From 7c2dadab17aea6e790ff9ff48fee5670cd15897e Mon Sep 17 00:00:00 2001 From: jsteube Date: Thu, 21 Sep 2017 20:08:24 +0200 Subject: [PATCH] no volatile required here --- OpenCL/inc_types.cl | 202 ++++++++++++++++++++++---------------------- 1 file changed, 101 insertions(+), 101 deletions(-) diff --git a/OpenCL/inc_types.cl b/OpenCL/inc_types.cl index 780fdcd5e..0c567cf1b 100644 --- a/OpenCL/inc_types.cl +++ b/OpenCL/inc_types.cl @@ -180,7 +180,7 @@ static u32 swap32_S (const u32 v) { u32 r; - __asm__ volatile ("V_PERM_B32 %0, 0, %1, %2;" : "=v"(r) : "v"(v), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, 0, %1, %2;" : "=v"(r) : "v"(v), "v"(0x00010203)); return r; } @@ -193,8 +193,8 @@ static u64 swap64_S (const u64 v) u32 t0; u32 t1; - __asm__ volatile ("V_PERM_B32 %0, 0, %1, %2;" : "=v"(t0) : "v"(v0), "v"(0x00010203)); - __asm__ volatile ("V_PERM_B32 %0, 0, %1, %2;" : "=v"(t1) : "v"(v1), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, 0, %1, %2;" : "=v"(t0) : "v"(v0), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, 0, %1, %2;" : "=v"(t1) : "v"(v1), "v"(0x00010203)); const u64 r = hl32_to_64_S (t1, t0); @@ -255,52 +255,52 @@ static u64x swap64 (const u64x v) u32x t1; #if VECT_SIZE == 1 - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0) : "v"(0), "v"(a0), "v"(0x00010203)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1) : "v"(0), "v"(a1), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0) : "v"(0), "v"(a0), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1) : "v"(0), "v"(a1), "v"(0x00010203)); #endif #if VECT_SIZE >= 2 - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.s0) : "v"(0), "v"(a0.s0), "v"(0x00010203)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.s0) : "v"(0), "v"(a1.s0), "v"(0x00010203)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.s1) : "v"(0), "v"(a0.s1), "v"(0x00010203)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.s1) : "v"(0), "v"(a1.s1), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.s0) : "v"(0), "v"(a0.s0), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.s0) : "v"(0), "v"(a1.s0), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.s1) : "v"(0), "v"(a0.s1), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.s1) : "v"(0), "v"(a1.s1), "v"(0x00010203)); #endif #if VECT_SIZE >= 4 - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.s2) : "v"(0), "v"(a0.s2), "v"(0x00010203)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.s2) : "v"(0), "v"(a1.s2), "v"(0x00010203)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.s3) : "v"(0), "v"(a0.s3), "v"(0x00010203)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.s3) : "v"(0), "v"(a1.s3), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.s2) : "v"(0), "v"(a0.s2), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.s2) : "v"(0), "v"(a1.s2), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.s3) : "v"(0), "v"(a0.s3), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.s3) : "v"(0), "v"(a1.s3), "v"(0x00010203)); #endif #if VECT_SIZE >= 8 - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.s4) : "v"(0), "v"(a0.s4), "v"(0x00010203)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.s4) : "v"(0), "v"(a1.s4), "v"(0x00010203)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.s5) : "v"(0), "v"(a0.s5), "v"(0x00010203)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.s5) : "v"(0), "v"(a1.s5), "v"(0x00010203)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.s6) : "v"(0), "v"(a0.s6), "v"(0x00010203)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.s6) : "v"(0), "v"(a1.s6), "v"(0x00010203)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.s7) : "v"(0), "v"(a0.s7), "v"(0x00010203)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.s7) : "v"(0), "v"(a1.s7), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.s4) : "v"(0), "v"(a0.s4), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.s4) : "v"(0), "v"(a1.s4), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.s5) : "v"(0), "v"(a0.s5), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.s5) : "v"(0), "v"(a1.s5), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.s6) : "v"(0), "v"(a0.s6), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.s6) : "v"(0), "v"(a1.s6), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.s7) : "v"(0), "v"(a0.s7), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.s7) : "v"(0), "v"(a1.s7), "v"(0x00010203)); #endif #if VECT_SIZE >= 16 - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.s8) : "v"(0), "v"(a0.s8), "v"(0x00010203)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.s8) : "v"(0), "v"(a1.s8), "v"(0x00010203)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.s9) : "v"(0), "v"(a0.s9), "v"(0x00010203)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.s9) : "v"(0), "v"(a1.s9), "v"(0x00010203)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.sa) : "v"(0), "v"(a0.sa), "v"(0x00010203)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.sa) : "v"(0), "v"(a1.sa), "v"(0x00010203)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.sb) : "v"(0), "v"(a0.sb), "v"(0x00010203)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.sb) : "v"(0), "v"(a1.sb), "v"(0x00010203)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.sc) : "v"(0), "v"(a0.sc), "v"(0x00010203)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.sc) : "v"(0), "v"(a1.sc), "v"(0x00010203)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.sd) : "v"(0), "v"(a0.sd), "v"(0x00010203)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.sd) : "v"(0), "v"(a1.sd), "v"(0x00010203)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.se) : "v"(0), "v"(a0.se), "v"(0x00010203)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.se) : "v"(0), "v"(a1.se), "v"(0x00010203)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.sf) : "v"(0), "v"(a0.sf), "v"(0x00010203)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.sf) : "v"(0), "v"(a1.sf), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.s8) : "v"(0), "v"(a0.s8), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.s8) : "v"(0), "v"(a1.s8), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.s9) : "v"(0), "v"(a0.s9), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.s9) : "v"(0), "v"(a1.s9), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.sa) : "v"(0), "v"(a0.sa), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.sa) : "v"(0), "v"(a1.sa), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.sb) : "v"(0), "v"(a0.sb), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.sb) : "v"(0), "v"(a1.sb), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.sc) : "v"(0), "v"(a0.sc), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.sc) : "v"(0), "v"(a1.sc), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.sd) : "v"(0), "v"(a0.sd), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.sd) : "v"(0), "v"(a1.sd), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.se) : "v"(0), "v"(a0.se), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.se) : "v"(0), "v"(a1.se), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t0.sf) : "v"(0), "v"(a0.sf), "v"(0x00010203)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(t1.sf) : "v"(0), "v"(a1.sf), "v"(0x00010203)); #endif const u64x r = hl32_to_64 (t1, t0); @@ -372,49 +372,49 @@ static u32x __byte_perm (const u32x a, const u32x b, const u32x c) u32x r; #if VECT_SIZE == 1 - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r) : "v"(b), "v"(a), "v"(c)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r) : "v"(b), "v"(a), "v"(c)); #endif #if VECT_SIZE >= 2 - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s0) : "v"(b.s0), "v"(a.s0), "v"(c.s0)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s1) : "v"(b.s1), "v"(a.s1), "v"(c.s1)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s0) : "v"(b.s0), "v"(a.s0), "v"(c.s0)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s1) : "v"(b.s1), "v"(a.s1), "v"(c.s1)); #endif #if VECT_SIZE >= 4 - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s0) : "v"(b.s0), "v"(a.s0), "v"(c.s0)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s1) : "v"(b.s1), "v"(a.s1), "v"(c.s1)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s2) : "v"(b.s2), "v"(a.s2), "v"(c.s2)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s3) : "v"(b.s3), "v"(a.s3), "v"(c.s3)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s0) : "v"(b.s0), "v"(a.s0), "v"(c.s0)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s1) : "v"(b.s1), "v"(a.s1), "v"(c.s1)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s2) : "v"(b.s2), "v"(a.s2), "v"(c.s2)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s3) : "v"(b.s3), "v"(a.s3), "v"(c.s3)); #endif #if VECT_SIZE >= 8 - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s0) : "v"(b.s0), "v"(a.s0), "v"(c.s0)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s1) : "v"(b.s1), "v"(a.s1), "v"(c.s1)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s2) : "v"(b.s2), "v"(a.s2), "v"(c.s2)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s3) : "v"(b.s3), "v"(a.s3), "v"(c.s3)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s4) : "v"(b.s4), "v"(a.s4), "v"(c.s4)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s5) : "v"(b.s5), "v"(a.s5), "v"(c.s5)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s6) : "v"(b.s6), "v"(a.s6), "v"(c.s6)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s7) : "v"(b.s7), "v"(a.s7), "v"(c.s7)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s0) : "v"(b.s0), "v"(a.s0), "v"(c.s0)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s1) : "v"(b.s1), "v"(a.s1), "v"(c.s1)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s2) : "v"(b.s2), "v"(a.s2), "v"(c.s2)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s3) : "v"(b.s3), "v"(a.s3), "v"(c.s3)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s4) : "v"(b.s4), "v"(a.s4), "v"(c.s4)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s5) : "v"(b.s5), "v"(a.s5), "v"(c.s5)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s6) : "v"(b.s6), "v"(a.s6), "v"(c.s6)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s7) : "v"(b.s7), "v"(a.s7), "v"(c.s7)); #endif #if VECT_SIZE >= 16 - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s0) : "v"(b.s0), "v"(a.s0), "v"(c.s0)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s1) : "v"(b.s1), "v"(a.s1), "v"(c.s1)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s2) : "v"(b.s2), "v"(a.s2), "v"(c.s2)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s3) : "v"(b.s3), "v"(a.s3), "v"(c.s3)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s4) : "v"(b.s4), "v"(a.s4), "v"(c.s4)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s5) : "v"(b.s5), "v"(a.s5), "v"(c.s5)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s6) : "v"(b.s6), "v"(a.s6), "v"(c.s6)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s7) : "v"(b.s7), "v"(a.s7), "v"(c.s7)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s8) : "v"(b.s8), "v"(a.s8), "v"(c.s8)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s9) : "v"(b.s9), "v"(a.s9), "v"(c.s9)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.sa) : "v"(b.sa), "v"(a.sa), "v"(c.sa)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.sb) : "v"(b.sb), "v"(a.sb), "v"(c.sb)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.sc) : "v"(b.sc), "v"(a.sc), "v"(c.sc)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.sd) : "v"(b.sd), "v"(a.sd), "v"(c.sd)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.se) : "v"(b.se), "v"(a.se), "v"(c.se)); - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.sf) : "v"(b.sf), "v"(a.sf), "v"(c.sf)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s0) : "v"(b.s0), "v"(a.s0), "v"(c.s0)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s1) : "v"(b.s1), "v"(a.s1), "v"(c.s1)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s2) : "v"(b.s2), "v"(a.s2), "v"(c.s2)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s3) : "v"(b.s3), "v"(a.s3), "v"(c.s3)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s4) : "v"(b.s4), "v"(a.s4), "v"(c.s4)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s5) : "v"(b.s5), "v"(a.s5), "v"(c.s5)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s6) : "v"(b.s6), "v"(a.s6), "v"(c.s6)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s7) : "v"(b.s7), "v"(a.s7), "v"(c.s7)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s8) : "v"(b.s8), "v"(a.s8), "v"(c.s8)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.s9) : "v"(b.s9), "v"(a.s9), "v"(c.s9)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.sa) : "v"(b.sa), "v"(a.sa), "v"(c.sa)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.sb) : "v"(b.sb), "v"(a.sb), "v"(c.sb)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.sc) : "v"(b.sc), "v"(a.sc), "v"(c.sc)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.sd) : "v"(b.sd), "v"(a.sd), "v"(c.sd)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.se) : "v"(b.se), "v"(a.se), "v"(c.se)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r.sf) : "v"(b.sf), "v"(a.sf), "v"(c.sf)); #endif return r; @@ -424,7 +424,7 @@ static u32 __byte_perm_S (const u32 a, const u32 b, const u32 c) { u32 r; - __asm__ volatile ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r) : "v"(b), "v"(a), "v"(c)); + __asm__ ("V_PERM_B32 %0, %1, %2, %3;" : "=v"(r) : "v"(b), "v"(a), "v"(c)); return r; } @@ -436,49 +436,49 @@ static u32x __add3 (const u32x a, const u32x b, const u32x c) u32x r; #if VECT_SIZE == 1 - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r) : "v"(b), "v"(a), "v"(c)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r) : "v"(b), "v"(a), "v"(c)); #endif #if VECT_SIZE >= 2 - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s0) : "v"(b.s0), "v"(a.s0), "v"(c.s0)); - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s1) : "v"(b.s1), "v"(a.s1), "v"(c.s1)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s0) : "v"(b.s0), "v"(a.s0), "v"(c.s0)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s1) : "v"(b.s1), "v"(a.s1), "v"(c.s1)); #endif #if VECT_SIZE >= 4 - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s0) : "v"(b.s0), "v"(a.s0), "v"(c.s0)); - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s1) : "v"(b.s1), "v"(a.s1), "v"(c.s1)); - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s2) : "v"(b.s2), "v"(a.s2), "v"(c.s2)); - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s3) : "v"(b.s3), "v"(a.s3), "v"(c.s3)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s0) : "v"(b.s0), "v"(a.s0), "v"(c.s0)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s1) : "v"(b.s1), "v"(a.s1), "v"(c.s1)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s2) : "v"(b.s2), "v"(a.s2), "v"(c.s2)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s3) : "v"(b.s3), "v"(a.s3), "v"(c.s3)); #endif #if VECT_SIZE >= 8 - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s0) : "v"(b.s0), "v"(a.s0), "v"(c.s0)); - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s1) : "v"(b.s1), "v"(a.s1), "v"(c.s1)); - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s2) : "v"(b.s2), "v"(a.s2), "v"(c.s2)); - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s3) : "v"(b.s3), "v"(a.s3), "v"(c.s3)); - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s4) : "v"(b.s4), "v"(a.s4), "v"(c.s4)); - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s5) : "v"(b.s5), "v"(a.s5), "v"(c.s5)); - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s6) : "v"(b.s6), "v"(a.s6), "v"(c.s6)); - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s7) : "v"(b.s7), "v"(a.s7), "v"(c.s7)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s0) : "v"(b.s0), "v"(a.s0), "v"(c.s0)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s1) : "v"(b.s1), "v"(a.s1), "v"(c.s1)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s2) : "v"(b.s2), "v"(a.s2), "v"(c.s2)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s3) : "v"(b.s3), "v"(a.s3), "v"(c.s3)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s4) : "v"(b.s4), "v"(a.s4), "v"(c.s4)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s5) : "v"(b.s5), "v"(a.s5), "v"(c.s5)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s6) : "v"(b.s6), "v"(a.s6), "v"(c.s6)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s7) : "v"(b.s7), "v"(a.s7), "v"(c.s7)); #endif #if VECT_SIZE >= 16 - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s0) : "v"(b.s0), "v"(a.s0), "v"(c.s0)); - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s1) : "v"(b.s1), "v"(a.s1), "v"(c.s1)); - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s2) : "v"(b.s2), "v"(a.s2), "v"(c.s2)); - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s3) : "v"(b.s3), "v"(a.s3), "v"(c.s3)); - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s4) : "v"(b.s4), "v"(a.s4), "v"(c.s4)); - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s5) : "v"(b.s5), "v"(a.s5), "v"(c.s5)); - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s6) : "v"(b.s6), "v"(a.s6), "v"(c.s6)); - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s7) : "v"(b.s7), "v"(a.s7), "v"(c.s7)); - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s8) : "v"(b.s8), "v"(a.s8), "v"(c.s8)); - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s9) : "v"(b.s9), "v"(a.s9), "v"(c.s9)); - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.sa) : "v"(b.sa), "v"(a.sa), "v"(c.sa)); - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.sb) : "v"(b.sb), "v"(a.sb), "v"(c.sb)); - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.sc) : "v"(b.sc), "v"(a.sc), "v"(c.sc)); - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.sd) : "v"(b.sd), "v"(a.sd), "v"(c.sd)); - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.se) : "v"(b.se), "v"(a.se), "v"(c.se)); - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.sf) : "v"(b.sf), "v"(a.sf), "v"(c.sf)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s0) : "v"(b.s0), "v"(a.s0), "v"(c.s0)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s1) : "v"(b.s1), "v"(a.s1), "v"(c.s1)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s2) : "v"(b.s2), "v"(a.s2), "v"(c.s2)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s3) : "v"(b.s3), "v"(a.s3), "v"(c.s3)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s4) : "v"(b.s4), "v"(a.s4), "v"(c.s4)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s5) : "v"(b.s5), "v"(a.s5), "v"(c.s5)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s6) : "v"(b.s6), "v"(a.s6), "v"(c.s6)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s7) : "v"(b.s7), "v"(a.s7), "v"(c.s7)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s8) : "v"(b.s8), "v"(a.s8), "v"(c.s8)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.s9) : "v"(b.s9), "v"(a.s9), "v"(c.s9)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.sa) : "v"(b.sa), "v"(a.sa), "v"(c.sa)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.sb) : "v"(b.sb), "v"(a.sb), "v"(c.sb)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.sc) : "v"(b.sc), "v"(a.sc), "v"(c.sc)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.sd) : "v"(b.sd), "v"(a.sd), "v"(c.sd)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.se) : "v"(b.se), "v"(a.se), "v"(c.se)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r.sf) : "v"(b.sf), "v"(a.sf), "v"(c.sf)); #endif return r; @@ -488,7 +488,7 @@ static u32 __add3_S (const u32 a, const u32 b, const u32 c) { u32 r; - __asm__ volatile ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r) : "v"(b), "v"(a), "v"(c)); + __asm__ ("V_ADD3_U32 %0, %1, %2, %3;" : "=v"(r) : "v"(b), "v"(a), "v"(c)); return r; }