[js/webgpu] Optimize matmulnbits (microsoft#22360)

### Description
<!-- Describe your changes. -->
This PR further optimizes matmulnbits specially for iGPUs. The phi3 demo
becomes ~12 tokens/second from ~8 tokens on iGPUs.

Some todos:
1. Make the optimization more general, Remove the blockSize = 32
limitation.
2. Tune the parameter, such as workgroupSize, components size (currently
only support components = 1), to see the performance change.

Author: qjia7

js/web/lib/wasm/jsep/webgpu/ops/common.ts

@@ -868,6 +868,7 @@ class ShaderHelperImpl implements ShaderHelper {
     const paramList = is1DimensionDispatch
       ? `@builtin(global_invocation_id) global_id : vec3<u32>,
     @builtin(workgroup_id) workgroup_id : vec3<u32>,
+    @builtin(local_invocation_index) local_idx : u32,
     @builtin(local_invocation_id) local_id : vec3<u32>`
       : `@builtin(global_invocation_id) global_id : vec3<u32>,
                                              @builtin(local_invocation_id) local_id : vec3<u32>,
@@ -876,7 +877,6 @@ class ShaderHelperImpl implements ShaderHelper {
     @builtin(num_workgroups) num_workgroups : vec3<u32>`;
     const globalIdxDefinition = is1DimensionDispatch
       ? `let global_idx = global_id.x;
-         let local_idx = local_id.x;
          let workgroup_index = workgroup_id.x;`
       : `let workgroup_index = workgroup_id.z * num_workgroups[0] * num_workgroups[1] +
              workgroup_id.y * num_workgroups[0] + workgroup_id.x;

js/web/lib/wasm/jsep/webgpu/ops/matmulnbits.ts

@@ -266,9 +266,185 @@ export const createMatMulNBitsProgramInfo = (
   };
 };
 
+// Currently, only support blockSize = 32.
+export const createMatMulNBitsBlockSize32ProgramInfo = (
+  inputs: readonly TensorView[],
+  attributes: MatMulNBitsAttributes,
+): ProgramInfo => {
+  const inputShape = inputs[0].dims;
+  const aRank = inputShape.length;
+  const dimAOuter = inputShape[aRank - 2];
+  const dimInner = attributes.k;
+  const dimBOuter = attributes.n;
+  const batchDims = inputShape.slice(0, aRank - 2);
+  const batchSize = ShapeUtil.size(batchDims);
+  const blobSize = inputs[1].dims[2];
+  const blobSizeInWords = blobSize / 4;
+  const dataType = inputs[0].dataType;
+  const aComponents = getMaxComponents(attributes.k);
+  const bComponents = getMaxComponents(blobSizeInWords);
+  const outputShape = batchDims.concat([dimAOuter, dimBOuter]);
+
+  const workgroupSize = 128;
+  const workgroupY = dimBOuter % 8 === 0 ? 8 : dimBOuter % 4 === 0 ? 4 : 1;
+  const workgroupX = workgroupSize / workgroupY;
+  const tileSize = workgroupX * bComponents * 8; // each uint32 has 8 data.
+  const aLengthPerTile = tileSize / aComponents;
+  const blocksPerTile = tileSize / attributes.blockSize;
+  const dispatchSize = ShapeUtil.size(outputShape) / workgroupY;
+
+  const programUniforms: ProgramUniform[] = [];
+  const inputShapeTemp = [batchSize, dimAOuter, dimInner / aComponents];
+  const bShape = ShapeUtil.convertShape(inputs[1].dims).slice();
+  bShape.splice(-1, 1, blobSizeInWords / bComponents);
+  programUniforms.push(...createTensorShapeVariables(inputShapeTemp));
+  programUniforms.push(...createTensorShapeVariables(bShape));
+  programUniforms.push(...createTensorShapeVariables(inputs[2].dims));
+  if (inputs.length === 4) {
+    programUniforms.push(...createTensorShapeVariables(ShapeUtil.convertShape(inputs[3].dims)));
+  }
+  const outputShapeTemp = [batchSize, dimAOuter, dimBOuter];
+  programUniforms.push(...createTensorShapeVariables(outputShapeTemp));
+
+  const getShaderSource = (shaderHelper: ShaderHelper) => {
+    const inputRank = inputShapeTemp.length;
+    const a = inputVariable('a', inputs[0].dataType, inputRank, aComponents);
+    const b = inputVariable('b', DataType.uint32, bShape.length, bComponents);
+    const scales = inputVariable('scales', inputs[2].dataType, inputs[2].dims.length);
+    const inputVariables = [a, b, scales];
+    const zeroPoints =
+      inputs.length === 4 ? inputVariable('zero_points', DataType.uint32, inputs[3].dims.length) : undefined;
+    if (zeroPoints) {
+      inputVariables.push(zeroPoints);
+    }
+    const outputRank = outputShapeTemp.length;
+    const output = outputVariable('output', inputs[0].dataType, outputRank);
+    const dataType = tensorTypeToWsglStorageType(inputs[0].dataType);
+    const readA = () => {
+      switch (aComponents) {
+        case 1:
+          return `
+          let a_data0 = vec4<${dataType}>(sub_a[word_offset], sub_a[word_offset + 1], sub_a[word_offset + 2], sub_a[word_offset + 3]);
+          let a_data1 = vec4<${dataType}>(sub_a[word_offset + 4], sub_a[word_offset + 5], sub_a[word_offset + 6], sub_a[word_offset + 7]);`;
+        case 2:
+          return `
+          let a_data0 = vec4<${dataType}>(sub_a[word_offset], sub_a[word_offset + 1]);
+          let a_data1 = vec4<${dataType}>(sub_a[word_offset + 2], sub_a[word_offset + 3]);`;
+        case 4:
+          return `
+          let a_data0 = sub_a[word_offset];
+          let a_data1 = sub_a[word_offset + 1];`;
+        default:
+          throw new Error(`${aComponents}-component is not supported.`);
+      }
+    };
+
+    return `
+        var<workgroup> sub_a: array<${a.type.value}, ${aLengthPerTile}>;
+        var<workgroup> inter_results: array<array<${output.type.value}, ${workgroupX}>, ${workgroupY}>;
+        ${shaderHelper.declareVariables(...inputVariables, output)}
+        ${shaderHelper.mainStart([workgroupX, workgroupY, 1])}
+          let output_indices = ${output.offsetToIndices(`workgroup_index * ${workgroupY}`)};
+          let col = output_indices[2];
+          let row = output_indices[1];
+          let batch = output_indices[0];
+          let n_blocks_per_col = uniforms.b_shape[1];
+          let num_tiles =  (n_blocks_per_col - 1) / ${blocksPerTile} + 1;
+
+          // Loop over shared dimension.
+          for (var tile: u32 = 0; tile < num_tiles; tile += 1) {
+            let a_col_start = tile * ${aLengthPerTile};
+            // load one tile A data into shared memory.
+            for (var a_offset = local_idx; a_offset < ${aLengthPerTile}; a_offset += ${workgroupSize})
+            {
+              let a_col = a_col_start + a_offset;
+              if (a_col < uniforms.a_shape[2])
+              {
+                sub_a[a_offset] = ${a.getByIndices(`${a.type.indices}(batch, row, a_col)`)};
+              } else {
+                sub_a[a_offset] = ${a.type.value}(0);
+              }
+            }
+            workgroupBarrier();
+
+            // each thread process one block
+            let b_row = col + local_id.y;
+            let block = tile * ${blocksPerTile} + local_id.x;
+            ${
+              zeroPoints
+                ? `
+            let zero_point_bytes_per_col = (n_blocks_per_col + 1) / 2;
+            let zero_point_byte_count = b_row * zero_point_bytes_per_col + (block >> 0x1u);
+            let zero_point_word_index = zero_point_byte_count >> 0x2u;
+            let zero_point_byte_offset = zero_point_byte_count & 0x3u;
+            let zero_point_nibble_offset: u32 = block & 0x1u;
+            let zero_point_bits_offset = (zero_point_byte_offset << 3) + (zero_point_nibble_offset << 2);
+            let zero_point_word = ${zeroPoints.getByOffset('zero_point_word_index')} >> zero_point_bits_offset;
+            let zero_point = ${dataType}((zero_point_word) & 0xFu);`
+                : `
+            // The default zero point is 8 for unsigned 4-bit quantization.
+            let zero_point = ${dataType}(${8.0});`
+            }
+            let scale = ${scales.getByOffset(`b_row * n_blocks_per_col + block`)};
+            let b_data = ${b.getByIndices(`${b.type.indices}(b_row, block, 0)`)};
+            var word_offset = local_id.x * ${attributes.blockSize / aComponents};
+            for (var i: u32 = 0; i < ${bComponents}; i++) {
+              ${readA()}
+              let b_value = ${bComponents === 1 ? `b_data` : `b_data[i]`};
+              let b_value_lower = unpack4xU8(b_value & 0x0F0F0F0Fu);
+              let b_value_upper = unpack4xU8((b_value >> 4) & 0x0F0F0F0Fu);
+              let b_quantized_values = mat2x4<${dataType}>(${Array.from(
+                { length: 4 },
+                (_, i) => `${dataType}(b_value_lower[${i}]), ${dataType}(b_value_upper[${i}])`,
+              ).join(', ')});
+              let b_dequantized_values = (b_quantized_values - mat2x4<${dataType}>(${Array(8).fill('zero_point').join(',')})) * scale;
+              inter_results[local_id.y][local_id.x] += ${Array.from(
+                { length: 2 },
+                (_, i) => `${`dot(a_data${i}, b_dequantized_values[${i}])`}`,
+              ).join(' + ')};
+              word_offset += ${8 / aComponents};
+            }
+            workgroupBarrier();
+          }
+
+          if (local_idx < ${workgroupY}) {
+            var output_value: ${output.type.value} = ${output.type.value}(0);
+            for (var b = 0u; b < ${workgroupX}; b++) {
+              output_value += inter_results[local_idx][b];
+            }
+            if (col + local_idx < uniforms.output_shape[2])
+            {
+              ${output.setByIndices(`${output.type.indices}(batch, row, col + local_idx)`, 'output_value')}
+            }
+          }
+        }`;
+  };
+  return {
+    name: 'BlockwiseMatMulNBits32',
+    shaderCache: {
+      hint: `${attributes.blockSize};${aComponents};${bComponents};${workgroupX};${workgroupY}`,
+      inputDependencies: Array(inputs.length).fill('rank'),
+    },
+    getRunData: () => ({
+      outputs: [{ dims: outputShape, dataType }],
+      dispatchGroup: { x: dispatchSize },
+      programUniforms,
+    }),
+    getShaderSource,
+  };
+};
+
 export const matMulNBits = (context: ComputeContext, attributes: MatMulNBitsAttributes): void => {
   validateInputs(context.inputs, attributes);
-  context.compute(createMatMulNBitsProgramInfo(context.inputs, attributes));
+  if (
+    attributes.blockSize === 32 &&
+    context.adapterInfo.isVendor('intel') &&
+    context.adapterInfo.isArchitecture('gen-12lp')
+  ) {
+    context.compute(createMatMulNBitsBlockSize32ProgramInfo(context.inputs, attributes));
+  } else {
+    context.compute(createMatMulNBitsProgramInfo(context.inputs, attributes));
+  }
 };
 
 export const parseMatMulNBitsAttributes = (attributes: Record<string, unknown>): MatMulNBitsAttributes =>

js/web/lib/wasm/jsep/webgpu/types.ts

@@ -15,7 +15,7 @@ export enum GpuDataType {
 }
 export type GpuDataId = number;
 
-export type GpuArchitecture = 'ampere';
+export type GpuArchitecture = 'ampere' | 'gen-12lp';
 export type GpuVendor = 'amd' | 'intel' | 'nvidia';
 export interface AdapterInfo {
   isArchitecture: (architecture: GpuArchitecture) => boolean;