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

Fix undefined variables in debug build, remove unused variables (#531)

Browse source
This commit is contained in:
Stephan Walter 2023-03-26 15:34:02 +00:00 committed by GitHub
parent 8c2ec5e21d
commit 939ad2d3a5
No known key found for this signature in database
GPG key ID: 4AEE18F83AFDEB23
1 changed files with 19 additions and 28 deletions
Show stats Download patch file Download diff file Expand all files Collapse all files

47
ggml.c
View file

@ -1698,8 +1698,6 @@ inline static void ggml_vec_dot_q4_0(const int n, float * restrict s, const void
// Horizontal sum of all lanes of the accumulator
sumf = _mm512_reduce_add_ps( acc0 ) + _mm512_reduce_add_ps( acc1 );
#elif defined(__AVX2__)
const size_t countBlocks = nb;
// Initialize accumulator with zeros
__m256 acc = _mm256_setzero_ps();
@ -5806,23 +5804,28 @@ static void ggml_compute_forward_mul_mat_f32(
const int ne02 = src0->ne[2];
const int ne03 = src0->ne[3];
#if defined(GGML_USE_ACCELERATE) || defined(GGML_USE_OPENBLAS)
const int ne10 = src1->ne[0];
#endif
const int ne11 = src1->ne[1];
//const int ne12 = src1->ne[2];
//const int ne13 = src1->ne[3];
#ifndef NDEBUG
const int ne12 = src1->ne[2];
const int ne13 = src1->ne[3];
//const int ne0 = dst->ne[0];
//const int ne1 = dst->ne[1];
//const int ne2 = dst->ne[2];
//const int ne3 = dst->ne[3];
//const int ne = ne0*ne1*ne2*ne3;
const int ne0 = dst->ne[0];
const int ne1 = dst->ne[1];
const int ne2 = dst->ne[2];
const int ne3 = dst->ne[3];
//const int nb00 = src0->nb[0];
const int nb00 = src0->nb[0];
#endif
const int nb01 = src0->nb[1];
const int nb02 = src0->nb[2];
const int nb03 = src0->nb[3];
#ifndef NDEBUG
const int nb10 = src1->nb[0];
#endif
const int nb11 = src1->nb[1];
const int nb12 = src1->nb[2];
const int nb13 = src1->nb[3];
@ -5840,8 +5843,9 @@ static void ggml_compute_forward_mul_mat_f32(
assert(ne2 == ne12);
assert(ne3 == ne13);
// TODO: we don't support permuted src0
// we don't support permuted src0 or src1
assert(nb00 == sizeof(float));
assert(nb10 == sizeof(float));
// dst cannot be transposed or permuted
assert(nb0 == sizeof(float));
@ -5859,8 +5863,6 @@ static void ggml_compute_forward_mul_mat_f32(
#if defined(GGML_USE_ACCELERATE) || defined(GGML_USE_OPENBLAS)
if (ggml_compute_forward_mul_mat_use_blas(src0, src1, dst)) {
GGML_ASSERT(nb10 == sizeof(float));
if (params->ith != 0) {
return;
}
@ -5903,9 +5905,6 @@ static void ggml_compute_forward_mul_mat_f32(
return;
}
// TODO: do not support transposed src1
assert(nb10 == sizeof(float));
// parallelize by src0 rows using ggml_vec_dot_f32
// total rows in src0
@ -6169,7 +6168,6 @@ static void ggml_compute_forward_mul_mat_q4_0_f32(
const int ne1 = dst->ne[1];
const int ne2 = dst->ne[2];
const int ne3 = dst->ne[3];
//const int ne = ne0*ne1*ne2*ne3;
const int nb00 = src0->nb[0];
const int nb01 = src0->nb[1];
@ -6194,8 +6192,9 @@ static void ggml_compute_forward_mul_mat_q4_0_f32(
GGML_ASSERT(ne2 == ne12);
GGML_ASSERT(ne3 == ne13);
// TODO: we don't support permuted src0
// we don't support permuted src0 or src1
GGML_ASSERT(nb00 == (int) GGML_TYPE_SIZE[GGML_TYPE_Q4_0]);
GGML_ASSERT(nb10 == sizeof(float));
// dst cannot be transposed or permuted
GGML_ASSERT(nb0 == sizeof(float));
@ -6213,8 +6212,6 @@ static void ggml_compute_forward_mul_mat_q4_0_f32(
#if defined(GGML_USE_ACCELERATE) || defined(GGML_USE_OPENBLAS)
if (ggml_compute_forward_mul_mat_use_blas(src0, src1, dst)) {
GGML_ASSERT(nb10 == sizeof(float));
if (params->ith != 0) {
return;
}
@ -6278,8 +6275,6 @@ static void ggml_compute_forward_mul_mat_q4_0_f32(
return;
}
// TODO: do not support transposed src1
// parallelize by src0 rows using ggml_vec_dot_q4_0
// total rows in src0
@ -6354,7 +6349,6 @@ static void ggml_compute_forward_mul_mat_q4_1_f32(
const int ne1 = dst->ne[1];
const int ne2 = dst->ne[2];
const int ne3 = dst->ne[3];
//const int ne = ne0*ne1*ne2*ne3;
const int nb00 = src0->nb[0];
const int nb01 = src0->nb[1];
@ -6379,8 +6373,9 @@ static void ggml_compute_forward_mul_mat_q4_1_f32(
GGML_ASSERT(ne2 == ne12);
GGML_ASSERT(ne3 == ne13);
// TODO: we don't support permuted src0
// we don't support permuted src0 or src1
GGML_ASSERT(nb00 == (int) GGML_TYPE_SIZE[GGML_TYPE_Q4_1]);
GGML_ASSERT(nb10 == sizeof(float));
// dst cannot be transposed or permuted
GGML_ASSERT(nb0 == sizeof(float));
@ -6398,8 +6393,6 @@ static void ggml_compute_forward_mul_mat_q4_1_f32(
#if defined(GGML_USE_ACCELERATE) || defined(GGML_USE_OPENBLAS)
if (ggml_compute_forward_mul_mat_use_blas(src0, src1, dst)) {
GGML_ASSERT(nb10 == sizeof(float));
if (params->ith != 0) {
return;
}
@ -6466,8 +6459,6 @@ static void ggml_compute_forward_mul_mat_q4_1_f32(
return;
}
// TODO: do not support transposed src1
// parallelize by src0 rows using ggml_vec_dot_q4_1
// total rows in src0