math/mpx-mul4-*.S: Use more portable type syntax for ambiguous instructions.

[catacomb] / math / mpx-mul4-amd64-sse2.S

diff --git a/math/mpx-mul4-amd64-sse2.S b/math/mpx-mul4-amd64-sse2.S
index da3e6d6..5a748c6 100644 (file)
--- a/math/mpx-mul4-amd64-sse2.S
+++ b/math/mpx-mul4-amd64-sse2.S
@@ -1251,9 +1251,9 @@ FUNC(mpxmont_redc4_amd64_sse2)
        // Continue carry propagation until the end of the buffer.
 0:     add     [rdi], C
        mov     C, 0                    // preserves flags
-       adcd    [rdi + 4], 0
-       adcd    [rdi + 8], 0
-       adcd    [rdi + 12], 0
+       adc     dword ptr [rdi + 4], 0
+       adc     dword ptr [rdi + 8], 0
+       adc     dword ptr [rdi + 12], 0
        adc     C, 0
        add     rdi, 16
        cmp     rdi, DVL4
@@ -1270,10 +1270,10 @@ FUNC(mpxmont_redc4_amd64_sse2)
        jb      7b
 
        // All done for this iteration.  Start the next.
-8:     mov     rdi, DV                 // -> Z = dv[i]
-       mov     rbx, NV                 // -> X = nv[0]
-       cmp     rdi, DVLO               // all done yet?
+       cmp     DV, DVLO                // all done yet?
        jae     9f
+       mov     rdi, DV                 // -> Z = dv[i]
+       mov     rbx, NV                 // -> X = nv[0]
        add     DV, 16
        call    mont4
        add     rdi, 16
@@ -1601,6 +1601,8 @@ FUNC(test_mmul4)
        testtop r11
        call    mmul4
        testtail
+       pshufd  xmm10, xmm10, SHUF(0, 2, 1, 3)
+       pshufd  xmm11, xmm11, SHUF(0, 2, 1, 3)
        movdqu  [r10 +  0], xmm10
        movdqu  [r10 + 16], xmm11
        testcarryout
@@ -1612,6 +1614,8 @@ FUNC(test_mmla4)
        testtop r11
        call    mmla4
        testtail
+       pshufd  xmm10, xmm10, SHUF(0, 2, 1, 3)
+       pshufd  xmm11, xmm11, SHUF(0, 2, 1, 3)
        movdqu  [r10 +  0], xmm10
        movdqu  [r10 + 16], xmm11
        testcarryout
@@ -1623,6 +1627,8 @@ FUNC(test_mont4)
        testtop
        call    mont4
        testtail
+       pshufd  xmm10, xmm10, SHUF(0, 2, 1, 3)
+       pshufd  xmm11, xmm11, SHUF(0, 2, 1, 3)
        movdqu  [r10 +  0], xmm10
        movdqu  [r10 + 16], xmm11
        testcarryout