GPU L10: Memory

00:01:31.572,00:01:34.572 Kshitij Bipin Deogade cs17b104: Spatial and temporal 00:01:33.934,00:01:36.934 Shagnik Pal ee17b147: temporal and spacial 00:09:34.074,00:09:37.074 Akash Haridas ae17b020: 2nd one has more spatial locality due to row major 00:09:39.892,00:09:42.892 SHETH DEV YASHPAL CS17B106: first: C is temporal local, A is spatial 00:09:44.580,00:09:47.580 Shubham Kashyapi: First is temporal, second has spatial 00:09:50.388,00:09:53.388 Anumala Venu Madhava Reddy cs18b051: tempral locality is not useful in 2nd one 00:10:02.832,00:10:05.832 SHETH DEV YASHPAL CS17B106: 2nd one has A has temporal local, B,C spatial local 00:24:59.045,00:25:02.045 Anumala Venu Madhava Reddy cs18b051: partial coalesced 00:25:05.736,00:25:08.736 Prasoon Mishra CS20S028: Semi-coalesced ? 00:25:16.409,00:25:19.409 Shagnik Pal ee17b147: Coalesced since overall they all are consecutive 00:25:19.530,00:25:22.530 Sumit Negi cs20m067: coalesced 00:25:19.677,00:25:22.677 Nishant Prabhu me17b084: coalesced 00:30:31.436,00:30:34.436 Rigved Sah cs20m053: Is memory coalescing only for mem read or for write also? 00:30:44.589,00:30:47.589 Sheera Shamsu CS20D001: So in the third case if we access some a[17], then all a[0]....a[31] will be brought to cache? And in Firat case, if we access a[0]....all a[0]...a[31] will be bring to cache? 00:35:41.257,00:35:44.257 SANKET NEEMA cs19m055: no sot 00:35:44.329,00:35:47.329 SANKET NEEMA cs19m055: sir* 00:35:52.942,00:35:55.942 Rigved Sah cs20m053: 32 00:35:53.846,00:35:56.846 Jeswant Krishna ae17b030: 32 00:35:54.336,00:35:57.336 Jeeva Keshav S ee20s057: 32? 00:35:54.428,00:35:57.428 Sumit Negi cs20m067: 32 00:35:54.982,00:35:57.982 Arihant Samar cs18b052: 32 00:35:54.996,00:35:57.996 Anumala Venu Madhava Reddy cs18b051: 32 00:40:01.804,00:40:04.804 Shubham Kashyapi: __global__ void dkernel(int degree, int* a){ int accesses = 33 - degree; for(int i=0; i lessthan accesses; i++){ a[32*i] = 1; } return; } 00:41:11.330,00:41:14.330 Patlolla Bharath Simha Reddy cs18b034: a[threadidx.x * num] 00:41:18.436,00:41:21.436 Harshit Kedia cs17b103: A[threadIdx.x * (33-DoC)] 00:41:54.189,00:41:57.189 SANKET NEEMA cs19m055: case 1: i=1 case 00:42:19.160,00:42:22.160 SANKET NEEMA cs19m055: case 1: i=1 00:42:23.896,00:42:26.896 SANKET NEEMA cs19m055: sorry sir 00:44:21.649,00:44:24.649 SANKET NEEMA cs19m055: case 1: i=1 for(j=0;j lessthan 32;j++) a[k+i*j]] case 2 i=2 and so on in 32 case 00:48:11.691,00:48:14.691 Shubham Kashyapi: Is Sir speaking? 00:48:20.740,00:48:23.740 Prasoon Mishra CS20S028: No 00:49:00.913,00:49:03.913 Jeeva Keshav S ee20s057: can you explain the stride part again? 00:49:19.688,00:49:22.688 Rupesh Nasre.: a[threadIdx.x * N] 00:50:58.794,00:51:01.794 Jeeva Keshav S ee20s057: thank you sir.