Refactor Vulkan backend to allow multiple contexts (ggerganov#7961)

* Refactor Vulkan backend to allow multiple contexts * Fix too many shader groups called validation error in llama3 on AMD and Intel GPUs * Fix Vulkan debug build error
arthw · Jun 30, 2024 · fdcfd66 · fdcfd66
1 parent f0a6455
commit fdcfd66
Show file tree

Hide file tree

Showing 8 changed files with 20,788 additions and 20,442 deletions.
diff --git a/ggml-vulkan-shaders.hpp b/ggml-vulkan-shaders.hpp
diff --git a/ggml-vulkan.cpp b/ggml-vulkan.cpp
diff --git a/vulkan-shaders/mul_mat_vec.comp b/vulkan-shaders/mul_mat_vec.comp
@@ -13,7 +13,7 @@ layout (constant_id = 0) const uint BLOCK_SIZE = 32;
 shared FLOAT_TYPE tmp[BLOCK_SIZE];
 
 void main() {
-    const uint row = gl_WorkGroupID.x;
+    const uint row = gl_WorkGroupID.x + gl_NumWorkGroups.x * gl_WorkGroupID.z;
     const uint tid = gl_LocalInvocationID.x;
 
     uint a_offset, b_offset, d_offset;

diff --git a/vulkan-shaders/mul_mat_vec_q2_k.comp b/vulkan-shaders/mul_mat_vec_q2_k.comp
@@ -7,7 +7,7 @@ layout(local_size_x = 32, local_size_y = 1, local_size_z = 1) in;
 shared FLOAT_TYPE tmp[32];
 
 void main() {
-    const uint row = gl_WorkGroupID.x;
+    const uint row = gl_WorkGroupID.x + gl_NumWorkGroups.x * gl_WorkGroupID.z;
 
     uint a_offset, b_offset, d_offset;
     get_offsets(a_offset, b_offset, d_offset);

diff --git a/vulkan-shaders/mul_mat_vec_q3_k.comp b/vulkan-shaders/mul_mat_vec_q3_k.comp
@@ -7,7 +7,7 @@ layout(local_size_x = 32, local_size_y = 1, local_size_z = 1) in;
 shared FLOAT_TYPE tmp[32];
 
 void main() {
-    const uint row = gl_WorkGroupID.x;
+    const uint row = gl_WorkGroupID.x + gl_NumWorkGroups.x * gl_WorkGroupID.z;
 
     uint a_offset, b_offset, d_offset;
     get_offsets(a_offset, b_offset, d_offset);

diff --git a/vulkan-shaders/mul_mat_vec_q4_k.comp b/vulkan-shaders/mul_mat_vec_q4_k.comp
@@ -7,7 +7,7 @@ layout(local_size_x = 32, local_size_y = 1, local_size_z = 1) in;
 shared FLOAT_TYPE tmp[32];
 
 void main() {
-    const uint row = gl_WorkGroupID.x;
+    const uint row = gl_WorkGroupID.x + gl_NumWorkGroups.x * gl_WorkGroupID.z;
 
     uint a_offset, b_offset, d_offset;
     get_offsets(a_offset, b_offset, d_offset);

diff --git a/vulkan-shaders/mul_mat_vec_q5_k.comp b/vulkan-shaders/mul_mat_vec_q5_k.comp
@@ -7,7 +7,7 @@ layout(local_size_x = 32, local_size_y = 1, local_size_z = 1) in;
 shared FLOAT_TYPE tmp[32];
 
 void main() {
-    const uint row = gl_WorkGroupID.x;
+    const uint row = gl_WorkGroupID.x + gl_NumWorkGroups.x * gl_WorkGroupID.z;
 
     uint a_offset, b_offset, d_offset;
     get_offsets(a_offset, b_offset, d_offset);

diff --git a/vulkan-shaders/mul_mat_vec_q6_k.comp b/vulkan-shaders/mul_mat_vec_q6_k.comp
@@ -7,7 +7,7 @@ layout(local_size_x = 32, local_size_y = 1, local_size_z = 1) in;
 shared FLOAT_TYPE tmp[32];
 
 void main() {
-    const uint row = gl_WorkGroupID.x;
+    const uint row = gl_WorkGroupID.x + gl_NumWorkGroups.x * gl_WorkGroupID.z;
 
     uint a_offset, b_offset, d_offset;
     get_offsets(a_offset, b_offset, d_offset);