SGX544MP/USSE ISA
< SGX544MP
Jump to navigation
Jump to search
What is USSE?
USSE = Universal Scalable Shader Engine (USSE is a TM)
USSE ISA
comments:
[SupportsNoSched] : result of analysis of HWInstSupportsNoSched
[SupportsSyncEnd] : result of analysis of HWInstSupportsSyncEnd
[ForcesDeschedule] : result of analysis of HWInstForcesDeschedule
[SupportsWriteMask] : result of analysis of HWInstSupportsWriteMask
[CanUseExtSrc0Banks] : result of analysis of HWInstCanUseExtSrc0Banks
[IsMOEControlInst] : result of analysis of HWInstIsMOEControlInst
branch instructions store 12-bit signed offset into low 12 bits of psHWInst->uWord0 : result of analysis of HWInstSetBranchOffset
invalid opcodes:
group 2:
1:31-27 0x02
1:10-9 3
group 3:
1:31-27 0x03
1:10-9 <2;3>
group 4:
1:31-27 0x04
1:10-9 <2;3>
group 6:
1:31-27 0x06
1:10-9 <1;3>
group 16:
1:31-27 0x16
1:21-20 1
group 1A:
1:31-27 0x1A
group 1B:
1:31-27 0x1B
group 1F_1:
1:31-27 0x1F
1:26-24 <6;7>
1:21-20 1
group 1F_3:
1:31-27 0x1F
1:26-24 {2,4,5,6,7}
1:21-20 3
common fields:
1:20 SyncStart
opcode "INVALID" [USP_OPCODE 0=0x0]
opcode "MAD" [USP_OPCODE 1=0x1]
1:31-27 0x00
#5 1:21 uRepeatMode
0: mask
1: repeat
#5 1:15-12 uRepeat
1:10-9 0
[SupportsNoSched]
[HWInstDecodeRepeat] => #5
opcode "ADM" [USP_OPCODE 2=0x2]
1:31-27 0x00
#5 1:21 uRepeatMode
0: mask
1: repeat
#5 1:15-12 uRepeat
1:10-9 1
[SupportsNoSched]
[HWInstDecodeRepeat] => #5
opcode "MSA" [USP_OPCODE 3=0x3]
1:31-27 0x00
#5 1:21 uRepeatMode
0: mask
1: repeat
#5 1:15-12 uRepeat
1:10-9 2
[SupportsNoSched]
[HWInstDecodeRepeat] => #5
opcode "FRC" [USP_OPCODE 4=0x4]
1:31-27 0x00
#5 1:21 uRepeatMode
0: mask
1: repeat
#5 1:15-12 uRepeat
1:10-9 3
[SupportsNoSched]
[HWInstDecodeRepeat] => #5
opcode "RCP" [USP_OPCODE 5=0x5]
1:31-27 0x01
#5 1:21 uRepeatMode
0: mask
1: repeat
#5 1:15-12 uRepeat
1:10-9 0
[SupportsNoSched]
[HWInstDecodeRepeat] => #5
opcode "RSQ" [USP_OPCODE 6=0x6]
1:31-27 0x01
#5 1:21 uRepeatMode
0: mask
1: repeat
#5 1:15-12 uRepeat
1:10-9 1
[SupportsNoSched]
[HWInstDecodeRepeat] => #5
opcode "LOG" [USP_OPCODE 7=0x7]
1:31-27 0x01
#5 1:21 uRepeatMode
0: mask
1: repeat
#5 1:15-12 uRepeat
1:10-9 2
[SupportsNoSched]
[HWInstDecodeRepeat] => #5
opcode "EXP" [USP_OPCODE 8=0x8]
1:31-27 0x01
#5 1:21 uRepeatMode
0: mask
1: repeat
#5 1:15-12 uRepeat
1:10-9 3
[SupportsNoSched]
[HWInstDecodeRepeat] => #5
opcode "DP" [USP_OPCODE 9=0x9]
1:31-27 0x02
#5 1:21 uRepeatMode
0: mask
1: repeat
#5 1:15-12 uRepeat
1:10-9 0
[SupportsNoSched]
[HWInstDecodeRepeat] => #5
opcode "DDP" [USP_OPCODE 10=0xa]
1:31-27 0x02
#5 1:21 uRepeatMode
0: mask
1: repeat
#5 1:15-12 uRepeat
1:10-9 1
[SupportsNoSched]
[HWInstDecodeRepeat] => #5
opcode "DDPC" [USP_OPCODE 11=0xb]
1:31-27 0x02
#5 1:21 uRepeatMode
0: mask
1: repeat
#5 1:15-12 uRepeat
1:10-9 2
[SupportsNoSched]
[HWInstDecodeRepeat] => #5
opcode "MIN" [USP_OPCODE 12=0xc]
1:31-27 0x03
#5 1:21 uRepeatMode
0: mask
1: repeat
#5 1:15-12 uRepeat
1:10-9 0
[SupportsNoSched]
[HWInstDecodeRepeat] => #5
opcode "MAX" [USP_OPCODE 13=0xd]
1:31-27 0x03
#5 1:21 uRepeatMode
0: mask
1: repeat
#5 1:15-12 uRepeat
1:10-9 1
[SupportsNoSched]
[HWInstDecodeRepeat] => #5
opcode "DSX" [USP_OPCODE 14=0xe]
1:31-27 0x04
1:10-9 0
[SupportsNoSched]
[HWInstDecodeRepeat] => does not support repeats
opcode "DSY" [USP_OPCODE 15=0xf]
1:31-27 0x04
1:10-9 1
[SupportsNoSched]
[HWInstDecodeRepeat] => does not support repeats
opcode "MOVC" [USP_OPCODE 16=0x10]
1:31-27 0x05
#5 1:21 uRepeatMode
0: mask
1: repeat
#5 1:15-12 uRepeat
#1 1:10-8 TestDataType
0: none
1: int8
2: int16
3: int32
4: float32
5: int10
[SupportsNoSched]
[HWInstDecodeMOVInstTestDataType] => #1
[HWInstDecodeRepeat] => #5
opcode "FMAD16" [USP_OPCODE 17=0x11]
1:31-27 0x06
#5 1:21 uRepeatMode
0: mask
1: repeat
#5 1:15-12 uRepeat
1:10-9 0
[SupportsNoSched]
[HWInstDecodeRepeat] => #5
opcode "EFO" [USP_OPCODE 18=0x12]
1:31-27 0x07
#5 1:13-12 uRepeat
[SupportsNoSched]
[HWInstDecodeRepeat] => #5, mode is "repeat"
opcode "PCKUNPCK" [USP_OPCODE 19=0x13]
1:31-27 0x08
#5 1:21 uRepeatMode
0: mask
1: repeat
#5 1:15-12 uRepeat
1:2-5 WriteMask
[SupportsNoSched]
[ForcesDeschedule]=SyncStart
[SupportsWriteMask]
[HWInstDecodeRepeat] => #5
opcode "TEST" [USP_OPCODE 20=0x14]
1:31-27 0x09
#5 1:15-12 uRepeat
[HWInstDecodeRepeat] => #5, mode is "mask"
opcode "AND" [USP_OPCODE 21=0x15]
1:31-27 0x0A
#5 1:21 uRepeatMode
0: mask
1: repeat
#5 1:15-12 uRepeat
1:3 0 ?? code in HWInstGetOpcode tests it and in both cases sets "AND" ??
[ForcesDeschedule]=SyncStart
[HWInstDecodeRepeat] => #5
opcode "OR" [USP_OPCODE 22=0x16]
1:31-27 0x0A
#5 1:21 uRepeatMode
0: mask
1: repeat
#5 1:15-12 uRepeat
1:3 1 ?? code in HWInstGetOpcode tests it and in both cases sets "AND" ?? shouldn't it set "OR" ??
[ForcesDeschedule]=SyncStart
[HWInstDecodeRepeat] => #5
opcode "XOR" [USP_OPCODE 23=0x17]
1:31-27 0x0B
#5 1:21 uRepeatMode
0: mask
1: repeat
#5 1:15-12 uRepeat
[SupportsNoSched]
[ForcesDeschedule]=SyncStart
[HWInstDecodeRepeat] => #5
opcode "SHL" [USP_OPCODE 24=0x18]
1:31-27 0x0C
#5 1:21 uRepeatMode
0: mask
1: repeat
#5 1:15-12 uRepeat
1:3 0
[ForcesDeschedule]=SyncStart
[HWInstDecodeRepeat] => #5
opcode "ROL" [USP_OPCODE 25=0x19]
1:31-27 0x0C
#5 1:21 uRepeatMode
0: mask
1: repeat
#5 1:15-12 uRepeat
1:3 1
[SupportsNoSched]
[ForcesDeschedule]=SyncStart
[HWInstDecodeRepeat] => #5
opcode "SHR" [USP_OPCODE 26=0x1a]
1:31-27 0x0D
#5 1:21 uRepeatMode
0: mask
1: repeat
#5 1:15-12 uRepeat
1:3 0
[ForcesDeschedule]=SyncStart
[HWInstDecodeRepeat] => #5
opcode "ASR" [USP_OPCODE 27=0x1b]
1:31-27 0x0D
#5 1:21 uRepeatMode
0: mask
1: repeat
#5 1:15-12 uRepeat
1:3 1
[SupportsNoSched]
[ForcesDeschedule]=SyncStart
[HWInstDecodeRepeat] => #5
opcode "RLP" [USP_OPCODE 28=0x1c]
1:31-27 0x0E
#5 1:21 uRepeatMode
0: mask
1: repeat
#5 1:15-12 uRepeat
[ForcesDeschedule]=SyncStart
[HWInstDecodeRepeat] => #5
opcode "TESTMASK" [USP_OPCODE 29=0x1d]
1:31-27 0x0F
#5 1:15-12 uRepeat
[SupportsNoSched]
[ForcesDeschedule]=SyncStart
[HWInstDecodeRepeat] => #5, mode is "mask"
opcode "SOP2" [USP_OPCODE 30=0x1e]
1:31-27 0x10
#5 1:14-12 uRepeat
[SupportsNoSched]
[HWInstDecodeRepeat] => #5, mode is "repeat"
opcode "SOP3" [USP_OPCODE 31=0x1f]
1:31-27 0x11
[SupportsNoSched]
[HWInstDecodeRepeat] => does not support repeats
opcode "SOPWM" [USP_OPCODE 32=0x20] SOP Write Mask?
1:31-27 0x12
1:14 WriteMask[2]
1:13 WriteMask[1]
1:12 WriteMask[0]
1:11 WriteMask[3]
[SupportsNoSched]
[SupportsWriteMask]
[HWInstDecodeRepeat] => does not support repeats
opcode "IMA8" [USP_OPCODE 33=0x21]
1:31-27 0x13
#5 1:14-12 uRepeat
[SupportsNoSched]
[HWInstDecodeRepeat] => #5, mode is "repeat"
opcode "IMA16" [USP_OPCODE 34=0x22]
1:31-27 0x14
#5 1:14-12 uRepeat
[SupportsNoSched]
[HWInstDecodeRepeat] => #5, mode is "repeat"
opcode "IMAE" [USP_OPCODE 35=0x23]
1:31-27 0x15
#5 1:14-12 uRepeat
[SupportsNoSched]
[HWInstDecodeRepeat] => #5, mode is "repeat"
opcode "ADIF" [USP_OPCODE 36=0x24]
1:31-27 0x16
1:21-20 0
#5 1:14-12 uRepeat
[SupportsNoSched]
[HWInstDecodeRepeat] => #5, mode is "repeat"
opcode "BILIN" [USP_OPCODE 37=0x25]
1:31-27 0x16
1:21-20 2
#5 1:14-12 uRepeat
[SupportsNoSched]
[HWInstDecodeRepeat] => #5, mode is "repeat"
opcode "FIRV" [USP_OPCODE 38=0x26]
1:31-27 0x16
1:21-20 3
#5 1:14-12 uRepeat
[SupportsNoSched]
[HWInstDecodeRepeat] => #5, mode is "repeat"
opcode "FIRH" [USP_OPCODE 39=0x27]
1:31-27 0x17
#5 1:13-12 uRepeat
[SupportsNoSched]
[HWInstDecodeRepeat] => #5, mode is "repeat"
opcode "DOT3" [USP_OPCODE 40=0x28]
1:31-27 0x18
1:24 0
#5 1:14-12 uRepeat
[SupportsNoSched]
[HWInstDecodeRepeat] => #5, mode is "repeat"
opcode "DOT4" [USP_OPCODE 41=0x29]
1:31-27 0x18
1:24 1
#5 1:14-12 uRepeat
[SupportsNoSched]
[HWInstDecodeRepeat] => #5, mode is "repeat"
opcode "FPMA" [USP_OPCODE 42=0x2a]
1:31-27 0x19
#5 1:14-12 uRepeat
[SupportsNoSched]
[HWInstDecodeRepeat] => #5, mode is "repeat"
opcode "SMP" [USP_OPCODE 43=0x2b]
1:31-27 0x1C
#3 1:11-10 uCoordDim
0: 1D
1: 2D
2: 3D
3: invalid
1:9-8 0
#2 1:7 DestRegType
0: TEMP
1: PA
1:1-0 DRC
#4 [ForcesDeschedule]=SyncStart
[CanUseExtSrc0Banks]
[HWInstDecodeDestBankAndNum] => #2
[HWInstDecodeSMPInstCoordDim] => #3
[HWInstDecodeSMPInstCoordFmt] => all sample coordinates are passed as F32
[HWInstDecodeSMPInstDRCNum] => #4
[HWInstDecodeRepeat] => does not support repeats
opcode "SMPBIAS" [USP_OPCODE 44=0x2c]
1:31-27 0x1C
#3 1:11-10 uCoordDim
0: 1D
1: 2D
2: 3D
3: invalid
1:9-8 1
#2 1:7 DestRegType
0: TEMP
1: PA
#4 1:1-0 DRC
[ForcesDeschedule]=SyncStart
[CanUseExtSrc0Banks]
[HWInstDecodeDestBankAndNum] => #2
[HWInstDecodeSMPInstCoordDim] => #3
[HWInstDecodeSMPInstCoordFmt] => all sample coordinates are passed as F32
[HWInstDecodeSMPInstDRCNum] => #4
[HWInstDecodeRepeat] => does not support repeats
opcode "SMPREPLACE" [USP_OPCODE 45=0x2d]
1:31-27 0x1C
#3 1:11-10 uCoordDim
0: 1D
1: 2D
2: 3D
3: invalid
1:9-8 2
#2 1:7 DestRegType
0: TEMP
1: PA
#4 1:1-0 DRC
[ForcesDeschedule]=SyncStart
[CanUseExtSrc0Banks]
[HWInstDecodeDestBankAndNum] => #2
[HWInstDecodeSMPInstCoordDim] => #3
[HWInstDecodeSMPInstCoordFmt] => all sample coordinates are passed as F32
[HWInstDecodeSMPInstDRCNum] => #4
[HWInstDecodeRepeat] => does not support repeats
opcode "SMPGRAD" [USP_OPCODE 46=0x2e]
1:31-27 0x1C
#3 1:11-10 uCoordDim
0: 1D
1: 2D
2: 3D
3: invalid
1:9-8 3
#2 1:7 DestRegType
0: TEMP
1: PA
#4 1:1-0 DRC
[ForcesDeschedule]=SyncStart
[CanUseExtSrc0Banks]
[HWInstDecodeDestBankAndNum] => #2
[HWInstDecodeSMPInstCoordDim] => #3
[HWInstDecodeSMPInstCoordFmt] => all sample coordinates are passed as F32
[HWInstDecodeSMPInstDRCNum] => #4
[HWInstDecodeRepeat] => does not support repeats
opcode "LD" [USP_OPCODE 47=0x2f]
1:31-27 0x1D
#5 1:21 uRepeatMode
0: fetch
1: repeat
#3,5 1:15-12 uFetchCount/uRepeat
#2 1:7 DestRegType
0: TEMP
1: PA
[ForcesDeschedule]=SyncStart
[CanUseExtSrc0Banks]
#2 analysis of HWInstDecodeDestBankAndNum
#3 analysis of HWInstEncodeLDInstFetchCount
[HWInstDecodeRepeat] => #5
opcode "ST" [USP_OPCODE 48=0x30]
1:31-27 0x1E
#5 1:21 uRepeatMode
0: fetch
1: repeat
#5 1:15-12 uRepeat
[ForcesDeschedule]=SyncStart
[CanUseExtSrc0Banks]
[HWInstDecodeRepeat] => #5
opcode "BA" [USP_OPCODE 49=0x31]
0:11-0 signed jump offset <FFFFF800;000007FF>
1:31-27 0x1F
1:23 SyncEnd
1:21-20 0
1:8-6 0
[SupportsSyncEnd]
[ForcesDeschedule]=true
[HWInstDecodeRepeat] => does not support repeats
opcode "BR" [USP_OPCODE 50=0x32]
0:11-0 signed jump offset <FFFFF800;000007FF>
1:31-27 0x1F
1:23 SyncEnd
1:21-20 0
1:8-6 1
[SupportsSyncEnd]
[ForcesDeschedule]=true
[HWInstDecodeRepeat] => does not support repeats
opcode "LAPC" [USP_OPCODE 51=0x33]
1:31-27 0x1F
1:23 SyncEnd
1:21-20 0
1:8-6 2
[SupportsSyncEnd]
[ForcesDeschedule]=true
[HWInstDecodeRepeat] => does not support repeats
opcode "SETL" [USP_OPCODE 52=0x34]
1:31-27 0x1F
1:23 SyncEnd
1:21-20 0
1:8-6 3
[SupportsSyncEnd]
[HWInstDecodeRepeat] => does not support repeats
opcode "SAVL" [USP_OPCODE 53=0x35]
1:31-27 0x1F
1:23 SyncEnd
1:21-20 0
1:8-6 4
[SupportsSyncEnd]
[HWInstDecodeRepeat] => does not support repeats
opcode "NOP" [USP_OPCODE 54=0x36]
1:31-27 0x1F
1:23 SyncStart
1:21-20 0
1:8-6 5
[SupportsNoSched]
[SupportsSyncEnd]
[ForcesDeschedule]=SyncStart
[HWInstDecodeRepeat] => does not support repeats
opcode "SMOA" [USP_OPCODE 55=0x37]
1:31-27 0x1F
1:26-24 0
1:21-20 1
[IsMOEControlInst]
[HWInstDecodeRepeat] => does not support repeats
opcode "SMR" [USP_OPCODE 56=0x38]
1:31-27 0x1F
1:26-24 1
1:21-20 1
[IsMOEControlInst]
[HWInstDecodeRepeat] => does not support repeats
opcode "SMLSI" [USP_OPCODE 57=0x39] Set MOE /???/ Swizle Increment ?
1:31-27 0x1F
1:26-24 2
1:21-20 1
#1 1:3 bUseSwizChan0
#1 1:2 bUseSwizChan1
#1 1:1 bUseSwizChan2
#1 1:0 bUseSwizChan3
#1 0:31-24 if bUseSwizChan0==1 uSwiz
if bUseSwizChan0==0 signed 8-bit MOE increment
#1 0:23-16 if bUseSwizChan1==1 uSwiz
if bUseSwizChan1==0 signed 8-bit MOE increment
#1 0:15-8 if bUseSwizChan2==1 uSwiz
if bUseSwizChan2==0 signed 8-bit MOE increment
#1 0: 7-0 if bUseSwizChan3==1 uSwiz
if bUseSwizChan3==0 signed 8-bit MOE increment
[IsMOEControlInst]
#1 =analysis of HWInstUpdateMOEState
[HWInstDecodeRepeat] => does not support repeats
opcode "SMBO" [USP_OPCODE 58=0x3a] Set MOE Base Offset ?
1:31-27 0x1F
1:26-24 3
1:21-20 1
#1 1:15-4 uBaseOffset[0] (bits 0..11)
#1 1:0-3 uBaseOffset[1] (bits 16..19) - seems to be an error in HWInstUpdateMOEState, this might be really (bits 8..11)
#1 0:31-24 uBaseOffset[1] (bits 0..7)
#1 0:23-12 uBaseOffset[2] (bits 0..11)
#1 0:11-0 uBaseOffset[3] (bits 0..11)
[IsMOEControlInst]
#1 =analysis of HWInstUpdateMOEState
[HWInstDecodeRepeat] => does not support repeats
opcode "IMO" [USP_OPCODE 59=0x3b]
1:31-27 0x1F
1:26-24 4
1:21-20 1
[IsMOEControlInst]
[HWInstDecodeRepeat] => does not support repeats
opcode "SETFC" [USP_OPCODE 60=0x3c] SET Format Control ?
1:31-27 0x1F
1:26-24 5
1:21-20 1
#6 0:8 bColFmtCtl
#6 0:0 EFOFmtCtl
[IsMOEControlInst]
[HWInstDecodeRepeat] => does not support repeats
[HWInstUpdateMOEState] => #6
opcode "IDF" [USP_OPCODE 61=0x3d]
1:31-27 0x1F
1:26-24 0
1:21-20 2
[HWInstDecodeRepeat] => does not support repeats
opcode "WDF" [USP_OPCODE 62=0x3e]
1:31-27 0x1F
1:26-24 1
1:21-20 2
[ForcesDeschedule]=true
[HWInstDecodeRepeat] => does not support repeats
opcode "SETM" [USP_OPCODE 63=0x3f]
1:31-27 0x1F
1:26-24 2
1:21-20 2
[SupportsNoSched]
[HWInstDecodeRepeat] => function returns error for this opcode
opcode "EMIT" [USP_OPCODE 64=0x40]
1:31-27 0x1F
1:26-24 3
1:21-20 2
[SupportsNoSched]
[CanUseExtSrc0Banks]
[HWInstDecodeRepeat] => does not support repeats
opcode "LIMM" [USP_OPCODE 65=0x41]
1:31-27 0x1F
1:26-24 4
1:21-20 2
[SupportsNoSched]
[HWInstDecodeRepeat] => does not support repeats
opcode "LOCK" [USP_OPCODE 66=0x42]
0:1 0
1:31-27 0x1F
1:26-24 5
1:21-20 2
[SupportsNoSched]
[ForcesDeschedule]=true
[HWInstDecodeRepeat] => does not support repeats
opcode "RELEASE" [USP_OPCODE 67=0x43]
0:1 1
1:31-27 0x1F
1:26-24 5
1:21-20 2
[SupportsNoSched]
[HWInstDecodeRepeat] => does not support repeats
opcode "LDR" [USP_OPCODE 68=0x44]
1:31-27 0x1F
1:26-24 6
1:21-20 2
1:19 0
#2 1:7 DestRegType
0: TEMP
1: PA
[SupportsNoSched]
#2 analysis of HWInstDecodeDestBankAndNum
[HWInstDecodeRepeat] => does not support repeats
opcode "STR" [USP_OPCODE 69=0x45]
1:31-27 0x1F
1:26-24 6
1:21-20 2
1:19 1
[SupportsNoSched]
[HWInstDecodeRepeat] => does not support repeats
opcode "WOP" [USP_OPCODE 70=0x46]
1:31-27 0x1F
1:26-24 7
1:21-20 2
[SupportsNoSched]
[ForcesDeschedule]=true
[HWInstDecodeRepeat] => does not support repeats
opcode "PCOEFF" [USP_OPCODE 71=0x47]
1:31-27 0x1F
1:26-24 0
1:21-20 3
1:15 0
[SupportsNoSched]
[CanUseExtSrc0Banks]
[HWInstDecodeRepeat] => does not support repeats
opcode "PTOFF" [USP_OPCODE 72=0x48]
1:31-27 0x1F
1:26-24 0
1:21-20 3
1:15 1
[SupportsNoSched]
[HWInstDecodeRepeat] => does not support repeats
opcode "ATST8" [USP_OPCODE 73=0x49]
1:31-27 0x1F
1:26-24 1
1:21-20 3
[SupportsNoSched]
[CanUseExtSrc0Banks]
[HWInstDecodeRepeat] => does not support repeats
opcode "DEPTHF" [USP_OPCODE 74=0x4a ]
1:31-27 0x1F
1:26-24 3
1:21-20 3
[SupportsNoSched]
[CanUseExtSrc0Banks]