ollama/llama/ggml-cuda/dmmv.cu
Gabe Goodhart f2890a4494
IBM granite/granitemoe architecture support (#6760)
* fix(ext_server): Port llama.cpp sampling refactors to ext_server

This was a fairly large changeset. I closely followed the changes here:
df270ef745

Branch: IBMGraniteArchitectureSupport

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* fix(server.cpp): Refactor server.cpp logging for llama.cpp overhaul

Branch: IBMGraniteArchitectureSupport

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* feat: Bump llama.cpp to the latest master with `granite` support

This does not yet have granite MoE support, but that can come in a
follow up PR

Branch: IBMGraniteArchitectureSupport

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* fix(patches): Update all patches (except solar-pro) to work with bumped llama.cpp

Branch: IBMGraniteArchitectureSupport

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* fix(solar): Update solar patch for llama.cpp bump

Branch: IBMGraniteArchitectureSupport

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* feat(llama.cpp): Bump llama.cpp for granitemoe support

Branch: IBMGraniteArchitectureSupport

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* feat(llama.cpp): Bump llama.cpp for granitemoe support

Branch: IBMGraniteArchitectureSupport

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* fix(solar): Update the solar-pro patch for latest llama.cpp bump

Branch: IBMGraniteArchitectureSupport

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* feat(llama.cpp): Bump to the latest master of llama.cpp

Branch: IBMGraniteArchitectureSupport

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* fix(patches): Update all patches for latest bump

Branch: IBMGraniteArchitectureSupport

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* feat(llama): Always run sync.sh from the right directory

Branch: IBMGraniteArchitectureSupport

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* fix(llama/patches): Update llama patches

Branch: IBMGraniteArchitectureSupport

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* feat(llama)!: Rough sync with llama.cpp submodule

There are a number of changes that will need to be propagated to llama.go
before any of this works!

Branch: IBMGraniteArchitectureSupport

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* fix(llama/patches): Add a patch and update for missing ggml-impl.h include

This include is where the ggml_cgraph struct is defined. It is included in
many of the .c files to define the forward declartion in ggml.h. It seems
that with the subset of code included here, the import was somehow lost (or
out-of-order) when building, so adding this include to llama.cpp fixes the
missing definition.

Branch: IBMGraniteArchitectureSupport

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* fix(llama/sync): Add missing ggml-cpu-impl.h copy-over in sync.sh

Branch: IBMGraniteArchitectureSupport

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* fix(llama): Add missing log.cpp

This was added as part of the logging overhaul done in llama.cpp

Branch: IBMGraniteArchitectureSupport

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* fix(llama): Overhaul use of sampling module for llama.cpp changes

The changes here reflect the changes made in the big llama.cpp sampling PR
https://github.com/ggerganov/llama.cpp/pull/9294

The sampling functionality is now broken into the base interface
(llama_sampler) and the generation implementation (gpt_sampler). The
changes here reflect that. Since the sampling.h/sampling.cpp code uses c++
STL headers, the sampling_ext.[h|cpp] wrapper is maintained to allow go to
access a pure-C interface.

Branch: IBMGraniteArchitectureSupport

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* fix(llama): Fix the impl of SampleTokenGreedy for new sampling

I don't think this method is currently used, so it could probably just be
removed so that all sampling goes through the GPT interface, but in the
interest of doing no harm, this should keep the method working as expected.

Branch: IBMGraniteArchitectureSupport

* fix(llama): Remove unused SampleTokenGreedy

Branch: IBMGraniteArchitectureSupport

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* fix(sync): Remove bash-specific change to sync.sh

Branch: IBMGraniteArchitectureSupport

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* chore(gofumpt): Format on llama.go to pass linting

Branch: IBMGraniteArchitectureSupport

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* fix(llm): Fix missing <thread> include in ext_server

Branch: IBMGraniteArchitectureSupport

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* fix(llama): Remove TODO about grammar_first

This feature was not used/needed previously so should be fine without
plumbing it through now.

Branch: IBMGraniteArchitectureSupport

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* fix(llama): Better naming for sampling wrapper and args

Branch: IBMGraniteArchitectureSupport

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* fix(llama): Fix patch 05 to use new wrapper api and re-sync

Branch: IBMGraniteArchitectureSupport

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* runner: Flush pending responses before returning

If there are any pending reponses (such as from potential stop
tokens) then we should send them back before ending the sequence.
Otherwise, we can be missing tokens at the end of a response.

Fixes #6707

* fix(llama/sampling): Use gpt_sampler with a forward declaration

Branch: IBMGraniteArchitectureSupport

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* fix(llama): Remove unnecessary patch for gguf impl header

This was caused by an earlier mistake in the embeddings patch that was
dereferencing the pointer instead of using the wrapper API.

Branch: IBMGraniteArchitectureSupport

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

* fix(llm): Remove use of deprecated --log-disable flag

Branch: IBMGraniteArchitectureSupport

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>

---------

Signed-off-by: Gabe Goodhart <ghart@us.ibm.com>
2024-10-17 11:59:52 -07:00

709 lines
29 KiB
Text
Vendored

/**
* llama.cpp - commit 3f1ae2e32cde00c39b96be6d01c2997c29bae555 - do not edit this file
*
* MIT License
*
* Copyright (c) 2023-2024 The ggml authors
*
* Permission is hereby granted, free of charge, to any person obtaining a copy
* of this software and associated documentation files (the "Software"), to deal
* in the Software without restriction, including without limitation the rights
* to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
* copies of the Software, and to permit persons to whom the Software is
* furnished to do so, subject to the following conditions:
*
* The above copyright notice and this permission notice shall be included in all
* copies or substantial portions of the Software.
*
* THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
* IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
* FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
* AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
* LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
* OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
* SOFTWARE.
*/
#include "dmmv.cuh"
#include "dequantize.cuh"
#include "convert.cuh"
#ifndef K_QUANTS_PER_ITERATION
#define K_QUANTS_PER_ITERATION 2
#else
static_assert(K_QUANTS_PER_ITERATION == 1 || K_QUANTS_PER_ITERATION == 2, "K_QUANTS_PER_ITERATION must be 1 or 2");
#endif
static __global__ void dequantize_mul_mat_vec_q2_k(const void * __restrict__ vx, const float * __restrict__ yy, float * __restrict__ dst, const int ncols, int nrows) {
static_assert(16%K_QUANTS_PER_ITERATION == 0, "16 must be divisible by K_QUANTS_PER_ITERATION");
const int row = blockIdx.x*blockDim.y + threadIdx.y;
if (row > nrows) return;
const int num_blocks_per_row = ncols / QK_K;
const int ib0 = row*num_blocks_per_row;
const block_q2_K * x = (const block_q2_K *)vx + ib0;
float tmp = 0; // partial sum for thread in warp
const int tid = threadIdx.x/K_QUANTS_PER_ITERATION; // 0...31 or 0...15
const int ix = threadIdx.x%K_QUANTS_PER_ITERATION; // 0 or 0,1
const int step = 16/K_QUANTS_PER_ITERATION;
const int im = tid/step; // 0 or 1. 0 computes 0..., 1 computes 128...
const int in = tid - step*im; // 0...15 or 0...7
const int l0 = K_QUANTS_PER_ITERATION*in; // 0...15 or 0...14 in steps of 2
const int q_offset = 32*im + l0;
const int s_offset = 8*im;
const int y_offset = 128*im + l0;
uint32_t aux[4];
const uint8_t * d = (const uint8_t *)aux;
const uint8_t * m = (const uint8_t *)(aux + 2);
for (int i = ix; i < num_blocks_per_row; i += K_QUANTS_PER_ITERATION) {
const float * y = yy + i * QK_K + y_offset;
const uint8_t * q = x[i].qs + q_offset;
const float dall = __low2half(x[i].dm);
const float dmin = __high2half(x[i].dm);
const uint32_t * a = (const uint32_t *)(x[i].scales + s_offset);
aux[0] = a[0] & 0x0f0f0f0f;
aux[1] = a[1] & 0x0f0f0f0f;
aux[2] = (a[0] >> 4) & 0x0f0f0f0f;
aux[3] = (a[1] >> 4) & 0x0f0f0f0f;
float sum1 = 0, sum2 = 0;
for (int l = 0; l < K_QUANTS_PER_ITERATION; ++l) {
sum1 += y[l+ 0] * d[0] * ((q[l+ 0] >> 0) & 3)
+ y[l+32] * d[2] * ((q[l+ 0] >> 2) & 3)
+ y[l+64] * d[4] * ((q[l+ 0] >> 4) & 3)
+ y[l+96] * d[6] * ((q[l+ 0] >> 6) & 3)
+ y[l+16] * d[1] * ((q[l+16] >> 0) & 3)
+ y[l+48] * d[3] * ((q[l+16] >> 2) & 3)
+ y[l+80] * d[5] * ((q[l+16] >> 4) & 3)
+y[l+112] * d[7] * ((q[l+16] >> 6) & 3);
sum2 += y[l+ 0] * m[0] + y[l+32] * m[2] + y[l+64] * m[4] + y[ l+96] * m[6]
+ y[l+16] * m[1] + y[l+48] * m[3] + y[l+80] * m[5] + y[l+112] * m[7];
}
tmp += dall * sum1 - dmin * sum2;
}
// sum up partial sums and write back result
tmp = warp_reduce_sum(tmp);
if (threadIdx.x == 0) {
dst[row] = tmp;
}
}
static __global__ void dequantize_mul_mat_vec_q3_k(const void * __restrict__ vx, const float * __restrict__ yy, float * __restrict__ dst, const int ncols, int nrows) {
const int row = blockIdx.x*blockDim.y + threadIdx.y;
if (row > nrows) return;
const int num_blocks_per_row = ncols / QK_K;
const int ib0 = row*num_blocks_per_row;
const block_q3_K * x = (const block_q3_K *)vx + ib0;
float tmp = 0; // partial sum for thread in warp
const uint16_t kmask1 = 0x0303;
const uint16_t kmask2 = 0x0f0f;
const int tid = threadIdx.x/K_QUANTS_PER_ITERATION; // 0...31 or 0...16
const int ix = threadIdx.x%K_QUANTS_PER_ITERATION; // 0 or 0,1
const int n = K_QUANTS_PER_ITERATION; // iterations in the inner loop
const int step = 16/K_QUANTS_PER_ITERATION;
const int im = tid/step; // 0 or 1. 0 computes 0..., 1 computes 128...
const int in = tid - step*im; // 0....15 or 0...7
const uint8_t m = 1 << (4*im);
const int l0 = n*in; // 0...15 or 0...14 in steps of 2
const int q_offset = 32*im + l0;
const int y_offset = 128*im + l0;
uint16_t utmp[4];
const int8_t * s = (const int8_t *)utmp;
const uint16_t s_shift = 4*im;
for (int i = ix; i < num_blocks_per_row; i += K_QUANTS_PER_ITERATION) {
const float * y = yy + i * QK_K + y_offset;
const uint8_t * q = x[i].qs + q_offset;
const uint8_t * h = x[i].hmask + l0;
const uint16_t * a = (const uint16_t *)x[i].scales;
utmp[0] = ((a[0] >> s_shift) & kmask2) | (((a[4] >> (s_shift + 0)) & kmask1) << 4);
utmp[1] = ((a[1] >> s_shift) & kmask2) | (((a[5] >> (s_shift + 0)) & kmask1) << 4);
utmp[2] = ((a[2] >> s_shift) & kmask2) | (((a[4] >> (s_shift + 2)) & kmask1) << 4);
utmp[3] = ((a[3] >> s_shift) & kmask2) | (((a[5] >> (s_shift + 2)) & kmask1) << 4);
const float d = x[i].d;
float sum = 0;
for (int l = 0; l < n; ++l) {
sum += y[l+ 0] * (s[0] - 32) * (((q[l] >> 0) & 3) - (h[l] & (m << 0) ? 0 : 4))
+ y[l+32] * (s[2] - 32) * (((q[l] >> 2) & 3) - (h[l] & (m << 1) ? 0 : 4))
+ y[l+64] * (s[4] - 32) * (((q[l] >> 4) & 3) - (h[l] & (m << 2) ? 0 : 4))
+ y[l+96] * (s[6] - 32) * (((q[l] >> 6) & 3) - (h[l] & (m << 3) ? 0 : 4));
sum += y[l+16] * (s[1] - 32) * (((q[l+16] >> 0) & 3) - (h[l+16] & (m << 0) ? 0 : 4))
+ y[l+48] * (s[3] - 32) * (((q[l+16] >> 2) & 3) - (h[l+16] & (m << 1) ? 0 : 4))
+ y[l+80] * (s[5] - 32) * (((q[l+16] >> 4) & 3) - (h[l+16] & (m << 2) ? 0 : 4))
+ y[l+112] * (s[7] - 32) * (((q[l+16] >> 6) & 3) - (h[l+16] & (m << 3) ? 0 : 4));
}
tmp += d * sum;
}
// sum up partial sums and write back result
tmp = warp_reduce_sum(tmp);
if (threadIdx.x == 0) {
dst[row] = tmp;
}
}
static __global__ void dequantize_mul_mat_vec_q4_k(const void * __restrict__ vx, const float * __restrict__ yy, float * __restrict__ dst, const int ncols, int nrows) {
const int row = blockIdx.x*blockDim.y + threadIdx.y;
if (row > nrows) return;
const int num_blocks_per_row = ncols / QK_K;
const int ib0 = row*num_blocks_per_row;
const block_q4_K * x = (const block_q4_K *)vx + ib0;
const uint16_t kmask1 = 0x3f3f;
const uint16_t kmask2 = 0x0f0f;
const uint16_t kmask3 = 0xc0c0;
const int tid = threadIdx.x/K_QUANTS_PER_ITERATION; // 0...31 or 0...16
const int ix = threadIdx.x%K_QUANTS_PER_ITERATION; // 0 or 0,1
const int step = 8/K_QUANTS_PER_ITERATION; // 8 or 4
const int il = tid/step; // 0...3
const int ir = tid - step*il; // 0...7 or 0...3
const int n = 2 * K_QUANTS_PER_ITERATION; // 2 or 4
const int im = il/2; // 0 or 1. 0 computes 0,32 + 128,160, 1 computes 64,96 + 192,224
const int in = il%2;
const int l0 = n*(2*ir + in);
const int q_offset = 32*im + l0;
const int y_offset = 64*im + l0;
uint16_t aux[4];
const uint8_t * sc = (const uint8_t *)aux;
#if K_QUANTS_PER_ITERATION == 2
uint32_t q32[4];
const uint8_t * q4 = (const uint8_t *)q32;
#else
uint16_t q16[4];
const uint8_t * q4 = (const uint8_t *)q16;
#endif
float tmp = 0; // partial sum for thread in warp
for (int i = ix; i < num_blocks_per_row; i += K_QUANTS_PER_ITERATION) {
const float * y1 = yy + i*QK_K + y_offset;
const float * y2 = y1 + 128;
const float dall = __low2half(x[i].dm);
const float dmin = __high2half(x[i].dm);
const uint16_t * a = (const uint16_t *)x[i].scales;
aux[0] = a[im+0] & kmask1;
aux[1] = a[im+2] & kmask1;
aux[2] = ((a[im+4] >> 0) & kmask2) | ((a[im+0] & kmask3) >> 2);
aux[3] = ((a[im+4] >> 4) & kmask2) | ((a[im+2] & kmask3) >> 2);
#if K_QUANTS_PER_ITERATION == 2
const uint32_t * q1 = (const uint32_t *)(x[i].qs + q_offset);
const uint32_t * q2 = q1 + 16;
q32[0] = q1[0] & 0x0f0f0f0f;
q32[1] = q1[0] & 0xf0f0f0f0;
q32[2] = q2[0] & 0x0f0f0f0f;
q32[3] = q2[0] & 0xf0f0f0f0;
float4 s = {0.f, 0.f, 0.f, 0.f};
float smin = 0;
for (int l = 0; l < 4; ++l) {
s.x += y1[l] * q4[l+0]; s.y += y1[l+32] * q4[l+ 4];
s.z += y2[l] * q4[l+8]; s.w += y2[l+32] * q4[l+12];
smin += y1[l] * sc[2] + y1[l+32] * sc[3] + y2[l] * sc[6] + y2[l+32] * sc[7];
}
tmp += dall * (s.x * sc[0] + s.y * sc[1] * 1.f/16.f + s.z * sc[4] + s.w * sc[5] * 1.f/16.f) - dmin * smin;
#else
const uint16_t * q1 = (const uint16_t *)(x[i].qs + q_offset);
const uint16_t * q2 = q1 + 32;
q16[0] = q1[0] & 0x0f0f;
q16[1] = q1[0] & 0xf0f0;
q16[2] = q2[0] & 0x0f0f;
q16[3] = q2[0] & 0xf0f0;
float4 s = {0.f, 0.f, 0.f, 0.f};
float smin = 0;
for (int l = 0; l < 2; ++l) {
s.x += y1[l] * q4[l+0]; s.y += y1[l+32] * q4[l+2];
s.z += y2[l] * q4[l+4]; s.w += y2[l+32] * q4[l+6];
smin += y1[l] * sc[2] + y1[l+32] * sc[3] + y2[l] * sc[6] + y2[l+32] * sc[7];
}
tmp += dall * (s.x * sc[0] + s.y * sc[1] * 1.f/16.f + s.z * sc[4] + s.w * sc[5] * 1.f/16.f) - dmin * smin;
#endif
}
// sum up partial sums and write back result
tmp = warp_reduce_sum(tmp);
if (tid == 0) {
dst[row] = tmp;
}
}
static __global__ void dequantize_mul_mat_vec_q5_k(const void * __restrict__ vx, const float * __restrict__ yy, float * __restrict__ dst, const int ncols) {
const int row = blockIdx.x;
const int num_blocks_per_row = ncols / QK_K;
const int ib0 = row*num_blocks_per_row;
const block_q5_K * x = (const block_q5_K *)vx + ib0;
float tmp = 0; // partial sum for thread in warp
const uint16_t kmask1 = 0x3f3f;
const uint16_t kmask2 = 0x0f0f;
const uint16_t kmask3 = 0xc0c0;
const int tid = threadIdx.x/2; // 0...15
const int ix = threadIdx.x%2;
const int il = tid/4; // 0...3
const int ir = tid - 4*il;// 0...3
const int n = 2;
const int im = il/2; // 0 or 1. 0 computes 0,32 + 128,160, 1 computes 64,96 + 192,224
const int in = il%2;
const int l0 = n*(2*ir + in);
const int q_offset = 32*im + l0;
const int y_offset = 64*im + l0;
const uint8_t hm1 = 1 << (2*im);
const uint8_t hm2 = hm1 << 4;
uint16_t aux[4];
const uint8_t * sc = (const uint8_t *)aux;
uint16_t q16[8];
const uint8_t * q4 = (const uint8_t *)q16;
for (int i = ix; i < num_blocks_per_row; i += 2) {
const uint8_t * ql1 = x[i].qs + q_offset;
const uint8_t * qh = x[i].qh + l0;
const float * y1 = yy + i*QK_K + y_offset;
const float * y2 = y1 + 128;
const float dall = __low2half(x[i].dm);
const float dmin = __high2half(x[i].dm);
const uint16_t * a = (const uint16_t *)x[i].scales;
aux[0] = a[im+0] & kmask1;
aux[1] = a[im+2] & kmask1;
aux[2] = ((a[im+4] >> 0) & kmask2) | ((a[im+0] & kmask3) >> 2);
aux[3] = ((a[im+4] >> 4) & kmask2) | ((a[im+2] & kmask3) >> 2);
float4 sum = {0.f, 0.f, 0.f, 0.f};
float smin = 0;
const uint16_t * q1 = (const uint16_t *)ql1;
const uint16_t * q2 = q1 + 32;
q16[0] = q1[0] & 0x0f0f;
q16[1] = q1[8] & 0x0f0f;
q16[2] = (q1[0] >> 4) & 0x0f0f;
q16[3] = (q1[8] >> 4) & 0x0f0f;
q16[4] = q2[0] & 0x0f0f;
q16[5] = q2[8] & 0x0f0f;
q16[6] = (q2[0] >> 4) & 0x0f0f;
q16[7] = (q2[8] >> 4) & 0x0f0f;
for (int l = 0; l < n; ++l) {
sum.x += y1[l+ 0] * (q4[l +0] + (qh[l+ 0] & (hm1 << 0) ? 16 : 0))
+ y1[l+16] * (q4[l +2] + (qh[l+16] & (hm1 << 0) ? 16 : 0));
sum.y += y1[l+32] * (q4[l +4] + (qh[l+ 0] & (hm1 << 1) ? 16 : 0))
+ y1[l+48] * (q4[l +6] + (qh[l+16] & (hm1 << 1) ? 16 : 0));
sum.z += y2[l+ 0] * (q4[l +8] + (qh[l+ 0] & (hm2 << 0) ? 16 : 0))
+ y2[l+16] * (q4[l+10] + (qh[l+16] & (hm2 << 0) ? 16 : 0));
sum.w += y2[l+32] * (q4[l+12] + (qh[l+ 0] & (hm2 << 1) ? 16 : 0))
+ y2[l+48] * (q4[l+14] + (qh[l+16] & (hm2 << 1) ? 16 : 0));
smin += (y1[l] + y1[l+16]) * sc[2] + (y1[l+32] + y1[l+48]) * sc[3]
+ (y2[l] + y2[l+16]) * sc[6] + (y2[l+32] + y2[l+48]) * sc[7];
}
tmp += dall * (sum.x * sc[0] + sum.y * sc[1] + sum.z * sc[4] + sum.w * sc[5]) - dmin * smin;
}
// sum up partial sums and write back result
tmp = warp_reduce_sum(tmp);
if (threadIdx.x == 0) {
dst[row] = tmp;
}
}
static __global__ void dequantize_mul_mat_vec_q6_k(const void * __restrict__ vx, const float * __restrict__ yy, float * __restrict__ dst, const int ncols, int nrows) {
static_assert(16%K_QUANTS_PER_ITERATION == 0, "16 must be divisible by K_QUANTS_PER_ITERATION");
const int row = blockIdx.x*blockDim.y + threadIdx.y;
if (row > nrows) return;
const int num_blocks_per_row = ncols / QK_K;
const int ib0 = row*num_blocks_per_row;
const block_q6_K * x = (const block_q6_K *)vx + ib0;
const int tid = threadIdx.x/K_QUANTS_PER_ITERATION; // 0...31 or 0...16
const int ix = threadIdx.x%K_QUANTS_PER_ITERATION; // 0 or 0, 1
const int step = 16/K_QUANTS_PER_ITERATION; // 16 or 8
const int im = tid/step; // 0 or 1. 0 computes 0..., 1 computes 128...
const int in = tid - step*im; // 0...15 or 0...7
#if K_QUANTS_PER_ITERATION == 1
const int l0 = K_QUANTS_PER_ITERATION*in; // 0...15
const int is = 0;
#else
const int l0 = 4 * in; // 0, 4, 8, ..., 28
const int is = in / 4;
#endif
const int ql_offset = 64*im + l0;
const int qh_offset = 32*im + l0;
const int s_offset = 8*im + is;
const int y_offset = 128*im + l0;
float tmp = 0; // partial sum for thread in warp
for (int i = ix; i < num_blocks_per_row; i += K_QUANTS_PER_ITERATION) {
const float * y = yy + i * QK_K + y_offset;
const uint8_t * ql = x[i].ql + ql_offset;
const uint8_t * qh = x[i].qh + qh_offset;
const int8_t * s = x[i].scales + s_offset;
const float d = x[i].d;
#if K_QUANTS_PER_ITERATION == 1
float sum = y[ 0] * s[0] * d * ((int8_t)((ql[ 0] & 0xF) | ((qh[ 0] & 0x03) << 4)) - 32)
+ y[16] * s[1] * d * ((int8_t)((ql[16] & 0xF) | ((qh[16] & 0x03) << 4)) - 32)
+ y[32] * s[2] * d * ((int8_t)((ql[32] & 0xF) | ((qh[ 0] & 0x0c) << 2)) - 32)
+ y[48] * s[3] * d * ((int8_t)((ql[48] & 0xF) | ((qh[16] & 0x0c) << 2)) - 32)
+ y[64] * s[4] * d * ((int8_t)((ql[ 0] >> 4) | ((qh[ 0] & 0x30) >> 0)) - 32)
+ y[80] * s[5] * d * ((int8_t)((ql[16] >> 4) | ((qh[16] & 0x30) >> 0)) - 32)
+ y[96] * s[6] * d * ((int8_t)((ql[32] >> 4) | ((qh[ 0] & 0xc0) >> 2)) - 32)
+y[112] * s[7] * d * ((int8_t)((ql[48] >> 4) | ((qh[16] & 0xc0) >> 2)) - 32);
tmp += sum;
#else
float sum = 0;
for (int l = 0; l < 4; ++l) {
sum += y[l+ 0] * s[0] * d * ((int8_t)((ql[l+ 0] & 0xF) | (((qh[l] >> 0) & 3) << 4)) - 32)
+ y[l+32] * s[2] * d * ((int8_t)((ql[l+32] & 0xF) | (((qh[l] >> 2) & 3) << 4)) - 32)
+ y[l+64] * s[4] * d * ((int8_t)((ql[l+ 0] >> 4) | (((qh[l] >> 4) & 3) << 4)) - 32)
+ y[l+96] * s[6] * d * ((int8_t)((ql[l+32] >> 4) | (((qh[l] >> 6) & 3) << 4)) - 32);
}
tmp += sum;
#endif
}
// sum up partial sums and write back result
tmp = warp_reduce_sum(tmp);
if (tid == 0) {
dst[row] = tmp;
}
}
static __device__ void convert_f16(const void * vx, const int64_t ib, const int iqs, dfloat2 & v){
const half * x = (const half *) vx;
// automatic half -> float type cast if dfloat == float
v.x = x[ib + iqs + 0];
v.y = x[ib + iqs + 1];
}
static constexpr __device__ dequantize_kernel_t get_dequantize_kernel(ggml_type type) {
return type == GGML_TYPE_Q4_0 ? dequantize_q4_0 :
type == GGML_TYPE_Q4_1 ? dequantize_q4_1 :
type == GGML_TYPE_Q5_0 ? dequantize_q5_0 :
type == GGML_TYPE_Q5_1 ? dequantize_q5_1 :
type == GGML_TYPE_Q8_0 ? dequantize_q8_0 :
type == GGML_TYPE_F16 ? convert_f16 :
nullptr;
}
template <ggml_type type>
static __global__ void dequantize_mul_mat_vec(const void * __restrict__ vx, const dfloat * __restrict__ y, float * __restrict__ dst, const int ncols, const int nrows) {
constexpr int qk = ggml_cuda_type_traits<type>::qk; // quantized weights per x block
constexpr int qr = ggml_cuda_type_traits<type>::qr; // number of quantized weights per data value in x block
constexpr dequantize_kernel_t dequantize_kernel = get_dequantize_kernel(type);
const int64_t row = (int64_t)blockIdx.x*blockDim.y + threadIdx.y;
if (row >= nrows) {
return;
}
const int tid = threadIdx.x;
const int iter_stride = 2*GGML_CUDA_DMMV_X;
const int vals_per_iter = iter_stride / WARP_SIZE; // num quantized vals per thread and i iter
const int y_offset = qr == 1 ? 1 : qk/2;
// partial sum for each thread
#ifdef GGML_CUDA_F16
half2 tmp = {0.0f, 0.0f}; // two sums for f16 to take advantage of half2 intrinsics
#else
float tmp = 0.0f;
#endif // GGML_CUDA_F16
for (int i = 0; i < ncols; i += iter_stride) {
const int col = i + vals_per_iter*tid;
const int64_t ib = ((int64_t)row*ncols + col)/qk; // x block index
const int iqs = (col%qk)/qr; // x quant index
const int iybs = col - col%qk; // y block start index
// processing >2 values per i iter is faster for fast GPUs
#pragma unroll
for (int j = 0; j < vals_per_iter; j += 2) {
// process 2 vals per j iter
// dequantize
// for qr = 2 the iqs needs to increase by 1 per j iter because 2 weights per data val
dfloat2 v;
dequantize_kernel(vx, ib, iqs + j/qr, v);
// matrix multiplication
// for qr = 2 the y index needs to increase by 1 per j iter because of y_offset = qk/2
#ifdef GGML_CUDA_F16
tmp += __hmul2(v, {
y[iybs + iqs + j/qr + 0],
y[iybs + iqs + j/qr + y_offset]
});
#else
tmp += v.x * y[iybs + iqs + j/qr + 0];
tmp += v.y * y[iybs + iqs + j/qr + y_offset];
#endif // GGML_CUDA_F16
}
}
// sum up partial sums and write back result
tmp = warp_reduce_sum(tmp);
if (tid == 0) {
#ifdef GGML_CUDA_F16
dst[row] = tmp.x + tmp.y;
#else
dst[row] = tmp;
#endif // GGML_CUDA_F16
}
}
static void dequantize_mul_mat_vec_q4_0_cuda(const void * vx, const dfloat * y, float * dst, const int ncols, const int nrows, cudaStream_t stream) {
GGML_ASSERT(ncols % (GGML_CUDA_DMMV_X*2) == 0);
const int block_num_y = (nrows + GGML_CUDA_MMV_Y - 1) / GGML_CUDA_MMV_Y;
// the number of rows may exceed maximum grid size in the y or z dimensions, use the x dimension instead
const dim3 block_nums(block_num_y, 1, 1);
const dim3 block_dims(WARP_SIZE, GGML_CUDA_MMV_Y, 1);
dequantize_mul_mat_vec<GGML_TYPE_Q4_0>
<<<block_nums, block_dims, 0, stream>>>(vx, y, dst, ncols, nrows);
}
static void dequantize_mul_mat_vec_q4_1_cuda(const void * vx, const dfloat * y, float * dst, const int ncols, const int nrows, cudaStream_t stream) {
GGML_ASSERT(ncols % (GGML_CUDA_DMMV_X*2) == 0);
const int block_num_y = (nrows + GGML_CUDA_MMV_Y - 1) / GGML_CUDA_MMV_Y;
const dim3 block_nums(block_num_y, 1, 1);
const dim3 block_dims(WARP_SIZE, GGML_CUDA_MMV_Y, 1);
dequantize_mul_mat_vec<GGML_TYPE_Q4_1>
<<<block_nums, block_dims, 0, stream>>>(vx, y, dst, ncols, nrows);
}
static void dequantize_mul_mat_vec_q5_0_cuda(const void * vx, const dfloat * y, float * dst, const int ncols, const int nrows, cudaStream_t stream) {
GGML_ASSERT(ncols % (GGML_CUDA_DMMV_X*2) == 0);
const int block_num_y = (nrows + GGML_CUDA_MMV_Y - 1) / GGML_CUDA_MMV_Y;
const dim3 block_nums(block_num_y, 1, 1);
const dim3 block_dims(WARP_SIZE, GGML_CUDA_MMV_Y, 1);
dequantize_mul_mat_vec<GGML_TYPE_Q5_0>
<<<block_nums, block_dims, 0, stream>>>(vx, y, dst, ncols, nrows);
}
static void dequantize_mul_mat_vec_q5_1_cuda(const void * vx, const dfloat * y, float * dst, const int ncols, const int nrows, cudaStream_t stream) {
GGML_ASSERT(ncols % (GGML_CUDA_DMMV_X*2) == 0);
const int block_num_y = (nrows + GGML_CUDA_MMV_Y - 1) / GGML_CUDA_MMV_Y;
const dim3 block_nums(block_num_y, 1, 1);
const dim3 block_dims(WARP_SIZE, GGML_CUDA_MMV_Y, 1);
dequantize_mul_mat_vec<GGML_TYPE_Q5_1>
<<<block_nums, block_dims, 0, stream>>>(vx, y, dst, ncols, nrows);
}
static void dequantize_mul_mat_vec_q8_0_cuda(const void * vx, const dfloat * y, float * dst, const int ncols, const int nrows, cudaStream_t stream) {
GGML_ASSERT(ncols % (GGML_CUDA_DMMV_X*2) == 0);
const int block_num_y = (nrows + GGML_CUDA_MMV_Y - 1) / GGML_CUDA_MMV_Y;
const dim3 block_nums(block_num_y, 1, 1);
const dim3 block_dims(WARP_SIZE, GGML_CUDA_MMV_Y, 1);
dequantize_mul_mat_vec<GGML_TYPE_Q8_0>
<<<block_nums, block_dims, 0, stream>>>(vx, y, dst, ncols, nrows);
}
static void dequantize_mul_mat_vec_q2_K_cuda(const void * vx, const float * y, float * dst, const int ncols, const int nrows, cudaStream_t stream) {
GGML_ASSERT(ncols % QK_K == 0);
const int ny = 2; // very slightly faster than 1 even when K_QUANTS_PER_ITERATION = 2
const int block_num_y = (nrows + ny - 1) / ny;
const dim3 block_nums(block_num_y, 1, 1);
const dim3 block_dims(32, ny, 1);
dequantize_mul_mat_vec_q2_k<<<block_nums, block_dims, 0, stream>>>(vx, y, dst, ncols, nrows);
}
static void dequantize_mul_mat_vec_q3_K_cuda(const void * vx, const float * y, float * dst, const int ncols, const int nrows, cudaStream_t stream) {
GGML_ASSERT(ncols % QK_K == 0);
const int ny = 2 / K_QUANTS_PER_ITERATION;
const int block_num_y = (nrows + ny - 1) / ny;
const dim3 block_nums(block_num_y, 1, 1);
const dim3 block_dims(32, ny, 1);
dequantize_mul_mat_vec_q3_k<<<block_nums, block_dims, 0, stream>>>(vx, y, dst, ncols, nrows);
}
static void dequantize_mul_mat_vec_q4_K_cuda(const void * vx, const float * y, float * dst, const int ncols, const int nrows, cudaStream_t stream) {
GGML_ASSERT(ncols % QK_K == 0);
const int ny = 2 / K_QUANTS_PER_ITERATION;
const int block_num_y = (nrows + ny - 1) / ny;
const dim3 block_nums(block_num_y, 1, 1);
const dim3 block_dims(32, ny, 1);
dequantize_mul_mat_vec_q4_k<<<block_nums, block_dims, 0, stream>>>(vx, y, dst, ncols, nrows);
}
static void dequantize_mul_mat_vec_q5_K_cuda(const void * vx, const float * y, float * dst, const int ncols, const int nrows, cudaStream_t stream) {
GGML_ASSERT(ncols % QK_K == 0);
const dim3 block_dims(32, 1, 1);
dequantize_mul_mat_vec_q5_k<<<nrows, block_dims, 0, stream>>>(vx, y, dst, ncols);
}
static void dequantize_mul_mat_vec_q6_K_cuda(const void * vx, const float * y, float * dst, const int ncols, const int nrows, cudaStream_t stream) {
GGML_ASSERT(ncols % QK_K == 0);
const int ny = 2 / K_QUANTS_PER_ITERATION;
const int block_num_y = (nrows + ny - 1) / ny;
const dim3 block_nums(block_num_y, 1, 1);
const dim3 block_dims(32, ny, 1);
dequantize_mul_mat_vec_q6_k<<<block_nums, block_dims, 0, stream>>>(vx, y, dst, ncols, nrows);
}
static void convert_mul_mat_vec_f16_cuda(const void * vx, const dfloat * y, float * dst, const int ncols, const int nrows, cudaStream_t stream) {
GGML_ASSERT(ncols % (GGML_CUDA_DMMV_X*2) == 0);
const int block_num_y = (nrows + GGML_CUDA_MMV_Y - 1) / GGML_CUDA_MMV_Y;
const dim3 block_nums(block_num_y, 1, 1);
const dim3 block_dims(WARP_SIZE, GGML_CUDA_MMV_Y, 1);
dequantize_mul_mat_vec<GGML_TYPE_F16>
<<<block_nums, block_dims, 0, stream>>>(vx, y, dst, ncols, nrows);
}
void ggml_cuda_op_dequantize_mul_mat_vec(
ggml_backend_cuda_context & ctx,
const ggml_tensor * src0, const ggml_tensor * src1, ggml_tensor * dst, const char * src0_dd_i, const float * src1_ddf_i,
const char * src1_ddq_i, float * dst_dd_i, const int64_t row_low, const int64_t row_high, const int64_t src1_ncols,
const int64_t src1_padded_row_size, cudaStream_t stream) {
GGML_UNUSED(ctx);
const int64_t ne00 = src0->ne[0];
const int64_t row_diff = row_high - row_low;
GGML_ASSERT(src1->type == GGML_TYPE_F32);
// on some GPUs it is faster to convert src1 to half and to use half precision intrinsics
#ifdef GGML_CUDA_F16
ggml_cuda_pool_alloc<half> src1_dfloat_a(ctx.pool());
half * src1_dfloat = nullptr; // dfloat == half
bool src1_convert_f16 =
src0->type == GGML_TYPE_Q4_0 || src0->type == GGML_TYPE_Q4_1 ||
src0->type == GGML_TYPE_Q5_0 || src0->type == GGML_TYPE_Q5_1 ||
src0->type == GGML_TYPE_Q8_0 || src0->type == GGML_TYPE_F16;
if (src1_convert_f16) {
src1_dfloat = src1_dfloat_a.alloc(ne00);
const to_fp16_cuda_t to_fp16_cuda = ggml_get_to_fp16_cuda(src1->type);
GGML_ASSERT(to_fp16_cuda != nullptr);
to_fp16_cuda(src1_ddf_i, src1_dfloat, ne00, stream);
}
#else
const dfloat * src1_dfloat = (const dfloat *) src1_ddf_i; // dfloat == float, no conversion
#endif // GGML_CUDA_F16
switch (src0->type) {
case GGML_TYPE_Q4_0:
dequantize_mul_mat_vec_q4_0_cuda(src0_dd_i, src1_dfloat, dst_dd_i, ne00, row_diff, stream);
break;
case GGML_TYPE_Q4_1:
dequantize_mul_mat_vec_q4_1_cuda(src0_dd_i, src1_dfloat, dst_dd_i, ne00, row_diff, stream);
break;
case GGML_TYPE_Q5_0:
dequantize_mul_mat_vec_q5_0_cuda(src0_dd_i, src1_dfloat, dst_dd_i, ne00, row_diff, stream);
break;
case GGML_TYPE_Q5_1:
dequantize_mul_mat_vec_q5_1_cuda(src0_dd_i, src1_dfloat, dst_dd_i, ne00, row_diff, stream);
break;
case GGML_TYPE_Q8_0:
dequantize_mul_mat_vec_q8_0_cuda(src0_dd_i, src1_dfloat, dst_dd_i, ne00, row_diff, stream);
break;
case GGML_TYPE_Q2_K:
dequantize_mul_mat_vec_q2_K_cuda(src0_dd_i, src1_ddf_i, dst_dd_i, ne00, row_diff, stream);
break;
case GGML_TYPE_Q3_K:
dequantize_mul_mat_vec_q3_K_cuda(src0_dd_i, src1_ddf_i, dst_dd_i, ne00, row_diff, stream);
break;
case GGML_TYPE_Q4_K:
dequantize_mul_mat_vec_q4_K_cuda(src0_dd_i, src1_ddf_i, dst_dd_i, ne00, row_diff, stream);
break;
case GGML_TYPE_Q5_K:
dequantize_mul_mat_vec_q5_K_cuda(src0_dd_i, src1_ddf_i, dst_dd_i, ne00, row_diff, stream);
break;
case GGML_TYPE_Q6_K:
dequantize_mul_mat_vec_q6_K_cuda(src0_dd_i, src1_ddf_i, dst_dd_i, ne00, row_diff, stream);
break;
case GGML_TYPE_F16:
convert_mul_mat_vec_f16_cuda(src0_dd_i, src1_dfloat, dst_dd_i, ne00, row_diff, stream);
break;
default:
GGML_ABORT("fatal error");
break;
}
GGML_UNUSED(src1);
GGML_UNUSED(dst);
GGML_UNUSED(src1_ddq_i);
GGML_UNUSED(src1_ncols);
GGML_UNUSED(src1_padded_row_size);
}
bool ggml_cuda_dmmv_type_supported(ggml_type src0_type) {
return src0_type == GGML_TYPE_Q4_0 || src0_type == GGML_TYPE_Q4_1 ||
src0_type == GGML_TYPE_Q5_0 || src0_type == GGML_TYPE_Q5_1 ||
src0_type == GGML_TYPE_Q8_0 || src0_type == GGML_TYPE_Q2_K ||
src0_type == GGML_TYPE_Q3_K || src0_type == GGML_TYPE_Q4_K ||
src0_type == GGML_TYPE_Q5_K || src0_type == GGML_TYPE_Q6_K ||
src0_type == GGML_TYPE_F16;
}