Revert "i965: Provide sse2 version for rgba8 <-> bgra8 swizzle"

This reverts commit ab30426e335116e29473faaafe8b57ec760516ee.

Apparently the memory isn't quite as aligned when this gets called
as it should be, causing crashes. (Albeit this looks independent
from this code, should crash just as well if ssse3 is enabled when
compiling without this patch.)
https://bugs.freedesktop.org/show_bug.cgi?id=93962

2 files changed, 12 insertions, 62 deletions

diff --git a/src/mesa/drivers/dri/i965/intel_mipmap_tree.c b/src/mesa/drivers/dri/i965/intel_mipmap_tree.c
index b52bbfb0f67..108dd87dd8b 100644
--- a/src/mesa/drivers/dri/i965/intel_mipmap_tree.c
+++ b/src/mesa/drivers/dri/i965/intel_mipmap_tree.c
@@ -2773,11 +2773,6 @@ intel_miptree_map(struct brw_context *brw,
    } else if (!(mode & GL_MAP_WRITE_BIT) &&
               !mt->compressed && cpu_has_sse4_1 &&
               (mt->pitch % 16 == 0)) {
-      /*
-       * XXX: without sse4_1, in some situations it would be beneficial
-       * to copy regardless (with an ordinary memcpy) as otherwise mesa
-       * may access uncached memory bytewise.
-       */
       intel_miptree_map_movntdqa(brw, mt, map, level, slice);
 #endif
    } else {
diff --git a/src/mesa/drivers/dri/i965/intel_tiled_memcpy.c b/src/mesa/drivers/dri/i965/intel_tiled_memcpy.c
index 671dc597955..2383401d14c 100644
--- a/src/mesa/drivers/dri/i965/intel_tiled_memcpy.c
+++ b/src/mesa/drivers/dri/i965/intel_tiled_memcpy.c
@@ -36,13 +36,10 @@
 #include "brw_context.h"
 #include "intel_tiled_memcpy.h"
 
-#if defined(__SSSE3__)
+#ifdef __SSSE3__
 #include <tmmintrin.h>
-#elif defined(__SSE2__)
-#include <emmintrin.h>
 #endif
 
-
 #define FILE_DEBUG_FLAG DEBUG_TEXTURE
 
 #define ALIGN_DOWN(a, b) ROUND_DOWN_TO(a, b)
@@ -59,65 +56,23 @@ static const uint32_t ytile_width = 128;
 static const uint32_t ytile_height = 32;
 static const uint32_t ytile_span = 16;
 
-#if defined(__SSSE3__)
+#ifdef __SSSE3__
 static const uint8_t rgba8_permutation[16] =
    { 2,1,0,3, 6,5,4,7, 10,9,8,11, 14,13,12,15 };
 
 /* NOTE: dst must be 16-byte aligned. src may be unaligned. */
-static inline void
-rgba8_copy_16_aligned_dst(void *dst, const void *src)
-{
-   _mm_store_si128((__m128i *)(dst),
-                   _mm_shuffle_epi8(_mm_loadu_si128((__m128i *)(src)),
-                                    *(__m128i *)rgba8_permutation));
-}
+#define rgba8_copy_16_aligned_dst(dst, src)                            \
+   _mm_store_si128((__m128i *)(dst),                                   \
+                   _mm_shuffle_epi8(_mm_loadu_si128((__m128i *)(src)), \
+                                    *(__m128i *) rgba8_permutation))
 
 /* NOTE: src must be 16-byte aligned. dst may be unaligned. */
-static inline void
-rgba8_copy_16_aligned_src(void *dst, const void *src)
-{
-   _mm_storeu_si128((__m128i *)(dst),
-                    _mm_shuffle_epi8(_mm_load_si128((__m128i *)(src)),
-                                     *(__m128i *)rgba8_permutation));
-}
-
-#elif defined(__SSE2__)
-static inline void
-rgba8_copy_16_aligned_dst(void *dst, const void *src)
-{
-   __m128i srcreg, dstreg, agmask, ag, rb, br;
-
-   agmask = _mm_set1_epi32(0xFF00FF00);
-   srcreg = _mm_loadu_si128((__m128i *)src);
-
-   rb = _mm_andnot_si128(agmask, srcreg);
-   ag = _mm_and_si128(agmask, srcreg);
-   br = _mm_shufflehi_epi16(_mm_shufflelo_epi16(rb, _MM_SHUFFLE(2, 3, 0, 1)),
-                            _MM_SHUFFLE(2, 3, 0, 1));
-   dstreg = _mm_or_si128(ag, br);
-
-   _mm_store_si128((__m128i *)dst, dstreg);
-}
-
-static inline void
-rgba8_copy_16_aligned_src(void *dst, const void *src)
-{
-   __m128i srcreg, dstreg, agmask, ag, rb, br;
-
-   agmask = _mm_set1_epi32(0xFF00FF00);
-   srcreg = _mm_load_si128((__m128i *)src);
-
-   rb = _mm_andnot_si128(agmask, srcreg);
-   ag = _mm_and_si128(agmask, srcreg);
-   br = _mm_shufflehi_epi16(_mm_shufflelo_epi16(rb, _MM_SHUFFLE(2, 3, 0, 1)),
-                            _MM_SHUFFLE(2, 3, 0, 1));
-   dstreg = _mm_or_si128(ag, br);
-
-   _mm_storeu_si128((__m128i *)dst, dstreg);
-}
+#define rgba8_copy_16_aligned_src(dst, src)                            \
+   _mm_storeu_si128((__m128i *)(dst),                                  \
+                    _mm_shuffle_epi8(_mm_load_si128((__m128i *)(src)), \
+                                     *(__m128i *) rgba8_permutation))
 #endif
 
-
 /**
  * Copy RGBA to BGRA - swap R and B, with the destination 16-byte aligned.
  */
@@ -127,7 +82,7 @@ rgba8_copy_aligned_dst(void *dst, const void *src, size_t bytes)
    uint8_t *d = dst;
    uint8_t const *s = src;
 
-#if defined(__SSSE3__) || defined(__SSE2__)
+#ifdef __SSSE3__
    if (bytes == 16) {
       assert(!(((uintptr_t)dst) & 0xf));
       rgba8_copy_16_aligned_dst(d+ 0, s+ 0);
@@ -165,7 +120,7 @@ rgba8_copy_aligned_src(void *dst, const void *src, size_t bytes)
    uint8_t *d = dst;
    uint8_t const *s = src;
 
-#if defined(__SSSE3__) || defined(__SSE2__)
+#ifdef __SSSE3__
    if (bytes == 16) {
       assert(!(((uintptr_t)src) & 0xf));
       rgba8_copy_16_aligned_src(d+ 0, s+ 0);