@@ -71,7 +71,7 @@ Hyesun Hong,
7171
7272* PIM/PNM technology enables computation directly on memory
7373* Prevents data movement improving performance and reducing consumption
74- * PIM operates directly on memory banks by reading and storing on rows and columns
74+ * Operates directly on memory banks by reading and storing on rows and columns
7575* Aquabolt-XL is the first demonstrator
7676* Can be drop in on any memory controller
7777* CXL-PNM is the CXL variant for PNM, can work with multiple PIM
@@ -116,10 +116,12 @@ SYCL Extension for PIM/PNM
116116
117117* Extension for PNM
118118
119- * Added new overloads of joint_exclusive_scan, joint_inclusive_scan, reduce_over_group
119+ * Added new overloads of joint_exclusive_scan,
120+ * joint_inclusive_scan, reduce_over_group
120121
121- * PNM standalone has less opportunity for parallelism, also limited by memory controller
122+ * PNM standalone has less opportunity for parallelism
122123
124+ * limited by memory controller
123125 * -> Combine PNM and PIM, PNM generates commands for PIM blocks
124126
125127* Two modes
151153
152154 * Can also be used with other hardware.
153155 * Semantics not PIM-specific, but translation of C++ to SYCL
154- * Can also map nicely to other types of hardware, for example vector processor
156+ * Can also map nicely to other types of hardware, e.g. vector processor
155157
156158* Why have the user explicitly specify a block-size?
157159
0 commit comments