[140984] trunk/dports/science/volk

michaelld at macports.org michaelld at macports.org
Wed Oct 7 06:40:31 PDT 2015


Revision: 140984
          https://trac.macports.org/changeset/140984
Author:   michaelld at macports.org
Date:     2015-10-07 06:40:30 -0700 (Wed, 07 Oct 2015)
Log Message:
-----------
volk-devel: update to 015b1e27 (20151006), adding in a temporary patch to fix a general build issue.

Modified Paths:
--------------
    trunk/dports/science/volk/Portfile

Added Paths:
-----------
    trunk/dports/science/volk/files/patch-kernels_volk_volk_32fc_s32fc_x2_rotator_32fc.h.diff

Removed Paths:
-------------
    trunk/dports/science/volk/files/patch-_mm_bsrli_si128.diff

Modified: trunk/dports/science/volk/Portfile
===================================================================
--- trunk/dports/science/volk/Portfile	2015-10-07 13:40:22 UTC (rev 140983)
+++ trunk/dports/science/volk/Portfile	2015-10-07 13:40:30 UTC (rev 140984)
@@ -36,20 +36,18 @@
 
 subport volk-devel {
 
-    github.setup    gnuradio volk a5f24520138d9e1b802722c42372526fad3a460a
-    version         20150915
+    github.setup    gnuradio volk 015b1e271ac32e100e50670fa1be6d2d4b1294c1
+    version         20151006
     checksums       \
-        rmd160 29a94831d5b7ed0805341853ab6baff06f0f9a6a \
-        sha256 c96b40e7f90a00994101313c68e87547bd2bfd8cef560a8225632c6748b0db96
+        rmd160 4f2253819bbe2cf8c059c6f4ea40a12586979c30 \
+        sha256 f4fd95c3f919130d71020d5e3e8e48af678a10b53f671789ddb4b3748bacab38
     conflicts       volk
     long_description ${long_description} ${subport} \
 provides the Volk GIT maint or master branch, depending on which is more advanced, and is typically updated every few days to week.
 
-    # temporary patch to do "s/_mm_bsrli_si128/_mm_srli_si128/g" in
-    # kernels/volk/volk_8u_x2_encodeframepolar_8u.h
+    # temporary patch to fix avx rotator build issue
+    patchfiles-append patch-kernels_volk_volk_32fc_s32fc_x2_rotator_32fc.h.diff
 
-    patchfiles-append patch-_mm_bsrli_si128.diff
-
 }
 
 # override githib PortGroup homepage setting

Deleted: trunk/dports/science/volk/files/patch-_mm_bsrli_si128.diff
===================================================================
--- trunk/dports/science/volk/files/patch-_mm_bsrli_si128.diff	2015-10-07 13:40:22 UTC (rev 140983)
+++ trunk/dports/science/volk/files/patch-_mm_bsrli_si128.diff	2015-10-07 13:40:30 UTC (rev 140984)
@@ -1,88 +0,0 @@
---- kernels/volk/volk_8u_x2_encodeframepolar_8u.h.orig
-+++ kernels/volk/volk_8u_x2_encodeframepolar_8u.h
-@@ -121,12 +121,12 @@ volk_8u_x2_encodeframepolar_8u_u_ssse3(unsigned char* frame, unsigned char* temp
-           r_temp1 = _mm_loadu_si128((__m128i *) temp_ptr);
-           temp_ptr += 16;
- 
--          shifted = _mm_bsrli_si128(r_temp0, 1);
-+          shifted = _mm_srli_si128(r_temp0, 1);
-           shifted = _mm_and_si128(shifted, mask_stage1);
-           r_temp0 = _mm_xor_si128(shifted, r_temp0);
-           r_temp0 = _mm_shuffle_epi8(r_temp0, shuffle_separate);
- 
--          shifted = _mm_bsrli_si128(r_temp1, 1);
-+          shifted = _mm_srli_si128(r_temp1, 1);
-           shifted = _mm_and_si128(shifted, mask_stage1);
-           r_temp1 = _mm_xor_si128(shifted, r_temp1);
-           r_temp1 = _mm_shuffle_epi8(r_temp1, shuffle_separate);
-@@ -170,7 +170,7 @@ volk_8u_x2_encodeframepolar_8u_u_ssse3(unsigned char* frame, unsigned char* temp
-     // shuffle once for bit-reversal.
-     r_temp0 = _mm_shuffle_epi8(r_temp0, shuffle_stage4);
- 
--    shifted = _mm_bsrli_si128(r_temp0, 8);
-+    shifted = _mm_srli_si128(r_temp0, 8);
-     shifted = _mm_and_si128(shifted, mask_stage4);
-     r_frame0 = _mm_xor_si128(shifted, r_temp0);
- 
-@@ -178,15 +178,15 @@ volk_8u_x2_encodeframepolar_8u_u_ssse3(unsigned char* frame, unsigned char* temp
-     r_temp0 = _mm_loadu_si128((__m128i*) temp_ptr);
-     temp_ptr += 16;
- 
--    shifted = _mm_bsrli_si128(r_frame0, 4);
-+    shifted = _mm_srli_si128(r_frame0, 4);
-     shifted = _mm_and_si128(shifted, mask_stage3);
-     r_frame0 = _mm_xor_si128(shifted, r_frame0);
- 
--    shifted = _mm_bsrli_si128(r_frame0, 2);
-+    shifted = _mm_srli_si128(r_frame0, 2);
-     shifted = _mm_and_si128(shifted, mask_stage2);
-     r_frame0 = _mm_xor_si128(shifted, r_frame0);
- 
--    shifted = _mm_bsrli_si128(r_frame0, 1);
-+    shifted = _mm_srli_si128(r_frame0, 1);
-     shifted = _mm_and_si128(shifted, mask_stage1);
-     r_frame0 = _mm_xor_si128(shifted, r_frame0);
- 
-@@ -243,12 +243,12 @@ volk_8u_x2_encodeframepolar_8u_a_ssse3(unsigned char* frame, unsigned char* temp
-           r_temp1 = _mm_load_si128((__m128i *) temp_ptr);
-           temp_ptr += 16;
- 
--          shifted = _mm_bsrli_si128(r_temp0, 1);
-+          shifted = _mm_srli_si128(r_temp0, 1);
-           shifted = _mm_and_si128(shifted, mask_stage1);
-           r_temp0 = _mm_xor_si128(shifted, r_temp0);
-           r_temp0 = _mm_shuffle_epi8(r_temp0, shuffle_separate);
- 
--          shifted = _mm_bsrli_si128(r_temp1, 1);
-+          shifted = _mm_srli_si128(r_temp1, 1);
-           shifted = _mm_and_si128(shifted, mask_stage1);
-           r_temp1 = _mm_xor_si128(shifted, r_temp1);
-           r_temp1 = _mm_shuffle_epi8(r_temp1, shuffle_separate);
-@@ -292,7 +292,7 @@ volk_8u_x2_encodeframepolar_8u_a_ssse3(unsigned char* frame, unsigned char* temp
-     // shuffle once for bit-reversal.
-     r_temp0 = _mm_shuffle_epi8(r_temp0, shuffle_stage4);
- 
--    shifted = _mm_bsrli_si128(r_temp0, 8);
-+    shifted = _mm_srli_si128(r_temp0, 8);
-     shifted = _mm_and_si128(shifted, mask_stage4);
-     r_frame0 = _mm_xor_si128(shifted, r_temp0);
- 
-@@ -300,15 +300,15 @@ volk_8u_x2_encodeframepolar_8u_a_ssse3(unsigned char* frame, unsigned char* temp
-     r_temp0 = _mm_load_si128((__m128i*) temp_ptr);
-     temp_ptr += 16;
- 
--    shifted = _mm_bsrli_si128(r_frame0, 4);
-+    shifted = _mm_srli_si128(r_frame0, 4);
-     shifted = _mm_and_si128(shifted, mask_stage3);
-     r_frame0 = _mm_xor_si128(shifted, r_frame0);
- 
--    shifted = _mm_bsrli_si128(r_frame0, 2);
-+    shifted = _mm_srli_si128(r_frame0, 2);
-     shifted = _mm_and_si128(shifted, mask_stage2);
-     r_frame0 = _mm_xor_si128(shifted, r_frame0);
- 
--    shifted = _mm_bsrli_si128(r_frame0, 1);
-+    shifted = _mm_srli_si128(r_frame0, 1);
-     shifted = _mm_and_si128(shifted, mask_stage1);
-     r_frame0 = _mm_xor_si128(shifted, r_frame0);
- 

Added: trunk/dports/science/volk/files/patch-kernels_volk_volk_32fc_s32fc_x2_rotator_32fc.h.diff
===================================================================
--- trunk/dports/science/volk/files/patch-kernels_volk_volk_32fc_s32fc_x2_rotator_32fc.h.diff	                        (rev 0)
+++ trunk/dports/science/volk/files/patch-kernels_volk_volk_32fc_s32fc_x2_rotator_32fc.h.diff	2015-10-07 13:40:30 UTC (rev 140984)
@@ -0,0 +1,36 @@
+--- kernels/volk/volk_32fc_s32fc_x2_rotator_32fc.h.orig
++++ kernels/volk/volk_32fc_s32fc_x2_rotator_32fc.h
+@@ -411,11 +411,11 @@ static inline void volk_32fc_s32fc_x2_rotator_32fc_a_avx(lv_32fc_t* outVector, c
+         cPtr += 4;
+     }
+     if (i) {
+-        tmp1 = _mm_mul_ps(phase_Val, phase_Val);
+-        tmp2 = _mm_hadd_ps(tmp1, tmp1);
+-        tmp1 = _mm_shuffle_ps(tmp2, tmp2, 0xD8);
+-        tmp2 = _mm_sqrt_ps(tmp1);
+-        phase_Val = _mm_div_ps(phase_Val, tmp2);
++        tmp1 = _mm256_mul_ps(phase_Val, phase_Val);
++        tmp2 = _mm256_hadd_ps(tmp1, tmp1);
++        tmp1 = _mm256_shuffle_ps(tmp2, tmp2, 0xD8);
++        tmp2 = _mm256_sqrt_ps(tmp1);
++        phase_Val = _mm256_div_ps(phase_Val, tmp2);
+     }
+ 
+     _mm256_storeu_ps((float*)phase_Ptr, phase_Val);
+@@ -517,11 +517,11 @@ static inline void volk_32fc_s32fc_x2_rotator_32fc_u_avx(lv_32fc_t* outVector, c
+         cPtr += 4;
+     }
+     if (i) {
+-        tmp1 = _mm_mul_ps(phase_Val, phase_Val);
+-        tmp2 = _mm_hadd_ps(tmp1, tmp1);
+-        tmp1 = _mm_shuffle_ps(tmp2, tmp2, 0xD8);
+-        tmp2 = _mm_sqrt_ps(tmp1);
+-        phase_Val = _mm_div_ps(phase_Val, tmp2);
++        tmp1 = _mm256_mul_ps(phase_Val, phase_Val);
++        tmp2 = _mm256_hadd_ps(tmp1, tmp1);
++        tmp1 = _mm256_shuffle_ps(tmp2, tmp2, 0xD8);
++        tmp2 = _mm256_sqrt_ps(tmp1);
++        phase_Val = _mm256_div_ps(phase_Val, tmp2);
+     }
+ 
+     _mm256_storeu_ps((float*)phase_Ptr, phase_Val);
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://lists.macosforge.org/pipermail/macports-changes/attachments/20151007/a187f654/attachment-0001.html>


More information about the macports-changes mailing list