decode1: Reduce number of single-issue instructions

This reduces the set of instructions marked as single-issue to just
attn and mtspr to "slow" SPRs (those that are not stored in the
register file).

The instructions that were previously single-issue are: isync, dcbf,
dcbst, dcbt, dcbtst, eieio, icbi, mfmsr, mtmsr, mtmsrd, mfspr to slow
SPRS, sync, tlbsync and wait.  The synchronization instructions are
mostly no-ops anyway due to the in-order nature of the core, and the
cache-management instructions are unimplemented (except for icbi).
The MSR ops don't need to be single-issue due to the in-order core and
the fact that MSR updates are effective on the following instruction.

Signed-off-by: Paul Mackerras <paulus@ozlabs.org>

diff --git a/decode1.vhdl b/decode1.vhdl
index fb92b9e6817815bc6762cf6d0cb38f8bed2bc6c6..3f3109f0fdb5e0dc302e069548574a8e870d78b6 100644 (file)
--- a/decode1.vhdl
+++ b/decode1.vhdl
@@ -170,7 +170,7 @@ architecture behaviour of decode1 is
         -- bclr, bcctr, bctar
         2#100#    =>       (ALU, NONE, OP_BCREG,     SPR,        SPR,         NONE, SPR,  '1', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '1', '0', NONE),
         -- isync
-        2#111#    =>       (ALU, NONE, OP_ISYNC,     NONE,       NONE,        NONE, NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '1', NONE),
+        2#111#    =>       (ALU, NONE, OP_ISYNC,     NONE,       NONE,        NONE, NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '0', NONE),
         -- rfid
         2#101#    =>       (ALU, NONE, OP_RFID,      SPR,        SPR,         NONE, NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '0', NONE),
         others   => illegal_inst
@@ -223,10 +223,10 @@ architecture behaviour of decode1 is
         2#1000111010#  =>       (ALU,  NONE, OP_CNTZ,      NONE,       NONE,        RS,   RA,   '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', RC,   '0', '0', NONE), -- cnttzd
         2#1000011010#  =>       (ALU,  NONE, OP_CNTZ,      NONE,       NONE,        RS,   RA,   '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '1', '0', RC,   '0', '0', NONE), -- cnttzw
         2#1011110011#  =>       (ALU,  NONE, OP_DARN,      NONE,       NONE,        NONE, RT,   '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '0', NONE), -- darn
-        2#0001010110#  =>       (ALU,  NONE, OP_DCBF,      NONE,       NONE,        NONE, NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '1', NONE), -- dcbf
-        2#0000110110#  =>       (ALU,  NONE, OP_DCBST,     NONE,       NONE,        NONE, NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '1', NONE), -- dcbst
-        2#0100010110#  =>       (ALU,  NONE, OP_DCBT,      NONE,       NONE,        NONE, NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '1', NONE), -- dcbt
-        2#0011110110#  =>       (ALU,  NONE, OP_DCBTST,    NONE,       NONE,        NONE, NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '1', NONE), -- dcbtst
+        2#0001010110#  =>       (ALU,  NONE, OP_DCBF,      NONE,       NONE,        NONE, NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '0', NONE), -- dcbf
+        2#0000110110#  =>       (ALU,  NONE, OP_DCBST,     NONE,       NONE,        NONE, NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '0', NONE), -- dcbst
+        2#0100010110#  =>       (ALU,  NONE, OP_DCBT,      NONE,       NONE,        NONE, NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '0', NONE), -- dcbt
+        2#0011110110#  =>       (ALU,  NONE, OP_DCBTST,    NONE,       NONE,        NONE, NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '0', NONE), -- dcbtst
         2#1111110110#  =>       (LDST, NONE, OP_DCBZ,      RA_OR_ZERO, RB,          NONE, NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '0', NONE), -- dcbz
         2#0110001001#  =>       (ALU,  NONE, OP_DIVE,      RA,         RB,          NONE, RT,   '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', RC,   '0', '0', NONE), -- divdeu
         2#1110001001#  =>       (ALU,  NONE, OP_DIVE,      RA,         RB,          NONE, RT,   '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', RC,   '0', '0', NONE), -- divdeuo
@@ -247,7 +247,7 @@ architecture behaviour of decode1 is
         2#1100110110#  =>       (ALU,  NONE, OP_NOP,       NONE,       NONE,        NONE, NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '0', NONE), -- dss
         2#0101010110#  =>       (ALU,  NONE, OP_NOP,       NONE,       NONE,        NONE, NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '0', NONE), -- dst
         2#0101110110#  =>       (ALU,  NONE, OP_NOP,       NONE,       NONE,        NONE, NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '0', NONE), -- dstst
-        2#1101010110#  =>       (ALU,  NONE, OP_NOP,       NONE,       NONE,        NONE, NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '1', NONE), -- eieio
+        2#1101010110#  =>       (ALU,  NONE, OP_NOP,       NONE,       NONE,        NONE, NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '0', NONE), -- eieio
         2#0100011100#  =>       (ALU,  NONE, OP_XOR,       NONE,       RB,          RS,   RA,   '0', '0', '0', '1', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', RC,   '0', '0', NONE), -- eqv
         2#1110111010#  =>       (ALU,  NONE, OP_EXTS,      NONE,       NONE,        RS,   RA,   '0', '0', '0', '0', ZERO, '0', is1B, '0', '0', '0', '0', '0', '0', RC,   '0', '0', NONE), -- extsb
         2#1110011010#  =>       (ALU,  NONE, OP_EXTS,      NONE,       NONE,        RS,   RA,   '0', '0', '0', '0', ZERO, '0', is2B, '0', '0', '0', '0', '0', '0', RC,   '0', '0', NONE), -- extsh
@@ -327,8 +327,8 @@ architecture behaviour of decode1 is
         2#1100001001#  =>       (ALU,  NONE, OP_MOD,       RA,         RB,          NONE, RT,   '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '1', NONE, '0', '0', NONE), -- modsd
         2#1100001011#  =>       (ALU,  NONE, OP_MOD,       RA,         RB,          NONE, RT,   '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '1', '1', NONE, '0', '0', NONE), -- modsw
         2#0010010000#  =>       (ALU,  NONE, OP_MTCRF,     NONE,       NONE,        RS,   NONE, '0', '1', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '0', NONE), -- mtcrf/mtocrf
-        2#0010010010#  =>       (ALU,  NONE, OP_MTMSRD,    NONE,       NONE,        RS,   NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '1', '0', NONE, '0', '1', NONE), -- mtmsr
-        2#0010110010#  =>       (ALU,  NONE, OP_MTMSRD,    NONE,       NONE,        RS,   NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '1', NONE), -- mtmsrd # ignore top bits and d
+        2#0010010010#  =>       (ALU,  NONE, OP_MTMSRD,    NONE,       NONE,        RS,   NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '1', '0', NONE, '0', '0', NONE), -- mtmsr
+        2#0010110010#  =>       (ALU,  NONE, OP_MTMSRD,    NONE,       NONE,        RS,   NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '0', NONE), -- mtmsrd # ignore top bits and d
         2#0111010011#  =>       (ALU,  NONE, OP_MTSPR,     NONE,       NONE,        RS,   SPR,  '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '0', NONE), -- mtspr
         2#0001001001#  =>       (ALU,  NONE, OP_MUL_H64,   RA,         RB,          NONE, RT,   '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '1', RC,   '0', '0', NONE), -- mulhd
         2#0000001001#  =>       (ALU,  NONE, OP_MUL_H64,   RA,         RB,          NONE, RT,   '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', RC,   '0', '0', NONE), -- mulhdu
@@ -409,13 +409,13 @@ architecture behaviour of decode1 is
         2#1011101000#  =>       (ALU,  NONE, OP_ADD,       RA,         CONST_M1,    NONE, RT,   '0', '0', '1', '0', CA,   '1', NONE, '0', '0', '0', '0', '0', '0', RC,   '0', '0', NONE), -- subfmeo
         2#0011001000#  =>       (ALU,  NONE, OP_ADD,       RA,         NONE,        NONE, RT,   '0', '0', '1', '0', CA,   '1', NONE, '0', '0', '0', '0', '0', '0', RC,   '0', '0', NONE), -- subfze
         2#1011001000#  =>       (ALU,  NONE, OP_ADD,       RA,         NONE,        NONE, RT,   '0', '0', '1', '0', CA,   '1', NONE, '0', '0', '0', '0', '0', '0', RC,   '0', '0', NONE), -- subfzeo
-        2#1001010110#  =>       (ALU,  NONE, OP_NOP,       NONE,       NONE,        NONE, NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '1', NONE), -- sync
+        2#1001010110#  =>       (ALU,  NONE, OP_NOP,       NONE,       NONE,        NONE, NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '0', NONE), -- sync
         2#0001000100#  =>       (ALU,  NONE, OP_TRAP,      RA,         RB,          NONE, NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '0', NONE), -- td
         2#0000000100#  =>       (ALU,  NONE, OP_TRAP,      RA,         RB,          NONE, NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '1', '0', NONE, '0', '0', NONE), -- tw
         2#0100110010#  =>       (LDST, NONE, OP_TLBIE,     NONE,       RB,          RS,   NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '0', NONE), -- tlbie
         2#0100010010#  =>       (LDST, NONE, OP_TLBIE,     NONE,       RB,          RS,   NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '0', NONE), -- tlbiel
-        2#1000110110#  =>       (ALU,  NONE, OP_NOP,       NONE,       NONE,        NONE, NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '1', NONE), -- tlbsync
-        2#0000011110#  =>       (ALU,  NONE, OP_NOP,       NONE,       NONE,        NONE, NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '1', NONE), -- wait
+        2#1000110110#  =>       (ALU,  NONE, OP_NOP,       NONE,       NONE,        NONE, NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '0', NONE), -- tlbsync
+        2#0000011110#  =>       (ALU,  NONE, OP_NOP,       NONE,       NONE,        NONE, NONE, '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', NONE, '0', '0', NONE), -- wait
         2#0100111100#  =>       (ALU,  NONE, OP_XOR,       NONE,       RB,          RS,   RA,   '0', '0', '0', '0', ZERO, '0', NONE, '0', '0', '0', '0', '0', '0', RC,   '0', '0', NONE), -- xor
         others => illegal_inst
        );
@@ -640,9 +640,9 @@ begin
 
             if std_match(f_in.insn(10 downto 1), "01-1010011") then
                 -- mfspr or mtspr
-                -- Make slow SPRs single issue
                 if is_fast_spr(v.ispr1) = '0' then
-                    vi.force_single := '1';
+                    -- Make mtspr to slow SPRs single issue
+                    vi.force_single := f_in.insn(8);
                     -- send MMU-related SPRs to loadstore1
                     case sprn is
                         when SPR_DAR | SPR_DSISR | SPR_PID | SPR_PTCR =>