In the image, a computer is set up on a large projection screen, displaying information for a group of people in a room. The projection screen is behind a young boy, who is likely giving the presentation. The presentation appears to be on a laptop, which is connected to the projector and displaying the content on the large screen. The boy is standing in front of the screen, facing the audience and gesturing with his hands, possibly answering a question or addressing his listeners. The room appears to be well-lit, with the focus on the young boy and the projected screen.
Text transcribed from the image:
Asynchrony: Overlapping GEMM and Softmax
Why overlapping?
Ex: headdim 128, block size 128 x 192
FP16 WGMMA: 2 x 2 x 128 x 192 x 128 = 12.6 MFLOPS, 4096 FLOPS/cycle -> 3072 cycles
MUFU.EX2: 128 x 192 = 24.6k OPS, 16 OPS/cycle -> 1536 cycles
A