performance refactoring - rebalanced stream buffers accross the device to enforce output buffering rule at compoments boudaries, finally resolved block ram R/W collusion discrepencies,
This commit is contained in:
@@ -42,9 +42,9 @@ module VX_ibuffer #(
|
||||
wire pop = reading && (size_r[i] != 1);
|
||||
|
||||
VX_generic_queue #(
|
||||
.DATAW(DATAW),
|
||||
.SIZE(SIZE),
|
||||
.BUFFERED(1)
|
||||
.DATAW (DATAW),
|
||||
.SIZE (SIZE),
|
||||
.FASTRAM (1)
|
||||
) queue (
|
||||
.clk (clk),
|
||||
.reset (reset),
|
||||
@@ -101,7 +101,7 @@ module VX_ibuffer #(
|
||||
end
|
||||
|
||||
// schedule the next instruction to issue
|
||||
// does round-robin scheduling when multiple warps are present
|
||||
// do round-robin when multiple warps are active
|
||||
always @(*) begin
|
||||
deq_valid_n = 0;
|
||||
deq_wid_n = 'x;
|
||||
|
||||
Reference in New Issue
Block a user