mirror of
https://github.com/tinygrad/tinygrad.git
synced 2026-01-09 15:08:02 -05:00
56fa5c1191
* dsp simulator * progress * fix * close on test tiny * working * less waste * line savings * Device DSP compiler * mock DSP at the bottom * DSP tests * docker caching * test update * need load * skip that test for CI DSP * last touch * ugh
28 lines
988 B
Python
28 lines
988 B
Python
from tinygrad import Device
|
|
|
|
# PATH=/opt/homebrew/opt/llvm/bin:$PATH python3 extra/dsp/opt.py
|
|
|
|
if __name__ == "__main__":
|
|
compiler = Device["DSP"].compiler
|
|
|
|
lib = compiler.compile("""
|
|
typedef long HVX_Vector __attribute__((__vector_size__(128))) __attribute__ ((aligned(128)));
|
|
typedef long HVX_VectorPair __attribute__((__vector_size__(256))) __attribute__ ((aligned(256)));
|
|
|
|
void test(unsigned char *c, unsigned char *a, unsigned char *b) {
|
|
HVX_Vector t0 = *(HVX_Vector*)a;
|
|
//HVX_VectorPair t1 = *((HVX_VectorPair*)b);
|
|
HVX_Vector acc = __builtin_HEXAGON_V6_vd0_128B();
|
|
for (int i = 0; i < 128; i++) {
|
|
//__builtin_HEXAGON_V6_lvsplatb_128B(t0[i])
|
|
//acc += __builtin_HEXAGON_V6_lvsplatb_128B(t0[i]) * t1;
|
|
//acc += t0[i] * t1;
|
|
unsigned int t1 = ((unsigned int *)b)[i];
|
|
//acc = __builtin_HEXAGON_V6_vrmpyub_acc_128B(acc, t0, t1);
|
|
acc = __builtin_HEXAGON_V6_vrmpybus_acc_128B(acc, t0, t1);
|
|
}
|
|
*((HVX_Vector*)c) = acc;
|
|
}""")
|
|
|
|
compiler.disassemble(lib)
|