whispercpp 1.3.4 → 1.3.6

data/ext/sources/ggml/src/ggml-opencl/kernels/gemv_moe_mxfp4_f32.cl

@@ -0,0 +1,156 @@
+#pragma OPENCL EXTENSION cl_khr_fp16 : enable
+#pragma OPENCL EXTENSION cl_khr_subgroups : enable
+#pragma OPENCL EXTENSION cl_qcom_reqd_sub_group_size : enable
+
+#define QK_MXFP4 32
+#define N_SIMDGROUP 4
+#define SIMDGROUP_WIDTH 64
+
+static inline half8 mxfp4_to_fp16_packed8(ushort2 fp4x8) { //, ushort 0x0E00, ushort 0x8000) {
+    ushort2 fp16_packed_a_0, fp16_packed_b_0, bias_a, bias_b, sign_a, sign_b;
+    fp16_packed_a_0.lo = (fp4x8.s0 << 9) & 0x0E00;
+    fp16_packed_a_0.hi = (fp4x8.s0 << 5) & 0x0E00;
+    fp16_packed_b_0.lo = (fp4x8.s0 << 1) & 0x0E00;
+    fp16_packed_b_0.hi = (fp4x8.s0 >> 3) & 0x0E00;
+
+    bias_a.lo = (fp16_packed_a_0.lo != 0) ? 0x3800 : 0x0;
+    bias_a.hi = (fp16_packed_a_0.hi != 0) ? 0x3800 : 0x0;
+    bias_b.lo = (fp16_packed_b_0.lo != 0) ? 0x3800 : 0x0;
+    bias_b.hi = (fp16_packed_b_0.hi != 0) ? 0x3800 : 0x0;
+
+    fp16_packed_a_0.lo = (fp16_packed_a_0.lo != 0x0200) ? fp16_packed_a_0.lo : 0x0;
+    fp16_packed_a_0.hi = (fp16_packed_a_0.hi != 0x0200) ? fp16_packed_a_0.hi : 0x0;
+    fp16_packed_b_0.lo = (fp16_packed_b_0.lo != 0x0200) ? fp16_packed_b_0.lo : 0x0;
+    fp16_packed_b_0.hi = (fp16_packed_b_0.hi != 0x0200) ? fp16_packed_b_0.hi : 0x0;
+
+    sign_a.lo = (fp4x8.s0 << 12) & 0x8000;
+    sign_a.hi = (fp4x8.s0 << 8) & 0x8000;
+    sign_b.lo = (fp4x8.s0 << 4) & 0x8000;
+    sign_b.hi = fp4x8.s0 & 0x8000;
+
+    fp16_packed_a_0 = sign_a + bias_a + fp16_packed_a_0;
+    fp16_packed_b_0 = sign_b + bias_b + fp16_packed_b_0;
+
+    ushort2 fp16_packed_a_1, fp16_packed_b_1;
+    fp16_packed_a_1.lo = (fp4x8.s1 << 9) & 0x0E00;
+    fp16_packed_a_1.hi = (fp4x8.s1 << 5) & 0x0E00;
+    fp16_packed_b_1.lo = (fp4x8.s1 << 1) & 0x0E00;
+    fp16_packed_b_1.hi = (fp4x8.s1 >> 3) & 0x0E00;
+
+    bias_a.lo = (fp16_packed_a_1.lo != 0) ? 0x3800 : 0x0;
+    bias_a.hi = (fp16_packed_a_1.hi != 0) ? 0x3800 : 0x0;
+    bias_b.lo = (fp16_packed_b_1.lo != 0) ? 0x3800 : 0x0;
+    bias_b.hi = (fp16_packed_b_1.hi != 0) ? 0x3800 : 0x0;
+
+    fp16_packed_a_1.lo = (fp16_packed_a_1.lo != 0x0200) ? fp16_packed_a_1.lo : 0x0;
+    fp16_packed_a_1.hi = (fp16_packed_a_1.hi != 0x0200) ? fp16_packed_a_1.hi : 0x0;
+    fp16_packed_b_1.lo = (fp16_packed_b_1.lo != 0x0200) ? fp16_packed_b_1.lo : 0x0;
+    fp16_packed_b_1.hi = (fp16_packed_b_1.hi != 0x0200) ? fp16_packed_b_1.hi : 0x0;
+
+    sign_a.lo = (fp4x8.s1 << 12) & 0x8000;
+    sign_a.hi = (fp4x8.s1 << 8) & 0x8000;
+    sign_b.lo = (fp4x8.s1 << 4) & 0x8000;
+    sign_b.hi = fp4x8.s1 & 0x8000;
+
+    fp16_packed_a_1 = sign_a + bias_a + fp16_packed_a_1;
+    fp16_packed_b_1 = sign_b + bias_b + fp16_packed_b_1;
+
+    return as_half8((ushort8)(fp16_packed_a_0, fp16_packed_b_0, fp16_packed_a_1, fp16_packed_b_1));
+}
+
+static inline float e8m0_to_fp32(uchar x) {
+    int bits;
+    bits = (x == 0) ? 0x00400000 : ((uint) x << 23);
+    return as_float(bits);
+}
+
+
+__attribute__((qcom_reqd_sub_group_size("half")))
+__kernel void kernel_gemv_moe_mxfp4_f32(
+    __global uint4 * src0_q,
+    __global uchar * src0_e,
+    __read_only image1d_buffer_t src1,
+    __global uint * src2,
+    __global float * dst,
+    ulong         offsetd,
+    int           ne00,
+    int           ne01,
+    int           ne11
+) {
+    uint i01  = get_global_id(0);
+    uint i20  = get_global_id(2);
+    uint sgid = get_local_id(1);
+    uint slid = get_sub_group_local_id();
+
+    uint i11 = i20 % ne11;
+
+    uint expert_id = src2[i20];
+    uint expert_offset = expert_id * ne00 * ne01 / 32;
+
+    __private float sum = 0.0f; // each thread calculate partial sum of one output
+
+    // loop along ne00 in block granularity, skip 4 blocks every iter
+    for (uint ib00 = sgid; ib00 < (ne00 / QK_MXFP4); ib00 += N_SIMDGROUP) {
+
+        // load one block of q
+        uint4 regQ = src0_q[expert_offset + ib00 * ne01 + i01];
+
+        uint offset = i11 * ne00 / 4 + ib00 * 8;
+
+        half8 fp16x8 = mxfp4_to_fp16_packed8(as_ushort2(regQ.s0));
+
+        float4 shared_y4;
+        shared_y4 = read_imagef(src1, (offset + 0));
+        float4 acc = shared_y4 * (float4)(fp16x8.s0, fp16x8.s2, fp16x8.s4, fp16x8.s6);
+
+        shared_y4 = read_imagef(src1, (offset + 4));
+        acc += shared_y4 * (float4)(fp16x8.s1, fp16x8.s3, fp16x8.s5, fp16x8.s7);
+
+
+        fp16x8 = mxfp4_to_fp16_packed8(as_ushort2(regQ.s1));
+
+        shared_y4 = read_imagef(src1, (offset + 1));
+        acc += shared_y4 * (float4)(fp16x8.s0, fp16x8.s2, fp16x8.s4, fp16x8.s6);
+
+        shared_y4 = read_imagef(src1, (offset + 5));
+        acc += shared_y4 * (float4)(fp16x8.s1, fp16x8.s3, fp16x8.s5, fp16x8.s7);
+
+
+        fp16x8 = mxfp4_to_fp16_packed8(as_ushort2(regQ.s2));
+
+        shared_y4 = read_imagef(src1, (offset + 2));
+        acc += shared_y4 * (float4)(fp16x8.s0, fp16x8.s2, fp16x8.s4, fp16x8.s6);
+
+        shared_y4 = read_imagef(src1, (offset + 6));
+        acc += shared_y4 * (float4)(fp16x8.s1, fp16x8.s3, fp16x8.s5, fp16x8.s7);
+
+
+        fp16x8 = mxfp4_to_fp16_packed8(as_ushort2(regQ.s3));
+
+        shared_y4 = read_imagef(src1, (offset + 3));
+        acc += shared_y4 * (float4)(fp16x8.s0, fp16x8.s2, fp16x8.s4, fp16x8.s6);
+
+        shared_y4 = read_imagef(src1, (offset + 7));
+        acc += shared_y4 * (float4)(fp16x8.s1, fp16x8.s3, fp16x8.s5, fp16x8.s7);
+
+        uchar regE = src0_e[ib00 * ne01 + i01 + expert_offset];
+        sum += e8m0_to_fp32(regE) * ((acc.s0 + acc.s1) + (acc.s2 + acc.s3));
+    }
+
+    // reduction in local memory, assumes #subgroups=4
+    __local float reduceLM[SIMDGROUP_WIDTH * (N_SIMDGROUP - 1)];
+    if (sgid == 1) reduceLM[SIMDGROUP_WIDTH * 0 + slid] = sum;
+    if (sgid == 2) reduceLM[SIMDGROUP_WIDTH * 1 + slid] = sum;
+    if (sgid == 3) reduceLM[SIMDGROUP_WIDTH * 2 + slid] = sum;
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (sgid == 0) sum += reduceLM[SIMDGROUP_WIDTH * 0 + slid];
+    if (sgid == 0) sum += reduceLM[SIMDGROUP_WIDTH * 1 + slid];
+    if (sgid == 0) sum += reduceLM[SIMDGROUP_WIDTH * 2 + slid];
+
+    // 1 outputs per thread in subgroup 0
+    if (sgid == 0) {
+        dst = dst + (offsetd >> 2);
+        dst[i01 + i20 * ne01] = sum;
+    }
+
+}

data/ext/sources/ggml/src/ggml-opencl/kernels/gemv_noshuffle_general_q8_0_f32.cl

@@ -0,0 +1,195 @@
+#pragma OPENCL EXTENSION cl_khr_fp16 : enable
+#pragma OPENCL EXTENSION cl_khr_subgroups : enable
+
+#ifdef cl_qcom_reqd_sub_group_size
+#pragma OPENCL EXTENSION cl_qcom_reqd_sub_group_size : enable
+#define ADRENO_GPU 1
+#define REQD_SUBGROUP_SIZE_64 __attribute__((qcom_reqd_sub_group_size("half")))
+#endif
+
+#define QK8_0 32
+#define N_SIMDGROUP 4
+
+#define dequantizeBlockAccum_ns_sgbroadcast_1(total_sums, bits8, scale, y) \
+    float shared_y; \
+    char elem; \
+                                             \
+    shared_y = sub_group_broadcast(y.s0, 0); \
+    elem = (char)(bits8.s0 & 0x000000FF); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+    shared_y = sub_group_broadcast(y.s1, 0); \
+    elem = (char)((bits8.s0 & 0x0000FF00) >> 8); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+    shared_y = sub_group_broadcast(y.s2, 0); \
+    elem = (char)((bits8.s0 & 0x00FF0000) >> 16); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+    shared_y = sub_group_broadcast(y.s3, 0); \
+    elem = (char)((bits8.s0 & 0xFF000000) >> 24); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+                                             \
+    shared_y = sub_group_broadcast(y.s4, 0); \
+    elem = (char)(bits8.s1 & 0x000000FF); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+    shared_y = sub_group_broadcast(y.s5, 0); \
+    elem = (char)((bits8.s1 & 0x0000FF00) >> 8); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+    shared_y = sub_group_broadcast(y.s6, 0); \
+    elem = (char)((bits8.s1 & 0x00FF0000) >> 16); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+    shared_y = sub_group_broadcast(y.s7, 0); \
+    elem = (char)((bits8.s1 & 0xFF000000) >> 24); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+                                             \
+    shared_y = sub_group_broadcast(y.s0, 1); \
+    elem = (char)(bits8.s2 & 0x000000FF); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+    shared_y = sub_group_broadcast(y.s1, 1); \
+    elem = (char)((bits8.s2 & 0x0000FF00) >> 8); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+    shared_y = sub_group_broadcast(y.s2, 1); \
+    elem = (char)((bits8.s2 & 0x00FF0000) >> 16); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+    shared_y = sub_group_broadcast(y.s3, 1); \
+    elem = (char)((bits8.s2 & 0xFF000000) >> 24); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+                                             \
+    shared_y = sub_group_broadcast(y.s4, 1); \
+    elem = (char)(bits8.s3 & 0x000000FF); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+    shared_y = sub_group_broadcast(y.s5, 1); \
+    elem = (char)((bits8.s3 & 0x0000FF00) >> 8); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+    shared_y = sub_group_broadcast(y.s6, 1); \
+    elem = (char)((bits8.s3 & 0x00FF0000) >> 16); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+    shared_y = sub_group_broadcast(y.s7, 1); \
+    elem = (char)((bits8.s3 & 0xFF000000) >> 24); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+                                             \
+    shared_y = sub_group_broadcast(y.s0, 2); \
+    elem = (char)(bits8.s4 & 0x000000FF); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+    shared_y = sub_group_broadcast(y.s1, 2); \
+    elem = (char)((bits8.s4 & 0x0000FF00) >> 8); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+    shared_y = sub_group_broadcast(y.s2, 2); \
+    elem = (char)((bits8.s4 & 0x00FF0000) >> 16); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+    shared_y = sub_group_broadcast(y.s3, 2); \
+    elem = (char)((bits8.s4 & 0xFF000000) >> 24); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+                                             \
+    shared_y = sub_group_broadcast(y.s4, 2); \
+    elem = (char)(bits8.s5 & 0x000000FF); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+    shared_y = sub_group_broadcast(y.s5, 2); \
+    elem = (char)((bits8.s5 & 0x0000FF00) >> 8); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+    shared_y = sub_group_broadcast(y.s6, 2); \
+    elem = (char)((bits8.s5 & 0x00FF0000) >> 16); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+    shared_y = sub_group_broadcast(y.s7, 2); \
+    elem = (char)((bits8.s5 & 0xFF000000) >> 24); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+                                             \
+    shared_y = sub_group_broadcast(y.s0, 3); \
+    elem = (char)(bits8.s6 & 0x000000FF); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+    shared_y = sub_group_broadcast(y.s1, 3); \
+    elem = (char)((bits8.s6 & 0x0000FF00) >> 8); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+    shared_y = sub_group_broadcast(y.s2, 3); \
+    elem = (char)((bits8.s6 & 0x00FF0000) >> 16); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+    shared_y = sub_group_broadcast(y.s3, 3); \
+    elem = (char)((bits8.s6 & 0xFF000000) >> 24); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+                                             \
+    shared_y = sub_group_broadcast(y.s4, 3); \
+    elem = (char)(bits8.s7 & 0x000000FF); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+    shared_y = sub_group_broadcast(y.s5, 3); \
+    elem = (char)((bits8.s7 & 0x0000FF00) >> 8); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+    shared_y = sub_group_broadcast(y.s6, 3); \
+    elem = (char)((bits8.s7 & 0x00FF0000) >> 16); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+    shared_y = sub_group_broadcast(y.s7, 3); \
+    elem = (char)((bits8.s7 & 0xFF000000) >> 24); \
+    total_sums += convert_int(elem) * scale * shared_y; \
+
+#ifdef ADRENO_GPU
+REQD_SUBGROUP_SIZE_64
+#endif
+__kernel void kernel_gemv_noshuffle_q8_0_f32(
+        __read_only  image1d_buffer_t src0_q,  // quantized A
+        global half  * src0_d,  // A scales
+        __read_only  image1d_buffer_t src1,    // B
+        ulong offset1,            // offset to B (0)
+        global float * dst,     // C
+        ulong offsetd,            // offset to C
+        int ne00,               // K
+        int ne01,               // M
+        int ne02,               // 1
+        int ne10,               // K
+        int ne12,               // 1
+        int ne0,                // M
+        int ne1,                // N
+        int r2,                 // 1
+        int r3)
+{
+    uint groupId = get_local_id(1);
+    uint gid     = get_global_id(0);
+    ushort slid    = get_sub_group_local_id();
+
+    uint K = ne00;
+    uint M = ne01;
+
+    uint LINE_STRIDE_A = M;
+    uint BLOCK_STRIDE_A = 8 * M;   // 32 / 4 = 8
+
+    __private uint8     regA;
+    __private half      regS;
+    __private float8    regB;
+
+    __private float totalSum = (float)(0.0f);
+
+    // loop along K in block granularity, skip 4 blocks every iter
+    #pragma unroll 1 /* tell compiler not to unroll */
+    for (uint k = groupId; k < (K / QK8_0); k += N_SIMDGROUP) {
+        regS = src0_d[gid + k * LINE_STRIDE_A]; // each fiber loads scale of one rows
+        // first 4 fibers in each wave load 8 B values to its private scope
+        if (slid < 4) {
+            regB.s0123 = read_imagef(src1, (slid * 2 + k * 8));
+            regB.s4567 = read_imagef(src1, (1 + slid * 2 + k * 8));
+        }
+
+        // load weights for one block in consecutive rows
+        regA.s0 = read_imageui(src0_q, (gid + k * BLOCK_STRIDE_A + LINE_STRIDE_A * 0)).x;
+        regA.s1 = read_imageui(src0_q, (gid + k * BLOCK_STRIDE_A + LINE_STRIDE_A * 1)).x;
+        regA.s2 = read_imageui(src0_q, (gid + k * BLOCK_STRIDE_A + LINE_STRIDE_A * 2)).x;
+        regA.s3 = read_imageui(src0_q, (gid + k * BLOCK_STRIDE_A + LINE_STRIDE_A * 3)).x;
+        regA.s4 = read_imageui(src0_q, (gid + k * BLOCK_STRIDE_A + LINE_STRIDE_A * 4)).x;
+        regA.s5 = read_imageui(src0_q, (gid + k * BLOCK_STRIDE_A + LINE_STRIDE_A * 5)).x;
+        regA.s6 = read_imageui(src0_q, (gid + k * BLOCK_STRIDE_A + LINE_STRIDE_A * 6)).x;
+        regA.s7 = read_imageui(src0_q, (gid + k * BLOCK_STRIDE_A + LINE_STRIDE_A * 7)).x;
+
+        dequantizeBlockAccum_ns_sgbroadcast_1(totalSum, regA, regS, regB);
+    }
+
+    // reduction in local memory, assumes #wave=4
+    __local float reduceLM[SIMDGROUP_WIDTH * 3];
+    if (groupId == 1) reduceLM[SIMDGROUP_WIDTH * 0 + slid] = totalSum;
+    if (groupId == 2) reduceLM[SIMDGROUP_WIDTH * 1 + slid] = totalSum;
+    if (groupId == 3) reduceLM[SIMDGROUP_WIDTH * 2 + slid] = totalSum;
+    barrier(CLK_LOCAL_MEM_FENCE);
+    if (groupId == 0) totalSum += reduceLM[SIMDGROUP_WIDTH * 0 + slid];
+    if (groupId == 0) totalSum += reduceLM[SIMDGROUP_WIDTH * 1 + slid];
+    if (groupId == 0) totalSum += reduceLM[SIMDGROUP_WIDTH * 2 + slid];
+
+    // 1 outputs per fiber in wave 0
+    if (groupId == 0) {
+        dst = (global float*)((global char*)dst + offsetd);
+        dst[gid] = totalSum;
+    }
+}

data/ext/sources/ggml/src/ggml-opencl/kernels/gemv_noshuffle_q4_1_f32.cl

@@ -0,0 +1,283 @@
+#pragma OPENCL EXTENSION cl_khr_fp16 : enable
+#pragma OPENCL EXTENSION cl_khr_subgroups : enable
+
+#ifdef cl_qcom_reqd_sub_group_size
+#pragma OPENCL EXTENSION cl_qcom_reqd_sub_group_size : enable
+#define ADRENO_GPU 1
+#define REQD_SUBGROUP_SIZE_64 __attribute__((qcom_reqd_sub_group_size("half")))
+#endif
+
+#define QK4_0 32
+#define NSUBGROUPS 4
+#define SUBGROUP_SIZE 64
+
+#define dequantizeBlockAccum_ns_sgbroadcast_1_hi(total_sums, bits4, scale, minv, y) \
+    float shared_y; \
+    shared_y = sub_group_broadcast(y.s0, 0); \
+    total_sums.s0 += ((bits4.s0 & 0x000F) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += ((bits4.s1 & 0x000F) * scale.s1 + minv.s1) * shared_y; \
+    shared_y = sub_group_broadcast(y.s1, 0); \
+    total_sums.s0 += (((bits4.s0 & 0x00F0) >> 4) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += (((bits4.s1 & 0x00F0) >> 4) * scale.s1 + minv.s1) * shared_y; \
+    shared_y = sub_group_broadcast(y.s2, 0); \
+    total_sums.s0 += (((bits4.s0 & 0x0F00) >> 8) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += (((bits4.s1 & 0x0F00) >> 8) * scale.s1 + minv.s1) * shared_y; \
+    shared_y = sub_group_broadcast(y.s3, 0); \
+    total_sums.s0 += (((bits4.s0 & 0xF000) >> 12) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += (((bits4.s1 & 0xF000) >> 12) * scale.s1 + minv.s1) * shared_y; \
+    shared_y = sub_group_broadcast(y.s4, 0); \
+    total_sums.s0 += ((bits4.s2 & 0x000F) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += ((bits4.s3 & 0x000F) * scale.s1 + minv.s1) * shared_y; \
+    shared_y = sub_group_broadcast(y.s5, 0); \
+    total_sums.s0 += (((bits4.s2 & 0x00F0) >> 4) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += (((bits4.s3 & 0x00F0) >> 4) * scale.s1 + minv.s1) * shared_y; \
+    shared_y = sub_group_broadcast(y.s6, 0); \
+    total_sums.s0 += (((bits4.s2 & 0x0F00) >> 8) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += (((bits4.s3 & 0x0F00) >> 8) * scale.s1 + minv.s1) * shared_y; \
+    shared_y = sub_group_broadcast(y.s7, 0); \
+    total_sums.s0 += (((bits4.s2 & 0xF000) >> 12) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += (((bits4.s3 & 0xF000) >> 12) * scale.s1 + minv.s1) * shared_y; \
+    shared_y = sub_group_broadcast(y.s0, 1); \
+    total_sums.s0 += ((bits4.s4 & 0x000F) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += ((bits4.s5 & 0x000F) * scale.s1 + minv.s1) * shared_y; \
+    shared_y = sub_group_broadcast(y.s1, 1); \
+    total_sums.s0 += (((bits4.s4 & 0x00F0) >> 4) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += (((bits4.s5 & 0x00F0) >> 4) * scale.s1 + minv.s1) * shared_y; \
+    shared_y = sub_group_broadcast(y.s2, 1); \
+    total_sums.s0 += (((bits4.s4 & 0x0F00) >> 8) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += (((bits4.s5 & 0x0F00) >> 8) * scale.s1 + minv.s1) * shared_y; \
+    shared_y = sub_group_broadcast(y.s3, 1); \
+    total_sums.s0 += (((bits4.s4 & 0xF000) >> 12) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += (((bits4.s5 & 0xF000) >> 12) * scale.s1 + minv.s1) * shared_y; \
+    shared_y = sub_group_broadcast(y.s4, 1); \
+    total_sums.s0 += ((bits4.s6 & 0x000F) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += ((bits4.s7 & 0x000F) * scale.s1 + minv.s1) * shared_y; \
+    shared_y = sub_group_broadcast(y.s5, 1); \
+    total_sums.s0 += (((bits4.s6 & 0x00F0) >> 4) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += (((bits4.s7 & 0x00F0) >> 4) * scale.s1 + minv.s1) * shared_y; \
+    shared_y = sub_group_broadcast(y.s6, 1); \
+    total_sums.s0 += (((bits4.s6 & 0x0F00) >> 8) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += (((bits4.s7 & 0x0F00) >> 8) * scale.s1 + minv.s1) * shared_y; \
+    shared_y = sub_group_broadcast(y.s7, 1); \
+    total_sums.s0 += (((bits4.s6 & 0xF000) >> 12) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += (((bits4.s7 & 0xF000) >> 12) * scale.s1 + minv.s1) * shared_y; \
+
+
+#define dequantizeBlockAccum_ns_sgbroadcast_1_lo(total_sums, bits4, scale, minv, y) \
+    shared_y = sub_group_broadcast(y.s0, 2); \
+    total_sums.s0 += ((bits4.s0 & 0x000F) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += ((bits4.s1 & 0x000F) * scale.s1 + minv.s1) * shared_y; \
+    shared_y = sub_group_broadcast(y.s1, 2); \
+    total_sums.s0 += (((bits4.s0 & 0x00F0) >> 4) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += (((bits4.s1 & 0x00F0) >> 4) * scale.s1 + minv.s1) * shared_y; \
+    shared_y = sub_group_broadcast(y.s2, 2); \
+    total_sums.s0 += (((bits4.s0 & 0x0F00) >> 8) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += (((bits4.s1 & 0x0F00) >> 8) * scale.s1 + minv.s1) * shared_y; \
+    shared_y = sub_group_broadcast(y.s3, 2); \
+    total_sums.s0 += (((bits4.s0 & 0xF000) >> 12) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += (((bits4.s1 & 0xF000) >> 12) * scale.s1 + minv.s1) * shared_y; \
+    shared_y = sub_group_broadcast(y.s4, 2); \
+    total_sums.s0 += ((bits4.s2 & 0x000F) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += ((bits4.s3 & 0x000F) * scale.s1 + minv.s1) * shared_y; \
+    shared_y = sub_group_broadcast(y.s5, 2); \
+    total_sums.s0 += (((bits4.s2 & 0x00F0) >> 4) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += (((bits4.s3 & 0x00F0) >> 4) * scale.s1 + minv.s1) * shared_y; \
+    shared_y = sub_group_broadcast(y.s6, 2); \
+    total_sums.s0 += (((bits4.s2 & 0x0F00) >> 8) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += (((bits4.s3 & 0x0F00) >> 8) * scale.s1 + minv.s1) * shared_y; \
+    shared_y = sub_group_broadcast(y.s7, 2); \
+    total_sums.s0 += (((bits4.s2 & 0xF000) >> 12) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += (((bits4.s3 & 0xF000) >> 12) * scale.s1 + minv.s1) * shared_y; \
+    shared_y = sub_group_broadcast(y.s0, 3); \
+    total_sums.s0 += ((bits4.s4 & 0x000F) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += ((bits4.s5 & 0x000F) * scale.s1 + minv.s1) * shared_y; \
+    shared_y = sub_group_broadcast(y.s1, 3); \
+    total_sums.s0 += (((bits4.s4 & 0x00F0) >> 4) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += (((bits4.s5 & 0x00F0) >> 4) * scale.s1 + minv.s1) * shared_y; \
+    shared_y = sub_group_broadcast(y.s2, 3); \
+    total_sums.s0 += (((bits4.s4 & 0x0F00) >> 8) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += (((bits4.s5 & 0x0F00) >> 8) * scale.s1 + minv.s1) * shared_y; \
+    shared_y = sub_group_broadcast(y.s3, 3); \
+    total_sums.s0 += (((bits4.s4 & 0xF000) >> 12) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += (((bits4.s5 & 0xF000) >> 12) * scale.s1 + minv.s1) * shared_y; \
+    shared_y = sub_group_broadcast(y.s4, 3); \
+    total_sums.s0 += ((bits4.s6 & 0x000F) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += ((bits4.s7 & 0x000F) * scale.s1 + minv.s1) * shared_y; \
+    shared_y = sub_group_broadcast(y.s5, 3); \
+    total_sums.s0 += (((bits4.s6 & 0x00F0) >> 4) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += (((bits4.s7 & 0x00F0) >> 4) * scale.s1 + minv.s1) * shared_y; \
+    shared_y = sub_group_broadcast(y.s6, 3); \
+    total_sums.s0 += (((bits4.s6 & 0x0F00) >> 8) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += (((bits4.s7 & 0x0F00) >> 8) * scale.s1 + minv.s1) * shared_y; \
+    shared_y = sub_group_broadcast(y.s7, 3); \
+    total_sums.s0 += (((bits4.s6 & 0xF000) >> 12) * scale.s0 + minv.s0) * shared_y; \
+    total_sums.s1 += (((bits4.s7 & 0xF000) >> 12) * scale.s1 + minv.s1) * shared_y; \
+
+
+#define dequantizeBlockAccum_ns_sgbroadcast_8_hi(total_sums, bits4, scale, minv, y) \
+    float8 shared_y; \
+    shared_y = sub_group_broadcast(y, 0); \
+    total_sums.s0 += ((bits4.s0 & 0x000F)         * scale.s0 + minv.s0) * shared_y.s0; \
+    total_sums.s0 += (((bits4.s0 & 0x00F0) >> 4)  * scale.s0 + minv.s0) * shared_y.s1; \
+    total_sums.s0 += (((bits4.s0 & 0x0F00) >> 8)  * scale.s0 + minv.s0) * shared_y.s2; \
+    total_sums.s0 += (((bits4.s0 & 0xF000) >> 12) * scale.s0 + minv.s0) * shared_y.s3; \
+    total_sums.s0 += ((bits4.s2 & 0x000F)         * scale.s0 + minv.s0) * shared_y.s4; \
+    total_sums.s0 += (((bits4.s2 & 0x00F0) >> 4)  * scale.s0 + minv.s0) * shared_y.s5; \
+    total_sums.s0 += (((bits4.s2 & 0x0F00) >> 8)  * scale.s0 + minv.s0) * shared_y.s6; \
+    total_sums.s0 += (((bits4.s2 & 0xF000) >> 12) * scale.s0 + minv.s0) * shared_y.s7; \
+    total_sums.s1 += ((bits4.s1 & 0x000F)         * scale.s1 + minv.s1) * shared_y.s0; \
+    total_sums.s1 += (((bits4.s1 & 0x00F0) >> 4)  * scale.s1 + minv.s1) * shared_y.s1; \
+    total_sums.s1 += (((bits4.s1 & 0x0F00) >> 8)  * scale.s1 + minv.s1) * shared_y.s2; \
+    total_sums.s1 += (((bits4.s1 & 0xF000) >> 12) * scale.s1 + minv.s1) * shared_y.s3; \
+    total_sums.s1 += ((bits4.s3 & 0x000F)         * scale.s1 + minv.s1) * shared_y.s4; \
+    total_sums.s1 += (((bits4.s3 & 0x00F0) >> 4)  * scale.s1 + minv.s1) * shared_y.s5; \
+    total_sums.s1 += (((bits4.s3 & 0x0F00) >> 8)  * scale.s1 + minv.s1) * shared_y.s6; \
+    total_sums.s1 += (((bits4.s3 & 0xF000) >> 12) * scale.s1 + minv.s1) * shared_y.s7; \
+    shared_y = sub_group_broadcast(y, 1); \
+    total_sums.s0 += ((bits4.s4 & 0x000F)         * scale.s0 + minv.s0) * shared_y.s0; \
+    total_sums.s0 += (((bits4.s4 & 0x00F0) >> 4)  * scale.s0 + minv.s0) * shared_y.s1; \
+    total_sums.s0 += (((bits4.s4 & 0x0F00) >> 8)  * scale.s0 + minv.s0) * shared_y.s2; \
+    total_sums.s0 += (((bits4.s4 & 0xF000) >> 12) * scale.s0 + minv.s0) * shared_y.s3; \
+    total_sums.s0 += ((bits4.s6 & 0x000F)         * scale.s0 + minv.s0) * shared_y.s4; \
+    total_sums.s0 += (((bits4.s6 & 0x00F0) >> 4)  * scale.s0 + minv.s0) * shared_y.s5; \
+    total_sums.s0 += (((bits4.s6 & 0x0F00) >> 8)  * scale.s0 + minv.s0) * shared_y.s6; \
+    total_sums.s0 += (((bits4.s6 & 0xF000) >> 12) * scale.s0 + minv.s0) * shared_y.s7; \
+    total_sums.s1 += ((bits4.s5 & 0x000F)         * scale.s1 + minv.s1) * shared_y.s0; \
+    total_sums.s1 += (((bits4.s5 & 0x00F0) >> 4)  * scale.s1 + minv.s1) * shared_y.s1; \
+    total_sums.s1 += (((bits4.s5 & 0x0F00) >> 8)  * scale.s1 + minv.s1) * shared_y.s2; \
+    total_sums.s1 += (((bits4.s5 & 0xF000) >> 12) * scale.s1 + minv.s1) * shared_y.s3; \
+    total_sums.s1 += ((bits4.s7 & 0x000F)         * scale.s1 + minv.s1) * shared_y.s4; \
+    total_sums.s1 += (((bits4.s7 & 0x00F0) >> 4)  * scale.s1 + minv.s1) * shared_y.s5; \
+    total_sums.s1 += (((bits4.s7 & 0x0F00) >> 8)  * scale.s1 + minv.s1) * shared_y.s6; \
+    total_sums.s1 += (((bits4.s7 & 0xF000) >> 12) * scale.s1 + minv.s1) * shared_y.s7; \
+
+
+#define dequantizeBlockAccum_ns_sgbroadcast_8_lo(total_sums, bits4, scale, minv, y) \
+    shared_y = sub_group_broadcast(y, 2); \
+    total_sums.s0 += ((bits4.s0 & 0x000F)         * scale.s0 + minv.s0) * shared_y.s0; \
+    total_sums.s0 += (((bits4.s0 & 0x00F0) >> 4)  * scale.s0 + minv.s0) * shared_y.s1; \
+    total_sums.s0 += (((bits4.s0 & 0x0F00) >> 8)  * scale.s0 + minv.s0) * shared_y.s2; \
+    total_sums.s0 += (((bits4.s0 & 0xF000) >> 12) * scale.s0 + minv.s0) * shared_y.s3; \
+    total_sums.s0 += ((bits4.s2 & 0x000F)         * scale.s0 + minv.s0) * shared_y.s4; \
+    total_sums.s0 += (((bits4.s2 & 0x00F0) >> 4)  * scale.s0 + minv.s0) * shared_y.s5; \
+    total_sums.s0 += (((bits4.s2 & 0x0F00) >> 8)  * scale.s0 + minv.s0) * shared_y.s6; \
+    total_sums.s0 += (((bits4.s2 & 0xF000) >> 12) * scale.s0 + minv.s0) * shared_y.s7; \
+    total_sums.s1 += ((bits4.s1 & 0x000F)         * scale.s1 + minv.s1) * shared_y.s0; \
+    total_sums.s1 += (((bits4.s1 & 0x00F0) >> 4)  * scale.s1 + minv.s1) * shared_y.s1; \
+    total_sums.s1 += (((bits4.s1 & 0x0F00) >> 8)  * scale.s1 + minv.s1) * shared_y.s2; \
+    total_sums.s1 += (((bits4.s1 & 0xF000) >> 12) * scale.s1 + minv.s1) * shared_y.s3; \
+    total_sums.s1 += ((bits4.s3 & 0x000F)         * scale.s1 + minv.s1) * shared_y.s4; \
+    total_sums.s1 += (((bits4.s3 & 0x00F0) >> 4)  * scale.s1 + minv.s1) * shared_y.s5; \
+    total_sums.s1 += (((bits4.s3 & 0x0F00) >> 8)  * scale.s1 + minv.s1) * shared_y.s6; \
+    total_sums.s1 += (((bits4.s3 & 0xF000) >> 12) * scale.s1 + minv.s1) * shared_y.s7; \
+    shared_y = sub_group_broadcast(y, 3); \
+    total_sums.s0 += ((bits4.s4 & 0x000F)         * scale.s0 + minv.s0) * shared_y.s0; \
+    total_sums.s0 += (((bits4.s4 & 0x00F0) >> 4)  * scale.s0 + minv.s0) * shared_y.s1; \
+    total_sums.s0 += (((bits4.s4 & 0x0F00) >> 8)  * scale.s0 + minv.s0) * shared_y.s2; \
+    total_sums.s0 += (((bits4.s4 & 0xF000) >> 12) * scale.s0 + minv.s0) * shared_y.s3; \
+    total_sums.s0 += ((bits4.s6 & 0x000F)         * scale.s0 + minv.s0) * shared_y.s4; \
+    total_sums.s0 += (((bits4.s6 & 0x00F0) >> 4)  * scale.s0 + minv.s0) * shared_y.s5; \
+    total_sums.s0 += (((bits4.s6 & 0x0F00) >> 8)  * scale.s0 + minv.s0) * shared_y.s6; \
+    total_sums.s0 += (((bits4.s6 & 0xF000) >> 12) * scale.s0 + minv.s0) * shared_y.s7; \
+    total_sums.s1 += ((bits4.s5 & 0x000F)         * scale.s1 + minv.s1) * shared_y.s0; \
+    total_sums.s1 += (((bits4.s5 & 0x00F0) >> 4)  * scale.s1 + minv.s1) * shared_y.s1; \
+    total_sums.s1 += (((bits4.s5 & 0x0F00) >> 8)  * scale.s1 + minv.s1) * shared_y.s2; \
+    total_sums.s1 += (((bits4.s5 & 0xF000) >> 12) * scale.s1 + minv.s1) * shared_y.s3; \
+    total_sums.s1 += ((bits4.s7 & 0x000F)         * scale.s1 + minv.s1) * shared_y.s4; \
+    total_sums.s1 += (((bits4.s7 & 0x00F0) >> 4)  * scale.s1 + minv.s1) * shared_y.s5; \
+    total_sums.s1 += (((bits4.s7 & 0x0F00) >> 8)  * scale.s1 + minv.s1) * shared_y.s6; \
+    total_sums.s1 += (((bits4.s7 & 0xF000) >> 12) * scale.s1 + minv.s1) * shared_y.s7; \
+
+#ifdef ADRENO_GPU
+REQD_SUBGROUP_SIZE_64
+#endif
+kernel void kernel_gemv_noshuffle_q4_1_f32(
+        read_only  image1d_buffer_t src0_q,
+        global half2  * src0_d,
+        global half2  * src0_m,
+        read_only  image1d_buffer_t src1,
+        global float * dst,
+        ulong offsetd,
+        int ne00,
+        int ne01)
+{
+    uint groupId = get_local_id(1);
+    uint gid     = get_global_id(0);
+    ushort slid    = get_sub_group_local_id();
+
+    uint K = ne00;
+    uint M = ne01;
+
+    uint LINE_STRIDE_A = M / 2;
+    uint BLOCK_STRIDE_A = NSUBGROUPS * M;
+
+    private uint4     regA;
+    private half2     regS;
+    private half2     regM;
+    private float8    regB;
+
+    private float2 totalSum = (float2)(0.0f);
+
+    // loop along K in block granularity, skip 4 blocks every iter
+    for (uint k = groupId; k < (K / QK4_0); k += NSUBGROUPS) {
+        regS = src0_d[gid + k * LINE_STRIDE_A]; // each fiber loads scale of two rows
+        regM = src0_m[gid + k * LINE_STRIDE_A]; // each fiber loads min of two rows
+        // first 4 fibers in each wave load 8 B values to its private scope
+        if (slid < 4) {
+            regB.s0123 = read_imagef(src1, (slid * 2 + k * 8));
+            regB.s4567 = read_imagef(src1, (1 + slid * 2 + k * 8));
+        }
+
+        // load half weights for two blocks in consecutive rows
+        regA.s0 = read_imageui(src0_q, (gid + k * BLOCK_STRIDE_A + LINE_STRIDE_A * 0)).x;
+        regA.s1 = read_imageui(src0_q, (gid + k * BLOCK_STRIDE_A + LINE_STRIDE_A * 1)).x;
+        regA.s2 = read_imageui(src0_q, (gid + k * BLOCK_STRIDE_A + LINE_STRIDE_A * 2)).x;
+        regA.s3 = read_imageui(src0_q, (gid + k * BLOCK_STRIDE_A + LINE_STRIDE_A * 3)).x;
+#ifdef VECTOR_SUB_GROUP_BROADCAT
+        dequantizeBlockAccum_ns_sgbroadcast_8_hi(totalSum, as_ushort8(regA), regS, regM, regB);
+#else
+        dequantizeBlockAccum_ns_sgbroadcast_1_hi(totalSum, as_ushort8(regA), regS, regM, regB);
+#endif // VECTOR_SUB_GROUP_BROADCAT
+
+        regA.s0 = read_imageui(src0_q, (gid + k * BLOCK_STRIDE_A + LINE_STRIDE_A * 4)).x;
+        regA.s1 = read_imageui(src0_q, (gid + k * BLOCK_STRIDE_A + LINE_STRIDE_A * 5)).x;
+        regA.s2 = read_imageui(src0_q, (gid + k * BLOCK_STRIDE_A + LINE_STRIDE_A * 6)).x;
+        regA.s3 = read_imageui(src0_q, (gid + k * BLOCK_STRIDE_A + LINE_STRIDE_A * 7)).x;
+#ifdef VECTOR_SUB_GROUP_BROADCAT
+        dequantizeBlockAccum_ns_sgbroadcast_8_lo(totalSum, as_ushort8(regA), regS, regM, regB);
+#else
+        dequantizeBlockAccum_ns_sgbroadcast_1_lo(totalSum, as_ushort8(regA), regS, regM, regB);
+#endif // VECTOR_SUB_GROUP_BROADCAT
+    }
+
+    // reduction in local memory, assumes #wave=4
+    local float2 reduceLM[SUBGROUP_SIZE * 3];
+    if (groupId == 1) {
+        reduceLM[SUBGROUP_SIZE * 0 + slid] = totalSum;
+    }
+    if (groupId == 2) {
+        reduceLM[SUBGROUP_SIZE * 1 + slid] = totalSum;
+    }
+    if (groupId == 3) {
+        reduceLM[SUBGROUP_SIZE * 2 + slid] = totalSum;
+    }
+
+    barrier(CLK_LOCAL_MEM_FENCE);
+
+    if (groupId == 0) {
+        totalSum += reduceLM[SUBGROUP_SIZE * 0 + slid];
+    }
+    if (groupId == 0) {
+        totalSum += reduceLM[SUBGROUP_SIZE * 1 + slid];
+    }
+    if (groupId == 0) {
+        totalSum += reduceLM[SUBGROUP_SIZE * 2 + slid];
+    }
+
+    // 2 outputs per fiber in wave 0
+    if (groupId == 0) {
+        dst = (global float*)((global char*)dst + offsetd);
+        vstore2(totalSum, 0, &(dst[gid * 2]));
+    }
+
+}