vbatts/llama.cpp
1
0
Fork 0
Code Issues Pull requests Projects Releases Packages Wiki Activity Actions

OpenCL: Pass src0 offset as kernel argument instead of global offset

Browse source
This commit is contained in:
shibe2 2023-10-29 12:18:46 +04:00
parent ff3bad83e2
commit 2d0826247b
1 changed files with 42 additions and 39 deletions
Download patch file Download diff file Expand all files Collapse all files

81
ggml-opencl.cpp
View file

@ -200,9 +200,9 @@ inline void get_scale_min_k4(int j, const __global uint8_t *q, uint8_t *d, uint8
}
}
__kernel void dequantize_block_q2_K(__global const struct block_q2_K *x, __global float *yy)
__kernel void dequantize_block_q2_K(__global const struct block_q2_K *x, uint x_offset, __global float *yy)
{
const int i = get_group_id(0) + get_global_offset(0);
const int i = get_group_id(0) + x_offset;
const int tid = get_local_id(0);
const int n = tid / 32;
const int l = tid - 32 * n;
@ -220,10 +220,10 @@ __kernel void dequantize_block_q2_K(__global const struct block_q2_K *x, __globa
y[l + 96] = dall * (x[i].scales[is + 6] & 0xF) * ((q >> 6) & 3) - dmin * (x[i].scales[is + 6] >> 4);
}
__kernel void dequantize_block_q3_K(__global const struct block_q3_K *x, __global float *yy)
__kernel void dequantize_block_q3_K(__global const struct block_q3_K *x, uint x_offset, __global float *yy)
{
int r = get_local_id(0) / 4;
int i = get_group_id(0) + get_global_offset(0);
int i = get_group_id(0) + x_offset;
int tid = r / 2;
int is0 = r % 2;
int l0 = 16 * is0 + 4 * (get_local_id(0) % 4);
@ -249,9 +249,9 @@ __kernel void dequantize_block_q3_K(__global const struct block_q3_K *x, __globa
y[l] = dl * ((int8_t)((q[l] >> shift) & 3) - ((hm[l] & m) ? 0 : 4));
}
__kernel void dequantize_block_q4_K(__global const struct block_q4_K *x, __global float *yy)
__kernel void dequantize_block_q4_K(__global const struct block_q4_K *x, uint x_offset, __global float *yy)
{
const int i = get_group_id(0) + get_global_offset(0);
const int i = get_group_id(0) + x_offset;
const int tid = get_local_id(0);
const int il = tid / 8;
const int ir = tid % 8;
@ -279,9 +279,9 @@ __kernel void dequantize_block_q4_K(__global const struct block_q4_K *x, __globa
}
}
__kernel void dequantize_block_q5_K(__global const struct block_q5_K *x, __global float *yy)
__kernel void dequantize_block_q5_K(__global const struct block_q5_K *x, uint x_offset, __global float *yy)
{
const int i = get_group_id(0) + get_global_offset(0);
const int i = get_group_id(0) + x_offset;
const int tid = get_local_id(0);
const int il = tid / 16;
const int ir = tid % 16;
@ -311,9 +311,9 @@ __kernel void dequantize_block_q5_K(__global const struct block_q5_K *x, __globa
y[33] = d2 * ((ql[1] >> 4) + (qh[1] & hm ? 16 : 0)) - m2;
}
__kernel void dequantize_block_q6_K(__global const struct block_q6_K *x, __global float *yy)
__kernel void dequantize_block_q6_K(__global const struct block_q6_K *x, uint x_offset, __global float *yy)
{
const int i = get_group_id(0) + get_global_offset(0);
const int i = get_group_id(0) + x_offset;
const int tid = get_local_id(0);
const int ip = tid / 32;
const int il = tid - 32 * ip;
@ -333,14 +333,14 @@ __kernel void dequantize_block_q6_K(__global const struct block_q6_K *x, __globa
y[96] = d * sc[6] * ((int8_t)((ql[32] >> 4) | (((qh >> 6) & 3) << 4)) - 32);
}
__kernel void dequantize_mul_mat_vec_q2_K(__global const struct block_q2_K * xx, __local float* tmp, __global float* yy, __global float* dst, const int ncols) {
__kernel void dequantize_mul_mat_vec_q2_K(__global const struct block_q2_K * xx, uint x_offset, __local float* tmp, __global float* yy, __global float* dst, const int ncols) {
const int row = get_group_id(0);
const int num_blocks_per_row = ncols / QK_K;
const int ib0 = row*num_blocks_per_row + get_global_offset(0);
const int ib0 = row*num_blocks_per_row;
__global const struct block_q2_K * x = xx + ib0;
__global const struct block_q2_K * x = xx + ib0 + x_offset;
const int tid = get_local_id(0)/K_QUANTS_PER_ITERATION; // 0...31 or 0...15
const int ix = get_local_id(0)%K_QUANTS_PER_ITERATION; // 0 or 0,1
@ -406,16 +406,16 @@ __kernel void dequantize_mul_mat_vec_q2_K(__global const struct block_q2_K * xx,
}
}
__kernel void dequantize_mul_mat_vec_q3_K(__global const struct block_q3_K * xx, __local float* tmp, __global float* yy, __global float* dst, const int ncols) {
__kernel void dequantize_mul_mat_vec_q3_K(__global const struct block_q3_K * xx, uint x_offset, __local float* tmp, __global float* yy, __global float* dst, const int ncols) {
const uint16_t kmask1 = 0x0303;
const uint16_t kmask2 = 0x0f0f;
const int row = get_group_id(0);
const int num_blocks_per_row = ncols / QK_K;
const int ib0 = row*num_blocks_per_row + get_global_offset(0);
const int ib0 = row*num_blocks_per_row;
__global const struct block_q3_K * x = xx + ib0;
__global const struct block_q3_K * x = xx + ib0 + x_offset;
const int tid = get_local_id(0)/K_QUANTS_PER_ITERATION; // 0...31 or 0...16
const int ix = get_local_id(0)%K_QUANTS_PER_ITERATION; // 0 or 0,1
@ -480,7 +480,7 @@ __kernel void dequantize_mul_mat_vec_q3_K(__global const struct block_q3_K * xx,
}
}
__kernel void dequantize_mul_mat_vec_q4_K(__global const struct block_q4_K * xx, __local float* tmp, __global float* yy, __global float* dst, const int ncols) {
__kernel void dequantize_mul_mat_vec_q4_K(__global const struct block_q4_K * xx, uint x_offset, __local float* tmp, __global float* yy, __global float* dst, const int ncols) {
//to rename it later, just to test now
const uint16_t kmask1 = 0x3f3f;
@ -489,7 +489,7 @@ __kernel void dequantize_mul_mat_vec_q4_K(__global const struct block_q4_K * xx,
const int row = get_group_id(0);
const int num_blocks_per_row = ncols / QK_K;
const int ib0 = row*num_blocks_per_row + get_global_offset(0);
const int ib0 = row*num_blocks_per_row;
const int tid = get_local_id(0)/K_QUANTS_PER_ITERATION; // 0...15
const int ix = get_local_id(0)%K_QUANTS_PER_ITERATION;
@ -510,7 +510,7 @@ __kernel void dequantize_mul_mat_vec_q4_K(__global const struct block_q4_K * xx,
uint16_t aux[4];
const uint8_t * sc = (const uint8_t *)aux;
__global const struct block_q4_K * x = xx + ib0;
__global const struct block_q4_K * x = xx + ib0 + x_offset;
tmp[16 * ix + tid] = 0;
@ -554,7 +554,7 @@ __kernel void dequantize_mul_mat_vec_q4_K(__global const struct block_q4_K * xx,
}
}
__kernel void dequantize_mul_mat_vec_q5_K(__global const struct block_q5_K * xx, __local float* tmp, __global float* yy, __global float* dst, const int ncols) {
__kernel void dequantize_mul_mat_vec_q5_K(__global const struct block_q5_K * xx, uint x_offset, __local float* tmp, __global float* yy, __global float* dst, const int ncols) {
const uint16_t kmask1 = 0x3f3f;
const uint16_t kmask2 = 0x0f0f;
@ -562,7 +562,7 @@ __kernel void dequantize_mul_mat_vec_q5_K(__global const struct block_q5_K * xx,
const int row = get_group_id(0);
const int num_blocks_per_row = ncols / QK_K;
const int ib0 = row*num_blocks_per_row + get_global_offset(0);
const int ib0 = row*num_blocks_per_row;
const int tid = get_local_id(0)/2; // 0...15
const int ix = get_local_id(0)%2;
@ -584,7 +584,7 @@ __kernel void dequantize_mul_mat_vec_q5_K(__global const struct block_q5_K * xx,
uint16_t aux[4];
const uint8_t * sc = (const uint8_t *)aux;
__global const struct block_q5_K * x = xx + ib0;
__global const struct block_q5_K * x = xx + ib0 + x_offset;
tmp[16 * ix + tid] = 0;
@ -636,14 +636,14 @@ __kernel void dequantize_mul_mat_vec_q5_K(__global const struct block_q5_K * xx,
}
}
__kernel void dequantize_mul_mat_vec_q6_K(__global const struct block_q6_K * xx, __local float* tmp, __global const float * yy, __global float * dst, const int ncols) {
__kernel void dequantize_mul_mat_vec_q6_K(__global const struct block_q6_K * xx, uint x_offset, __local float* tmp, __global const float * yy, __global float * dst, const int ncols) {
const int row = get_group_id(0);
const int num_blocks_per_row = ncols / QK_K;
const int ib0 = row*num_blocks_per_row + get_global_offset(0);
const int ib0 = row*num_blocks_per_row;
__global const struct block_q6_K * x = xx + ib0;
__global const struct block_q6_K * x = xx + ib0 + x_offset;
const int tid = get_local_id(0)/K_QUANTS_PER_ITERATION; // 0...31 or 0...16
const int ix = get_local_id(0)%K_QUANTS_PER_ITERATION; // 0 or 0, 1
@ -720,7 +720,7 @@ __kernel void dequantize_mul_mat_vec_q6_K(__global const struct block_q6_K * xx,
std::string dequant_template = MULTILINE_QUOTE(
__kernel void KERNEL_NAME(__global X_TYPE* x, __global float* y) {
__kernel void KERNEL_NAME(__global X_TYPE* x, uint x_offset, __global float* y) {
const int i = get_group_id(0)*get_local_size(0) + get_local_id(0)*2;
if (i >= get_global_size(0)) {
@ -730,21 +730,21 @@ __kernel void KERNEL_NAME(__global X_TYPE* x, __global float* y) {
const uint qk = QUANT_K;
const uint qr = QUANT_R;
const int ib = i/qk + get_global_offset(0); // block index
const int ib = i/qk; // block index
const int iqs = (i%qk)/qr; // quant index
const int iybs = i - i%qk; // y block start index
const int y_offset = qr == 1 ? 1 : qk/2;
// dequantize
float v0, v1;
DEQUANT_FUNC(x, ib, iqs, &v0, &v1);
DEQUANT_FUNC(x + x_offset, ib, iqs, &v0, &v1);
y[iybs + iqs + 0] = v0;
y[iybs + iqs + y_offset] = v1;
}
);
std::string dequant_mul_mat_vec_template = MULTILINE_QUOTE(
__kernel void KERNEL_NAME(__global X_TYPE* x, __local float* tmp, __global float* y, __global float* dst, const int ncols) {
__kernel void KERNEL_NAME(__global X_TYPE* x, uint x_offset, __local float* tmp, __global float* y, __global float* dst, const int ncols) {
const int local_size = get_local_size(0);
const int row = get_group_id(0);
const int tid = get_local_id(0);
@ -755,7 +755,7 @@ __kernel void KERNEL_NAME(__global X_TYPE* x, __local float* tmp, __global float
const int col_step = local_size * 2;
const int y_offset = qr == 1 ? 1 : qk/2;
x += get_global_offset(0);
x += x_offset;
tmp[tid] = 0;
@ -1714,6 +1714,8 @@ static void ggml_cl_mul_mat_q_f32(const ggml_tensor * src0, const ggml_tensor *
size_t ev_idx = 0;
std::vector<cl_event> events;
cl_uint x_offset = 0;
for (int64_t i03 = 0; i03 < ne03; i03++) {
// TODO: copy and dequantize src0 here when r3>1
for (int64_t i13 = i03 * r3, e13 = i13 + r3; i13 < e13; i13++) {
@ -1724,6 +1726,7 @@ static void ggml_cl_mul_mat_q_f32(const ggml_tensor * src0, const ggml_tensor *
CL_CHECK(ggml_cl_h2d_tensor_2d(queue, d_Q, 0, src0, i03, i02, events.data() + ev_idx++));
} else if (src0->backend == GGML_BACKEND_GPU) {
d_Q = (cl_mem) src0->extra;
x_offset = (i03 * ne02 + i02) * x_bps;
} else {
GGML_ASSERT(false);
}
@ -1731,10 +1734,10 @@ static void ggml_cl_mul_mat_q_f32(const ggml_tensor * src0, const ggml_tensor *
if (!mul_mat_vec) {
// convert src0 to fp32 on device
const size_t global = x_ne / global_denom;
const size_t offset = src0->backend == GGML_BACKEND_GPU ? (i03 * ne02 + i02) * x_bps : 0;
CL_CHECK(clSetKernelArg(*to_fp32_cl, 0, sizeof(cl_mem), &d_Q));
CL_CHECK(clSetKernelArg(*to_fp32_cl, 1, sizeof(cl_mem), &d_X));
CL_CHECK(clEnqueueNDRangeKernel(queue, *to_fp32_cl, 1, &offset, &global, local > 0 ? &local : NULL, events.size(), !events.empty() ? events.data() : NULL, NULL));
CL_CHECK(clSetKernelArg(*to_fp32_cl, 1, sizeof(cl_uint), &x_offset));
CL_CHECK(clSetKernelArg(*to_fp32_cl, 2, sizeof(cl_mem), &d_X));
CL_CHECK(clEnqueueNDRangeKernel(queue, *to_fp32_cl, 1, NULL, &global, local > 0 ? &local : NULL, events.size(), !events.empty() ? events.data() : NULL, NULL));
}
for (int64_t i12 = i02 * r2, e12 = i12 + r2; i12 < e12; i12++) {
@ -1745,15 +1748,15 @@ static void ggml_cl_mul_mat_q_f32(const ggml_tensor * src0, const ggml_tensor *
// compute
const size_t global = ne01 * local;
const size_t offset = src0->backend == GGML_BACKEND_GPU ? (i03 * ne02 + i02) * x_bps : 0;
const cl_int ncols = ne00;
events.emplace_back();
CL_CHECK(clSetKernelArg(*dmmv, 0, sizeof(cl_mem), &d_Q));
CL_CHECK(clSetKernelArg(*dmmv, 1, sizeof(float) * local, NULL));
CL_CHECK(clSetKernelArg(*dmmv, 2, sizeof(cl_mem), &d_Y));
CL_CHECK(clSetKernelArg(*dmmv, 3, sizeof(cl_mem), &d_D));
CL_CHECK(clSetKernelArg(*dmmv, 4, sizeof(cl_int), &ncols));
CL_CHECK(clEnqueueNDRangeKernel(queue, *dmmv, 1, &offset, &global, &local, events.size() - 1, events.data(), events.data() + ev_idx++));
CL_CHECK(clSetKernelArg(*dmmv, 1, sizeof(cl_uint), &x_offset));
CL_CHECK(clSetKernelArg(*dmmv, 2, sizeof(float) * local, NULL));
CL_CHECK(clSetKernelArg(*dmmv, 3, sizeof(cl_mem), &d_Y));
CL_CHECK(clSetKernelArg(*dmmv, 4, sizeof(cl_mem), &d_D));
CL_CHECK(clSetKernelArg(*dmmv, 5, sizeof(cl_int), &ncols));
CL_CHECK(clEnqueueNDRangeKernel(queue, *dmmv, 1, NULL, &global, &local, events.size() - 1, events.data(), events.data() + ev_idx++));
} else { // CLBlast matrix matrix multiplication
// copy src1 to device
CL_CHECK(ggml_cl_h2d_tensor_2d(queue, d_Y, 0, src1, i13, i12, NULL));