[Pkg-opencl-devel] #767961: unblock pre-approval: beignet/0.9.3~really.0.8+dfsg-1
Rebecca N. Palmer
rebecca_palmer at zoho.com
Wed Nov 12 15:46:58 UTC 2014
Control: retitle -1 unblock pre-approval: beignet/0.9.3~really.0.8+dfsg-1
This version does also fix #768090 (see there for test results);
should we upload this, or drop beignet from jessie?
* Revert to 0.8 to comply with freeze policy.
* Remove non-DFSG tests. (Closes: #767387)
* Revert to LLVM/Clang 3.4, update versioned-llvm-tools.patch.
(Closes: #764930)
* Replace broken pow(n), rootn, erf(c), tgamma. (Closes: #768090)
* Document in the description what hardware this supports.
(While I started the recent debian-devel thread "Should fast-evolving
packages be backports-only?" with this package in mind, I did so
believing that backports would be on by default, which it won't:
https://lists.debian.org/debian-devel/2014/11/msg00425.html .)
> at least #745363 should be RC, too. (unconditional
> exit(1) on unsupported hardware
I don't consider that RC (it would be if we'd done the "install all
ICDs by default" proposal, but I dropped that after nobody applied
my fix for this bug; currently python(3)-pyopencl defaults to
mesa-opencl-icd and explicit "apt-get install opencl-icd" (the virtual
package) asks which provider you want, so it's hard for beignet to
get installed by accident), but if others do my original patch is
against 0.8, so can easily be included.
The following debdiff is the /debian directory; in addition, the
following non-DFSG files are deleted from the orig.tar.gz:
kernels/lenna128x128.bmp
kernels/compiler_box_blur_float_ref.bmp
kernels/compiler_box_blur_ref.bmp
kernels/compiler_chocolux.cl
kernels/compiler_chocolux_ref.bmp
kernels/compiler_clod.cl
kernels/compiler_clod_function_call.cl
kernels/compiler_clod_ref.bmp
kernels/compiler_julia.cl
kernels/compiler_julia_function_call.cl
kernels/compiler_julia_no_break.cl
kernels/compiler_julia_no_break_ref.bmp
kernels/compiler_julia_ref.bmp
kernels/compiler_menger_sponge.cl
kernels/compiler_menger_sponge_no_shadow.cl
kernels/compiler_menger_sponge_no_shadow_ref.bmp
kernels/compiler_menger_sponge_ref.bmp
kernels/compiler_nautilus.cl
kernels/compiler_nautilus_ref.bmp
kernels/compiler_ribbon.cl
kernels/compiler_ribbon_ref.bmp
diff -upNr /home/rnpalmer/Debian/sourcepkgs/beignet-0.8/debian/changelog /home/rnpalmer/Debian/sourcepkgs/beignet-0.8+dfsg/debian/changelog
--- /home/rnpalmer/Debian/sourcepkgs/beignet-0.8/debian/changelog 2014-09-12 17:11:43.000000000 +0100
+++ /home/rnpalmer/Debian/sourcepkgs/beignet-0.8+dfsg/debian/changelog 2014-11-12 12:35:07.559222226 +0000
@@ -1,3 +1,42 @@
+beignet (0.9.3~really.0.8+dfsg-1) unstable; urgency=medium
+
+ * Revert to 0.8 to comply with freeze policy.
+ * Remove non-DFSG tests. (Closes: #767387)
+ * Revert to LLVM/Clang 3.4, update versioned-llvm-tools.patch.
+ (Closes: #764930)
+ * Replace broken pow(n), rootn, erf(c), tgamma. (Closes: #768090)
+ * Document in the description what hardware this supports.
+
+ -- Rebecca N. Palmer <rebecca_palmer at zoho.com> Wed, 12 Nov 2014 12:34:41 +0000
+
+beignet (0.9.3~dfsg-1) unstable; urgency=medium
+
+ [ Julian Wollrath ]
+ * New upstream release. (Closes: #759707, #745363, #745767) (LP: #1372889)
+ + Supports llvm 3.5. (Closes: #764930) (LP: #1350773)
+ * Add B-D on libedit-dev and zlib1g-dev.
+ * Add debian/watch file.
+ * Update debian/copyright.
+
+ [ Andreas Beckmann ]
+ * Set Maintainer to "Debian OpenCL Maintainers" with Simon's permission.
+ * Add Simon Richter, Rebecca N. Palmer and myself to Uploaders.
+ * Repack upstream tarball to remove non-distributable Len(n)a images and
+ CC-BY-NC-SA licensed parts from the test suite. (Closes: #767387)
+ * 0001-fix-some-typos.patch: New.
+ * 0002-use-env-to-set-environment-variables-for-GBE_BIN_GEN.patch: New, fix
+ FTBFS of "~dfsg" versions in pbuilder etc.
+ * 0003-GBE-fix-one-compilation-warning.patch: New. Cherry-picked from
+ upstream 0.9.x branch.
+ * Skip-deleted-tests.patch: New. Thanks to Rebecca N. Palmer.
+ * Simplify using system OpenCL headers.
+ * Use-Khronos-Group-headers.patch: Removed.
+ * d/control: Fix some issues found by lintian.
+ * Bump Standards-Version to 3.9.6 (no changes needed).
+ * Import debian/ packaging history into GIT. Add Vcs-* URLs.
+
+ -- Andreas Beckmann <anbe at debian.org> Mon, 03 Nov 2014 14:23:48 +0100
+
beignet (0.8-1.1) unstable; urgency=medium
* Non-maintainer upload.
diff -upNr /home/rnpalmer/Debian/sourcepkgs/beignet-0.8/debian/control /home/rnpalmer/Debian/sourcepkgs/beignet-0.8+dfsg/debian/control
--- /home/rnpalmer/Debian/sourcepkgs/beignet-0.8/debian/control 2014-09-11 16:43:33.000000000 +0100
+++ /home/rnpalmer/Debian/sourcepkgs/beignet-0.8+dfsg/debian/control 2014-11-01 14:01:06.182243870 +0000
@@ -1,12 +1,16 @@
Source: beignet
Priority: extra
-Maintainer: Simon Richter <sjr at debian.org>
+Maintainer: Debian OpenCL Maintainers <pkg-opencl-devel at lists.alioth.debian.org>
+Uploaders:
+ Simon Richter <sjr at debian.org>,
+ Rebecca N. Palmer <rebecca_palmer at zoho.com>,
+ Andreas Beckmann <anbe at debian.org>,
Build-Depends: debhelper (>= 9), cmake, pkg-config, python-minimal,
ocl-icd-dev, ocl-icd-opencl-dev,
libdrm-dev, libxfixes-dev, libxext-dev,
- llvm-dev (>= 1:3.4),
- clang (>= 1:3.4),
- libclang-dev (>= 1:3.4),
+ llvm-3.4-dev,
+ clang-3.4,
+ libclang-3.4-dev,
libgl1-mesa-dev (>= 9) [!kfreebsd-any],
libegl1-mesa-dev (>= 9) [!kfreebsd-any],
libgbm-dev (>= 9) [!kfreebsd-any],
@@ -34,9 +38,13 @@ Depends: ${shlibs:Depends}, ${misc:Depen
Conflicts: beignet0.0.1
Replaces: beignet0.0.1
Provides: opencl-icd
-Description: Intel OpenCL library
+Description: OpenCL library for Intel Ivy Bridge GPUs
OpenCL (Open Computing Language) is a multivendor open standard for
general-purpose parallel programming of heterogeneous systems that include
CPUs, GPUs and other processors.
.
This package contains the shared library for the Intel implementation.
+ .
+ This version of the package supports only Ivy Bridge GPUs
+ (HD Graphics 2500/4000, Core ix-3xxx); versions supporting new hardware
+ will be made available in -backports.
diff -upNr /home/rnpalmer/Debian/sourcepkgs/beignet-0.8/debian/patches/Fix-pow-erf-tgamma.patch /home/rnpalmer/Debian/sourcepkgs/beignet-0.8+dfsg/debian/patches/Fix-pow-erf-tgamma.patch
--- /home/rnpalmer/Debian/sourcepkgs/beignet-0.8/debian/patches/Fix-pow-erf-tgamma.patch 1970-01-01 01:00:00.000000000 +0100
+++ /home/rnpalmer/Debian/sourcepkgs/beignet-0.8+dfsg/debian/patches/Fix-pow-erf-tgamma.patch 2014-11-12 10:47:41.579280974 +0000
@@ -0,0 +1,567 @@
+Description: Replace broken pow(n), rootn, erf(c), tgamma with working ones
+
+Don't ignore the first argument's sign in pow/pown/rootn.
+Use a proper erf/erfc implementation, instead of a power series that
+diverges above about 2.
+Make tgamma actually be tgamma instead of lgamma=log(fabs(tgamma)).
+
+Author: Rebecca Palmer <rebecca_palmer at zoho.com>, Zhigang Gong <zhigang.gong at intel.com>; contains erf(c) code from glibc http://sources.debian.net/src/glibc/2.19-12/sysdeps/ieee754/flt-32/s_erff.c/
+Bug-Debian: https://bugs.debian.org/768090
+
+diff --git a/backend/src/builtin_vector_proto.def b/backend/src/builtin_vector_proto.def
+index 18d23ca..811ab6c 100644
+--- a/backend/src/builtin_vector_proto.def
++++ b/backend/src/builtin_vector_proto.def
+@@ -94,8 +94,7 @@ floatn pown (floatn x, intn y)
+ float pown (float x, int y)
+ doublen pown (doublen x, intn y)
+ double pown (double x, int y)
+-#XXX we define powr as pow
+-#gentype powr (gentype x, gentype y)
++gentype powr (gentype x, gentype y)
+ gentype remainder (gentype x, gentype y)
+ floatn remquo (floatn x, floatn y, __global intn *quo)
+ floatn remquo (floatn x, floatn y, __local intn *quo)
+diff --git a/backend/src/ocl_stdlib.tmpl.h b/backend/src/ocl_stdlib.tmpl.h
+index f648a8c..d6900ee 100755
+--- a/backend/src/ocl_stdlib.tmpl.h
++++ b/backend/src/ocl_stdlib.tmpl.h
+@@ -1545,188 +1545,6 @@ INLINE_OVERLOADABLE float native_log2(float x) { return __gen_ocl_log(x); }
+ INLINE_OVERLOADABLE float native_log(float x) {
+ return native_log2(x) * 0.6931472002f;
+ }
+-INLINE_OVERLOADABLE float tgamma(float x) {
+-/*
+- * ====================================================
+- * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
+- *
+- * Developed at SunPro, a Sun Microsystems, Inc. business.
+- * Permission to use, copy, modify, and distribute this
+- * software is freely granted, provided that this notice
+- * is preserved.
+- * ====================================================
+- */
+- float pi = 3.1415927410e+00,
+- a0 = 7.7215664089e-02,
+- a1 = 3.2246702909e-01,
+- a2 = 6.7352302372e-02,
+- a3 = 2.0580807701e-02,
+- a4 = 7.3855509982e-03,
+- a5 = 2.8905137442e-03,
+- a6 = 1.1927076848e-03,
+- a7 = 5.1006977446e-04,
+- a8 = 2.2086278477e-04,
+- a9 = 1.0801156895e-04,
+- a10 = 2.5214456400e-05,
+- a11 = 4.4864096708e-05,
+- tc = 1.4616321325e+00,
+- tf = -1.2148628384e-01,
+- tt = 6.6971006518e-09,
+- t0 = 4.8383611441e-01,
+- t1 = -1.4758771658e-01,
+- t2 = 6.4624942839e-02,
+- t3 = -3.2788541168e-02,
+- t4 = 1.7970675603e-02,
+- t5 = -1.0314224288e-02,
+- t6 = 6.1005386524e-03,
+- t7 = -3.6845202558e-03,
+- t8 = 2.2596477065e-03,
+- t9 = -1.4034647029e-03,
+- t10 = 8.8108185446e-04,
+- t11 = -5.3859531181e-04,
+- t12 = 3.1563205994e-04,
+- t13 = -3.1275415677e-04,
+- t14 = 3.3552918467e-04,
+- u0 = -7.7215664089e-02,
+- u1 = 6.3282704353e-01,
+- u2 = 1.4549225569e+00,
+- u3 = 9.7771751881e-01,
+- u4 = 2.2896373272e-01,
+- u5 = 1.3381091878e-02,
+- v1 = 2.4559779167e+00,
+- v2 = 2.1284897327e+00,
+- v3 = 7.6928514242e-01,
+- v4 = 1.0422264785e-01,
+- v5 = 3.2170924824e-03,
+- s0 = -7.7215664089e-02,
+- s1 = 2.1498242021e-01,
+- s2 = 3.2577878237e-01,
+- s3 = 1.4635047317e-01,
+- s4 = 2.6642270386e-02,
+- s5 = 1.8402845599e-03,
+- s6 = 3.1947532989e-05,
+- r1 = 1.3920053244e+00,
+- r2 = 7.2193557024e-01,
+- r3 = 1.7193385959e-01,
+- r4 = 1.8645919859e-02,
+- r5 = 7.7794247773e-04,
+- r6 = 7.3266842264e-06,
+- w0 = 4.1893854737e-01,
+- w1 = 8.3333335817e-02,
+- w2 = -2.7777778450e-03,
+- w3 = 7.9365057172e-04,
+- w4 = -5.9518753551e-04,
+- w5 = 8.3633989561e-04,
+- w6 = -1.6309292987e-03;
+- float t, y, z, nadj, p, p1, p2, p3, q, r, w;
+- int i, hx, ix;
+- nadj = 0;
+- hx = *(int *) (&x);
+- ix = hx & 0x7fffffff;
+- if (ix >= 0x7f800000)
+- return x * x;
+- if (ix == 0)
+- return INFINITY;
+- if (ix < 0x1c800000) {
+- if (hx < 0) {
+- return - native_log(-x);
+- } else
+- return - native_log(x);
+- }
+- if (hx < 0) {
+- if (ix >= 0x4b000000)
+- return INFINITY;
+- t = __gen_ocl_internal_sinpi(x);
+- if (__gen_ocl_fabs(t) < 1e-8f)
+- return INFINITY;
+- nadj = native_log(M_PI_F / __gen_ocl_fabs(t * x));
+- x = -x;
+- }
+-
+- if (ix == 0x3f800000 || ix == 0x40000000)
+- r = 0;
+- else if (ix < 0x40000000) {
+- if (ix <= 0x3f666666) {
+- r = - native_log(x);
+- if (ix >= 0x3f3b4a20) {
+- y = 1 - x;
+- i = 0;
+- } else if (ix >= 0x3e6d3308) {
+- y = x - (tc - 1);
+- i = 1;
+- } else {
+- y = x;
+- i = 2;
+- }
+- } else {
+- r = 0;
+- if (ix >= 0x3fdda618) {
+- y = 2 - x;
+- i = 0;
+- } else if (ix >= 0x3F9da620) {
+- y = x - tc;
+- i = 1;
+- } else {
+- y = x - 1;
+- i = 2;
+- }
+- }
+- switch (i) {
+- case 0:
+- z = y * y;
+- p1 = a0 + z * (a2 + z * (a4 + z * (a6 + z * (a8 + z * a10))));
+- p2 = z * (a1 + z * (a3 + z * (a5 + z * (a7 + z * (a9 + z * a11)))));
+- p = y * p1 + p2;
+- r += (p - .5f * y);
+- break;
+- case 1:
+- z = y * y;
+- w = z * y;
+- p1 = t0 + w * (t3 + w * (t6 + w * (t9 + w * t12)));
+- p2 = t1 + w * (t4 + w * (t7 + w * (t10 + w * t13)));
+- p3 = t2 + w * (t5 + w * (t8 + w * (t11 + w * t14)));
+- p = z * p1 - (tt - w * (p2 + y * p3));
+- r += (tf + p);
+- break;
+- case 2:
+- p1 = y * (u0 + y * (u1 + y * (u2 + y * (u3 + y * (u4 + y * u5)))));
+- p2 = 1 + y * (v1 + y * (v2 + y * (v3 + y * (v4 + y * v5))));
+- r += (-.5f * y + p1 / p2);
+- }
+- } else if (ix < 0x41000000) {
+- i = x;
+- t = 0;
+- y = x - i;
+- p = y*(s0+y*(s1+y*(s2+y*(s3+y*(s4+y*(s5+y*s6))))));
+- q = 1 + y * (r1 + y * (r2 + y * (r3 + y * (r4 + y * (r5 + y * r6)))));
+- r = .5f * y + p / q;
+- z = 1;
+- switch (i) {
+- case 7:
+- z *= (y + 6.f);
+- case 6:
+- z *= (y + 5.f);
+- case 5:
+- z *= (y + 4.f);
+- case 4:
+- z *= (y + 3.f);
+- case 3:
+- z *= (y + 2.f);
+- r += native_log(z);
+- break;
+- }
+- } else if (ix < 0x5c800000) {
+- t = native_log(x);
+- z = 1 / x;
+- y = z * z;
+- w = w0 + z * (w1 + y * (w2 + y * (w3 + y * (w4 + y * (w5 + y * w6)))));
+- r = (x - .5f) * (t - 1) + w;
+- } else
+- r = x * (native_log(x) - 1);
+- if (hx < 0)
+- r = nadj - r;
+- return r;
+-}
+
+ INLINE_OVERLOADABLE float lgamma(float x) {
+ /*
+@@ -2454,12 +2454,6 @@ INLINE_OVERLOADABLE float __gen_ocl_internal_atan(float x) {
+ INLINE_OVERLOADABLE float __gen_ocl_internal_atanpi(float x) {
+ return __gen_ocl_internal_atan(x) / M_PI_F;
+ }
+-INLINE_OVERLOADABLE float __gen_ocl_internal_erf(float x) {
+- return M_2_SQRTPI_F * (x - __gen_ocl_pow(x, 3) / 3 + __gen_ocl_pow(x, 5) / 10 - __gen_ocl_pow(x, 7) / 42 + __gen_ocl_pow(x, 9) / 216);
+-}
+-INLINE_OVERLOADABLE float __gen_ocl_internal_erfc(float x) {
+- return 1 - __gen_ocl_internal_erf(x);
+-}
+
+ // XXX work-around PTX profile
+ #define sqrt native_sqrt
+@@ -2697,6 +2691,295 @@ INLINE_OVERLOADABLE float __gen_ocl_internal_exp(float x) {
+ return y*twom100;
+ }
+ }
++
++INLINE_OVERLOADABLE float tgamma(float x) {
++ float y;
++ int s;
++ y=lgamma_r(x,&s);
++ return __gen_ocl_internal_exp(y)*s;
++}
++
++/* erf,erfc from glibc s_erff.c -- float version of s_erf.c.
++ * Conversion to float by Ian Lance Taylor, Cygnus Support, ian at cygnus.com.
++ */
++
++/*
++ * ====================================================
++ * Copyright (C) 1993 by Sun Microsystems, Inc. All rights reserved.
++ *
++ * Developed at SunPro, a Sun Microsystems, Inc. business.
++ * Permission to use, copy, modify, and distribute this
++ * software is freely granted, provided that this notice
++ * is preserved.
++ * ====================================================
++ */
++
++INLINE_OVERLOADABLE float __gen_ocl_internal_erf(float x) {
++/*...*/
++const float
++tiny = 1.0e-30,
++half_val= 5.0000000000e-01, /* 0x3F000000 */
++one = 1.0000000000e+00, /* 0x3F800000 */
++two = 2.0000000000e+00, /* 0x40000000 */
++ /* c = (subfloat)0.84506291151 */
++erx = 8.4506291151e-01, /* 0x3f58560b */
++/*
++ * Coefficients for approximation to erf on [0,0.84375]
++ */
++efx = 1.2837916613e-01, /* 0x3e0375d4 */
++efx8= 1.0270333290e+00, /* 0x3f8375d4 */
++pp0 = 1.2837916613e-01, /* 0x3e0375d4 */
++pp1 = -3.2504209876e-01, /* 0xbea66beb */
++pp2 = -2.8481749818e-02, /* 0xbce9528f */
++pp3 = -5.7702702470e-03, /* 0xbbbd1489 */
++pp4 = -2.3763017452e-05, /* 0xb7c756b1 */
++qq1 = 3.9791721106e-01, /* 0x3ecbbbce */
++qq2 = 6.5022252500e-02, /* 0x3d852a63 */
++qq3 = 5.0813062117e-03, /* 0x3ba68116 */
++qq4 = 1.3249473704e-04, /* 0x390aee49 */
++qq5 = -3.9602282413e-06, /* 0xb684e21a */
++/*
++ * Coefficients for approximation to erf in [0.84375,1.25]
++ */
++pa0 = -2.3621185683e-03, /* 0xbb1acdc6 */
++pa1 = 4.1485610604e-01, /* 0x3ed46805 */
++pa2 = -3.7220788002e-01, /* 0xbebe9208 */
++pa3 = 3.1834661961e-01, /* 0x3ea2fe54 */
++pa4 = -1.1089469492e-01, /* 0xbde31cc2 */
++pa5 = 3.5478305072e-02, /* 0x3d1151b3 */
++pa6 = -2.1663755178e-03, /* 0xbb0df9c0 */
++qa1 = 1.0642088205e-01, /* 0x3dd9f331 */
++qa2 = 5.4039794207e-01, /* 0x3f0a5785 */
++qa3 = 7.1828655899e-02, /* 0x3d931ae7 */
++qa4 = 1.2617121637e-01, /* 0x3e013307 */
++qa5 = 1.3637083583e-02, /* 0x3c5f6e13 */
++qa6 = 1.1984500103e-02, /* 0x3c445aa3 */
++ /*
++ * Coefficients for approximation to erfc in [1.25,1/0.35]
++ */ra0 = -9.8649440333e-03, /* 0xbc21a093 */
++ra1 = -6.9385856390e-01, /* 0xbf31a0b7 */
++ra2 = -1.0558626175e+01, /* 0xc128f022 */
++ra3 = -6.2375331879e+01, /* 0xc2798057 */
++ra4 = -1.6239666748e+02, /* 0xc322658c */
++ra5 = -1.8460508728e+02, /* 0xc3389ae7 */
++ra6 = -8.1287437439e+01, /* 0xc2a2932b */
++ra7 = -9.8143291473e+00, /* 0xc11d077e */
++sa1 = 1.9651271820e+01, /* 0x419d35ce */
++sa2 = 1.3765776062e+02, /* 0x4309a863 */
++sa3 = 4.3456588745e+02, /* 0x43d9486f */
++sa4 = 6.4538726807e+02, /* 0x442158c9 */
++sa5 = 4.2900814819e+02, /* 0x43d6810b */
++sa6 = 1.0863500214e+02, /* 0x42d9451f */
++sa7 = 6.5702495575e+00, /* 0x40d23f7c */
++sa8 = -6.0424413532e-02, /* 0xbd777f97 */
++/*
++ * Coefficients for approximation to erfc in [1/.35,28]
++ */
++rb0 = -9.8649431020e-03, /* 0xbc21a092 */
++rb1 = -7.9928326607e-01, /* 0xbf4c9dd4 */
++rb2 = -1.7757955551e+01, /* 0xc18e104b */
++rb3 = -1.6063638306e+02, /* 0xc320a2ea */
++rb4 = -6.3756646729e+02, /* 0xc41f6441 */
++rb5 = -1.0250950928e+03, /* 0xc480230b */
++rb6 = -4.8351919556e+02, /* 0xc3f1c275 */
++sb1 = 3.0338060379e+01, /* 0x41f2b459 */
++sb2 = 3.2579251099e+02, /* 0x43a2e571 */
++sb3 = 1.5367296143e+03, /* 0x44c01759 */
++sb4 = 3.1998581543e+03, /* 0x4547fdbb */
++sb5 = 2.5530502930e+03, /* 0x451f90ce */
++sb6 = 4.7452853394e+02, /* 0x43ed43a7 */
++sb7 = -2.2440952301e+01; /* 0xc1b38712 */
++
++ int hx,ix,i;
++ float R,S,P,Q,s,y,z,r;
++ GEN_OCL_GET_FLOAT_WORD(hx,x);
++ ix = hx&0x7fffffff;
++ if(ix>=0x7f800000) { /* erf(nan)=nan */
++ i = ((unsigned int)hx>>31)<<1;
++ return (float)(1-i)+one/x; /* erf(+-inf)=+-1 */
++ }
++
++ if(ix < 0x3f580000) { /* |x|<0.84375 */
++ if(ix < 0x31800000) { /* |x|<2**-28 */
++ if (ix < 0x04000000)
++ /*avoid underflow */
++ return (float)0.125*((float)8.0*x+efx8*x);
++ return x + efx*x;
++ }
++ z = x*x;
++ r = pp0+z*(pp1+z*(pp2+z*(pp3+z*pp4)));
++ s = one+z*(qq1+z*(qq2+z*(qq3+z*(qq4+z*qq5))));
++ y = r/s;
++ return x + x*y;
++ }
++ if(ix < 0x3fa00000) { /* 0.84375 <= |x| < 1.25 */
++ s = __gen_ocl_internal_fabs(x)-one;
++ P = pa0+s*(pa1+s*(pa2+s*(pa3+s*(pa4+s*(pa5+s*pa6)))));
++ Q = one+s*(qa1+s*(qa2+s*(qa3+s*(qa4+s*(qa5+s*qa6)))));
++ if(hx>=0) return erx + P/Q; else return -erx - P/Q;
++ }
++ if (ix >= 0x40c00000) { /* inf>|x|>=6 */
++ if(hx>=0) return one-tiny; else return tiny-one;
++ }
++ x = __gen_ocl_internal_fabs(x);
++ s = one/(x*x);
++ if(ix< 0x4036DB6E) { /* |x| < 1/0.35 */
++ R=ra0+s*(ra1+s*(ra2+s*(ra3+s*(ra4+s*(
++ ra5+s*(ra6+s*ra7))))));
++ S=one+s*(sa1+s*(sa2+s*(sa3+s*(sa4+s*(
++ sa5+s*(sa6+s*(sa7+s*sa8)))))));
++ } else { /* |x| >= 1/0.35 */
++ R=rb0+s*(rb1+s*(rb2+s*(rb3+s*(rb4+s*(
++ rb5+s*rb6)))));
++ S=one+s*(sb1+s*(sb2+s*(sb3+s*(sb4+s*(
++ sb5+s*(sb6+s*sb7))))));
++ }
++ GEN_OCL_GET_FLOAT_WORD(ix,x);
++ GEN_OCL_SET_FLOAT_WORD(z,ix&0xfffff000);
++ r = __gen_ocl_internal_exp(-z*z-(float)0.5625)*__gen_ocl_internal_exp((z-x)*(z+x)+R/S);
++ if(hx>=0) return one-r/x; else return r/x-one;
++}
++INLINE_OVERLOADABLE float __gen_ocl_internal_erfc(float x) {
++/*...*/
++const float
++tiny = 1.0e-30,
++half_val= 5.0000000000e-01, /* 0x3F000000 */
++one = 1.0000000000e+00, /* 0x3F800000 */
++two = 2.0000000000e+00, /* 0x40000000 */
++ /* c = (subfloat)0.84506291151 */
++erx = 8.4506291151e-01, /* 0x3f58560b */
++/*
++ * Coefficients for approximation to erf on [0,0.84375]
++ */
++efx = 1.2837916613e-01, /* 0x3e0375d4 */
++efx8= 1.0270333290e+00, /* 0x3f8375d4 */
++pp0 = 1.2837916613e-01, /* 0x3e0375d4 */
++pp1 = -3.2504209876e-01, /* 0xbea66beb */
++pp2 = -2.8481749818e-02, /* 0xbce9528f */
++pp3 = -5.7702702470e-03, /* 0xbbbd1489 */
++pp4 = -2.3763017452e-05, /* 0xb7c756b1 */
++qq1 = 3.9791721106e-01, /* 0x3ecbbbce */
++qq2 = 6.5022252500e-02, /* 0x3d852a63 */
++qq3 = 5.0813062117e-03, /* 0x3ba68116 */
++qq4 = 1.3249473704e-04, /* 0x390aee49 */
++qq5 = -3.9602282413e-06, /* 0xb684e21a */
++/*
++ * Coefficients for approximation to erf in [0.84375,1.25]
++ */
++pa0 = -2.3621185683e-03, /* 0xbb1acdc6 */
++pa1 = 4.1485610604e-01, /* 0x3ed46805 */
++pa2 = -3.7220788002e-01, /* 0xbebe9208 */
++pa3 = 3.1834661961e-01, /* 0x3ea2fe54 */
++pa4 = -1.1089469492e-01, /* 0xbde31cc2 */
++pa5 = 3.5478305072e-02, /* 0x3d1151b3 */
++pa6 = -2.1663755178e-03, /* 0xbb0df9c0 */
++qa1 = 1.0642088205e-01, /* 0x3dd9f331 */
++qa2 = 5.4039794207e-01, /* 0x3f0a5785 */
++qa3 = 7.1828655899e-02, /* 0x3d931ae7 */
++qa4 = 1.2617121637e-01, /* 0x3e013307 */
++qa5 = 1.3637083583e-02, /* 0x3c5f6e13 */
++qa6 = 1.1984500103e-02, /* 0x3c445aa3 */
++ /*
++ * Coefficients for approximation to erfc in [1.25,1/0.35]
++ */ra0 = -9.8649440333e-03, /* 0xbc21a093 */
++ra1 = -6.9385856390e-01, /* 0xbf31a0b7 */
++ra2 = -1.0558626175e+01, /* 0xc128f022 */
++ra3 = -6.2375331879e+01, /* 0xc2798057 */
++ra4 = -1.6239666748e+02, /* 0xc322658c */
++ra5 = -1.8460508728e+02, /* 0xc3389ae7 */
++ra6 = -8.1287437439e+01, /* 0xc2a2932b */
++ra7 = -9.8143291473e+00, /* 0xc11d077e */
++sa1 = 1.9651271820e+01, /* 0x419d35ce */
++sa2 = 1.3765776062e+02, /* 0x4309a863 */
++sa3 = 4.3456588745e+02, /* 0x43d9486f */
++sa4 = 6.4538726807e+02, /* 0x442158c9 */
++sa5 = 4.2900814819e+02, /* 0x43d6810b */
++sa6 = 1.0863500214e+02, /* 0x42d9451f */
++sa7 = 6.5702495575e+00, /* 0x40d23f7c */
++sa8 = -6.0424413532e-02, /* 0xbd777f97 */
++/*
++ * Coefficients for approximation to erfc in [1/.35,28]
++ */
++rb0 = -9.8649431020e-03, /* 0xbc21a092 */
++rb1 = -7.9928326607e-01, /* 0xbf4c9dd4 */
++rb2 = -1.7757955551e+01, /* 0xc18e104b */
++rb3 = -1.6063638306e+02, /* 0xc320a2ea */
++rb4 = -6.3756646729e+02, /* 0xc41f6441 */
++rb5 = -1.0250950928e+03, /* 0xc480230b */
++rb6 = -4.8351919556e+02, /* 0xc3f1c275 */
++sb1 = 3.0338060379e+01, /* 0x41f2b459 */
++sb2 = 3.2579251099e+02, /* 0x43a2e571 */
++sb3 = 1.5367296143e+03, /* 0x44c01759 */
++sb4 = 3.1998581543e+03, /* 0x4547fdbb */
++sb5 = 2.5530502930e+03, /* 0x451f90ce */
++sb6 = 4.7452853394e+02, /* 0x43ed43a7 */
++sb7 = -2.2440952301e+01; /* 0xc1b38712 */
++ int hx,ix;
++ float R,S,P,Q,s,y,z,r;
++ GEN_OCL_GET_FLOAT_WORD(hx,x);
++ ix = hx&0x7fffffff;
++ if(ix>=0x7f800000) { /* erfc(nan)=nan */
++ /* erfc(+-inf)=0,2 */
++ return (float)(((unsigned int)hx>>31)<<1)+one/x;
++ }
++
++ if(ix < 0x3f580000) { /* |x|<0.84375 */
++ if(ix < 0x23800000) /* |x|<2**-56 */
++ return one-x;
++ z = x*x;
++ r = pp0+z*(pp1+z*(pp2+z*(pp3+z*pp4)));
++ s = one+z*(qq1+z*(qq2+z*(qq3+z*(qq4+z*qq5))));
++ y = r/s;
++ if(hx < 0x3e800000) { /* x<1/4 */
++ return one-(x+x*y);
++ } else {
++ r = x*y;
++ r += (x-half_val);
++ return half_val - r ;
++ }
++ }
++ if(ix < 0x3fa00000) { /* 0.84375 <= |x| < 1.25 */
++ s = __gen_ocl_internal_fabs(x)-one;
++ P = pa0+s*(pa1+s*(pa2+s*(pa3+s*(pa4+s*(pa5+s*pa6)))));
++ Q = one+s*(qa1+s*(qa2+s*(qa3+s*(qa4+s*(qa5+s*qa6)))));
++ if(hx>=0) {
++ z = one-erx; return z - P/Q;
++ } else {
++ z = erx+P/Q; return one+z;
++ }
++ }
++ if (ix < 0x41e00000) { /* |x|<28 */
++ x = __gen_ocl_internal_fabs(x);
++ s = one/(x*x);
++ if(ix< 0x4036DB6D) { /* |x| < 1/.35 ~ 2.857143*/
++ R=ra0+s*(ra1+s*(ra2+s*(ra3+s*(ra4+s*(
++ ra5+s*(ra6+s*ra7))))));
++ S=one+s*(sa1+s*(sa2+s*(sa3+s*(sa4+s*(
++ sa5+s*(sa6+s*(sa7+s*sa8)))))));
++ } else { /* |x| >= 1/.35 ~ 2.857143 */
++ if(hx<0&&ix>=0x40c00000) return two-tiny;/* x < -6 */
++ R=rb0+s*(rb1+s*(rb2+s*(rb3+s*(rb4+s*(
++ rb5+s*rb6)))));
++ S=one+s*(sb1+s*(sb2+s*(sb3+s*(sb4+s*(
++ sb5+s*(sb6+s*sb7))))));
++ }
++ GEN_OCL_GET_FLOAT_WORD(ix,x);
++ GEN_OCL_SET_FLOAT_WORD(z,ix&0xffffe000);
++ r = __gen_ocl_internal_exp(-z*z-(float)0.5625)*
++ __gen_ocl_internal_exp((z-x)*(z+x)+R/S);
++ if(hx>0) {
++ float ret = r/x;
++ return ret;
++ } else
++ return two-r/x;
++ } else {
++ if(hx>0) {
++ return tiny*tiny;
++ } else
++ return two-tiny;
++ }
++}
++
+ INLINE_OVERLOADABLE float __gen_ocl_internal_fmod (float x, float y) {
+ //return x-y*__gen_ocl_rndz(x/y);
+ float one = 1.0;
+@@ -3161,7 +3161,6 @@ INLINE_OVERLOADABLE float __gen_ocl_internal_exp10(float x){
+ #define atan2pi __gen_ocl_internal_atan2pi
+ #define atanpi __gen_ocl_internal_atanpi
+ #define atanh __gen_ocl_internal_atanh
+-#define pow powr
+ #define cbrt __gen_ocl_internal_cbrt
+ #define rint __gen_ocl_internal_rint
+ #define copysign __gen_ocl_internal_copysign
+@@ -3476,11 +3475,30 @@ INLINE_OVERLOADABLE float remquo(float x, float y, private int *quo) { BODY; }
+ #undef BODY
+ INLINE_OVERLOADABLE float native_divide(float x, float y) { return x/y; }
+ INLINE_OVERLOADABLE float pown(float x, int n) {
+- if (x == 0 && n == 0)
+- return 1;
++ if (x == 0.f && n == 0)
++ return 1.f;
++ if (x < 0.f && (n&1) )
++ return -powr(-x, n);
+ return powr(x, n);
+ }
++INLINE_OVERLOADABLE float pow(float x, float y) {
++ int n;
++ if (x == 0.f && y == 0.f)
++ return 1.f;
++ if (x >= 0.f)
++ return powr(x, y);
++ n = y;
++ if ((float)n == y)//is exact integer
++ return pown(x, n);
++ return NAN;
++}
++
+ INLINE_OVERLOADABLE float rootn(float x, int n) {
++ if (x < 0.f) {
++ if ( n&1 )
++ return -powr(-x, 1.f / n);
++ return NAN;
++ }
+ return powr(x, 1.f / n);
+ }
+
diff -upNr /home/rnpalmer/Debian/sourcepkgs/beignet-0.8/debian/patches/series /home/rnpalmer/Debian/sourcepkgs/beignet-0.8+dfsg/debian/patches/series
--- /home/rnpalmer/Debian/sourcepkgs/beignet-0.8/debian/patches/series 2014-09-12 17:03:28.000000000 +0100
+++ /home/rnpalmer/Debian/sourcepkgs/beignet-0.8+dfsg/debian/patches/series 2014-11-12 12:33:06.139223333 +0000
@@ -5,3 +5,4 @@ deprecated-in-utest
versioned-llvm-tools
terminfo
fix_license_issue
+Fix-pow-erf-tgamma.patch
diff -upNr /home/rnpalmer/Debian/sourcepkgs/beignet-0.8/debian/patches/versioned-llvm-tools /home/rnpalmer/Debian/sourcepkgs/beignet-0.8+dfsg/debian/patches/versioned-llvm-tools
--- /home/rnpalmer/Debian/sourcepkgs/beignet-0.8/debian/patches/versioned-llvm-tools 2014-04-19 18:54:55.000000000 +0100
+++ /home/rnpalmer/Debian/sourcepkgs/beignet-0.8+dfsg/debian/patches/versioned-llvm-tools 2014-11-01 13:28:17.650301427 +0000
@@ -1,9 +1,20 @@
Description: Use versioned LLVM tools
-Author: Simon Richter <sjr at debian.org>
-Last-Update: 2014-04-19
+Author: Simon Richter <sjr at debian.org>, Rebecca N. Palmer <rebecca_palmer at zoho.com>
+Bug-Debian: https://bugs.debian.org/759933,https://bugs.debian.org/764930
--- beignet-0.8.orig/backend/src/CMakeLists.txt
+++ beignet-0.8/backend/src/CMakeLists.txt
+@@ -58,8 +58,8 @@ set (clang_cmd ${clang_cmd} -fno-builtin
+ add_custom_command(
+ OUTPUT ${pch_object}
+ COMMAND rm -f ${pch_object}
+- COMMAND clang ${clang_cmd} --relocatable-pch -emit-pch -isysroot ${CMAKE_CURRENT_BINARY_DIR} ${ocl_blob_file} -o ${pch_object}
+- COMMAND clang ${clang_cmd} -emit-pch ${ocl_blob_file} -o ${local_pch_object}
++ COMMAND clang-3.4 ${clang_cmd} --relocatable-pch -emit-pch -isysroot ${CMAKE_CURRENT_BINARY_DIR} ${ocl_blob_file} -o ${pch_object}
++ COMMAND clang-3.4 ${clang_cmd} -emit-pch ${ocl_blob_file} -o ${local_pch_object}
+ DEPENDS ${ocl_blob_file}
+ )
+
@@ -71,14 +71,14 @@ macro(ll_add_library ll_lib ll_sources)
add_custom_command(
OUTPUT ${ll}.bc
More information about the Pkg-opencl-devel
mailing list