X-Git-Url: https://git.libre-soc.org/?a=blobdiff_plain;f=src%2Fgallium%2Fdrivers%2Fllvmpipe%2Flp_tile_soa.py;h=d548ad845c7e2f3b44ebc742457f37443fbdc623;hb=351d3c59f2a1153047d45fcdb23cc487f231683d;hp=e49f9c62fe7c777ac96fd18f1a892ab25834f88b;hpb=4926c5748028d33da4808f8a5473aa7b2f2bdc62;p=mesa.git

diff --git a/src/gallium/drivers/llvmpipe/lp_tile_soa.py b/src/gallium/drivers/llvmpipe/lp_tile_soa.py
index e49f9c62fe7..d548ad845c7 100644
--- a/src/gallium/drivers/llvmpipe/lp_tile_soa.py
+++ b/src/gallium/drivers/llvmpipe/lp_tile_soa.py
@@ -1,6 +1,6 @@
 #!/usr/bin/env python
 
-'''
+CopyRight = '''
 /**************************************************************************
  *
  * Copyright 2009 VMware, Inc.
@@ -51,6 +51,12 @@ def is_format_supported(format):
 
     # FIXME: Ideally we would support any format combination here.
 
+    if format.name == 'PIPE_FORMAT_R11G11B10_FLOAT':
+        return True;
+
+    if format.name == 'PIPE_FORMAT_R9G9B9E5_FLOAT':
+        return True;
+
     if format.layout != PLAIN:
         return False
 
@@ -75,7 +81,7 @@ def generate_format_read(format, dst_channel, dst_native_type, dst_suffix):
     src_native_type = native_type(format)
 
     print 'static void'
-    print 'lp_tile_%s_swizzle_%s(%s *dst, const uint8_t *src, unsigned src_stride, unsigned x0, unsigned y0)' % (name, dst_suffix, dst_native_type)
+    print 'lp_tile_%s_swizzle_%s(%s * restrict dst, const uint8_t * restrict src, unsigned src_stride, unsigned x0, unsigned y0)' % (name, dst_suffix, dst_native_type)
     print '{'
     print '   unsigned x, y;'
     print '   const uint8_t *src_row = src + y0*src_stride;'
@@ -98,7 +104,19 @@ def generate_format_read(format, dst_channel, dst_native_type, dst_suffix):
     else:
         assert False
 
-    if format.layout == PLAIN:
+    if format.name == 'PIPE_FORMAT_R11G11B10_FLOAT':
+        print '         float tmp[3];'
+        print '         uint8_t r, g, b;'
+        print '         r11g11b10f_to_float3(*src_pixel++, tmp);'
+        for i in range(3):
+            print '         %s = tmp[%d] * 0xff;' % (names[i], i)
+    elif format.name == 'PIPE_FORMAT_R9G9B9E5_FLOAT':
+        print '         float tmp[3];'
+        print '         uint8_t r, g, b;'
+        print '         rgb9e5_to_float3(*src_pixel++, tmp);'
+        for i in range(3):
+            print '         %s = tmp[%d] * 0xff;' % (names[i], i)
+    elif format.layout == PLAIN:
         if not format.is_array():
             print '         %s pixel = *src_pixel++;' % src_native_type
             shift = 0;
@@ -235,7 +253,17 @@ def emit_tile_pixel_unswizzle_code(format, src_channel):
     print '      %s *dst_pixel = (%s *)(dst_row + x0*%u);' % (dst_native_type, dst_native_type, format.stride())
     print '      for (x = 0; x < TILE_SIZE; ++x) {'
 
-    if format.layout == PLAIN:
+    if format.name == 'PIPE_FORMAT_R11G11B10_FLOAT':
+        print '         float tmp[3];'
+        for i in range(3):
+            print '         tmp[%d] = ubyte_to_float(TILE_PIXEL(src, x, y, %u));' % (i, inv_swizzle[i])
+        print '         *dst_pixel++ = float3_to_r11g11b10f(tmp);'
+    elif format.name == 'PIPE_FORMAT_R9G9B9E5_FLOAT':
+        print '         float tmp[3];'
+        for i in range(3):
+            print '         tmp[%d] = ubyte_to_float(TILE_PIXEL(src, x, y, %u));' % (i, inv_swizzle[i])
+        print '         *dst_pixel++ = float3_to_rgb9e5(tmp);'
+    elif format.layout == PLAIN:
         if not format.is_array():
             print '         %s pixel = 0;' % dst_native_type
             shift = 0;
@@ -273,7 +301,7 @@ def generate_format_write(format, src_channel, src_native_type, src_suffix):
     name = format.short_name()
 
     print 'static void'
-    print 'lp_tile_%s_unswizzle_%s(const %s *src, uint8_t *dst, unsigned dst_stride, unsigned x0, unsigned y0)' % (name, src_suffix, src_native_type)
+    print 'lp_tile_%s_unswizzle_%s(const %s * restrict src, uint8_t * restrict dst, unsigned dst_stride, unsigned x0, unsigned y0)' % (name, src_suffix, src_native_type)
     print '{'
     if format.layout == PLAIN \
         and format.colorspace == 'rgb' \
@@ -423,6 +451,70 @@ lp_tile_b8g8r8a8_unorm_unswizzle_4ub_sse2(const uint8_t * restrict src,
    }
 }
 
+static void
+lp_tile_b8g8r8x8_unorm_swizzle_4ub_sse2(uint8_t * restrict dst,
+                                        const uint8_t * restrict src, unsigned src_stride,
+                                        unsigned x0, unsigned y0)
+{
+   __m128i *dst128 = (__m128i *) dst;
+   unsigned x, y;
+
+   src += y0 * src_stride;
+   src += x0 * sizeof(uint32_t);
+
+   for (y = 0; y < TILE_SIZE; y += 4) {
+      const uint8_t *src_row = src;
+
+      for (x = 0; x < TILE_SIZE; x += 4) {
+         swz4((const __m128i *) (src_row + 0 * src_stride),
+              (const __m128i *) (src_row + 1 * src_stride),
+              (const __m128i *) (src_row + 2 * src_stride),
+              (const __m128i *) (src_row + 3 * src_stride),
+              dst128 + 2,     /* b */
+              dst128 + 1,     /* g */
+              dst128 + 0,     /* r */
+              dst128 + 3);    /* a */
+
+         dst128 += 4;
+         src_row += sizeof(__m128i);
+      }
+
+      src += 4 * src_stride;
+   }
+}
+
+static void
+lp_tile_b8g8r8x8_unorm_unswizzle_4ub_sse2(const uint8_t * restrict src,
+                                          uint8_t * restrict dst, unsigned dst_stride,
+                                          unsigned x0, unsigned y0)
+{
+   unsigned int x, y;
+   const __m128i *src128 = (const __m128i *) src;
+
+   dst += y0 * dst_stride;
+   dst += x0 * sizeof(uint32_t);
+
+   for (y = 0; y < TILE_SIZE; y += 4) {
+      const uint8_t *dst_row = dst;
+
+      for (x = 0; x < TILE_SIZE; x += 4) {
+         unswz4( &src128[2],     /* b */
+                 &src128[1],     /* g */
+                 &src128[0],     /* r */
+                 &src128[3],     /* a */
+                 (__m128i *) (dst_row + 0 * dst_stride),
+                 (__m128i *) (dst_row + 1 * dst_stride),
+                 (__m128i *) (dst_row + 2 * dst_stride),
+                 (__m128i *) (dst_row + 3 * dst_stride));
+
+         src128 += 4;
+         dst_row += sizeof(__m128i);;
+      }
+
+      dst += 4 * dst_stride;
+   }
+}
+
 #endif /* PIPE_ARCH_SSE */
 '''
 
@@ -437,7 +529,7 @@ def generate_swizzle(formats, dst_channel, dst_native_type, dst_suffix):
     print 'void'
     print 'lp_tile_swizzle_%s(enum pipe_format format, %s *dst, const void *src, unsigned src_stride, unsigned x, unsigned y)' % (dst_suffix, dst_native_type)
     print '{'
-    print '   void (*func)(%s *dst, const uint8_t *src, unsigned src_stride, unsigned x0, unsigned y0);' % dst_native_type
+    print '   void (*func)(%s * restrict dst, const uint8_t * restrict src, unsigned src_stride, unsigned x0, unsigned y0);' % dst_native_type
     print '#ifdef DEBUG'
     print '   lp_tile_swizzle_count += 1;'
     print '#endif'
@@ -446,7 +538,7 @@ def generate_swizzle(formats, dst_channel, dst_native_type, dst_suffix):
         if is_format_supported(format):
             print '   case %s:' % format.name
             func_name = 'lp_tile_%s_swizzle_%s' % (format.short_name(), dst_suffix)
-            if format.name == 'PIPE_FORMAT_B8G8R8A8_UNORM':
+            if format.name == 'PIPE_FORMAT_B8G8R8A8_UNORM' or format.name == 'PIPE_FORMAT_B8G8R8X8_UNORM':
                 print '#ifdef PIPE_ARCH_SSE'
                 print '      func = util_cpu_caps.has_sse2 ? %s_sse2 : %s;' % (func_name, func_name)
                 print '#else'
@@ -475,7 +567,7 @@ def generate_unswizzle(formats, src_channel, src_native_type, src_suffix):
     print 'lp_tile_unswizzle_%s(enum pipe_format format, const %s *src, void *dst, unsigned dst_stride, unsigned x, unsigned y)' % (src_suffix, src_native_type)
     
     print '{'
-    print '   void (*func)(const %s *src, uint8_t *dst, unsigned dst_stride, unsigned x0, unsigned y0);' % src_native_type
+    print '   void (*func)(const %s * restrict src, uint8_t * restrict dst, unsigned dst_stride, unsigned x0, unsigned y0);' % src_native_type
     print '#ifdef DEBUG'
     print '   lp_tile_unswizzle_count += 1;'
     print '#endif'
@@ -484,7 +576,7 @@ def generate_unswizzle(formats, src_channel, src_native_type, src_suffix):
         if is_format_supported(format):
             print '   case %s:' % format.name
             func_name = 'lp_tile_%s_unswizzle_%s' % (format.short_name(), src_suffix)
-            if format.name == 'PIPE_FORMAT_B8G8R8A8_UNORM':
+            if format.name == 'PIPE_FORMAT_B8G8R8A8_UNORM' or format.name == 'PIPE_FORMAT_B8G8R8X8_UNORM':
                 print '#ifdef PIPE_ARCH_SSE'
                 print '      func = util_cpu_caps.has_sse2 ? %s_sse2 : %s;' % (func_name, func_name)
                 print '#else'
@@ -510,11 +602,13 @@ def main():
     print '/* This file is autogenerated by lp_tile_soa.py from u_format.csv. Do not edit directly. */'
     print
     # This will print the copyright message on the top of this file
-    print __doc__.strip()
+    print CopyRight.strip()
     print
     print '#include "pipe/p_compiler.h"'
-    print '#include "util/u_format.h"'
     print '#include "util/u_math.h"'
+    print '#include "util/u_format.h"'
+    print '#include "util/u_format_r11g11b10f.h"'
+    print '#include "util/u_format_rgb9e5.h"'
     print '#include "util/u_half.h"'
     print '#include "util/u_cpu_detect.h"'
     print '#include "lp_tile_soa.h"'
@@ -548,7 +642,7 @@ def main():
 
     generate_sse2()
 
-    channel = Channel(UNSIGNED, True, 8)
+    channel = Channel(UNSIGNED, True, False, 8)
     native_type = 'uint8_t'
     suffix = '4ub'