@FIR-770 - LLama.cpp: Adding Perf Status for all GGML Operation

[akapoor3518]

Perf Summary:
=== GGML Perf Summary ===
ADD : 172 runs, 63454 us total, avg 368.92 us
MUL : 133 runs, 6807770 us total, avg 51186.24 us
RMS_NORM : 178 runs, 3009 us total, avg 16.90 us
MUL_MAT : 676 runs, 6675856 us total, avg 9875.53 us
CPY : 168 runs, 1123 us total, avg 6.68 us
CONT : 87 runs, 344 us total, avg 3.95 us
RESHAPE : 316 runs, 284 us total, avg 0.90 us
VIEW : 294 runs, 67 us total, avg 0.23 us
PERMUTE : 294 runs, 105 us total, avg 0.36 us
TRANSPOSE : 68 runs, 31 us total, avg 0.46 us
GET_ROWS : 9 runs, 127 us total, avg 14.11 us
SOFT_MAX : 88 runs, 5991 us total, avg 68.08 us
ROPE : 175 runs, 3593 us total, avg 20.53 us
UNARY : 22 runs, 8535113 us total, avg 387959.68 us

Detailed information is written to the file, including the size (shape) of each tensor node. Below is a sample output:
cat ggml_perf.log
ggml_graph_compute_perf: total compute time: 22096.867 ms

• BACKEND:CPU OP:GET_ROWS: total 0.110 ms over 4 runs (avg 0.028 ms) [shape=2048,6,1]
• BACKEND:CPU OP:RMS_NORM: total 0.148 ms over 4 runs (avg 0.037 ms) [shape=2048,6,1]
• BACKEND:TSAVORITE OP:MUL: total 143.046 ms over 1 runs (avg 143.046 ms) [shape=2048,6,1]
• BACKEND:CPU OP:MUL_MAT: total 34.345 ms over 4 runs (avg 8.586 ms) [shape=2048,6,1]
• BACKEND:CPU OP:RESHAPE: total 0.005 ms over 2 runs (avg 0.003 ms) [shape=64,32,6]
• BACKEND:CPU OP:ROPE: total 0.329 ms over 4 runs (avg 0.082 ms) [shape=64,32,6]
• BACKEND:CPU OP:MUL_MAT: total 4.916 ms over 4 runs (avg 1.229 ms) [shape=256,6,1]
• BACKEND:CPU OP:RESHAPE: total 0.002 ms over 4 runs (avg 0.001 ms) [shape=64,4,6]
• BACKEND:CPU OP:ROPE: total 0.032 ms over 4 runs (avg 0.008 ms) [shape=64,4,6]
• BACKEND:CPU OP:MUL_MAT: total 4.840 ms over 4 runs (avg 1.210 ms) [shape=256,6,1]
• BACKEND:CPU OP:RESHAPE: total 0.002 ms over 3 runs (avg 0.001 ms) [shape=64,4,6]
• BACKEND:CPU OP:VIEW: total 0.007 ms over 4 runs (avg 0.002 ms) [shape=1536,1,1]
• BACKEND:CPU OP:CPY: total 0.021 ms over 4 runs (avg 0.005 ms) [shape=1536,1,1]
• BACKEND:CPU OP:RESHAPE: total 0.000 ms over 3 runs (avg 0.000 ms) [shape=256,6,1]
• BACKEND:CPU OP:TRANSPOSE: total 0.002 ms over 4 runs (avg 0.001 ms) [shape=6,256,1]
• BACKEND:CPU OP:VIEW: total 0.000 ms over 4 runs (avg 0.000 ms) [shape=6,256,1]
• BACKEND:CPU OP:CPY: total 0.040 ms over 4 runs (avg 0.010 ms) [shape=6,256,1]
• BACKEND:CPU OP:VIEW: total 0.001 ms over 4 runs (avg 0.000 ms) [shape=32,4,64]
• BACKEND:CPU OP:PERMUTE: total 0.004 ms over 4 runs (avg 0.001 ms) [shape=32,64,4]

[LewisLui777]

Great job this looks good.

[atrivedi-tsavoritesi]

Most of the changes look good, just few minor indentation to be taken care of. Feel free to push the changes once indentation is addressed.