Skip to content

WIP: DP4AMatMul fix matmul for subgoup size 64 GPUs #23637

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
guschmue merged 1 commit into from
Feb 13, 2025
Merged

WIP: DP4AMatMul fix matmul for subgoup size 64 GPUs #23637

Changes from all commits
Commits
Show all changes
1 commit
Select commit
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
45 changes: 13 additions & 32 deletions onnxruntime/contrib_ops/webgpu/quantization/matmul_nbits.cc
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -535,42 +535,23 @@ Status DP4AMatMulQuantizeProgram::GenerateShaderCode(ShaderHelper& shader) const
shader.AddOutput("output", ShaderUsage::UseUniform);
shader.AddOutput("scales", ShaderUsage::UseUniform);

shader.AdditionalImplementation() << R"ADDNL_FN(
var<workgroup> max_values : array<input_a_element_t, 4>;
)ADDNL_FN";

shader.MainFunctionBody() << R"MAIN_FN(
var local_a = input_a[global_idx];
var max_val = subgroupMax(abs(local_a));
var max_temp = max(max_val.xy, max_val.zw);
var scale = max(max_temp[0], max_temp[1]);
if (local_idx % sg_size == 0) {
max_values[local_idx / sg_size] = scale;
}
workgroupBarrier();

if (sg_size == 8)
var local_a : array<vec4<input_a_element_t>, 32>;
var max_value:vec4<input_a_element_t> = vec4<input_a_element_t>(0);
for (var idx:u32=0;idx<32;idx+=1)
{
scale = max(max_values[0], max_values[1]);
scale = max(scale, max_values[2]);
scale = max(scale, max_values[3]);
local_a[idx] = input_a[workgroup_id.x*32 + idx];
max_value = max(max_value, abs(local_a[idx]));
}
else if (sg_size == 16)
{
scale = max(max_values[0], max_values[1]);
}
else
{
scale = max_values[0];
}

var norm_a = local_a/scale;
output[global_idx] = pack4x8snorm(vec4<f32>(norm_a));
if (local_idx == 0)
var scale = max(max_value.x, max_value.y);
scale = max(scale, max_value.z);
scale = max(scale, max_value.w);
for (var idx:u32=0;idx<32;idx+=1)
{
// 127 is the max value of signed int8 [-127,127] used by pack4x8snorm for 1.0f.
scales[workgroup_idx] = scale/127;
output[workgroup_id.x*32+idx] = pack4x8snorm(vec4<f32>(local_a[idx]/scale));
}
// 127 is the max value of signed int8 [-127,127] used by pack4x8snorm for 1.0f.
scales[workgroup_id.x] = scale/127;
)MAIN_FN";
return Status::OK();
}
Expand Down Expand Up @@ -838,7 +819,7 @@ Status MatMulNBits::ComputeInternal(onnxruntime::webgpu::ComputeContext& context

constexpr uint32_t kBlockSizeA = 128;
DP4AMatMulQuantizeProgram quantize_program;
quantize_program.SetWorkgroupSize(32);
quantize_program.SetWorkgroupSize(1);
quantize_program.SetDispatchGroupSize(M * K / kBlockSizeA, 1, 1);
TensorShape a_quant_shape{1, M, K / kU32Components};
Tensor a_quant = context.CreateGPUTensor(DataTypeImpl::GetType<uint32_t>(), a_quant_shape);
Expand Down
Loading