From 45b8a6a0db3a2dd47ec7aa0129215f7b4813a234 Mon Sep 17 00:00:00 2001 From: Luke Kenneth Casson Leighton Date: Sat, 16 Dec 2023 08:37:31 +0000 Subject: [PATCH] bug 1236: add extra argument to svstep: RA. --- openpower/isa/simplev.mdwn | 4 +- openpower/isatables/RM-1P-1D.csv | 1 - openpower/isatables/RM-2P-1S1D.csv | 1 + openpower/isatables/minor_22.csv | 2 +- src/openpower/decoder/isa/caller.py | 2 +- .../decoder/isa/test_caller_setvl.py | 74 +++++++++---------- .../decoder/isa/test_caller_svp64_bc.py | 2 +- .../decoder/isa/test_caller_svp64_chacha20.py | 4 +- .../decoder/isa/test_caller_svp64_dct.py | 4 +- .../decoder/isa/test_caller_svp64_fft.py | 10 +-- .../decoder/isa/test_caller_svp64_pack.py | 18 ++--- .../decoder/isa/test_caller_svstate.py | 8 +- src/openpower/sv/sv_analysis.py | 9 ++- 13 files changed, 70 insertions(+), 69 deletions(-) diff --git a/openpower/isa/simplev.mdwn b/openpower/isa/simplev.mdwn index 33a02e66..3172bed4 100644 --- a/openpower/isa/simplev.mdwn +++ b/openpower/isa/simplev.mdwn @@ -5,8 +5,8 @@ SVL-Form -* svstep RT,SVi,vf (Rc=0) -* svstep. RT,SVi,vf (Rc=1) +* svstep RT,RA,SVi,vf (Rc=0) +* svstep. RT,RA,SVi,vf (Rc=1) Pseudo-code: diff --git a/openpower/isatables/RM-1P-1D.csv b/openpower/isatables/RM-1P-1D.csv index c13f9a6f..fc397899 100644 --- a/openpower/isatables/RM-1P-1D.csv +++ b/openpower/isatables/RM-1P-1D.csv @@ -1,3 +1,2 @@ insn,mode,CONDITIONS,Ptype,Etype,SM,0,1,2,3,in1,in2,in3,out,CR in,CR out,out2 fmvis,NORMAL,,1P,EXTRA3,NO,d:FRS,0,0,0,0,0,0,FRS,0,0,0 -svstep,NORMAL,,1P,EXTRA3,NO,d:RT;d:CR0,0,0,0,0,0,0,RT,0,CR0,0 diff --git a/openpower/isatables/RM-2P-1S1D.csv b/openpower/isatables/RM-2P-1S1D.csv index b24b4384..8216fdaa 100644 --- a/openpower/isatables/RM-2P-1S1D.csv +++ b/openpower/isatables/RM-2P-1S1D.csv @@ -34,6 +34,7 @@ xori,NORMAL,,2P,EXTRA3,EN,d:RA,s:RS,0,0,RS,0,0,RA,0,0,0 xoris,NORMAL,,2P,EXTRA3,EN,d:RA,s:RS,0,0,RS,0,0,RA,0,0,0 subfic,NORMAL,,2P,EXTRA3,EN,d:RT,s:RA,0,0,RA,0,0,RT,0,0,0 fishmv,NORMAL,,2P,EXTRA3,EN,d:FRS,s:FRS,0,0,FRS,0,0,FRS,0,0,0 +svstep,NORMAL,,2P,EXTRA3,EN,d:RT;d:CR0,s:RA,0,0,RA_OR_ZERO,0,0,RT,0,CR0,0 setvl,NORMAL,,2P,EXTRA3,EN,d:RT,s:RA,0,0,RA_OR_ZERO,0,0,RT_OR_ZERO,0,CR0,0 cntlzw,NORMAL,,2P,EXTRA3,EN,d:RA;d:CR0,s:RS,0,0,RS,0,0,RA,0,CR0,0 cntlzd,NORMAL,,2P,EXTRA3,EN,d:RA;d:CR0,s:RS,0,0,RS,0,0,RA,0,CR0,0 diff --git a/openpower/isatables/minor_22.csv b/openpower/isatables/minor_22.csv index 7a2e5c80..4dbe045e 100644 --- a/openpower/isatables/minor_22.csv +++ b/openpower/isatables/minor_22.csv @@ -4,7 +4,7 @@ opcode,unit,internal op,in1,in2,in3,out,CR in,CR out,inv A,inv out,cry in,cry ou -----11011-,VL,OP_SETVL,RA_OR_ZERO,NONE,NONE,RT_OR_ZERO,NONE,CR0,0,0,ZERO,0,NONE,0,0,0,0,0,0,RC_ONLY,0,0,setvl,SVL,,1,unofficial until submitted and approved/renumbered by the opf isa wg -----101001,VL,OP_SVINDEX,NONE,NONE,NONE,NONE,NONE,NONE,0,0,ZERO,0,NONE,0,0,0,0,0,0,NONE,0,0,svindex,SVI,,1,unofficial until submitted and approved/renumbered by the opf isa wg -----111001,VL,OP_SVREMAP,NONE,NONE,NONE,NONE,NONE,NONE,0,0,ZERO,0,NONE,0,0,0,0,0,0,NONE,0,0,svremap,SVRM,,1,unofficial until submitted and approved/renumbered by the opf isa wg ------10011-,VL,OP_SVSTEP,NONE,NONE,NONE,RT,NONE,CR0,0,0,ZERO,0,NONE,0,0,0,0,0,0,RC_ONLY,0,0,svstep,SVL,,1,unofficial until submitted and approved/renumbered by the opf isa wg +-----10011-,VL,OP_SVSTEP,RA_OR_ZERO,NONE,NONE,RT,NONE,CR0,0,0,ZERO,0,NONE,0,0,0,0,0,0,RC_ONLY,0,0,svstep,SVL,,1,unofficial until submitted and approved/renumbered by the opf isa wg # svshape svshape2 fit same space, using 4-bits 21-24. svshape2 takes up two 4-bit patterns. # bit 25 is "vf" in svshape, and "sk" in svshape2 hence the "-" in between, in "NNNN-011001" 0000-011001,VL,OP_SVSHAPE,NONE,NONE,NONE,NONE,NONE,NONE,0,0,ZERO,0,NONE,0,0,0,0,0,0,NONE,0,0,svshape,SVM,,1,unofficial until submitted and approved/renumbered by the opf isa wg diff --git a/src/openpower/decoder/isa/caller.py b/src/openpower/decoder/isa/caller.py index 8e476a28..7ed040e2 100644 --- a/src/openpower/decoder/isa/caller.py +++ b/src/openpower/decoder/isa/caller.py @@ -3015,7 +3015,7 @@ class ISACaller(ISACallerHelper, ISAFPHelpers, StepLoop): self.svstate.vfirst = 0 return nia_update - def SVSTATE_NEXT(self, mode, submode): + def SVSTATE_NEXT(self, mode, submode, RA=None): """explicitly moves srcstep/dststep on to next element, for "Vertical-First" mode. this function is called from setvl pseudo-code, as a pseudo-op "svstep" diff --git a/src/openpower/decoder/isa/test_caller_setvl.py b/src/openpower/decoder/isa/test_caller_setvl.py index a0661aa1..4d033698 100644 --- a/src/openpower/decoder/isa/test_caller_setvl.py +++ b/src/openpower/decoder/isa/test_caller_setvl.py @@ -296,8 +296,8 @@ class DecoderTestCase(FHDLTestCase): def test_svstep_1(self): lst = SVP64Asm(["setvl 0, 0, 10, 1, 1, 1", # actual setvl (VF mode) - "svstep 0, 1, 0", # svstep - "svstep 0, 1, 0", # svstep + "svstep 0, 0, 1, 0", # svstep + "svstep 0, 0, 1, 0", # svstep ]) lst = list(lst) @@ -327,8 +327,8 @@ class DecoderTestCase(FHDLTestCase): """tests svstep when it reaches VL """ lst = SVP64Asm(["setvl 0, 0, 2, 1, 1, 1", # actual setvl (VF mode) - "svstep. 0, 1, 0", # svstep (Rc=1) - "svstep. 0, 1, 0", # svstep (Rc=1) + "svstep. 0, 0, 1, 0", # svstep (Rc=1) + "svstep. 0, 0, 1, 0", # svstep (Rc=1) ]) lst = list(lst) @@ -365,8 +365,8 @@ class DecoderTestCase(FHDLTestCase): """tests svstep when it *doesn't* reach VL """ lst = SVP64Asm(["setvl 0, 0, 3, 1, 1, 1", # actual setvl (VF mode) - "svstep. 0, 1, 0", # svstep (Rc=1) - "svstep. 0, 1, 0", # svstep (Rc=1) + "svstep. 0, 0, 1, 0", # svstep (Rc=1) + "svstep. 0, 0, 1, 0", # svstep (Rc=1) ]) lst = list(lst) @@ -580,12 +580,12 @@ class DecoderTestCase(FHDLTestCase): # set triple butterfly mode with persistent "REMAP" "svshape 8, 1, 1, 1, 1", "svremap 31, 1, 0, 2, 0, 1, 1", - "svstep. 0, 2, 0", # svstep (Rc=1) - "svstep. 0, 2, 0", # svstep (Rc=1) - "svstep. 0, 2, 0", # svstep (Rc=1) - "svstep. 0, 2, 0", # svstep (Rc=1) - "svstep. 0, 2, 0", # svstep (Rc=1) - "svstep. 2, 2, 0", # svstep (Rc=1) + "svstep. 0, 0, 2, 0", # svstep (Rc=1) + "svstep. 0, 0, 2, 0", # svstep (Rc=1) + "svstep. 0, 0, 2, 0", # svstep (Rc=1) + "svstep. 0, 0, 2, 0", # svstep (Rc=1) + "svstep. 0, 0, 2, 0", # svstep (Rc=1) + "svstep. 2, 0, 2, 0", # svstep (Rc=1) ]) lst = list(lst) @@ -624,9 +624,9 @@ class DecoderTestCase(FHDLTestCase): # set triple butterfly mode with persistent "REMAP" "svshape 8, 1, 1, 1, 1", "svremap 31, 1, 0, 2, 0, 1, 1", - "svstep. 0, 2, 0", # svstep (Rc=1) - "svstep. 0, 2, 0", # svstep (Rc=1) - "svstep. 0, 2, 0", # svstep (Rc=1) + "svstep. 0, 0, 2, 0", # svstep (Rc=1) + "svstep. 0, 0, 2, 0", # svstep (Rc=1) + "svstep. 0, 0, 2, 0", # svstep (Rc=1) ]) lst = list(lst) @@ -665,10 +665,10 @@ class DecoderTestCase(FHDLTestCase): # set triple butterfly mode with persistent "REMAP" "svshape 8, 1, 1, 1, 1", "svremap 31, 1, 0, 2, 0, 1, 1", - "svstep. 0, 2, 0", # svstep (Rc=1) - "svstep. 0, 2, 0", # svstep (Rc=1) - "svstep. 0, 2, 0", # svstep (Rc=1) - "svstep. 0, 2, 0", # svstep (Rc=1) + "svstep. 0, 0, 2, 0", # svstep (Rc=1) + "svstep. 0, 0, 2, 0", # svstep (Rc=1) + "svstep. 0, 0, 2, 0", # svstep (Rc=1) + "svstep. 0, 0, 2, 0", # svstep (Rc=1) ]) lst = list(lst) @@ -708,10 +708,10 @@ class DecoderTestCase(FHDLTestCase): # set triple butterfly mode with persistent "REMAP" "svshape 8, 1, 1, 1, 1", "svremap 31, 1, 0, 2, 0, 1, 1", - "svstep. 0, 2, 0", # svstep (Rc=1) - "svstep. 0, 2, 0", # svstep (Rc=1) - "svstep. 0, 2, 0", # svstep (Rc=1) - "svstep. 2, 2, 0", # svstep (Rc=1) + "svstep. 0, 0, 2, 0", # svstep (Rc=1) + "svstep. 0, 0, 2, 0", # svstep (Rc=1) + "svstep. 0, 0, 2, 0", # svstep (Rc=1) + "svstep. 2, 0, 2, 0", # svstep (Rc=1) ]) lst = list(lst) @@ -760,8 +760,8 @@ class DecoderTestCase(FHDLTestCase): # set FFT triple butterfly mode with persistent "REMAP" "svshape 8, 1, 1, 1, 0", "svremap 0, 0, 0, 2, 0, 1, 1", - "sv.svstep *0, 1, 1", # svstep get vector of j+halfsize - "sv.svstep *16, 3, 1", # svstep get vector of k in exptable[k] + "sv.svstep *0, 0, 1, 1", # svstep get vector of j+halfsize + "sv.svstep *16, 0, 3, 1", # svstep get vector of k in exptable[k] ]) lst = list(lst) @@ -828,10 +828,10 @@ class DecoderTestCase(FHDLTestCase): """tests svstep with an add, when it reaches VL lst = SVP64Asm(["setvl 0, 0, 2, 1, 1, 1", 'sv.add *1, *5, *9', - "svstep. 0, 1, 0", # svstep (Rc=1) + "svstep. 0, 0, 1, 0", # svstep (Rc=1) "setvl. 0, 0, 1, 1, 0, 0", 'sv.add *1, *5, *9', - "svstep. 3, 1, 0", # svstep (Rc=1) + "svstep. 3, 0, 1, 0", # svstep (Rc=1) "setvl 4, 0, 1, 0, 0, 0", # getvl ]) sequence is as follows: @@ -846,9 +846,9 @@ class DecoderTestCase(FHDLTestCase): """ lst = SVP64Asm(["setvl 0, 0, 2, 1, 1, 1", 'sv.add *1, *5, *9', - "svstep. 0, 1, 0", # svstep (Rc=1) + "svstep. 0, 0, 1, 0", # svstep (Rc=1) 'sv.add *1, *5, *9', - "svstep. 3, 1, 0", # svstep (Rc=1) + "svstep. 3, 0, 1, 0", # svstep (Rc=1) "setvl 4, 0, 1, 0, 0, 0", # getvl ]) lst = list(lst) @@ -900,7 +900,7 @@ class DecoderTestCase(FHDLTestCase): """tests svstep with a branch. lst = SVP64Asm(["setvl 0, 0, 2, 1, 1, 1", 'sv.add *1, *5, *9', - "svstep. 0, 1, 0", # svstep (Rc=1) + "svstep. 0, 0, 1, 0", # svstep (Rc=1) "bc 6, 3, -0xc" ]) sequence is as follows: @@ -926,7 +926,7 @@ class DecoderTestCase(FHDLTestCase): """ lst = SVP64Asm(["setvl 0, 0, 2, 1, 1, 1", 'sv.add *1, *5, *9', - "svstep. 0, 1, 0", # svstep (Rc=1) + "svstep. 0, 0, 1, 0", # svstep (Rc=1) "bc 6, 3, -0xc" # branch to add (64-bit op so -0xc!) ]) lst = list(lst) @@ -981,9 +981,9 @@ class DecoderTestCase(FHDLTestCase): lst = SVP64Asm(["setvl 0, 0, 5, 1, 1, 1", 'sv.cmp 0, 1, *4, 14', # r8 contains the temp 'sv.isel 14,*4,14,1', # copy if cmp was greater - "svstep. 12, 6, 0", # get srcstep + "svstep. 12, 0, 6, 0", # get srcstep 'sv.isel 10,12,10,1', # copy if cmp was greater - "svstep. 0, 1, 0", # svstep (Rc=1) + "svstep. 0, 0, 1, 0", # svstep (Rc=1) "bc 6, 3, -0x24" # branch to cmp ]) lst = list(lst) @@ -1061,8 +1061,8 @@ class DecoderTestCase(FHDLTestCase): """tests svstep "straight", placing srcstep, dststep into vector """ lst = SVP64Asm(["setvl 0, 0, 4, 0, 1, 1", - "sv.svstep *0, 5, 1", # svstep get vector srcstep - "sv.svstep. *4, 6, 1", # svstep get vector dststep + "sv.svstep *0, 0, 5, 1", # svstep get vector srcstep + "sv.svstep. *4, 0, 6, 1", # svstep get vector dststep ]) lst = list(lst) @@ -1100,8 +1100,8 @@ class DecoderTestCase(FHDLTestCase): """tests svstep "straight", placing srcstep, dststep into vector """ lst = SVP64Asm(["setvl 0, 0, 5, 0, 1, 1", - "sv.svstep/m=r30 *0, 5, 1", # svstep get vector srcstep - "sv.svstep./m=r30 *8, 6, 1", # svstep get vector dststep + "sv.svstep/m=r30 *0,0,5,1", # svstep get vector srcstep + "sv.svstep./m=r30 *8,0,6,1", # svstep get vector dststep ]) lst = list(lst) diff --git a/src/openpower/decoder/isa/test_caller_svp64_bc.py b/src/openpower/decoder/isa/test_caller_svp64_bc.py index 93689ded..6915ea16 100644 --- a/src/openpower/decoder/isa/test_caller_svp64_bc.py +++ b/src/openpower/decoder/isa/test_caller_svp64_bc.py @@ -317,7 +317,7 @@ class DecoderTestCase(FHDLTestCase): "setvl 0, 0, %d, 1, 1, 1" % maxvl, # VL = MAXVL = 2, vf=1 "sv.cmpi *cr0, 1, *10, 0x10", # compare reg val with immediate "sv.bc 0, *2, 0x10", # jmp if CTR!=0 AND reg not equal to imm - "svstep. 27, 1, 0", + "svstep. 27, 0, 1, 0", "bc 4, 3, -0x14", # CR_BI=0, jump to start of loop (sv.cmpi) "or 0, 0, 0", # jump to here if terminate VF loop early ] diff --git a/src/openpower/decoder/isa/test_caller_svp64_chacha20.py b/src/openpower/decoder/isa/test_caller_svp64_chacha20.py index 01c5edbf..1fd9a4d0 100644 --- a/src/openpower/decoder/isa/test_caller_svp64_chacha20.py +++ b/src/openpower/decoder/isa/test_caller_svp64_chacha20.py @@ -162,8 +162,8 @@ class SVSTATETestCase(FHDLTestCase): 'sv.xor/w=32 *%d, *%d, *%d' % (block, block, block), 'svremap 31, 0, 3, 2, 2, 0, 0', # RA=2, RB=3, RS=2 (0b01110) 'sv.rldcl/w=32 *%d, *%d, *%d, 0' % (block, block, shifts), - 'svstep. %d, 1, 0' % ctr, # step to next in-regs element - 'bc 6, 3, -0x28', # svstep. Rc=1 loop-end-condition? + 'svstep. %d, 0, 1, 0' % ctr, # step to next in-regs element + 'bc 6, 3, -0x28', # svstep. Rc=1 loop-end-condition? # inner-loop done: outer loop standard CTR-decrement to setvl again 'bc 16, 0, -0x30', ]) diff --git a/src/openpower/decoder/isa/test_caller_svp64_dct.py b/src/openpower/decoder/isa/test_caller_svp64_dct.py index 581b3522..96674d4a 100644 --- a/src/openpower/decoder/isa/test_caller_svp64_dct.py +++ b/src/openpower/decoder/isa/test_caller_svp64_dct.py @@ -438,8 +438,8 @@ class DCTTestCase(FHDLTestCase): """ lst = SVP64Asm(["svshape 8, 1, 1, 5, 0", "svremap 0, 0, 0, 2, 0, 1, 1", - "sv.svstep *4, 3, 1", # svstep get vector of ci - "sv.svstep *16, 2, 1", # svstep get vector of step + "sv.svstep *4, 0, 3, 1", # svstep get vector of ci + "sv.svstep *16, 0, 2, 1", # svstep get vector of step "addi 1, 0, 0x0000", "setvl 0, 0, 7, 0, 1, 1", "sv.std *4, 0(1)", diff --git a/src/openpower/decoder/isa/test_caller_svp64_fft.py b/src/openpower/decoder/isa/test_caller_svp64_fft.py index bc201c31..fdaefff0 100644 --- a/src/openpower/decoder/isa/test_caller_svp64_fft.py +++ b/src/openpower/decoder/isa/test_caller_svp64_fft.py @@ -260,7 +260,7 @@ class FFTTestCase(FHDLTestCase): "svshape 8, 1, 1, 1, 1", "svremap 31, 1, 0, 2, 0, 1, 0", "sv.ffmadds *0, *8, *0", - "svstep. 12, 1, 0", + "svstep. 12, 0, 1, 0", "bc 6, 3, -16" ]) runs a full in-place O(N log2 N) butterfly schedule for @@ -275,7 +275,7 @@ class FFTTestCase(FHDLTestCase): "svshape 8, 1, 1, 1, 1", "svremap 31, 1, 0, 2, 0, 1, 0", "sv.ffmadds *0, *8, *0", - "svstep. 27, 1, 0", + "svstep. 27, 0, 1, 0", "bc 6, 3, -16" ]) lst = list(lst) @@ -347,7 +347,7 @@ class FFTTestCase(FHDLTestCase): # RA: scal RB: jl (S0) RC: n/a RT: jl (S0) EA: jh (S1) "svremap 26, 0, 0, 0, 0, 1, 1", "sv.ffadds *0, 24, *0", - "svstep. 27, 1, 0", + "svstep. 27, 0, 1, 0", "bc 6, 3, -28" ]) @@ -386,7 +386,7 @@ class FFTTestCase(FHDLTestCase): # RA: scal RB: jl (S0) RC: n/a RT: jl (S0) EA: jh (S1) "svremap 26, 0, 0, 0, 0, 1, 0", "sv.ffadds *0, 24, *0", - "svstep. 27, 1, 0", + "svstep. 27, 0, 1, 0", "bc 6, 3, -28" ]) lst = list(lst) @@ -622,7 +622,7 @@ class FFTTestCase(FHDLTestCase): "sv.ffadds *8, 26, *8", # vh/vl +- tpim # svstep loop - "svstep. 27, 1, 0", + "svstep. 27, 0, 1, 0", "bc 6, 3, -56" ]) lst = list(lst) diff --git a/src/openpower/decoder/isa/test_caller_svp64_pack.py b/src/openpower/decoder/isa/test_caller_svp64_pack.py index 0e83b9eb..fb46a8b0 100644 --- a/src/openpower/decoder/isa/test_caller_svp64_pack.py +++ b/src/openpower/decoder/isa/test_caller_svp64_pack.py @@ -20,11 +20,11 @@ class DecoderTestCase(FHDLTestCase): """tests pack mode """ lst = SVP64Asm(["setvl 0, 0, 4, 0, 1, 1", - "svstep 0, 15, 0", # set dst-pack - "sv.svstep./vec2 *0, 5, 1", # svstep get vector srcstep - "sv.svstep./vec2 *8, 6, 1", # svstep get vector dststep - "sv.svstep./vec2 *16, 7, 1", # svstep get src substep - "sv.svstep./vec2 *24, 8, 1", # svstep get dst substep + "svstep 0, 0, 15, 0", # set dst-pack + "sv.svstep./vec2 *0,0,5,1", # svstep get vector srcstep + "sv.svstep./vec2 *8,0,6,1", # svstep get vector dststep + "sv.svstep./vec2 *16,0,7,1", # svstep get src substep + "sv.svstep./vec2 *24,0,8,1", # svstep get dst substep ]) lst = list(lst) @@ -70,7 +70,7 @@ class DecoderTestCase(FHDLTestCase): the vector output in the order of their own values. """ lst = SVP64Asm(["setvl 0, 0, 4, 0, 1, 1", - "svstep 0, 14, 0", # set src-pack + "svstep 0, 0, 14, 0", # set src-pack "sv.ori/vec2 *0, *32, 0", # copy 01234567 to new order ]) lst = list(lst) @@ -108,7 +108,7 @@ class DecoderTestCase(FHDLTestCase): """tests pack mode """ lst = SVP64Asm(["setvl 0, 0, 4, 0, 1, 1", - "svstep 0, 15, 0", # set dst-pack + "svstep 0, 0, 15, 0", # set dst-pack "sv.ori/vec2 *0, *32, 0", # copy 01234567 to new order ]) lst = list(lst) @@ -146,7 +146,7 @@ class DecoderTestCase(FHDLTestCase): """tests pack mode with a predicate """ lst = SVP64Asm(["setvl 0, 0, 4, 0, 1, 1", - "svstep 0, 15, 0", # set dst-pack + "svstep 0, 0, 15, 0", # set dst-pack "sv.ori/vec2/m=r3 *8, *16, 0", ]) lst = list(lst) @@ -203,7 +203,7 @@ class DecoderTestCase(FHDLTestCase): """tests unpack mode with predicate """ lst = SVP64Asm(["setvl 0, 0, 4, 0, 1, 1", - "svstep 0, 14, 0", # set src-pack + "svstep 0, 0, 14, 0", # set src-pack "sv.ori/vec2/m=r3 *8, *16, 0", # copy with mask ]) lst = list(lst) diff --git a/src/openpower/decoder/isa/test_caller_svstate.py b/src/openpower/decoder/isa/test_caller_svstate.py index 48c4b6f2..88e77a7d 100644 --- a/src/openpower/decoder/isa/test_caller_svstate.py +++ b/src/openpower/decoder/isa/test_caller_svstate.py @@ -60,10 +60,10 @@ class SVSTATETestCase(FHDLTestCase): lst = SVP64Asm(["setvl 0, 0, 2, 1, 1, 1", 'sv.add 1, *5, *9', 'sv.addi *12, 1, 1', - "svstep. 0, 1, 0", + "svstep. 0, 0, 1, 0", 'sv.add 1, *5, *9', 'sv.addi *12, 1, 1', - "svstep. 0, 1, 0", + "svstep. 0, 0, 1, 0", ]) sequence is as follows: @@ -92,10 +92,10 @@ class SVSTATETestCase(FHDLTestCase): lst = SVP64Asm(["setvl 0, 0, 2, 1, 1, 1", 'sv.add 1, *5, *9', # scalar dest (into r1) 'sv.addi *12, 1, 1', # scalar src (from r1) - "svstep. 0, 1, 0", # svstep + "svstep. 0, 0, 1, 0", # svstep 'sv.add 1, *5, *9', # again, scalar dest 'sv.addi *12, 1, 1', # but vector dest - "svstep. 0, 1, 0", # svstep (end: sets CR0.SO) + "svstep. 0, 0, 1, 0", # svstep (end: sets CR0.SO) ]) lst = list(lst) diff --git a/src/openpower/sv/sv_analysis.py b/src/openpower/sv/sv_analysis.py index 21778ad0..06351e29 100644 --- a/src/openpower/sv/sv_analysis.py +++ b/src/openpower/sv/sv_analysis.py @@ -539,8 +539,11 @@ def extra_classifier(insn_name, value, name, res, regs): res['0'] = 'd:FRS' # FRS: Rdest1_EXTRA3 res['1'] = 's:FRS' # FRS: Rsrc1_EXTRA3 elif insn_name == 'setvl': - res['0'] = 'd:RT' # RT: Rdest1_EXTRA3 - res['1'] = 's:RA' # RS: Rsrc1_EXTRA3 + res['0'] = 'd:RT' # RT: Rdest1_EXTRA3 + res['1'] = 's:RA' # RS: Rsrc1_EXTRA3 + elif insn_name == 'svstep': + res['0'] = 'd:RT;d:CR0' # RT,CR0: Rdest1_EXTRA3 + res['1'] = 's:RA' # RA: Rsrc1_EXTRA3 else: raise NotImplementedError(insn_name) @@ -649,8 +652,6 @@ def extra_classifier(insn_name, value, name, res, regs): elif value == 'RM-1P-1D': res['Etype'] = 'EXTRA3' # RM EXTRA3 type - if insn_name == 'svstep': - res['0'] = 'd:RT;d:CR0' # RT,CR0: Rdest1_EXTRA3 if insn_name == 'fmvis': res['0'] = 'd:FRS' # FRS: Rdest1_EXTRA3 -- 2.30.2