AI/rocm-systems
2
0
Fork 0
Code Issues Pull Requests Actions Packages Projects Releases Wiki Activity

SWDEV-432174 - Change the fillBuffer kernel

Browse Source 
- Add the new fillBuffer kernel, which allows to launch a limited
number of workgroups for memory fill operation
- Switch fill memory to 16 bytes write by default
- Allow to limit the workgroups with DEBUG_CLR_LIMIT_BLIT_WG

Change-Id: Ibad1822f2d42b2fc71bcfc1917c31409c0623e8e


[ROCm/clr commit: f1dc81f427]
This commit is contained in:
German Andryeyev
2023-11-14 12:49:17 -05:00
parent e1889b77b4
commit e390ec044f
13 changed files with 283 additions and 214 deletions
Download Patch File Download Diff File Expand all files Collapse all files
projects/clr/rocclr/device/blitcl.cpp
+51 -10
View File
@@ -47,16 +47,57 @@ const char* BlitLinearSourceCode = BLIT_KERNELS(
extern void __ockl_gws_init(uint nwm1, uint rid);
// Implementation
__kernel void __amd_rocclr_fillBufferAligned(__global uchar* bufUChar,
__global ushort* bufUShort,
__global uint* bufUInt,
__global ulong* bufULong,
__constant uchar* pattern,
uint patternSize, ulong offset,
ulong size) {
__amd_fillBufferAligned(bufUChar, bufUShort, bufUInt, bufULong,
pattern, patternSize, offset, size);
__kernel void __amd_rocclr_fillBufferAligned(
__global uchar* bufUChar, __global ushort* bufUShort, __global uint* bufUInt,
__global ulong* bufULong, __global ulong2* bufULong2, __constant uchar* pattern,
uint pattern_size, ulong offset, ulong end_ptr, uint next_chunk) {
int id = get_global_id(0);
long cur_id = offset + id * pattern_size;
if (bufULong2) {
__global ulong2* element = &bufULong2[cur_id];
__constant ulong2* pt = (__constant ulong2*)pattern;
while ((ulong)element < end_ptr) {
for (uint i = 0; i < pattern_size; ++i) {
element[i] = pt[i];
}
element += next_chunk;
}
} else if (bufULong) {
__global ulong* element = &bufULong[cur_id];
__constant ulong* pt = (__constant ulong*)pattern;
while ((ulong)element < end_ptr) {
for (uint i = 0; i < pattern_size; ++i) {
element[i] = pt[i];
}
element += next_chunk;
}
} else if (bufUInt) {
__global uint* element = &bufUInt[cur_id];
__constant uint* pt = (__constant uint*)pattern;
while ((ulong)element < end_ptr) {
for (uint i = 0; i < pattern_size; ++i) {
element[i] = pt[i];
}
element += next_chunk;
}
} else if (bufUShort) {
__global ushort* element = &bufUShort[cur_id];
__constant ushort* pt = (__constant ushort*)pattern;
while ((ulong)element < end_ptr) {
for (uint i = 0; i < pattern_size; ++i) {
element[i] = pt[i];
}
element += next_chunk;
}
} else {
__global uchar* element = &bufUChar[cur_id];
while ((ulong)element < end_ptr) {
for (uint i = 0; i < pattern_size; ++i) {
element[i] = pattern[i];
}
element += next_chunk;
}
}
}
__kernel void __amd_rocclr_fillBufferAligned2D(__global uchar* bufUChar,