Skip to content

More optimizations on metal #2959

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Merged
merged 8 commits into from
Sep 3, 2023
Merged

More optimizations on metal #2959

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
8 commits
Select commit Hold shift + click to select a range
2cb47e0
Very minor speedup via simd-group synchronization in f16 x f32
Kawrakow Sep 1, 2023
e3ff8c2
Another very minor speedup on metal
Kawrakow Sep 1, 2023
2b60170
Quite significant PP speedup on metal
Kawrakow Sep 1, 2023
b557bc3
Another attempt
Kawrakow Sep 1, 2023
74df0de
Minor
Kawrakow Sep 1, 2023
01eed46
Merge branch 'master' into ik/more_metal_optimizations
ggerganov Sep 2, 2023
363f0bf
Massive improvement for TG for fp16
Kawrakow Sep 2, 2023
6af0bab
~4-5% improvement for Q8_0 TG on metal
Kawrakow Sep 3, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
22 changes: 16 additions & 6 deletions ggml-metal.m
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -76,6 +76,7 @@
GGML_METAL_DECL_KERNEL(rms_norm);
GGML_METAL_DECL_KERNEL(norm);
GGML_METAL_DECL_KERNEL(mul_mat_f16_f32);
GGML_METAL_DECL_KERNEL(mul_mat_f16_f32_1row);
GGML_METAL_DECL_KERNEL(mul_mat_q4_0_f32);
GGML_METAL_DECL_KERNEL(mul_mat_q4_1_f32);
GGML_METAL_DECL_KERNEL(mul_mat_q8_0_f32);
Expand Down Expand Up @@ -205,6 +206,7 @@ @implementation GGMLMetalClass
GGML_METAL_ADD_KERNEL(rms_norm);
GGML_METAL_ADD_KERNEL(norm);
GGML_METAL_ADD_KERNEL(mul_mat_f16_f32);
GGML_METAL_ADD_KERNEL(mul_mat_f16_f32_1row);
GGML_METAL_ADD_KERNEL(mul_mat_q4_0_f32);
GGML_METAL_ADD_KERNEL(mul_mat_q4_1_f32);
GGML_METAL_ADD_KERNEL(mul_mat_q8_0_f32);
Expand Down Expand Up @@ -270,6 +272,7 @@ void ggml_metal_free(struct ggml_metal_context * ctx) {
GGML_METAL_DEL_KERNEL(rms_norm);
GGML_METAL_DEL_KERNEL(norm);
GGML_METAL_DEL_KERNEL(mul_mat_f16_f32);
GGML_METAL_DEL_KERNEL(mul_mat_f16_f32_1row);
GGML_METAL_DEL_KERNEL(mul_mat_q4_0_f32);
GGML_METAL_DEL_KERNEL(mul_mat_q4_1_f32);
GGML_METAL_DEL_KERNEL(mul_mat_q8_0_f32);
Expand Down Expand Up @@ -854,7 +857,11 @@ void ggml_metal_graph_compute(
{
nth0 = 32;
nth1 = 1;
[encoder setComputePipelineState:ctx->pipeline_mul_mat_f16_f32];
if (ne11 * ne12 < 4) {
[encoder setComputePipelineState:ctx->pipeline_mul_mat_f16_f32_1row];
} else {
[encoder setComputePipelineState:ctx->pipeline_mul_mat_f16_f32];
}
} break;
case GGML_TYPE_Q4_0:
{
Expand Down Expand Up @@ -906,8 +913,8 @@ void ggml_metal_graph_compute(
GGML_ASSERT(ne02 == 1);
GGML_ASSERT(ne12 == 1);

nth0 = 2;
nth1 = 32;
nth0 = 4; //1;
nth1 = 8; //32;
[encoder setComputePipelineState:ctx->pipeline_mul_mat_q4_K_f32];
} break;
case GGML_TYPE_Q5_K:
Expand Down Expand Up @@ -955,9 +962,12 @@ void ggml_metal_graph_compute(
[encoder setBytes:&gqa length:sizeof(gqa) atIndex:17];

if (src0t == GGML_TYPE_Q4_0 || src0t == GGML_TYPE_Q4_1 || src0t == GGML_TYPE_Q8_0 ||
src0t == GGML_TYPE_Q2_K || src0t == GGML_TYPE_Q4_K) {
src0t == GGML_TYPE_Q2_K) {// || src0t == GGML_TYPE_Q4_K) {
[encoder dispatchThreadgroups:MTLSizeMake((ne01 + 7)/8, ne11, ne12) threadsPerThreadgroup:MTLSizeMake(nth0, nth1, 1)];
}
else if (src0t == GGML_TYPE_Q4_K) {
[encoder dispatchThreadgroups:MTLSizeMake((ne01 + 3)/4, ne11, ne12) threadsPerThreadgroup:MTLSizeMake(nth0, nth1, 1)];
}
else if (src0t == GGML_TYPE_Q3_K) {
#ifdef GGML_QKK_64
[encoder dispatchThreadgroups:MTLSizeMake((ne01 + 1)/2, ne11, ne12) threadsPerThreadgroup:MTLSizeMake(nth0, nth1, 1)];
Expand All @@ -971,8 +981,8 @@ void ggml_metal_graph_compute(
else if (src0t == GGML_TYPE_Q6_K) {
[encoder dispatchThreadgroups:MTLSizeMake((ne01 + 1)/2, ne11, ne12) threadsPerThreadgroup:MTLSizeMake(nth0, nth1, 1)];
} else {
[encoder setThreadgroupMemoryLength:nth0*sizeof(float) atIndex:0];
[encoder dispatchThreadgroups:MTLSizeMake(ne01, ne11, ne12) threadsPerThreadgroup:MTLSizeMake(nth0, nth1, 1)];
int64_t ny = (ne11 + 3)/4;
[encoder dispatchThreadgroups:MTLSizeMake(ne01, ny, ne12) threadsPerThreadgroup:MTLSizeMake(nth0, nth1, 1)];
}
}
} break;
Expand Down
Loading