Changes between Version 27 and Version 28 of GcnTimings
- Timestamp:
- 12/29/17 14:00:36 (6 years ago)
Legend:
- Unmodified
- Added
- Removed
- Modified
-
GcnTimings
v27 v28 141 141 <li>only the first 3 dwords in the 32-byte block incur no penalty. Any 2-dword 142 142 instruction outside these first 3 dwords adds a single penalty.</li> 143 <li>if instructions is longer (more than four cycles) thenlast cycles/4 dwords are free</li>143 <li>if the instructions is longer (more than four cycles) then the last cycles/4 dwords are free</li> 144 144 <li>if 16 or more cycle 2-dword instruction and 2-dword instruction in 4 dword, then there is 145 no penalty for second 2-dword instruction.</li>146 <li>best place to jump is 5 first dwords in 32-byte block. Jump to rest ofdwords causes147 1-3 penalties, depending on number of dwords (N-4, where N is adword number). This rule145 no penalty for the second 2-dword instruction.</li> 146 <li>best place to jump is the 5 first dwords in the 32-byte block. Jump to rest of the dwords causes 147 1-3 penalties, depending on number of dwords (N-4, where N is the dword number). This rule 148 148 does not apply to backward jumps (???)</li> 149 <li>any conditional jump instruction should be in first half of 32-byte block, otherwise149 <li>any conditional jump instruction should be in first half of the 32-byte block, otherwise 150 150 1-4 penalties are added if jump is not taken, depending on dword number (N-3, where N is dword number).</li> 151 151 </ul> 152 <p>IMPORTANT: If the occupancy is greater than 1 wave per compute unit, then penalties,152 <p>IMPORTANT: If the occupancy is greater than 1 wave per compute unit, then the penalties, 153 153 branches, and scalar instructions will be masked while executing 154 154 more waves than 4*CUs. For best results is recommended to execute many waves … … 1541 1541 </table> 1542 1542 <p>About bank conflicts: The LDS memory is partitioned in 32 banks. The bank number is in 1543 bits 2-6 of the address. A bank conflict occurs when two addresses hit same1544 bank, but addresses are different starting from7bit1543 bits 2-6 of the address. A bank conflict occurs when two addresses hit the same 1544 bank, but the addresses are different starting from the 7bit 1545 1545 (the first 2 bits of the address doesn't matter). 1546 1546 Any bank conflict adds penalty to timing and throughput. In the worst case, the throughput