From e074c1c23307649bd2d240d40a73d3829b06a4b4 Mon Sep 17 00:00:00 2001
From: Geoffrey Mainland <gmainlan@microsoft.com>
Date: Thu, 14 Feb 2013 18:52:00 +0000
Subject: Add support for 256-bit-wide vectors.

---
 compiler/codeGen/StgCmmArgRep.hs | 11 ++++++++---
 compiler/codeGen/StgCmmLayout.hs |  1 +
 2 files changed, 9 insertions(+), 3 deletions(-)

(limited to 'compiler/codeGen')

diff --git a/compiler/codeGen/StgCmmArgRep.hs b/compiler/codeGen/StgCmmArgRep.hs
index bd228d4617..26569cffdd 100644
--- a/compiler/codeGen/StgCmmArgRep.hs
+++ b/compiler/codeGen/StgCmmArgRep.hs
@@ -47,6 +47,7 @@ data ArgRep = P   -- GC Ptr
             | F   -- Float
             | D   -- Double
             | V16 -- 16-byte (128-bit) vectors of Float/Double/Int8/Word32/etc.
+            | V32 -- 32-byte (256-bit) vectors of Float/Double/Int8/Word32/etc.
 instance Outputable ArgRep where ppr = text . argRepString
 
 argRepString :: ArgRep -> String
@@ -57,6 +58,7 @@ argRepString V = "V"
 argRepString F = "F"
 argRepString D = "D"
 argRepString V16 = "V16"
+argRepString V32 = "V32"
 
 toArgRep :: PrimRep -> ArgRep
 toArgRep VoidRep           = V
@@ -68,9 +70,10 @@ toArgRep Int64Rep          = L
 toArgRep Word64Rep         = L
 toArgRep FloatRep          = F
 toArgRep DoubleRep         = D
-toArgRep (VecRep len elem)
-    | len*primElemRepSizeB elem == 16 = V16
-    | otherwise                       = error "toArgRep: bad vector primrep"
+toArgRep (VecRep len elem) = case len*primElemRepSizeB elem of
+                               16 -> V16
+                               32 -> V32
+                               _  -> error "toArgRep: bad vector primrep"
 
 isNonV :: ArgRep -> Bool
 isNonV V = False
@@ -84,6 +87,7 @@ argRepSizeW dflags L   = wORD64_SIZE        `quot` wORD_SIZE dflags
 argRepSizeW dflags D   = dOUBLE_SIZE dflags `quot` wORD_SIZE dflags
 argRepSizeW _      V   = 0
 argRepSizeW dflags V16 = 16                 `quot` wORD_SIZE dflags
+argRepSizeW dflags V32 = 32                 `quot` wORD_SIZE dflags
 
 idArgRep :: Id -> ArgRep
 idArgRep = toArgRep . idPrimRep
@@ -132,4 +136,5 @@ slowCallPattern (F: _)                = (fsLit "stg_ap_f", 1)
 slowCallPattern (D: _)                = (fsLit "stg_ap_d", 1)
 slowCallPattern (L: _)                = (fsLit "stg_ap_l", 1)
 slowCallPattern (V16: _)              = (fsLit "stg_ap_v16", 1)
+slowCallPattern (V32: _)              = (fsLit "stg_ap_v32", 1)
 slowCallPattern []                    = (fsLit "stg_ap_0", 0)
diff --git a/compiler/codeGen/StgCmmLayout.hs b/compiler/codeGen/StgCmmLayout.hs
index b52d4e57df..30c2d36ff2 100644
--- a/compiler/codeGen/StgCmmLayout.hs
+++ b/compiler/codeGen/StgCmmLayout.hs
@@ -385,6 +385,7 @@ stdPattern reps
         [D]   -> Just ARG_D
         [L]   -> Just ARG_L
         [V16] -> Just ARG_V16
+        [V32] -> Just ARG_V32
 
         [N,N] -> Just ARG_NN
         [N,P] -> Just ARG_NP
-- 
cgit v1.2.1