whispercpp 1.3.4 → 1.3.5

data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_norm.comp

@@ -1,41 +1,14 @@
 #version 450
 
-#include "rope_head.comp"
+#include "rope_head.glsl"
+#include "rope_funcs.glsl"
 
 void main() {
     const uint i0 = 2*gl_GlobalInvocationID.y;
-    uint ne0 = p.ncols;
-    uint ne1 = p.p_delta_rows;
-
-    if (i0 >= ne0) {
+    // i1 is actually i2*nb2+i1, but the rows are contiguous
+    const uint i1 = gl_GlobalInvocationID.x + 32768 * gl_GlobalInvocationID.z;
+    if (i1 >= pc.nrows) {
         return;
     }
-
-    const uint row_dst = gl_GlobalInvocationID.x;
-
-    const uint row_x     = row_dst % ne1;
-    const uint channel_x = row_dst / ne1;
-
-    const uint idst = row_dst*ne0 + i0;
-    const uint ix   = channel_x*p.s2 + row_x*p.s1 + i0;
-
-    if (i0 >= p.n_dims) {
-        data_d[idst + 0] = data_a[ix + 0];
-        data_d[idst + 1] = data_a[ix + 1];
-
-        return;
-    }
-
-    const float theta_base = data_pos[channel_x] * pow(p.theta_scale, i0/2.0f);
-
-    const float freq_factor = p.has_ff != 0 ? data_ff[i0/2] : 1.0f;
-
-    float cos_theta, sin_theta;
-    rope_yarn(theta_base / freq_factor, i0, cos_theta, sin_theta);
-
-    const float x0 = float(data_a[ix + 0]);
-    const float x1 = float(data_a[ix + 1]);
-
-    data_d[idst + 0] = D_TYPE(x0*cos_theta - x1*sin_theta);
-    data_d[idst + 1] = D_TYPE(x0*sin_theta + x1*cos_theta);
+    rope_norm(i0, i1, pc);
 }

data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_params.glsl

@@ -0,0 +1,28 @@
+#if !defined(GGML_ROPE_PARAMS)
+#define GGML_ROPE_PARAMS
+
+#include "rte.glsl"
+
+struct rope_params {
+    uint rope_mode;
+    uint ncols;
+    uint nrows;
+    uint n_dims;
+    float freq_scale;
+    uint p_delta_rows;
+    float freq_base;
+    float ext_factor;
+    float attn_factor;
+    float corr_dims[2];
+    float theta_scale;
+    uint has_ff;
+    uint ne02;
+    uint nb01;
+    uint nb02;
+    int sections[4];
+    uint is_imrope;
+    uint is_back;
+    uint set_rows_stride;
+};
+
+#endif // !defined(GGML_ROPE_PARAMS)

data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/rope_vision.comp

@@ -1,47 +1,14 @@
 #version 450
 
-#include "rope_head.comp"
+#include "rope_head.glsl"
+#include "rope_funcs.glsl"
 
 void main() {
     const uint i0 = 2*gl_GlobalInvocationID.y;
-    uint ne0 = p.ncols;
-    uint ne1 = p.p_delta_rows;
-    uint ne2 = p.ne02;
-
-    if (i0 >= ne0) {
+    // i1 is actually i2*nb2+i1, but the rows are contiguous
+    const uint i1 = gl_GlobalInvocationID.x + 32768 * gl_GlobalInvocationID.z;
+    if (i1 >= pc.nrows) {
         return;
     }
-
-    const uint row_dst = gl_GlobalInvocationID.x;
-
-    const uint row_x     = row_dst % ne1;
-    const uint channel_x = row_dst / ne1;
-
-    const uint idst = row_dst*ne0 + i0/2;
-    const uint ix   = channel_x*p.s2 + row_x*p.s1 + i0/2;
-
-    const int sect_dims = p.sections[0] + p.sections[1];
-    const int sec_w = p.sections[1] + p.sections[0];
-    const uint sector = (i0 / 2) % sect_dims;
-
-    float theta_base = 0.0;
-    if (sector < p.sections[0]) {
-        const uint p0 = sector;
-        theta_base = data_pos[channel_x]*pow(p.theta_scale, p0);
-    }
-    else if (sector >= p.sections[0] && sector < sec_w) {
-        const uint p0 = sector - p.sections[0];
-        theta_base = data_pos[channel_x + ne2]*pow(p.theta_scale, p0);
-    }
-
-    const float freq_factor = p.has_ff != 0 ? data_ff[i0/2] : 1.0f;
-
-    float cos_theta, sin_theta;
-    rope_yarn(theta_base / freq_factor, i0, cos_theta, sin_theta);
-
-    const float x0 = float(data_a[ix + 0]);
-    const float x1 = float(data_a[ix + p.n_dims]);
-
-    data_d[idst + 0]        = D_TYPE(x0*cos_theta - x1*sin_theta);
-    data_d[idst + p.n_dims] = D_TYPE(x0*sin_theta + x1*cos_theta);
+    rope_vision(i0, i1, pc);
 }

data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/round.comp

@@ -0,0 +1,29 @@
+#version 450
+
+#include "generic_head.glsl"
+#include "types.glsl"
+
+#extension GL_EXT_control_flow_attributes : enable
+
+layout(local_size_x = 512, local_size_y = 1, local_size_z = 1) in;
+
+layout (binding = 0) readonly buffer X {A_TYPE data_a[];};
+layout (binding = 1) writeonly buffer D {D_TYPE data_d[];};
+
+void main() {
+    const uint i = gl_GlobalInvocationID.z * 262144 + gl_GlobalInvocationID.y * 512 + gl_GlobalInvocationID.x;
+
+    if (i >= p.KX) {
+        return;
+    }
+
+    const float x = float(data_a[i]);
+    float result;
+    // Round halfway cases away from zero as roundf does.
+    if (x >= 0.0) {
+        result = floor(x + 0.5);
+    } else {
+        result = ceil(x - 0.5);
+    }
+    data_d[i] = D_TYPE(result);
+}

data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/scale.comp

@@ -1,7 +1,7 @@
 #version 450
 
-#include "types.comp"
-#include "generic_unary_head.comp"
+#include "types.glsl"
+#include "generic_unary_head.glsl"
 
 const uint num_threads = 128;
 

data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/sigmoid.comp

@@ -1,7 +1,7 @@
 #version 450
 
-#include "generic_head.comp"
-#include "types.comp"
+#include "generic_head.glsl"
+#include "types.glsl"
 
 #extension GL_EXT_control_flow_attributes : enable
 

data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/silu.comp

@@ -1,7 +1,7 @@
 #version 450
 
-#include "generic_head.comp"
-#include "types.comp"
+#include "generic_head.glsl"
+#include "types.glsl"
 
 #extension GL_EXT_control_flow_attributes : enable
 

data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/silu_back.comp

@@ -1,7 +1,7 @@
 #version 450
 
-#include "generic_head.comp"
-#include "types.comp"
+#include "generic_head.glsl"
+#include "types.glsl"
 
 #extension GL_EXT_control_flow_attributes : enable
 

data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/sin.comp

@@ -1,7 +1,7 @@
 #version 450
 
-#include "types.comp"
-#include "generic_unary_head.comp"
+#include "types.glsl"
+#include "generic_unary_head.glsl"
 
 layout(local_size_x = 512, local_size_y = 1, local_size_z = 1) in;
 

data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/soft_max.comp

@@ -23,7 +23,7 @@ layout (push_constant) uniform parameter
     uint has_sinks;
 } p;
 
-#include "types.comp"
+#include "types.glsl"
 
 layout(constant_id = 0) const uint BLOCK_SIZE = 32;
 layout(local_size_x_id = 0, local_size_y = 1, local_size_z = 1) in;

data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/soft_max_back.comp

@@ -2,8 +2,8 @@
 
 #extension GL_EXT_control_flow_attributes : enable
 
-#include "generic_head.comp"
-#include "types.comp"
+#include "generic_head.glsl"
+#include "types.glsl"
 
 layout(constant_id = 0) const uint BLOCK_SIZE = 32;
 layout(local_size_x_id = 0, local_size_y = 1, local_size_z = 1) in;

data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/soft_max_large1.comp

@@ -0,0 +1,62 @@
+#version 450
+
+#include "soft_max_large_common.glsl"
+
+void main() {
+    const uint tid = gl_LocalInvocationID.x;
+    const uint rowx = gl_WorkGroupID.y;
+    const uint wg_start = gl_WorkGroupID.x * BLOCK_SIZE * num_iters;
+
+    const uint32_t i03 = rowx / (p.ne01 * p.ne02);
+    const uint32_t i02 = (rowx - i03 * p.ne01 * p.ne02) / p.ne01;
+    const uint32_t i01 = rowx % p.ne01;
+
+    uint rowy_start = 0;
+    if (p.KY > 0) {
+        rowy_start = i01 * p.nb11 + (i02 % p.ne12) * p.nb12 + (i03 % p.ne13) * p.nb13;
+    }
+
+    if (rowx >= p.nrows_x) {
+        return;
+    }
+
+    float slope = get_slope(rowx);
+
+    // Find max
+    FLOAT_TYPE max_val = p.has_sinks == 0 ? uintBitsToFloat(0xFF800000) : data_c[i02];
+
+    [[unroll]] for (uint col0 = wg_start, idx = 0; idx < num_iters; col0 += BLOCK_SIZE, ++idx) {
+        const uint col = col0 + tid;
+
+        FLOAT_TYPE a = FLOAT_TYPE(0);
+        if (col < p.KX) {
+            a = data_a[rowx * p.KX + col];
+        }
+
+        FLOAT_TYPE b = FLOAT_TYPE(0);
+        if (p.KY > 0 && col < p.KX) {
+            b = data_b[rowy_start + col];
+        }
+
+        FLOAT_TYPE v = a * p.scale + slope * b;
+
+        if (col < p.KX) {
+            max_val = max(max_val, v);
+        }
+    }
+
+    // reduce across the workgroup
+    vals[tid] = max_val;
+    barrier();
+    [[unroll]] for (uint s = BLOCK_SIZE / 2; s > 0; s >>= 1) {
+        if (tid < s) {
+            vals[tid] = max(vals[tid], vals[tid + s]);
+        }
+        barrier();
+    }
+
+    if (tid == 0) {
+        max_val = vals[0];
+        data_m[rowx * gl_NumWorkGroups.x + gl_WorkGroupID.x] = max_val;
+    }
+}

data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/soft_max_large2.comp

@@ -0,0 +1,79 @@
+#version 450
+
+#include "soft_max_large_common.glsl"
+
+void main() {
+    const uint tid = gl_LocalInvocationID.x;
+    const uint rowx = gl_WorkGroupID.y;
+    const uint wg_start = gl_WorkGroupID.x * BLOCK_SIZE * num_iters;
+
+    const uint32_t i03 = rowx / (p.ne01 * p.ne02);
+    const uint32_t i02 = (rowx - i03 * p.ne01 * p.ne02) / p.ne01;
+    const uint32_t i01 = rowx % p.ne01;
+
+    uint rowy_start = 0;
+    if (p.KY > 0) {
+        rowy_start = i01 * p.nb11 + (i02 % p.ne12) * p.nb12 + (i03 % p.ne13) * p.nb13;
+    }
+
+    if (rowx >= p.nrows_x) {
+        return;
+    }
+
+    float slope = get_slope(rowx);
+
+    // Find max
+    FLOAT_TYPE max_val = p.has_sinks == 0 ? uintBitsToFloat(0xFF800000) : data_c[i02];
+
+    [[unroll]] for (uint i = 0; i < gl_NumWorkGroups.x; i += BLOCK_SIZE) {
+        if (i + tid < gl_NumWorkGroups.x) {
+            max_val = max(max_val, data_m[rowx * gl_NumWorkGroups.x + i + tid]);
+        }
+    }
+
+    // reduce across the workgroup
+    vals[tid] = max_val;
+    barrier();
+    [[unroll]] for (uint s = BLOCK_SIZE / 2; s > 0; s >>= 1) {
+        if (tid < s) {
+            vals[tid] = max(max_val, vals[tid + s]);
+        }
+        barrier();
+    }
+
+    max_val = vals[0];
+    barrier();
+
+    FLOAT_TYPE sum = FLOAT_TYPE(0.0f);
+
+    // Compute sum{exp(x - max)}
+    [[unroll]] for (uint col0 = wg_start, idx = 0; idx < num_iters; col0 += BLOCK_SIZE, ++idx) {
+        const uint col = col0 + tid;
+
+        if (col >= p.KX) {
+            break;
+        }
+
+        // compute exp(a*scale+b*slope), add it to sum
+        const uint i = rowx * p.KX + col;
+        FLOAT_TYPE val;
+        val = exp(FLOAT_TYPE(data_a[i]) * p.scale + (p.KY > 0 ? slope * FLOAT_TYPE(data_b[rowy_start + col]) : FLOAT_TYPE(0.0f)) - max_val);
+        sum += val;
+        data_d[i] = D_TYPE(val);
+    }
+
+    // reduce across the workgroup
+    vals[tid] = sum;
+    barrier();
+    [[unroll]] for (uint s = BLOCK_SIZE / 2; s > 0; s >>= 1) {
+        if (tid < s) {
+            vals[tid] += vals[tid + s];
+        }
+        barrier();
+    }
+
+    if (tid == 0) {
+        sum = vals[0];
+        data_s[rowx * gl_NumWorkGroups.x + gl_WorkGroupID.x] = sum;
+    }
+}

data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/soft_max_large3.comp

@@ -0,0 +1,65 @@
+#version 450
+
+#include "soft_max_large_common.glsl"
+
+shared FLOAT_TYPE sumsh[BLOCK_SIZE];
+
+void main() {
+    const uint tid = gl_LocalInvocationID.x;
+    const uint rowx = gl_WorkGroupID.y;
+    const uint wg_start = gl_WorkGroupID.x * BLOCK_SIZE * num_iters;
+
+    const uint32_t i03 = rowx / (p.ne01 * p.ne02);
+    const uint32_t i02 = (rowx - i03 * p.ne01 * p.ne02) / p.ne01;
+    const uint32_t i01 = rowx % p.ne01;
+
+    uint rowy_start = 0;
+    if (p.KY > 0) {
+        rowy_start = i01 * p.nb11 + (i02 % p.ne12) * p.nb12 + (i03 % p.ne13) * p.nb13;
+    }
+
+    if (rowx >= p.nrows_x) {
+        return;
+    }
+
+    FLOAT_TYPE max_val = p.has_sinks == 0 ? uintBitsToFloat(0xFF800000) : data_c[i02];
+    FLOAT_TYPE sum = FLOAT_TYPE(0.0f);
+
+    [[unroll]] for (uint i = 0; i < gl_NumWorkGroups.x; i += BLOCK_SIZE) {
+        if (i + tid < gl_NumWorkGroups.x) {
+            max_val = max(max_val, data_m[rowx * gl_NumWorkGroups.x + i + tid]);
+            sum += data_s[rowx * gl_NumWorkGroups.x + i + tid];
+        }
+    }
+
+    // reduce across the workgroup
+    vals[tid] = max_val;
+    sumsh[tid] = sum;
+    barrier();
+    [[unroll]] for (uint s = BLOCK_SIZE / 2; s > 0; s >>= 1) {
+        if (tid < s) {
+            vals[tid] = max(max_val, vals[tid + s]);
+            sumsh[tid] += sumsh[tid + s];
+        }
+        barrier();
+    }
+
+    max_val = vals[0];
+    sum = sumsh[0];
+
+    if (p.has_sinks != 0) {
+        sum += FLOAT_TYPE(exp(FLOAT_TYPE(data_c[i02]) - max_val));
+    }
+
+    FLOAT_TYPE rcpdivisor = 1.0/sum;
+
+    [[unroll]] for (uint col0 = wg_start, idx = 0; idx < num_iters; col0 += BLOCK_SIZE, ++idx) {
+        const uint col = col0 + tid;
+
+        if (col >= p.KX) {
+            continue;
+        }
+
+        data_d[rowx*p.KX + col] *= D_TYPE(rcpdivisor);
+    }
+}

data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/soft_max_large_common.glsl

@@ -0,0 +1,53 @@
+#extension GL_EXT_control_flow_attributes : enable
+
+layout (push_constant) uniform parameter
+{
+    uint KX;
+    uint KY;
+    uint ne00;
+    uint ne01;
+    uint ne02;
+    uint ne12;
+    uint ne13;
+    uint nb11;
+    uint nb12;
+    uint nb13;
+    float scale;
+    float max_bias;
+    float m0;
+    float m1;
+    uint n_head_log2;
+    uint nrows_x;
+    uint has_sinks;
+} p;
+
+#include "types.glsl"
+
+layout(constant_id = 0) const uint BLOCK_SIZE = 128;
+layout(local_size_x_id = 0, local_size_y = 1, local_size_z = 1) in;
+layout(constant_id = 1) const uint num_iters = 4;
+
+layout (binding = 0) readonly buffer X {A_TYPE data_a[];};
+layout (binding = 1) readonly buffer Y {B_TYPE data_b[];};
+layout (binding = 2) readonly buffer Z {float data_c[];};
+layout (binding = 3) buffer D {D_TYPE data_d[];};
+layout (binding = 4) buffer M {float data_m[];};
+layout (binding = 5) buffer S {float data_s[];};
+
+shared FLOAT_TYPE vals[BLOCK_SIZE];
+
+float get_slope(uint rowx) {
+    float slope = 1.0f;
+
+    // ALiBi
+    if (p.max_bias > 0.0f) {
+        const uint h = (rowx / p.ne01) % p.ne02; // head index
+
+        const float base = h < p.n_head_log2 ? p.m0 : p.m1;
+        const uint   exp = h < p.n_head_log2 ? h + 1 : 2*(h - p.n_head_log2) + 1;
+
+        slope = pow(base, exp);
+    }
+
+    return slope;
+}

data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/softplus.comp

@@ -0,0 +1,23 @@
+#version 450
+
+#include "generic_head.glsl"
+#include "types.glsl"
+
+#extension GL_EXT_control_flow_attributes : enable
+
+layout(local_size_x = 512, local_size_y = 1, local_size_z = 1) in;
+
+layout (binding = 0) readonly buffer X {A_TYPE data_a[];};
+layout (binding = 1) writeonly buffer D {D_TYPE data_d[];};
+
+void main() {
+    const uint i = gl_GlobalInvocationID.z * 262144 + gl_GlobalInvocationID.y * 512 + gl_GlobalInvocationID.x;
+
+    if (i >= p.KX) {
+        return;
+    }
+
+    const float x = float(data_a[i]);
+    const float result = (x > 20.0f) ? x : log(1.0f + exp(x));
+    data_d[i] = D_TYPE(result);
+}

data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/solve_tri.comp

@@ -0,0 +1,81 @@
+#version 450
+
+#include "types.glsl"
+#include "generic_binary_head.glsl"
+
+layout (constant_id = 1) const uint N = 64;
+layout (constant_id = 2) const uint K = 32;
+layout (constant_id = 3) const uint BATCH_N = 32;
+
+layout(local_size_x_id = 4, local_size_y = 1, local_size_z = 1) in;
+
+uint a_base, b_base, x_base;
+
+FLOAT_TYPE get_a(uint r, uint c) {
+    return FLOAT_TYPE(data_a[a_base + r * p.nb01 + c * p.nb00]);
+}
+
+FLOAT_TYPE get_b(uint r, uint c) {
+    return FLOAT_TYPE(data_b[b_base + r * p.nb11 + c * p.nb10]);
+}
+
+void store_x(uint r, uint c, FLOAT_TYPE v) {
+    data_d[x_base + r * p.nb21 + c * p.nb20] = D_TYPE(v);
+}
+
+shared FLOAT_TYPE shA[BATCH_N * N];
+shared FLOAT_TYPE shB[BATCH_N * K];
+
+void main() {
+    const uint batch = gl_WorkGroupID.z * 262144 + gl_WorkGroupID.y * 512 + gl_WorkGroupID.x;
+    const uint tid = gl_LocalInvocationID.x;
+
+    if (batch >= p.ne02 * p.ne03) {
+        return;
+    }
+
+    const uint i3 = batch / p.ne22;
+    const uint i2 = batch % p.ne22;
+    a_base = get_aoffset() + i2 * p.nb02 + i3 * p.nb03;
+    b_base = get_boffset() + i2 * p.nb12 + i3 * p.nb13;
+    x_base = get_doffset() + i2 * p.nb22 + i3 * p.nb23;
+
+    FLOAT_TYPE X[N];
+
+    // Loop over batches of rows
+    [[unroll]] for (uint row_base = 0; row_base < N; row_base += BATCH_N) {
+        const uint cur_N = min(BATCH_N, N - row_base);
+
+        // Load the A matrix batch into shA
+        [[unroll]] for (uint i = 0; i < cur_N * N; i += gl_WorkGroupSize.x) {
+            uint idx = i + tid;
+            if (((cur_N * N) % gl_WorkGroupSize.x == 0) || idx < cur_N * N) {
+                shA[idx] = get_a(row_base + idx / N, idx % N);
+            }
+        }
+        // Load the B matrix batch into shB
+        [[unroll]] for (uint i = 0; i < cur_N * K; i += gl_WorkGroupSize.x) {
+            uint idx = i + tid;
+            if (((cur_N * K) % gl_WorkGroupSize.x == 0) || idx < cur_N * K) {
+                shB[idx] = get_b(row_base + idx / K, idx % K);
+            }
+        }
+        barrier();
+
+        // Each thread solves one column
+        if (tid < K) {
+            [[unroll]] for (uint row_offset = 0; row_offset < cur_N; ++row_offset) {
+                uint r = row_base + row_offset;
+                FLOAT_TYPE b = shB[row_offset * K + tid];
+                // Compute x[r,c] = (b[r,c] - sum(a[r,c]*x[c])) / a[r,r]
+                [[unroll]] for (int c = 0; c < r; ++c) {
+                    b -= shA[row_offset * N + c] * X[c];
+                }
+                FLOAT_TYPE x = b / shA[row_offset * N + r];
+                X[r] = x;
+                store_x(r, tid, x);
+            }
+        }
+        barrier();
+    }
+}

data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/sqrt.comp

@@ -1,7 +1,7 @@
 #version 450
 
-#include "types.comp"
-#include "generic_unary_head.comp"
+#include "types.glsl"
+#include "generic_unary_head.glsl"
 
 layout(local_size_x = 512, local_size_y = 1, local_size_z = 1) in;
 

data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/square.comp

@@ -1,7 +1,7 @@
 #version 450
 
-#include "types.comp"
-#include "generic_unary_head.comp"
+#include "types.glsl"
+#include "generic_unary_head.glsl"
 
 layout(local_size_x = 512, local_size_y = 1, local_size_z = 1) in;
 

data/ext/sources/ggml/src/ggml-vulkan/vulkan-shaders/ssm_conv.comp

@@ -0,0 +1,44 @@
+#version 450
+
+#extension GL_EXT_control_flow_attributes : require
+
+#include "types.glsl"
+
+layout(constant_id = 0) const uint BLOCK_SIZE = 32;
+
+layout(local_size_x_id = 0, local_size_y = 1, local_size_z = 1) in;
+
+layout(binding = 0) readonly buffer Src0 { float src0[]; };
+layout(binding = 1) readonly buffer Src1 { float src1[]; };
+layout(binding = 2) buffer Dst { float dst[]; };
+
+layout(push_constant) uniform PushConstants {
+    uint nb01; uint nb02;
+    uint nb11;
+    uint dst_nb0; uint dst_nb1; uint dst_nb2;
+    uint nc; uint ncs; uint nr; uint n_t; uint n_s;
+};
+
+void main() {
+    const uint global_thread_id = gl_GlobalInvocationID.x;
+    const uint i2 = gl_WorkGroupID.y;
+    const uint i3 = gl_WorkGroupID.z;
+
+    if (global_thread_id >= nr || i2 >= n_t || i3 >= n_s) {
+        return;
+    }
+
+    const uint i1 = global_thread_id;
+    const uint src0_base = i3 * (nb02 / 4) + i2 + i1 * (nb01 / 4);
+    const uint src1_base = i1 * (nb11 / 4);
+    const uint dst_idx = i3 * (dst_nb2 / 4) + i2 * (dst_nb1 / 4) + i1;
+
+    float sum = 0.0;
+    [[unroll]] for (uint i0 = 0; i0 < nc; i0++) {
+        const uint src0_idx = src0_base + i0;
+        const uint src1_idx = src1_base + i0;
+        sum += src0[src0_idx] * src1[src1_idx];
+    }
+
+    dst[dst_idx] = sum;
+}