads / WARP_SIZE;
assert(nwarps % WARP_SIZE == 0);
sycl::float2 mean_var = sycl::float2(0.f, 0.f);
for (int col = tid; col < ncols; col += block_size) {
@ -55,7 +54,6 @@ static void group_norm_f32(const float* x, float* dst, const int group_size, con
int end = start + group_size;
const int nthreads = item_ct1.get_local_range(2);
const int nwarps = nthreads / WARP_SIZE;
assert(nwarps % WARP_SIZE == 0);
start += item_ct1.get_local_id(2);
int nreduce = nwarps / WARP_SIZE;
@ -144,7 +142,6 @@ static void rms_norm_f32(const float* x, float* dst, const int ncols, const floa
const int tid = item_ct1.get_local_id(2);
const int nthreads = item_ct1.get_local_range(2);
const int nwarps = nthreads / WARP_SIZE;
assert(nwarps % WARP_SIZE == 0);
float tmp = 0.0f; // partial sum for thread in warp
for (int col = tid; col < ncols; col += block_size) {
@ -202,6 +199,7 @@ static void norm_f32_sycl(const float* x, float* dst, const int ncols,
}
else {
const int work_group_size = ggml_sycl_info().max_work_group_sizes[device];
assert(work_group_size % (WARP_SIZE * WARP_SIZE) == 0);
const sycl::range<3> block_dims(1, 1, work_group_size);
/*
DPCT1049:17: The work-group size passed to the SYCL kernel may exceed
@ -244,6 +242,7 @@ static void group_norm_f32_sycl(const float* x, float* dst,
}
else {
const int work_group_size = ggml_sycl_info().max_work_group_sizes[device];
assert(work_group_size % (WARP_SIZE * WARP_SIZE) == 0);
const sycl::range<3> block_dims(1, 1, work_group_size);
/*
DPCT1049:18: The work-group size passed to the SYCL kernel may exceed
@ -290,6 +289,7 @@ static void rms_norm_f32_sycl(const float* x, float* dst, const int ncols,
}
else {
const int work_group_size = ggml_sycl_info().max_work_group_sizes[device];
assert(work_group_size % (WARP_SIZE * WARP_SIZE) == 0);
const sycl::range<3> block_dims(1, 1, work_group_size);
/*
DPCT1049:19: The work-group size passed to the SYCL kernel may exceed