gallium: Fix llvmpipe on big-endian machines

Squashed commit of the following:

commit 0857a7e105bfcbc4d1431b2cc56612094c747ca3
Author: Richard Sandiford <[email protected]>
Date:   Tue Jun 18 12:25:07 2013 -0400

    gallivm: Fix lp_build_rgba8_to_fi32_soa for big endian

    Reviewed-by: Adam Jackson <[email protected]>
    Signed-off-by: Richard Sandiford <[email protected]>

commit 0d65131649a8aa140e2db228ba779d685c4333e3
Author: Richard Sandiford <[email protected]>
Date:   Tue Jun 18 12:25:07 2013 -0400

    gallivm: Fix big-endian machines

    This adds a bit-shift count to the format table, and adds the concept of
    vector or bitwise alignment on gathers.

    Reviewed-by: Adam Jackson <[email protected]>
    Signed-off-by: Richard Sandiford <[email protected]>

commit 9740bda9b7dc894b629ed38be9b51059ce90818f
Author: Richard Sandiford <[email protected]>
Date:   Tue Jun 18 12:25:07 2013 -0400

    llvmpipe: Fix convert_to_blend_type on big-endian

    Reviewed-by: Adam Jackson <[email protected]>
    Signed-off-by: Richard Sandiford <[email protected]>

commit ae037c2de0f029e4e99371c0de25560484f0d8df
Author: Richard Sandiford <[email protected]>
Date:   Tue Jun 18 12:25:06 2013 -0400

    util: Convert color pack to packed formats

    This fixes them on big-endian.

    Reviewed-by: Adam Jackson <[email protected]>
    Signed-off-by: Richard Sandiford <[email protected]>

commit 5b05ac0c89ae092ea8ba5bba9f739708d7396b5c
Author: Richard Sandiford <[email protected]>
Date:   Tue Jun 18 12:25:06 2013 -0400

    graw-xlib: Convert to packed formats

    Reviewed-by: Adam Jackson <[email protected]>
    Signed-off-by: Richard Sandiford <[email protected]>

commit 51396e7d098cb6ff794391cf11afe4dbf86dbea0
Author: Richard Sandiford <[email protected]>
Date:   Tue Jun 18 12:25:06 2013 -0400

    format: Convert to packed formats

    Reviewed-by: Adam Jackson <[email protected]>
    Signed-off-by: Richard Sandiford <[email protected]>

commit 417b60bc66eb450e68a92ab0e47f76e292b385e6
Author: Adam Jackson <[email protected]>
Date:   Tue Jun 18 12:25:06 2013 -0400

    st/dri: Convert to packed formats

    Reviewed-by: Adam Jackson <[email protected]>
    Signed-off-by: Richard Sandiford <[email protected]>

commit 0934b2e022a5e0847d312c40734e2b44cac52fd8
Author: Richard Sandiford <[email protected]>
Date:   Tue Jun 18 12:25:06 2013 -0400

    st/xlib: Convert to packed formats

    Reviewed-by: Adam Jackson <[email protected]>
    Signed-off-by: Richard Sandiford <[email protected]>

commit a307ea3c3716a706963acce7966b5e405ba11db9
Author: Richard Sandiford <[email protected]>
Date:   Tue Jun 18 12:25:06 2013 -0400

    gbm: Convert to packed formats

    Reviewed-by: Adam Jackson <[email protected]>
    Signed-off-by: Richard Sandiford <[email protected]>

commit 53eebdd253e1960a645ea278f31d7ef6a6cf4aeb
Author: Richard Sandiford <[email protected]>
Date:   Tue Jun 18 12:25:06 2013 -0400

    tests: Convert to packed formats

    Reviewed-by: Adam Jackson <[email protected]>
    Signed-off-by: Richard Sandiford <[email protected]>

commit 2f77fe3ee524945eacd546efcac34f7799fb3124
Author: Adam Jackson <[email protected]>
Date:   Tue Jun 18 13:07:37 2013 -0400

    gallium: Document packed formats

    Signed-off-by: Adam Jackson <[email protected]>

commit 1f1017159ce951f922210a430de9229f91f62714
Author: Richard Sandiford <[email protected]>
Date:   Tue Jun 18 12:25:06 2013 -0400

    gallium: Introduce 32-bit packed format names

    These are for interacting with buffers natively described in terms of
    bit shifts, like X11 visuals:

        uint32_t xyzw8888 = (x << 0) | (y << 8) | (z << 16) | (w << 24);

    Define these in terms of (endian-dependent) aliases to the array-style
    format names.

    Reviewed-by: Adam Jackson <[email protected]>
    Signed-off-by: Richard Sandiford <[email protected]>

commit 6cc7ab1ee66ed668da78c1d951dfd7782b4e786a
Author: Adam Jackson <[email protected]>
Date:   Mon Jun 3 12:10:32 2013 -0400

    gallium: Document format name conventions

    v2:
    - Fix a channel name thinko (Michel Dänzer)
    - Elaborate on SCALED versus INT
    - Add links to DirectX and FOURCC docs

    Signed-off-by: Adam Jackson <[email protected]>

commit df4d269e7fb62051a3c029b84147465001e5776e
Author: Adam Jackson <[email protected]>
Date:   Tue Jun 18 12:25:06 2013 -0400

    gallivm: Remove all notion of byte-swapping

    Signed-off-by: Adam Jackson <[email protected]>

Signed-off-by: Adam Jackson <[email protected]>

10 files changed, 104 insertions, 208 deletions

diff --git a/src/gallium/auxiliary/gallivm/lp_bld_conv.c b/src/gallium/auxiliary/gallivm/lp_bld_conv.c
index f11361ac06b..cbea966882a 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_conv.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_conv.c
@@ -80,82 +80,6 @@
 
 
 /**
- * Byte swap on element. It will construct a call to intrinsic llvm.bswap
- * based on the type.
- *
- * @param res           element to byte swap.
- * @param type          int16_t, int32_t, int64_t, float or double
- * @param 
- */
-LLVMValueRef
-lp_build_bswap(struct gallivm_state *gallivm,
-               LLVMValueRef res,
-               struct lp_type type)
-{
-   LLVMTypeRef int_type = LLVMIntTypeInContext(gallivm->context,
-                                               type.width);
-   const char *intrinsic = NULL;
-   if (type.width == 8)
-      return res;
-   if (type.width == 16)
-      intrinsic = "llvm.bswap.i16";
-   else if (type.width == 32)
-     intrinsic = "llvm.bswap.i32";
-   else if (type.width == 64)
-      intrinsic = "llvm.bswap.i64";
-
-   assert (intrinsic != NULL);
-
-   /* In case of a floating-point type cast to a int of same size and then
-    * cast back to fp type.
-    */
-   if (type.floating)
-      res = LLVMBuildBitCast(gallivm->builder, res, int_type, "");
-   res = lp_build_intrinsic_unary(gallivm->builder, intrinsic, int_type, res);
-   if (type.floating)
-      res = LLVMBuildBitCast(gallivm->builder, res,
-                             lp_build_elem_type(gallivm, type), "");
-   return res;
-}
-
-
-/**
- * Byte swap every element in the vector.
- *
- * @param packed        <vector> to convert
- * @param src_type      <vector> type of int16_t, int32_t, int64_t, float or
- *                      double
- * @param dst_type      <vector> type to return
- */
-LLVMValueRef
-lp_build_bswap_vec(struct gallivm_state *gallivm,
-                   LLVMValueRef packed,
-                   struct lp_type src_type_vec,
-                   struct lp_type dst_type_vec)
-{
-   LLVMBuilderRef builder = gallivm->builder;
-   LLVMTypeRef dst_type = lp_build_elem_type(gallivm, dst_type_vec);
-   LLVMValueRef res;
-
-   if (src_type_vec.length == 1) {
-      res = lp_build_bswap(gallivm, packed, src_type_vec);
-      res = LLVMBuildBitCast(gallivm->builder, res, dst_type, "");
-   } else {
-      unsigned i;
-      res = LLVMGetUndef(lp_build_vec_type(gallivm, dst_type_vec));
-      for (i = 0; i < src_type_vec.length; ++i) {
-         LLVMValueRef index = lp_build_const_int32(gallivm, i);
-         LLVMValueRef elem = LLVMBuildExtractElement(builder, packed, index, "");
-         elem = lp_build_bswap(gallivm, elem, src_type_vec);
-         elem = LLVMBuildBitCast(gallivm->builder, elem, dst_type, "");
-         res = LLVMBuildInsertElement(gallivm->builder, res, elem, index, "");
-      }
-   }
-   return res;
-}
-
-
-/**
  * Converts int16 half-float to float32
  * Note this can be performed in 1 instruction if vcvtph2ps exists (f16c/cvt16)
  * [llvm.x86.vcvtph2ps / _mm_cvtph_ps]
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_conv.h b/src/gallium/auxiliary/gallivm/lp_bld_conv.h
index d7dfed85187..42a11137473 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_conv.h
+++ b/src/gallium/auxiliary/gallivm/lp_bld_conv.h
@@ -43,17 +43,6 @@
 struct lp_type;
 
 LLVMValueRef
-lp_build_bswap(struct gallivm_state *gallivm,
-               LLVMValueRef res,
-               struct lp_type type);
-
-LLVMValueRef
-lp_build_bswap_vec(struct gallivm_state *gallivm,
-                   LLVMValueRef packed,
-                   struct lp_type src_type,
-                   struct lp_type dst_type);
-
-LLVMValueRef
 lp_build_half_to_float(struct gallivm_state *gallivm,
                        LLVMValueRef src);
 
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_format_aos.c b/src/gallium/auxiliary/gallivm/lp_bld_format_aos.c
index 6a1bf6765d7..af755d460c1 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_format_aos.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_format_aos.c
@@ -139,12 +139,12 @@ format_matches_type(const struct util_format_description *desc,
 
 
 /**
- * Unpack a single pixel into its RGBA components.
+ * Unpack a single pixel into its XYZW components.
  *
  * @param desc  the pixel format for the packed pixel value
  * @param packed integer pixel in a format such as PIPE_FORMAT_B8G8R8A8_UNORM
  *
- * @return RGBA in a float[4] or ubyte[4] or ushort[4] vector.
+ * @return XYZW in a float[4] or ubyte[4] or ushort[4] vector.
  */
 static INLINE LLVMValueRef
 lp_build_unpack_arith_rgba_aos(struct gallivm_state *gallivm,
@@ -159,7 +159,6 @@ lp_build_unpack_arith_rgba_aos(struct gallivm_state *gallivm,
 
    boolean normalized;
    boolean needs_uitofp;
-   unsigned shift;
    unsigned i;
 
    /* TODO: Support more formats */
@@ -172,10 +171,6 @@ lp_build_unpack_arith_rgba_aos(struct gallivm_state *gallivm,
     * matches floating point size */
    assert (LLVMTypeOf(packed) == LLVMInt32TypeInContext(gallivm->context));
 
-#ifdef PIPE_ARCH_BIG_ENDIAN
-   packed = lp_build_bswap(gallivm, packed, lp_type_uint(32));
-#endif
-
    /* Broadcast the packed value to all four channels
     * before: packed = BGRA
     * after: packed = {BGRA, BGRA, BGRA, BGRA}
@@ -194,11 +189,11 @@ lp_build_unpack_arith_rgba_aos(struct gallivm_state *gallivm,
    /* Initialize vector constants */
    normalized = FALSE;
    needs_uitofp = FALSE;
-   shift = 0;
 
    /* Loop over 4 color components */
    for (i = 0; i < 4; ++i) {
       unsigned bits = desc->channel[i].size;
+      unsigned shift = desc->channel[i].shift;
 
       if (desc->channel[i].type == UTIL_FORMAT_TYPE_VOID) {
          shifts[i] = LLVMGetUndef(LLVMInt32TypeInContext(gallivm->context));
@@ -224,12 +219,10 @@ lp_build_unpack_arith_rgba_aos(struct gallivm_state *gallivm,
          else
             scales[i] =  lp_build_const_float(gallivm, 1.0);
       }
-
-      shift += bits;
    }
 
-   /* Ex: convert packed = {BGRA, BGRA, BGRA, BGRA}
-    * into masked = {B, G, R, A}
+   /* Ex: convert packed = {XYZW, XYZW, XYZW, XYZW}
+    * into masked = {X, Y, Z, W}
     */
    shifted = LLVMBuildLShr(builder, packed, LLVMConstVector(shifts, 4), "");
    masked = LLVMBuildAnd(builder, shifted, LLVMConstVector(masks, 4), "");
@@ -276,7 +269,6 @@ lp_build_pack_rgba_aos(struct gallivm_state *gallivm,
    LLVMValueRef shifts[4];
    LLVMValueRef scales[4];
    boolean normalized;
-   unsigned shift;
    unsigned i, j;
 
    assert(desc->layout == UTIL_FORMAT_LAYOUT_PLAIN);
@@ -302,9 +294,9 @@ lp_build_pack_rgba_aos(struct gallivm_state *gallivm,
                                        LLVMConstVector(swizzles, 4), "");
 
    normalized = FALSE;
-   shift = 0;
    for (i = 0; i < 4; ++i) {
       unsigned bits = desc->channel[i].size;
+      unsigned shift = desc->channel[i].shift;
 
       if (desc->channel[i].type == UTIL_FORMAT_TYPE_VOID) {
          shifts[i] = LLVMGetUndef(LLVMInt32TypeInContext(gallivm->context));
@@ -325,8 +317,6 @@ lp_build_pack_rgba_aos(struct gallivm_state *gallivm,
          else
             scales[i] = lp_build_const_float(gallivm, 1.0);
       }
-
-      shift += bits;
    }
 
    if (normalized)
@@ -410,16 +400,11 @@ lp_build_fetch_rgba_aos(struct gallivm_state *gallivm,
 
       packed = lp_build_gather(gallivm, type.length/4,
                                format_desc->block.bits, type.width*4,
-                               base_ptr, offset);
+                               base_ptr, offset, TRUE);
 
       assert(format_desc->block.bits <= vec_len);
 
       packed = LLVMBuildBitCast(gallivm->builder, packed, dst_vec_type, "");
-#ifdef PIPE_ARCH_BIG_ENDIAN
-      if (type.floating)
-         packed = lp_build_bswap_vec(gallivm, packed, type,
-                                    lp_type_float_vec(type.width, vec_len));
-#endif
       return lp_build_format_swizzle_aos(format_desc, &bld, packed);
    }
 
@@ -453,7 +438,7 @@ lp_build_fetch_rgba_aos(struct gallivm_state *gallivm,
 
          packed = lp_build_gather_elem(gallivm, num_pixels,
                                        format_desc->block.bits, 32,
-                                       base_ptr, offset, k);
+                                       base_ptr, offset, k, FALSE);
 
          tmps[k] = lp_build_unpack_arith_rgba_aos(gallivm,
                                                   format_desc,
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_format_aos_array.c b/src/gallium/auxiliary/gallivm/lp_bld_format_aos_array.c
index 3402a0b1530..ee3ca86c000 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_format_aos_array.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_format_aos_array.c
@@ -40,58 +40,6 @@
 #include "pipe/p_state.h"
 
 
-#ifdef PIPE_ARCH_BIG_ENDIAN
-static LLVMValueRef
-lp_build_read_int_bswap(struct gallivm_state *gallivm,
-                        LLVMValueRef base_ptr,
-                        unsigned src_width,
-                        LLVMTypeRef src_type,
-                        unsigned i,
-                        LLVMTypeRef dst_type)
-{
-   LLVMBuilderRef builder = gallivm->builder;
-   LLVMValueRef index = lp_build_const_int32(gallivm, i);
-   LLVMValueRef ptr = LLVMBuildGEP(builder, base_ptr, &index, 1, "");
-   LLVMValueRef res = LLVMBuildLoad(builder, ptr, "");
-   res = lp_build_bswap(gallivm, res, lp_type_uint(src_width));
-   return LLVMBuildBitCast(builder, res, dst_type, "");
-}
-
-static LLVMValueRef
-lp_build_fetch_read_big_endian(struct gallivm_state *gallivm,
-                               struct lp_type src_type,
-                               LLVMValueRef base_ptr)
-{
-   LLVMBuilderRef builder = gallivm->builder;
-   unsigned src_width = src_type.width;
-   unsigned length = src_type.length;
-   LLVMTypeRef src_elem_type = LLVMIntTypeInContext(gallivm->context, src_width);
-   LLVMTypeRef dst_elem_type = lp_build_elem_type (gallivm, src_type);
-   LLVMTypeRef src_ptr_type = LLVMPointerType(src_elem_type, 0);
-   LLVMValueRef res;
-
-   base_ptr = LLVMBuildPointerCast(builder, base_ptr, src_ptr_type, "");
-   if (length == 1) {
-      /* Scalar */
-      res = lp_build_read_int_bswap(gallivm, base_ptr, src_width, src_elem_type,
-                                    0, dst_elem_type);
-   } else {
-      /* Vector */
-      LLVMTypeRef dst_vec_type = LLVMVectorType(dst_elem_type, length);
-      unsigned i;
-
-      res = LLVMGetUndef(dst_vec_type);
-      for (i = 0; i < length; ++i) {
-         LLVMValueRef index = lp_build_const_int32(gallivm, i);
-         LLVMValueRef elem = lp_build_read_int_bswap(gallivm, base_ptr, src_width,
-                                                     src_elem_type, i, dst_elem_type);
-         res = LLVMBuildInsertElement(builder, res, elem, index, "");
-      }
-   }
-
-   return res;
-}
-#endif
 
 /**
  * @brief lp_build_fetch_rgba_aos_array
@@ -124,13 +72,9 @@ lp_build_fetch_rgba_aos_array(struct gallivm_state *gallivm,
 
    /* Read whole vector from memory, unaligned */
    ptr = LLVMBuildGEP(builder, base_ptr, &offset, 1, "");
-#ifdef PIPE_ARCH_BIG_ENDIAN
-   res = lp_build_fetch_read_big_endian(gallivm, src_type, ptr);
-#else
    ptr = LLVMBuildPointerCast(builder, ptr, LLVMPointerType(src_vec_type, 0), "");
    res = LLVMBuildLoad(builder, ptr, "");
    lp_set_load_alignment(res, src_type.width / 8);
-#endif
 
    /* Truncate doubles to float */
    if (src_type.floating && src_type.width == 64) {
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_format_soa.c b/src/gallium/auxiliary/gallivm/lp_bld_format_soa.c
index edc24424289..4c6bd81047a 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_format_soa.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_format_soa.c
@@ -115,7 +115,6 @@ lp_build_unpack_rgba_soa(struct gallivm_state *gallivm,
    LLVMBuilderRef builder = gallivm->builder;
    struct lp_build_context bld;
    LLVMValueRef inputs[4];
-   unsigned start;
    unsigned chan;
 
    assert(format_desc->layout == UTIL_FORMAT_LAYOUT_PLAIN);
@@ -128,9 +127,9 @@ lp_build_unpack_rgba_soa(struct gallivm_state *gallivm,
    lp_build_context_init(&bld, gallivm, type);
 
    /* Decode the input vector components */
-   start = 0;
    for (chan = 0; chan < format_desc->nr_channels; ++chan) {
       const unsigned width = format_desc->channel[chan].size;
+      const unsigned start = format_desc->channel[chan].shift;
       const unsigned stop = start + width;
       LLVMValueRef input;
 
@@ -256,8 +255,6 @@ lp_build_unpack_rgba_soa(struct gallivm_state *gallivm,
       }
 
       inputs[chan] = input;
-
-      start = stop;
    }
 
    lp_build_format_swizzle_soa(format_desc, &bld, inputs, rgba_out);
@@ -291,7 +288,11 @@ lp_build_rgba8_to_fi32_soa(struct gallivm_state *gallivm,
 
    /* Decode the input vector components */
    for (chan = 0; chan < 4; ++chan) {
+#ifdef PIPE_ARCH_LITTLE_ENDIAN
       unsigned start = chan*8;
+#else
+      unsigned start = (3-chan)*8;
+#endif
       unsigned stop = start + 8;
       LLVMValueRef input;
 
@@ -360,13 +361,14 @@ lp_build_fetch_rgba_soa(struct gallivm_state *gallivm,
 
       /*
        * gather the texels from the texture
-       * Ex: packed = {BGRA, BGRA, BGRA, BGRA}.
+       * Ex: packed = {XYZW, XYZW, XYZW, XYZW}
        */
+      assert(format_desc->block.bits <= type.width);
       packed = lp_build_gather(gallivm,
                                type.length,
                                format_desc->block.bits,
                                type.width,
-                               base_ptr, offset);
+                               base_ptr, offset, FALSE);
 
       /*
        * convert texels to float rgba
@@ -391,7 +393,8 @@ lp_build_fetch_rgba_soa(struct gallivm_state *gallivm,
 
       packed = lp_build_gather(gallivm, type.length,
                                format_desc->block.bits,
-                               type.width, base_ptr, offset);
+                               type.width, base_ptr, offset,
+			       FALSE);
       if (format_desc->format == PIPE_FORMAT_R11G11B10_FLOAT) {
          lp_build_r11g11b10_to_float(gallivm, packed, rgba_out);
       }
@@ -418,14 +421,14 @@ lp_build_fetch_rgba_soa(struct gallivm_state *gallivm,
          LLVMValueRef s_offset = lp_build_const_int_vec(gallivm, type, 4);
          offset = LLVMBuildAdd(builder, offset, s_offset, "");
          packed = lp_build_gather(gallivm, type.length,
-                                  32, type.width, base_ptr, offset);
+                                  32, type.width, base_ptr, offset, FALSE);
          packed = LLVMBuildAnd(builder, packed,
                                lp_build_const_int_vec(gallivm, type, mask), "");
       }
       else {
          assert (format_desc->format == PIPE_FORMAT_Z32_FLOAT_S8X24_UINT);
          packed = lp_build_gather(gallivm, type.length,
-                                  32, type.width, base_ptr, offset);
+                                  32, type.width, base_ptr, offset, TRUE);
          packed = LLVMBuildBitCast(builder, packed,
                                    lp_build_vec_type(gallivm, type), "");
       }
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_format_yuv.c b/src/gallium/auxiliary/gallivm/lp_bld_format_yuv.c
index e542abce0cc..873f354c040 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_format_yuv.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_format_yuv.c
@@ -497,7 +497,7 @@ lp_build_fetch_subsampled_rgba_aos(struct gallivm_state *gallivm,
    assert(format_desc->block.width == 2);
    assert(format_desc->block.height == 1);
 
-   packed = lp_build_gather(gallivm, n, 32, 32, base_ptr, offset);
+   packed = lp_build_gather(gallivm, n, 32, 32, base_ptr, offset, FALSE);
 
    (void)j;
 
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_gather.c b/src/gallium/auxiliary/gallivm/lp_bld_gather.c
index eefb23a946f..9155d811c06 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_gather.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_gather.c
@@ -78,7 +78,8 @@ lp_build_gather_elem(struct gallivm_state *gallivm,
                      unsigned dst_width,
                      LLVMValueRef base_ptr,
                      LLVMValueRef offsets,
-                     unsigned i)
+                     unsigned i,
+                     boolean vector_justify)
 {
    LLVMTypeRef src_type = LLVMIntTypeInContext(gallivm->context, src_width);
    LLVMTypeRef src_ptr_type = LLVMPointerType(src_type, 0);
@@ -97,10 +98,12 @@ lp_build_gather_elem(struct gallivm_state *gallivm,
       res = LLVMBuildTrunc(gallivm->builder, res, dst_elem_type, "");
    } else if (src_width < dst_width) {
       res = LLVMBuildZExt(gallivm->builder, res, dst_elem_type, "");
+      if (vector_justify) {
 #ifdef PIPE_ARCH_BIG_ENDIAN
-      res = LLVMBuildShl(gallivm->builder, res,
-                         LLVMConstInt(dst_elem_type, dst_width - src_width, 0), "");
+         res = LLVMBuildShl(gallivm->builder, res,
+                            LLVMConstInt(dst_elem_type, dst_width - src_width, 0), "");
 #endif
+      }
    }
 
    return res;
@@ -112,11 +115,20 @@ lp_build_gather_elem(struct gallivm_state *gallivm,
  * Use for fetching texels from a texture.
  * For SSE, typical values are length=4, src_width=32, dst_width=32.
  *
+ * When src_width < dst_width, the return value can be justified in
+ * one of two ways:
+ * "integer justification" is used when the caller treats the destination
+ * as a packed integer bitmask, as described by the channels' "shift" and
+ * "width" fields;
+ * "vector justification" is used when the caller casts the destination
+ * to a vector and needs channel X to be in vector element 0.
+ *
  * @param length length of the offsets
  * @param src_width src element width in bits
  * @param dst_width result element width in bits (src will be expanded to fit)
  * @param base_ptr base pointer, should be a i8 pointer type.
  * @param offsets vector with offsets
+ * @param vector_justify select vector rather than integer justification
  */
 LLVMValueRef
 lp_build_gather(struct gallivm_state *gallivm,
@@ -124,7 +136,8 @@ lp_build_gather(struct gallivm_state *gallivm,
                 unsigned src_width,
                 unsigned dst_width,
                 LLVMValueRef base_ptr,
-                LLVMValueRef offsets)
+                LLVMValueRef offsets,
+                boolean vector_justify)
 {
    LLVMValueRef res;
 
@@ -132,7 +145,7 @@ lp_build_gather(struct gallivm_state *gallivm,
       /* Scalar */
       return lp_build_gather_elem(gallivm, length,
                                   src_width, dst_width,
-                                  base_ptr, offsets, 0);
+                                  base_ptr, offsets, 0, vector_justify);
    } else {
       /* Vector */
 
@@ -146,7 +159,7 @@ lp_build_gather(struct gallivm_state *gallivm,
          LLVMValueRef elem;
          elem = lp_build_gather_elem(gallivm, length,
                                      src_width, dst_width,
-                                     base_ptr, offsets, i);
+                                     base_ptr, offsets, i, vector_justify);
          res = LLVMBuildInsertElement(gallivm->builder, res, elem, index, "");
       }
    }
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_gather.h b/src/gallium/auxiliary/gallivm/lp_bld_gather.h
index 8e4c07d82b8..ee694732d36 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_gather.h
+++ b/src/gallium/auxiliary/gallivm/lp_bld_gather.h
@@ -47,7 +47,8 @@ lp_build_gather_elem(struct gallivm_state *gallivm,
                      unsigned dst_width,
                      LLVMValueRef base_ptr,
                      LLVMValueRef offsets,
-                     unsigned i);
+                     unsigned i,
+                     boolean vector_justify);
 
 LLVMValueRef
 lp_build_gather(struct gallivm_state *gallivm,
@@ -55,7 +56,8 @@ lp_build_gather(struct gallivm_state *gallivm,
                 unsigned src_width,
                 unsigned dst_width,
                 LLVMValueRef base_ptr,
-                LLVMValueRef offsets);
+                LLVMValueRef offsets,
+                boolean vector_justify);
 
 LLVMValueRef
 lp_build_gather_values(struct gallivm_state * gallivm,
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_sample_aos.c b/src/gallium/auxiliary/gallivm/lp_bld_sample_aos.c
index c31b05d7022..104c24d83b9 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_sample_aos.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_sample_aos.c
@@ -531,7 +531,7 @@ lp_build_sample_fetch_image_nearest(struct lp_build_sample_context *bld,
                               bld->texel_type.length,
                               bld->format_desc->block.bits,
                               bld->texel_type.width,
-                              data_ptr, offset);
+                              data_ptr, offset, TRUE);
 
       rgba8 = LLVMBuildBitCast(builder, rgba8, u8n_vec_type, "");
    }
@@ -893,7 +893,7 @@ lp_build_sample_fetch_image_linear(struct lp_build_sample_context *bld,
                                        bld->texel_type.length,
                                        bld->format_desc->block.bits,
                                        bld->texel_type.width,
-                                       data_ptr, offset[k][j][i]);
+                                       data_ptr, offset[k][j][i], TRUE);
 
                rgba8 = LLVMBuildBitCast(builder, rgba8, u8n_vec_type, "");
             }
diff --git a/src/gallium/auxiliary/gallivm/lp_bld_swizzle.c b/src/gallium/auxiliary/gallivm/lp_bld_swizzle.c
index 08d817a28f5..79116bcf391 100644
--- a/src/gallium/auxiliary/gallivm/lp_bld_swizzle.c
+++ b/src/gallium/auxiliary/gallivm/lp_bld_swizzle.c
@@ -217,6 +217,20 @@ lp_build_swizzle_scalar_aos(struct lp_build_context *bld,
 
       a = LLVMBuildBitCast(builder, a, lp_build_vec_type(bld->gallivm, type2), "");
 
+      /*
+       * Vector element 0 is always channel X.
+       *
+       *                        76 54 32 10 (array numbering)
+       * Little endian reg in:  YX YX YX YX
+       * Little endian reg out: YY YY YY YY if shift right (shift == -1)
+       *                        XX XX XX XX if shift left (shift == 1)
+       *
+       *                        01 23 45 67 (array numbering)
+       * Big endian reg in:     XY XY XY XY
+       * Big endian reg out:    YY YY YY YY if shift left (shift == 1)
+       *                        XX XX XX XX if shift right (shift == -1)
+       *
+       */
 #ifdef PIPE_ARCH_LITTLE_ENDIAN
       shift = channel == 0 ? 1 : -1;
 #else
@@ -240,10 +254,23 @@ lp_build_swizzle_scalar_aos(struct lp_build_context *bld,
       /*
        * Bit mask and recursive shifts
        *
+       * Little-endian registers:
+       *
+       *   7654 3210
+       *   WZYX WZYX .... WZYX  <= input
+       *   00Y0 00Y0 .... 00Y0  <= mask
+       *   00YY 00YY .... 00YY  <= shift right 1 (shift amount -1)
+       *   YYYY YYYY .... YYYY  <= shift left 2 (shift amount 2)
+       *
+       * Big-endian registers:
+       *
+       *   0123 4567
        *   XYZW XYZW .... XYZW  <= input
-       *   0Y00 0Y00 .... 0Y00
-       *   YY00 YY00 .... YY00
-       *   YYYY YYYY .... YYYY  <= output
+       *   0Y00 0Y00 .... 0Y00  <= mask
+       *   YY00 YY00 .... YY00  <= shift left 1 (shift amount 1)
+       *   YYYY YYYY .... YYYY  <= shift right 2 (shift amount -2)
+       *
+       * shifts[] gives little-endian shift amounts; we need to negate for big-endian.
        */
       struct lp_type type4;
       const int shifts[4][2] = {
@@ -274,14 +301,15 @@ lp_build_swizzle_scalar_aos(struct lp_build_context *bld,
          LLVMValueRef tmp = NULL;
          int shift = shifts[channel][i];
 
-#ifdef PIPE_ARCH_LITTLE_ENDIAN
+         /* See endianness diagram above */
+#ifdef PIPE_ARCH_BIG_ENDIAN
          shift = -shift;
 #endif
 
          if(shift > 0)
-            tmp = LLVMBuildLShr(builder, a, lp_build_const_int_vec(bld->gallivm, type4, shift*type.width), "");
+            tmp = LLVMBuildShl(builder, a, lp_build_const_int_vec(bld->gallivm, type4, shift*type.width), "");
          if(shift < 0)
-            tmp = LLVMBuildShl(builder, a, lp_build_const_int_vec(bld->gallivm, type4, -shift*type.width), "");
+            tmp = LLVMBuildLShr(builder, a, lp_build_const_int_vec(bld->gallivm, type4, -shift*type.width), "");
 
          assert(tmp);
          if(tmp)
@@ -474,21 +502,39 @@ lp_build_swizzle_aos(struct lp_build_context *bld,
 
       /*
        * Mask and shift the channels, trying to group as many channels in the
-       * same shift as possible
+       * same shift as possible.  The shift amount is positive for shifts left
+       * and negative for shifts right.
        */
       for (shift = -3; shift <= 3; ++shift) {
          uint64_t mask = 0;
 
          assert(type4.width <= sizeof(mask)*8);
 
+         /*
+          * Vector element numbers follow the XYZW order, so 0 is always X, etc.
+          * After widening 4 times we have:
+          *
+          *                                3210
+          * Little-endian register layout: WZYX
+          *
+          *                                0123
+          * Big-endian register layout:    XYZW
+          *
+          * For little-endian, higher-numbered channels are obtained by a shift right
+          * (negative shift amount) and lower-numbered channels by a shift left
+          * (positive shift amount).  The opposite is true for big-endian.
+          */
          for (chan = 0; chan < 4; ++chan) {
-            /* FIXME: big endian */
-            if (swizzles[chan] < 4 &&
-                chan - swizzles[chan] == shift) {
+            if (swizzles[chan] < 4) {
+               /* We need to move channel swizzles[chan] into channel chan */
 #ifdef PIPE_ARCH_LITTLE_ENDIAN
-               mask |= ((1ULL << type.width) - 1) << (swizzles[chan] * type.width);
+               if (swizzles[chan] - chan == -shift) {
+                  mask |= ((1ULL << type.width) - 1) << (swizzles[chan] * type.width);
+               }
 #else
-               mask |= ((1ULL << type.width) - 1) << (type4.width - type.width) >> (swizzles[chan] * type.width);
+               if (swizzles[chan] - chan == shift) {
+                  mask |= ((1ULL << type.width) - 1) << (type4.width - type.width) >> (swizzles[chan] * type.width);
+               }
 #endif
             }
          }
@@ -502,21 +548,11 @@ lp_build_swizzle_aos(struct lp_build_context *bld,
             masked = LLVMBuildAnd(builder, a,
                                   lp_build_const_int_vec(bld->gallivm, type4, mask), "");
             if (shift > 0) {
-#ifdef PIPE_ARCH_LITTLE_ENDIAN
                shifted = LLVMBuildShl(builder, masked,
                                       lp_build_const_int_vec(bld->gallivm, type4, shift*type.width), "");
-#else
-               shifted = LLVMBuildLShr(builder, masked,
-                                       lp_build_const_int_vec(bld->gallivm, type4, shift*type.width), "");
-#endif
             } else if (shift < 0) {
-#ifdef PIPE_ARCH_LITTLE_ENDIAN
                shifted = LLVMBuildLShr(builder, masked,
                                        lp_build_const_int_vec(bld->gallivm, type4, -shift*type.width), "");
-#else
-               shifted = LLVMBuildShl(builder, masked,
-                                      lp_build_const_int_vec(bld->gallivm, type4, -shift*type.width), "");
-#endif
             } else {
                shifted = masked;
             }