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

upgraded clblast

Browse source
This commit is contained in:
Concedo 2023-05-25 10:18:12 +08:00
parent 37a34deaa0
commit d2da155661
5 changed files with 8 additions and 13 deletions
Download patch file Download diff file Expand all files Collapse all files

BIN
clblast.dll
View file

Binary file not shown.

6
include/clblast.h
View file

@ -39,10 +39,10 @@
#define PUBLIC_API #define PUBLIC_API
#endif #endif
// Version numbering (v1.5.3) // Version numbering (v1.6.0)
#define CLBLAST_VERSION_MAJOR 1 #define CLBLAST_VERSION_MAJOR 1
#define CLBLAST_VERSION_MINOR 5 #define CLBLAST_VERSION_MINOR 6
#define CLBLAST_VERSION_PATCH 3 #define CLBLAST_VERSION_PATCH 0
namespace clblast { namespace clblast {
// ================================================================================================= // =================================================================================================

6
include/clblast_c.h
View file

@ -34,10 +34,10 @@
#define PUBLIC_API #define PUBLIC_API
#endif #endif
// Version numbering (v1.5.3) // Version numbering (v1.6.0)
#define CLBLAST_VERSION_MAJOR 1 #define CLBLAST_VERSION_MAJOR 1
#define CLBLAST_VERSION_MINOR 5 #define CLBLAST_VERSION_MINOR 6
#define CLBLAST_VERSION_PATCH 3 #define CLBLAST_VERSION_PATCH 0
// The C interface // The C interface
#ifdef __cplusplus #ifdef __cplusplus

9
include/clblast_half.h
View file

@ -18,11 +18,6 @@
#ifndef CLBLAST_HALF_H_ #ifndef CLBLAST_HALF_H_
#define CLBLAST_HALF_H_ #define CLBLAST_HALF_H_
// MSVC 2013 doesn't fully support C99
#ifdef _MSC_VER
#define inline __inline
#endif
// ================================================================================================= // =================================================================================================
// The host data-type for half-precision floating-point (16-bit) is based on the `cl_half` OpenCL // The host data-type for half-precision floating-point (16-bit) is based on the `cl_half` OpenCL
@ -40,7 +35,7 @@ typedef union ConversionBits_ {
// Converts a IEEE-compliant single-precision value to half-precision floating-point. This function // Converts a IEEE-compliant single-precision value to half-precision floating-point. This function
// applies simple truncation (round toward zero, but with overflows set to infinity) as rounding // applies simple truncation (round toward zero, but with overflows set to infinity) as rounding
// mode. // mode.
inline half FloatToHalf(const float value) { static half FloatToHalf(const float value) {
static const unsigned short base_table[512] = { static const unsigned short base_table[512] = {
0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000,
0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000, 0x0000,
@ -101,7 +96,7 @@ inline half FloatToHalf(const float value) {
} }
// Converts a half-precision value to IEEE-compliant single-precision floating-point // Converts a half-precision value to IEEE-compliant single-precision floating-point
inline float HalfToFloat(const half value) { static float HalfToFloat(const half value) {
static const unsigned int mantissa_table[2048] = { static const unsigned int mantissa_table[2048] = {
0x00000000, 0x33800000, 0x34000000, 0x34400000, 0x34800000, 0x34A00000, 0x34C00000, 0x34E00000, 0x35000000, 0x35100000, 0x35200000, 0x35300000, 0x35400000, 0x35500000, 0x35600000, 0x35700000, 0x00000000, 0x33800000, 0x34000000, 0x34400000, 0x34800000, 0x34A00000, 0x34C00000, 0x34E00000, 0x35000000, 0x35100000, 0x35200000, 0x35300000, 0x35400000, 0x35500000, 0x35600000, 0x35700000,
0x35800000, 0x35880000, 0x35900000, 0x35980000, 0x35A00000, 0x35A80000, 0x35B00000, 0x35B80000, 0x35C00000, 0x35C80000, 0x35D00000, 0x35D80000, 0x35E00000, 0x35E80000, 0x35F00000, 0x35F80000, 0x35800000, 0x35880000, 0x35900000, 0x35980000, 0x35A00000, 0x35A80000, 0x35B00000, 0x35B80000, 0x35C00000, 0x35C80000, 0x35D00000, 0x35D80000, 0x35E00000, 0x35E80000, 0x35F00000, 0x35F80000,

BIN
lib/clblast.lib
View file

Binary file not shown.