2024-03-27 16:55:10 +00:00
|
|
|
#include "quantize.cuh"
|
2024-06-09 07:42:25 +00:00
|
|
|
#include <cstdint>
|
2024-03-27 16:55:10 +00:00
|
|
|
|
2024-06-09 07:42:25 +00:00
|
|
|
static __global__ void quantize_q8_1(const float * __restrict__ x, void * __restrict__ vy, const int64_t kx, const int64_t kx0_padded) {
|
|
|
|
const int64_t ix0 = (int64_t)blockDim.x*blockIdx.x + threadIdx.x;
|
2024-03-27 16:55:10 +00:00
|
|
|
|
2024-06-09 07:42:25 +00:00
|
|
|
if (ix0 >= kx0_padded) {
|
2024-03-27 16:55:10 +00:00
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
2024-06-09 07:42:25 +00:00
|
|
|
const int64_t ix1 = blockIdx.y;
|
2024-03-27 16:55:10 +00:00
|
|
|
|
2024-06-09 07:42:25 +00:00
|
|
|
const int64_t i_padded = ix1*kx0_padded + ix0;
|
2024-03-27 16:55:10 +00:00
|
|
|
|
|
|
|
block_q8_1 * y = (block_q8_1 *) vy;
|
|
|
|
|
2024-04-09 08:16:13 +00:00
|
|
|
const int64_t ib = i_padded / QK8_1; // block index
|
|
|
|
const int64_t iqs = i_padded % QK8_1; // quant index
|
2024-03-27 16:55:10 +00:00
|
|
|
|
2024-06-09 07:42:25 +00:00
|
|
|
const float xi = ix0 < kx ? x[ix1*kx + ix0] : 0.0f;
|
2024-03-27 16:55:10 +00:00
|
|
|
float amax = fabsf(xi);
|
|
|
|
float sum = xi;
|
|
|
|
|
|
|
|
amax = warp_reduce_max(amax);
|
|
|
|
sum = warp_reduce_sum(sum);
|
|
|
|
|
|
|
|
const float d = amax / 127;
|
|
|
|
const int8_t q = amax == 0.0f ? 0 : roundf(xi / d);
|
|
|
|
|
|
|
|
y[ib].qs[iqs] = q;
|
|
|
|
|
|
|
|
if (iqs > 0) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
reinterpret_cast<half&>(y[ib].ds.x) = d;
|
|
|
|
reinterpret_cast<half&>(y[ib].ds.y) = sum;
|
|
|
|
}
|
|
|
|
|
2024-06-09 07:42:25 +00:00
|
|
|
template <bool need_sum>
|
|
|
|
static __global__ void quantize_mmq_q8_1(
|
|
|
|
const float * __restrict__ x, void * __restrict__ vy, const int64_t kx0, const int64_t kx1, const int64_t kx0_padded) {
|
|
|
|
|
|
|
|
const int64_t ix0 = (int64_t)blockDim.x*blockIdx.x + threadIdx.x;
|
|
|
|
|
|
|
|
if (ix0 >= kx0_padded) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
const int64_t ix1 = kx1*blockIdx.z + blockIdx.y;
|
|
|
|
|
|
|
|
block_q8_1_mmq * y = (block_q8_1_mmq *) vy;
|
|
|
|
|
|
|
|
const int64_t ib0 = blockIdx.z*(gridDim.y*gridDim.x*blockDim.x/(4*QK8_1)); // first block of channel
|
|
|
|
const int64_t ib = ib0 + (ix0 / (4*QK8_1))*kx1 + blockIdx.y; // block index in channel
|
|
|
|
const int64_t iqs = ix0 % (4*QK8_1); // quant index in block
|
|
|
|
|
|
|
|
const float xi = ix0 < kx0 ? x[ix1*kx0 + ix0] : 0.0f;
|
|
|
|
float amax = fabsf(xi);
|
|
|
|
|
|
|
|
amax = warp_reduce_max(amax);
|
|
|
|
|
|
|
|
float sum;
|
|
|
|
if (need_sum) {
|
|
|
|
sum = warp_reduce_sum(xi);
|
|
|
|
}
|
|
|
|
|
|
|
|
const float d = amax / 127;
|
|
|
|
const int8_t q = amax == 0.0f ? 0 : roundf(xi / d);
|
|
|
|
|
|
|
|
y[ib].qs[iqs] = q;
|
|
|
|
|
|
|
|
if (iqs % QK8_1 != 0) {
|
|
|
|
return;
|
|
|
|
}
|
|
|
|
|
|
|
|
if (need_sum) {
|
|
|
|
y[ib].ds[iqs/QK8_1] = make_half2(d, sum);
|
|
|
|
} else {
|
|
|
|
((float *) y[ib].ds)[iqs/QK8_1] = d;
|
|
|
|
}
|
|
|
|
}
|
|
|
|
|
|
|
|
void quantize_row_q8_1_cuda(
|
|
|
|
const float * x, void * vy, const int64_t kx0, const int64_t kx1, const int64_t channels,
|
|
|
|
const int64_t kx0_padded, const ggml_type type_x, cudaStream_t stream) {
|
|
|
|
|
|
|
|
GGML_ASSERT(kx0_padded % QK8_1 == 0);
|
|
|
|
|
|
|
|
const int64_t block_num_x = (kx0_padded + CUDA_QUANTIZE_BLOCK_SIZE - 1) / CUDA_QUANTIZE_BLOCK_SIZE;
|
|
|
|
const dim3 num_blocks(block_num_x, kx1*channels, 1);
|
2024-03-27 16:55:10 +00:00
|
|
|
const dim3 block_size(CUDA_QUANTIZE_BLOCK_SIZE, 1, 1);
|
2024-06-09 07:42:25 +00:00
|
|
|
quantize_q8_1<<<num_blocks, block_size, 0, stream>>>(x, vy, kx0, kx0_padded);
|
|
|
|
|
|
|
|
GGML_UNUSED(type_x);
|
2024-03-27 16:55:10 +00:00
|
|
|
}
|
|
|
|
|
2024-06-09 07:42:25 +00:00
|
|
|
void quantize_mmq_q8_1_cuda(
|
|
|
|
const float * x, void * vy, const int64_t kx0, const int64_t kx1, const int64_t channels,
|
|
|
|
const int64_t kx0_padded, const ggml_type type_x, cudaStream_t stream) {
|
|
|
|
|
|
|
|
GGML_ASSERT(kx0_padded % (4*QK8_1) == 0);
|
|
|
|
|
|
|
|
const int64_t block_num_x = (kx0_padded + CUDA_QUANTIZE_BLOCK_SIZE - 1) / CUDA_QUANTIZE_BLOCK_SIZE;
|
|
|
|
const dim3 num_blocks(block_num_x, kx1, channels);
|
|
|
|
const dim3 block_size(CUDA_QUANTIZE_BLOCK_SIZE, 1, 1);
|
|
|
|
if (mmq_need_sum(type_x)) {
|
|
|
|
quantize_mmq_q8_1<true><<<num_blocks, block_size, 0, stream>>>(x, vy, kx0, kx1, kx0_padded);
|
|
|
|
} else {
|
|
|
|
quantize_mmq_q8_1<false><<<num_blocks, block_size, 0, stream>>>(x, vy, kx0, kx1, kx0_padded);
|
|
|
|
}
|
|
|
|
}
|