`
`-
`
`fix for dropped real time flag: moved export_pos bit into the "flags" field
`(it was using a bit that was set aside for the extra in field)
`
`Change 140350 on 2003/12/31 by rramsey@rramsey_xenos linux orl
`
`Fix a bug that was allowing the texconst mem to be written when full (no phys addr
`available)
`
`Change 140331 on 2003/12/31 by mearl@mearl xenos linux orl
`
`Fixed bug; was using thread type instead of fetch type.
`
`Change 140330 on 2003/12/31 by tien@tien r500 emu
`
`Bug fix for sec dim calcs
`
`Change 140316 on 2003/12/31 by mmantor@FL mmantorLT r400 win
`
`<special build with a sq_alu.h provided by tom to address dot lsb mismatches>
`
`Change 140313 on 2003/12/31 by jcarroll@jcarroll r400 win
`
`added latest status; picked new test
`
`Change 140284 on 2003/12/30 by dclifton@dclifton xenos linux orl
`
`Fixed scalar MOVA FLOOR opcode;
`
`Change 140270 on 2003/12/30 by danh@danh r400 win
`
`Status update.
`
`Change 140266 on 2003/12/30 by smoss@smoss crayola linux orl regress
`
`changes for simple register indirect test (register read)
`
`Change 140247 on 2003/12/30 by vbhatia@vbhatia r400 linux marlboro
`
`Reverted fetch multisample test clamp, for the Team believes that it might be best to
`support it
`
`Change 140219 on 2003/12/30 by mdesai@mdesai r400 linux
`
`During tfetch, using msb of texel offset_pre
`
`Change 140218 on 2003/12/30 by vbhatia@vbhatia r400 linux marlboro
`
`AMO CONFIDENTIAL BUSINESS INFORMATION - SUBJECT TO THE PROTECTIVE ORDER
`
`AMD1044_0179855
`
`ATI Ex. 2036
`
`ATI Ex. 2036
`IPR2023-00922
`Page 1 of 337
`
`
`
`Optimised away fetch multi sample tests to be overridded as tfetches, not supported
`
`Change 140205 on 2003/12/30 by rramsey@RRAMSEY_P4 r400 win
`
`update test status, pick another one
`
`Change 140203 on 2003/12/30 by danh@danh r400 win
`
`Status update.
`
`Change 140180 on 2003/12/30 by vbhatia@vbhatia r400 linux marlboro
`
`Added clamp so as to test only point sampling when doing vertex fetches
`
`Change 140174 on 2003/12/30 by rramsey@RRAMSEY_P4 r400 win
`
`update status
`
`Change 140155 on 2003/12/30 by rramsey@RRAMSEY_P4 r400 win
`
`update with 12/30 regression results
`
`Change 140117 on 2003/12/30 by danh@danh r400 win
`
`Status update.
`
`Change 140102 on 2003/12/30 by mmantor@FL mmantorLT r400 win
`
`<special build with sq_alu.h#79 for debug>
`
`Change 140051 on 2003/12/29 by mearl@mearl r400 win
`
`update status
`
`Change 140050 on 2003/12/29 by vromaker@vromaker r400 linux marlboro
`
`status update - emulator fix was made for scalar const opcodes
`
`Change 140036 on 2003/12/29 by jcarroll@jcarroll r400 win
`
`Updated jcarroll status
`
`Change 139967 on 2003/12/29 by tien@tien r400 devel marlboro
`
`Added another delay on compare to mirror added pipestage in RTL
`
`Change 139957 on 2003/12/29 by tien@tien r500 emu
`
`AMO CONFIDENTIAL BUSINESS INFORMATION - SUBJECT TO THE PROTECTIVE ORDER
`
`AMD1044_0179856
`
`ATI Ex. 2036
`
`ATI Ex. 2036
`IPR2023-00922
`Page 2 of 337
`
`
`
`Recoded log pitches going to TCF for better timing
`Closing GetBorderColorFraction bug .. verified tests passing
`Added stage to tp addresser and refreshed .bvrls to be safe (tp_parameters dependency)
`
`Change 139909 on 2003/12/28 by mmantor@mmantor xenos linux orl
`
`<timing fixes>
`
`Change 139836 on 2003/12/24 by jayw@jayw_r400 linux_marlboro
`
`Depth code clean up part 1/38103. Fix for ignoring SXl rtr. Non-hiz depth still has
`bug with expansion shown by rare gc r400rb zwave failures.
`
`Change 139795 on 2003/12/23 by danh@danh r400 win
`
`Status update.
`
`Change 139793 on 2003/12/23 by vbhatia@vbhatia r400 linux marlboro
`
`Added clamp for no loci grad when vertex fetches
`
`Change 139789 on 2003/12/23 by mdesai@mdesai r400 linux
`
`Fixed Addresser random test genRandll (large denorms with offsets)
`
`Change 139383 on 2003/12/23 by ctaylor@ctaylor xenos linux orl
`
`Fixed bug in control flow sequencer where when thread was put back onto thread
`buffer due to alloc cfi, the no-serialize bit was being taken from bit 40 of the cfs
`opcode instead of the execute state machine opcode so it was the right bit from the
`
`wrong instruction. Things have been working mostly due to the fact that bit 40 of most
`of the other CFI opcodes is reserved and therefore 0.
`
`Change 139373 on 2003/12/23 by llefebvr@llefebvr r400 emu montreal
`
`updated status for SX->PA missmatches.
`
`Change 139356 on 2003/12/23 by bhankins@bhankins xenos linux orl
`
`Move some assignments out of process where they shouldn't be. No functional change.
`
`Change 139338 on 2003/12/23 by vromaker@vromaker r400 linux marlboro
`
`-
`
`fix for scalar const ops: y and x swizzle fields used for gpr address bits [5:4]
`and [3:2] were swapped
`
`Change 139327 on 2003/12/23 by rramsey@rramsey_xenos linux orl
`
`AMO CONFIDENTIAL BUSINESS INFORMATION - SUBJECT TO THE PROTECTIVE ORDER
`
`AMD1044_0179857
`
`ATI Ex. 2036
`
`ATI Ex. 2036
`IPR2023-00922
`Page 3 of 337
`
`
`
`add simd_id to spout mismatch message
`make sx rb color tracker multi-threaded per sx/rb interface
`
`Change 139310 on 2003/12/23 by vromaker@vromaker r400 linux marlboro
`
`status update -
`
`took another test
`
`Change 139296 on 2003/12/23 by tien@tien r500 emu
`
`Fixed GetBorderColorFraction DATA FORMAT override
`
`Change 139280 on 2003/12/23 by rmanapat@rmanapat r400 release
`
`Fix for bug 3126
`
`Change 139223 on 2003/12/22 by llefebvr@llefebvr r400 emu montreal
`
`Now working on SX->PA missmatches.
`
`Change 139193 on 2003/12/22 by rramsey@RRAMSEY_P4 r400 win
`
`update status
`
`Change 139142 on 2003/12/22 by llefebvr@llefebvr r400 emu montreal
`
`The SX->RB tracker is bad. Added a comment to explain the situation.
`
`Change 139097 on 2003/12/22 by ctaylor@ctaylor xenos linux orl
`
`Fixed bug related to clamping of GPR addresses which are out of range. Old code
`
`clamped to absolute zero instead of the base for the current thread.
`
`Change 139066 on 2003/12/22 by bhankins@bhankins xenos linux orl
`
`Add tbtrk sx_bc quad tracker (Cl version only)
`
`Change 139057 on 2003/12/22 by rramsey@RRAMSEY_P4 r400 win
`
`update more tests, pick another one
`
`Change 139050 on 2003/12/22 by rramsey@RRAMSEY_P4 r400 win
`
`Update with new regression results, pick a test
`
`Change 139045 on 2003/12/22 by rramsey@rramsey_xenos linux orl
`
`Fix phasing of thread count in sq_ais output.
`
`AMO CONFIDENTIAL BUSINESS INFORMATION - SUBJECT TO THE PROTECTIVE ORDER
`
`AMD1044_0179858
`
`ATI Ex. 2036
`
`ATI Ex. 2036
`IPR2023-00922
`Page 4 of 337
`
`
`
`Fix o gprsm_busy from sq_vtx ctl and change thread counter to only
`reset on RST VTX CNT event.
`
`Change 139044 on 2003/12/22 by bhankins@bhankins xenos linux orl
`
`Initial add to r400 branch
`
`Change 138986 on 2003/12/20 by mmantor@mmantor xenos linux orl
`
`<changed csim to only make one pass for param gen and gen index and write the dump
`files correctly, fixed a timing loop in pix tthread buffer>
`
`Change 138964 on 2003/12/20 by donaldl@donaldl xenos linux orl
`
`Redundancy bug fixes --
`1. Added RSP output muxes to TP SP data valid[3:0] in tp output.v.
`2. Created I/O port TPC TP simd_id[l:0]. Needed so simd id lines up
`
`with TP outputs going to SPs and RSP.
`
`Change 138960 on 2003/12/20 by bhankins@bhankins xenos linux orl
`
`Forgot to make the same change to the BCl side. Has no affect on r400 version.
`
`Change 138958 on 2003/12/20 by bhankins@bhankins xenos linux orl
`
`fix miswiring of quad_x and quad_y outputs of detailed quad fifo
`
`Change 138921 on 2003/12/19 by tien@tien r400 devel marlboro
`
`Man if this breaks anything I'll freak.
`Changed the format comp* mux control for data format=61
`
`(FMT ... AS 1 1 1 1)
`
`Change 138662 on 2003/12/19 by mearl@mearl r400 win
`
`update status
`
`Change 138650 on 2003/12/19 by danh@danh r400 win
`
`Status update.
`
`Change 138647 on 2003/12/19 by bhankins@bhankins xenos linux orl
`
`Add include of sx defines.v. No affect on r400 version.
`
`Change 138609 on 2003/12/19 by tien@tien r500 emu
`
`Fixes location of ANISO BIAS and DIM fields in emu and rtl
`Closing out GetWeights bug, which I fixed last checkin
`
`AMO CONFIDENTIAL BUSINESS INFORMATION - SUBJECT TO THE PROTECTIVE ORDER
`
`AMD1044_0179859
`
`ATI Ex. 2036
`
`ATI Ex. 2036
`IPR2023-00922
`Page 5 of 337
`
`
`
`Change 138588 on 2003/12/19 by vromaker@vromaker r400 linux marlboro
`
`-
`
`fixed a copy-paste error in the new code that generates src c sel for scalar const
`
`ops
`
`Change 138586 on 2003/12/19 by mearl@mearl r400 win
`
`update status
`
`Change 138585 on 2003/12/19 by mearl@mearl xenos linux orl
`
`Fixed bug in tracker.
`
`Change 138491 on 2003/12/18 by llefebvr@llefebvr r400 emu montreal
`
`I'll leave the const add test to Vic since he is working on it with Dan. I'll focus
`
`instead on coissue frac 01.
`
`Change 138489 on 2003/12/18 by mearl@mearl r400 win
`
`update status
`
`Change 138486 on 2003/12/18 by danh@danh r400 win
`
`Status update.
`
`Change 138469 on 2003/12/18 by cbrennan@cbrennan r400 release
`
`Try again to make a better performing miss stall.
`
`Change 138467 on 2003/12/18 by bhankins@bhankins xenos linux orl
`
`Increase detailed quad fifo skid size to account for input pipeline. Only affects Cl
`version.
`
`Change 138455 on 2003/12/18 by mearl@mearl r400 win
`
`update status
`
`Change 138414 on 2003/12/18 by tien@tien r400 devel marlboro
`
`Timing fix in tp lod_aniso: put tri juice in parallel with min/max clamp
`went from -0.3 to +0.4 slack coming out of MC (tp lod_aniso)
`
`GetWeights fix (tp tsel)
`Connected to vertex state bits properly (tp lod_fifo/tp input)
`
`Change 138380 on 2003/12/18 by mearl@mearl xenos linux orl
`
`AMO CONFIDENTIAL BUSINESS INFORMATION - SUBJECT TO THE PROTECTIVE ORDER
`
`AMD1044_0179860
`
`ATI Ex. 2036
`
`ATI Ex. 2036
`IPR2023-00922
`Page 6 of 337
`
`
`
`Updated the tracker to open all streams so a false PASS does not occur.
`
`Change 138375 on 2003/12/18 by jayw@jayw_r400 linux_marlboro
`
`Fix for hiz failures. Wrong hiz calculated. For Allen.
`
`Change 138369 on 2003/12/18 by kevino@kevino r400 release
`
`tcb fetch gen_walker timing fix
`
`Change 138309 on 2003/12/18 by danh@danh r400 win
`
`Status update.
`
`Change 138307 on 2003/12/18 by bhankins@bhankins xenos linux orl
`
`fix the way export memory is read in Cl mode
`
`Change 138289 on 2003/12/17 by rramsey@rramsey_xenos linux orl
`
`Fix a bug with pred_override that can occur when a clause starts
`with two predicated alu instructions. pred_override needs to use
`the isr version of the pred bits in this case because the pred register
`can't be intit'ed until after the the last instr of the prev clause has
`
`a chance to return its pred values and push them back to the thread buffer.
`This fixes r400sc sp sample cntl 47 and hopefully many more.
`
`Change 138288 on 2003/12/17 by vromaker@vromaker r400 linux marlboro
`
`-
`
`fix for scalar const opcodes: src c sel and gpr read_en logic was updated
`
`Change 138266 on 2003/12/17 by vbhatia@vbhatia r400 linux marlboro
`
`Standalone addresser testbench updates
`
`Change 138212 on 2003/12/17 by mearl@mearl r400 win
`
`update status
`
`Change 138152 on 2003/12/17 by danh@danh r400 win
`
`Status Update.
`
`Change 138138 on 2003/12/17 by bhankins@bhankins xenos linux orl
`
`1. Modify detailed quad fifo to keep mrt quads and mem export quads together
`2. Change NEW BC defines to Cl
`
`AMO CONFIDENTIAL BUSINESS INFORMATION - SUBJECT TO THE PROTECTIVE ORDER
`
`AMD1044_0179861
`
`ATI Ex. 2036
`
`ATI Ex. 2036
`IPR2023-00922
`Page 7 of 337
`
`
`
`Change 138120 on 2003/12/17 by mearl@mearl xenos linux orl
`
`Added ability to track exports that are not ordered.
`
`Change 138119 on 2003/12/17 by cbrennan@cbrennan r400 emu
`
`Change to miss stall generation that may increase perf in worst latency case by about
`5% up to 10%.
`Basically after a miss stall cycle, the next stall is calculated off of more accurate
`info which may let it go sooner.
`
`Change 138105 on 2003/12/17 by mdesai@mdesai r400 linux
`
`Resolved all hardware issues on bug3077.
`Fixed Y & Z overflow case
`
`Change 138087 on 2003/12/17 by rramsey@RRAMSEY_P4 r400 win
`
`latest regression status
`
`Change 138053 on 2003/12/17 by chammer@chammer r400 linux
`
`Changed XENOS flag to Cl to correspond to the other blocks
`
`Change 137864 on 2003/12/16 by rramsey@rramsey_xenos linux orl
`
`Change emulator so param cache reads for params not exported by the VS
`still show up in sq_sx_pcaddr.
`Fix cf resource change logic in the cfs so it catches the clause boundary
`where a cf instr with only tex instr gets sent to the alu cfs.
`
`Change 137839 on 2003/12/16 by amys@amys xenos linux orl
`
`fix read vc cntl signal
`
`Change 137833 on 2003/12/16 by vromaker@vromaker r400 linux marlboro
`
`changed OR to AND on rb rtr inputs of sx rb if module
`
`Change 137805 on 2003/12/16 by smoss@smoss crayola linux orl regress
`
`warnings for r400 too.
`
`Change 137787 on 2003/12/16 by rmanapat@rmanapat r400 release
`
`Fix for bug3106
`
`AMO CONFIDENTIAL BUSINESS INFORMATION - SUBJECT TO THE PROTECTIVE ORDER
`
`AMD1044_0179862
`
`ATI Ex. 2036
`
`ATI Ex. 2036
`IPR2023-00922
`Page 8 of 337
`
`
`
`Change 137773 on 2003/12/16 by vromaker@vromaker r400 linux marlboro
`
`-
`
`-
`
`fixed a predicate override bug: pred_overide is driven from the done bits if
`the previous operation was a waterfall, but it must only be driven for the first
`
`instruction following a waterfall. The bug occurred on back-to-back waterfalls
`where the pred_override was being driven for all cycles of the second waterfall.
`this fix caused r400sq_gpr index 01 to pass
`
`Change 137753 on 2003/12/16 by danh@danh r400 win
`
`Status update.
`
`Change 137701 on 2003/12/15 by rramsey@rramsey_xenos linux orl
`
`Add new sf (single-file) versions of PLI routines that allow trackers
`to only open their dump files one time.
`Modify a few trackers and models to use the new sf routines to verify
`
`they are working.
`Fix a problem with the cfsm not ignoring clause boundaries for unexecuted
`predicate control flow instr.
`
`Change 137604 on 2003/12/15 by dclifton@dclifton xenos linux orl
`
`Fixed scalar RSQ and RCP clamp to zero for -0 input.
`Fixed scalar MUL clamp to one for exponent overflow.
`Fixed scalar MUL sign on exponent underflow.
`Fixed scalar FRACT for 1 < X < 2.
`Fixed scalar FRACT for X all integer.
`Fixed scalar SUB with shift of 24.
`Fixed scalar ADD with clamp and exp underflow.
`Fixed scalar SIN sign with input 0.
`
`Enabled adder in sp_macc for vector PRED SET ops.
`Fixed more CUBE face id cases in sp comp opcodes with ma= -0.
`Fixed PRED SETGT and PRED SETGE conditions in sp_macc32 to account for nans.
`Redefined data flow in sp_macc32 for PRED SET ops.
`Enabled denorms and -0 pass-thru in sp_macc32 on DST op.
`Forced clearing of mantisa and sign in sp_macc32 for CUBE ma result.
`
`Change 137569 on 2003/12/15 by donaldl@donaldl xenos linux orl
`
`Removed dependency of bp[15:0] bits on vc-to-sp valid bits. Done so can
`still track SP data even when the RSP is used during redundancy testing.
`
`Change 137568 on 2003/12/15 by tien@tien r400 devel marlboro
`
`Expanded genl00/101 for aniso testbench
`Put in fix for Get/Set opcodes to completely match EMU ...
`including don't care fields
`
`AMO CONFIDENTIAL BUSINESS INFORMATION - SUBJECT TO THE PROTECTIVE ORDER
`
`AMD1044_0179863
`
`ATI Ex. 2036
`
`ATI Ex. 2036
`IPR2023-00922
`Page 9 of 337
`
`
`
`Change 137566 on 2003/12/15 by donaldl@donaldl xenos linux orl
`
`Added tracker for RSP to SX data.
`
`Change 137563 on 2003/12/15 by donaldl@donaldl xenos linux orl
`
`Fixed sensitivity list bug.
`
`Change 137560 on 2003/12/15 by danh@danh r400 win
`
`Updated status.
`
`Change 137474 on 2003/12/15 by hartogs@fl hartogs xenos win
`
`Integrated changes for addition of "VGT_MH_pglb clear" from Xenos to R400.
`
`Change 137470 on 2003/12/15 by rmanapat@rmanapat r400 sun_marlboro
`
`Changes for TCF, TCR, TCM chicken registers
`
`Change 137425 on 2003/12/15 by cbrennan@cbrennan r400 release
`
`Changed ifndef to ifdef+else because v2k doesnt work in some environments.
`
`Change 137415 on 2003/12/15 by jcarroll@jcarroll r400 win
`
`Picked tests
`
`Change 137401 on 2003/12/15 by rramsey@RRAMSEY_P4 r400 win
`
`update with weekend's results
`
`Change 137386 on 2003/12/15 by mmantor@mmantor xenos linux orl
`
`<synchronize sx0 and sxl paths to the rb's for xenos ordering within a tile>
`
`Change 137363 on 2003/12/13 by smoss@smoss crayola linux orl regress
`
`<Orlando Hardware Regression Results>
`
`Change 137313 on 2003/12/12 by mmantor@FL mmantorLT r400 win
`
`<update for standalone vp test>
`
`Change 137238 on 2003/12/12 by vromaker@vromaker r400 linux marlboro
`
`took a few more tests
`
`AMO CONFIDENTIAL BUSINESS INFORMATION - SUBJECT TO THE PROTECTIVE ORDER
`
`AMD1044_0179864
`
`ATI Ex. 2036
`
`ATI Ex. 2036
`IPR2023-00922
`Page 10 of 337
`
`
`
`Change 137212 on 2003/12/12 by tien@tien r400 devel marlboro
`
`Fix to sign determination for dni input sub (need to XOR sign of
`
`add result into input sign)
`Enhancements to the deriv/aniso testbenches
`
`Change 137206 on 2003/12/12 by cbrennan@cbrennan r400 release
`
`If it makes
`Its too generous sometimes and causes a hang.
`Revert this optimization..
`any material difference a much more timing messy optimization could be put in place.
`
`Change 137205 on 2003/12/12 by mearl@mearl r400 win
`
`updated status
`
`Change 137203 on 2003/12/12 by kevino@kevino r400 release
`
`Switched over to the low lat fifos in tea, fetch fifo, and tcd.
`Added tcd_ipbuf fifo top.v and switched overe to using 2 16x141 mems instead a 1
`32x141 mems.
`Put latency params in for latency fifo prog depth testcases
`
`Change 137188 on 2003/12/12 by rramsey@RRAMSEY_P4 r400 win
`
`update with latest regression results
`
`Change 137182 on 2003/12/12 by smburu@smburu r400 linux marlboro
`
`Fix for 16 EXPAND border color.
`
`Change 137166 on 2003/12/12 by vromaker@vromaker r400 linux marlboro
`
`-
`
`increased the depth of the sq-vc request fifo; this is a temporary fix while the
`mini and mega dee signals from the VC are added to the vc rp sp dump file
`
`Change 137165 on 2003/12/12 by mearl@mearl r400 win
`
`updated status
`
`Change 137146 on 2003/12/12 by vromaker@vromaker r400 linux marlboro
`
`updated status
`
`Change 137128 on 2003/12/12 by kevino@kevino r400 release
`
`Removed extra debug fifo that was causing problems.
`
`AMO CONFIDENTIAL BUSINESS INFORMATION - SUBJECT TO THE PROTECTIVE ORDER
`
`AMD1044_0179865
`
`ATI Ex. 2036
`
`ATI Ex. 2036
`IPR2023-00922
`Page 11 of 337
`
`
`
`Change 137115 on 2003/12/12 by bhankins@bhankins xenos linux orl
`
`1. fix some bugs in sx-bc interface
`2. remove obsolete, commented code
`
`3. start to add support for grouping quads to be
`
`Change 137114 on 2003/12/12 by dclifton@dclifton xenos linux orl
`
`Fixed clamp on scalar LOG(inf).
`Fixed sign on scalar RECIP(-inf) w/o clamp.
`Fixed scalar FRACT of large negative (integer) nums
`Fixed CUBE in sp_macc32 and sp comp opcodes with Y max at -0.
`Fixed DST with -0 for output as w or z.
`
`Change 137105 on 2003/12/12 by mmantor@mmantor xenos linux orl
`
`<fixed bug in the emu for redundancy control, added new input to the sq called
`
`sx sp alloc table free >
`
`Change 137104 on 2003/12/12 by mmantor@mmantor xenos linux orl
`
`<This changed changed SQ and SX top level ports by added thread_type from sq_alloc
`through the sx so tracker at sx rb works correct and fixed some other minor bugs>
`
`Change 137099 on 2003/12/12 by jayw@jayw_r400 linux_marlboro
`
`For Allen and Rex; fixes for depth and increased quad FIFO depth.
`
`Change 137079 on 2003/12/11 by mmantor@FL mmantorLT r400 win
`
`<updated for Laurent's cube changes>
`
`Change 137041 on 2003/12/11 by cbrennan@cbrennan r400 release
`
`Tiny cleanup of miss stall condition. Might save a cycle or two over a long test.
`
`Change 137030 on 2003/12/11 by cbrennan@cbrennan r400 emu
`
`Flipped nibble order of DXT3A AS 1 1 1 1 to really match dx spec this time.
`
`Change 136926 on 2003/12/11 by cbrennan@cbrennan r400 release
`
`Changed Cl to map MCs to requestors differently than R400
`Increased the length of the mh arb fifos from 8 to 24.
`
`Change 136917 on 2003/12/11 by mearl@mearl r400 win
`
`updated status
`
`AMO CONFIDENTIAL BUSINESS INFORMATION - SUBJECT TO THE PROTECTIVE ORDER
`
`AMD1044_0179866
`
`ATI Ex. 2036
`
`ATI Ex. 2036
`IPR2023-00922
`Page 12 of 337
`
`
`
`Change 136910 on 2003/12/11 by bhankins@bhankins xenos linux orl
`
`Fix logic that adds hook for a write of a quad event/null indicator to the detailed
`
`quad fifo.
`
`Change 136909 on 2003/12/11 by dclifton@dclifton xenos linux orl
`
`Fixed scalar FRACT of negative all-integer number.
`Fixed MOVA in sp comp opcodes to work like MAX.
`Fixed MIN(0,0) in sp comp opcodes
`Fixed CUBE in sp comp opcodes for ma
`
`nan or inf
`
`Change 136893 on 2003/12/11 by mearl@mearl r400 win
`
`updated status
`
`Change 136888 on 2003/12/11 by bhankins@bhankins xenos linux orl
`
`Change defined "Cl" switch to "NEW BC" in sx rtl and related vcpp files
`
`Change 136887 on 2003/12/11 by vhopkins@vhopkins xenos linux orl
`
`added test_plusargs for deactivation of fsdb
`
`Change 136871 on 2003/12/11 by mearl@mearl r400 win
`
`updated status
`
`Change 136867 on 2003/12/11 by rramsey@rramsey_xenos linux orl
`
`don't reset current context at eo rt load
`
`Change 136854 on 2003/12/11 by dclifton@dclifton xenos linux orl
`
`Fix to sp_macc32 max clamp
`Fix to sp_macc32 MULADD produce ati nan logic.
`Fix to sp_macc32 TRUNC and FLOOR -denorms.
`Fix to sp comp opcodes MIN(0,denorm)
`
`Change 136832 on 2003/12/10 by mmantor@FL mmantorLT r400 win
`
`<added all ops and provide pred opcode swizzle changes and kill in the standalone
`vectpipetest>
`
`Change 136805 on 2003/12/10 by kmeekins@kmeekins xenos linux orl
`
`Corrected logic that use to take advantage of the implied channel.
`
`AMO CONFIDENTIAL BUSINESS INFORMATION - SUBJECT TO THE PROTECTIVE ORDER
`
`AMD1044_0179867
`
`ATI Ex. 2036
`
`ATI Ex. 2036
`IPR2023-00922
`Page 13 of 337
`
`
`
`Change 136794 on 2003/12/10 by kmeekins@kmeekins xenos linux orl
`
`Added compile directives to make Xenos channel selection changes specific only
`
`to Xenos.
`
`Change 136782 on 2003/12/10 by kmeekins@kmeekins xenos linux orl
`
`Changed the channel selection from bit 6 to bit 7 of the fetch address.
`
`Change 136773 on 2003/12/10 by danh@danh r400 win
`
`Updated status.
`
`Change 136758 on 2003/12/10 by bhankins@bhankins xenos linux orl
`
`fix ifdef/endif mismatch
`
`Change 136749 on 2003/12/10 by smburu@smburu Cl synth
`
`Changed the reset going to TPC to be the hard reset.
`
`Change 136733 on 2003/12/10 by dclifton@dclifton xenos linux orl
`
`Fixed FLOOR and FRACT of negative denorms.
`Fixed 0*inf + -inf on MULADD
`
`Change 136727 on 2003/12/10 by mmantor@FL mmantorLT r400 win
`
`<added all vector ops>
`
`Change 136713 on 2003/12/10 by vromaker@vromaker r400 linux marlboro
`
`updated status
`
`Change 136691 on 2003/12/10 by bhankins@bhankins xenos linux orl
`
`1. Add ability for both r400 and xenos versions of sx to coexist
`2. Rewrite memory read mux select logic in sx_bc if.v for better synthesis
`3. Add quad_x and quad_y signals to BC interface.
`4. Update 'copy_virage ' files to reflect memory updates
`5. Change
`'ENABLE SX TO BC' compile switch to 'Cl'
`6. Remove obsolete code (sx export_buffers common.v logic is now in sx rb if.v)
`7. Update virage .cnt files
`
`Change 136682 on 2003/12/10 by kevino@kevino r400 release
`
`Added low latency fifo controllers to some fifos, but have disabled them with
`
`AMO CONFIDENTIAL BUSINESS INFORMATION - SUBJECT TO THE PROTECTIVE ORDER
`
`AMD1044_0179868
`
`ATI Ex. 2036
`
`ATI Ex. 2036
`IPR2023-00922
`Page 14 of 337
`
`
`
`ifdefs until the controller works for all cases.
`fifos along with corresponding test cases
`
`Added programmable depth for latency
`
`Change 136672 on 2003/12/10 by mearl@mearl r400 win
`
`updated status
`
`Change 136616 on 2003/12/10 by dclifton@dclifton xenos linux orl
`
`Scalar compare ops getting ati nan if opposite sign inf input.
`Scalar compare ops always output b input with nan input.
`Removed mantissa clear on zero for scalar inputs.
`Fixed scalar MAX(-0, 0);
`
`Fixed scalar MAX(x, inf);
`Fixed vector MIN
`Fixed vector FLOOR, neg x > -1
`Fixed inf and nan detect on dot product
`Fixed ati nan for MULADD -inf+ +inf
`Fixed MULADD for O * inf or nan
`Fixed final vector clamp logic
`
`Change 136596 on 2003/12/09 by vromaker@vromaker r400 linux marlboro
`
`added a couple wire names for ppb read data in cfs
`- added fsdb dump for tbtrk sq_vtx rs input in tb sqsp
`- changed checking of predicate to registered version in above trk to fix false
`mismatch
`bit 95 of vc/tp instruction was wired to O causing a mismatch, so
`I changed it to the actual instruction bit 95 (which is only used by the sq)
`
`Change 136589 on 2003/12/09 by donaldl@donaldl xenos linux orl
`
`Bug fix - connected parameter cache write data from output of SX redundancy
`input muxes (as opposed to directly from SP input registers).
`
`Change 136574 on 2003/12/09 by danh@danh r400 win
`
`Updated r400sc rts * status
`
`Change 136557 on 2003/12/09 by mmantor@mmantor xenos linux orl
`
`<fixed allocation counter for ea and cleaned up controls for rest of the counters and
`fixed a bug in the spi sp tracker by removing delay on sq_sp simd_id because of
`pipelining the vertex and pixel input data>
`
`Change 136555 on 2003/12/09 by mmantor@FL mmantorLT r400 win
`
`add VectPipeTest that test the vector pipe of the sp.
`
`AMO CONFIDENTIAL BUSINESS INFORMATION - SUBJECT TO THE PROTECTIVE ORDER
`
`AMD1044_0179869
`
`ATI Ex. 2036
`
`ATI Ex. 2036
`IPR2023-00922
`Page 15 of 337
`
`
`
`Change 136470 on 2003/12/09 by cbrennan@cbrennan r400 emu
`
`Fix a perfmon counter. Wasnt paying attention to the stall signal.
`
`Change 136457 on 2003/12/09 by georgev@devel georgev_r400 lin2_marlboro coverage tc
`
`Put under perforce control.
`
`Change 136442 on 2003/12/09 by kmeekins@kmeekins xenos linux orl
`
`randvc.pl
`
`Created script for running random tests on the VC and recording the results.
`
`runvc
`
`Corrected incomplete test detection.
`
`Change 136424 on 2003/12/09 by mearl@mearl r400 win
`
`update status
`
`Change 136403 on 2003/12/09 by jayw@jayw_r400 linux_marlboro
`
`Fix for Rex performance fifo resizing!
`
`Change 136398 on 2003/12/09 by rmanapat@rmanapat r400 release
`
`More changes to get the soft reset wire correctly
`
`Change 136395 on 2003/12/09 by amys@amys xenos linux orl
`
`connect w0 output registers correctly
`
`Change 136391 on 2003/12/09 by mmantor@FL mmantorLT r400 win
`
`<src special swizz moved to the sq>
`
`Change 136358 on 2003/12/09 by vromaker@vromaker r400 linux marlboro
`
`updated status for r400sq_auto_wrapping memories 01 (test issue)
`
`Change 136336 on 2003/12/09 by smburu@smburu Cl synth
`
`Extraneous comma in module port list.
`
`Change 136334 on 2003/12/09 by rramsey@RRAMSEY_P4 r400 win
`
`AMO CONFIDENTIAL BUSINESS INFORMATION - SUBJECT TO THE PROTECTIVE ORDER
`
`AMD1044_0179870
`
`ATI Ex. 2036
`
`ATI Ex. 2036
`IPR2023-00922
`Page 16 of 337
`
`
`
`update dot2add status, take more tests
`
`Change 136332 on 2003/12/09 by mmantor@FL mmantorLT r400 win
`
`<took test with >2 exports>
`
`Change 136330 on 2003/12/09 by rmanapat@rmanapat r400 release
`
`Wired in soft reset for tcf tcr and tern following the way Tien
`did it for the tpc and tp
`
`Change 136326 on 2003/12/09 by mearl@mearl r400 win
`
`took a few tests
`
`Change 136235 on 2003/12/08 by cbrennan@cbrennan r400 emu
`
`Make fix for X's in comparitors for small rectangle shaped textures for mipmapped
`stacks and cubes.
`
`Change 136192 on 2003/12/08 by mearl@mearl r400 win
`
`Removed more SC pipe disable tests.
`
`Change 136174 on 2003/12/08 by danh@danh r400 win
`
`Updated r400sc * status
`
`Change 136166 on 2003/12/08 by llefebvr@llefebvre laptop r400 emu
`
`working on r400sp coissue add_0l.cpp
`
`Change 136158 on 2003/12/08 by mearl@mearl xenos linux orl
`
`IFDEFed ports TMl and TM2 for XENOS.
`
`Change 136153 on 2003/12/08 by vromaker@vromaker r400 linux marlboro
`
`added my name by a few tests
`
`Change 136141 on 2003/12/08 by mearl@mearl r400 win
`
`Removed pipe disable tests, renamed and moved to the ROM block
`
`Change 136135 on 2003/12/08 by rramsey@rramsey_xenos linux orl
`
`Add a bit to pix thread counter to handle larger thread buffer.
`
`AMO CONFIDENTIAL BUSINESS INFORMATION - SUBJECT TO THE PROTECTIVE ORDER
`
`AMD1044_0179871
`
`ATI Ex. 2036
`
`ATI Ex. 2036
`IPR2023-00922
`Page 17 of 337
`
`
`
`Change 136107 on 2003/12/08 by georgev@devel georgev_r400 lin2_marlboro coverage tc
`
`Streamlined TC for current directory.
`
`Change 136102 on 2003/12/08 by rramsey@RRAMSEY_P4 r400 win
`
`update 'sorted by type' page with latest results
`
`Change 136070 on 2003/12/08 by dclifton@dclifton xenos linux orl
`
`Fixes to pass denorms on max or compare ops
`
`Change 136063 on 2003/12/08 by mmantor@mmantor xenos linux orl
`
`<another synthesis issue>
`
`Change 136048 on 2003/12/08 by mmantor@FL mmantorLT r400 win
`
`<update for more ops>
`
`Change 135995 on 2003/12/08 by danh@danh r400 win
`
`Updated r400sc sp sample cntl* status
`
`Change 135990 on 2003/12/08 by mmantor@FL mmantorLT r400 win
`
`<added executable for vectorpipe test bench vector generation>
`
`Change 135987 on 2003/12/08 by rramsey@RRAMSEY_P4 r400 win
`
`update with status from 12/8/2003
`
`Change 135984 on 2003/12/08 by mmantor@FL mmantorLT r400 win
`
`<update for tb vector>
`
`Change 135983 on 2003/12/08 by dclifton@dclifton r400
`
`Updated for new sq rams
`
`Change 135978 on 2003/12/08 by dclifton@dclifton xenos linux orl
`
`Fixed clamp of scalar mul with overflow.
`
`lxl < 1.
`Fixed trunc of neg numbers
`Reworked scalar fract of numbers
`lxl > 1.
`
`Change 135975 on 2003/12/08 by mmantor@mmantor xenos linux orl
`
`AMO CONFIDENTIAL BUSINESS INFORMATION - SUBJECT TO THE PROTECTIVE ORDER
`
`AMD1044_0179872
`
`ATI Ex. 2036
`
`ATI Ex. 2036
`IPR2023-00922
`Page 18 of 337
`
`
`
`<fixed leda errors for synthesis>
`
`Change 135943 on 2003/12/07 by vromaker@vromaker r400 linux marlboro
`
`- connected resource management register to thread buffers
`(programmable thread buffer size)
`fixed typo and leda error in sq_vtx ctl
`
`-
`
`Change 135932 on 2003/12/07 by rramsey@rramsey_xenos linux orl
`
`fix a problem with vizq_start events and how they cause state locks in the tb.
`this should fix the vgt event tests
`
`Change 135879 on 2003/12/05 by mmantor@mmantor xenos linux orl
`
`<fixed a synthesis problem during elaboration in the sq_input arb.v and fixed a problem
`
`with redunancy so that both vertex and pixel input controllers would send simd_id with
`there respective request to the spi. This change renamed a top level port between the
`sq and sp sq_sp interp simd_id changed to sq_sp simd_id >
`
`Change 135795 on 2003/12/05 by rmanapat@rmanapat r400 release
`
`Fixes a bug related to src address and 12sets and fetch3d
`
`Change 135747 on 2003/12/05 by vbhatia@vbhatia r400 linux marlboro
`
`Added support for TP Track DisplayLog to suppress display logging in default
`
`case
`
`for diskspace usage reasons.
`
`Change 135715 on 2003/12/05 by mearl@mearl xenos linux orl
`
`Fixed a bug in the multi-pass logic.
`
`Change 135678 on 2003/12/05 by kmeekins@kmeekins xenos linux orl
`
`buildtb
`tbasrt rg.v
`
`Added assertion testing for the RG.
`
`runvc
`
`- Added more testing for incomplete tests
`- Collect the runvc command line arguments and write them to the FAIL file
`
`on failing tests.
`- Moving .dmp files from the random directory to the test directory to assist
`
`AMO CONFIDENTIAL BUSINESS INFORMATION - SUBJECT TO THE PROTECTIVE ORDER
`
`AMD1044_0179873
`
`ATI Ex. 2036
`
`ATI Ex. 2036
`IPR2023-00922
`Page 19 of 337
`
`
`
`in file clean-up on passing tests.
`
`vc.v
`
`Changed I/O definitions to prevent erroneous mis-compares in netlist generation
`scripts.
`
`vcrg.cpp
`vcrg.v
`vcrg tag gen.v
`
`Added exception logic to handle vertex buffer size of zero.
`
`Change 135600 on 2003/12/05 by bhankins@bhankins xenos linux orl
`
`1. add behavioral support for sx to be interface. Disabled.
`2. fixed bug in alloc/dealloc block to hold off resetting alloc bit until the last
`
`bank of memory is read for a particular address.
`3. fixed bug in alloc/dealloc block where free logic was searching all 256 locations
`of the buffer when only 128 are enabled.
`4. connect SX SQ free export address_buf to indicate last quad of memory export has
`been created and written to the detailed quad fifo.
`5. fix minor bug in sx-rb interface logic that would have shown up with larger
`export buffer.
`
`Change 135598 on 2003/12/05 by smoss@smoss crayola linux orl regress
`
`removed reference to internal tracker
`
`Change 135590 on 2003/12/05 by dclifton@dclifton r400
`
`Update for new rams
`
`Change 135584 on 2003/12/05 by rramsey@rramsey_xenos linux orl
`
`absolute address mode (const addr_mode
`
`3'b001) should apply to all src constants
`
`Change 135537 on 2003/12/04 by vromaker@vromaker r400 linux marlboro
`
`increased size of thread buffers: vtx from 16 to 32 threads, pix from 48 to 64 thread
`fixed gpr dealloc bug that resulted in reduced performance
`-
`testbench and tracker changes we