Signed-off-by: Jared Van Bortel <jared@nomic.ai> Co-authored-by: niansa <anton-sa@web.de> Co-authored-by: Adam Treat <treat.adam@gmail.com> Co-authored-by: Aaron Miller <apage43@ninjawhale.com> Co-authored-by: ToKiNoBug <tokinobug@163.com> Co-authored-by: Georgi Gerganov <ggerganov@gmail.com> Co-authored-by: slaren <slarengh@gmail.com>
18 lines
609 B
Plaintext
18 lines
609 B
Plaintext
void main() {
|
|
const uint i = gl_WorkGroupID.x;
|
|
const int r = inB[i + pcs.inBOff];
|
|
|
|
int z = 0;
|
|
for (uint ind = gl_LocalInvocationID.x; ind < pcs.ne00/16; ind += gl_WorkGroupSize.x) {
|
|
const uint inIndex = (r * pcs.nb01 + pcs.inAOff) + ind/NL * SIZE_OF_BLOCK;
|
|
const mat4 result = dequantize_block(inIndex, ind%NL);
|
|
for (uint j = 0; j < 4; ++j) {
|
|
for (uint k = 0; k < 4; ++k) {
|
|
const uint outIndex = i * pcs.nb1/BYTES_FOR_TYPE + pcs.outOff + z;
|
|
out_[outIndex] = result[j][k];
|
|
++z;
|
|
}
|
|
}
|
|
}
|
|
}
|