diff options
| author | Tim Dettmers <tim.dettmers@gmail.com> | 2022-07-26 19:38:01 -0700 |
|---|---|---|
| committer | Tim Dettmers <tim.dettmers@gmail.com> | 2022-07-26 19:38:01 -0700 |
| commit | 5737f2b027a1e0ec8540a3aa914632d44ad9c62d (patch) | |
| tree | b288c905eaba75dc6b43a8bcebc82720c16e4816 /csrc/pythonInterface.c | |
| parent | 47a73d94c3d3284f6073b0ff189ed5bc9e3a8762 (diff) | |
| parent | dc8c9efdb33130f960adc864916b67d0cb744dbb (diff) | |
Merge branch 'patch_merge' into extract_outliers
Diffstat (limited to 'csrc/pythonInterface.c')
| -rw-r--r-- | csrc/pythonInterface.c | 6 |
1 files changed, 3 insertions, 3 deletions
diff --git a/csrc/pythonInterface.c b/csrc/pythonInterface.c index 2ecbaae..7356c11 100644 --- a/csrc/pythonInterface.c +++ b/csrc/pythonInterface.c @@ -82,7 +82,6 @@ void quantizeBlockwise_stochastic_fp32(float * code, float *A, float *absmax, un void dequantizeBlockwise_fp16(float *code, unsigned char *A, float *absmax, half *out, int blocksize, const int n){ dequantizeBlockwise<half>(code, A, absmax, out, blocksize, n); } \ void dequantizeBlockwise_fp32(float *code, unsigned char *A, float *absmax, float *out, int blocksize, const int n){ dequantizeBlockwise<float>(code, A, absmax, out, blocksize, n); } -#endif #define MAKE_FUNC_TRANSFORM(fbits, fsrc, ftrgt, ftranspose, dtype, src, target, transpose, bits) \ void transform_##fbits##_##fsrc##_to_##ftrgt##_##ftranspose(cublasLtHandle_t ltHandle, dtype *A, dtype *out, int dim1, int dim2) \ @@ -132,10 +131,11 @@ void spmm_coo_very_sparse_naive_fp16(int *max_count, int *max_idx, int *offset_r void spmm_coo_very_sparse_naive_int8(int *max_count, int *max_idx, int *offset_rowidx, int *rowidx, int *colidx, half *values, signed char *B, half *out, float *dequant_stats, int nnz_rows, int nnz, int rowsA, int rowsB, int colsB) { spmm_coo_very_sparse_naive<signed char, 8>(max_count, max_idx, offset_rowidx, rowidx, colidx, values, B, out, dequant_stats, nnz_rows, nnz, rowsA, rowsB, colsB); } +#endif extern "C" { - #if BUILD_CUDA +#if BUILD_CUDA void cestimate_quantiles_fp32(float *A, float *code, float offset, int n){ estimateQuantiles_fp32(A, code, offset, n); } void cestimate_quantiles_fp16(half *A, float *code, float offset, int n){ estimateQuantiles_fp16(A, code, offset, n); } void cquantize(float *code, float *A, unsigned char *out, int n){ quantize(code, A, out, n); } @@ -231,7 +231,7 @@ extern "C" { return igemmlt_ampere_8_rowscale((cublasLtHandle_t) context->m_handle, m, n, k, A, B, C, row_scale, lda, ldb, ldc); } int cigemmlt_ampere_8(Context *context, int m, int n, int k, const int8_t *A, const int8_t *B, void *C, float *row_scale, int lda, int ldb, int ldc) - { return igemmlt_ampere_8_rowscale((cublasLtHandle_t) context->m_handle, m, n, k, A, B, C, row_scale, lda, ldb, ldc); } + { return igemmlt_ampere_8((cublasLtHandle_t) context->m_handle, m, n, k, A, B, C, row_scale, lda, ldb, ldc); } #define MAKE_FUNC_CTRANSFORM(fbits, fsrc, ftrgt, ftranspose, dtype, src, target, transpose, bits) \ void ctransform_##fbits##_##fsrc##_to_##ftrgt##_##ftranspose(Context *context, dtype *A, dtype *out, int dim1, int dim2) \ |