Pass 256-bit-wide vectors in registers.

7 files changed, 27 insertions, 0 deletions

diff --git a/compiler/cmm/CmmCallConv.hs b/compiler/cmm/CmmCallConv.hs
index de10d56490..de86c4ed33 100644
--- a/compiler/cmm/CmmCallConv.hs
+++ b/compiler/cmm/CmmCallConv.hs
@@ -67,6 +67,7 @@ assignArgumentsPos dflags off conv arg_ty reps = (stk_off, assignments)
                                     | otherwise      = int
         where vec = case (w, regs) of
                       (W128, (vs, fs, ds, ls, s:ss)) -> k (RegisterParam (XmmReg s), (vs, fs, ds, ls, ss))
+                      (W256, (vs, fs, ds, ls, s:ss)) -> k (RegisterParam (YmmReg s), (vs, fs, ds, ls, ss))
                       _ -> (assts, (r:rs))
               float = case (w, regs) of
                         (W32, (vs, fs, ds, ls, s:ss))
diff --git a/compiler/cmm/CmmExpr.hs b/compiler/cmm/CmmExpr.hs
index d3624dac6b..02126905c8 100644
--- a/compiler/cmm/CmmExpr.hs
+++ b/compiler/cmm/CmmExpr.hs
@@ -343,6 +343,9 @@ data GlobalReg
   | XmmReg                      -- 128-bit SIMD vector register
         {-# UNPACK #-} !Int     -- its number
 
+  | YmmReg                      -- 256-bit SIMD vector register 
+        {-# UNPACK #-} !Int     -- its number
+
   -- STG registers
   | Sp                  -- Stack ptr; points to last occupied stack location.
   | SpLim               -- Stack limit
@@ -379,6 +382,7 @@ instance Eq GlobalReg where
    DoubleReg i == DoubleReg j = i==j
    LongReg i == LongReg j = i==j
    XmmReg i == XmmReg j = i==j
+   YmmReg i == YmmReg j = i==j
    Sp == Sp = True
    SpLim == SpLim = True
    Hp == Hp = True
@@ -401,6 +405,7 @@ instance Ord GlobalReg where
    compare (DoubleReg i) (DoubleReg j) = compare i j
    compare (LongReg i)   (LongReg   j) = compare i j
    compare (XmmReg i)    (XmmReg    j) = compare i j
+   compare (YmmReg i)    (YmmReg    j) = compare i j
    compare Sp Sp = EQ
    compare SpLim SpLim = EQ
    compare Hp Hp = EQ
@@ -424,6 +429,8 @@ instance Ord GlobalReg where
    compare _ (LongReg _)      = GT
    compare (XmmReg _) _       = LT
    compare _ (XmmReg _)       = GT
+   compare (YmmReg _) _       = LT
+   compare _ (YmmReg _)       = GT
    compare Sp _ = LT
    compare _ Sp = GT
    compare SpLim _ = LT
@@ -467,6 +474,7 @@ globalRegType _      (FloatReg _)      = cmmFloat W32
 globalRegType _      (DoubleReg _)     = cmmFloat W64
 globalRegType _      (LongReg _)       = cmmBits W64
 globalRegType _      (XmmReg _)        = cmmVec 4 (cmmBits W32)
+globalRegType _      (YmmReg _)        = cmmVec 8 (cmmBits W32)
 
 globalRegType dflags Hp                = gcWord dflags
                                             -- The initialiser for all
@@ -479,4 +487,5 @@ isArgReg (FloatReg {})   = True
 isArgReg (DoubleReg {})  = True
 isArgReg (LongReg {})    = True
 isArgReg (XmmReg {})     = True
+isArgReg (YmmReg {})     = True
 isArgReg _               = False
diff --git a/compiler/cmm/PprCmmExpr.hs b/compiler/cmm/PprCmmExpr.hs
index d1128b07d3..c96dee2800 100644
--- a/compiler/cmm/PprCmmExpr.hs
+++ b/compiler/cmm/PprCmmExpr.hs
@@ -256,6 +256,7 @@ pprGlobalReg gr
         DoubleReg  n   -> char 'D' <> int n
         LongReg    n   -> char 'L' <> int n
         XmmReg     n   -> ptext (sLit "XMM") <> int n
+        YmmReg     n   -> ptext (sLit "YMM") <> int n
         Sp             -> ptext (sLit "Sp")
         SpLim          -> ptext (sLit "SpLim")
         Hp             -> ptext (sLit "Hp")
diff --git a/compiler/codeGen/CgUtils.hs b/compiler/codeGen/CgUtils.hs
index c06dd60cb1..2cf4cb3483 100644
--- a/compiler/codeGen/CgUtils.hs
+++ b/compiler/codeGen/CgUtils.hs
@@ -56,6 +56,13 @@ baseRegOffset dflags (XmmReg 4)          = oFFSET_StgRegTable_rXMM4 dflags
 baseRegOffset dflags (XmmReg 5)          = oFFSET_StgRegTable_rXMM5 dflags
 baseRegOffset dflags (XmmReg 6)          = oFFSET_StgRegTable_rXMM6 dflags
 baseRegOffset _      (XmmReg n)          = panic ("Registers above XMM6 are not supported (tried to use XMM" ++ show n ++ ")")
+baseRegOffset dflags (YmmReg 1)          = oFFSET_StgRegTable_rYMM1 dflags
+baseRegOffset dflags (YmmReg 2)          = oFFSET_StgRegTable_rYMM2 dflags
+baseRegOffset dflags (YmmReg 3)          = oFFSET_StgRegTable_rYMM3 dflags
+baseRegOffset dflags (YmmReg 4)          = oFFSET_StgRegTable_rYMM4 dflags
+baseRegOffset dflags (YmmReg 5)          = oFFSET_StgRegTable_rYMM5 dflags
+baseRegOffset dflags (YmmReg 6)          = oFFSET_StgRegTable_rYMM6 dflags
+baseRegOffset _      (YmmReg n)          = panic ("Registers above YMM6 are not supported (tried to use YMM" ++ show n ++ ")")
 baseRegOffset dflags Sp                  = oFFSET_StgRegTable_rSp dflags
 baseRegOffset dflags SpLim               = oFFSET_StgRegTable_rSpLim dflags
 baseRegOffset dflags (LongReg 1)         = oFFSET_StgRegTable_rL1 dflags
diff --git a/compiler/llvmGen/LlvmCodeGen/Base.hs b/compiler/llvmGen/LlvmCodeGen/Base.hs
index 6ae3c4252d..9142c35064 100644
--- a/compiler/llvmGen/LlvmCodeGen/Base.hs
+++ b/compiler/llvmGen/LlvmCodeGen/Base.hs
@@ -155,6 +155,7 @@ llvmFunArgs dflags live =
           isSSE (FloatReg _)  = True
           isSSE (DoubleReg _) = True
           isSSE (XmmReg _)    = True
+          isSSE (YmmReg _)    = True
           isSSE _             = False
 
 -- | Llvm standard fun attributes
diff --git a/compiler/llvmGen/LlvmCodeGen/CodeGen.hs b/compiler/llvmGen/LlvmCodeGen/CodeGen.hs
index c52640b17f..08aa45fa71 100644
--- a/compiler/llvmGen/LlvmCodeGen/CodeGen.hs
+++ b/compiler/llvmGen/LlvmCodeGen/CodeGen.hs
@@ -1533,6 +1533,7 @@ funEpilogue live = do
         isSSE (FloatReg _)  = True
         isSSE (DoubleReg _) = True
         isSSE (XmmReg _)    = True
+        isSSE (YmmReg _)    = True
         isSSE _             = False
 
     -- Set to value or "undef" depending on whether the register is
diff --git a/compiler/llvmGen/LlvmCodeGen/Regs.hs b/compiler/llvmGen/LlvmCodeGen/Regs.hs
index 1b87929499..16d1237c23 100644
--- a/compiler/llvmGen/LlvmCodeGen/Regs.hs
+++ b/compiler/llvmGen/LlvmCodeGen/Regs.hs
@@ -62,6 +62,12 @@ lmGlobalReg dflags suf reg
         XmmReg 4       -> xmmGlobal $ "XMM4" ++ suf
         XmmReg 5       -> xmmGlobal $ "XMM5" ++ suf
         XmmReg 6       -> xmmGlobal $ "XMM6" ++ suf
+        YmmReg 1       -> ymmGlobal $ "YMM1" ++ suf
+        YmmReg 2       -> ymmGlobal $ "YMM2" ++ suf
+        YmmReg 3       -> ymmGlobal $ "YMM3" ++ suf
+        YmmReg 4       -> ymmGlobal $ "YMM4" ++ suf
+        YmmReg 5       -> ymmGlobal $ "YMM5" ++ suf
+        YmmReg 6       -> ymmGlobal $ "YMM6" ++ suf
         _other         -> panic $ "LlvmCodeGen.Reg: GlobalReg (" ++ (show reg)
                                 ++ ") not supported!"
         -- LongReg, HpLim, CCSS, CurrentTSO, CurrentNusery, HpAlloc
@@ -72,6 +78,7 @@ lmGlobalReg dflags suf reg
         floatGlobal  name = LMNLocalVar (fsLit name) LMFloat
         doubleGlobal name = LMNLocalVar (fsLit name) LMDouble
         xmmGlobal    name = LMNLocalVar (fsLit name) (LMVector 4 (LMInt 32))
+        ymmGlobal    name = LMNLocalVar (fsLit name) (LMVector 8 (LMInt 32))
 
 -- | A list of STG Registers that should always be considered alive
 alwaysLive :: [GlobalReg]