From d3d504af718e41719d57131e6fd1a23aba991b65 Mon Sep 17 00:00:00 2001 From: Luke Kenneth Casson Leighton Date: Fri, 28 Dec 2018 02:45:24 +0000 Subject: [PATCH] add notes --- 3d_gpu/microarchitecture.mdwn | 18 ++++++++++++++++++ 1 file changed, 18 insertions(+) diff --git a/3d_gpu/microarchitecture.mdwn b/3d_gpu/microarchitecture.mdwn index 4ab8ce1c2..e68fc9c18 100644 --- a/3d_gpu/microarchitecture.mdwn +++ b/3d_gpu/microarchitecture.mdwn @@ -418,6 +418,24 @@ as you need in 4 cycles for the last operand, then write as much as you can for the result. This simply requires flip-flops to capture the width and then deliver operands in parallel (serial to parallel converter) and similarly for writing. + +---- + +* + +discussion of how to do dest-latches rather than src-latches. + +also includes need for forwarding to achieve it (synonymous with +Tomasulo CDB). + +also, assigning a result number at issue time allows multiple results +to be stored-and-forwarded, meaning that multiplying up the FUs is +not needed. + +also, discussion of how to have multiple instructions issued even with +the same dest reg: drop the reg-store and effectively rename them +to "R.FU#". exceptions under discussion. + # Design Layout ok,so continuing some thoughts-in-order notes: -- 2.30.2