The "rm" basm constraint doesn't work with my version of gcc (4.5.2),
not even in a simple example. Since we usually only have 5 registers
available on i386, force it to be memory on that architecture.

I've reported an Ubuntu bug on the "rm" constraint issue [1], and it seems to apply to a lot of gcc versions. In the meantime I recommend this patch to be applied to both master and stable-queue. I've got some positive feedback from testing in the Ubuntu community as well.

@Colin, I haven't forgot that there might still be issues if you completely silence one channel, it just hasn't been time/priority to deal with that yet, feel free to beat me to it if you like :-)

--
David Henningsson, Canonical Ltd.
http://launchpad.net/~diwic

[1] https://bugs.launchpad.net/ubuntu/+source/gcc-4.5/+bug/789031
>From 75d0e5f5ad7764eba560810c958c5dfa4a248f81 Mon Sep 17 00:00:00 2001
From: David Henningsson <[email protected]>
Date: Fri, 27 May 2011 09:28:57 +0200
Subject: [PATCH] SSE/MMX: Fix problem with highpitched noise on i386

The "rm" basm constraint doesn't work with my version of gcc (4.5.2),
not even in a simple example. Since we usually only have 5 registers
available on i386, force it to be memory on that architecture.

Signed-off-by: David Henningsson <[email protected]>
---
 src/pulsecore/svolume_mmx.c |   12 ++++++++++--
 src/pulsecore/svolume_sse.c |   12 ++++++++++--
 2 files changed, 20 insertions(+), 4 deletions(-)

diff --git a/src/pulsecore/svolume_mmx.c b/src/pulsecore/svolume_mmx.c
index a71a39b..129d7f0 100644
--- a/src/pulsecore/svolume_mmx.c
+++ b/src/pulsecore/svolume_mmx.c
@@ -155,7 +155,11 @@ pa_volume_s16ne_mmx (int16_t *samples, int32_t *volumes, unsigned channels, unsi
         " emms                          \n\t"
 
         : "+r" (samples), "+r" (volumes), "+r" (length), "=D" ((pa_reg_x86)channel), "=&r" (temp)
-        : "rm" ((pa_reg_x86)channels)
+#if defined (__i386__)
+        : "m" ((pa_reg_x86)channels)
+#else
+        : "r" ((pa_reg_x86)channels)
+#endif
         : "cc"
     );
 }
@@ -232,7 +236,11 @@ pa_volume_s16re_mmx (int16_t *samples, int32_t *volumes, unsigned channels, unsi
         " emms                          \n\t"
 
         : "+r" (samples), "+r" (volumes), "+r" (length), "=D" ((pa_reg_x86)channel), "=&r" (temp)
-        : "rm" ((pa_reg_x86)channels)
+#if defined (__i386__)
+        : "m" ((pa_reg_x86)channels)
+#else
+        : "r" ((pa_reg_x86)channels)
+#endif
         : "cc"
     );
 }
diff --git a/src/pulsecore/svolume_sse.c b/src/pulsecore/svolume_sse.c
index 5983164..ccf45e7 100644
--- a/src/pulsecore/svolume_sse.c
+++ b/src/pulsecore/svolume_sse.c
@@ -154,7 +154,11 @@ pa_volume_s16ne_sse2 (int16_t *samples, int32_t *volumes, unsigned channels, uns
         "8:                             \n\t"
 
         : "+r" (samples), "+r" (volumes), "+r" (length), "=D" (channel), "=&r" (temp)
-        : "rm" ((pa_reg_x86)channels)
+#if defined (__i386__)
+        : "m" ((pa_reg_x86)channels)
+#else
+        : "r" ((pa_reg_x86)channels)
+#endif
         : "cc"
     );
 }
@@ -244,7 +248,11 @@ pa_volume_s16re_sse2 (int16_t *samples, int32_t *volumes, unsigned channels, uns
         "8:                             \n\t"
 
         : "+r" (samples), "+r" (volumes), "+r" (length), "=D" (channel), "=&r" (temp)
-        : "rm" ((pa_reg_x86)channels)
+#if defined (__i386__)
+        : "m" ((pa_reg_x86)channels)
+#else
+        : "r" ((pa_reg_x86)channels)
+#endif
         : "cc"
     );
 }
-- 
1.7.4.1

_______________________________________________
pulseaudio-discuss mailing list
[email protected]
http://lists.freedesktop.org/mailman/listinfo/pulseaudio-discuss

Reply via email to